xhci: Handle command completion and timeout race
authorMathias Nyman <mathias.nyman@linux.intel.com>
Tue, 3 Jan 2017 16:28:48 +0000 (18:28 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 3 Jan 2017 16:37:32 +0000 (17:37 +0100)
commita5a1b9514154437aa1ed35c291191f82fd3e941a
treed166091e509be9a6b07ed40103deb29c797afbb4
parent2a7cfdf37b7c08ac29df4c62ea5ccb01474b6597
xhci: Handle command completion and timeout race

If we get a command completion event at the same time as the command
timeout work starts on another cpu we might end up aborting the wrong
command.

If the command completion takes the xhci lock before the timeout work, it
will handle the command, pick the next command, mark it as current_cmd, and
re-queue the timeout work. When the timeout work finally gets the lock
It will start aborting the wrong command.

This case can be resolved by checking if the timeout work is pending inside
the timeout function itself. A new timeout work can only be pending if the
command completed and a new command was queued.

If there are no more commands pending then command completion will set
the current_cmd to NULL, which is already handled in the timeout work.

Cc: <stable@vger.kernel.org>
Reported-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/usb/host/xhci-ring.c