USB stress test failure on AMD SB600

Linux USB guys submit another patch to make things better:
http://marc.info/?l=linux-usb&m=120469059715031&w=2

And that new patch can also replace this one:
http://lkml.org/lkml/2008/2/19/546

So, I backported and attach this new patch here.
Can you apply it as well as the above patch?
You need apply the above patch at first, then this patch.

Please also note that these two patch is needed if your kernel <= 2.6.24
because they will be included from kernel 2.6.25

Thanks

I'm sorry that I attached another patch by mistake just now,
please use this one to replace the above one "Another USB patch to make things better"

Hi Shane,

Thanks for the upstream references, it's very helpful. I've reassigned this to the kernel team for their consideration. Thanks.

 @Shane

thanks for the references. We try to get those integrated with Hardy, however this might be after beta.

The third patch http://marc.info/?l=linux-usb&m=120469059715031&w=2
has been added to linus source tree too, which is:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=e82cc1288fa57857c6af8c57f3d07096d4bcd9d9

But there is some small difference between them,
the latter(git commit) also modified another line as below:

@@ -781,7 +811,7 @@ static int ehci_urb_enqueue (
 static void unlink_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
 {
        /* failfast */
- if (!HC_IS_RUNNING(ehci_to_hcd(ehci)->state))
+ if (!HC_IS_RUNNING(ehci_to_hcd(ehci)->state) && ehci->reclaim)
                end_unlink_async(ehci);

        /* if it's not linked then there's nothing to do */

Can you also modify it? I will NOT do any special backport for this line alone.

Thanks

Again, two patches plus one line modification should be added to fix the bug:
1. "backported patch to fix USB stress test failure"
2. "to replace the above attached patch by mistake"
3.
- if (!HC_IS_RUNNING(ehci_to_hcd(ehci)->state))
+ if (!HC_IS_RUNNING(ehci_to_hcd(ehci)->state) && ehci->reclaim)

If Hardy will use final stable 2.6.25 in the future, all these patches will be
included automatically, we need not backport any of them.

Just found this issue when I was about to report my ongoing problems with SB600 USB support (symptoms as described, with an AMD 690G board, 32-bit install with powernowd enabled, and fglrx and VMWare modules installed -- all of which had me losing hope that I'd ever track down the bug). Please let me confirm that this is a major issue for users and let me thank all of you (and all involved on the Kernel.org bug and LKML threads) for working on it.

FWIW, the misbehavior is being observed with 2.6.22-14-generic #1 SMP Tue Feb 12 07:42:25 UTC 2008 i686 GNU/Linux on 7.10.

Cherry picked the upstream patches provided by Shane,

Just double check on this issue: Have you added all these three patches into Hardy?

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=07d29b63ef6b39963ab37818653284d861cf55af

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=f8fa7571a928d6d0e1b7444b0ea69ec7dc7db3b6

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=e82cc1288fa57857c6af8c57f3d07096d4bcd9d9

Thanks

Yes, I applied all three.

Stefan

This bug was fixed in the package linux - 2.6.24-13.23

---------------
linux (2.6.24-13.23) hardy; urgency=low

  [Alessio Igor Bogani]

  * rt: Updated configuration files

  [Ben Collins]

  * openvz: New custom flavour for OpenVZ
  * config: Disable IDE AMD driver in favor of PATA version
    - LP: #181561
  * config: Disable IDE VIA driver in favor of PATA version
    - LP: #181561
  * drivers/video: Restore gutsy backlight dimming behavior
    - LP: #205261
  * build/config: Enable CONFIG_CIFS_WEAK_PW_HASH
    - LP: #202445

  [Colin Ian King]

  * SAUCE: Add support for version 4 of Chelsio NICs in cxgb3 driver
    - LP: #201893

  [Kees Cook]

  * AppArmor: re-add missing "type" field in syslog reports.
    - LP: #202888
  * kvm: reset TSS on x86_64 to avoid ioperm bitmap corruption
    - LP: #144900

  [Stefan Bader]

  * USB: EHCI: add separate IAA watchdog timer
    - LP: #198619
  * SAUCE: Always use SCO protocol (disable eSCO support)
    - LP: #39414
  * PM: Introduce PM_EVENT_HIBERNATE callback state
    - LP: #201086

  [Tim Gardner]

  * Disable DRM suspend/resume on pre-915 Intel chips
    - LP: #207496
  * frame buffer regression - screen blank except for blinking cursor after fbcon
    vtswitch
    - LP: #201591

 -- Tim Gardner <email address hidden> Wed, 19 Mar 2008 10:05:05 -0400