Date: Fri, 8 Sep 2000 20:33:45 +0200
From: Torben Mathiasen <tmm@image.dk>
How about converting the cli() code to spin_locks while we are at
it?
There are many protocols which are of this nature, I did appletalk
spinlocking for example. I didn't even finish cleaning appltalk up,
it still is a "stupid" protocol in that it still changes packet
contents during input packet processing, for example. Only after
that would appletalk be %100 "new-style".
I encourage people to do the spinlocking and receive processing fixups
for these protocols, but it is pain in the ass because many of these
protocols are hard to test without the correct kit being available.
(The work is also a bit depressing, doing SMP locking really shows how
buggy and full of races these lesser used protocols are.)
We have wrappers etc. to handle these non-converted protocols so that
they function in 2.4.x as properly as they did in 2.2.x, so it's not
really a matter of them being less correct until spinlocked. The only
large gain we would achieve from this work would be cleaner code plus
the ability to remove the wrappers and the "stupid protocol receive
packet mangling" support in netif_rx.
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.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Sep 15 2000 - 21:00:11 EST