August 2017 - dovecot

Strange folders
by Marti Markov 31 Aug '17

31 Aug '17

Hi all, I’ve been experiencing the “magical” creation of random folders on my mail server. I managed to catch the creation of one last night. Here is the log: Aug 30 01:38:13 mail dovecot: imap(f.lastName(a)domain.com): Debug: Namespace : /home/vmail/domain.com/f.lastName/Maildir/.wkkwgfrubwioniohkcyqttugdtdabuyn doesn't exist yet, using default permissions Aug 30 01:38:13 mail dovecot: imap(f.lastName(a)domain.com): Debug: Namespace : Using permissions from /home/vmail/domain.com/f.lastName/Maildir: mode=0700 gid=default Aug 30 01:38:13 mail dovecot: imap(f.lastName(a)domain.com): Debug: acl vfile: file /home/vmail/domain.com/f.lastName/Maildir/.wkkwgfrubwioniohkcyqttugdtdabuyn/dovecot-acl not found I have substituted only the username and the domain name. The folder name is as it was created: wkkwgfrubwioniohkcyqttugdtdabuyn There has been no activity logged on the exim4 side at that time. What would be the best course of action to determine what is happening?

2 1

Segmentation fault with doveadm search
by Stefán Tamás 30 Aug '17

30 Aug '17

Hi, We have a Ubuntu 14.04 server with dovecot 2.2.9 If I try to run /usr/bin/doveadm search -u email(a)exmaple.com mailbox Junk savedbefore 30d It will dump a core file. We have a similar config on an other server and there is no problem there. Any ideas how to prevent the core dump? root@triws2:~# gdb --args doveadm search -u email(a)example.com mailbox Junk savedbefore 30d Reading symbols from doveadm...Reading symbols from /usr/lib/debug//usr/bin/doveadm...done. done. (gdb) run Starting program: /usr/bin/doveadm search -u email(a)example.com mailbox Junk savedbefore 30d Program received signal SIGSEGV, Segmentation fault. 0x0000000000000000 in ?? () (gdb) where #0 0x0000000000000000 in ?? () #1 0x00007ffff76ef1b5 in mailbox_get_metadata (box=box@entry=0x5555557e1c70, items=items@entry=MAILBOX_METADATA_GUID, metadata_r=metadata_r@entry=0x7fffffffe2f0) at mail-storage.c:1544 #2 0x0000555555570918 in cmd_search_box (info=<optimized out>, ctx=0x5555557c8230) at doveadm-mail-search.c:28 #3 cmd_search_run (ctx=0x5555557c8230, user=<optimized out>) at doveadm-mail-search.c:58 #4 0x0000555555569b5b in doveadm_mail_next_user (ctx=0x5555557c8230, input=<optimized out>, error_r=0x7fffffffe3f8) at doveadm-mail.c:314 #5 0x000055555556a8ac in doveadm_mail_cmd (argv=0x5555557c33b0, argc=7, cmd=0x5555557c7408) at doveadm-mail.c:525 #6 doveadm_mail_try_run (cmd_name=<optimized out>, argc=7, argv=0x5555557c3398) at doveadm-mail.c:617 #7 0x00005555555696d4 in main (argc=7, argv=0x5555557c3398) at doveadm.c:398 Üdvözlettel Stefán Tamás --------------------------------- domain > email > web >>> siker Numex Informatika Kft. Mobil: +36 20 956 0233, Tel: +36 1 205 3915 http://numex.hu

2 6

Proper way to enable plugin for some users
by Ian Bobbitt 29 Aug '17

29 Aug '17

What's the best way to enable a plugin (in this case, mail_log) for only some users? I have this as a field returned in my sql user_query that seems to be doing the job: if(debug_log != 0 and '%s' = 'imap', 'listescape acl notify replication mail_log', 'listescape acl notify replication') as mail_plugins But it feels clunky and error prone having to replicate the entire list of plugins in my userdb. Is this the best way to do it, or is there something I should be doing differently? OS: CentOS 7 Dovecot: 2.2.31, GF RPMs

1 0

unexpected delivery location
by lists 29 Aug '17

29 Aug '17

Hi, We're running dovecot 2.2.13, virtual users, with postfix. We have an olddomain and a new domainname. To 'translate' *@olddomain into *@newdomain, I have configured: > cat /etc/postfix/canonical > @olddomain.com @newdomain.com While this seems to work, lately we have noticed that dovecot occasionally creates mailboxes for non-existent users, like: > /var/vmail/...username > /var/vmail/20username This only seems to be happening for mails sent to *olddomain* Looking at the source of these mails, they are indeed sent to 20username(a)olddomain.com and ...username(a)olddomain.com Mail headers sample: > Delivered-To: 20username(a)newdomain.com > Received: from server1.newdomain.com (server2.newdomain.com [x.y.z.q]) > by server3.newdomain.com (Postfix) with ESMTPS id 067B4812CF29E > for <20username(a)olddomain.com>; Thu, 13 Jul 2017 19:05:02 +0200 (CEST) Why doesn't the emailadress 20username(a)newdomain.com in the above generate an error, but instead are delivered to a newly created mailbox? My feeling is that something is wrong with the way I created the alias for olddomain -> newdomain with the canonical-file . Does that make sense? Our goal is to accept *(a)olddomain.com like it was sent to *(a)newdomain.com, including delivery failures for nonexistent mailboxes like 20username We configured postfix like: > canonical_maps = hash:/etc/postfix/canonical > virtual_alias_maps = ldap:/etc/postfix/ad-mailboxes.cf, ldap:/etc/postfix/ad-groups.cf > virtual_uid_maps = static:5000 > virtual_gid_maps = static:5000 > virtual_transport = dovecot Verified that this config works for newdomain, using: > root@mail:/etc/postfix# postmap -q alias(a)newdomain.com ldap:/etc/postfix/ad-mailboxes.cf > userid(a)newdomain.com > root@mail:/etc/postfix# postmap -q alias(a)olddomain.com ldap:/etc/postfix/ad-mailboxes.cf > root@mail:/etc/postfix# olddomain gives no result. Then, from master.cf: > dovecot unix - n n - - pipe > flags=DRhu user=vmail:vmail argv=/usr/lib/dovecot/dovecot-lda -f ${sender} -a ${recipient} -d ${user}@${nexthop} Finally, the mail location as defined in dovecot: > mail_location = maildir:/var/vmail/%Ln/Maildir:LAYOUT=fs:DIRNAME=mAildir So, can anyone explain why for mails to olddomain, dovecot creates new mailboxes, instead of simply generating an error?

3 6

dovecot-2.2.32/src/doveadm/doveadm-mail-fetch.c: 4 * suspicious expression ?
by David Binderman 29 Aug '17

29 Aug '17

Hello there, dovecot-2.2.32/src/doveadm/doveadm-mail-fetch.c:130]: (style) Suspicious condition (assignment + comparison); Clarify expression with parentheses. Source code is if ((ret = doveadm_print_istream(input) < 0)) Maybe better code if ((ret = doveadm_print_istream(input)) < 0) Some duplicates: [dovecot-2.2.32/src/doveadm/doveadm-mail-fetch.c:220]: (style) Suspicious condition (assignment + comparison); Clarify expression with parentheses. [dovecot-2.2.32/src/doveadm/doveadm-mail-fetch.c:237]: (style) Suspicious condition (assignment + comparison); Clarify expression with parentheses. [dovecot-2.2.32/src/doveadm/doveadm-mail-fetch.c:261]: (style) Suspicious condition (assignment + comparison); Clarify expression with parentheses. Regards David Binderman

2 2

vacation problem with SRS
by Kacper Guzik 28 Aug '17

28 Aug '17

Hi i have similiar problem like here: http://www.iredmail.org/forum/topic11833-iredmail-support-vocation-respone-… email send : from : web123(a)gmail.com to : jacka(a)mail.com vocation sent back from : jacka(a)mail.com to : srs0=hmc8=v7=gmail.com=web123(a)mail.com postsrsd changing return-path from web123(a)gmail.com to srs0=hmc8=v7= gmail.com=web123(a)mail.com this is no problem for me but sieve: Vacation's messages are always addressed to the Return-Path address it is possible some how change this ugly to header to normal? I can't find anything on dovecot mailing lists

2 7

[PATCH] Add support for lower TLS version than default
by Sebastian Andrzej Siewior 27 Aug '17

27 Aug '17

The openssl library in Debian unstable (targeting Buster) supports TLS1.2 by default. The library itself supports also TLS1.1 and TLS1.0. If the admin decides to also support TLS1.[01] users he can then enable the lower protocol version in case the users can't update their system. Signed-off-by: Sebastian Andrzej Siewior <sebastian(a)breakpoint.cc> --- src/config/all-settings.c | 1 + src/lib-master/master-service-ssl-settings.c | 2 ++ src/lib-master/master-service-ssl-settings.h | 1 + src/login-common/ssl-proxy-openssl.c | 15 ++++++++++++++- 4 files changed, 18 insertions(+), 1 deletion(-) --- a/src/config/all-settings.c +++ b/src/config/all-settings.c @@ -308,6 +308,7 @@ struct master_service_ssl_settings { const char *ssl_cert_username_field; const char *ssl_crypto_device; const char *ssl_options; + const char *ssl_lowest_version; bool ssl_verify_client_cert; bool ssl_require_crl; --- a/src/lib-master/master-service-ssl-settings.c +++ b/src/lib-master/master-service-ssl-settings.c @@ -26,6 +26,7 @@ static const struct setting_define maste DEF(SET_STR, ssl_protocols), DEF(SET_STR, ssl_cert_username_field), DEF(SET_STR, ssl_crypto_device), + DEF(SET_STR, ssl_lowest_version), DEF(SET_BOOL, ssl_verify_client_cert), DEF(SET_BOOL, ssl_require_crl), DEF(SET_BOOL, verbose_ssl), @@ -54,6 +55,7 @@ static const struct master_service_ssl_s .ssl_protocols = "!SSLv3", #endif .ssl_cert_username_field = "commonName", + .ssl_lowest_version = NULL, .ssl_crypto_device = "", .ssl_verify_client_cert = FALSE, .ssl_require_crl = TRUE, --- a/src/lib-master/master-service-ssl-settings.h +++ b/src/lib-master/master-service-ssl-settings.h @@ -16,6 +16,7 @@ struct master_service_ssl_settings { const char *ssl_cert_username_field; const char *ssl_crypto_device; const char *ssl_options; + const char *ssl_lowest_version; bool ssl_verify_client_cert; bool ssl_require_crl; --- a/src/login-common/ssl-proxy-openssl.c +++ b/src/login-common/ssl-proxy-openssl.c @@ -1302,7 +1302,20 @@ ssl_server_context_init(const struct log if (ctx->prefer_server_ciphers) SSL_CTX_set_options(ssl_ctx, SSL_OP_CIPHER_SERVER_PREFERENCE); SSL_CTX_set_options(ssl_ctx, openssl_get_protocol_options(ctx->protocols)); - +#if OPENSSL_VERSION_NUMBER >= 0x10100000 + if (ssl_set->ssl_lowest_version) { + if (!strcmp(ssl_set->ssl_lowest_version, "TLS1.0")) + SSL_CTX_set_min_proto_version(ssl_ctx, TLS1_VERSION); + else if (!strcmp(ssl_set->ssl_lowest_version, "TLS1.1")) + SSL_CTX_set_min_proto_version(ssl_ctx, TLS1_1_VERSION); + else if (!strcmp(ssl_set->ssl_lowest_version, "TLS1.2")) + SSL_CTX_set_min_proto_version(ssl_ctx, TLS1_2_VERSION); + else + i_fatal("TLS min version: '%s' is invalid. Only " + "'TLS1.0' and 'TLS1.1' is supported", + ssl_set->ssl_lowest_version); + } +#endif if (ctx->pri.cert != NULL && *ctx->pri.cert != '\0' && ssl_proxy_ctx_use_certificate_chain(ctx->ctx, ctx->pri.cert) != 1) { i_fatal("Can't load ssl_cert: %s",

3 6

Debian packages for 2.2.32?
by Harald Leithner 27 Aug '17

27 Aug '17

Hi, It seams that IX didn't Motive Themen Version change. Thx Harald -- Diese Nachricht wurde von meinem Mobiltelefon gesendet.

3 6

Released Pigeonhole v0.4.20 for Dovecot v2.2.32.
by Stephan Bosch 27 Aug '17

27 Aug '17

Hello Dovecot users, Here's the definitive 0.4.20 release. There were no changes since the release candidate. Changelog v0.4.20: + Made the retention period for redirect duplicate identifiers configurable. For accounts that perform many redirects, the lda-dupes database could grow to impractical sizes. Changed the default retention period from 24 to 12 hours. - sieve-filter: Fixed memory leak: forgot to clean up script binary at end of execution. Normally, this would merely be an inconsequential memory leak. However, when the script comes from an LDAP storage, this would cause io leak warnings. - managesieve-login: Fixed handling of AUTHENTICATE command. A second authenticate command would be parsed wrong. This problem was caused by changes in the previous release. - LDA Sieve plugin: Fixed minor memory leak caused by not cleaning up the sieve_discard script. The release is available as follows: https://pigeonhole.dovecot.org/releases/2.2/dovecot-2.2-pigeonhole-0.4.20.t… https://pigeonhole.dovecot.org/releases/2.2/dovecot-2.2-pigeonhole-0.4.20.t… Refer to http://pigeonhole.dovecot.org and the Dovecot v2.x wiki for more information. Have fun testing this release and don't hesitate to notify me when there are any problems. Regards, -- Stephan Bosch stephan(a)rename-it.nl

1 0

Re: a bit further along - OpenSSL - Re: trouble compiling Dovecot 2.2.31 on Solaris 10 SPARC - libssl_iostream_openssl.so is not portable!
by Joseph Tam 27 Aug '17

27 Aug '17

Aki Tuomi <aki.tuomi(a)dovecot.fi> writes: > > I simulated your compile arguments as best as I could, and compared the libtool > > command in my compile logs with your's. > > > > (mine) > > Mine: -lssl -lcrypto -lrt -lnsl -lsocket -lsendfile > > > > (your's) > > -lcrypto -lrt -lnsl -lsocket -lsendfile -lcrypt > > > > You're missing "-lssl" but have a spurious "-lcrypt". Probably all your missing symbols are > > actually in libssl.so. > > OpenSSL usually requires both libssl and libcrypto, the -lcrypt probably is for > crypt() Makes sense. So the OP can confirm this is the problem by manually running the libtool command but adding "-lssl", and if that works, then something screwed up Makefile or other build files. Joseph Tam <jtam.home(a)gmail.com>

1 0