At first we pulled the hub but still had the problem.
Then compiled and ran 2.6.12-rc5 with patches noted
below and the USB2 storage device now seems to be
working. On the other hand, we've only moved about
35G (3 DV files) so far. I'll post a follow up note if things
break while running with the patches, but, for now,
hopeful that everything is fixed. THANKS!
It is hard to find out what the actual device is in
a store box item, e.g. PCI USB card. Could someone post
a couple of names of card/model that would have the
mentioned NEC chip set? If current setup breaks
that's my next move.
Pieter, cc'ing you because these patches may solve
the problem you researched awhile back and wrote up
on your web page.
Alan, from my perspective, if your patches fix this
problem then those patches belong in the production
kernel ASAP, i.e. 2.6.12. If not 2.6.12 official, then all
of the distros should roll the patch in. The bug
(and I understand the problem is in out-of-spec hardware)
is nasty when it bites.
What we are doing:
turn on usb2 drive
it auto mounts (ubuntu)
umount it
fsck.ext3 -f it
mount it
cp -p files over to it
umount it
fsck.ext3 -f it
if it is clean then rm the files that were copied
etc.
Have to until we've managed to move 3 or 4 hundred G
back and forth without a problem. Then I'll start to relax.
This is a hassle. But we cannot afford to lose the files.
If an enterprise got bit by this problem and lost data
they aren't going to leave Linux on their system. Hence
my urging the kernel gods to roll the patch in quickly.
And, thanks again for your help.
Heitzso
Moving lots of very large 10G and 12G files back and forth
from ubuntu (current) system to 3 SilverRiver (thermaltake)
USB2 enclosures with new 300G WD caviar drives ....
formatted ext3 ... ends up with corrupted filesystems.
Everything may work well for 100G or 150G before it breaks.
Or it may break after 25G worth of file(s).
If I take the same drive and put it inside the case connected
up directly it works fine.
There are a number of kernel bugs reported, patches floating
around, etc.
The one easy fix is to avoid using the high speed connection,
but dropping back to a slower speed and moving 150G
worth of files is a no-op. (though better than getting beaten
to a pulp by a corrupted file system)
Does anyone know ... what kernel(s) have the best chance
of fixing this? I can arrange to compile 2.6.12-rc5 on this
box if it has a chance of fixing the problem. Ditto 2.6.11.11.
These are current stable and current rc as of date of this
email.
I'm monitoring 2 kernel bugs that flag this, or related,
problem(s) and they're not showing closure.
Advice greatly appreciated.
As you say, a lot of people have been having these sorts of problems.
It's hard to pin it down to a software bug because in many cases changing
one of the hardware components solves the problem. For example, a $10
add-on PCI USB controller card has been known to make a big difference
(ones based on the NEC chipset are said to be quite reliable).
Many USB 2.0 - IDE interface chips are slightly buggy. Ditto for many USB
2.0 host controller chips. Or if not buggy, then they have rather tight
tolerances for, or rather loose control over, low-level things like bit
timings.
Some people experience only occasional problems, but it's enough to
disrupt their workflow. Other people reliably experience problems almost
as soon as they start using their USB device. It's hard to know exactly
what's going wrong; in none of the cases has anyone been in a position to
use a hardware analyzer to see exactly what's happening. The exact
details are probably different for each case, anyway.
Given that we don't know how to prevent errors, the best we can do is to
try and recover from them as quickly as possible (provided they don't
occur too often). A pair of patches for usb-storage may help with error
recovery:
https://lists.one-eyed-alien.net/pipermail/usb-storage/2005-April/001565.html
https://lists.one-eyed-alien.net/pipermail/usb-storage/2005-April/001558.html
These will apply to 2.6.12-rc5. They've been known to help some people a
lot, others not so much. Hopefully they will be accepted into the kernel
when 2.6.13 comes out.
-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games. How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
linux-usb-devel@lists.sourceforge.net
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
