[Dovecot] Some crashes with 1.1.1
I'm getting some cra shes with 1.1.1, but even with these start options:
#!/bin/sh mkdir /var/core chmod 1777 /var/core echo /var/core/%p > /proc/sys/kernel/core_pattern ulimit -n 65535 ulimit -c unlimited /usr/local/sbin/dovecot -F
I'm not getting any corefiles - why? It definitely used to work :)
Jun 22 00:05:53 postamt dovecot: imap-login: *** glibc detected *** imap-login: free(): invalid pointer: 0x08cbd600 *** Jun 22 00:05:53 postamt dovecot: child 28879 (login) killed with signal 6
Jun 22 00:29:58 postamt dovecot: imap-login: *** glibc detected *** imap-login: malloc(): memory corruption: 0x09db3d88 *** Jun 22 00:29:58 postamt dovecot: child 28883 (login) killed with signal 6
Jun 22 00:47:07 postamt dovecot: imap-login: *** glibc detected *** imap-login: free(): invalid pointer: 0x090693d8 *** Jun 22 00:47:07 postamt dovecot: child 28876 (login) killed with signal 6
Jun 22 02:47:27 postamt dovecot: child 28885 (login) killed with signal 11
Jun 22 10:06:19 postamt dovecot: imap-login: *** glibc detected *** imap-login: malloc(): memory corruption: 0x091ee0a0 *** Jun 22 10:06:19 postamt dovecot: child 28887 (login) killed with signal 6
Jun 22 15:20:21 postamt dovecot: imap-login: *** glibc detected *** imap-login: malloc(): memory corruption: 0x09f84718 *** Jun 22 15:20:21 postamt dovecot: child 12530 (login) killed with signal 6
-- Ralf Hildebrandt (Ralf.Hildebrandt@charite.de) snickebo@charite.de Postfix - Einrichtung, Betrieb und Wartung Tel. +49 (0)30-450 570-155 http://www.arschkrebs.de All the features in the world only do more to make me believe that Sendmail is hopelessly baroque, like some huge construction with chrome and tailfins sticking out in every direction -- like some monstrocity sculpted by people not the least bit interested in correctness.
On Jun 22, 2008, at 5:38 PM, Ralf Hildebrandt wrote:
I'm getting some cra shes with 1.1.1, but even with these start
options:#!/bin/sh mkdir /var/core chmod 1777 /var/core echo /var/core/%p > /proc/sys/kernel/core_pattern ulimit -n 65535 ulimit -c unlimited /usr/local/sbin/dovecot -F
I'm not getting any corefiles - why? It definitely used to work :)
Also from login processes? One problem with it is that it calls
setuid() after exec(), so kernel thinks it's a setuid-program and
doesn't dump its core. But I don't know if changing core_pattern is
supposed to change that behavior. The other problem is that it's
chrooted, but I guess that doesn't matter then.
Jun 22 00:05:53 postamt dovecot: imap-login: *** glibc detected ***
imap-login: free(): invalid pointer: 0x08cbd600 ***
Well, these really aren't good and there's a good chance that cores
won't help finding out the cause. The best way would be to run via
valgrind:
login_executable = /usr/bin/valgrind /usr/local/libexec/dovecot/ imap-login
I don't really have any good guesses as to why these could be
happening, but could you post your dovecot -n output? Maybe there are
some less common settings..
- Timo Sirainen <tss@iki.fi>:
Well, these really aren't good and there's a good chance that cores won't help finding out the cause. The best way would be to run via valgrind:
login_executable = /usr/bin/valgrind /usr/local/libexec/dovecot/ imap-login
I can try that.
I don't really have any good guesses as to why these could be happening, but could you post your dovecot -n output? Maybe there are some less common settings..
attached
-- Ralf Hildebrandt (Ralf.Hildebrandt@charite.de) snickebo@charite.de Postfix - Einrichtung, Betrieb und Wartung Tel. +49 (0)30-450 570-155 http://www.arschkrebs.de "Looking at the proliferation of personal web pages on the net, it looks like very soon everyone on earth will have 15 Megabytes of fame." -MG Siriam
On Jun 22, 2008, at 6:11 PM, Ralf Hildebrandt wrote:
* Timo Sirainen <tss@iki.fi>:
Well, these really aren't good and there's a good chance that cores won't help finding out the cause. The best way would be to run via valgrind:
login_executable = /usr/bin/valgrind /usr/local/libexec/dovecot/ imap-login
I can try that.
I don't really have any good guesses as to why these could be happening, but could you post your dovecot -n output? Maybe there are some less common settings..
attached
Didn't seem to have anything special. You could also try if the patch below changes anything. Although I haven't heard other people getting heap corruption in v1.1, so it shouldn't be that common problem.. diff -r 65c19e970618 src/login-common/main.c --- a/src/login-common/main.c Sun Jun 22 14:02:54 2008 +0300 +++ b/src/login-common/main.c Sun Jun 22 19:37:45 2008 +0300 @@ -407,8 +407,8 @@ processes pretty safe to reuse for new connections since the attacker won't be able to find anything interesting from the memory. */ - default_pool = system_clean_pool; - data_stack_set_clean_after_pop(TRUE); + /*default_pool = system_clean_pool; + data_stack_set_clean_after_pop(TRUE);*/ /* NOTE: we start rooted, so keep the code minimal until restrict_access_by_env() is called */
participants (2)
-
Ralf Hildebrandt
-
Timo Sirainen