Re: Possible IDE corruption bug?

[John Goerzen <jgoerzen%complete.org@localhost>]

On Tue, Oct 27 2026, Martin Husemann wrote:

> On Tue, Oct 27, 2026 at 12:57:43AM -0400, Jeffrey S. Worley wrote:
>> The onboard ata driver is broken on NetBSD 11.  It is moribund on 10.1
>> and below.  It is unfenced, Alpha requires memory barriers for ordered
>> IO, the coder did not get that memo.
>> 
>> I boot my Alpha from the onboard ATA controller just to get kernel into
>> memory, then pivot to an SI3124 SATA controller.  It runs a
>> 100meg@sec!, and is bulletproof.
>
> Someone (tm) please file a PR with the dmesg section of the controller and
> disks in question.
>
> For the split boot case: you can modify your kernel config to force root
> on some other device by default, that makes testing of the broken controller
> easier than removing it completely.

Currently it looks like the HDD is working, and the CD isn't.  I still
have some SYM53C875 cards floating around and may even have a SCSI CD
drive.  But really, I don't need the CD for anything once installed.

Some sleuthing suggests that perhaps this is a CMD646 issue, such as
https://en.wikipedia.org/wiki/CMD640 .  Some sources are suggesting
these bugs were more likely on ATAPI.  I don't recall running into this
on Linux, though I may be blocking something out of my memory from 10
years ago, or it's possible Linux had a workaround or I never used an
IDE CD drive on this system.

In any case, some further investigation suggested that "flags cmdide0
0x00f0" would disable DMA for the IDE channel the CD drive was on, but I
struggled and failed to make it take effect (no combination of SRM boot
arguments could get me into UserConf, /boot.cfg appears to be ignored,
and I wasn't going to compile my own kernel just to test a theory).

Incidentally, it would be great to have
https://cdn.netbsd.org/pub/NetBSD/NetBSD-11.0/alpha/INSTALL.html cover a
bit more ground on post-install boot.  I had utterly failed to remember
how to tell SRM what the default boot device should be, and dug that up
elsewhere.  -flags c and -flags "-c" both get a complaint from NetBSD as
being unrecognized and drop me into single-user mode rather than
userconf.

I could make this system entirely SCSI.  I still have two UW SCSI
drives, but both are exceptionally noisy, and only the noisiest (a
Cheetah) has no bad sectors.  I was annoyed at that noise for years, so
maybe I'll just buy a BlueSCSI or something if needed.

I'm not running it to win any performance records, so that's probably
sufficient.  The Cheetah maxxed out at 29MB/s anyhow!

John