January 2007 - dovecot

[Dovecot] dovecotpw/sql crypt scheme core dump with rc17-19
by Andre Dalle 24 Jan '07

24 Jan '07

Platform is Solaris 8 on a 280R. I'm setting up a new installation of Dovecot, and I ran into some troubles that I've partly tracked down. I setup auth/user dbs with LDAP initially, and things worked well. Passwords are stored as "{crypt}zxcv..." in LDAP. I setup SQL, and began getting this: dovecot: Jan 23 16:37:47 Error: child 8718 (auth-worker) killed with signal 11 Passwords are stored as "zxcv..." in SQL. I changed default_pass_scheme from CRYPT to PLAIN, and no troubles. I then tried dovecotpw: dovecotpw -p 123 -s PLAIN-MD5 {PLAIN-MD5}202cb962ac59075b964b07152d234b70 dovecotpw -p 123 -s CRYPT Segmentation Fault (core dumped) So I then tried this: truss dovecotpw -p 123 -s crypt open64("/dev/urandom", O_RDONLY) = 3 read(3, "F80E82 c", 4) = 4 fcntl(3, F_GETFD, 0x00000000) = 0 fcntl(3, F_SETFD, 0x00000001) = 0 open64("/usr/local/lib/dovecot/auth/password", O_RDONLY|O_NDELAY) Err#2 ENOENT read(3, "8382", 2) = 2 Incurred fault #6, FLTBOUNDS %pc = 0xFF0E3304 siginfo: SIGSEGV SEGV_MAPERR addr=0x00000005 Received signal #11, SIGSEGV [default] siginfo: SIGSEGV SEGV_MAPERR addr=0x00000005 *** process killed *** Suffice to say, there is no "/usr/local/lib/dovecot/auth/password". I'm guessing that the dovecot auth-worker process is failing for the same reason but I don't know how to debug that one. I have core files for the dovecotpw process. -- André Dalle Systems Administrator NCF Internet

2 2

[Dovecot] All index file bugs fixed(?)
by Timo Sirainen 23 Jan '07

23 Jan '07

I fixed the nightly snapshot building. Now it's again being built every night (if there have been any changes). Please test it before I'll make rc18 release. http://dovecot.org/nightly/dovecot-latest.tar.gz I've also been stress testing a lot with http://dovecot.org/tools/imaptest.c. Last night I ran it for 9 hours with mmap_disable=yes and it picked up only one race condition, which is also now fixed. I'll run the next night with mmap_disable=no and see if anything shows up with it. I'm hopeful that there aren't any important index file related bugs left anymore. This is actually the first time in Dovecot's existence that I'm really not aware of *any* bugs in index file handling. v1.0 release should be really near now, unless you figure out new problems :) Here's a list of bugs that should have been fixed since rc17: "Corrupted index file dovecot.index: Record field keywords points outside record size (12 < 12+2)" "Corrupted Index file dovecot.index: Record [1].uid=0" Crashes - http://dovecot.org/list/dovecot-cvs/2007-January/007395.html "open() failed with file dovecot.index.log: No such file or directory" - http://dovecot.org/list/dovecot-cvs/2007-January/007405.html "Fixed index file dovecot.index: seen_messages_count 8 -> 9" "View synchronization from transaction log failed: Deleted counter wrong" etc. - http://dovecot.org/list/dovecot-cvs/2007-January/007403.html Corrupted transaction log errors: - http://dovecot.org/list/dovecot-cvs/2007-January/007389.html - http://dovecot.org/list/dovecot-cvs/2007-January/007393.html Corrupted view errors: - http://dovecot.org/list/dovecot-cvs/2007-January/007366.html - http://dovecot.org/list/dovecot-cvs/2007-January/007368.html Corrupted cache file errors: - http://dovecot.org/list/dovecot-cvs/2007-January/007364.html Fixed a lot of crashes whenever something went wrong (I deleted index files every 3 seconds until it crashed, last run went for several minutes before I stopped trying): - http://dovecot.org/list/dovecot-cvs/2007-January/007397.html

5 15

[Dovecot] 1.0.rc18 released
by Timo Sirainen 23 Jan '07

23 Jan '07

http://dovecot.org/releases/dovecot-1.0.rc18.tar.gz http://dovecot.org/releases/dovecot-1.0.rc18.tar.gz.sig I think we're quite near v1.0 now. * ACL plugin + Maildir: Moved dovecot-acl file from control directory to maildir. To prevent accidents caused by this change, Dovecot kills itself if it finds dovecot-acl file from the control directory. * When opening a maildir, check if tmp/'s atime is over 8h old. If it is, delete files in it with ctime older than 36h. However if atime - ctime > 36h, it means that there's nothing to be deleted and the scanning isn't done. We update atime ourself if filesystem is mounted with noatime. * base_dir doesn't need to be group-readable, don't force it. * mail_read_mmaped setting is deprecated and possibly broken. It's now removed from dovecot-example.conf, but it still works for now. * Removed also umask setting from dovecot-example.conf since currently it doesn't do what it's supposed to. + Authentication cache caches now also userdb data. + Added mail_log plugin to log various mail operations. Currently it logs mail copies, deletions, expunges and mailbox deletions. - dict quota: messages=n parameter actually changed storage limit. - A lot of fixes to handling index files. This should fix almost all of the problems ever reported. - LDAP: auth_bind=yes was more or less broken. - Saved mails and dovecot-keywords file didn't set the group from dovecot-shared file. - Fixed potential assert-crash while searching messages - Fixed some crashes with invalid X-UID headers in mboxes - If you didn't have a namespace with empty prefix, giving STATUS command for a non-existing namespace caused the connection to give "NO Unknown namespace" errors for all the future commands.

4 6

[Dovecot] Dovecot-1.0.rc17 on Solaris10 = assert_error
by Kevin Richter 23 Jan '07

23 Jan '07

Hi, i have a problem with dovecot rc17 on Solaris10. I have compiled Dovecot and openSSL in 32bit: CFLAGS="-I/opt/openssl-0.9.7l/include" LDFLAGS="-L/opt/openssl-0.9.7l/lib -R/opt/openssl-0.9.7l/lib" ./configure --prefix=/opt/dovecot-1.0.rc17 --with-ssl=openssl --with-ssldir=/opt/openssl-0.9.7l --enable-shared --disable-ipv6 --with-file-offset-size=32 make and make install ran without errors or warnings. The /opt/dovecot-1.0.rc17/var/log says: dovecot: Jan 16 11:20:05 Error: IMAP(kric0999): file lib.c: line 13 (nearest_power): assertion failed: (num <= ((size_t)1 << (BITS_IN_SIZE _T-1))) The error occures whenever i start my mail client and the authorization is successful. Cam anyone help me? Regards, Kevin Richter

2 6

[Dovecot] Current state of mulit-server NFS setups
by Justin McAleer 23 Jan '07

23 Jan '07

Hi list, I've been perusing the list archives, but I couldn't find a recent, definitive answer to what I'd like to know. I'm looking to set up a new batch of servers to migrate ~160,000 accounts, spread across several domains, onto (from CommuniGate). Last year I worked with the earlier beta/RC versions of Dovecot, so I'm generally familiar with the setup required. However, a lot of development has been done since I was last dealing with it. So, what I'd like to end up with is a load balanced group of servers providing pop and imap access with quota support using an NFS filesystem. We have a new Netapp device (forget the actual model) that can do SAN/iSCSI/NFS. We tried using various clustered filesystems previously, but the performance/stability issues were a nightmare, so I've fallen back to NFS (or perhaps SAN if users can't hit any server in the group). So, what does my dovecot setup need to be? I've seen some people say they disable indexing in the LDA, but allow pop/imap to index. Some say the attribute cache needs to be disabled, which I'd think would be worse than just disabling indexes. Some say indexes should be kept on local disk (dedicated server for users), and others seem to report problems even when doing so. Is there a general consensus? I was also wondering if dovecot's sanity checking on the indexes would allow each server in a balanced group to keep indexes on local disk (so if the maildir has changed on another server, dovecot rebuilds it). If so, does that take away all performance gains from having indexes at all? Thanks for any info you all can provide.

2 1

[Dovecot] Maildir Quota backend
by Ioannis Zachariadis 23 Jan '07

23 Jan '07

Hi, I've noticed that in the maildir backend for the quota plugin the quota limit is being read in kilobytes. In the quota-maildir.c at maildir_quota_init() , line 550 (rc18) if (strncmp(*args, "storage=", 8) == 0) { size = strtoull(*args + 8, NULL, 10) * 1024; if (size != 0) root->message_bytes_limit = size; root->master_message_limits = TRUE; } The value that's being read from storage is supposed to be in kB, at least that's what I understand from the wiki, and then the value of size is stored in bytes. thanks -- giannis.

2 1

[Dovecot] LDAP authentication stops working...
by Adrian Close 23 Jan '07

23 Jan '07

Hi all, I'm running dovecot-1.0.rc17 on OpenBSD 3.9, using userdb and passdb methods of "ldap" (SSL on 636/tcp) in addition to "passwd". Occasionally (generally after a few hours of operation, but not always), LDAP-based logins stop working (e.g. hang/timeout after POP3 PASS command). Accounts with local passwords (as opposed to accounts with a password field of "x") still work fine at this point. 'tcpdump' shows the persisient 636/tcp connection still with bidirectional traffic (PUSH/ACKs and ACKs like normal). After restarting Dovecot, things are fine until next time. I guess I could try and get login_ldap working with bsdauth, but if there's an issue in the Dovecot LDAP support I'd like to see it fixed. Any ideas? Thanks, Adrian Close P.S. I note the LDAP timeout logging patch but it won't apply easily to rc17 and I'm hoping the issue has progressed since then anyway...

7 14

[Dovecot] X-UID: 4085788205 != 2147483647
by Stephen Gowing 23 Jan '07

23 Jan '07

Hi, Dovecot (at least versions 0.99.10.4, 0.99.11, 1.0rc15, rc17 and rc18) allows message Unique Identifiers to be an unsigned 32 bit value, so something like "X-UID: 4085788205" is allowed. Some IMAP clients I've tested (Thunderbird 1.5.0.9, SquirrelMail 1.4.9a/PHP 4.3.2) appear to use signed 32 bit identifiers and request the above message as UID 2147483647 (max signed 32 bit integer). Evolution 2.0.2 fetches 4085788205 without complaint. Is there any chance Dovecot could decide to invalidate and renumber a folder when identifiers greater than 2147483647 are found? sbg

2 1

[Dovecot] mysteriously disappearing RECENT flag
by Vadim Zeitlin 22 Jan '07

22 Jan '07

Hello, I've decided to switch from UW-IMAP to Dovecot and so installed 1.0.rc15-2 on a Debian/Etch system and started using it with the existing MBOX folders. My configuration is pretty much the default one, in particular I didn't touch to any of mbox_xxx options. The (hoped for) performance improvements were indeed quite noticeable and mostly it works just fine but I seem to have one strange problem with the recent messages: Dovecot seems to reset the RECENT flag on the message right *before* the mailbox is opened. So even if a message was just copied to a folder (e.g. by email client filtering), it doesn't have RECENT flag when the mailbox is opened. Here is an example of what I see: % telnet localhost 143 Trying 127.0.0.1... Connected to localhost. Escape character is '^]'. * OK [CAPABILITY IMAP4rev1 SASL-IR SORT THREAD=REFERENCES MULTIAPPEND UNSELECT LITERAL+ IDLE CHILDREN NAMESPACE LOGIN-REFERRALS STARTTLS AUTH=PLAIN] Dovecot ready. 1 login user password 1 OK Logged in. 2 status test (messages recent unseen) * STATUS "test" (MESSAGES 41 RECENT 0 UNSEEN 0) 2 OK Status completed. [--- switch to another window and copy a message to the test folder ---] 3 status test (messages recent unseen) * STATUS "test" (MESSAGES 42 RECENT 1 UNSEEN 0) 3 OK Status completed. 4 select test * FLAGS (\Answered \Flagged \Deleted \Seen \Draft) * OK [PERMANENTFLAGS (\Answered \Flagged \Deleted \Seen \Draft \*)] Flags permitted. * 42 EXISTS * 0 RECENT * OK [UIDVALIDITY 1048987099] UIDs valid * OK [UIDNEXT 407] Predicted next UID 4 OK [READ-WRITE] Select completed. 5 status test (messages recent unseen) * STATUS "test" (MESSAGES 42 RECENT 0 UNSEEN 0) 5 OK Status completed. 6 fetch 42 flags * 42 FETCH (FLAGS (\Seen)) 7 logout * BYE Logging out 7 OK Logout completed. Connection closed by foreign host. Why does Dovecot send "0 RECENT" when the mailbox is opened when it had just told that there was a recent message in it? And why is the message effectively old (no \Recent in the FETCH reply)? Notice that if I look at the MBOX file I see "Status: R" until command number 7, it becomes "Status: RO" only when I log out -- but internally Dovecot clearly considers it to be old long before that. Am I doing something wrong? Or is this the expected behaviour and, in this case, is there any way to change it to the more expected one (UW-IMAP-like)? Thanks in advance, VZ

2 3

[Dovecot] RC18 core dumps
by Cassidy B. Larson 22 Jan '07

22 Jan '07

Still seeing a few core dumps via the deliver process and corrupted transaction log files with RC18: Jan 22 10:46:36 zahara deliver(user(a)host.com): Corrupted transaction log file /data/mail/host.com/us/user(a)home.com/dovecot.index.log: start_offset (71680) > file size (71520) Jan 22 10:46:36 zahara deliver(user(a)host.com): seq = 6, rec->uid = 0, first_new_seq = 6, records = 5 Jan 22 10:46:36 zahara kernel: pid 41186 (deliver), uid 127: exited on signal 6 (core dumped) Jan 22 10:50:17 zahara deliver(user(a)host.com): Corrupted transaction log file /data/mail/host.com/us/user(a)host.com/dovecot.index.log: Append with UID 512, but next_uid = 513 However, if the user gets another email, it appears deliver doesnt core dump again. Thoughts? -c

2 1