I'm seeing a lot of these: May 29 18:59:48 fold err: POLLERR occurred while idle; fd='25' [...] May 29 19:04:13 fold err: EOF occurred while idle; fd='25' (It also just coredumped, but this led to an autoreboot and the log was not preserved across the reboot :( ) -- NULL && (void)
On 29 May 2011, nix@esperi.org.uk outgrape:
(It also just coredumped, but this led to an autoreboot and the log was not preserved across the reboot :( )
I have several hundred coredumps on another system, all with the same backtrace: #0 0x00007f3ad7c71aca in affile_dd_open_writer () from /usr/lib/syslog-ng/libaffile.so #1 0x00007f3ad7c73d16 in affile_dd_queue () from /usr/lib/syslog-ng/libaffile.so #2 0x00007f3ad9725850 in log_dest_group_queue () from /usr/lib/libsyslog-ng.so.0 #3 0x00007f3ad972b8b5 in log_multiplexer_queue () from /usr/lib/libsyslog-ng.so.0 #4 0x00007f3ad972b924 in log_multiplexer_queue () from /usr/lib/libsyslog-ng.so.0 #5 0x00007f3ad972b8b5 in log_multiplexer_queue () from /usr/lib/libsyslog-ng.so.0 #6 0x00007f3ad973fc87 in log_source_group_queue () from /usr/lib/libsyslog-ng.so.0 #7 0x00007f3ad9734d37 in log_source_queue () from /usr/lib/libsyslog-ng.so.0 #8 0x00007f3ad973310e in log_reader_work_perform () from /usr/lib/libsyslog-ng.so.0 #9 0x00007f3ad9733231 in log_reader_io_process_input () from /usr/lib/libsyslog-ng.so.0 #10 0x00007f3ad9752e04 in iv_main () from /usr/lib/libsyslog-ng.so.0 #11 0x00007f3ad973a40b in main_loop_run () from /usr/lib/libsyslog-ng.so.0 #12 0x00000000004016e9 in main () I have reverted to 3.0.x (which I was previously running) for now. -- NULL && (void)
On Sun, 2011-05-29 at 22:02 +0100, Nix wrote:
On 29 May 2011, nix@esperi.org.uk outgrape:
(It also just coredumped, but this led to an autoreboot and the log was not preserved across the reboot :( )
I have several hundred coredumps on another system, all with the same backtrace:
#0 0x00007f3ad7c71aca in affile_dd_open_writer () from /usr/lib/syslog-ng/libaffile.so #1 0x00007f3ad7c73d16 in affile_dd_queue () from /usr/lib/syslog-ng/libaffile.so #2 0x00007f3ad9725850 in log_dest_group_queue () from /usr/lib/libsyslog-ng.so.0 #3 0x00007f3ad972b8b5 in log_multiplexer_queue () from /usr/lib/libsyslog-ng.so.0 #4 0x00007f3ad972b924 in log_multiplexer_queue () from /usr/lib/libsyslog-ng.so.0 #5 0x00007f3ad972b8b5 in log_multiplexer_queue () from /usr/lib/libsyslog-ng.so.0 #6 0x00007f3ad973fc87 in log_source_group_queue () from /usr/lib/libsyslog-ng.so.0 #7 0x00007f3ad9734d37 in log_source_queue () from /usr/lib/libsyslog-ng.so.0 #8 0x00007f3ad973310e in log_reader_work_perform () from /usr/lib/libsyslog-ng.so.0 #9 0x00007f3ad9733231 in log_reader_io_process_input () from /usr/lib/libsyslog-ng.so.0 #10 0x00007f3ad9752e04 in iv_main () from /usr/lib/libsyslog-ng.so.0 #11 0x00007f3ad973a40b in main_loop_run () from /usr/lib/libsyslog-ng.so.0 #12 0x00000000004016e9 in main ()
I have reverted to 3.0.x (which I was previously running) for now.
Hmm... sorry to have caused disruptions. The problem with this backtrace, is that it lacks a lot of information, debug symbols would be needed to properly analyze it. Can you help in that regard? Thanks. -- Bazsi
On 4 Jun 2011, Balazs Scheidler told this:
Hmm... sorry to have caused disruptions. The problem with this backtrace, is that it lacks a lot of information, debug symbols would be needed to properly analyze it. Can you help in that regard?
Will do. Sorry, there are problems with the so-alpha-it-bleeds GDB I'm using here. I'll give you better info an a few days. -- NULL && (void)
On Sat, 2011-06-04 at 23:51 +0100, Nix wrote:
On 4 Jun 2011, Balazs Scheidler told this:
Hmm... sorry to have caused disruptions. The problem with this backtrace, is that it lacks a lot of information, debug symbols would be needed to properly analyze it. Can you help in that regard?
Will do. Sorry, there are problems with the so-alpha-it-bleeds GDB I'm using here. I'll give you better info an a few days.
Thanks for your efforts, waiting for the inforamtion. :) -- Bazsi
Hi, I think we got this problem fixed in this patch: commit 33de4c3b4de40493544cee3cf3de8bd0e4d76d34 Author: Balazs Scheidler <bazsi@balabit.hu> Date: Sun Jun 12 10:38:18 2011 +0200 [affile] fixed a reference counting problemin the file-open code A reference was leaked whenever the same file got opened from multiple threads at about the same time, causing use-after-free problems. This patch also fixes a possible NULL deref that could happen if the writer couldn't be initialized. Signed-off-by: Viktor Juhasz <jviktor@balabit.hu> Signed-off-by: Balazs Scheidler <bazsi@balabit.hu> So it'd make sense to retry with a recent git snapshot. I'll try to create another release in the not too distant future. Thanks for the report. On Sat, 2011-06-04 at 23:51 +0100, Nix wrote:
On 4 Jun 2011, Balazs Scheidler told this:
Hmm... sorry to have caused disruptions. The problem with this backtrace, is that it lacks a lot of information, debug symbols would be needed to properly analyze it. Can you help in that regard?
Will do. Sorry, there are problems with the so-alpha-it-bleeds GDB I'm using here. I'll give you better info an a few days.
-- Bazsi
On 12 Jun 2011, Balazs Scheidler verbalised:
Hi,
I think we got this problem fixed in this patch:
Confirmed (finally I have time to look at syslog-ng).
On Tue, 2011-07-19 at 11:52 +0100, Nix wrote:
On 12 Jun 2011, Balazs Scheidler verbalised:
Hi,
I think we got this problem fixed in this patch:
Confirmed (finally I have time to look at syslog-ng).
Great, thanks for reporting back. -- Bazsi
participants (2)
-
Balazs Scheidler
-
Nix