Re: [REGRESSION] qcom PAS TZ API migration breaks GPU and modem on TrustZone without PAS (sc7180 trogdor)

[Paul Hollinsky]

On 8/21/26 08:33, Thorsten Leemhuis wrote:
> Just wondering: who will submit the proper fix to ideally fix this
> before the culprit hits mainline

I have now. I sent them as two separate patches, because the two culprits
are in different places:

  drm/msm/adreno: Only check for PAS when a zap shader is present
  https://lore.kernel.org/r/20260821081325.89088-1-phollinsky@xxxxxxxxxxxxxx
    0be72be03ca7 has in fact already reached mainline, in this merge
    window, via the soc-drivers-7.3 pull (Linus' merge 368cf60c36a3).
    So this one is based on Linus' tree rather than on msm-fixes, which
    still predates the culprit - it went in through the qcom tree, not
    through msm. It is addressed to the msm maintainers, but it can just
    as easily go back through the tree the culprit came from.

  remoteproc: qcom_q6v5_mss: Don't require PAS for memory protection
  https://lore.kernel.org/r/20260821081540.89137-1-phollinsky@xxxxxxxxxxxxxx
    f3b1357673dd is still only in remoteproc/rproc-next, so this one is
    based on that branch and would ideally be picked up there before it
    is sent on.

Both are exactly the shapes Konrad agreed with in this thread, and both
are tested on the affected machine (Lenovo IdeaPad Duet 3,
sc7180-trogdor-wormdingler).

#regzbot monitor: https://lore.kernel.org/r/20260821081325.89088-1-phollinsky@xxxxxxxxxxxxxx
#regzbot monitor: https://lore.kernel.org/r/20260821081540.89137-1-phollinsky@xxxxxxxxxxxxxx

All the best,
Paul Hollinsky