Re: [linux-usb-devel] Re: PATCH: usb-storage: make internal structsmore consistent

[David Brownell]

I saw changes in how 2.5.51 (*) behaved:

- One of the "buffer layer error" messages vanished,
but the fs/buffer.c:1163 one still seems to happen.

- New "scsi HBA driver usb-storage didn't set max_sectors,
please fix the template" message on device probe.

- SCSI shows differently in sysfs, so the hotplug
scsi.agent needs to change a bit.

- "mkfs -c -c" (destructive badblocks) is still wierd.

Big delays ... maybe this is a version of what you saw.
I'm thinking that the "oops" mode morphed into this.

* a "sync" during the badblocks took 15+ minutes;
same thing later, while mkfs said it was writing
out the filesystem data, took even longer.

* oops hasn't happened, yet, but somethings really
wierd when "ls -l" on a directory hangs ... the
contents should all be in the cache, but then
"ls" showed the directory was wrongly empty.
(10+ minutes since a "tar xfj ..." returned.)

* in fact something was clearly wrong in the "mount
after 'mkfs -c -c'" scenario ... I ^C the test;
something in the kernel seems to be waiting a lot
for timeouts, with no I/O pending!

* at which point the totally wierd thing is that the
files in /dev/sda1 no longer showed ... seems like
something's clearly broken above usb-storage, since
even the _mounted_ filesystem root is empty.

* "less async" (to see the ehci control/bulk queues)
acted wierd during 'badblocks'. Sometimes it doesn't
even try to produce a 4KB page, several times it
segfaulted, once it logged me out (!). Maybe sysfs
isn't handling memory shortages right, or maybe it's
just some kind of memory corruption (512 MBytes).

Debug printks showed me that the only errors this disk
was getting (mostly different sectors than last time)
were flagged by about ten stalled reads of 13 bytes.
Repeated quite often.

That sounds like request status, and retrying, which
seems like what I'd expect to happen on disk errors,
except that they printed as five pairs of two.

- Dave

(*) Actually 2.5.51 with the 'core-1209' and 'sync-1209'
patches I've sent you, plus a diagnostics tweak to
usb-storage. Behavior of usb-storage and ehci should
not have changed.




-------------------------------------------------------
This sf.net email is sponsored by:
With Great Power, Comes Great Responsibility Learn to use your power at OSDN's High Performance Computing Channel
http://hpc.devchannel.org/
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel