Re: sata_via: write errors on SATA drive connected to VT6421

[Jeff Garzik <jgarzik@pobox.com>]

Landry Breuil wrote:
> On Tue, Feb 03, 2009 at 10:59:10AM -0500, Jeff Garzik wrote:
>> Landry Breuil wrote:
>>> Hi,
>>>
>>> i encounter a very similar problem as reported in
>>> http://marc.info/?l=linux-kernel&m=118764381013943&w=2, but with a SATA
>>> disk. Trying to transfer 250Gb (mostly small files) from a SATA disk
>>> (ata1/sda) to another (ata2/sdb), both plugged on the VT6421 card, after
>>> some dozens of seconds the controler bails out and degrades write performance
>>> (UDMA133->UDMA100->UDMA33...)
>>>
>>> kernel: ata2.00: limiting speed to UDMA/33:PIO4
>>> kernel: ata2.00: exception Emask 0x12 SAct 0x0 SErr 0x1400500 action 0x2 frozen
>>> kernel: ata2: SError: { UnrecovData Proto Handshk TrStaTrns }
>>> kernel: ata2.00: cmd 35/00:00:07:c6:f4/00:04:07:00:00/e0 tag 0 dma 524288 out
>>> kernel:          res 40/00:b0:27:1d:dc/84:00:07:00:00/e0 Emask 0x16 (ATA bus error)
>> Like Robert noted, this is the hardware complaining that it has found a  
>> hardware error.
>>
>> "SErr" stands for SATA SError register value.  That is a value read  
>> directly from hardware, which indicates one or more problems.  Here is  
>> some more info:
>> http://ata.wiki.kernel.org/index.php/Libata_error_messages#SATA_SError_expansion
>>
>> When libata sees a bunch of hardware problems, it tries the only things  
>> it can do to attempt progress in that situation:  reset the hardware,  
>> and slow down data transfer in an attempt to avoid errors again.
> 
> Okay, that makes sense. I swapped disks on the controller, and the
> errors still pointed at the same disk. Will try with another disk..
> the weird thing is that transferring files at a slow pace (from a
> usb2 disk) shows no failure...

That is surprisingly normal, actually.  Flaky hardware can often behave 
100% normally, until you stress it.

	Jeff