Re: libata Promise driver regression 2.6.5->2.6.6

[Brad Campbell]

Against my better judgment I'm going to send this again as I have been having problems with a 
particularly faulty mail server randomly blackholing outgoing mail and I have not seen this in the 
archives after about 20 hours. Apologies if this arrives twice. I have temporarily subscribed to the 
list.

Jeff Garzik wrote:
> Interesting.
>
> Well since it's not global behavior, but isolated to one port or card, I 
> still worry about non-libata things:
> 1) is a SATA cable bad, or not plugged in well?  I'm finding that it's 
> easier to screw up SATA cabling than PATA.  It's more-convenient design 
> is also less rugged.

Nup, all seated fine and working perfectly well (and I mean raid-5 across all 10 disks and spraying
data at maximum PCI speed for 8 hours at a time well) with 2.6.5, and they are Supermicro cables
SATA cables with a pair of 5 bay Supermicro Hotswap bays so the quality is pretty good.

> 2) is a PCI slot bad, or not busmastering like it should?  have you 
> tried moving the card to another PCI slot?

I hadn't, but I just gutted the system and did a heap of card shuffles.. it ain't that.

> 3) is it always the 9th port, regardless of the ordering of the PCI 
> cards in PCI slots?

Nup, but it's always appears to be on the 3rd card. Pull one of the cards and it's sweet.


> I apologize if you answered some of these in a previous message.

I hadn't, but I have now :p)

Under all these circumstances, 2.6.5 would boot fine. These tests were done with 2.6.5 and the patch
you supplied me.

I can even pull all the drives from boards 1&2 (drives 0-7) and it will lock hard on drive 8.

If I pull the 3rd card then it all behaves perfectly no matter what I do to it. I have relocated the
3rd card in other slots to change the detection order and it always *seems* to die on the 3rd
initialised card, no matter what slot.

Strange, No?

Brad
(I have knocked the other guys from the CC: as it does not appear to be ACPI related)

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/