problem s HW ?

[martin hudec]

Ahoj,

On Fri, Dec 17, 2004 at 03:51:35PM +0100 or thereabouts, Divacky Roman wrote:
> dost silne pochybuju ze je to chyba hw. spis fbsd implementace ata neco
> neumi/dela blbe...
> 
> rozhodne 5.2.1-R je tezkej pr*ser (z hlediska ata i jinak) a tak zkuste 5.3R
> treba to pomuze... nejspis nepomuze ale skodit to nemuze ;)
> 

Priklanam sa k tomuto nazoru. Podobny problem zazivam raz za cas (priblizne
2x po sebe raz za dva tyzdne) na FreeBSD 5.2.1 s vinum v raid1 a Western
Digital a Maxtor diskami. Nevinil by som za to vinum ani hardware ako taky.
Cez sviatky snad bude cas to otestovat na 5.3.

Hlasky (okrem kernel panicu :)):

Dec  6 16:11:53 office kernel: ad0: WARNING - WRITE_DMA interrupt was seen but taskqueue stalled LBA=8388799
Dec  6 16:11:53 office kernel: ad1: WARNING - WRITE_DMA interrupt was seen but timeout fired LBA=109142664
Dec  6 16:11:53 office kernel: ad1: WARNING - WRITE_DMA interrupt was seen but timeout fired LBA=109142664
Dec  6 16:11:53 office kernel: ad1: WARNING - WRITE_DMA interrupt was seen but taskqueue stalled LBA=109142664

Pan Kirk Strausser mi odpovedal:

I recently became one of the growing number of users seeing
WRITE_DMA-related panics on a regular basis any time I loaded down my
system.  I'd tried the workaround of running my main 120GB drive in PIO4
mode, but it was unbearably slow and the system crawled during large
transfers.

I'd been installing FreeBSD on vinum concat volumes for several years for
the flexibility in allocating storage (want to grow /usr?  Just tack another
plex to the end and growfs it).  However, I'd noticed that my storage
requirements had pretty well stabilized over the last couple of years, and
I decided to remove my vinum setup in favor of traditional partitions.

When I did that, my WRITE_DMA problems disappeared completely under all
circumstances that I've been able to test.  I have been totally unable to
panic my system (or get any kind of DMA warnings) since the migration.

I'm not ready to completely blame vinum for destabilizing my ATA system - I
wouldn't testify in court that I didn't inadvertently change any seemingly
minor but actually important setting - but the circumstantial evidence seems
pretty strong for me.


-- 
martin hudec


   * 421 907 303 393
   * corwin at aeternal.net
   * http://www.aeternal.net

"Nothing travels faster than the speed of light with the possible 
exception of bad news, which obeys its own special laws."

   Douglas Adams, "The Hitchhiker's Guide to the Galaxy"