Re: [Linux-graphics-maintainer] [PATCH 2/3] drm/ttm: move memory accounting into vmwgfx v3

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Just had a quick peek. The issue is that you can’t attach to the drm device (card0) because it hasn’t been registered yet (drm device registration is last in the vmw_probe in vmwgfx_drv.c via the drm_dev_register). So dev->primary->kdev->kobj that you’re using as argument to kobject_init_and_add in ttm_mem_global_init hasn’t been initialized yet. So that particular sysfs code would likely have to be refactored out of ttm_mem_global_init to another function that could be called after drm registraction. I could take this on but not until Friday or so.

z


> On Feb 2, 2021, at 12:42, Zack Rusin <zackr@xxxxxxxxxx> wrote:
> 
> Ah, yes, sorry, I missed that. I just double checked and it fails with:
> 
> kobject_add_internal failed for memory_accounting (error: -2 parent: card0)
> 
> which breaks the probe and the driver won’t load. I won’t have time to look into it until tomorrow though.
> 
> z
> 
>> On Feb 2, 2021, at 10:16, Christian König <christian.koenig@xxxxxxx> wrote:
>> 
>> Hi Zack,
>> 
>> can you also give it a quick smoke test?
>> 
>> I'm not sure if I wired up all the sysfs magic correctly inside vmwgfx, but I currently don't have a setup where I can test this.
>> 
>> Thanks,
>> Christian.
>> 
>> Am 02.02.21 um 16:14 schrieb Zack Rusin:
>>> Looks good. There’s probably not much reason to call it ttm_memory anymore as it only deals with ttm_mem_glob, we’ll likely fold it in after you submit. Thanks.
>>> 
>>> Reviewed-by: Zack Rusin <zackr@xxxxxxxxxx>
>>> 
>>> z
>>> 
>>>> On Feb 2, 2021, at 08:04, Christian König <christian.koenig@xxxxxxx> wrote:
>>>> 
>>>> Ping?
>>>> 
>>>> Especially Roland and Zack do you have any objections to this?
>>>> 
>>>> Regards,
>>>> Christian.
>>>> 
>>>> Am 28.01.21 um 14:16 schrieb Christian König:
>>>>> This is just another feature which is only used by VMWGFX, so move
>>>>> it into the driver instead.
>>>>> 
>>>>> I've tried to add the accounting sysfs file to the kobject of the drm
>>>>> minor, but I'm not 100% sure if this works as expected.
>>>>> 
>>>>> v2: fix typo in KFD and avoid 64bit divide
>>>>> v3: fix init order in VMWGFX
>>>>> 
>>>>> Signed-off-by: Christian König <christian.koenig@xxxxxxx>
>>>>> ---
>>>>> .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c  | 16 ++++++---
>>>>> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c    |  8 ++---
>>>>> drivers/gpu/drm/drm_gem_vram_helper.c         |  6 ++--
>>>>> drivers/gpu/drm/nouveau/nouveau_bo.c          |  7 ++--
>>>>> drivers/gpu/drm/nouveau/nouveau_drv.h         |  1 -
>>>>> drivers/gpu/drm/qxl/qxl_object.c              |  4 +--
>>>>> drivers/gpu/drm/radeon/radeon_object.c        |  8 ++---
>>>>> drivers/gpu/drm/ttm/Makefile                  |  7 ++--
>>>>> drivers/gpu/drm/ttm/ttm_bo.c                  | 33 +------------------
>>>>> drivers/gpu/drm/ttm/ttm_bo_util.c             |  1 -
>>>>> drivers/gpu/drm/ttm/ttm_device.c              | 22 ++++++++++---
>>>>> drivers/gpu/drm/ttm/ttm_pool.c                | 13 +-------
>>>>> drivers/gpu/drm/vmwgfx/Makefile               |  2 +-
>>>>> drivers/gpu/drm/{ttm => vmwgfx}/ttm_memory.c  | 19 ++++-------
>>>>> .../gpu/drm/vmwgfx}/ttm_memory.h              |  5 +--
>>>>> drivers/gpu/drm/vmwgfx/ttm_object.h           |  3 +-
>>>>> drivers/gpu/drm/vmwgfx/vmwgfx_bo.c            | 22 ++++++++++---
>>>>> drivers/gpu/drm/vmwgfx/vmwgfx_drv.c           |  5 +++
>>>>> drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c    | 28 +++++++++++++++-
>>>>> include/drm/ttm/ttm_bo_api.h                  | 13 ++------
>>>>> include/drm/ttm/ttm_bo_driver.h               |  1 -
>>>>> include/drm/ttm/ttm_tt.h                      |  1 +
>>>>> 22 files changed, 110 insertions(+), 115 deletions(-)
>>>>> rename drivers/gpu/drm/{ttm => vmwgfx}/ttm_memory.c (97%)
>>>>> rename {include/drm/ttm => drivers/gpu/drm/vmwgfx}/ttm_memory.h (97%)
>>>>> 
>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>>>> index 0849b68e784f..e440af37dde8 100644
>>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>>>> @@ -118,6 +118,16 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
>>>>>  */
>>>>> #define ESTIMATE_PT_SIZE(mem_size) ((mem_size) >> 14)
>>>>> +static size_t amdgpu_amdkfd_acc_size(uint64_t size)
>>>>> +{
>>>>> +size >>= PAGE_SHIFT;
>>>>> +size *= sizeof(dma_addr_t) + sizeof(void *);
>>>>> +
>>>>> +return __roundup_pow_of_two(sizeof(struct amdgpu_bo)) +
>>>>> +__roundup_pow_of_two(sizeof(struct ttm_tt)) +
>>>>> +PAGE_ALIGN(size);
>>>>> +}
>>>>> +
>>>>> static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
>>>>> uint64_t size, u32 domain, bool sg)
>>>>> {
>>>>> @@ -126,8 +136,7 @@ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
>>>>> size_t acc_size, system_mem_needed, ttm_mem_needed, vram_needed;
>>>>> int ret = 0;
>>>>> -acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size,
>>>>> -       sizeof(struct amdgpu_bo));
>>>>> +acc_size = amdgpu_amdkfd_acc_size(size);
>>>>>   vram_needed = 0;
>>>>> if (domain == AMDGPU_GEM_DOMAIN_GTT) {
>>>>> @@ -174,8 +183,7 @@ static void unreserve_mem_limit(struct amdgpu_device *adev,
>>>>> {
>>>>> size_t acc_size;
>>>>> -acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size,
>>>>> -       sizeof(struct amdgpu_bo));
>>>>> +acc_size = amdgpu_amdkfd_acc_size(size);
>>>>>   spin_lock(&kfd_mem_limit.mem_limit_lock);
>>>>> if (domain == AMDGPU_GEM_DOMAIN_GTT) {
>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>>>> index 6cc9919b12cc..599c9a132eb6 100644
>>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>>>> @@ -523,7 +523,6 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev,
>>>>> };
>>>>> struct amdgpu_bo *bo;
>>>>> unsigned long page_align, size = bp->size;
>>>>> -size_t acc_size;
>>>>> int r;
>>>>>   /* Note that GDS/GWS/OA allocates 1 page per byte/resource. */
>>>>> @@ -546,9 +545,6 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev,
>>>>>   *bo_ptr = NULL;
>>>>> -acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size,
>>>>> -       sizeof(struct amdgpu_bo));
>>>>> -
>>>>> bo = kzalloc(sizeof(struct amdgpu_bo), GFP_KERNEL);
>>>>> if (bo == NULL)
>>>>> return -ENOMEM;
>>>>> @@ -577,8 +573,8 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev,
>>>>> bo->tbo.priority = 1;
>>>>>   r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, bp->type,
>>>>> - &bo->placement, page_align, &ctx, acc_size,
>>>>> - NULL, bp->resv, &amdgpu_bo_destroy);
>>>>> + &bo->placement, page_align, &ctx,  NULL,
>>>>> + bp->resv, &amdgpu_bo_destroy);
>>>>> if (unlikely(r != 0))
>>>>> return r;
>>>>> diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
>>>>> index 0b13c8507688..a0992f0b8afd 100644
>>>>> --- a/drivers/gpu/drm/drm_gem_vram_helper.c
>>>>> +++ b/drivers/gpu/drm/drm_gem_vram_helper.c
>>>>> @@ -189,7 +189,6 @@ struct drm_gem_vram_object *drm_gem_vram_create(struct drm_device *dev,
>>>>> struct drm_vram_mm *vmm = dev->vram_mm;
>>>>> struct ttm_device *bdev;
>>>>> int ret;
>>>>> -size_t acc_size;
>>>>>   if (WARN_ONCE(!vmm, "VRAM MM not initialized"))
>>>>> return ERR_PTR(-EINVAL);
>>>>> @@ -216,7 +215,6 @@ struct drm_gem_vram_object *drm_gem_vram_create(struct drm_device *dev,
>>>>> }
>>>>>   bdev = &vmm->bdev;
>>>>> -acc_size = ttm_bo_dma_acc_size(bdev, size, sizeof(*gbo));
>>>>>   gbo->bo.bdev = bdev;
>>>>> drm_gem_vram_placement(gbo, DRM_GEM_VRAM_PL_FLAG_SYSTEM);
>>>>> @@ -226,8 +224,8 @@ struct drm_gem_vram_object *drm_gem_vram_create(struct drm_device *dev,
>>>>>  * to release gbo->bo.base and kfree gbo.
>>>>>  */
>>>>> ret = ttm_bo_init(bdev, &gbo->bo, size, ttm_bo_type_device,
>>>>> -  &gbo->placement, pg_align, false, acc_size,
>>>>> -  NULL, NULL, ttm_buffer_object_destroy);
>>>>> +  &gbo->placement, pg_align, false, NULL, NULL,
>>>>> +  ttm_buffer_object_destroy);
>>>>> if (ret)
>>>>> return ERR_PTR(ret);
>>>>> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
>>>>> index c177940d6e2c..ca2a8ae1938e 100644
>>>>> --- a/drivers/gpu/drm/nouveau/nouveau_bo.c
>>>>> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
>>>>> @@ -300,18 +300,15 @@ nouveau_bo_init(struct nouveau_bo *nvbo, u64 size, int align, u32 domain,
>>>>> struct sg_table *sg, struct dma_resv *robj)
>>>>> {
>>>>> int type = sg ? ttm_bo_type_sg : ttm_bo_type_device;
>>>>> -size_t acc_size;
>>>>> int ret;
>>>>> -acc_size = ttm_bo_dma_acc_size(nvbo->bo.bdev, size, sizeof(*nvbo));
>>>>> -
>>>>> nvbo->bo.mem.num_pages = size >> PAGE_SHIFT;
>>>>> nouveau_bo_placement_set(nvbo, domain, 0);
>>>>> INIT_LIST_HEAD(&nvbo->io_reserve_lru);
>>>>>   ret = ttm_bo_init(nvbo->bo.bdev, &nvbo->bo, size, type,
>>>>> -  &nvbo->placement, align >> PAGE_SHIFT, false,
>>>>> -  acc_size, sg, robj, nouveau_bo_del_ttm);
>>>>> +  &nvbo->placement, align >> PAGE_SHIFT, false, sg,
>>>>> +  robj, nouveau_bo_del_ttm);
>>>>> if (ret) {
>>>>> /* ttm will call nouveau_bo_del_ttm if it fails.. */
>>>>> return ret;
>>>>> diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
>>>>> index edf9d1ee9d58..a491c2c1c56e 100644
>>>>> --- a/drivers/gpu/drm/nouveau/nouveau_drv.h
>>>>> +++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
>>>>> @@ -54,7 +54,6 @@
>>>>> #include <drm/ttm/ttm_bo_api.h>
>>>>> #include <drm/ttm/ttm_bo_driver.h>
>>>>> #include <drm/ttm/ttm_placement.h>
>>>>> -#include <drm/ttm/ttm_memory.h>
>>>>>   #include <drm/drm_audio_component.h>
>>>>> diff --git a/drivers/gpu/drm/qxl/qxl_object.c b/drivers/gpu/drm/qxl/qxl_object.c
>>>>> index ceebc5881f68..705b51535492 100644
>>>>> --- a/drivers/gpu/drm/qxl/qxl_object.c
>>>>> +++ b/drivers/gpu/drm/qxl/qxl_object.c
>>>>> @@ -138,8 +138,8 @@ int qxl_bo_create(struct qxl_device *qdev,
>>>>> qxl_ttm_placement_from_domain(bo, domain);
>>>>>   r = ttm_bo_init_reserved(&qdev->mman.bdev, &bo->tbo, size, type,
>>>>> - &bo->placement, 0, &ctx, size,
>>>>> - NULL, NULL, &qxl_ttm_bo_destroy);
>>>>> + &bo->placement, 0, &ctx, NULL, NULL,
>>>>> + &qxl_ttm_bo_destroy);
>>>>> if (unlikely(r != 0)) {
>>>>> if (r != -ERESTARTSYS)
>>>>> dev_err(qdev->ddev.dev,
>>>>> diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
>>>>> index 6a336284466f..804f7a427be7 100644
>>>>> --- a/drivers/gpu/drm/radeon/radeon_object.c
>>>>> +++ b/drivers/gpu/drm/radeon/radeon_object.c
>>>>> @@ -159,7 +159,6 @@ int radeon_bo_create(struct radeon_device *rdev,
>>>>> struct radeon_bo *bo;
>>>>> enum ttm_bo_type type;
>>>>> unsigned long page_align = roundup(byte_align, PAGE_SIZE) >> PAGE_SHIFT;
>>>>> -size_t acc_size;
>>>>> int r;
>>>>>   size = ALIGN(size, PAGE_SIZE);
>>>>> @@ -173,9 +172,6 @@ int radeon_bo_create(struct radeon_device *rdev,
>>>>> }
>>>>> *bo_ptr = NULL;
>>>>> -acc_size = ttm_bo_dma_acc_size(&rdev->mman.bdev, size,
>>>>> -       sizeof(struct radeon_bo));
>>>>> -
>>>>> bo = kzalloc(sizeof(struct radeon_bo), GFP_KERNEL);
>>>>> if (bo == NULL)
>>>>> return -ENOMEM;
>>>>> @@ -230,8 +226,8 @@ int radeon_bo_create(struct radeon_device *rdev,
>>>>> /* Kernel allocation are uninterruptible */
>>>>> down_read(&rdev->pm.mclk_lock);
>>>>> r = ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type,
>>>>> -&bo->placement, page_align, !kernel, acc_size,
>>>>> -sg, resv, &radeon_ttm_bo_destroy);
>>>>> +&bo->placement, page_align, !kernel, sg, resv,
>>>>> +&radeon_ttm_bo_destroy);
>>>>> up_read(&rdev->pm.mclk_lock);
>>>>> if (unlikely(r != 0)) {
>>>>> return r;
>>>>> diff --git a/drivers/gpu/drm/ttm/Makefile b/drivers/gpu/drm/ttm/Makefile
>>>>> index 8e6437eadabe..40e5e9da7953 100644
>>>>> --- a/drivers/gpu/drm/ttm/Makefile
>>>>> +++ b/drivers/gpu/drm/ttm/Makefile
>>>>> @@ -2,10 +2,9 @@
>>>>> #
>>>>> # Makefile for the drm device driver.  This driver provides support for the
>>>>> -ttm-y := ttm_memory.o ttm_tt.o ttm_bo.o \
>>>>> -ttm_bo_util.o ttm_bo_vm.o ttm_module.o \
>>>>> -ttm_execbuf_util.o ttm_range_manager.o \
>>>>> -ttm_resource.o ttm_pool.o ttm_device.o
>>>>> +ttm-y := ttm_tt.o ttm_bo.o ttm_bo_util.o ttm_bo_vm.o ttm_module.o \
>>>>> +ttm_execbuf_util.o ttm_range_manager.o ttm_resource.o ttm_pool.o \
>>>>> +ttm_device.o
>>>>> ttm-$(CONFIG_AGP) += ttm_agp_backend.o
>>>>>   obj-$(CONFIG_DRM_TTM) += ttm.o
>>>>> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
>>>>> index 643befc1a6f2..e38102282fd5 100644
>>>>> --- a/drivers/gpu/drm/ttm/ttm_bo.c
>>>>> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
>>>>> @@ -425,7 +425,6 @@ static void ttm_bo_release(struct kref *kref)
>>>>> struct ttm_buffer_object *bo =
>>>>>     container_of(kref, struct ttm_buffer_object, kref);
>>>>> struct ttm_device *bdev = bo->bdev;
>>>>> -size_t acc_size = bo->acc_size;
>>>>> int ret;
>>>>>   if (!bo->deleted) {
>>>>> @@ -485,7 +484,6 @@ static void ttm_bo_release(struct kref *kref)
>>>>> if (!ttm_bo_uses_embedded_gem_object(bo))
>>>>> dma_resv_fini(&bo->base._resv);
>>>>> bo->destroy(bo);
>>>>> -ttm_mem_global_free(&ttm_mem_glob, acc_size);
>>>>> }
>>>>>   void ttm_bo_put(struct ttm_buffer_object *bo)
>>>>> @@ -1046,25 +1044,13 @@ int ttm_bo_init_reserved(struct ttm_device *bdev,
>>>>>  struct ttm_placement *placement,
>>>>>  uint32_t page_alignment,
>>>>>  struct ttm_operation_ctx *ctx,
>>>>> - size_t acc_size,
>>>>>  struct sg_table *sg,
>>>>>  struct dma_resv *resv,
>>>>>  void (*destroy) (struct ttm_buffer_object *))
>>>>> {
>>>>> -struct ttm_mem_global *mem_glob = &ttm_mem_glob;
>>>>> bool locked;
>>>>> int ret = 0;
>>>>> -ret = ttm_mem_global_alloc(mem_glob, acc_size, ctx);
>>>>> -if (ret) {
>>>>> -pr_err("Out of kernel memory\n");
>>>>> -if (destroy)
>>>>> -(*destroy)(bo);
>>>>> -else
>>>>> -kfree(bo);
>>>>> -return -ENOMEM;
>>>>> -}
>>>>> -
>>>>> bo->destroy = destroy ? destroy : ttm_bo_default_destroy;
>>>>>   kref_init(&bo->kref);
>>>>> @@ -1081,7 +1067,6 @@ int ttm_bo_init_reserved(struct ttm_device *bdev,
>>>>> bo->mem.bus.addr = NULL;
>>>>> bo->moving = NULL;
>>>>> bo->mem.placement = 0;
>>>>> -bo->acc_size = acc_size;
>>>>> bo->pin_count = 0;
>>>>> bo->sg = sg;
>>>>> if (resv) {
>>>>> @@ -1142,7 +1127,6 @@ int ttm_bo_init(struct ttm_device *bdev,
>>>>> struct ttm_placement *placement,
>>>>> uint32_t page_alignment,
>>>>> bool interruptible,
>>>>> -size_t acc_size,
>>>>> struct sg_table *sg,
>>>>> struct dma_resv *resv,
>>>>> void (*destroy) (struct ttm_buffer_object *))
>>>>> @@ -1151,8 +1135,7 @@ int ttm_bo_init(struct ttm_device *bdev,
>>>>> int ret;
>>>>>   ret = ttm_bo_init_reserved(bdev, bo, size, type, placement,
>>>>> -   page_alignment, &ctx, acc_size,
>>>>> -   sg, resv, destroy);
>>>>> +   page_alignment, &ctx, sg, resv, destroy);
>>>>> if (ret)
>>>>> return ret;
>>>>> @@ -1163,20 +1146,6 @@ int ttm_bo_init(struct ttm_device *bdev,
>>>>> }
>>>>> EXPORT_SYMBOL(ttm_bo_init);
>>>>> -size_t ttm_bo_dma_acc_size(struct ttm_device *bdev,
>>>>> -   unsigned long bo_size,
>>>>> -   unsigned struct_size)
>>>>> -{
>>>>> -unsigned npages = (PAGE_ALIGN(bo_size)) >> PAGE_SHIFT;
>>>>> -size_t size = 0;
>>>>> -
>>>>> -size += ttm_round_pot(struct_size);
>>>>> -size += ttm_round_pot(npages * (2*sizeof(void *) + sizeof(dma_addr_t)));
>>>>> -size += ttm_round_pot(sizeof(struct ttm_tt));
>>>>> -return size;
>>>>> -}
>>>>> -EXPORT_SYMBOL(ttm_bo_dma_acc_size);
>>>>> -
>>>>> /*
>>>>>  * buffer object vm functions.
>>>>>  */
>>>>> diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
>>>>> index db0f2661d504..031e5819fec4 100644
>>>>> --- a/drivers/gpu/drm/ttm/ttm_bo_util.c
>>>>> +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
>>>>> @@ -309,7 +309,6 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
>>>>>   kref_init(&fbo->base.kref);
>>>>> fbo->base.destroy = &ttm_transfered_destroy;
>>>>> -fbo->base.acc_size = 0;
>>>>> fbo->base.pin_count = 0;
>>>>> if (bo->type != ttm_bo_type_sg)
>>>>> fbo->base.base.resv = &fbo->base.base._resv;
>>>>> diff --git a/drivers/gpu/drm/ttm/ttm_device.c b/drivers/gpu/drm/ttm/ttm_device.c
>>>>> index ac0903c9e60a..6bde344e5da7 100644
>>>>> --- a/drivers/gpu/drm/ttm/ttm_device.c
>>>>> +++ b/drivers/gpu/drm/ttm/ttm_device.c
>>>>> @@ -27,9 +27,12 @@
>>>>>   #define pr_fmt(fmt) "[TTM DEVICE] " fmt
>>>>> +#include <linux/mm.h>
>>>>> +
>>>>> #include <drm/ttm/ttm_device.h>
>>>>> -#include <drm/ttm/ttm_memory.h>
>>>>> +#include <drm/ttm/ttm_tt.h>
>>>>> #include <drm/ttm/ttm_placement.h>
>>>>> +#include <drm/ttm/ttm_bo_api.h>
>>>>>   #include "ttm_module.h"
>>>>> @@ -49,9 +52,11 @@ static void ttm_global_release(void)
>>>>> if (--ttm_glob_use_count > 0)
>>>>> goto out;
>>>>> +ttm_pool_mgr_fini();
>>>>> +ttm_tt_mgr_fini();
>>>>> +
>>>>> kobject_del(&glob->kobj);
>>>>> kobject_put(&glob->kobj);
>>>>> -ttm_mem_global_release(&ttm_mem_glob);
>>>>> __free_page(glob->dummy_read_page);
>>>>> memset(glob, 0, sizeof(*glob));
>>>>> out:
>>>>> @@ -61,6 +66,8 @@ static void ttm_global_release(void)
>>>>> static int ttm_global_init(void)
>>>>> {
>>>>> struct ttm_global *glob = &ttm_glob;
>>>>> +unsigned long num_pages;
>>>>> +struct sysinfo si;
>>>>> int ret = 0;
>>>>> unsigned i;
>>>>> @@ -68,9 +75,14 @@ static int ttm_global_init(void)
>>>>> if (++ttm_glob_use_count > 1)
>>>>> goto out;
>>>>> -ret = ttm_mem_global_init(&ttm_mem_glob);
>>>>> -if (ret)
>>>>> -goto out;
>>>>> +si_meminfo(&si);
>>>>> +
>>>>> +/* Limit the number of pages in the pool to about 50% of the total
>>>>> + * system memory.
>>>>> + */
>>>>> +num_pages = ((u64)si.totalram * si.mem_unit) >> PAGE_SHIFT;
>>>>> +ttm_pool_mgr_init(num_pages * 50 / 100);
>>>>> +ttm_tt_mgr_init();
>>>>>   spin_lock_init(&glob->lru_lock);
>>>>> glob->dummy_read_page = alloc_page(__GFP_ZERO | GFP_DMA32);
>>>>> diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
>>>>> index e0617717113f..6b0f957d63d5 100644
>>>>> --- a/drivers/gpu/drm/ttm/ttm_pool.c
>>>>> +++ b/drivers/gpu/drm/ttm/ttm_pool.c
>>>>> @@ -404,16 +404,10 @@ int ttm_pool_alloc(struct ttm_pool *pool, struct ttm_tt *tt,
>>>>> caching = pages + (1 << order);
>>>>> }
>>>>> -r = ttm_mem_global_alloc_page(&ttm_mem_glob, p,
>>>>> -      (1 << order) * PAGE_SIZE,
>>>>> -      ctx);
>>>>> -if (r)
>>>>> -goto error_free_page;
>>>>> -
>>>>> if (dma_addr) {
>>>>> r = ttm_pool_map(pool, order, p, &dma_addr);
>>>>> if (r)
>>>>> -goto error_global_free;
>>>>> +goto error_free_page;
>>>>> }
>>>>>   num_pages -= 1 << order;
>>>>> @@ -427,9 +421,6 @@ int ttm_pool_alloc(struct ttm_pool *pool, struct ttm_tt *tt,
>>>>>   return 0;
>>>>> -error_global_free:
>>>>> -ttm_mem_global_free_page(&ttm_mem_glob, p, (1 << order) * PAGE_SIZE);
>>>>> -
>>>>> error_free_page:
>>>>> ttm_pool_free_page(pool, tt->caching, order, p);
>>>>> @@ -464,8 +455,6 @@ void ttm_pool_free(struct ttm_pool *pool, struct ttm_tt *tt)
>>>>>   order = ttm_pool_page_order(pool, p);
>>>>> num_pages = 1ULL << order;
>>>>> -ttm_mem_global_free_page(&ttm_mem_glob, p,
>>>>> - num_pages * PAGE_SIZE);
>>>>> if (tt->dma_address)
>>>>> ttm_pool_unmap(pool, tt->dma_address[i], num_pages);
>>>>> diff --git a/drivers/gpu/drm/vmwgfx/Makefile b/drivers/gpu/drm/vmwgfx/Makefile
>>>>> index cc4cdca7176e..8c02fa5852e7 100644
>>>>> --- a/drivers/gpu/drm/vmwgfx/Makefile
>>>>> +++ b/drivers/gpu/drm/vmwgfx/Makefile
>>>>> @@ -9,7 +9,7 @@ vmwgfx-y := vmwgfx_execbuf.o vmwgfx_gmr.o vmwgfx_kms.o vmwgfx_drv.o \
>>>>>     vmwgfx_cotable.o vmwgfx_so.o vmwgfx_binding.o vmwgfx_msg.o \
>>>>>     vmwgfx_simple_resource.o vmwgfx_va.o vmwgfx_blit.o \
>>>>>     vmwgfx_validation.o vmwgfx_page_dirty.o vmwgfx_streamoutput.o \
>>>>> -    ttm_object.o ttm_lock.o
>>>>> +    ttm_object.o ttm_lock.o ttm_memory.o
>>>>>   vmwgfx-$(CONFIG_TRANSPARENT_HUGEPAGE) += vmwgfx_thp.o
>>>>> obj-$(CONFIG_DRM_VMWGFX) := vmwgfx.o
>>>>> diff --git a/drivers/gpu/drm/ttm/ttm_memory.c b/drivers/gpu/drm/vmwgfx/ttm_memory.c
>>>>> similarity index 97%
>>>>> rename from drivers/gpu/drm/ttm/ttm_memory.c
>>>>> rename to drivers/gpu/drm/vmwgfx/ttm_memory.c
>>>>> index 634a85c2dc4c..1306d9e0f095 100644
>>>>> --- a/drivers/gpu/drm/ttm/ttm_memory.c
>>>>> +++ b/drivers/gpu/drm/vmwgfx/ttm_memory.c
>>>>> @@ -28,7 +28,6 @@
>>>>>   #define pr_fmt(fmt) "[TTM] " fmt
>>>>> -#include <drm/ttm/ttm_memory.h>
>>>>> #include <linux/spinlock.h>
>>>>> #include <linux/sched.h>
>>>>> #include <linux/wait.h>
>>>>> @@ -36,10 +35,11 @@
>>>>> #include <linux/module.h>
>>>>> #include <linux/slab.h>
>>>>> #include <linux/swap.h>
>>>>> -#include <drm/ttm/ttm_pool.h>
>>>>> -#include <drm/ttm/ttm_tt.h>
>>>>> -#include "ttm_module.h"
>>>>> +#include <drm/drm_device.h>
>>>>> +#include <drm/drm_file.h>
>>>>> +
>>>>> +#include "ttm_memory.h"
>>>>>   #define TTM_MEMORY_ALLOC_RETRIES 4
>>>>> @@ -414,7 +414,7 @@ static int ttm_mem_init_dma32_zone(struct ttm_mem_global *glob,
>>>>> }
>>>>> #endif
>>>>> -int ttm_mem_global_init(struct ttm_mem_global *glob)
>>>>> +int ttm_mem_global_init(struct ttm_mem_global *glob, struct drm_device *dev)
>>>>> {
>>>>> struct sysinfo si;
>>>>> int ret;
>>>>> @@ -425,7 +425,8 @@ int ttm_mem_global_init(struct ttm_mem_global *glob)
>>>>> glob->swap_queue = create_singlethread_workqueue("ttm_swap");
>>>>> INIT_WORK(&glob->work, ttm_shrink_work);
>>>>> ret = kobject_init_and_add(
>>>>> -&glob->kobj, &ttm_mem_glob_kobj_type, ttm_get_kobj(), "memory_accounting");
>>>>> +&glob->kobj, &ttm_mem_glob_kobj_type, &dev->primary->kdev->kobj,
>>>>> +"memory_accounting");
>>>>> if (unlikely(ret != 0)) {
>>>>> kobject_put(&glob->kobj);
>>>>> return ret;
>>>>> @@ -453,8 +454,6 @@ int ttm_mem_global_init(struct ttm_mem_global *glob)
>>>>> pr_info("Zone %7s: Available graphics memory: %llu KiB\n",
>>>>> zone->name, (unsigned long long)zone->max_mem >> 10);
>>>>> }
>>>>> -ttm_pool_mgr_init(glob->zone_kernel->max_mem/(2*PAGE_SIZE));
>>>>> -ttm_tt_mgr_init();
>>>>> return 0;
>>>>> out_no_zone:
>>>>> ttm_mem_global_release(glob);
>>>>> @@ -466,10 +465,6 @@ void ttm_mem_global_release(struct ttm_mem_global *glob)
>>>>> struct ttm_mem_zone *zone;
>>>>> unsigned int i;
>>>>> -/* let the page allocator first stop the shrink work. */
>>>>> -ttm_pool_mgr_fini();
>>>>> -ttm_tt_mgr_fini();
>>>>> -
>>>>> flush_workqueue(glob->swap_queue);
>>>>> destroy_workqueue(glob->swap_queue);
>>>>> glob->swap_queue = NULL;
>>>>> diff --git a/include/drm/ttm/ttm_memory.h b/drivers/gpu/drm/vmwgfx/ttm_memory.h
>>>>> similarity index 97%
>>>>> rename from include/drm/ttm/ttm_memory.h
>>>>> rename to drivers/gpu/drm/vmwgfx/ttm_memory.h
>>>>> index c1f167881e33..850ee6c867da 100644
>>>>> --- a/include/drm/ttm/ttm_memory.h
>>>>> +++ b/drivers/gpu/drm/vmwgfx/ttm_memory.h
>>>>> @@ -35,7 +35,8 @@
>>>>> #include <linux/errno.h>
>>>>> #include <linux/kobject.h>
>>>>> #include <linux/mm.h>
>>>>> -#include "ttm_bo_api.h"
>>>>> +
>>>>> +#include <drm/ttm/ttm_bo_api.h>
>>>>>   /**
>>>>>  * struct ttm_mem_global - Global memory accounting structure.
>>>>> @@ -79,7 +80,7 @@ extern struct ttm_mem_global {
>>>>> #endif
>>>>> } ttm_mem_glob;
>>>>> -int ttm_mem_global_init(struct ttm_mem_global *glob);
>>>>> +int ttm_mem_global_init(struct ttm_mem_global *glob, struct drm_device *dev);
>>>>> void ttm_mem_global_release(struct ttm_mem_global *glob);
>>>>> int ttm_mem_global_alloc(struct ttm_mem_global *glob, uint64_t memory,
>>>>>  struct ttm_operation_ctx *ctx);
>>>>> diff --git a/drivers/gpu/drm/vmwgfx/ttm_object.h b/drivers/gpu/drm/vmwgfx/ttm_object.h
>>>>> index ede26df87c93..49b064f0cb19 100644
>>>>> --- a/drivers/gpu/drm/vmwgfx/ttm_object.h
>>>>> +++ b/drivers/gpu/drm/vmwgfx/ttm_object.h
>>>>> @@ -43,7 +43,8 @@
>>>>> #include <linux/rcupdate.h>
>>>>>   #include <drm/drm_hashtab.h>
>>>>> -#include <drm/ttm/ttm_memory.h>
>>>>> +
>>>>> +#include "ttm_memory.h"
>>>>>   /**
>>>>>  * enum ttm_ref_type
>>>>> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>>>> index 6b3bfd8c678a..50e529a01677 100644
>>>>> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>>>> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>>>> @@ -507,11 +507,16 @@ int vmw_bo_create_kernel(struct vmw_private *dev_priv, unsigned long size,
>>>>> acc_size = ttm_round_pot(sizeof(*bo));
>>>>> acc_size += ttm_round_pot(npages * sizeof(void *));
>>>>> acc_size += ttm_round_pot(sizeof(struct ttm_tt));
>>>>> +
>>>>> +ret = ttm_mem_global_alloc(&ttm_mem_glob, acc_size, &ctx);
>>>>> +if (unlikely(ret))
>>>>> +goto error_free;
>>>>> +
>>>>> ret = ttm_bo_init_reserved(&dev_priv->bdev, bo, size,
>>>>>    ttm_bo_type_device, placement, 0,
>>>>> -   &ctx, acc_size, NULL, NULL, NULL);
>>>>> +   &ctx, NULL, NULL, NULL);
>>>>> if (unlikely(ret))
>>>>> -goto error_free;
>>>>> +goto error_account;
>>>>>   ttm_bo_pin(bo);
>>>>> ttm_bo_unreserve(bo);
>>>>> @@ -519,6 +524,9 @@ int vmw_bo_create_kernel(struct vmw_private *dev_priv, unsigned long size,
>>>>>   return 0;
>>>>> +error_account:
>>>>> +ttm_mem_global_free(&ttm_mem_glob, acc_size);
>>>>> +
>>>>> error_free:
>>>>> kfree(bo);
>>>>> return ret;
>>>>> @@ -558,11 +566,17 @@ int vmw_bo_init(struct vmw_private *dev_priv,
>>>>> vmw_bo->base.priority = 3;
>>>>> vmw_bo->res_tree = RB_ROOT;
>>>>> +ret = ttm_mem_global_alloc(&ttm_mem_glob, acc_size, &ctx);
>>>>> +if (unlikely(ret))
>>>>> +return ret;
>>>>> +
>>>>> ret = ttm_bo_init_reserved(bdev, &vmw_bo->base, size,
>>>>>    ttm_bo_type_device, placement,
>>>>> -   0, &ctx, acc_size, NULL, NULL, bo_free);
>>>>> -if (unlikely(ret))
>>>>> +   0, &ctx, NULL, NULL, bo_free);
>>>>> +if (unlikely(ret)) {
>>>>> +ttm_mem_global_free(&ttm_mem_glob, acc_size);
>>>>> return ret;
>>>>> +}
>>>>>   if (pin)
>>>>> ttm_bo_pin(&vmw_bo->base);
>>>>> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
>>>>> index 710ba5169a74..6c0ca1011629 100644
>>>>> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
>>>>> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
>>>>> @@ -1268,6 +1268,7 @@ static void vmw_remove(struct pci_dev *pdev)
>>>>> {
>>>>> struct drm_device *dev = pci_get_drvdata(pdev);
>>>>> +ttm_mem_global_release(&ttm_mem_glob);
>>>>> drm_dev_unregister(dev);
>>>>> vmw_driver_unload(dev);
>>>>> }
>>>>> @@ -1518,6 +1519,10 @@ static int vmw_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>   pci_set_drvdata(pdev, &vmw->drm);
>>>>> +ret = ttm_mem_global_init(&ttm_mem_glob, &vmw->drm);
>>>>> +if (ret)
>>>>> +return ret;
>>>>> +
>>>>> ret = vmw_driver_load(vmw, ent->device);
>>>>> if (ret)
>>>>> return ret;
>>>>> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>>>> index d1bfa59579f1..63f10c865061 100644
>>>>> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>>>> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>>>> @@ -576,11 +576,31 @@ static void vmw_ttm_destroy(struct ttm_device *bdev, struct ttm_tt *ttm)
>>>>> static int vmw_ttm_populate(struct ttm_device *bdev,
>>>>>     struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
>>>>> {
>>>>> +unsigned int i;
>>>>> +int ret;
>>>>> +
>>>>> /* TODO: maybe completely drop this ? */
>>>>> if (ttm_tt_is_populated(ttm))
>>>>> return 0;
>>>>> -return ttm_pool_alloc(&bdev->pool, ttm, ctx);
>>>>> +ret = ttm_pool_alloc(&bdev->pool, ttm, ctx);
>>>>> +if (ret)
>>>>> +return ret;
>>>>> +
>>>>> +for (i = 0; i < ttm->num_pages; ++i) {
>>>>> +ret = ttm_mem_global_alloc_page(&ttm_mem_glob, ttm->pages[i],
>>>>> +PAGE_SIZE, ctx);
>>>>> +if (ret)
>>>>> +goto error;
>>>>> +}
>>>>> +return 0;
>>>>> +
>>>>> +error:
>>>>> +while (i--)
>>>>> +ttm_mem_global_free_page(&ttm_mem_glob, ttm->pages[i],
>>>>> + PAGE_SIZE);
>>>>> +ttm_pool_free(&bdev->pool, ttm);
>>>>> +return ret;
>>>>> }
>>>>>   static void vmw_ttm_unpopulate(struct ttm_device *bdev,
>>>>> @@ -588,6 +608,7 @@ static void vmw_ttm_unpopulate(struct ttm_device *bdev,
>>>>> {
>>>>> struct vmw_ttm_tt *vmw_tt = container_of(ttm, struct vmw_ttm_tt,
>>>>>  dma_ttm);
>>>>> +unsigned int i;
>>>>>   if (vmw_tt->mob) {
>>>>> vmw_mob_destroy(vmw_tt->mob);
>>>>> @@ -595,6 +616,11 @@ static void vmw_ttm_unpopulate(struct ttm_device *bdev,
>>>>> }
>>>>>   vmw_ttm_unmap_dma(vmw_tt);
>>>>> +
>>>>> +for (i = 0; i < ttm->num_pages; ++i)
>>>>> +ttm_mem_global_free_page(&ttm_mem_glob, ttm->pages[i],
>>>>> + PAGE_SIZE);
>>>>> +
>>>>> ttm_pool_free(&bdev->pool, ttm);
>>>>> }
>>>>> diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h
>>>>> index 1297a8fb7ccb..4fb523dfab32 100644
>>>>> --- a/include/drm/ttm/ttm_bo_api.h
>>>>> +++ b/include/drm/ttm/ttm_bo_api.h
>>>>> @@ -88,7 +88,6 @@ struct ttm_tt;
>>>>>  * @type: The bo type.
>>>>>  * @destroy: Destruction function. If NULL, kfree is used.
>>>>>  * @num_pages: Actual number of pages.
>>>>> - * @acc_size: Accounted size for this object.
>>>>>  * @kref: Reference count of this buffer object. When this refcount reaches
>>>>>  * zero, the object is destroyed or put on the delayed delete list.
>>>>>  * @mem: structure describing current placement.
>>>>> @@ -125,7 +124,6 @@ struct ttm_buffer_object {
>>>>> struct ttm_device *bdev;
>>>>> enum ttm_bo_type type;
>>>>> void (*destroy) (struct ttm_buffer_object *);
>>>>> -size_t acc_size;
>>>>>   /**
>>>>> * Members not needing protection.
>>>>> @@ -357,10 +355,6 @@ void ttm_bo_unlock_delayed_workqueue(struct ttm_device *bdev, int resched);
>>>>> bool ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
>>>>>       const struct ttm_place *place);
>>>>> -size_t ttm_bo_dma_acc_size(struct ttm_device *bdev,
>>>>> -   unsigned long bo_size,
>>>>> -   unsigned struct_size);
>>>>> -
>>>>> /**
>>>>>  * ttm_bo_init_reserved
>>>>>  *
>>>>> @@ -371,7 +365,6 @@ size_t ttm_bo_dma_acc_size(struct ttm_device *bdev,
>>>>>  * @flags: Initial placement flags.
>>>>>  * @page_alignment: Data alignment in pages.
>>>>>  * @ctx: TTM operation context for memory allocation.
>>>>> - * @acc_size: Accounted size for this object.
>>>>>  * @resv: Pointer to a dma_resv, or NULL to let ttm allocate one.
>>>>>  * @destroy: Destroy function. Use NULL for kfree().
>>>>>  *
>>>>> @@ -402,8 +395,7 @@ int ttm_bo_init_reserved(struct ttm_device *bdev,
>>>>>  struct ttm_placement *placement,
>>>>>  uint32_t page_alignment,
>>>>>  struct ttm_operation_ctx *ctx,
>>>>> - size_t acc_size, struct sg_table *sg,
>>>>> - struct dma_resv *resv,
>>>>> + struct sg_table *sg, struct dma_resv *resv,
>>>>>  void (*destroy) (struct ttm_buffer_object *));
>>>>>   /**
>>>>> @@ -421,7 +413,6 @@ int ttm_bo_init_reserved(struct ttm_device *bdev,
>>>>>  * holds a pointer to a persistent shmem object. Typically, this would
>>>>>  * point to the shmem object backing a GEM object if TTM is used to back a
>>>>>  * GEM user interface.
>>>>> - * @acc_size: Accounted size for this object.
>>>>>  * @resv: Pointer to a dma_resv, or NULL to let ttm allocate one.
>>>>>  * @destroy: Destroy function. Use NULL for kfree().
>>>>>  *
>>>>> @@ -446,7 +437,7 @@ int ttm_bo_init_reserved(struct ttm_device *bdev,
>>>>> int ttm_bo_init(struct ttm_device *bdev, struct ttm_buffer_object *bo,
>>>>> size_t size, enum ttm_bo_type type,
>>>>> struct ttm_placement *placement,
>>>>> -uint32_t page_alignment, bool interrubtible, size_t acc_size,
>>>>> +uint32_t page_alignment, bool interrubtible,
>>>>> struct sg_table *sg, struct dma_resv *resv,
>>>>> void (*destroy) (struct ttm_buffer_object *));
>>>>> diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
>>>>> index 1c9bf993e252..8959c0075cfd 100644
>>>>> --- a/include/drm/ttm/ttm_bo_driver.h
>>>>> +++ b/include/drm/ttm/ttm_bo_driver.h
>>>>> @@ -40,7 +40,6 @@
>>>>> #include <drm/ttm/ttm_device.h>
>>>>>   #include "ttm_bo_api.h"
>>>>> -#include "ttm_memory.h"
>>>>> #include "ttm_placement.h"
>>>>> #include "ttm_tt.h"
>>>>> #include "ttm_pool.h"
>>>>> diff --git a/include/drm/ttm/ttm_tt.h b/include/drm/ttm/ttm_tt.h
>>>>> index cce57fb49e2c..069f8130241a 100644
>>>>> --- a/include/drm/ttm/ttm_tt.h
>>>>> +++ b/include/drm/ttm/ttm_tt.h
>>>>> @@ -30,6 +30,7 @@
>>>>> #include <linux/types.h>
>>>>> #include <drm/ttm/ttm_caching.h>
>>>>> +struct ttm_bo_device;
>>>>> struct ttm_tt;
>>>>> struct ttm_resource;
>>>>> struct ttm_buffer_object;
>> 
> 
> 
> 
> ----------
> 
> You're receiving this message because you're a member of the Linux-graphics-maintainer group from VMware, Inc..
> 
> Leave group:
> https://outlook.office365.com/owa/Linux-graphics-maintainer@xxxxxxxxxx/groupsubscription.ashx?source=EscalatedMessage&action=leave&GuestId=69d3bf6f-5242-4be4-b863-b7949752f363
> _______________________________________________
> Sent to linux-graphics-maintainer@xxxxxxxxxx

_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel




[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux