January 2022 - dovecot

Patch: assertion failed in doveadm fts lookup
by John Fawcett 09 Jan '22

09 Jan '22

Hi I'm reposting this patch, which I have been applying locally since I originally posted it. Hopefully it can be considered for inclusion in the official release. It was tested on dovecot 2.3.16 but applies against the latest version 2.3.17.1. This is the "assertion failed" that it addresses doveadm fts lookup -u john(a)voipsupport.it body "text to look for" doveadm(john(a)voipsupport.it): Panic: file mail-storage.c: line 2108 (mailbox_get_open_status): assertion failed: (box->opened) doveadm(john(a)voipsupport.it): Error: Raw backtrace: /usr/local/lib/dovecot/libdovecot.so.0(backtrace_append+0x42) [0x7ff624fa7072] -> /usr/local/lib/dovecot/libdovecot.so.0(backtrace_get+0x1e) [0x7ff624fa718e] -> /usr/local/lib/dovecot/libdovecot.so.0(+0xffc3b) [0x7ff624fb3c3b] -> /usr/local/lib/dovecot/libdovecot.so.0(+0xffc71) [0x7ff624fb3c71] -> /usr/local/lib/dovecot/libdovecot.so.0(+0x5536f) [0x7ff624f0936f] -> /usr/local/lib/dovecot/libdovecot-storage.so.0(+0x3fc31) [0x7ff6250bcc31] -> /usr/local/lib/dovecot/lib21_fts_solr_plugin.so(+0x540d) [0x7ff62244f40d] -> /usr/local/lib/dovecot/lib20_fts_plugin.so(fts_backend_lookup+0x4d) [0x7ff6246b624d] -> /usr/local/lib/dovecot/doveadm/lib20_doveadm_fts_plugin.so(+0x2e8c) [0x7ff62240ce8c] -> doveadm(+0x31ded) [0x55f0c7f3aded] -> doveadm(+0x32862) [0x55f0c7f3b862] -> doveadm(doveadm_cmd_ver2_to_mail_cmd_wrapper+0x22d) [0x55f0c7f3c6fd] -> doveadm(doveadm_cmd_run_ver2+0x4d8) [0x55f0c7f4d158] -> doveadm(doveadm_cmd_try_run_ver2+0x3a) [0x55f0c7f4d1aa] -> doveadm(main+0x1f6) [0x55f0c7f2b606] -> /lib64/libc.so.6(__libc_start_main+0xd5) [0x7ff624bc8b75] -> doveadm(_start+0x2e) [0x55f0c7f2ba0e] Aborted (core dumped) Here's the patch: diff -ur dovecot-2.3.17.1-orig/src/plugins/fts/doveadm-fts.c dovecot-2.3.17.1-new/src/plugins/fts/doveadm-fts.c --- dovecot-2.3.17.1-orig/src/plugins/fts/doveadm-fts.c 2021-12-03 12:48:47.000000000 +0100 +++ dovecot-2.3.17.1-new/src/plugins/fts/doveadm-fts.c 2022-01-09 01:33:23.396342025 +0100 @@ -47,6 +47,14 @@ i_array_init(&result.scores, 16); box = mailbox_alloc(info->ns->list, info->vname, 0); + mailbox_set_reason(box,"fts search"); + if (mailbox_open(box) < 0) { + i_error("Couldn't open mailbox: %s", + mailbox_get_last_internal_error(box, NULL)); + doveadm_mail_failed_error(ctx, MAIL_ERROR_TEMP); + return -1; + } + if (fts_backend_lookup(backend, box, ctx->search_args->args, FTS_LOOKUP_FLAG_AND_ARGS, &result) < 0) { i_error("fts lookup failed"); John

2 1

Non-user logins?
by Ken Wright 09 Jan '22

09 Jan '22

My Dovecot issues continue. Right now I see at least two issues: first, my logs consistently show non-users trying (and failing) to log in, and I'm still unable to log in from my email client (Evolution or Roundcube, either one). I'll post about the second issue later; right now I wonder why I'm getting so many non-users trying to log in. Am I the subject of concerted hacking attacks, or is there something else going on? Some of the attempted logins are more-or-less random names claiming to be @mydomain, but at least one is a username that's really on my server, to wit: Jan 7 22:52:01 grace dovecot: lmtp(776281): Error: lmtp-server: conn unix:pid=776262,uid=117 [3]: rcpt www-data(a)mydomain.com: Failed to lookup user www-data(a)mydomain.com: Internal error occurred. Refer to server log for more information. (Another quick question: which server log should I check?) So, if anyone can tell me what's going on with all these logins, I'd be much obliged! Ken

9 16

Patch: enhancements for solr/tika integration
by John Fawcett 09 Jan '22

09 Jan '22

Hi here's a patch with some enhancements that I am applying locally for solr/tika integration. Hopefully this can be considered for inclusion. I've tested up to 2.3.16 and this patch applies against latest version 2.3.17.1. The contents are: 1. Allow username and password in tika configuration (as is already supported for solr). For example: fts_tika = https://user:password@tika_host:tika_port/tika/tika This means that tika can be more easily located on a different server to dovecot. Locating Solr and Tika on separate servers can isolate the Dovecot server from issues that may happen on Solr or Tika (like out of memory). 2. Introduce an optional configurable limit on the size of emails whose message bodies will be indexed with solr via a new setting fts_max_size, to be put in the plugin section. For example plugin { ... fts_max_size = 2M ... } Message body indexing is skipped for messages that are larger than this setting. 3. Introduce an optional configurable limit on the size of emails whose attachments will be sent to tika for keyword extraction, via a new setting fts_max_size_tika, to be put in the plugin section. For example: plugin { ... fts_max_size_tika = 1M } I noticed that some very large attachments are being sent to tika and these often do not have any useful content to index. For simplicity the check is done on message size not individual or total attachment size. Here's the patch: diff -ur dovecot-2.3.17.1-orig/src/plugins/fts/fts-build-mail.c dovecot-2.3.17.1-new/src/plugins/fts/fts-build-mail.c --- dovecot-2.3.17.1-orig/src/plugins/fts/fts-build-mail.c 2021-12-03 12:48:47.000000000 +0100 +++ dovecot-2.3.17.1-new/src/plugins/fts/fts-build-mail.c 2022-01-09 01:33:23.398341998 +0100 @@ -17,6 +17,7 @@ #include "fts-filter.h" #include "fts-api-private.h" #include "fts-build-mail.h" +#include "settings-parser.h" /* there are other characters as well, but this doesn't have to be exact */ #define IS_WORD_WHITESPACE(c) \ @@ -25,6 +26,8 @@ wherever */ #define MAX_WORD_SIZE 1024 +static bool debug = FALSE; + struct fts_mail_build_context { struct mail *mail; struct fts_backend_update_context *update_ctx; @@ -34,6 +37,7 @@ buffer_t *word_buf, *pending_input; struct fts_user_language *cur_user_lang; + bool oversized_tika; }; static int fts_build_data(struct fts_mail_build_context *ctx, @@ -223,7 +227,7 @@ parser_context.user = mail_storage_get_user(storage); parser_context.content_disposition = ctx->content_disposition; - + parser_context.oversized_tika = ctx->oversized_tika; if (fts_parser_init(&parser_context, &ctx->body_parser)) { /* extract text using the the returned parser */ *binary_body_r = TRUE; @@ -496,7 +500,32 @@ bool binary_body; const char *error; int ret; - + uoff_t msg_size; + uoff_t fts_max_size = 0; + uoff_t fts_max_size_tika = 0; + const char * fts_max_size_setting; + const char * fts_max_size_tika_setting; + bool oversized_msg; + bool oversized_tika; + + fts_max_size_setting = mail_user_plugin_getenv(update_ctx->backend->ns->user, "fts_max_size"); + if (fts_max_size_setting != NULL) { + if (debug) i_debug("fts_max_size %s",fts_max_size_setting); + if (settings_get_size(fts_max_size_setting, &fts_max_size, &error) < 0) { + i_error("%s",error); + fts_max_size = 0; + } + if (debug) i_debug("fts_max_size (value) %"PRIuUOFF_T,fts_max_size); + } + fts_max_size_tika_setting = mail_user_plugin_getenv(update_ctx->backend->ns->user, "fts_max_size_tika"); + if (fts_max_size_tika_setting != NULL) { + if (debug) i_debug("fts_max_size_tika %s",fts_max_size_tika_setting); + if (settings_get_size(fts_max_size_tika_setting, &fts_max_size_tika, &error) < 0) { + i_error("%s",error); + fts_max_size_tika = 0; + } + if (debug) i_debug("fts_max_size_tika (value) %"PRIuUOFF_T,fts_max_size_tika); + } *may_need_retry_r = FALSE; if (mail_get_stream_because(mail, NULL, NULL, "fts indexing", &input) < 0) { if (mail->expunged) @@ -505,10 +534,21 @@ mailbox_get_last_internal_error(mail->box, NULL)); return -1; } - + oversized_msg = FALSE; + oversized_tika = FALSE; + i_stream_get_size(input,TRUE,&msg_size); + if (fts_max_size > 0 && msg_size > fts_max_size) { + i_info("Skipping message body indexing because size %"PRIuUOFF_T" exceeds setting fts_max_size %s",msg_size,fts_max_size_setting); + oversized_msg = TRUE; + } + if (fts_max_size_tika > 0 && msg_size > fts_max_size_tika) { + i_info("Skipping message attachment indexing because size %"PRIuUOFF_T" exceeds setting fts_max_size_tika %s",msg_size,fts_max_size_tika_setting); + oversized_tika = TRUE; + } i_zero(&ctx); ctx.update_ctx = update_ctx; ctx.mail = mail; + ctx.oversized_tika = oversized_tika; if ((update_ctx->backend->flags & FTS_BACKEND_FLAG_TOKENIZED_INPUT) != 0) ctx.pending_input = buffer_create_dynamic(default_pool, 128); @@ -563,7 +603,7 @@ message_decoder_set_return_binary(decoder, TRUE); body_part = TRUE; } else { - if (skip_body) + if (skip_body ||oversized_msg) continue; } @@ -597,7 +637,7 @@ else (void)fts_parser_deinit(&ctx.body_parser, NULL); } - if (ret == 0 && body_part && !skip_body && !body_added) { + if (ret == 0 && body_part && !skip_body && !oversized_msg && !body_added) { /* make sure body is added even when it doesn't exist */ block.data = NULL; block.size = 0; ret = fts_build_body_block(&ctx, &block, TRUE); diff -ur dovecot-2.3.17.1-orig/src/plugins/fts/fts-parser.h dovecot-2.3.17.1-new/src/plugins/fts/fts-parser.h --- dovecot-2.3.17.1-orig/src/plugins/fts/fts-parser.h 2021-12-03 12:48:47.000000000 +0100 +++ dovecot-2.3.17.1-new/src/plugins/fts/fts-parser.h 2022-01-09 01:33:23.399341985 +0100 @@ -10,6 +10,7 @@ /* Can't be NULL */ const char *content_type; const char *content_disposition; + bool oversized_tika; }; struct fts_parser_vfuncs { diff -ur dovecot-2.3.17.1-orig/src/plugins/fts/fts-parser-tika.c dovecot-2.3.17.1-new/src/plugins/fts/fts-parser-tika.c --- dovecot-2.3.17.1-orig/src/plugins/fts/fts-parser-tika.c 2021-12-03 12:48:47.000000000 +0100 +++ dovecot-2.3.17.1-new/src/plugins/fts/fts-parser-tika.c 2022-01-09 01:33:23.400341972 +0100 @@ -57,7 +57,7 @@ tuser = p_new(user->pool, struct fts_parser_tika_user, 1); MODULE_CONTEXT_SET(user, fts_parser_tika_user_module, tuser); - if (http_url_parse(url, NULL, 0, user->pool, + if (http_url_parse(url, NULL, HTTP_URL_ALLOW_USERINFO_PART, user->pool, &tuser->http_url, &error) < 0) { i_error("fts_tika: Failed to parse HTTP url %s: %s", url, error); return -1; @@ -148,6 +148,10 @@ if (tika_get_http_client_url(parser_context->user, &http_url) < 0) return NULL; + if (parser_context->oversized_tika) { + i_info("skipping tika parser due to oversized message"); + return NULL; + } if (http_url->path == NULL) http_url->path = "/"; @@ -159,6 +163,11 @@ http_url->host.name, t_strconcat(http_url->path, http_url->enc_query, NULL), fts_tika_parser_response, parser); + if (http_url->user != NULL) { + http_client_request_set_auth_simple( + http_req, http_url->user, http_url->password); + } + http_client_request_set_port(http_req, http_url->port); http_client_request_set_ssl(http_req, http_url->have_ssl); if (parser_context->content_type != NULL) John

1 0

Patch: safeguard against too large value for uid being sent to solr for single mailbox searches
by John Fawcett 09 Jan '22

09 Jan '22

Hi I'm reposting a patch for solr integration which I have been applying locally. It applies against 2.3.17.1. Dovecot already has a mechanism when doing solr fts searches on multiple mailboxes that prevents a too large value for maximum rows being sent to solr. #define SOLR_MAX_MULTI_ROWS 100000 (defined and used in plugins/fts-solr/fts-backend-solr.c) However on single mailbox searches Dovecot uses the last uid of the mailbox, which can be too large for solr. My suggested patch is to also limit maximum rows to return on single mailbox searches. Here is the patch: diff -ur dovecot-2.3.17.1-orig/src/plugins/fts-solr/fts-backend-solr.c dovecot-2.3.17.1-new/src/plugins/fts-solr/fts-backend-solr.c --- dovecot-2.3.17.1-orig/src/plugins/fts-solr/fts-backend-solr.c 2021-12-03 12:48:47.000000000 +0100 +++ dovecot-2.3.17.1-new/src/plugins/fts-solr/fts-backend-solr.c 2022-01-09 01:33:23.401341959 +0100 @@ -837,7 +837,7 @@ str = t_str_new(256); str_printfa(str, "wt=xml&fl=uid,score&rows=%u&sort=uid+asc&q=%%7b!lucene+q.op%%3dAND%%7d", - status.uidnext); + I_MIN(status.uidnext,SOLR_MAX_MULTI_ROWS)); prefix_len = str_len(str); if (solr_add_definite_query_args(str, args, and_args)) { John

1 0

Can't log in from Evolution or Roundcube
by Ken Wright 09 Jan '22

09 Jan '22

When I try to log in to my mail server (ubuntu 20.04, Postfix 3.4.13, Dovecot 2.3.7.2) I get a response saying "Source stream returned no data”. At least to me, that's not particularly informative. Is it any more informative to anyone else? Ken

3 10

Sv: banning
by Sebastian Nielsen 08 Jan '22

08 Jan '22

I would say, lock accounts to for example IP address, ASN or GeoIP. This can be accomplished simply by a custom login handler, which also checks IP against database. And first time users, and those who change country/ISP/IP have to simply logon to a web interface (where 2FA can be required and also Captcha) and add their IPs/ASNs/Geo's. For a larger user base, I would recommend GeoIP and no web interface, save country of first login to database, and subsequent logins must originate from same country. Users that want to reset have to contact support. If you are a web hotel who only sell service to a specific country at all, lock the ports in firewall to that GeoIP. For smaller user base, like 50-100 users, I would recommend locking to ASN and providing a web interface where multiple ASN can be added. So people syncing from mobile and home can succeed. For very small user base, like 10's of users, just plan lock to IP. By connecting IP to accounts, you greatly reduce the attack surface.

1 0

TLS connection closed unexpectedly
by Ken Wright 08 Jan '22

08 Jan '22

I reinstalled Dovecot this evening, because I haven't been able to receive emails recently. After configuring and adjusting for the use of PostfixAdmin, I get the following error message in Evolution:; Failed to open folder. The reported error was "Failed to authenticate: TLS connection closed unexpectedly". Does anyone recognize this error? Ken

6 8

Update from 2.3.16 to 2.3.17 Received invalid SSL certificate: unable to get certificate CRL
by Salatiel Filho 08 Jan '22

08 Jan '22

Hi, after updating from dovecot.x86_64 2:2.3.16-2 to 2:2.3.17.1-1 on AlmaLinux 8 my dovecot replication stopped working. Log is full of Error: doveadm server disconnected before handshake: Received invalid SSL certificate: unable to get certificate CRL: /CN=my.domain.redacted (check ssl_client_ca_* settings?) No configuration has changed on my side. My certificate is a wildcard certificate ( not sure if related ). I am also replicating over tcps. Anyone else facing this issue? Atenciosamente/Kind regards, Salatiel

1 0

Sieve scripts igniored
by Sam R 07 Jan '22

07 Jan '22

Hello, I have an issue with Sieve plugin, I can drive rules remotely with sieve:// myserver.fr:4190/?tls=YES, the rules are well created under /home/vmail/name.firstname, but these are never executed. I have also read https://doc.dovecot.org/configuration_manual/sieve/troubleshooting/, but I don't see any evident mistake. here is my dovecot -n : # 2.3.13 (89f716dc2): /etc/dovecot/dovecot.conf # Pigeonhole version 0.5.13 (cdd19fe3) # OS: Linux 5.10.0-8-amd64 x86_64 Debian 11.0 ext4 # Hostname: mail.xxx.fr auth_debug = yes auth_gssapi_hostname = $ALL auth_krb5_keytab = /etc/dovecot/dovecot.keytab auth_mechanisms = login plain gssapi mail_location = maildir:/home/vmail/%n/Maildir/ mail_privileged_group = mail managesieve_notify_capability = mailto managesieve_sieve_capability = fileinto reject envelope encoded-character vacation subaddress comparator-i;ascii-numeric relationa l regex imap4flags copy include variables body enotify environment mailbox date index ihave duplicate mime foreverypart extracttex t 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 } passdb { driver = pam } plugin { sieve = file:~/sieve;active=~/.dovecot.sieve } protocols = " imap lmtp sieve pop3" service lmtp { executable = lmtp -L unix_listener /var/spool/postfix/private/dovecot-lmtp { group = postfix mode = 0666 user = postfix } } service managesieve-login { inet_listener sieve { port = 4190 } } ssl = required ssl_cert = </etc/letsencrypt/live/mail.xxx.fr/fullchain.pem ssl_client_ca_dir = /etc/ssl/certs ssl_client_ca_file = /etc/letsencrypt/live/mail.xxx.fr/chain.pem ssl_dh = # hidden, use -P to show it ssl_key = # hidden, use -P to show it userdb { driver = passwd } userdb { args = uid=vmail gid=vmail home=/home/vmail/%u allow_all_users=yes driver = static } userdb { driver = passwd } protocol lmtp { mail_plugins = " sieve" } protocol lda { mail_plugins = " sieve" } Thanks for helping. kind regards. Samuel

3 3

feature-request: make received-header on submission optional or at least drop the ip in it
by dc-ml＠dvl.werbittewas.de 07 Jan '22

07 Jan '22

hi. we're just trying to re-structure a system and want to use dovcots submission-ability. because the github-repository have no entry for any issues at all, we're sending this feature-fequest to this ML. unfortunately there doesn't seem to be any option to drop the received-header at all or at least hide the IP of the user. because the client-ip is covered by GDPR as personal data and so it should never shown to others without a certain reason, we want to hide it, but there's no "submission_add_received_header"-option like for lmtp and it doesn't seem to be any other option to hide this information. maybe this is possible to implement in near future? thanks a lot. best regards d.

8 18