Promise PDC20378 errors, hangs

[soulreaver1]

Hello,

I've a problem with my PCI sata controller based on PDC20378 chip. Freebsd 8.2 x86 is currently installed system. The problem is somtimes on heavy load (high transfer rate) system hangs. After reboot i've ran

cat messages | grep ad6

command to find out what's going on:

Oct 16 14:57:17 theone root: ZFS: checksum mismatch, zpool=magazyn1 path=/dev/ad6 offset=56065068032 size=1024
Oct 16 14:58:08 theone kernel: ad6: WARNING - SETFEATURES SET TRANSFER MODE taskqueue timeout - completing request directly
Oct 16 14:58:48 theone kernel: ad6: WARNING - SETFEATURES ENABLE RCACHE taskqueue timeout - completing request directly
Oct 16 14:59:28 theone kernel: ad6: WARNING - SETFEATURES ENABLE WCACHE taskqueue timeout - completing request directly
Oct 16 15:00:08 theone root: ZFS: checksum mismatch, zpool=magazyn1 path=/dev/ad6 offset=56065067008 size=1024
Oct 16 15:00:08 theone kernel: ad6: WARNING - SET_MULTI taskqueue timeout - completing request directly
Oct 16 15:00:08 theone kernel: ad6: TIMEOUT - READ_DMA retrying (1 retry left) LBA=46376704
Oct 16 15:01:12 theone root: ZFS: checksum mismatch, zpool=magazyn1 path=/dev/ad6 offset=4206080 size=512
Oct 16 15:19:17 theone kernel: ad6: 238475MB <Seagate ST3250410AS 3.AAC> at ata3-master UDMA100 SATA 1.5Gb/s

pool: magazyn1
 state: ONLINE
status: One or more devices are faulted in response to IO failures.
action: Make sure the affected devices are connected, then run 'zpool clear'.
   see: http://www.sun.com/msg/ZFS-8000-HC
 scrub: scrub in progress for 0h27m, 97.54% done, 0h0m to go
config:

        NAME        STATE     READ WRITE CKSUM
        magazyn1    ONLINE       0    14     0
          ad6       ONLINE       3    37     0

errors: Permanent errors have been detected in the following files:

        <metadata>:<0x0>
        <metadata>:<0x39>
        <metadata>:<0x54>

 zpool clear magazyn1
cannot clear errors for magazyn1: I/O error

Any ideas?

 

[gkontos]

Run smartmontools to see if your disk has problems. If your disk looks good then upgrade to 8.2-STABLE, there was a patch regarding promise controllers a couple of months ago. If all that fails then join the club.

George

 

[soulreaver1]

Thank you for your response @gkontos. Yesterday, the machine has completly freezed/locked. After that the system hangs at boot time, while loading the ZFS module. The only way to avoid this was taking the pool offline (removing the disk before boot), disabling zfs in the rc.conf file or physically removing the device (disk). I've bought a used, old SATA controller based on sil3112 and all problems have disappeared, so far.