-
Notifications
You must be signed in to change notification settings - Fork 598
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.
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
interfaces: Add more kernel fusion driver files to opengl #15068
base: master
Are you sure you want to change the base?
interfaces: Add more kernel fusion driver files to opengl #15068
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #15068 +/- ##
=========================================
Coverage ? 78.06%
=========================================
Files ? 1180
Lines ? 157753
Branches ? 0
=========================================
Hits ? 123153
Misses ? 26949
Partials ? 7651
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
Tue Feb 18 18:34:48 UTC 2025 Failures:Preparing:
Executing:
Restoring:
|
interfaces/builtin/opengl.go
Outdated
@@ -198,6 +198,11 @@ unix (send, receive) type=dgram peer=(addr="@var/run/nvidia-xdriver-*"), | |||
# Kernel Fusion Driver for AMD GPUs | |||
/dev/kfd rw, | |||
/sys/module/amdgpu/initstate r, | |||
/sys/devices/virtual/kfd/kfd/dev r, | |||
/sys/devices/virtual/kfd/kfd/uevent rw, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is write access definitely needed here?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
IIRC write is useful if you want to synthesize events from userspace. With how crazy the drivers are I wouldn't be surprised it's actually used for something. In any case, getting a conformation for needing write access would be nice.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I added them as the rocminfo
utility seemed to access them. I added these all based on the app armor allows when using it in devmode
.
I am off today and Monday, but if I get a chance tonight, I can re run it and capture the logs if you'd like.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hm, actually I don't see it in the logs when I run rocm-validation-suite
or rocminfo
, perhaps this isn't needed (at least in the usecase of rvs
or rocminfo
). I can bump that to read-only for now.
I've attached logs from a full run of rocminfo
and a full run of rocm-validation-suite
(which runs various different test programs that the Certification team uses for GPGPU testing).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM! Thanks for confirming that write permission is not needed.
This is a follow-up of #14685.
I am trying to create a snap of
rocminfo
, and I noticed thatrocminfo
is accessing the Kernel Fusion Driver via the/sys/devices/virtual/kfd
path to retrieve the GPUs' properties. I've added the files thatrocminfo
was specifically accessing, but there are other subdirectories/files that are not included here.These denials seem to also be blocking
rocm-validation-suite
from querying these properties when running some of its tests.I also noticed that the Mir team is also interested in including
rocminfo
in their graphics-test-tools, so I'm sure they'd run into these issues eventually.Let me know if you need anything or if I missed anything.