[PATCH 3/4] drm/ttm: add input parameter force_alloc for ttm_bo_evict_mm
He, Roger
Hongbo.He at amd.com
Fri Feb 9 04:34:48 UTC 2018
I can't think of an use case when we don't want this to succeed.
That is true. seems I can simplify more here.
Thanks
Roger(Hongbo.He)
-----Original Message-----
From: Koenig, Christian
Sent: Thursday, February 08, 2018 8:58 PM
To: He, Roger <Hongbo.He at amd.com>; amd-gfx at lists.freedesktop.org; dri-devel at lists.freedesktop.org
Subject: Re: [PATCH 3/4] drm/ttm: add input parameter force_alloc for ttm_bo_evict_mm
Am 08.02.2018 um 10:06 schrieb Roger He:
> if true, allocate TTM pages regardless of zone global memory account
> limit. For suspend, We should avoid TTM memory allocate failure then
> result in suspend failure.
Why the extra parameter for amdgpu_bo_evict_vram ?
I can't think of an use case when we don't want this to succeed.
Christian.
>
> Signed-off-by: Roger He <Hongbo.He at amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 2 +-
> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++--
> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 4 ++--
> drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 2 +-
> drivers/gpu/drm/nouveau/nouveau_drm.c | 2 +-
> drivers/gpu/drm/qxl/qxl_object.c | 4 ++--
> drivers/gpu/drm/radeon/radeon_device.c | 6 +++---
> drivers/gpu/drm/radeon/radeon_object.c | 5 +++--
> drivers/gpu/drm/radeon/radeon_object.h | 3 ++-
> drivers/gpu/drm/ttm/ttm_bo.c | 16 ++++++++++------
> drivers/gpu/drm/vmwgfx/vmwgfx_drv.c | 6 +++---
> include/drm/ttm/ttm_bo_api.h | 5 ++++-
> 12 files changed, 34 insertions(+), 25 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
> index ee76b46..59ee12c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
> @@ -763,7 +763,7 @@ static int amdgpu_debugfs_evict_vram(struct seq_file *m, void *data)
> struct drm_device *dev = node->minor->dev;
> struct amdgpu_device *adev = dev->dev_private;
>
> - seq_printf(m, "(%d)\n", amdgpu_bo_evict_vram(adev));
> + seq_printf(m, "(%d)\n", amdgpu_bo_evict_vram(adev, true));
> return 0;
> }
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index e3fa3d7..3c5f9ca 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -2168,7 +2168,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon)
> }
> }
> /* evict vram memory */
> - amdgpu_bo_evict_vram(adev);
> + amdgpu_bo_evict_vram(adev, true);
>
> amdgpu_fence_driver_suspend(adev);
>
> @@ -2178,7 +2178,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon)
> * This second call to evict vram is to evict the gart page table
> * using the CPU.
> */
> - amdgpu_bo_evict_vram(adev);
> + amdgpu_bo_evict_vram(adev, true);
>
> pci_save_state(dev->pdev);
> if (suspend) {
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> index 0338ef6..db813f9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> @@ -803,14 +803,14 @@ int amdgpu_bo_unpin(struct amdgpu_bo *bo)
> return r;
> }
>
> -int amdgpu_bo_evict_vram(struct amdgpu_device *adev)
> +int amdgpu_bo_evict_vram(struct amdgpu_device *adev, bool
> +force_alloc)
> {
> /* late 2.6.33 fix IGP hibernate - we need pm ops to do this correct */
> if (0 && (adev->flags & AMD_IS_APU)) {
> /* Useless to evict on IGP chips */
> return 0;
> }
> - return ttm_bo_evict_mm(&adev->mman.bdev, TTM_PL_VRAM);
> + return ttm_bo_evict_mm(&adev->mman.bdev, TTM_PL_VRAM, force_alloc);
> }
>
> static const char *amdgpu_vram_names[] = { diff --git
> a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> index c2b02f5..6724cdc 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> @@ -227,7 +227,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
> u64 min_offset, u64 max_offset,
> u64 *gpu_addr);
> int amdgpu_bo_unpin(struct amdgpu_bo *bo); -int
> amdgpu_bo_evict_vram(struct amdgpu_device *adev);
> +int amdgpu_bo_evict_vram(struct amdgpu_device *adev, bool
> +force_alloc);
> int amdgpu_bo_init(struct amdgpu_device *adev);
> void amdgpu_bo_fini(struct amdgpu_device *adev);
> int amdgpu_bo_fbdev_mmap(struct amdgpu_bo *bo, diff --git
> a/drivers/gpu/drm/nouveau/nouveau_drm.c
> b/drivers/gpu/drm/nouveau/nouveau_drm.c
> index 8d4a5be..c9627ef 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_drm.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_drm.c
> @@ -702,7 +702,7 @@ nouveau_do_suspend(struct drm_device *dev, bool runtime)
> }
>
> NV_DEBUG(drm, "evicting buffers...\n");
> - ttm_bo_evict_mm(&drm->ttm.bdev, TTM_PL_VRAM);
> + ttm_bo_evict_mm(&drm->ttm.bdev, TTM_PL_VRAM, true);
>
> NV_DEBUG(drm, "waiting for kernel channels to go idle...\n");
> if (drm->cechan) {
> diff --git a/drivers/gpu/drm/qxl/qxl_object.c
> b/drivers/gpu/drm/qxl/qxl_object.c
> index f6b80fe..d8d26c8 100644
> --- a/drivers/gpu/drm/qxl/qxl_object.c
> +++ b/drivers/gpu/drm/qxl/qxl_object.c
> @@ -350,10 +350,10 @@ int qxl_bo_check_id(struct qxl_device *qdev,
> struct qxl_bo *bo)
>
> int qxl_surf_evict(struct qxl_device *qdev)
> {
> - return ttm_bo_evict_mm(&qdev->mman.bdev, TTM_PL_PRIV);
> + return ttm_bo_evict_mm(&qdev->mman.bdev, TTM_PL_PRIV, true);
> }
>
> int qxl_vram_evict(struct qxl_device *qdev)
> {
> - return ttm_bo_evict_mm(&qdev->mman.bdev, TTM_PL_VRAM);
> + return ttm_bo_evict_mm(&qdev->mman.bdev, TTM_PL_VRAM, true);
> }
> diff --git a/drivers/gpu/drm/radeon/radeon_device.c
> b/drivers/gpu/drm/radeon/radeon_device.c
> index 8d3e3d2..c11ee06 100644
> --- a/drivers/gpu/drm/radeon/radeon_device.c
> +++ b/drivers/gpu/drm/radeon/radeon_device.c
> @@ -1522,7 +1522,7 @@ void radeon_device_fini(struct radeon_device *rdev)
> DRM_INFO("radeon: finishing device.\n");
> rdev->shutdown = true;
> /* evict vram memory */
> - radeon_bo_evict_vram(rdev);
> + radeon_bo_evict_vram(rdev, true);
> radeon_fini(rdev);
> if (!pci_is_thunderbolt_attached(rdev->pdev))
> vga_switcheroo_unregister_client(rdev->pdev);
> @@ -1607,7 +1607,7 @@ int radeon_suspend_kms(struct drm_device *dev, bool suspend,
> }
> }
> /* evict vram memory */
> - radeon_bo_evict_vram(rdev);
> + radeon_bo_evict_vram(rdev, true);
>
> /* wait for gpu to finish processing current batch */
> for (i = 0; i < RADEON_NUM_RINGS; i++) { @@ -1626,7 +1626,7 @@ int
> radeon_suspend_kms(struct drm_device *dev, bool suspend,
> * This second call to evict vram is to evict the gart page table
> * using the CPU.
> */
> - radeon_bo_evict_vram(rdev);
> + radeon_bo_evict_vram(rdev, true);
>
> radeon_agp_suspend(rdev);
>
> diff --git a/drivers/gpu/drm/radeon/radeon_object.c
> b/drivers/gpu/drm/radeon/radeon_object.c
> index 15404af..99a9a45 100644
> --- a/drivers/gpu/drm/radeon/radeon_object.c
> +++ b/drivers/gpu/drm/radeon/radeon_object.c
> @@ -420,7 +420,8 @@ int radeon_bo_unpin(struct radeon_bo *bo)
> return r;
> }
>
> -int radeon_bo_evict_vram(struct radeon_device *rdev)
> +int
> +radeon_bo_evict_vram(struct radeon_device *rdev, bool force_alloc)
> {
> /* late 2.6.33 fix IGP hibernate - we need pm ops to do this correct */
> if (0 && (rdev->flags & RADEON_IS_IGP)) { @@ -428,7 +429,7 @@ int
> radeon_bo_evict_vram(struct radeon_device *rdev)
> /* Useless to evict on IGP chips */
> return 0;
> }
> - return ttm_bo_evict_mm(&rdev->mman.bdev, TTM_PL_VRAM);
> + return ttm_bo_evict_mm(&rdev->mman.bdev, TTM_PL_VRAM, force_alloc);
> }
>
> void radeon_bo_force_delete(struct radeon_device *rdev) diff --git
> a/drivers/gpu/drm/radeon/radeon_object.h
> b/drivers/gpu/drm/radeon/radeon_object.h
> index 9ffd821..757ba88 100644
> --- a/drivers/gpu/drm/radeon/radeon_object.h
> +++ b/drivers/gpu/drm/radeon/radeon_object.h
> @@ -136,7 +136,8 @@ extern int radeon_bo_pin(struct radeon_bo *bo, u32 domain, u64 *gpu_addr);
> extern int radeon_bo_pin_restricted(struct radeon_bo *bo, u32 domain,
> u64 max_offset, u64 *gpu_addr);
> extern int radeon_bo_unpin(struct radeon_bo *bo); -extern int
> radeon_bo_evict_vram(struct radeon_device *rdev);
> +extern int radeon_bo_evict_vram(struct radeon_device *rdev,
> + bool force_alloc);
> extern void radeon_bo_force_delete(struct radeon_device *rdev);
> extern int radeon_bo_init(struct radeon_device *rdev);
> extern void radeon_bo_fini(struct radeon_device *rdev); diff --git
> a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index
> a907311..31d10f1 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
> @@ -1342,15 +1342,17 @@ int ttm_bo_create(struct ttm_bo_device *bdev,
> EXPORT_SYMBOL(ttm_bo_create);
>
> static int ttm_bo_force_list_clean(struct ttm_bo_device *bdev,
> - unsigned mem_type)
> + unsigned mem_type, bool force_alloc)
> {
> - struct ttm_operation_ctx ctx = { false, false };
> + struct ttm_operation_ctx ttm_opt_ctx = { false, false };
> struct ttm_mem_type_manager *man = &bdev->man[mem_type];
> struct ttm_bo_global *glob = bdev->glob;
> struct dma_fence *fence;
> int ret;
> unsigned i;
>
> + if (force_alloc)
> + ttm_opt_ctx.flags = TTM_OPT_FLAG_FORCE_ALLOC;
> /*
> * Can't use standard list traversal since we're unlocking.
> */
> @@ -1359,7 +1361,8 @@ static int ttm_bo_force_list_clean(struct ttm_bo_device *bdev,
> for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
> while (!list_empty(&man->lru[i])) {
> spin_unlock(&glob->lru_lock);
> - ret = ttm_mem_evict_first(bdev, mem_type, NULL, &ctx);
> + ret = ttm_mem_evict_first(bdev, mem_type, NULL,
> + &ttm_opt_ctx);
> if (ret)
> return ret;
> spin_lock(&glob->lru_lock);
> @@ -1403,7 +1406,7 @@ int ttm_bo_clean_mm(struct ttm_bo_device *bdev,
> unsigned mem_type)
>
> ret = 0;
> if (mem_type > 0) {
> - ret = ttm_bo_force_list_clean(bdev, mem_type);
> + ret = ttm_bo_force_list_clean(bdev, mem_type, true);
> if (ret) {
> pr_err("Cleanup eviction failed\n");
> return ret;
> @@ -1419,7 +1422,8 @@ int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type)
> }
> EXPORT_SYMBOL(ttm_bo_clean_mm);
>
> -int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type)
> +int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type,
> + bool force_alloc)
> {
> struct ttm_mem_type_manager *man = &bdev->man[mem_type];
>
> @@ -1433,7 +1437,7 @@ int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type)
> return 0;
> }
>
> - return ttm_bo_force_list_clean(bdev, mem_type);
> + return ttm_bo_force_list_clean(bdev, mem_type, force_alloc);
> }
> EXPORT_SYMBOL(ttm_bo_evict_mm);
>
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
> b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
> index 184340d..28f8e4f 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
> @@ -430,7 +430,7 @@ static int vmw_request_device(struct vmw_private *dev_priv)
> if (dev_priv->cman)
> vmw_cmdbuf_remove_pool(dev_priv->cman);
> if (dev_priv->has_mob) {
> - (void) ttm_bo_evict_mm(&dev_priv->bdev, VMW_PL_MOB);
> + (void) ttm_bo_evict_mm(&dev_priv->bdev, VMW_PL_MOB, true);
> vmw_otables_takedown(dev_priv);
> }
> if (dev_priv->cman)
> @@ -463,7 +463,7 @@ static void vmw_release_device_early(struct vmw_private *dev_priv)
> vmw_cmdbuf_remove_pool(dev_priv->cman);
>
> if (dev_priv->has_mob) {
> - ttm_bo_evict_mm(&dev_priv->bdev, VMW_PL_MOB);
> + ttm_bo_evict_mm(&dev_priv->bdev, VMW_PL_MOB, true);
> vmw_otables_takedown(dev_priv);
> }
> }
> @@ -1342,7 +1342,7 @@ void vmw_svga_disable(struct vmw_private *dev_priv)
> if (dev_priv->bdev.man[TTM_PL_VRAM].use_type) {
> dev_priv->bdev.man[TTM_PL_VRAM].use_type = false;
> spin_unlock(&dev_priv->svga_lock);
> - if (ttm_bo_evict_mm(&dev_priv->bdev, TTM_PL_VRAM))
> + if (ttm_bo_evict_mm(&dev_priv->bdev, TTM_PL_VRAM, true))
> DRM_ERROR("Failed evicting VRAM buffers.\n");
> vmw_write(dev_priv, SVGA_REG_ENABLE,
> SVGA_REG_ENABLE_HIDE |
> diff --git a/include/drm/ttm/ttm_bo_api.h
> b/include/drm/ttm/ttm_bo_api.h index 2142639..6b5db9c 100644
> --- a/include/drm/ttm/ttm_bo_api.h
> +++ b/include/drm/ttm/ttm_bo_api.h
> @@ -636,6 +636,8 @@ int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type);
> *
> * @bdev: Pointer to a ttm_bo_device struct.
> * @mem_type: The memory type.
> + * @force_alloc: if true allow ttm pages allocation always
> + * regardless of zone memory account limit
> *
> * Evicts all buffers on the lru list of the memory type.
> * This is normally part of a VT switch or an @@ -649,7 +651,8 @@
> int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type);
> * -ERESTARTSYS: The call was interrupted by a signal while waiting to
> * evict a buffer.
> */
> -int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type);
> +int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type,
> + bool force_alloc);
>
> /**
> * ttm_kmap_obj_virtual
More information about the amd-gfx
mailing list