Thanks for helping me debug this!

On Mon, Jun 30, 2003 at 03:02:32PM -0400, Alan Stern wrote:
> > http://bostoncoop.net/adam/temp/kernel-log-2.5 
> I looked at this.  It doesn't show any problems at all.  Did you see 
> anything go wrong at the time this log was written?

I was experiencing some of the problems when that log was written, but I
think I left out some important things by grepping only for usb.  

When I mount or unmount the device, I see this in syslog (and on the
console): 

Jun 28 00:32:30 joehill kernel: SCSI error: host 1 id 0 lun 0 return code = 8000002
Jun 28 00:32:30 joehill kernel: ^ISense class 7, sense error 0, extended sense 0

As long as I don't read or write to the device, I can mount and unmount
(in 2.5.73) repeatedly with no problems other than this error message.  

> How about sending the debugging output for when you try to unmount the 
> drive?

I'm having some trouble figuring out what constitutes an abnormal message
amongst the verbose debugging messages in syslog.  

I just tried copying a bunch of data to the device. After 50M or so, it
slowed to a crawl. Initially, I thought it was crashed but eventually it
started moving again, very slowly (probably around 10K-100K / sec).  The
log from this operation is here: 

http://bostoncoop.net/adam/temp/log_while_copying

Here is the log from the point where I tried unsuccessfully to unmount
the device: 

http://bostoncoop.net/adam/temp/unmount_log

Interestingly, while the unmount process was stuck, I ran df, and got: 

Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/hda2             18313076  17338432     44388 100% /
/dev/sda1             18313076  17338432     44388 100% /mnt/neuros

/dev/sda1, the USB drive, is now returning to same size as the hard drive
(in reality, it's a 128M drive, which is the capacity it returned before
I started the unmount).  

After several minutes where the unmount process was stuck, I finally
removed the module, which caused lots of Buffer I/O errors, a filesystem
panic, etc.. There is a call trace log as well.  Here is the log from
the point of removing the module until the time it was finally done: 

http://bostoncoop.net/adam/temp/remove_module_log 

> Don't worry about data corruption for now; it might just be a side-effect 
> from this problem.  Or it might be something else entirely unrelated; I've 
> heard of at least one device where it turned out the manufacturer 
> acknowledged the firmware had bugs that would cause data corruption.

Okay, corruption doesn't really hurt anyway--it's just holding mp3 files
stored elsewhere.

Incidentally, if are able to isolate this to a firmware bug, my
expectation is that Neuros would be open to suggestions.

> > Also, frequently the drive appears to be full when it's not.  Although
> > it's a 128M memory card, it usually maxes out around 60-70M.  
> What do you mean, appeared to be full?  What did 'df' show?

df showed 0M free... I can't replicate it now, so I don't remember what
it thought the capacity was at the time. I suspect this is related to
filesystem corruption on the drive rather than anything else... Space is
occupied by files that were copied to the device but don't actually show
up.

> Again, debugging output and the kernel log from the segfault would be 
> useful.

Hopefully the logs above will provide more insight into the situation.
Is there debugging output I should send other than what usb-storage is
putting in syslog?

--Adam Kessel

Attachment: pgp00000.pgp
Description: PGP signature

Reply via email to