Hi everyone,
I've been having a problem for a while now with a strange lockup induced
under heavy SCSI I/O (particularly when I write directly to block devices
with dd of=/dev/sd?, but also when writing to a filesystem on that
device). I'm writing around 50-100mb/sec over FC (qlogic 2200) under
Linux 2.4 (tried 2.4.3 through 2.4.10).
It seems that some vm or I/O related spinlock is being taken and held, but
not released (?). There is no oops or BUG() or anything (no messages at
all in fact).. all I/O just stops. I can still invoke sysrq, type
characters at the console, etc. In fact, usually top continues to run and
display kswapd as the dominant process.
Any ideas?
- Gord R. Lamb ([email protected])
Maximum Throughput, Inc.
Sr. Systems Architect
On Mon, 12 Nov 2001, Gord R. Lamb wrote:
> Hi everyone,
>
> I've been having a problem for a while now with a strange lockup induced
> under heavy SCSI I/O (particularly when I write directly to block devices
> with dd of=/dev/sd?, but also when writing to a filesystem on that
> device). I'm writing around 50-100mb/sec over FC (qlogic 2200) under
> Linux 2.4 (tried 2.4.3 through 2.4.10).
>
> It seems that some vm or I/O related spinlock is being taken and held, but
> not released (?). There is no oops or BUG() or anything (no messages at
> all in fact).. all I/O just stops. I can still invoke sysrq, type
> characters at the console, etc. In fact, usually top continues to run and
> display kswapd as the dominant process.
If kswapd is eating the CPU, its probably a VM problem.
Please try the newest 2.4.xx which has VM updates and if the problem
happens again, tell us.
Well, Arjan just told me that qla2x00 is doing weird shit. He seems to be
fixing it...
On Mon, 12 Nov 2001, Marcelo Tosatti wrote:
>
>
> On Mon, 12 Nov 2001, Gord R. Lamb wrote:
>
> > Hi everyone,
> >
> > I've been having a problem for a while now with a strange lockup induced
> > under heavy SCSI I/O (particularly when I write directly to block devices
> > with dd of=/dev/sd?, but also when writing to a filesystem on that
> > device). I'm writing around 50-100mb/sec over FC (qlogic 2200) under
> > Linux 2.4 (tried 2.4.3 through 2.4.10).
> >
> > It seems that some vm or I/O related spinlock is being taken and held, but
> > not released (?). There is no oops or BUG() or anything (no messages at
> > all in fact).. all I/O just stops. I can still invoke sysrq, type
> > characters at the console, etc. In fact, usually top continues to run and
> > display kswapd as the dominant process.
>
> If kswapd is eating the CPU, its probably a VM problem.
>
> Please try the newest 2.4.xx which has VM updates and if the problem
> happens again, tell us.
>
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>