the patch below fixes two yield()-usage bugs introduced by the new
scheduler. In the first O(1) scheduler iterations i've added TASK_YIELDED,
which made the setting of p->state unnecessery before calling yield(). But
later i cleaned up yield() further, again enabling the yielding of
non-running tasks. But the places which need to set p->state were not
reverted to the pre-TASK_YIELDED behavior.
the bug was seen live by Arjan van de Ven (in the 2.4 kernel), it causes
'stuck' shells under heavy IO and VM load.
Ingo
--- linux/fs/buffer.c.orig Fri Jan 25 12:40:11 2002
+++ linux/fs/buffer.c Fri Jan 25 12:40:18 2002
@@ -724,6 +724,7 @@
wakeup_bdflush();
try_to_free_pages(zone, GFP_NOFS, 0);
run_task_queue(&tq_disk);
+ __set_current_state(TASK_RUNNING);
yield();
}
--- linux/mm/page_alloc.c.orig Fri Jan 25 12:43:42 2002
+++ linux/mm/page_alloc.c Fri Jan 25 12:43:50 2002
@@ -394,6 +394,7 @@
return NULL;
/* Yield for kswapd, and try again */
+ __set_current_state(TASK_RUNNING);
yield();
goto rebalance;
}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Jan 31 2002 - 21:00:31 EST