September 2024 - dovecot

Missing IMAP METADATA Plugin
by Adam Miller 23 Aug '25

23 Aug '25

OS: Ubuntu 22.04.1 Dovecot: 2.3.16 (7e2e900c1a) When trying to enable the “imap_metadata” plugin, I get the following error: Plugin 'imap_metadata' not found from directory /usr/lib/dovecot/modules I have searched high and low via Google, StackOverflow and even ChatGPT (haha) but I cannot seem to find the package that provides the plugin. Does Dovecot have to be built from source for this plugin to be included? Any help is appreciated. Thank you!

8 8

[Patch] Allow sieve redirects to specified domains only
by Kadlecsik József 10 Aug '25

10 Aug '25

Hello, The use case is simple: we have to forbid sieve redirects to outside, but should be able to redirect inside. The patch below adds a new sieve configuration option sieve_allowed_redirects to list the domains to which redirects are permitted and otherwise are not allowed. Please consider it. diff --git a/pigeonhole/doc/example-config/conf.d/90-sieve.conf b/pigeonhole/doc/example-config/conf.d/90-sieve.conf index 238bcf4..3b4ac65 100644 --- a/pigeonhole/doc/example-config/conf.d/90-sieve.conf +++ b/pigeonhole/doc/example-config/conf.d/90-sieve.conf @@ -126,6 +126,10 @@ plugin { # script execution. If set to 0, no redirect actions are allowed. #sieve_max_redirects = 4 + # The comma/space separated list of email domains where redirection is allowed. + # If the list is empty, there is no restriction. + sieve_allowed_redirects = + # The maximum number of personal Sieve scripts a single user can have. If set # to 0, no limit on the number of scripts is enforced. # (Currently only relevant for ManageSieve) diff --git a/pigeonhole/src/lib-sieve/cmd-redirect.c b/pigeonhole/src/lib-sieve/cmd-redirect.c index 6a3b0a4..17e9031 100644 --- a/pigeonhole/src/lib-sieve/cmd-redirect.c +++ b/pigeonhole/src/lib-sieve/cmd-redirect.c @@ -107,6 +107,24 @@ const struct sieve_action_def act_redirect = { .commit = act_redirect_commit, }; +static bool +cmd_redirect_allowed(struct sieve_instance *svinst, + const struct smtp_address *to_address) +{ + const char **domain; + bool result = TRUE; + + if (!svinst->allowed_redirects) + return TRUE; + for (domain = svinst->allowed_redirects; *domain != NULL; ++domain) { + if (strcasecmp(*domain, to_address->domain) == 0) { + return TRUE; + } + result = FALSE; + } + return result; +} + /* * Validation */ @@ -127,6 +145,12 @@ cmd_redirect_validate(struct sieve_validator *validator, if (!sieve_validator_argument_activate(validator, cmd, arg, FALSE)) return FALSE; + if (svinst->max_redirects == 0) { + sieve_command_validate_error(validator, cmd, + "local policy prohibits the use of a redirect action"); + return FALSE; + } + /* We can only assess the validity of the outgoing address when it is * a string literal. For runtime-generated strings this needs to be * done at runtime. @@ -148,14 +172,17 @@ cmd_redirect_validate(struct sieve_validator *validator, } } T_END; + if (!result && + !cmd_redirect_allowed(svinst, + sieve_address_parse_str(raw_address, &error))) { + sieve_command_validate_error(validator, cmd, + "local policy prohibits the use of a redirect action"); + return FALSE; + } + return result; } - if (svinst->max_redirects == 0) { - sieve_command_validate_error(validator, cmd, - "local policy prohibits the use of a redirect action"); - return FALSE; - } return TRUE; } @@ -238,6 +265,11 @@ cmd_redirect_operation_execute(const struct sieve_runtime_env *renv, "local policy prohibits the use of a redirect action"); return SIEVE_EXEC_FAILURE; } + if (!cmd_redirect_allowed(svinst, to_address)) { + sieve_runtime_error(renv, NULL, + "local policy prohibits the use of a redirect action"); + return SIEVE_EXEC_FAILURE; + } if (sieve_runtime_trace_active(renv, SIEVE_TRLVL_ACTIONS)) { sieve_runtime_trace(renv, 0, "redirect action"); diff --git a/pigeonhole/src/lib-sieve/sieve-common.h b/pigeonhole/src/lib-sieve/sieve-common.h index e79fb4d..d980c9d 100644 --- a/pigeonhole/src/lib-sieve/sieve-common.h +++ b/pigeonhole/src/lib-sieve/sieve-common.h @@ -209,6 +209,7 @@ struct sieve_instance { const struct smtp_address *user_email, *user_email_implicit; struct sieve_address_source redirect_from; unsigned int redirect_duplicate_period; + const char **allowed_redirects; }; /* diff --git a/pigeonhole/src/lib-sieve/sieve-settings.c b/pigeonhole/src/lib-sieve/sieve-settings.c index 47f70da..70addc7 100644 --- a/pigeonhole/src/lib-sieve/sieve-settings.c +++ b/pigeonhole/src/lib-sieve/sieve-settings.c @@ -267,4 +267,10 @@ void sieve_settings_load(struct sieve_instance *svinst) svinst->user_email = address; } } + svinst->allowed_redirects = NULL; + str_setting = sieve_setting_get(svinst, "sieve_allowed_redirects"); + if (str_setting != NULL && *str_setting != '\0') { + svinst->allowed_redirects = + (const char **)p_strsplit_spaces(svinst->pool, str_setting, ", "); + } } Best regards, Jozsef -- E-mail : kadlecsik.jozsef(a)wigner.hun-ren.hu PGP key: https://wigner.hu/~kadlec/pgp_public_key.txt Address: Wigner Research Centre for Physics H-1525 Budapest 114, POB. 49, Hungary

2 1

Panic with FTS_BACKEND_FLAG_TOKENIZED_INPUT and virtual mailboxes: file index-search-result.c: line 132
by markus-dovecot＠doits.de 30 May '25

30 May '25

Originally posted and debugged here: https://github.com/slusarz/dovecot-fts-flatcurve/issues/66 I have to following panic with fts flatcurve enabled: ``` Panic: file index-search-result.c: line 132 (index_search_result_update_flags): assertion failed: (result->search_args->args == &search_arg) Error: Raw backtrace: /usr/lib/dovecot/libdovecot.so.0(backtrace_append+0x46) [0x74e9890a2846] -> /usr/lib/dovecot/libdovecot.so.0(backtrace_get+0x22) [0x74e9890a2962] -> /usr/lib/dovecot/libdovecot.so.0(+0x10d81b) [0x74e9890af81b] -> /usr/lib/dovecot/libdovecot.so.0(+0x10d8b7) [0x74e9890af8b7] -> /usr/lib/dovecot/libdovecot.so.0(+0x5e2e2) [0x74e9890002e2] -> /usr/lib/dovecot/libdovecot-storage.so.0(+0x518a4) [0x74e9891d38a4] -> /usr/lib/dovecot/modules/lib20_virtual_plugin.so(virtual_storage_sync_init+0x274e) [0x74e988bf8a9e] -> /usr/lib/dovecot/libdovecot-storage.so.0(mailbox_sync_init+0x5c) [0x74e9891eff9c] -> /usr/lib/dovecot/libdovecot-storage.so.0(mailbox_sync+0x39) [0x74e9891f0039] -> dovecot/imap(cmd_select_full+0x1df) [0x62354bd8326f] -> dovecot/imap(command_exec+0xa4) [0x62354bd8b444] -> dovecot/imap(+0x22392) [0x62354bd89392] -> dovecot/imap(+0x22444) [0x62354bd89444] -> dovecot/imap(client_handle_input+0x1bd) [0x62354bd8989d] -> dovecot/imap(client_input+0x74) [0x62354bd89e64] -> /usr/lib/dovecot/libdovecot.so.0(io_loop_call_io+0x6d) [0x74e9890c603d] -> /usr/lib/dovecot/libdovecot.so.0(io_loop_handler_run_internal+0x13a) [0x74e9890c77aa] -> /usr/lib/dovecot/libdovecot.so.0(io_loop_handler_run+0x54) [0x74e9890c60e4] -> /usr/lib/dovecot/libdovecot.so.0(io_loop_run+0x40) [0x74e9890c62a0] -> /usr/lib/dovecot/libdovecot.so.0(master_service_run+0x17) [0x74e989036ad7] -> dovecot/imap(main+0x570) [0x62354bd7ad20] -> /lib/x86_64-linux-gnu/libc.so.6(+0x29d90) [0x74e988c29d90] -> /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0x80) [0x74e988c29e40] -> dovecot/imap(_start+0x25) [0x62354bd7ade5] ``` It happens when accessing virtual folder with filters like this: ``` something-*/excep-dev* header x-abc-exception "1" ``` The error happens on subsequent access of the folder: - Removing dovecot.index{,.log} files allows to access the virtual folder again - dovecot.index gets recreated on access and the folder can be accessed multiple times - Marking an e-mail read triggers the error, though! Now each next access raises the error - Removing dovecot.index{,.log} restores access again and the e-mails are still marked as read (so this information is not lost), until the next e-mail is marked as read: then the folder cannot be accessed again Slusarz debugged the backtrace and wrote a detailed description at https://github.com/slusarz/dovecot-fts-flatcurve/issues/66#issuecomment-229… which I'll quote here: > This is not a flatcurve bug - it is a bug within Dovecot's indexing/fts handling. > > (Detailed debug below, so that core team can understand the problem.) > > As discussed above, this bug is triggered with virtual mailboxes on subsequent accesses. The initial access will work fine; the assert crash only happens when the Dovecot lib-storage index code reads the indexes created for the virtual mailboxes. > > `index_search_result_update_appends()` in lib-storage/index/index-search-result.c > > * Inside this function, a "temporary search parameter to limit the search only to the new messages" is added to the mail search args. Later, it is checked that the mail_search args points to this same search parameter after the search is executed. It is this assert check that is failing, because the mail args is changing during the course of the search. > > * `i_assert(result->search_args->args == &search_arg);` > > > The search args are being altered in plugins/fts/fts-search.args.c, `fts_search_args_expand()`. Specifically, in that function, either `fts_search_args_expand_language_top_level()` or `fts_search_args_expand_tree()` is being called, which can replace the current args parameter with a new pointer. > > `fts_search_args_expand()` is being called in plugins//fts/fts-search.c by `fts_search_try_lookup()` when the FTS backend has the FTS_BACKEND_FLAG_TOKENIZED_INPUT flag set. Flatcurve has this flag set. > > This flag is defined in plugins/fts/fts-api-private.h. Comments say: > > ``` > /* Tokenize all the input. update_build_more() will be called a single > directly indexable token at a time. Searching will modify the search > args so that lookup() sees only tokens that can be directly > searched. */ > ``` > > Thus, the comments explicitly say the search args will be modified if this flag is set. This modification is incompatible with the assert check in `index_search_result_update_appends()`, which requires that the search args not be modified. > > The core team will have to determine whether this behavior is correct, or if the assert check is invalid. You can find some more discussion at https://github.com/slusarz/dovecot-fts-flatcurve/issues/66, but I think the most important part (the very detailed analysis from Slusarz, thanks again!) might be enough to understand and maybe fix the problem. Relevant information about the mail system: ``` # 2.3.21 (47349e2482): /etc/dovecot/dovecot.conf # Pigeonhole version 0.5.21 (f6cd4b8e) # OS: Linux 6.5.0-45-generic x86_64 Ubuntu 22.04.4 LTS [...] namespace { location = virtual:~/virtual:LAYOUT=maildir++ prefix = _/ separator = / } [...] plugin { acl = vfile acl_shared_dict = file:/var/lib/dovecot/db/shared fts = flatcurve fts_autoindex = yes fts_filters = normalizer-icu snowball stopwords fts_filters_en = lowercase snowball stopwords fts_languages = de en fts_tokenizer_email_address = maxlen=100 fts_tokenizer_generic = algorithm=simple maxlen=30 fts_tokenizers = generic email-address sieve = file:~/sieve;active=~/.dovecot.sieve } [...] ``` I can provide more information if you need. Thanks for looking into it!

2 2

Quota wrongfully reported as exceeded on shared mailbox
by adminkram＠tripelspark.de 07 May '25

07 May '25

Hi, I've been scratching my head without luck with the following problem: I've shared a mailbox from one user "COMPANY" to another user "JOE". Both users have quotas and both quotas are less than 50% used. When the user JOE tries to post a message into the shared folder, dovecot complains that the user would be over quota: bash# /usr/lib/dovecot/imap -u JOE 1 append shared/COMPANY/Sent/JOE {10} a NO [OVERQUOTA] Quota exceeded (mailbox for user is full) (0.057 + 0.000 + 0.057 secs). However, nobody is over quota: bash# doveadm quota get -u COMPANY Quota name Type Value Limit % User quota STORAGE 4754996 10485760 45 User quota MESSAGE 33695 - 0 bash# doveadm quota get -u JOE Quota name Type Value Limit % User quota STORAGE 14 1048576 0 User quota MESSAGE 3 - 0 Some more infos: 1. Dovecot-2.3.21 (+Debian updates) 2.3.21 (47349e2482) (NOTE: dovecot.org/mailing-lists says: Use dovecot –version to get it. The command is dovecot --version, though!) 2. I've performed "doveadm quota recalc" on both users, no change. 3. I've checked sharing ACL permissions, if I give just "l" permission, the response will be "NO [NOPERM] Permission denied". It does not seem to be an ACL permission reported wrongly. 4. I've tried increasing the quota size beyond reason (100 Gbytes for JOE), no change. 5. I've restarted dovecot after each change. 6. When trying to move a message into the shared Sent/joe folder in Thunderbird, same problem happens, Thunderbird displays the over quota message. Thunderbird cannot display the quota info for the shared folder, though. 7. I tried to enable rawlog (rawlog_dir in protocol imap). It does nothing helpful. It logs the manual commands from my pre-auth sessions (imap -u JOE) but it will stubbornly not log any commands issued by Thunderbird. ?! Any idea how to further debug this or what could be wrong? Thank you, Wolfgang

2 1

Stats-writer permission denied
by Ken Wright 15 Apr '25

15 Apr '25

I've got a new Dovecot error (well, new to me, at least)! I ran tail /var/log/mail.err and got the following: net_connect_unix(/var/run/dovecot/stats-writer) failed: Permission denied I checked the file, and it's owned by www-data and the www-data group, with read and write permissions for owner and group... just like I have specified in /etc/dovecot/conf.d/10-master.conf. I can't understand why permission is denied when the file permissions appear to be correct. Can anyone shed any light on this for me? Ubuntu 20.04 Dovecot 2.3.7.2 Ken

3 2

[Dovecot] auth: Error: net_connect_unix(anvil-auth-penalty) failed: Permission denied
by mailing＠securitylabs.it 03 Apr '25

03 Apr '25

Hello, just upgraded one of my testing machine from dovecot 1.2.15 to 2.0.7. After some trouble with permissions now all seems working, but when I stop dovecot I have these errors in log: Nov 9 17:08:07 in dovecot: master: Warning: Killed with signal 15 (by pid=20362 uid=0 code=kill) Nov 9 17:08:07 in dovecot: pop3-login: Error: read(anvil) failed: EOF Nov 9 17:08:07 in dovecot: imap-login: Error: read(anvil) failed: EOF Nov 9 17:08:07 in dovecot: imap(mailing(a)securitylabs.it): Server shutting down. bytes=410/59361 Nov 9 17:08:07 in dovecot: imap(mailing(a)securitylabs.it): Server shutting down. bytes=79/26605 Nov 9 17:08:07 in dovecot: imap(mailing(a)securitylabs.it): Server shutting down. bytes=802/55708 Nov 9 17:08:07 in dovecot: imap(mailing(a)securitylabs.it): Server shutting down. bytes=258/1242 Nov 9 17:08:07 in dovecot: auth: Error: read(anvil-auth-penalty) failed: EOF Nov 9 17:08:07 in dovecot: auth: Error: net_connect_unix(anvil-auth-penalty) failed: Permission denied dovecot -n: # 2.0.7: /usr/local/etc/dovecot/dovecot.conf # OS: Linux 2.6.32-5-xen-686 i686 Debian squeeze/sid auth_cache_negative_ttl = 2 mins auth_cache_size = 1000 M auth_cache_ttl = 2 mins auth_master_user_separator = * auth_username_chars = abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ01234567890.-_@& disable_plaintext_auth = no dotlock_use_excl = yes first_valid_uid = 100 last_valid_uid = 105 lock_method = dotlock log_timestamp = "%Y-%m-%d %H:%M:%S " login_greeting = Server ready. mail_fsync = never mail_gid = Debian-exim mail_location = maildir:~/Maildir:INDEX=/var/indexes/%d/%n mail_privileged_group = Debian-exim mail_uid = Debian-exim 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 namespace { inbox = yes location = prefix = separator = / type = private } namespace { location = maildir:~/Maildir/expunged prefix = EXPUNGED/ separator = / type = private } namespace { location = maildir:~/Maildir/deleted prefix = DELETED/ separator = / type = private } namespace { location = maildir:~/Maildir/deleted/expunged prefix = DELETED/EXPUNGED/ separator = / type = private } passdb { args = /usr/local/etc/dovecot-sql.conf driver = sql } passdb { args = /usr/local/etc/passwd.masterusers driver = passwd-file master = yes pass = yes } plugin { lazy_expunge = EXPUNGED/ DELETED/ DELETED/EXPUNGED/ mail_log_events = delete expunge mail_log_group_events = quota = maildir quota_warning = storage=80%% /usr/local/bin/dovecot-quota-warning.sh 80 quota_warning2 = storage=90%% /usr/local/bin/dovecot-quota-warning.sh 90 sieve = ~/.dovecot.sieve } pop3_no_flag_updates = yes protocols = imap pop3 service auth { unix_listener auth-userdb { group = Debian-exim mode = 0600 user = Debian-exim } } service imap-login { client_limit = 256 process_limit = 128 process_min_avail = 3 service_count = 1 } service imap { drop_priv_before_exec = yes process_limit = 256 vsz_limit = 256 B } service pop3-login { client_limit = 256 process_limit = 128 process_min_avail = 3 service_count = 1 } service pop3 { drop_priv_before_exec = yes process_limit = 256 vsz_limit = 256 B } ssl_cert = </usr/local/etc/dovecot.crt ssl_key = </usr/local/etc/dovecot.key userdb { args = /usr/local/etc/dovecot-sql.conf driver = sql } protocol imap { mail_max_userip_connections = 10 mail_plugins = " notify quota imap_quota mail_log lazy_expunge" } protocol pop3 { mail_max_userip_connections = 3 mail_plugins = " notify quota mail_log" pop3_uidl_format = %08Xu%08Xv } protocol lda { mail_plugins = quota sieve postmaster_address = postmaster(a)securitylabs.it } My /usr/local/var/run/dovecot directory: 4 drwxr-sr-x 4 root staff 4096 Nov 9 17:08 . 4 drwxrwsrwx 3 root staff 4096 Nov 9 17:05 .. 0 srw------- 1 root staff 0 Nov 9 17:05 anvil 0 srw------- 1 root staff 0 Nov 9 17:05 anvil-auth-penalty 0 srw------- 1 root staff 0 Nov 9 17:05 auth-client 0 srw------- 1 dovecot staff 0 Nov 9 17:05 auth-login 0 srw------- 1 root staff 0 Nov 9 17:05 auth-master 0 srw------- 1 Debian-exim Debian-exim 0 Nov 9 17:05 auth-userdb 0 srw------- 1 dovecot staff 0 Nov 9 17:05 auth-worker 0 srw------- 1 root staff 0 Nov 9 17:05 config 0 srw------- 1 root staff 0 Nov 9 17:05 dict 0 srw------- 1 root staff 0 Nov 9 17:05 director-admin 0 srw-rw-rw- 1 root staff 0 Nov 9 17:05 dns-client 0 srw------- 1 root staff 0 Nov 9 17:05 doveadm-server 0 lrwxrwxrwx 1 root staff 35 Nov 9 17:05 dovecot.conf -> /usr/local/etc/dovecot/dovecot.conf 4 drwxr-xr-x 2 root root 4096 Nov 9 17:05 empty 4 drwxr-x--- 2 root dovenull 4096 Nov 9 17:05 login My emails are delivered as user Debian-exim and users are virtual in a MySQL DB. This system is behind a proxy (so the master user)

3 2

Bug: Logout counters reset after recreating hibernated process
by Eirik Rye 12 Mar '25

12 Mar '25

Hello, # dovecot --version 2.3.20 (80a5ac675d) When using hibernation (`imap_hibernate_timeout = 10s`), counters such as deleted/expunged messages, appear to be reset after the user process is recreated. Steps to reproduce: 1. Configure IMAP hibernation 2. Login with IMAP client 3. Delete+expunge a message 4. IDLE, and wait 10s for process to be hibernated 5. Unhibernate process (`DONE`), and log out 6. Observe the logged line: 2023-09-05T17:36:26+0200 imap(<redacted>)<776933><jA2nYJ4EPdcqAVtAAADwAAutur4AAAAB>: Info: Disconnected: Logged out in=9 out=88 deleted=0 expunged=0 trashed=0 hdr_count=0 hdr_bytes=0 body_count=0 body_bytes=0 The expected output is "deleted=1 expunged=1", not "deleted=0 expunged=0". If you follow the steps above, but skip the IDLE/hibernate step, the output is as expected. Interestingly, if the client is disconnected during hibernation (without recreating the process), the output from `imap-hibernate` is still as expected: 2023-09-05T17:45:15+0200 imap-hibernate(<redacted>)<770540><0qE0gJ4Emt8qAVtAAADwAAutur4AAAAB>: Info: Disconnected: Connection closed in=97 out=1020 deleted=1 expunged=1 trashed=0 hdr_count=0 hdr_bytes=0 body_count=0 body_bytes=0 This leads me to believe the counters are lost during process recreation, not during hibernation. This is an issue as customers will often come to us asking why messages have disappeared in their inbox, or asking for help to identify which client is expunging messages. Right now, this information is lost if hibernation is used and the client is hibernated after expunging. Best regards, Eirik Rye

2 2

2.3.21.1 and oauth
by v＠minakov.pro 14 Nov '24

14 Nov '24

Hello! I'm using dovecot with keycloak for oauth authentication. My config is: client_id = dovecot client_secret = MY_SECRET introspection_url = https://MY_KEYCLOAK/realms/master/protocol/openid-connect/token/introspect introspection_mode = post pass_attrs = pass=%{oauth2:access_token} Everything worked great on version 2.3.20. After upgrading to version 2.3.21.1 oauth stopped working with errors: auth: Debug: http-client: conn [::1]:443 [1]: Got 401 response for request [Req1: POST https://MY_KEYCLOAK/realms/master/protocol/openid-connect/token/introspect] Unauthorized (took 5 ms + >Aug 16 00:23:58 auth: Error: oauth2(MY_EMAIL,127.0.0.1,<HADwdb8f+PN/AAAB>): oauth2 failed: Introspection failed: No username returned I tried all combination of configurations, debugging and versions of keycloak. Seems something was broken in this version of dovecot.

3 3

What is the appropriate action for out of memory error for exceeding vsz_limit in my situation?
by Steve Dondley 08 Nov '24

08 Nov '24

Hi, I have sporadic messages on my server about an account getting an out of memory error: Fatal: master: service(imap): child 17910 returned error 83 (Out of memory (service imap { vsz_limit=325 MB }, you may need to increase it) - set CORE_OUTOFMEM=1 environment to get core dump) In response, I raised the limit from 256MB up to 325. That didn’t seem to help. The server has only 512 MB of RAM so I’m afraid to go much higher. I googled around a bit and some people are saying that when the dovecot.index.cache grows too large, it can cause these errors. The imap account has a dovecot.index.cache file that is 167 MB which seems to be well below the limit (but maybe it is compressed?). So I’m not entirely sure as to what the underlying problem is. So before upgrading to the server to 1024 MB of ram and going on a wild goose chase, I just want to get input on whether I have positively identified the problem and whether there might be something else I can do to manage this problem.

2 3

sieve crashes when two optional include scripts are missing
by Michal Hlavinka 27 Oct '24

27 Oct '24

Hi, we've found an issue with sieve plugin. When script includes two missing optional scripts, it causes crash. For reproducing the issue, script names must belong to same hash table node, so either two scripts with same name or scripts where 'hash(name) % table->size' results in same value, like missing_a and missing_aa in reproducer below. Reduced reproducer: cat >dovecot.min.conf <<EOF mail_location = maildir:~/Maildir plugin { sieve = file:/tmp;active=/tmp/dovecot.sieve } EOF cat >reproducer.sieve <<EOF require "include"; include :personal :optional "missing_a"; include :personal :optional "missing_aa"; EOF echo -n > empty sieve-test -c dovecot.min.conf reproducer.sieve empty any test mail would do, but empty file is enough for reproducer # return sblock->id causes the crash as sblock is NULL backtrace: (gdb) bt #0 sieve_binary_block_get_id (sblock=0x0) at src/lib-sieve/sieve-binary.c:421 #1 ext_include_execute_include (renv=0x5555555bbed8, include_id=1, flags=EXT_INCLUDE_FLAG_OPTIONAL) at plugins/include/ext-include-common.c:696 #2 opc_include_execute (renv=0x5555555bbed8, address=0x5555555bbf10) at plugins/include/cmd-include.c:399 #3 sieve_interpreter_operation_execute (interp=0x5555555bbe88) at src/lib-sieve/sieve-interpreter.c:901 #4 sieve_interpreter_continue (interp=0x5555555bbe88, interrupted=0x0) at src/lib-sieve/sieve-interpreter.c:959 #5 sieve_interpreter_start (interp=0x5555555bbe88, result=0x5555555b9230, interrupted=0x0) at src/lib-sieve/sieve-interpreter.c:1049 #6 sieve_interpreter_run (interp=0x5555555bbe88, result=0x5555555b9230) at src/lib-sieve/sieve-interpreter.c:1057 #7 sieve_run (sbin=0x55555558ee18, result=0x5555555b9230, eenv=0x7fffffffe890, ehandler=0x55555558c338) at src/lib-sieve/sieve.c:357 #8 sieve_test (sbin=0x55555558ee18, msgdata=0x7fffffffe9f0, senv=0x7fffffffea30, ehandler=0x55555558c338, stream=0x5555555b2cd0, flags=SIEVE_EXECUTE_FLAG_LOG_RESULT) at src/lib-sieve/sieve.c:598 #9 main (argc=5, argv=0x555555567a10) at src/sieve-tools/sieve-test.c:391 issue seems to be caused by lib-sieve/plugins/include/ext-include-common.c: ext_include_generate_include on first pass (script) the check for script already compiled into binary at 529: included = ext_include_binary_script_get_include_info(binctx, script); returns NULL, so code goes through 'else' block where it hits sieve_script_is_open(script) check. It includes script with NULL block, returns 0 so caller cmd_include_generate does not emit anything. On second pass the above 'included' check returns first missing script, it goes through true block where it just checks flags and returns non-zero, so caller this time goes through the emit code block and the previously included NULL sblock gets referenced causing crash later. Note: despite scripts have different name, they pass the "same script" check because sieve_file_script_equals compares 0==0 dev & inode numbers as scripts do not exist. This is why scripts do not have to have same names. Side effect of this is that as missing_a and missing_aa are treated like the same script, only first one is linked in compiled svbin. Which means that if missing_aa reappears later, svbin won't get recompiled. Another manifestation of the issue is when included missing_a script is present when main script is compiled and later missina_a is removed again. During execution, script is not recompiled and it aborts for hash table dupe check: sieve_binary_open->ext_include_binary_script_include->hash_table_insert There are several ways how to solve this. I've tried a few, but there were other corner cases that appeared. One tried way was for example ext_include_generate_include in the 'if (included)' replicate the check sieve_script_is_open(...) from else block, just check flags, no include and set result to 0. Another way to solve this is to return from ext_include_execute_include early if sblock is NULL just before it tries to derefernce it, similarly as it does if it fails 'once' check. Same for dumping in opc_include_dump. Whilst the first approach seems a bit better, it won't stop crashing in existing environments as it won't trigger svbin recompilation and svbin would fail. This means that the second soultion is sort of needed anyway. This still won't solve the problem with different missing scripts as only first one would be able to trigger recompilation when file becomes available. Just including both in the mentioned function won't work either as they would still be evaluated as same and hash table anti-dupe check would trigger abort. So some changes in script comparison seem needed anyway. This is what I've used in my approach. Return early from ext_include_execute_include and opc_include_dump if sblock is NULL and change sieve_file_script_equals to compare also script->location for missing scripts (when inode is 0). Let me know if you need more information. Cheers, Michal Hlavinka

2 1