manoj 99/09/17 16:26:54
Modified:pthreads/src/main http_main.c
Log:
Instead of blocking all signals in child_main, we were unblocking them;
this fixes that. Reported by Dave Colasurdo <[EMAIL PROTECTED]>.
Revision ChangesPath
1.98 +1 -1 apache-apr/pthreads/s
manoj 99/07/29 13:11:43
Modified:pthreads/src/include httpd.h
pthreads/src/main http_main.c
Log:
Support non-threaded operation, for the most part. The code still
references pthread calls, but they are essentially no-ops in the
non-threaded case; these can be #de
manoj 99/07/18 19:02:02
Modified:pthreads/src/main http_main.c
Log:
We want to deal with SIGTERM properly until we've set up the sigwait
thread, so put back the standard signal handler. This will also be
useful once support for unthreaded mode is put back in.
Revision Cha
manoj 99/07/16 12:45:49
Modified:pthreads/src/main http_main.c
Log:
Ralf noted that accept_mutex_child_init() has to be done before changing
the child's UID for platforms using the flock() mutex. Fix it in this
tree.
Revision ChangesPath
1.93 +2 -1 apache-
manoj 99/06/08 20:14:59
Modified:pthreads/src/include http_conf_globals.h
pthreads/src/main http_main.c
Log:
Clean up some unused variables
Revision ChangesPath
1.9 +0 -2 apache-apr/pthreads/src/include/http_conf_globals.h
Index: http_conf
manoj 99/05/26 00:12:32
Modified:pthreads/src/main http_main.c
Log:
Cleanup and dead code removal.
Revision ChangesPath
1.88 +15 -73apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
==
manoj 99/05/26 00:09:19
Modified:pthreads/src/main http_main.c
Log:
Move a little piece of code so it isn't run as root
Revision ChangesPath
1.87 +1 -2 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
=
manoj 99/05/26 00:06:32
Modified:pthreads/src/main http_main.c
Log:
When a child dies, we now mark it dead after wait_or_timeout() in the
parent; marking it dead in the child itself is somewhat incorrect and
unnecessary.
Revision ChangesPath
1.86 +0 -7 apa
manoj 99/05/24 13:33:35
Modified:pthreads/src/main http_main.c
Log:
Since pthread_join isn't used anymore in the server, create all of our
threads in detached state.
Revision ChangesPath
1.85 +5 -1 apache-apr/pthreads/src/main/http_main.c
Index: http_ma
manoj 99/05/23 23:29:45
Modified:pthreads/src/main http_main.c
Log:
Minor cleanup
Revision ChangesPath
1.83 +2 -5 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
===
RCS file:
manoj 99/05/23 20:07:03
Modified:pthreads/src/main http_main.c
Log:
Get rid of the last vestiges of SIGWINCH handling in child processes.
Now, each worker thread knows when it's the last one to exit, and in
that case, sends SIGTERM to kick the sigwait thread.
Revision Cha
manoj 99/05/19 21:33:33
Modified:pthreads/src/main http_main.c
Log:
Bug fix. Make sure that perform_idle_server_maintanence runs through
whole scoreboard.
Revision ChangesPath
1.80 +2 -2 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
manoj 99/04/29 13:09:45
Modified:pthreads/src/include http_accept.h
pthreads/src/main http_main.c
Log:
Add a per-thread pool, and make all worker threads use it instead of
pchild, for thread-safety
Revision ChangesPath
1.7 +1 -0 apache-apr/pt
manoj 99/04/21 21:14:41
Modified:pthreads/src/main http_main.c
Log:
We were setting max_daemons_limit based on the last fully living
children. Instead, we should set it based on the last non-dead child.
Revision ChangesPath
1.77 +3 -1 apache-apr/pthreads/src
manoj 99/04/19 18:43:05
Modified:pthreads/src/main http_main.c
Log:
Fix comment wrapping.
Revision ChangesPath
1.76 +5 -5 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
===
R
manoj 99/04/19 18:31:04
Modified:pthreads/src/main http_main.c
Log:
Bug fix for scoreboard handling. Before, if only a single child thread
was dead, we assumed that the whole child was dead and reused its slot,
even if there are other threads still running in that process slot.
manoj 99/04/16 22:10:19
Modified:pthreads/src/main http_main.c
Log:
Note in the scoreboard when a child process dies abnormally, so that a
substitute can be started.
Revision ChangesPath
1.74 +3 -3 apache-apr/pthreads/src/main/http_main.c
Index: http_ma
manoj 99/04/15 13:01:43
Modified:pthreads/src/main http_main.c http_accept.c
Log:
Enable graceful restart and MaxRequestsPerChild for the multiaccept
model, using a pipe watched by poll() to wake up the server at the
child's graceful shutdown time.
Revision ChangesPat
manoj 99/03/22 12:38:04
Modified:pthreads/src/main http_main.c
Log:
Revision ChangesPath
1.66 +22 -12apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
===
RCS file: /home/cvs/ap
rbb 99/03/22 12:10:42
Modified:pthreads/src/main http_main.c
Log:
Small abstraction patch. This also closes a small hole. We want to kill
accept threads before we kill worker threads.
Revision ChangesPath
1.65 +29 -8 apache-apr/pthreads/src/main/http_main
manoj 99/03/21 02:51:46
Modified:pthreads/src/main http_main.c http_protocol.c
Log:
Add a comment related to the previous patch, and rewrap another one.
Revision ChangesPath
1.64 +2 -0 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
manoj 99/03/21 02:42:36
Modified:pthreads/src/main http_main.c http_protocol.c
Log:
Make MaxRequestsPerChild work in the threaded server again (well at
least it works like the process server works). This consists of a few
changes:
- We need to actually count requests. In t
rbb 99/03/19 06:02:04
Modified:pthreads/src/main http_main.c
Log:
Added some error logging logic.
Revision ChangesPath
1.62 +5 -5 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
=
manoj 99/03/16 13:26:27
Modified:pthreads/src/main http_main.c
Log:
Allow the server to exit in single-process (-X) mode when we hit Ctrl-C.
This involves making the server respond to SIGINT like it does with
SIGTERM in the signal-handling thread.
Revision ChangesPath
rbb 99/02/23 12:37:27
Modified:pthreads/src/main http_main.c
Log:
When I get a graceful restart, I have to kick acceptors out of blocking on the
full queue. Otherwise, they will sit there, and they won't quit quickly
enough.
Revision ChangesPath
1.54 +3 -0
rbb 99/02/19 12:35:42
Modified:pthreads/src/main http_main.c
Log:
Logic to serialize per socket, instead of one mutex for the whole server.
Probably not optimal yet, but this will still improve our performance.
Revision ChangesPath
1.50 +245 -178 apache-apr/pt
rbb 99/02/18 10:40:57
Modified:pthreads/src/main http_main.c
Log:
Just putting back in a void in a function def. Minor change, but easy enough
to put it back.
Revision ChangesPath
1.48 +1 -1 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
rbb 99/02/18 10:17:42
Modified:pthreads/src/main http_main.c
Log:
This patch keeps our server accepting connections over restarts. It basically
puts back in logic from 1.3 that was taken out because of other bugs getting
in the way of our restarts keeping their sockets avail
rbb 99/02/18 10:06:51
Modified:pthreads/src/main http_main.c
Log:
Logic for single thread signal handler.
Submitted by: Manoj Kasichainula
Revision ChangesPath
1.46 +78 -59apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
=
manoj 99/02/17 21:43:52
Modified:pthreads/src/main http_main.c
Log:
More SIGUSR1 -> SIGWINCH changes.
Revision ChangesPath
1.45 +4 -4 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
===
manoj 99/02/17 21:39:53
Modified:pthreads/src/main http_main.c
Log:
During graceful restart we close our listeners to kick our threads out of
accept(), we seem to be missing the first file descriptor, probably because
of an errant cut-and-paste.
I think this bug didn't sho
rbb 99/02/17 14:39:44
Modified:pthreads/src/main http_main.c
Log:
This fixes a stupid mistake I made earlier. Too much coding the last few days
Revision ChangesPath
1.43 +2 -2 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
==
manoj 99/02/17 12:28:57
Modified:pthreads/src/main http_main.c
Log:
First, fix unbalanced locking and unlocking of the exit_mutex (which
will hopefully go away when we switch to a single signal-handling
thread). Also get rid of a malloc-free pair.
Revision ChangesPath
rbb 99/02/17 10:42:03
Modified:pthreads/src/main http_main.c
Log:
This patch fixes a few different things. The first, is that if
MaxRequestsPerChild was set to 0 in the config file, the child would start
and stop immediately. This fixes that. The rest are mostly syntax err
rbb 99/02/17 08:28:04
Modified:pthreads/src/main http_main.c
Log:
Part two of the restart patches. This fixes the underlying problem of the
thread getting stuck in the accept_mutex code, by allowing the user out if
errno == EINTR. This patch also separates clean_child_exit
rbb 99/02/17 07:29:52
Modified:pthreads/src/main http_main.c
Log:
Removing Manoj's two patches from last night, because they go after a symptom,
not the cause of the problem. This is the first of a series of commits to
fix the underlying problem of graceful restarts. This i
manoj 99/02/17 00:13:16
Modified:.STATUS
pthreads/src/main http_main.c
Log:
Start to remove pthread calls from clean_child_exit to make it legal for
a signal handler. There is still a pthread_cancel loop, which seems to
work sometimes for killing hanging
manoj 99/02/16 21:32:02
Modified:pthreads/src/main http_main.c
Log:
pthread calls while handling a signal are bad voodoo. This is another way
to allow cleanups to run after all threads exit.
Revision ChangesPath
1.37 +11 -10apache-apr/pthreads/src/main/ht
manoj 99/02/16 16:03:43
Modified:pthreads/src/main http_main.c
Log:
Get rid of some more debugging messages which were clogging the error log.
Revision ChangesPath
1.36 +0 -2 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
==
manoj 99/02/16 16:01:54
Modified:pthreads/src/main http_main.c
Log:
Eep. ptrans existed in the process scope. Now there is one ptrans per
thread, and it is sub-pooled off the correct "super-pool" now. This
fixes some very confusing coredumps, which were confusing because there
rbb 99/02/16 13:37:33
Modified:pthreads/src/main http_main.c
Log:
Added back the accept_mutex_init line. Don't know how I forgot it in my first
patch for inter-process serialization.
Revision ChangesPath
1.34 +1 -0 apache-apr/pthreads/src/main/http_main.c
rbb 99/02/16 10:42:02
Modified:pthreads/src/main http_main.c
Log:
Fix compile errors associated with last patch.
Submitted by: Manoj and Ryan
Revision ChangesPath
1.33 +2 -4 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
==
rbb 99/02/16 10:31:58
Modified:pthreads/src/main http_main.c
Log:
Code to properly finish a child process. This makes sure the child process
calls the proper module functions, and sets the scoreboard to SERVER_DEAD
for all threads.
Submitted by: Manoj Kasichainula and Rya
rbb 99/02/16 07:34:04
Modified:pthreads/src/main http_main.c
Log:
First pass at accept loop serialization. Right now, it is one mutex for the
whole server. This should be one mutex per socket, but I just wanted to get
the basic code in, I'll clean it up later.
Revision
manoj 99/02/15 23:10:29
Modified:pthreads/src/main http_main.c
Log:
Clean up a nasty memory leak. Now apache-apr uses a per-connection pool
(ptrans) just like Apache 1.3. Previously, pchild was used, and in my
tests, there was anywhere from 12-24k of memory lost per request.
manoj 99/02/15 22:09:47
Modified:pthreads/src/main http_main.c
Log:
When we close the accept socket, we should also kill the cleanup that
goes with it.
Revision ChangesPath
1.28 +1 -1 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
===
rbb 99/02/15 13:04:22
Modified:pthreads/src/main http_main.c
Log:
This change sets up the acceptor threads to get the thread-id when they become
ready. More importantly, the scoreboard is more accurate, because it tells
us when the thread is starting, not just ready to accep
rbb 99/02/15 12:39:02
Modified:pthreads/src/include http_conf_globals.h httpd.h
scoreboard.h
pthreads/src/main http_main.c scoreboard.c
Log:
Scoreboard updates. We now store the number of workers and acceptors in the
scoreboard file. I a
rbb 99/02/15 10:42:51
Modified:pthreads/src/main http_main.c scoreboard.c
Log:
A cuple of fixes for those static variables from the last scoreboard patch.
This creates functions that return the value of those variables, and sets
them back to static.
Revision Changes
manoj 99/02/14 23:48:37
Modified:pthreads/src/main http_main.c
Log:
Fix a buglet. Worker threads didn't note in the scoreboard when they
were back in SERVER_READY state.
Revision ChangesPath
1.22 +4 -3 apache-apr/pthreads/src/main/http_main.c
Index: htt
rbb 99/02/14 19:12:02
Modified:pthreads/src/main http_main.c
Log:
This change should solve the problem of spawning new child server's at the
first opportunity. I have tested it as much as I can tonight, but I will try
to break it again tomorrow morning when I have more comput
manoj 99/02/09 23:59:40
Modified:pthreads/src/main http_main.c
Log:
Two pieces that are closely intertwined (or at least seem to be at 3AM).
First, there is now a means for the parent to kill off a child process
at will. One thread is assigned the duty of listening to signals a
manoj 99/02/09 23:24:34
Modified:pthreads/src/main http_main.c
Log:
Put back ap_sync_scoreboard_image. A no-op in most cases, but it'll be
needed.
Revision ChangesPath
1.13 +9 -0 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
rbb 99/02/08 05:58:38
Modified:pthreads/src/main http_main.c
Log:
Make the final acceptor thread be the main thread for the process. This
change
means we have one fewer thread per process than we used to, but the thread we
are removing was a wasted thread anyway.
Revis
manoj 99/02/07 22:47:43
Modified:pthreads/src/main http_main.c
Log:
Put back -t and -S functionality (config test and virtual host config dump)
Revision ChangesPath
1.10 +9 -1 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
=
rbb 99/02/05 12:23:12
Modified:pthreads/src/main http_main.c
Log:
Added back the code to make the server detach from the console.
Revision ChangesPath
1.8 +69 -0 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
=
stoddard99/02/03 10:29:30
Modified:pthreads/src/main http_main.c
Log:
Get rid of stupid comment
Revision ChangesPath
1.6 +1 -1 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
===
rbb 99/01/28 13:54:08
Modified:pthreads/src/main http_main.c
Log:
Ignore SIG_PIPE, so server doesn't go down when a user stops their browser.
Revision ChangesPath
1.4 +4 -0 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
=
rbb 99/01/26 08:59:02
Modified:pthreads/src/main http_main.c
Log:
Removed printf debugging logic
Revision ChangesPath
1.3 +0 -18 apache-apr/pthreads/src/main/http_main.c
Index: http_main.c
==
stoddard99/01/22 08:10:56
Modified:pthreads/src/include scoreboard.h
pthreads/src/main http_main.c http_config.c
Log:
Clean-up some comments.
Revision ChangesPath
1.2 +1 -1 apache-apr/pthreads/src/include/scoreboard.h
Index: scoreboard.h
=
60 matches
Mail list logo
