Re: [patch] fixed both processes in D state and the /proc/ oopses

Tom Holroyd (tomh@nibh.go.jp)
Thu, 28 Jan 1999 18:45:44 +0900 (JST)


Well, I guess there were *two* reasons why procs were getting stuck in D.
Andrea's patches both (spinlock and not-spinlock) fix the case for top
getting stuck in D, but on an Alpha you _also_ need Richard's semaphore.S
patch -- without that the makes get stuck in D. I didn't notice the
difference before, and Richard's patch has no effect on the tops getting
stuck in D.

Re: to spinlock or not to spinlock; either patch + Richard's seems stable,
but the interactive feel is different under high load. The spinlock
version seems to freeze while it's doing big writes, but seems more
responsive at the mouse (except when it freezes). The non-spinlock
version seems to have poorer overall interactivness. Performance is really
similar though, the make MAKE="make -j5" dep takes just as long in either
case. And maybe it's just randomness. They seem quite close, considering
the load (up to 20). I did get both to oops with the down debug patch,
but that's probably 'cause the timeout's only 20 sec., and I did an rm;
sync of a 100 meg file during the make... It takes a while.

Time for other things. I was just invited to participate in a
beer-tasting experiment (really! they gave me a form to fill out
and everything!).

Dr. Tom Holroyd -- my email address changed; previous server got hosed
I would dance and be merry,
Life would be a ding-a-derry,
If I only had a brain.
-- The Scarecrow

-
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/