190 Commits

Author SHA1 Message Date
Patrick J Cherry
102738d9ad Updated logging output during readloop() and writeloop() failures
There's a handy SHOW_ERRNO macro we can use to get consistent logging
for system call failures from readloop() and writeloop().
2018-04-27 10:45:42 +01:00
James Carter
3e0d30f6b9 Merge branch 'reinstate-sync-after-every-write' into 'develop'
Reinstate sync after every write

See merge request open-source/flexnbd-c!51
2018-04-24 12:02:53 +01:00
Patrick J Cherry
3b1a150315 Updated changelgo 2018-04-24 10:27:46 +01:00
Patrick J Cherry
ead6328d80 Force sync after every write 2018-04-24 10:27:02 +01:00
James Carter
20b4f069c8 Merge branch 'release-to-master' into 'develop'
Merge back to develop

See merge request open-source/flexnbd-c!49
2018-02-20 11:52:37 +00:00
Patrick J Cherry
331ca4be14 Updated changelog for release 2018-02-20 11:45:42 +00:00
James Carter
fb5714765c Merge branch 'fix-formatting' into 'develop'
Formatted all code using `indent`

See merge request open-source/flexnbd-c!47
2018-02-20 11:42:25 +00:00
Patrick J Cherry
af3bb16ff7 Merge branch 'develop' into fix-formatting 2018-02-20 11:06:58 +00:00
Patrick J Cherry
9cbcc7c95a Added note about the test file formatting 2018-02-20 11:05:36 +00:00
Patrick J Cherry
8893cd06c4 Re-formatted tests with a bit of tinkering by hand 2018-02-20 11:02:33 +00:00
James Carter
166db9b1f7 Merge branch 'enable-flags-test' into 'develop'
Enable request flags test

See merge request open-source/flexnbd-c!48
2018-02-20 10:23:42 +00:00
Patrick J Cherry
103bd7ad5b Undo formatting on test suite -- it wasn't right 2018-02-20 10:13:42 +00:00
Patrick J Cherry
7bee1aadfe Enable request flags test
Missed this out when I wrote the test!
2018-02-20 10:11:38 +00:00
Patrick J Cherry
f47f56d4c4 Formatted all code using indent 2018-02-20 10:05:35 +00:00
James Carter
19a1127bde Merge branch 'fix-correct-num-clients-status' into 'develop'
Call the thread cleanup code when requesting `status`

See merge request open-source/flexnbd-c!46
2018-02-20 09:51:37 +00:00
James Carter
073a4ac0fa Merge branch '35-incorrect-struct-type-used-in-readwrite-c' into 'develop'
Resolve "Incorrect struct type used in readwrite.c"

Closes #35

See merge request open-source/flexnbd-c!41
2018-02-20 09:50:25 +00:00
Patrick J Cherry
623007bfff Remove last reference to removed test_gets_num_clients 2018-02-19 10:22:01 +00:00
Patrick J Cherry
27a94a807e Remove the test_gets_num_clients test from the C unit tests
This test was causing problems by using dummy pointers to simulate
connections.  When calling the cleanup code, these pointers were
thought to be real, and the code attemtped to clean up threads
referenced by those pointers, causing a segfault.

I've reimplemented the test in the ruby acceptance suite.
2018-02-16 13:46:31 +00:00
Patrick J Cherry
1407407ff4 Updated changelog 2018-02-16 13:00:31 +00:00
Patrick J Cherry
d0439dab88 Call the thread cleanup code when requesting status
This ensures the correct number of connected clients is returned when
the status command is issued.

Previously the thread pool would only be cleaned up on a new connection.
2018-02-16 12:58:03 +00:00
James F. Carter
9f56f38f42 Merge branch 'rationalise-ld-preload-tests' into develop 2018-02-14 16:48:57 +00:00
Chris Elsworth
370d04d971 Merge branch 'take-request-response-size-into-malloc' into 'develop'
Update proxy malloc to add the struct size onto the request/response buffer

See merge request open-source/flexnbd-c!45
2018-02-14 05:28:24 +00:00
Patrick J Cherry
099e29de91 Merge branch 'develop' into 'take-request-response-size-into-malloc'
# Conflicts:
#   debian/changelog
2018-02-13 17:06:41 +00:00
Patrick J Cherry
2e17e8955f Added tests for NBD_MAX_SIZE
This constant is only used in the proxy, so the tests only cover proxy
mode.
2018-02-13 17:04:51 +00:00
Patrick J Cherry
bb1f6ecdf5 Updated changelog 2018-02-13 15:51:09 +00:00
Patrick J Cherry
158379ba7a Use correct constant name. 2018-02-12 19:11:24 +00:00
Patrick J Cherry
1c66b56af1 Update proxy malloc to add the struct size onto the request/response buffer
This alters the meaning of NBD_MAX_SIZE to be the actual max request size
we'll accept over nbd.  Previously it was *nearly* the max size we'd
accept depending on the size of the struct.
2018-02-12 19:04:29 +00:00
Ian Chilton
03d9eb01b5 Merge branch 'increase-log-level-for-readloop-failures' into 'develop'
Increase log level for readloop failures, which might help with diagnosis

See merge request open-source/flexnbd-c!44
2018-02-09 15:38:48 +00:00
Patrick J Cherry
cdcd527544 Refactored read_reply to compare the network-byte-ordered handle 2018-02-09 12:18:34 +00:00
Patrick J Cherry
169d40f575 Increase log level for readloop failures, which might help with diagnosis 2018-02-09 11:57:07 +00:00
Patrick J Cherry
21f384e343 Updated changelog 2018-02-09 11:44:28 +00:00
Patrick J Cherry
9817fd7b0a Final tidies, comments etc. 2018-02-09 11:42:25 +00:00
Patrick J Cherry
195de41d86 Remove extra line 2018-02-09 11:32:26 +00:00
Patrick J Cherry
5b350e10e5 Merge branch 'develop' into '35-incorrect-struct-type-used-in-readwrite-c'
# Conflicts:
#   debian/changelog
2018-02-09 11:29:48 +00:00
Patrick J Cherry
b75a6529d0 Move LdPreload include to correct place 2018-02-09 10:41:24 +00:00
Patrick J Cherry
8e67180999 Check that TCP_NODELAY is set on upstream sockets on reconnection
Also rationalize the test to see if a function has been called.  Still
not great, but getting there :)
2018-02-09 10:26:08 +00:00
Patrick J Cherry
c053a54faa Added test to cover setsockopt for tcpkeepalive 2018-02-08 23:07:17 +00:00
Patrick J Cherry
ebacf738bc Tidy up ld preload hacks 2018-02-08 22:28:34 +00:00
James Carter
c4bab3f81f Merge branch 'truncate-odd-sized-discs' into 'develop'
Discs must be sized in multiples of 512 bytes or odd things happen

See merge request open-source/flexnbd-c!42
2018-02-08 16:49:36 +00:00
Patrick J Cherry
a19267b377 Adjust block-rounding line to match in serve.c 2018-02-08 16:37:36 +00:00
Patrick J Cherry
23d9ff587e Updated changelog 2018-02-08 16:36:20 +00:00
Patrick J Cherry
347b7978e4 Discs must be sized in multiples of 512 bytes or odd things happen
In #36 some of the odd errors were due to seeks beyond the end of the
disc.  This was because the disc was "specially crafted" to be 25GB + 1
byte, which doesn't fit into the normal 512 byte sectors expected of a
disc.  This lead to reads going beyond the end of the disc etc.

If a similarly evil disc is used with `losetup`, it just ignores the
last bytes of the disc that don't fit into 512 chunks.  This is what
that patch does, logging an error at the same time.
2018-02-08 16:31:28 +00:00
Patrick J Cherry
f8fec5f57e Alter struct types to reflect reality, avoiding mixing "host" and "raw" structs 2018-02-08 15:46:34 +00:00
James Carter
1672b4b88b Merge branch '36-breaks-when-trying-to-install-debian-from-cd' into 'develop'
Resolve "breaks when trying to install debian from CD"

Closes #36

See merge request open-source/flexnbd-c!40
2018-02-08 13:59:12 +00:00
Patrick J Cherry
5e9dbbd626 Updated changelgo 2018-02-08 13:32:10 +00:00
Patrick J Cherry
8beb3f0af6 Allow proxy to pass NBD protocol errors downstream; server returns EINVAL/ENOSPC appropriately
Previously the proxy would just disconnect when it saw an NBD protocol
error, and retry the operation it was in the middle of.

Additionally, the server needs to return the correct error types when
this happens.
2018-02-08 13:19:51 +00:00
James Carter
806de13024 Merge branch 'try-flags' into 'develop'
Set flags to show we can accept FUA and FLUSH commands

See merge request open-source/flexnbd-c!38
2018-02-08 11:18:31 +00:00
Patrick J Cherry
f71b872622 Only set up LD_PRELOAD for tests that actually need it. 2018-02-07 22:05:07 +00:00
Patrick J Cherry
79181b3153 Added LD_PRELOAD library to monitor msync calls in testing 2018-02-07 21:45:20 +00:00
Patrick J Cherry
55548cc969 Change ordering of @env configuration/start so we can alter the blocksize.
argh.
2018-02-06 10:24:54 +00:00
Patrick J Cherry
9bf3b52d54 Call proxy_finish_connect_to_upstream when reconnecting, setting
TCP_NODELAY
2018-02-06 10:02:16 +00:00
Patrick J Cherry
da35187af0 Allow blocksize to be changed in Environment
This number is peppered all over the test suite, so changing @blocksize
for everything is not a goer, when we really only need to change it for
one test.
2018-02-06 09:55:32 +00:00
Patrick J Cherry
7704f9e5c8 Fix tests to reflect new filesize. 2018-02-06 07:57:40 +00:00
Patrick J Cherry
3a86870c9f Use sysconf to determine actual page size for msync
Also added comments in tests around testing for msync offsets/lengths.
2018-02-06 07:32:58 +00:00
Patrick J Cherry
6d6948af09 Fix offset calculation for partial msyncs to go to nearest 4k block
Previously they were always set to zero.
2018-02-05 23:05:00 +00:00
Patrick J Cherry
c423900f02 Fix typo 2018-02-05 17:04:23 +00:00
Patrick J Cherry
afa1bb0efb Use msync rather than fsync to flush the entire disc
This involves storing the size of the mapped disc in the client struct,
and then supplying that to the msync command.
2018-02-05 17:01:32 +00:00
Patrick J Cherry
ad2014ac9d Fixed long-standing bug with h2r functions being back to front
h2r seemd to be using beXXtoh functions instead of htobeXX.  Foruntately
ROT13 works symmetrically on our systems..!
2018-02-05 16:16:17 +00:00
Patrick J Cherry
d1dc7392c2 Open file with O_NOATIME, not O_SYNC
O_SYNC is not necessary as we're not doing direct writes to the file.
O_NOATIME might give some speed boost.
2018-02-05 16:15:36 +00:00
Patrick J Cherry
ba59a4c03f Updated changelog 2018-02-05 08:15:56 +00:00
Patrick J Cherry
2b58468800 Added test for FUA acceptance.
Although I think this might be a bit useless as servers normally just
ingore flags.
2018-02-03 20:29:15 +00:00
Patrick J Cherry
4d9db4d6e9 Added basic FLUSH test 2018-02-03 20:10:47 +00:00
Patrick J Cherry
d6057a4244 Use 'English' in ruby 2018-02-02 21:41:07 +00:00
Patrick J Cherry
1d98ba1d3e Further rubocopping 2018-02-02 21:36:30 +00:00
Patrick J Cherry
9c48da82cc Rubocop 2018-02-02 21:34:14 +00:00
Patrick J Cherry
1b7b688f7a Tidied up nbd init test 2018-02-02 21:30:55 +00:00
Patrick J Cherry
3410ccd4c5 Fixed up commenting around our advertised flags. 2018-02-02 20:50:48 +00:00
Patrick J Cherry
051576df6d Remove warnings about Object#timeout 2018-02-02 20:46:46 +00:00
Patrick J Cherry
9eb7072f49 Removed some extra spaces I'd added 2018-02-02 20:46:25 +00:00
Patrick J Cherry
6aa5907f5e Tidied constants up a bit 2018-02-02 20:34:49 +00:00
Patrick J Cherry
72c8c6f757 Altered test to check for type as a 16-bit uint; added flags test 2018-02-02 20:30:39 +00:00
Patrick J Cherry
b22b99d9b9 Fix fill_request to set flags as well as type. 2018-02-02 20:28:00 +00:00
Patrick J Cherry
ad001cb83c Tidy comments 2018-02-02 16:17:01 +00:00
Patrick J Cherry
f37e4438c8 Merge branch 'develop' into try-flags 2018-02-02 16:05:57 +00:00
Chris Elsworth
084d429961 Merge branch 'update-changelog-for-mr35' into 'develop'
Updated changelog for !35

See merge request open-source/flexnbd-c!39
2018-02-02 14:57:58 +00:00
Patrick J Cherry
1883bee43c Updated changelog for !35 2018-02-02 14:52:26 +00:00
Patrick J Cherry
68a196e93d Allow the proxy connection to pass through flags from upstream. 2018-02-02 10:30:40 +00:00
Patrick J Cherry
1f0ef0aad6 Implement FLUSH command and honour FUA flag
I changed the request struct to break the 32 bits reserved for the
request type into two.  The first part of this is used for the flags
(such as FUA), and the second part for the command type.  Previously
we'd masked the top two bytes, thus ignoring any flags.
2018-02-01 22:13:59 +00:00
Patrick J Cherry
25cc084108 First steps towards implementing flags as part of oldstyle negotiation 2018-02-01 19:25:36 +00:00
Patrick J Cherry
f2fa00260b Merge branch 'avoid-crash-on-timeout' into 'develop'
avoid fatal error on client connection timeout

See merge request open-source/flexnbd-c!36
2018-01-26 16:04:51 +00:00
James F. Carter
b2007c9dad debian: uodate changelog 2018-01-26 15:06:26 +00:00
James F. Carter
9b1781164a avoid fatal error on client connection timeout 2018-01-26 15:03:44 +00:00
Ian Chilton
1f99929589 Merge branch 'develop' into 'develop'
Develop

See merge request open-source/flexnbd-c!35
2018-01-24 12:42:49 +00:00
Chris Cottam
c37627a5b9 not high enough, trying 32MB 2018-01-18 17:08:32 +00:00
Chris Cottam
ceb3328261 increasing the NBD max size to see if it fixes an issue with qemu-2.11.0 2018-01-18 16:52:24 +00:00
Patrick J Cherry
61940bdfc5 Merge branch '34-logging-should-include-the-id-of-the-disc-that-is-being-served' into 'develop'
add a log_context, a string output as part of any log message

Closes #34

See merge request open-source/flexnbd-c!34
2018-01-11 10:35:45 +00:00
James F. Carter
6d96d751d8 debian: update changelog 2018-01-11 10:06:03 +00:00
James F. Carter
fa75de0a8b proxy sets the upstream address and port as its log context 2018-01-11 10:04:18 +00:00
James F. Carter
1cb11bfd38 serve sets the disc's backing file as its log context 2018-01-11 10:03:16 +00:00
James F. Carter
2702e73a26 add a log_context, a string output as part of any log message 2018-01-11 10:01:42 +00:00
Patrick J Cherry
dbf50046a8 Merge branch '33-tcp-keepalive-should-be-applied-to-connection-so-that-dead-connections-can-be-properly-reaped' into 'develop'
apply tcp keepalive to serving sockets

Closes #33

See merge request open-source/flexnbd-c!33
2018-01-10 17:51:02 +00:00
James F. Carter
d62b069ce4 debian: update changelog 2018-01-10 13:58:11 +00:00
James F. Carter
884a714744 whitespace fix 2018-01-10 13:55:05 +00:00
James F. Carter
0c668f1776 remember how || works in C 2018-01-10 13:54:26 +00:00
James F. Carter
1d5b315f17 apply tcp keepalive to serving sockets 2018-01-10 13:49:22 +00:00
Patrick J Cherry
24f1e62a73 Merge branch 'release' into 'develop'
Merge changelog back to develop

See merge request !32
2017-07-14 17:41:51 +01:00
Chris Elsworth
5c37cba39b New release 2017-07-14 17:03:56 +01:00
James F. Carter
59f264184b Merge pull request #1 from BytemarkHosting/better-stats
Calculate and return bytes_left in migration statistics
2017-07-14 16:36:50 +01:00
Chris Elsworth
42d206cfb7 Update test 2017-07-14 16:26:25 +01:00
Chris Elsworth
ab3106202a Also return migration_bytes_left 2017-07-14 16:18:34 +01:00
James Carter
e04dead5ce Merge branch 'update-changelog' into 'develop'
Updated changelog.

See merge request !30
2017-04-13 12:52:00 +01:00
Patrick J Cherry
88bc5f0643 Updated changelog. 2017-04-13 12:49:55 +01:00
James Carter
e89c87e2b9 Merge branch 'fix-compiler-flags' into 'develop'
Remove lots of per-cpu compiler flags.

See merge request !28
2017-02-23 12:11:25 +00:00
Patrick J Cherry
9d2ac3f403 Remove lots of per-cpu compiler flags.
These flags appear to cause SIGILL when flexnbd starts on some CPUs.
2017-02-22 17:52:52 +00:00
James Carter
67823bf85b Merge branch '32-package-and-publish-in-gitlab-ci-retire-maker2-job' into 'master'
Resolve "package and publish in gitlab-ci - retire maker2 job"

Closes #32 and #21

See merge request !27
2017-01-23 14:04:43 +00:00
Patrick J Cherry
17d30b86ad Updated build-deps to have libsubunit and ruby-test-unit 2017-01-23 14:00:09 +00:00
Patrick J Cherry
b97bcd6f51 Don't test separately from packaging. Also use correct source "format" 2017-01-23 13:58:04 +00:00
Patrick J Cherry
4d3c15a4d0 Switch to native from quilted packaging 2017-01-23 13:52:22 +00:00
Patrick J Cherry
83d6872a8d Add ruby test dependency 2017-01-23 13:48:19 +00:00
Patrick J Cherry
ab8470aef3 Modernise gitlab-ci 2017-01-23 13:46:42 +00:00
Patrick J Cherry
716df32fd6 Merge remote-tracking branch 'origin/debian' into 32-package-and-publish-in-gitlab-ci-retire-maker2-job 2017-01-23 13:44:44 +00:00
Michel Pollet
1a768d5e9c Merge branch '29-fix-linker-issue' into 'master'
Link against subunit for testing.

This fixes the problems in Debian stretch+.

Closes #29

See merge request !26
2016-10-13 16:47:37 +01:00
Patrick J Cherry
72992c76ac Added libsubunit to the gitlab-ci 2016-10-13 16:42:21 +01:00
Patrick J Cherry
cace8123f4 Link against subunit for testing.
This fixes the problems in Debian stretch+.
2016-10-13 16:39:20 +01:00
Patrick J Cherry
c3b241464a Updated changelog 2016-10-07 12:26:52 +01:00
Patrick J Cherry
4f956e4b9d Merge branch 'master' of gitlab.bytemark.co.uk:open-source/flexnbd-c into debian 2016-10-07 12:24:51 +01:00
James Carter
b4cb2d9240 Merge branch 'fix-wrong-handle-type' into 'master'
Fix up "wrong" handle type from char* to uint64_t

Following from the NBD handle comparison simplifications.

See merge request !25
2016-10-07 10:20:35 +01:00
James Carter
1efb7bada6 Merge branch 'fix-unsigned-longs-in-bitset-test' into 'master'
fix check_bitset test on 32-bit platforms

The use of `unsigned long` and `UL` suffices caused this test to fail
on 32 bit platforms, where these are just 4, not 8 bits long.

```
tests/unit/check_bitset.c:73:F:bit:test_bit_ranges:0: longs[32] = 0 SHOULD BE ffffffff
```

See merge request !24
2016-10-07 10:20:08 +01:00
James Carter
6bc2a4c0b9 Merge branch 'fix-cast-from-pointer-to-wrong-size-integer-in-serve' into 'master'
This fixes the compiler warning pointer-to-int-cast in serve.c

```
In file included from src/server/bitset.h:4:0,
                 from src/server/mirror.h:8,
                 from src/server/flexnbd.h:5,
                 from src/server/serve.h:8,
                 from src/server/serve.c:1:
src/server/serve.c: In function 'tryjoin_client_thread':
src/server/serve.c:258:6: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
      (uint64_t)status);
      ^
```

See merge request !23
2016-10-07 09:59:24 +01:00
James Carter
59de76c50c Merge branch 'skip-large-file-test-on-i386' into 'master'
Skip large file test on 32-bit platforms

This test cannot run on 32-bit machines as they cannot access files
large than 2G.  Makes flexnbd on 32-bit a bit useless really..

See merge request !22
2016-10-07 09:57:09 +01:00
Patrick J Cherry
209da655b3 Skip large file test on 32-bit platforms
This test cannot run on 32-bit machines as they cannot access files
large than 2G.  Makes flexnbd on 32-bit a bit useless really..
2016-10-06 21:42:52 +01:00
Patrick J Cherry
52b45e6b40 fix check_bitset test on 32-bit platforms
The use of `unsigned long` and `UL` suffices caused this test to fail
on 32 bit platforms, where these are just 4, not 8 bits long.

```
tests/unit/check_bitset.c:73:F:bit:test_bit_ranges:0: longs[32] = 0 SHOULD BE ffffffff
```
2016-10-06 21:22:53 +01:00
Patrick J Cherry
d279eb7570 Fix up "wrong" handle type from char* to uint64_t
Following from the NBD handle comparison simplifications.
2016-10-06 21:19:15 +01:00
Patrick J Cherry
c07df76ede This fixes the compiler warning pointer-to-int-cast in serve.c
```
In file included from src/server/bitset.h:4:0,
                 from src/server/mirror.h:8,
                 from src/server/flexnbd.h:5,
                 from src/server/serve.h:8,
                 from src/server/serve.c:1:
src/server/serve.c: In function 'tryjoin_client_thread':
src/server/serve.c:258:6: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
      (uint64_t)status);
      ^
```
2016-10-06 21:16:07 +01:00
Patrick J Cherry
e7e99b099c Updated debian packaging, adding in new build-deps. 2016-10-06 16:02:15 +01:00
Patrick J Cherry
b2edd0734a Merge branch 'master' of gitlab.bytemark.co.uk:open-source/flexnbd-c into debian 2016-10-06 16:00:14 +01:00
James Carter
e19d005636 Merge branch '26-fix-function-definition' into 'master'
OK removed the cast and fixed the function def in the test

This should definitely clear the warning.

Closes #26

See merge request !21
2016-10-06 15:59:57 +01:00
Patrick J Cherry
d1e6e835c4 OK removed the cast and fixed the function def in the test
This should definitely clear the warning.
2016-10-06 15:56:57 +01:00
Patrick J Cherry
8fed794fe7 Merge branch 'master' of gitlab.bytemark.co.uk:open-source/flexnbd-c into debian 2016-10-06 15:47:25 +01:00
James Carter
e24efa9864 Merge branch '26-fix-compiler-warning' into 'master'
Resolve "tests/unit/check_readwrite.c causes compiler warnings"

Closes #26

See merge request !20
2016-10-06 15:46:41 +01:00
James Carter
3134d619ef Merge branch '27-fix-make-test' into 'master'
Update Makefile to specify dependencies properly for tests

Closes #27

See merge request !19
2016-10-06 15:44:46 +01:00
Patrick J Cherry
898f3f6c7e Reinstated char * cast to remove compiler warning 2016-10-06 15:43:20 +01:00
Patrick J Cherry
5a1bc21088 Update Makefile to specify dependencies properly for tests 2016-10-06 15:40:15 +01:00
James Carter
deb8f2c53b Merge branch 'fix-check-nbdtypes' into 'master'
Fix up nbdtypes test to correctly use htobe64

Previous change hadn't taken this into account, and hopefully this makes
our test a little clearer.

See merge request !18
2016-10-06 14:50:03 +01:00
Patrick J Cherry
1338d9e910 Fix up nbdtypes test to correctly use htobe64
Previous change hadn't taken this into account, and hopefully this makes
our test a little clearer.
2016-10-06 14:46:29 +01:00
James Carter
47c05174b6 Merge branch 'fix-check-readwrite' into 'master'
Fix check readwrite segfault

Little slip corrected :)

See merge request !17
2016-10-06 14:10:22 +01:00
Patrick J Cherry
191b3bc72c Merge branch 'master' of gitlab.bytemark.co.uk:open-source/flexnbd-c into fix-check-readwrite 2016-10-06 14:06:21 +01:00
James Carter
770ca0d0e5 Merge branch 'fix-test-names' into 'master'
Fixed up internal test names (copy/pasta?)

The test names output by `make check` now reflect reality.

See merge request !16
2016-10-06 14:04:54 +01:00
Patrick J Cherry
6505588f25 Fixed check_readwrite test to pass correct handle to fd_write_reply
The (char*) cast to resp->received.handle.b was causing a segfault
2016-10-06 14:01:47 +01:00
Patrick J Cherry
957707bcfc Fixed up internal test names (copy/pasta?)
The test names output by `make check` now reflect reality.
2016-10-06 13:44:20 +01:00
James Carter
3f01b77221 Merge branch 'update-manpages-again' into 'master'
Updated manpages, replaces a2x with txt2man

This simplifies the build-deps for Debian packages a little, and brings
the docs up to date.

See merge request !15
2016-10-06 13:43:53 +01:00
Patrick J Cherry
0dbea7f8fe Removed extra tabs
Oops
2016-10-06 13:11:07 +01:00
Patrick J Cherry
091aacd16d Updated manpages, replaces a2x with txt2man
This simplifies the build-deps for Debian packages a little, and brings
the docs up to date.
2016-10-06 12:55:05 +01:00
Patrick J Cherry
04b6637451 Merge branch 'failed-tests-cause-error' into 'master'
failures in make check now result in an error



See merge request !13
2016-10-05 17:20:43 +01:00
James F. Carter
7d2eda6cea failures in make check now result in an error 2016-10-05 16:28:27 +01:00
Patrick J Cherry
7e152ca4f2 Merge branch '24-tests-in-gitlab' into 'master'
Resolve "tests should be run in gitlab-ci"

Closes #24

See merge request !12
2016-10-05 15:08:22 +01:00
James F. Carter
fe0125efbc Merge branch 'master' into 24-tests-in-gitlab 2016-10-05 14:27:56 +01:00
James Carter
ebaaa6d671 Merge branch '25-retire-rake' into 'master'
Moved tasks from Rake to Make

The rake file was obsolete, apart from one invocation of ruby in a shell!

Closes #25

See merge request !11
2016-10-05 14:26:06 +01:00
James F. Carter
8cc8588744 Merge branch '24-tests-in-gitlab' of gitlab.bytemark.co.uk:open-source/flexnbd-c into 24-tests-in-gitlab 2016-10-05 13:06:36 +01:00
James F. Carter
5da77ea39a remove unnecessary step in gitlab-ci 2016-10-05 12:52:32 +01:00
James F. Carter
a744965c67 add missing deps on server object files when building check binaries 2016-10-05 12:51:58 +01:00
Patrick J Cherry
d07659f694 Merge branch '24-tests-in-gitlab' of gitlab.bytemark.co.uk:open-source/flexnbd-c into 25-retire-rake 2016-10-05 12:49:53 +01:00
Patrick J Cherry
30562ed900 Added dpkg-dev to requirements
Allows dpkg-architecture to run.
2016-10-05 12:49:25 +01:00
Patrick J Cherry
93c0fa2e92 Merged in gitlab-ci.yml and fixed to use Make
The CI should now use Make instead of Rake
2016-10-05 12:47:24 +01:00
Patrick J Cherry
8dc491fb89 Merge branch '24-tests-in-gitlab' of gitlab.bytemark.co.uk:open-source/flexnbd-c into 25-retire-rake 2016-10-05 12:46:47 +01:00
Patrick J Cherry
ea7cd64fc2 Moved tasks from Rake to Make
The rake file was obsolete, apart from one invocation of ruby in a
shell!
2016-10-05 12:36:06 +01:00
James F. Carter
35d3340708 avoid need for slow-to-install asciidoc in gitlab-ci 2016-10-05 12:10:21 +01:00
James F. Carter
d47a44a204 install asciidoc in gitlab-ci 2016-10-05 12:07:24 +01:00
James F. Carter
d6968d8242 explicitly compile before running tests in gitlab-ci 2016-10-05 12:06:11 +01:00
James F. Carter
bf85e329a0 clean build environment before running tests in gitlab-ci 2016-10-05 12:03:23 +01:00
James F. Carter
edcaef532c revert gitlab-ci to ruby2.1 2016-10-05 11:58:50 +01:00
James F. Carter
cb920e4e9d Merge branch 'master' into 24-tests-in-gitlab 2016-10-05 11:58:21 +01:00
Patrick J Cherry
91d85633b6 Merge branch 'force-encoding-for-ruby21' into 'master'
force binary encoding in a ruby2.1-compatible way



See merge request !10
2016-10-05 11:57:14 +01:00
James Carter
7c516b85a6 Merge branch 'makefile-fixes' into 'master'
Makefile fixes



See merge request !2
2016-10-05 11:55:54 +01:00
James F. Carter
679fa6dbf8 force binary encoding in a ruby2.1-compatible way 2016-10-05 11:54:09 +01:00
James F. Carter
50708326ec try ruby2.3 in gitlab-ci 2016-10-05 11:42:32 +01:00
James F. Carter
d907025d71 try a newer version of ruby in gitlab-ci 2016-10-05 11:39:56 +01:00
James F. Carter
e4d398a078 install net-tools in gitlab-ci 2016-10-05 11:29:42 +01:00
James F. Carter
8de0780125 install libev-dev in gitlab-ci 2016-10-05 11:26:46 +01:00
James F. Carter
0fd16822ea run tests in gitlab-ci 2016-10-05 11:12:39 +01:00
Patrick J Cherry
1e3c61b541 Merge branch '23-fix-unit-tests' into 'master'
update tests to reflect changes in handle storage

Closes #23

See merge request !9
2016-10-05 11:08:21 +01:00
James F. Carter
a09e14b2d4 whitespace fix 2016-10-05 11:06:39 +01:00
James F. Carter
a6710b6c32 update tests to reflect changes in handle storage 2016-10-05 10:57:52 +01:00
Patrick J Cherry
ed3995303f Reinstate doc to all 2016-10-05 10:41:23 +01:00
James Carter
f5de8fb12b Merge branch '20-fix-encoding-failures' into 'master'
Use a BINARY encoded string when doing read/write comparisons.

This is a bit of a cheat really, but `#read` returns an ASCII encoded
string, where as our ruby generates UTF-8 encoded strings, causing
assertion failures.

Closes #20

See merge request !8
2016-10-05 10:32:05 +01:00
James F. Carter
99a5f79a52 fixed typo 2016-10-05 10:30:44 +01:00
Patrick J Cherry
356e1fd6a1 Use a BINARY encoded string when doing read/write comparisons.
This is a bit of a cheat really, but `#read` returns an ASCII encoded
string, where as our ruby generates UTF-8 encoded strings, causing
assertion failures.

Fixes #20
2016-10-05 10:01:15 +01:00
James Carter
67dcea207d Merge branch '19-fix-double-definition-warnings' into 'master'
Fixes "double-definition of constants" warning

Looks like `#constants.include?` doesn't work as well as `#const_defined?`.

Closes #19

See merge request !6
2016-10-05 10:00:50 +01:00
Patrick J Cherry
d3762162db Fixes "double-definition of constants" warning
Looks like `#constants.include?` doesn't work as well as
`#const_defined?`.
2016-10-05 09:29:07 +01:00
Patrick J Cherry
3571d3f82e Added net-tools to the build-deps for testing
Fixes #21
2016-10-05 09:27:10 +01:00
Patrick J Cherry
4cd7e764bb Updated changelog 2016-10-04 21:22:07 +01:00
Patrick J Cherry
4f535fbb02 Merge branch 'master' of gitlab.bytemark.co.uk:open-source/flexnbd-c into debian 2016-10-04 21:14:26 +01:00
James Carter
218c55fb63 Merge branch 'simplify-nbd-handles-part-deux' into 'master'
Simplified NBD handle comparisons

8 bytes, therefore a uing64_t to compare to, no need for memcmp()

Signed-off-by: Michel Pollet <buserror@gmail.com>

See merge request !5
2016-10-04 15:49:07 +01:00
Michel Pollet
956a602475 Simplified NBD handle comparisons
8 bytes, therefore a uing64_t to compare to, no need for memcmp()

Signed-off-by: Michel Pollet <buserror@gmail.com>
2016-10-04 15:41:48 +01:00
James Carter
26a0a82f9d Merge branch '12-fix-bind' into 'master'
Attempt at fixing bind() bug

This will prevent the bind() wrapper to loop forever in some cases. I
could nor reproduce the issue, but this removes the only infinite loop I
could find.

Closes #12

See merge request !3
2016-10-04 15:41:37 +01:00
Michel Pollet
76e0476113 Attempt at fixing bind() bug
This will prevent the bind() wrapper to loop forever in some cases. I
could nor reproduc the issue, but this removes the only infinite loop I
could find.

Signed-off-by: Michel Pollet <buserror@gmail.com>
2016-10-04 15:36:46 +01:00
Michel Pollet
d9651a038c Makefile: don't include *.d's before 'all'
Include any .d file from the build directory, and do that after all the
other targets

Signed-off-by: Michel Pollet <buserror@gmail.com>
2016-10-04 15:32:56 +01:00
Michel Pollet
fcd3d33498 Simplified Makefile
gcc and clang can generate dep files as well as compiling in a single
pass, no need for two.

Signed-off-by: Michel Pollet <buserror@gmail.com>
2016-10-04 15:32:49 +01:00
James Carter
e3360a3a1b Merge branch 'cherry-pick-41f25408' into 'master'
Close socket fix, might relate to migration crashing

This was listed as a bug, and was immediatelly picked the static
analyzer anyway, this is very likely the cause for the
migration-cancel-crash bug.

closes #10 and possibly closes #11

See merge request !1
2016-09-14 11:29:12 +01:00
Michel Pollet
1fefe1a669 Close socket fix, might relate to migration crashing
This was listed as a bug, and was immediatelly picked the static
analyzer anyway, this is very likely the cause for the
migration-cancel-crash bug.

Signed-off-by: Michel Pollet <buserror@gmail.com>
2016-09-14 10:45:49 +01:00
114 changed files with 9156 additions and 8796 deletions

27
.gitlab-ci.yml Normal file
View File

@@ -0,0 +1,27 @@
stages:
- package
- publish
package:jessie: &package
stage: package
image: $CI_REGISTRY/docker-images/layers:$DISTRO-deb
variables:
DISTRO: jessie
script:
- package
artifacts:
paths:
- pkg/
package:stretch:
<<: *package
variables:
DISTRO: stretch
publish:
stage: publish
tags:
- shell
script:
- publish

24
CONTRIBUTING.md Normal file
View File

@@ -0,0 +1,24 @@
# Contribution guide
The code is formatted using the K&R style of "indent".
```
indent -kr <files go here>
```
The C unit tests have also been indented in the same way, but manually adjsted
such that the functions follow the normal libcheck layout.
```c
START_TEST( ... ) {
}
END TEST
```
Indent tends to mangle the `END_TEST` macro, so that will need adjusting if
`indent` is run over the test files again.

View File

@@ -11,28 +11,9 @@ ifdef DEBUG
else
CFLAGS_EXTRA=-O2
endif
CFLAGS_EXTRA += -fPIC --std=gnu99
LDFLAGS_EXTRA += -Wl,--relax,--gc-sections
TOOLCHAIN := $(shell $(CC) --version|awk '/Debian/ {print "debian";exit;}')
#
# This bit adds extra flags depending of the distro, and the
# architecture. To make sure debian packages have the right
# set of 'native' flags on them
#
ifeq ($(TOOLCHAIN),debian)
DEBARCH := $(shell dpkg-architecture -qDEB_BUILD_ARCH)
ifeq ($(DEBARCH),$(filter $(DEBARCH),amd64 i386))
CFLAGS_EXTRA += -march=native
endif
ifeq ($(DEBARCH),armhf)
CFLAGS_EXTRA += -march=armv7-a -mtune=cortex-a8 -mfpu=neon
endif
LDFLAGS_EXTRA += -L$(LIB) -Wl,-rpath,${shell readlink -f ${LIB}}
else
LDFLAGS_EXTRA += -L$(LIB) -Wl,-rpath-link,$(LIB)
endif
LDFLAGS_EXTRA += -Wl,--relax,--gc-sections -L$(LIB) -Wl,-rpath-link,$(LIB)
# The -Wunreachable-code warning is only implemented in clang, but it
# doesn't break anything for gcc to see it.
@@ -42,23 +23,19 @@ WARNINGS=-Wall \
-Wstrict-prototypes \
-Wno-missing-field-initializers \
-Wunreachable-code
CCFLAGS=-D_GNU_SOURCE=1 $(WARNINGS) $(CFLAGS_EXTRA) $(CFLAGS)
LLDFLAGS=-lm -lrt -lev $(LDFLAGS_EXTRA) $(LDFLAGS)
CC?=gcc
LIBS=-lpthread
INC=-I/usr/include/libev -Isrc/common -Isrc/server -Isrc/proxy
COMPILE=$(CC) $(INC) -c $(CCFLAGS)
SAVEDEP=$(CC) $(INC) -MM $(CCFLAGS)
COMPILE=$(CC) -MMD $(INC) -c $(CCFLAGS)
LINK=$(CC) $(LLDFLAGS) -Isrc $(LIBS)
LIB=build/
EXISTING_OBJS := $(wildcard build/*.o)
-include $(EXISTING_OBJS:.o=.d)
COMMON_SRC := $(wildcard src/common/*.c)
SERVER_SRC := $(wildcard src/server/*.c)
PROXY_SRC := $(wildcard src/proxy/*.c)
@@ -71,12 +48,13 @@ SRCS := $(COMMON_SRC) $(SERVER_SRC) $(PROXY_SRC)
OBJS := $(COMMON_OBJ) $(SERVER_OBJ) $(PROXY_OBJ)
all: build/flexnbd build/flexnbd-proxy doc
all: build doc
build: server proxy
build/%.o: %.c
mkdir -p $(dir $@)
$(COMPILE) $< -o $@
$(SAVEDEP) $< > build/$*.d
objs: $(OBJS)
@@ -87,41 +65,42 @@ build/flexnbd-proxy: $(COMMON_OBJ) $(PROXY_OBJ) build/proxy-main.o
$(LINK) $^ -o $@
server: build/flexnbd
proxy: build/flexnbd-proxy
CHECK_SRC := $(wildcard tests/unit/*.c)
CHECK_OBJ := $(CHECK_SRC:tests/unit/%.c=build/tests/%.o)
CHECK_OBJ := $(CHECK_SRC:tests/unit/%.c=build/%.o)
# Why can't we reuse the build/%.o rule above? Not sure.
build/tests/%.o: tests/unit/%.c
mkdir -p $(dir $@)
$(COMPILE) $< -o $@
$(SAVEDEP) $< > build/tests/$*.d
CHECK_BINS := $(CHECK_OBJ:build/tests/%.o=build/tests/%)
build/tests/%: build/tests/%.o $(OBJS)
$(LINK) $^ -o $@ -lcheck
CHECK_BINS := $(CHECK_SRC:tests/unit/%.c=build/%)
build/check_%: build/check_%.o
$(LINK) $^ -o $@ $(COMMON_OBJ) $(SERVER_OBJ) -lcheck -lsubunit
check_objs: $(CHECK_OBJ)
check_bins: $(CHECK_BINS)
check: $(CHECK_BINS)
for bin in $^; do $$bin; done
check: $(OBJS) $(CHECK_BINS)
r=true ; for bin in $(CHECK_BINS); do $$bin || r=false; done ; $$r
acceptance: build
cd tests/acceptance && RUBYOPT='-I.' ruby nbd_scenarios -v
test: check acceptance
build/flexnbd.1: README.txt
a2x --destination-dir build --format manpage $<
txt2man -t flexnbd -s 1 $< > $@
build/flexnbd-proxy.1: README.proxy.txt
a2x --destination-dir build --format manpage $<
txt2man -t flexnbd-proxy -s 1 $< > $@
# If we don't pipe to file, gzip clobbers the original, causing make
# to rebuild each time
%.1.gz: %.1
gzip -c -f $< > $@
server-man: build/flexnbd.1.gz
proxy-man: build/flexnbd-proxy.1.gz
doc: server-man proxy-man
doc: build/flexnbd.1.gz build/flexnbd-proxy.1.gz
install:
mkdir -p $(INSTALLDIR)
@@ -130,5 +109,7 @@ install:
clean:
rm -rf build/*
.PHONY: clean objs check_objs all server proxy check_bins check doc build test acceptance
.PHONY: clean objs check_objs all server proxy check_bins check server-man proxy-man doc
# Include extra dependencies at the end, NOT before 'all'
-include $(wildcard build/*.d)

View File

@@ -1,19 +1,14 @@
FLEXNBD-PROXY(1)
================
:doctype: manpage
NAME
----
flexnbd-proxy - A simple NBD proxy
flexnbd-proxy - A simple NBD proxy
SYNOPSIS
--------
*flexnbd-proxy* ['OPTIONS']
flexnbd-proxy --addr ADDR [--port PORT] --conn-addr ADDR
--conn-port PORT [--bind ADDR] [--cache[=CACHE_BYTES]]
[--help] [--verbose] [--quiet]
DESCRIPTION
-----------
flexnbd-proxy is a simple NBD proxy server that implements resilient
connection logic for the client. It connects to an upstream NBD server
@@ -25,11 +20,6 @@ of view of the client) reconnects and retransmits the request, before
returning the response to the client.
USAGE
-----
$ flexnbd-proxy --addr <ADDR> [ --port <PORT> ]
--conn-addr <ADDR> --conn-port <PORT>
[--bind <ADDR>] [--cache[=<CACHE_BYTES>]] [option]*
Proxy requests from an NBD client to an NBD server, resiliently. Only one
client can be connected at a time, and ACLs cannot be applied to the client, as they
@@ -58,75 +48,73 @@ Only one request may be in-flight at a time under the current architecture; that
doesn't seem to slow things down much relative to alternative options, but may
be changed in the future if it becomes an issue.
Options
~~~~~~~
OPTIONS
*--addr, -l ADDR*:
--addr, -l ADDR
The address to listen on. If this begins with a '/', it is assumed to be
a UNIX domain socket to create. Otherwise, it should be an IPv4 or IPv6
address.
*--port, -p PORT*:
--port, -p PORT
The port to listen on, if --addr is not a UNIX socket.
*--conn-addr, -C ADDR*:
--conn-addr, -C ADDR
The address of the NBD server to connect to. Required.
*--conn-port, -P PORT*:
--conn-port, -P PORT
The port of the NBD server to connect to. Required.
*--cache, -c=CACHE_BYTES*:
--cache, -c=CACHE_BYTES
If given, the size in bytes of read cache to use. CACHE_BYTES
defaults to 4096.
*--help, -h* :
--help, -h
Show command or global help.
*--verbose, -v* :
--verbose, -v
Output all available log information to STDERR.
*--quiet, -q* :
--quiet, -q
Output as little log information as possible to STDERR.
LOGGING
-------
Log output is sent to STDERR. If --quiet is set, no output will be seen
unless the program termintes abnormally. If neither --quiet nor
Log output is sent to STDERR. If --quiet is set, no output will be
seen unless the program termintes abnormally. If neither --quiet nor
--verbose are set, no output will be seen unless something goes wrong
with a specific request. If --verbose is given, every available log
message will be seen (which, for a debug build, is many). It is not an
error to set both --verbose and --quiet. The last one wins.
with a specific request. If --verbose is given, every available log
message will be seen (which, for a debug build, is many). It is not an
error to set both --verbose and --quiet. The last one wins.
The log line format is:
<TIMESTAMP>:<LEVEL>:<PID> <THREAD> <SOURCEFILE>:<SOURCELINE>: <MSG>
<TIMESTAMP>:<LEVEL>:<PID> <THREAD> <SOURCEFILE>:<SOURCELINE>: <MSG>
*TIMESTAMP*:
<TIMESTAMP>
Time the log entry was made. This is expressed in terms of monotonic ms
*LEVEL*:
<LEVEL>
This will be one of 'D', 'I', 'W', 'E', 'F' in increasing order of
severity. If flexnbd is started with the --quiet flag, only 'F' will be
seen. If it is started with the --verbose flag, any from 'I' upwards
will be seen. Only if you have a debug build and start it with
--verbose will you see 'D' entries.
severity. If flexnbd is started with the --quiet flag, only 'F' will
be seen. If it is started with the --verbose flag, any from 'I'
upwards will be seen. Only if you have a debug build and start it
with --verbose will you see 'D' entries.
*PID*:
<PID>
This is the process ID.
*THREAD*:
flexnbd-proxy is currently single-threaded, so this should be the same
for all lines. That may not be the case in the future.
<THREAD>
flexnbd-proxy is currently single-threaded, so this should be the
same for all lines. That may not be the case in the future.
*SOURCEFILE:SOURCELINE*:
<SOURCEFILE:SOURCELINE>
Identifies where in the source code this log line can be found.
*MSG*:
<MSG>
A short message describing what's happening, how it's being done, or
if you're very lucky *why* it's going on.
if you're very lucky why it's going on.
Proxying
~~~~~~~~
EXAMPLES
The main point of the proxy mode is to allow clients that would otherwise break
when the NBD server goes away (during a migration, for instance) to see a
@@ -160,53 +148,59 @@ The data in myfile has been moved between physical servers without the nbd
client process having to be disturbed at all.
READ CACHE
----------
If the --cache option is given at the command line, either without an
argument or with an argument greater than 0, flexnbd-proxy will use a
read-ahead cache. The cache as currently implemented doubles each read
read-ahead cache. The cache as currently implemented doubles each read
request size, up to a maximum of 2xCACHE_BYTES, and retains the latter
half in a buffer. If the next read request from the client exactly
half in a buffer. If the next read request from the client exactly
matches the region held in the buffer, flexnbd-proxy responds from the
cache without making a request to the server.
This pattern is designed to match sequential reads, such as those
performed by a booting virtual machine.
Note: If specifying a cache size, you *must* use this form:
Note: If specifying a cache size, you must use this form:
nbd-client$ flexnbd-proxy --cache=XXXX
That is, the '=' is required. This is a limitation of getopt-long.
That is, the '=' is required. This is a limitation of getopt-long.
If no cache size is given, a size of 4096 bytes is assumed. Caching can
be explicitly disabled by setting a size of 0.
BUGS
----
Should be reported to nick@bytemark.co.uk.
Should be reported via GitHub.
* https://github.com/BytemarkHosting/flexnbd-c/issues
Current issues include:
* Only old-style NBD negotiation is supported
* Only one request may be in-flight at a time
* All I/O is blocking, and signals terminate the process immediately
* UNIX socket support is limited to the listen address
* FLUSH and TRIM commands, and the FUA flag, are not supported
* DISCONNECT requests do not get passed through to the NBD server
* No active timeout-retry of requests - we trust the kernel's idea of failure
* only old-style NBD negotiation is supported;
* only one request may be in-flight at a time;
* all I/O is blocking, and signals terminate the process immediately;
* UNIX socket support is limited to the listen address;
* FLUSH and TRIM commands, and the FUA flag, are not supported;
* DISCONNECT requests do not get passed through to the NBD server;
* no active timeout-retry of requests - we trust the kernel's idea of
failure.
AUTHOR
------
Written by Alex Young <alex@bytemark.co.uk>.
Originally written by Alex Young <alex@blackkettle.org>.
Original concept and core code by Matthew Bloch <matthew@bytemark.co.uk>.
Proxy mode written by Nick Thomas <nick@bytemark.co.uk>
Proxy mode written by Nick Thomas <me@ur.gs>.
COPYING
-------
The full commit history is available on GitHub.
Copyright (c) 2012 Bytemark Hosting Ltd. Free use of this software is
granted under the terms of the GNU General Public License version 3 or
later.
SEE ALSO
flexnbd(1), nbd-client(8), xnbd-server(8), xnbd-client(8)
COPYRIGHT
Copyright (c) 2012-2016 Bytemark Hosting Ltd. Free use of this
software is granted under the terms of the GNU General Public License
version 3 or later.

View File

@@ -1,17 +1,36 @@
FLEXNBD(1)
==========
:doctype: manpage
NAME
----
flexnbd - A fast NBD server
SYNOPSIS
--------
*flexnbd* 'COMMAND' ['OPTIONS']
flexnbd MODE [ ARGS ]
flexnbd serve --addr ADDR --port PORT --file FILE [--sock SOCK]
[--default-deny] [--killswitch] [global_option]* [acl_entry]*
flexnbd listen --addr ADDR --port PORT --file FILE [--sock SOCK]
[--default-deny] [global_option]* [acl_entry]*
flexnbd mirror --addr ADDR --port PORT --sock SOCK [--unlink]
[--bind BIND_ADDR] [global_option]*
flexnbd acl --sock SOCK [acl_entry]+ [global_option]*
flexnbd break --sock SOCK [global_option]*
flexnbd status --sock SOCK [global_option]*
flexnbd read --addr ADDR --port PORT --from OFFSET --size SIZE
[--bind BIND_ADDR] [global_option]*
flexnbd write --addr ADDR --port PORT --from OFFSET --size SIZE
[--bind BIND_ADDR] [global_option]*
flexnbd help [mode] [global_option]*
DESCRIPTION
-----------
Flexnbd is a fast NBD server which supports live migration. Live
migration is performed by writing the data to a new server. A failed
migration will be invisible to any connected clients.
@@ -19,304 +38,290 @@ migration will be invisible to any connected clients.
Flexnbd tries quite hard to preserve sparsity of files it is serving,
even across migrations.
COMMANDS
--------
SERVE MODE
Serve a file.
serve
~~~~~
$ flexnbd serve --addr <ADDR> --port <PORT> --file <FILE>
[--sock <SOCK>] [--default-deny] [-k] [global option]* [acl entry]*
[--sock <SOCK>] [--default-deny] [-k] [global_option]*
[acl_entry]*
Serve a file. If any ACL entries are given (which should be IP
If any ACL entries are given (which should be IP
addresses), only those clients listed will be permitted to connect.
flexnbd will continue to serve until a SIGINT, SIGQUIT, or a successful
migration.
Options
^^^^^^^
OPTIONS
*--addr, -l ADDR*:
--addr, -l ADDR
The address to listen on. Required.
*--port, -p PORT*:
--port, -p PORT
The port to listen on. Required.
*--file, -f FILE*:
--file, -f FILE
The file to serve. Must already exist. Required.
*--sock, -s SOCK*:
Path to a control socket to open. You will need this if you want to
--sock, -s SOCK
Path to a control socket to open. You will need this if you want to
migrate, get the current status, or manipulate the access control
list.
*--default-deny, -d*:
How to interpret an empty ACL. If --default-deny is given, an
empty ACL will let no clients connect. If it is not given, an
--default-deny, -d
How to interpret an empty ACL. If --default-deny is given, an
empty ACL will let no clients connect. If it is not given, an
empty ACL will let any client connect.
*--killswitch, -k*:
--killswitch, -k
If set, we implement a 2-minute timeout on NBD requests and
responses. If a request takes longer than that to complete,
the client is disconnected. This is useful to keep broken
clients from breaking migrations, among other things.
listen
~~~~~~
$ flexnbd listen --addr <ADDR> --port <PORT> --file <FILE>
[--sock <SOCK>] [--default-deny] [global option]* [acl entry]*
LISTEN MODE
Listen for an inbound migration, and quit with a status of 0 on
completion.
$ flexnbd listen --addr ADDR --port PORT --file FILE
[--sock SOCK] [--default-deny] [global_option]*
[acl_entry]*
flexnbd will wait for a successful migration, and then quit. The file
to write the inbound migration data to must already exist before you
run 'flexnbd listen'.
Only one sender may connect to send data, and if the sender
disconnects part-way through the migration, the destination will
expect it to reconnect and retry the whole migration. It isn't safe
expect it to reconnect and retry the whole migration. It isn't safe
to assume that a partial migration can be resumed because the
destination has no knowledge of whether a client has made a write to
the source in the interim.
If the migration fails for a reason which the `flexnbd listen` process
If the migration fails for a reason which the 'flexnbd listen' process
can't fix (say, a failed local write), it will exit with an error
status. In this case, the sender will continually retry the migration
until it succeeds, and you will need to restart the `flexnbd listen`
status. In this case, the sender will continually retry the migration
until it succeeds, and you will need to restart the 'flexnbd listen'
process to allow that to happen.
Options
^^^^^^^
As for 'serve'.
OPTIONS
mirror
~~~~~~
As for serve.
$ flexnbd mirror --addr <ADDR> --port <PORT> --sock SOCK
[--unlink] [--bind <BIND-ADDR>] [global option]*
MIRROR MODE
Start a migration from the server with control socket SOCK to the server
listening at ADDR:PORT.
$ flexnbd mirror --addr ADDR --port PORT --sock SOCK [--unlink]
[--bind BIND_ADDR] [global_option]*
Migration can be a slow process. Rather than block the 'flexnbd mirror'
process until it completes, it will exit with a message of "Migration
started" once it has confirmation that the local server was able to
connect to ADDR:PORT and got an NBD header back. To check on the
connect to ADDR:PORT and got an NBD header back. To check on the
progress of a running migration, use 'flexnbd status'.
If the destination unexpectedly disconnects part-way through the
migration, the source will attempt to reconnect and start the migration
again. It is not safe to resume the migration from where it left off
again. It is not safe to resume the migration from where it left off
because the source can't see that the backing store behind the
destination is intact, or even on the same machine.
If the `--unlink` option is given, the local file will be deleted
immediately before the mirror connection is terminated. This allows
If the --unlink option is given, the local file will be deleted
immediately before the mirror connection is terminated. This allows
an otherwise-ambiguous situation to be resolved: if you don't unlink
the file and the flexnbd process at either end is terminated, it's not
possible to tell which copy of the data is canonical. Since the
possible to tell which copy of the data is canonical. Since the
unlink happens as soon as the sender knows that it has transmitted all
the data, there can be no ambiguity.
Note: files smaller than 4096 bytes cannot be mirrored.
Options
^^^^^^^
OPTIONS
*--addr, -l ADDR*:
The address of the remote server to migrate to. Required.
--addr, -l ADDR
The address of the remote server to migrate to. Required.
*--port, -p PORT*:
The port of the remote server to migrate to. Required.
--port, -p PORT
The port of the remote server to migrate to. Required.
*--sock, -s SOCK*:
The control socket of the local server to migrate from. Required.
--sock, -s SOCK
The control socket of the local server to migrate from. Required.
*--unlink, -u*:
Unlink the served file from the local filesystem after successfully
mirroring.
--unlink, -u
Unlink the served file from the local filesystem after
successfully mirroring.
*--bind, -b BIND-ADDR*:
The local address to bind to. You may need this if the remote server
is using an access control list.
--bind, -b BIND_ADDR
The local address to bind to. You may need this if the remote
server is using an access control list.
break
~~~~~
$ flexnbd mirror --sock SOCK [global option]*
BREAK MODE
Stop a running migration.
Options
^^^^^^^
$ flexnbd break --sock SOCK [global_option]*
*--sock, -s SOCK*:
The control socket of the local server whose emigration to stop.
Required.
OPTIONS
--sock, -s SOCK
The control socket of the local server whose migration to stop.
Required.
acl
~~~
$ flexnbd acl --sock <SOCK> [acl entry]+ [global option]*
ACL MODE
Set the access control list of the server with the control socket SOCK
to the given access control list entries.
$ flexnbd acl --sock SOCK [acl_entry]+ [global_option]*
ACL entries are given as IP addresses.
Options
^^^^^^^
OPTIONS
*--sock, -s SOCK*:
The control socket of the server whose ACL to replace.
--sock, -s SOCK
The control socket of the server whose ACL to replace. Required
status
~~~~~~
$ flexnbd status --sock <SOCK> [global option]*
STATUS MODE
Get the current status of the server with control socket SOCK.
The status will be printed to STDOUT. It is a space-separated list of
key=value pairs. The space character will never appear in a key or
value. Currently reported values are:
$ flexnbd status --sock SOCK [global_option]*
*pid*:
The status will be printed to STDOUT. It is a space-separated list of
key=value pairs. The space character will never appear in a key or
value. Currently reported values are:
pid
The process id of the server listening on SOCK.
*is_mirroring*:
is_mirroring
'true' if this server is sending migration data, 'false' otherwise.
*has_control*:
has_control
'false' if this server was started in 'listen' mode. 'true' otherwise.
read
~~~~
OPTIONS
$ flexnbd read --addr <ADDR> --port <PORT> --from <OFFSET>
--size <SIZE> [--bind BIND-ADDR] [global option]*
--sock, -s SOCK
The control socket of the server of interest. Required.
READ MODE
Connect to the server at ADDR:PORT, and read SIZE bytes starting at
OFFSET in a single NBD query. The returned data will be echoed to
STDOUT. In case of a remote ACL, set the local source address to
BIND-ADDR.
OFFSET in a single NBD query.
Options
^^^^^^^
$ flexnbd read --addr ADDR --port PORT --from OFFSET --size SIZE
[--bind BIND_ADDR] [global_option]*
*--addr, -l ADDR*:
The address of the remote server. Required.
The returned data will be echoed to STDOUT. In case of a remote ACL,
set the local source address to BIND_ADDR.
*--port, -p PORT*:
The port of the remote server. Required.
OPTIONS
*--from, -F OFFSET*:
The byte offset to start reading from. Required. Maximum 2^62.
--addr, -l ADDR
The address of the remote server. Required.
*--size, -S SIZE*:
The number of bytes to read. Required. Maximum 2^30.
--port, -p PORT
The port of the remote server. Required.
*--bind, -b BIND-ADDR*:
The local address to bind to. You may need this if the remote server
is using an access control list.
--from, -F OFFSET
The byte offset to start reading from. Required. Maximum 2^62.
write
~~~~~
--size, -S SIZE
The number of bytes to read. Required. Maximum 2^30.
$ cat ... | flexnbd write --addr <ADDR> --port <PORT> --from <OFFSET>
--size <SIZE> [--bind BIND-ADDR] [global option]*
--bind, -b BIND_ADDR
The local address to bind to. You may need this if the remote
server is using an access control list.
WRITE MODE
Connect to the server at ADDR:PORT, and write SIZE bytes from STDIN
starting at OFFSET in a single NBD query. In case of a remote ACL, set
the local source address to BIND-ADDR.
starting at OFFSET in a single NBD query.
Options
^^^^^^^
$ cat ... | flexnbd write --addr ADDR --port PORT --from OFFSET
--size SIZE [--bind BIND_ADDR] [global_option]*
*--addr, -l ADDR*:
The address of the remote server. Required.
In case of a remote ACL, set the local source address to BIND_ADDR.
*--port, -p PORT*:
The port of the remote server. Required.
OPTIONS
*--from, -F OFFSET*:
The byte offset to start writing from. Required. Maximum 2^62.
--addr, -l ADDR
The address of the remote server. Required.
*--size, -S SIZE*:
The number of bytes to write. Required. Maximum 2^30.
--port, -p PORT
The port of the remote server. Required.
*--bind, -b BIND-ADDR*:
The local address to bind to. You may need this if the remote server
is using an access control list.
--from, -F OFFSET
The byte offset to start writing from. Required. Maximum 2^62.
help
~~~~
--size, -S SIZE
The number of bytes to write. Required. Maximum 2^30.
$ flexnbd help [command] [global option]*
--bind, -b BIND_ADDR
The local address to bind to. You may need this if the remote
server is using an access control list.
Without 'command', show the list of available commands. With 'command',
show help for that command.
HELP MODE
$ flexnbd help [mode] [global_option]*
Without mode, show the list of available modes. With mode, show help for that mode.
GLOBAL OPTIONS
--------------
*--help, -h* :
Show command or global help.
--help, -h Show mode or global help.
*--verbose, -v* :
Output all available log information to STDERR.
*--quiet, -q* :
Output as little log information as possible to STDERR.
--verbose, -v Output all available log information to STDERR.
--quiet, -q Output as little log information as possible to STDERR.
LOGGING
-------
Log output is sent to STDERR. If --quiet is set, no output will be seen
unless the program termintes abnormally. If neither --quiet nor
Log output is sent to STDERR. If --quiet is set, no output will be
seen unless the program termintes abnormally. If neither --quiet nor
--verbose are set, no output will be seen unless something goes wrong
with a specific request. If --verbose is given, every available log
message will be seen (which, for a debug build, is many). It is not an
error to set both --verbose and --quiet. The last one wins.
with a specific request. If --verbose is given, every available log
message will be seen (which, for a debug build, is many). It is not an
error to set both --verbose and --quiet. The last one wins.
The log line format is:
<TIMESTAMP>:<LEVEL>:<PID> <THREAD> <SOURCEFILE>:<SOURCELINE>: <MSG>
<TIMESTAMP>:<LEVEL>:<PID> <THREAD> <SOURCEFILE:SOURCELINE>: <MSG>
*TIMESTAMP*:
Time the log entry was made. This is expressed in terms of monotonic ms.
<TIMESTAMP>
Time the log entry was made. This is expressed in terms of monotonic
ms.
*LEVEL*:
<LEVEL>
This will be one of 'D', 'I', 'W', 'E', 'F' in increasing order of
severity. If flexnbd is started with the --quiet flag, only 'F' will be
seen. If it is started with the --verbose flag, any from 'I' upwards
will be seen. Only if you have a debug build and start it with
--verbose will you see 'D' entries.
severity. If flexnbd is started with the --quiet flag, only 'F'
will be seen. If it is started with the --verbose flag, any from 'I'
upwards will be seen. Only if you have a debug build and start it
with --verbose will you see 'D' entries.
*PID*:
<PID>
This is the process ID.
*THREAD*:
There are several pthreads per flexnbd process: a main thread, a serve
thread, a thread per client, and possibly a pair of mirror threads and a
control thread. This field identifies which thread was responsible for
the log line.
<THREAD>
There are several pthreads per flexnbd process: a main thread, a
serve thread, a thread per client, and possibly a pair of mirror
threads and a control thread. This field identifies which thread was
responsible for the log line.
*SOURCEFILE:SOURCELINE*:
<SOURCEFILE:SOURCELINE>
Identifies where in the source code this log line can be found.
*MSG*:
<MSG>
A short message describing what's happening, how it's being done, or
if you're very lucky *why* it's going on.
if you're very lucky why it's going on.
EXAMPLES
--------
Serving a file
~~~~~~~~~~~~~~
SERVING A FILE
The simplest case is serving a file on the default nbd port:
@@ -326,8 +331,7 @@ The simplest case is serving a file on the default nbd port:
root:x:
$
Reading server status
~~~~~~~~~~~~~~~~~~~~~
READING SERVER STATUS
In order to read a server's status, we need it to open a control socket.
@@ -335,13 +339,12 @@ In order to read a server's status, we need it to open a control socket.
--sock /tmp/flexnbd.sock
$ flexnbd status --sock /tmp/flexnbd.sock
pid=9635 is_mirroring=false has_control=true
$
Note that the status output is newline-terminated.
Migrating
~~~~~~~~~
MIGRATING
To migrate, we need to provide a destination file of the right size.
@@ -367,8 +370,8 @@ With this knowledge in hand, we can start the migration:
$ flexnbd mirror --addr 127.0.0.1 --port 4779 \
--sock /tmp/flex-source.sock
Migration started
[1] + 9648 done build/flexnbd serve --addr 0.0.0.0 --port 4778
[2] + 9651 done build/flexnbd listen --addr 0.0.0.0 --port 4779
[1] + 9648 done flexnbd serve --addr 0.0.0.0 --port 4778
[2] + 9651 done flexnbd listen --addr 0.0.0.0 --port 4779
$
Note that because the file is so small in this case, we see the source
@@ -376,21 +379,25 @@ server quit soon after we start the migration, and the destination
exited at roughly the same time.
BUGS
----
Should be reported to alex@bytemark.co.uk.
Should be reported on GitHub at
* https://github.com/BytemarkHosting/flexnbd-c/issues
AUTHOR
------
Written by Alex Young <alex@bytemark.co.uk>.
Originally written by Alex Young <alex@blackkettle.org>.
Original concept and core code by Matthew Bloch <matthew@bytemark.co.uk>.
Some additions by Nick Thomas <nick@bytemark.co.uk>
Proxy mode written by Nick Thomas <me@ur.gs>.
COPYING
-------
The full commit history is available on GitHub.
Copyright (c) 2012 Bytemark Hosting Ltd. Free use of this software is
granted under the terms of the GNU General Public License version 3 or
later.
SEE ALSO
flexnbd-proxy(1), nbd-client(8), xnbd-server(8), xnbd-client(8)
COPYRIGHT
Copyright (c) 2012-2016 Bytemark Hosting Ltd. Free use of this
software is granted under the terms of the GNU General Public License
version 3 or later.

View File

@@ -1,60 +0,0 @@
# encoding: utf-8
def make(*targets)
sh "make #{targets.map{|t| t.to_s}.join(" ")}"
end
def maketask( opts )
case opts
when Symbol
maketask opts => opts
else
opts.each do |name, targets|
task( name ){make *[*targets]}
end
end
end
desc "Build the binary and man page"
maketask :build => [:all, :doc]
desc "Build just the flexnbd binary"
maketask :flexnbd => [:server]
file "build/flexnbd" => :flexnbd
desc "Build just the flexnbd-proxy binary"
maketask :flexnbd_proxy => [:proxy]
file "build/flexnbd-proxy" => :flexnbd_proxy
desc "Build just the man page"
maketask :man => :doc
namespace "test" do
desc "Run all tests"
task 'run' => ["unit", "scenarios"]
desc "Build C tests"
maketask :build => :check_bins
desc "Run C tests"
maketask :unit => :check
desc "Run NBD test scenarios"
task 'scenarios' => ["build/flexnbd", "build/flexnbd-proxy"] do
sh "cd tests/acceptance && RUBYOPT='-I.' ruby nbd_scenarios -v"
end
end
desc "Remove all build targets, binaries and temporary files"
maketask :clean
file "debian/changelog" do
FileUtils.mkdir_p "debian"
sh "hg log --style=changelog.template > debian/changelog"
end
desc "Generate the changelog"
task :changelog => "debian/changelog"

101
debian/changelog vendored
View File

@@ -1,3 +1,104 @@
flexnbd (0.2.1) UNRELEASED; urgency=medium
* Force a msync after every write, ignoring FUA flag, or lack thereof.
-- Patrick J Cherry <patrick@bytemark.co.uk> Tue, 24 Apr 2018 10:27:12 +0100
flexnbd (0.2.0) stable; urgency=medium
[ James Carter ]
* Set TCP keepalive on sockets so broken connections are reaped (#33, !33,
!36)
* Add a context to logs to make debugging problems easier (#34, !34)
[ Chris Cottam ]
* Increased NBD_MAX_SIZE from 1MB to 32MB for qemu 2.11 (!35)
[ Patrick J Cherry ]
* Added FLUSH and FUA support (!38)
* Server returns ENOSPC in response to writes beyond the end of the
filesystem, and EINVAL to unknown commands. (#36, !40)
* Proxy passes all NBD protocol errors through to the client instead of
disconnecting and retrying (#36, !40)
* Fix struct types in readwrite.c (#35, !41)
* Ignore ends of discs that stray outside of 512-byte sector sizes (!42).
* Tweak logging for readloop failures (!44)
* Alter semantics of NBD_MAX_BLOCK_SIZE to remove struct size overheads when
calculating if a request exceeds the max block size (!45)
* Added tests for setting TCP_NODELAY on upstream-reconnections in the
proxy, and refactored the other LD_PRELOAD tests (!43)
* Clean up dead threads before calculating the number of connected clients
on the status command (!46)
-- Patrick J Cherry <patrick@bytemark.co.uk> Tue, 20 Feb 2018 11:43:22 +0000
flexnbd (0.1.7) stable; urgency=medium
* Return bytes_left in migration statistics.
-- Chris Elsworth <chris.elsworth@bytemark.co.uk> Fri, 14 Jul 2017 17:00:38 +0100
flexnbd (0.1.6) stable; urgency=medium
* Remove lots of per-cpu compiler flags, notably march=native.
-- Patrick J Cherry <patrick@bytemark.co.uk> Thu, 13 Apr 2017 12:48:23 +0100
flexnbd (0.1.5) stable; urgency=medium
* Switched to native packaging.
-- Patrick J Cherry <patrick@bytemark.co.uk> Mon, 23 Jan 2017 13:51:58 +0000
flexnbd (0.1.5-1) stable; urgency=medium
* This fixes the compiler warning pointer-to-int-cast in serve.c.
* Fix up "wrong" handle type from char* to uint64_t in check_readwrite test.
* fix check_bitset test on 32-bit platforms.
* Skip large file test on 32-bit platforms.
-- Patrick J Cherry <patrick@bytemark.co.uk> Fri, 07 Oct 2016 12:25:36 +0100
flexnbd (0.1.4-1) stable; urgency=medium
[ Michel Pollet ]
* Simplified Makefile
* Makefile: don't include *.d's before 'all'
[ Patrick J Cherry ]
* Added net-tools to the build-deps for testing
* Fixed "double-definition of constants" warning
* Use a BINARY encoded string when doing read/write comparisons.
* Moved all tasks from Rake to Make
* Added dpkg-dev to build-deps
* Updated manpages, replace a2x with txt2man
* Fixed up internal test names (copy/pasta?)
* Fixed check_readwrite test to pass correct handle to fd_write_reply
* Fix up nbdtypes test to correctly use htobe64
* Update Makefile to specify dependencies properly for tests
[ James F. Carter ]
* update tests to reflect changes in handle storage
* run tests in gitlab-ci
* force binary encoding in a ruby2.1-compatible way
* add missing deps on server object files when building check binaries
* remove unnecessary step in gitlab-ci
* failures in make check now result in an error
-- Patrick J Cherry <patrick@bytemark.co.uk> Thu, 06 Oct 2016 15:49:33 +0100
flexnbd (0.1.3-1) stable; urgency=medium
[ Michel Pollet ]
* Close socket fix, might relate to migration crashing
* Attempt at fixing bind() bug
* Simplified NBD handle comparisons
[ Patrick J Cherry ]
* Non-maintainer upload.
-- Patrick J Cherry <patrick@bytemark.co.uk> Tue, 04 Oct 2016 21:20:27 +0100
flexnbd (0.1.2-1) stable; urgency=medium
* Non-maintainer upload.

3
debian/control vendored
View File

@@ -2,8 +2,7 @@ Source: flexnbd
Section: web
Priority: extra
Maintainer: Patrick J Cherry <patrick@bytemark.co.uk>
Build-Depends: debhelper (>= 7.0.50), ruby, rake, gcc, libev-dev,
asciidoc, libxml2-utils, xsltproc, xmlto, check
Build-Depends: debhelper (>= 7.0.50), ruby, gcc, libev-dev, txt2man, check, net-tools, libsubunit-dev, ruby-test-unit
Standards-Version: 3.8.1
Homepage: https://github.com/BytemarkHosting/flexnbd-c

View File

@@ -1,5 +1,3 @@
build/flexnbd usr/bin
build/flexnbd-proxy usr/bin
build/flexnbd.1.gz usr/share/man/man1
build/flexnbd-proxy.1.gz usr/share/man/man1

2
debian/flexnbd.manpages vendored Normal file
View File

@@ -0,0 +1,2 @@
build/flexnbd.1.gz
build/flexnbd-proxy.1.gz

View File

@@ -1 +1 @@
3.0 (quilt)
3.0 (native)

View File

@@ -13,219 +13,233 @@
#include "ioutil.h"
int build_allocation_map(struct bitset * allocation_map, int fd)
int build_allocation_map(struct bitset *allocation_map, int fd)
{
/* break blocking ioctls down */
const unsigned long max_length = 100*1024*1024;
const unsigned int max_extents = 1000;
/* break blocking ioctls down */
const unsigned long max_length = 100 * 1024 * 1024;
const unsigned int max_extents = 1000;
unsigned long offset = 0;
unsigned long offset = 0;
struct {
struct fiemap fiemap;
struct fiemap_extent extents[max_extents];
} fiemap_static;
struct fiemap* fiemap = (struct fiemap*) &fiemap_static;
struct {
struct fiemap fiemap;
struct fiemap_extent extents[max_extents];
} fiemap_static;
struct fiemap *fiemap = (struct fiemap *) &fiemap_static;
memset(&fiemap_static, 0, sizeof(fiemap_static));
memset(&fiemap_static, 0, sizeof(fiemap_static));
for (offset = 0; offset < allocation_map->size; ) {
for (offset = 0; offset < allocation_map->size;) {
fiemap->fm_start = offset;
fiemap->fm_start = offset;
fiemap->fm_length = max_length;
if ( offset + max_length > allocation_map->size ) {
fiemap->fm_length = allocation_map->size-offset;
}
fiemap->fm_flags = FIEMAP_FLAG_SYNC;
fiemap->fm_extent_count = max_extents;
fiemap->fm_mapped_extents = 0;
if ( ioctl( fd, FS_IOC_FIEMAP, fiemap ) < 0 ) {
debug( "Couldn't get fiemap, returning no allocation_map" );
return 0; /* it's up to the caller to free the map */
}
else {
for ( unsigned int i = 0; i < fiemap->fm_mapped_extents; i++ ) {
bitset_set_range( allocation_map,
fiemap->fm_extents[i].fe_logical,
fiemap->fm_extents[i].fe_length );
}
/* must move the offset on, but careful not to jump max_length
* if we've actually hit max_offsets.
*/
if (fiemap->fm_mapped_extents > 0) {
struct fiemap_extent *last = &fiemap->fm_extents[
fiemap->fm_mapped_extents-1
];
offset = last->fe_logical + last->fe_length;
}
else {
offset += fiemap->fm_length;
}
}
fiemap->fm_length = max_length;
if (offset + max_length > allocation_map->size) {
fiemap->fm_length = allocation_map->size - offset;
}
info("Successfully built allocation map");
return 1;
fiemap->fm_flags = FIEMAP_FLAG_SYNC;
fiemap->fm_extent_count = max_extents;
fiemap->fm_mapped_extents = 0;
if (ioctl(fd, FS_IOC_FIEMAP, fiemap) < 0) {
debug("Couldn't get fiemap, returning no allocation_map");
return 0; /* it's up to the caller to free the map */
} else {
for (unsigned int i = 0; i < fiemap->fm_mapped_extents; i++) {
bitset_set_range(allocation_map,
fiemap->fm_extents[i].fe_logical,
fiemap->fm_extents[i].fe_length);
}
/* must move the offset on, but careful not to jump max_length
* if we've actually hit max_offsets.
*/
if (fiemap->fm_mapped_extents > 0) {
struct fiemap_extent *last =
&fiemap->fm_extents[fiemap->fm_mapped_extents - 1];
offset = last->fe_logical + last->fe_length;
} else {
offset += fiemap->fm_length;
}
}
}
info("Successfully built allocation map");
return 1;
}
int open_and_mmap(const char* filename, int* out_fd, uint64_t *out_size, void **out_map)
int open_and_mmap(const char *filename, int *out_fd, uint64_t * out_size,
void **out_map)
{
/*
* size and out_size are intentionally of different types.
* lseek64() uses off64_t to signal errors in the sign bit.
* Since we check for these errors before trying to assign to
* *out_size, we know *out_size can never go negative.
*/
off64_t size;
/*
* size and out_size are intentionally of different types.
* lseek64() uses off64_t to signal errors in the sign bit.
* Since we check for these errors before trying to assign to
* *out_size, we know *out_size can never go negative.
*/
off64_t size;
/* O_DIRECT should not be used with mmap() */
*out_fd = open(filename, O_RDWR | O_SYNC );
/* O_DIRECT should not be used with mmap() */
*out_fd = open(filename, O_RDWR | O_NOATIME);
if (*out_fd < 1) {
warn("open(%s) failed: does it exist?", filename);
return *out_fd;
}
if (*out_fd < 1) {
warn("open(%s) failed: does it exist?", filename);
return *out_fd;
}
size = lseek64(*out_fd, 0, SEEK_END);
if (size < 0) {
warn("lseek64() failed");
return size;
}
if (out_size) {
*out_size = size;
}
size = lseek64(*out_fd, 0, SEEK_END);
if (size < 0) {
warn("lseek64() failed");
return size;
}
if (out_map) {
*out_map = mmap64(NULL, size, PROT_READ|PROT_WRITE, MAP_SHARED,
*out_fd, 0);
if (((long) *out_map) == -1) {
warn("mmap64() failed");
return -1;
}
debug("opened %s size %ld on fd %d @ %p", filename, size, *out_fd, *out_map);
}
else {
debug("opened %s size %ld on fd %d", filename, size, *out_fd);
}
/* If discs are not in multiples of 512, then odd things happen,
* resulting in reads/writes past the ends of files.
*/
if (size != (size & (~0x1ff))) {
warn("file does not fit into 512-byte sectors; the end of the file will be ignored.");
size &= ~0x1ff;
}
return 0;
if (out_size) {
*out_size = size;
}
if (out_map) {
*out_map = mmap64(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED,
*out_fd, 0);
if (((long) *out_map) == -1) {
warn("mmap64() failed");
return -1;
}
debug("opened %s size %ld on fd %d @ %p", filename, size, *out_fd,
*out_map);
} else {
debug("opened %s size %ld on fd %d", filename, size, *out_fd);
}
return 0;
}
int writeloop(int filedes, const void *buffer, size_t size)
{
size_t written=0;
while (written < size) {
ssize_t result = write(filedes, buffer+written, size-written);
if (result == -1) {
if ( errno == EINTR || errno == EAGAIN || errno == EWOULDBLOCK ) {
continue; // busy-wait
}
return -1; // failure
}
written += result;
size_t written = 0;
while (written < size) {
ssize_t result = write(filedes, buffer + written, size - written);
if (result == -1) {
if (errno == EINTR || errno == EAGAIN || errno == EWOULDBLOCK) {
continue; // busy-wait
}
return -1; // failure
}
return 0;
written += result;
}
return 0;
}
int readloop(int filedes, void *buffer, size_t size)
{
size_t readden=0;
while (readden < size) {
ssize_t result = read(filedes, buffer+readden, size-readden);
size_t readden = 0;
while (readden < size) {
ssize_t result = read(filedes, buffer + readden, size - readden);
if ( result == 0 /* EOF */ ) {
warn( "end-of-file detected while reading after %i bytes", readden );
return -1;
}
if ( result == -1 ) {
if ( errno == EINTR || errno == EAGAIN || errno == EWOULDBLOCK ) {
continue; // busy-wait
}
return -1; // failure
}
readden += result;
if (result == 0 /* EOF */ ) {
return -1;
}
return 0;
if (result == -1) {
if (errno == EINTR || errno == EAGAIN || errno == EWOULDBLOCK) {
continue; // busy-wait
}
return -1; // failure
}
readden += result;
}
return 0;
}
int sendfileloop(int out_fd, int in_fd, off64_t *offset, size_t count)
int sendfileloop(int out_fd, int in_fd, off64_t * offset, size_t count)
{
size_t sent=0;
while (sent < count) {
ssize_t result = sendfile64(out_fd, in_fd, offset, count-sent);
debug("sendfile64(out_fd=%d, in_fd=%d, offset=%p, count-sent=%ld) = %ld", out_fd, in_fd, offset, count-sent, result);
size_t sent = 0;
while (sent < count) {
ssize_t result = sendfile64(out_fd, in_fd, offset, count - sent);
debug
("sendfile64(out_fd=%d, in_fd=%d, offset=%p, count-sent=%ld) = %ld",
out_fd, in_fd, offset, count - sent, result);
if (result == -1) {
debug( "%s (%i) calling sendfile64()", strerror(errno), errno );
return -1;
}
sent += result;
debug("sent=%ld, count=%ld", sent, count);
if (result == -1) {
debug("%s (%i) calling sendfile64()", strerror(errno), errno);
return -1;
}
debug("exiting sendfileloop");
return 0;
sent += result;
debug("sent=%ld, count=%ld", sent, count);
}
debug("exiting sendfileloop");
return 0;
}
#include <errno.h>
ssize_t spliceloop(int fd_in, loff_t *off_in, int fd_out, loff_t *off_out, size_t len, unsigned int flags2)
ssize_t spliceloop(int fd_in, loff_t * off_in, int fd_out,
loff_t * off_out, size_t len, unsigned int flags2)
{
const unsigned int flags = SPLICE_F_MORE|SPLICE_F_MOVE|flags2;
size_t spliced=0;
const unsigned int flags = SPLICE_F_MORE | SPLICE_F_MOVE | flags2;
size_t spliced = 0;
//debug("spliceloop(%d, %ld, %d, %ld, %ld)", fd_in, off_in ? *off_in : 0, fd_out, off_out ? *off_out : 0, len);
//debug("spliceloop(%d, %ld, %d, %ld, %ld)", fd_in, off_in ? *off_in : 0, fd_out, off_out ? *off_out : 0, len);
while (spliced < len) {
ssize_t result = splice(fd_in, off_in, fd_out, off_out, len, flags);
if (result < 0) {
//debug("result=%ld (%s), spliced=%ld, len=%ld", result, strerror(errno), spliced, len);
if (errno == EAGAIN && (flags & SPLICE_F_NONBLOCK) ) {
return spliced;
}
else {
return -1;
}
} else {
spliced += result;
//debug("result=%ld (%s), spliced=%ld, len=%ld", result, strerror(errno), spliced, len);
}
while (spliced < len) {
ssize_t result =
splice(fd_in, off_in, fd_out, off_out, len, flags);
if (result < 0) {
//debug("result=%ld (%s), spliced=%ld, len=%ld", result, strerror(errno), spliced, len);
if (errno == EAGAIN && (flags & SPLICE_F_NONBLOCK)) {
return spliced;
} else {
return -1;
}
} else {
spliced += result;
//debug("result=%ld (%s), spliced=%ld, len=%ld", result, strerror(errno), spliced, len);
}
}
return spliced;
return spliced;
}
int splice_via_pipe_loop(int fd_in, int fd_out, size_t len)
{
int pipefd[2]; /* read end, write end */
size_t spliced=0;
int pipefd[2]; /* read end, write end */
size_t spliced = 0;
if (pipe(pipefd) == -1) {
return -1;
if (pipe(pipefd) == -1) {
return -1;
}
while (spliced < len) {
ssize_t run = len - spliced;
ssize_t s2, s1 = spliceloop(fd_in, NULL, pipefd[1], NULL, run,
SPLICE_F_NONBLOCK);
/*if (run > 65535)
run = 65535; */
if (s1 < 0) {
break;
}
while (spliced < len) {
ssize_t run = len-spliced;
ssize_t s2, s1 = spliceloop(fd_in, NULL, pipefd[1], NULL, run, SPLICE_F_NONBLOCK);
/*if (run > 65535)
run = 65535;*/
if (s1 < 0) { break; }
s2 = spliceloop(pipefd[0], NULL, fd_out, NULL, s1, 0);
if (s2 < 0) { break; }
spliced += s2;
s2 = spliceloop(pipefd[0], NULL, fd_out, NULL, s1, 0);
if (s2 < 0) {
break;
}
close(pipefd[0]);
close(pipefd[1]);
spliced += s2;
}
close(pipefd[0]);
close(pipefd[1]);
return spliced < len ? -1 : 0;
return spliced < len ? -1 : 0;
}
/* Reads single bytes from fd until either an EOF or a newline appears.
@@ -235,117 +249,123 @@ int splice_via_pipe_loop(int fd_in, int fd_out, size_t len)
* Returns the number of read bytes: the length of the line without the
* newline, plus the trailing null.
*/
int read_until_newline(int fd, char* buf, int bufsize)
int read_until_newline(int fd, char *buf, int bufsize)
{
int cur;
int cur;
for (cur=0; cur < bufsize; cur++) {
int result = read(fd, buf+cur, 1);
if (result <= 0) { return -1; }
if (buf[cur] == 10) {
buf[cur] = '\0';
break;
}
for (cur = 0; cur < bufsize; cur++) {
int result = read(fd, buf + cur, 1);
if (result <= 0) {
return -1;
}
if (buf[cur] == 10) {
buf[cur] = '\0';
break;
}
}
return cur+1;
return cur + 1;
}
int read_lines_until_blankline(int fd, int max_line_length, char ***lines)
{
int lines_count = 0;
char line[max_line_length+1];
*lines = NULL;
int lines_count = 0;
char line[max_line_length + 1];
*lines = NULL;
memset(line, 0, max_line_length+1);
memset(line, 0, max_line_length + 1);
while (1) {
int readden = read_until_newline(fd, line, max_line_length);
/* readden will be:
* 1 for an empty line
* -1 for an eof
* -1 for a read error
*/
if (readden <= 1) { return lines_count; }
*lines = xrealloc(*lines, (lines_count+1) * sizeof(char*));
(*lines)[lines_count] = strdup(line);
if ((*lines)[lines_count][0] == 0) {
return lines_count;
}
lines_count++;
while (1) {
int readden = read_until_newline(fd, line, max_line_length);
/* readden will be:
* 1 for an empty line
* -1 for an eof
* -1 for a read error
*/
if (readden <= 1) {
return lines_count;
}
*lines = xrealloc(*lines, (lines_count + 1) * sizeof(char *));
(*lines)[lines_count] = strdup(line);
if ((*lines)[lines_count][0] == 0) {
return lines_count;
}
lines_count++;
}
}
int fd_is_closed( int fd_in )
int fd_is_closed(int fd_in)
{
int errno_old = errno;
int result = fcntl( fd_in, F_GETFL ) < 0;
errno = errno_old;
return result;
int errno_old = errno;
int result = fcntl(fd_in, F_GETFL) < 0;
errno = errno_old;
return result;
}
static inline int io_errno_permanent(void)
{
return ( errno != EAGAIN && errno != EWOULDBLOCK && errno != EINTR );
return (errno != EAGAIN && errno != EWOULDBLOCK && errno != EINTR);
}
/* Returns -1 if the operation failed, or the number of bytes read if all is
* well. Note that 0 bytes may be returned. Unlike read(), this is not an EOF! */
ssize_t iobuf_read(int fd, struct iobuf *iobuf, size_t default_size )
ssize_t iobuf_read(int fd, struct iobuf * iobuf, size_t default_size)
{
size_t left;
ssize_t count;
size_t left;
ssize_t count;
if ( iobuf->needle == 0 ) {
iobuf->size = default_size;
}
if (iobuf->needle == 0) {
iobuf->size = default_size;
}
left = iobuf->size - iobuf->needle;
debug( "Reading %"PRIu32" of %"PRIu32" bytes from fd %i", left, iobuf->size, fd );
left = iobuf->size - iobuf->needle;
debug("Reading %" PRIu32 " of %" PRIu32 " bytes from fd %i", left,
iobuf->size, fd);
count = read( fd, iobuf->buf + iobuf->needle, left );
count = read(fd, iobuf->buf + iobuf->needle, left);
if ( count > 0 ) {
iobuf->needle += count;
debug( "read() returned %"PRIu32" bytes", count );
} else if ( count == 0 ) {
warn( "read() returned EOF on fd %i", fd );
errno = 0;
return -1;
} else if ( count == -1 ) {
if ( io_errno_permanent() ) {
warn( SHOW_ERRNO( "read() failed on fd %i", fd ) );
} else {
debug( SHOW_ERRNO( "read() returned 0 bytes" ) );
count = 0;
}
}
return count;
}
ssize_t iobuf_write( int fd, struct iobuf *iobuf )
{
size_t left = iobuf->size - iobuf->needle;
ssize_t count;
debug( "Writing %"PRIu32" of %"PRIu32" bytes to fd %i", left, iobuf->size, fd );
count = write( fd, iobuf->buf + iobuf->needle, left );
if ( count >= 0 ) {
iobuf->needle += count;
debug( "write() returned %"PRIu32" bytes", count );
if (count > 0) {
iobuf->needle += count;
debug("read() returned %" PRIu32 " bytes", count);
} else if (count == 0) {
warn("read() returned EOF on fd %i", fd);
errno = 0;
return -1;
} else if (count == -1) {
if (io_errno_permanent()) {
warn(SHOW_ERRNO("read() failed on fd %i", fd));
} else {
if ( io_errno_permanent() ) {
warn( SHOW_ERRNO( "write() failed on fd %i", fd ) );
} else {
debug( SHOW_ERRNO( "write() returned 0 bytes" ) );
count = 0;
}
debug(SHOW_ERRNO("read() returned 0 bytes"));
count = 0;
}
}
return count;
return count;
}
ssize_t iobuf_write(int fd, struct iobuf * iobuf)
{
size_t left = iobuf->size - iobuf->needle;
ssize_t count;
debug("Writing %" PRIu32 " of %" PRIu32 " bytes to fd %i", left,
iobuf->size, fd);
count = write(fd, iobuf->buf + iobuf->needle, left);
if (count >= 0) {
iobuf->needle += count;
debug("write() returned %" PRIu32 " bytes", count);
} else {
if (io_errno_permanent()) {
warn(SHOW_ERRNO("write() failed on fd %i", fd));
} else {
debug(SHOW_ERRNO("write() returned 0 bytes"));
count = 0;
}
}
return count;
}

View File

@@ -3,16 +3,16 @@
#include <sys/types.h>
struct iobuf {
unsigned char *buf;
size_t size;
size_t needle;
unsigned char *buf;
size_t size;
size_t needle;
};
ssize_t iobuf_read( int fd, struct iobuf* iobuf, size_t default_size );
ssize_t iobuf_write( int fd, struct iobuf* iobuf );
ssize_t iobuf_read(int fd, struct iobuf *iobuf, size_t default_size);
ssize_t iobuf_write(int fd, struct iobuf *iobuf);
#include "serve.h"
struct bitset; /* don't need whole of bitset.h here */
struct bitset; /* don't need whole of bitset.h here */
/** Scan the file opened in ''fd'', set bits in ''allocation_map'' that
* correspond to which blocks are physically allocated on disc (or part-
@@ -20,7 +20,7 @@ struct bitset; /* don't need whole of bitset.h here */
* than you've asked for, any block or part block will count as "allocated"
* with the corresponding bit set. Returns 1 if successful, 0 otherwise.
*/
int build_allocation_map(struct bitset * allocation_map, int fd);
int build_allocation_map(struct bitset *allocation_map, int fd);
/** Repeat a write() operation that succeeds partially until ''size'' bytes
* are written, or an error is returned, when it returns -1 as usual.
@@ -35,10 +35,11 @@ int readloop(int filedes, void *buffer, size_t size);
/** Repeat a sendfile() operation that succeeds partially until ''size'' bytes
* are written, or an error is returned, when it returns -1 as usual.
*/
int sendfileloop(int out_fd, int in_fd, off64_t *offset, size_t count);
int sendfileloop(int out_fd, int in_fd, off64_t * offset, size_t count);
/** Repeat a splice() operation until we have 'len' bytes. */
ssize_t spliceloop(int fd_in, loff_t *off_in, int fd_out, loff_t *off_out, size_t len, unsigned int flags2);
ssize_t spliceloop(int fd_in, loff_t * off_in, int fd_out,
loff_t * off_out, size_t len, unsigned int flags2);
/** Copy ''len'' bytes from ''fd_in'' to ''fd_out'' by creating a temporary
* pipe and using the Linux splice call repeatedly until it has transferred
@@ -50,7 +51,7 @@ int splice_via_pipe_loop(int fd_in, int fd_out, size_t len);
* until an LF character is received, which is written to the buffer at a zero
* byte. Returns -1 on error, or the number of bytes written to the buffer.
*/
int read_until_newline(int fd, char* buf, int bufsize);
int read_until_newline(int fd, char *buf, int bufsize);
/** Read a number of lines using read_until_newline, until an empty line is
* received (i.e. the sequence LF LF). The data is read from ''fd'' and
@@ -65,12 +66,12 @@ int read_lines_until_blankline(int fd, int max_line_length, char ***lines);
* ''out_size'' and the address of the mmap in ''out_map''. If anything goes
* wrong, returns -1 setting errno, otherwise 0.
*/
int open_and_mmap( const char* filename, int* out_fd, uint64_t* out_size, void **out_map);
int open_and_mmap(const char *filename, int *out_fd, uint64_t * out_size,
void **out_map);
/** Check to see whether the given file descriptor is closed.
*/
int fd_is_closed( int fd_in );
int fd_is_closed(int fd_in);
#endif

View File

@@ -2,7 +2,7 @@
#define MODE_H
void mode(char* mode, int argc, char **argv);
void mode(char *mode, int argc, char **argv);
#include <getopt.h>
@@ -68,9 +68,9 @@ void mode(char* mode, int argc, char **argv);
"\t--" OPT_VERBOSE ",-" SOPT_VERBOSE "\t\tOutput debug information.\n"
#ifdef DEBUG
# define VERBOSE_LOG_LEVEL 0
#define VERBOSE_LOG_LEVEL 0
#else
# define VERBOSE_LOG_LEVEL 1
#define VERBOSE_LOG_LEVEL 1
#endif
#define QUIET_LOG_LEVEL 4
@@ -91,7 +91,6 @@ void mode(char* mode, int argc, char **argv);
#define MAX_SPEED_LINE \
"\t--" OPT_MAX_SPEED ",-m <bps>\tMaximum speed of the migration, in bytes/sec.\n"
char * help_help_text;
char *help_help_text;
#endif

View File

@@ -8,51 +8,54 @@
* We intentionally ignore the reserved 128 bytes at the end of the
* request, since there's nothing we can do with them.
*/
void nbd_r2h_init( struct nbd_init_raw * from, struct nbd_init * to )
void nbd_r2h_init(struct nbd_init_raw *from, struct nbd_init *to)
{
memcpy( to->passwd, from->passwd, 8 );
to->magic = be64toh( from->magic );
to->size = be64toh( from->size );
memcpy(to->passwd, from->passwd, 8);
to->magic = be64toh(from->magic);
to->size = be64toh(from->size);
to->flags = be32toh(from->flags);
}
void nbd_h2r_init( struct nbd_init * from, struct nbd_init_raw * to)
void nbd_h2r_init(struct nbd_init *from, struct nbd_init_raw *to)
{
memcpy( to->passwd, from->passwd, 8 );
to->magic = htobe64( from->magic );
to->size = htobe64( from->size );
memcpy(to->passwd, from->passwd, 8);
to->magic = htobe64(from->magic);
to->size = htobe64(from->size);
to->flags = htobe32(from->flags);
}
void nbd_r2h_request( struct nbd_request_raw *from, struct nbd_request * to )
void nbd_r2h_request(struct nbd_request_raw *from, struct nbd_request *to)
{
to->magic = htobe32( from->magic );
to->type = htobe32( from->type );
memcpy( to->handle, from->handle, 8 );
to->from = htobe64( from->from );
to->len = htobe32( from->len );
to->magic = be32toh(from->magic);
to->flags = be16toh(from->flags);
to->type = be16toh(from->type);
to->handle.w = from->handle.w;
to->from = be64toh(from->from);
to->len = be32toh(from->len);
}
void nbd_h2r_request( struct nbd_request * from, struct nbd_request_raw * to )
void nbd_h2r_request(struct nbd_request *from, struct nbd_request_raw *to)
{
to->magic = be32toh( from->magic );
to->type = be32toh( from->type );
memcpy( to->handle, from->handle, 8 );
to->from = be64toh( from->from );
to->len = be32toh( from->len );
to->magic = htobe32(from->magic);
to->flags = htobe16(from->flags);
to->type = htobe16(from->type);
to->handle.w = from->handle.w;
to->from = htobe64(from->from);
to->len = htobe32(from->len);
}
void nbd_r2h_reply( struct nbd_reply_raw * from, struct nbd_reply * to )
void nbd_r2h_reply(struct nbd_reply_raw *from, struct nbd_reply *to)
{
to->magic = htobe32( from->magic );
to->error = htobe32( from->error );
memcpy( to->handle, from->handle, 8 );
to->magic = be32toh(from->magic);
to->error = be32toh(from->error);
to->handle.w = from->handle.w;
}
void nbd_h2r_reply( struct nbd_reply * from, struct nbd_reply_raw * to )
void nbd_h2r_reply(struct nbd_reply *from, struct nbd_reply_raw *to)
{
to->magic = be32toh( from->magic );
to->error = be32toh( from->error );
memcpy( to->handle, from->handle, 8 );
to->magic = htobe32(from->magic);
to->error = htobe32(from->error);
to->handle.w = from->handle.w;
}

View File

@@ -7,16 +7,39 @@
#define INIT_MAGIC 0x0000420281861253
#define REQUEST_MAGIC 0x25609513
#define REPLY_MAGIC 0x67446698
#define REQUEST_READ 0
#define REQUEST_WRITE 1
#define REQUEST_DISCONNECT 2
#define REQUEST_FLUSH 3
/* The top 2 bytes of the type field are overloaded and can contain flags */
#define REQUEST_MASK 0x0000ffff
/* values for transmission flag field */
#define FLAG_HAS_FLAGS (1 << 0) /* Flags are there */
#define FLAG_SEND_FLUSH (1 << 2) /* Send FLUSH */
#define FLAG_SEND_FUA (1 << 3) /* Send FUA (Force Unit Access) */
/* values for command flag field */
#define CMD_FLAG_FUA (1 << 0)
#if 0
/* Not yet implemented by flexnbd */
#define REQUEST_TRIM 4
#define REQUEST_WRITE_ZEROES 6
#define FLAG_READ_ONLY (1 << 1) /* Device is read-only */
#define FLAG_ROTATIONAL (1 << 4) /* Use elevator algorithm - rotational media */
#define FLAG_SEND_TRIM (1 << 5) /* Send TRIM (discard) */
#define FLAG_SEND_WRITE_ZEROES (1 << 6) /* Send NBD_CMD_WRITE_ZEROES */
#define FLAG_CAN_MULTI_CONN (1 << 8) /* multiple connections are okay */
#define CMD_FLAG_NO_HOLE (1 << 1)
#endif
/* 1MiB is the de-facto standard for maximum size of header + data */
#define NBD_MAX_SIZE ( 1024 * 1024 )
/* 32 MiB is the maximum qemu will send you:
* https://github.com/qemu/qemu/blob/v2.11.0/include/block/nbd.h#L183
*/
#define NBD_MAX_SIZE ( 32 * 1024 * 1024 )
#define NBD_REQUEST_SIZE ( sizeof( struct nbd_request_raw ) )
#define NBD_REPLY_SIZE ( sizeof( struct nbd_reply_raw ) )
@@ -24,62 +47,68 @@
#include <linux/types.h>
#include <inttypes.h>
typedef union nbd_handle_t {
uint8_t b[8];
uint64_t w;
} nbd_handle_t;
/* The _raw types are the types as they appear on the wire. Non-_raw
* types are in host-format.
* Conversion functions are _r2h_ for converting raw to host, and _h2r_
* for converting host to raw.
*/
struct nbd_init_raw {
char passwd[8];
__be64 magic;
__be64 size;
char reserved[128];
char passwd[8];
__be64 magic;
__be64 size;
__be32 flags;
char reserved[124];
};
struct nbd_request_raw {
__be32 magic;
__be32 type; /* == READ || == WRITE */
char handle[8];
__be64 from;
__be32 len;
} __attribute__((packed));
__be32 magic;
__be16 flags;
__be16 type; /* == READ || == WRITE || == FLUSH */
nbd_handle_t handle;
__be64 from;
__be32 len;
} __attribute__ ((packed));
struct nbd_reply_raw {
__be32 magic;
__be32 error; /* 0 = ok, else error */
char handle[8]; /* handle you got from request */
__be32 magic;
__be32 error; /* 0 = ok, else error */
nbd_handle_t handle; /* handle you got from request */
};
struct nbd_init {
char passwd[8];
uint64_t magic;
uint64_t size;
char reserved[128];
char passwd[8];
uint64_t magic;
uint64_t size;
uint32_t flags;
char reserved[124];
};
struct nbd_request {
uint32_t magic;
uint32_t type; /* == READ || == WRITE || == DISCONNECT */
char handle[8];
uint64_t from;
uint32_t len;
} __attribute__((packed));
uint32_t magic;
uint16_t flags;
uint16_t type; /* == READ || == WRITE || == DISCONNECT || == FLUSH */
nbd_handle_t handle;
uint64_t from;
uint32_t len;
} __attribute__ ((packed));
struct nbd_reply {
uint32_t magic;
uint32_t error; /* 0 = ok, else error */
char handle[8]; /* handle you got from request */
uint32_t magic;
uint32_t error; /* 0 = ok, else error */
nbd_handle_t handle; /* handle you got from request */
};
void nbd_r2h_init( struct nbd_init_raw * from, struct nbd_init * to );
void nbd_r2h_request( struct nbd_request_raw *from, struct nbd_request * to );
void nbd_r2h_reply( struct nbd_reply_raw * from, struct nbd_reply * to );
void nbd_r2h_init(struct nbd_init_raw *from, struct nbd_init *to);
void nbd_r2h_request(struct nbd_request_raw *from, struct nbd_request *to);
void nbd_r2h_reply(struct nbd_reply_raw *from, struct nbd_reply *to);
void nbd_h2r_init( struct nbd_init * from, struct nbd_init_raw * to);
void nbd_h2r_request( struct nbd_request * from, struct nbd_request_raw * to );
void nbd_h2r_reply( struct nbd_reply * from, struct nbd_reply_raw * to );
void nbd_h2r_init(struct nbd_init *from, struct nbd_init_raw *to);
void nbd_h2r_request(struct nbd_request *from, struct nbd_request_raw *to);
void nbd_h2r_reply(struct nbd_reply *from, struct nbd_reply_raw *to);
#endif

View File

@@ -10,118 +10,116 @@ int atoi(const char *nptr);
)
/* FIXME: should change this to return negative on error like everything else */
int parse_ip_to_sockaddr(struct sockaddr* out, char* src)
int parse_ip_to_sockaddr(struct sockaddr *out, char *src)
{
NULLCHECK( out );
NULLCHECK( src );
NULLCHECK(out);
NULLCHECK(src);
char temp[64];
struct sockaddr_in *v4 = (struct sockaddr_in *) out;
struct sockaddr_in6 *v6 = (struct sockaddr_in6 *) out;
char temp[64];
struct sockaddr_in *v4 = (struct sockaddr_in *) out;
struct sockaddr_in6 *v6 = (struct sockaddr_in6 *) out;
/* allow user to start with [ and end with any other invalid char */
{
int i=0, j=0;
if (src[i] == '[') { i++; }
for (; i<64 && IS_IP_VALID_CHAR(src[i]); i++) {
temp[j++] = src[i];
}
temp[j] = 0;
/* allow user to start with [ and end with any other invalid char */
{
int i = 0, j = 0;
if (src[i] == '[') {
i++;
}
if (temp[0] == '0' && temp[1] == '\0') {
v4->sin_family = AF_INET;
v4->sin_addr.s_addr = INADDR_ANY;
return 1;
for (; i < 64 && IS_IP_VALID_CHAR(src[i]); i++) {
temp[j++] = src[i];
}
temp[j] = 0;
}
if (inet_pton(AF_INET, temp, &v4->sin_addr) == 1) {
out->sa_family = AF_INET;
return 1;
}
if (temp[0] == '0' && temp[1] == '\0') {
v4->sin_family = AF_INET;
v4->sin_addr.s_addr = INADDR_ANY;
return 1;
}
if (inet_pton(AF_INET6, temp, &v6->sin6_addr) == 1) {
out->sa_family = AF_INET6;
return 1;
}
if (inet_pton(AF_INET, temp, &v4->sin_addr) == 1) {
out->sa_family = AF_INET;
return 1;
}
return 0;
if (inet_pton(AF_INET6, temp, &v6->sin6_addr) == 1) {
out->sa_family = AF_INET6;
return 1;
}
return 0;
}
int parse_to_sockaddr(struct sockaddr* out, char* address)
int parse_to_sockaddr(struct sockaddr *out, char *address)
{
struct sockaddr_un* un = (struct sockaddr_un*) out;
struct sockaddr_un *un = (struct sockaddr_un *) out;
NULLCHECK( address );
NULLCHECK(address);
if ( address[0] == '/' ) {
un->sun_family = AF_UNIX;
strncpy( un->sun_path, address, 108 ); /* FIXME: linux only */
return 1;
}
if (address[0] == '/') {
un->sun_family = AF_UNIX;
strncpy(un->sun_path, address, 108); /* FIXME: linux only */
return 1;
}
return parse_ip_to_sockaddr( out, address );
return parse_ip_to_sockaddr(out, address);
}
int parse_acl(struct ip_and_mask (**out)[], int max, char **entries)
{
struct ip_and_mask* list;
int i;
struct ip_and_mask *list;
int i;
if (max == 0) {
*out = NULL;
return 0;
}
else {
list = xmalloc(max * sizeof(struct ip_and_mask));
*out = (struct ip_and_mask (*)[])list;
debug("acl alloc: %p", *out);
}
if (max == 0) {
*out = NULL;
return 0;
} else {
list = xmalloc(max * sizeof(struct ip_and_mask));
*out = (struct ip_and_mask(*)[]) list;
debug("acl alloc: %p", *out);
}
for (i = 0; i < max; i++) {
int j;
struct ip_and_mask* outentry = &list[i];
for (i = 0; i < max; i++) {
int j;
struct ip_and_mask *outentry = &list[i];
# define MAX_MASK_BITS (outentry->ip.family == AF_INET ? 32 : 128)
if (parse_ip_to_sockaddr(&outentry->ip.generic, entries[i]) == 0) {
return i;
}
if (parse_ip_to_sockaddr(&outentry->ip.generic, entries[i]) == 0) {
return i;
}
for (j=0; entries[i][j] && entries[i][j] != '/'; j++)
; // increment j!
for (j = 0; entries[i][j] && entries[i][j] != '/'; j++); // increment j!
if (entries[i][j] == '/') {
outentry->mask = atoi(entries[i]+j+1);
if (outentry->mask < 1 || outentry->mask > MAX_MASK_BITS) {
return i;
}
}
else {
outentry->mask = MAX_MASK_BITS;
}
if (entries[i][j] == '/') {
outentry->mask = atoi(entries[i] + j + 1);
if (outentry->mask < 1 || outentry->mask > MAX_MASK_BITS) {
return i;
}
} else {
outentry->mask = MAX_MASK_BITS;
}
# undef MAX_MASK_BITS
debug("acl ptr[%d]: %p %d",i, outentry, outentry->mask);
}
debug("acl ptr[%d]: %p %d", i, outentry, outentry->mask);
}
for (i=0; i < max; i++) {
debug("acl entry %d @ %p has mask %d", i, list[i], list[i].mask);
}
for (i = 0; i < max; i++) {
debug("acl entry %d @ %p has mask %d", i, list[i], list[i].mask);
}
return max;
return max;
}
void parse_port( char *s_port, struct sockaddr_in *out )
void parse_port(char *s_port, struct sockaddr_in *out)
{
NULLCHECK( s_port );
NULLCHECK(s_port);
int raw_port;
int raw_port;
raw_port = atoi( s_port );
if ( raw_port < 0 || raw_port > 65535 ) {
fatal( "Port number must be >= 0 and <= 65535" );
}
out->sin_port = htobe16( raw_port );
raw_port = atoi(s_port);
if (raw_port < 0 || raw_port > 65535) {
fatal("Port number must be >= 0 and <= 65535");
}
out->sin_port = htobe16(raw_port);
}

View File

@@ -8,22 +8,21 @@
#include <unistd.h>
union mysockaddr {
unsigned short family;
struct sockaddr generic;
struct sockaddr_in v4;
struct sockaddr_in6 v6;
struct sockaddr_un un;
unsigned short family;
struct sockaddr generic;
struct sockaddr_in v4;
struct sockaddr_in6 v6;
struct sockaddr_un un;
};
struct ip_and_mask {
union mysockaddr ip;
int mask;
union mysockaddr ip;
int mask;
};
int parse_ip_to_sockaddr(struct sockaddr* out, char* src);
int parse_to_sockaddr(struct sockaddr* out, char* src);
int parse_ip_to_sockaddr(struct sockaddr *out, char *src);
int parse_to_sockaddr(struct sockaddr *out, char *src);
int parse_acl(struct ip_and_mask (**out)[], int max, char **entries);
void parse_port( char *s_port, struct sockaddr_in *out );
void parse_port(char *s_port, struct sockaddr_in *out);
#endif

View File

@@ -8,214 +8,226 @@
#include <string.h>
#include <sys/socket.h>
int socket_connect(struct sockaddr* to, struct sockaddr* from)
int socket_connect(struct sockaddr *to, struct sockaddr *from)
{
int fd = socket(to->sa_family == AF_INET ? PF_INET : PF_INET6, SOCK_STREAM, 0);
if( fd < 0 ){
warn( "Couldn't create client socket");
return -1;
}
int fd =
socket(to->sa_family == AF_INET ? PF_INET : PF_INET6, SOCK_STREAM,
0);
if (fd < 0) {
warn("Couldn't create client socket");
return -1;
}
if (NULL != from) {
if ( 0 > bind( fd, from, sizeof(struct sockaddr_in6 ) ) ){
warn( SHOW_ERRNO( "bind() to source address failed" ) );
if ( 0 > close( fd ) ) { /* Non-fatal leak */
warn( SHOW_ERRNO( "Failed to close fd %i", fd ) );
}
return -1;
}
if (NULL != from) {
if (0 > bind(fd, from, sizeof(struct sockaddr_in6))) {
warn(SHOW_ERRNO("bind() to source address failed"));
if (0 > close(fd)) { /* Non-fatal leak */
warn(SHOW_ERRNO("Failed to close fd %i", fd));
}
return -1;
}
}
if ( 0 > sock_try_connect( fd, to, sizeof( struct sockaddr_in6 ), 15 ) ) {
warn( SHOW_ERRNO( "connect failed" ) );
if ( 0 > close( fd ) ) { /* Non-fatal leak */
warn( SHOW_ERRNO( "Failed to close fd %i", fd ) );
}
return -1;
if (0 > sock_try_connect(fd, to, sizeof(struct sockaddr_in6), 15)) {
warn(SHOW_ERRNO("connect failed"));
if (0 > close(fd)) { /* Non-fatal leak */
warn(SHOW_ERRNO("Failed to close fd %i", fd));
}
return -1;
}
if ( sock_set_tcp_nodelay( fd, 1 ) == -1 ) {
warn( SHOW_ERRNO( "Failed to set TCP_NODELAY" ) );
}
if (sock_set_tcp_nodelay(fd, 1) == -1) {
warn(SHOW_ERRNO("Failed to set TCP_NODELAY"));
}
return fd;
return fd;
}
int nbd_check_hello( struct nbd_init_raw* init_raw, uint64_t* out_size )
int nbd_check_hello(struct nbd_init_raw *init_raw, uint64_t * out_size,
uint32_t * out_flags)
{
if ( strncmp( init_raw->passwd, INIT_PASSWD, 8 ) != 0 ) {
warn( "wrong passwd" );
goto fail;
}
if ( be64toh( init_raw->magic ) != INIT_MAGIC ) {
warn( "wrong magic (%x)", be64toh( init_raw->magic ) );
goto fail;
}
if (strncmp(init_raw->passwd, INIT_PASSWD, 8) != 0) {
warn("wrong passwd");
goto fail;
}
if (be64toh(init_raw->magic) != INIT_MAGIC) {
warn("wrong magic (%x)", be64toh(init_raw->magic));
goto fail;
}
if ( NULL != out_size ) {
*out_size = be64toh( init_raw->size );
}
if (NULL != out_size) {
*out_size = be64toh(init_raw->size);
}
return 1;
fail:
if (NULL != out_flags) {
*out_flags = be32toh(init_raw->flags);
}
return 1;
fail:
return 0;
}
int socket_nbd_read_hello(int fd, uint64_t * out_size,
uint32_t * out_flags)
{
struct nbd_init_raw init_raw;
if (0 > readloop(fd, &init_raw, sizeof(init_raw))) {
warn(SHOW_ERRNO("Couldn't read init"));
return 0;
}
return nbd_check_hello(&init_raw, out_size, out_flags);
}
int socket_nbd_read_hello( int fd, uint64_t* out_size )
void nbd_hello_to_buf(struct nbd_init_raw *buf, off64_t out_size,
uint32_t out_flags)
{
struct nbd_init_raw init_raw;
struct nbd_init init;
memcpy(&init.passwd, INIT_PASSWD, 8);
init.magic = INIT_MAGIC;
init.size = out_size;
init.flags = out_flags;
if ( 0 > readloop( fd, &init_raw, sizeof(init_raw) ) ) {
warn( "Couldn't read init" );
return 0;
}
memset(buf, 0, sizeof(struct nbd_init_raw)); // ensure reserved is 0s
nbd_h2r_init(&init, buf);
return nbd_check_hello( &init_raw, out_size );
return;
}
void nbd_hello_to_buf( struct nbd_init_raw *buf, off64_t out_size )
int socket_nbd_write_hello(int fd, off64_t out_size, uint32_t out_flags)
{
struct nbd_init init;
struct nbd_init_raw init_raw;
nbd_hello_to_buf(&init_raw, out_size, out_flags);
memcpy( &init.passwd, INIT_PASSWD, 8 );
init.magic = INIT_MAGIC;
init.size = out_size;
memset( buf, 0, sizeof( struct nbd_init_raw ) ); // ensure reserved is 0s
nbd_h2r_init( &init, buf );
return;
if (0 > writeloop(fd, &init_raw, sizeof(init_raw))) {
warn(SHOW_ERRNO("failed to write hello to socket"));
return 0;
}
return 1;
}
int socket_nbd_write_hello(int fd, off64_t out_size)
void fill_request(struct nbd_request_raw *request_raw, uint16_t type,
uint16_t flags, uint64_t from, uint32_t len)
{
struct nbd_init_raw init_raw;
nbd_hello_to_buf( &init_raw, out_size );
if ( 0 > writeloop( fd, &init_raw, sizeof( init_raw ) ) ) {
warn( SHOW_ERRNO( "failed to write hello to socket" ) );
return 0;
}
return 1;
request_raw->magic = htobe32(REQUEST_MAGIC);
request_raw->type = htobe16(type);
request_raw->flags = htobe16(flags);
request_raw->handle.w =
(((uint64_t) rand()) << 32) | ((uint64_t) rand());
request_raw->from = htobe64(from);
request_raw->len = htobe32(len);
}
void fill_request(struct nbd_request *request, int type, uint64_t from, uint32_t len)
void read_reply(int fd, uint64_t request_raw_handle,
struct nbd_reply *reply)
{
request->magic = htobe32(REQUEST_MAGIC);
request->type = htobe32(type);
uint32_t * randa = (uint32_t*)request->handle;
randa[0] = rand();
randa[1] = rand();
request->from = htobe64(from);
request->len = htobe32(len);
struct nbd_reply_raw reply_raw;
ERROR_IF_NEGATIVE(readloop
(fd, &reply_raw, sizeof(struct nbd_reply_raw)),
SHOW_ERRNO("Couldn't read reply"));
nbd_r2h_reply(&reply_raw, reply);
if (reply->magic != REPLY_MAGIC) {
error("Reply magic incorrect (%x)", reply->magic);
}
if (reply->error != 0) {
error("Server replied with error %d", reply->error);
}
if (request_raw_handle != reply_raw.handle.w) {
error("Did not reply with correct handle");
}
}
void read_reply(int fd, struct nbd_request *request, struct nbd_reply *reply)
void wait_for_data(int fd, int timeout_secs)
{
struct nbd_reply_raw reply_raw;
fd_set fds;
struct timeval tv = { timeout_secs, 0 };
int selected;
ERROR_IF_NEGATIVE(readloop(fd, &reply_raw, sizeof(struct nbd_reply_raw)),
"Couldn't read reply");
FD_ZERO(&fds);
FD_SET(fd, &fds);
nbd_r2h_reply( &reply_raw, reply );
selected =
sock_try_select(FD_SETSIZE, &fds, NULL, NULL,
timeout_secs >= 0 ? &tv : NULL);
if (reply->magic != REPLY_MAGIC) {
error("Reply magic incorrect (%x)", reply->magic);
}
if (reply->error != 0) {
error("Server replied with error %d", reply->error);
}
if (strncmp(request->handle, reply->handle, 8) != 0) {
error("Did not reply with correct handle");
}
}
void wait_for_data( int fd, int timeout_secs )
{
fd_set fds;
struct timeval tv = { timeout_secs, 0 };
int selected;
FD_ZERO( &fds );
FD_SET( fd, &fds );
selected = sock_try_select(
FD_SETSIZE, &fds, NULL, NULL, timeout_secs >=0 ? &tv : NULL
);
FATAL_IF( -1 == selected, "Select failed" );
ERROR_IF( 0 == selected, "Timed out waiting for reply" );
FATAL_IF(-1 == selected, "Select failed");
ERROR_IF(0 == selected, "Timed out waiting for reply");
}
void socket_nbd_read(int fd, uint64_t from, uint32_t len, int out_fd, void* out_buf, int timeout_secs)
void socket_nbd_read(int fd, uint64_t from, uint32_t len, int out_fd,
void *out_buf, int timeout_secs)
{
struct nbd_request request;
struct nbd_reply reply;
struct nbd_request_raw request_raw;
struct nbd_reply reply;
fill_request(&request, REQUEST_READ, from, len);
FATAL_IF_NEGATIVE(writeloop(fd, &request, sizeof(request)),
"Couldn't write request");
fill_request(&request_raw, REQUEST_READ, 0, from, len);
FATAL_IF_NEGATIVE(writeloop(fd, &request_raw, sizeof(request_raw)),
SHOW_ERRNO("Couldn't write request"));
wait_for_data( fd, timeout_secs );
read_reply(fd, &request, &reply);
if (out_buf) {
FATAL_IF_NEGATIVE(readloop(fd, out_buf, len),
"Read failed");
}
else {
FATAL_IF_NEGATIVE(
splice_via_pipe_loop(fd, out_fd, len),
"Splice failed"
);
}
wait_for_data(fd, timeout_secs);
read_reply(fd, request_raw.handle.w, &reply);
if (out_buf) {
FATAL_IF_NEGATIVE(readloop(fd, out_buf, len),
SHOW_ERRNO("Read failed"));
} else {
FATAL_IF_NEGATIVE(splice_via_pipe_loop(fd, out_fd, len),
"Splice failed");
}
}
void socket_nbd_write(int fd, uint64_t from, uint32_t len, int in_fd, void* in_buf, int timeout_secs)
void socket_nbd_write(int fd, uint64_t from, uint32_t len, int in_fd,
void *in_buf, int timeout_secs)
{
struct nbd_request request;
struct nbd_reply reply;
struct nbd_request_raw request_raw;
struct nbd_reply reply;
fill_request(&request, REQUEST_WRITE, from, len);
ERROR_IF_NEGATIVE(writeloop(fd, &request, sizeof(request)),
"Couldn't write request");
fill_request(&request_raw, REQUEST_WRITE, 0, from, len);
ERROR_IF_NEGATIVE(writeloop(fd, &request_raw, sizeof(request_raw)),
SHOW_ERRNO("Couldn't write request"));
if (in_buf) {
ERROR_IF_NEGATIVE(writeloop(fd, in_buf, len),
"Write failed");
}
else {
ERROR_IF_NEGATIVE(
splice_via_pipe_loop(in_fd, fd, len),
"Splice failed"
);
}
if (in_buf) {
ERROR_IF_NEGATIVE(writeloop(fd, in_buf, len),
SHOW_ERRNO("Write failed"));
} else {
ERROR_IF_NEGATIVE(splice_via_pipe_loop(in_fd, fd, len),
"Splice failed");
}
wait_for_data( fd, timeout_secs );
read_reply(fd, &request, &reply);
wait_for_data(fd, timeout_secs);
read_reply(fd, request_raw.handle.w, &reply);
}
int socket_nbd_disconnect( int fd )
int socket_nbd_disconnect(int fd)
{
int success = 1;
struct nbd_request request;
int success = 1;
struct nbd_request_raw request_raw;
fill_request( &request, REQUEST_DISCONNECT, 0, 0 );
/* FIXME: This shouldn't be a FATAL error. We should just drop
* the mirror without affecting the main server.
*/
FATAL_IF_NEGATIVE( writeloop( fd, &request, sizeof( request ) ),
"Failed to write the disconnect request." );
return success;
fill_request(&request_raw, REQUEST_DISCONNECT, 0, 0, 0);
/* FIXME: This shouldn't be a FATAL error. We should just drop
* the mirror without affecting the main server.
*/
FATAL_IF_NEGATIVE(writeloop(fd, &request_raw, sizeof(request_raw)),
SHOW_ERRNO
("Failed to write the disconnect request."));
return success;
}
#define CHECK_RANGE(error_type) { \
uint64_t size;\
int success = socket_nbd_read_hello(params->client, &size); \
uint32_t flags;\
int success = socket_nbd_read_hello(params->client, &size, &flags); \
if ( success ) {\
uint64_t endpoint = params->from + params->len; \
if (endpoint > size || \
@@ -231,23 +243,26 @@ int socket_nbd_disconnect( int fd )
}\
}
void do_read(struct mode_readwrite_params* params)
void do_read(struct mode_readwrite_params *params)
{
params->client = socket_connect(&params->connect_to.generic, &params->connect_from.generic);
FATAL_IF_NEGATIVE( params->client, "Couldn't connect." );
CHECK_RANGE("read");
socket_nbd_read(params->client, params->from, params->len,
params->data_fd, NULL, 10);
close(params->client);
params->client =
socket_connect(&params->connect_to.generic,
&params->connect_from.generic);
FATAL_IF_NEGATIVE(params->client, "Couldn't connect.");
CHECK_RANGE("read");
socket_nbd_read(params->client, params->from, params->len,
params->data_fd, NULL, 10);
close(params->client);
}
void do_write(struct mode_readwrite_params* params)
void do_write(struct mode_readwrite_params *params)
{
params->client = socket_connect(&params->connect_to.generic, &params->connect_from.generic);
FATAL_IF_NEGATIVE( params->client, "Couldn't connect." );
CHECK_RANGE("write");
socket_nbd_write(params->client, params->from, params->len,
params->data_fd, NULL, 10);
close(params->client);
params->client =
socket_connect(&params->connect_to.generic,
&params->connect_from.generic);
FATAL_IF_NEGATIVE(params->client, "Couldn't connect.");
CHECK_RANGE("write");
socket_nbd_write(params->client, params->from, params->len,
params->data_fd, NULL, 10);
close(params->client);
}

View File

@@ -6,18 +6,21 @@
#include <sys/socket.h>
#include "nbdtypes.h"
int socket_connect(struct sockaddr* to, struct sockaddr* from);
int socket_nbd_read_hello(int fd, uint64_t* size);
int socket_nbd_write_hello(int fd, uint64_t size);
void socket_nbd_read(int fd, uint64_t from, uint32_t len, int out_fd, void* out_buf, int timeout_secs);
void socket_nbd_write(int fd, uint64_t from, uint32_t len, int out_fd, void* out_buf, int timeout_secs);
int socket_nbd_disconnect( int fd );
int socket_connect(struct sockaddr *to, struct sockaddr *from);
int socket_nbd_read_hello(int fd, uint64_t * size, uint32_t * flags);
int socket_nbd_write_hello(int fd, uint64_t size, uint32_t flags);
void socket_nbd_read(int fd, uint64_t from, uint32_t len, int out_fd,
void *out_buf, int timeout_secs);
void socket_nbd_write(int fd, uint64_t from, uint32_t len, int out_fd,
void *out_buf, int timeout_secs);
int socket_nbd_disconnect(int fd);
/* as you can see, we're slowly accumulating code that should really be in an
* NBD library */
void nbd_hello_to_buf( struct nbd_init_raw* buf, uint64_t out_size );
int nbd_check_hello( struct nbd_init_raw* init_raw, uint64_t* out_size );
void nbd_hello_to_buf(struct nbd_init_raw *buf, uint64_t out_size,
uint32_t out_flags);
int nbd_check_hello(struct nbd_init_raw *init_raw, uint64_t * out_size,
uint32_t * out_flags);
#endif

View File

@@ -4,64 +4,62 @@
#include <stdlib.h>
#include <sys/un.h>
static const int max_response=1024;
static const int max_response = 1024;
void print_response( const char * response )
void print_response(const char *response)
{
char * response_text;
FILE * out;
int exit_status;
char *response_text;
FILE *out;
int exit_status;
NULLCHECK( response );
NULLCHECK(response);
exit_status = atoi(response);
response_text = strchr( response, ':' );
exit_status = atoi(response);
response_text = strchr(response, ':');
FATAL_IF_NULL( response_text,
"Error parsing server response: '%s'", response );
FATAL_IF_NULL(response_text,
"Error parsing server response: '%s'", response);
out = exit_status > 0 ? stderr : stdout;
fprintf(out, "%s\n", response_text + 2);
out = exit_status > 0 ? stderr : stdout;
fprintf(out, "%s\n", response_text + 2);
}
void do_remote_command(char* command, char* socket_name, int argc, char** argv)
void do_remote_command(char *command, char *socket_name, int argc,
char **argv)
{
char newline=10;
int i;
debug( "connecting to run remote command %s", command );
int remote = socket(AF_UNIX, SOCK_STREAM, 0);
struct sockaddr_un address;
char response[max_response];
char newline = 10;
int i;
debug("connecting to run remote command %s", command);
int remote = socket(AF_UNIX, SOCK_STREAM, 0);
struct sockaddr_un address;
char response[max_response];
memset(&address, 0, sizeof(address));
memset(&address, 0, sizeof(address));
FATAL_IF_NEGATIVE(remote, "Couldn't create client socket");
FATAL_IF_NEGATIVE(remote, "Couldn't create client socket");
address.sun_family = AF_UNIX;
strncpy(address.sun_path, socket_name, sizeof(address.sun_path));
address.sun_family = AF_UNIX;
strncpy(address.sun_path, socket_name, sizeof(address.sun_path));
FATAL_IF_NEGATIVE(
connect(remote, (struct sockaddr*) &address, sizeof(address)),
"Couldn't connect to %s", socket_name
);
FATAL_IF_NEGATIVE(connect
(remote, (struct sockaddr *) &address,
sizeof(address)), "Couldn't connect to %s",
socket_name);
write(remote, command, strlen(command));
write(remote, &newline, 1);
for (i=0; i<argc; i++) {
if ( NULL != argv[i] ) {
write(remote, argv[i], strlen(argv[i]));
}
write(remote, &newline, 1);
write(remote, command, strlen(command));
write(remote, &newline, 1);
for (i = 0; i < argc; i++) {
if (NULL != argv[i]) {
write(remote, argv[i], strlen(argv[i]));
}
write(remote, &newline, 1);
}
write(remote, &newline, 1);
FATAL_IF_NEGATIVE(
read_until_newline(remote, response, max_response),
"Couldn't read response from %s", socket_name
);
FATAL_IF_NEGATIVE(read_until_newline(remote, response, max_response),
"Couldn't read response from %s", socket_name);
print_response( response );
print_response(response);
exit(atoi(response));
exit(atoi(response));
}

View File

@@ -24,18 +24,18 @@
#include "util.h"
#include "self_pipe.h"
#define ERR_MSG_PIPE "Couldn't open a pipe for signaling."
#define ERR_MSG_PIPE "Couldn't open a pipe for signaling."
#define ERR_MSG_FCNTL "Couldn't set a signalling pipe non-blocking."
#define ERR_MSG_WRITE "Couldn't write to a signaling pipe."
#define ERR_MSG_READ "Couldn't read from a signaling pipe."
void self_pipe_server_error( int err, char *msg )
void self_pipe_server_error(int err, char *msg)
{
char errbuf[1024] = {0};
char errbuf[1024] = { 0 };
strerror_r( err, errbuf, 1024 );
strerror_r(err, errbuf, 1024);
fatal( "%s\t%d (%s)", msg, err, errbuf );
fatal("%s\t%d (%s)", msg, err, errbuf);
}
/**
@@ -47,33 +47,36 @@ void self_pipe_server_error( int err, char *msg )
* Remember to call self_pipe_destroy when you're done with the return
* value.
*/
struct self_pipe * self_pipe_create(void)
struct self_pipe *self_pipe_create(void)
{
struct self_pipe *sig = xmalloc( sizeof( struct self_pipe ) );
int fds[2];
if ( NULL == sig ) { return NULL; }
struct self_pipe *sig = xmalloc(sizeof(struct self_pipe));
int fds[2];
if ( pipe( fds ) ) {
free( sig );
self_pipe_server_error( errno, ERR_MSG_PIPE );
return NULL;
}
if (NULL == sig) {
return NULL;
}
if ( fcntl( fds[0], F_SETFL, O_NONBLOCK ) || fcntl( fds[1], F_SETFL, O_NONBLOCK ) ) {
int fcntl_err = errno;
while( close( fds[0] ) == -1 && errno == EINTR );
while( close( fds[1] ) == -1 && errno == EINTR );
free( sig );
self_pipe_server_error( fcntl_err, ERR_MSG_FCNTL );
if (pipe(fds)) {
free(sig);
self_pipe_server_error(errno, ERR_MSG_PIPE);
return NULL;
}
return NULL;
}
if (fcntl(fds[0], F_SETFL, O_NONBLOCK)
|| fcntl(fds[1], F_SETFL, O_NONBLOCK)) {
int fcntl_err = errno;
while (close(fds[0]) == -1 && errno == EINTR);
while (close(fds[1]) == -1 && errno == EINTR);
free(sig);
self_pipe_server_error(fcntl_err, ERR_MSG_FCNTL);
sig->read_fd = fds[0];
sig->write_fd = fds[1];
return NULL;
}
return sig;
sig->read_fd = fds[0];
sig->write_fd = fds[1];
return sig;
}
@@ -83,19 +86,19 @@ struct self_pipe * self_pipe_create(void)
* Returns 1 on success. Can fail if weirdness happened to the write fd
* of the pipe in the self_pipe struct.
*/
int self_pipe_signal( struct self_pipe * sig )
int self_pipe_signal(struct self_pipe *sig)
{
NULLCHECK( sig );
FATAL_IF( 1 == sig->write_fd, "Shouldn't be writing to stdout" );
FATAL_IF( 2 == sig->write_fd, "Shouldn't be writing to stderr" );
NULLCHECK(sig);
FATAL_IF(1 == sig->write_fd, "Shouldn't be writing to stdout");
FATAL_IF(2 == sig->write_fd, "Shouldn't be writing to stderr");
int written = write( sig->write_fd, "X", 1 );
if ( written != 1 ) {
self_pipe_server_error( errno, ERR_MSG_WRITE );
return 0;
}
int written = write(sig->write_fd, "X", 1);
if (written != 1) {
self_pipe_server_error(errno, ERR_MSG_WRITE);
return 0;
}
return 1;
return 1;
}
@@ -106,11 +109,11 @@ int self_pipe_signal( struct self_pipe * sig )
* Returns the number of bytes read, which will be 1 on success and 0 if
* there was no signal.
*/
int self_pipe_signal_clear( struct self_pipe *sig )
int self_pipe_signal_clear(struct self_pipe *sig)
{
char buf[1];
char buf[1];
return 1 == read( sig->read_fd, buf, 1 );
return 1 == read(sig->read_fd, buf, 1);
}
@@ -118,30 +121,30 @@ int self_pipe_signal_clear( struct self_pipe *sig )
* Close the pipe and free the self_pipe. Do not try to use the
* self_pipe struct after calling this, the innards are mush.
*/
int self_pipe_destroy( struct self_pipe * sig )
int self_pipe_destroy(struct self_pipe *sig)
{
NULLCHECK(sig);
NULLCHECK(sig);
while( close( sig->read_fd ) == -1 && errno == EINTR );
while( close( sig->write_fd ) == -1 && errno == EINTR );
while (close(sig->read_fd) == -1 && errno == EINTR);
while (close(sig->write_fd) == -1 && errno == EINTR);
/* Just in case anyone *does* try to use this after free,
* we should set the memory locations to an error value
*/
sig->read_fd = -1;
sig->write_fd = -1;
/* Just in case anyone *does* try to use this after free,
* we should set the memory locations to an error value
*/
sig->read_fd = -1;
sig->write_fd = -1;
free( sig );
return 1;
free(sig);
return 1;
}
int self_pipe_fd_set( struct self_pipe * sig, fd_set * fds)
int self_pipe_fd_set(struct self_pipe *sig, fd_set * fds)
{
FD_SET( sig->read_fd, fds );
return 1;
FD_SET(sig->read_fd, fds);
return 1;
}
int self_pipe_fd_isset( struct self_pipe * sig, fd_set * fds)
int self_pipe_fd_isset(struct self_pipe *sig, fd_set * fds)
{
return FD_ISSET( sig->read_fd, fds );
return FD_ISSET(sig->read_fd, fds);
}

View File

@@ -4,16 +4,16 @@
#include <sys/select.h>
struct self_pipe {
int read_fd;
int write_fd;
int read_fd;
int write_fd;
};
struct self_pipe * self_pipe_create(void);
int self_pipe_signal( struct self_pipe * sig );
int self_pipe_signal_clear( struct self_pipe *sig );
int self_pipe_destroy( struct self_pipe * sig );
int self_pipe_fd_set( struct self_pipe * sig, fd_set * fds );
int self_pipe_fd_isset( struct self_pipe *sig, fd_set *fds );
struct self_pipe *self_pipe_create(void);
int self_pipe_signal(struct self_pipe *sig);
int self_pipe_signal_clear(struct self_pipe *sig);
int self_pipe_destroy(struct self_pipe *sig);
int self_pipe_fd_set(struct self_pipe *sig, fd_set * fds);
int self_pipe_fd_isset(struct self_pipe *sig, fd_set * fds);
#endif

View File

@@ -9,245 +9,287 @@
#include "sockutil.h"
#include "util.h"
size_t sockaddr_size( const struct sockaddr* sa )
size_t sockaddr_size(const struct sockaddr * sa)
{
struct sockaddr_un* un = (struct sockaddr_un*) sa;
size_t ret = 0;
struct sockaddr_un *un = (struct sockaddr_un *) sa;
size_t ret = 0;
switch( sa->sa_family ) {
case AF_INET:
ret = sizeof( struct sockaddr_in );
break;
case AF_INET6:
ret = sizeof( struct sockaddr_in6 );
break;
case AF_UNIX:
ret = sizeof( un->sun_family ) + SUN_LEN( un );
break;
}
switch (sa->sa_family) {
case AF_INET:
ret = sizeof(struct sockaddr_in);
break;
case AF_INET6:
ret = sizeof(struct sockaddr_in6);
break;
case AF_UNIX:
ret = sizeof(un->sun_family) + SUN_LEN(un);
break;
}
return ret;
return ret;
}
const char* sockaddr_address_string( const struct sockaddr* sa, char* dest, size_t len )
const char *sockaddr_address_string(const struct sockaddr *sa, char *dest,
size_t len)
{
NULLCHECK( sa );
NULLCHECK( dest );
NULLCHECK(sa);
NULLCHECK(dest);
struct sockaddr_in* in = ( struct sockaddr_in* ) sa;
struct sockaddr_in6* in6 = ( struct sockaddr_in6* ) sa;
struct sockaddr_un* un = ( struct sockaddr_un* ) sa;
struct sockaddr_in *in = (struct sockaddr_in *) sa;
struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) sa;
struct sockaddr_un *un = (struct sockaddr_un *) sa;
unsigned short real_port = ntohs( in->sin_port ); // common to in and in6
const char* ret = NULL;
unsigned short real_port = ntohs(in->sin_port); // common to in and in6
const char *ret = NULL;
memset( dest, 0, len );
memset(dest, 0, len);
if ( sa->sa_family == AF_INET ) {
ret = inet_ntop( AF_INET, &in->sin_addr, dest, len );
} else if ( sa->sa_family == AF_INET6 ) {
ret = inet_ntop( AF_INET6, &in6->sin6_addr, dest, len );
} else if ( sa->sa_family == AF_UNIX ) {
ret = strncpy( dest, un->sun_path, SUN_LEN( un ) );
}
if (sa->sa_family == AF_INET) {
ret = inet_ntop(AF_INET, &in->sin_addr, dest, len);
} else if (sa->sa_family == AF_INET6) {
ret = inet_ntop(AF_INET6, &in6->sin6_addr, dest, len);
} else if (sa->sa_family == AF_UNIX) {
ret = strncpy(dest, un->sun_path, SUN_LEN(un));
}
if ( ret == NULL ) {
strncpy( dest, "???", len );
}
if (ret == NULL) {
strncpy(dest, "???", len);
}
if ( NULL != ret && real_port > 0 && sa->sa_family != AF_UNIX ) {
size_t size = strlen( dest );
snprintf( dest + size, len - size, " port %d", real_port );
}
if (NULL != ret && real_port > 0 && sa->sa_family != AF_UNIX) {
size_t size = strlen(dest);
snprintf(dest + size, len - size, " port %d", real_port);
}
return ret;
return ret;
}
int sock_set_reuseaddr( int fd, int optval )
int sock_set_reuseaddr(int fd, int optval)
{
return setsockopt( fd, SOL_SOCKET, SO_REUSEADDR, &optval, sizeof(optval) );
return setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &optval,
sizeof(optval));
}
int sock_set_keepalive_params(int fd, int time, int intvl, int probes)
{
if (sock_set_keepalive(fd, 1) ||
sock_set_tcp_keepidle(fd, time) ||
sock_set_tcp_keepintvl(fd, intvl) ||
sock_set_tcp_keepcnt(fd, probes)) {
return -1;
}
return 0;
}
int sock_set_keepalive(int fd, int optval)
{
return setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, &optval,
sizeof(optval));
}
int sock_set_tcp_keepidle(int fd, int optval)
{
return setsockopt(fd, IPPROTO_TCP, TCP_KEEPIDLE, &optval,
sizeof(optval));
}
int sock_set_tcp_keepintvl(int fd, int optval)
{
return setsockopt(fd, IPPROTO_TCP, TCP_KEEPINTVL, &optval,
sizeof(optval));
}
int sock_set_tcp_keepcnt(int fd, int optval)
{
return setsockopt(fd, IPPROTO_TCP, TCP_KEEPCNT, &optval,
sizeof(optval));
}
/* Set the tcp_nodelay option */
int sock_set_tcp_nodelay( int fd, int optval )
int sock_set_tcp_nodelay(int fd, int optval)
{
return setsockopt( fd, IPPROTO_TCP, TCP_NODELAY, &optval, sizeof(optval) );
return setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &optval,
sizeof(optval));
}
int sock_set_tcp_cork( int fd, int optval )
int sock_set_tcp_cork(int fd, int optval)
{
return setsockopt( fd, IPPROTO_TCP, TCP_CORK, &optval, sizeof(optval) );
return setsockopt(fd, IPPROTO_TCP, TCP_CORK, &optval, sizeof(optval));
}
int sock_set_nonblock( int fd, int optval )
int sock_set_nonblock(int fd, int optval)
{
int flags = fcntl( fd, F_GETFL );
int flags = fcntl(fd, F_GETFL);
if ( flags == -1 ) {
return -1;
}
if (flags == -1) {
return -1;
}
if ( optval ) {
flags = flags | O_NONBLOCK;
if (optval) {
flags = flags | O_NONBLOCK;
} else {
flags = flags & (~O_NONBLOCK);
}
return fcntl(fd, F_SETFL, flags);
}
int sock_try_bind(int fd, const struct sockaddr *sa)
{
int bind_result;
char s_address[256];
int retry = 10;
sockaddr_address_string(sa, &s_address[0], 256);
do {
bind_result = bind(fd, sa, sockaddr_size(sa));
if (0 == bind_result) {
info("Bound to %s", s_address);
break;
} else {
flags = flags & (~O_NONBLOCK);
warn(SHOW_ERRNO("Couldn't bind to %s", s_address));
switch (errno) {
/* bind() can give us EACCES, EADDRINUSE, EADDRNOTAVAIL, EBADF,
* EINVAL, ENOTSOCK, EFAULT, ELOOP, ENAMETOOLONG, ENOENT,
* ENOMEM, ENOTDIR, EROFS
*
* Any of these other than EADDRINUSE & EADDRNOTAVAIL signify
* that there's a logic error somewhere.
*
* EADDRINUSE is fatal: if there's something already where we
* want to be listening, we have no guarantees that any clients
* will cope with it.
*/
case EADDRNOTAVAIL:
retry--;
if (retry) {
debug("retrying");
sleep(1);
}
continue;
case EADDRINUSE:
warn("%s in use, giving up.", s_address);
retry = 0;
break;
default:
warn("giving up");
retry = 0;
}
}
} while (retry);
return bind_result;
}
int sock_try_select(int nfds, fd_set * readfds, fd_set * writefds,
fd_set * exceptfds, struct timeval *timeout)
{
int result;
do {
result = select(nfds, readfds, writefds, exceptfds, timeout);
if (errno != EINTR) {
break;
}
return fcntl( fd, F_SETFL, flags );
} while (result == -1);
return result;
}
int sock_try_bind( int fd, const struct sockaddr* sa )
int sock_try_connect(int fd, struct sockaddr *to, socklen_t addrlen,
int wait)
{
int bind_result;
char s_address[256];
int retry = 1;
fd_set fds;
struct timeval tv = { wait, 0 };
int result = 0;
sockaddr_address_string( sa, &s_address[0], 256 );
if (sock_set_nonblock(fd, 1) == -1) {
warn(SHOW_ERRNO
("Failed to set socket non-blocking for connect()"));
return connect(fd, to, addrlen);
}
do {
bind_result = bind( fd, sa, sockaddr_size( sa ) );
if ( 0 == bind_result ) {
info( "Bound to %s", s_address );
break;
}
else {
warn( SHOW_ERRNO( "Couldn't bind to %s", s_address ) );
FD_ZERO(&fds);
FD_SET(fd, &fds);
switch ( errno ) {
/* bind() can give us EACCES, EADDRINUSE, EADDRNOTAVAIL, EBADF,
* EINVAL, ENOTSOCK, EFAULT, ELOOP, ENAMETOOLONG, ENOENT,
* ENOMEM, ENOTDIR, EROFS
*
* Any of these other than EADDRINUSE & EADDRNOTAVAIL signify
* that there's a logic error somewhere.
*
* EADDRINUSE is fatal: if there's something already where we
* want to be listening, we have no guarantees that any clients
* will cope with it.
*/
case EADDRNOTAVAIL:
debug( "retrying" );
sleep( 1 );
continue;
case EADDRINUSE:
warn( "%s in use, giving up.", s_address );
retry = 0;
break;
default:
warn( "giving up" );
retry = 0;
}
}
} while ( retry );
do {
result = connect(fd, to, addrlen);
return bind_result;
}
int sock_try_select(int nfds, fd_set *readfds, fd_set *writefds, fd_set *exceptfds, struct timeval *timeout)
{
int result;
do {
result = select(nfds, readfds, writefds, exceptfds, timeout);
if ( errno != EINTR ) {
break;
}
} while ( result == -1 );
return result;
}
int sock_try_connect( int fd, struct sockaddr* to, socklen_t addrlen, int wait )
{
fd_set fds;
struct timeval tv = { wait, 0 };
int result = 0;
if ( sock_set_nonblock( fd, 1 ) == -1 ) {
warn( SHOW_ERRNO( "Failed to set socket non-blocking for connect()" ) );
return connect( fd, to, addrlen );
}
FD_ZERO( &fds );
FD_SET( fd, &fds );
do {
result = connect( fd, to, addrlen );
if ( result == -1 ) {
switch( errno ) {
case EINPROGRESS:
result = 0;
break; /* success */
case EAGAIN:
case EINTR:
/* Try connect() again. This only breaks out of the switch,
* not the do...while loop. since result == -1, we go again.
*/
break;
default:
warn( SHOW_ERRNO( "Failed to connect()" ) );
goto out;
}
}
} while ( result == -1 );
if ( -1 == sock_try_select( FD_SETSIZE, NULL, &fds, NULL, &tv) ) {
warn( SHOW_ERRNO( "failed to select() on non-blocking connect" ) );
result = -1;
if (result == -1) {
switch (errno) {
case EINPROGRESS:
result = 0;
break; /* success */
case EAGAIN:
case EINTR:
/* Try connect() again. This only breaks out of the switch,
* not the do...while loop. since result == -1, we go again.
*/
break;
default:
warn(SHOW_ERRNO("Failed to connect()"));
goto out;
}
}
} while (result == -1);
if ( !FD_ISSET( fd, &fds ) ) {
result = -1;
errno = ETIMEDOUT;
goto out;
}
if (-1 == sock_try_select(FD_SETSIZE, NULL, &fds, NULL, &tv)) {
warn(SHOW_ERRNO("failed to select() on non-blocking connect"));
result = -1;
goto out;
}
int scratch;
socklen_t s_size = sizeof( scratch );
if ( getsockopt( fd, SOL_SOCKET, SO_ERROR, &scratch, &s_size ) == -1 ) {
result = -1;
warn( SHOW_ERRNO( "getsockopt() failed" ) );
goto out;
}
if (!FD_ISSET(fd, &fds)) {
result = -1;
errno = ETIMEDOUT;
goto out;
}
if ( scratch == EINPROGRESS ) {
scratch = ETIMEDOUT;
}
int scratch;
socklen_t s_size = sizeof(scratch);
if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &scratch, &s_size) == -1) {
result = -1;
warn(SHOW_ERRNO("getsockopt() failed"));
goto out;
}
result = scratch ? -1 : 0;
errno = scratch;
if (scratch == EINPROGRESS) {
scratch = ETIMEDOUT;
}
out:
if ( sock_set_nonblock( fd, 0 ) == -1 ) {
warn( SHOW_ERRNO( "Failed to make socket blocking after connect()" ) );
return -1;
}
result = scratch ? -1 : 0;
errno = scratch;
debug( "sock_try_connect: %i", result );
return result;
out:
if (sock_set_nonblock(fd, 0) == -1) {
warn(SHOW_ERRNO("Failed to make socket blocking after connect()"));
return -1;
}
debug("sock_try_connect: %i", result);
return result;
}
int sock_try_close( int fd )
int sock_try_close(int fd)
{
int result;
int result;
do {
result = close( fd );
do {
result = close(fd);
if ( result == -1 ) {
if ( EINTR == errno ) {
continue; /* retry EINTR */
} else {
warn( SHOW_ERRNO( "Failed to close() fd %i", fd ) );
break; /* Other errors get reported */
}
}
if (result == -1) {
if (EINTR == errno) {
continue; /* retry EINTR */
} else {
warn(SHOW_ERRNO("Failed to close() fd %i", fd));
break; /* Other errors get reported */
}
}
} while( 0 );
} while (0);
return result;
return result;
}

View File

@@ -7,16 +7,32 @@
#include <sys/select.h>
/* Returns the size of the sockaddr, or 0 on error */
size_t sockaddr_size(const struct sockaddr* sa);
size_t sockaddr_size(const struct sockaddr *sa);
/* Convert a sockaddr into an address. Like inet_ntop, it returns dest if
* successful, NULL otherwise. In the latter case, dest will contain "???"
*/
const char* sockaddr_address_string(const struct sockaddr* sa, char* dest, size_t len);
const char *sockaddr_address_string(const struct sockaddr *sa, char *dest,
size_t len);
/* Configure TCP keepalive on a socket */
int sock_set_keepalive_params(int fd, int time, int intvl, int probes);
/* Set the SOL_KEEPALIVE otion */
int sock_set_keepalive(int fd, int optval);
/* Set the SOL_REUSEADDR otion */
int sock_set_reuseaddr(int fd, int optval);
/* Set the tcp_keepidle option */
int sock_set_tcp_keepidle(int fd, int optval);
/* Set the tcp_keepintvl option */
int sock_set_tcp_keepintvl(int fd, int optval);
/* Set the tcp_keepcnt option */
int sock_set_tcp_keepcnt(int fd, int optval);
/* Set the tcp_nodelay option */
int sock_set_tcp_nodelay(int fd, int optval);
@@ -26,16 +42,17 @@ int sock_set_tcp_cork(int fd, int optval);
int sock_set_nonblock(int fd, int optval);
/* Attempt to bind the fd to the sockaddr, retrying common transient failures */
int sock_try_bind(int fd, const struct sockaddr* sa);
int sock_try_bind(int fd, const struct sockaddr *sa);
/* Try to call select(), retrying EINTR */
int sock_try_select(int nfds, fd_set *readfds, fd_set *writefds, fd_set *exceptfds, struct timeval *timeout);
int sock_try_select(int nfds, fd_set * readfds, fd_set * writefds,
fd_set * exceptfds, struct timeval *timeout);
/* Try to call connect(), timing out after wait seconds */
int sock_try_connect( int fd, struct sockaddr* to, socklen_t addrlen, int wait );
int sock_try_connect(int fd, struct sockaddr *to, socklen_t addrlen,
int wait);
/* Try to call close(), retrying EINTR */
int sock_try_close( int fd );
int sock_try_close(int fd);
#endif

View File

@@ -13,75 +13,79 @@
pthread_key_t cleanup_handler_key;
int log_level = 2;
char *log_context = "";
void error_init(void)
{
pthread_key_create(&cleanup_handler_key, free);
pthread_key_create(&cleanup_handler_key, free);
}
void error_handler(int fatal)
{
DECLARE_ERROR_CONTEXT(context);
DECLARE_ERROR_CONTEXT(context);
if (context) {
longjmp(context->jmp, fatal ? 1 : 2 );
}
else {
if ( fatal ) { abort(); }
else { pthread_exit((void*) 1); }
if (context) {
longjmp(context->jmp, fatal ? 1 : 2);
} else {
if (fatal) {
abort();
} else {
pthread_exit((void *) 1);
}
}
}
void exit_err( const char *msg )
void exit_err(const char *msg)
{
fprintf( stderr, "%s\n", msg );
exit( 1 );
fprintf(stderr, "%s\n", msg);
exit(1);
}
void mylog(int line_level, const char* format, ...)
void mylog(int line_level, const char *format, ...)
{
if (line_level < log_level) { return; }
if (line_level < log_level) {
return;
}
va_list argptr;
va_list argptr;
va_start(argptr, format);
vfprintf(stderr, format, argptr);
va_end(argptr);
va_start(argptr, format);
vfprintf(stderr, format, argptr);
va_end(argptr);
}
uint64_t monotonic_time_ms()
{
struct timespec ts;
uint64_t seconds_ms, nanoseconds_ms;
struct timespec ts;
uint64_t seconds_ms, nanoseconds_ms;
FATAL_IF_NEGATIVE(
clock_gettime(CLOCK_MONOTONIC, &ts),
SHOW_ERRNO( "clock_gettime failed" )
FATAL_IF_NEGATIVE(clock_gettime(CLOCK_MONOTONIC, &ts),
SHOW_ERRNO("clock_gettime failed")
);
seconds_ms = ts.tv_sec;
seconds_ms = seconds_ms * 1000;
seconds_ms = ts.tv_sec;
seconds_ms = seconds_ms * 1000;
nanoseconds_ms = ts.tv_nsec;
nanoseconds_ms = nanoseconds_ms / 1000000;
nanoseconds_ms = ts.tv_nsec;
nanoseconds_ms = nanoseconds_ms / 1000000;
return seconds_ms + nanoseconds_ms;
return seconds_ms + nanoseconds_ms;
}
void* xrealloc(void* ptr, size_t size)
void *xrealloc(void *ptr, size_t size)
{
void* p = realloc(ptr, size);
FATAL_IF_NULL(p, "couldn't xrealloc %d bytes", ptr ? "realloc" : "malloc", size);
return p;
void *p = realloc(ptr, size);
FATAL_IF_NULL(p, "couldn't xrealloc %d bytes",
ptr ? "realloc" : "malloc", size);
return p;
}
void* xmalloc(size_t size)
void *xmalloc(size_t size)
{
void* p = xrealloc(NULL, size);
memset(p, 0, size);
return p;
void *p = xrealloc(NULL, size);
memset(p, 0, size);
return p;
}

View File

@@ -10,10 +10,11 @@
#include <unistd.h>
#include <inttypes.h>
void* xrealloc(void* ptr, size_t size);
void* xmalloc(size_t size);
void *xrealloc(void *ptr, size_t size);
void *xmalloc(size_t size);
typedef void (cleanup_handler)(void* /* context */, int /* is fatal? */);
typedef void (cleanup_handler) (void * /* context */ ,
int /* is fatal? */ );
/* set from 0 - 5 depending on what level of verbosity you want */
extern int log_level;
@@ -21,16 +22,19 @@ extern int log_level;
/* set up the error globals */
void error_init(void);
/* some context for the overall process that appears on each log line */
extern char *log_context;
void exit_err( const char * );
void exit_err(const char *);
/* error_set_handler must be a macro not a function due to setjmp stack rules */
#include <setjmp.h>
struct error_handler_context {
jmp_buf jmp;
cleanup_handler* handler;
void* data;
jmp_buf jmp;
cleanup_handler *handler;
void *data;
};
#define DECLARE_ERROR_CONTEXT(name) \
@@ -84,7 +88,7 @@ extern pthread_key_t cleanup_handler_key;
void error_handler(int fatal);
/* mylog a line at the given level (0 being most verbose) */
void mylog(int line_level, const char* format, ...);
void mylog(int line_level, const char *format, ...);
/* Returns the current time, in milliseconds, from CLOCK_MONOTONIC */
uint64_t monotonic_time_ms(void);
@@ -92,12 +96,12 @@ uint64_t monotonic_time_ms(void);
#define levstr(i) (i==0?'D':(i==1?'I':(i==2?'W':(i==3?'E':'F'))))
#define myloglev(level, msg, ...) mylog( level, "%"PRIu64":%c:%d %p %s:%d: "msg"\n", monotonic_time_ms(), levstr(level), getpid(),pthread_self(), __FILE__, __LINE__, ##__VA_ARGS__ )
#define myloglev(level, msg, ...) mylog( level, "%"PRIu64":%c:%d %p %s %s:%d: "msg"\n", monotonic_time_ms(), levstr(level), getpid(),pthread_self(), log_context, __FILE__, __LINE__, ##__VA_ARGS__ )
#ifdef DEBUG
# define debug(msg, ...) myloglev(0, msg, ##__VA_ARGS__)
#define debug(msg, ...) myloglev(0, msg, ##__VA_ARGS__)
#else
# define debug(msg, ...) /* no-op */
#define debug(msg, ...) /* no-op */
#endif
/* informational message, not expected to be compiled out */
@@ -159,4 +163,3 @@ uint64_t monotonic_time_ms(void);
#define WARN_IF_NEGATIVE( value, msg, ... ) if ( value < 0 ) { warn( msg, ##__VA_ARGS__ ); }
#endif

View File

@@ -5,18 +5,17 @@
#include <stdlib.h>
#include <time.h>
int main(int argc, char** argv)
int main(int argc, char **argv)
{
signal(SIGPIPE, SIG_IGN); /* calls to splice() unhelpfully throw this */
error_init();
signal(SIGPIPE, SIG_IGN); /* calls to splice() unhelpfully throw this */
error_init();
srand(time(NULL));
srand(time(NULL));
if (argc < 2) {
exit_err( help_help_text );
}
mode(argv[1], argc-1, argv+1); /* never returns */
if (argc < 2) {
exit_err(help_help_text);
}
mode(argv[1], argc - 1, argv + 1); /* never returns */
return 0;
return 0;
}

View File

@@ -8,164 +8,158 @@
static struct option proxy_options[] = {
GETOPT_HELP,
GETOPT_ADDR,
GETOPT_PORT,
GETOPT_CONNECT_ADDR,
GETOPT_CONNECT_PORT,
GETOPT_BIND,
GETOPT_CACHE,
GETOPT_QUIET,
GETOPT_VERBOSE,
{0}
GETOPT_HELP,
GETOPT_ADDR,
GETOPT_PORT,
GETOPT_CONNECT_ADDR,
GETOPT_CONNECT_PORT,
GETOPT_BIND,
GETOPT_CACHE,
GETOPT_QUIET,
GETOPT_VERBOSE,
{0}
};
static char proxy_short_options[] = "hl:p:C:P:b:" SOPT_QUIET SOPT_VERBOSE;
static char proxy_help_text[] =
"Usage: flexnbd-proxy <options>\n\n"
"Resiliently proxy an NBD connection between client and server\n"
"We can listen on TCP or UNIX socket, but only connect to TCP servers.\n\n"
HELP_LINE
"\t--" OPT_ADDR ",-l <ADDR>\tThe address we will bind to as a proxy.\n"
"\t--" OPT_PORT ",-p <PORT>\tThe port we will bind to as a proxy, if required.\n"
"\t--" OPT_CONNECT_ADDR ",-C <ADDR>\tAddress of the proxied server.\n"
"\t--" OPT_CONNECT_PORT ",-P <PORT>\tPort of the proxied server.\n"
"\t--" OPT_BIND ",-b <ADDR>\tThe address we connect from, as a proxy.\n"
"\t--" OPT_CACHE ",-c[=<CACHE-BYTES>]\tUse a RAM read cache of the given size.\n"
QUIET_LINE
VERBOSE_LINE;
"Usage: flexnbd-proxy <options>\n\n"
"Resiliently proxy an NBD connection between client and server\n"
"We can listen on TCP or UNIX socket, but only connect to TCP servers.\n\n"
HELP_LINE
"\t--" OPT_ADDR ",-l <ADDR>\tThe address we will bind to as a proxy.\n"
"\t--" OPT_PORT
",-p <PORT>\tThe port we will bind to as a proxy, if required.\n"
"\t--" OPT_CONNECT_ADDR ",-C <ADDR>\tAddress of the proxied server.\n"
"\t--" OPT_CONNECT_PORT ",-P <PORT>\tPort of the proxied server.\n"
"\t--" OPT_BIND
",-b <ADDR>\tThe address we connect from, as a proxy.\n" "\t--"
OPT_CACHE
",-c[=<CACHE-BYTES>]\tUse a RAM read cache of the given size.\n"
QUIET_LINE VERBOSE_LINE;
static char proxy_default_cache_size[] = "4096";
void read_proxy_param(
int c,
char **downstream_addr,
char **downstream_port,
char **upstream_addr,
char **upstream_port,
char **bind_addr,
char **cache_bytes)
void read_proxy_param(int c,
char **downstream_addr,
char **downstream_port,
char **upstream_addr,
char **upstream_port,
char **bind_addr, char **cache_bytes)
{
switch( c ) {
case 'h' :
fprintf( stdout, "%s\n", proxy_help_text );
exit( 0 );
case 'l':
*downstream_addr = optarg;
break;
case 'p':
*downstream_port = optarg;
break;
case 'C':
*upstream_addr = optarg;
break;
case 'P':
*upstream_port = optarg;
break;
case 'b':
*bind_addr = optarg;
break;
case 'c':
*cache_bytes = optarg ? optarg : proxy_default_cache_size;
break;
case 'q':
log_level = QUIET_LOG_LEVEL;
break;
case 'v':
log_level = VERBOSE_LOG_LEVEL;
break;
default:
exit_err( proxy_help_text );
break;
}
switch (c) {
case 'h':
fprintf(stdout, "%s\n", proxy_help_text);
exit(0);
case 'l':
*downstream_addr = optarg;
break;
case 'p':
*downstream_port = optarg;
break;
case 'C':
*upstream_addr = optarg;
break;
case 'P':
*upstream_port = optarg;
break;
case 'b':
*bind_addr = optarg;
break;
case 'c':
*cache_bytes = optarg ? optarg : proxy_default_cache_size;
break;
case 'q':
log_level = QUIET_LOG_LEVEL;
break;
case 'v':
log_level = VERBOSE_LOG_LEVEL;
break;
default:
exit_err(proxy_help_text);
break;
}
}
struct proxier * proxy = NULL;
struct proxier *proxy = NULL;
void my_exit(int signum)
{
info( "Exit signalled (%i)", signum );
if ( NULL != proxy ) {
proxy_cleanup( proxy );
};
exit( 0 );
info("Exit signalled (%i)", signum);
if (NULL != proxy) {
proxy_cleanup(proxy);
};
exit(0);
}
int main( int argc, char *argv[] )
int main(int argc, char *argv[])
{
int c;
char *downstream_addr = NULL;
char *downstream_port = NULL;
char *upstream_addr = NULL;
char *upstream_port = NULL;
char *bind_addr = NULL;
char *cache_bytes = NULL;
int success;
int c;
char *downstream_addr = NULL;
char *downstream_port = NULL;
char *upstream_addr = NULL;
char *upstream_port = NULL;
char *bind_addr = NULL;
char *cache_bytes = NULL;
int success;
sigset_t mask;
struct sigaction exit_action;
sigset_t mask;
struct sigaction exit_action;
sigemptyset( &mask );
sigaddset( &mask, SIGTERM );
sigaddset( &mask, SIGQUIT );
sigaddset( &mask, SIGINT );
sigemptyset(&mask);
sigaddset(&mask, SIGTERM);
sigaddset(&mask, SIGQUIT);
sigaddset(&mask, SIGINT);
exit_action.sa_handler = my_exit;
exit_action.sa_mask = mask;
exit_action.sa_flags = 0;
exit_action.sa_handler = my_exit;
exit_action.sa_mask = mask;
exit_action.sa_flags = 0;
srand(time(NULL));
srand(time(NULL));
while (1) {
c = getopt_long( argc, argv, proxy_short_options, proxy_options, NULL );
if ( -1 == c ) { break; }
read_proxy_param( c,
&downstream_addr,
&downstream_port,
&upstream_addr,
&upstream_port,
&bind_addr,
&cache_bytes
);
while (1) {
c = getopt_long(argc, argv, proxy_short_options, proxy_options,
NULL);
if (-1 == c) {
break;
}
read_proxy_param(c,
&downstream_addr,
&downstream_port,
&upstream_addr,
&upstream_port, &bind_addr, &cache_bytes);
}
if ( NULL == downstream_addr ){
fprintf( stderr, "--addr is required.\n" );
exit_err( proxy_help_text );
} else if ( NULL == upstream_addr || NULL == upstream_port ){
fprintf( stderr, "both --conn-addr and --conn-port are required.\n" );
exit_err( proxy_help_text );
}
if (NULL == downstream_addr) {
fprintf(stderr, "--addr is required.\n");
exit_err(proxy_help_text);
} else if (NULL == upstream_addr || NULL == upstream_port) {
fprintf(stderr,
"both --conn-addr and --conn-port are required.\n");
exit_err(proxy_help_text);
}
proxy = proxy_create(
downstream_addr,
downstream_port,
upstream_addr,
upstream_port,
bind_addr,
cache_bytes
);
proxy = proxy_create(downstream_addr,
downstream_port,
upstream_addr,
upstream_port, bind_addr, cache_bytes);
/* Set these *after* proxy has been assigned to */
sigaction(SIGTERM, &exit_action, NULL);
sigaction(SIGQUIT, &exit_action, NULL);
sigaction(SIGINT, &exit_action, NULL);
signal(SIGPIPE, SIG_IGN); /* calls to splice() unhelpfully throw this */
/* Set these *after* proxy has been assigned to */
sigaction(SIGTERM, &exit_action, NULL);
sigaction(SIGQUIT, &exit_action, NULL);
sigaction(SIGINT, &exit_action, NULL);
signal(SIGPIPE, SIG_IGN); /* calls to splice() unhelpfully throw this */
if ( NULL != downstream_port ) {
info(
"Proxying between %s %s (downstream) and %s %s (upstream)",
downstream_addr, downstream_port, upstream_addr, upstream_port
);
} else {
info(
"Proxying between %s (downstream) and %s %s (upstream)",
downstream_addr, upstream_addr, upstream_port
);
}
if (NULL != downstream_port) {
info("Proxying between %s %s (downstream) and %s %s (upstream)",
downstream_addr, downstream_port, upstream_addr,
upstream_port);
} else {
info("Proxying between %s (downstream) and %s %s (upstream)",
downstream_addr, upstream_addr, upstream_port);
}
success = do_proxy( proxy );
proxy_destroy( proxy );
success = do_proxy(proxy);
proxy_destroy(proxy);
return success ? 0 : 1;
return success ? 0 : 1;
}

View File

@@ -2,67 +2,77 @@
#include "util.h"
struct prefetch* prefetch_create( size_t size_bytes ){
struct prefetch *prefetch_create(size_t size_bytes)
{
struct prefetch* out = xmalloc( sizeof( struct prefetch ) );
NULLCHECK( out );
out->buffer = xmalloc( size_bytes );
NULLCHECK( out->buffer );
struct prefetch *out = xmalloc(sizeof(struct prefetch));
NULLCHECK(out);
out->size = size_bytes;
out->is_full = 0;
out->from = 0;
out->len = 0;
out->buffer = xmalloc(size_bytes);
NULLCHECK(out->buffer);
return out;
out->size = size_bytes;
out->is_full = 0;
out->from = 0;
out->len = 0;
return out;
}
void prefetch_destroy( struct prefetch *prefetch ) {
if( prefetch ) {
free( prefetch->buffer );
free( prefetch );
}
void prefetch_destroy(struct prefetch *prefetch)
{
if (prefetch) {
free(prefetch->buffer);
free(prefetch);
}
}
size_t prefetch_size( struct prefetch *prefetch){
if ( prefetch ) {
return prefetch->size;
} else {
return 0;
}
size_t prefetch_size(struct prefetch *prefetch)
{
if (prefetch) {
return prefetch->size;
} else {
return 0;
}
}
void prefetch_set_is_empty( struct prefetch *prefetch ){
prefetch_set_full( prefetch, 0 );
void prefetch_set_is_empty(struct prefetch *prefetch)
{
prefetch_set_full(prefetch, 0);
}
void prefetch_set_is_full( struct prefetch *prefetch ){
prefetch_set_full( prefetch, 1 );
void prefetch_set_is_full(struct prefetch *prefetch)
{
prefetch_set_full(prefetch, 1);
}
void prefetch_set_full( struct prefetch *prefetch, int val ){
if( prefetch ) {
prefetch->is_full = val;
}
void prefetch_set_full(struct prefetch *prefetch, int val)
{
if (prefetch) {
prefetch->is_full = val;
}
}
int prefetch_is_full( struct prefetch *prefetch ){
if( prefetch ) {
return prefetch->is_full;
} else {
return 0;
}
int prefetch_is_full(struct prefetch *prefetch)
{
if (prefetch) {
return prefetch->is_full;
} else {
return 0;
}
}
int prefetch_contains( struct prefetch *prefetch, uint64_t from, uint32_t len ){
NULLCHECK( prefetch );
return from >= prefetch->from &&
from + len <= prefetch->from + prefetch->len;
int prefetch_contains(struct prefetch *prefetch, uint64_t from,
uint32_t len)
{
NULLCHECK(prefetch);
return from >= prefetch->from &&
from + len <= prefetch->from + prefetch->len;
}
char *prefetch_offset( struct prefetch *prefetch, uint64_t from ){
NULLCHECK( prefetch );
return prefetch->buffer + (from - prefetch->from);
char *prefetch_offset(struct prefetch *prefetch, uint64_t from)
{
NULLCHECK(prefetch);
return prefetch->buffer + (from - prefetch->from);
}

View File

@@ -7,27 +7,28 @@
#define PREFETCH_BUFSIZE 4096
struct prefetch {
/* True if there is data in the buffer. */
int is_full;
/* The start point of the current content of buffer */
uint64_t from;
/* The length of the current content of buffer */
uint32_t len;
/* True if there is data in the buffer. */
int is_full;
/* The start point of the current content of buffer */
uint64_t from;
/* The length of the current content of buffer */
uint32_t len;
/* The total size of the buffer, in bytes. */
size_t size;
/* The total size of the buffer, in bytes. */
size_t size;
char *buffer;
char *buffer;
};
struct prefetch* prefetch_create( size_t size_bytes );
void prefetch_destroy( struct prefetch *prefetch );
size_t prefetch_size( struct prefetch *);
void prefetch_set_is_empty( struct prefetch *prefetch );
void prefetch_set_is_full( struct prefetch *prefetch );
void prefetch_set_full( struct prefetch *prefetch, int val );
int prefetch_is_full( struct prefetch *prefetch );
int prefetch_contains( struct prefetch *prefetch, uint64_t from, uint32_t len );
char *prefetch_offset( struct prefetch *prefetch, uint64_t from );
struct prefetch *prefetch_create(size_t size_bytes);
void prefetch_destroy(struct prefetch *prefetch);
size_t prefetch_size(struct prefetch *);
void prefetch_set_is_empty(struct prefetch *prefetch);
void prefetch_set_is_full(struct prefetch *prefetch);
void prefetch_set_full(struct prefetch *prefetch, int val);
int prefetch_is_full(struct prefetch *prefetch);
int prefetch_contains(struct prefetch *prefetch, uint64_t from,
uint32_t len);
char *prefetch_offset(struct prefetch *prefetch, uint64_t from);
#endif

File diff suppressed because it is too large Load Diff

View File

@@ -10,7 +10,7 @@
#include "self_pipe.h"
#ifdef PREFETCH
#include "prefetch.h"
#include "prefetch.h"
#endif
/** UPSTREAM_TIMEOUT
@@ -21,77 +21,77 @@
struct proxier {
/** address/port to bind to */
union mysockaddr listen_on;
union mysockaddr listen_on;
/** address/port to connect to */
union mysockaddr connect_to;
union mysockaddr connect_to;
/** address to bind to when making outgoing connections */
union mysockaddr connect_from;
int bind; /* Set to true if we should use it */
union mysockaddr connect_from;
int bind; /* Set to true if we should use it */
/* The socket we listen() on and accept() against */
int listen_fd;
/* The socket we listen() on and accept() against */
int listen_fd;
/* The socket returned by accept() that we receive requests from and send
* responses to
*/
int downstream_fd;
/* The socket returned by accept() that we receive requests from and send
* responses to
*/
int downstream_fd;
/* The socket returned by connect() that we send requests to and receive
* responses from
*/
int upstream_fd;
/* The socket returned by connect() that we send requests to and receive
* responses from
*/
int upstream_fd;
/* This is the size we advertise to the downstream server */
uint64_t upstream_size;
/* This is the size we advertise to the downstream server */
uint64_t upstream_size;
/* We transform the raw request header into here */
struct nbd_request req_hdr;
/* These are the transmission flags sent as part of the handshake */
uint32_t upstream_flags;
/* We transform the raw request header into here */
struct nbd_request req_hdr;
/* We transform the raw reply header into here */
struct nbd_reply rsp_hdr;
struct nbd_reply rsp_hdr;
/* Used for our non-blocking negotiation with upstream. TODO: maybe use
* for downstream as well ( we currently overload rsp ) */
struct iobuf init;
/* Used for our non-blocking negotiation with upstream. TODO: maybe use
* for downstream as well ( we currently overload rsp ) */
struct iobuf init;
/* The current NBD request from downstream */
struct iobuf req;
/* The current NBD request from downstream */
struct iobuf req;
/* The current NBD reply from upstream */
struct iobuf rsp;
/* The current NBD reply from upstream */
struct iobuf rsp;
/* It's starting to feel like we need an object for a single proxy session.
* These two track how many requests we've sent so far, and whether the
* NBD_INIT code has been sent to the client yet.
*/
uint64_t req_count;
int hello_sent;
/* It's starting to feel like we need an object for a single proxy session.
* These two track how many requests we've sent so far, and whether the
* NBD_INIT code has been sent to the client yet.
*/
uint64_t req_count;
int hello_sent;
/** These are only used if we pass --cache on the command line */
/* While the in-flight request has been munged by prefetch, these two are
* set to true, and the original length of the request, respectively */
int is_prefetch_req;
uint32_t prefetch_req_orig_len;
/* While the in-flight request has been munged by prefetch, these two are
* set to true, and the original length of the request, respectively */
int is_prefetch_req;
uint32_t prefetch_req_orig_len;
/* And here, we actually store the prefetched data once it's returned */
struct prefetch *prefetch;
/* And here, we actually store the prefetched data once it's returned */
struct prefetch *prefetch;
/** */
};
struct proxier* proxy_create(
char* s_downstream_address,
char* s_downstream_port,
char* s_upstream_address,
char* s_upstream_port,
char* s_upstream_bind,
char* s_cache_bytes);
int do_proxy( struct proxier* proxy );
void proxy_cleanup( struct proxier* proxy );
void proxy_destroy( struct proxier* proxy );
struct proxier *proxy_create(char *s_downstream_address,
char *s_downstream_port,
char *s_upstream_address,
char *s_upstream_port,
char *s_upstream_bind, char *s_cache_bytes);
int do_proxy(struct proxier *proxy);
void proxy_cleanup(struct proxier *proxy);
void proxy_destroy(struct proxier *proxy);
#endif

View File

@@ -6,103 +6,104 @@
#include "acl.h"
struct acl * acl_create( int len, char ** lines, int default_deny )
struct acl *acl_create(int len, char **lines, int default_deny)
{
struct acl * acl;
struct acl *acl;
acl = (struct acl *)xmalloc( sizeof( struct acl ) );
acl->len = parse_acl( &acl->entries, len, lines );
acl->default_deny = default_deny;
return acl;
acl = (struct acl *) xmalloc(sizeof(struct acl));
acl->len = parse_acl(&acl->entries, len, lines);
acl->default_deny = default_deny;
return acl;
}
static int testmasks[9] = { 0,128,192,224,240,248,252,254,255 };
static int testmasks[9] = { 0, 128, 192, 224, 240, 248, 252, 254, 255 };
/** Test whether AF_INET or AF_INET6 sockaddr is included in the given access
* control list, returning 1 if it is, and 0 if not.
*/
static int is_included_in_acl(int list_length, struct ip_and_mask (*list)[], union mysockaddr* test)
static int is_included_in_acl(int list_length,
struct ip_and_mask (*list)[],
union mysockaddr *test)
{
NULLCHECK( test );
NULLCHECK(test);
int i;
int i;
for (i=0; i < list_length; i++) {
struct ip_and_mask *entry = &(*list)[i];
int testbits;
unsigned char *raw_address1 = NULL, *raw_address2 = NULL;
for (i = 0; i < list_length; i++) {
struct ip_and_mask *entry = &(*list)[i];
int testbits;
unsigned char *raw_address1 = NULL, *raw_address2 = NULL;
debug("checking acl entry %d (%d/%d)", i, test->generic.sa_family, entry->ip.family);
debug("checking acl entry %d (%d/%d)", i, test->generic.sa_family,
entry->ip.family);
if (test->generic.sa_family != entry->ip.family) {
continue;
}
if (test->generic.sa_family == AF_INET) {
debug("it's an AF_INET");
raw_address1 = (unsigned char*) &test->v4.sin_addr;
raw_address2 = (unsigned char*) &entry->ip.v4.sin_addr;
}
else if (test->generic.sa_family == AF_INET6) {
debug("it's an AF_INET6");
raw_address1 = (unsigned char*) &test->v6.sin6_addr;
raw_address2 = (unsigned char*) &entry->ip.v6.sin6_addr;
}
else {
fatal( "Can't check an ACL for this address type." );
}
debug("testbits=%d", entry->mask);
for (testbits = entry->mask; testbits > 0; testbits -= 8) {
debug("testbits=%d, c1=%02x, c2=%02x", testbits, raw_address1[0], raw_address2[0]);
if (testbits >= 8) {
if (raw_address1[0] != raw_address2[0]) { goto no_match; }
}
else {
if ((raw_address1[0] & testmasks[testbits%8]) !=
(raw_address2[0] & testmasks[testbits%8]) ) {
goto no_match;
}
}
raw_address1++;
raw_address2++;
}
return 1;
no_match: ;
debug("no match");
if (test->generic.sa_family != entry->ip.family) {
continue;
}
return 0;
}
int acl_includes( struct acl * acl, union mysockaddr * addr )
{
NULLCHECK( acl );
if ( 0 == acl->len ) {
return !( acl->default_deny );
if (test->generic.sa_family == AF_INET) {
debug("it's an AF_INET");
raw_address1 = (unsigned char *) &test->v4.sin_addr;
raw_address2 = (unsigned char *) &entry->ip.v4.sin_addr;
} else if (test->generic.sa_family == AF_INET6) {
debug("it's an AF_INET6");
raw_address1 = (unsigned char *) &test->v6.sin6_addr;
raw_address2 = (unsigned char *) &entry->ip.v6.sin6_addr;
} else {
fatal("Can't check an ACL for this address type.");
}
else {
return is_included_in_acl( acl->len, acl->entries, addr );
debug("testbits=%d", entry->mask);
for (testbits = entry->mask; testbits > 0; testbits -= 8) {
debug("testbits=%d, c1=%02x, c2=%02x", testbits,
raw_address1[0], raw_address2[0]);
if (testbits >= 8) {
if (raw_address1[0] != raw_address2[0]) {
goto no_match;
}
} else {
if ((raw_address1[0] & testmasks[testbits % 8]) !=
(raw_address2[0] & testmasks[testbits % 8])) {
goto no_match;
}
}
raw_address1++;
raw_address2++;
}
return 1;
no_match:;
debug("no match");
}
return 0;
}
int acl_default_deny( struct acl * acl )
int acl_includes(struct acl *acl, union mysockaddr *addr)
{
NULLCHECK( acl );
return acl->default_deny;
NULLCHECK(acl);
if (0 == acl->len) {
return !(acl->default_deny);
} else {
return is_included_in_acl(acl->len, acl->entries, addr);
}
}
void acl_destroy( struct acl * acl )
int acl_default_deny(struct acl *acl)
{
free( acl->entries );
acl->len = 0;
acl->entries = NULL;
free( acl );
NULLCHECK(acl);
return acl->default_deny;
}
void acl_destroy(struct acl *acl)
{
free(acl->entries);
acl->len = 0;
acl->entries = NULL;
free(acl);
}

View File

@@ -4,9 +4,9 @@
#include "parse.h"
struct acl {
int len;
int default_deny;
struct ip_and_mask (*entries)[];
int len;
int default_deny;
struct ip_and_mask (*entries)[];
};
/** Allocate a new acl structure, parsing the given lines to sockaddr
@@ -17,21 +17,21 @@ struct acl {
* default_deny controls the behaviour of an empty list: if true, all
* requests will be denied. If true, all requests will be accepted.
*/
struct acl * acl_create( int len, char **lines, int default_deny );
struct acl *acl_create(int len, char **lines, int default_deny);
/** Check to see whether an address is allowed by an acl.
* See acl_create for how the default_deny setting affects this.
*/
int acl_includes( struct acl *, union mysockaddr *);
int acl_includes(struct acl *, union mysockaddr *);
/** Get the default_deny status */
int acl_default_deny( struct acl * );
int acl_default_deny(struct acl *);
/** Free the acl structure and the internal acl entries table.
*/
void acl_destroy( struct acl * );
void acl_destroy(struct acl *);
#endif

View File

@@ -12,8 +12,8 @@
* poking at the bits directly without using these
* accessors/macros
*/
typedef uint64_t bitfield_word_t;
typedef bitfield_word_t * bitfield_p;
typedef uint64_t bitfield_word_t;
typedef bitfield_word_t *bitfield_p;
#define BITFIELD_WORD_SIZE sizeof(bitfield_word_t)
#define BITS_PER_WORD (BITFIELD_WORD_SIZE * 8)
@@ -30,65 +30,78 @@ typedef bitfield_word_t * bitfield_p;
((_bytes + (BITFIELD_WORD_SIZE-1)) / BITFIELD_WORD_SIZE)
/** Return the bit value ''idx'' in array ''b'' */
static inline int bit_get(bitfield_p b, uint64_t idx) {
return (BIT_WORD(b, idx) >> (idx & (BITS_PER_WORD-1))) & 1;
static inline int bit_get(bitfield_p b, uint64_t idx)
{
return (BIT_WORD(b, idx) >> (idx & (BITS_PER_WORD - 1))) & 1;
}
/** Return 1 if the bit at ''idx'' in array ''b'' is set */
static inline int bit_is_set(bitfield_p b, uint64_t idx) {
return bit_get(b, idx);
static inline int bit_is_set(bitfield_p b, uint64_t idx)
{
return bit_get(b, idx);
}
/** Return 1 if the bit at ''idx'' in array ''b'' is clear */
static inline int bit_is_clear(bitfield_p b, uint64_t idx) {
return !bit_get(b, idx);
static inline int bit_is_clear(bitfield_p b, uint64_t idx)
{
return !bit_get(b, idx);
}
/** Tests whether the bit at ''idx'' in array ''b'' has value ''value'' */
static inline int bit_has_value(bitfield_p b, uint64_t idx, int value) {
return bit_get(b, idx) == !!value;
static inline int bit_has_value(bitfield_p b, uint64_t idx, int value)
{
return bit_get(b, idx) == ! !value;
}
/** Sets the bit ''idx'' in array ''b'' */
static inline void bit_set(bitfield_p b, uint64_t idx) {
BIT_WORD(b, idx) |= BIT_MASK(idx);
static inline void bit_set(bitfield_p b, uint64_t idx)
{
BIT_WORD(b, idx) |= BIT_MASK(idx);
}
/** Clears the bit ''idx'' in array ''b'' */
static inline void bit_clear(bitfield_p b, uint64_t idx) {
BIT_WORD(b, idx) &= ~BIT_MASK(idx);
static inline void bit_clear(bitfield_p b, uint64_t idx)
{
BIT_WORD(b, idx) &= ~BIT_MASK(idx);
}
/** Sets ''len'' bits in array ''b'' starting at offset ''from'' */
static inline void bit_set_range(bitfield_p b, uint64_t from, uint64_t len)
{
for ( ; (from % BITS_PER_WORD) != 0 && len > 0 ; len-- ) {
bit_set( b, from++ );
}
for (; (from % BITS_PER_WORD) != 0 && len > 0; len--) {
bit_set(b, from++);
}
if (len >= BITS_PER_WORD) {
memset(&BIT_WORD(b, from), 0xff, len / 8 );
from += len;
len = len % BITS_PER_WORD;
from -= len;
}
if (len >= BITS_PER_WORD) {
memset(&BIT_WORD(b, from), 0xff, len / 8);
from += len;
len = len % BITS_PER_WORD;
from -= len;
}
for ( ; len > 0 ; len-- ) {
bit_set( b, from++ );
}
for (; len > 0; len--) {
bit_set(b, from++);
}
}
/** Clears ''len'' bits in array ''b'' starting at offset ''from'' */
static inline void bit_clear_range(bitfield_p b, uint64_t from, uint64_t len)
static inline void bit_clear_range(bitfield_p b, uint64_t from,
uint64_t len)
{
for ( ; (from % BITS_PER_WORD) != 0 && len > 0 ; len-- ) {
bit_clear( b, from++ );
}
for (; (from % BITS_PER_WORD) != 0 && len > 0; len--) {
bit_clear(b, from++);
}
if (len >= BITS_PER_WORD) {
memset(&BIT_WORD(b, from), 0, len / 8 );
from += len;
len = len % BITS_PER_WORD;
from -= len;
}
if (len >= BITS_PER_WORD) {
memset(&BIT_WORD(b, from), 0, len / 8);
from += len;
len = len % BITS_PER_WORD;
from -= len;
}
for ( ; len > 0 ; len-- ) {
bit_clear( b, from++ );
}
for (; len > 0; len--) {
bit_clear(b, from++);
}
}
/** Counts the number of contiguous bits in array ''b'', starting at ''from''
@@ -96,52 +109,54 @@ static inline void bit_clear_range(bitfield_p b, uint64_t from, uint64_t len)
* bits that are the same as the first one specified. If ''run_is_set'' is
* non-NULL, the value of that bit is placed into it.
*/
static inline uint64_t bit_run_count(bitfield_p b, uint64_t from, uint64_t len, int *run_is_set) {
uint64_t count = 0;
int first_value = bit_get(b, from);
bitfield_word_t word_match = first_value ? -1 : 0;
static inline uint64_t bit_run_count(bitfield_p b, uint64_t from,
uint64_t len, int *run_is_set)
{
uint64_t count = 0;
int first_value = bit_get(b, from);
bitfield_word_t word_match = first_value ? -1 : 0;
if ( run_is_set != NULL ) {
*run_is_set = first_value;
if (run_is_set != NULL) {
*run_is_set = first_value;
}
for (; ((from + count) % BITS_PER_WORD) != 0 && len > 0; len--) {
if (bit_has_value(b, from + count, first_value)) {
count++;
} else {
return count;
}
}
for ( ; ((from + count) % BITS_PER_WORD) != 0 && len > 0; len--) {
if (bit_has_value(b, from + count, first_value)) {
count++;
} else {
return count;
}
for (; len >= BITS_PER_WORD; len -= BITS_PER_WORD) {
if (BIT_WORD(b, from + count) == word_match) {
count += BITS_PER_WORD;
} else {
break;
}
}
for ( ; len >= BITS_PER_WORD ; len -= BITS_PER_WORD ) {
if (BIT_WORD(b, from + count) == word_match) {
count += BITS_PER_WORD;
} else {
break;
}
for (; len > 0; len--) {
if (bit_has_value(b, from + count, first_value)) {
count++;
}
}
for ( ; len > 0; len-- ) {
if ( bit_has_value(b, from + count, first_value) ) {
count++;
}
}
return count;
return count;
}
enum bitset_stream_events {
BITSET_STREAM_UNSET = 0,
BITSET_STREAM_SET = 1,
BITSET_STREAM_ON = 2,
BITSET_STREAM_OFF = 3
BITSET_STREAM_UNSET = 0,
BITSET_STREAM_SET = 1,
BITSET_STREAM_ON = 2,
BITSET_STREAM_OFF = 3
};
#define BITSET_STREAM_EVENTS_ENUM_SIZE 4
struct bitset_stream_entry {
enum bitset_stream_events event;
uint64_t from;
uint64_t len;
enum bitset_stream_events event;
uint64_t from;
uint64_t len;
};
/** Limit the stream size to 1MB for now.
@@ -152,14 +167,14 @@ struct bitset_stream_entry {
#define BITSET_STREAM_SIZE ( ( 1024 * 1024 ) / sizeof( struct bitset_stream_entry ) )
struct bitset_stream {
struct bitset_stream_entry entries[BITSET_STREAM_SIZE];
int in;
int out;
int size;
pthread_mutex_t mutex;
pthread_cond_t cond_not_full;
pthread_cond_t cond_not_empty;
uint64_t queued_bytes[BITSET_STREAM_EVENTS_ENUM_SIZE];
struct bitset_stream_entry entries[BITSET_STREAM_SIZE];
int in;
int out;
int size;
pthread_mutex_t mutex;
pthread_cond_t cond_not_full;
pthread_cond_t cond_not_empty;
uint64_t queued_bytes[BITSET_STREAM_EVENTS_ENUM_SIZE];
};
@@ -169,47 +184,49 @@ struct bitset_stream {
* written reliably by multiple threads.
*/
struct bitset {
pthread_mutex_t lock;
uint64_t size;
int resolution;
struct bitset_stream *stream;
int stream_enabled;
bitfield_word_t bits[];
pthread_mutex_t lock;
uint64_t size;
int resolution;
struct bitset_stream *stream;
int stream_enabled;
bitfield_word_t bits[];
};
/** Allocate a bitset for a file of the given size, and chunks of the
* given resolution.
*/
static inline struct bitset *bitset_alloc( uint64_t size, int resolution )
static inline struct bitset *bitset_alloc(uint64_t size, int resolution)
{
// calculate a size to allocate that is a multiple of the size of the
// bitfield word
size_t bitfield_size =
BIT_WORDS_FOR_SIZE((( size + resolution - 1 ) / resolution)) * sizeof( bitfield_word_t );
struct bitset *bitset = xmalloc(sizeof( struct bitset ) + ( bitfield_size / 8 ) );
// calculate a size to allocate that is a multiple of the size of the
// bitfield word
size_t bitfield_size =
BIT_WORDS_FOR_SIZE(((size + resolution -
1) / resolution)) * sizeof(bitfield_word_t);
struct bitset *bitset =
xmalloc(sizeof(struct bitset) + (bitfield_size / 8));
bitset->size = size;
bitset->resolution = resolution;
/* don't actually need to call pthread_mutex_destroy '*/
pthread_mutex_init(&bitset->lock, NULL);
bitset->stream = xmalloc( sizeof( struct bitset_stream ) );
pthread_mutex_init( &bitset->stream->mutex, NULL );
bitset->size = size;
bitset->resolution = resolution;
/* don't actually need to call pthread_mutex_destroy ' */
pthread_mutex_init(&bitset->lock, NULL);
bitset->stream = xmalloc(sizeof(struct bitset_stream));
pthread_mutex_init(&bitset->stream->mutex, NULL);
/* Technically don't need to call pthread_cond_destroy either */
pthread_cond_init( &bitset->stream->cond_not_full, NULL );
pthread_cond_init( &bitset->stream->cond_not_empty, NULL );
/* Technically don't need to call pthread_cond_destroy either */
pthread_cond_init(&bitset->stream->cond_not_full, NULL);
pthread_cond_init(&bitset->stream->cond_not_empty, NULL);
return bitset;
return bitset;
}
static inline void bitset_free( struct bitset * set )
static inline void bitset_free(struct bitset *set)
{
/* TODO: free our mutex... */
/* TODO: free our mutex... */
free( set->stream );
set->stream = NULL;
free(set->stream);
set->stream = NULL;
free( set );
free(set);
}
#define INT_FIRST_AND_LAST \
@@ -224,215 +241,201 @@ static inline void bitset_free( struct bitset * set )
FATAL_IF_NEGATIVE(pthread_mutex_unlock(&set->lock), "Error unlocking bitset")
static inline void bitset_stream_enqueue(
struct bitset * set,
enum bitset_stream_events event,
uint64_t from,
uint64_t len
)
static inline void bitset_stream_enqueue(struct bitset *set,
enum bitset_stream_events event,
uint64_t from, uint64_t len)
{
struct bitset_stream * stream = set->stream;
struct bitset_stream *stream = set->stream;
pthread_mutex_lock( &stream->mutex );
pthread_mutex_lock(&stream->mutex);
while ( stream->size == BITSET_STREAM_SIZE ) {
pthread_cond_wait( &stream->cond_not_full, &stream->mutex );
}
while (stream->size == BITSET_STREAM_SIZE) {
pthread_cond_wait(&stream->cond_not_full, &stream->mutex);
}
stream->entries[stream->in].event = event;
stream->entries[stream->in].from = from;
stream->entries[stream->in].len = len;
stream->queued_bytes[event] += len;
stream->entries[stream->in].event = event;
stream->entries[stream->in].from = from;
stream->entries[stream->in].len = len;
stream->queued_bytes[event] += len;
stream->size++;
stream->in++;
stream->in %= BITSET_STREAM_SIZE;
stream->size++;
stream->in++;
stream->in %= BITSET_STREAM_SIZE;
pthread_mutex_unlock( & stream->mutex );
pthread_cond_signal( &stream->cond_not_empty );
pthread_mutex_unlock(&stream->mutex);
pthread_cond_signal(&stream->cond_not_empty);
return;
return;
}
static inline void bitset_stream_dequeue(
struct bitset * set,
struct bitset_stream_entry * out
)
static inline void bitset_stream_dequeue(struct bitset *set,
struct bitset_stream_entry *out)
{
struct bitset_stream * stream = set->stream;
struct bitset_stream_entry * dequeued;
struct bitset_stream *stream = set->stream;
struct bitset_stream_entry *dequeued;
pthread_mutex_lock( &stream->mutex );
pthread_mutex_lock(&stream->mutex);
while ( stream->size == 0 ) {
pthread_cond_wait( &stream->cond_not_empty, &stream->mutex );
}
while (stream->size == 0) {
pthread_cond_wait(&stream->cond_not_empty, &stream->mutex);
}
dequeued = &stream->entries[stream->out];
dequeued = &stream->entries[stream->out];
if ( out != NULL ) {
out->event = dequeued->event;
out->from = dequeued->from;
out->len = dequeued->len;
}
if (out != NULL) {
out->event = dequeued->event;
out->from = dequeued->from;
out->len = dequeued->len;
}
stream->queued_bytes[dequeued->event] -= dequeued->len;
stream->size--;
stream->out++;
stream->out %= BITSET_STREAM_SIZE;
stream->queued_bytes[dequeued->event] -= dequeued->len;
stream->size--;
stream->out++;
stream->out %= BITSET_STREAM_SIZE;
pthread_mutex_unlock( &stream->mutex );
pthread_cond_signal( &stream->cond_not_full );
pthread_mutex_unlock(&stream->mutex);
pthread_cond_signal(&stream->cond_not_full);
return;
return;
}
static inline size_t bitset_stream_size( struct bitset * set )
static inline size_t bitset_stream_size(struct bitset *set)
{
size_t size;
size_t size;
pthread_mutex_lock( &set->stream->mutex );
size = set->stream->size;
pthread_mutex_unlock( &set->stream->mutex );
pthread_mutex_lock(&set->stream->mutex);
size = set->stream->size;
pthread_mutex_unlock(&set->stream->mutex);
return size;
return size;
}
static inline uint64_t bitset_stream_queued_bytes(
struct bitset * set,
enum bitset_stream_events event
)
static inline uint64_t bitset_stream_queued_bytes(struct bitset *set,
enum bitset_stream_events
event)
{
uint64_t total;
uint64_t total;
pthread_mutex_lock( &set->stream->mutex );
total = set->stream->queued_bytes[event];
pthread_mutex_unlock( &set->stream->mutex );
pthread_mutex_lock(&set->stream->mutex);
total = set->stream->queued_bytes[event];
pthread_mutex_unlock(&set->stream->mutex);
return total;
return total;
}
static inline void bitset_enable_stream( struct bitset * set )
static inline void bitset_enable_stream(struct bitset *set)
{
BITSET_LOCK;
set->stream_enabled = 1;
bitset_stream_enqueue( set, BITSET_STREAM_ON, 0, set->size );
BITSET_UNLOCK;
BITSET_LOCK;
set->stream_enabled = 1;
bitset_stream_enqueue(set, BITSET_STREAM_ON, 0, set->size);
BITSET_UNLOCK;
}
static inline void bitset_disable_stream( struct bitset * set )
static inline void bitset_disable_stream(struct bitset *set)
{
BITSET_LOCK;
bitset_stream_enqueue( set, BITSET_STREAM_OFF, 0, set->size );
set->stream_enabled = 0;
BITSET_UNLOCK;
BITSET_LOCK;
bitset_stream_enqueue(set, BITSET_STREAM_OFF, 0, set->size);
set->stream_enabled = 0;
BITSET_UNLOCK;
}
/** Set the bits in a bitset which correspond to the given bytes in the larger
* file.
*/
static inline void bitset_set_range(
struct bitset * set,
uint64_t from,
uint64_t len)
static inline void bitset_set_range(struct bitset *set,
uint64_t from, uint64_t len)
{
INT_FIRST_AND_LAST;
BITSET_LOCK;
bit_set_range(set->bits, first, bitlen);
INT_FIRST_AND_LAST;
BITSET_LOCK;
bit_set_range(set->bits, first, bitlen);
if ( set->stream_enabled ) {
bitset_stream_enqueue( set, BITSET_STREAM_SET, from, len );
}
if (set->stream_enabled) {
bitset_stream_enqueue(set, BITSET_STREAM_SET, from, len);
}
BITSET_UNLOCK;
BITSET_UNLOCK;
}
/** Set every bit in the bitset. */
static inline void bitset_set( struct bitset * set )
static inline void bitset_set(struct bitset *set)
{
bitset_set_range(set, 0, set->size);
bitset_set_range(set, 0, set->size);
}
/** Clear the bits in a bitset which correspond to the given bytes in the
* larger file.
*/
static inline void bitset_clear_range(
struct bitset * set,
uint64_t from,
uint64_t len)
static inline void bitset_clear_range(struct bitset *set,
uint64_t from, uint64_t len)
{
INT_FIRST_AND_LAST;
BITSET_LOCK;
bit_clear_range(set->bits, first, bitlen);
INT_FIRST_AND_LAST;
BITSET_LOCK;
bit_clear_range(set->bits, first, bitlen);
if ( set->stream_enabled ) {
bitset_stream_enqueue( set, BITSET_STREAM_UNSET, from, len );
}
if (set->stream_enabled) {
bitset_stream_enqueue(set, BITSET_STREAM_UNSET, from, len);
}
BITSET_UNLOCK;
BITSET_UNLOCK;
}
/** Clear every bit in the bitset. */
static inline void bitset_clear( struct bitset * set )
static inline void bitset_clear(struct bitset *set)
{
bitset_clear_range(set, 0, set->size);
bitset_clear_range(set, 0, set->size);
}
/** As per bitset_run_count but also tells you whether the run it found was set
* or unset, atomically.
*/
static inline uint64_t bitset_run_count_ex(
struct bitset * set,
uint64_t from,
uint64_t len,
int* run_is_set
)
static inline uint64_t bitset_run_count_ex(struct bitset *set,
uint64_t from,
uint64_t len, int *run_is_set)
{
uint64_t run;
uint64_t run;
/* Clip our requests to the end of the bitset, avoiding uint underflow. */
if ( from > set->size ) {
return 0;
}
len = ( len + from ) > set->size ? ( set->size - from ) : len;
/* Clip our requests to the end of the bitset, avoiding uint underflow. */
if (from > set->size) {
return 0;
}
len = (len + from) > set->size ? (set->size - from) : len;
INT_FIRST_AND_LAST;
INT_FIRST_AND_LAST;
BITSET_LOCK;
run = bit_run_count(set->bits, first, bitlen, run_is_set) * set->resolution;
run -= (from % set->resolution);
BITSET_UNLOCK;
BITSET_LOCK;
run =
bit_run_count(set->bits, first, bitlen,
run_is_set) * set->resolution;
run -= (from % set->resolution);
BITSET_UNLOCK;
return run;
return run;
}
/** Counts the number of contiguous bytes that are represented as a run in
* the bit field.
*/
static inline uint64_t bitset_run_count(
struct bitset * set,
uint64_t from,
uint64_t len)
static inline uint64_t bitset_run_count(struct bitset *set,
uint64_t from, uint64_t len)
{
return bitset_run_count_ex( set, from, len, NULL );
return bitset_run_count_ex(set, from, len, NULL);
}
/** Tests whether the bit field is clear for the given file offset.
*/
static inline int bitset_is_clear_at( struct bitset * set, uint64_t at )
static inline int bitset_is_clear_at(struct bitset *set, uint64_t at)
{
return bit_is_clear(set->bits, at/set->resolution);
return bit_is_clear(set->bits, at / set->resolution);
}
/** Tests whether the bit field is set for the given file offset.
*/
static inline int bitset_is_set_at( struct bitset * set, uint64_t at )
static inline int bitset_is_set_at(struct bitset *set, uint64_t at)
{
return bit_is_set(set->bits, at/set->resolution);
return bit_is_set(set->bits, at / set->resolution);
}
#endif

File diff suppressed because it is too large Load Diff

View File

@@ -3,6 +3,7 @@
#include <signal.h>
#include <time.h>
#include <inttypes.h>
/** CLIENT_HANDLER_TIMEOUT
* This is the length of time (in seconds) any request can be outstanding for.
@@ -18,39 +19,40 @@
struct client {
/* When we call pthread_join, if the thread is already dead
* we can get an ESRCH. Since we have no other way to tell
* if that ESRCH is from a dead thread or a thread that never
* existed, we use a `stopped` flag to indicate a thread which
* did exist, but went away. Only check this after a
* pthread_join call.
*/
int stopped;
int socket;
/* When we call pthread_join, if the thread is already dead
* we can get an ESRCH. Since we have no other way to tell
* if that ESRCH is from a dead thread or a thread that never
* existed, we use a `stopped` flag to indicate a thread which
* did exist, but went away. Only check this after a
* pthread_join call.
*/
int stopped;
int socket;
int fileno;
char* mapped;
int fileno;
char *mapped;
struct self_pipe * stop_signal;
uint64_t mapped_size;
struct server* serve; /* FIXME: remove above duplication */
struct self_pipe *stop_signal;
/* Have we seen a REQUEST_DISCONNECT message? */
int disconnect;
struct server *serve; /* FIXME: remove above duplication */
/* kill the whole server if a request has been outstanding too long,
* assuming use_killswitch is set in serve
*/
timer_t killswitch;
/* Have we seen a REQUEST_DISCONNECT message? */
int disconnect;
/* kill the whole server if a request has been outstanding too long,
* assuming use_killswitch is set in serve
*/
timer_t killswitch;
};
void client_killswitch_hit(int signal, siginfo_t *info, void *ptr);
void client_killswitch_hit(int signal, siginfo_t * info, void *ptr);
void* client_serve(void* client_uncast);
struct client * client_create( struct server * serve, int socket );
void client_destroy( struct client * client );
void client_signal_stop( struct client * client );
void *client_serve(void *client_uncast);
struct client *client_create(struct server *serve, int socket);
void client_destroy(struct client *client);
void client_signal_stop(struct client *client);
#endif

View File

@@ -44,590 +44,570 @@
#include <unistd.h>
struct control * control_create(
struct flexnbd * flexnbd,
const char * csn)
struct control *control_create(struct flexnbd *flexnbd, const char *csn)
{
struct control * control = xmalloc( sizeof( struct control ) );
struct control *control = xmalloc(sizeof(struct control));
NULLCHECK( csn );
NULLCHECK(csn);
control->flexnbd = flexnbd;
control->socket_name = csn;
control->open_signal = self_pipe_create();
control->close_signal = self_pipe_create();
control->mirror_state_mbox = mbox_create();
control->flexnbd = flexnbd;
control->socket_name = csn;
control->open_signal = self_pipe_create();
control->close_signal = self_pipe_create();
control->mirror_state_mbox = mbox_create();
return control;
return control;
}
void control_signal_close( struct control * control)
void control_signal_close(struct control *control)
{
NULLCHECK( control );
self_pipe_signal( control->close_signal );
NULLCHECK(control);
self_pipe_signal(control->close_signal);
}
void control_destroy( struct control * control )
void control_destroy(struct control *control)
{
NULLCHECK( control );
NULLCHECK(control);
mbox_destroy( control->mirror_state_mbox );
self_pipe_destroy( control->close_signal );
self_pipe_destroy( control->open_signal );
free( control );
mbox_destroy(control->mirror_state_mbox);
self_pipe_destroy(control->close_signal);
self_pipe_destroy(control->open_signal);
free(control);
}
struct control_client * control_client_create(
struct flexnbd * flexnbd,
int client_fd ,
struct mbox * state_mbox )
struct control_client *control_client_create(struct flexnbd *flexnbd,
int client_fd,
struct mbox *state_mbox)
{
NULLCHECK( flexnbd );
NULLCHECK(flexnbd);
struct control_client * control_client =
xmalloc( sizeof( struct control_client ) );
struct control_client *control_client =
xmalloc(sizeof(struct control_client));
control_client->socket = client_fd;
control_client->flexnbd = flexnbd;
control_client->mirror_state_mbox = state_mbox;
return control_client;
control_client->socket = client_fd;
control_client->flexnbd = flexnbd;
control_client->mirror_state_mbox = state_mbox;
return control_client;
}
void control_client_destroy( struct control_client * client )
void control_client_destroy(struct control_client *client)
{
NULLCHECK( client );
free( client );
NULLCHECK(client);
free(client);
}
void control_respond(struct control_client * client);
void control_respond(struct control_client *client);
void control_handle_client( struct control * control, int client_fd )
void control_handle_client(struct control *control, int client_fd)
{
NULLCHECK( control );
NULLCHECK( control->flexnbd );
struct control_client * control_client =
control_client_create(
control->flexnbd,
client_fd ,
control->mirror_state_mbox);
NULLCHECK(control);
NULLCHECK(control->flexnbd);
struct control_client *control_client =
control_client_create(control->flexnbd,
client_fd,
control->mirror_state_mbox);
/* We intentionally don't spawn a thread for the client here.
* This is to avoid having more than one thread potentially
* waiting on the migration commit status.
*/
control_respond( control_client );
/* We intentionally don't spawn a thread for the client here.
* This is to avoid having more than one thread potentially
* waiting on the migration commit status.
*/
control_respond(control_client);
}
void control_accept_client( struct control * control )
void control_accept_client(struct control *control)
{
int client_fd;
union mysockaddr client_address;
socklen_t addrlen = sizeof( union mysockaddr );
int client_fd;
union mysockaddr client_address;
socklen_t addrlen = sizeof(union mysockaddr);
client_fd = accept( control->control_fd, &client_address.generic, &addrlen );
FATAL_IF( -1 == client_fd, "control accept failed" );
client_fd =
accept(control->control_fd, &client_address.generic, &addrlen);
FATAL_IF(-1 == client_fd, "control accept failed");
control_handle_client( control, client_fd );
control_handle_client(control, client_fd);
}
int control_accept( struct control * control )
int control_accept(struct control *control)
{
NULLCHECK( control );
NULLCHECK(control);
fd_set fds;
fd_set fds;
FD_ZERO( &fds );
FD_SET( control->control_fd, &fds );
self_pipe_fd_set( control->close_signal, &fds );
debug("Control thread selecting");
FATAL_UNLESS( 0 < select( FD_SETSIZE, &fds, NULL, NULL, NULL ),
"Control select failed." );
FD_ZERO(&fds);
FD_SET(control->control_fd, &fds);
self_pipe_fd_set(control->close_signal, &fds);
debug("Control thread selecting");
FATAL_UNLESS(0 < select(FD_SETSIZE, &fds, NULL, NULL, NULL),
"Control select failed.");
if ( self_pipe_fd_isset( control->close_signal, &fds ) ){
return 0;
}
if (self_pipe_fd_isset(control->close_signal, &fds)) {
return 0;
}
if ( FD_ISSET( control->control_fd, &fds ) ) {
control_accept_client( control );
}
return 1;
if (FD_ISSET(control->control_fd, &fds)) {
control_accept_client(control);
}
return 1;
}
void control_accept_loop( struct control * control )
void control_accept_loop(struct control *control)
{
while( control_accept( control ) );
while (control_accept(control));
}
int open_control_socket( const char * socket_name )
int open_control_socket(const char *socket_name)
{
struct sockaddr_un bind_address;
int control_fd;
struct sockaddr_un bind_address;
int control_fd;
if (!socket_name) {
fatal( "Tried to open a control socket without a socket name" );
}
if (!socket_name) {
fatal("Tried to open a control socket without a socket name");
}
control_fd = socket(AF_UNIX, SOCK_STREAM, 0);
FATAL_IF_NEGATIVE(control_fd ,
"Couldn't create control socket");
control_fd = socket(AF_UNIX, SOCK_STREAM, 0);
FATAL_IF_NEGATIVE(control_fd, "Couldn't create control socket");
memset(&bind_address, 0, sizeof(struct sockaddr_un));
bind_address.sun_family = AF_UNIX;
strncpy(bind_address.sun_path, socket_name, sizeof(bind_address.sun_path)-1);
memset(&bind_address, 0, sizeof(struct sockaddr_un));
bind_address.sun_family = AF_UNIX;
strncpy(bind_address.sun_path, socket_name,
sizeof(bind_address.sun_path) - 1);
//unlink(socket_name); /* ignore failure */
//unlink(socket_name); /* ignore failure */
FATAL_IF_NEGATIVE(
bind(control_fd , &bind_address, sizeof(bind_address)),
"Couldn't bind control socket to %s: %s",
socket_name, strerror( errno )
FATAL_IF_NEGATIVE(bind
(control_fd, &bind_address, sizeof(bind_address)),
"Couldn't bind control socket to %s: %s",
socket_name, strerror(errno)
);
FATAL_IF_NEGATIVE(
listen(control_fd , 5),
"Couldn't listen on control socket"
);
return control_fd;
FATAL_IF_NEGATIVE(listen(control_fd, 5),
"Couldn't listen on control socket");
return control_fd;
}
void control_listen(struct control* control)
void control_listen(struct control *control)
{
NULLCHECK( control );
control->control_fd = open_control_socket( control->socket_name );
NULLCHECK(control);
control->control_fd = open_control_socket(control->socket_name);
}
void control_wait_for_open_signal( struct control * control )
void control_wait_for_open_signal(struct control *control)
{
fd_set fds;
FD_ZERO( &fds );
self_pipe_fd_set( control->open_signal, &fds );
FATAL_IF_NEGATIVE( select( FD_SETSIZE, &fds, NULL, NULL, NULL ),
"select() failed" );
fd_set fds;
FD_ZERO(&fds);
self_pipe_fd_set(control->open_signal, &fds);
FATAL_IF_NEGATIVE(select(FD_SETSIZE, &fds, NULL, NULL, NULL),
"select() failed");
self_pipe_signal_clear( control->open_signal );
self_pipe_signal_clear(control->open_signal);
}
void control_serve( struct control * control )
void control_serve(struct control *control)
{
NULLCHECK( control );
NULLCHECK(control);
control_wait_for_open_signal( control );
control_listen( control );
while( control_accept( control ) );
control_wait_for_open_signal(control);
control_listen(control);
while (control_accept(control));
}
void control_cleanup(
struct control * control,
int fatal __attribute__((unused)) )
void control_cleanup(struct control *control,
int fatal __attribute__ ((unused)))
{
NULLCHECK( control );
unlink( control->socket_name );
close( control->control_fd );
NULLCHECK(control);
unlink(control->socket_name);
close(control->control_fd);
}
void * control_runner( void * control_uncast )
void *control_runner(void *control_uncast)
{
debug("Control thread");
NULLCHECK( control_uncast );
struct control * control = (struct control *)control_uncast;
debug("Control thread");
NULLCHECK(control_uncast);
struct control *control = (struct control *) control_uncast;
error_set_handler( (cleanup_handler*)control_cleanup, control );
error_set_handler((cleanup_handler *) control_cleanup, control);
control_serve( control );
control_serve(control);
control_cleanup( control, 0 );
pthread_exit( NULL );
control_cleanup(control, 0);
pthread_exit(NULL);
}
#define write_socket(msg) write(client_fd, (msg "\n"), strlen((msg))+1)
void control_write_mirror_response( enum mirror_state mirror_state, int client_fd )
void control_write_mirror_response(enum mirror_state mirror_state,
int client_fd)
{
switch (mirror_state) {
case MS_INIT:
case MS_UNKNOWN:
write_socket( "1: Mirror failed to initialise" );
fatal( "Impossible mirror state: %d", mirror_state );
case MS_FAIL_CONNECT:
write_socket( "1: Mirror failed to connect");
break;
case MS_FAIL_REJECTED:
write_socket( "1: Mirror was rejected" );
break;
case MS_FAIL_NO_HELLO:
write_socket( "1: Remote server failed to respond");
break;
case MS_FAIL_SIZE_MISMATCH:
write_socket( "1: Remote size does not match local size" );
break;
case MS_ABANDONED:
write_socket( "1: Mirroring abandoned" );
break;
case MS_GO:
case MS_DONE: /* Yes, I know we know better, but it's simpler this way */
write_socket( "0: Mirror started" );
break;
default:
fatal( "Unhandled mirror state: %d", mirror_state );
}
switch (mirror_state) {
case MS_INIT:
case MS_UNKNOWN:
write_socket("1: Mirror failed to initialise");
fatal("Impossible mirror state: %d", mirror_state);
case MS_FAIL_CONNECT:
write_socket("1: Mirror failed to connect");
break;
case MS_FAIL_REJECTED:
write_socket("1: Mirror was rejected");
break;
case MS_FAIL_NO_HELLO:
write_socket("1: Remote server failed to respond");
break;
case MS_FAIL_SIZE_MISMATCH:
write_socket("1: Remote size does not match local size");
break;
case MS_ABANDONED:
write_socket("1: Mirroring abandoned");
break;
case MS_GO:
case MS_DONE: /* Yes, I know we know better, but it's simpler this way */
write_socket("0: Mirror started");
break;
default:
fatal("Unhandled mirror state: %d", mirror_state);
}
}
#undef write_socket
/* Call this in the thread where you want to receive the mirror state */
enum mirror_state control_client_mirror_wait(
struct control_client* client)
enum mirror_state control_client_mirror_wait(struct control_client *client)
{
NULLCHECK( client );
NULLCHECK( client->mirror_state_mbox );
NULLCHECK(client);
NULLCHECK(client->mirror_state_mbox);
struct mbox * mbox = client->mirror_state_mbox;
enum mirror_state mirror_state;
enum mirror_state * contents;
struct mbox *mbox = client->mirror_state_mbox;
enum mirror_state mirror_state;
enum mirror_state *contents;
contents = (enum mirror_state*)mbox_receive( mbox );
NULLCHECK( contents );
contents = (enum mirror_state *) mbox_receive(mbox);
NULLCHECK(contents);
mirror_state = *contents;
mirror_state = *contents;
free( contents );
free(contents);
return mirror_state;
return mirror_state;
}
#define write_socket(msg) write(client->socket, (msg "\n"), strlen((msg))+1)
/** Command parser to start mirror process from socket input */
int control_mirror(struct control_client* client, int linesc, char** lines)
int control_mirror(struct control_client *client, int linesc, char **lines)
{
NULLCHECK( client );
NULLCHECK(client);
struct flexnbd * flexnbd = client->flexnbd;
union mysockaddr *connect_to = xmalloc( sizeof( union mysockaddr ) );
union mysockaddr *connect_from = NULL;
uint64_t max_Bps = UINT64_MAX;
int action_at_finish;
int raw_port;
struct flexnbd *flexnbd = client->flexnbd;
union mysockaddr *connect_to = xmalloc(sizeof(union mysockaddr));
union mysockaddr *connect_from = NULL;
uint64_t max_Bps = UINT64_MAX;
int action_at_finish;
int raw_port;
if (linesc < 2) {
write_socket("1: mirror takes at least two parameters");
return -1;
if (linesc < 2) {
write_socket("1: mirror takes at least two parameters");
return -1;
}
if (parse_ip_to_sockaddr(&connect_to->generic, lines[0]) == 0) {
write_socket("1: bad IP address");
return -1;
}
raw_port = atoi(lines[1]);
if (raw_port < 0 || raw_port > 65535) {
write_socket("1: bad IP port number");
return -1;
}
connect_to->v4.sin_port = htobe16(raw_port);
action_at_finish = ACTION_EXIT;
if (linesc > 2) {
if (strcmp("exit", lines[2]) == 0) {
action_at_finish = ACTION_EXIT;
} else if (strcmp("unlink", lines[2]) == 0) {
action_at_finish = ACTION_UNLINK;
} else if (strcmp("nothing", lines[2]) == 0) {
action_at_finish = ACTION_NOTHING;
} else {
write_socket("1: action must be 'exit' or 'nothing'");
return -1;
}
}
if (linesc > 3) {
connect_from = xmalloc(sizeof(union mysockaddr));
if (parse_ip_to_sockaddr(&connect_from->generic, lines[3]) == 0) {
write_socket("1: bad bind address");
return -1;
}
}
if (linesc > 4) {
errno = 0;
max_Bps = strtoull(lines[4], NULL, 10);
if (errno == ERANGE) {
write_socket("1: max_bps out of range");
return -1;
} else if (errno != 0) {
write_socket("1: max_bps couldn't be parsed");
return -1;
}
}
if (linesc > 5) {
write_socket("1: unrecognised parameters to mirror");
return -1;
}
struct server *serve = flexnbd_server(flexnbd);
server_lock_start_mirror(serve);
{
if (server_mirror_can_start(serve)) {
serve->mirror_super = mirror_super_create(serve->filename,
connect_to,
connect_from,
max_Bps,
action_at_finish,
client->
mirror_state_mbox);
serve->mirror = serve->mirror_super->mirror;
server_prevent_mirror_start(serve);
} else {
if (serve->mirror_super) {
warn("Tried to start a second mirror run");
write_socket("1: mirror already running");
} else {
warn("Cannot start mirroring, shutting down");
write_socket("1: shutting down");
}
}
if (parse_ip_to_sockaddr(&connect_to->generic, lines[0]) == 0) {
write_socket("1: bad IP address");
return -1;
}
}
server_unlock_start_mirror(serve);
raw_port = atoi(lines[1]);
if (raw_port < 0 || raw_port > 65535) {
write_socket("1: bad IP port number");
return -1;
}
connect_to->v4.sin_port = htobe16(raw_port);
/* Do this outside the lock to minimise the length of time the
* sighandler can block the serve thread
*/
if (serve->mirror_super) {
FATAL_IF(0 != pthread_create(&serve->mirror_super->thread,
NULL,
mirror_super_runner,
serve),
"Failed to create mirror thread");
action_at_finish = ACTION_EXIT;
if (linesc > 2) {
if (strcmp("exit", lines[2]) == 0) {
action_at_finish = ACTION_EXIT;
}
else if (strcmp( "unlink", lines[2]) == 0 ) {
action_at_finish = ACTION_UNLINK;
}
else if (strcmp("nothing", lines[2]) == 0) {
action_at_finish = ACTION_NOTHING;
}
else {
write_socket("1: action must be 'exit' or 'nothing'");
return -1;
}
}
debug("Control thread mirror super waiting");
enum mirror_state state = control_client_mirror_wait(client);
debug("Control thread writing response");
control_write_mirror_response(state, client->socket);
}
if (linesc > 3) {
connect_from = xmalloc( sizeof( union mysockaddr ) );
if (parse_ip_to_sockaddr(&connect_from->generic, lines[3]) == 0) {
write_socket("1: bad bind address");
return -1;
}
}
debug("Control thread going away.");
if (linesc > 4) {
errno = 0;
max_Bps = strtoull( lines[4], NULL, 10 );
if ( errno == ERANGE ) {
write_socket( "1: max_bps out of range" );
return -1;
} else if ( errno != 0 ) {
write_socket( "1: max_bps couldn't be parsed" );
return -1;
}
}
if (linesc > 5) {
write_socket("1: unrecognised parameters to mirror");
return -1;
}
struct server * serve = flexnbd_server(flexnbd);
server_lock_start_mirror( serve );
{
if ( server_mirror_can_start( serve ) ) {
serve->mirror_super = mirror_super_create(
serve->filename,
connect_to,
connect_from,
max_Bps ,
action_at_finish,
client->mirror_state_mbox );
serve->mirror = serve->mirror_super->mirror;
server_prevent_mirror_start( serve );
} else {
if ( serve->mirror_super ) {
warn( "Tried to start a second mirror run" );
write_socket( "1: mirror already running" );
} else {
warn( "Cannot start mirroring, shutting down" );
write_socket( "1: shutting down" );
}
}
}
server_unlock_start_mirror( serve );
/* Do this outside the lock to minimise the length of time the
* sighandler can block the serve thread
*/
if ( serve->mirror_super ) {
FATAL_IF( 0 != pthread_create(
&serve->mirror_super->thread,
NULL,
mirror_super_runner,
serve
),
"Failed to create mirror thread"
);
debug("Control thread mirror super waiting");
enum mirror_state state =
control_client_mirror_wait( client );
debug("Control thread writing response");
control_write_mirror_response( state, client->socket );
}
debug( "Control thread going away." );
return 0;
return 0;
}
int control_mirror_max_bps( struct control_client* client, int linesc, char** lines )
int control_mirror_max_bps(struct control_client *client, int linesc,
char **lines)
{
NULLCHECK( client );
NULLCHECK( client->flexnbd );
NULLCHECK(client);
NULLCHECK(client->flexnbd);
struct server* serve = flexnbd_server( client->flexnbd );
uint64_t max_Bps;
struct server *serve = flexnbd_server(client->flexnbd);
uint64_t max_Bps;
if ( !serve->mirror_super ) {
write_socket( "1: Not currently mirroring" );
return -1;
}
if (!serve->mirror_super) {
write_socket("1: Not currently mirroring");
return -1;
}
if ( linesc != 1 ) {
write_socket( "1: Bad format" );
return -1;
}
if (linesc != 1) {
write_socket("1: Bad format");
return -1;
}
errno = 0;
max_Bps = strtoull( lines[0], NULL, 10 );
if ( errno == ERANGE ) {
write_socket( "1: max_bps out of range" );
return -1;
} else if ( errno != 0 ) {
write_socket( "1: max_bps couldn't be parsed" );
return -1;
}
errno = 0;
max_Bps = strtoull(lines[0], NULL, 10);
if (errno == ERANGE) {
write_socket("1: max_bps out of range");
return -1;
} else if (errno != 0) {
write_socket("1: max_bps couldn't be parsed");
return -1;
}
serve->mirror->max_bytes_per_second = max_Bps;
write_socket( "0: updated" );
serve->mirror->max_bytes_per_second = max_Bps;
write_socket("0: updated");
return 0;
return 0;
}
#undef write_socket
/** Command parser to alter access control list from socket input */
int control_acl(struct control_client* client, int linesc, char** lines)
int control_acl(struct control_client *client, int linesc, char **lines)
{
NULLCHECK( client );
NULLCHECK( client->flexnbd );
struct flexnbd * flexnbd = client->flexnbd;
NULLCHECK(client);
NULLCHECK(client->flexnbd);
struct flexnbd *flexnbd = client->flexnbd;
int default_deny = flexnbd_default_deny( flexnbd );
struct acl * new_acl = acl_create( linesc, lines, default_deny );
int default_deny = flexnbd_default_deny(flexnbd);
struct acl *new_acl = acl_create(linesc, lines, default_deny);
if (new_acl->len != linesc) {
warn("Bad ACL spec: %s", lines[new_acl->len] );
write(client->socket, "1: bad spec: ", 13);
write(client->socket, lines[new_acl->len],
strlen(lines[new_acl->len]));
write(client->socket, "\n", 1);
acl_destroy( new_acl );
}
else {
flexnbd_replace_acl( flexnbd, new_acl );
info("ACL set");
write( client->socket, "0: updated\n", 11);
}
if (new_acl->len != linesc) {
warn("Bad ACL spec: %s", lines[new_acl->len]);
write(client->socket, "1: bad spec: ", 13);
write(client->socket, lines[new_acl->len],
strlen(lines[new_acl->len]));
write(client->socket, "\n", 1);
acl_destroy(new_acl);
} else {
flexnbd_replace_acl(flexnbd, new_acl);
info("ACL set");
write(client->socket, "0: updated\n", 11);
}
return 0;
return 0;
}
int control_break(
struct control_client* client,
int linesc __attribute__ ((unused)),
char** lines __attribute__((unused))
)
int control_break(struct control_client *client,
int linesc __attribute__ ((unused)),
char **lines __attribute__ ((unused))
)
{
NULLCHECK( client );
NULLCHECK( client->flexnbd );
NULLCHECK(client);
NULLCHECK(client->flexnbd);
int result = 0;
struct flexnbd* flexnbd = client->flexnbd;
int result = 0;
struct flexnbd *flexnbd = client->flexnbd;
struct server * serve = flexnbd_server( flexnbd );
struct server *serve = flexnbd_server(flexnbd);
server_lock_start_mirror( serve );
{
if ( server_is_mirroring( serve ) ) {
server_lock_start_mirror(serve);
{
if (server_is_mirroring(serve)) {
info( "Signaling to abandon mirror" );
server_abandon_mirror( serve );
debug( "Abandon signaled" );
info("Signaling to abandon mirror");
server_abandon_mirror(serve);
debug("Abandon signaled");
if ( server_is_closed( serve ) ) {
info( "Mirror completed while canceling" );
write( client->socket,
"1: mirror completed\n", 20 );
}
else {
info( "Mirror successfully stopped." );
write( client->socket,
"0: mirror stopped\n", 18 );
result = 1;
}
if (server_is_closed(serve)) {
info("Mirror completed while canceling");
write(client->socket, "1: mirror completed\n", 20);
} else {
info("Mirror successfully stopped.");
write(client->socket, "0: mirror stopped\n", 18);
result = 1;
}
} else {
warn( "Not mirroring." );
write( client->socket, "1: not mirroring\n", 17 );
}
} else {
warn("Not mirroring.");
write(client->socket, "1: not mirroring\n", 17);
}
server_unlock_start_mirror( serve );
}
server_unlock_start_mirror(serve);
return result;
return result;
}
/** FIXME: add some useful statistics */
int control_status(
struct control_client* client,
int linesc __attribute__ ((unused)),
char** lines __attribute__((unused))
)
int control_status(struct control_client *client,
int linesc __attribute__ ((unused)),
char **lines __attribute__ ((unused))
)
{
NULLCHECK( client );
NULLCHECK( client->flexnbd );
struct status * status = flexnbd_status_create( client->flexnbd );
NULLCHECK(client);
NULLCHECK(client->flexnbd);
struct status *status = flexnbd_status_create(client->flexnbd);
write( client->socket, "0: ", 3 );
status_write( status, client->socket );
status_destroy( status );
write(client->socket, "0: ", 3);
status_write(status, client->socket);
status_destroy(status);
return 0;
return 0;
}
void control_client_cleanup(struct control_client* client,
int fatal __attribute__ ((unused)) )
void control_client_cleanup(struct control_client *client,
int fatal __attribute__ ((unused)))
{
if (client->socket) { close(client->socket); }
if (client->socket) {
close(client->socket);
}
/* This is wrongness */
if ( server_acl_locked( client->flexnbd->serve ) ) { server_unlock_acl( client->flexnbd->serve ); }
/* This is wrongness */
if (server_acl_locked(client->flexnbd->serve)) {
server_unlock_acl(client->flexnbd->serve);
}
control_client_destroy( client );
control_client_destroy(client);
}
/** Master command parser for control socket connections, delegates quickly */
void control_respond(struct control_client * client)
void control_respond(struct control_client *client)
{
char **lines = NULL;
char **lines = NULL;
error_set_handler((cleanup_handler*) control_client_cleanup, client);
error_set_handler((cleanup_handler *) control_client_cleanup, client);
int i, linesc;
linesc = read_lines_until_blankline(client->socket, 256, &lines);
int i, linesc;
linesc = read_lines_until_blankline(client->socket, 256, &lines);
if (linesc < 1)
{
write(client->socket, "9: missing command\n", 19);
/* ignore failure */
if (linesc < 1) {
write(client->socket, "9: missing command\n", 19);
/* ignore failure */
} else if (strcmp(lines[0], "acl") == 0) {
info("acl command received");
if (control_acl(client, linesc - 1, lines + 1) < 0) {
debug("acl command failed");
}
else if (strcmp(lines[0], "acl") == 0) {
info("acl command received" );
if (control_acl(client, linesc-1, lines+1) < 0) {
debug("acl command failed");
}
} else if (strcmp(lines[0], "mirror") == 0) {
info("mirror command received");
if (control_mirror(client, linesc - 1, lines + 1) < 0) {
debug("mirror command failed");
}
else if (strcmp(lines[0], "mirror") == 0) {
info("mirror command received" );
if (control_mirror(client, linesc-1, lines+1) < 0) {
debug("mirror command failed");
}
} else if (strcmp(lines[0], "break") == 0) {
info("break command received");
if (control_break(client, linesc - 1, lines + 1) < 0) {
debug("break command failed");
}
else if (strcmp(lines[0], "break") == 0) {
info( "break command received" );
if ( control_break( client, linesc-1, lines+1) < 0) {
debug( "break command failed" );
}
} else if (strcmp(lines[0], "status") == 0) {
info("status command received");
if (control_status(client, linesc - 1, lines + 1) < 0) {
debug("status command failed");
}
else if (strcmp(lines[0], "status") == 0) {
info("status command received" );
if (control_status(client, linesc-1, lines+1) < 0) {
debug("status command failed");
}
} else if ( strcmp( lines[0], "mirror_max_bps" ) == 0 ) {
info( "mirror_max_bps command received" );
if( control_mirror_max_bps( client, linesc-1, lines+1 ) < 0 ) {
debug( "mirror_max_bps command failed" );
}
}
else {
write(client->socket, "10: unknown command\n", 23);
} else if (strcmp(lines[0], "mirror_max_bps") == 0) {
info("mirror_max_bps command received");
if (control_mirror_max_bps(client, linesc - 1, lines + 1) < 0) {
debug("mirror_max_bps command failed");
}
} else {
write(client->socket, "10: unknown command\n", 23);
}
for (i=0; i<linesc; i++) {
free(lines[i]);
}
free(lines);
for (i = 0; i < linesc; i++) {
free(lines[i]);
}
free(lines);
control_client_cleanup(client, 0);
debug("control command handled" );
control_client_cleanup(client, 0);
debug("control command handled");
}

View File

@@ -13,47 +13,46 @@ struct server;
#include "mbox.h"
struct control {
struct flexnbd * flexnbd;
int control_fd;
const char * socket_name;
struct flexnbd *flexnbd;
int control_fd;
const char *socket_name;
pthread_t thread;
pthread_t thread;
struct self_pipe * open_signal;
struct self_pipe * close_signal;
struct self_pipe *open_signal;
struct self_pipe *close_signal;
/* This is owned by the control object, and used by a
* mirror_super to communicate the state of a mirror attempt as
* early as feasible. It can't be owned by the mirror_super
* object because the mirror_super object can be freed at any
* time (including while the control_client is waiting on it),
* whereas the control object lasts for the lifetime of the
* process (and we can only have a mirror thread if the control
* thread has started it).
*/
struct mbox * mirror_state_mbox;
/* This is owned by the control object, and used by a
* mirror_super to communicate the state of a mirror attempt as
* early as feasible. It can't be owned by the mirror_super
* object because the mirror_super object can be freed at any
* time (including while the control_client is waiting on it),
* whereas the control object lasts for the lifetime of the
* process (and we can only have a mirror thread if the control
* thread has started it).
*/
struct mbox *mirror_state_mbox;
};
struct control_client{
int socket;
struct flexnbd * flexnbd;
struct control_client {
int socket;
struct flexnbd *flexnbd;
/* Passed in on creation. We know it's all right to do this
* because we know there's only ever one control_client.
*/
struct mbox * mirror_state_mbox;
/* Passed in on creation. We know it's all right to do this
* because we know there's only ever one control_client.
*/
struct mbox *mirror_state_mbox;
};
struct control * control_create(
struct flexnbd *,
const char * control_socket_name );
void control_signal_close( struct control * );
void control_destroy( struct control * );
struct control *control_create(struct flexnbd *,
const char *control_socket_name);
void control_signal_close(struct control *);
void control_destroy(struct control *);
void * control_runner( void * );
void *control_runner(void *);
void accept_control_connection(struct server* params, int client_fd, union mysockaddr* client_address);
void serve_open_control_socket(struct server* params);
void accept_control_connection(struct server *params, int client_fd,
union mysockaddr *client_address);
void serve_open_control_socket(struct server *params);
#endif

View File

@@ -43,223 +43,206 @@
int flexnbd_build_signal_fd(void)
{
sigset_t mask;
int sfd;
sigset_t mask;
int sfd;
sigemptyset( &mask );
sigaddset( &mask, SIGTERM );
sigaddset( &mask, SIGQUIT );
sigaddset( &mask, SIGINT );
sigemptyset(&mask);
sigaddset(&mask, SIGTERM);
sigaddset(&mask, SIGQUIT);
sigaddset(&mask, SIGINT);
FATAL_UNLESS( 0 == pthread_sigmask( SIG_BLOCK, &mask, NULL ),
"Signal blocking failed" );
FATAL_UNLESS(0 == pthread_sigmask(SIG_BLOCK, &mask, NULL),
"Signal blocking failed");
sfd = signalfd( -1, &mask, 0 );
FATAL_IF( -1 == sfd, "Failed to get a signal fd" );
sfd = signalfd(-1, &mask, 0);
FATAL_IF(-1 == sfd, "Failed to get a signal fd");
return sfd;
return sfd;
}
void flexnbd_create_shared(
struct flexnbd * flexnbd,
const char * s_ctrl_sock)
void flexnbd_create_shared(struct flexnbd *flexnbd,
const char *s_ctrl_sock)
{
NULLCHECK( flexnbd );
if ( s_ctrl_sock ){
flexnbd->control =
control_create( flexnbd, s_ctrl_sock );
}
else {
flexnbd->control = NULL;
}
NULLCHECK(flexnbd);
if (s_ctrl_sock) {
flexnbd->control = control_create(flexnbd, s_ctrl_sock);
} else {
flexnbd->control = NULL;
}
flexnbd->signal_fd = flexnbd_build_signal_fd();
flexnbd->signal_fd = flexnbd_build_signal_fd();
}
struct flexnbd * flexnbd_create_serving(
char* s_ip_address,
char* s_port,
char* s_file,
char* s_ctrl_sock,
int default_deny,
int acl_entries,
char** s_acl_entries,
int max_nbd_clients,
int use_killswitch)
struct flexnbd *flexnbd_create_serving(char *s_ip_address,
char *s_port,
char *s_file,
char *s_ctrl_sock,
int default_deny,
int acl_entries,
char **s_acl_entries,
int max_nbd_clients,
int use_killswitch)
{
struct flexnbd * flexnbd = xmalloc( sizeof( struct flexnbd ) );
flexnbd->serve = server_create(
flexnbd,
s_ip_address,
s_port,
s_file,
default_deny,
acl_entries,
s_acl_entries,
max_nbd_clients,
use_killswitch,
1);
flexnbd_create_shared( flexnbd, s_ctrl_sock );
struct flexnbd *flexnbd = xmalloc(sizeof(struct flexnbd));
flexnbd->serve = server_create(flexnbd,
s_ip_address,
s_port,
s_file,
default_deny,
acl_entries,
s_acl_entries,
max_nbd_clients, use_killswitch, 1);
flexnbd_create_shared(flexnbd, s_ctrl_sock);
// Beats installing one handler per client instance
if ( use_killswitch ) {
struct sigaction act = {
.sa_sigaction = client_killswitch_hit,
.sa_flags = SA_RESTART | SA_SIGINFO
};
// Beats installing one handler per client instance
if (use_killswitch) {
struct sigaction act = {
.sa_sigaction = client_killswitch_hit,
.sa_flags = SA_RESTART | SA_SIGINFO
};
FATAL_UNLESS(
0 == sigaction( CLIENT_KILLSWITCH_SIGNAL, &act, NULL ),
"Installing client killswitch signal failed"
);
}
FATAL_UNLESS(0 == sigaction(CLIENT_KILLSWITCH_SIGNAL, &act, NULL),
"Installing client killswitch signal failed");
}
return flexnbd;
return flexnbd;
}
struct flexnbd * flexnbd_create_listening(
char* s_ip_address,
char* s_port,
char* s_file,
char* s_ctrl_sock,
int default_deny,
int acl_entries,
char** s_acl_entries )
struct flexnbd *flexnbd_create_listening(char *s_ip_address,
char *s_port,
char *s_file,
char *s_ctrl_sock,
int default_deny,
int acl_entries,
char **s_acl_entries)
{
struct flexnbd * flexnbd = xmalloc( sizeof( struct flexnbd ) );
flexnbd->serve = server_create(
flexnbd,
s_ip_address,
s_port,
s_file,
default_deny,
acl_entries,
s_acl_entries,
1, 0, 0);
flexnbd_create_shared( flexnbd, s_ctrl_sock );
struct flexnbd *flexnbd = xmalloc(sizeof(struct flexnbd));
flexnbd->serve = server_create(flexnbd,
s_ip_address,
s_port,
s_file,
default_deny,
acl_entries, s_acl_entries, 1, 0, 0);
flexnbd_create_shared(flexnbd, s_ctrl_sock);
// listen can't use killswitch, as mirror may pause on sending things
// for a very long time.
// listen can't use killswitch, as mirror may pause on sending things
// for a very long time.
return flexnbd;
return flexnbd;
}
void flexnbd_spawn_control(struct flexnbd * flexnbd )
void flexnbd_spawn_control(struct flexnbd *flexnbd)
{
NULLCHECK( flexnbd );
NULLCHECK( flexnbd->control );
NULLCHECK(flexnbd);
NULLCHECK(flexnbd->control);
pthread_t * control_thread = &flexnbd->control->thread;
pthread_t *control_thread = &flexnbd->control->thread;
FATAL_UNLESS( 0 == pthread_create(
control_thread,
NULL,
control_runner,
flexnbd->control ),
"Couldn't create the control thread" );
FATAL_UNLESS(0 == pthread_create(control_thread,
NULL,
control_runner,
flexnbd->control),
"Couldn't create the control thread");
}
void flexnbd_stop_control( struct flexnbd * flexnbd )
void flexnbd_stop_control(struct flexnbd *flexnbd)
{
NULLCHECK( flexnbd );
NULLCHECK( flexnbd->control );
NULLCHECK(flexnbd);
NULLCHECK(flexnbd->control);
control_signal_close( flexnbd->control );
pthread_t tid = flexnbd->control->thread;
FATAL_UNLESS( 0 == pthread_join( tid, NULL ),
"Failed joining the control thread" );
debug( "Control thread %p pthread_join returned", tid );
control_signal_close(flexnbd->control);
pthread_t tid = flexnbd->control->thread;
FATAL_UNLESS(0 == pthread_join(tid, NULL),
"Failed joining the control thread");
debug("Control thread %p pthread_join returned", tid);
}
int flexnbd_signal_fd( struct flexnbd * flexnbd )
int flexnbd_signal_fd(struct flexnbd *flexnbd)
{
NULLCHECK( flexnbd );
return flexnbd->signal_fd;
NULLCHECK(flexnbd);
return flexnbd->signal_fd;
}
void flexnbd_destroy( struct flexnbd * flexnbd )
void flexnbd_destroy(struct flexnbd *flexnbd)
{
NULLCHECK( flexnbd );
if ( flexnbd->control ) {
control_destroy( flexnbd->control );
}
NULLCHECK(flexnbd);
if (flexnbd->control) {
control_destroy(flexnbd->control);
}
close( flexnbd->signal_fd );
free( flexnbd );
close(flexnbd->signal_fd);
free(flexnbd);
}
struct server * flexnbd_server( struct flexnbd * flexnbd )
struct server *flexnbd_server(struct flexnbd *flexnbd)
{
NULLCHECK( flexnbd );
return flexnbd->serve;
NULLCHECK(flexnbd);
return flexnbd->serve;
}
void flexnbd_replace_acl( struct flexnbd * flexnbd, struct acl * acl )
void flexnbd_replace_acl(struct flexnbd *flexnbd, struct acl *acl)
{
NULLCHECK( flexnbd );
server_replace_acl( flexnbd_server(flexnbd), acl );
NULLCHECK(flexnbd);
server_replace_acl(flexnbd_server(flexnbd), acl);
}
struct status * flexnbd_status_create( struct flexnbd * flexnbd )
struct status *flexnbd_status_create(struct flexnbd *flexnbd)
{
NULLCHECK( flexnbd );
struct status * status;
NULLCHECK(flexnbd);
struct status *status;
status = status_create( flexnbd_server( flexnbd ) );
return status;
status = status_create(flexnbd_server(flexnbd));
return status;
}
void flexnbd_set_server( struct flexnbd * flexnbd, struct server * serve )
void flexnbd_set_server(struct flexnbd *flexnbd, struct server *serve)
{
NULLCHECK( flexnbd );
flexnbd->serve = serve;
NULLCHECK(flexnbd);
flexnbd->serve = serve;
}
/* Get the default_deny of the current server object. */
int flexnbd_default_deny( struct flexnbd * flexnbd )
int flexnbd_default_deny(struct flexnbd *flexnbd)
{
NULLCHECK( flexnbd );
return server_default_deny( flexnbd->serve );
NULLCHECK(flexnbd);
return server_default_deny(flexnbd->serve);
}
void make_writable( const char * filename )
void make_writable(const char *filename)
{
NULLCHECK( filename );
FATAL_IF_NEGATIVE( chmod( filename, S_IWUSR ),
"Couldn't chmod %s: %s",
filename,
strerror( errno ) );
NULLCHECK(filename);
FATAL_IF_NEGATIVE(chmod(filename, S_IWUSR),
"Couldn't chmod %s: %s", filename, strerror(errno));
}
int flexnbd_serve( struct flexnbd * flexnbd )
int flexnbd_serve(struct flexnbd *flexnbd)
{
NULLCHECK( flexnbd );
int success;
struct self_pipe * open_signal = NULL;
NULLCHECK(flexnbd);
int success;
struct self_pipe *open_signal = NULL;
if ( flexnbd->control ){
debug( "Spawning control thread" );
flexnbd_spawn_control( flexnbd );
open_signal = flexnbd->control->open_signal;
}
if (flexnbd->control) {
debug("Spawning control thread");
flexnbd_spawn_control(flexnbd);
open_signal = flexnbd->control->open_signal;
}
success = do_serve( flexnbd->serve, open_signal );
debug("do_serve success is %d", success );
success = do_serve(flexnbd->serve, open_signal);
debug("do_serve success is %d", success);
if ( flexnbd->control ) {
debug( "Stopping control thread" );
flexnbd_stop_control( flexnbd );
debug("Control thread stopped");
}
if (flexnbd->control) {
debug("Stopping control thread");
flexnbd_stop_control(flexnbd);
debug("Control thread stopped");
}
return success;
return success;
}

View File

@@ -13,54 +13,51 @@
/* Carries the "globals". */
struct flexnbd {
/* Our serve pointer should never be dereferenced outside a
* flexnbd_switch_lock/unlock pair.
*/
struct server * serve;
/* Our serve pointer should never be dereferenced outside a
* flexnbd_switch_lock/unlock pair.
*/
struct server *serve;
/* We only have a control object if a control socket name was
* passed on the command line.
*/
struct control * control;
/* We only have a control object if a control socket name was
* passed on the command line.
*/
struct control *control;
/* File descriptor for a signalfd(2) signal stream. */
int signal_fd;
/* File descriptor for a signalfd(2) signal stream. */
int signal_fd;
};
struct flexnbd * flexnbd_create(void);
struct flexnbd * flexnbd_create_serving(
char* s_ip_address,
char* s_port,
char* s_file,
char* s_ctrl_sock,
int default_deny,
int acl_entries,
char** s_acl_entries,
int max_nbd_clients,
int use_killswitch);
struct flexnbd *flexnbd_create(void);
struct flexnbd *flexnbd_create_serving(char *s_ip_address,
char *s_port,
char *s_file,
char *s_ctrl_sock,
int default_deny,
int acl_entries,
char **s_acl_entries,
int max_nbd_clients,
int use_killswitch);
struct flexnbd * flexnbd_create_listening(
char* s_ip_address,
char* s_port,
char* s_file,
char* s_ctrl_sock,
int default_deny,
int acl_entries,
char** s_acl_entries );
struct flexnbd *flexnbd_create_listening(char *s_ip_address,
char *s_port,
char *s_file,
char *s_ctrl_sock,
int default_deny,
int acl_entries,
char **s_acl_entries);
void flexnbd_destroy( struct flexnbd * );
void flexnbd_destroy(struct flexnbd *);
enum mirror_state;
enum mirror_state flexnbd_get_mirror_state( struct flexnbd * );
int flexnbd_default_deny( struct flexnbd * );
void flexnbd_set_server( struct flexnbd * flexnbd, struct server * serve );
int flexnbd_signal_fd( struct flexnbd * flexnbd );
enum mirror_state flexnbd_get_mirror_state(struct flexnbd *);
int flexnbd_default_deny(struct flexnbd *);
void flexnbd_set_server(struct flexnbd *flexnbd, struct server *serve);
int flexnbd_signal_fd(struct flexnbd *flexnbd);
int flexnbd_serve( struct flexnbd * flexnbd );
int flexnbd_proxy( struct flexnbd * flexnbd );
struct server * flexnbd_server( struct flexnbd * flexnbd );
void flexnbd_replace_acl( struct flexnbd * flexnbd, struct acl * acl );
struct status * flexnbd_status_create( struct flexnbd * flexnbd );
int flexnbd_serve(struct flexnbd *flexnbd);
int flexnbd_proxy(struct flexnbd *flexnbd);
struct server *flexnbd_server(struct flexnbd *flexnbd);
void flexnbd_replace_acl(struct flexnbd *flexnbd, struct acl *acl);
struct status *flexnbd_status_create(struct flexnbd *flexnbd);
#endif

View File

@@ -4,72 +4,70 @@
#include <pthread.h>
struct flexthread_mutex * flexthread_mutex_create(void)
struct flexthread_mutex *flexthread_mutex_create(void)
{
struct flexthread_mutex * ftm =
xmalloc( sizeof( struct flexthread_mutex ) );
struct flexthread_mutex *ftm =
xmalloc(sizeof(struct flexthread_mutex));
FATAL_UNLESS( 0 == pthread_mutex_init( &ftm->mutex, NULL ),
"Mutex initialisation failed" );
return ftm;
FATAL_UNLESS(0 == pthread_mutex_init(&ftm->mutex, NULL),
"Mutex initialisation failed");
return ftm;
}
void flexthread_mutex_destroy( struct flexthread_mutex * ftm )
void flexthread_mutex_destroy(struct flexthread_mutex *ftm)
{
NULLCHECK( ftm );
NULLCHECK(ftm);
if( flexthread_mutex_held( ftm ) ) {
flexthread_mutex_unlock( ftm );
}
else if ( (pthread_t)NULL != ftm->holder ) {
/* This "should never happen": if we can try to destroy
* a mutex currently held by another thread, there's a
* logic bug somewhere. I know the test here is racy,
* but there's not a lot we can do about it at this
* point.
*/
fatal( "Attempted to destroy a flexthread_mutex"\
" held by another thread!" );
}
if (flexthread_mutex_held(ftm)) {
flexthread_mutex_unlock(ftm);
} else if ((pthread_t) NULL != ftm->holder) {
/* This "should never happen": if we can try to destroy
* a mutex currently held by another thread, there's a
* logic bug somewhere. I know the test here is racy,
* but there's not a lot we can do about it at this
* point.
*/
fatal("Attempted to destroy a flexthread_mutex"
" held by another thread!");
}
FATAL_UNLESS( 0 == pthread_mutex_destroy( &ftm->mutex ),
"Mutex destroy failed" );
free( ftm );
FATAL_UNLESS(0 == pthread_mutex_destroy(&ftm->mutex),
"Mutex destroy failed");
free(ftm);
}
int flexthread_mutex_lock( struct flexthread_mutex * ftm )
int flexthread_mutex_lock(struct flexthread_mutex *ftm)
{
NULLCHECK( ftm );
NULLCHECK(ftm);
int failure = pthread_mutex_lock( &ftm->mutex );
if ( 0 == failure ) {
ftm->holder = pthread_self();
}
int failure = pthread_mutex_lock(&ftm->mutex);
if (0 == failure) {
ftm->holder = pthread_self();
}
return failure;
return failure;
}
int flexthread_mutex_unlock( struct flexthread_mutex * ftm )
int flexthread_mutex_unlock(struct flexthread_mutex *ftm)
{
NULLCHECK( ftm );
NULLCHECK(ftm);
pthread_t orig = ftm->holder;
ftm->holder = (pthread_t)NULL;
int failure = pthread_mutex_unlock( &ftm->mutex );
if ( 0 != failure ) {
ftm->holder = orig;
}
return failure;
pthread_t orig = ftm->holder;
ftm->holder = (pthread_t) NULL;
int failure = pthread_mutex_unlock(&ftm->mutex);
if (0 != failure) {
ftm->holder = orig;
}
return failure;
}
int flexthread_mutex_held( struct flexthread_mutex * ftm )
int flexthread_mutex_held(struct flexthread_mutex *ftm)
{
NULLCHECK( ftm );
return pthread_self() == ftm->holder;
NULLCHECK(ftm);
return pthread_self() == ftm->holder;
}

View File

@@ -15,15 +15,15 @@
*/
struct flexthread_mutex {
pthread_mutex_t mutex;
pthread_t holder;
pthread_mutex_t mutex;
pthread_t holder;
};
struct flexthread_mutex * flexthread_mutex_create(void);
void flexthread_mutex_destroy( struct flexthread_mutex * );
struct flexthread_mutex *flexthread_mutex_create(void);
void flexthread_mutex_destroy(struct flexthread_mutex *);
int flexthread_mutex_lock( struct flexthread_mutex * );
int flexthread_mutex_unlock( struct flexthread_mutex * );
int flexthread_mutex_held( struct flexthread_mutex * );
int flexthread_mutex_lock(struct flexthread_mutex *);
int flexthread_mutex_unlock(struct flexthread_mutex *);
int flexthread_mutex_held(struct flexthread_mutex *);
#endif

View File

@@ -3,75 +3,75 @@
#include <pthread.h>
struct mbox * mbox_create( void )
struct mbox *mbox_create(void)
{
struct mbox * mbox = xmalloc( sizeof( struct mbox ) );
FATAL_UNLESS( 0 == pthread_cond_init( &mbox->filled_cond, NULL ),
"Failed to initialise a condition variable" );
FATAL_UNLESS( 0 == pthread_cond_init( &mbox->emptied_cond, NULL ),
"Failed to initialise a condition variable" );
FATAL_UNLESS( 0 == pthread_mutex_init( &mbox->mutex, NULL ),
"Failed to initialise a mutex" );
return mbox;
struct mbox *mbox = xmalloc(sizeof(struct mbox));
FATAL_UNLESS(0 == pthread_cond_init(&mbox->filled_cond, NULL),
"Failed to initialise a condition variable");
FATAL_UNLESS(0 == pthread_cond_init(&mbox->emptied_cond, NULL),
"Failed to initialise a condition variable");
FATAL_UNLESS(0 == pthread_mutex_init(&mbox->mutex, NULL),
"Failed to initialise a mutex");
return mbox;
}
void mbox_post( struct mbox * mbox, void * contents )
void mbox_post(struct mbox *mbox, void *contents)
{
pthread_mutex_lock( &mbox->mutex );
{
if (mbox->full){
pthread_cond_wait( &mbox->emptied_cond, &mbox->mutex );
}
mbox->contents = contents;
mbox->full = 1;
while( 0 != pthread_cond_signal( &mbox->filled_cond ) );
pthread_mutex_lock(&mbox->mutex);
{
if (mbox->full) {
pthread_cond_wait(&mbox->emptied_cond, &mbox->mutex);
}
pthread_mutex_unlock( &mbox->mutex );
mbox->contents = contents;
mbox->full = 1;
while (0 != pthread_cond_signal(&mbox->filled_cond));
}
pthread_mutex_unlock(&mbox->mutex);
}
void * mbox_contents( struct mbox * mbox )
void *mbox_contents(struct mbox *mbox)
{
return mbox->contents;
return mbox->contents;
}
int mbox_is_full( struct mbox * mbox )
int mbox_is_full(struct mbox *mbox)
{
return mbox->full;
return mbox->full;
}
void * mbox_receive( struct mbox * mbox )
void *mbox_receive(struct mbox *mbox)
{
NULLCHECK( mbox );
void * result;
NULLCHECK(mbox);
void *result;
pthread_mutex_lock( &mbox->mutex );
{
if ( !mbox->full ) {
pthread_cond_wait( &mbox->filled_cond, &mbox->mutex );
}
mbox->full = 0;
result = mbox->contents;
mbox->contents = NULL;
while( 0 != pthread_cond_signal( &mbox->emptied_cond));
pthread_mutex_lock(&mbox->mutex);
{
if (!mbox->full) {
pthread_cond_wait(&mbox->filled_cond, &mbox->mutex);
}
pthread_mutex_unlock( &mbox->mutex );
mbox->full = 0;
result = mbox->contents;
mbox->contents = NULL;
return result;
while (0 != pthread_cond_signal(&mbox->emptied_cond));
}
pthread_mutex_unlock(&mbox->mutex);
return result;
}
void mbox_destroy( struct mbox * mbox )
void mbox_destroy(struct mbox *mbox)
{
NULLCHECK( mbox );
NULLCHECK(mbox);
while( 0 != pthread_cond_destroy( &mbox->emptied_cond ) );
while( 0 != pthread_cond_destroy( &mbox->filled_cond ) );
while (0 != pthread_cond_destroy(&mbox->emptied_cond));
while (0 != pthread_cond_destroy(&mbox->filled_cond));
while( 0 != pthread_mutex_destroy( &mbox->mutex ) );
while (0 != pthread_mutex_destroy(&mbox->mutex));
free( mbox );
free(mbox);
}

View File

@@ -14,42 +14,42 @@
struct mbox {
void * contents;
void *contents;
/** Marker to tell us if there's content in the box.
* Keeping this separate allows us to use NULL for the contents.
*/
int full;
int full;
/** This gets signaled by mbox_post, and waited on by
* mbox_receive */
pthread_cond_t filled_cond;
pthread_cond_t filled_cond;
/** This is signaled by mbox_receive, and waited on by mbox_post */
pthread_cond_t emptied_cond;
pthread_mutex_t mutex;
pthread_cond_t emptied_cond;
pthread_mutex_t mutex;
};
/* Create an mbox. */
struct mbox * mbox_create(void);
struct mbox *mbox_create(void);
/* Put something in the mbox, blocking if it's already full.
* That something can be NULL if you want.
*/
void mbox_post( struct mbox *, void *);
void mbox_post(struct mbox *, void *);
/* See what's in the mbox. This isn't thread-safe. */
void * mbox_contents( struct mbox *);
void *mbox_contents(struct mbox *);
/* See if anything has been put into the mbox. This isn't thread-safe.
* */
int mbox_is_full( struct mbox *);
int mbox_is_full(struct mbox *);
/* Get the contents from the mbox, blocking if there's nothing there. */
void * mbox_receive( struct mbox *);
void *mbox_receive(struct mbox *);
/* Free the mbox and destroy the associated pthread bits. */
void mbox_destroy( struct mbox *);
void mbox_destroy(struct mbox *);
#endif

File diff suppressed because it is too large Load Diff

View File

@@ -58,65 +58,65 @@ enum mirror_state;
#define MS_REQUEST_LIMIT_SECS_F 60.0
enum mirror_finish_action {
ACTION_EXIT,
ACTION_UNLINK,
ACTION_NOTHING
ACTION_EXIT,
ACTION_UNLINK,
ACTION_NOTHING
};
enum mirror_state {
MS_UNKNOWN,
MS_INIT,
MS_GO,
MS_ABANDONED,
MS_DONE,
MS_FAIL_CONNECT,
MS_FAIL_REJECTED,
MS_FAIL_NO_HELLO,
MS_FAIL_SIZE_MISMATCH
MS_UNKNOWN,
MS_INIT,
MS_GO,
MS_ABANDONED,
MS_DONE,
MS_FAIL_CONNECT,
MS_FAIL_REJECTED,
MS_FAIL_NO_HELLO,
MS_FAIL_SIZE_MISMATCH
};
struct mirror {
pthread_t thread;
pthread_t thread;
/* Signal to this then join the thread if you want to abandon mirroring */
struct self_pipe * abandon_signal;
/* Signal to this then join the thread if you want to abandon mirroring */
struct self_pipe *abandon_signal;
union mysockaddr * connect_to;
union mysockaddr * connect_from;
int client;
const char * filename;
union mysockaddr *connect_to;
union mysockaddr *connect_from;
int client;
const char *filename;
/* Limiter, used to restrict migration speed Only dirty bytes (those going
* over the network) are considered */
uint64_t max_bytes_per_second;
/* Limiter, used to restrict migration speed Only dirty bytes (those going
* over the network) are considered */
uint64_t max_bytes_per_second;
enum mirror_finish_action action_at_finish;
enum mirror_finish_action action_at_finish;
char *mapped;
char *mapped;
/* We need to send every byte at least once; we do so by */
uint64_t offset;
/* We need to send every byte at least once; we do so by */
uint64_t offset;
enum mirror_state commit_state;
enum mirror_state commit_state;
/* commit_signal is sent immediately after attempting to connect
* and checking the remote size, whether successful or not.
*/
struct mbox * commit_signal;
/* commit_signal is sent immediately after attempting to connect
* and checking the remote size, whether successful or not.
*/
struct mbox *commit_signal;
/* The time (from monotonic_time_ms()) the migration was started. Can be
* used to calculate bps, etc. */
uint64_t migration_started;
/* The time (from monotonic_time_ms()) the migration was started. Can be
* used to calculate bps, etc. */
uint64_t migration_started;
/* Running count of all bytes we've transferred */
uint64_t all_dirty;
/* Running count of all bytes we've transferred */
uint64_t all_dirty;
};
struct mirror_super {
struct mirror * mirror;
pthread_t thread;
struct mbox * state_mbox;
struct mirror *mirror;
pthread_t thread;
struct mbox *state_mbox;
};
@@ -127,15 +127,13 @@ struct mirror_super {
struct server;
struct flexnbd;
struct mirror_super * mirror_super_create(
const char * filename,
union mysockaddr * connect_to,
union mysockaddr * connect_from,
uint64_t max_Bps,
enum mirror_finish_action action_at_finish,
struct mbox * state_mbox
);
void * mirror_super_runner( void * serve_uncast );
struct mirror_super *mirror_super_create(const char *filename,
union mysockaddr *connect_to,
union mysockaddr *connect_from,
uint64_t max_Bps,
enum mirror_finish_action
action_at_finish,
struct mbox *state_mbox);
void *mirror_super_runner(void *serve_uncast);
#endif

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@@ -3,165 +3,165 @@
#include <sys/types.h>
#include <unistd.h>
#include <signal.h> /* for sig_atomic_t */
#include <signal.h> /* for sig_atomic_t */
#include "flexnbd.h"
#include "parse.h"
#include "acl.h"
static const int block_allocation_resolution = 4096;//128<<10;
static const int block_allocation_resolution = 4096; //128<<10;
struct client_tbl_entry {
pthread_t thread;
union mysockaddr address;
struct client * client;
pthread_t thread;
union mysockaddr address;
struct client *client;
};
#define MAX_NBD_CLIENTS 16
#define CLIENT_KEEPALIVE_TIME 30
#define CLIENT_KEEPALIVE_INTVL 10
#define CLIENT_KEEPALIVE_PROBES 3
struct server {
/* The flexnbd wrapper this server is attached to */
struct flexnbd * flexnbd;
/* The flexnbd wrapper this server is attached to */
struct flexnbd *flexnbd;
/** address/port to bind to */
union mysockaddr bind_to;
union mysockaddr bind_to;
/** (static) file name to serve */
char* filename;
char *filename;
/** TCP backlog for listen() */
int tcp_backlog;
int tcp_backlog;
/** (static) file name of UNIX control socket (or NULL if none) */
char* control_socket_name;
char *control_socket_name;
/** size of file */
uint64_t size;
uint64_t size;
/** to interrupt accept loop and clients, write() to close_signal[1] */
struct self_pipe * close_signal;
struct self_pipe *close_signal;
/** access control list */
struct acl * acl;
struct acl *acl;
/** acl_updated_signal will be signalled after the acl struct
* has been replaced
*/
struct self_pipe * acl_updated_signal;
struct self_pipe *acl_updated_signal;
/* Claimed around any updates to the ACL. */
struct flexthread_mutex * l_acl;
/* Claimed around any updates to the ACL. */
struct flexthread_mutex *l_acl;
/* Claimed around starting a mirror so that it doesn't race with
* shutting down on a SIGTERM. */
struct flexthread_mutex * l_start_mirror;
/* Claimed around starting a mirror so that it doesn't race with
* shutting down on a SIGTERM. */
struct flexthread_mutex *l_start_mirror;
struct mirror* mirror;
struct mirror_super * mirror_super;
/* This is used to stop the mirror from starting after we
* receive a SIGTERM */
int mirror_can_start;
struct mirror *mirror;
struct mirror_super *mirror_super;
/* This is used to stop the mirror from starting after we
* receive a SIGTERM */
int mirror_can_start;
int server_fd;
int control_fd;
int server_fd;
int control_fd;
/* the allocation_map keeps track of which blocks in the backing file
* have been allocated, or part-allocated on disc, with unallocated
* blocks presumed to contain zeroes (i.e. represented as sparse files
* by the filesystem). We can use this information when receiving
* incoming writes, and avoid writing zeroes to unallocated sections
* of the file which would needlessly increase disc usage. This
* bitmap will start at all-zeroes for an empty file, and tend towards
* all-ones as the file is written to (i.e. we assume that allocated
* blocks can never become unallocated again, as is the case with ext3
* at least).
*/
struct bitset * allocation_map;
/* when starting up, this thread builds the allocation_map */
pthread_t allocation_map_builder_thread;
/* the allocation_map keeps track of which blocks in the backing file
* have been allocated, or part-allocated on disc, with unallocated
* blocks presumed to contain zeroes (i.e. represented as sparse files
* by the filesystem). We can use this information when receiving
* incoming writes, and avoid writing zeroes to unallocated sections
* of the file which would needlessly increase disc usage. This
* bitmap will start at all-zeroes for an empty file, and tend towards
* all-ones as the file is written to (i.e. we assume that allocated
* blocks can never become unallocated again, as is the case with ext3
* at least).
*/
struct bitset *allocation_map;
/* when starting up, this thread builds the allocation_map */
pthread_t allocation_map_builder_thread;
/* when the thread has finished, it sets this to 1 */
volatile sig_atomic_t allocation_map_built;
volatile sig_atomic_t allocation_map_not_built;
/* when the thread has finished, it sets this to 1 */
volatile sig_atomic_t allocation_map_built;
volatile sig_atomic_t allocation_map_not_built;
int max_nbd_clients;
struct client_tbl_entry *nbd_client;
int max_nbd_clients;
struct client_tbl_entry *nbd_client;
/** Should clients use the killswitch? */
int use_killswitch;
int use_killswitch;
/** If this isn't set, newly accepted clients will be closed immediately */
int allow_new_clients;
int allow_new_clients;
/* Marker for whether this server has control over the data in
* the file, or if we're waiting to receive it from an inbound
* migration which hasn't yet finished.
*
* It's the value which controls the exit status of a serve or
* listen process.
*/
int success;
/* Marker for whether this server has control over the data in
* the file, or if we're waiting to receive it from an inbound
* migration which hasn't yet finished.
*
* It's the value which controls the exit status of a serve or
* listen process.
*/
int success;
};
struct server * server_create(
struct flexnbd * flexnbd,
char* s_ip_address,
char* s_port,
char* s_file,
int default_deny,
int acl_entries,
char** s_acl_entries,
int max_nbd_clients,
int use_killswitch,
int success );
void server_destroy( struct server * );
int server_is_closed(struct server* serve);
void serve_signal_close( struct server *serve );
void serve_wait_for_close( struct server * serve );
void server_replace_acl( struct server *serve, struct acl * acl);
void server_control_arrived( struct server *serve );
int server_is_in_control( struct server *serve );
int server_default_deny( struct server * serve );
int server_acl_locked( struct server * serve );
void server_lock_acl( struct server *serve );
void server_unlock_acl( struct server *serve );
void server_lock_start_mirror( struct server *serve );
void server_unlock_start_mirror( struct server *serve );
int server_is_mirroring( struct server * serve );
struct server *server_create(struct flexnbd *flexnbd,
char *s_ip_address,
char *s_port,
char *s_file,
int default_deny,
int acl_entries,
char **s_acl_entries,
int max_nbd_clients,
int use_killswitch, int success);
void server_destroy(struct server *);
int server_is_closed(struct server *serve);
void serve_signal_close(struct server *serve);
void serve_wait_for_close(struct server *serve);
void server_replace_acl(struct server *serve, struct acl *acl);
void server_control_arrived(struct server *serve);
int server_is_in_control(struct server *serve);
int server_default_deny(struct server *serve);
int server_acl_locked(struct server *serve);
void server_lock_acl(struct server *serve);
void server_unlock_acl(struct server *serve);
void server_lock_start_mirror(struct server *serve);
void server_unlock_start_mirror(struct server *serve);
int server_is_mirroring(struct server *serve);
uint64_t server_mirror_bytes_remaining( struct server * serve );
uint64_t server_mirror_eta( struct server * serve );
uint64_t server_mirror_bps( struct server * serve );
uint64_t server_mirror_bytes_remaining(struct server *serve);
uint64_t server_mirror_eta(struct server *serve);
uint64_t server_mirror_bps(struct server *serve);
void server_abandon_mirror( struct server * serve );
void server_prevent_mirror_start( struct server *serve );
void server_allow_mirror_start( struct server *serve );
int server_mirror_can_start( struct server *serve );
void server_abandon_mirror(struct server *serve);
void server_prevent_mirror_start(struct server *serve);
void server_allow_mirror_start(struct server *serve);
int server_mirror_can_start(struct server *serve);
/* These three functions are used by mirror around the final pass, to close
* existing clients and prevent new ones from being around
*/
void server_forbid_new_clients( struct server *serve );
void server_close_clients( struct server *serve );
void server_join_clients( struct server *serve );
void server_allow_new_clients( struct server *serve );
void server_forbid_new_clients(struct server *serve);
void server_close_clients(struct server *serve);
void server_join_clients(struct server *serve);
void server_allow_new_clients(struct server *serve);
/* Returns a count (ish) of the number of currently-running client threads */
int server_count_clients( struct server *params );
int server_count_clients(struct server *params);
void server_unlink( struct server * serve );
void server_unlink(struct server *serve);
int do_serve( struct server *, struct self_pipe * );
int do_serve(struct server *, struct self_pipe *);
struct mode_readwrite_params {
union mysockaddr connect_to;
union mysockaddr connect_from;
union mysockaddr connect_to;
union mysockaddr connect_from;
uint64_t from;
uint32_t len;
uint64_t from;
uint32_t len;
int data_fd;
int client;
int data_fd;
int client;
};
#endif

View File

@@ -2,40 +2,44 @@
#include "serve.h"
#include "util.h"
struct status * status_create( struct server * serve )
struct status *status_create(struct server *serve)
{
NULLCHECK( serve );
struct status * status;
NULLCHECK(serve);
struct status *status;
status = xmalloc( sizeof( struct status ) );
status->pid = getpid();
status->size = serve->size;
status->has_control = serve->success;
status = xmalloc(sizeof(struct status));
status->pid = getpid();
status->size = serve->size;
status->has_control = serve->success;
status->clients_allowed = serve->allow_new_clients;
status->num_clients = server_count_clients( serve );
status->clients_allowed = serve->allow_new_clients;
status->num_clients = server_count_clients(serve);
server_lock_start_mirror( serve );
server_lock_start_mirror(serve);
status->is_mirroring = NULL != serve->mirror;
if ( status->is_mirroring ) {
status->migration_duration = monotonic_time_ms();
status->is_mirroring = NULL != serve->mirror;
if (status->is_mirroring) {
status->migration_duration = monotonic_time_ms();
if ( ( serve->mirror->migration_started ) < status->migration_duration ) {
status->migration_duration -= serve->mirror->migration_started;
} else {
status->migration_duration = 0;
}
status->migration_duration /= 1000;
status->migration_speed = server_mirror_bps( serve );
status->migration_speed_limit = serve->mirror->max_bytes_per_second;
status->migration_seconds_left = server_mirror_eta( serve );
if ((serve->mirror->migration_started) <
status->migration_duration) {
status->migration_duration -= serve->mirror->migration_started;
} else {
status->migration_duration = 0;
}
status->migration_duration /= 1000;
status->migration_speed = server_mirror_bps(serve);
status->migration_speed_limit =
serve->mirror->max_bytes_per_second;
server_unlock_start_mirror( serve );
status->migration_seconds_left = server_mirror_eta(serve);
status->migration_bytes_left =
server_mirror_bytes_remaining(serve);
}
return status;
server_unlock_start_mirror(serve);
return status;
}
@@ -47,32 +51,32 @@ struct status * status_create( struct server * serve )
#define PRINT_UINT64( var ) \
do{dprintf( fd, #var "=%"PRIu64" ", status->var );}while(0)
int status_write( struct status * status, int fd )
int status_write(struct status *status, int fd)
{
PRINT_INT( pid );
PRINT_UINT64( size );
PRINT_BOOL( is_mirroring );
PRINT_BOOL( clients_allowed );
PRINT_INT( num_clients );
PRINT_BOOL( has_control );
PRINT_INT(pid);
PRINT_UINT64(size);
PRINT_BOOL(is_mirroring);
PRINT_BOOL(clients_allowed);
PRINT_INT(num_clients);
PRINT_BOOL(has_control);
if ( status->is_mirroring ) {
PRINT_UINT64( migration_speed );
PRINT_UINT64( migration_duration );
PRINT_UINT64( migration_seconds_left );
if ( status->migration_speed_limit < UINT64_MAX ) {
PRINT_UINT64( migration_speed_limit );
};
}
if (status->is_mirroring) {
PRINT_UINT64(migration_speed);
PRINT_UINT64(migration_duration);
PRINT_UINT64(migration_seconds_left);
PRINT_UINT64(migration_bytes_left);
if (status->migration_speed_limit < UINT64_MAX) {
PRINT_UINT64(migration_speed_limit);
};
}
dprintf(fd, "\n");
return 1;
dprintf(fd, "\n");
return 1;
}
void status_destroy( struct status * status )
void status_destroy(struct status *status)
{
NULLCHECK( status );
free( status );
NULLCHECK(status);
free(status);
}

View File

@@ -64,6 +64,8 @@
* Our current best estimate of how many seconds are left before the migration
* migration is finished.
*
* migration_bytes_left:
* The number of bytes remaining to migrate.
*/
@@ -73,29 +75,29 @@
#include <unistd.h>
struct status {
pid_t pid;
uint64_t size;
int has_control;
int clients_allowed;
int num_clients;
int is_mirroring;
pid_t pid;
uint64_t size;
int has_control;
int clients_allowed;
int num_clients;
int is_mirroring;
uint64_t migration_duration;
uint64_t migration_speed;
uint64_t migration_speed_limit;
uint64_t migration_seconds_left;
uint64_t migration_duration;
uint64_t migration_speed;
uint64_t migration_speed_limit;
uint64_t migration_seconds_left;
uint64_t migration_bytes_left;
};
/** Create a status object for the given server. */
struct status * status_create( struct server * );
struct status *status_create(struct server *);
/** Output the given status object to the given file descriptot */
int status_write( struct status *, int fd );
int status_write(struct status *, int fd);
/** Free the status object */
void status_destroy( struct status * );
void status_destroy(struct status *);
#endif

View File

@@ -1,39 +1,42 @@
# encoding: utf-8
require 'flexnbd'
require 'file_writer'
class Environment
attr_reader( :blocksize, :filename1, :filename2, :ip,
:port1, :port2, :nbd1, :nbd2, :file1, :file2 )
attr_reader(:blocksize, :filename1, :filename2, :ip,
:port1, :port2, :nbd1, :nbd2, :file1, :file2)
def initialize
@blocksize = 1024
@filename1 = "/tmp/.flexnbd.test.#{$$}.#{Time.now.to_i}.1"
@filename2 = "/tmp/.flexnbd.test.#{$$}.#{Time.now.to_i}.2"
@ip = "127.0.0.1"
@available_ports = [*40000..41000] - listening_ports
@filename1 = "/tmp/.flexnbd.test.#{$PROCESS_ID}.#{Time.now.to_i}.1"
@filename2 = "/tmp/.flexnbd.test.#{$PROCESS_ID}.#{Time.now.to_i}.2"
@ip = '127.0.0.1'
@available_ports = [*40_000..41_000] - listening_ports
@port1 = @available_ports.shift
@port2 = @available_ports.shift
@nbd1 = FlexNBD::FlexNBD.new("../../build/flexnbd", @ip, @port1)
@nbd2 = FlexNBD::FlexNBD.new("../../build/flexnbd", @ip, @port2)
@nbd1 = FlexNBD::FlexNBD.new('../../build/flexnbd', @ip, @port1)
@nbd2 = FlexNBD::FlexNBD.new('../../build/flexnbd', @ip, @port2)
@fake_pid = nil
end
def blocksize=(b)
raise RuntimeError, "Unable to change blocksize after files have been opened" if @file1 or @file2
@blocksize = b
end
def prefetch_proxy!
@nbd1.prefetch_proxy = true
@nbd2.prefetch_proxy = true
end
def proxy1(port=@port2)
def proxy1(port = @port2)
@nbd1.proxy(@ip, port)
end
def proxy2(port=@port1)
def proxy2(port = @port1)
@nbd2.proxy(@ip, port)
end
def serve1(*acl)
@nbd1.serve(@filename1, *acl)
end
@@ -42,29 +45,26 @@ class Environment
@nbd2.serve(@filename2, *acl)
end
def listen1( *acl )
@nbd1.listen( @filename1, *(acl.empty? ? @acl1: acl) )
def listen1(*acl)
@nbd1.listen(@filename1, *(acl.empty? ? @acl1 : acl))
end
def listen2( *acl )
@nbd2.listen( @filename2, *acl )
def listen2(*acl)
@nbd2.listen(@filename2, *acl)
end
def break1
@nbd1.break
end
def acl1( *acl )
@nbd1.acl( *acl )
def acl1(*acl)
@nbd1.acl(*acl)
end
def acl2( *acl )
@nbd2.acl( *acl )
def acl2(*acl)
@nbd2.acl(*acl)
end
def status1
@nbd1.status.first
end
@@ -73,23 +73,20 @@ class Environment
@nbd2.status.first
end
def mirror12
@nbd1.mirror( @nbd2.ip, @nbd2.port )
@nbd1.mirror(@nbd2.ip, @nbd2.port)
end
def mirror12_unchecked
@nbd1.mirror_unchecked( @nbd2.ip, @nbd2.port, nil, nil, 10 )
@nbd1.mirror_unchecked(@nbd2.ip, @nbd2.port, nil, nil, 10)
end
def mirror12_unlink
@nbd1.mirror_unlink( @nbd2.ip, @nbd2.port, 2 )
@nbd1.mirror_unlink(@nbd2.ip, @nbd2.port, 2)
end
def write1( data )
@nbd1.write( 0, data )
def write1(data)
@nbd1.write(0, data)
end
def writefile1(data)
@@ -100,63 +97,54 @@ class Environment
@file2 = FileWriter.new(@filename2, @blocksize).write(data)
end
def truncate1( size )
def truncate1(size)
system "truncate -s #{size} #{@filename1}"
end
def listening_ports
`netstat -ltn`.
split("\n").
map { |x| x.split(/\s+/) }[2..-1].
map { |l| l[3].split(":")[-1].to_i }
`netstat -ltn`
.split("\n")
.map { |x| x.split(/\s+/) }[2..-1]
.map { |l| l[3].split(':')[-1].to_i }
end
def cleanup
if @fake_pid
begin
Process.waitpid2( @fake_pid )
Process.waitpid2(@fake_pid)
rescue Errno::ESRCH
end
end
@nbd1.can_die(0)
@nbd1.kill
@nbd2.kill
[@filename1, @filename2].each do |f|
File.unlink(f) if File.exists?(f)
File.unlink(f) if File.exist?(f)
end
end
def run_fake( name, addr, port, sock=nil )
fakedir = File.join( File.dirname( __FILE__ ), "fakes" )
fakeglob = File.join( fakedir, name ) + "*"
fake = Dir[fakeglob].sort.find { |fn|
File.executable?( fn )
}
def run_fake(name, addr, port, sock = nil)
fakedir = File.join(File.dirname(__FILE__), 'fakes')
fakeglob = File.join(fakedir, name) + '*'
fake = Dir[fakeglob].sort.find do |fn|
File.executable?(fn)
end
raise "no fake executable at #{fakeglob}" unless fake
raise "no addr" unless addr
raise "no port" unless port
raise 'no addr' unless addr
raise 'no port' unless port
@fake_pid = fork do
exec [fake, addr, port, @nbd1.pid, sock].map{|x| x.to_s}.join(" ")
exec [fake, addr, port, @nbd1.pid, sock].map(&:to_s).join(' ')
end
sleep(0.5)
end
def fake_reports_success
_,status = Process.waitpid2( @fake_pid )
_, status = Process.waitpid2(@fake_pid)
@fake_pid = nil
status.success?
end
end # class Environment

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# Open a server, accept a client, then cancel the migration by issuing
# a break command.
@@ -8,28 +6,27 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port, src_pid, sock = *ARGV
server = FakeDest.new( addr, port )
server = FakeDest.new(addr, port)
client = server.accept
ctrl = UNIXSocket.open( sock )
ctrl = UNIXSocket.open(sock)
Process.kill("STOP", src_pid.to_i)
ctrl.write( "break\n" )
Process.kill('STOP', src_pid.to_i)
ctrl.write("break\n")
ctrl.close_write
client.write_hello
Process.kill("CONT", src_pid.to_i)
Process.kill('CONT', src_pid.to_i)
fail "Unexpected control response" unless
raise 'Unexpected control response' unless
ctrl.read =~ /0: mirror stopped/
client2 = nil
begin
client2 = server.accept( "Expected timeout" )
fail "Unexpected reconnection"
client2 = server.accept('Expected timeout')
raise 'Unexpected reconnection'
rescue Timeout::Error
# expected
end
client.close
exit(0)

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# Receive a mirror, and disconnect after sending the entrust reply but
# before it can send the disconnect signal.
#
@@ -11,26 +9,25 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port, src_pid = *ARGV
server = FakeDest.new( addr, port )
server = FakeDest.new(addr, port)
client = server.accept
client.write_hello
while (req = client.read_request; req[:type] == 1)
client.read_data( req[:len] )
client.write_reply( req[:handle] )
while req = client.read_request; req[:type] == 1
client.read_data(req[:len])
client.write_reply(req[:handle])
end
system "kill -STOP #{src_pid}"
client.write_reply( req[:handle] )
client.write_reply(req[:handle])
client.close
system "kill -CONT #{src_pid}"
sleep( 0.25 )
client2 = server.accept( "Timed out waiting for a reconnection" )
sleep(0.25)
client2 = server.accept('Timed out waiting for a reconnection')
client2.close
server.close
$stderr.puts "done"
warn 'done'
exit(0)

View File

@@ -10,12 +10,12 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port = *ARGV
server = FakeDest.new( addr, port )
client = server.accept( "Timed out waiting for a connection" )
server = FakeDest.new(addr, port)
client = server.accept('Timed out waiting for a connection')
client.write_hello
client.close
new_client = server.accept( "Timed out waiting for a reconnection" )
new_client = server.accept('Timed out waiting for a reconnection')
new_client.close
server.close

View File

@@ -11,13 +11,13 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port = *ARGV
server = FakeDest.new( addr, port )
client = server.accept( "Timed out waiting for a connection" )
server = FakeDest.new(addr, port)
client = server.accept('Timed out waiting for a connection')
client.write_hello
client.read_request
client.close
new_client = server.accept( "Timed out waiting for a reconnection" )
new_client = server.accept('Timed out waiting for a reconnection')
new_client.close
server.close

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# Open a server, accept a client, then we expect a single write
# followed by an entrust. However, we disconnect after the write so
# the entrust will fail. We don't expect a reconnection: the sender
@@ -10,16 +8,16 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port, src_pid = *ARGV
server = FakeDest.new( addr, port )
server = FakeDest.new(addr, port)
client = server.accept
client.write_hello
req = client.read_request
data = client.read_data( req[:len] )
data = client.read_data(req[:len])
Process.kill("STOP", src_pid.to_i)
client.write_reply( req[:handle], 0 )
Process.kill('STOP', src_pid.to_i)
client.write_reply(req[:handle], 0)
client.close
Process.kill("CONT", src_pid.to_i)
Process.kill('CONT', src_pid.to_i)
exit(0)

View File

@@ -1,19 +1,16 @@
#!/usr/bin/env ruby
# encoding: utf-8
require 'flexnbd/fake_dest'
include FlexNBD
addr, port = *ARGV
server = FakeDest.new( addr, port )
server = FakeDest.new(addr, port)
client = server.accept
client.write_hello
handle = client.read_request[:handle]
client.write_error( handle )
client.write_error(handle)
client2 = server.accept( "Timed out waiting for a reconnection" )
client2 = server.accept('Timed out waiting for a reconnection')
client.close
client2.close

View File

@@ -14,8 +14,8 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port = *ARGV
server = FakeDest.new( addr, port )
client = server.accept( "Client didn't make a connection" )
server = FakeDest.new(addr, port)
client = server.accept("Client didn't make a connection")
# Sleep for one second past the timeout (a bit of slop in case ruby
# doesn't launch things quickly)
@@ -26,10 +26,10 @@ client.close
# Invert the sense of the timeout exception, since we *don't* want a
# connection attempt
begin
server.accept( "Expected timeout" )
fail "Unexpected reconnection"
server.accept('Expected timeout')
raise 'Unexpected reconnection'
rescue Timeout::Error
# expected
# expected
end
server.close

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# Open a socket, say hello, receive a write, then sleep for >
# MS_REQUEST_LIMIT_SECS seconds. This should tell the source that the
# write has gone MIA, and we expect a reconnect.
@@ -9,24 +7,24 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port = *ARGV
server = FakeDest.new( addr, port )
client1 = server.accept( server )
server = FakeDest.new(addr, port)
client1 = server.accept(server)
client1.write_hello
client1.read_request
t = Thread.start do
client2 = server.accept( "Timed out waiting for a reconnection",
FlexNBD::MS_REQUEST_LIMIT_SECS + 2 )
client2 = server.accept('Timed out waiting for a reconnection',
FlexNBD::MS_REQUEST_LIMIT_SECS + 2)
client2.close
end
sleep_time = if ENV.has_key?('FLEXNBD_MS_REQUEST_LIMIT_SECS')
ENV['FLEXNBD_MS_REQUEST_LIMIT_SECS'].to_f
else
FlexNBD::MS_REQUEST_LIMIT_SECS
sleep_time = if ENV.key?('FLEXNBD_MS_REQUEST_LIMIT_SECS')
ENV['FLEXNBD_MS_REQUEST_LIMIT_SECS'].to_f
else
FlexNBD::MS_REQUEST_LIMIT_SECS
end
sleep( sleep_time + 2.0 )
sleep(sleep_time + 2.0)
client1.close
t.join

View File

@@ -7,21 +7,21 @@ include FlexNBD
Thread.abort_on_exception
addr, port = *ARGV
server = FakeDest.new( addr, port )
server = FakeDest.new(addr, port)
client1 = server.accept
# We don't expect a reconnection attempt.
t = Thread.new do
begin
client2 = server.accept( "Timed out waiting for a reconnection",
FlexNBD::MS_RETRY_DELAY_SECS + 1 )
fail "Unexpected reconnection"
client2 = server.accept('Timed out waiting for a reconnection',
FlexNBD::MS_RETRY_DELAY_SECS + 1)
raise 'Unexpected reconnection'
rescue Timeout::Error
#expected
# expected
end
end
client1.write_hello( :magic => :wrong )
client1.write_hello(magic: :wrong)
t.join

View File

@@ -9,7 +9,7 @@ include FlexNBD
Thread.abort_on_exception = true
addr, port = *ARGV
server = FakeDest.new( addr, port )
server = FakeDest.new(addr, port)
client = server.accept
t = Thread.new do
@@ -18,21 +18,21 @@ t = Thread.new do
# so it makes no sense to continue. This means we have to invert the
# sense of the exception.
begin
client2 = server.accept( "Timed out waiting for a reconnection",
FlexNBD::MS_RETRY_DELAY_SECS + 1 )
client2 = server.accept('Timed out waiting for a reconnection',
FlexNBD::MS_RETRY_DELAY_SECS + 1)
client2.close
fail "Unexpected reconnection."
raise 'Unexpected reconnection.'
rescue Timeout::Error
end
end
client.write_hello( :size => :wrong )
client.write_hello(size: :wrong)
t.join
# Now check that the source closed the first socket (yes, this was an
# actual bug)
fail "Didn't close socket" unless client.disconnected?
raise "Didn't close socket" unless client.disconnected?
exit 0

View File

@@ -7,18 +7,16 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port = *ARGV
server = FakeDest.new( addr, port )
server = FakeDest.new(addr, port)
server.accept.close
begin
server.accept
fail "Unexpected reconnection"
raise 'Unexpected reconnection'
rescue Timeout::Error
# expected
end
server.close
exit(0)

View File

@@ -8,8 +8,8 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port, pid = *ARGV
server = FakeDest.new( addr, port )
client = server.accept( "Timed out waiting for a connection" )
server = FakeDest.new(addr, port)
client = server.accept('Timed out waiting for a connection')
client.write_hello
Process.kill(15, pid.to_i)

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# Accept a connection, write hello, wait for a write request, read the
# data, then write back a reply with a bad magic field. We then
# expect a reconnect.
@@ -9,13 +7,13 @@ require 'flexnbd/fake_dest'
include FlexNBD
addr, port = *ARGV
server = FakeDest.new( addr, port )
server = FakeDest.new(addr, port)
client = server.accept
client.write_hello
req = client.read_request
client.read_data( req[:len] )
client.write_reply( req[:handle], 0, :magic => :wrong )
client.read_data(req[:len])
client.write_reply(req[:handle], 0, magic: :wrong)
client2 = server.accept
client.close

View File

@@ -11,13 +11,13 @@ include FlexNBD
addr, port = *ARGV
FakeSource.new( addr, port, "Failed to connect" ).close
# Sleep to be sure we don't try to connect too soon. That wouldn't
# be a problem for the destination, but it would prevent us from
# determining success or failure here in the case where we try to
# reconnect before the destination has tidied up after the first
# thread went away.
FakeSource.new(addr, port, 'Failed to connect').close
# Sleep to be sure we don't try to connect too soon. That wouldn't
# be a problem for the destination, but it would prevent us from
# determining success or failure here in the case where we try to
# reconnect before the destination has tidied up after the first
# thread went away.
sleep(0.5)
FakeSource.new( addr, port, "Failed to reconnect" ).close
FakeSource.new(addr, port, 'Failed to reconnect').close
exit 0

View File

@@ -11,10 +11,10 @@ include FlexNBD
addr, port, srv_pid = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting" )
client = FakeSource.new(addr, port, 'Timed out connecting')
client.read_hello
client.write_write_request( 0, 8 )
client.write_data( "12345678" )
client.write_write_request(0, 8)
client.write_data('12345678')
# Use system "kill" rather than Process.kill because Process.kill
# doesn't seem to work
@@ -25,12 +25,11 @@ client.close
system "kill -CONT #{srv_pid}"
sleep(0.25)
begin
client2 = FakeSource.new( addr, port, "Expected timeout" )
fail "Unexpected reconnection"
client2 = FakeSource.new(addr, port, 'Expected timeout')
raise 'Unexpected reconnection'
rescue Timeout::Error
# expected
end

View File

@@ -10,10 +10,10 @@ include FlexNBD
addr, port, srv_pid = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting" )
client = FakeSource.new(addr, port, 'Timed out connecting')
client.read_hello
client.write_write_request( 0, 8 )
client.write_data( "12345678" )
client.write_write_request(0, 8)
client.write_data('12345678')
client.write_entrust_request
client.read_response
@@ -21,13 +21,11 @@ client.close
sleep(0.25)
begin
client2 = FakeSource.new( addr, port, "Expected timeout" )
fail "Unexpected reconnection"
client2 = FakeSource.new(addr, port, 'Expected timeout')
raise 'Unexpected reconnection'
rescue Timeout::Error
# expected
end
exit(0)

View File

@@ -12,13 +12,12 @@ include FlexNBD
addr, port = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting." )
client = FakeSource.new(addr, port, 'Timed out connecting.')
client.read_hello
client.close
sleep(0.2)
FakeSource.new( addr, port, "Timed out reconnecting." )
FakeSource.new(addr, port, 'Timed out reconnecting.')
exit(0)

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# We connect, pause the server, issue a write request, disconnect,
# then cont the server. This ensures that our disconnect happens
# while the server is trying to read the write data.
@@ -10,11 +8,11 @@ include FlexNBD
addr, port, srv_pid = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting" )
client = FakeSource.new(addr, port, 'Timed out connecting')
client.read_hello
system "kill -STOP #{srv_pid}"
client.write_write_request( 0, 8 )
client.write_write_request(0, 8)
client.close
system "kill -CONT #{srv_pid}"
@@ -24,7 +22,7 @@ system "kill -CONT #{srv_pid}"
sleep(0.25)
# ...and can we reconnect?
client2 = FakeSource.new( addr, port, "Timed out connecting" )
client2 = FakeSource.new(addr, port, 'Timed out connecting')
client2.close
exit(0)

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# We connect, pause the server, issue a write request, send data,
# disconnect, then cont the server. This ensures that our disconnect
# happens before the server can try to write the reply.
@@ -10,13 +8,13 @@ include FlexNBD
addr, port, srv_pid = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting" )
client = FakeSource.new(addr, port, 'Timed out connecting')
client.read_hello
system "kill -STOP #{srv_pid}"
client.write_write_request( 0, 8 )
client.write_data( "12345678" )
client.write_write_request(0, 8)
client.write_data('12345678')
client.close
system "kill -CONT #{srv_pid}"
@@ -27,7 +25,7 @@ system "kill -CONT #{srv_pid}"
sleep(0.25)
# ...and can we reconnect?
client2 = FakeSource.new( addr, port, "Timed out reconnecting" )
client2 = FakeSource.new(addr, port, 'Timed out reconnecting')
client2.close
exit(0)

View File

@@ -8,10 +8,9 @@ include FlexNBD
addr, port, srv_pid, newaddr, newport = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting" )
client.write_read_request( 0, 8 )
client.read_raw( 4 )
client = FakeSource.new(addr, port, 'Timed out connecting')
client.write_read_request(0, 8)
client.read_raw(4)
client.close
exit(0)

View File

@@ -10,9 +10,9 @@ include FlexNBD
addr, port = *ARGV
client1 = FakeSource.new( addr, port, "Timed out connecting" )
client1 = FakeSource.new(addr, port, 'Timed out connecting')
sleep(0.25)
client2 = FakeSource.new( addr, port, "Timed out connecting a second time" )
client2 = FakeSource.new(addr, port, 'Timed out connecting a second time')
# This is the expected source crashing after connect
client1.close

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# We connect from a local address which should be blocked, sleep for a
# bit, then try to read from the socket. We should get an instant EOF
# as we've been cut off by the destination.
@@ -11,10 +9,9 @@ include FlexNBD
addr, port = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting", "127.0.0.6" )
sleep( 0.25 )
client = FakeSource.new(addr, port, 'Timed out connecting', '127.0.0.6')
sleep(0.25)
rsp = client.disconnected? ? 0 : 1
client.close
exit(rsp)

View File

@@ -7,10 +7,10 @@
# listening for an incoming migration.
addr, port = *ARGV
require "flexnbd/fake_source"
require 'flexnbd/fake_source'
include FlexNBD
client = FakeSource.new( addr, port, "Timed out connecting" )
client = FakeSource.new(addr, port, 'Timed out connecting')
client.read_hello
# Now we do two things:
@@ -24,16 +24,16 @@ client.read_hello
kidpid = fork do
client.close
new_client = nil
sleep( FlexNBD::CLIENT_MAX_WAIT_SECS + 1 )
new_client = FakeSource.new( addr, port, "Timed out reconnecting." )
sleep(FlexNBD::CLIENT_MAX_WAIT_SECS + 1)
new_client = FakeSource.new(addr, port, 'Timed out reconnecting.')
new_client.read_hello
exit 0
end
# Sleep for longer than the child, to give the flexnbd process a bit
# of slop
sleep( FlexNBD::CLIENT_MAX_WAIT_SECS + 3 )
sleep(FlexNBD::CLIENT_MAX_WAIT_SECS + 3)
client.close
_,status = Process.waitpid2( kidpid )
_, status = Process.waitpid2(kidpid)
exit status.exitstatus

View File

@@ -9,10 +9,10 @@ include FlexNBD
addr, port, pid = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting." )
client = FakeSource.new(addr, port, 'Timed out connecting.')
client.read_hello
Process.kill( "TERM", pid.to_i )
Process.kill('TERM', pid.to_i)
sleep(0.2)
client.close

View File

@@ -9,10 +9,9 @@ include FlexNBD
addr, port, srv_pid, newaddr, newport = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting" )
client.send_mirror()
client = FakeSource.new(addr, port, 'Timed out connecting')
client.send_mirror
sleep(1)
exit( 0 )
exit(0)

View File

@@ -1,6 +1,4 @@
#!/usr/bin/env ruby
# encoding: utf-8
# Connect, read the hello then make a write request with an impossible
# (from,len) pair. We expect an error response, and not to be
# disconnected.
@@ -13,20 +11,20 @@ include FlexNBD
addr, port = *ARGV
client = FakeSource.new( addr, port, "Timed out connecting" )
client = FakeSource.new(addr, port, 'Timed out connecting')
hello = client.read_hello
client.write_write_request( hello[:size]+1, 32, "myhandle" )
client.write_data("1"*32)
client.write_write_request(hello[:size] + 1, 32, 'myhandle')
client.write_data('1' * 32)
response = client.read_response
fail "Not an error" if response[:error] == 0
fail "Wrong handle" unless "myhandle" == response[:handle]
raise 'Not an error' if response[:error] == 0
raise 'Wrong handle' unless response[:handle] == 'myhandle'
client.write_write_request( 0, 32 )
client.write_data( "2"*32 )
client.write_write_request(0, 32)
client.write_data('2' * 32)
success_response = client.read_response
fail "Second write failed" unless success_response[:error] == 0
raise 'Second write failed' unless success_response[:error] == 0
client.close
exit(0)

View File

@@ -3,13 +3,13 @@
#
class FileWriter
def initialize(filename, blocksize)
@fh = File.open(filename, "w+")
@fh = File.open(filename, 'w+')
@blocksize = blocksize
@pattern = ""
@pattern = ''
end
def size
@blocksize * @pattern.split("").size
@blocksize * @pattern.split('').size
end
# We write in fixed block sizes, given by "blocksize"
@@ -20,8 +20,8 @@ class FileWriter
def write(data)
@pattern += data
data.split("").each do |code|
if code == "_"
data.split('').each do |code|
if code == '_'
@fh.seek(@blocksize, IO::SEEK_CUR)
else
@fh.write(data(code))
@@ -31,15 +31,14 @@ class FileWriter
self
end
# Returns what the data ought to be at the given offset and length
#
def read_original( off, len )
patterns = @pattern.split( "" )
patterns.zip( (0...patterns.length).to_a ).
map { |blk, blk_off|
def read_original(off, len)
patterns = @pattern.split('')
patterns.zip((0...patterns.length).to_a)
.map do |blk, blk_off|
data(blk, blk_off)
}.join[off...(off+len)]
end.join[off...(off + len)]
end
# Read what's actually in the file
@@ -60,68 +59,66 @@ class FileWriter
protected
def data(code, at=@fh.tell)
def data(code, at = @fh.tell)
case code
when "0", "_"
"\0" * @blocksize
when "X"
"X" * @blocksize
when "f"
r = ""
(@blocksize/4).times do
r += [at].pack("I")
at += 4
end
r
else
raise "Unknown character '#{block}'"
when '0', '_'
"\0" * @blocksize
when 'X'
'X' * @blocksize
when 'f'
r = ''
(@blocksize / 4).times do
r += [at].pack('I')
at += 4
end
r
else
raise "Unknown character '#{block}'"
end
end
end
if __FILE__==$0
if $PROGRAM_NAME == __FILE__
require 'tempfile'
require 'test/unit'
class FileWriterTest < Test::Unit::TestCase
def test_read_original_zeros
Tempfile.open("test_read_original_zeros") do |tempfile|
Tempfile.open('test_read_original_zeros') do |tempfile|
tempfile.close
file = FileWriter.new( tempfile.path, 4096 )
file.write( "0" )
assert_equal file.read( 0, 4096 ), file.read_original( 0, 4096 )
assert( file.untouched?(0,4096) , "Untouched file was touched." )
file = FileWriter.new(tempfile.path, 4096)
file.write('0')
assert_equal file.read(0, 4096), file.read_original(0, 4096)
assert(file.untouched?(0, 4096), 'Untouched file was touched.')
end
end
def test_read_original_offsets
Tempfile.open("test_read_original_offsets") do |tempfile|
Tempfile.open('test_read_original_offsets') do |tempfile|
tempfile.close
file = FileWriter.new( tempfile.path, 4096 )
file.write( "f" )
assert_equal file.read( 0, 4096 ), file.read_original( 0, 4096 )
assert( file.untouched?(0,4096) , "Untouched file was touched." )
file = FileWriter.new(tempfile.path, 4096)
file.write('f')
assert_equal file.read(0, 4096), file.read_original(0, 4096)
assert(file.untouched?(0, 4096), 'Untouched file was touched.')
end
end
def test_file_size
Tempfile.open("test_file_size") do |tempfile|
Tempfile.open('test_file_size') do |tempfile|
tempfile.close
file = FileWriter.new( tempfile.path, 4096 )
file.write( "f" )
assert_equal 4096, File.stat( tempfile.path ).size
file = FileWriter.new(tempfile.path, 4096)
file.write('f')
assert_equal 4096, File.stat(tempfile.path).size
end
end
def test_read_original_size
Tempfile.open("test_read_original_offsets") do |tempfile|
Tempfile.open('test_read_original_offsets') do |tempfile|
tempfile.close
file = FileWriter.new( tempfile.path, 4)
file.write( "f"*4 )
file = FileWriter.new(tempfile.path, 4)
file.write('f' * 4)
assert_equal 4, file.read_original(0, 4).length
end
end
end
end

View File

@@ -4,28 +4,26 @@ require 'open3'
require 'timeout'
require 'rexml/document'
require 'rexml/streamlistener'
require 'English'
Thread.abort_on_exception = true
class Executor
attr_reader :pid
def run( cmd )
@pid = fork do exec cmd end
def run(cmd)
@pid = fork { exec cmd }
end
end # class Executor
class ValgrindExecutor
attr_reader :pid
def run( cmd )
@pid = fork do exec "valgrind --track-origins=yes --suppressions=custom.supp #{cmd}" end
def run(cmd)
@pid = fork { exec "valgrind --track-origins=yes --suppressions=custom.supp #{cmd}" }
end
end # class ValgrindExecutor
class ValgrindKillingExecutor
attr_reader :pid
@@ -33,10 +31,10 @@ class ValgrindKillingExecutor
attr_accessor :what, :kind, :pid
attr_reader :backtrace
def initialize
@backtrace=[]
@what = ""
@kind = ""
@pid = ""
@backtrace = []
@what = ''
@kind = ''
@pid = ''
end
def add_frame
@@ -56,115 +54,104 @@ class ValgrindKillingExecutor
end
def to_s
([@what + " (#{@kind}) in #{@pid}"] + @backtrace.map{|h| "#{h[:file]}:#{h[:line]} #{h[:fn]}" }).join("\n")
([@what + " (#{@kind}) in #{@pid}"] + @backtrace.map { |h| "#{h[:file]}:#{h[:line]} #{h[:fn]}" }).join("\n")
end
end # class Error
class ErrorListener
include REXML::StreamListener
def initialize( killer )
def initialize(killer)
@killer = killer
@error = Error.new
@found = false
end
def text( text )
def text(text)
@text = text
end
def tag_start(tag, attrs)
def tag_start(tag, _attrs)
case tag.to_s
when "error"
when 'error'
@found = true
when "frame"
when 'frame'
@error.add_frame
end
end
def tag_end(tag)
case tag.to_s
when "what"
when 'what'
@error.what = @text if @found
@text = ""
when "kind"
@text = ''
when 'kind'
@error.kind = @text if @found
when "file"
@error.add_file( @text ) if @found
when "fn"
@error.add_fn( @text ) if @found
when "line"
@error.add_line( @text ) if @found
when "error", "stack"
if @found
@killer.call( @error )
end
when "pid"
@error.pid=@text
when 'file'
@error.add_file(@text) if @found
when 'fn'
@error.add_fn(@text) if @found
when 'line'
@error.add_line(@text) if @found
when 'error', 'stack'
@killer.call(@error) if @found
when 'pid'
@error.pid = @text
end
end
end # class ErrorListener
class DebugErrorListener < ErrorListener
def text( txt )
def text(txt)
print txt
super( txt )
super(txt)
end
def tag_start( tag, attrs )
def tag_start(tag, attrs)
print "<#{tag}>"
super( tag, attrs )
super(tag, attrs)
end
def tag_end( tag )
def tag_end(tag)
print "</#{tag}>"
super( tag )
super(tag)
end
end
def initialize
@pid = nil
end
def run( cmd )
def run(cmd)
@io_r, io_w = IO.pipe
@pid = fork do exec( "valgrind --suppressions=custom.supp --xml=yes --xml-fd=#{io_w.fileno} " + cmd ) end
launch_watch_thread( @pid, @io_r )
@pid = fork { exec("valgrind --suppressions=custom.supp --xml=yes --xml-fd=#{io_w.fileno} " + cmd) }
launch_watch_thread(@pid, @io_r)
@pid
end
def call( err )
$stderr.puts "*"*72
$stderr.puts "* Valgrind error spotted:"
$stderr.puts err.to_s.split("\n").map{|s| " #{s}"}
$stderr.puts "*"*72
Process.kill( "KILL", @pid )
def call(err)
warn '*' * 72
warn '* Valgrind error spotted:'
warn err.to_s.split("\n").map { |s| " #{s}" }
warn '*' * 72
Process.kill('KILL', @pid)
exit(1)
end
private
def pick_listener
ENV['DEBUG'] ? DebugErrorListener : ErrorListener
end
def launch_watch_thread(pid, io_r)
def launch_watch_thread(_pid, io_r)
Thread.start do
io_source = REXML::IOSource.new( io_r )
listener = pick_listener.new( self )
REXML::Document.parse_stream( io_source, listener )
io_source = REXML::IOSource.new(io_r)
listener = pick_listener.new(self)
REXML::Document.parse_stream(io_source, listener)
end
end
end # class ValgrindExecutor
module FlexNBD
# Noddy test class to exercise FlexNBD from the outside for testing.
#
@@ -173,7 +160,7 @@ module FlexNBD
class << self
def counter
Dir['tmp/*'].select{|f| File.file?(f)}.length + 1
Dir['tmp/*'].select { |f| File.file?(f) }.length + 1
end
end
@@ -189,19 +176,18 @@ module FlexNBD
end
end
def build_debug_opt
if @do_debug
"--verbose"
'--verbose'
else
"--quiet"
'--quiet'
end
end
attr_accessor :prefetch_proxy
def initialize( bin, ip, port )
@bin = bin
def initialize(bin, ip, port)
@bin = bin
@do_debug = ENV['DEBUG']
@debug = build_debug_opt
raise "#{bin} not executable" unless File.executable?(bin)
@@ -209,21 +195,20 @@ module FlexNBD
@ctrl = "/tmp/.flexnbd.ctrl.#{Time.now.to_i}.#{rand}"
@ip = ip
@port = port
@pid = @wait_thread = nil
@kill = []
@prefetch_proxy = false
end
def debug?
!!@do_debug
end
def debug( msg )
$stderr.puts msg if debug?
def debug(msg)
warn msg if debug?
end
def serve_cmd( file, acl )
def serve_cmd(file, acl)
"#{bin} serve "\
"--addr #{ip} "\
"--port #{port} "\
@@ -233,8 +218,7 @@ module FlexNBD
"#{acl.join(' ')}"
end
def listen_cmd( file, acl )
def listen_cmd(file, acl)
"#{bin} listen "\
"--addr #{ip} "\
"--port #{port} "\
@@ -244,18 +228,17 @@ module FlexNBD
"#{acl.join(' ')}"
end
def proxy_cmd( connect_ip, connect_port )
def proxy_cmd(connect_ip, connect_port)
"#{bin}-proxy "\
"--addr #{ip} "\
"--port #{port} "\
"--conn-addr #{connect_ip} "\
"--conn-port #{connect_port} "\
"#{prefetch_proxy ? "--cache " : ""}"\
"#{prefetch_proxy ? '--cache ' : ''}"\
"#{@debug}"
end
def read_cmd( offset, length )
def read_cmd(offset, length)
"#{bin} read "\
"--addr #{ip} "\
"--port #{port} "\
@@ -264,8 +247,7 @@ module FlexNBD
"--size #{length}"
end
def write_cmd( offset, data )
def write_cmd(offset, data)
"#{bin} write "\
"--addr #{ip} "\
"--port #{port} "\
@@ -274,30 +256,29 @@ module FlexNBD
"--size #{data.length}"
end
def base_mirror_opts( dest_ip, dest_port )
def base_mirror_opts(dest_ip, dest_port)
"--addr #{dest_ip} "\
"--port #{dest_port} "\
"--sock #{ctrl} "\
end
def unlink_mirror_opts( dest_ip, dest_port )
"#{base_mirror_opts( dest_ip, dest_port )} "\
"--unlink "
def unlink_mirror_opts(dest_ip, dest_port)
"#{base_mirror_opts(dest_ip, dest_port)} "\
'--unlink '
end
def base_mirror_cmd( opts )
def base_mirror_cmd(opts)
"#{@bin} mirror "\
"#{opts} "\
"#{@debug}"
end
def mirror_cmd(dest_ip, dest_port)
base_mirror_cmd( base_mirror_opts( dest_ip, dest_port ) )
base_mirror_cmd(base_mirror_opts(dest_ip, dest_port))
end
def mirror_unlink_cmd( dest_ip, dest_port )
base_mirror_cmd( unlink_mirror_opts( dest_ip, dest_port ) )
def mirror_unlink_cmd(dest_ip, dest_port)
base_mirror_cmd(unlink_mirror_opts(dest_ip, dest_port))
end
def break_cmd
@@ -312,58 +293,64 @@ module FlexNBD
"#{@debug}"
end
def acl_cmd( *acl )
def acl_cmd(*acl)
"#{@bin} acl " \
"--sock #{ctrl} "\
"#{@debug} "\
"#{acl.join " "}"
"#{acl.join ' '}"
end
def run_serve_cmd(cmd)
File.unlink(ctrl) if File.exists?(ctrl)
debug( cmd )
File.unlink(ctrl) if File.exist?(ctrl)
debug(cmd)
@pid = @executor.run( cmd )
@pid = @executor.run(cmd)
while !File.socket?(ctrl)
until File.socket?(ctrl)
pid, status = Process.wait2(@pid, Process::WNOHANG)
raise "server did not start (#{cmd})" if pid
sleep 0.1
end
start_wait_thread( @pid )
start_wait_thread(@pid)
at_exit { kill }
end
private :run_serve_cmd
def serve( file, *acl)
cmd = serve_cmd( file, acl )
run_serve_cmd( cmd )
sleep( 0.2 ) until File.exists?( ctrl )
def serve(file, *acl)
cmd = serve_cmd(file, acl)
run_serve_cmd(cmd)
sleep(0.2) until File.exist?(ctrl)
end
def listen(file, *acl)
run_serve_cmd( listen_cmd( file, acl ) )
run_serve_cmd(listen_cmd(file, acl))
end
def tcp_server_open?
# raises if the other side doesn't accept()
sock = TCPSocket.new(ip, port) rescue nil
sock = begin
TCPSocket.new(ip, port)
rescue StandardError
nil
end
success = !!sock
( sock.close rescue nil) if sock
if sock
(begin
sock.close
rescue StandardError
nil
end)
end
success
end
def proxy( connect_ip, connect_port )
cmd = proxy_cmd( connect_ip, connect_port )
debug( cmd )
def proxy(connect_ip, connect_port)
cmd = proxy_cmd(connect_ip, connect_port)
debug(cmd)
@pid = @executor.run( cmd )
@pid = @executor.run(cmd)
until tcp_server_open?
pid, status = Process.wait2(@pid, Process::WNOHANG)
@@ -371,31 +358,29 @@ module FlexNBD
sleep 0.1
end
start_wait_thread( @pid )
start_wait_thread(@pid)
at_exit { kill }
end
def start_wait_thread( pid )
def start_wait_thread(pid)
@wait_thread = Thread.start do
_, status = Process.waitpid2( pid )
_, status = Process.waitpid2(pid)
if @kill
if status.signaled?
fail "flexnbd quit with a bad signal: #{status.inspect}" unless
raise "flexnbd quit with a bad signal: #{status.inspect}" unless
@kill.include? status.termsig
else
fail "flexnbd quit with a bad status: #{status.inspect}" unless
raise "flexnbd quit with a bad status: #{status.inspect}" unless
@kill.include? status.exitstatus
end
else
$stderr.puts "flexnbd #{self.pid} quit"
fail "flexnbd #{self.pid} quit early with status #{status.to_i}"
warn "flexnbd #{self.pid} quit"
raise "flexnbd #{self.pid} quit early with status #{status.to_i}"
end
end
end
def can_die(*status)
status = [0] if status.empty?
@kill += status
@@ -407,7 +392,7 @@ module FlexNBD
can_die(1)
if @pid
begin
Process.kill("INT", @pid)
Process.kill('INT', @pid)
rescue Errno::ESRCH => e
# already dead. Presumably this means it went away after a
# can_die() call.
@@ -417,63 +402,60 @@ module FlexNBD
end
def read(offset, length)
cmd = read_cmd( offset, length )
debug( cmd )
cmd = read_cmd(offset, length)
debug(cmd)
IO.popen(cmd) do |fh|
return fh.read
end
raise IOError.new "NBD read failed" unless $?.success?
raise IOError, 'NBD read failed' unless $CHILD_STATUS.success?
out
end
def write(offset, data)
cmd = write_cmd( offset, data )
debug( cmd )
cmd = write_cmd(offset, data)
debug(cmd)
IO.popen(cmd, "w") do |fh|
IO.popen(cmd, 'w') do |fh|
fh.write(data)
end
raise IOError.new "NBD write failed" unless $?.success?
raise IOError, 'NBD write failed' unless $CHILD_STATUS.success?
nil
end
def join
@wait_thread.join
end
def mirror_unchecked(dest_ip, dest_port, _bandwidth = nil, _action = nil, timeout = nil)
cmd = mirror_cmd(dest_ip, dest_port)
debug(cmd)
def mirror_unchecked( dest_ip, dest_port, bandwidth=nil, action=nil, timeout=nil )
cmd = mirror_cmd( dest_ip, dest_port)
debug( cmd )
maybe_timeout( cmd, timeout )
maybe_timeout(cmd, timeout)
end
def mirror_unlink(dest_ip, dest_port, timeout = nil)
cmd = mirror_unlink_cmd(dest_ip, dest_port)
debug(cmd)
def mirror_unlink( dest_ip, dest_port, timeout=nil )
cmd = mirror_unlink_cmd( dest_ip, dest_port )
debug( cmd )
maybe_timeout( cmd, timeout )
maybe_timeout(cmd, timeout)
end
def maybe_timeout(cmd, timeout=nil )
stdout, stderr = "",""
def maybe_timeout(cmd, timeout = nil)
stdout = ''
stderr = ''
stat = nil
run = Proc.new do
run = proc do
# Ruby 1.9 changed the popen3 api. instead of 3 args, the block
# gets 4. Not only that, but it no longer sets $?, so we have to
# go elsewhere for the process' exit status.
Open3.popen3( cmd ) do |io_in, io_out, io_err, maybe_thr|
Open3.popen3(cmd) do |io_in, io_out, io_err, maybe_thr|
io_in.close
stdout.replace io_out.read
stderr.replace io_err.read
stat = maybe_thr.value if maybe_thr
end
stat ||= $?
stat ||= $CHILD_STATUS
end
if timeout
@@ -485,85 +467,73 @@ module FlexNBD
[stdout, stderr, stat]
end
def mirror(dest_ip, dest_port, bandwidth=nil, action=nil)
stdout, stderr, status = mirror_unchecked( dest_ip, dest_port, bandwidth, action )
raise IOError.new( "Migrate command failed\n" + stderr) unless status.success?
def mirror(dest_ip, dest_port, bandwidth = nil, action = nil)
stdout, stderr, status = mirror_unchecked(dest_ip, dest_port, bandwidth, action)
raise IOError, "Migrate command failed\n" + stderr unless status.success?
stdout
end
def break(timeout=nil)
def break(timeout = nil)
cmd = break_cmd
debug( cmd )
debug(cmd)
maybe_timeout( cmd, timeout )
maybe_timeout(cmd, timeout)
end
def acl(*acl)
cmd = acl_cmd( *acl )
debug( cmd )
cmd = acl_cmd(*acl)
debug(cmd)
maybe_timeout( cmd, 2 )
maybe_timeout(cmd, 2)
end
def status(timeout = nil)
cmd = status_cmd
debug(cmd)
def status( timeout = nil )
cmd = status_cmd()
debug( cmd )
o,e = maybe_timeout( cmd, timeout )
o, e = maybe_timeout(cmd, timeout)
[parse_status(o), e]
end
def launched?
!!@pid
end
def paused
Process.kill( "STOP", @pid )
Process.kill('STOP', @pid)
yield
ensure
Process.kill( "CONT", @pid )
Process.kill('CONT', @pid)
end
protected
def control_command(*args)
raise "Server not running" unless @pid
raise 'Server not running' unless @pid
args = args.compact
UNIXSocket.open(@ctrl) do |u|
u.write(args.join("\n") + "\n")
code, message = u.readline.split(": ", 2)
code, message = u.readline.split(': ', 2)
return [code, message]
end
end
def parse_status( status )
def parse_status(status)
hsh = {}
status.split(" ").each do |part|
status.split(' ').each do |part|
next if part.strip.empty?
a,b = part.split("=")
a, b = part.split('=')
b.strip!
b = true if b == "true"
b = false if b == "false"
b = true if b == 'true'
b = false if b == 'false'
hsh[a.strip] = b
end
hsh
end
end
end

View File

@@ -1,10 +1,7 @@
# encoding: utf-8
module FlexNBD
def self.binary( str )
def self.binary(str)
if str.respond_to? :force_encoding
str.force_encoding "ASCII-8BIT"
str.force_encoding 'ASCII-8BIT'
else
str
end
@@ -13,36 +10,33 @@ module FlexNBD
# eeevil is his one and only name...
def self.read_constants
parents = []
current = File.expand_path(".")
while current != "/"
current = File.expand_path('.')
while current != '/'
parents << current
current = File.expand_path( File.join( current, ".." ) )
current = File.expand_path(File.join(current, '..'))
end
source_root = parents.find do |dirname|
File.directory?( File.join( dirname, "src" ) )
File.directory?(File.join(dirname, 'src'))
end
fail "No source root!" unless source_root
raise 'No source root!' unless source_root
headers = Dir[File.join( source_root, "src", "{common,proxy,server}","*.h" ) ]
headers = Dir[File.join(source_root, 'src', '{common,proxy,server}', '*.h')]
headers.each do |header_filename|
txt_lines = File.readlines( header_filename )
txt_lines = File.readlines(header_filename)
txt_lines.each do |line|
if line =~ /^#\s*define\s+([A-Z0-9_]+)\s+(\d+)\s*$/
# Bodge until I can figure out what to do with #ifdefs
const_set($1, $2.to_i) unless constants.include?( $1 )
const_set(Regexp.last_match(1), Regexp.last_match(2).to_i) unless const_defined?(Regexp.last_match(1))
end
end
end
end
read_constants()
read_constants
REQUEST_MAGIC = binary("\x25\x60\x95\x13") unless defined?(REQUEST_MAGIC)
REPLY_MAGIC = binary("\x67\x44\x66\x98") unless defined?(REPLY_MAGIC)
end # module FlexNBD

View File

@@ -1,5 +1,3 @@
# encoding: utf-8
require 'socket'
require 'timeout'
@@ -7,114 +5,104 @@ require 'flexnbd/constants'
module FlexNBD
class FakeDest
class Client
def initialize( sock )
def initialize(sock)
@sock = sock
end
def write_hello( opts = {} )
@sock.write( "NBDMAGIC" )
def write_hello(opts = {})
@sock.write('NBDMAGIC')
if opts[:magic] == :wrong
write_rand( @sock, 8 )
write_rand(@sock, 8)
else
@sock.write( "\x00\x00\x42\x02\x81\x86\x12\x53" )
@sock.write("\x00\x00\x42\x02\x81\x86\x12\x53")
end
if opts[:size] == :wrong
write_rand( @sock, 8 )
write_rand(@sock, 8)
else
@sock.write( "\x00\x00\x00\x00\x00\x00\x10\x00" )
@sock.write("\x00\x00\x00\x00\x00\x00\x10\x00")
end
@sock.write( "\x00" * 128 )
@sock.write("\x00" * 128)
end
def write_rand( sock, len )
len.times do sock.write( rand(256).chr ) end
def write_rand(sock, len)
len.times { sock.write(rand(256).chr) }
end
def read_request()
def read_request
req = @sock.read(28)
magic_s = req[0 ... 4 ]
type_s = req[4 ... 8 ]
handle_s = req[8 ... 16]
from_s = req[16 ... 24]
len_s = req[24 ... 28]
magic_s = req[0...4]
type_s = req[4...8]
handle_s = req[8...16]
from_s = req[16...24]
len_s = req[24...28]
{
:magic => magic_s,
:type => type_s.unpack("N").first,
:handle => handle_s,
:from => self.class.parse_be64( from_s ),
:len => len_s.unpack( "N").first
magic: magic_s,
type: type_s.unpack('N').first,
handle: handle_s,
from: self.class.parse_be64(from_s),
len: len_s.unpack('N').first
}
end
def write_error( handle )
write_reply( handle, 1 )
def write_error(handle)
write_reply(handle, 1)
end
def disconnected?
begin
Timeout.timeout(2) do
@sock.read(1) == nil
end
rescue Timeout::Error
return false
Timeout.timeout(2) do
@sock.read(1).nil?
end
rescue Timeout::Error
return false
end
def write_reply( handle, err=0, opts={} )
def write_reply(handle, err = 0, opts = {})
if opts[:magic] == :wrong
write_rand( @sock, 4 )
write_rand(@sock, 4)
else
@sock.write( ::FlexNBD::REPLY_MAGIC )
@sock.write(::FlexNBD::REPLY_MAGIC)
end
@sock.write( [err].pack("N") )
@sock.write( handle )
@sock.write([err].pack('N'))
@sock.write(handle)
end
def close
@sock.close
end
def read_data( len )
@sock.read( len )
def read_data(len)
@sock.read(len)
end
def write_data( len )
@sock.write( len )
def write_data(len)
@sock.write(len)
end
def self.parse_be64(str)
raise "String is the wrong length: 8 bytes expected (#{str.length} received)" unless
str.length == 8
top, bottom = str.unpack("NN")
top, bottom = str.unpack('NN')
(top << 32) + bottom
end
def receive_mirror( opts = {} )
write_hello()
def receive_mirror(opts = {})
write_hello
loop do
req = read_request
case req[:type]
when 1
read_data( req[:len] )
write_reply( req[:handle] )
when 65536
write_reply( req[:handle], opts[:err] == :entrust ? 1 : 0 )
read_data(req[:len])
write_reply(req[:handle])
when 65_536
write_reply(req[:handle], opts[:err] == :entrust ? 1 : 0)
break
else
raise "Unexpected request: #{req.inspect}"
@@ -129,16 +117,13 @@ module FlexNBD
raise "Not a disconnect: #{req.inspect}"
end
end
end # class Client
def initialize( addr, port )
@sock = TCPServer.new( addr, port )
def initialize(addr, port)
@sock = TCPServer.new(addr, port)
end
def accept( err_msg = "Timed out waiting for a connection", timeout = 5)
def accept(err_msg = 'Timed out waiting for a connection', timeout = 5)
client_sock = nil
begin
@@ -146,21 +131,16 @@ module FlexNBD
client_sock = @sock.accept
end
rescue Timeout::Error
raise Timeout::Error.new(err_msg)
raise Timeout::Error, err_msg
end
client_sock
Client.new( client_sock )
Client.new(client_sock)
end
def close
@sock.close
end
end # module FakeDest
end # module FlexNBD

View File

@@ -1,112 +1,117 @@
# encoding: utf-8
require 'socket'
require "timeout"
require 'timeout'
require 'flexnbd/constants'
module FlexNBD
class FakeSource
def initialize( addr, port, err_msg, source_addr=nil, source_port=0 )
timing_out( 2, err_msg ) do
def initialize(addr, port, err_msg, source_addr = nil, source_port = 0)
timing_out(2, err_msg) do
begin
@sock = if source_addr
TCPSocket.new( addr, port, source_addr, source_port )
TCPSocket.new(addr, port, source_addr, source_port)
else
TCPSocket.new( addr, port )
TCPSocket.new(addr, port)
end
rescue Errno::ECONNREFUSED
$stderr.puts "Connection refused, retrying"
warn 'Connection refused, retrying'
sleep(0.2)
retry
end
end
end
def close
@sock.close
end
def read_hello
timing_out(::FlexNBD::MS_HELLO_TIME_SECS,
'Timed out waiting for hello.') do
raise 'No hello.' unless (hello = @sock.read(152)) &&
hello.length == 152
def read_hello()
timing_out( ::FlexNBD::MS_HELLO_TIME_SECS,
"Timed out waiting for hello." ) do
fail "No hello." unless (hello = @sock.read( 152 )) &&
hello.length==152
passwd_s = hello[0..7]
magic = hello[8..15].unpack('Q>').first
size = hello[16..23].unpack('Q>').first
flags = hello[24..27].unpack('L>').first
reserved = hello[28..-1]
magic_s = hello[0..7]
ignore_s= hello[8..15]
size_s = hello[16..23]
size_h, size_l = size_s.unpack("NN")
size = (size_h << 32) + size_l
return { :magic => magic_s, :size => size }
return { passwd: passwd_s, magic: magic, size: size, flags: flags, reserved: reserved }
end
end
def send_request(type, handle = 'myhandle', from = 0, len = 0, magic = REQUEST_MAGIC, flags = 0)
raise 'Bad handle' unless handle.length == 8
def send_request( type, handle="myhandle", from=0, len=0, magic=REQUEST_MAGIC )
fail "Bad handle" unless handle.length == 8
@sock.write( magic )
@sock.write( [type].pack( 'N' ) )
@sock.write( handle )
@sock.write( [n64( from )].pack( 'q' ) )
@sock.write( [len].pack( 'N' ) )
@sock.write(magic)
@sock.write([flags].pack('n'))
@sock.write([type].pack('n'))
@sock.write(handle)
@sock.write([n64(from)].pack('q'))
@sock.write([len].pack('N'))
end
def write_write_request( from, len, handle="myhandle" )
send_request( 1, handle, from, len )
def write_write_request(from, len, handle = 'myhandle')
send_request(1, handle, from, len)
end
def write_entrust_request( handle="myhandle" )
send_request( 65536, handle )
def write_write_request_with_fua(from, len, handle = 'myhandle')
send_request(1, handle, from, len, REQUEST_MAGIC, 1)
end
def write_disconnect_request( handle="myhandle" )
send_request( 2, handle )
def write_flush_request(handle = 'myhandle')
send_request(3, handle, 0, 0)
end
def write_read_request( from, len, handle="myhandle" )
send_request( 0, "myhandle", from, len )
def write_entrust_request(handle = 'myhandle')
send_request(65_536, handle)
end
def write_data( data )
@sock.write( data )
def write_disconnect_request(handle = 'myhandle')
send_request(2, handle)
end
def write_read_request(from, len, _handle = 'myhandle')
send_request(0, 'myhandle', from, len)
end
def write_data(data)
@sock.write(data)
end
# Handy utility
def read( from, len )
timing_out( 2, "Timed out reading" ) do
send_request( 0, "myhandle", from, len )
read_raw( len )
def read(from, len)
timing_out(2, 'Timed out reading') do
send_request(0, 'myhandle', from, len)
read_raw(len)
end
end
def read_raw( len )
@sock.read( len )
def read_raw(len)
@sock.read(len)
end
def send_mirror
read_hello()
write( 0, "12345678" )
read_response()
write_disconnect_request()
close()
read_hello
write(0, '12345678')
read_response
write_disconnect_request
close
end
def write( from, data )
write_write_request( from, data.length )
write_data( data )
def write(from, data)
write_write_request(from, data.length)
write_data(data)
end
def write_with_fua(from, data)
write_write_request_with_fua(from, data.length)
write_data(data)
end
def flush
write_flush_request
end
def read_response
magic = @sock.read(4)
@@ -114,29 +119,25 @@ module FlexNBD
handle = @sock.read(8)
{
:magic => magic,
:error => error_s.unpack("N").first,
:handle => handle
magic: magic,
error: error_s.unpack('N').first,
handle: handle
}
end
def disconnected?
result = nil
Timeout.timeout( 2 ) { result = ( @sock.read(1) == nil ) }
Timeout.timeout(2) { result = @sock.read(1).nil? }
result
end
def timing_out( time, msg )
begin
Timeout.timeout( time ) do
yield
end
rescue Timeout::Error
$stderr.puts msg
exit 1
def timing_out(time, msg)
Timeout.timeout(time) do
yield
end
rescue Timeout::Error
warn msg
exit 1
end
private
@@ -146,15 +147,13 @@ module FlexNBD
# )
def n64(b)
((b & 0xff00000000000000) >> 56) |
((b & 0x00ff000000000000) >> 40) |
((b & 0x0000ff0000000000) >> 24) |
((b & 0x000000ff00000000) >> 8) |
((b & 0x00000000ff000000) << 8) |
((b & 0x0000000000ff0000) << 24) |
((b & 0x000000000000ff00) << 40) |
((b & 0x00000000000000ff) << 56)
((b & 0x00ff000000000000) >> 40) |
((b & 0x0000ff0000000000) >> 24) |
((b & 0x000000ff00000000) >> 8) |
((b & 0x00000000ff000000) << 8) |
((b & 0x0000000000ff0000) << 24) |
((b & 0x000000000000ff00) << 40) |
((b & 0x00000000000000ff) << 56)
end
end # class FakeSource
end # module FlexNBD

View File

@@ -0,0 +1,55 @@
require 'tempfile'
#
# LdPreload is a little wrapper for using LD_PRELOAD loggers to pick up system
# calls when testing flexnbd.
#
module LdPreload
#
# This takes an object name, sets up a temporary log file, whose name is
# recorded in the environment as OUTPUT_obj_name, where obj_name is the
# name of the preload module to build and load.
def with_ld_preload(obj_name)
@ld_preload_logs ||= {}
flunk 'Can only load a preload module once!' if @ld_preload_logs[obj_name]
system("make -C ld_preloads/ #{obj_name}.o > /dev/null") ||
flunk("Failed to build object #{obj_name}")
orig_env = ENV['LD_PRELOAD']
ENV['LD_PRELOAD'] = [orig_env, File.expand_path("./ld_preloads/#{obj_name}.o")].compact.join(' ')
# Open the log, and stick it in a hash
@ld_preload_logs[obj_name] = Tempfile.new(obj_name)
ENV['OUTPUT_' + obj_name] = @ld_preload_logs[obj_name].path
yield
ensure
if @ld_preload_logs[obj_name]
@ld_preload_logs[obj_name].close
@ld_preload_logs.delete(obj_name)
end
ENV['LD_PRELOAD'] = orig_env
end
def read_ld_preload_log(obj_name)
lines = []
lines << @ld_preload_logs[obj_name].readline.chomp until
@ld_preload_logs[obj_name].eof?
lines
end
#
# The next to methods assume the log file has one entry per line, and that
# each entry is a series of values separated by colons.
#
def parse_ld_preload_logs(obj_name)
read_ld_preload_log(obj_name).map do |l|
l.split(':').map { |i| i =~ /^\d+$/ ? i.to_i : i }
end
end
def assert_func_call(loglines, args, msg)
re = Regexp.new('^' + args.join(':'))
assert(loglines.any? { |l| l.match(re) }, msg)
end
end

View File

@@ -0,0 +1,13 @@
SRC := $(wildcard *.c)
OBJS := $(SRC:%.c=%.o)
all: $(OBJS)
clean:
$(RM) $(OBJS)
%.o: %.c
gcc -shared -fPIC -ldl -o $@ $<
.PHONY: all clean

View File

@@ -0,0 +1,33 @@
#define _GNU_SOURCE
#include <dlfcn.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
typedef int (*real_msync_t)(void *addr, size_t length, int flags);
int real_msync(void *addr, size_t length, int flags) {
return ((real_msync_t)dlsym(RTLD_NEXT, "msync"))(addr, length, flags);
}
/*
* Noddy LD_PRELOAD wrapper to catch msync calls, and log them to a file.
*/
int msync(void *addr, size_t length, int flags) {
FILE *fd;
char *fn;
int retval;
retval = real_msync(addr, length, flags);
fn = getenv("OUTPUT_msync_logger");
if ( fn != NULL ) {
fd = fopen(fn,"a");
fprintf(fd,"msync:%d:%i:%i:%i\n", addr, length, flags, retval);
fclose(fd);
}
return retval;
}

View File

@@ -0,0 +1,38 @@
#define _GNU_SOURCE
#include <dlfcn.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/socket.h>
/*
* Noddy LD_PRELOAD wrapper to catch setsockopt calls, and log them to a file.
*/
typedef int (*real_setsockopt_t)(int sockfd, int level, int optname, const void *optval, socklen_t optlen);
int real_setsockopt(int sockfd, int level, int optname, const void *optval, socklen_t optlen)
{
return ((real_setsockopt_t)dlsym(RTLD_NEXT, "setsockopt"))(sockfd, level, optname, optval, optlen);
}
int setsockopt(int sockfd, int level, int optname, const void *optval, socklen_t optlen)
{
FILE *fd;
char *fn;
int retval;
retval = real_setsockopt(sockfd, level, optname, optval, optlen);
fn = getenv("OUTPUT_setsockopt_logger");
/*
* Only interested in catching non-null 4-byte (integer) values
*/
if ( fn != NULL && optval != NULL && optlen == 4) {
fd = fopen(fn,"a");
fprintf(fd,"setsockopt:%i:%i:%i:%i:%i\n", sockfd, level, optname, *(int *)optval, retval);
fclose(fd);
}
return retval;
}

2
tests/acceptance/nbd_scenarios Normal file → Executable file
View File

@@ -1,6 +1,6 @@
#!/usr/bin/ruby
test_files = Dir[File.dirname( __FILE__ ) + "/test*.rb"]
test_files = Dir[File.dirname(__FILE__) + '/test*.rb']
for filename in test_files
require filename
end

View File

@@ -1,22 +1,33 @@
# encoding: utf-8
require 'flexnbd/fake_source'
require 'flexnbd/fake_dest'
require 'ld_preload'
module ProxyTests
def with_proxied_client( override_size = nil )
include LdPreload
def b
"\xFF".b
end
def with_proxied_client(override_size = nil)
@env.serve1 unless @server_up
@env.proxy2 unless @proxy_up
@env.nbd2.can_die(0)
client = FlexNBD::FakeSource.new(@env.ip, @env.port2, "Couldn't connect to proxy")
begin
result = client.read_hello
assert_equal "NBDMAGIC", result[:magic]
assert_equal 'NBDMAGIC', result[:passwd]
assert_equal override_size || @env.file1.size, result[:size]
yield client
ensure
client.close rescue nil
begin
client.close
rescue StandardError
nil
end
end
end
@@ -28,11 +39,11 @@ module ProxyTests
with_proxied_client do |client|
(0..3).each do |n|
offset = n * 4096
client.write_read_request(offset, 4096, "myhandle")
client.write_read_request(offset, 4096, 'myhandle')
rsp = client.read_response
assert_equal ::FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal "myhandle", rsp[:handle]
assert_equal 'myhandle', rsp[:handle]
assert_equal 0, rsp[:error]
orig_data = @env.file1.read(offset, 4096)
@@ -41,8 +52,8 @@ module ProxyTests
assert_equal 4096, orig_data.size
assert_equal 4096, data.size
assert_equal( orig_data, data,
"Returned data does not match on request #{n+1}" )
assert_equal(orig_data, data,
"Returned data does not match on request #{n + 1}")
end
end
end
@@ -51,20 +62,34 @@ module ProxyTests
with_proxied_client do |client|
(0..3).each do |n|
offset = n * 4096
client.write(offset, "\xFF" * 4096)
client.write(offset, b * 4096)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal "myhandle", rsp[:handle]
assert_equal 'myhandle', rsp[:handle]
assert_equal 0, rsp[:error]
data = @env.file1.read(offset, 4096)
assert_equal( ( "\xFF" * 4096 ), data, "Data not written correctly (offset is #{n})" )
assert_equal((b * 4096), data, "Data not written correctly (offset is #{n})")
end
end
end
def test_write_request_past_end_of_disc_returns_to_client
with_proxied_client do |client|
n = 1000
offset = n * 4096
client.write(offset, b * 4096)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 'myhandle', rsp[:handle]
# NBD protocol say ENOSPC (28) in this situation
assert_equal 28, rsp[:error]
end
end
def make_fake_server
server = FlexNBD::FakeDest.new(@env.ip, @env.port1)
@server_up = true
@@ -74,7 +99,7 @@ module ProxyTests
sc = server.accept # just tell the supervisor we're up
sc.write_hello
[ server, sc ]
[server, sc]
end
end
@@ -85,7 +110,7 @@ module ProxyTests
server, sc1 = maker.value
# Send the read request to the proxy
client.write_read_request( 0, 4096 )
client.write_read_request(0, 4096)
# ensure we're given the read request
req1 = sc1.read_request
@@ -106,85 +131,110 @@ module ProxyTests
assert_equal req1, req2
# The reply should be proxied back to the client.
sc2.write_reply( req2[:handle] )
sc2.write_data( "\xFF" * 4096 )
sc2.write_reply(req2[:handle])
sc2.write_data(b * 4096)
# Check it to make sure it's correct
rsp = timeout(15) { client.read_response }
rsp = Timeout.timeout(15) { client.read_response }
assert_equal ::FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
assert_equal req1[:handle], rsp[:handle]
data = client.read_raw( 4096 )
assert_equal( ("\xFF" * 4096), data, "Wrong data returned" )
data = client.read_raw(4096)
assert_equal((b * 4096), data, 'Wrong data returned')
sc2.close
server.close
end
end
def test_write_request_retried_when_upstream_dies_partway
maker = make_fake_server
with_proxied_client(4096) do |client|
server, sc1 = maker.value
with_ld_preload('setsockopt_logger') do
with_proxied_client(4096) do |client|
server, sc1 = maker.value
# Send the read request to the proxy
client.write( 0, ( "\xFF" * 4096 ) )
# Send the read request to the proxy
client.write(0, (b * 4096))
# ensure we're given the read request
req1 = sc1.read_request
assert_equal ::FlexNBD::REQUEST_MAGIC, req1[:magic]
assert_equal ::FlexNBD::REQUEST_WRITE, req1[:type]
assert_equal 0, req1[:from]
assert_equal 4096, req1[:len]
data1 = sc1.read_data( 4096 )
assert_equal( ( "\xFF" * 4096 ), data1, "Data not proxied successfully" )
# ensure we're given the read request
req1 = sc1.read_request
assert_equal ::FlexNBD::REQUEST_MAGIC, req1[:magic]
assert_equal ::FlexNBD::REQUEST_WRITE, req1[:type]
assert_equal 0, req1[:from]
assert_equal 4096, req1[:len]
data1 = sc1.read_data(4096)
assert_equal((b * 4096), data1, 'Data not proxied successfully')
# Kill the server again, now we're sure the read request has been sent once
sc1.close
# Read the setsockopt logs, so we can check that TCP_NODELAY is re-set
# later
read_ld_preload_log('setsockopt_logger')
# We expect the proxy to reconnect without our client doing anything.
sc2 = server.accept
sc2.write_hello
# Kill the server again, now we're sure the read request has been sent once
sc1.close
# And once reconnected, it should resend an identical request.
req2 = sc2.read_request
assert_equal req1, req2
data2 = sc2.read_data( 4096 )
assert_equal data1, data2
# We expect the proxy to reconnect without our client doing anything.
sc2 = server.accept
sc2.write_hello
# The reply should be proxied back to the client.
sc2.write_reply( req2[:handle] )
# And once reconnected, it should resend an identical request.
req2 = sc2.read_request
assert_equal req1, req2
data2 = sc2.read_data(4096)
assert_equal data1, data2
# Check it to make sure it's correct
rsp = timeout(15) { client.read_response }
assert_equal ::FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
assert_equal req1[:handle], rsp[:handle]
# The reply should be proxied back to the client.
sc2.write_reply(req2[:handle])
sc2.close
server.close
# Check it to make sure it's correct
rsp = Timeout.timeout(15) { client.read_response }
assert_equal ::FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
assert_equal req1[:handle], rsp[:handle]
sc2.close
server.close
# Check TCP_NODELAY was set on the upstream socket
log = read_ld_preload_log('setsockopt_logger')
assert_func_call(log,
['setsockopt', 3,
Socket::SOL_TCP, Socket::TCP_NODELAY, 1, 0],
'TCP_NODELAY not set on upstream fd 3')
end
end
end
def test_only_one_client_can_connect_to_proxy_at_a_time
with_proxied_client do |client|
with_proxied_client do |_client|
c2 = nil
assert_raises(Timeout::Error) do
timeout(1) do
Timeout.timeout(1) do
c2 = FlexNBD::FakeSource.new(@env.ip, @env.port2, "Couldn't connect to proxy (2)")
c2.read_hello
end
end
c2.close rescue nil if c2
if c2
begin
c2.close
rescue StandardError
nil
end
end
end
end
def test_maximum_write_request_size
# Defined in src/common/nbdtypes.h
nbd_max_block_size = 32 * 1024 * 1024
@env.writefile1('0' * 40 * 1024)
with_proxied_client do |client|
# This will crash with EPIPE if the proxy dies.
client.write(0, b * nbd_max_block_size)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
end
end
end

View File

@@ -1,13 +1,10 @@
# encoding: utf-8
require 'test/unit'
require 'environment'
class TestDestErrorHandling < Test::Unit::TestCase
class TestDestErrorHandling < Test::Unit::TestCase
def setup
@env = Environment.new
@env.writefile1( "0" * 4 )
@env.writefile1('0' * 4)
@env.listen1
end
@@ -15,89 +12,77 @@ class TestDestErrorHandling < Test::Unit::TestCase
@env.cleanup
end
def test_hello_blocked_by_disconnect_causes_error_not_fatal
run_fake( "source/close_after_connect" )
run_fake('source/close_after_connect')
assert_no_control
end
=begin
# This is disabled while CLIENT_MAX_WAIT_SECS is removed
def test_hello_goes_astray_causes_timeout_error
run_fake( "source/hang_after_hello" )
assert_no_control
end
=end
# # This is disabled while CLIENT_MAX_WAIT_SECS is removed
# def test_hello_goes_astray_causes_timeout_error
# run_fake( "source/hang_after_hello" )
# assert_no_control
# end
def test_sigterm_has_bad_exit_status
@env.nbd1.can_die(1)
run_fake( "source/sigterm_after_hello" )
run_fake('source/sigterm_after_hello')
end
def test_disconnect_after_hello_causes_error_not_fatal
run_fake( "source/close_after_hello" )
run_fake('source/close_after_hello')
assert_no_control
end
def test_partial_read_causes_error
run_fake( "source/close_mid_read" )
run_fake('source/close_mid_read')
end
def test_double_connect_during_hello
run_fake( "source/connect_during_hello" )
run_fake('source/connect_during_hello')
end
def test_acl_rejection
@env.acl1("127.0.0.1")
run_fake( "source/connect_from_banned_ip")
@env.acl1('127.0.0.1')
run_fake('source/connect_from_banned_ip')
end
def test_bad_write
run_fake( "source/write_out_of_range" )
run_fake('source/write_out_of_range')
end
def test_disconnect_before_write_data_causes_error
run_fake( "source/close_after_write" )
run_fake('source/close_after_write')
end
def test_disconnect_before_write_reply_causes_error
# Note that this is an odd case: writing the reply doesn't fail.
# The test passes because the next attempt by flexnbd to read a
# request returns EOF.
run_fake( "source/close_after_write_data" )
run_fake('source/close_after_write_data')
end
def test_straight_migration
@env.nbd1.can_die(0)
run_fake( "source/successful_transfer" )
run_fake('source/successful_transfer')
end
private
def run_fake( name )
@env.run_fake( name, @env.ip, @env.port1 )
def run_fake(name)
@env.run_fake(name, @env.ip, @env.port1)
assert @env.fake_reports_success, "#{name} failed."
end
def status
stat, _ = @env.status1
stat, = @env.status1
stat
end
def assert_no_control
assert !status['has_control'], "Thought it had control"
assert !status['has_control'], 'Thought it had control'
end
def assert_control
assert status['has_control'], "Didn't think it had control"
end
end # class TestDestErrorHandling

View File

@@ -1,5 +1,3 @@
# encoding: utf-8
require 'test/unit'
require 'environment'
require 'flexnbd/constants'
@@ -19,20 +17,18 @@ class TestHappyPath < Test::Unit::TestCase
@env.cleanup
end
def test_read1
@env.writefile1("f"*64)
@env.writefile1('f' * 64)
@env.serve1
[0, 12, 63].each do |num|
assert_equal(
bin( @env.nbd1.read(num*@env.blocksize, @env.blocksize) ),
bin( @env.file1.read(num*@env.blocksize, @env.blocksize) )
bin(@env.nbd1.read(num * @env.blocksize, @env.blocksize)),
bin(@env.file1.read(num * @env.blocksize, @env.blocksize))
)
end
[124, 1200, 10028, 25488].each do |num|
[124, 1200, 10_028, 25_488].each do |num|
assert_equal(bin(@env.nbd1.read(num, 4)), bin(@env.file1.read(num, 4)))
end
end
@@ -40,14 +36,14 @@ class TestHappyPath < Test::Unit::TestCase
# Check that we're not
#
def test_writeread1
@env.writefile1("0"*64)
@env.writefile1('0' * 64)
@env.serve1
[0, 12, 63].each do |num|
data = "X"*@env.blocksize
@env.nbd1.write(num*@env.blocksize, data)
assert_equal(data, @env.file1.read(num*@env.blocksize, data.size))
assert_equal(data, @env.nbd1.read(num*@env.blocksize, data.size))
data = 'X' * @env.blocksize
@env.nbd1.write(num * @env.blocksize, data)
assert_equal(data, @env.file1.read(num * @env.blocksize, data.size))
assert_equal(data, @env.nbd1.read(num * @env.blocksize, data.size))
end
end
@@ -55,110 +51,105 @@ class TestHappyPath < Test::Unit::TestCase
# up.
#
def test_writeread2
@env.writefile1("0"*1024)
@env.writefile1('0' * 1024)
@env.serve1
d0 = "\0"*@env.blocksize
d1 = "X"*@env.blocksize
d0 = "\0" * @env.blocksize
d1 = 'X' * @env.blocksize
(0..63).each do |num|
@env.nbd1.write(num*@env.blocksize*2, d1)
@env.nbd1.write(num * @env.blocksize * 2, d1)
end
(0..63).each do |num|
assert_equal(d0, @env.nbd1.read(((2*num)+1)*@env.blocksize, d0.size))
assert_equal(d0, @env.nbd1.read(((2 * num) + 1) * @env.blocksize, d0.size))
end
end
def setup_to_mirror
@env.writefile1( "f"*4 )
@env.writefile1('f' * 4)
@env.serve1
@env.writefile2( "0"*4 )
@env.writefile2('0' * 4)
@env.listen2
end
def test_mirror
@env.nbd1.can_die
@env.nbd2.can_die(0)
setup_to_mirror()
setup_to_mirror
stdout, stderr = @env.mirror12
@env.nbd1.join
@env.nbd2.join
assert( File.file?( @env.filename1 ),
"The source file was incorrectly deleted")
assert_equal(@env.file1.read_original( 0, @env.blocksize ),
@env.file2.read( 0, @env.blocksize ) )
assert(File.file?(@env.filename1),
'The source file was incorrectly deleted')
assert_equal(@env.file1.read_original(0, @env.blocksize),
@env.file2.read(0, @env.blocksize))
end
def test_mirror_unlink
@env.nbd1.can_die(0)
@env.nbd2.can_die(0)
setup_to_mirror()
setup_to_mirror
assert File.file?( @env.filename1 )
assert File.file?(@env.filename1)
stdout, stderr = @env.mirror12_unlink
assert_no_match( /unrecognized/, stderr )
assert_no_match(/unrecognized/, stderr)
Timeout.timeout(10) { @env.nbd1.join }
Timeout.timeout(10) do @env.nbd1.join end
assert !File.file?( @env.filename1 )
assert !File.file?(@env.filename1)
end
def test_write_to_high_block
#
# This test does not work on 32 bit platforms.
#
skip('Not relevant on 32-bit platforms') if ['a'].pack('p').size < 8
# Create a large file, then try to write to somewhere after the 2G boundary
@env.truncate1 "4G"
@env.truncate1 '4G'
@env.serve1
@env.nbd1.write( 2**31+2**29, "12345678" )
@env.nbd1.write(2**31 + 2**29, '12345678')
sleep(1)
assert_equal "12345678", @env.nbd1.read( 2**31+2**29, 8 )
assert_equal '12345678', @env.nbd1.read(2**31 + 2**29, 8)
end
def test_set_acl
# Just check that we get sane feedback here
@env.writefile1( "f"*4 )
@env.writefile1('f' * 4)
@env.serve1
_,stderr = @env.acl1("127.0.0.1")
assert_no_match( /^(F|E):/, stderr )
_, stderr = @env.acl1('127.0.0.1')
assert_no_match(/^(F|E):/, stderr)
end
def test_write_more_than_one_run
one_mb = 2**20
data = "\0" * 256 * one_mb
File.open(@env.filename1, "wb") do |f| f.write( "1" * 256 * one_mb ) end
File.open(@env.filename1, 'wb') { |f| f.write('1' * 256 * one_mb) }
@env.serve1
sleep 5
@env.write1( data )
@env.write1(data)
@env.nbd1.can_die(0)
@env.nbd1.kill
i = 0
File.open(@env.filename1, "rb") do |f|
while mb = f.read( one_mb )
unless "\0"*one_mb == mb
msg = "Read non-zeros after offset %x:\n"%(i * one_mb)
File.open(@env.filename1, 'rb') do |f|
while mb = f.read(one_mb)
unless "\0" * one_mb == mb
msg = format("Read non-zeros after offset %x:\n", (i * one_mb))
msg += `hexdump #{@env.filename1} | head -n5`
fail msg
raise msg
end
i += 1
end
end
end
end

View File

@@ -2,7 +2,6 @@ require 'test/unit'
require 'environment'
require 'proxy_tests'
class TestPrefetchProxyMode < Test::Unit::TestCase
include ProxyTests
@@ -10,7 +9,7 @@ class TestPrefetchProxyMode < Test::Unit::TestCase
super
@env = Environment.new
@env.prefetch_proxy!
@env.writefile1( "f" * 16 )
@env.writefile1('f' * 16)
end
def teardown
@@ -18,5 +17,3 @@ class TestPrefetchProxyMode < Test::Unit::TestCase
super
end
end

View File

@@ -2,14 +2,13 @@ require 'test/unit'
require 'environment'
require 'proxy_tests'
class TestProxyMode < Test::Unit::TestCase
include ProxyTests
def setup
super
@env = Environment.new
@env.writefile1( "f" * 16 )
@env.writefile1('f' * 16)
end
def teardown
@@ -17,4 +16,3 @@ class TestProxyMode < Test::Unit::TestCase
super
end
end

View File

@@ -1,14 +1,15 @@
require 'test/unit'
require 'environment'
require 'flexnbd/fake_source'
require 'ld_preload'
class TestServeMode < Test::Unit::TestCase
include LdPreload
def setup
super
@b = "\xFF".b
@env = Environment.new
@env.writefile1( "0" )
@env.serve1
end
def teardown
@@ -17,69 +18,78 @@ class TestServeMode < Test::Unit::TestCase
end
def connect_to_server
client = FlexNBD::FakeSource.new(@env.ip, @env.port1, "Connecting to server failed")
@env.writefile1('0')
@env.serve1
client = FlexNBD::FakeSource.new(@env.ip, @env.port1, 'Connecting to server failed')
begin
result = client.read_hello
assert_equal "NBDMAGIC", result[:magic]
assert_equal 'NBDMAGIC', result[:passwd]
assert_equal 0x00420281861253, result[:magic]
assert_equal @env.file1.size, result[:size]
# See src/common/nbdtypes.h for the various flags. At the moment we
# support HAS_FLAGS (1), SEND_FLUSH (4), SEND_FUA (8)
assert_equal (1 | 4 | 8), result[:flags]
assert_equal "\x0" * 124, result[:reserved]
yield client
ensure
client.close rescue nil
begin
client.close
rescue StandardError
nil
end
end
end
def test_bad_request_magic_receives_error_response
connect_to_server do |client|
# replace REQUEST_MAGIC with all 0s to make it look bad
client.send_request( 0, "myhandle", 0, 0, "\x00\x00\x00\x00" )
client.send_request(0, 'myhandle', 0, 0, "\x00\x00\x00\x00")
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal "myhandle", rsp[:handle]
assert_equal 'myhandle', rsp[:handle]
assert rsp[:error] != 0, "Server sent success reply back: #{rsp[:error]}"
# The client should be disconnected now
assert client.disconnected?, "Server not disconnected"
assert client.disconnected?, 'Server not disconnected'
end
end
def test_long_write_on_top_of_short_write_is_respected
connect_to_server do |client|
# Start with a file of all-zeroes.
client.write( 0, "\x00" * @env.file1.size )
client.write(0, "\x00" * @env.file1.size)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
client.write( 0, "\xFF" )
client.write(0, @b)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
client.write( 0, "\xFF\xFF" )
client.write(0, @b * 2)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
end
assert_equal "\xFF\xFF", @env.file1.read( 0, 2 )
assert_equal @b * 2, @env.file1.read(0, 2)
end
def test_read_request_out_of_bounds_receives_error_response
connect_to_server do |client|
client.write_read_request( @env.file1.size, 4096 )
client.write_read_request(@env.file1.size, 4096)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal "myhandle", rsp[:handle]
assert rsp[:error] != 0, "Server sent success reply back: #{rsp[:error]}"
assert_equal 'myhandle', rsp[:handle]
# NBD protocol suggests ENOSPC (28) is returned
assert_equal 28, rsp[:error], 'Server sent incorrect response'
# Ensure we're not disconnected by sending a request. We don't care about
# whether the reply is good or not, here.
client.write_read_request( 0, 4096 )
client.write_read_request(0, 4096)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
end
@@ -87,23 +97,127 @@ class TestServeMode < Test::Unit::TestCase
def test_write_request_out_of_bounds_receives_error_response
connect_to_server do |client|
client.write( @env.file1.size, "\x00" * 4096 )
client.write(@env.file1.size, "\x00" * 4096)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal "myhandle", rsp[:handle]
assert rsp[:error] != 0, "Server sent success reply back: #{rsp[:error]}"
assert_equal 'myhandle', rsp[:handle]
# NBD protocol suggests ENOSPC (28) is returned
assert_equal 28, rsp[:error], 'Server sent incorrect response'
# Ensure we're not disconnected by sending a request. We don't care about
# whether the reply is good or not, here.
client.write( 0, "\x00" * @env.file1.size )
client.write(0, "\x00" * @env.file1.size)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
end
end
def test_unknown_command_receives_error_response
connect_to_server do |client|
client.send_request(123)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 'myhandle', rsp[:handle]
# NBD protocol suggests EINVAL (22) is returned
assert_equal 22, rsp[:error], 'Server sent incorrect response'
# Ensure we're not disconnected by sending a request. We don't care about
# whether the reply is good or not, here.
client.write(0, "\x00" * @env.file1.size)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
end
end
def test_flush_is_accepted
with_ld_preload('msync_logger') do
connect_to_server do |client|
client.flush
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
end
op = parse_ld_preload_logs('msync_logger')
assert_equal 1, op.count, 'Only one msync expected'
assert_equal @env.blocksize, op.first[2], 'msync length wrong'
assert_equal 6, op.first[3], 'msync called with incorrect flags'
end
end
def test_write_with_fua_is_accepted
with_ld_preload('msync_logger') do
page_size = Integer(`getconf PAGESIZE`)
@env.blocksize = page_size * 10
connect_to_server do |client|
# Write somewhere in the third page
pos = page_size * 3 + 100
client.write_with_fua(pos, "\x00" * 33)
rsp = client.read_response
assert_equal FlexNBD::REPLY_MAGIC, rsp[:magic]
assert_equal 0, rsp[:error]
end
op = parse_ld_preload_logs('msync_logger')
assert_equal 1, op.count, 'Only one msync expected'
# Should be 100 + 33, as we've started writing 100 bytes into a page, for
# 33 bytes
assert_equal 133, op.first[2], 'msync length wrong'
assert_equal 6, op.first[3], 'msync called with incorrect flags'
end
end
def test_odd_size_discs_are_truncated_to_nearest_512
# This should get rounded down to 1024
@env.blocksize = 1024 + 511
@env.writefile1('0')
@env.serve1
client = FlexNBD::FakeSource.new(@env.ip, @env.port1, 'Connecting to server failed')
begin
result = client.read_hello
assert_equal 'NBDMAGIC', result[:passwd]
assert_equal 0x00420281861253, result[:magic]
assert_equal 1024, result[:size]
client.close
end
end
def test_server_sets_tcpkeepalive
with_ld_preload('setsockopt_logger') do
connect_to_server(&:close)
op = read_ld_preload_log('setsockopt_logger')
assert_func_call(op,
['setsockopt', '\d+',
Socket::SOL_SOCKET, Socket::SO_KEEPALIVE, 1, 0],
'TCP Keepalive not successfully set')
assert_func_call(op,
['setsockopt', '\d+',
Socket::SOL_TCP, Socket::TCP_KEEPIDLE, 30, 0],
'TCP Keepalive idle timeout not set to 30s')
assert_func_call(op,
['setsockopt', '\d+',
Socket::SOL_TCP, Socket::TCP_KEEPINTVL, 10, 0],
'TCP keepalive retry time not set to 10s')
assert_func_call(op,
['setsockopt', '\d+',
Socket::SOL_TCP, Socket::TCP_KEEPCNT, 3, 0],
'TCP keepalive count not set to 3')
end
end
def test_status_returns_correct_client_count
@env.writefile1('0')
@env.serve1
assert_equal('0', @env.status1['num_clients'])
client = FlexNBD::FakeSource.new(@env.ip, @env.port1, 'Connecting to server failed')
assert_equal('1', @env.status1['num_clients'])
client2 = FlexNBD::FakeSource.new(@env.ip, @env.port1, 'Connecting to server failed')
assert_equal('2', @env.status1['num_clients'])
client2.close
client.close
assert_equal('0', @env.status1['num_clients'])
end
end

View File

@@ -1,126 +1,105 @@
# encoding: utf-8
require 'test/unit'
require 'environment'
class TestSourceErrorHandling < Test::Unit::TestCase
def setup
@old_env = ENV['FLEXNBD_MS_REQUEST_LIMIT_SECS']
ENV['FLEXNBD_MS_REQUEST_LIMIT_SECS'] = "4.0"
ENV['FLEXNBD_MS_REQUEST_LIMIT_SECS'] = '4.0'
@env = Environment.new
@env.writefile1( "f" * 4 )
@env.writefile1('f' * 4)
@env.serve1
end
def teardown
@env.nbd1.can_die(0)
@env.cleanup
ENV['FLEXNBD_MS_REQUEST_LIMIT_SECS'] = @old_env
end
def expect_term_during_migration
@env.nbd1.can_die(1,9)
@env.nbd1.can_die(1, 9)
end
def test_failure_to_connect_reported_in_mirror_cmd_response
stdout, stderr = @env.mirror12_unchecked
expect_term_during_migration
assert_match( /failed to connect/, stderr )
assert_match(/failed to connect/, stderr)
end
def test_sigterm_after_hello_quits_with_status_of_1
expect_term_during_migration
run_fake( "dest/sigterm_after_hello" )
run_fake('dest/sigterm_after_hello')
end
def test_destination_hangs_after_connect_reports_error_at_source
run_fake( "dest/hang_after_connect",
:err => /Remote server failed to respond/ )
run_fake('dest/hang_after_connect',
err: /Remote server failed to respond/)
end
def test_destination_rejects_connection_reports_error_at_source
run_fake( "dest/reject_acl",
:err => /Mirror was rejected/ )
run_fake('dest/reject_acl',
err: /Mirror was rejected/)
end
def test_wrong_size_causes_disconnect
run_fake( "dest/hello_wrong_size",
:err => /Remote size does not match local size/ )
run_fake('dest/hello_wrong_size',
err: /Remote size does not match local size/)
end
def test_wrong_magic_causes_disconnect
expect_term_during_migration
run_fake( "dest/hello_wrong_magic",
:err => /Mirror was rejected/ )
run_fake('dest/hello_wrong_magic',
err: /Mirror was rejected/)
end
def test_disconnect_after_hello_causes_retry
expect_term_during_migration
run_fake( "dest/close_after_hello",
:out => /Mirror started/ )
run_fake('dest/close_after_hello',
out: /Mirror started/)
end
def test_write_times_out_causes_retry
expect_term_during_migration
run_fake( "dest/hang_after_write" )
run_fake('dest/hang_after_write')
end
def test_rejected_write_causes_retry
expect_term_during_migration
run_fake( "dest/error_on_write" )
run_fake('dest/error_on_write')
end
def test_disconnect_before_write_reply_causes_retry
expect_term_during_migration
run_fake( "dest/close_after_write" )
run_fake('dest/close_after_write')
end
def test_bad_write_reply_causes_retry
expect_term_during_migration
run_fake( "dest/write_wrong_magic" )
run_fake('dest/write_wrong_magic')
end
def test_pre_entrust_disconnect_causes_retry
expect_term_during_migration
run_fake( "dest/close_after_writes" )
run_fake('dest/close_after_writes')
end
def test_cancel_migration
run_fake( "dest/break_after_hello" )
run_fake('dest/break_after_hello')
end
private
def run_fake(name, opts = {})
@env.run_fake( name, @env.ip, @env.port2, @env.nbd1.ctrl )
@env.run_fake(name, @env.ip, @env.port2, @env.nbd1.ctrl)
stdout, stderr = @env.mirror12_unchecked
assert_success
assert_match( opts[:err], stderr ) if opts[:err]
assert_match( opts[:out], stdout ) if opts[:out]
return stdout, stderr
assert_match(opts[:err], stderr) if opts[:err]
assert_match(opts[:out], stdout) if opts[:out]
[stdout, stderr]
end
def assert_success( msg=nil )
assert @env.fake_reports_success, msg || "Fake failed"
def assert_success(msg = nil)
assert @env.fake_reports_success, msg || 'Fake failed'
end
end # class TestSourceErrorHandling

121
tests/acceptance/test_write_during_migration.rb Executable file → Normal file
View File

@@ -9,102 +9,98 @@ require 'tmpdir'
Thread.abort_on_exception = true
class TestWriteDuringMigration < Test::Unit::TestCase
def setup
@flexnbd = File.expand_path("../../build/flexnbd")
@flexnbd = File.expand_path('../../build/flexnbd')
raise "No binary!" unless File.executable?( @flexnbd )
raise 'No binary!' unless File.executable?(@flexnbd)
@size = 20*1024*1024 # 20MB
@write_data = "foo!" * 2048 # 8K write
@size = 20 * 1024 * 1024 # 20MB
@write_data = 'foo!' * 2048 # 8K write
@source_port = 9990
@dest_port = 9991
@source_sock = "src.sock"
@dest_sock = "dst.sock"
@source_file = "src.file"
@dest_file = "dst.file"
@source_sock = 'src.sock'
@dest_sock = 'dst.sock'
@source_file = 'src.file'
@dest_file = 'dst.file'
end
def teardown
[@dst_proc, @src_proc].each do |pid|
if pid
Process.kill( "KILL", pid ) rescue nil
next unless pid
begin
Process.kill('KILL', pid)
rescue StandardError
nil
end
end
end
def debug_arg
ENV['DEBUG'] ? "--verbose" : ""
ENV['DEBUG'] ? '--verbose' : ''
end
def launch_servers
@dst_proc = fork() {
@dst_proc = fork do
cmd = "#{@flexnbd} listen -l 127.0.0.1 -p #{@dest_port} -f #{@dest_file} -s #{@dest_sock} #{debug_arg}"
exec cmd
}
end
@src_proc = fork() {
@src_proc = fork do
cmd = "#{@flexnbd} serve -l 127.0.0.1 -p #{@source_port} -f #{@source_file} -s #{@source_sock} #{debug_arg}"
exec cmd
}
end
begin
awaiting = nil
Timeout.timeout(10) do
awaiting = :source
sleep 0.1 while !File.exists?( @source_sock )
sleep 0.1 until File.exist?(@source_sock)
awaiting = :dest
sleep 0.1 while !File.exists?( @dest_sock )
sleep 0.1 until File.exist?(@dest_sock)
end
rescue Timeout::Error
case awaiting
when :source
fail "Couldn't get a source socket."
raise "Couldn't get a source socket."
when :dest
fail "Couldn't get a destination socket."
raise "Couldn't get a destination socket."
else
fail "Something went wrong I don't understand."
raise "Something went wrong I don't understand."
end
end
end
def make_files()
def make_files
FileUtils.touch(@source_file)
File.truncate(@source_file, @size)
FileUtils.touch(@dest_file)
File.truncate(@dest_file, @size)
File.open(@source_file, "wb"){|f| f.write "a"*@size }
File.open(@source_file, 'wb') { |f| f.write 'a' * @size }
end
def start_mirror
UNIXSocket.open(@source_sock) {|sock|
sock.write(["mirror", "127.0.0.1", @dest_port.to_s, "exit"].join("\x0A") + "\x0A\x0A")
UNIXSocket.open(@source_sock) do |sock|
sock.write(['mirror', '127.0.0.1', @dest_port.to_s, 'exit'].join("\x0A") + "\x0A\x0A")
sock.flush
rsp = sock.readline
}
end
end
def wait_for_quit()
Timeout.timeout( 10 ) do
start_time = Time.now
dst_result = Process::waitpid2(@dst_proc)
src_result = Process::waitpid2(@src_proc)
def wait_for_quit
Timeout.timeout(10) do
start_time = Time.now
dst_result = Process.waitpid2(@dst_proc)
src_result = Process.waitpid2(@src_proc)
end
end
def source_writer
client = FlexNBD::FakeSource.new( "127.0.0.1", @source_port, "Timed out connecting" )
offsets = Range.new(0, (@size - @write_data.size) / 4096 ).to_a
client = FlexNBD::FakeSource.new('127.0.0.1', @source_port, 'Timed out connecting')
offsets = Range.new(0, (@size - @write_data.size) / 4096).to_a
loop do
begin
client.write(offsets[rand(offsets.size)] * 4096, @write_data)
rescue => err
rescue StandardError => err
# We expect a broken write at some point, so ignore it
break
end
@@ -115,32 +111,32 @@ class TestWriteDuringMigration < Test::Unit::TestCase
# puts `md5sum #{@source_file} #{@dest_file}`
# Ensure each block matches
File.open(@source_file, "r") do |source|
File.open(@dest_file, "r") do |dest|
0.upto( @size / 4096 ) do |block_num|
s_data = source.read( 4096 )
d_data = dest.read( 4096 )
File.open(@source_file, 'r') do |source|
File.open(@dest_file, 'r') do |dest|
0.upto(@size / 4096) do |block_num|
s_data = source.read(4096)
d_data = dest.read(4096)
assert s_data == d_data, "Block #{block_num} mismatch!"
source.seek( 4096, IO::SEEK_CUR )
dest.seek( 4096, IO::SEEK_CUR )
source.seek(4096, IO::SEEK_CUR)
dest.seek(4096, IO::SEEK_CUR)
end
end
end
end
def test_write_during_migration
Dir.mktmpdir() do |tmpdir|
Dir.chdir( tmpdir ) do
make_files()
Dir.mktmpdir do |tmpdir|
Dir.chdir(tmpdir) do
make_files
launch_servers()
launch_servers
src_writer = Thread.new { source_writer }
start_mirror()
wait_for_quit()
start_mirror
wait_for_quit
src_writer.join
assert_both_sides_identical
end
@@ -148,24 +144,21 @@ class TestWriteDuringMigration < Test::Unit::TestCase
end
def test_many_clients_during_migration
Dir.mktmpdir() do |tmpdir|
Dir.chdir( tmpdir ) do
make_files()
Dir.mktmpdir do |tmpdir|
Dir.chdir(tmpdir) do
make_files
launch_servers()
launch_servers
src_writers_1 = (1..5).collect { Thread.new { source_writer } }
start_mirror()
start_mirror
src_writers_2 = (1..5).collect { Thread.new { source_writer } }
wait_for_quit()
( src_writers_1 + src_writers_2 ).each {|t| t.join }
wait_for_quit
(src_writers_1 + src_writers_2).each(&:join)
assert_both_sides_identical
end
end end
end end
end

View File

@@ -4,226 +4,229 @@
#include "acl.h"
#include "util.h"
START_TEST( test_null_acl )
START_TEST(test_null_acl)
{
struct acl *acl = acl_create( 0,NULL, 0 );
struct acl *acl = acl_create(0, NULL, 0);
fail_if( NULL == acl, "No acl alloced." );
fail_unless( 0 == acl->len, "Incorrect length" );
fail_if(NULL == acl, "No acl alloced.");
fail_unless(0 == acl->len, "Incorrect length");
}
END_TEST
START_TEST( test_parses_single_line )
START_TEST(test_parses_single_line)
{
char *lines[] = {"127.0.0.1"};
struct acl * acl = acl_create( 1, lines, 0 );
char *lines[] = { "127.0.0.1" };
struct acl *acl = acl_create(1, lines, 0);
fail_unless( 1 == acl->len, "Incorrect length." );
fail_if( NULL == acl->entries, "No entries present." );
fail_unless(1 == acl->len, "Incorrect length.");
fail_if(NULL == acl->entries, "No entries present.");
}
END_TEST
START_TEST( test_parses_multiple_lines )
START_TEST(test_parses_multiple_lines)
{
char *lines[] = {"127.0.0.1", "::1"};
struct acl * acl = acl_create( 2, lines, 0 );
union mysockaddr e0, e1;
char *lines[] = { "127.0.0.1", "::1" };
struct acl *acl = acl_create(2, lines, 0);
union mysockaddr e0, e1;
parse_ip_to_sockaddr( &e0.generic, lines[0] );
parse_ip_to_sockaddr( &e1.generic, lines[1] );
parse_ip_to_sockaddr(&e0.generic, lines[0]);
parse_ip_to_sockaddr(&e1.generic, lines[1]);
fail_unless( acl->len == 2, "Multiple lines not parsed" );
fail_unless(acl->len == 2, "Multiple lines not parsed");
struct ip_and_mask *entry;
entry = &(*acl->entries)[0];
fail_unless(entry->ip.family == e0.family, "entry 0 has wrong family!");
entry = &(*acl->entries)[1];
fail_unless(entry->ip.family == e1.family, "entry 1 has wrong family!");
struct ip_and_mask *entry;
entry = &(*acl->entries)[0];
fail_unless(entry->ip.family == e0.family,
"entry 0 has wrong family!");
entry = &(*acl->entries)[1];
fail_unless(entry->ip.family == e1.family,
"entry 1 has wrong family!");
}
END_TEST
START_TEST( test_destroy_doesnt_crash )
START_TEST(test_destroy_doesnt_crash)
{
char *lines[] = {"127.0.0.1"};
struct acl * acl = acl_create( 1, lines, 0 );
char *lines[] = { "127.0.0.1" };
struct acl *acl = acl_create(1, lines, 0);
acl_destroy( acl );
acl_destroy(acl);
}
END_TEST
START_TEST( test_includes_single_address )
START_TEST(test_includes_single_address)
{
char *lines[] = {"127.0.0.1"};
struct acl * acl = acl_create( 1, lines, 0 );
union mysockaddr x;
char *lines[] = { "127.0.0.1" };
struct acl *acl = acl_create(1, lines, 0);
union mysockaddr x;
parse_ip_to_sockaddr( &x.generic, "127.0.0.1" );
parse_ip_to_sockaddr(&x.generic, "127.0.0.1");
fail_unless( acl_includes( acl, &x ), "Included address wasn't covered" );
fail_unless(acl_includes(acl, &x), "Included address wasn't covered");
}
END_TEST
START_TEST( test_includes_single_address_when_netmask_specified_ipv4 )
START_TEST(test_includes_single_address_when_netmask_specified_ipv4)
{
char *lines[] = {"127.0.0.1/24"};
struct acl * acl = acl_create( 1, lines, 0 );
union mysockaddr x;
char *lines[] = { "127.0.0.1/24" };
struct acl *acl = acl_create(1, lines, 0);
union mysockaddr x;
parse_ip_to_sockaddr( &x.generic, "127.0.0.0" );
fail_unless( acl_includes( acl, &x ), "Included address wasn't covered" );
parse_ip_to_sockaddr(&x.generic, "127.0.0.0");
fail_unless(acl_includes(acl, &x), "Included address wasn't covered");
parse_ip_to_sockaddr( &x.generic, "127.0.0.1" );
fail_unless( acl_includes( acl, &x ), "Included address wasn't covered" );
parse_ip_to_sockaddr(&x.generic, "127.0.0.1");
fail_unless(acl_includes(acl, &x), "Included address wasn't covered");
parse_ip_to_sockaddr( &x.generic, "127.0.0.255" );
fail_unless( acl_includes( acl, &x ), "Included address wasn't covered" );
parse_ip_to_sockaddr(&x.generic, "127.0.0.255");
fail_unless(acl_includes(acl, &x), "Included address wasn't covered");
}
END_TEST
START_TEST( test_includes_single_address_when_netmask_specified_ipv6 )
START_TEST(test_includes_single_address_when_netmask_specified_ipv6)
{
char *lines[] = {"fe80::/10"};
struct acl * acl = acl_create( 1, lines, 0 );
union mysockaddr x;
char *lines[] = { "fe80::/10" };
struct acl *acl = acl_create(1, lines, 0);
union mysockaddr x;
parse_ip_to_sockaddr( &x.generic, "fe80::1" );
fail_unless( acl_includes( acl, &x ), "Included address wasn't covered" );
parse_ip_to_sockaddr(&x.generic, "fe80::1");
fail_unless(acl_includes(acl, &x), "Included address wasn't covered");
parse_ip_to_sockaddr( &x.generic, "fe80::2" );
fail_unless( acl_includes( acl, &x ), "Included address wasn't covered" );
parse_ip_to_sockaddr(&x.generic, "fe80::2");
fail_unless(acl_includes(acl, &x), "Included address wasn't covered");
parse_ip_to_sockaddr( &x.generic, "fe80:ffff:ffff::ffff" );
fail_unless( acl_includes( acl, &x ), "Included address wasn't covered" );
parse_ip_to_sockaddr(&x.generic, "fe80:ffff:ffff::ffff");
fail_unless(acl_includes(acl, &x), "Included address wasn't covered");
}
END_TEST
START_TEST( test_includes_single_address_when_multiple_entries_exist )
START_TEST(test_includes_single_address_when_multiple_entries_exist)
{
char *lines[] = {"127.0.0.1", "::1"};
struct acl * acl = acl_create( 2, lines, 0 );
union mysockaddr e0;
union mysockaddr e1;
char *lines[] = { "127.0.0.1", "::1" };
struct acl *acl = acl_create(2, lines, 0);
union mysockaddr e0;
union mysockaddr e1;
parse_ip_to_sockaddr( &e0.generic, "127.0.0.1" );
parse_ip_to_sockaddr( &e1.generic, "::1" );
parse_ip_to_sockaddr(&e0.generic, "127.0.0.1");
parse_ip_to_sockaddr(&e1.generic, "::1");
fail_unless( acl_includes( acl, &e0 ), "Included address 0 wasn't covered" );
fail_unless( acl_includes( acl, &e1 ), "Included address 1 wasn't covered" );
fail_unless(acl_includes(acl, &e0),
"Included address 0 wasn't covered");
fail_unless(acl_includes(acl, &e1),
"Included address 1 wasn't covered");
}
END_TEST
START_TEST( test_doesnt_include_other_address )
START_TEST(test_doesnt_include_other_address)
{
char *lines[] = {"127.0.0.1"};
struct acl * acl = acl_create( 1, lines, 0 );
union mysockaddr x;
char *lines[] = { "127.0.0.1" };
struct acl *acl = acl_create(1, lines, 0);
union mysockaddr x;
parse_ip_to_sockaddr( &x.generic, "127.0.0.2" );
fail_if( acl_includes( acl, &x ), "Excluded address was covered." );
parse_ip_to_sockaddr(&x.generic, "127.0.0.2");
fail_if(acl_includes(acl, &x), "Excluded address was covered.");
}
END_TEST
START_TEST( test_doesnt_include_other_address_when_netmask_specified )
START_TEST(test_doesnt_include_other_address_when_netmask_specified)
{
char *lines[] = {"127.0.0.1/32"};
struct acl * acl = acl_create( 1, lines, 0 );
union mysockaddr x;
char *lines[] = { "127.0.0.1/32" };
struct acl *acl = acl_create(1, lines, 0);
union mysockaddr x;
parse_ip_to_sockaddr( &x.generic, "127.0.0.2" );
fail_if( acl_includes( acl, &x ), "Excluded address was covered." );
parse_ip_to_sockaddr(&x.generic, "127.0.0.2");
fail_if(acl_includes(acl, &x), "Excluded address was covered.");
}
END_TEST
START_TEST( test_doesnt_include_other_address_when_multiple_entries_exist )
START_TEST(test_doesnt_include_other_address_when_multiple_entries_exist)
{
char *lines[] = {"127.0.0.1", "::1"};
struct acl * acl = acl_create( 2, lines, 0 );
union mysockaddr e0;
union mysockaddr e1;
char *lines[] = { "127.0.0.1", "::1" };
struct acl *acl = acl_create(2, lines, 0);
union mysockaddr e0;
union mysockaddr e1;
parse_ip_to_sockaddr( &e0.generic, "127.0.0.2" );
parse_ip_to_sockaddr( &e1.generic, "::2" );
parse_ip_to_sockaddr(&e0.generic, "127.0.0.2");
parse_ip_to_sockaddr(&e1.generic, "::2");
fail_if( acl_includes( acl, &e0 ), "Excluded address 0 was covered." );
fail_if( acl_includes( acl, &e1 ), "Excluded address 1 was covered." );
fail_if(acl_includes(acl, &e0), "Excluded address 0 was covered.");
fail_if(acl_includes(acl, &e1), "Excluded address 1 was covered.");
}
END_TEST
START_TEST( test_default_deny_rejects )
START_TEST(test_default_deny_rejects)
{
struct acl * acl = acl_create( 0, NULL, 1 );
union mysockaddr x;
struct acl *acl = acl_create(0, NULL, 1);
union mysockaddr x;
parse_ip_to_sockaddr( &x.generic, "127.0.0.1" );
parse_ip_to_sockaddr(&x.generic, "127.0.0.1");
fail_if( acl_includes( acl, &x ), "Default deny accepted." );
fail_if(acl_includes(acl, &x), "Default deny accepted.");
}
END_TEST
START_TEST( test_default_accept_rejects )
START_TEST(test_default_accept_rejects)
{
struct acl * acl = acl_create( 0, NULL, 0 );
union mysockaddr x;
struct acl *acl = acl_create(0, NULL, 0);
union mysockaddr x;
parse_ip_to_sockaddr( &x.generic, "127.0.0.1" );
parse_ip_to_sockaddr(&x.generic, "127.0.0.1");
fail_unless( acl_includes( acl, &x ), "Default accept rejected." );
fail_unless(acl_includes(acl, &x), "Default accept rejected.");
}
END_TEST
Suite* acl_suite(void)
Suite * acl_suite(void)
{
Suite *s = suite_create("acl");
TCase *tc_create = tcase_create("create");
TCase *tc_includes = tcase_create("includes");
TCase *tc_destroy = tcase_create("destroy");
Suite *s = suite_create("acl");
TCase *tc_create = tcase_create("create");
TCase *tc_includes = tcase_create("includes");
TCase *tc_destroy = tcase_create("destroy");
tcase_add_test(tc_create, test_null_acl);
tcase_add_test(tc_create, test_parses_single_line);
tcase_add_test(tc_includes, test_parses_multiple_lines);
tcase_add_test(tc_create, test_null_acl);
tcase_add_test(tc_create, test_parses_single_line);
tcase_add_test(tc_includes, test_parses_multiple_lines);
tcase_add_test(tc_includes, test_includes_single_address);
tcase_add_test(tc_includes, test_includes_single_address_when_netmask_specified_ipv4);
tcase_add_test(tc_includes, test_includes_single_address_when_netmask_specified_ipv6);
tcase_add_test(tc_includes, test_includes_single_address);
tcase_add_test(tc_includes,
test_includes_single_address_when_netmask_specified_ipv4);
tcase_add_test(tc_includes,
test_includes_single_address_when_netmask_specified_ipv6);
tcase_add_test(tc_includes, test_includes_single_address_when_multiple_entries_exist);
tcase_add_test(tc_includes,
test_includes_single_address_when_multiple_entries_exist);
tcase_add_test(tc_includes, test_doesnt_include_other_address);
tcase_add_test(tc_includes, test_doesnt_include_other_address_when_netmask_specified);
tcase_add_test(tc_includes, test_doesnt_include_other_address_when_multiple_entries_exist);
tcase_add_test(tc_includes, test_doesnt_include_other_address);
tcase_add_test(tc_includes,
test_doesnt_include_other_address_when_netmask_specified);
tcase_add_test(tc_includes,
test_doesnt_include_other_address_when_multiple_entries_exist);
tcase_add_test(tc_includes, test_default_deny_rejects);
tcase_add_test(tc_includes, test_default_accept_rejects);
tcase_add_test(tc_includes, test_default_deny_rejects);
tcase_add_test(tc_includes, test_default_accept_rejects);
tcase_add_test(tc_destroy, test_destroy_doesnt_crash);
tcase_add_test(tc_destroy, test_destroy_doesnt_crash);
suite_add_tcase(s, tc_create);
suite_add_tcase(s, tc_includes);
suite_add_tcase(s, tc_destroy);
suite_add_tcase(s, tc_create);
suite_add_tcase(s, tc_includes);
suite_add_tcase(s, tc_destroy);
return s;
return s;
}
int main(void)
{
#ifdef DEBUG
log_level = 0;
log_level = 0;
#else
log_level = 2;
log_level = 2;
#endif
int number_failed;
Suite *s = acl_suite();
SRunner *sr = srunner_create(s);
srunner_run_all(sr, CK_NORMAL);
log_level = 0;
number_failed = srunner_ntests_failed(sr);
srunner_free(sr);
return (number_failed == 0) ? 0 : 1;
int number_failed;
Suite *s = acl_suite();
SRunner *sr = srunner_create(s);
srunner_run_all(sr, CK_NORMAL);
log_level = 0;
number_failed = srunner_ntests_failed(sr);
srunner_free(sr);
return (number_failed == 0) ? 0 : 1;
}

View File

@@ -9,492 +9,492 @@
START_TEST(test_bit_set)
{
uint64_t num = 0;
bitfield_p bits = (bitfield_p) &num;
uint64_t num = 0;
bitfield_p bits = (bitfield_p) & num;
#define TEST_BIT_SET(bit, newvalue) \
bit_set(bits, (bit)); \
fail_unless(num == (newvalue), "num was %x instead of %x", num, (newvalue));
TEST_BIT_SET(0, 1);
TEST_BIT_SET(1, 3);
TEST_BIT_SET(2, 7);
TEST_BIT_SET(7, 0x87);
TEST_BIT_SET(63, 0x8000000000000087);
TEST_BIT_SET(0, 1);
TEST_BIT_SET(1, 3);
TEST_BIT_SET(2, 7);
TEST_BIT_SET(7, 0x87);
TEST_BIT_SET(63, 0x8000000000000087);
}
END_TEST
START_TEST(test_bit_clear)
{
uint64_t num = 0xffffffffffffffff;
bitfield_p bits = (bitfield_p) &num;
uint64_t num = 0xffffffffffffffff;
bitfield_p bits = (bitfield_p) & num;
#define TEST_BIT_CLEAR(bit, newvalue) \
bit_clear(bits, (bit)); \
fail_unless(num == (newvalue), "num was %x instead of %x", num, (newvalue));
TEST_BIT_CLEAR(0, 0xfffffffffffffffe);
TEST_BIT_CLEAR(1, 0xfffffffffffffffc);
TEST_BIT_CLEAR(2, 0xfffffffffffffff8);
TEST_BIT_CLEAR(7, 0xffffffffffffff78);
TEST_BIT_CLEAR(63,0x7fffffffffffff78);
TEST_BIT_CLEAR(0, 0xfffffffffffffffe);
TEST_BIT_CLEAR(1, 0xfffffffffffffffc);
TEST_BIT_CLEAR(2, 0xfffffffffffffff8);
TEST_BIT_CLEAR(7, 0xffffffffffffff78);
TEST_BIT_CLEAR(63, 0x7fffffffffffff78);
}
END_TEST
START_TEST(test_bit_tests)
{
uint64_t num = 0x5555555555555555;
bitfield_p bits = (bitfield_p) &num;
uint64_t num = 0x5555555555555555;
bitfield_p bits = (bitfield_p) & num;
fail_unless(bit_has_value(bits, 0, 1), "bit_has_value malfunction");
fail_unless(bit_has_value(bits, 1, 0), "bit_has_value malfunction");
fail_unless(bit_has_value(bits, 63, 0), "bit_has_value malfunction");
fail_unless(bit_is_set(bits, 0), "bit_is_set malfunction");
fail_unless(bit_is_clear(bits, 1), "bit_is_clear malfunction");
fail_unless(bit_is_set(bits, 62), "bit_is_set malfunction");
fail_unless(bit_is_clear(bits, 63), "bit_is_clear malfunction");
fail_unless(bit_has_value(bits, 0, 1), "bit_has_value malfunction");
fail_unless(bit_has_value(bits, 1, 0), "bit_has_value malfunction");
fail_unless(bit_has_value(bits, 63, 0), "bit_has_value malfunction");
fail_unless(bit_is_set(bits, 0), "bit_is_set malfunction");
fail_unless(bit_is_clear(bits, 1), "bit_is_clear malfunction");
fail_unless(bit_is_set(bits, 62), "bit_is_set malfunction");
fail_unless(bit_is_clear(bits, 63), "bit_is_clear malfunction");
}
END_TEST
START_TEST(test_bit_ranges)
{
bitfield_word_t buffer[BIT_WORDS_FOR_SIZE(4160)];
uint64_t *longs = (unsigned long*) buffer;
uint64_t i;
bitfield_word_t buffer[BIT_WORDS_FOR_SIZE(4160)];
uint64_t *longs = (uint64_t *) buffer;
uint64_t i;
memset(buffer, 0, 4160);
memset(buffer, 0, 4160);
for (i=0; i<64; i++) {
bit_set_range(buffer, i*64, i);
fail_unless(
longs[i] == (1UL<<i)-1,
"longs[%ld] = %lx SHOULD BE %lx",
i, longs[i], (1L<<i)-1
);
for (i = 0; i < 64; i++) {
bit_set_range(buffer, i * 64, i);
fail_unless(longs[i] == (1ULL << i) - 1,
"longs[%ld] = %lx SHOULD BE %lx",
i, longs[i], (1ULL << i) - 1);
fail_unless(longs[i+1] == 0, "bit_set_range overshot at i=%d", i);
}
fail_unless(longs[i + 1] == 0, "bit_set_range overshot at i=%d",
i);
}
for (i=0; i<64; i++) {
bit_clear_range(buffer, i*64, i);
fail_unless(longs[i] == 0, "bit_clear_range didn't work at i=%d", i);
}
for (i = 0; i < 64; i++) {
bit_clear_range(buffer, i * 64, i);
fail_unless(longs[i] == 0, "bit_clear_range didn't work at i=%d",
i);
}
}
END_TEST
START_TEST(test_bit_runs)
{
bitfield_word_t buffer[BIT_WORDS_FOR_SIZE(256)];
int i, ptr=0, runs[] = {
56,97,22,12,83,1,45,80,85,51,64,40,63,67,75,64,94,81,79,62
};
bitfield_word_t buffer[BIT_WORDS_FOR_SIZE(256)];
int i, ptr = 0, runs[] = {
56, 97, 22, 12, 83, 1, 45, 80, 85, 51, 64, 40, 63, 67, 75, 64, 94,
81, 79, 62
};
memset(buffer,0,256);
memset(buffer, 0, 256);
for (i=0; i < 20; i += 2) {
ptr += runs[i];
bit_set_range(buffer, ptr, runs[i+1]);
ptr += runs[i+1];
}
for (i = 0; i < 20; i += 2) {
ptr += runs[i];
bit_set_range(buffer, ptr, runs[i + 1]);
ptr += runs[i + 1];
}
ptr = 0;
ptr = 0;
for (i=0; i < 20; i += 1) {
int run = bit_run_count(buffer, ptr, 2048-ptr, NULL);
fail_unless(
run == runs[i],
"run %d should have been %d, was %d",
i, runs[i], run
);
ptr += runs[i];
}
for (i = 0; i < 20; i += 1) {
int run = bit_run_count(buffer, ptr, 2048 - ptr, NULL);
fail_unless(run == runs[i],
"run %d should have been %d, was %d", i, runs[i], run);
ptr += runs[i];
}
}
END_TEST
START_TEST(test_bitset)
{
struct bitset * map;
uint64_t *num;
struct bitset *map;
uint64_t *num;
map = bitset_alloc(6400, 100);
num = (uint64_t*) map->bits;
map = bitset_alloc(6400, 100);
num = (uint64_t *) map->bits;
bitset_set_range(map,0,50);
ck_assert_int_eq(1, *num);
bitset_set_range(map,99,1);
ck_assert_int_eq(1, *num);
bitset_set_range(map,100,1);
ck_assert_int_eq(3, *num);
bitset_set_range(map,0,800);
ck_assert_int_eq(255, *num);
bitset_set_range(map,1499,2);
ck_assert_int_eq(0xc0ff, *num);
bitset_clear_range(map,1499,2);
ck_assert_int_eq(255, *num);
bitset_set_range(map, 0, 50);
ck_assert_int_eq(1, *num);
bitset_set_range(map, 99, 1);
ck_assert_int_eq(1, *num);
bitset_set_range(map, 100, 1);
ck_assert_int_eq(3, *num);
bitset_set_range(map, 0, 800);
ck_assert_int_eq(255, *num);
bitset_set_range(map, 1499, 2);
ck_assert_int_eq(0xc0ff, *num);
bitset_clear_range(map, 1499, 2);
ck_assert_int_eq(255, *num);
*num = 0;
bitset_set_range(map, 1499, 2);
bitset_clear_range(map, 1300, 200);
ck_assert_int_eq(0x8000, *num);
*num = 0;
bitset_set_range(map, 1499, 2);
bitset_clear_range(map, 1300, 200);
ck_assert_int_eq(0x8000, *num);
*num = 0;
bitset_set_range(map, 0, 6400);
ck_assert_int_eq(0xffffffffffffffff, *num);
bitset_clear_range(map, 3200, 400);
ck_assert_int_eq(0xfffffff0ffffffff, *num);
*num = 0;
bitset_set_range(map, 0, 6400);
ck_assert_int_eq(0xffffffffffffffff, *num);
bitset_clear_range(map, 3200, 400);
ck_assert_int_eq(0xfffffff0ffffffff, *num);
}
END_TEST
START_TEST( test_bitset_set )
START_TEST(test_bitset_set)
{
struct bitset * map;
uint64_t run;
struct bitset *map;
uint64_t run;
map = bitset_alloc(64, 1);
map = bitset_alloc(64, 1);
assert_bitset_is( map, 0x0000000000000000 );
bitset_set( map );
assert_bitset_is( map, 0xffffffffffffffff );
bitset_free( map );
assert_bitset_is(map, 0x0000000000000000);
bitset_set(map);
assert_bitset_is(map, 0xffffffffffffffff);
bitset_free(map);
map = bitset_alloc( 6400, 100 );
assert_bitset_is( map, 0x0000000000000000 );
bitset_set( map );
assert_bitset_is( map, 0xffffffffffffffff );
bitset_free( map );
map = bitset_alloc(6400, 100);
assert_bitset_is(map, 0x0000000000000000);
bitset_set(map);
assert_bitset_is(map, 0xffffffffffffffff);
bitset_free(map);
// Now do something large and representative
map = bitset_alloc( 53687091200, 4096 );
bitset_set( map );
// Now do something large and representative
map = bitset_alloc(53687091200, 4096);
bitset_set(map);
run = bitset_run_count( map, 0, 53687091200 );
ck_assert_int_eq( run, 53687091200 );
bitset_free( map );
run = bitset_run_count(map, 0, 53687091200);
ck_assert_int_eq(run, 53687091200);
bitset_free(map);
}
END_TEST
START_TEST( test_bitset_clear )
START_TEST(test_bitset_clear)
{
struct bitset * map;
uint64_t *num;
uint64_t run;
struct bitset *map;
uint64_t *num;
uint64_t run;
map = bitset_alloc(64, 1);
num = (uint64_t*) map->bits;
map = bitset_alloc(64, 1);
num = (uint64_t *) map->bits;
ck_assert_int_eq( 0x0000000000000000, *num );
bitset_set( map );
bitset_clear( map );
ck_assert_int_eq( 0x0000000000000000, *num );
ck_assert_int_eq(0x0000000000000000, *num);
bitset_set(map);
bitset_clear(map);
ck_assert_int_eq(0x0000000000000000, *num);
bitset_free( map );
bitset_free(map);
// Now do something large and representative
map = bitset_alloc( 53687091200, 4096 );
bitset_set( map );
bitset_clear( map );
run = bitset_run_count( map, 0, 53687091200 );
ck_assert_int_eq( run, 53687091200 );
bitset_free( map );
// Now do something large and representative
map = bitset_alloc(53687091200, 4096);
bitset_set(map);
bitset_clear(map);
run = bitset_run_count(map, 0, 53687091200);
ck_assert_int_eq(run, 53687091200);
bitset_free(map);
}
END_TEST
START_TEST( test_bitset_set_range )
START_TEST(test_bitset_set_range)
{
struct bitset* map = bitset_alloc( 64, 1 );
assert_bitset_is( map, 0x0000000000000000 );
struct bitset *map = bitset_alloc(64, 1);
assert_bitset_is(map, 0x0000000000000000);
bitset_set_range( map, 8, 8 );
assert_bitset_is( map, 0x000000000000ff00 );
bitset_set_range(map, 8, 8);
assert_bitset_is(map, 0x000000000000ff00);
bitset_clear( map );
assert_bitset_is( map, 0x0000000000000000 );
bitset_set_range( map, 0, 0 );
assert_bitset_is( map, 0x0000000000000000 );
bitset_clear(map);
assert_bitset_is(map, 0x0000000000000000);
bitset_set_range(map, 0, 0);
assert_bitset_is(map, 0x0000000000000000);
bitset_free( map );
bitset_free(map);
}
END_TEST
START_TEST( test_bitset_clear_range )
START_TEST(test_bitset_clear_range)
{
struct bitset* map = bitset_alloc( 64, 1 );
bitset_set( map );
assert_bitset_is( map, 0xffffffffffffffff );
struct bitset *map = bitset_alloc(64, 1);
bitset_set(map);
assert_bitset_is(map, 0xffffffffffffffff);
bitset_clear_range( map, 8, 8 );
assert_bitset_is( map, 0xffffffffffff00ff );
bitset_clear_range(map, 8, 8);
assert_bitset_is(map, 0xffffffffffff00ff);
bitset_set( map );
assert_bitset_is( map, 0xffffffffffffffff );
bitset_clear_range( map, 0, 0 );
assert_bitset_is( map, 0xffffffffffffffff );
bitset_set(map);
assert_bitset_is(map, 0xffffffffffffffff);
bitset_clear_range(map, 0, 0);
assert_bitset_is(map, 0xffffffffffffffff);
bitset_free( map );
bitset_free(map);
}
END_TEST
START_TEST( test_bitset_run_count )
START_TEST(test_bitset_run_count)
{
struct bitset* map = bitset_alloc( 64, 1 );
uint64_t run;
struct bitset *map = bitset_alloc(64, 1);
uint64_t run;
assert_bitset_is( map, 0x0000000000000000 );
assert_bitset_is(map, 0x0000000000000000);
run = bitset_run_count( map, 0, 64 );
ck_assert_int_eq( 64, run );
run = bitset_run_count(map, 0, 64);
ck_assert_int_eq(64, run);
bitset_set_range( map, 0, 32 );
assert_bitset_is( map, 0x00000000ffffffff );
bitset_set_range(map, 0, 32);
assert_bitset_is(map, 0x00000000ffffffff);
run = bitset_run_count( map, 0, 64 );
ck_assert_int_eq( 32, run );
run = bitset_run_count(map, 0, 64);
ck_assert_int_eq(32, run);
run = bitset_run_count( map, 0, 16 );
ck_assert_int_eq( 16, run );
run = bitset_run_count(map, 0, 16);
ck_assert_int_eq(16, run);
run = bitset_run_count( map, 16, 64 );
ck_assert_int_eq( 16, run );
run = bitset_run_count(map, 16, 64);
ck_assert_int_eq(16, run);
run = bitset_run_count( map, 31, 64 );
ck_assert_int_eq( 1, run );
run = bitset_run_count(map, 31, 64);
ck_assert_int_eq(1, run);
run = bitset_run_count( map, 32, 64 );
ck_assert_int_eq( 32, run );
run = bitset_run_count(map, 32, 64);
ck_assert_int_eq(32, run);
run = bitset_run_count( map, 32, 32 );
ck_assert_int_eq( 32, run );
run = bitset_run_count(map, 32, 32);
ck_assert_int_eq(32, run);
run = bitset_run_count( map, 32, 16 );
ck_assert_int_eq( 16, run );
run = bitset_run_count(map, 32, 16);
ck_assert_int_eq(16, run);
bitset_free( map );
bitset_free(map);
map = bitset_alloc( 6400, 100 );
assert_bitset_is( map, 0x0000000000000000 );
map = bitset_alloc(6400, 100);
assert_bitset_is(map, 0x0000000000000000);
run = bitset_run_count( map, 0, 6400 );
ck_assert_int_eq( 6400, run );
run = bitset_run_count(map, 0, 6400);
ck_assert_int_eq(6400, run);
bitset_set_range( map, 0, 3200 );
bitset_set_range(map, 0, 3200);
run = bitset_run_count( map, 0, 6400 );
ck_assert_int_eq( 3200, run );
run = bitset_run_count(map, 0, 6400);
ck_assert_int_eq(3200, run);
run = bitset_run_count( map, 1, 6400 );
ck_assert_int_eq( 3199, run );
run = bitset_run_count(map, 1, 6400);
ck_assert_int_eq(3199, run);
run = bitset_run_count( map, 3200, 6400 );
ck_assert_int_eq( 3200, run );
run = bitset_run_count(map, 3200, 6400);
ck_assert_int_eq(3200, run);
run = bitset_run_count( map, 6500, 6400 );
ck_assert_int_eq( 0, run );
bitset_free( map );
run = bitset_run_count(map, 6500, 6400);
ck_assert_int_eq(0, run);
bitset_free(map);
// Now do something large and representative
map = bitset_alloc( 53687091200, 4096 );
bitset_set( map );
run = bitset_run_count( map, 0, 53687091200 );
ck_assert_int_eq( run, 53687091200 );
// Now do something large and representative
map = bitset_alloc(53687091200, 4096);
bitset_set(map);
run = bitset_run_count(map, 0, 53687091200);
ck_assert_int_eq(run, 53687091200);
bitset_free( map );
bitset_free(map);
}
END_TEST
START_TEST( test_bitset_set_range_doesnt_push_to_stream )
START_TEST(test_bitset_set_range_doesnt_push_to_stream)
{
struct bitset *map = bitset_alloc( 64, 1 );
bitset_set_range( map, 0, 64 );
ck_assert_int_eq( 0, bitset_stream_size( map ) );
bitset_free( map );
struct bitset *map = bitset_alloc(64, 1);
bitset_set_range(map, 0, 64);
ck_assert_int_eq(0, bitset_stream_size(map));
bitset_free(map);
}
END_TEST
START_TEST( test_bitset_clear_range_doesnt_push_to_stream )
START_TEST(test_bitset_clear_range_doesnt_push_to_stream)
{
struct bitset *map = bitset_alloc( 64, 1 );
bitset_clear_range( map, 0, 64 );
ck_assert_int_eq( 0, bitset_stream_size( map ) );
bitset_free( map );
struct bitset *map = bitset_alloc(64, 1);
bitset_clear_range(map, 0, 64);
ck_assert_int_eq(0, bitset_stream_size(map));
bitset_free(map);
}
END_TEST
START_TEST(test_bitset_enable_stream)
{
struct bitset *map = bitset_alloc( 64, 1 );
struct bitset_stream_entry result;
memset( &result, 0, sizeof( result ) );
struct bitset *map = bitset_alloc(64, 1);
struct bitset_stream_entry result;
memset(&result, 0, sizeof(result));
bitset_enable_stream( map );
bitset_enable_stream(map);
ck_assert_int_eq( 1, map->stream_enabled );
ck_assert_int_eq(1, map->stream_enabled);
bitset_stream_dequeue( map, &result );
bitset_stream_dequeue(map, &result);
ck_assert_int_eq( BITSET_STREAM_ON, result.event );
ck_assert_int_eq( 0, result.from );
ck_assert_int_eq( 64, result.len );
ck_assert_int_eq(BITSET_STREAM_ON, result.event);
ck_assert_int_eq(0, result.from);
ck_assert_int_eq(64, result.len);
bitset_free( map );
bitset_free(map);
}
END_TEST
START_TEST(test_bitset_disable_stream)
{
struct bitset *map = bitset_alloc( 64, 1 );
struct bitset_stream_entry result;
memset( &result, 0, sizeof( result ) );
struct bitset *map = bitset_alloc(64, 1);
struct bitset_stream_entry result;
memset(&result, 0, sizeof(result));
bitset_enable_stream( map );
bitset_disable_stream( map );
bitset_enable_stream(map);
bitset_disable_stream(map);
ck_assert_int_eq( 0, map->stream_enabled );
ck_assert_int_eq( 2, bitset_stream_size( map ) );
ck_assert_int_eq(0, map->stream_enabled);
ck_assert_int_eq(2, bitset_stream_size(map));
bitset_stream_dequeue( map, NULL ); // ON
bitset_stream_dequeue( map, &result ); // OFF
bitset_stream_dequeue(map, NULL); // ON
bitset_stream_dequeue(map, &result); // OFF
ck_assert_int_eq( BITSET_STREAM_OFF, result.event );
ck_assert_int_eq( 0, result.from );
ck_assert_int_eq( 64, result.len );
ck_assert_int_eq(BITSET_STREAM_OFF, result.event);
ck_assert_int_eq(0, result.from);
ck_assert_int_eq(64, result.len);
bitset_free( map );
bitset_free(map);
}
END_TEST
START_TEST(test_bitset_stream_with_set_range)
{
struct bitset *map = bitset_alloc( 64, 1 );
struct bitset_stream_entry result;
memset( &result, 0, sizeof( result ) );
struct bitset *map = bitset_alloc(64, 1);
struct bitset_stream_entry result;
memset(&result, 0, sizeof(result));
bitset_enable_stream( map );
bitset_set_range( map, 0, 32 );
bitset_enable_stream(map);
bitset_set_range(map, 0, 32);
ck_assert_int_eq( 2, bitset_stream_size( map ) );
ck_assert_int_eq(2, bitset_stream_size(map));
bitset_stream_dequeue( map, NULL ); // ON
bitset_stream_dequeue( map, &result ); // SET
bitset_stream_dequeue(map, NULL); // ON
bitset_stream_dequeue(map, &result); // SET
ck_assert_int_eq( BITSET_STREAM_SET, result.event );
ck_assert_int_eq( 0, result.from );
ck_assert_int_eq( 32, result.len );
ck_assert_int_eq(BITSET_STREAM_SET, result.event);
ck_assert_int_eq(0, result.from);
ck_assert_int_eq(32, result.len);
bitset_free( map );
bitset_free(map);
}
END_TEST
START_TEST(test_bitset_stream_with_clear_range)
{
struct bitset *map = bitset_alloc( 64, 1 );
struct bitset_stream_entry result;
memset( &result, 0, sizeof( result ) );
struct bitset *map = bitset_alloc(64, 1);
struct bitset_stream_entry result;
memset(&result, 0, sizeof(result));
bitset_enable_stream( map );
bitset_clear_range( map, 0, 32 );
ck_assert_int_eq( 2, bitset_stream_size( map ) );
bitset_enable_stream(map);
bitset_clear_range(map, 0, 32);
ck_assert_int_eq(2, bitset_stream_size(map));
bitset_stream_dequeue( map, NULL ); // ON
bitset_stream_dequeue( map, &result ); // UNSET
bitset_stream_dequeue(map, NULL); // ON
bitset_stream_dequeue(map, &result); // UNSET
ck_assert_int_eq( BITSET_STREAM_UNSET, result.event );
ck_assert_int_eq( 0, result.from );
ck_assert_int_eq( 32, result.len );
ck_assert_int_eq(BITSET_STREAM_UNSET, result.event);
ck_assert_int_eq(0, result.from);
ck_assert_int_eq(32, result.len);
bitset_free( map );
bitset_free(map);
}
END_TEST
START_TEST(test_bitset_stream_size)
{
struct bitset *map = bitset_alloc( 64, 1 );
bitset_enable_stream( map );
bitset_set_range( map, 0, 32 );
bitset_set_range( map, 16, 32 );
bitset_set_range( map, 7, 16 );
struct bitset *map = bitset_alloc(64, 1);
bitset_enable_stream(map);
bitset_set_range(map, 0, 32);
bitset_set_range(map, 16, 32);
bitset_set_range(map, 7, 16);
bitset_clear_range( map, 0, 32 );
bitset_clear_range( map, 16, 32 );
bitset_clear_range( map, 48, 16 );
bitset_disable_stream( map );
bitset_clear_range(map, 0, 32);
bitset_clear_range(map, 16, 32);
bitset_clear_range(map, 48, 16);
bitset_disable_stream(map);
ck_assert_int_eq( 8, bitset_stream_size( map ) );
ck_assert_int_eq(8, bitset_stream_size(map));
bitset_free( map );
bitset_free(map);
}
END_TEST
START_TEST(test_bitset_stream_queued_bytes)
{
struct bitset *map = bitset_alloc( 64, 1 );
bitset_enable_stream( map );
bitset_set_range( map, 0, 32 );
bitset_set_range( map, 16, 32 );
bitset_set_range( map, 7, 16 );
struct bitset *map = bitset_alloc(64, 1);
bitset_enable_stream(map);
bitset_set_range(map, 0, 32);
bitset_set_range(map, 16, 32);
bitset_set_range(map, 7, 16);
bitset_clear_range( map, 0, 32 );
bitset_clear_range( map, 16, 32 );
bitset_clear_range( map, 48, 16 );
bitset_clear_range( map, 0, 2 );
bitset_disable_stream( map );
bitset_clear_range(map, 0, 32);
bitset_clear_range(map, 16, 32);
bitset_clear_range(map, 48, 16);
bitset_clear_range(map, 0, 2);
bitset_disable_stream(map);
ck_assert_int_eq( 64, bitset_stream_queued_bytes( map, BITSET_STREAM_ON ) );
ck_assert_int_eq( 80, bitset_stream_queued_bytes( map, BITSET_STREAM_SET ) );
ck_assert_int_eq( 82, bitset_stream_queued_bytes( map, BITSET_STREAM_UNSET ) );
ck_assert_int_eq( 64, bitset_stream_queued_bytes( map, BITSET_STREAM_OFF ) );
bitset_free( map );
ck_assert_int_eq(64,
bitset_stream_queued_bytes(map, BITSET_STREAM_ON));
ck_assert_int_eq(80,
bitset_stream_queued_bytes(map, BITSET_STREAM_SET));
ck_assert_int_eq(82,
bitset_stream_queued_bytes(map, BITSET_STREAM_UNSET));
ck_assert_int_eq(64,
bitset_stream_queued_bytes(map, BITSET_STREAM_OFF));
bitset_free(map);
}
END_TEST
Suite* bitset_suite(void)
Suite * bitset_suite(void)
{
Suite *s = suite_create("bitset");
Suite *s = suite_create("bitset");
TCase *tc_bit = tcase_create("bit");
tcase_add_test(tc_bit, test_bit_set);
tcase_add_test(tc_bit, test_bit_clear);
tcase_add_test(tc_bit, test_bit_tests);
tcase_add_test(tc_bit, test_bit_ranges);
tcase_add_test(tc_bit, test_bit_runs);
suite_add_tcase(s, tc_bit);
TCase *tc_bit = tcase_create("bit");
tcase_add_test(tc_bit, test_bit_set);
tcase_add_test(tc_bit, test_bit_clear);
tcase_add_test(tc_bit, test_bit_tests);
tcase_add_test(tc_bit, test_bit_ranges);
tcase_add_test(tc_bit, test_bit_runs);
suite_add_tcase(s, tc_bit);
TCase *tc_bitset = tcase_create("bitset");
tcase_add_test(tc_bitset, test_bitset);
tcase_add_test(tc_bitset, test_bitset_set);
tcase_add_test(tc_bitset, test_bitset_clear);
tcase_add_test(tc_bitset, test_bitset_run_count);
tcase_add_test(tc_bitset, test_bitset_set_range);
tcase_add_test(tc_bitset, test_bitset_clear_range);
tcase_add_test(tc_bitset, test_bitset_set_range_doesnt_push_to_stream);
tcase_add_test(tc_bitset, test_bitset_clear_range_doesnt_push_to_stream);
suite_add_tcase(s, tc_bitset);
TCase *tc_bitset = tcase_create("bitset");
tcase_add_test(tc_bitset, test_bitset);
tcase_add_test(tc_bitset, test_bitset_set);
tcase_add_test(tc_bitset, test_bitset_clear);
tcase_add_test(tc_bitset, test_bitset_run_count);
tcase_add_test(tc_bitset, test_bitset_set_range);
tcase_add_test(tc_bitset, test_bitset_clear_range);
tcase_add_test(tc_bitset, test_bitset_set_range_doesnt_push_to_stream);
tcase_add_test(tc_bitset,
test_bitset_clear_range_doesnt_push_to_stream);
suite_add_tcase(s, tc_bitset);
TCase *tc_bitset_stream = tcase_create("bitset_stream");
tcase_add_test(tc_bitset_stream, test_bitset_enable_stream);
tcase_add_test(tc_bitset_stream, test_bitset_disable_stream);
tcase_add_test(tc_bitset_stream, test_bitset_stream_with_set_range);
tcase_add_test(tc_bitset_stream, test_bitset_stream_with_clear_range);
tcase_add_test(tc_bitset_stream, test_bitset_stream_size);
tcase_add_test(tc_bitset_stream, test_bitset_stream_queued_bytes);
suite_add_tcase(s, tc_bitset_stream);
TCase *tc_bitset_stream = tcase_create("bitset_stream");
tcase_add_test(tc_bitset_stream, test_bitset_enable_stream);
tcase_add_test(tc_bitset_stream, test_bitset_disable_stream);
tcase_add_test(tc_bitset_stream, test_bitset_stream_with_set_range);
tcase_add_test(tc_bitset_stream, test_bitset_stream_with_clear_range);
tcase_add_test(tc_bitset_stream, test_bitset_stream_size);
tcase_add_test(tc_bitset_stream, test_bitset_stream_queued_bytes);
suite_add_tcase(s, tc_bitset_stream);
return s;
return s;
}
int main(void)
{
int number_failed;
Suite *s = bitset_suite();
SRunner *sr = srunner_create(s);
srunner_run_all(sr, CK_NORMAL);
number_failed = srunner_ntests_failed(sr);
srunner_free(sr);
return (number_failed == 0) ? 0 : 1;
int number_failed;
Suite *s = bitset_suite();
SRunner *sr = srunner_create(s);
srunner_run_all(sr, CK_NORMAL);
number_failed = srunner_ntests_failed(sr);
srunner_free(sr);
return (number_failed == 0) ? 0 : 1;
}

Some files were not shown because too many files have changed in this diff Show More