The major goal is to simplify certificate management in ejabberd.
Currently it requires some effort from a user to configure certficates,
especially in the situation where a lot of virtual domains are hosted.
The task is splitted in several sub-tasks:
* Implement basic certificate validator. The validator should check all
configured certificates for existence, validity, duration and so on. The
validator should not perform any actions in the case of errors except
logging an error message. This is actually implemented by this commit.
* All certificates should be configured inside a single section (something
like 'certfiles') where ejabberd should parse them, check the full-chain,
find the corresponding private keys and, if needed, resort chains and
split the certficates into separate files for easy to use by fast_tls.
* Options like 'domain_certfile', 'c2s_certfile' or 's2s_certfile' should
probably be deprecated, since the process of matching certificates with the
corresponding virtual hosts should be done automatically and these options
only introduce configuration errors without any meaningful purpose.
The changes are very similar to those from previous commit:
* Now there is no need to pass validating function in
gen_mod:get_opt() and gen_mod:get_module_opt() functions,
because the modules' configuration keeps already validated values.
* New functions gen_mod:get_opt/2 and gen_mod:get_module_opt/3 are
introduced.
* Functions gen_mod:get_opt/4 and get_module_opt/5 are deprecated.
If the functions are still called, the "function" argument is
simply ignored.
* Validating callback Mod:listen_opt_type/1 is introduced to validate
listening options at startup.
The commit introduces the following changes:
* Now there is no need to pass validating function in
ejabberd_config:get_option() functions, because the configuration
keeps already validated values.
* New function ejabberd_config:get_option/1 is introduced
* Function ejabberd_config:get_option/3 is deprecated. If the function
is still called, the second argument (validating function) is simply
ignored.
* The second argument for ejabberd_config:get_option/2 is now
a default value, not a validating function.
Don't let ejabberd_c2s close the session and unset presence if a
'c2s_terminated' callback stops hook execution, as is done in
mod_stream_mgmt:c2s_terminated/2 on resumption.
Fixes#1680.
Now 'From' and 'To' arguments must be omitted in functions
and structures related to routing.
The commit deprecates the following functions:
ejabberd_router:route/3 in favor of ejabberd_router:route/1
ejabberd_router:route_error/4 in favor of ejabberd_router:route_error/2
ejabberd_local:route_iq/4 in favor of ejabberd_local:route_iq/2
ejabberd_local:route_iq/5 in favor of ejabberd_local:route_iq/3
The format of {route, From, To, Packet} is changed in favor of {route, Packet}
If set to 'true' (this is the default), new processes spawned by
ejabberd_listener will be attached to the corresponding supervisor.
No such processes will be attached to a supervisor otherwise.
Setting this to 'false' will improve performance of high loaded
systems where new C2S/S2S processes are spawned very rapidly.
Check whether the 'from' and 'to' attributes are valid before bouncing
or resending a stanza from the stream management queue. They might be
invalid in certain corner cases.
Thanks to Evgeniy for spotting this.
Close the connection if a stream management client fails to respond to
an acknowledgement request within 60 seconds. This number of seconds
can be changed with the new "ack_timeout" option, and the mechanism can
be disabled by specifying 'infinity'.
As a side effect of this change, a new acknowledgement is no longer
requested before the response to the previous request is received.
The CSI state is always set to 'active' when a stream management session
is resumed; so there's no need to apply the CSI state of the old c2s
process, first.
If session resumption failed because requesting the #state from the old
c2s process took too long, the new c2s process will usually receive the
response. Let the new process handle that case gracefully.
During stream resumption, the #state is transferred from the old c2s
process to the new one. This is usually very fast, but under certain
conditions, it can take longer than five seconds.
If a message stanza is blocked as per XEP-0016 or XEP-0191 and the
stanza is marked as a private MUC message, don't return an error. This
makes sure users won't be kicked from MUC rooms when blocking other
participants.
Let mod_client_state handle the queueing of stanzas, not just their
classification. This simplifies the ejabberd_c2s code and gives
(custom) CSI modules more flexibility.
When stanzas are bounced from the stream management queue (because the
session timed out or was closed for some other reason), use a different
error message so that this situation can be distinguished from other
cases.
As per XEP-0016 and XEP-0191, return a service-unavailable error when an
incoming last activity query was blocked by a privacy list (just as we
do for other IQ requests).
If the "resend_on_timeout" option is set to 'if_offline' and a pending
stream management session is terminated because a new session is opened
by the same resource (while no other resource is online), resend
unacknowledged messages rather than bouncing error messages.
If a stream management session times out for a user who appears to be
using MAM, drop any unacknowledged messages rather than resending or
bouncing them. This avoids duplicates or bogus error messages.
However, this is only done if the new mod_mam option "assume_mam_usage"
is set to 'if_enabled' or 'on_request'. In the former case, a user is
assumed to be using MAM if archiving is enabled for his account. In the
latter case, MAM usage is assumed only if archiving was explicitly
requested by the client, or if archiving was enabled by means of
mod_mam's "request_activates_archiving" option.
