2006-02-02 14:36:28

by Eric Buddington

[permalink] [raw]
Subject: 2.6.15-mm3 hang with raid-1

Running 2.6.15-mm3 PREEMPT on an Athlon XP, Asus motherboard with
VT8366/A/7 chipset.

I have a 240GB raid0 on hda3 and hdc1 containing a reiser3
filesystem. It works fine.

I have a 40GB raid1 on hda2 and hdd1, which is giving me problems. I
can write to all of hda2 and hdd1 (before creating the array) without
error, and mdadm seems to create the array all right.

However, just after the array finishes synchronizing (either after
creation or messy reboot), all new commands hang (I can still type at
existing shells, but as soon as I try to run a program, it becomes
unresponsive). While hung, disk activity light is off. On reboot, the
array is detected as clean, no synchronization is done, and my system
works (until I try to use the raid1 array).

mkreiserfs /dev/md0 also causes the hang, when it syscalls fsync().

My system disk is hda1, so I don't know if all disks are hung up, or
just all of hda.

I'm doing all this at the console, and I see no error messages.

The computer keeps performing NAT, and alt-sysrq-b works.

What can I do to fix or help debug this problem?

-Eric


Attachments:
(No filename) (1.07 kB)
(No filename) (189.00 B)
Download all attachments

2006-02-05 17:26:31

by Eric Buddington

[permalink] [raw]
Subject: Re: 2.6.15-mm3 hang with raid-1

On Sun, Feb 05, 2006 at 12:14:24AM -0800, Andrew Morton wrote:
> Eric Buddington <[email protected]> wrote:
> >
> > Running 2.6.15-mm3 PREEMPT on an Athlon XP, Asus motherboard with
> > VT8366/A/7 chipset.
> >
> > I have a 240GB raid0 on hda3 and hdc1 containing a reiser3
> > filesystem. It works fine.
> >
> > I have a 40GB raid1 on hda2 and hdd1, which is giving me problems. I
> > can write to all of hda2 and hdd1 (before creating the array) without
> > error, and mdadm seems to create the array all right.
> >
> > However, just after the array finishes synchronizing (either after
> > creation or messy reboot), all new commands hang (I can still type at
> > existing shells, but as soon as I try to run a program, it becomes
> > unresponsive). While hung, disk activity light is off. On reboot, the
> > array is detected as clean, no synchronization is done, and my system
> > works (until I try to use the raid1 array).
> >
> > mkreiserfs /dev/md0 also causes the hang, when it syscalls fsync().
> >
> > My system disk is hda1, so I don't know if all disks are hung up, or
> > just all of hda.
> >
> > I'm doing all this at the console, and I see no error messages.
> >
> > The computer keeps performing NAT, and alt-sysrq-b works.
> >
> > What can I do to fix or help debug this problem?
> >
>
> Could you test -mm5 please?

using 2.6.16-rc1-mm5:
- creation of reiserfs on resyncing raid1 works.
- resync completes without hanging the system.
- creating of reiserfs on resynced raid1 works

Looks good - thanks!

-Eric