The most recent failure I had was with the kernel in an Ubuntu
edgy install.  I was trying to copy files to a 2.5 inch drive
in a USB interface box.  My friend (quoted below) suggested
the fail sounded like what he has seen.

------------------------------------------------------------------------
Hi Allen,

My experience has been that all kernels are bad, just some are worse than
others (sometimes much worse).  The type of access has always been rsync
backups of my server hard drive to the USB disk.  Most of the time the
external drive had most of the data on it already, but usually about 10GB
would be transfered each time, and of course the entire drive directory
structure would be traversed at the start of the rsync.  In most cases
where the drive locked up, it would just trash the USB mass storage
subsystem, but in some cases it would lock up the kernel as well leaving
the system dead.  It does make a difference which external drive you use
as the lockups are at least partly due to how the drive's USB interface is
implemented, I have two different (older) external drives is much more
likely to lock up the USB subsystem, while the other seems to be more
likely to lock up the kernel.  I have had problems that I can recall with
the following kernels: 2.4.22, 2.4.27, 2.6.8, 2.6.10, 2.6.12, 2.6.15,
2.6.16, and 2.6.18.  I don't recall which ones had a kernel lockup occur
as that happened far less often than USB subsystem lockups, however, I
know that it occurs with 2.6.18 as it happened just a couple weeks ago.
If I had to estimate, I would say that problems occurred about once every
100GB of data transferred overall (virtually all writes for my usage),
though on some kernels a few gigabytes was enough (needless to say I
didn't stick with those kernels for very long).

FWIW.

[name deleted]
-----------------------------------------------------------------------

His experience is more extensive than mine.  My attempt to send
10s of GB to that 2.5 inch drive was more than I had ever written
to USB before.
-- 
Allen Brown
http://brown.armoredpenguin.com/~abrown

> Hmmm, Odd.  I do a lot with both external USB 2.0 HDDs and I read/write
> a lot of 1-2GB CF cards.  I've never experienced this sort of lockup.
> Maybe I've just been lucky.  I've done this with Suse, Ubuntu, & Fedora
> on a variety of boxes ranging from my own company's USB 2.0
> implimentation, to Nforce based MB chipsets, to whatever flavor of the
> month Dell decided to put in the box I've got to work on this week.  Is
> there a specific pattern of activity you could follow to get the failure
> to occur?  I'd love to test my HW with it if so.
>                       -Mike
>
> On Thu, 2007-12-13 at 17:36 -0800, Allen Brown wrote:
>> That doesn't match my experience.  Yes, UHCI should be a complete
>> description.  But not every manufacturer's UHCI is created equal.
>>
>> But I've painfully noticed the problem you mention about USB hubs.
>> At least with 2.0.  I don't have a working USB 2.0 hub.
>>
>> The problem with USB was that the storage device would randomly
>> lock up after awhile.  With older kernels you would be lucky to
>> transfer a few gigabytes.  I don't know what the mean time to fail
>> on the newer kernels are.
>>
>> The lockups were so bad that you had to reboot.  You couldn't
>> remove the module and reload.
>



-- 
Allen Brown
http://brown.armoredpenguin.com/~abrown

_______________________________________________
EUGLUG mailing list
[email protected]
http://www.euglug.org/mailman/listinfo/euglug

Reply via email to