Bump the @version pragmas to 3.6 #2
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Since the default version was switched to 3.6 earlier, bump the @Version of the default config and the in the func-tests too. Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
This has been merged independently, closing. |
talien
pushed a commit
to talien/syslog-ng
that referenced
this pull request
Jan 10, 2014
Instead of being tied to GStrings only, support registering new kinds of stacks, which will be automatiaclly free'd on thread shutdown. For convenience, wrappers are provided for GString-based scratch buffer acquire and release, and that stack is automatically registered and free'd by the library itself. This fixes syslog-ng#2. Signed-off-by: Gergely Nagy <algernon@madhouse-project.org>
talien
pushed a commit
to talien/syslog-ng
that referenced
this pull request
Jan 10, 2014
Instead of being tied to GStrings only, support registering new kinds of stacks, which will be automatiaclly free'd on thread shutdown. For convenience, wrappers are provided for GString-based scratch buffer acquire and release, and that stack is automatically registered and free'd by the library itself. This fixes syslog-ng#2. Signed-off-by: Gergely Nagy <algernon@madhouse-project.org>
Closed
bazsi
added a commit
that referenced
this pull request
Nov 17, 2015
There's a strict ordering requirement between suppress_lock and interacting with the main loop. The reason is that the main thread (running the main loop) sometimes acquires suppress_lock (at suppress timer expiration) and while blocking on suppress_lock it cannot service main_loop_calls() This patch makes it sure that ml_batched_timer_update/cancel calls are only done with the suppress lock released. If we do this, we might have a few unfortunate side effects due to races that this patch also tries to handle: Two messages race, one of these matches the recorded last message, the other doesn't. In this case, moving the update on the suppress_timer outside of the lock region might cause two different races: 1) matching message comes first, then non-matching This case the suppress_lock protected region decides that the suppress timer needs to fire (#1) and then the other decides that it needs to be cancelled. (#2) If these are processed in order, then we are the same as if the two was also protected by the mutex (which is ok) If they are reversed, e.g. we first cancels the timer and the second arms it, then we might have a timer wakeup which will find no suppressed messages to report (as the non-matching message will set last_msg_count to zero). This spurious wakeup should be handled by the expiration callback. 1) non-matching message comes first, then matching This is simply a message reordering case, e.g. we don't want any suppressions to be emitted. In this case the locked regions finds that neither messages matched the recorded one, thus both times they decide to cancel the timer, which is ok. Timer cancellation can be reordered as they will have the same effect anyway. Signed-off-by: Balazs Scheidler <balazs.scheidler@balabit.com>
bazsi
added a commit
that referenced
this pull request
Nov 17, 2015
There's a strict ordering requirement between suppress_lock and interacting with the main loop. The reason is that the main thread (running the main loop) sometimes acquires suppress_lock (at suppress timer expiration) and while blocking on suppress_lock it cannot service main_loop_calls() This patch makes it sure that ml_batched_timer_update/cancel calls are only done with the suppress lock released. If we do this, we might have a few unfortunate side effects due to races that this patch also tries to handle: Two messages race, one of these matches the recorded last message, the other doesn't. In this case, moving the update on the suppress_timer outside of the lock region might cause two different races: 1) matching message comes first, then non-matching This case the suppress_lock protected region decides that the suppress timer needs to fire (#1) and then the other decides that it needs to be cancelled. (#2) If these are processed in order, then we are the same as if the two was also protected by the mutex (which is ok) If they are reversed, e.g. we first cancels the timer and the second arms it, then we might have a timer wakeup which will find no suppressed messages to report (as the non-matching message will set last_msg_count to zero). This spurious wakeup should be handled by the expiration callback. 1) non-matching message comes first, then matching This is simply a message reordering case, e.g. we don't want any suppressions to be emitted. In this case the locked regions finds that neither messages matched the recorded one, thus both times they decide to cancel the timer, which is ok. Timer cancellation can be reordered as they will have the same effect anyway. Signed-off-by: Balazs Scheidler <balazs.scheidler@balabit.com>
MrAnno
referenced
this pull request
in MrAnno/syslog-ng
Feb 11, 2016
There's a strict ordering requirement between suppress_lock and interacting with the main loop. The reason is that the main thread (running the main loop) sometimes acquires suppress_lock (at suppress timer expiration) and while blocking on suppress_lock it cannot service main_loop_calls() This patch makes it sure that ml_batched_timer_update/cancel calls are only done with the suppress lock released. If we do this, we might have a few unfortunate side effects due to races that this patch also tries to handle: Two messages race, one of these matches the recorded last message, the other doesn't. In this case, moving the update on the suppress_timer outside of the lock region might cause two different races: 1) matching message comes first, then non-matching This case the suppress_lock protected region decides that the suppress timer needs to fire (#1) and then the other decides that it needs to be cancelled. (#2) If these are processed in order, then we are the same as if the two was also protected by the mutex (which is ok) If they are reversed, e.g. we first cancels the timer and the second arms it, then we might have a timer wakeup which will find no suppressed messages to report (as the non-matching message will set last_msg_count to zero). This spurious wakeup should be handled by the expiration callback. 1) non-matching message comes first, then matching This is simply a message reordering case, e.g. we don't want any suppressions to be emitted. In this case the locked regions finds that neither messages matched the recorded one, thus both times they decide to cancel the timer, which is ok. Timer cancellation can be reordered as they will have the same effect anyway. Signed-off-by: Balazs Scheidler <balazs.scheidler@balabit.com>
bazsi
added a commit
that referenced
this pull request
Feb 21, 2016
syslog-ng was crashing in my development environment at startup if the systemd-journal() driver was used, in a backtrace like this: Program received signal SIGSEGV, Segmentation fault. g_module_symbol (module=0x6771b0, symbol_name=0x7ffff4942a65 "sd_journal_open", symbol=0x7ffff7f838f0 <sd_journal_open>) at /build/glib2.0-ajuDY6/glib2.0-2.46.1/./gmodule/gmodule.c:818 818 /build/glib2.0-ajuDY6/glib2.0-2.46.1/./gmodule/gmodule.c: No such file or directory. (gdb) bt #0 g_module_symbol (module=0x6771b0, symbol_name=0x7ffff4942a65 "sd_journal_open", symbol=0x7ffff7f838f0 <sd_journal_open>) at /build/glib2.0-ajuDY6/glib2.0-2.46.1/./gmodule/gmodule.c:818 #1 0x00007ffff493e6cd in load_journald_subsystem () at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/modules/systemd-journal/journald-subsystem.c:105 #2 0x00007ffff493cd0c in systemd_journal_module_init (cfg=0x63a2a0, args=0x0) at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/modules/systemd-journal/systemd-journal-plugin.c:44 #3 0x00007ffff7b4a9d3 in plugin_load_module (module_name=0x6572a0 "sdjournal", cfg=0x63a2a0, args=0x0) at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/lib/plugin.c:385 #4 0x00007ffff7b4a18c in plugin_find (cfg=0x63a2a0, plugin_type=3, plugin_name=0x675c80 "systemd-journal") at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/lib/plugin.c:155 #5 0x00007ffff7b5ee8e in main_parse (lexer=0x63f010, dummy=0x7fffffffd0c8, arg=0x0) at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/lib/cfg-grammar.y:580 #6 0x00007ffff7b2a7a5 in cfg_parser_parse (self=0x7ffff7dd4ee0 <main_parser>, lexer=0x63f010, instance=0x7fffffffd0c8, arg=0x0) at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/lib/cfg-parser.h:83 #7 0x00007ffff7b2b7f7 in cfg_run_parser (self=0x63a2a0, lexer=0x63f010, parser=0x7ffff7dd4ee0 <main_parser>, result=0x7fffffffd0c8, arg=0x0) at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/lib/cfg.c:430 #8 0x00007ffff7b2ba3a in cfg_read_config (self=0x63a2a0, fname=0x60e940 "etc/syslog-ng.conf", syntax_only=0, preprocess_into=0x0) at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/lib/cfg.c:502 #9 0x00007ffff7b45368 in main_loop_read_and_init_config () at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/lib/mainloop.c:447 #10 0x0000000000401b31 in main (argc=1, argv=0x7fffffffd238) at /home/bazsi/zwa/projects/syslog-ng-ose-3.6/source/syslog-ng/syslog-ng/main.c:261 My belief is that the root cause was that the global variables in journald-subsystem.c are named the same as the symbols in libsystemd.so, and both are symbols with global visibility. My theory was that due to the name clash, the actual pointer passed to g_module_symbol() was pointing to the actual code of the functions, instead of the pointer we wanted to store the values. I didn't completely confirm this theory, but adding "static" qualifiers to the variables resolved the issue for me. Honestly, I don't really like how journal-systemd.c is organized, but this was only a distraction for me and I was satisfied with resolving the problem for now. Signed-off-by: Balazs Scheidler <balazs.scheidler@balabit.com>
Kokan
added a commit
to Kokan/syslog-ng
that referenced
this pull request
Aug 29, 2020
Parameterized tests are the worst in Criterion. Specially with pointers and ASLR.
Causes test to fail when executed with ASAN:
```
==3181004==ERROR: AddressSanitizer: SEGV on unknown address 0x7faeee7938a8 (pc 0x7f0f053880e7 bp 0x7faeee7938a0 sp 0x7ffe7fe59510 T0)
==3181004==The signal is caused by a READ memory access.
#0 0x7f0f053880e7 in PyObject_Call (/usr/lib/libpython3.8.so.1.0+0x1420e7)
#1 0x4f7f7c in python_ack_tracker_test_ack_tracker_factory_impl /home/kokan/src/syslog-ng/build/../modules/python/tests/test_python_ack_tracker.c:124:42
syslog-ng#2 0x7f0f05225e3e in criterion_internal_test_main (/usr/lib/libcriterion.so.3+0x8e3e)
syslog-ng#3 0x4f7e34 in python_ack_tracker_test_ack_tracker_factory_jmp /home/kokan/src/syslog-ng/build/../modules/python/tests/test_python_ack_tracker.c:119:1
syslog-ng#4 0x7f0f05225308 (/usr/lib/libcriterion.so.3+0x8308)
syslog-ng#5 0x7f0f05236b4c (/usr/lib/libcriterion.so.3+0x19b4c)
syslog-ng#6 0x7f0f04626151 in __libc_start_main (/usr/lib/libc.so.6+0x28151)
syslog-ng#7 0x41f43d in _start (/home/kokan/src/syslog-ng/build/modules/python/tests/Debug/test_python_ack_tracker+0x41f43d)
```
Signed-off-by: Kokan <kokaipeter@gmail.com>
Kokan
added a commit
to Kokan/syslog-ng
that referenced
this pull request
Oct 2, 2020
Every time a key created a counter names is allocated. But it is not
freed. The only time it is freed when a new stats-cluster is created
from it (first time looked up) and the counter_names owenership
transfared to the stats-cluster.
With this patch the allocation is delayed to the ownership change, thus
only allocated if there is still a pointer to the memory.
Reproduction:
1. syslog-ng configuration with at least one network source
2. start syslog-ng with a leak detector
(2.1 optional do restart (in order to have the exact leak as above))
3. stop syslog-ng
=================================================================
==2095056==ERROR: LeakSanitizer: detected memory leaks
Direct leak of 8 byte(s) in 1 object(s) allocated from:
#0 0x55a4f86db781 in calloc (/tmp/install/sbin/syslog-ng+0xc5781)
#1 0x7f343f03a941 in g_malloc0 (/usr/lib/libglib-2.0.so.0+0x50941)
syslog-ng#2 0x7f343acb37c7 in _make_connection_conter_stats_queryable /home/user/src/syslog-ng/build/../modules/afsocket/afsocket-source.c:1007:9
syslog-ng#3 0x7f343acb30f3 in afsocket_sd_init_method /home/user/src/syslog-ng/build/../modules/afsocket/afsocket-source.c:1060:5
syslog-ng#4 0x7f343acbc358 in afinet_sd_init /home/user/src/syslog-ng/build/../modules/afsocket/afinet-source.c:103:8
syslog-ng#5 0x7f343f248e1a in log_pipe_init /home/user/src/syslog-ng/build/../lib/logpipe.h:301:23
syslog-ng#6 0x7f343f248c88 in cfg_tree_start /home/user/src/syslog-ng/build/../lib/cfg-tree.c:1419:12
syslog-ng#7 0x7f343f23af71 in cfg_init /home/user/src/syslog-ng/build/../lib/cfg.c:368:8
syslog-ng#8 0x7f343f270b3c in main_loop_reload_config_apply /home/user/src/syslog-ng/build/../lib/mainloop.c:276:41
syslog-ng#9 0x7f343f273162 in _consume_action /home/user/src/syslog-ng/build/../lib/mainloop-worker.c:240:3
syslog-ng#10 0x7f343f27321f in _invoke_sync_call_actions /home/user/src/syslog-ng/build/../lib/mainloop-worker.c:250:7
syslog-ng#11 0x7f343f2736ba in main_loop_worker_sync_call /home/user/src/syslog-ng/build/../lib/mainloop-worker.c:396:7
syslog-ng#12 0x7f343f2709d1 in main_loop_reload_config_commence /home/user/src/syslog-ng/build/../lib/mainloop.c:340:3
syslog-ng#13 0x7f343f275852 in control_connection_reload /home/user/src/syslog-ng/build/../lib/mainloop-control.c:167:3
syslog-ng#14 0x7f343f290c80 in control_connection_io_input /home/user/src/syslog-ng/build/../lib/control/control-server.c:378:3
syslog-ng#15 0x7f343efd4b2c (/usr/lib/libivykis.so.0+0x6b2c)
syslog-ng#16 0x7f343f239c50 in run_application_hook /home/user/src/syslog-ng/build/../lib/apphook.c:125:11
Direct leak of 8 byte(s) in 1 object(s) allocated from:
#0 0x55a4f86db781 in calloc (/tmp/install/sbin/syslog-ng+0xc5781)
#1 0x7f343f03a941 in g_malloc0 (/usr/lib/libglib-2.0.so.0+0x50941)
syslog-ng#2 0x7f343acb3e27 in _stop_connection_counter_stats_queryable /home/user/src/syslog-ng/build/../modules/afsocket/afsocket-source.c:1027:9
syslog-ng#3 0x7f343acb38bd in afsocket_sd_deinit_method /home/user/src/syslog-ng/build/../modules/afsocket/afsocket-source.c:1073:3
syslog-ng#4 0x7f343f2491d4 in log_pipe_deinit /home/user/src/syslog-ng/build/../lib/logpipe.h:316:25
syslog-ng#5 0x7f343f2490e5 in cfg_tree_stop /home/user/src/syslog-ng/build/../lib/cfg-tree.c:1439:12
syslog-ng#6 0x7f343f23b2d1 in cfg_deinit /home/user/src/syslog-ng/build/../lib/cfg.c:387:10
syslog-ng#7 0x7f343f271ec2 in main_loop_exit_finish /home/user/src/syslog-ng/build/../lib/mainloop.c:427:3
syslog-ng#8 0x7f343f273162 in _consume_action /home/user/src/syslog-ng/build/../lib/mainloop-worker.c:240:3
syslog-ng#9 0x7f343f27321f in _invoke_sync_call_actions /home/user/src/syslog-ng/build/../lib/mainloop-worker.c:250:7
syslog-ng#9 0x7f343f27321f in _invoke_sync_call_actions /home/user/src/syslog-ng/build/../lib/mainloop-worker.c:250:7
syslog-ng#10 0x7f343f2736ba in main_loop_worker_sync_call /home/user/src/syslog-ng/build/../lib/mainloop-worker.c:396:7
syslog-ng#11 0x7f343f2709d1 in main_loop_reload_config_commence /home/user/src/syslog-ng/build/../lib/mainloop.c:340:3
syslog-ng#12 0x7f343f275852 in control_connection_reload /home/user/src/syslog-ng/build/../lib/mainloop-control.c:167:3
syslog-ng#13 0x7f343f290c80 in control_connection_io_input /home/user/src/syslog-ng/build/../lib/control/control-server.c:378:3
syslog-ng#14 0x7f343efd4b2c (/usr/lib/libivykis.so.0+0x6b2c)
syslog-ng#15 0x7f343f239c50 in run_application_hook /home/user/src/syslog-ng/build/../lib/apphook.c:125:11
Signed-off-by: Kokan <kokaipeter@gmail.com>
szemere
pushed a commit
that referenced
this pull request
Dec 15, 2020
This commit accumulates synthetic messages in an array, and forwards them after releasing the GroupingBy parser's lock. It is NOT allowed to call log_pipe_forward_msg() when locks are held, because it would slow down the entire pipeline, and it could also result in deadlock, for example: 1. A new message is produced in syslog-ng's worker pool, which goes through the parser, so it acquires GroupingBy::lock, and starts processing the message (continues in step 3). 2. The main thread schedules grouping_by_timer_tick(), which needs to hold GroupingBy::lock, so it blocks and waits for the other thread. 3. A new synthetic message is emitted in the worker thread (step 1). stateful_parser_emit_synthetic() is called in grouping_by_expire_entry(), which is protected by GroupingBy::lock, The message goes through the pipeline with this lock held, reaches a destination, where main_loop_call(wait=TRUE) is called. Worker waits for the main loop call to be scheduled and finished, main thread waits for GroupingBy::lock, which will never be released. A deadlock situation requires at least 2 locks: - one was GroupingBy::lock; - the other was main_task_lock and its GCond in main_loop_call(). Main thread: ``` #0 syscall () #1 g_mutex_lock_slowpath () #2 _grouping_by_timer_tick () #3 grouping_by_timer_tick () #4 iv_run_timers () #5 iv_main () #6 main_loop_run () #7 main () ``` Worker thread: ``` #0 syscall () #1 g_cond_wait () #2 main_loop_call () #3 ml_batched_timer_postpone () #4 log_writer_queue () #5 log_dest_driver_queue_method () #6 log_multiplexer_queue () #7 log_multiplexer_queue () #8 log_filter_pipe_queue () #9 stateful_parser_emit_synthetic () #10 grouping_by_expire_entry () #11 timer_wheel_set_time () #12 grouping_by_set_time () #13 grouping_by_process () ... ``` Important note: The unlock trick wouldn't have worked in the timer expire callback case, as we would be iterating through a non-locked data structure (TimerWheel) that could be modified by other threads. This commit is based on a00164c, which is a similar deadlock fix for PatternDB. Signed-off-by: László Várady <laszlo.varady@protonmail.com>
Closed
sparlane
added a commit
to sparlane/syslog-ng
that referenced
this pull request
Apr 29, 2022
log_proto_server_reset_error blindly attempts to deref the pointer, which results in a crash when it is NULL. Since this is trying to reset errors, if the object is NULL it definitely wont have an errors to clear. Crash is seen as: Thread 1 (Thread 0x77fa5020 (LWP 405)): #0 0x0ff59618 in log_proto_server_reset_error (s=0x0) at ./lib/logproto/logproto-server.h:164 syslog-ng#1 log_reader_work_finished (s=0x10053800) at lib/logreader.c:426 syslog-ng#2 0x0fc75b98 in iv_fd_poll_and_run (st=st@entry=0x100270a0, abs=<optimized out>) at iv_fd.c:219 syslog-ng#3 0x0fc770b4 in iv_main () at iv_main_posix.c:112 syslog-ng#4 0x0ff5e060 in main_loop_run (self=0xfff629c <main_loop>) at lib/mainloop.c:657 syslog-ng#5 0x10001264 in main (argc=<optimized out>, argv=<optimized out>) at syslog-ng/main.c:316 Change-Id: I8d63a342e74527e6f056afc44a60de05803d1d6e Reviewed-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Signed-off-by: Anthony Lineham <anthony.lineham@alliedtelesis.co.nz>
HofiOne
pushed a commit
to HofiOne/syslog-ng
that referenced
this pull request
Jun 10, 2024
docker/ci: fix Alpine snapshot package versioning
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Since the default version was switched to 3.6 earlier, bump the @Version
of the default config and the in the func-tests too.
Signed-off-by: Gergely Nagy algernon@balabit.hu