[Public]
Hi Christian
I ping Leslie offline with no response.
How about this patch.
Looks like more reasonable to put doorbell_fini with mem_scratch_fini/
free_static_csa together.
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index d4b01ef00025..ff48096c1d80 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -3508,6 +3508,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device
*adev)
amdgpu_device_mem_scratch_fini(adev);
amdgpu_ib_pool_fini(adev);
amdgpu_seq64_fini(adev);
+ amdgpu_doorbell_fini(adev);
}
if (adev->ip_blocks[i].version->funcs->sw_fini) {
r =
adev->ip_blocks[i].version->funcs->sw_fini(&adev->ip_blocks[i]);
@@ -4851,7 +4852,6 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev)
iounmap(adev->rmmio);
adev->rmmio = NULL;
- amdgpu_doorbell_fini(adev);
drm_dev_exit(idx);
}
Best,
Zhenguo
Cloud-GPU Core team, SRDC
-----Original Message-----
From: Yin, ZhenGuo (Chris)
Sent: Tuesday, April 1, 2025 11:41 AM
To: Koenig, Christian <christian.koe...@amd.com>;
amd-gfx@lists.freedesktop.org; Shi, Leslie <yuliang....@amd.com>
Cc: Sharma, Shashank <shashank.sha...@amd.com>; Deucher, Alexander
<alexander.deuc...@amd.com>
Subject: RE: [PATCH] drm/amdgpu: fix warning of drm_mm_clean
Hi Leslie
Any comment on this?
Looks like you add this code 3 years ago on patch: drm/amdgpu: Unmap MMIO
mappings when device is not unplugged
Best,
Zhenguo
Cloud-GPU Core team, SRDC
-----Original Message-----
From: Koenig, Christian <christian.koe...@amd.com>
Sent: Monday, March 31, 2025 8:07 PM
To: Yin, ZhenGuo (Chris) <zhenguo....@amd.com>; amd-gfx@lists.freedesktop.org
Cc: Sharma, Shashank <shashank.sha...@amd.com>; Deucher, Alexander
<alexander.deuc...@amd.com>
Subject: Re: [PATCH] drm/amdgpu: fix warning of drm_mm_clean
Am 31.03.25 um 13:26 schrieb ZhenGuo Yin:
> Kernel doorbell BOs needs to be freed before ttm_fini.
Good catch, but does anybody remember why we have the
drm_dev_enter()/drm_dev_exit() here?
It's clearly for hotplug, but it doesn't make sense to have the
amdgpu_doorbell_fini() under it as far as I can see.
Regards,
Christian.
>
> Fixes: 54c30d2a8def ("drm/amdgpu: create kernel doorbell pages")
> Signed-off-by: ZhenGuo Yin <zhenguo....@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 15 +++++++--------
> 1 file changed, 7 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index d4b01ef00025..b49604ade6c2 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -4807,6 +4807,13 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev)
> int i, idx;
> bool px;
>
> + if (drm_dev_enter(adev_to_drm(adev), &idx)) {
> + iounmap(adev->rmmio);
> + adev->rmmio = NULL;
> + amdgpu_doorbell_fini(adev);
> + drm_dev_exit(idx);
> + }
> +
> amdgpu_device_ip_fini(adev);
> amdgpu_fence_driver_sw_fini(adev);
> amdgpu_ucode_release(&adev->firmware.gpu_info_fw);
> @@ -4847,14 +4854,6 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev)
> if ((adev->pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA)
> vga_client_unregister(adev->pdev);
>
> - if (drm_dev_enter(adev_to_drm(adev), &idx)) {
> -
> - iounmap(adev->rmmio);
> - adev->rmmio = NULL;
> - amdgpu_doorbell_fini(adev);
> - drm_dev_exit(idx);
> - }
> -
> if (IS_ENABLED(CONFIG_PERF_EVENTS))
> amdgpu_pmu_fini(adev);
> if (adev->mman.discovery_bin)
