[Bug 220069] [6.13.9] regression USB controller dies

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=220069

Mathias Nyman (mathias.nyman@linux.intel.com) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mathias.nyman@linux.intel.c
                   |                            |om

--- Comment #5 from Mathias Nyman (mathias.nyman@linux.intel.com) ---
The symptoms still hint to be the regression caused by:

36b972d4b7ce usb: xhci: improve xhci_clear_command_ring()

Which  was fixed by Michal in:
c7c1f3b05c67 usb: xhci: Fix host controllers "dying" after suspend and resume

Regression was caused by command cycle confusion, meaning controller stops
fetching commands from the ringbuffer as it incorrectly assumes that commands
queued after cycle mismatch are old commands from last cycle on ring buffer.
Driver keeps queuing new commands and hardware keeps ignoring them.

Regression was first noticed when a dedicated timeout timer for "stop endpoint"
command did not complete in time, and timer handler assumed host is dead.

In this case it's the generic command timer that times out before a "stop
endpoint" timer. xhci driver tries to recover by aborting the failing command,
but abort likely fails as well as hardware isn't really processing any commands
at the moment, and we end up assuming host is dead.

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

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