dovecot - Jun 2006 - Help! Uidlist files are gone and won't come back; imap keeps coredumping!

Folks -

   Some time ago as I migrated to dovecot 1.0, I ran into all sorts of problems
with file locks, and I tried various solutions.  Apparently, in the process, I
got a bit careless and removed all my dovecot-uidlist files.  They're gone.

   My understanding, however, is that they should come back, but this
doesn't happen.

   I've been forced to run with index=MEMORY in 1.0 for the entire time
I've been using it, but I would really like to get back to a more standard
configuration.  Yet, anything I do to try and actually use index files, whether
(locally) on /var or (via NFS) in /home, I run into problems.  The log files
say:

dovecot: Jun 01 07:36:36 Info: imap-login: Login: user=<rvvk40>,
method=plain, rip=10.69.75.131, lip=172.16.26.151
dovecot: Jun 01 07:36:36 Info: IMAP(rvvk40): Effective uid=10019173, gid=21
dovecot: Jun 01 07:36:36 Info: IMAP(rvvk40): maildir:
data=/home/mario/.mail/Maildir
dovecot: Jun 01 07:36:36 Info: IMAP(rvvk40): maildir:
root=/home/mario/.mail/Maildir, index=/home/mario/.mail/Maildir, control=,
inboxdovecot: Jun 01 07:36:36 Error: child 21207 (imap) killed with signal 11
dovecot: Jun 01 07:36:36 Error: child 21208 (imap) killed with signal 11
dovecot: Jun 01 07:36:36 Error: child 21211 (imap) killed with signal 11
dovecot: Jun 01 07:36:36 Error: child 21212 (imap) killed with signal 11

and a truss says:

29146:  execve("/home/mario/tools/SunOS/libexec/dovecot/imap",
0x000407D8, 0x00052850)  argc = 1
29146:   argv: imap
...
29146:  write(1, " *   1 2 4 5   F E T C H".., 2091)    = 2091
29146:  write(1, " *   1 2 9 5   F E T C H".., 2059)    = 2059
29146:  write(1, " *   1 3 4 6   F E T C H".., 2056)    = 2056
29146:  poll(0x000AEC90, 3, 9912)                       = 1
29146:  read(0, " 4   U I D   f e t c h  ".., 4037)     = 173
29146: 
open64("/home/mario/.mail/Maildir/.Freescale.ATIC.CMOS65/dovecot-uidlist",
O_RDONLY) Err#2 ENOENT
29146:      Incurred fault #6, FLTBOUNDS  %pc = 0xFF133218
29146:        siginfo: SIGSEGV SEGV_MAPERR addr=0x00000000
29146:      Received signal #11, SIGSEGV [default]      
29146:        siginfo: SIGSEGV SEGV_MAPERR addr=0x00000000
29146:          *** process killed ***                  

Some background info:

dovecot is running only for me, so it runs under my UID, from my home directory.

% tools/SunOS/sbin/dovecot --version
1.0.beta8

% tools/SunOS/sbin/dovecot --build-options          
Build options: ioloop=poll openssl
Passdb: checkpassword pam passwd passwd-file shadow 
Userdb: checkpassword passwd passwd-file static 

% uname -a
SunOS velocity 5.8 Generic_117350-06 sun4u sparc


What do I have to do to recreate the uidlist files?

					Mario

-- 
I don't need a name; my number's just fine. |  
Mario.Nigrovic@freescale.com
It's nobody else's -- just mine, all mine.  |                  
480-413-3578
                                                           Internal Use Only

On Thu, 2006-06-01 at 07:59 -0700, Mario Nigrovic-rvvk40
wrote:
> 29146: 
open64("/home/mario/.mail/Maildir/.Freescale.ATIC.CMOS65/dovecot-uidlist",
O_RDONLY) Err#2 ENOENT
> 29146:      Incurred fault #6, FLTBOUNDS  %pc = 0xFF133218
> 29146:        siginfo: SIGSEGV SEGV_MAPERR addr=0x00000000
The missing uidlist isn't directly the problem. Rather it crashes while
trying to start maildir syncing (or something).

Since this works for most other people, it's difficult to start guessing
why it could crash. If you could get gdb backtrace of the crash it would
help a lot in fixing it. See http://dovecot.org/bugreport.html how to do
it.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 191 bytes
Desc: This is a digitally signed message part
Url :
http://dovecot.org/pipermail/dovecot/attachments/20060601/8f5b8107/attachment.pgp

Timo -

   The only behavior change that I notice is that when I exit the telnet session
the truss also exits.  Previously, the truss hung around until I did the
"pkill dovecot", after exiting the telnet session.  So, that's
weird.  But, hold on!  Now I do seem to have a uidlist file, again!

% ls -l
total 308
drwx--S---   2 rvvk40   lowv        98304 Jun 19 07:44 cur/
-rw-------   1 rvvk40   lowv          260 Jun 19 07:47 dovecot-keywords
-rw-------   1 rvvk40   lowv        65752 Jun 19 07:47 dovecot-uidlist
-rw-------   1 rvvk40   lowv        29560 Jun 11 08:42 dovecot.index
drwx--S---   2 rvvk40   lowv        32768 Jun 19 07:43 new/
-rw-------   1 rvvk40   lowv         1188 May  1 07:01 subscriptions
drwx--S---   2 rvvk40   lowv        61440 Jun 19 07:42 tmp/

				Mario

Timo Sirainen wrote:
> Try if the attached patch changes anything.
> 
> 
> 
> ------------------------------------------------------------------------
> 
> Index: src/lib/file-dotlock.c
> ==================================================================> RCS
file: /var/lib/cvs/dovecot/src/lib/file-dotlock.c,v
> retrieving revision 1.38
> diff -u -r1.38 file-dotlock.c
> --- src/lib/file-dotlock.c	8 Jun 2006 16:42:40 -0000	1.38
> +++ src/lib/file-dotlock.c	19 Jun 2006 12:55:51 -0000
> @@ -664,6 +664,12 @@
>  		}
>  	}
>  
> +	if (dotlock->fd != -1) {
> +		if (close(dotlock->fd) < 0)
> +			i_error("close(%s) failed: %m", dotlock->path);
> +		dotlock->fd = -1;
> +	}
> +
>  	if (rename(lock_path, dotlock->path) < 0) {
>  		i_error("rename(%s, %s) failed: %m", lock_path,
dotlock->path);
>  		file_dotlock_free(dotlock);

-- 
I don't need a name; my number's just fine. |  
Mario.Nigrovic@freescale.com
It's nobody else's -- just mine, all mine.  |                  
480-413-3578
                                                           Internal Use Only