You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Please confirm this issue does not happen with the proprietary driver (of the same version). This issue tracker is only for bugs specific to the open kernel driver.
I confirm that this does not happen with the proprietary driver package.
Operating System and Version
Arch Linux
Kernel Release
6.12.8
Please confirm you are running a stable release kernel (e.g. not a -rc). We do not accept bug reports for unreleased kernels.
In mutter we are noticing that screen changes, e.g. caused by moving the mouse cursor or even spinner animations running, is causing rather high CPU usage located in drmModeCloseFB. Using sysprof to analyze the root, it ends up being due to nv_drm_framebuffer_destroy and its callees:
From the profile it seems as if every close of the framebuffer a buffer is immediately being destroyed. I'm not familiar with the callees in the profile, but the destroy may or may not be a red herring as it looks as if this code ends up relying on some kind of lock across threads underneath that blocks for longer than expected, resulting in high CPU usage.
This happens on single-GPU NVIDIA systems as well as hybrid systems with a monitor attached to the NVIDIA dGPU. As can be seen in the discussions on the ticket we did some testing on various NVIDIA drivers and it seems less problematic with the GSP firmware disabled on the closed driver. It's also not happening on other graphics drivers and seems to be unique to this driver.
To Reproduce
Ensure you are on an empty desktop in GNOME (47.4).
Start a terminal, start top.
Observe baseline of gnome-shell CPU use. For me it is about 0% to 1%.
Keep moving the mouse over the background wallpaper whilst observing top.
Notice gnome-shell spiking. For me it goes to 15% or higher.
Stop moving the cursor, notice CPU use going back to what it was.
(Having a clean desktop/workspace ensures damage of applications that you are moving the mouse over is not influencing the result.)
Gert-dev
changed the title
Expensive drmModeCloseFB due to framebuffers being destroyed immediately
High CPU use in drmModeCloseFB on common actions such as mouse movement
Jan 16, 2025
NVIDIA Open GPU Kernel Modules Version
565.77
Please confirm this issue does not happen with the proprietary driver (of the same version). This issue tracker is only for bugs specific to the open kernel driver.
Operating System and Version
Arch Linux
Kernel Release
6.12.8
Please confirm you are running a stable release kernel (e.g. not a -rc). We do not accept bug reports for unreleased kernels.
Hardware: GPU
NVIDIA GeForce RTX 4090 Laptop GPU
Describe the bug
(This is an upstream report of mutter issue 3579.)
In mutter we are noticing that screen changes, e.g. caused by moving the mouse cursor or even spinner animations running, is causing rather high CPU usage located in
drmModeCloseFB
. Using sysprof to analyze the root, it ends up being due tonv_drm_framebuffer_destroy
and its callees:From the profile it seems as if every close of the framebuffer a buffer is immediately being destroyed. I'm not familiar with the callees in the profile, but the destroy may or may not be a red herring as it looks as if this code ends up relying on some kind of lock across threads underneath that blocks for longer than expected, resulting in high CPU usage.
This happens on single-GPU NVIDIA systems as well as hybrid systems with a monitor attached to the NVIDIA dGPU. As can be seen in the discussions on the ticket we did some testing on various NVIDIA drivers and it seems less problematic with the GSP firmware disabled on the closed driver. It's also not happening on other graphics drivers and seems to be unique to this driver.
To Reproduce
top
.gnome-shell
CPU use. For me it is about 0% to 1%.top
.(Having a clean desktop/workspace ensures damage of applications that you are moving the mouse over is not influencing the result.)
Bug Incidence
Always
nvidia-bug-report.log.gz
nvidia-bug-report.log.gz
More Info
No response
The text was updated successfully, but these errors were encountered: