Promise SATA 300 TX4

[Arno J. Klaassen]

Jeremie Le Hen <jeremie at le-hen.org> writes:

> Nathan,
> 
> On Thu, Aug 23, 2007 at 10:19:02PM +0200, Arno J. Klaassen wrote:
> > I re-read the above email a bit more attentively (?) and noticed
> > that, as Ulf, I use Adaptec SCSI (ahc or ahd) to boot from and
> > have nothing on standard irq-14/15 IDE subsystem.
> > 
> > Once again, I experience the problems on -STABLE-amd64, but can
> > dig up a good-old (not so fast) SCSI-disk to test the (almost) same
> > hardware setup under -CURRENT.
> 
> Which arch are you running?  amd64 or i386?  Also do you use an
> Adapter SCSI controller in the same time?

just a quick 'state-of-the-art' :

1) I replaced TX4 (Promise PDC40718 SATA300) with some cheap Sil3114
   (SATA 150) based 4-ports cards : -stable-amd64 (including last
   sys/vm MFC but rev. 1.268.2.4 for vm_pageout.c) is rock-solid while
   doing moderate heavy geom_raid5 testing of a 6-disk system
   including hot-swapping disk, continuous swap on an Adaptec 2940
   (320 disk connected by 50->68 convertor) and continuous NFS. About
   4000 irq/seconds and not the slightest problem for three days
   (including a lot of more or less brutal reboots)

2) same hardware with TX4 cards (-stable of a couple of days
   ago) crashes when only lookin at it.
   I switched SATA-cables, SCSI-cards, no difference

3) I just realised our main server has a TX2 (Promise PDC20378 SATA150)
   which is rock-stable bur running -stable-i386 of Aug 12th
   (and booting from IDE, no SCSI involved)

voili-voila, not a clear picture, but amd64 versus i386, ATA-only
versus SCSI+ATA, and/or SATA150 versus SATA300 seem credible
suspects.

Arno

NB, /me votes in favour of including geom_raid5 in -current
when code-freeze is over. IMHO this piece of code deserves
wider audience and testing.