Bug 248783 - USB input death then system freeze
Summary: USB input death then system freeze
Keywords:
Status: CLOSED RAWHIDE
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel
Version: rawhide
Hardware: All
OS: Linux
medium
high
Target Milestone: ---
Assignee: Kernel Maintainer List
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
: 248556 (view as bug list)
Depends On:
Blocks: F8Blocker
TreeView+ depends on / blocked
 
Reported: 2007-07-18 19:10 UTC by Nicolas Mailhot
Modified: 2007-11-30 22:12 UTC (History)
2 users (show)

Fixed In Version: 2.6.23-0.174.rc6.fc8
Clone Of:
Environment:
Last Closed: 2007-10-03 20:18:16 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Linux Kernel 8535 0 None None None Never

Description Nicolas Mailhot 2007-07-18 19:10:17 UTC
All the info is in the upstream bug

Comment 1 Tom London 2007-07-19 13:48:08 UTC
*** Bug 248556 has been marked as a duplicate of this bug. ***

Comment 2 Tom London 2007-07-21 18:07:13 UTC
I continue to have USB loss and eventual system freeze with all kernels after
kernel-PAE-2.6.23-0.15.rc0.git1.fc8, including today's
kernel-PAE-2.6.23-0.41.rc0.git14.fc8.  (0.15.rc0.git1 is the last stable kernel
for me).

I do not see any pattern in /var/log messages.

My system is a Thinkpad X60.

What info/debugging can I provide?

Comment 3 Nicolas Mailhot 2007-07-21 22:49:09 UTC
upstream devs want usbmon traces. look at the bugzilla.kernel.org bug

Comment 4 Tom London 2007-07-22 04:37:09 UTC
OK.  Read instructions....

Will setup and collect....

Comment 5 Tom London 2007-08-08 14:01:55 UTC
I've been applying the patch filed upstream for the past several kernel
packages. 'Fixes for me'.

For me, the issue seems to be 'slow USB' devices (e.g., mouse/keyboard) on EHCI
ports.

The patch refers to nVidia controllers, although I believe I have 'slow devices'
connected to Intel ICH7 EHCI through the Thinkpad X60 dock.

Both upstream and Fedora patches attached here:
http://bugzilla.kernel.org/show_bug.cgi?id=8535

I'm not sure what more to do to push this upstream, or if this is already
working its way through the process.

Comment 6 Tom London 2007-08-23 16:43:04 UTC
This appears to 'work for me' with kernel-2.6.23-0.129.rc3.git4.fc8.

That likely, or am I just seeing random stuff....

Comment 7 Nicolas Mailhot 2007-08-23 17:03:36 UTC
I don't think you're seeing stuff, that's the first post 2.6.22 rawhide kernel
that seems to work here (but this bug didn't always trigger fast on my system,
so it's a bit early to declare the bug closed)

Comment 8 Pete Zaitcev 2007-08-23 18:01:58 UTC
The matter is being worked upon by Stuart. Linus backed out his original
fix for EHCI losing data when CPU frequency changed, so this is probably
what you two are seeing.

Start around here and navigate down thread:
http://lkml.org/lkml/2007/8/20/369

Comment 9 Will Woods 2007-10-03 19:58:58 UTC
I hit something similar to this recently, on a 2.6.23-rc7 kernel. The upstream
bug is months old and still in NEW. Is this thought to be fixed upstream, or are
we still waiting for a patch to land?

Comment 10 Will Woods 2007-10-03 20:03:37 UTC
Err, of course, there is another possible explanation: I am seeing something
entirely different and should file that bug elsewhere. I checked the logs and
the telltale "drivers/hid/usbhid/hid-core.c: input irq status -75 received"
message is not there. So probably I'm seeing something else.

Still, I'd like to know if this is supposed to be CLOSED UPSTREAM or if it's
still an open issue.

Comment 11 Nicolas Mailhot 2007-10-03 20:10:56 UTC
IMHO it's closed upstream or at least I'm not hitting it anymore on my system
(both on mm and fedora kernels)

Comment 12 Pete Zaitcev 2007-10-03 20:18:16 UTC
Will, this bug tracks what was happening to Nicolas, e.g. the fallout from
Stuart's attempts to fix EHCI coughing up a lung upon a CPU frequence change.
If you hit a specific issue with USB, please file a new bug. Likely the root
cause is different.

I'll close this bug now to avoid confusion.


Note You need to log in before you can comment on or make changes to this bug.