Another IDE UDMA issue

Jamie Lokier (lkd@tantalophile.demon.co.uk)
Tue, 28 Jul 1998 12:31:17 +0100


Noting all the discussion about IDE DMA, I thought I'd mention:

My BIOS enables UDMA, it is configured that way.

Whenever I run Linux 2.1.x (pre 2.1.111 presumably) with default
settings, I can expect a complete system freeze within a few days.

If I explicitly disable DMA with ide0=nodma, I never get a freeze but
IDE data throughput drops to half its former rate (4.5MB/s vs. 9MB/s
approx). And there's the CPU load issue. I can't say I've noticed the
difference in "feel" though. Presumably it's more critical when there's
lots of paging.

Recently following an idea on this list, I tried `hdparm -X34 /dev/hda'
instead of disabling DMA. I think this changes the DMA timings on the
driver, to something faster (so it's not a cable issue). And now the
system is reliable again -- I've had no freezes with this setting with
several weeks use. The throughput is back to the UDMA rate.

I'm under the impression that this is a timing thing that the kernel
can't possibly work out reliably. But I thought I'd mention that I do
have UDMA problems that can be cured without turning off UDMA.

I have one of the supposedly "good at UDMA" Quantum drives -- a Quantum
Fireball ST 6.4GB.

My chipset is a VIA VP2/97.

In not setting the drive right, the BIOS is presumably one that sucks. :-)

-- Jamie

-
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