Brad Campbell wrote:
[...]
Actually, I'm not sure without the libata dev patch as that removes SMART support, and I'm not convinced that my smartd polling every 20 minutes does not have something to do with it. All I know is the older kernel seems to cope. We'll see. 320 minutes left on this rebuild. I expect it will be done in the morning if all goes according to plan. (With the 2.6.11 kernel it never survived past about 25% rebuilt)
Can you find time to try it without smartd active ? You report running a uni-processor system, and I have only seen PIO problems with (fast) SMP systems in my testing, but I am forming the opinion that libata-PIO functions are in need of a minor overhaul.
I have seen issues where port activity monopolised data-paths/arbitration inside chipsets such that the PIO operations would appear to time out. Since you're doing a raid rebuild, perhaps the I/O load is causing something similar to occur.
I suspect that is exactly the problem I'm seeing. I can see the smart polling on the drive lights when it occurs. It slows the rebuild quite significantly for that period.
I hit the problem using 2.6.10-bk10 also. It's just much harder to hit with that kernel. I'm now trying 2.6.11-rc4 with all the libata patches (the kernel I was using before), but smartd disabled. I have a sneaking suspicion that SMART is the root cause here however I don't see it on the other machine because
A) I'm using RAID-5 and not 6, thus my CPU usage during a rebuild is a LOT lower
B) I have more cards/drives in this machine and a RAID-6 rebuild across 15 drives appears to be quite taxing on the hardware.
Anyway, rebuild started. We will see in 12 hours.
I did note when I get an ata timeout in 2.6.10 it handles it normally. In 2.6.11 it hardlocks the machine. No alt-srq or anything else.
Regards, Brad -- "Human beings, who are almost unique in having the ability to learn from the experience of others, are also remarkable for their apparent disinclination to do so." -- Douglas Adams - To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
