- dovecot - dovecot.org

Re: [Dovecot] Dovecot LDA
by Steven F Siirila 09 Oct '06

09 Oct '06

On Sun, Oct 08, 2006 at 10:25:53PM +0300, Timo Sirainen wrote: > On Fri, 2006-10-06 at 21:00 -0500, Steven F Siirila wrote: > > I am running Dovecot 1.0-rc6 along with dovecot-sieve from 2006-10-05 CVS. > > I am calling dovecot deliver from Postfix (main.cf): > > > > mailbox_command = /opt/dovecot/libexec/dovecot/deliver > > > > Upon switching to dovecot deliver, I observed two problems: > > > > 1) An extraneous "From " header now appears at the top of each message > > delivered, e.g.: > > > > From dovecot.deliver Fri Oct 06 19:59:57 2006 > > From sfs(a)tc.umn.edu Fri Oct 6 19:59:57 2006 > > Return-Path: <sfs(a)tc.umn.edu> > > > > Using Postfix delivery the first of these lines does not appear. > > This, of course, confuses MUAs. > > Don't give F flag to Postfix's pipe command. The messages in question are accepted by the Postfix smtpd process and processed by Postfix, and handed off to dovecot deliver. I am not specifying an F flag anywhere. If I remove the "mailbox_command" I specified above, Postfix delivers the mail with only the one "From " header -- with the mailbox_command above (no options specified), I get two "From " headers. > Although this reminds me that the "dovecot.deliver" is wrong in there.. > Hmm. I think I'll add a maildrop-compatible -f <envelope> parameter > here. How will this help me? I don't even want the added "From " header. If Postfix is not supposed to be writing one, then how do I tell it not to and how does Postfix tell dovecot deliver what to put in the "From " header it writes? (It must be the proper return address seen in the original SMTP transaction as the MAIL FROM value) > There, done in CVS. I'll also make 1.0rc8 release today. > > > 2) The "Content-Length:" header appears after a blank line, essentially > > putting it into the RFC822 body, e.g.: > > > > User-Agent: Mutt/1.4.2.1i > > X-UID: 11510 > > Status: > > X-Keywords: > > > > Content-Length: 200 > > > > test > > Are you sure it's a blank line? Dovecot adds lots of spaces after > X-Keywords which could just make it look like that. You're correct; the line prior to the "Content-Length:" header was indeed an "X-Keywords:" header padded out to exactly 80 characters, making it appear that there was an extra blank line after it (my xterm window is set to 80 columns). -- Steven F. Siirila Office: Lind Hall, Room 130B Internet Services E-mail: sfs(a)umn.edu Office of Information Technology Voice: (612) 626-0244 University of Minnesota Fax: (612) 626-7593

1 0

[Dovecot] which e-mail client can work with recent dovecot?
by Sergey Ivanov 09 Oct '06

09 Oct '06

Hi, I'm trying to use dovecot-1.0.rc7. Before I've used dovecot-1.0 from nightly cvs build of August 08, 2006. It worked perfectly, on Linux 64-bit, you can see spec and patches at http://www.sisyphus.ru/srpm/dovecot/ When I tried to upgrade to rc7 and to recent nightly cvs builds with the same configuration and patches (provided necessary adaptations), it builds O.K, but can not move messages from folder to folder. I've tried the following mail clients: thunderbird-1.5.0.7, evolution-2.0.2, sylpheed-claws-2.0 and KMail-1.7.1 (all from Centos-4.4). KMail refuses to copy (or to move) multiple selected messages, saying:" Error while copying messages.Connection to Server Closed Unexpectedly ...". Evolution is not able to open folders at 3-rd level (like dovecot/2006/09), or it is so slow that I was not able to test moving/copying messages in it. Sylpheed-claws died whan I tried this message copying. Thunderbird tries to move infinetly in a loop, filling the dovecot's log with lines: --- Sep 28 12:56:25 alt64 dovecot: imap-login: Login: user=<seriv>, method=plain, rip=66.80.117.2, lip=192.168.10.8, TLS Sep 28 12:56:25 alt64 dovecot: IMAP(seriv): file mail-cache-transaction.c: line 709 (mail_cache_add): assertion failed: (fixed_size == (unsigned int)-1 || fixed_size == data_size) Sep 28 12:56:25 alt64 dovecot: child 25177 (imap) killed with signal 6 Sep 28 12:56:25 alt64 dovecot: imap-login: Login: user=<seriv>, method=plain, rip=66.80.117.2, lip=192.168.10.8, TLS Sep 28 12:56:25 alt64 dovecot: IMAP(seriv): file mail-cache-transaction.c: line 709 (mail_cache_add): assertion failed: (fixed_size == (unsigned int)-1 || fixed_size == data_size) Sep 28 12:56:25 alt64 dovecot: child 25183 (imap) killed with signal 6 --- (approximitely 2 cicles per second). And when I reinstalled back the previously worked version of dovecot, the problem remains. -- With best regards, Sergey Ivanov.

3 3

[Dovecot] 1.0rc8 - not quite yet
by Timo Sirainen 09 Oct '06

09 Oct '06

OK, I know I promised rc8 today in my recent mails, but it's way too late and writing changes to NEWS file takes time. :) So I guess it'll happen tomorrow then. In the mean time it would be nice if you'd test that I didn't break anything badly with my recent changes. Especially when I changed the password verifications to detect if the password is in hex or base64 encoding based on the string length. This also allowed me to make LDAP-MD5 and PLAIN-MD5 use the same password verification function. I like this change. :) So, the almost-rc8 snapshot is in: http://dovecot.org/nightly/dovecot-latest.tar.gz I also started building dovecot-sieve snapshots: http://dovecot.org/nightly/sieve/

1 0

[Dovecot] Maildir++ quota inconsistency
by Anand Buddhdev 09 Oct '06

09 Oct '06

I've found what I think is an inconsistency in the Maildir quota implementation. According to the Maildir++ quota specification at: http://www.inter7.com/courierimap/README.maildirquota.html Maildir++ quota can be specified as maximum size, or maximum number of messages, or even both. The size specification is supposed to be in bytes. So I could use '10000000S' to set a quota of around 10MB. Now, the Exim MTA has an independent Maildir++ quota implementation, and it follows the specification, so if I set the quota in Exim to 10000000, it will work perfectly with sqwebmail and courier-imap. However, it doesn't play well with Dovecot, because Dovecot chooses to interpret the quota as kilobytes, rather than bytes. An example: If I set the quota to 10240 in Dovecot, it creates a maildirsize file with the value 10240000 in it. If I then set the quota to 10M in Exim, Exim calculates that as 10485760. It detects that the value in the maildirsize file is wrong, and recreates the file with a quota of 10485760 bytes. Exim's behaviour is consistent with the courier family of packages, as well as the specification, so I'm venturing to say that Dovecot's implementation is slightly non-standard, and should be fixed so that the quota setting is interpreted as bytes, and not as kilobytes, to avoid a situation like the above. -- Anand

4 4

[Dovecot] Password mismatch - SHA1 authentication problem.
by guard 09 Oct '06

09 Oct '06

Hello, I have problem with authentication. I'm using SHA1 passwords stored in MySQL. >From debug: Info: auth-worker(default): sql(test@mail, IP): query: SELECT password FROM vmail WHERE mail = 'test@mail' Info: auth-worker(default): sql(test@mail ,IP): Password mismatch Info: auth-worker(default): sql(test@mail ,IP): SHA1(dupadupa) != '2bca9dbe8a14811c687bb8c51c0e1cd14db8a018' I tried convert password to upper case, but it didn't help. Then I tried authentication in plain text and everything works fine... So, does dovecot crypt SHA1 using other method? I also use same SHA1 passwords to authenticate users in exim and it doesn't have any problem with that. I'm using mysql 5.0.24 and dovecot 1.0 rc7. Best regards, Tomasz Krynicki

2 1

[Dovecot] Dovecot behaviour when the server's file system fails (possible feature request)
by Timo Neuvonen 09 Oct '06

09 Oct '06

I've met a couple of times a situation, that there is something wrong with the mail server's operating system / file system, but Dovecot is still working, although it cannot any more access the mailbox files. The result is, that the imap mail client in the workstation automatically synchronises with the improperly empty-looking mailbox, effectively deleting everything previously loaded into the workstation's mail client (eg. Outlook Express) when using imap protocol. This is very annoying, especially when the workstation isn't connected with a high-speed-enough connection, since after the server is properly brought up again, reloading all the contents of the improperly emptied mail client may take considerable amount of time. This is especially the case, if the workstation has kind of mobile connection (eg. gprs) with very limited bandwith -in that case the sychronizing of eg. 20MB account may take at least from half an hour to more than one hour, and in some cases (eg. gprs roaming) also cost considerable amount of money (gprs data transfer may cost from eg. 5 to over 10 euros or dollars per megabyte, when roaming outside the phone operator's own network). So, is there currently a way to configure Dovecot in a way it would recognize that the mailbox actually is not accessible, and in that case it would return an error message to the mail client instead of returning "empty mailbox"? Of course, a smart enough mail client might also ask something from the user, but that seems not be the case either... If this behavious cannot be changed with the current configuration options, then a kind feature request to the developer: Would it be possible to change the behaviour of Dovecot when it actually can not access the mailbox files? Although the actual problem in this case is outside Dovecot, this kind of improvement would greatly improve the error tolerance of the whole system. -- TiN

2 1

[Dovecot] fd leak in 1.0rc7?
by Quentin Garnier 09 Oct '06

09 Oct '06

Hi, I just upgraded a 0.99 installation to 1.0rc7 (going from mbox to maildir along the way, but that's another story), and I'm seeing a rather fast fd quick; I already had to restart dovecot once, after a few hours. What is leaked are pipes, left open on the master dovecot process while the other end is long gone. I can't tell right now if it happens for every single login process spawned (either pop3 or imap), but at least a lot of them. FWIW, I'm running dovecot 1.0rc7 on NetBSD 2.0.3_STABLE. I'll try debugging tomorrow, but if anyone has an idea what happens in the meantime, I'm all hears. -- Quentin Garnier - cube(a)cubidou.net - cube(a)NetBSD.org "When I find the controls, I'll go where I like, I'll know where I want to be, but maybe for now I'll stay right here on a silent sea." KT Tunstall, Silent Sea, Eye to the Telescope, 2004.

2 3

[Dovecot] Pop3s duplicates
by Mark Adams 09 Oct '06

09 Oct '06

Hi, dovecot 1.0 rc6 Can someone please advise on my interpretation of the following logs, Users are reporting duplicate email coming through to their inbox. (user@domain replaced actual username and domain) There are also reports of the connection seemingly freezing whilst the emails are being downloaded (pausing at a certain amount of kB then continuing after a minute or 2). info: pop3-login: Login: user=<user@domain>, method=PLAIN, rip=89.192.*.**, lip=172.16.22.10, TLS info: POP3(user@domain): Disconnected top=0/0, retr=4/183443, del=0/313, size=83779942 info: pop3-login: Login: user=<user@domain>, method=PLAIN, rip=89.192.*.**, lip=172.16.22.10, TLS info: POP3(user@domain): Disconnected top=0/0, retr=3/183443, del=0/322, size=84201570 In the first instance the user downloaded 4 emails totalling 183443 bytes leaving 313 messages on the server. (outlook 2003 set to leave a copy of messages on the server) second instance user downloaded 3 emails, exactly the same amount of bytes and left 322 messages on the server. Does this mean the users connection was interrupted when downloading the original emails and they attempted to be sent again a second time? how could it be that there are 9 emails left on the server if only 3 extra were downloaded? Any help on this would be much appreciated. Thanks, Mark

3 5

[Dovecot] coding techniques
by David Lee 09 Oct '06

09 Oct '06

Whilst debugging a problem in dovecot LDA over the last few days, I came across two different issues in coding techniques. (Note: Please don't take this negatively; my intent is both positive and constructive!) The front page of the website www.dovecot.org says: "it uses several coding techniques to avoid most of the common pitfalls" So I hope it is OK to follow the spirit of that with some suggestions. The two issues I encountered both seem to be fairly easy holes to fall into, and yet equally easy to avoid in the first place by using established C practices. Firstly, many 'if' statements having a single statement (whether in then or else part) are currently written in a short-cut fashion: if (condition) statement 1; else statement 2; This, though technically legal, is potentially lethal when inserting (say) debugging statements. Without diligence it is all to easy to produce: if (condition) debug 1a; statement 1; debug 1b; else debug 2a; statement 2; debug 2b; A simple edit that looks, layout-wise, as if it would do the right thing (and do in Python!); but actually radically changes the execution flow. (This is left as an exercise to the reader... to demonstrate the point.) Could I suggest that the relevant brackets be always used as a matter of course? Something like: if (condition) { statement 1; } else { statement 2; } Then the insertion of debug statements doesn't inflict surprise. (No, I didn't fall into this hole. But it is a trap waiting to happen.) Secondly, code duplication. To give one example: 'get_var_expand_table()' appears in both 'deliver/deliver.c' and 'master/mail-process.c'. (There is also a similar looking function in 'auth/auth-request.c'.) Perhaps consideration should be given to merging such instances. For instance, my woes with "%i"-expansion being treated differently in mail-reading and mail-delivery would likely never have arisen if common code had been used. I hope that helps (constructively!). Best wishes. -- : David Lee I.T. Service : : Senior Systems Programmer Computer Centre : : Durham University : : http://www.dur.ac.uk/t.d.lee/ South Road : : Durham DH1 3LE : : Phone: +44 191 334 2752 U.K. :

2 1

[Dovecot] rc7 crashing under heavy load
by Jeff A. Earickson 09 Oct '06

09 Oct '06

Help! Our students are back and they are pounding my dovecot server to death (ie "Login process died too early - shutting down" or "too many open files" type crash). I've had my third crash today. My setup: Solaris 10, imap and imaps, 3K users coming in via horde webmail. My current connection settings are: login_max_processes_count = 2048 login_max_connections = 4096 Q: Does raising these limits actually make my problem worse or better? Are there any other variables that I should change to prevent the "too many open files"? Suggestions? Jeff Earickson Colby College

6 9