As part of the S3 suspend sequence dGPUs will evict VRAM. If there is high memory pressure at this time, there is a chance this fails. systemd has a policy to try to "fall back" from S3 to s2idle and see if that works. When under high memory pressure this also fails, and harder. Really we don't want this flow to be possible. Fail the sequence if the dGPU won't be suspended using either method. Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3476 Link: https://github.com/systemd/systemd/issues/25151 Signed-off-by: Mario Limonciello <mario.limonciello@xxxxxxx> --- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 802debd8d9f0..6a5a3e132319 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -2562,7 +2562,7 @@ static int amdgpu_pmops_suspend(struct device *dev) else if (amdgpu_acpi_is_s3_active(adev)) adev->in_s3 = true; if (!adev->in_s0ix && !adev->in_s3) - return 0; + return -EINVAL; return amdgpu_device_suspend(drm_dev, true); } -- 2.45.2
