Various trouble in 2.5.31

Various trouble in 2.5.31

[Helge Hafting]

Some problems and oddities I have seen:

UP, no preempt
* nfs gets stuck sometimes.  klogd and rpciod shares the cpu,
  umounting is impossible.  cat'ing a nfs file shows the
  contents but the cat process won't die.  Stuck on close?
  This also happens with 2.5.30

* ide-cd is broken as module

SMP, with and without preempt.
* During boot I get atkbd not present(ed), and the keyboard 
  is disabled.  It wakes up later in the boot process, so
  no big trouble here.

* raid has trouble, even with 4k maximum bio's.
  The bootup e2fsck complains that it can't find a valid
  superblock on my raid-1 and raid-0 devices, although
  mount has no trouble.  I get logged complaints about
  e2fsck using outdated ioctl's for md.  Strange that
  fsck should needs to know about raid devices.
  Possibly a problem with e2fsck 1.27?

Helge Hafting

Various trouble in 2.5.31

[Trond Myklebust]

>>>>> " " == Helge Hafting <helgehaf@aitel.hist.no> writes:

     > Some problems and oddities I have seen: UP, no preempt
     > * nfs gets stuck sometimes.  klogd and rpciod shares the cpu,
     >   umounting is impossible.  cat'ing a nfs file shows the
     >   contents but the cat process won't die.  Stuck on close?
     >   This also happens with 2.5.30

Could you get me a log from tcpdump and 'echo 1 >/proc/sys/sunrpc/rpc_debug'
when this happens?

Cheers,
  Trond

Re: Various trouble in 2.5.31

[Neil Brown]

On Tuesday August 13, helgehaf@aitel.hist.no wrote:
> Some problems and oddities I have seen:
> 
> * raid has trouble, even with 4k maximum bio's.
>   The bootup e2fsck complains that it can't find a valid
>   superblock on my raid-1 and raid-0 devices, although
>   mount has no trouble.  I get logged complaints about
>   e2fsck using outdated ioctl's for md.  Strange that
>   fsck should needs to know about raid devices.
>   Possibly a problem with e2fsck 1.27?

The warning about ioctl's is fairly bogus.  It gets reported for *any*
unknown ioctl, whether it looks like it might be raid related or not.
So I wouldn't worry about that.

I have a new collection of patches for md which break if fairly
thoroughly.  I am about to started testing/fixing and will
make sure that mkfs/fsck/mount work for raid{linear,0,1,5}.

Thanks,
NeilBrown