WRITE_DMA UDMA ICRC errors with my ata drive(s)/controller

[Kendall Gifford]

Greetings. I recently upgraded my server from 4.9-RELEASE to
5.2.1-RELEASE and immediately started to get the following error
logged whenever there is significant disk activity:

ad0: WARNING - WRITE_DMA UDMA ICRC error (retrying request) LBA=137690696
ad0: FAILURE - WRITE_DMA status=51<READY,DSC,ERROR> 
error=84<ICRC,ABORTED> LBA=137690696

[there are slight, insignificant variations, of course]

Anyhow, the first thing I did was to install and run smartctl (part of
the smartmontools package) since both my drives are S.M.A.R.T. enabled.
Aside from having the same write dma errors in their internal logs, the
drives appear to be fine (all values are well above their threshold).
Also, all disk activity that sparked said errors always succeeded as all
copied files were uncorrupted.

So, I did a little searching online and found that besides the misc.
person having this problem and being told to check their drives or
cables or ide controllers, I came across the following two threads:

http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&threadm=20040409042611.GA68595_binary.net%40ns.sol.net&rnum=1&prev=/groups%3Fq%3Dfreebsd%2Bwrite_dma%2Budma%2Bicrc%2Berror%26hl%3Den%26lr%3D%26ie%3DUTF-8%26selm%3D20040409042611.GA68595_binary.net%2540ns.sol.net%26rnum%3D1

http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&threadm=20040414024308.GA6468%40binary.net&rnum=4&prev=/groups%3Fq%3Dfreebsd%2Bwrite_dma%2Budma%2Bicrc%2Berror%26hl%3Den%26lr%3D%26ie%3DUTF-8%26selm%3D20040414024308.GA6468%2540binary.net%26rnum%3D4

The first thing I noted was that I am also using a board with the VIA
8235: [dmesg]
atapci0: <VIA 8235 UDMA133 controller> port 0xdc00-0xdc0f at device 17.1 
on pci0

And I also have fairly large drives, two Maxtor 120GB drives: [dmesg]
ad0: 117246MB <Maxtor 6Y120P0> [238216/16/63] at ata0-master UDMA133
acd0: CDRW <LITE-ON LTR-48246S> at ata0-slave BIOSPIO
ad2: 117246MB <Maxtor 6Y120P0> [238216/16/63] at ata1-master UDMA133

Also of note is that these two drives worked just fine in UDMA133 mode
in my 4.9-RELEASE system. Due to this and the newness of the two drives,
the motherboard, and the ATA133 round cables as well as having
everything check out in my S.M.A.R.T. test, and the fact that despite
these errors, the disk I/O transactions always seem to succeed (as far 
as I've been able to test) and the system otherwise functions normally
even with both drives in UDMA133 mode, I wonder if there is possibly
either some hardware bug with the VIA 8235 or some incompatibility
between the ata driver in 5.2.1-RELEASE and the VIA 8235 or if someone,
after reading my message and the two threads I linked to, has yet
another idea?

P.S. - Just as was described in the threads above, if I also put my
problem drive in PIO mode, I get no more UDMA ICRC errors.

P.P.S. - While, for reasons I've describe above, I suspect my hardware
is not at fault, I know that I certainly can't guarantee that this is
the case since I don't have the resources (a.k.a. extra hardware and
time) to swap parts around to verify, for sure, it isn't at fault.

-- 
Kendall Gifford
========================================
WEB:   http://kendall.jedis.com/
EMAIL: REPLY TO LIST
========================================