On some processors (such as arch/x86), accessing a page via a WC PAT is bypassed if the page is physically tagged in the CPU cache, and the access is serviced by the cache instead -- which leads to incoherency should the physical page itself be accessed using DMA. In order to prevent the false cache sharing of the physical pages, we need to explicitly flush the cache lines of those pages.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk --- drivers/gpu/drm/drm_gem.c | 8 ++++++-- drivers/gpu/drm/drm_gem_shmem_helper.c | 8 +++++++- drivers/gpu/drm/etnaviv/etnaviv_gem.c | 2 +- drivers/gpu/drm/gma500/gtt.c | 2 +- drivers/gpu/drm/msm/msm_gem.c | 2 +- drivers/gpu/drm/omapdrm/omap_gem.c | 2 +- drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 2 +- drivers/gpu/drm/tegra/gem.c | 2 +- drivers/gpu/drm/vgem/vgem_drv.c | 2 +- drivers/gpu/drm/vkms/vkms_gem.c | 2 +- drivers/gpu/drm/xen/xen_drm_front_gem.c | 2 +- include/drm/drm_gem.h | 2 +- 12 files changed, 23 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c index 1da67d34e55d..1948855d69e6 100644 --- a/drivers/gpu/drm/drm_gem.c +++ b/drivers/gpu/drm/drm_gem.c @@ -40,6 +40,7 @@ #include <linux/pagevec.h>
#include <drm/drm.h> +#include <drm/drm_cache.h> #include <drm/drm_device.h> #include <drm/drm_drv.h> #include <drm/drm_file.h> @@ -525,6 +526,7 @@ static void drm_gem_check_release_pagevec(struct pagevec *pvec) * drm_gem_get_pages - helper to allocate backing pages for a GEM object * from shmem * @obj: obj in question + * @clflush: whether to clear any CPU caches associated with the backing store * * This reads the page-array of the shmem-backing storage of the given gem * object. An array of pages is returned. If a page is not allocated or @@ -546,14 +548,13 @@ static void drm_gem_check_release_pagevec(struct pagevec *pvec) * drm_gem_object_init(), but not for those initialized with * drm_gem_private_object_init() only. */ -struct page **drm_gem_get_pages(struct drm_gem_object *obj) +struct page **drm_gem_get_pages(struct drm_gem_object *obj, bool clflush) { struct address_space *mapping; struct page *p, **pages; struct pagevec pvec; int i, npages;
- if (WARN_ON(!obj->filp)) return ERR_PTR(-EINVAL);
@@ -589,6 +590,9 @@ struct page **drm_gem_get_pages(struct drm_gem_object *obj) (page_to_pfn(p) >= 0x00100000UL)); }
+ if (clflush) + drm_clflush_pages(pages, npages); + return pages;
fail: diff --git a/drivers/gpu/drm/drm_gem_shmem_helper.c b/drivers/gpu/drm/drm_gem_shmem_helper.c index fb11df7aced5..78a2eb77802b 100644 --- a/drivers/gpu/drm/drm_gem_shmem_helper.c +++ b/drivers/gpu/drm/drm_gem_shmem_helper.c @@ -152,7 +152,13 @@ static int drm_gem_shmem_get_pages_locked(struct drm_gem_shmem_object *shmem) if (shmem->pages_use_count++ > 0) return 0;
- pages = drm_gem_get_pages(obj); + /* + * On some processors (such as arch/x86), accessing a page via a WC PAT + * is bypassed if the page is physically tagged in the CPU cache, and + * the access is serviced by the cache instead -- which leads to + * incoherency should the physical page itself be accessed using DMA. + */ + pages = drm_gem_get_pages(obj, !shmem->map_cached); if (IS_ERR(pages)) { DRM_DEBUG_KMS("Failed to get pages (%ld)\n", PTR_ERR(pages)); shmem->pages_use_count = 0; diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem.c b/drivers/gpu/drm/etnaviv/etnaviv_gem.c index 67d9a2b9ea6a..d8279ea363b3 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gem.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem.c @@ -58,7 +58,7 @@ static void etnaviv_gem_scatterlist_unmap(struct etnaviv_gem_object *etnaviv_obj static int etnaviv_gem_shmem_get_pages(struct etnaviv_gem_object *etnaviv_obj) { struct drm_device *dev = etnaviv_obj->base.dev; - struct page **p = drm_gem_get_pages(&etnaviv_obj->base); + struct page **p = drm_gem_get_pages(&etnaviv_obj->base, false);
if (IS_ERR(p)) { dev_dbg(dev->dev, "could not get pages: %ld\n", PTR_ERR(p)); diff --git a/drivers/gpu/drm/gma500/gtt.c b/drivers/gpu/drm/gma500/gtt.c index 9278bcfad1bf..ada56aec7e68 100644 --- a/drivers/gpu/drm/gma500/gtt.c +++ b/drivers/gpu/drm/gma500/gtt.c @@ -197,7 +197,7 @@ static int psb_gtt_attach_pages(struct gtt_range *gt)
WARN_ON(gt->pages);
- pages = drm_gem_get_pages(>->gem); + pages = drm_gem_get_pages(>->gem, false); if (IS_ERR(pages)) return PTR_ERR(pages);
diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c index c79828d31822..a7a67ef4e27e 100644 --- a/drivers/gpu/drm/msm/msm_gem.c +++ b/drivers/gpu/drm/msm/msm_gem.c @@ -102,7 +102,7 @@ static struct page **get_pages(struct drm_gem_object *obj) int npages = obj->size >> PAGE_SHIFT;
if (use_pages(obj)) - p = drm_gem_get_pages(obj); + p = drm_gem_get_pages(obj, false); else p = get_pages_vram(obj, npages);
diff --git a/drivers/gpu/drm/omapdrm/omap_gem.c b/drivers/gpu/drm/omapdrm/omap_gem.c index d8e09792793a..1ff272c4be33 100644 --- a/drivers/gpu/drm/omapdrm/omap_gem.c +++ b/drivers/gpu/drm/omapdrm/omap_gem.c @@ -236,7 +236,7 @@ static int omap_gem_attach_pages(struct drm_gem_object *obj) if (!(omap_obj->flags & OMAP_BO_MEM_SHMEM) || omap_obj->pages) return 0;
- pages = drm_gem_get_pages(obj); + pages = drm_gem_get_pages(obj, false); if (IS_ERR(pages)) { dev_err(obj->dev->dev, "could not get pages: %ld\n", PTR_ERR(pages)); return PTR_ERR(pages); diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c index 7d5ebb10323b..003b76c3e623 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c @@ -80,7 +80,7 @@ static int rockchip_gem_get_pages(struct rockchip_gem_object *rk_obj) int ret, i; struct scatterlist *s;
- rk_obj->pages = drm_gem_get_pages(&rk_obj->base); + rk_obj->pages = drm_gem_get_pages(&rk_obj->base, false); if (IS_ERR(rk_obj->pages)) return PTR_ERR(rk_obj->pages);
diff --git a/drivers/gpu/drm/tegra/gem.c b/drivers/gpu/drm/tegra/gem.c index 26af8daa9a16..f075f005ecbb 100644 --- a/drivers/gpu/drm/tegra/gem.c +++ b/drivers/gpu/drm/tegra/gem.c @@ -289,7 +289,7 @@ static int tegra_bo_get_pages(struct drm_device *drm, struct tegra_bo *bo) { int err;
- bo->pages = drm_gem_get_pages(&bo->gem); + bo->pages = drm_gem_get_pages(&bo->gem, false); if (IS_ERR(bo->pages)) return PTR_ERR(bo->pages);
diff --git a/drivers/gpu/drm/vgem/vgem_drv.c b/drivers/gpu/drm/vgem/vgem_drv.c index fa54a6d1403d..f38dd590fa45 100644 --- a/drivers/gpu/drm/vgem/vgem_drv.c +++ b/drivers/gpu/drm/vgem/vgem_drv.c @@ -272,7 +272,7 @@ static struct page **vgem_pin_pages(struct drm_vgem_gem_object *bo) if (bo->pages_pin_count++ == 0) { struct page **pages;
- pages = drm_gem_get_pages(&bo->base); + pages = drm_gem_get_pages(&bo->base, false); if (IS_ERR(pages)) { bo->pages_pin_count--; mutex_unlock(&bo->pages_lock); diff --git a/drivers/gpu/drm/vkms/vkms_gem.c b/drivers/gpu/drm/vkms/vkms_gem.c index 19a0e260a4df..1d17b0ef56c7 100644 --- a/drivers/gpu/drm/vkms/vkms_gem.c +++ b/drivers/gpu/drm/vkms/vkms_gem.c @@ -166,7 +166,7 @@ static struct page **_get_pages(struct vkms_gem_object *vkms_obj) struct drm_gem_object *gem_obj = &vkms_obj->gem;
if (!vkms_obj->pages) { - struct page **pages = drm_gem_get_pages(gem_obj); + struct page **pages = drm_gem_get_pages(gem_obj, false);
if (IS_ERR(pages)) return pages; diff --git a/drivers/gpu/drm/xen/xen_drm_front_gem.c b/drivers/gpu/drm/xen/xen_drm_front_gem.c index 4f34ef34ba60..520a15d75eb6 100644 --- a/drivers/gpu/drm/xen/xen_drm_front_gem.c +++ b/drivers/gpu/drm/xen/xen_drm_front_gem.c @@ -132,7 +132,7 @@ static struct xen_gem_object *gem_create(struct drm_device *dev, size_t size) * with the backend */ xen_obj->num_pages = DIV_ROUND_UP(size, PAGE_SIZE); - xen_obj->pages = drm_gem_get_pages(&xen_obj->base); + xen_obj->pages = drm_gem_get_pages(&xen_obj->base, false); if (IS_ERR(xen_obj->pages)) { ret = PTR_ERR(xen_obj->pages); xen_obj->pages = NULL; diff --git a/include/drm/drm_gem.h b/include/drm/drm_gem.h index c38dd35da00b..118f13b1bb29 100644 --- a/include/drm/drm_gem.h +++ b/include/drm/drm_gem.h @@ -395,7 +395,7 @@ void drm_gem_free_mmap_offset(struct drm_gem_object *obj); int drm_gem_create_mmap_offset(struct drm_gem_object *obj); int drm_gem_create_mmap_offset_size(struct drm_gem_object *obj, size_t size);
-struct page **drm_gem_get_pages(struct drm_gem_object *obj); +struct page **drm_gem_get_pages(struct drm_gem_object *obj, bool clflush); void drm_gem_put_pages(struct drm_gem_object *obj, struct page **pages, bool dirty, bool accessed);
vgem is a minimalistic driver that provides shmemfs objects to userspace that may then be used as an in-memory surface and transported across dma-buf to other drivers. Since vgem's introduction, drm_gem_shmem_helper now provides the same shmemfs facilities and so we can trim vgem to wrap the helper.
v2: The gem prime mmap helper does not handle exchanging the vma->file, so skip the conversion to the common prime helper and leave the patch only converting to the common shmem helper.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk Reviewed-by: Daniel Vetter daniel.vetter@ffwll.ch #v1 --- drivers/gpu/drm/Kconfig | 1 + drivers/gpu/drm/vgem/vgem_drv.c | 284 +++----------------------------- drivers/gpu/drm/vgem/vgem_drv.h | 11 -- 3 files changed, 25 insertions(+), 271 deletions(-)
diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig index 147d61b9674e..db2ff76638cd 100644 --- a/drivers/gpu/drm/Kconfig +++ b/drivers/gpu/drm/Kconfig @@ -278,6 +278,7 @@ source "drivers/gpu/drm/i915/Kconfig" config DRM_VGEM tristate "Virtual GEM provider" depends on DRM + select DRM_GEM_SHMEM_HELPER help Choose this option to get a virtual graphics memory manager, as used by Mesa's software renderer for enhanced performance. diff --git a/drivers/gpu/drm/vgem/vgem_drv.c b/drivers/gpu/drm/vgem/vgem_drv.c index f38dd590fa45..6811ac518509 100644 --- a/drivers/gpu/drm/vgem/vgem_drv.c +++ b/drivers/gpu/drm/vgem/vgem_drv.c @@ -38,6 +38,7 @@
#include <drm/drm_drv.h> #include <drm/drm_file.h> +#include <drm/drm_gem_shmem_helper.h> #include <drm/drm_ioctl.h> #include <drm/drm_managed.h> #include <drm/drm_prime.h> @@ -50,87 +51,11 @@ #define DRIVER_MAJOR 1 #define DRIVER_MINOR 0
-static const struct drm_gem_object_funcs vgem_gem_object_funcs; - static struct vgem_device { struct drm_device drm; struct platform_device *platform; } *vgem_device;
-static void vgem_gem_free_object(struct drm_gem_object *obj) -{ - struct drm_vgem_gem_object *vgem_obj = to_vgem_bo(obj); - - kvfree(vgem_obj->pages); - mutex_destroy(&vgem_obj->pages_lock); - - if (obj->import_attach) - drm_prime_gem_destroy(obj, vgem_obj->table); - - drm_gem_object_release(obj); - kfree(vgem_obj); -} - -static vm_fault_t vgem_gem_fault(struct vm_fault *vmf) -{ - struct vm_area_struct *vma = vmf->vma; - struct drm_vgem_gem_object *obj = vma->vm_private_data; - /* We don't use vmf->pgoff since that has the fake offset */ - unsigned long vaddr = vmf->address; - vm_fault_t ret = VM_FAULT_SIGBUS; - loff_t num_pages; - pgoff_t page_offset; - page_offset = (vaddr - vma->vm_start) >> PAGE_SHIFT; - - num_pages = DIV_ROUND_UP(obj->base.size, PAGE_SIZE); - - if (page_offset >= num_pages) - return VM_FAULT_SIGBUS; - - mutex_lock(&obj->pages_lock); - if (obj->pages) { - get_page(obj->pages[page_offset]); - vmf->page = obj->pages[page_offset]; - ret = 0; - } - mutex_unlock(&obj->pages_lock); - if (ret) { - struct page *page; - - page = shmem_read_mapping_page( - file_inode(obj->base.filp)->i_mapping, - page_offset); - if (!IS_ERR(page)) { - vmf->page = page; - ret = 0; - } else switch (PTR_ERR(page)) { - case -ENOSPC: - case -ENOMEM: - ret = VM_FAULT_OOM; - break; - case -EBUSY: - ret = VM_FAULT_RETRY; - break; - case -EFAULT: - case -EINVAL: - ret = VM_FAULT_SIGBUS; - break; - default: - WARN_ON(PTR_ERR(page)); - ret = VM_FAULT_SIGBUS; - break; - } - - } - return ret; -} - -static const struct vm_operations_struct vgem_gem_vm_ops = { - .fault = vgem_gem_fault, - .open = drm_gem_vm_open, - .close = drm_gem_vm_close, -}; - static int vgem_open(struct drm_device *dev, struct drm_file *file) { struct vgem_file *vfile; @@ -159,33 +84,15 @@ static void vgem_postclose(struct drm_device *dev, struct drm_file *file) kfree(vfile); }
-static struct drm_vgem_gem_object *__vgem_gem_create(struct drm_device *dev, - unsigned long size) +static struct drm_gem_shmem_object *__vgem_gem_create(struct drm_device *dev, + unsigned long size) { - struct drm_vgem_gem_object *obj; - int ret; - - obj = kzalloc(sizeof(*obj), GFP_KERNEL); - if (!obj) - return ERR_PTR(-ENOMEM); - - obj->base.funcs = &vgem_gem_object_funcs; - - ret = drm_gem_object_init(dev, &obj->base, roundup(size, PAGE_SIZE)); - if (ret) { - kfree(obj); - return ERR_PTR(ret); - } - - mutex_init(&obj->pages_lock); - - return obj; -} - -static void __vgem_gem_destroy(struct drm_vgem_gem_object *obj) -{ - drm_gem_object_release(&obj->base); - kfree(obj); + /* + * How should we use map_cached? prime vmaps should be write-combining + * so that device coherency can be assumed, but ordinary user mmaps + * would benefit from being write-back. + */ + return drm_gem_shmem_create(dev, round_up(size, PAGE_SIZE)); }
static struct drm_gem_object *vgem_gem_create(struct drm_device *dev, @@ -193,7 +100,7 @@ static struct drm_gem_object *vgem_gem_create(struct drm_device *dev, unsigned int *handle, unsigned long size) { - struct drm_vgem_gem_object *obj; + struct drm_gem_shmem_object *obj; int ret;
obj = __vgem_gem_create(dev, size); @@ -234,154 +141,6 @@ static int vgem_gem_dumb_create(struct drm_file *file, struct drm_device *dev, return 0; }
-static struct drm_ioctl_desc vgem_ioctls[] = { - DRM_IOCTL_DEF_DRV(VGEM_FENCE_ATTACH, vgem_fence_attach_ioctl, DRM_RENDER_ALLOW), - DRM_IOCTL_DEF_DRV(VGEM_FENCE_SIGNAL, vgem_fence_signal_ioctl, DRM_RENDER_ALLOW), -}; - -static int vgem_mmap(struct file *filp, struct vm_area_struct *vma) -{ - unsigned long flags = vma->vm_flags; - int ret; - - ret = drm_gem_mmap(filp, vma); - if (ret) - return ret; - - /* Keep the WC mmaping set by drm_gem_mmap() but our pages - * are ordinary and not special. - */ - vma->vm_flags = flags | VM_DONTEXPAND | VM_DONTDUMP; - return 0; -} - -static const struct file_operations vgem_driver_fops = { - .owner = THIS_MODULE, - .open = drm_open, - .mmap = vgem_mmap, - .poll = drm_poll, - .read = drm_read, - .unlocked_ioctl = drm_ioctl, - .compat_ioctl = drm_compat_ioctl, - .release = drm_release, -}; - -static struct page **vgem_pin_pages(struct drm_vgem_gem_object *bo) -{ - mutex_lock(&bo->pages_lock); - if (bo->pages_pin_count++ == 0) { - struct page **pages; - - pages = drm_gem_get_pages(&bo->base, false); - if (IS_ERR(pages)) { - bo->pages_pin_count--; - mutex_unlock(&bo->pages_lock); - return pages; - } - - bo->pages = pages; - } - mutex_unlock(&bo->pages_lock); - - return bo->pages; -} - -static void vgem_unpin_pages(struct drm_vgem_gem_object *bo) -{ - mutex_lock(&bo->pages_lock); - if (--bo->pages_pin_count == 0) { - drm_gem_put_pages(&bo->base, bo->pages, true, true); - bo->pages = NULL; - } - mutex_unlock(&bo->pages_lock); -} - -static int vgem_prime_pin(struct drm_gem_object *obj) -{ - struct drm_vgem_gem_object *bo = to_vgem_bo(obj); - long n_pages = obj->size >> PAGE_SHIFT; - struct page **pages; - - pages = vgem_pin_pages(bo); - if (IS_ERR(pages)) - return PTR_ERR(pages); - - /* Flush the object from the CPU cache so that importers can rely - * on coherent indirect access via the exported dma-address. - */ - drm_clflush_pages(pages, n_pages); - - return 0; -} - -static void vgem_prime_unpin(struct drm_gem_object *obj) -{ - struct drm_vgem_gem_object *bo = to_vgem_bo(obj); - - vgem_unpin_pages(bo); -} - -static struct sg_table *vgem_prime_get_sg_table(struct drm_gem_object *obj) -{ - struct drm_vgem_gem_object *bo = to_vgem_bo(obj); - - return drm_prime_pages_to_sg(obj->dev, bo->pages, bo->base.size >> PAGE_SHIFT); -} - -static struct drm_gem_object* vgem_prime_import(struct drm_device *dev, - struct dma_buf *dma_buf) -{ - struct vgem_device *vgem = container_of(dev, typeof(*vgem), drm); - - return drm_gem_prime_import_dev(dev, dma_buf, &vgem->platform->dev); -} - -static struct drm_gem_object *vgem_prime_import_sg_table(struct drm_device *dev, - struct dma_buf_attachment *attach, struct sg_table *sg) -{ - struct drm_vgem_gem_object *obj; - int npages; - - obj = __vgem_gem_create(dev, attach->dmabuf->size); - if (IS_ERR(obj)) - return ERR_CAST(obj); - - npages = PAGE_ALIGN(attach->dmabuf->size) / PAGE_SIZE; - - obj->table = sg; - obj->pages = kvmalloc_array(npages, sizeof(struct page *), GFP_KERNEL); - if (!obj->pages) { - __vgem_gem_destroy(obj); - return ERR_PTR(-ENOMEM); - } - - obj->pages_pin_count++; /* perma-pinned */ - drm_prime_sg_to_page_addr_arrays(obj->table, obj->pages, NULL, - npages); - return &obj->base; -} - -static void *vgem_prime_vmap(struct drm_gem_object *obj) -{ - struct drm_vgem_gem_object *bo = to_vgem_bo(obj); - long n_pages = obj->size >> PAGE_SHIFT; - struct page **pages; - - pages = vgem_pin_pages(bo); - if (IS_ERR(pages)) - return NULL; - - return vmap(pages, n_pages, 0, pgprot_writecombine(PAGE_KERNEL)); -} - -static void vgem_prime_vunmap(struct drm_gem_object *obj, void *vaddr) -{ - struct drm_vgem_gem_object *bo = to_vgem_bo(obj); - - vunmap(vaddr); - vgem_unpin_pages(bo); -} - static int vgem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) { @@ -405,16 +164,21 @@ static int vgem_prime_mmap(struct drm_gem_object *obj, return 0; }
-static const struct drm_gem_object_funcs vgem_gem_object_funcs = { - .free = vgem_gem_free_object, - .pin = vgem_prime_pin, - .unpin = vgem_prime_unpin, - .get_sg_table = vgem_prime_get_sg_table, - .vmap = vgem_prime_vmap, - .vunmap = vgem_prime_vunmap, - .vm_ops = &vgem_gem_vm_ops, +static struct drm_ioctl_desc vgem_ioctls[] = { + DRM_IOCTL_DEF_DRV(VGEM_FENCE_ATTACH, vgem_fence_attach_ioctl, DRM_RENDER_ALLOW), + DRM_IOCTL_DEF_DRV(VGEM_FENCE_SIGNAL, vgem_fence_signal_ioctl, DRM_RENDER_ALLOW), };
+DEFINE_DRM_GEM_FOPS(vgem_driver_fops); + +static struct drm_gem_object *vgem_prime_import(struct drm_device *dev, + struct dma_buf *dma_buf) +{ + struct vgem_device *vgem = container_of(dev, typeof(*vgem), drm); + + return drm_gem_prime_import_dev(dev, dma_buf, &vgem->platform->dev); +} + static struct drm_driver vgem_driver = { .driver_features = DRIVER_GEM | DRIVER_RENDER, .open = vgem_open, @@ -428,7 +192,7 @@ static struct drm_driver vgem_driver = { .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, .gem_prime_import = vgem_prime_import, - .gem_prime_import_sg_table = vgem_prime_import_sg_table, + .gem_prime_import_sg_table = drm_gem_shmem_prime_import_sg_table, .gem_prime_mmap = vgem_prime_mmap,
.name = DRIVER_NAME, diff --git a/drivers/gpu/drm/vgem/vgem_drv.h b/drivers/gpu/drm/vgem/vgem_drv.h index 0ed300317f87..34cf63e6fb3d 100644 --- a/drivers/gpu/drm/vgem/vgem_drv.h +++ b/drivers/gpu/drm/vgem/vgem_drv.h @@ -39,17 +39,6 @@ struct vgem_file { struct mutex fence_mutex; };
-#define to_vgem_bo(x) container_of(x, struct drm_vgem_gem_object, base) -struct drm_vgem_gem_object { - struct drm_gem_object base; - - struct page **pages; - unsigned int pages_pin_count; - struct mutex pages_lock; - - struct sg_table *table; -}; - int vgem_fence_open(struct vgem_file *file); int vgem_fence_attach_ioctl(struct drm_device *dev, void *data,
On Mon, Oct 12, 2020 at 11:51:30AM +0100, Chris Wilson wrote:
On some processors (such as arch/x86), accessing a page via a WC PAT is bypassed if the page is physically tagged in the CPU cache, and the access is serviced by the cache instead -- which leads to incoherency should the physical page itself be accessed using DMA. In order to prevent the false cache sharing of the physical pages, we need to explicitly flush the cache lines of those pages.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk
Hm I'd leave this out for now. dma-api/cache flushing, and especially on x86 is kinda a mess. I'd just land v1 of your patch meanwhile for vgem. -Daniel
drivers/gpu/drm/drm_gem.c | 8 ++++++-- drivers/gpu/drm/drm_gem_shmem_helper.c | 8 +++++++- drivers/gpu/drm/etnaviv/etnaviv_gem.c | 2 +- drivers/gpu/drm/gma500/gtt.c | 2 +- drivers/gpu/drm/msm/msm_gem.c | 2 +- drivers/gpu/drm/omapdrm/omap_gem.c | 2 +- drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 2 +- drivers/gpu/drm/tegra/gem.c | 2 +- drivers/gpu/drm/vgem/vgem_drv.c | 2 +- drivers/gpu/drm/vkms/vkms_gem.c | 2 +- drivers/gpu/drm/xen/xen_drm_front_gem.c | 2 +- include/drm/drm_gem.h | 2 +- 12 files changed, 23 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c index 1da67d34e55d..1948855d69e6 100644 --- a/drivers/gpu/drm/drm_gem.c +++ b/drivers/gpu/drm/drm_gem.c @@ -40,6 +40,7 @@ #include <linux/pagevec.h>
#include <drm/drm.h> +#include <drm/drm_cache.h> #include <drm/drm_device.h> #include <drm/drm_drv.h> #include <drm/drm_file.h> @@ -525,6 +526,7 @@ static void drm_gem_check_release_pagevec(struct pagevec *pvec)
- drm_gem_get_pages - helper to allocate backing pages for a GEM object
- from shmem
- @obj: obj in question
- @clflush: whether to clear any CPU caches associated with the backing store
- This reads the page-array of the shmem-backing storage of the given gem
- object. An array of pages is returned. If a page is not allocated or
@@ -546,14 +548,13 @@ static void drm_gem_check_release_pagevec(struct pagevec *pvec)
- drm_gem_object_init(), but not for those initialized with
- drm_gem_private_object_init() only.
*/ -struct page **drm_gem_get_pages(struct drm_gem_object *obj) +struct page **drm_gem_get_pages(struct drm_gem_object *obj, bool clflush) { struct address_space *mapping; struct page *p, **pages; struct pagevec pvec; int i, npages;
- if (WARN_ON(!obj->filp)) return ERR_PTR(-EINVAL);
@@ -589,6 +590,9 @@ struct page **drm_gem_get_pages(struct drm_gem_object *obj) (page_to_pfn(p) >= 0x00100000UL)); }
- if (clflush)
drm_clflush_pages(pages, npages);- return pages;
fail: diff --git a/drivers/gpu/drm/drm_gem_shmem_helper.c b/drivers/gpu/drm/drm_gem_shmem_helper.c index fb11df7aced5..78a2eb77802b 100644 --- a/drivers/gpu/drm/drm_gem_shmem_helper.c +++ b/drivers/gpu/drm/drm_gem_shmem_helper.c @@ -152,7 +152,13 @@ static int drm_gem_shmem_get_pages_locked(struct drm_gem_shmem_object *shmem) if (shmem->pages_use_count++ > 0) return 0;
- pages = drm_gem_get_pages(obj);
- /*
* On some processors (such as arch/x86), accessing a page via a WC PAT* is bypassed if the page is physically tagged in the CPU cache, and* the access is serviced by the cache instead -- which leads to* incoherency should the physical page itself be accessed using DMA.*/- pages = drm_gem_get_pages(obj, !shmem->map_cached); if (IS_ERR(pages)) { DRM_DEBUG_KMS("Failed to get pages (%ld)\n", PTR_ERR(pages)); shmem->pages_use_count = 0;
diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem.c b/drivers/gpu/drm/etnaviv/etnaviv_gem.c index 67d9a2b9ea6a..d8279ea363b3 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gem.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem.c @@ -58,7 +58,7 @@ static void etnaviv_gem_scatterlist_unmap(struct etnaviv_gem_object *etnaviv_obj static int etnaviv_gem_shmem_get_pages(struct etnaviv_gem_object *etnaviv_obj) { struct drm_device *dev = etnaviv_obj->base.dev;
- struct page **p = drm_gem_get_pages(&etnaviv_obj->base);
struct page **p = drm_gem_get_pages(&etnaviv_obj->base, false);
if (IS_ERR(p)) { dev_dbg(dev->dev, "could not get pages: %ld\n", PTR_ERR(p));
diff --git a/drivers/gpu/drm/gma500/gtt.c b/drivers/gpu/drm/gma500/gtt.c index 9278bcfad1bf..ada56aec7e68 100644 --- a/drivers/gpu/drm/gma500/gtt.c +++ b/drivers/gpu/drm/gma500/gtt.c @@ -197,7 +197,7 @@ static int psb_gtt_attach_pages(struct gtt_range *gt)
WARN_ON(gt->pages);
- pages = drm_gem_get_pages(>->gem);
- pages = drm_gem_get_pages(>->gem, false); if (IS_ERR(pages)) return PTR_ERR(pages);
diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c index c79828d31822..a7a67ef4e27e 100644 --- a/drivers/gpu/drm/msm/msm_gem.c +++ b/drivers/gpu/drm/msm/msm_gem.c @@ -102,7 +102,7 @@ static struct page **get_pages(struct drm_gem_object *obj) int npages = obj->size >> PAGE_SHIFT;
if (use_pages(obj))
p = drm_gem_get_pages(obj);
p = drm_gem_get_pages(obj, false);diff --git a/drivers/gpu/drm/omapdrm/omap_gem.c b/drivers/gpu/drm/omapdrm/omap_gem.c index d8e09792793a..1ff272c4be33 100644 --- a/drivers/gpu/drm/omapdrm/omap_gem.c +++ b/drivers/gpu/drm/omapdrm/omap_gem.c @@ -236,7 +236,7 @@ static int omap_gem_attach_pages(struct drm_gem_object *obj) if (!(omap_obj->flags & OMAP_BO_MEM_SHMEM) || omap_obj->pages) return 0;
- pages = drm_gem_get_pages(obj);
- pages = drm_gem_get_pages(obj, false); if (IS_ERR(pages)) { dev_err(obj->dev->dev, "could not get pages: %ld\n", PTR_ERR(pages)); return PTR_ERR(pages);
diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c index 7d5ebb10323b..003b76c3e623 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c @@ -80,7 +80,7 @@ static int rockchip_gem_get_pages(struct rockchip_gem_object *rk_obj) int ret, i; struct scatterlist *s;
- rk_obj->pages = drm_gem_get_pages(&rk_obj->base);
- rk_obj->pages = drm_gem_get_pages(&rk_obj->base, false); if (IS_ERR(rk_obj->pages)) return PTR_ERR(rk_obj->pages);
diff --git a/drivers/gpu/drm/tegra/gem.c b/drivers/gpu/drm/tegra/gem.c index 26af8daa9a16..f075f005ecbb 100644 --- a/drivers/gpu/drm/tegra/gem.c +++ b/drivers/gpu/drm/tegra/gem.c @@ -289,7 +289,7 @@ static int tegra_bo_get_pages(struct drm_device *drm, struct tegra_bo *bo) { int err;
- bo->pages = drm_gem_get_pages(&bo->gem);
- bo->pages = drm_gem_get_pages(&bo->gem, false); if (IS_ERR(bo->pages)) return PTR_ERR(bo->pages);
diff --git a/drivers/gpu/drm/vgem/vgem_drv.c b/drivers/gpu/drm/vgem/vgem_drv.c index fa54a6d1403d..f38dd590fa45 100644 --- a/drivers/gpu/drm/vgem/vgem_drv.c +++ b/drivers/gpu/drm/vgem/vgem_drv.c @@ -272,7 +272,7 @@ static struct page **vgem_pin_pages(struct drm_vgem_gem_object *bo) if (bo->pages_pin_count++ == 0) { struct page **pages;
pages = drm_gem_get_pages(&bo->base);
pages = drm_gem_get_pages(&bo->base, false);diff --git a/drivers/gpu/drm/vkms/vkms_gem.c b/drivers/gpu/drm/vkms/vkms_gem.c index 19a0e260a4df..1d17b0ef56c7 100644 --- a/drivers/gpu/drm/vkms/vkms_gem.c +++ b/drivers/gpu/drm/vkms/vkms_gem.c @@ -166,7 +166,7 @@ static struct page **_get_pages(struct vkms_gem_object *vkms_obj) struct drm_gem_object *gem_obj = &vkms_obj->gem;
if (!vkms_obj->pages) {
struct page **pages = drm_gem_get_pages(gem_obj);
struct page **pages = drm_gem_get_pages(gem_obj, false);if (IS_ERR(pages)) return pages;
diff --git a/drivers/gpu/drm/xen/xen_drm_front_gem.c b/drivers/gpu/drm/xen/xen_drm_front_gem.c index 4f34ef34ba60..520a15d75eb6 100644 --- a/drivers/gpu/drm/xen/xen_drm_front_gem.c +++ b/drivers/gpu/drm/xen/xen_drm_front_gem.c @@ -132,7 +132,7 @@ static struct xen_gem_object *gem_create(struct drm_device *dev, size_t size) * with the backend */ xen_obj->num_pages = DIV_ROUND_UP(size, PAGE_SIZE);
- xen_obj->pages = drm_gem_get_pages(&xen_obj->base);
- xen_obj->pages = drm_gem_get_pages(&xen_obj->base, false); if (IS_ERR(xen_obj->pages)) { ret = PTR_ERR(xen_obj->pages); xen_obj->pages = NULL;
diff --git a/include/drm/drm_gem.h b/include/drm/drm_gem.h index c38dd35da00b..118f13b1bb29 100644 --- a/include/drm/drm_gem.h +++ b/include/drm/drm_gem.h @@ -395,7 +395,7 @@ void drm_gem_free_mmap_offset(struct drm_gem_object *obj); int drm_gem_create_mmap_offset(struct drm_gem_object *obj); int drm_gem_create_mmap_offset_size(struct drm_gem_object *obj, size_t size);
-struct page **drm_gem_get_pages(struct drm_gem_object *obj); +struct page **drm_gem_get_pages(struct drm_gem_object *obj, bool clflush); void drm_gem_put_pages(struct drm_gem_object *obj, struct page **pages, bool dirty, bool accessed);
-- 2.20.1
dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
dri-devel@lists.freedesktop.org
-
Chris Wilson -
Daniel Vetter