USB hotplugging broken on resume

[Corey Burger]

As of Sep 1, with a fully updated machine, USB devices no longer work on resume.
They receive power, but dmesg does not show them plugging in, nor do they work.

[Matthew Garrett (mjg59)]

Is this resume from RAM or resume from disk?

[Corey Burger (corey.burger)]

Suspend from disk.

[Matthew Garrett (mjg59)]

*** Bug 20892 has been marked as a duplicate of this bug. ***

[Matthew Garrett (mjg59)]

*** Bug 21105 has been marked as a duplicate of this bug. ***

[Matthew Garrett (mjg59)]

*** Bug 20929 has been marked as a duplicate of this bug. ***

[Corey Burger (corey.burger)]

Ok, further testing reveals that this bug only appears 1 in ~5-6 times.

[Martin Pitt (pitti)]

Confirmed on my system. Some more interesting things:

 * my USB devices do not get power after resume, replugging does not help

 * rmmod ohci_hcd works, modprobe ohci_hcd hangs

 * hibernate/resume works perfectly if I do "echo disk | sudo tail
/sys/power/state". When using "sudo pmi action hibernate", I have broken USB and
ethernet after resuming (I file a separate report for the ethernet). Corey, all
the others, maybe you can also try this?

[Martin Pitt (pitti)]

(In reply to comment #7)
> Confirmed on my system. Some more interesting things:
>
> * my USB devices do not get power after resume, replugging does not help
>
> * rmmod ohci_hcd works, modprobe ohci_hcd hangs
>
> * hibernate/resume works perfectly if I do "echo disk | sudo tail
> /sys/power/state". When using "sudo pmi action hibernate", I have broken USB and

Crap. Of course this is supposed to be "echo disk | sudo tee /sys/power/state".

[Arthur Peters (amp)]

I think there are several different problems behind this. For one, my failure
occures every time (not every ~5-6 times as reported by Corey Burger), also my
tests were on AMD64, also Martin Pitt's report appears to be on a MAC which may
be quite different because of the different archatecture.

IMHO, this means that the problem is either very big (effecting many targets and
such) or there are several small problems.

Matthew, what kind of debugging could we do that would help narrow this down to
a solvable problem before breezy's release?

Thanks
-Arthur

[Matthew Garrett (mjg59)]

This should be fixed in the next initramfs-tools upload.

[Matthew Garrett (mjg59)]

Fixed now, with luck.

[Mika Fischer (zoop)]

Fixed for me by initramfs-tools 0.28