[PATCH 09/16] drm/amdgpu: create kernel doorbell page

Shashank Sharma shashank.sharma at amd.com
Thu Mar 30 14:48:42 UTC 2023


On 30/03/2023 16:42, Christian König wrote:
> Am 30.03.23 um 16:39 schrieb Alex Deucher:
>> On Thu, Mar 30, 2023 at 7:49 AM Shashank Sharma 
>> <shashank.sharma at amd.com> wrote:
>>>
>>> On 30/03/2023 13:30, Christian König wrote:
>>>>
>>>> Am 29.03.23 um 17:47 schrieb Shashank Sharma:
>>>>> From: Shashank Sharma <contactshashanksharma at gmail.com>
>>>>>
>>>>> This patch:
>>>>> - creates a doorbell page for graphics driver usages.
>>>>> - removes the adev->doorbell.ptr variable, replaces it with
>>>>>     kernel-doorbell-bo's cpu address.
>>>>>
>>>>> Cc: Alex Deucher <alexander.deucher at amd.com>
>>>>> Cc: Christian Koenig <christian.koenig at amd.com>
>>>>> Signed-off-by: Shashank Sharma <shashank.sharma at amd.com>
>>>>> ---
>>>>>    drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h  | 16 ++++++-
>>>>>    .../gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c  | 44 
>>>>> +++++++++++++++----
>>>>>    drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c       |  7 +++
>>>>>    3 files changed, 57 insertions(+), 10 deletions(-)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
>>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
>>>>> index 6581b78fe438..10a9bb10e974 100644
>>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
>>>>> @@ -49,10 +49,13 @@ struct amdgpu_doorbell {
>>>>>        /* doorbell mmio */
>>>>>        resource_size_t        base;
>>>>>        resource_size_t        size;
>>>>> -    u32 __iomem        *ptr;
>>>>> +    u32    __iomem        *ptr;
>>>> This one can probably go away if we use the pointer from
>>>> amdgpu_bo_create_kernel().
>>> We started like that, but later realized that the cpu_addr from
>>> create_kernel() will just limit us
>>>
>>> to that object only, whereas we are keeping this ptr to ioremap the
>>> whole doorbell space in one shot.
>> Why do we need that?  For the kernel driver, we'd only need to mmap
>> the page used for kernel doorbells.  Then each user app would mmap its
>> doorbell page.
>
> Yes, that is exactly my concern as well.
>
> The kernel needs a fixed number of doorbells allocated for its 
> internal use. Everything else should probably use the normal BO API.
>
> For KFD we can use the BO API internal in the kernel, but that is 
> certainly completely different to the kernel allocations.
>
There are 3 different kernel doorbell clients here:

- graphics driver

- mes (for aggregated doorbells and kernel ring test)

- kfd (for kernel ring test and KIQ)


The fix num_doorbells are just for kernel graphics driver, but We are 
allocating doorbell pages for each of those, and they all need to be 
mapped.

Please see patch 12-16 for these details.

- Shashank

> Christian.
>
>>
>> Alex
>>
>>> - Shashank
>>>
>>>>>          /* Number of doorbells reserved for amdgpu kernel driver */
>>>>>        u32 num_kernel_doorbells;
>>>>> +
>>>>> +    /* For kernel doorbell pages */
>>>>> +    struct amdgpu_doorbell_obj kernel_doorbells;
>>>>>    };
>>>>>      /* Reserved doorbells for amdgpu (including multimedia).
>>>>> @@ -369,6 +372,17 @@ void amdgpu_doorbell_free_page(struct
>>>>> amdgpu_device *adev,
>>>>>    int amdgpu_doorbell_alloc_page(struct amdgpu_device *adev,
>>>>>                    struct amdgpu_doorbell_obj *db_obj);
>>>>>    +/**
>>>>> + * amdgpu_doorbell_create_kernel_doorbells - Create kernel doorbells
>>>>> for graphics
>>>>> + *
>>>>> + * @adev: amdgpu_device pointer
>>>>> + *
>>>>> + * Creates doorbells for graphics driver
>>>>> + *
>>>>> + * returns 0 on success, error otherwise.
>>>>> + */
>>>>> +int amdgpu_doorbell_create_kernel_doorbells(struct amdgpu_device
>>>>> *adev);
>>>>> +
>>>>>    #define RDOORBELL32(index) amdgpu_mm_rdoorbell(adev, (index))
>>>>>    #define WDOORBELL32(index, v) amdgpu_mm_wdoorbell(adev, 
>>>>> (index), (v))
>>>>>    #define RDOORBELL64(index) amdgpu_mm_rdoorbell64(adev, (index))
>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
>>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
>>>>> index 8be15b82b545..b46fe8b1378d 100644
>>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
>>>>> @@ -160,6 +160,38 @@ int amdgpu_doorbell_alloc_page(struct
>>>>> amdgpu_device *adev,
>>>>>        return 0;
>>>>>    }
>>>>>    +/**
>>>>> + * amdgpu_doorbell_create_kernel_doorbells - Create kernel doorbells
>>>>> for graphics
>>>>> + *
>>>>> + * @adev: amdgpu_device pointer
>>>>> + *
>>>>> + * Creates doorbells for graphics driver
>>>>> + *
>>>>> + * returns 0 on success, error otherwise.
>>>>> + */
>>>>> +int amdgpu_doorbell_create_kernel_doorbells(struct amdgpu_device 
>>>>> *adev)
>>>>> +{
>>>>> +    int r;
>>>>> +    struct amdgpu_doorbell_obj *kernel_doorbells =
>>>>> &adev->doorbell.kernel_doorbells;
>>>>> +
>>>>> +    kernel_doorbells->doorbell_bitmap =
>>>>> bitmap_zalloc(adev->doorbell.num_kernel_doorbells,
>>>>> +                              GFP_KERNEL);
>>>>> +    if (!kernel_doorbells->doorbell_bitmap) {
>>>>> +        DRM_ERROR("Failed to create kernel doorbell bitmap\n");
>>>>> +        return -ENOMEM;
>>>>> +    }
>>>>> +
>>>>> +    kernel_doorbells->size = adev->doorbell.num_kernel_doorbells *
>>>>> sizeof(u32);
>>>>> +    r = amdgpu_doorbell_alloc_page(adev, kernel_doorbells);
>>>>> +    if (r) {
>>>>> + bitmap_free(kernel_doorbells->doorbell_bitmap);
>>>>> +        DRM_ERROR("Failed to allocate kernel doorbells, 
>>>>> err=%d\n", r);
>>>>> +        return r;
>>>>> +    }
>>>>> +
>>>>> +    return 0;
>>>>> +}
>>>>> +
>>>>>    /*
>>>>>     * GPU doorbell aperture helpers function.
>>>>>     */
>>>>> @@ -179,7 +211,6 @@ int amdgpu_device_doorbell_init(struct
>>>>> amdgpu_device *adev)
>>>>>            adev->doorbell.base = 0;
>>>>>            adev->doorbell.size = 0;
>>>>>            adev->doorbell.num_kernel_doorbells = 0;
>>>>> -        adev->doorbell.ptr = NULL;
>>>>>            return 0;
>>>>>        }
>>>>>    @@ -208,12 +239,7 @@ int amdgpu_device_doorbell_init(struct
>>>>> amdgpu_device *adev)
>>>>>        if (adev->asic_type >= CHIP_VEGA10)
>>>>>            adev->doorbell.num_kernel_doorbells += 0x400;
>>>>>    -    adev->doorbell.ptr = ioremap(adev->doorbell.base,
>>>>> - adev->doorbell.num_kernel_doorbells *
>>>>> -                     sizeof(u32));
>>>>> -    if (adev->doorbell.ptr == NULL)
>>>>> -        return -ENOMEM;
>>>>> -
>>>>> +    adev->doorbell.ptr = ioremap(adev->doorbell.base,
>>>>> adev->doorbell.size);
>>>>>        return 0;
>>>>>    }
>>>>>    @@ -226,6 +252,6 @@ int amdgpu_device_doorbell_init(struct
>>>>> amdgpu_device *adev)
>>>>>     */
>>>>>    void amdgpu_device_doorbell_fini(struct amdgpu_device *adev)
>>>>>    {
>>>>> -    iounmap(adev->doorbell.ptr);
>>>>> -    adev->doorbell.ptr = NULL;
>>>>> + bitmap_free(adev->doorbell.kernel_doorbells.doorbell_bitmap);
>>>>> +    amdgpu_doorbell_free_page(adev, 
>>>>> &adev->doorbell.kernel_doorbells);
>>>>>    }
>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>>>> index 203d77a20507..75c6852845c4 100644
>>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>>>> @@ -1866,6 +1866,13 @@ int amdgpu_ttm_init(struct amdgpu_device 
>>>>> *adev)
>>>>>            return r;
>>>>>        }
>>>>>    +    /* Create a boorbell page for kernel usages */
>>>>> +    r = amdgpu_doorbell_create_kernel_doorbells(adev);
>>>>> +    if (r) {
>>>>> +        DRM_ERROR("Failed to initialize kernel doorbells. \n");
>>>>> +        return r;
>>>>> +    }
>>>>> +
>>>>>        /* Initialize preemptible memory pool */
>>>>>        r = amdgpu_preempt_mgr_init(adev);
>>>>>        if (r) {
>


More information about the amd-gfx mailing list