November 2016 - dovecot

doveadm service: verify client cert
by Matwey V. Kornilov 20 Nov '16

20 Nov '16

Hello, I want to open a socket for connecting doveadm using tcps. I do the following: service doveadm { inet_listener { port = 5001 ssl = yes } } At the same time, I would like to verify client certificates for connections goes to port 5001. I am trying to do the following, but it doesn't work: protocol doveadm { ssl_require_crl = yes ssl_verify_client_cert = yes } How could I achieve required behavior?

1 0

Compound Literal - xlc and gcc differences can be patched
by Michael Felt 19 Nov '16

19 Nov '16

Since I so miserably misspelled packaging - a new thread specific to the issue at hand. I found a "workaround". In short, xlc does not accept arrays of nested array - the bottom line (best message) is: 1506-995 (S) An aggregate containing a flexible array member cannot be used as a member of a structure or as an array element. At the core - all the other messages come because this (I shall call it nesting). For the test-http-auth.c file I have a patch (attached). make ran for while with no issue, then it started running into several files with this same style of specification - doveadm-dict.c being the first one. I ran "make -i" and I notice several files with this - for xlc - "unaccepted" syntax. Please look at the patch - what it basically is - and let me know whether you would consider accepting patches in this form (e.g., verify gcc (I assume) will accept it as well). If your answer is yes, I shall proceed with the additional files (learn how your MACROS work!) and send them to you. Again, initially I would just send one, e.g., doveadm-dict.c - to be sure your regular compiler also builds this alternate specification. Thank you for your consideration! Michael

2 5

Re: Good email client to use with Dovecot?
by Joseph Tam 19 Nov '16

19 Nov '16

Steve Litt asks > * And most important, where can I find the best, most unambiguous > Alpine documentation, that doesn't assume I know anything about > Alpine? The ~/.pinerc config file is well documented. There is a bewildering number of options, and you can tweak it just about anyway you like. Choice of mail readers tend to a religious issue. Users of a particular reader will only change when you pry their cold dead fingers off of it. Alpine is not a bad choice (I've been using it and it's predeccesor for decades), and I go through mail faster than most people I know as I can type faster than they can click buttons, scroll, and drag windows around. However, it sort of sucks when you read formatted mail HTML with lots of graphics). Joseph Tam <jtam.home(a)gmail.com>

1 0

Generalized looping possible in pigeonhole sieve?
by Jeff Kletsky 18 Nov '16

18 Nov '16

I'm struggling to achieve the kind of filtering with sieve that I was able to do with procmail. TL;DR I'd like a way to loop through a set of (address, destination) pairs in sieve so that I can maintain the (address, destination) pairs in one place and not have to explicitly write scores of nearly identical [...] if address :matches ["From", "Sender", "To", "Cc"] "address53" { fileinto "destination53" } if address :matches ["From", "Sender", "To", "Cc"] "address54" { fileinto "destination54" } [...] Longer: I've been using a table-based dispatch approach with procmail for years and it was working reasonably well. The approach uses list of pairs of address and destination boxes and doing a lookup based on the message to be delivered. With procmail, I collected the list of addresses and passed them to external scripts to do the lookup. If I just had a dozen boxes that I deal with, that would be straightforward to implement and maintain in sieve. As the list is an order of magnitude greater than that, maintaining the list of pairs in an external file and programmatically looping through it is a lot easier. I've looked at vnd.dovecot.execute and it certainly can return a destination box name given the input information. Unfortunately, the construct of (for example) address :matches "To" "*"' only returns the first address, not the list of addresses. I can't see a good way to get all the addressees from sieve to pass to the external program. If I have to pass the message (or at least the interesting headers) to the external program and then parse the address list myself, then I'm nearly all the way to just using a full-on external program to parse the whole message. While I'm not adverse to writing the whole thing in Python (or the like) and just using sieve as a thin shell, I'd like to make sure I'm not missing something in sieve. One way to achieve this would be to be able to loop through the pairs and using variable substitution for each iteration. I haven't seen anything in Dovecot Pigeonhole sieve that allows this. If you have any ideas on how to harness sieve for this, I'd appreciate it! Jeff

3 2

BUG: nopassword doesn't work with CRAM-MD5
by Arkadiusz Miśkiewicz 18 Nov '16

18 Nov '16

Hello. dovecot 2.2.26.0 When testing nopassword extra field (http://wiki2.dovecot.org/PasswordDatabase/ExtraFields) with CRAM-MD5 dovecot doesn't allow any password (while it should) and returns " Authentication failed" while in logs: Nov 17 08:22:34 auth-worker(1551): Info: sql(pepe,127.0.0.1,<Y8amDXpBptV/AAAB>): Requested CRAM-MD5 scheme, but we have a NULL password NULL is there because our sql query returns empty password just like wiki says "nopassword: you want to allow all passwords, use an empty password and this field. " If password is returned in sql query then it fails, too: Nov 17 09:00:49 auth-worker(2206): Error: sql(pepe,127.0.0.1,<eO5vlnpBtNd/AAAB>): nopassword set but password is non- empty So looks to be a bug. -- Arkadiusz Miśkiewicz, arekm / ( maven.pl | pld-linux.org )

3 6

Any way to simplify repeated use of same string-list?
by Jeff Kletsky 18 Nov '16

18 Nov '16

I often use the same set of header names when doing an 'address' comparison. I would prefer to be able to specify that once, then re-use it in several places. As an example set "address_mailman_fields" "\"To\", \"Cc\", \"From\", \"Errors-To\""; if address :domain :matches ["${address_mailman_fields}"] "example.com" { [...] } if address :domain :matches ["${address_mailman_fields}"] "dovecot.org" { [...] } or the like would be quite helpful. From what I can tell, even though the variable interpolation *looks* like a string-list, it isn't parsed as such. For example, from sieve-test output: sieve-test(jeff): Debug: Mailbox <filename> Opened mail UID=1 because: header "To", "Cc", "From", "Errors-To" (Cache file is unusable) testing: line 27: info: DEBUG: address ["${address_mailman_fields}"] didn't match. sieve-test(jeff): Debug: Mailbox <filename> Opened mail UID=1 because: header ["To", "Cc", "From", "Errors-To"] (Cache file is unusable) testing: line 34: info: DEBUG: address "[${address_mailman_fields}]" didn't match. Is there any trick available to specify a string-list once, then re-use it throughout the script? Thanks, Jeff

1 0

dovecot pre-install issue
by soumitri＠iitk.ac.in 17 Nov '16

17 Nov '16

Hello all, I am going for a dovecot director based setup (2 director+ 2 imap), more imap servers will be added later depending on demand/load. Presently I have 12000+ dovecot users with Maildir quota varying from 1 GB to 20GB. (peak hour IOPS 50000+) I am having 2 options in choosing dovecot version. 1) Old stable release. I.e RHEL, with prebuilt binary. This will be having less trouble in managing. Why RHEL still using version 2.2.10. 2) Latest release with best features and lesser known bugs. I.e CentOS7 with with latest compiled version. I have to be more involved if a bug is found. I will prefer a less admin work after the setup, with all/most features working. If you have a recent similar setup/dovecot gurus, Pl. suggest. With thanks & regards, -- Soumitri Mishra http://home.iitk.ac.in/~soumitri/

3 5

New Gmail Android App and Dovecot
by David Flanigan 16 Nov '16

16 Nov '16

Dovecot Users, I have a sudden new problem that has cropped up just recently and I am stumped. My implementation has been working for over a year with no issues. Version Data: Dovecot: v2.2.10 (using self signed certificates) openSSL: v1.01e Gmail Android App: v6.10.23 November 3, 2016 The issue: I use the iphone Gmail App to access my e-mail from phone and tablet. It is simple and supports multiple accounts very nicely. This has worked for better than 18 months without issue. This week I upgraded my Android app to the latest version of Gmail and I stopped getting incoming e-mail. I can still pick up via outlook and other tools I use, but just not the android version of Gmail. When I attempt to pick up mail i see the below error in my log file. _Nov 16 14:18:56 ns1 dovecot: imap-login: Disconnected (no auth attempts in 0 secs): user=<>, rip=10.10.10.198, lip=10.1.1.10, TLS: SSL_read() failed: error:1408F10B:SSL routines:SSL3_GET_RECORD:wrong version number, session=<73vF8W9BdgCmrL7G>_ Disconnect and no new e-mail. There is no error in the Gmail app that I can find. The Gmail app works fine with my other accounts (Gmail, work, etc) so I am wondering if it is a configuration issue on my side or Gmail App stopped supporting something I was using. My SSl config is RPM standard other than SSL=yes and my self signed certificates. Any thoughts would be appreciated. -- Kind Regards, David

3 2

Crashing when run against OpenSSL 1.1.0c
by Michael Marley 15 Nov '16

15 Nov '16

Hi, I am running Dovecot 2.2.26.0 compiled against OpenSSL 1.1 and, since upgrading to OpenSSL 1.1.0c, the "lmtp" process has been crashing with SIGSEGV whenever it receives SIGINT. This always happens a minute or so after the lmtp process handles a message. It can also be manually reproduced by sending SIGINT to one of the running lmtp processes. I am compiling and running on an Ubuntu 17.04 x86_64 system using GCC 6.2. Here is the output of me reproducing it with gdb: (gdb) signal SIGINT Continuing with signal SIGINT. Program received signal SIGSEGV, Segmentation fault. 0x00007f6748cc2fb0 in ?? () (gdb) bt #0 0x00007f6748cc2fb0 in ?? () #1 0x00007f674872ac60 in ossl_init_thread_stop (locals=<optimized out>) at crypto/init.c:336 #2 0x00007f674872aee4 in OPENSSL_cleanup () at crypto/init.c:391 #3 0x00007f67491052e0 in __run_exit_handlers (status=0, listp=0x7f674948c5d8 <__exit_funcs>, run_list_atexit=run_list_atexit@entry=true, run_dtors=run_dtors@entry=true) at exit.c:83 #4 0x00007f674910533a in __GI_exit (status=<optimized out>) at exit.c:105 #5 0x00007f67490eb3f8 in __libc_start_main (main=0x555b35fbfbc0 <main>, argc=1, argv=0x7ffd4ede3588, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7ffd4ede3578) at ../csu/libc-start.c:325 #6 0x0000555b35fbfe3a in _start () Here is the output of "doveconf -n": # 2.2.26.0 (23d1de6): /etc/dovecot/dovecot.conf # Pigeonhole version 0.4.devel (623ae77) # OS: Linux 4.8.7-040807-generic x86_64 Ubuntu Zesty Zapus (development branch) auth_mechanisms = plain login auth_username_format = %Ln mail_location = mdbox:~/mdbox mailbox_list_index = yes managesieve_notify_capability = mailto managesieve_sieve_capability = fileinto reject envelope encoded-character vacation subaddress comparator-i;ascii-numeric relational regex imap4flags copy include variables body enotify environment mailbox date index ihave duplicate mime foreverypart extracttext namespace inbox { inbox = yes location = mailbox Drafts { special_use = \Drafts } mailbox Junk { special_use = \Junk } mailbox Sent { special_use = \Sent } mailbox "Sent Messages" { special_use = \Sent } mailbox Trash { special_use = \Trash } prefix = } passdb { driver = pam } plugin { sieve = file:~/sieve;active=~/.dovecot.sieve } postmaster_address = michael(a)michaelmarley.com protocols = imap sieve lmtp service auth { client_limit = 1624 unix_listener /var/spool/postfix/private/dovecot-auth { group = postfix mode = 0660 user = postfix } } service imap-login { inet_listener imaps { port = 0 } } service lmtp { process_min_avail = 5 unix_listener /var/spool/postfix/private/dovecot-lmtp { group = postfix mode = 0600 user = postfix } } ssl = required ssl_ca = </etc/ssl/private/COMODORSADomainValidationSecureServerCA.crt ssl_cert = </etc/ssl/private/michaelmarley.com.crt ssl_cipher_list = HIGH !RC4 !aNULL !eNULL !3DES @STRENGTH ssl_dh_parameters_length = 4096 ssl_key = # hidden, use -P to show it ssl_protocols = !SSLv3 userdb { driver = passwd } protocol lmtp { mail_plugins = " sieve" } protocol lda { mail_plugins = " sieve" } local_name matthewtmarley.com { ssl_ca = </etc/ssl/private/COMODORSADomainValidationSecureServerCA.crt ssl_cert = </etc/ssl/private/matthewtmarley.com.crt ssl_key = # hidden, use -P to show it } I tried recompiling Dovecot to see if that might do any good, but I got the same result. Also, I noticed that the "test-crypto" test in src/lib-dcrypt also crashes on exit (despite the fact that all the tests succeeded) with a very similar stacktrace: (gdb) run Starting program: /home/michael/Source/dovecot/core/src/lib-dcrypt/test-crypto [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". test_cipher_test_vectors ............................................. : ok test_cipher_aead_test_vectors ........................................ : ok test_hmac_test_vectors ............................................... : ok test_load_v1_keys .................................................... : ok test_load_v1_key ..................................................... : ok test_load_v1_public_key .............................................. : ok test_load_v2_key ..................................................... : ok test_load_v2_public_key .............................................. : ok test_get_info_v2_key ................................................. : ok test_gen_and_get_info_rsa_pem ........................................ : ok test_get_info_rsa_private_key ........................................ : ok test_get_info_invalid_keys ........................................... : ok test_get_info_key_encrypted .......................................... : ok test_get_info_pw_encrypted ........................................... : ok test_password_change ................................................. : ok test_load_invalid_keys ............................................... : ok 0 / 16 tests failed Program received signal SIGSEGV, Segmentation fault. 0x00007ffff75a5480 in ?? () (gdb) bt #0 0x00007ffff75a5480 in ?? () #1 0x00007ffff6ff7c60 in ossl_init_thread_stop (locals=<optimized out>) at crypto/init.c:336 #2 0x00007ffff6ff7ee4 in OPENSSL_cleanup () at crypto/init.c:391 #3 0x00007ffff78472e0 in ?? () from /lib/x86_64-linux-gnu/libc.so.6 #4 0x00007ffff784733a in exit () from /lib/x86_64-linux-gnu/libc.so.6 #5 0x00007ffff782d3f8 in __libc_start_main () from /lib/x86_64-linux-gnu/libc.so.6 #6 0x000055555555f68a in _start () If there is anything I have missed or if there is any other way I can help, please let me know. Thanks, Michael Marley

3 3

lazy-load SNI?
by Felipe Gasper 15 Nov '16

15 Nov '16

Hello, We’re rolling out large SNI deployments for our mail servers. Each domain gets an entry like this in the config: local_name mail.foo.com { ssl_cert = </ssl/domain_tls/*.foo.com/combined ssl_key = </ssl/domain_tls/*.foo.com/combined } There are a couple problems we’re finding with this approach: 1) Dovecot wants to load everything at once, which has some machines taking up many GiB of memory just for Dovecot. Is there any way to defer loading of an SSL cert until a client actually requests it? 2) Any time we add or remove a domain, Dovecot’s SNI config matrix needs to be rebuilt. Is there a way to handle SNI requests dynamically via some sort of configuration plugin, so we wouldn’t need to rebuild the config on domain add/remove? I looked through the docs but couldn’t see a way to do this. Thank you in advance! -Felipe Gasper Mississauga, ON

4 12