RE: ATA driver races with interrupts

I wrote:

> After 3+ days of uptime, including continuous SMART monitoring (which
> without patch triggers an immediate disconnect, and with Sörens patch
> triggered a disconnect after ~16 hours) and a fair amount of 
> disc activity on both channels it looks like maybe the serialization
> patch from Ville-Pertti is enough to make my SATA controller work 
> properly. Of course, it also slows things down, but right now that is
> better than the alternative.

Ha, so much for thinking it was stable. Earlier today (after 4+ days of
uptime) the machine disconnected one of the SATA discs despite the fact that
I'm running with Ville-Pertti's serialization patch.

I have two 250GB SATA discs, one Western Digital and one Maxtor. After
looking at the logs from today, and grep'ing some old logs, I noticed that
it seems more common for the channel with the WD disc to lock up. My logs
don't go back far enough that I can say it has always been like that, but it
looks that way from the logs right now (but I only switched channels for the
discs once, so the sampling is very limited and the result could be entirely
coincidental).

I have updated my system to sources from 2004.08.09.13.00.00 without any
patches (because I saw some commits to the ATA code in the last couple of
days). The base code now allows me to at least start the SMART monitor and
run some stress tests. I'll report back when/if it fails. By next week I'll
probably have a Promise SATA150 TX4 that I will use instead. It will be very
interesting to see if that one also will give me problems with the WD disc.

/Daniel Eriksson