Caveh Jalali 4a8b509020 usb_pd: only service interrupts on enabled ports. ...

as it turns out, we're pretty aggressive in iterating over all TCPCs
checking for alerts when any of them generate an interrupt or their
interrupt line is low.  this can cause unfortunate behavior if the
driver hasn't initialized itself (and the chip) yet for interrupts to
be handled or we've released (disconnected) the driver so we can do a
TCPC firmware update.  so, check the PD task state to see if it makes
sense to service the port's interrupt.

note: there seems to be a quirk with the ps8751 in that it holds its
ALERT# (interrupt) line low during firmware update.  this line is
supposed to be falling edge triggered, so it's technically not
interrupting, but since we also poll the line level, we think there's
a continuous interrupt that isn't acutally there.  we get away with
this because pd_exchange_status() has a 5ms delay in its polling loop
to avoid spinning.

the particular test case was to unplug the PD power brick during TCPC
firmware update (over i2c).  the interrupt handler would be called,
accessing the TCPC over i2c and causing all sorts of havoc.

TEST=tested with follow-up CLs and verified ps8751 firmware update
	works on electro.

BRANCH=none

BUG=b:35586896

Change-Id: I880cff49e0e9637256efa9003bcc46274536e631
Signed-off-by: Caveh Jalali <caveh@google.com>
Reviewed-on: https://chromium-review.googlesource.com/544661
Reviewed-by: Stefan Reinauer <reinauer@chromium.org>
Reviewed-by: Shawn N <shawnn@chromium.org>

2017-06-30 16:02:50 -07:00

47 KiB

Raw Blame History

View Raw