Skip to contentSkip to conversation

disk I/O hang in 2.6.0-test4-mm5

DISCUSSION Dave Olien 3 participants 5 messages Sep 4, 2003

lore
Share & tools
lore

Dave Olien

author
#0 lore raw

I'm seeing a mkfs.ext2 that never completes under 2.6.0-test4-mm5. I ran 4 mkfs.ext2's concurrntly, each on a seperate partition on the same disk. Three of the completed. Here's the sysrq stack trace from the one that didn't.

This doesn't occur on mm4.

mkfs.ext2 D 00000000 16474 16431 (NOTLB)
f19cbef0 00000086 f6200ce0 00000000 00000003 c0127852 00000000 c3671bc0

       c3671bc0 0000000d f19cbf28 f19cbefc c011fa08 c184ab20 c360d09c c014007c
       c184ab20 00000000 f7a8a040 c0120d70 f19cbf34 f19cbf34 f7a8a040 ffffffff
Call Trace:
 [<c0127852>] tasklet_action+0x72/0xd0
 [<c011fa08>] io_schedule+0x28/0x40
 [<c014007c>] wait_on_page_bit_wq+0xcc/0x100
 [<c0120d70>] autoremove_wake_function+0x0/0x50
 [<c0120d70>] autoremove_wake_function+0x0/0x50
 [<c013fda3>] filemap_fdatawait+0x1e3/0x240
 [<c0162819>] sync_blockdev+0x39/0x50
 [<c0162b52>] sys_fsync+0xc2/0x100
 [<c03106ea>] sysenter_past_esp+0x43/0x65

This is on a system with a DAC960 driver. But we've seen it here on a system with scsi disk as well.

#1 lore raw
Quoted 7 lines · Dave Olien

I'm seeing a mkfs.ext2 that never completes under 2.6.0-test4-mm5. I ran 4 mkfs.ext2's concurrntly, each on a seperate partition on the same disk. Three of the completed. Here's the sysrq stack trace from the one that didn't.

This doesn't occur on mm4.

Could you please revert

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/broken-out/elv-insertion-fix.patch

and retest?

author
#2 lore raw

Sure, I'll give it a shot and get back to you.

Quoted 14 lines · Andrew Morton
Quoted 7 lines · Dave Olien

I'm seeing a mkfs.ext2 that never completes under 2.6.0-test4-mm5. I ran 4 mkfs.ext2's concurrntly, each on a seperate partition on the same disk. Three of the completed. Here's the sysrq stack trace from the one that didn't.

This doesn't occur on mm4.

Could you please revert

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/broken-out/elv-insertion-fix.patch

and retest?

author
#3 lore raw
In reply to Andrew Morton

Yes, reverting this patch fixes the I/O hang I was observing.

Quoted 14 lines · Andrew Morton
Quoted 7 lines · Dave Olien

I'm seeing a mkfs.ext2 that never completes under 2.6.0-test4-mm5. I ran 4 mkfs.ext2's concurrntly, each on a seperate partition on the same disk. Three of the completed. Here's the sysrq stack trace from the one that didn't.

This doesn't occur on mm4.

Could you please revert

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/broken-out/elv-insertion-fix.patch

and retest?

#4 lore raw

How to get a backtrace (sysrq-t) on a specific task?

In reply to Dave Olien
Quoted 5 lines · Dave Olien

I'm seeing a mkfs.ext2 that never completes under 2.6.0-test4-mm5. I ran 4 mkfs.ext2's concurrntly, each on a seperate partition on the same disk. Three of the completed. Here's the sysrq stack trace from the one that didn't.

How do you do a sysrq-T on a specific task???

[snip]

Greets, Antonio.