[Dovecot] rc25: need_space assert, core
Timo,
I see where at least one other person reported this, but here goes. I went from rc24 to rc25 this morning, and I got an assert and core from my own mailbox withing five minutes:
Mar 2 06:52:26 karst dovecot: [ID 107833 mail.error] IMAP(jaearick): file mbox-sync-rewrite.c: line 408: assertion failed: (need_space == (uoff_t)-mails[idx].space) Mar 2 06:52:26 karst dovecot: [ID 107833 mail.error] IMAP(jaearick): Raw backtrace: 0x89720 -> 0x46e50 -> 0x47410 -> 0x426d4 -> 0x42d60 -> 0x438b0 -> 0x43b38 -> 0x2a828 -> 0x2ada4 -> 0x231dc -> 0x23368 -> 0x23450 -> 0x8fe14 -> 0x8f638 -> 0x2cd64 -> 0x1dad8
Unfortunately, I switched compilers at rc24 and rc25. Before I used to compile with gcc -g -O so I could poke at cores. Now I use (Sun Studio 11) cc -O, so no useful information on a gdb traceback:
(gdb) backtrace #0 0xff1c16e0 in _lwp_kill () from /lib/libc.so.1 #1 0xff15ff38 in raise () from /lib/libc.so.1 #2 0xff140160 in abort () from /lib/libc.so.1 #3 0x00089cd4 in i_internal_panic_handler () #4 0x00089728 in i_panic () #5 0x00046e58 in mbox_sync_read_and_move () #6 0x00047418 in mbox_sync_rewrite () #7 0x000426dc in mbox_sync_handle_eof_updates () #8 0x00042d68 in mbox_sync_do () #9 0x000438b8 in mbox_sync () #10 0x00043b40 in mbox_storage_sync_init () #11 0x0002a830 in imap_sync_init () #12 0x0002adac in cmd_sync () #13 0x000231e4 in client_handle_input () #14 0x00023370 in client_handle_input () #15 0x00023458 in _client_input () #16 0x0008fe1c in io_loop_handler_run () #17 0x0008f640 in io_loop_run () #18 0x0002cd6c in main ()
My setup: Solaris 10, homedirs (imap cache files) on a local zfs filesystem, mailboxes NFS mounted from another Sun system (NFS v4) which is also zfs on the mail server. Mbox format, no disk quotas anyplace, no plugins.
I have rolled back to rc24.
Jeff Earickson Colby College
On Fri 02 Mar 2007 at 08:30AM, Jeff A. Earickson wrote:
Timo,
I see where at least one other person reported this, but here goes. I went from rc24 to rc25 this morning, and I got an assert and core from my own mailbox withing five minutes:
Mar 2 06:52:26 karst dovecot: [ID 107833 mail.error] IMAP(jaearick): file mbox-sync-rewrite.c: line 408: assertion failed: (need_space == (uoff_t)-mails[idx].space) Mar 2 06:52:26 karst dovecot: [ID 107833 mail.error] IMAP(jaearick): Raw backtrace: 0x89720 -> 0x46e50 -> 0x47410 -> 0x426d4 -> 0x42d60 -> 0x438b0 -> 0x43b38 -> 0x2a828 -> 0x2ada4 -> 0x231dc -> 0x23368 -> 0x23450 -> 0x8fe14 -> 0x8f638 -> 0x2cd64 -> 0x1dad8
Me too. I deployed rc25 last night, and this morning we've got 100 core files. They look similar:
core 'core.imap.595749.10785' of 595749: imap ff1c4310 _lwp_kill (6, 0, 5, 6, ffffffff, 6) + 8 ff151288 abort (a1f10, 1, 6, ff215c80, c1b2c, 0) + 108 00080bc4 i_internal_panic_handler (982a0, ffbfed88, 0, a, 0, 80b88) + 3c 00080670 i_panic (982a0, 982d0, 198, 98530, 985c8, 0) + 24 0004952c mbox_sync_rewrite (ffbff4f8, e1, 0, 1974534, 0, 0) + 7e4 00044118 mbox_sync_do (ffbff4f8, 688, 0, 1975bff, 78d, ffbff3a0) + 93c 00045ed0 mbox_sync (1, 18, 3, c94c0, 0, ca8f1) + 36c 000462f4 mbox_storage_sync_init (c94c0, 3, 0, 0, 0, b595c) + 54 000723dc mailbox_sync_init (c94c0, 3, 0, 65642e00, 0, 8855c) + c 00030368 imap_sync_init (ca8b8, c94c0, 0, 3, 0, 8855c) + 34 00030904 cmd_sync (c35b4, 3, 0, 91b68, 80808080, 1010101) + 64 000259bc cmd_check (1, c2c68, 0, 8, 2a404, c2c68) + 2c 00029dd4 client_handle_input (c35b4, c38a0, b6180, be198, 7f00, 6000000) + 138 00029d4c client_handle_input (0, 45e85361, b6180, be198, 45e85361, 4000000) + b0 00029f48 _client_input (c3570, c3570, 2c4, 0, 1, 0) + 84 000869a8 io_loop_handler_run (c0d68, 0, 0, ffbffa3c, 4c, 80000000) + 140 0008625c io_loop_run (c0d68, ff212cb0, 1, b615c, bae80, ff215dbc) + 1c 00032334 main (ffbffc3d, b4c00, b61ec, b615c, c1314, ff3a0180) + 3f8 00024d20 _start (0, 0, 0, 0, 0, 0) + 5c
dovecot: Mar 02 10:51:20 Error: IMAP(brittle): file mbox-sync-rewrite.c: line 40 8 (mbox_sync_read_and_move): assertion failed: (need_space == (uoff_t)-mails[idx ].space) dovecot: Mar 02 10:51:20 Error: IMAP(brittle): Raw backtrace: 0x80670 -> 0x4952c -> 0x44118 -> 0x45ed0 -> 0x462f4 -> 0x723dc -> 0x30368 -> 0x30904 -> 0x259bc -> 0x29dd4 -> 0x29d4c -> 0x29f48 -> 0x869a8 -> 0x8625c -> 0x32334 -> 0x24d20 dovecot: Mar 02 10:51:21 Error: child 621975 (imap) killed with signal 6
BTW, Timo, we should try to get symbolic backtraces working in Solaris. I'll try to find a few minutes to figure that out.
-dp-- Daniel Price - Solaris Kernel Engineering - dp@eng.sun.com - blogs.sun.com/dp
Dan Price wrote:
On Fri 02 Mar 2007 at 08:30AM, Jeff A. Earickson wrote:
Timo,
I see where at least one other person reported this, but here goes. I went from rc24 to rc25 this morning, and I got an assert and core from my own mailbox withing five minutes:
Mar 2 06:52:26 karst dovecot: [ID 107833 mail.error] IMAP(jaearick): file mbox-sync-rewrite.c: line 408: assertion failed: (need_space == (uoff_t)-mails[idx].space) Mar 2 06:52:26 karst dovecot: [ID 107833 mail.error] IMAP(jaearick): Raw backtrace: 0x89720 -> 0x46e50 -> 0x47410 -> 0x426d4 -> 0x42d60 -> 0x438b0 -> 0x43b38 -> 0x2a828 -> 0x2ada4 -> 0x231dc -> 0x23368 -> 0x23450 -> 0x8fe14 -> 0x8f638 -> 0x2cd64 -> 0x1dad8
Me too. I deployed rc25 last night, and this morning we've got 100 core files. They look similar: ...
Me too as well. I've gone back to the previous version we were using and the errors have disappeared. We are in "pilot project" mode, so I'd be happy to do some testing/etc to get information on how to fix this.
-- Ethan Sommer UNIX Systems Administrator Gustavus Adolphus College Saint Peter, MN 507-933-7042
On Tue, 2007-03-06 at 09:05 -0600, Ethan Sommer wrote:
Mar 2 06:52:26 karst dovecot: [ID 107833 mail.error] IMAP(jaearick): file mbox-sync-rewrite.c: line 408: assertion failed: (need_space == (uoff_t)-mails[idx].space)
I don't understand how I've managed to break this same fix so many times already. I think this is the 3rd or 4th time I've changed this code: http://dovecot.org/list/dovecot-cvs/2007-March/007925.html
Hopefully that's correct and rc26 will finally fix the mbox problems. :)
participants (4)
-
Dan Price
-
Ethan Sommer
-
Jeff A. Earickson
-
Timo Sirainen