mpm_event: follow up to r1666468 and r1666618. We don't need to return 0 in the compare function, but for debugging purpose which we could implement later if necessary (in a separate function). For now, keep the function simple as in 2.4.x to ease backport, and add a comment about why we never return 0 here.

mpm_event: follow up to r1666468. We only need one compare function for add semantic with apr_skiplist_insert() and unique timers (pointers). It also should work with apr_skiplist_remove() and apr_skiplist_find(), be they used some day.

mpm_event: Allow for timer events duplicates. Meanwhile ap[r]_skiplist_add()...

core: If explicitly configured, use the KeepaliveTimeout value of the virtual host which handled the latest request on the connection, or by default the one of the first virtual host bound to the same IP:port. For non-async MPMs, use either r->server's or c->base_server's value in ap_process_http_sync_connection() depending on a new server_rec's flag called keep_alive_timeout_set and determined at config time. For event MPM, use a queue per timeout value, chaining the queues per type (keepalive wrt KeepAliveTimeout, write completion wrt to Timeout) so that maintenance can be done on all the queues from the head, and such that insertions/maintenance remain in O(1). A server config is created and pointing to the queue of each vhost at post_config time, hence the config can be associated to the connection state (cs) at post_read_request time (keep_alive_timeout_set is used to determine r->server vs c->base_server here), and we can simply insert with TO_QUEUE_INSERT(cs->sc->q, cs). PR56226. While at it, since each queue now embeds it own timeout and hence the expiration_time of the cs has changed to a queue_timestamp (the time it was queued), we can detect clock skews and expire entries immediatly if the system is set (eg. far) in the past during runtime and we want to avoid waiting for (eg.) centuries before the current logic kills them. Any entry which is registered above now + q->timeout is concerned, and is now cleaned from the queue when encountered. PR57374.

mpm_event: follow up to r1538490, r1545736, r1604350, r1639614 and r1642868. Clear and restore sbh in notify_suspend() and notify_resume() respectively.

mpm_event: follow up to r1493741 and r1662440. Really fix style by not introducing tabs...

mpm_event: follow up to r1493741. Fix typo (and style) for event_unregister_socket_callback()'s final result.

Update conn_rec.id when a new thread begins working on a connection, because the old thread may work on a new connection and assign the same ID in parallel. Submitted By: Michael Thorpe Committed By: covener

mpm_event(opt): avoid casts/comparisons from unsigned to signed (atomics).

mpm_event: follow up to r1639614 (trunk) and r1642858 (2.4.x). Notify suspend in any case when starting lingering close (blocking), and outside the lingering queue's critical section.

mpm_event(opt), mpm_worker, mpm_prefork: follow up to r1635521, r1640161. Retain num_buckets and max_buckets accross restarts so that we can determine whether new buckets were allocated and set their idle_spawn_rate at the same level as the existing ones (max). Also, adjust ap_daemons_limit and ap_daemons_to_start lower bounds at mpm_run() time, once num_buckets is available for the current generation (previously done at check_config time, hence before num_buckets is computed, and then with the previous generation's value).

mpm_event(opt), mpm_worker: follow up to r1635521. Allocate retained idle_spawn_rate vector on the process pool.

event, eventopt: follow up to r1638879 and r1639960. Clear the pool in ap_push_pool() before recycling.

don't call notify_suspend() in a worker thread after start_lingering_close_common may have put the socket back into the pollset. If it becomes readable too quickly, cs can be free'ed or accessed concurrently.

revert r1638691, more comprehensive followup planned shortly.

avoid dereferencing a recently apr_pool_clear()'ed event_conn_state_t *cs in several paths where ptrans is being recycled at the end of a request.

Fix a bug in r1604350 that can lead to crashes in the event MPM under load. if start_lingering_close_blocking() returns 0, notify_suspend() will write a 1 into some potentially recently free'd memory from ptrans (cs->suspended).

MPMs, core: make duplicated listeners (SO_REUSEPORT) introduced in r1599531 less intrusive. Submitted by: Yingqi Lu <yingqi.lu@intel.com> Modified/Committed by: ylavic Add ListenCoresBucketsRatio which is a configurable ratio between the number of CPU cores (online) and the number of listeners buckets to create, defaulting to zero (so that listeners buckets become an opt-in, ie. ncpus / ratio > 1). This could also be made an opt-out by using the previous hardcoded value (8) as default. Make ap_close_listeners() act on all the listeners (including duplicated ones), since the function is also called externally (eg. mod_cgid, mod_ssl_ct and possibly any third party module) to cleanup opened descriptors when a process is forked (the duplicated listeners are kept in a scoped/static variable). Add ap_close_listeners_ex() to close a single bucket of listeners, used by the children to close unused duplicates and internally by ap_close_listeners(). Make ap_duplicate_listeners() compute the number of buckets to be used, instead of each MPM. This number is now based on the above ratio and will not change unless asked to (given *num_buckets < 1, that is when the MPM does not run in one-process mode nor after a graceful restart). Remove some global variables (mpm_listen, enable_default_listeners) previously used to communicate between MPMs and ap_listen, since ap_duplicate_listeners() API can now be used to do so. Also rename num_buckets as ap_num_listen_buckets, and prefix have_so_reuseport with ap_ (both printed by ap_log_common(), hence kept global). Detect ap_have_so_reuseport once only at startup. Restore dummy_connection() as before r1599531 since sending POD signals should not depend on the number of listeners buckets (there is still one single socket receiving the connections). For each MPM (concerned), move the bucket data (pod, listeners and eventually accept mutex) into a struct and instanciate an array of them (sized by the number of buckets), for each child to use its own data according to its bucket index, and the parent to maintain the whole.

core: follow up to r1629909: fix min_spare_threads lower bound and check wrt num_buckets.

core: ensure that MPMs return an error on runtime failure and hence that httpd's main process also exits with an error.

mpms: enforce assertion that unreachable code is not reached.

core: Use process scoreboard to store each child's listener bucket, and silently adjust the configured number of processes/threads to be above the computed number of listener buckets (depending on the CPU cores).

event: Fix worker-listener deadlock in graceful restart caused by get_worker() allocating new worker after ap_queue_info_term(), but not setting the have_idle_worker variable. PR 56960. Submitted By: Zin UDA Committed By: jkaluza

event: have_idle_worker must not be cleared in every listener_thread iteration. Fixes bug when workers were not stopped after graceful restart (introduced in r1605328).

mod_proxy_wstunnel: Concurrent websockets messages could be lost or delayed with ProxyWebsocketAsync enabled. Submitted By: Edward Lu Committed By: covener

Avoid C99 idiom. Followup to r1604466.

Revert r1601943, r1602058, r1605307 (socket callback returning pollfds) Revert r1605369 (wstunnel refactoring) per discassion on dev@httpd.a.o thread Message-ID: <CALK=YjN9HfThP_k_rF9iJPcUhcQk9sNOaDPxZNG+HcJNWS4ZAQ@mail.gmail.com>

event: minify local variables scope.

follow up r1601943: Each apr_pollset_poll modified *out_pfd, so we need to copy it.

followup to r1604350, move the c->sbh assignment to immediately before the socket is added to the timeout queue. Technically not needed because we hold the lock on the timeout queue, but more consistent with other blocks that leave the thread. Submitted By: Edward Lu <Chaosed0 gmail com> Committed By: covener

event: Avoid timer's mutex two successive locks/unlocks with unrelated (and optional) work in between. Also saves declaration/usage of pointer 'ep' by reusing available 'te' instead, and removes duplicated 'te' declaration in a local block.

event: follow up to r1604347. Don't leak the timer event when it is cancelled.

missed a case in r1538490: PR56639 Always NULL c->sbh before putting a connection back in a pollset or queue. We can't NULL c->sbh at the bottom of process_socket() after putting a socket back on the event_pollset or having it go into lingering close, because the listener or a worker thread could A) continue on the connection or B) free and allocate the same conn_rec pointer before we get to the bottom of process_socket(). Submitted By: Edward Lu Committed By: covener

Do not call a timeout callback after the asociated socket woke and the socket callback was called. PR 56642

Mark all unused/duplicated/closed child listeners as inactive. Fix unused "ap_listen_rec *lr;" in worker_run().

style fix, no functional change.

eventMPM: fix SEGV in the case socket callback was registered and no timeout function was registered.

Socket event callback now takes pollfd. User (e.g. mod_proxy_wstunnel) can know which socket is ready.

mpm_event[opt]: Send the SSL close notify alert when the KeepAliveTimeout expires. PR54998.

-MPMs event & eventopt: child listeners are used in ONE_PROCESS mode too.

Break out common code to be share-able

fold in performance hack from eventopt

Now allow MPMs to call ap_log_common()

Optimize w/ duplicated listeners and use of SO_REUSEPORT where available.

Fix some comment typos

PR56333: Add an API to resume a connection that a handler has previously suspended. Submitted by: Artem <artemciy gmail.com>, Edward Lu <Chaosed0 gmail.com> Committed by: covener

Extend the socket callbacks in event to allow a timeout on the I/O callback. When a socket callback has a timeout, an associated timer event is used to remove the sockets from the pollset and call a timeout function. * This includes a noteworthy change to the main event loop. Previously, we would call epoll, then process the timer events, then iterate through the poll results. After this patch, the timer events are processed before the poll() a _non-queued_ action can change the pollset conents (a users timed callback function conversely could easily sit in a queue while the main thread continues down into epoll) * timer events can now have sockets associated with them, those sockets are removed from the pollset when the timer event fires w/o a queue to the worker. * timer events now have a canceled flag that can be toggled without locking the timer list. * Drop the severity of some wstunnel messages from DEBUG to TRACE1 * Lift the restriction on using asynchronous websockets connections but having an idle timeout

r1494157 from eventopt for event, per sf's review comments. use a subpool of pchild since skiplist allocations will happen across threads and are only protected from other skiplist operations.

follow-up to r1546759: remove an inadvertently committed testcase

Add suspend_connection and resume_connection hooks to notify modules when the thread/connection relationship changes. (Currently implemented only for the Event MPM; should be implemented for all async MPMs.)

fix whitespace in a debug message

Follow-up to r1538490: Ensure that conn_rec->sbh is always cleared when returning from process_socket().

Use a normalized offset point for idlers... still need to worry that atomics work as "expected", in this case that a add32 of a -1 is the "same" as dec32 (as far as effect on idlers)

revert 1544876

Just critically warn people... but keep pushing through.

c->sbh can be unexpectedly NULL when the thread that pulls the ready keepalive connection out of the queue laps the thread that put it on the queue.

axe unused autodata+heap (clang scan-build)

skiplist is now in apr 1.5 and above

In event's process_lingering_close, close the socket and remove it from the ring as soon as we see any status that isn't APR_EAGAIN or APR_SUCCESS, instead of waiting specifically for EOF.

Fold in Eric Covener's socket callback

update log tags replace duplicate, add missing

core, mod_ssl: Lift the restriction that prevents mod_ssl taking full advantage of the event MPM. Enable the ability for a module to reverse the sense of a poll event from a read to a write or vice versa.

Remove some useless declarations that were shadowing other local variables of the same name.

rv uninitialized -- the failing checks don't set rv, so don't pass it to ap_log_error.

As long as we have skiplist, use it.

Log a warning if ThreadStackSize has an inappropriate value. PR : 54311 Submitted by: Tianyin Xu <tixu cs ucsd edu> Only event.c has been compiled on my system. The 2 others are just cut and paste. I have left the same value for APLOGNO in event and eventopt has for the other logged messages. I don't know if it is the right way to do.

Check that AsyncRequestWorkerFactor is not negative PR :54254 Submitted by: Jackie Zhang <jackie qq zhang gmail com>

fdq expects a certain behavior from atomics... ensure that the event mpms check this.

Prefer higher-mode polling, but allow for drop down to default

no longer needed

Pull out the "extended" pod functions used by event and worker to core, since it will be used by Simple and Lean and likely other MPMs. Avoid duplication.

Fix spelling mistakes.

Follow up to r1363557: Refactor with some existing code to slightly optimize thread tracking. Fix an implementation problem affecting statically linked MPMs which would break the ability to fail restart if a resource shortage was found immediately. Found by: rjung

mpm_event, mpm_worker: Remain active amidst prevalent child process resource shortages. This is a somewhat different direction than r168182 ("transient thread creation errors shouldn't take down the whole server"). r168182: If APEXIT_CHILDSICK is received and there aren't any active children at the time, exit. Now: If APEXIT_CHILDSICK is received and we never successfully initialized a child, exit. The issue seen with the r168182 handling is that it is rather easy to be left with no active child processes (which causes the server to exit completely) during a resource shortage that lasts for some measurable period of time, as contrasted with a resource shortage that results in only a handful of allocation failures. Now the server will remain active, though as long as the resource shortage exists children may continually fail and the parent will try once per second to create a replacement. The existing logic to reduce the spawn rate after such errors will prevent the parent from trying to create children more rapidly.

mpm_event, mpm_worker: Fix cases where the spawn rate wasn't reduced after child process resource shortages. The broken scenario: child X exits with APEXIT_CHILDSICK another child Y is created and reuses child X's scoreboard slot child X's exit status is processed The case that worked is when child X's scoreboard slot hadn't been reused by the time that its exit status was processed.

Don't count connections in lingering close state when calculating how many additional connections may be accepted

Keep track of the number of clogged, lingering, and suspended connections

Simplify handling of MaxConnectionsPerChild If we decrement the counter at accept time and not at close time, we only need to access the counter in the listener thread and the code gets much simpler.

Fix MaxConnectionsPerChild This was broken when the handling of lingering close was moved into the listener thread. - Make the connection counting thread safe. - Do the counting in the connection pool cleanup to ensure that it gets also executed if the listener thread closes the connection. - Add a trace log message when a process is recycled.

rename requests_this_child to conns_this_child, which is more accurate

Take care not to call ap_start_lingering_close from the listener thread, because it may block when flushing data to the client. From the listener thread, do a lingering close without flushing. This is OK because we only do this if there has been an error during write completion or if our send buffers are empty because we are in keep-alive. PR: 52229

Prevent listener thread from ever updating a worker's scoreboard slot The worker may be doing something else by now. This should take care of slots staying in "L" state in the scoreboard

Revert Paul's lock free circular queue changes done on MPM-event. The changes have been moved to branches/mpm-event-optimization. This puts trunk's MPM-event back to the state of 2.4.x, except for serf support. The following commits are reverted: ------------------------------------------------------------------------ r1203404 | trawick | 2011-11-17 23:48:35 +0100 event no longer requires APR_POLLSET_THREADSAFE, or any other non-standard feature (all but OS/2 have APR_POLLSET_WAKEABLE) config-foo: don't require thread-safe pollset in order to build event event.c: don't stress APR_ENOTIMPL in apr_pollset_create errors; just give the standard "check system or user limits" message ------------------------------------------------------------------------ r1203858 | trawick | 2011-11-18 22:39:33 +0100 follow up to r1202257 -- perform normal wakeup processing when APR_EINTR is seen from apr_pollset_poll(), with expectation that it was triggered by apr_pollset_wakeup() ------------------------------------------------------------------------ r1202395 | trawick | 2011-11-15 20:38:31 +0100 spellcheck r1202258 ------------------------------------------------------------------------ r1202329 | pquerna | 2011-11-15 18:47:33 +0100 Calculate the power of two size of the buffer before allocating it. Spotted by: Rüdiger Plüm ------------------------------------------------------------------------ r1202259 | pquerna | 2011-11-15 16:52:59 +0100 Use apr_pollset_wakeup to ensure that the listener thread will process most enqueue'd events quickly ------------------------------------------------------------------------ r1202258 | pquerna | 2011-11-15 16:52:00 +0100 Because the pollset is now only mutated from the event thread, we no longer need the APR_POLLSET_THREADSAFE flag for the event_pollset ------------------------------------------------------------------------ r1202257 | pquerna | 2011-11-15 16:51:03 +0100 Create a new lock free circular queue, and use it in the EventMPM to remove the timeout mutex that was wrapping both timeout queue operations and pollset operations. ------------------------------------------------------------------------ r1202256 | pquerna | 2011-11-15 16:50:09 +0100 Instead of disabling the listening sockets from the pollset when under load, just stop calling the accept call, but leave the sockets in the pollset.

Fix the case where AP_DEBUG is unset.

Fix warning: variable 'rv' set but not used.

Add lots of unique tags to error log messages

Remove MPM-private stuff from conn_state_t

follow up to r1202257 -- perform normal wakeup processing when APR_EINTR is seen from apr_pollset_poll(), with expectation that it was triggered by apr_pollset_wakeup()

garbage C's in server-status and extra child processes with async lingering close support in Event. these are set by the listener due to a residual c->sbh for a worker thread no longer associated with the connection. the associated worker threads are really idle and blocked in ap_queue_pop_something and will serve new work normally. but since they are marked as SERVER_CLOSING rather than SERVER_READY in the scoreboard, the parent process thinks they are busy and forks extra children. reproduction recipe: start Event with server-status enabled. hit localhost/server-status?refresh=5 and watch the C's trail behind the lone W.

event no longer requires APR_POLLSET_THREADSAFE, or any other non-standard feature (all but OS/2 have APR_POLLSET_WAKEABLE) config-foo: don't require thread-safe pollset in order to build event event.c: don't stress APR_ENOTIMPL in apr_pollset_create errors; just give the standard "check system or user limits" message

spellcheck r1202258

Use apr_pollset_wakeup to ensure that the listener thread will process most enqueue'd events quickly

Because the pollset is now only mutated from the event thread, we no longer need the APR_POLLSET_THREADSAFE flag for the event_pollset

Create a new lock free circular queue, and use it in the EventMPM to remove the timeout mutex that was wrapping both timeout queue operations and pollset operations.

Instead of disabling the listening sockets from the pollset when under load, just stop calling the accept call, but leave the sockets in the pollset.

Fix assertion failure during very high load by preventing race condition between appending to the timeout queues and adding to the pollset. We don't add additional locking calls but only extend the present calls to include the apr_pollset_add. Therefore this hopefully should not cause too much performance regression. Add some comments Replace two AP_DEBUG_ASSERTS with better error handling

clean up unused ap_my_generation variable (presence is odd, since this variable as an API was supposedly removed long ago, and a similarly named static variable was removed in r1078286)

Downgrade log message which is triggered when client disconnects

If a child is created just before graceful restart and then exits because of a missing lock file, don't shutdown the whole server. PR: 39311 Submitted by: Shawn Michael <smichael rightnow com>

Check the return value from ap_run_create_connection in mpm_event. Add a todo in mpm_simple PR: 41194 Submitted by: Davi Arnaut

Cleanup effort in prep for GA push: Trim trailing whitespace... no func change

Add wrappers for malloc, calloc, realloc that check for out of memory situations. Use them in most places where malloc, and friends are used. This results in clean error messages in an out of memory situation instead of segfaulting or silently malfunctioning. In some places, it just allows to remove some logging code. PR 51568, PR 51569, PR 51571.

Install signal handlers for crashes before detaching. Previously, there was a time span where crashes would be neither visible on the console nor logged in the error log.

Reduce period during startup and restart where a successive signal may be lost PR: 43696 Submitted by: Arun Bhalla <arun shme net>

Log the deprecation warning at level info, not warn

Fix search'n'replace error: Actually warn if MaxClients is used

New AsyncRequestWorkerFactor directive to influence how many connections will be accepted per process

Remove dead assignment, as pointed out by Chris Wilson <cwilson vigilantsw com>

Rename MaxClients to MaxRequestWorkers which describes more accurately what it does.

Some improvements for handling of many connections for MPM event: - Process lingering close asynchronously instead of tying up worker threads (based on patch by Jeff Trawick). - If the number of connections of a process is above threads_per_child + WORKER_OVERCOMMIT * (idle_workers - 1) (WORKER_OVERCOMMIT is fixed at 2, at the moment), or if all workers are busy, don't accept new connections in that process. Such a dynamic connection limit is necessary because we may have both async and non-async (ssl) connections. WORKER_OVERCOMMIT should be a config option. - Don't count idle workers of not-accepting processes against MinSpareThreads, so that the parent will spawn new processes when necessary. - If we receive a keep-alive request while all workers are busy, don't block but close the connection immediately so that the client will re-connect to a different process. Related changes: - Log what is going on at trace loglevels. - Remove the bypass_push poll type flag, this code cannot be hit anymore (if it ever could?). - Add some macro helpers for dealing with timeout queues.

If MaxMemFree is set, limit the number of recycled pools that is kept

Some logging tweaks

Formatting changes, reuse 'now' variable instead of declaring a new one in several blocks. No significant code change.

Fix graceful restart aborting connections PR: 43359 Submitted by: Takashi Sato <takashi lans-tv com>

Add child_status hook for tracking creation/termination of MPM child processes. Add end_generation hook for notification when the last MPM child of a generation exits. end_generation is implemented completely by core using the child_status hook run by the MPM. simple and mpmt_os2 MPMs don't currently run the child_status hook, so neither hook is invoked with those MPMs.

Simplify the interface to ap_reclaim_child_processes() and ap_relieve_child_processes(): instead of requiring the MPM to implement an otherwise-useless hook, just use a callback function. As I don't expect third-party MPM devs are following our day to day progress, the API changes are considered part of yesterday's MMN change.

mpm_common.c implements a handful of config directives; the related state has to be re-initialized to the default values at pre-config time, but that was handled inconsistently by the MPMs, resulting in unexpected behavior when some of the directives were removed across restart move that necessary initialization from the MPMs to common code run from core's pre-config; MPMs that need to override defaults can do so by running after core's pre-config (the NetWare MPM now does that) the DEFAULT_MAX_REQUESTS_PER_CHILD compile setting wasn't useful enough to keep the simple MPM wasn't consistent in which of these directives were respected and which weren't, and that hasn't changed (see procmgr.max_requests_per_child vs. ap_max_requests_per_child)

consolidate logic to remove the pidfile in a new API, ap_remove_pid(), related to existing ap_log_pid() and ap_read_pid() presumably this is useful to third-party MPMs as well

* server/mpm/prefork/prefork.c (set_signals): Ignore SIGXFSZ; have a write() beyond 2GB fail with E2BIG rather than killing the process in the now hopefully rare case of non-LFS file usage on LFS systems. * server/mpm/worker/worker.c (set_signals): Likewise. * server/mpm/event/event.c (set_signals): Likewise.

worker, event, prefork: Correct several issues when built as DSOs; most notably, the scoreboard was reinitialized during graceful restart, such that processes of the previous generation were not observable. Other noticeable ones: * the generation number did not increment properly * worker's check-config hook didn't detect startup properly, so messages about inconsistent directives could have been sent to the wrong place max_daemons_limit wasn't tracked across restart, but as the scoreboard was always cleared this was a non-issue. Reinitialization of idle_spawn_rate, hold_off_on_exponential_spawning, and maxclients-message-reported may or may not have been features, but the big picture is that the behaviour shouldn't have differed. These changes should have been part of r758185.

zap needless reinitialization of is_graceful on the 2nd call of the open-logs hook

sanity check use of volatile in prefork, worker, and event

event: child_main() never returns, so remove some dead code after one of the calls

Rename MaxRequestsPerChild to MaxConnectionsPerChild, which describes more accurately what the directive does. The old name still works but logs a warning.

Fix crash accessing pollset on worker thread when child process is exiting. The timeout mutex and pollset were allocated from the listener thread pool. During child process shutdown, the listener thread exits first while any outstanding requests finish. These objects need to be allocated from pchild since the lifetime extends until the last worker thread has finished. Switch to pchild, and move init of these objects to the same place as other thread-independent objects.

Re-add AP_DEBUG_ASSERTs that were wrongly removed in r982050

remove more unused variables revove some obsolete comments netware file is untested

Fix MPM event/libserf build problems: If libserf was installed during build, MPM event was picking up the headers but not linking with libserf, causing unresolved symbol errors. Now build MPM event with or without serf support depending on --with-serf being specified or not.

Use the new APLOG_USE_MODULE/AP_DECLARE_MODULE macros everywhere to take advantage of per-module loglevels

odd assortment of spelling fixes

all signs indicate that status value is always -1 on error

reduce the warning level in the bindprocessor() call used to make sure child processes aren't bound to the CPU the parent is running on. In recent levels of AIX, the parent isn't implicitly bound to a CPU (so the children don't inherit it), and when the AIX WPAR feature is enabled this call can return EPERM, even though the child process is already unbound.

Log command line on startup, so there's a record of command line arguments like -f. Suggested by Shaya Potter. [Dan Poirier] PR: 48752

axe some vestiges of Event's accept mutex

Enhance ap_hook_monitor to pass along a server_rec (in general the ap_server_conf) and tuck away some storage in there which may be useful as an opaque data pointer.

* Do better checking of pollset operations in various places to avoid segfaults. PR: 46467 Submitted by: Stefan Fritsch <sf sfritsch.de> Reviewed by: rpluem

no longer amusing. I'll still pay off if somebody does it legitimately, but I haven't been to Apachecon in ages so we'll have to work out a different delivery mechanism. With the stability of the 2.0.x and 2.2.x releases and Moore's Law in effect for the last decade, it might be possible...

main() can use ap_run_mpm() directly, so axe the old ap_mpm_run() function change the mpm hooks to return OK/DONE instead of 0/1

clean up apr_time_from_msec() compatibility macros * where needed, base on APR_VERSION_AT_LEAST() so we remember what this is about * where not, axe

mpm-query hook: distinguish between DECLINED and APR_ENOTIMPL so that . ap_mpm_query() can return APR_EGENERAL if called too early (for debugging a module) . some hypothetical module which implements the mpm-query hook can bypass the MPM with APR_ENOTIMPL

rename the new retained-data APIs to ap_retained_data_create()/ap_retained_data_get(), for better consistency with existing APIs (e.g., slotmem) don't bother changing the MMN, which doesn't reflect recent -dev changes anyway

the mpm_get_child_pid hook is unnecessary, as was the per-MPM MPM_CHILD_PID() macro which it replaced axe this new hook, and use ap_get_scoreboard_process() instead

Core can specify a platform-specific rewrite args hook. Use that on Unix. Windows continues to specify this in the MPM, at least until someone can separate out the MPM-specific bits.

Register rewrite_args hook. Now, dunno if that was intended, but at least allow to start the httpd

If serf is available, compile in driving the serf event loop from inside the Event MPM. Add a new MPM Query to determine if an MPM supports this.

The Event MPM is no longer experimental; move it out of that ghetto. Yank the few bits that implement the concept of an experimental MPM.