More Intel ICH5 SATA150 hangs

[Jeff Doolittle]

Louis LeBlanc wrote:

>I know I've asked this question before, and I sure hate to start
>sounding like a broken record, but I wasn't able to get a solution
>last time around.  The problem is with an Intel ICH5 SATA150 Disk
>Controller.  I'm still getting errors when the disk is placed under
>load:
>
>Oct 29 15:02:24 key2 kernel: ad4: TIMEOUT - WRITE_DMA retrying (2 retries left) LBA=62023103
>Oct 29 15:02:29 key2 kernel: ad4: TIMEOUT - WRITE_DMA retrying (2 retries left) LBA=61992511
>
>I'll get a couple of these and then the whole system freezes,
>requiring a hard boot.  Recovery always involves manually running fsck
>on one or more partitions, and sometimes lost data.
>  
>
Yes, I've been there before with my ASUS P4C800-E....

>I've included various bits of info in past postings.  I'll re-collect
>whatever info might be helpful if anyone wants, but the scanpci output
>describing the controller is here:
>
>pci bus 0x0000 cardnum 0x1f function 0x00: vendor 0x8086 device 0x24d0
> Intel Corp. 82801EB LPC Interface Controller
>
>The system is a fairly new (3 months old) Dell 8300; 3Ghz Pentium with
>HT enabled.  The disk controller is an Intel ICH5 SATA controller, as
>mentioned above.  The disk is a WDC WD1600JD-75HBB0 (Western Digital).
>
>Last time I asked about this, it was suggested that the drive might be
>bad.  The bios had a test that passed the drive.  It was then
>suggested that I use the WD test utility.  I wasn't able to do this at
>the time because I don't have a floppy on this system, and creating a
>bootable CD from the image didn't work (I think it had a: hardcoded
>in).  Well, I finally broke down and hijacked the floppy from another
>machine and ran the WD diags.  Passed with flying colors, both the
>quick test and the extended test.  The drive is fine.
>
>So today I was building OpenOffice and everything came to a screeching
>halt again.  See the log entries above.
>
>I did a lot of googling to see if there was anything about the
>controller and the WRITE_DMA message out there - even associated with
>Linux or any other *nix.  Very little, but there were suggestions to
>change the CPU to compatible mode.  I tried this and it wouldn't boot.
>  
>
CPU to "compatible mode"?  What worked for my ASUS was to change the 
Serial ATA from enhanced mode to compatible mode which results in the 
loss of either your Primary or Secondary IDE channel.  I've been running 
this way for months without a problem.  I use the onboard Promise 
controller for my "live" data and do a weekly dump to the WDC drives on 
the ICH5 "compatible" mode enabled drives.   The only reboots of the box 
have been for newer versions of CURRENT and now STABLE.

>Same result the other poor schmuck got when he tried it.  There was
>also a suggestion that HT be turned off.  Kinda defeats the whole
>purpose of spending the extra $100 when I bought this system, so I
>didn't bother trying it - especially since there were folks out there
>that said it wasn't related - they had no problems with the Disk
>controller and HT.
>
>  
>
For the record I'm still running with HT enabled although I might turn 
it off after reading many messages about how system performance is impacted.

>So, I'm back to the 5.2.1 kernel and/or driver.  There's also the fact
>that this particular controller isn't mentioned in the 5.2.1 hardware
>support list on the FreeBSD site, but it is mentioned in the 4.10
>hardware support list.
>
>I'd rather stick with 5.x, of course, but I don't know if my hardware
>is dropping off the support list.
>
>If I were to put the system in place as my server, and it locked up,
>it would be a real pain in the ass, with me getting all grumpy over
>it.  On the other hand, I really want to get the venerable old 400Mhz
>system out of here.
>
>Anyone have any other suggestions, info, whatever?  I'll try turning
>off HT if that's the only way to go for now, but I don't see how that
>could be the problem - unless there's a locking problem in the kernel
>- which I'd hope gets fixed in 5.3.
>
>Thanks in advance for all feedback.
>
>Lou
>  
>
Hopefully the above is of some help.

    Good luck

       - Jeff