[Bug 221073] xHCI host controller dies on resume from s2idle on AMD Strix Halo [1022:1587]

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=221073

--- Comment #34 from Michał Pecio (michal.pecio@gmail.com) ---
Created attachment 309583
  --> https://bugzilla.kernel.org/attachment.cgi?id=309583&action=edit
debug suspend/resume

There are some error flags set on DevSta and changes to MSI capabilities of
other functions on c4:00 device, but IDK if it means anything. Though it would
be suspicious if those changes only occur during problematic resumes.

Mathias suggested that CMD_INTE may be getting cleared inadvertently. I don't
see how it could happen, but just in case here's patch which logs xHCI
registers relevant to interrupts during suspend, resume and on command timeout.

It also checks for pending commands on suspend (shouldn't happen, but who
knows) and explicitly disables and reenables IRQ generation at xHCI layer. I
gave it a quick test and it seems to be OK, but you could also try without
those two blocks which manipulate CMD_EIE in xhci->op_regs->command.

I also tried removing only the part which enables interrupts on resume and got
a failure similar to yours, as expected.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.