[Dovecot] doveadm mailbox status 2.0.12 crashes
My gdb won't cooperate and produce a backtrace so I hope this dbx stack trace gives sufficient clues as to why it crashed. It happended when I was testing a doveadm command for a user which a large collection of mboxs:
# dbx doveadm
(dbx) run mailbox status -u user messages \*
... long list ...
igel messages=1
INBOX messages=231
doveadm(user): Panic: Trying to deinit storage before freeing its objects
doveadm(user): Error: Raw backtrace: 0xff1bf0e4 -> 0xff1bf34c -> 0xff2a42a0 -> 0xff29d68c -> 0xff29e494 -> 0xff2a7bc4 -> 0xff2a7e3c -> 0x16894 -> 0x169a0 -> 0x16e18 -> 0x17234 -> 0x17304 -> 0x1c0a0 -> 0x1629c
signal ABRT (Abort) in __lwp_kill at 0xfefcd6ac
0xfefcd6ac: __lwp_kill+0x0008: bcc,a,pt %icc,__lwp_kill+0x18 ! 0xfefcd6bc
Current function is default_fatal_finish
dbx: warning: can't find file "/build/dovecot-2.0.12/src/lib/failures.c"
(dbx) where
[1] __lwp_kill(0x0, 0x6, 0x0, 0x6, 0xffbffeff, 0x0), at 0xfefcd6ac
[2] raise(0x6, 0x0, 0x0, 0xfefacd88, 0xffffffff, 0x6), at 0xfef669ac
[3] abort(0xff1ea980, 0x1, 0x0, 0xee4cc, 0xff0334d8, 0x0), at 0xfef42010
=>[4] default_fatal_finish(type = LOG_TYPE_PANIC, status = 0), line 187 in "failures.c"
[5] default_fatal_handler(ctx = 0xffbff598, format = (nil), args = 0xffbff600), line 201 in "failures.c"
[6] i_panic(format = 0xff30d3b8 "Trying to deinit storage before freeing its objects", ...), line 259 in "failures.c"
[7] mail_storage_unref(_storage = 0x6b9ec), line 379 in "mail-storage.c"
[8] mail_namespace_free(ns = 0x6b9b8), line 52 in "mail-namespace.c"
[9] mail_namespaces_deinit(_namespaces = 0x66efc), line 438 in "mail-namespace.c"
[10] mail_user_deinit_base(user = 0x66ec0), line 28 in "mail-user.c"
[11] mail_user_unref(_user = 0x55450), line 140 in "mail-user.c"
[12] doveadm_mail_next_user(ctx = 0x55438, input = 0x55450, error_r = 0xffbff8c4), line 221 in "doveadm-mail.c"
[13] doveadm_mail_single_user(ctx = 0x55438, argv = 0xffbffc08, username = 0xffbffd1f "user", service_flags = -4196156), line 246 in "doveadm-mail.c"
[14] doveadm_mail_cmd(cmd = 0x544f0, argc = 5, argv = 0xffbffc08), line 416 in "doveadm-mail.c"
[15] doveadm_mail_try_run_multi_word(cmd = 0x544f0, cmdname = 0x2ade0 "status", argc = 6, argv = 0xffbffbf8), line 457 in "doveadm-mail.c"
[16] doveadm_mail_try_run(cmd_name = 0xffbffd0d "mailbox", argc = 6, argv = 0xffbffbf8), line 480 in "doveadm-mail.c"
[17] main(argc = 6, argv = 0xffbffbf8), line 183 in "doveadm.c"
Joseph Tam jtam.home@gmail.com
On Sat, 2011-04-16 at 02:49 -0700, Joseph Tam wrote:
My gdb won't cooperate and produce a backtrace so I hope this dbx stack trace gives sufficient clues as to why it crashed. It happended when I was testing a doveadm command for a user which a large collection of mboxs: .. # dbx doveadm (dbx) run mailbox status -u user messages \* ... long list ... igel messages=1 INBOX messages=231 doveadm(user): Panic: Trying to deinit storage before freeing its objects
A backtrace can't really show why it crashed with this. Something wasn't freed for some reason. Can you still reproduce this? I'd like to know of a way to reproduce this. Maybe you can copy this problematic user's mailboxes to a test user and keep deleting stuff until you see what triggers it?
participants (2)
-
Joseph Tam
-
Timo Sirainen