I noticed a few problems while reviewing the code in ip_fragment.c,
and have attached a patch that should fix things up.
Nice work Bill. I made one change to your patch.
As it turns out, I noticed after scanning all the code paths that we
never get into the defragmentation code from anywhere which is not in
a BH context. So I removed all the BH locking altogether.
If you look at all the callers into that code, it is input IP packet
processing or timers. If this were not within BH then TCP just
wouldn't work and would crash all over the place.
Therefore, since this is non-trivial to understand for people reading
the code, I've put appropriate commentary where we used to do the BH
locking.
Later,
David S. Miller
davem@dm.cobaltmicro.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu