qf/quakeforge
0
0
Fork 0
mirror of https://git.code.sf.net/p/quake/quakeforge synced 2024-11-10 15:22:04 +00:00
Code Issues Projects Releases Packages Wiki Activity

[qwaq] Close up race-condition hole

Browse source 
After a lot of thought, I have come to the conclusion that the weird
crash the other day was caused by a race while the command ring buffer
had just been emptied: the command submission code opened up space for
writing, threads switched, and command processing saw the available data
and pounced on it before the submit code could write valid data. Thus
include the while header in the lock, and move the loop-end release
outside the lock. It may be a little confusing, but it seems to work.
This commit is contained in:
Bill Currie 2020-03-25 16:18:14 +09:00
parent 4cef9792f4
commit 585b33c111
1 changed files with 9 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

12
ruamoko/qwaq/qwaq-curses.c
View file

@ -640,10 +640,14 @@ process_commands (qwaq_resources_t *res)
ret = pthread_cond_timedwait (&res->command_cond.rcond,
&res->command_cond.mut, &timeout);
}
pthread_mutex_unlock (&res->command_cond.mut);
// The mutex is unlocked at the top of the loop and locked again
// (for broadcast) at the bottom, then finally unlocked after the loop.
// It should be only the data availability check that's not threadsafe
// as the mutex is not released until after the data has been written to
// the buffer.
while ((avail = RB_DATA_AVAILABLE (res->command_queue)) >= 2
&& avail >= (len = RB_PEEK_DATA (res->command_queue, 1))) {
pthread_mutex_unlock (&res->command_cond.mut);
dump_command (res, len);
qwaq_commands cmd = RB_PEEK_DATA (res->command_queue, 0);
switch (cmd) {
@ -738,8 +742,10 @@ process_commands (qwaq_resources_t *res)
pthread_mutex_lock (&res->command_cond.mut);
RB_DROP_DATA (res->command_queue, len);
pthread_cond_broadcast (&res->command_cond.wcond);
pthread_mutex_unlock (&res->command_cond.mut);
// unlocked at top of top if there's more data, otherwise just after
// the loop
}
pthread_mutex_unlock (&res->command_cond.mut);
}
static int