linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* latency while writing to pc-card disks/tapes
@ 2002-08-04 11:01 Julian Bradfield
  2002-08-04 13:57 ` Alan Cox
  0 siblings, 1 reply; 2+ messages in thread
From: Julian Bradfield @ 2002-08-04 11:01 UTC (permalink / raw)
  To: linux-kernel

I've tried searching, and asked on general lists, but so far without
result, so I'd like to ask here to find out what causes the issue:

I have a notebook (HP Omnibook 6000, but same happens with others)
running Linux 2.4.18 and pcmcia-cs 3.2.0.

When I make backups to PCMCIA devices, whether hard drives (standard
PC-card hard drive) or tapes (SCSI tape attached to Adaptec PC-card),
system response drops through the floor -- we're talking seconds to
respond to a mouse movement.

With the tape drive, this is constant; with the disk, the buffer cache
fills up, then the kernel starts writing to the disk and the rest of
the system crawls until the writes stop and the next batch of reads
happens.

Another thing that surprises me is that this runs the CPU at
full stretch, since the fan comes on. I find it hard to believe that
writing to a rather slow tape drive can fully exercise a 1GHz
processor continuously.
(It may be relevant for the tape that scsi dis/reconnect is disabled,
as it doesn't work for my card.)

Is there something I can do about this? I tried the kernel preemption
patch and the lock-breaking patch, with no result.
(With the disk, I have set the interrupt unmask via hdparm, but this
also makes no difference.)

I will follow the list, but would appreciate being cc'ed on any
replies.
Thanks for any advice!


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: latency while writing to pc-card disks/tapes
  2002-08-04 11:01 latency while writing to pc-card disks/tapes Julian Bradfield
@ 2002-08-04 13:57 ` Alan Cox
  0 siblings, 0 replies; 2+ messages in thread
From: Alan Cox @ 2002-08-04 13:57 UTC (permalink / raw)
  To: Julian Bradfield; +Cc: linux-kernel

> Another thing that surprises me is that this runs the CPU at
> full stretch, since the fan comes on. I find it hard to believe that
> writing to a rather slow tape drive can fully exercise a 1GHz
> processor continuously.
> (It may be relevant for the tape that scsi dis/reconnect is disabled,
> as it doesn't work for my card.)

PCMCIA is PIO only. Your processor is hard at work sitting waiting for
out instructions to complete across the low speed (8Mhz or so) PCMCIA
bus. Cardbus might help immensely if there are suitable cardbus adapters
around because a cardbus slot with a cardbus card (not a pcmica card) in
it is basically pure PCI



^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2002-08-04 12:36 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-08-04 11:01 latency while writing to pc-card disks/tapes Julian Bradfield
2002-08-04 13:57 ` Alan Cox

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).