Re: [PATCH 0/3] platform/x86/intel/pmt: Fix crashlog regressions and add completion uevent

[Ilpo Järvinen]

On Thu, 1 Oct 2026, David E. Box wrote:

> This series adds a uevent so userspace can find out when an Intel PMT
> crashlog is available without polling. It also fixes two regressions
> from the v7.1 vsec rework that prevent the crashlog path from being
> used at all.
> 
> Patches 1 and 2 are fixes, tagged for stable:
> 
>  - Patch 1 fixes a NULL pointer dereference on any read of a crashlog
>    sysfs data file. Commit 353042d54d82 ("platform/x86/intel/vsec:
>    Switch exported helpers from pci_dev to device") started using
>    entry->ep->dev, but only telemetry entries allocate an endpoint.
> 
>  - Patch 2 fixes an inverted return value test of
>    intel_vsec_walk_header() in intel_vsec_get_features(). On devices
>    without DVSEC/VSEC capabilities that describe their features through
>    device_data, probe fails with -ENODEV and the auxiliary devices it
>    created are torn down again.
> 
> Patch 3 is the new feature. It emits a KOBJ_CHANGE uevent on the
> crashlog<x> device when a completed crashlog is found at probe, or when
> a manually triggered capture completes. The event carries
> INTEL_PMT_CRASHLOG_EVENT=PRESENT and INTEL_PMT_CRASHLOG_COMPLETE=1, and
> the ABI documentation is updated to match.
> 
> The fixes are sent together with the feature because patch 3 cannot be
> tested without them. Crashlog data cannot be read without patch 1.
> Upcoming hardware with crashlog support has no DVSEC/VSEC capabilities
> and describes its features through device_data, so it hits the bug
> fixed by patch 2. Patches 1 and 2 do not depend on patch 3 and can be
> applied to the fixes branch on their own if preferred.
> 
> Testing: the series was tested in an emulated (QEMU) environment,
> covering both the probe-time uevent with a completed crashlog present
> and the uevent after a manual trigger completes. Hardware with this
> crashlog support is available, but testing on it is still pending.
> 
> David E. Box (3):
>   platform/x86/intel/pmt: Fix NULL dereference when reading crashlog
>     data
>   platform/x86/intel/vsec: Fix inverted walk_header() test in
>     get_features()
>   platform/x86/intel/pmt: Notify userspace when crashlogs complete

I took the 2nd patch but the others seem to need more discussion/work 
given the recent changes.

-- 
 i.