New and improved 2.0.36 crash, or is this new?

david parsons (o.r.c@p.e.l.l.p.o.r.t.l.a.n.d.o.r.u.s)
11 Dec 1998 02:32:56 -0800


I've got a lightly patched 2.0.36 kernel (VIA apollo UDMA patches, Linux
logo patches) with a slight problem. I'm building up a fileserver
(largish AMD K6, 400mb of core, 2 x 13gb IBM UDMA disk, motherboard
with a VIA Apollo chipset) and was running some routine death tests
against it and got a pair of null pointer oopses inside bonnie [-s 800],
after which I could no longer read anything large from the disk.

I tracked down the source of the oops, fully expecting to see it inside
the VIA apollo driver, but to my intense surprise I found that the oops
was happening inside try_to_read_ahead(), with a call chain of

system_call()
sys_read()
try_to_read_ahead()

I thought that it might be hardware problems with the VIA apollo, so I
built a 2.0.28 kernel (is it just me, or has gas changed recently so
that `;' doesn't work anymore?) with an older less-desirable patch
(that only activates DMA on ide0, and that only works for the master
device) for the chipset. This patched kernel worked happily, even
under the crowd-pleasing

(cd /tmp; bonnie -s 1000) & (cd /home; bonnie -s 1000) &

watch me die test.

Have I missed something that's obvious, or is it something unhappy
with 2.0.36?

____
david parsons \bi/ It's difficult to buy Triton-chipsetted
\/ 8050x boards that accept EDO Dimms.

-
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/