Re: patch-ikd+profile+stack-meter against 109

Andrea Arcangeli (arcangeli@mbox.queen.it)
Sat, 18 Jul 1998 15:58:33 +0200 (CEST)


On Sat, 18 Jul 1998, Michael L. Galbraith wrote:

>I pushed my machine hard enough for it to take half an hour to regain
>control again (300+ MB into swap gets kinda slugish:) and came up with..
>
> 4500 -1071470953 = 0xc022a696 = aic7xxx_allocate_scb()
^^^^
Nice!

There was 1page + 404byte free of stack. It seems that as now we need
really 8kbyte of stack to be quiet.

As I guessed _probably_ the scsi layer tend to eat a lot of stack. Some
time ago I noticed that if the scsi_done() function was recalled using the
same stack of the caller of queuecommand() there was no way to avoid a
stack overflow... (I hope to remeber well...).

>I was trying to max out SB16, PSAUX, NE2000 and AHA2940 interrupts.
>
>BTW, not one fork failure despite 2 concurrent make -j bzImage. With
>earlier kernels this always eventually led to fork failures. (K=109)

Nice! BTW I don' t think that a make -j is the best thing to use to
stress the stack...

Andrea[s] Arcangeli

-
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.altern.org/andrebalsa/doc/lkml-faq.html