Functional Description
236 Datasheet
5.21.2.9 Case for Considerations
Below are some corner cases and some possible actions that the external controller
could take.
Note that a 1-byte sequence number is available to the data read by the external
controller. Each time the PCH updates the thermal information it will increment the
sequence number. The external controller can use this value as an indication that the
thermal FW is actually operating. Note that the sequence number will roll over to 00h
when it reaches FFh.
1. Power on:
The PCH will not respond to any SMBus activity (on SMLink1 interface) until it
has loaded the thermal Firmware (FW), which in general would take 1–4
seconds. During this period, the PCH will NACK any SMBus transaction from the
external controller.
The load should take 1-4 seconds, but the external controller should design for
30 seconds based on long delays for S4 resume which takes longer than normal
power up. This would be an extreme case, but for larger memory footprints and
non-optimized recovery times, 30 seconds is a safe number to use for the time-
out.
Recover/Failsafe: if the PCH has not responded within 30 seconds, the external
controller can assume that the system has had a major error and the external
controller should ramp the fans to some reasonably high value.
The only recover from this is an internal reset on the PCH, which is not visible
to the external controller. Therefore the external controller might choose to poll
every 10-60 seconds (some fairly long period) hereafter to see if the PCH's
thermal reporting has come alive.
2. The PCH Thermal FW hangs and requires an internal reset which is not visible to
the external controller.
The PCH will NACK any SMBus transaction from the external controller. The PCH
may not be able to respond for up to 30 seconds while the FW is being reset and
reconfigured.
The external controller could choose to poll every 1-10 seconds to see if the
thermal FW has been successfully reset and is now providing data.
General recovery for this case is about 1 second, but 30 seconds should be used
by the external controller at the time-out.
Recovery/Failsafe: same as in case #1.
3. Fatal PCH error, causes a global reset of all components.
When there is a fatal PCH error, a global reset may occur, and then case #1
applies.
The external controller can observe, if desired, PLTRST# assertion as an
indication of this event.
4. The PCH thermal FW fails or is hung, but no reset occurs
The sequence number will not be updated, so the external controller knows to
go to failsafe after some number of reads (8 or so) return the same sequence
number.
The external controller could choose to poll every 1-10 seconds to see if the
thermal FW has been successfully reset and working again.
In the absence of other errors, the updates for the sequence number should
never be longer than 400 ms, so the number of reads needed to indicate that
there is a hang should be at around 2 seconds. But when there is an error, the
sequence number may not get updated for seconds. In the case that the
To Next Page
Loading...
Need help?
General
