[PATCH 2/2] drm/amdgpu: use real_vram_size in ttm_vram_fops
Christian König
ckoenig.leichtzumerken at gmail.com
Wed Jun 22 12:38:25 UTC 2022
Am 22.06.22 um 12:07 schrieb Pierre-Eric Pelloux-Prayer:
> If amdgpu.vramlimit=XXXX is used, amdgpu_gmc_vram_location will update
> real_vram_size based on this value.
> mc_vram_size is the real amount of VRAM, initialized in gmc_..._mc_init.
Thinking more about it I came to the conclusion that this patch here is
not correct.
Even when we restricted the driver to use only a fraction of the VRAM we
can still read/write all of it through the debugging interfaces.
That could be useful if we try to track down things like random memory
corruption etc..
Regards,
Christian.
>
> Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer at amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 952e99e6d07e..8f245e9f8f7c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -2252,10 +2252,10 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf,
> if (size & 0x3 || *pos & 0x3)
> return -EINVAL;
>
> - if (*pos >= adev->gmc.mc_vram_size)
> + if (*pos >= adev->gmc.real_vram_size)
> return -ENXIO;
>
> - size = min(size, (size_t)(adev->gmc.mc_vram_size - *pos));
> + size = min(size, (size_t)(adev->gmc.real_vram_size - *pos));
> while (size) {
> size_t bytes = min(size, AMDGPU_TTM_VRAM_MAX_DW_READ * 4);
> uint32_t value[AMDGPU_TTM_VRAM_MAX_DW_READ];
> @@ -2288,13 +2288,13 @@ static ssize_t amdgpu_ttm_vram_write(struct file *f, const char __user *buf,
> if (size & 0x3 || *pos & 0x3)
> return -EINVAL;
>
> - if (*pos >= adev->gmc.mc_vram_size)
> + if (*pos >= adev->gmc.real_vram_size)
> return -ENXIO;
>
> while (size) {
> uint32_t value;
>
> - if (*pos >= adev->gmc.mc_vram_size)
> + if (*pos >= adev->gmc.real_vram_size)
> return result;
>
> r = get_user(value, (uint32_t *)buf);
> @@ -2442,7 +2442,7 @@ void amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
> struct dentry *root = minor->debugfs_root;
>
> debugfs_create_file_size("amdgpu_vram", 0444, root, adev,
> - &amdgpu_ttm_vram_fops, adev->gmc.mc_vram_size);
> + &amdgpu_ttm_vram_fops, adev->gmc.real_vram_size);
> debugfs_create_file("amdgpu_iomem", 0444, root, adev,
> &amdgpu_ttm_iomem_fops);
> debugfs_create_file("amdgpu_vram_mm", 0444, root, adev,
More information about the amd-gfx
mailing list