Re: [Nouveau] [PATCH 3/8] drm/amdgpu: Implement mmap as GEM object function

["Christian König" <christian.koenig@amd.com>]

Hi Thomas,

Am 06.04.21 um 12:38 schrieb Thomas Zimmermann:
> Hi
>
> Am 06.04.21 um 11:35 schrieb Christian König:
>> Am 06.04.21 um 11:08 schrieb Thomas Zimmermann:
>>> Moving the driver-specific mmap code into a GEM object function allows
>>> for using DRM helpers for various mmap callbacks.
>>>
>>> This change resolves several inconsistencies between regular mmap and
>>> prime-based mmap. The vm_ops field in vma is now set for all mmap'ed
>>> areas. Previously it way only set for regular mmap calls, prime-based
>>> mmap used TTM's default vm_ops. The check for kfd_bo has been taken
>>> from amdgpu_verify_access(), which is not called any longer and has
>>> been removed.
>>>
>>> As a side effect, amdgpu_ttm_vm_ops and amdgpu_ttm_fault() are now
>>> implemented in amdgpu's GEM code.
>>>
>>> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
>>> ---
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 46 -------------
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.h |  2 -
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c     |  4 +-
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c     | 64 +++++++++++++++++++
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c     | 71 
>>> ---------------------
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h     |  1 -
>>>   6 files changed, 66 insertions(+), 122 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
>>> index e0c4f7c7f1b9..19c5ab08d9ec 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
>>> @@ -42,52 +42,6 @@
>>>   #include <linux/pci-p2pdma.h>
>>>   #include <linux/pm_runtime.h>
>>> -/**
>>> - * amdgpu_gem_prime_mmap - &drm_driver.gem_prime_mmap implementation
>>> - * @obj: GEM BO
>>> - * @vma: Virtual memory area
>>> - *
>>> - * Sets up a userspace mapping of the BO's memory in the given
>>> - * virtual memory area.
>>> - *
>>> - * Returns:
>>> - * 0 on success or a negative error code on failure.
>>> - */
>>> -int amdgpu_gem_prime_mmap(struct drm_gem_object *obj,
>>> -              struct vm_area_struct *vma)
>>> -{
>>> -    struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
>>> -    struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
>>> -    unsigned asize = amdgpu_bo_size(bo);
>>> -    int ret;
>>> -
>>> -    if (!vma->vm_file)
>>> -        return -ENODEV;
>>> -
>>> -    if (adev == NULL)
>>> -        return -ENODEV;
>>> -
>>> -    /* Check for valid size. */
>>> -    if (asize < vma->vm_end - vma->vm_start)
>>> -        return -EINVAL;
>>> -
>>> -    if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) ||
>>> -        (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) {
>>> -        return -EPERM;
>>> -    }
>>> -    vma->vm_pgoff += amdgpu_bo_mmap_offset(bo) >> PAGE_SHIFT;
>>> -
>>> -    /* prime mmap does not need to check access, so allow here */
>>> -    ret = drm_vma_node_allow(&obj->vma_node, 
>>> vma->vm_file->private_data);
>>> -    if (ret)
>>> -        return ret;
>>> -
>>> -    ret = ttm_bo_mmap(vma->vm_file, vma, &adev->mman.bdev);
>>> -    drm_vma_node_revoke(&obj->vma_node, vma->vm_file->private_data);
>>> -
>>> -    return ret;
>>> -}
>>> -
>>>   static int
>>>   __dma_resv_make_exclusive(struct dma_resv *obj)
>>>   {
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.h 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.h
>>> index 39b5b9616fd8..3e93b9b407a9 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.h
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.h
>>> @@ -31,8 +31,6 @@ struct drm_gem_object 
>>> *amdgpu_gem_prime_import(struct drm_device *dev,
>>>                           struct dma_buf *dma_buf);
>>>   bool amdgpu_dmabuf_is_xgmi_accessible(struct amdgpu_device *adev,
>>>                         struct amdgpu_bo *bo);
>>> -int amdgpu_gem_prime_mmap(struct drm_gem_object *obj,
>>> -              struct vm_area_struct *vma);
>>>   extern const struct dma_buf_ops amdgpu_dmabuf_ops;
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
>>> index 76f48f79c70b..e96d2758f4bb 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
>>> @@ -1656,7 +1656,7 @@ static const struct file_operations 
>>> amdgpu_driver_kms_fops = {
>>>       .flush = amdgpu_flush,
>>>       .release = drm_release,
>>>       .unlocked_ioctl = amdgpu_drm_ioctl,
>>> -    .mmap = amdgpu_mmap,
>>> +    .mmap = drm_gem_mmap,
>>>       .poll = drm_poll,
>>>       .read = drm_read,
>>>   #ifdef CONFIG_COMPAT
>>> @@ -1719,7 +1719,7 @@ static const struct drm_driver 
>>> amdgpu_kms_driver = {
>>>       .prime_handle_to_fd = drm_gem_prime_handle_to_fd,
>>>       .prime_fd_to_handle = drm_gem_prime_fd_to_handle,
>>>       .gem_prime_import = amdgpu_gem_prime_import,
>>> -    .gem_prime_mmap = amdgpu_gem_prime_mmap,
>>> +    .gem_prime_mmap = drm_gem_prime_mmap,
>>>       .name = DRIVER_NAME,
>>>       .desc = DRIVER_DESC,
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>>> index fb7171e5507c..fe93faad05f2 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>>> @@ -41,6 +41,36 @@
>>>   static const struct drm_gem_object_funcs amdgpu_gem_object_funcs;
>>> +static vm_fault_t amdgpu_ttm_fault(struct vm_fault *vmf)
>>
>> Please name that function amdgpu_gem_fault or amdgpu_gem_object_fault
>>
>>> +{
>>> +    struct ttm_buffer_object *bo = vmf->vma->vm_private_data;
>>> +    vm_fault_t ret;
>>> +
>>> +    ret = ttm_bo_vm_reserve(bo, vmf);
>>> +    if (ret)
>>> +        return ret;
>>> +
>>> +    ret = amdgpu_bo_fault_reserve_notify(bo);
>>> +    if (ret)
>>> +        goto unlock;
>>> +
>>> +    ret = ttm_bo_vm_fault_reserved(vmf, vmf->vma->vm_page_prot,
>>> +                       TTM_BO_VM_NUM_PREFAULT, 1);
>>> +    if (ret == VM_FAULT_RETRY && !(vmf->flags & 
>>> FAULT_FLAG_RETRY_NOWAIT))
>>> +        return ret;
>>> +
>>> +unlock:
>>> +    dma_resv_unlock(bo->base.resv);
>>> +    return ret;
>>> +}
>>> +
>>> +static const struct vm_operations_struct amdgpu_ttm_vm_ops = {
>>> +    .fault = amdgpu_ttm_fault,
>>> +    .open = ttm_bo_vm_open,
>>> +    .close = ttm_bo_vm_close,
>>> +    .access = ttm_bo_vm_access
>>> +};
>>> +
>>>   static void amdgpu_gem_object_free(struct drm_gem_object *gobj)
>>>   {
>>>       struct amdgpu_bo *robj = gem_to_amdgpu_bo(gobj);
>>> @@ -201,6 +231,38 @@ static void amdgpu_gem_object_close(struct 
>>> drm_gem_object *obj,
>>>       ttm_eu_backoff_reservation(&ticket, &list);
>>>   }
>>> +static int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct 
>>> vm_area_struct *vma)
>>> +{
>>> +    struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
>>> +    struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
>>> +    unsigned long asize = amdgpu_bo_size(bo);
>>> +
>>> +    if (!vma->vm_file)
>>> +        return -ENODEV;
>>> +
>>> +    if (!adev)
>>> +        return -ENODEV;
>>> +
>>> +    /* Check for valid size. */
>>> +    if (asize < vma->vm_end - vma->vm_start)
>>> +        return -EINVAL;
>>
>>> +
>>> +    /*
>>> +     * Don't verify access for KFD BOs. They don't have a GEM
>>> +     * object associated with them.
>>> +     */
>>> +    if (bo->kfd_bo)
>>> +        goto out;
>>
>> Who does the access verification now?
>
> This is somewhat confusing.
>
> I took this check as-is, including the comment, from amdgpu's 
> verify_access function. The verify_access function was called by 
> ttm_bo_mmap. It returned 0 and ttm_bo_mmap did the mapping.
>
>>
>> Christian.
>>
>>> +
>>> +    if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) ||
>>> +        (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) {
>>> +        return -EPERM;
>>> +    }
>>> +
>>> +out:
>>> +    return drm_gem_ttm_mmap(obj, vma);
>>> +}
>>> +
>>>   static const struct drm_gem_object_funcs amdgpu_gem_object_funcs = {
>>>       .free = amdgpu_gem_object_free,
>>>       .open = amdgpu_gem_object_open,
>>> @@ -208,6 +270,8 @@ static const struct drm_gem_object_funcs 
>>> amdgpu_gem_object_funcs = {
>>>       .export = amdgpu_gem_prime_export,
>>>       .vmap = drm_gem_ttm_vmap,
>>>       .vunmap = drm_gem_ttm_vunmap,
>>> +    .mmap = amdgpu_gem_prime_mmap,
>>> +    .vm_ops = &amdgpu_ttm_vm_ops,
>>>   };
>>>   /*
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> index 1c6131489a85..d9de91a517c6 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> @@ -152,32 +152,6 @@ static void amdgpu_evict_flags(struct 
>>> ttm_buffer_object *bo,
>>>       *placement = abo->placement;
>>>   }
>>> -/**
>>> - * amdgpu_verify_access - Verify access for a mmap call
>>> - *
>>> - * @bo:    The buffer object to map
>>> - * @filp: The file pointer from the process performing the mmap
>>> - *
>>> - * This is called by ttm_bo_mmap() to verify whether a process
>>> - * has the right to mmap a BO to their process space.
>>> - */
>>> -static int amdgpu_verify_access(struct ttm_buffer_object *bo, 
>>> struct file *filp)
>>> -{
>>> -    struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo);
>>> -
>>> -    /*
>>> -     * Don't verify access for KFD BOs. They don't have a GEM
>>> -     * object associated with them.
>>> -     */
>>> -    if (abo->kfd_bo)
>>> -        return 0;
>>> -
>>> -    if (amdgpu_ttm_tt_get_usermm(bo->ttm))
>>> -        return -EPERM;
>>> -    return drm_vma_node_verify_access(&abo->tbo.base.vma_node,
>>> -                      filp->private_data);
>>> -}
>
> Here's the orignal verification code. It gives a free pass to KFD.
>
>>> -
>>>   /**
>>>    * amdgpu_ttm_map_buffer - Map memory into the GART windows
>>>    * @bo: buffer object to map
>>> @@ -1531,7 +1505,6 @@ static struct ttm_device_funcs 
>>> amdgpu_bo_driver = {
>>>       .eviction_valuable = amdgpu_ttm_bo_eviction_valuable,
>>>       .evict_flags = &amdgpu_evict_flags,
>>>       .move = &amdgpu_bo_move,
>>> -    .verify_access = &amdgpu_verify_access,
>>>       .delete_mem_notify = &amdgpu_bo_delete_mem_notify,
>>>       .release_notify = &amdgpu_bo_release_notify,
>>>       .io_mem_reserve = &amdgpu_ttm_io_mem_reserve,
>>> @@ -1906,50 +1879,6 @@ void 
>>> amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool 
>>> enable)
>>>       adev->mman.buffer_funcs_enabled = enable;
>>>   }
>>> -static vm_fault_t amdgpu_ttm_fault(struct vm_fault *vmf)
>>> -{
>>> -    struct ttm_buffer_object *bo = vmf->vma->vm_private_data;
>>> -    vm_fault_t ret;
>>> -
>>> -    ret = ttm_bo_vm_reserve(bo, vmf);
>>> -    if (ret)
>>> -        return ret;
>>> -
>>> -    ret = amdgpu_bo_fault_reserve_notify(bo);
>>> -    if (ret)
>>> -        goto unlock;
>>> -
>>> -    ret = ttm_bo_vm_fault_reserved(vmf, vmf->vma->vm_page_prot,
>>> -                       TTM_BO_VM_NUM_PREFAULT, 1);
>>> -    if (ret == VM_FAULT_RETRY && !(vmf->flags & 
>>> FAULT_FLAG_RETRY_NOWAIT))
>>> -        return ret;
>>> -
>>> -unlock:
>>> -    dma_resv_unlock(bo->base.resv);
>>> -    return ret;
>>> -}
>>> -
>>> -static const struct vm_operations_struct amdgpu_ttm_vm_ops = {
>>> -    .fault = amdgpu_ttm_fault,
>>> -    .open = ttm_bo_vm_open,
>>> -    .close = ttm_bo_vm_close,
>>> -    .access = ttm_bo_vm_access
>>> -};
>>> -
>>> -int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma)
>>> -{
>>> -    struct drm_file *file_priv = filp->private_data;
>>> -    struct amdgpu_device *adev = drm_to_adev(file_priv->minor->dev);
>>> -    int r;
>>> -
>>> -    r = ttm_bo_mmap(filp, vma, &adev->mman.bdev);
>>> -    if (unlikely(r != 0))
>>> -        return r;
>>> -
>>> -    vma->vm_ops = &amdgpu_ttm_vm_ops;
>>> -    return 0;
>>> -}
>
> And this was the mmap callback in struct file_operations. It calls 
> ttm_bo_mmap(), which skips verification for KFD BOs. To the best of my 
> knowledge, there was no additional verification for these KFD BOs.
>
> The original code in amdgpu_gem_prime_mmap() did seom verification, 
> but didn't handle KFD specially. I guess, PRIME needs GEM and KFD BOs 
> wouldn't quailify.
>
> In the end I went with the semantics I found in amdgpu_mmap() and 
> handled KFD specially. Let me know if this requires to be changed.

Well the question is where is the call to drm_vma_node_verify_access() 
now? Cause that needs to be skipped for KFD BOs.

Regards,
Christian.

>
> Best regards
> Thomas
>
>>> -
>>>   int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
>>>                  uint64_t dst_offset, uint32_t byte_count,
>>>                  struct dma_resv *resv,
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
>>> index dec0db8b0b13..6e51faad7371 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
>>> @@ -146,7 +146,6 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
>>>               struct dma_resv *resv,
>>>               struct dma_fence **fence);
>>> -int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma);
>>>   int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo);
>>>   int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo);
>>>   uint64_t amdgpu_ttm_domain_start(struct amdgpu_device *adev, 
>>> uint32_t type);
>>
>> _______________________________________________
>> dri-devel mailing list
>> dri-devel@lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/dri-devel
>

_______________________________________________
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau