On specific revisions of the ASUS KGPE-D16 (> 1.03G) there is a
high (< 1:10) chance of lockup from spurious HW monitor IRQs
during LPC configuration. This was originally erroneously identified
as a bug within the SP5100 southbridge due to serial console buffering
moving the hang slightly before HW monitor setup. It is currently
unknown how changing the CBFS layout / code size was able to alter
the frequency of the lockup occuring; this odd characteristic made
debugging extremely difficult, and it also indicates testing
across multiple PCB revisions will be neded to verify that the
bug has been completely resolved.
It is highly likely that the KCMA-D8 is also affected. As there
does not seem to be a reason to keep the HW monitor IRQ enabled,
simply disable it on both mainboards.
This configuration has passed burn-on power cycle testing with
no lockups noted. All other tests noted a lockup in under 25
power cycles or so, with failure typically occuring in under 5
power cycles; the affected Rev. 1.04 KGPE-D16 has cycled 25 times
times using this patch with only one failure finally noted. This
final failure may have in fact been related to SP5100 Erratum 18
as the frequency is more in line with the errata document guidelines.
Change-Id: Ie9f4f37d2c7dfad0a02daff8b75cd2a1e6f1b09a
Signed-off-by: Timothy Pearson <tpearson@raptorengineeringinc.com>
Reviewed-on: https://review.coreboot.org/14333
Tested-by: build bot (Jenkins)
Tested-by: Raptor Engineering Automated Test Stand <noreply@raptorengineeringinc.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Martin Roth <martinroth@google.com>
7f53b98112