Re: [PATCH] make rwlocks SHAREABLE + cleanups

I have a question regarding the following changes in pth_mutex_release():

+    if (!mutex->mx_shared.pshared && mutex->mx_owner != current) return
EPERM;

Why this check is not needed for the pshared case? The mx_owner is always
set to
current in pth_mutex_acquire() for both normal and pshared mutexes.

Also,  if (!(mutex->mx_state & PTH_MUTEX_INITIALIZED)) return EINVAL;
is not needed, because the caller always check for valid mx_mutex.

Thanks,
- - - - -
Saurabh Desai
POSIX Threading for Linux
IBM  Linux Technology Center
e-mail: [email protected] OR [email protected]
phone: 512-838-2655, T/L: 678-2655
http://oss.software.ibm.com/developerworks/opensource/pthreads



[email protected]@www-124.southbury.usf.ibm.com on 08/29/2002 12:13:14 PM

Sent by:    [email protected]


To:    [email protected]
cc:
Subject:    [pthreads-devel] [PATCH] make rwlocks SHAREABLE + cleanups


a) I made a patch to support PROCESS_SHARED rwlocks.

b) while doing this I had to supply the patch in pth_sched.c and the
dopoll flag
             - therefore it's included:
@@ -840,7 +840,7 @@
                             this_occurred = FALSE;

                             /* Filedescriptor I/O */
-                            if (ev->ev_type == PTH_EVENT_FD && !dopoll &&
(ds == descr)) {
+                            if (ev->ev_type == PTH_EVENT_FD && (ds ==
descr)) {
                                     /* filedescriptors are checked later
all at once.
                                        Here we only assemble them in the
fd sets */
                                     if (ev->ev_goal &
PTH_UNTIL_FD_READABLE)
@@ -1035,7 +1035,7 @@
      FD_SET(descr->sigpipe[0], &rfds);

      /* If dopoll is set, handle the thread from ready queue... */
-    if (dopoll)
+    if (dopoll && m_to_n)
             return;


c) declaration pthread_kill_all_threads() removed from td_manager.c but
still
             does not compile when configuring --enable-thread_db

d) fixed a memory leak in pthread_cond_init()
-            rc = pth_cond_init(cn);
+            if (!(rc = pth_cond_init(cn)))
+                        pth_free(cn);

e) replaced EDEADLK with EPERM or EINVAL
             pthread_mutex_lock or pthread_rwlock_*lock should return
EDEADLK
             when the same thread already holds the lock (and it's not
recursive)

             the checking has to be improved on this (i.e. when pshared)