[Dovecot] Segfault in dovecot-lda 2.0.5
Hi,
I've recently started using dovecot-lda to handle my local maildir deliveries from Exim. Unfortunately I've been experiencing the occasional failed delivery - My Exim logs were showing:
Child process of virtual_dovecot_delivery transport (running
command "/usr/local/libexec/dovecot/dovecot-lda -d $local_part@$domain -f $sender_address -a $original_local_part@$original_domain") was terminated by signal 11 (Segmentation fault)
I originally thought a specific mail was causing the crash, but calling the command myself delivered the mail without any issues. However, I found if I ran a script that delivered the same mail 10 times, I could reliably cause multiple segfaults all with the following backtrace:
#0 0x005581e9 in maildir_sync_set_new_msgs_count (ctx=0x0, count=1) at maildir-sync-index.c:50 No locals. #1 0x0055f3c0 in maildir_save_move_files_to_newcur (_ctx=0x85afe60) at maildir-save.c:857 files = {arr = {buffer = 0x8555a08, element_size = 4}, v = 0x8555a08, v_modifiable = 0x8555a08} mfp = 0x8555a2c new_changed = false mf = 0x85affa8 newdir = true #2 maildir_transaction_save_commit_pre (_ctx=0x85afe60) at maildir-save.c:951 _data_stack_cur_id = 3 ctx = <value optimized out> _t = 0x85af500 sync_flags = <value optimized out> ret = <value optimized out> __FUNCTION__ = "maildir_transaction_save_commit_pre" #3 0x00555e8f in index_transaction_index_commit (index_trans=0x85af540, result_r=0xbfd78510) at index-transaction.c:27 it = 0x85af500 ibox = 0x858a108 ret = <value optimized out> __FUNCTION__ = "index_transaction_index_commit" #4 0x00595c11 in mail_index_transaction_commit_full (_t=0xbfd78528, result_r=0xbfd78510) at mail-index-transaction.c:237 t = 0x85af540 index = 0x85835e0 index_undeleted = false #5 0x00556284 in index_transaction_commit (t=0x85af500, changes_r=0xbfd785a8) at index-transaction.c:123 box = 0x8589ef0 itrans = 0x0 result = {log_file_seq = 0, log_file_offset = 0, commit_size = 0, ignored_modseq_changes = 0} ret = <value optimized out> #6 0x00533122 in mailbox_transaction_commit_get_changes (_t=0xbfd785d4, changes_r=0xbfd785a8) at mail-storage.c:1215 _data_stack_cur_id = 2 t = 0x85af500 ret = <value optimized out> #7 0x00a7eaef in mail_deliver_save (ctx=0xbfd786f8, mailbox=0x804ab55 "INBOX", flags=0, keywords=0x0, storage_r=0xbfd78794) at mail-deliver.c:219 open_ctx = {user = 0x85781b0, lda_mailbox_autocreate = false, lda_mailbox_autosubscribe = false} box = 0x8589ef0 trans_flags = <value optimized out> t = 0x0 save_ctx = 0x0 kw = 0x0 error = MAIL_ERROR_NONE mailbox_name = 0x804ab55 "INBOX" errstr = <value optimized out> changes = {pool = 0x85b5998, uid_validity = 0, saved_uids = {arr = {buffer = 0x85b59a8, element_size = 8}, v = 0x85b59a8, v_modifiable = 0x85b59a8}, ignored_modseq_changes = 0} range = <value optimized out> ret = 0 __FUNCTION__ = "mail_deliver_save" #8 0x00a7edf8 in mail_deliver (ctx=0xbfd786f8, storage_r=0xbfd78794) at mail-deliver.c:289 ret = -1 #9 0x0804a124 in main (argc=139843496, argv=0x855d1b8) at main.c:429 set_roots = {0x804b390, 0x0} ctx = {pool = 0x855d7a8, set = 0x855ef48, dup_ctx = 0x0, session_id = 0x0, src_mail = 0x85856e0, src_envelope_sender = 0x855d7b8 "sean@cardus.com", dest_user = 0x85781b0, dest_addr = 0x855d228 "sean@cardus.com", final_dest_addr = 0x855d228 "sean@cardus.com", dest_mailbox_name = 0x804ab55 "INBOX", dest_mail = 0x0, tried_default_save = true, saved_mail = false, save_dest_mail = false} service_flags = 1027 user = 0x1
errstr = <value optimized out> path = 0x0 storage_service = 0x855e3b0 service_user = 0x855ec70 service_input = {module = 0x804ab3c "lda", service = 0x804ab3c "lda", username = 0x855d202 "sean@cardus.com", local_ip = {family = 0, u = {ip6 = {in6_u = { u6_addr8 = '\000'Any ideas?
Thanks in advance, Sean
On 10/10/2010 16:30, Sean Cardus wrote:
I originally thought a specific mail was causing the crash, but calling the command myself delivered the mail without any issues.
However, I found if I ran a script that delivered the same mail 10 times, I could reliably cause multiple segfaults all with the following backtrace:#0 0x005581e9 in maildir_sync_set_new_msgs_count (ctx=0x0, count=1) at maildir-sync-index.c:50
Apologies for the noise - It appears the patch at http://hg.dovecot.org/dovecot-2.0/rev/e2f9baa436f2 has fixed the problem!
Sean
Op 10-10-2010 17:30, Sean Cardus schreef:
Hi,
I've recently started using dovecot-lda to handle my local maildir deliveries from Exim. Unfortunately I've been experiencing the occasional failed delivery - My Exim logs were showing:
Child process of virtual_dovecot_delivery transport (running
command "/usr/local/libexec/dovecot/dovecot-lda -d $local_part@$domain -f $sender_address -a $original_local_part@$original_domain") was terminated by signal 11 (Segmentation fault)
I originally thought a specific mail was causing the crash, but calling the command myself delivered the mail without any issues.
However, I found if I ran a script that delivered the same mail 10 times, I could reliably cause multiple segfaults all with the following backtrace:#0 0x005581e9 in maildir_sync_set_new_msgs_count (ctx=0x0, count=1) at maildir-sync-index.c:50 No locals. #1 0x0055f3c0 in maildir_save_move_files_to_newcur (_ctx=0x85afe60) at maildir-save.c:857 files = {arr = {buffer = 0x8555a08, element_size = 4}, v = 0x8555a08, v_modifiable = 0x8555a08} mfp = 0x8555a2c new_changed = false mf = 0x85affa8 newdir = true [...]
Any ideas?
Yes, this was reported several times already. First thread:
http://www.dovecot.org/list/dovecot/2010-October/053475.html
Apply this change:
http://hg.dovecot.org/dovecot-2.0/rev/e2f9baa436f2
That should fix it.
Regards,
Stephan.
participants (2)
-
Sean Cardus
-
Stephan Bosch