Re: Bad apache perfomance wtih linux SMP

Bjorn Wesen (bjorn@sparta.lu.se)
Wed, 19 May 1999 14:12:53 +0200 (MET DST)

On Tue, 18 May 1999, Andi Kleen wrote:
> > > in tcp.c:tcp_do_sendmsg with
> > >
> > > unlock_kernel();
> > > skb->csum = csum_and_copy_from_user(from,
> > > skb_put(skb, copy), copy, 0, &err);
> > > lock_kernel();

On a related note, if one would, say, rewrite csum_and_copy to use DMA HW
based copying + IP checksumming, one would want to sleep the process until
the DMA is finished (and not busy wait for the DMA). If the above is ok,
is it also safe to explicitely sleep the process (like, sleeping on a
waitqueue awoken by the DMA done IRQ), and would it work on 2.0 (non-SMP)?

> > > The patch does not violate any locking requirements in the kernel, because
> > > the kerne lock could have been dropped at any time anyways when the copy_from_user
> > > slept to swap a page in.

regards,
/Bjorn

-
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/