this patch against 2.3.16 should fix the SMP race:
--- 2.3.16-flushpage/fs/buffer.c.~1~ Wed Sep 1 09:30:03 1999
+++ 2.3.16-flushpage/fs/buffer.c Wed Sep 1 16:21:54 1999
@@ -1231,16 +1231,10 @@
*/
if (offset <= curr_off) {
if (buffer_mapped(bh)) {
- atomic_inc(&bh->b_count);
- wait_on_buffer(bh);
if (bh->b_dev == B_FREE)
BUG();
mark_buffer_clean(bh);
- clear_bit(BH_Uptodate, &bh->b_state);
- clear_bit(BH_Mapped, &bh->b_state);
- clear_bit(BH_Req, &bh->b_state);
- bh->b_blocknr = 0;
- atomic_dec(&bh->b_count);
+ wait_on_buffer(bh);
}
}
curr_off = next_off;
I removed also the atomic_dec/inc since they are useless.
Maybe a better idea is to sleep until we can obtain the buffer locked
ourselves, this %100 guarentees that ll_rw_block will not do anything
with the buffer.
On the otherhand, I cannot convince myself that your change is safe
and %100 fixes the race and also does not create new ones.
Later,
David S. Miller
davem@redhat.com
-
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/