Skip to content

hwpmc: Avoid using AMD PMCs if in use by firmware #2277

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mashtizadeh wants to merge 1 commit into
base: main
Choose a base branch
from
+30 −0
Open

hwpmc: Avoid using AMD PMCs if in use by firmware #2277

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 30 additions & 0 deletions sys/dev/hwpmc/hwpmc_amd.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -907,6 +907,36 @@
return (NULL);
}

/*
* Some PC vendors enable the core counters in firmware to track
* performance. The best guess is that this is being used to control
* power management from within the SMM mode. We shouldn't just take
* over the PMCs in this case. The user should try disabling any
* performance monitoring or power management functions in the BIOS to
* safely make use of the counters.
*/
for (i = 0; i < AMD_PMC_CORE_DEFAULT; i++) {

@mhorne mhorne Jun 11, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should these tests come after we have queried and set amd_core_npmcs? Or is it only important to check the first 6 counters?

@mashtizadeh mashtizadeh Jun 12, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's fine were it is, we should only see this on older machines. HWP and other mechanisms are providing alternative ways for them to continue to implement similar functionality without using the counters anymore. I can't repro it on any recent Intel or AMD machines I bought from those brands that I know have done this in the past.

Don't forget it needs to come above the VM check as that will touch the first PMC.

@mashtizadeh mashtizadeh Jun 12, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

As for the helper function, I'll make that change and update this diff tomorrow or Saturday.

@mashtizadeh mashtizadeh Jun 12, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I also owe you a patch to clean up the SYSTEM/USER flag it's going to work as you suggested, but I want to implement a best effort filtering. IBS isn't like other counters we get accurate samples but still suffer from skid during delivery.

if ((amd_feature2 & AMDID2_PCXC) != 0) {
error = rdmsr_safe(AMD_PMC_CORE_BASE + 2 * i, &reg);
} else {
if (i >= AMD_NPMCS_K8)
break;

error = rdmsr_safe(AMD_PMC_EVSEL_0 + i, &reg);
}

if (error != 0) {
printf("hwpmc: AMD evsel %d rdmsr failed!\n", i);
return (NULL);
}

if ((reg & AMD_PMC_ENABLE) != 0) {
printf("hwpmc: PMCs maybe in use by firmware!\n");
printf("hwpmc: Disable the PMC use in the BIOS before loading\n");

Check warning on line 935 in sys/dev/hwpmc/hwpmc_amd.c

View workflow job for this annotation

GitHub Actions / Style Checker 

line over 80 characters
return (NULL);
}
}

/*
* Unforunately, there is no way to communicate that the original four
* core counters are disabled through CPUIDs alone. We attempt to
Expand Down
Loading