stm32/usb: Handle spurious wakes in usb_resume

Sometimes, usb_resume gets called, but the D+/D- lines do not
indicate an actual resume event (e.g. during ESD discharge).

Reference manual tells we should go back to sleep if state is 10 or
11. However, setting FSUSP and LP_MODE in this interrupt routine
seems to lock the USB controller (see b/35775088 and b/71688150).
Instead, we do it in a deferred routine. The host must assert the
reset condition for 20ms, so reading D+/D- after ~3ms should be safe
(there is no chance we end up sampling during a bus transaction).

BRANCH=none
BUG=b:35775088
BUG=b:67766202
BUG=b:71688150
TEST=On staff, with USB autosuspend enabled on soraka, short
     D+/D- lines, which causes a spurious wake event. After that
     remote wake-up still works.
TEST=Repeat test with ESD discharge.
TEST=Repeat test with plugging/unplugging of USB-C monitor.

Signed-off-by: Nicolas Boichat <drinkcat@chromium.org>
Change-Id: I0f2697d1fa5b68356fd8a4fc16eaab5eadad9086
Reviewed-on: https://chromium-review.googlesource.com/868093
Reviewed-by: Vincent Palatin <vpalatin@chromium.org>

0 files changed, 0 insertions, 0 deletions