[patch] 2.3.18ac7 nr_threads fixes

Andrea Arcangeli (andrea@suse.de)
Wed, 22 Sep 1999 15:40:04 +0200 (CEST)

In 2.3.18ac7 there's still the SMP race where release() can decrease the
nr_threads counter in racy SMP beahviour.

Also we was decreasing nr_threads in the failpath while nr_threads is
increased in fork() only if fork will be succesfully (so we was wrong in
decreasing nr_threas there).

Here it is the fix against 2.3.18ac7:

diff -urN 2.3.18ac7/include/linux/sched.h 2.3.18ac7-nr_threads/include/linux/sched.h
--- 2.3.18ac7/include/linux/sched.h Tue Sep 21 16:46:50 1999
+++ 2.3.18ac7-nr_threads/include/linux/sched.h Wed Sep 22 15:31:53 1999
@@ -825,8 +825,8 @@
extern inline void unhash_process(struct task_struct *p)
{
if (task_on_runqueue(p)) BUG();
- nr_threads--;
write_lock_irq(&tasklist_lock);
+ nr_threads--;
unhash_pid(p);
REMOVE_LINKS(p);
write_unlock_irq(&tasklist_lock);
diff -urN 2.3.18ac7/kernel/fork.c 2.3.18ac7-nr_threads/kernel/fork.c
--- 2.3.18ac7/kernel/fork.c Tue Sep 21 16:46:51 1999
+++ 2.3.18ac7-nr_threads/kernel/fork.c Wed Sep 22 15:36:40 1999
@@ -617,8 +617,9 @@
}

/*
- * Counter atomicity is protected by
- * the kernel lock
+ * Counter increases are protected by
+ * the kernel lock so nr_threads can't
+ * increase under us (but it may decrease).
*/
if (nr_threads >= max_threads)
goto bad_fork_cleanup_count;
@@ -718,9 +719,9 @@
write_lock_irq(&tasklist_lock);
SET_LINKS(p);
hash_pid(p);
+ nr_threads++;
write_unlock_irq(&tasklist_lock);

- nr_threads++;
wake_up_process(p); /* do this last */
++total_forks;

@@ -742,8 +743,6 @@
__MOD_DEC_USE_COUNT(p->exec_domain->module);
if (p->binfmt && p->binfmt->module)
__MOD_DEC_USE_COUNT(p->binfmt->module);
-
- nr_threads--;
bad_fork_cleanup_count:
if (p->user)
free_uid(p);

Andrea

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/