2.4.21-rc6 ide-scsi bug?

2.4.21-rc6 ide-scsi bug?

[Alexandre Pereira Nunes]

Hi,

My system (athlon on an asus mb with via kt133 chipset) is currently 
running 2.4.21-rc6 (but I had similar behaviour with 2.4.21-rc2-ac2), 
and I map my ide cdrom to ide-scsi by passing option ide-cd ignore in 
modules.conf (devfs is enabled, so that the modules load at /dev/cdrom 
request).

Sometimes my cd drive (actually a mdma2 enabled combo dvd reader/cdrw 
writer lg 4120B) seems to 'sleep' after a long time of inactivity and 
waking it up (i.e. by trying to mount the unit) takes a time, so 
eventually the system has to do a "soft" bus reset so that it comes to 
life again (maybe it actually does that because the drive takes too long 
to wake up by itself, but I don't know). If I use ide-cd instead of 
ide-scsi, it happens as described and the system works ok (the kernel 
prints dma disabled but if I enable it back it works fine), but with 
ide-scsi, after the kernel prints that atapi reset and dma disabled 
stuff, the system hangs. nothing else is printed, and my caps lock and 
scroll lock keys starts blinking, so it seems to be a serious kernel 
panic, but nothing about that is printed on the screen even when I'm at 
console.

Older kernels i've tried (2.4.20 and 2.4.21-pre7) seems to behave ok.

I work around by using ide-cd until I need ide-scsi for something (like 
tunning the rpc2 features on the drive to change region settings for dvd 
playing), and replacing the driver by hand after making sure the drive 
is ok, but that actually sucks :-)

Thanks in advance,

Alex

Re: 2.4.21-rc6 ide-scsi bug?

[Byron Stanoszek]

On Sat, 31 May 2003, Alexandre Pereira Nunes wrote:

> Hi,
> 
> My system (athlon on an asus mb with via kt133 chipset) is currently 
> running 2.4.21-rc6 (but I had similar behaviour with 2.4.21-rc2-ac2), 
> and I map my ide cdrom to ide-scsi by passing option ide-cd ignore in 
> modules.conf (devfs is enabled, so that the modules load at /dev/cdrom 
> request).
> 
> Sometimes my cd drive (actually a mdma2 enabled combo dvd reader/cdrw 
> writer lg 4120B) seems to 'sleep' after a long time of inactivity and 
> waking it up (i.e. by trying to mount the unit) takes a time, so 
> eventually the system has to do a "soft" bus reset so that it comes to 
> life again (maybe it actually does that because the drive takes too long 
> to wake up by itself, but I don't know). If I use ide-cd instead of 
> ide-scsi, it happens as described and the system works ok (the kernel 
> prints dma disabled but if I enable it back it works fine), but with 
> ide-scsi, after the kernel prints that atapi reset and dma disabled 
> stuff, the system hangs. nothing else is printed, and my caps lock and 
> scroll lock keys starts blinking, so it seems to be a serious kernel 
> panic, but nothing about that is printed on the screen even when I'm at 
> console.
> 
> Older kernels i've tried (2.4.20 and 2.4.21-pre7) seems to behave ok.
> 
> I work around by using ide-cd until I need ide-scsi for something (like 
> tunning the rpc2 features on the drive to change region settings for dvd 
> playing), and replacing the driver by hand after making sure the drive 
> is ok, but that actually sucks :-)

I've just had this problem twice today when trying to cat /dev/scd0 past the
end of the DVD. It would stop after sector 716744 (1024-byte sectors) and hang
completely. After 1 minute, the device would do a soft reset and instantly I
get an oops, and the caps/scroll lock keys would start blinking. Couldn't save
the oops because of the panic in the interrupt handler.

This happened because I had written a 700MB file raw to the cd prior (716800
sectors total) and I wanted to read the whole thing back to make sure it was
stored successfully.  Now I'm out in the mkisofs code trying to figure out what
kind of padding I need at the end of the 700MB so it'll read the rest of it.

Note that a non ide-scsi device will just report I/O Error after reading the
end of the CD, however with an ide-scsi device, it will either hang or crash.

I'm currently using 2.4.20-pre5-ac1 with Andre Hedrick's IDE Patch applied. So,
I assume this has been around for some time. Thanks for sending your email--now
I know not to upgrade just yet.

I hope this is of help to someone. :)

 -Byron

-- 
Byron Stanoszek                         Ph: (330) 644-3059
Systems Programmer                      Fax: (330) 644-8110
Commercial Timesharing Inc.             Email: byron@comtime.com

Re: 2.4.21-rc6 ide-scsi bug?

[Mike Fedyk]

On Sat, May 31, 2003 at 03:57:11PM -0400, Byron Stanoszek wrote:
> I'm currently using 2.4.20-pre5-ac1 with Andre Hedrick's IDE Patch applied. So,

Do you need this patch?  Why isn't Alan's code enough?

If Andre isn't sending his patches or Alan hasn't integrated them then maybe
Alan has some plans for it.

Also, maybe the patch is incompatible with the latest -ac?

Re: 2.4.21-rc6 ide-scsi bug?

[Byron Stanoszek]

On Sat, 31 May 2003, Mike Fedyk wrote:

> On Sat, May 31, 2003 at 03:57:11PM -0400, Byron Stanoszek wrote:
> > I'm currently using 2.4.20-pre5-ac1 with Andre Hedrick's IDE Patch applied. So,
> 
> Do you need this patch?  Why isn't Alan's code enough?

Um, no, I just never upgraded my kernel since 2.4.20-pre5-ac1. At that time,
ide-scsi didn't work in the -ac series, and Andre Hedrick had initially
released his patches for it to work.

I recommend sticking with the latest -ac version regardless. :) It has most of
Andre's patches in anyway, with everything else fixed--except for this one
problem.

 -Byron

> 
> If Andre isn't sending his patches or Alan hasn't integrated them then maybe
> Alan has some plans for it.
> 
> Also, maybe the patch is incompatible with the latest -ac?
> 

-- 
Byron Stanoszek                         Ph: (330) 644-3059
Systems Programmer                      Fax: (330) 644-8110
Commercial Timesharing Inc.             Email: byron@comtime.com

Re: 2.4.21-rc6 ide-scsi bug?

[Alan Cox]

On Sul, 2003-06-01 at 00:00, Byron Stanoszek wrote:
> I recommend sticking with the latest -ac version regardless. :) It has most of
> Andre's patches in anyway, with everything else fixed--except for this one
> problem.

To do anything about it I need a call trace, that means causing the crash in console
mode I suspect

Re: 2.4.21-rc6 ide-scsi bug?

[Roland Mas]

Alan Cox, 2003-06-01 00:06:35 +0100 :

> On Sul, 2003-06-01 at 00:00, Byron Stanoszek wrote:
>> I recommend sticking with the latest -ac version regardless. :) It has most of
>> Andre's patches in anyway, with everything else fixed--except for this one
>> problem.
>
> To do anything about it I need a call trace, that means causing the crash in console
> mode I suspect

I seem to have the same problem, or at least a very similar one.  I
reported it in a message with subject "[OOPS] ide-ops:1262 in
2.4.21-rc3", posted to LKML last Sunday.  You might like to read that
message (it includes a ksymoops'ed panic trace).  I have since then
been able to reproduce with 2.4.21-rc6.  I will try to get a more
detailed analysis today, and post it tonight.

Roland.
-- 
Roland Mas

[...] Des fois, y'a des trous. -- (Ptiluc)
  -- Signatures à collectionner, série n°2, partie 3/3.

Re: 2.4.21-rc6 ide-scsi bug?

[Roland Mas]

Roland Mas, 2003-06-01 13:09:58 +0200 :

[...]

> I seem to have the same problem, or at least a very similar one.  I
> reported it in a message with subject "[OOPS] ide-ops:1262 in
> 2.4.21-rc3", posted to LKML last Sunday.  You might like to read that
> message (it includes a ksymoops'ed panic trace).  I have since then
> been able to reproduce with 2.4.21-rc6.  I will try to get a more
> detailed analysis today, and post it tonight.

Done, in a message with subject "[OOPS] ide-scsi panic in 2.4.21-rc2,
-rc3, -rc6, -rc6-ac1", Cc:ed to various people.

Roland.
-- 
Roland Mas

C   c e    e l    m  re q   j  l   a l  l  iè e  .
  -- Signatures à collectionner, série n°1, partie 3/3.