Hello,
I'm sending the fifth version of my patch series to abstract vma handling from the various media drivers. The patches got some review from mm people and testing from device driver guys so unless someone objects, patches will be queued in media tree for the next merge window.
After this patch set drivers have to know much less details about vmas, their types, and locking. Also quite some code is removed from them. As a bonus drivers get automatically VM_FAULT_RETRY handling. The primary motivation for this series is to remove knowledge about mmap_sem locking from as many places a possible so that we can change it with reasonable effort.
The core of the series is the new helper get_vaddr_frames() which is given a virtual address and it fills in PFNs / struct page pointers (depending on VMA type) into the provided array. If PFNs correspond to normal pages it also grabs references to these pages. The difference from get_user_pages() is that this function can also deal with pfnmap, and io mappings which is what the media drivers need.
I have tested the patches with vivid driver so at least vb2 code got some exposure. Conversion of other drivers was just compile-tested (for x86 so e.g. exynos driver which is only for Samsung platform is completely untested).
Honza Changes since v4: * Minor cleanups and fixes pointed out by Mel and Vlasta * Added Acked-by tags
Changes since v3: * Added include <linux/vmalloc.h> into mm/gup.c as it's needed for some archs * Fixed error path for exynos driver
Changes since v2: * Renamed functions and structures as Mel suggested * Other minor changes suggested by Mel * Rebased on top of 4.1-rc2 * Changed functions to get pointer to array of pages / pfns to perform conversion if necessary. This fixes possible issue in the omap I may have introduced in v2 and generally makes the API less errorprone.
Currently vb2 core acquires mmap_sem just around call to __qbuf_userptr(). However since commit f035eb4e976ef5 (videobuf2: fix lockdep warning) it isn't necessary to acquire it so early as we no longer have to drop queue mutex before acquiring mmap_sem. So push acquisition of mmap_sem down into .get_userptr and .put_userptr memops so that the semaphore is acquired for a shorter time and it is clearer what it is needed for.
Signed-off-by: Jan Kara jack@suse.cz --- drivers/media/v4l2-core/videobuf2-core.c | 2 -- drivers/media/v4l2-core/videobuf2-dma-contig.c | 7 +++++++ drivers/media/v4l2-core/videobuf2-dma-sg.c | 6 ++++++ drivers/media/v4l2-core/videobuf2-vmalloc.c | 6 +++++- 4 files changed, 18 insertions(+), 3 deletions(-)
diff --git a/drivers/media/v4l2-core/videobuf2-core.c b/drivers/media/v4l2-core/videobuf2-core.c index 66ada01c796c..20cdbc0900ea 100644 --- a/drivers/media/v4l2-core/videobuf2-core.c +++ b/drivers/media/v4l2-core/videobuf2-core.c @@ -1657,9 +1657,7 @@ static int __buf_prepare(struct vb2_buffer *vb, const struct v4l2_buffer *b) ret = __qbuf_mmap(vb, b); break; case V4L2_MEMORY_USERPTR: - down_read(¤t->mm->mmap_sem); ret = __qbuf_userptr(vb, b); - up_read(¤t->mm->mmap_sem); break; case V4L2_MEMORY_DMABUF: ret = __qbuf_dmabuf(vb, b); diff --git a/drivers/media/v4l2-core/videobuf2-dma-contig.c b/drivers/media/v4l2-core/videobuf2-dma-contig.c index 644dec73d220..620c4aa78881 100644 --- a/drivers/media/v4l2-core/videobuf2-dma-contig.c +++ b/drivers/media/v4l2-core/videobuf2-dma-contig.c @@ -532,7 +532,9 @@ static void vb2_dc_put_userptr(void *buf_priv) sg_free_table(sgt); kfree(sgt); } + down_read(¤t->mm->mmap_sem); vb2_put_vma(buf->vma); + up_read(¤t->mm->mmap_sem); kfree(buf); }
@@ -616,6 +618,7 @@ static void *vb2_dc_get_userptr(void *alloc_ctx, unsigned long vaddr, goto fail_buf; }
+ down_read(¤t->mm->mmap_sem); /* current->mm->mmap_sem is taken by videobuf2 core */ vma = find_vma(current->mm, vaddr); if (!vma) { @@ -642,6 +645,7 @@ static void *vb2_dc_get_userptr(void *alloc_ctx, unsigned long vaddr, if (ret) { unsigned long pfn; if (vb2_dc_get_user_pfn(start, n_pages, vma, &pfn) == 0) { + up_read(¤t->mm->mmap_sem); buf->dma_addr = vb2_dc_pfn_to_dma(buf->dev, pfn); buf->size = size; kfree(pages); @@ -651,6 +655,7 @@ static void *vb2_dc_get_userptr(void *alloc_ctx, unsigned long vaddr, pr_err("failed to get user pages\n"); goto fail_vma; } + up_read(¤t->mm->mmap_sem);
sgt = kzalloc(sizeof(*sgt), GFP_KERNEL); if (!sgt) { @@ -713,10 +718,12 @@ fail_get_user_pages: while (n_pages) put_page(pages[--n_pages]);
+ down_read(¤t->mm->mmap_sem); fail_vma: vb2_put_vma(buf->vma);
fail_pages: + up_read(¤t->mm->mmap_sem); kfree(pages); /* kfree is NULL-proof */
fail_buf: diff --git a/drivers/media/v4l2-core/videobuf2-dma-sg.c b/drivers/media/v4l2-core/videobuf2-dma-sg.c index 45c708e463b9..afd4b514affc 100644 --- a/drivers/media/v4l2-core/videobuf2-dma-sg.c +++ b/drivers/media/v4l2-core/videobuf2-dma-sg.c @@ -263,6 +263,7 @@ static void *vb2_dma_sg_get_userptr(void *alloc_ctx, unsigned long vaddr, if (!buf->pages) goto userptr_fail_alloc_pages;
+ down_read(¤t->mm->mmap_sem); vma = find_vma(current->mm, vaddr); if (!vma) { dprintk(1, "no vma for address %lu\n", vaddr); @@ -301,6 +302,7 @@ static void *vb2_dma_sg_get_userptr(void *alloc_ctx, unsigned long vaddr, 1, /* force */ buf->pages, NULL); + up_read(¤t->mm->mmap_sem);
if (num_pages_from_user != buf->num_pages) goto userptr_fail_get_user_pages; @@ -328,8 +330,10 @@ userptr_fail_get_user_pages: if (!vma_is_io(buf->vma)) while (--num_pages_from_user >= 0) put_page(buf->pages[num_pages_from_user]); + down_read(¤t->mm->mmap_sem); vb2_put_vma(buf->vma); userptr_fail_find_vma: + up_read(¤t->mm->mmap_sem); kfree(buf->pages); userptr_fail_alloc_pages: kfree(buf); @@ -362,7 +366,9 @@ static void vb2_dma_sg_put_userptr(void *buf_priv) put_page(buf->pages[i]); } kfree(buf->pages); + down_read(¤t->mm->mmap_sem); vb2_put_vma(buf->vma); + up_read(¤t->mm->mmap_sem); kfree(buf); }
diff --git a/drivers/media/v4l2-core/videobuf2-vmalloc.c b/drivers/media/v4l2-core/videobuf2-vmalloc.c index 657ab302a5cf..0ba40be21ebd 100644 --- a/drivers/media/v4l2-core/videobuf2-vmalloc.c +++ b/drivers/media/v4l2-core/videobuf2-vmalloc.c @@ -89,7 +89,7 @@ static void *vb2_vmalloc_get_userptr(void *alloc_ctx, unsigned long vaddr, offset = vaddr & ~PAGE_MASK; buf->size = size;
- + down_read(¤t->mm->mmap_sem); vma = find_vma(current->mm, vaddr); if (vma && (vma->vm_flags & VM_PFNMAP) && (vma->vm_pgoff)) { if (vb2_get_contig_userptr(vaddr, size, &vma, &physp)) @@ -121,6 +121,7 @@ static void *vb2_vmalloc_get_userptr(void *alloc_ctx, unsigned long vaddr, if (!buf->vaddr) goto fail_get_user_pages; } + up_read(¤t->mm->mmap_sem);
buf->vaddr += offset; return buf; @@ -133,6 +134,7 @@ fail_get_user_pages: kfree(buf->pages);
fail_pages_array_alloc: + up_read(¤t->mm->mmap_sem); kfree(buf);
return NULL; @@ -144,6 +146,7 @@ static void vb2_vmalloc_put_userptr(void *buf_priv) unsigned long vaddr = (unsigned long)buf->vaddr & PAGE_MASK; unsigned int i;
+ down_read(¤t->mm->mmap_sem); if (buf->pages) { if (vaddr) vm_unmap_ram((void *)vaddr, buf->n_pages); @@ -157,6 +160,7 @@ static void vb2_vmalloc_put_userptr(void *buf_priv) vb2_put_vma(buf->vma); iounmap((__force void __iomem *)buf->vaddr); } + up_read(¤t->mm->mmap_sem); kfree(buf); }
Provide new function get_vaddr_frames(). This function maps virtual addresses from given start and fills given array with page frame numbers of the corresponding pages. If given start belongs to a normal vma, the function grabs reference to each of the pages to pin them in memory. If start belongs to VM_IO | VM_PFNMAP vma, we don't touch page structures. Caller must make sure pfns aren't reused for anything else while he is using them.
This function is created for various drivers to simplify handling of their buffers.
Acked-by: Mel Gorman mgorman@suse.de Acked-by: Vlastimil Babka vbabka@suse.cz Signed-off-by: Jan Kara jack@suse.cz --- include/linux/mm.h | 44 +++++++++++ mm/gup.c | 226 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 270 insertions(+)
diff --git a/include/linux/mm.h b/include/linux/mm.h index 0755b9fd03a7..dcd1f02a78e9 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -20,6 +20,7 @@ #include <linux/shrinker.h> #include <linux/resource.h> #include <linux/page_ext.h> +#include <linux/err.h>
struct mempolicy; struct anon_vma; @@ -1197,6 +1198,49 @@ long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm, int write, int force, struct page **pages); int get_user_pages_fast(unsigned long start, int nr_pages, int write, struct page **pages); + +/* Container for pinned pfns / pages */ +struct frame_vector { + unsigned int nr_allocated; /* Number of frames we have space for */ + unsigned int nr_frames; /* Number of frames stored in ptrs array */ + bool got_ref; /* Did we pin pages by getting page ref? */ + bool is_pfns; /* Does array contain pages or pfns? */ + void *ptrs[0]; /* Array of pinned pfns / pages. Use + * pfns_vector_pages() or pfns_vector_pfns() + * for access */ +}; + +struct frame_vector *frame_vector_create(unsigned int nr_frames); +void frame_vector_destroy(struct frame_vector *vec); +int get_vaddr_frames(unsigned long start, unsigned int nr_pfns, + bool write, bool force, struct frame_vector *vec); +void put_vaddr_frames(struct frame_vector *vec); +int frame_vector_to_pages(struct frame_vector *vec); +void frame_vector_to_pfns(struct frame_vector *vec); + +static inline unsigned int frame_vector_count(struct frame_vector *vec) +{ + return vec->nr_frames; +} + +static inline struct page **frame_vector_pages(struct frame_vector *vec) +{ + if (vec->is_pfns) { + int err = frame_vector_to_pages(vec); + + if (err) + return ERR_PTR(err); + } + return (struct page **)(vec->ptrs); +} + +static inline unsigned long *frame_vector_pfns(struct frame_vector *vec) +{ + if (!vec->is_pfns) + frame_vector_to_pfns(vec); + return (unsigned long *)(vec->ptrs); +} + struct kvec; int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, struct page **pages); diff --git a/mm/gup.c b/mm/gup.c index 6297f6bccfb1..9d7f4fde30cb 100644 --- a/mm/gup.c +++ b/mm/gup.c @@ -8,6 +8,7 @@ #include <linux/rmap.h> #include <linux/swap.h> #include <linux/swapops.h> +#include <linux/vmalloc.h>
#include <linux/sched.h> #include <linux/rwsem.h> @@ -936,6 +937,231 @@ int __mm_populate(unsigned long start, unsigned long len, int ignore_errors) return ret; /* 0 or negative error code */ }
+/* + * get_vaddr_frames() - map virtual addresses to pfns + * @start: starting user address + * @nr_frames: number of pages / pfns from start to map + * @write: whether pages will be written to by the caller + * @force: whether to force write access even if user mapping is + * readonly. See description of the same argument of + get_user_pages(). + * @vec: structure which receives pages / pfns of the addresses mapped. + * It should have space for at least nr_frames entries. + * + * This function maps virtual addresses from @start and fills @vec structure + * with page frame numbers or page pointers to corresponding pages (choice + * depends on the type of the vma underlying the virtual address). If @start + * belongs to a normal vma, the function grabs reference to each of the pages + * to pin them in memory. If @start belongs to VM_IO | VM_PFNMAP vma, we don't + * touch page structures and the caller must make sure pfns aren't reused for + * anything else while he is using them. + * + * The function returns number of pages mapped which may be less than + * @nr_frames. In particular we stop mapping if there are more vmas of + * different type underlying the specified range of virtual addresses. + * When the function isn't able to map a single page, it returns error. + * + * This function takes care of grabbing mmap_sem as necessary. + */ +int get_vaddr_frames(unsigned long start, unsigned int nr_frames, + bool write, bool force, struct frame_vector *vec) +{ + struct mm_struct *mm = current->mm; + struct vm_area_struct *vma; + int ret = 0; + int err; + int locked; + + if (nr_frames == 0) + return 0; + + if (WARN_ON_ONCE(nr_frames > vec->nr_allocated)) + nr_frames = vec->nr_allocated; + + down_read(&mm->mmap_sem); + locked = 1; + vma = find_vma_intersection(mm, start, start + 1); + if (!vma) { + ret = -EFAULT; + goto out; + } + if (!(vma->vm_flags & (VM_IO | VM_PFNMAP))) { + vec->got_ref = true; + vec->is_pfns = false; + ret = get_user_pages_locked(current, mm, start, nr_frames, + write, force, (struct page **)(vec->ptrs), &locked); + goto out; + } + + vec->got_ref = false; + vec->is_pfns = true; + do { + unsigned long *nums = frame_vector_pfns(vec); + + while (ret < nr_frames && start + PAGE_SIZE <= vma->vm_end) { + err = follow_pfn(vma, start, &nums[ret]); + if (err) { + if (ret == 0) + ret = err; + goto out; + } + start += PAGE_SIZE; + ret++; + } + /* + * We stop if we have enough pages or if VMA doesn't completely + * cover the tail page. + */ + if (ret >= nr_frames || start < vma->vm_end) + break; + vma = find_vma_intersection(mm, start, start + 1); + } while (vma && vma->vm_flags & (VM_IO | VM_PFNMAP)); +out: + if (locked) + up_read(&mm->mmap_sem); + if (!ret) + ret = -EFAULT; + if (ret > 0) + vec->nr_frames = ret; + return ret; +} +EXPORT_SYMBOL(get_vaddr_frames); + +/** + * put_vaddr_frames() - drop references to pages if get_vaddr_frames() acquired + * them + * @vec: frame vector to put + * + * Drop references to pages if get_vaddr_frames() acquired them. We also + * invalidate the frame vector so that it is prepared for the next call into + * get_vaddr_frames(). + */ +void put_vaddr_frames(struct frame_vector *vec) +{ + int i; + struct page **pages; + + if (!vec->got_ref) + goto out; + pages = frame_vector_pages(vec); + /* + * frame_vector_pages() might needed to do a conversion when + * get_vaddr_frames() got pages but vec was later converted to pfns. + * But it shouldn't really fail to convert pfns back... + */ + if (WARN_ON(IS_ERR(pages))) + goto out; + for (i = 0; i < vec->nr_frames; i++) + put_page(pages[i]); + vec->got_ref = false; +out: + vec->nr_frames = 0; +} +EXPORT_SYMBOL(put_vaddr_frames); + +/** + * frame_vector_to_pages - convert frame vector to contain page pointers + * @vec: frame vector to convert + * + * Convert @vec to contain array of page pointers. If the conversion is + * successful, return 0. Otherwise return an error. Note that we do not grab + * page references for the page structures. + */ +int frame_vector_to_pages(struct frame_vector *vec) +{ + int i; + unsigned long *nums; + struct page **pages; + + if (!vec->is_pfns) + return 0; + nums = frame_vector_pfns(vec); + for (i = 0; i < vec->nr_frames; i++) + if (!pfn_valid(nums[i])) + return -EINVAL; + pages = (struct page **)nums; + for (i = 0; i < vec->nr_frames; i++) + pages[i] = pfn_to_page(nums[i]); + vec->is_pfns = false; + return 0; +} +EXPORT_SYMBOL(frame_vector_to_pages); + +/** + * frame_vector_to_pfns - convert frame vector to contain pfns + * @vec: frame vector to convert + * + * Convert @vec to contain array of pfns. + */ +void frame_vector_to_pfns(struct frame_vector *vec) +{ + int i; + unsigned long *nums; + struct page **pages; + + if (vec->is_pfns) + return; + pages = (struct page **)(vec->ptrs); + nums = (unsigned long *)pages; + for (i = 0; i < vec->nr_frames; i++) + nums[i] = page_to_pfn(pages[i]); + vec->is_pfns = true; +} +EXPORT_SYMBOL(frame_vector_to_pfns); + +/** + * frame_vector_create() - allocate & initialize structure for pinned pfns + * @nr_frames: number of pfns slots we should reserve + * + * Allocate and initialize struct pinned_pfns to be able to hold @nr_pfns + * pfns. + */ +struct frame_vector *frame_vector_create(unsigned int nr_frames) +{ + struct frame_vector *vec; + int size = sizeof(struct frame_vector) + sizeof(void *) * nr_frames; + + if (WARN_ON_ONCE(nr_frames == 0)) + return NULL; + /* + * This is absurdly high. It's here just to avoid strange effects when + * arithmetics overflows. + */ + if (WARN_ON_ONCE(nr_frames > INT_MAX / sizeof(void *) / 2)) + return NULL; + /* + * Avoid higher order allocations, use vmalloc instead. It should + * be rare anyway. + */ + if (size <= PAGE_SIZE) + vec = kmalloc(size, GFP_KERNEL); + else + vec = vmalloc(size); + if (!vec) + return NULL; + vec->nr_allocated = nr_frames; + vec->nr_frames = 0; + return vec; +} +EXPORT_SYMBOL(frame_vector_create); + +/** + * frame_vector_destroy() - free memory allocated to carry frame vector + * @vec: Frame vector to free + * + * Free structure allocated by frame_vector_create() to carry frames. + */ +void frame_vector_destroy(struct frame_vector *vec) +{ + /* Make sure put_vaddr_frames() got called properly... */ + VM_BUG_ON(vec->nr_frames > 0); + if (!is_vmalloc_addr(vec)) + kfree(vec); + else + vfree(vec); +} +EXPORT_SYMBOL(frame_vector_destroy); + /** * get_dump_page() - pin user page in memory while writing it to core dump * @addr: user address
On Wed, 13 May 2015 15:08:08 +0200 Jan Kara jack@suse.cz wrote:
Provide new function get_vaddr_frames(). This function maps virtual addresses from given start and fills given array with page frame numbers of the corresponding pages. If given start belongs to a normal vma, the function grabs reference to each of the pages to pin them in memory. If start belongs to VM_IO | VM_PFNMAP vma, we don't touch page structures. Caller must make sure pfns aren't reused for anything else while he is using them.
This function is created for various drivers to simplify handling of their buffers.
Acked-by: Mel Gorman mgorman@suse.de Acked-by: Vlastimil Babka vbabka@suse.cz Signed-off-by: Jan Kara jack@suse.cz
include/linux/mm.h | 44 +++++++++++ mm/gup.c | 226 +++++++++++++++++++++++++++++++++++++++++++++++++++++
That's a lump of new code which many kernels won't be needing. Can we put all this in a new .c file and select it within drivers/media Kconfig?
On Thu 28-05-15 16:24:02, Andrew Morton wrote:
On Wed, 13 May 2015 15:08:08 +0200 Jan Kara jack@suse.cz wrote:
Provide new function get_vaddr_frames(). This function maps virtual addresses from given start and fills given array with page frame numbers of the corresponding pages. If given start belongs to a normal vma, the function grabs reference to each of the pages to pin them in memory. If start belongs to VM_IO | VM_PFNMAP vma, we don't touch page structures. Caller must make sure pfns aren't reused for anything else while he is using them.
This function is created for various drivers to simplify handling of their buffers.
Acked-by: Mel Gorman mgorman@suse.de Acked-by: Vlastimil Babka vbabka@suse.cz Signed-off-by: Jan Kara jack@suse.cz
include/linux/mm.h | 44 +++++++++++ mm/gup.c | 226 +++++++++++++++++++++++++++++++++++++++++++++++++++++
That's a lump of new code which many kernels won't be needing. Can we put all this in a new .c file and select it within drivers/media Kconfig?
Yeah, makes sense. I'll write a patch. Hans, is it OK with you if I just create a patch on top of the series you have in your tree?
Honza
On 06/01/2015 02:40 PM, Jan Kara wrote:
On Thu 28-05-15 16:24:02, Andrew Morton wrote:
On Wed, 13 May 2015 15:08:08 +0200 Jan Kara jack@suse.cz wrote:
Provide new function get_vaddr_frames(). This function maps virtual addresses from given start and fills given array with page frame numbers of the corresponding pages. If given start belongs to a normal vma, the function grabs reference to each of the pages to pin them in memory. If start belongs to VM_IO | VM_PFNMAP vma, we don't touch page structures. Caller must make sure pfns aren't reused for anything else while he is using them.
This function is created for various drivers to simplify handling of their buffers.
Acked-by: Mel Gorman mgorman@suse.de Acked-by: Vlastimil Babka vbabka@suse.cz Signed-off-by: Jan Kara jack@suse.cz
include/linux/mm.h | 44 +++++++++++ mm/gup.c | 226 +++++++++++++++++++++++++++++++++++++++++++++++++++++
That's a lump of new code which many kernels won't be needing. Can we put all this in a new .c file and select it within drivers/media Kconfig?
Yeah, makes sense. I'll write a patch. Hans, is it OK with you if I just create a patch on top of the series you have in your tree?
No problem.
Regards,
Hans
On Thu 28-05-15 16:24:02, Andrew Morton wrote:
On Wed, 13 May 2015 15:08:08 +0200 Jan Kara jack@suse.cz wrote:
Provide new function get_vaddr_frames(). This function maps virtual addresses from given start and fills given array with page frame numbers of the corresponding pages. If given start belongs to a normal vma, the function grabs reference to each of the pages to pin them in memory. If start belongs to VM_IO | VM_PFNMAP vma, we don't touch page structures. Caller must make sure pfns aren't reused for anything else while he is using them.
This function is created for various drivers to simplify handling of their buffers.
Acked-by: Mel Gorman mgorman@suse.de Acked-by: Vlastimil Babka vbabka@suse.cz Signed-off-by: Jan Kara jack@suse.cz
include/linux/mm.h | 44 +++++++++++ mm/gup.c | 226 +++++++++++++++++++++++++++++++++++++++++++++++++++++
That's a lump of new code which many kernels won't be needing. Can we put all this in a new .c file and select it within drivers/media Kconfig?
So the attached patch should do what you had in mind. OK?
Honza
On Tue, 2 Jun 2015 17:23:00 +0200 Jan Kara jack@suse.cz wrote:
That's a lump of new code which many kernels won't be needing. Can we put all this in a new .c file and select it within drivers/media Kconfig?
So the attached patch should do what you had in mind. OK?
lgtm.
drivers/gpu/drm/exynos/Kconfig | 1 + drivers/media/platform/omap/Kconfig | 1 + drivers/media/v4l2-core/Kconfig | 1 + mm/Kconfig | 3 + mm/Makefile | 1 + mm/frame-vec.c | 233 ++++++++++++++++++++++++++++++++++++
But frame_vector.c would be a more pleasing name. For `struct frame_vector'.
On Tue 02-06-15 15:29:12, Andrew Morton wrote:
On Tue, 2 Jun 2015 17:23:00 +0200 Jan Kara jack@suse.cz wrote:
That's a lump of new code which many kernels won't be needing. Can we put all this in a new .c file and select it within drivers/media Kconfig?
So the attached patch should do what you had in mind. OK?
lgtm.
drivers/gpu/drm/exynos/Kconfig | 1 + drivers/media/platform/omap/Kconfig | 1 + drivers/media/v4l2-core/Kconfig | 1 + mm/Kconfig | 3 + mm/Makefile | 1 + mm/frame-vec.c | 233 ++++++++++++++++++++++++++++++++++++
But frame_vector.c would be a more pleasing name. For `struct frame_vector'.
OK, makes sense. Updated patch attached.
Honza
Convert omap_vout_uservirt_to_phys() to use get_vaddr_pfns() instead of hand made mapping of virtual address to physical address. Also the function leaked page reference from get_user_pages() so fix that by properly release the reference when omap_vout_buffer_release() is called.
Signed-off-by: Jan Kara jack@suse.cz --- drivers/media/platform/omap/omap_vout.c | 67 +++++++++++++++------------------ 1 file changed, 31 insertions(+), 36 deletions(-)
diff --git a/drivers/media/platform/omap/omap_vout.c b/drivers/media/platform/omap/omap_vout.c index 17b189a81ec5..0e4b3cfacc5d 100644 --- a/drivers/media/platform/omap/omap_vout.c +++ b/drivers/media/platform/omap/omap_vout.c @@ -195,46 +195,34 @@ static int omap_vout_try_format(struct v4l2_pix_format *pix) }
/* - * omap_vout_uservirt_to_phys: This inline function is used to convert user - * space virtual address to physical address. + * omap_vout_get_userptr: Convert user space virtual address to physical + * address. */ -static unsigned long omap_vout_uservirt_to_phys(unsigned long virtp) +static int omap_vout_get_userptr(struct videobuf_buffer *vb, u32 virtp, + u32 *physp) { - unsigned long physp = 0; - struct vm_area_struct *vma; - struct mm_struct *mm = current->mm; + struct frame_vector *vec; + int ret;
/* For kernel direct-mapped memory, take the easy way */ - if (virtp >= PAGE_OFFSET) - return virt_to_phys((void *) virtp); - - down_read(¤t->mm->mmap_sem); - vma = find_vma(mm, virtp); - if (vma && (vma->vm_flags & VM_IO) && vma->vm_pgoff) { - /* this will catch, kernel-allocated, mmaped-to-usermode - addresses */ - physp = (vma->vm_pgoff << PAGE_SHIFT) + (virtp - vma->vm_start); - up_read(¤t->mm->mmap_sem); - } else { - /* otherwise, use get_user_pages() for general userland pages */ - int res, nr_pages = 1; - struct page *pages; + if (virtp >= PAGE_OFFSET) { + *physp = virt_to_phys((void *)virtp); + return 0; + }
- res = get_user_pages(current, current->mm, virtp, nr_pages, 1, - 0, &pages, NULL); - up_read(¤t->mm->mmap_sem); + vec = frame_vector_create(1); + if (!vec) + return -ENOMEM;
- if (res == nr_pages) { - physp = __pa(page_address(&pages[0]) + - (virtp & ~PAGE_MASK)); - } else { - printk(KERN_WARNING VOUT_NAME - "get_user_pages failed\n"); - return 0; - } + ret = get_vaddr_frames(virtp, 1, true, false, vec); + if (ret != 1) { + frame_vector_destroy(vec); + return -EINVAL; } + *physp = __pfn_to_phys(frame_vector_pfns(vec)[0]); + vb->priv = vec;
- return physp; + return 0; }
/* @@ -788,11 +776,15 @@ static int omap_vout_buffer_prepare(struct videobuf_queue *q, * address of the buffer */ if (V4L2_MEMORY_USERPTR == vb->memory) { + int ret; + if (0 == vb->baddr) return -EINVAL; /* Physical address */ - vout->queued_buf_addr[vb->i] = (u8 *) - omap_vout_uservirt_to_phys(vb->baddr); + ret = omap_vout_get_userptr(vb, vb->baddr, + (u32 *)&vout->queued_buf_addr[vb->i]); + if (ret < 0) + return ret; } else { unsigned long addr, dma_addr; unsigned long size; @@ -841,9 +833,12 @@ static void omap_vout_buffer_release(struct videobuf_queue *q, struct omap_vout_device *vout = q->priv_data;
vb->state = VIDEOBUF_NEEDS_INIT; + if (vb->memory == V4L2_MEMORY_USERPTR && vb->priv) { + struct frame_vector *vec = vb->priv;
- if (V4L2_MEMORY_MMAP != vout->memory) - return; + put_vaddr_frames(vec); + frame_vector_destroy(vec); + } }
/*
Provide simple helper functions to map virtual address range into an array of pfns / pages.
Acked-by: Marek Szyprowski m.szyprowski@samsung.com Tested-by: Marek Szyprowski m.szyprowski@samsung.com Signed-off-by: Jan Kara jack@suse.cz --- drivers/media/v4l2-core/videobuf2-memops.c | 58 ++++++++++++++++++++++++++++++ include/media/videobuf2-memops.h | 5 +++ 2 files changed, 63 insertions(+)
diff --git a/drivers/media/v4l2-core/videobuf2-memops.c b/drivers/media/v4l2-core/videobuf2-memops.c index 81c1ad8b2cf1..0ec186d41b9b 100644 --- a/drivers/media/v4l2-core/videobuf2-memops.c +++ b/drivers/media/v4l2-core/videobuf2-memops.c @@ -137,6 +137,64 @@ int vb2_get_contig_userptr(unsigned long vaddr, unsigned long size, EXPORT_SYMBOL_GPL(vb2_get_contig_userptr);
/** + * vb2_create_framevec() - map virtual addresses to pfns + * @start: Virtual user address where we start mapping + * @length: Length of a range to map + * @write: Should we map for writing into the area + * + * This function allocates and fills in a vector with pfns corresponding to + * virtual address range passed in arguments. If pfns have corresponding pages, + * page references are also grabbed to pin pages in memory. The function + * returns pointer to the vector on success and error pointer in case of + * failure. Returned vector needs to be freed via vb2_destroy_pfnvec(). + */ +struct frame_vector *vb2_create_framevec(unsigned long start, + unsigned long length, + bool write) +{ + int ret; + unsigned long first, last; + unsigned long nr; + struct frame_vector *vec; + + first = start >> PAGE_SHIFT; + last = (start + length - 1) >> PAGE_SHIFT; + nr = last - first + 1; + vec = frame_vector_create(nr); + if (!vec) + return ERR_PTR(-ENOMEM); + ret = get_vaddr_frames(start, nr, write, 1, vec); + if (ret < 0) + goto out_destroy; + /* We accept only complete set of PFNs */ + if (ret != nr) { + ret = -EFAULT; + goto out_release; + } + return vec; +out_release: + put_vaddr_frames(vec); +out_destroy: + frame_vector_destroy(vec); + return ERR_PTR(ret); +} +EXPORT_SYMBOL(vb2_create_framevec); + +/** + * vb2_destroy_framevec() - release vector of mapped pfns + * @vec: vector of pfns / pages to release + * + * This releases references to all pages in the vector @vec (if corresponding + * pfns are backed by pages) and frees the passed vector. + */ +void vb2_destroy_framevec(struct frame_vector *vec) +{ + put_vaddr_frames(vec); + frame_vector_destroy(vec); +} +EXPORT_SYMBOL(vb2_destroy_framevec); + +/** * vb2_common_vm_open() - increase refcount of the vma * @vma: virtual memory region for the mapping * diff --git a/include/media/videobuf2-memops.h b/include/media/videobuf2-memops.h index f05444ca8c0c..2f0564ff5f31 100644 --- a/include/media/videobuf2-memops.h +++ b/include/media/videobuf2-memops.h @@ -15,6 +15,7 @@ #define _MEDIA_VIDEOBUF2_MEMOPS_H
#include <media/videobuf2-core.h> +#include <linux/mm.h>
/** * vb2_vmarea_handler - common vma refcount tracking handler @@ -36,5 +37,9 @@ int vb2_get_contig_userptr(unsigned long vaddr, unsigned long size, struct vm_area_struct *vb2_get_vma(struct vm_area_struct *vma); void vb2_put_vma(struct vm_area_struct *vma);
+struct frame_vector *vb2_create_framevec(unsigned long start, + unsigned long length, + bool write); +void vb2_destroy_framevec(struct frame_vector *vec);
#endif
Acked-by: Marek Szyprowski m.szyprowski@samsung.com Tested-by: Marek Szyprowski m.szyprowski@samsung.com Signed-off-by: Jan Kara jack@suse.cz --- drivers/media/v4l2-core/videobuf2-dma-sg.c | 97 +++++------------------------- 1 file changed, 15 insertions(+), 82 deletions(-)
diff --git a/drivers/media/v4l2-core/videobuf2-dma-sg.c b/drivers/media/v4l2-core/videobuf2-dma-sg.c index afd4b514affc..4ee1b3fbfe2a 100644 --- a/drivers/media/v4l2-core/videobuf2-dma-sg.c +++ b/drivers/media/v4l2-core/videobuf2-dma-sg.c @@ -38,6 +38,7 @@ struct vb2_dma_sg_buf { struct device *dev; void *vaddr; struct page **pages; + struct frame_vector *vec; int offset; enum dma_data_direction dma_dir; struct sg_table sg_table; @@ -51,7 +52,6 @@ struct vb2_dma_sg_buf { unsigned int num_pages; atomic_t refcount; struct vb2_vmarea_handler handler; - struct vm_area_struct *vma;
struct dma_buf_attachment *db_attach; }; @@ -224,25 +224,17 @@ static void vb2_dma_sg_finish(void *buf_priv) dma_sync_sg_for_cpu(buf->dev, sgt->sgl, sgt->nents, buf->dma_dir); }
-static inline int vma_is_io(struct vm_area_struct *vma) -{ - return !!(vma->vm_flags & (VM_IO | VM_PFNMAP)); -} - static void *vb2_dma_sg_get_userptr(void *alloc_ctx, unsigned long vaddr, unsigned long size, enum dma_data_direction dma_dir) { struct vb2_dma_sg_conf *conf = alloc_ctx; struct vb2_dma_sg_buf *buf; - unsigned long first, last; - int num_pages_from_user; - struct vm_area_struct *vma; struct sg_table *sgt; DEFINE_DMA_ATTRS(attrs); + struct frame_vector *vec;
dma_set_attr(DMA_ATTR_SKIP_CPU_SYNC, &attrs); - buf = kzalloc(sizeof *buf, GFP_KERNEL); if (!buf) return NULL; @@ -253,63 +245,19 @@ static void *vb2_dma_sg_get_userptr(void *alloc_ctx, unsigned long vaddr, buf->offset = vaddr & ~PAGE_MASK; buf->size = size; buf->dma_sgt = &buf->sg_table; + vec = vb2_create_framevec(vaddr, size, buf->dma_dir == DMA_FROM_DEVICE); + if (IS_ERR(vec)) + goto userptr_fail_pfnvec; + buf->vec = vec;
- first = (vaddr & PAGE_MASK) >> PAGE_SHIFT; - last = ((vaddr + size - 1) & PAGE_MASK) >> PAGE_SHIFT; - buf->num_pages = last - first + 1; - - buf->pages = kzalloc(buf->num_pages * sizeof(struct page *), - GFP_KERNEL); - if (!buf->pages) - goto userptr_fail_alloc_pages; - - down_read(¤t->mm->mmap_sem); - vma = find_vma(current->mm, vaddr); - if (!vma) { - dprintk(1, "no vma for address %lu\n", vaddr); - goto userptr_fail_find_vma; - } - - if (vma->vm_end < vaddr + size) { - dprintk(1, "vma at %lu is too small for %lu bytes\n", - vaddr, size); - goto userptr_fail_find_vma; - } - - buf->vma = vb2_get_vma(vma); - if (!buf->vma) { - dprintk(1, "failed to copy vma\n"); - goto userptr_fail_find_vma; - } - - if (vma_is_io(buf->vma)) { - for (num_pages_from_user = 0; - num_pages_from_user < buf->num_pages; - ++num_pages_from_user, vaddr += PAGE_SIZE) { - unsigned long pfn; - - if (follow_pfn(vma, vaddr, &pfn)) { - dprintk(1, "no page for address %lu\n", vaddr); - break; - } - buf->pages[num_pages_from_user] = pfn_to_page(pfn); - } - } else - num_pages_from_user = get_user_pages(current, current->mm, - vaddr & PAGE_MASK, - buf->num_pages, - buf->dma_dir == DMA_FROM_DEVICE, - 1, /* force */ - buf->pages, - NULL); - up_read(¤t->mm->mmap_sem); - - if (num_pages_from_user != buf->num_pages) - goto userptr_fail_get_user_pages; + buf->pages = frame_vector_pages(vec); + if (IS_ERR(buf->pages)) + goto userptr_fail_sgtable; + buf->num_pages = frame_vector_count(vec);
if (sg_alloc_table_from_pages(buf->dma_sgt, buf->pages, buf->num_pages, buf->offset, size, 0)) - goto userptr_fail_alloc_table_from_pages; + goto userptr_fail_sgtable;
sgt = &buf->sg_table; /* @@ -323,19 +271,9 @@ static void *vb2_dma_sg_get_userptr(void *alloc_ctx, unsigned long vaddr,
userptr_fail_map: sg_free_table(&buf->sg_table); -userptr_fail_alloc_table_from_pages: -userptr_fail_get_user_pages: - dprintk(1, "get_user_pages requested/got: %d/%d]\n", - buf->num_pages, num_pages_from_user); - if (!vma_is_io(buf->vma)) - while (--num_pages_from_user >= 0) - put_page(buf->pages[num_pages_from_user]); - down_read(¤t->mm->mmap_sem); - vb2_put_vma(buf->vma); -userptr_fail_find_vma: - up_read(¤t->mm->mmap_sem); - kfree(buf->pages); -userptr_fail_alloc_pages: +userptr_fail_sgtable: + vb2_destroy_framevec(vec); +userptr_fail_pfnvec: kfree(buf); return NULL; } @@ -362,13 +300,8 @@ static void vb2_dma_sg_put_userptr(void *buf_priv) while (--i >= 0) { if (buf->dma_dir == DMA_FROM_DEVICE) set_page_dirty_lock(buf->pages[i]); - if (!vma_is_io(buf->vma)) - put_page(buf->pages[i]); } - kfree(buf->pages); - down_read(¤t->mm->mmap_sem); - vb2_put_vma(buf->vma); - up_read(¤t->mm->mmap_sem); + vb2_destroy_framevec(buf->vec); kfree(buf); }
Convert vb2_vmalloc_get_userptr() to use frame vector infrastructure. When we are doing that there's no need to allocate page array and some code can be simplified.
Acked-by: Marek Szyprowski m.szyprowski@samsung.com Tested-by: Marek Szyprowski m.szyprowski@samsung.com Signed-off-by: Jan Kara jack@suse.cz --- drivers/media/v4l2-core/videobuf2-vmalloc.c | 94 +++++++++++------------------ 1 file changed, 36 insertions(+), 58 deletions(-)
diff --git a/drivers/media/v4l2-core/videobuf2-vmalloc.c b/drivers/media/v4l2-core/videobuf2-vmalloc.c index 0ba40be21ebd..d2ce81fa2cdf 100644 --- a/drivers/media/v4l2-core/videobuf2-vmalloc.c +++ b/drivers/media/v4l2-core/videobuf2-vmalloc.c @@ -23,11 +23,9 @@
struct vb2_vmalloc_buf { void *vaddr; - struct page **pages; - struct vm_area_struct *vma; + struct frame_vector *vec; enum dma_data_direction dma_dir; unsigned long size; - unsigned int n_pages; atomic_t refcount; struct vb2_vmarea_handler handler; struct dma_buf *dbuf; @@ -76,10 +74,8 @@ static void *vb2_vmalloc_get_userptr(void *alloc_ctx, unsigned long vaddr, enum dma_data_direction dma_dir) { struct vb2_vmalloc_buf *buf; - unsigned long first, last; - int n_pages, offset; - struct vm_area_struct *vma; - dma_addr_t physp; + struct frame_vector *vec; + int n_pages, offset, i;
buf = kzalloc(sizeof(*buf), GFP_KERNEL); if (!buf) @@ -88,53 +84,36 @@ static void *vb2_vmalloc_get_userptr(void *alloc_ctx, unsigned long vaddr, buf->dma_dir = dma_dir; offset = vaddr & ~PAGE_MASK; buf->size = size; - - down_read(¤t->mm->mmap_sem); - vma = find_vma(current->mm, vaddr); - if (vma && (vma->vm_flags & VM_PFNMAP) && (vma->vm_pgoff)) { - if (vb2_get_contig_userptr(vaddr, size, &vma, &physp)) - goto fail_pages_array_alloc; - buf->vma = vma; - buf->vaddr = (__force void *)ioremap_nocache(physp, size); - if (!buf->vaddr) - goto fail_pages_array_alloc; + vec = vb2_create_framevec(vaddr, size, dma_dir == DMA_FROM_DEVICE); + if (IS_ERR(vec)) + goto fail_pfnvec_create; + buf->vec = vec; + n_pages = frame_vector_count(vec); + if (frame_vector_to_pages(vec) < 0) { + unsigned long *nums = frame_vector_pfns(vec); + + /* + * We cannot get page pointers for these pfns. Check memory is + * physically contiguous and use direct mapping. + */ + for (i = 1; i < n_pages; i++) + if (nums[i-1] + 1 != nums[i]) + goto fail_map; + buf->vaddr = (__force void *) + ioremap_nocache(nums[0] << PAGE_SHIFT, size); } else { - first = vaddr >> PAGE_SHIFT; - last = (vaddr + size - 1) >> PAGE_SHIFT; - buf->n_pages = last - first + 1; - buf->pages = kzalloc(buf->n_pages * sizeof(struct page *), - GFP_KERNEL); - if (!buf->pages) - goto fail_pages_array_alloc; - - /* current->mm->mmap_sem is taken by videobuf2 core */ - n_pages = get_user_pages(current, current->mm, - vaddr & PAGE_MASK, buf->n_pages, - dma_dir == DMA_FROM_DEVICE, - 1, /* force */ - buf->pages, NULL); - if (n_pages != buf->n_pages) - goto fail_get_user_pages; - - buf->vaddr = vm_map_ram(buf->pages, buf->n_pages, -1, + buf->vaddr = vm_map_ram(frame_vector_pages(vec), n_pages, -1, PAGE_KERNEL); - if (!buf->vaddr) - goto fail_get_user_pages; } - up_read(¤t->mm->mmap_sem);
+ if (!buf->vaddr) + goto fail_map; buf->vaddr += offset; return buf;
-fail_get_user_pages: - pr_debug("get_user_pages requested/got: %d/%d]\n", n_pages, - buf->n_pages); - while (--n_pages >= 0) - put_page(buf->pages[n_pages]); - kfree(buf->pages); - -fail_pages_array_alloc: - up_read(¤t->mm->mmap_sem); +fail_map: + vb2_destroy_framevec(vec); +fail_pfnvec_create: kfree(buf);
return NULL; @@ -145,22 +124,21 @@ static void vb2_vmalloc_put_userptr(void *buf_priv) struct vb2_vmalloc_buf *buf = buf_priv; unsigned long vaddr = (unsigned long)buf->vaddr & PAGE_MASK; unsigned int i; + struct page **pages; + unsigned int n_pages;
- down_read(¤t->mm->mmap_sem); - if (buf->pages) { + if (!buf->vec->is_pfns) { + n_pages = frame_vector_count(buf->vec); + pages = frame_vector_pages(buf->vec); if (vaddr) - vm_unmap_ram((void *)vaddr, buf->n_pages); - for (i = 0; i < buf->n_pages; ++i) { - if (buf->dma_dir == DMA_FROM_DEVICE) - set_page_dirty_lock(buf->pages[i]); - put_page(buf->pages[i]); - } - kfree(buf->pages); + vm_unmap_ram((void *)vaddr, n_pages); + if (buf->dma_dir == DMA_FROM_DEVICE) + for (i = 0; i < n_pages; i++) + set_page_dirty_lock(pages[i]); } else { - vb2_put_vma(buf->vma); iounmap((__force void __iomem *)buf->vaddr); } - up_read(¤t->mm->mmap_sem); + vb2_destroy_framevec(buf->vec); kfree(buf); }
Convert vb2_dc_get_userptr() to use frame vector infrastructure. When we are doing that there's no need to allocate page array and some code can be simplified.
Acked-by: Marek Szyprowski m.szyprowski@samsung.com Tested-by: Marek Szyprowski m.szyprowski@samsung.com Signed-off-by: Jan Kara jack@suse.cz --- drivers/media/v4l2-core/videobuf2-dma-contig.c | 214 ++++--------------------- 1 file changed, 34 insertions(+), 180 deletions(-)
diff --git a/drivers/media/v4l2-core/videobuf2-dma-contig.c b/drivers/media/v4l2-core/videobuf2-dma-contig.c index 620c4aa78881..e6cea452302b 100644 --- a/drivers/media/v4l2-core/videobuf2-dma-contig.c +++ b/drivers/media/v4l2-core/videobuf2-dma-contig.c @@ -32,15 +32,13 @@ struct vb2_dc_buf { dma_addr_t dma_addr; enum dma_data_direction dma_dir; struct sg_table *dma_sgt; + struct frame_vector *vec;
/* MMAP related */ struct vb2_vmarea_handler handler; atomic_t refcount; struct sg_table *sgt_base;
- /* USERPTR related */ - struct vm_area_struct *vma; - /* DMABUF related */ struct dma_buf_attachment *db_attach; }; @@ -49,24 +47,6 @@ struct vb2_dc_buf { /* scatterlist table functions */ /*********************************************/
- -static void vb2_dc_sgt_foreach_page(struct sg_table *sgt, - void (*cb)(struct page *pg)) -{ - struct scatterlist *s; - unsigned int i; - - for_each_sg(sgt->sgl, s, sgt->orig_nents, i) { - struct page *page = sg_page(s); - unsigned int n_pages = PAGE_ALIGN(s->offset + s->length) - >> PAGE_SHIFT; - unsigned int j; - - for (j = 0; j < n_pages; ++j, ++page) - cb(page); - } -} - static unsigned long vb2_dc_get_contiguous_size(struct sg_table *sgt) { struct scatterlist *s; @@ -429,92 +409,12 @@ static struct dma_buf *vb2_dc_get_dmabuf(void *buf_priv, unsigned long flags) /* callbacks for USERPTR buffers */ /*********************************************/
-static inline int vma_is_io(struct vm_area_struct *vma) -{ - return !!(vma->vm_flags & (VM_IO | VM_PFNMAP)); -} - -static int vb2_dc_get_user_pfn(unsigned long start, int n_pages, - struct vm_area_struct *vma, unsigned long *res) -{ - unsigned long pfn, start_pfn, prev_pfn; - unsigned int i; - int ret; - - if (!vma_is_io(vma)) - return -EFAULT; - - ret = follow_pfn(vma, start, &pfn); - if (ret) - return ret; - - start_pfn = pfn; - start += PAGE_SIZE; - - for (i = 1; i < n_pages; ++i, start += PAGE_SIZE) { - prev_pfn = pfn; - ret = follow_pfn(vma, start, &pfn); - - if (ret) { - pr_err("no page for address %lu\n", start); - return ret; - } - if (pfn != prev_pfn + 1) - return -EINVAL; - } - - *res = start_pfn; - return 0; -} - -static int vb2_dc_get_user_pages(unsigned long start, struct page **pages, - int n_pages, struct vm_area_struct *vma, - enum dma_data_direction dma_dir) -{ - if (vma_is_io(vma)) { - unsigned int i; - - for (i = 0; i < n_pages; ++i, start += PAGE_SIZE) { - unsigned long pfn; - int ret = follow_pfn(vma, start, &pfn); - - if (!pfn_valid(pfn)) - return -EINVAL; - - if (ret) { - pr_err("no page for address %lu\n", start); - return ret; - } - pages[i] = pfn_to_page(pfn); - } - } else { - int n; - - n = get_user_pages(current, current->mm, start & PAGE_MASK, - n_pages, dma_dir == DMA_FROM_DEVICE, 1, pages, NULL); - /* negative error means that no page was pinned */ - n = max(n, 0); - if (n != n_pages) { - pr_err("got only %d of %d user pages\n", n, n_pages); - while (n) - put_page(pages[--n]); - return -EFAULT; - } - } - - return 0; -} - -static void vb2_dc_put_dirty_page(struct page *page) -{ - set_page_dirty_lock(page); - put_page(page); -} - static void vb2_dc_put_userptr(void *buf_priv) { struct vb2_dc_buf *buf = buf_priv; struct sg_table *sgt = buf->dma_sgt; + int i; + struct page **pages;
if (sgt) { DEFINE_DMA_ATTRS(attrs); @@ -526,15 +426,15 @@ static void vb2_dc_put_userptr(void *buf_priv) */ dma_unmap_sg_attrs(buf->dev, sgt->sgl, sgt->orig_nents, buf->dma_dir, &attrs); - if (!vma_is_io(buf->vma)) - vb2_dc_sgt_foreach_page(sgt, vb2_dc_put_dirty_page); - + pages = frame_vector_pages(buf->vec); + /* sgt should exist only if vector contains pages... */ + BUG_ON(IS_ERR(pages)); + for (i = 0; i < frame_vector_count(buf->vec); i++) + set_page_dirty_lock(pages[i]); sg_free_table(sgt); kfree(sgt); } - down_read(¤t->mm->mmap_sem); - vb2_put_vma(buf->vma); - up_read(¤t->mm->mmap_sem); + vb2_destroy_framevec(buf->vec); kfree(buf); }
@@ -574,13 +474,10 @@ static void *vb2_dc_get_userptr(void *alloc_ctx, unsigned long vaddr, { struct vb2_dc_conf *conf = alloc_ctx; struct vb2_dc_buf *buf; - unsigned long start; - unsigned long end; + struct frame_vector *vec; unsigned long offset; - struct page **pages; - int n_pages; + int n_pages, i; int ret = 0; - struct vm_area_struct *vma; struct sg_table *sgt; unsigned long contig_size; unsigned long dma_align = dma_get_cache_alignment(); @@ -606,75 +503,43 @@ static void *vb2_dc_get_userptr(void *alloc_ctx, unsigned long vaddr, buf->dev = conf->dev; buf->dma_dir = dma_dir;
- start = vaddr & PAGE_MASK; offset = vaddr & ~PAGE_MASK; - end = PAGE_ALIGN(vaddr + size); - n_pages = (end - start) >> PAGE_SHIFT; - - pages = kmalloc(n_pages * sizeof(pages[0]), GFP_KERNEL); - if (!pages) { - ret = -ENOMEM; - pr_err("failed to allocate pages table\n"); + vec = vb2_create_framevec(vaddr, size, dma_dir == DMA_FROM_DEVICE); + if (IS_ERR(vec)) { + ret = PTR_ERR(vec); goto fail_buf; } + buf->vec = vec; + n_pages = frame_vector_count(vec); + ret = frame_vector_to_pages(vec); + if (ret < 0) { + unsigned long *nums = frame_vector_pfns(vec);
- down_read(¤t->mm->mmap_sem); - /* current->mm->mmap_sem is taken by videobuf2 core */ - vma = find_vma(current->mm, vaddr); - if (!vma) { - pr_err("no vma for address %lu\n", vaddr); - ret = -EFAULT; - goto fail_pages; - } - - if (vma->vm_end < vaddr + size) { - pr_err("vma at %lu is too small for %lu bytes\n", vaddr, size); - ret = -EFAULT; - goto fail_pages; - } - - buf->vma = vb2_get_vma(vma); - if (!buf->vma) { - pr_err("failed to copy vma\n"); - ret = -ENOMEM; - goto fail_pages; - } - - /* extract page list from userspace mapping */ - ret = vb2_dc_get_user_pages(start, pages, n_pages, vma, dma_dir); - if (ret) { - unsigned long pfn; - if (vb2_dc_get_user_pfn(start, n_pages, vma, &pfn) == 0) { - up_read(¤t->mm->mmap_sem); - buf->dma_addr = vb2_dc_pfn_to_dma(buf->dev, pfn); - buf->size = size; - kfree(pages); - return buf; - } - - pr_err("failed to get user pages\n"); - goto fail_vma; + /* + * Failed to convert to pages... Check the memory is physically + * contiguous and use direct mapping + */ + for (i = 1; i < n_pages; i++) + if (nums[i-1] + 1 != nums[i]) + goto fail_pfnvec; + buf->dma_addr = vb2_dc_pfn_to_dma(buf->dev, nums[0]); + goto out; } - up_read(¤t->mm->mmap_sem);
sgt = kzalloc(sizeof(*sgt), GFP_KERNEL); if (!sgt) { pr_err("failed to allocate sg table\n"); ret = -ENOMEM; - goto fail_get_user_pages; + goto fail_pfnvec; }
- ret = sg_alloc_table_from_pages(sgt, pages, n_pages, + ret = sg_alloc_table_from_pages(sgt, frame_vector_pages(vec), n_pages, offset, size, GFP_KERNEL); if (ret) { pr_err("failed to initialize sg table\n"); goto fail_sgt; }
- /* pages are no longer needed */ - kfree(pages); - pages = NULL; - /* * No need to sync to the device, this will happen later when the * prepare() memop is called. @@ -696,8 +561,9 @@ static void *vb2_dc_get_userptr(void *alloc_ctx, unsigned long vaddr, }
buf->dma_addr = sg_dma_address(sgt->sgl); - buf->size = size; buf->dma_sgt = sgt; +out: + buf->size = size;
return buf;
@@ -706,25 +572,13 @@ fail_map_sg: buf->dma_dir, &attrs);
fail_sgt_init: - if (!vma_is_io(buf->vma)) - vb2_dc_sgt_foreach_page(sgt, put_page); sg_free_table(sgt);
fail_sgt: kfree(sgt);
-fail_get_user_pages: - if (pages && !vma_is_io(buf->vma)) - while (n_pages) - put_page(pages[--n_pages]); - - down_read(¤t->mm->mmap_sem); -fail_vma: - vb2_put_vma(buf->vma); - -fail_pages: - up_read(¤t->mm->mmap_sem); - kfree(pages); /* kfree is NULL-proof */ +fail_pfnvec: + vb2_destroy_framevec(vec);
fail_buf: kfree(buf);
Conversion to the use of pinned pfns made some functions unused. Remove them. Also there's no need to lock mmap_sem in __buf_prepare() anymore.
Acked-by: Marek Szyprowski m.szyprowski@samsung.com Tested-by: Marek Szyprowski m.szyprowski@samsung.com Signed-off-by: Jan Kara jack@suse.cz --- drivers/media/v4l2-core/videobuf2-memops.c | 114 ----------------------------- include/media/videobuf2-memops.h | 6 -- 2 files changed, 120 deletions(-)
diff --git a/drivers/media/v4l2-core/videobuf2-memops.c b/drivers/media/v4l2-core/videobuf2-memops.c index 0ec186d41b9b..48c6a49c4928 100644 --- a/drivers/media/v4l2-core/videobuf2-memops.c +++ b/drivers/media/v4l2-core/videobuf2-memops.c @@ -23,120 +23,6 @@ #include <media/videobuf2-memops.h>
/** - * vb2_get_vma() - acquire and lock the virtual memory area - * @vma: given virtual memory area - * - * This function attempts to acquire an area mapped in the userspace for - * the duration of a hardware operation. The area is "locked" by performing - * the same set of operation that are done when process calls fork() and - * memory areas are duplicated. - * - * Returns a copy of a virtual memory region on success or NULL. - */ -struct vm_area_struct *vb2_get_vma(struct vm_area_struct *vma) -{ - struct vm_area_struct *vma_copy; - - vma_copy = kmalloc(sizeof(*vma_copy), GFP_KERNEL); - if (vma_copy == NULL) - return NULL; - - if (vma->vm_ops && vma->vm_ops->open) - vma->vm_ops->open(vma); - - if (vma->vm_file) - get_file(vma->vm_file); - - memcpy(vma_copy, vma, sizeof(*vma)); - - vma_copy->vm_mm = NULL; - vma_copy->vm_next = NULL; - vma_copy->vm_prev = NULL; - - return vma_copy; -} -EXPORT_SYMBOL_GPL(vb2_get_vma); - -/** - * vb2_put_userptr() - release a userspace virtual memory area - * @vma: virtual memory region associated with the area to be released - * - * This function releases the previously acquired memory area after a hardware - * operation. - */ -void vb2_put_vma(struct vm_area_struct *vma) -{ - if (!vma) - return; - - if (vma->vm_ops && vma->vm_ops->close) - vma->vm_ops->close(vma); - - if (vma->vm_file) - fput(vma->vm_file); - - kfree(vma); -} -EXPORT_SYMBOL_GPL(vb2_put_vma); - -/** - * vb2_get_contig_userptr() - lock physically contiguous userspace mapped memory - * @vaddr: starting virtual address of the area to be verified - * @size: size of the area - * @res_paddr: will return physical address for the given vaddr - * @res_vma: will return locked copy of struct vm_area for the given area - * - * This function will go through memory area of size @size mapped at @vaddr and - * verify that the underlying physical pages are contiguous. If they are - * contiguous the virtual memory area is locked and a @res_vma is filled with - * the copy and @res_pa set to the physical address of the buffer. - * - * Returns 0 on success. - */ -int vb2_get_contig_userptr(unsigned long vaddr, unsigned long size, - struct vm_area_struct **res_vma, dma_addr_t *res_pa) -{ - struct mm_struct *mm = current->mm; - struct vm_area_struct *vma; - unsigned long offset, start, end; - unsigned long this_pfn, prev_pfn; - dma_addr_t pa = 0; - - start = vaddr; - offset = start & ~PAGE_MASK; - end = start + size; - - vma = find_vma(mm, start); - - if (vma == NULL || vma->vm_end < end) - return -EFAULT; - - for (prev_pfn = 0; start < end; start += PAGE_SIZE) { - int ret = follow_pfn(vma, start, &this_pfn); - if (ret) - return ret; - - if (prev_pfn == 0) - pa = this_pfn << PAGE_SHIFT; - else if (this_pfn != prev_pfn + 1) - return -EFAULT; - - prev_pfn = this_pfn; - } - - /* - * Memory is contigous, lock vma and return to the caller - */ - *res_vma = vb2_get_vma(vma); - if (*res_vma == NULL) - return -ENOMEM; - - *res_pa = pa + offset; - return 0; -} -EXPORT_SYMBOL_GPL(vb2_get_contig_userptr); - -/** * vb2_create_framevec() - map virtual addresses to pfns * @start: Virtual user address where we start mapping * @length: Length of a range to map diff --git a/include/media/videobuf2-memops.h b/include/media/videobuf2-memops.h index 2f0564ff5f31..830b5239fd8b 100644 --- a/include/media/videobuf2-memops.h +++ b/include/media/videobuf2-memops.h @@ -31,12 +31,6 @@ struct vb2_vmarea_handler {
extern const struct vm_operations_struct vb2_common_vm_ops;
-int vb2_get_contig_userptr(unsigned long vaddr, unsigned long size, - struct vm_area_struct **res_vma, dma_addr_t *res_pa); - -struct vm_area_struct *vb2_get_vma(struct vm_area_struct *vma); -void vb2_put_vma(struct vm_area_struct *vma); - struct frame_vector *vb2_create_framevec(unsigned long start, unsigned long length, bool write);
Convert g2d_userptr_get_dma_addr() to pin pages using get_vaddr_frames(). This removes the knowledge about vmas and mmap_sem locking from exynos driver. Also it fixes a problem that the function has been mapping user provided address without holding mmap_sem.
Signed-off-by: Jan Kara jack@suse.cz --- drivers/gpu/drm/exynos/exynos_drm_g2d.c | 89 ++++++++++-------------------- drivers/gpu/drm/exynos/exynos_drm_gem.c | 97 --------------------------------- 2 files changed, 29 insertions(+), 157 deletions(-)
diff --git a/drivers/gpu/drm/exynos/exynos_drm_g2d.c b/drivers/gpu/drm/exynos/exynos_drm_g2d.c index 81a250830808..265519c0fe2d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_g2d.c +++ b/drivers/gpu/drm/exynos/exynos_drm_g2d.c @@ -190,10 +190,8 @@ struct g2d_cmdlist_userptr { dma_addr_t dma_addr; unsigned long userptr; unsigned long size; - struct page **pages; - unsigned int npages; + struct frame_vector *vec; struct sg_table *sgt; - struct vm_area_struct *vma; atomic_t refcount; bool in_pool; bool out_of_list; @@ -363,6 +361,7 @@ static void g2d_userptr_put_dma_addr(struct drm_device *drm_dev, { struct g2d_cmdlist_userptr *g2d_userptr = (struct g2d_cmdlist_userptr *)obj; + struct page **pages;
if (!obj) return; @@ -382,19 +381,21 @@ out: exynos_gem_unmap_sgt_from_dma(drm_dev, g2d_userptr->sgt, DMA_BIDIRECTIONAL);
- exynos_gem_put_pages_to_userptr(g2d_userptr->pages, - g2d_userptr->npages, - g2d_userptr->vma); + pages = frame_vector_pages(g2d_userptr->vec); + if (!IS_ERR(pages)) { + int i;
- exynos_gem_put_vma(g2d_userptr->vma); + for (i = 0; i < frame_vector_count(g2d_userptr->vec); i++) + set_page_dirty_lock(pages[i]); + } + put_vaddr_frames(g2d_userptr->vec); + frame_vector_destroy(g2d_userptr->vec);
if (!g2d_userptr->out_of_list) list_del_init(&g2d_userptr->list);
sg_free_table(g2d_userptr->sgt); kfree(g2d_userptr->sgt); - - drm_free_large(g2d_userptr->pages); kfree(g2d_userptr); }
@@ -413,6 +414,7 @@ static dma_addr_t *g2d_userptr_get_dma_addr(struct drm_device *drm_dev, struct vm_area_struct *vma; unsigned long start, end; unsigned int npages, offset; + struct frame_vector *vec; int ret;
if (!size) { @@ -456,65 +458,37 @@ static dma_addr_t *g2d_userptr_get_dma_addr(struct drm_device *drm_dev, return ERR_PTR(-ENOMEM);
atomic_set(&g2d_userptr->refcount, 1); + g2d_userptr->size = size;
start = userptr & PAGE_MASK; offset = userptr & ~PAGE_MASK; end = PAGE_ALIGN(userptr + size); npages = (end - start) >> PAGE_SHIFT; - g2d_userptr->npages = npages; - - pages = drm_calloc_large(npages, sizeof(struct page *)); - if (!pages) { - DRM_ERROR("failed to allocate pages.\n"); - ret = -ENOMEM; + vec = g2d_userptr->vec = frame_vector_create(npages); + if (!vec) goto err_free; - }
- down_read(¤t->mm->mmap_sem); - vma = find_vma(current->mm, userptr); - if (!vma) { - up_read(¤t->mm->mmap_sem); - DRM_ERROR("failed to get vm region.\n"); + ret = get_vaddr_frames(start, npages, 1, 1, vec); + if (ret != npages) { + DRM_ERROR("failed to get user pages from userptr.\n"); + if (ret < 0) + goto err_destroy_framevec; ret = -EFAULT; - goto err_free_pages; + goto err_put_framevec; } - - if (vma->vm_end < userptr + size) { - up_read(¤t->mm->mmap_sem); - DRM_ERROR("vma is too small.\n"); + if (frame_vector_to_pages(vec) < 0) { ret = -EFAULT; - goto err_free_pages; + goto err_put_framevec; }
- g2d_userptr->vma = exynos_gem_get_vma(vma); - if (!g2d_userptr->vma) { - up_read(¤t->mm->mmap_sem); - DRM_ERROR("failed to copy vma.\n"); - ret = -ENOMEM; - goto err_free_pages; - } - - g2d_userptr->size = size; - - ret = exynos_gem_get_pages_from_userptr(start & PAGE_MASK, - npages, pages, vma); - if (ret < 0) { - up_read(¤t->mm->mmap_sem); - DRM_ERROR("failed to get user pages from userptr.\n"); - goto err_put_vma; - } - - up_read(¤t->mm->mmap_sem); - g2d_userptr->pages = pages; - sgt = kzalloc(sizeof(*sgt), GFP_KERNEL); if (!sgt) { ret = -ENOMEM; - goto err_free_userptr; + goto err_put_framevec; }
- ret = sg_alloc_table_from_pages(sgt, pages, npages, offset, - size, GFP_KERNEL); + ret = sg_alloc_table_from_pages(sgt, frame_vector_pages(vec), npages, + offset, size, GFP_KERNEL); if (ret < 0) { DRM_ERROR("failed to get sgt from pages.\n"); goto err_free_sgt; @@ -549,16 +523,11 @@ err_sg_free_table: err_free_sgt: kfree(sgt);
-err_free_userptr: - exynos_gem_put_pages_to_userptr(g2d_userptr->pages, - g2d_userptr->npages, - g2d_userptr->vma); - -err_put_vma: - exynos_gem_put_vma(g2d_userptr->vma); +err_put_framevec: + put_vaddr_frames(vec);
-err_free_pages: - drm_free_large(pages); +err_destroy_framevec: + frame_vector_destroy(vec);
err_free: kfree(g2d_userptr); diff --git a/drivers/gpu/drm/exynos/exynos_drm_gem.c b/drivers/gpu/drm/exynos/exynos_drm_gem.c index 0d5b9698d384..47068ae44ced 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_gem.c +++ b/drivers/gpu/drm/exynos/exynos_drm_gem.c @@ -378,103 +378,6 @@ int exynos_drm_gem_get_ioctl(struct drm_device *dev, void *data, return 0; }
-struct vm_area_struct *exynos_gem_get_vma(struct vm_area_struct *vma) -{ - struct vm_area_struct *vma_copy; - - vma_copy = kmalloc(sizeof(*vma_copy), GFP_KERNEL); - if (!vma_copy) - return NULL; - - if (vma->vm_ops && vma->vm_ops->open) - vma->vm_ops->open(vma); - - if (vma->vm_file) - get_file(vma->vm_file); - - memcpy(vma_copy, vma, sizeof(*vma)); - - vma_copy->vm_mm = NULL; - vma_copy->vm_next = NULL; - vma_copy->vm_prev = NULL; - - return vma_copy; -} - -void exynos_gem_put_vma(struct vm_area_struct *vma) -{ - if (!vma) - return; - - if (vma->vm_ops && vma->vm_ops->close) - vma->vm_ops->close(vma); - - if (vma->vm_file) - fput(vma->vm_file); - - kfree(vma); -} - -int exynos_gem_get_pages_from_userptr(unsigned long start, - unsigned int npages, - struct page **pages, - struct vm_area_struct *vma) -{ - int get_npages; - - /* the memory region mmaped with VM_PFNMAP. */ - if (vma_is_io(vma)) { - unsigned int i; - - for (i = 0; i < npages; ++i, start += PAGE_SIZE) { - unsigned long pfn; - int ret = follow_pfn(vma, start, &pfn); - if (ret) - return ret; - - pages[i] = pfn_to_page(pfn); - } - - if (i != npages) { - DRM_ERROR("failed to get user_pages.\n"); - return -EINVAL; - } - - return 0; - } - - get_npages = get_user_pages(current, current->mm, start, - npages, 1, 1, pages, NULL); - get_npages = max(get_npages, 0); - if (get_npages != npages) { - DRM_ERROR("failed to get user_pages.\n"); - while (get_npages) - put_page(pages[--get_npages]); - return -EFAULT; - } - - return 0; -} - -void exynos_gem_put_pages_to_userptr(struct page **pages, - unsigned int npages, - struct vm_area_struct *vma) -{ - if (!vma_is_io(vma)) { - unsigned int i; - - for (i = 0; i < npages; i++) { - set_page_dirty_lock(pages[i]); - - /* - * undo the reference we took when populating - * the table. - */ - put_page(pages[i]); - } - } -} - int exynos_gem_map_sgt_with_dma(struct drm_device *drm_dev, struct sg_table *sgt, enum dma_data_direction dir)
Hi,
On 2015년 05월 13일 22:08, Jan Kara wrote:
Convert g2d_userptr_get_dma_addr() to pin pages using get_vaddr_frames(). This removes the knowledge about vmas and mmap_sem locking from exynos driver. Also it fixes a problem that the function has been mapping user provided address without holding mmap_sem.
Signed-off-by: Jan Kara jack@suse.cz
drivers/gpu/drm/exynos/exynos_drm_g2d.c | 89 ++++++++++-------------------- drivers/gpu/drm/exynos/exynos_drm_gem.c | 97 --------------------------------- 2 files changed, 29 insertions(+), 157 deletions(-)
diff --git a/drivers/gpu/drm/exynos/exynos_drm_g2d.c b/drivers/gpu/drm/exynos/exynos_drm_g2d.c index 81a250830808..265519c0fe2d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_g2d.c +++ b/drivers/gpu/drm/exynos/exynos_drm_g2d.c @@ -190,10 +190,8 @@ struct g2d_cmdlist_userptr { dma_addr_t dma_addr; unsigned long userptr; unsigned long size;
- struct page **pages;
- unsigned int npages;
- struct frame_vector *vec; struct sg_table *sgt;
- struct vm_area_struct *vma; atomic_t refcount; bool in_pool; bool out_of_list;
@@ -363,6 +361,7 @@ static void g2d_userptr_put_dma_addr(struct drm_device *drm_dev, { struct g2d_cmdlist_userptr *g2d_userptr = (struct g2d_cmdlist_userptr *)obj;
struct page **pages;
if (!obj) return;
@@ -382,19 +381,21 @@ out: exynos_gem_unmap_sgt_from_dma(drm_dev, g2d_userptr->sgt, DMA_BIDIRECTIONAL);
- exynos_gem_put_pages_to_userptr(g2d_userptr->pages,
g2d_userptr->npages,g2d_userptr->vma);
- pages = frame_vector_pages(g2d_userptr->vec);
- if (!IS_ERR(pages)) {
int i;
- exynos_gem_put_vma(g2d_userptr->vma);
for (i = 0; i < frame_vector_count(g2d_userptr->vec); i++)set_page_dirty_lock(pages[i]);}
put_vaddr_frames(g2d_userptr->vec);
frame_vector_destroy(g2d_userptr->vec);
if (!g2d_userptr->out_of_list) list_del_init(&g2d_userptr->list);
sg_free_table(g2d_userptr->sgt); kfree(g2d_userptr->sgt);
- drm_free_large(g2d_userptr->pages); kfree(g2d_userptr);
}
@@ -413,6 +414,7 @@ static dma_addr_t *g2d_userptr_get_dma_addr(struct drm_device *drm_dev, struct vm_area_struct *vma; unsigned long start, end; unsigned int npages, offset;
struct frame_vector *vec; int ret;
if (!size) {
@@ -456,65 +458,37 @@ static dma_addr_t *g2d_userptr_get_dma_addr(struct drm_device *drm_dev, return ERR_PTR(-ENOMEM);
atomic_set(&g2d_userptr->refcount, 1);
g2d_userptr->size = size;
start = userptr & PAGE_MASK; offset = userptr & ~PAGE_MASK; end = PAGE_ALIGN(userptr + size); npages = (end - start) >> PAGE_SHIFT;
- g2d_userptr->npages = npages;
- pages = drm_calloc_large(npages, sizeof(struct page *));
The declaration to pages isn't needed anymore because you removed it.
- if (!pages) {
DRM_ERROR("failed to allocate pages.\n");ret = -ENOMEM;
- vec = g2d_userptr->vec = frame_vector_create(npages);
I think you can use g2d_userptr->vec so it seems that vec isn't needed.
- if (!vec) goto err_free;
}
down_read(¤t->mm->mmap_sem);
vma = find_vma(current->mm, userptr);
For vma, ditto.
Thanks, Inki Dae
[--SNIP--]
On Thu 14-05-15 19:51:23, Inki Dae wrote:
Hi,
On 2015년 05월 13일 22:08, Jan Kara wrote:
Convert g2d_userptr_get_dma_addr() to pin pages using get_vaddr_frames(). This removes the knowledge about vmas and mmap_sem locking from exynos driver. Also it fixes a problem that the function has been mapping user provided address without holding mmap_sem.
Signed-off-by: Jan Kara jack@suse.cz
drivers/gpu/drm/exynos/exynos_drm_g2d.c | 89 ++++++++++-------------------- drivers/gpu/drm/exynos/exynos_drm_gem.c | 97 --------------------------------- 2 files changed, 29 insertions(+), 157 deletions(-)
diff --git a/drivers/gpu/drm/exynos/exynos_drm_g2d.c b/drivers/gpu/drm/exynos/exynos_drm_g2d.c index 81a250830808..265519c0fe2d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_g2d.c +++ b/drivers/gpu/drm/exynos/exynos_drm_g2d.c @@ -190,10 +190,8 @@ struct g2d_cmdlist_userptr { dma_addr_t dma_addr; unsigned long userptr; unsigned long size;
- struct page **pages;
- unsigned int npages;
- struct frame_vector *vec; struct sg_table *sgt;
- struct vm_area_struct *vma; atomic_t refcount; bool in_pool; bool out_of_list;
@@ -363,6 +361,7 @@ static void g2d_userptr_put_dma_addr(struct drm_device *drm_dev, { struct g2d_cmdlist_userptr *g2d_userptr = (struct g2d_cmdlist_userptr *)obj;
struct page **pages;
if (!obj) return;
@@ -382,19 +381,21 @@ out: exynos_gem_unmap_sgt_from_dma(drm_dev, g2d_userptr->sgt, DMA_BIDIRECTIONAL);
- exynos_gem_put_pages_to_userptr(g2d_userptr->pages,
g2d_userptr->npages,g2d_userptr->vma);
- pages = frame_vector_pages(g2d_userptr->vec);
- if (!IS_ERR(pages)) {
int i;
- exynos_gem_put_vma(g2d_userptr->vma);
for (i = 0; i < frame_vector_count(g2d_userptr->vec); i++)set_page_dirty_lock(pages[i]);}
put_vaddr_frames(g2d_userptr->vec);
frame_vector_destroy(g2d_userptr->vec);
if (!g2d_userptr->out_of_list) list_del_init(&g2d_userptr->list);
sg_free_table(g2d_userptr->sgt); kfree(g2d_userptr->sgt);
- drm_free_large(g2d_userptr->pages); kfree(g2d_userptr);
}
@@ -413,6 +414,7 @@ static dma_addr_t *g2d_userptr_get_dma_addr(struct drm_device *drm_dev, struct vm_area_struct *vma; unsigned long start, end; unsigned int npages, offset;
struct frame_vector *vec; int ret;
if (!size) {
@@ -456,65 +458,37 @@ static dma_addr_t *g2d_userptr_get_dma_addr(struct drm_device *drm_dev, return ERR_PTR(-ENOMEM);
atomic_set(&g2d_userptr->refcount, 1);
g2d_userptr->size = size;
start = userptr & PAGE_MASK; offset = userptr & ~PAGE_MASK; end = PAGE_ALIGN(userptr + size); npages = (end - start) >> PAGE_SHIFT;
- g2d_userptr->npages = npages;
- pages = drm_calloc_large(npages, sizeof(struct page *));
The declaration to pages isn't needed anymore because you removed it.
- if (!pages) {
DRM_ERROR("failed to allocate pages.\n");ret = -ENOMEM;
- vec = g2d_userptr->vec = frame_vector_create(npages);
I think you can use g2d_userptr->vec so it seems that vec isn't needed.
- if (!vec) goto err_free;
}
down_read(¤t->mm->mmap_sem);
vma = find_vma(current->mm, userptr);
For vma, ditto.
Thanks for review! Attached is a new version of the patch.
Honza
For unclear reasons my SoB was missing in my pull request. So add it now:
Signed-off-by: Hans Verkuil hans.verkuil@cisco.com
Regards,
Hans
On 05/13/15 15:08, Jan Kara wrote:
Convert g2d_userptr_get_dma_addr() to pin pages using get_vaddr_frames(). This removes the knowledge about vmas and mmap_sem locking from exynos driver. Also it fixes a problem that the function has been mapping user provided address without holding mmap_sem.
Signed-off-by: Jan Kara jack@suse.cz
drivers/gpu/drm/exynos/exynos_drm_g2d.c | 89 ++++++++++-------------------- drivers/gpu/drm/exynos/exynos_drm_gem.c | 97 --------------------------------- 2 files changed, 29 insertions(+), 157 deletions(-)
diff --git a/drivers/gpu/drm/exynos/exynos_drm_g2d.c b/drivers/gpu/drm/exynos/exynos_drm_g2d.c index 81a250830808..265519c0fe2d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_g2d.c +++ b/drivers/gpu/drm/exynos/exynos_drm_g2d.c @@ -190,10 +190,8 @@ struct g2d_cmdlist_userptr { dma_addr_t dma_addr; unsigned long userptr; unsigned long size;
- struct page **pages;
- unsigned int npages;
- struct frame_vector *vec; struct sg_table *sgt;
- struct vm_area_struct *vma; atomic_t refcount; bool in_pool; bool out_of_list;
@@ -363,6 +361,7 @@ static void g2d_userptr_put_dma_addr(struct drm_device *drm_dev, { struct g2d_cmdlist_userptr *g2d_userptr = (struct g2d_cmdlist_userptr *)obj;
struct page **pages;
if (!obj) return;
@@ -382,19 +381,21 @@ out: exynos_gem_unmap_sgt_from_dma(drm_dev, g2d_userptr->sgt, DMA_BIDIRECTIONAL);
- exynos_gem_put_pages_to_userptr(g2d_userptr->pages,
g2d_userptr->npages,g2d_userptr->vma);
- pages = frame_vector_pages(g2d_userptr->vec);
- if (!IS_ERR(pages)) {
int i;
- exynos_gem_put_vma(g2d_userptr->vma);
for (i = 0; i < frame_vector_count(g2d_userptr->vec); i++)set_page_dirty_lock(pages[i]);}
put_vaddr_frames(g2d_userptr->vec);
frame_vector_destroy(g2d_userptr->vec);
if (!g2d_userptr->out_of_list) list_del_init(&g2d_userptr->list);
sg_free_table(g2d_userptr->sgt); kfree(g2d_userptr->sgt);
- drm_free_large(g2d_userptr->pages); kfree(g2d_userptr);
}
@@ -413,6 +414,7 @@ static dma_addr_t *g2d_userptr_get_dma_addr(struct drm_device *drm_dev, struct vm_area_struct *vma; unsigned long start, end; unsigned int npages, offset;
struct frame_vector *vec; int ret;
if (!size) {
@@ -456,65 +458,37 @@ static dma_addr_t *g2d_userptr_get_dma_addr(struct drm_device *drm_dev, return ERR_PTR(-ENOMEM);
atomic_set(&g2d_userptr->refcount, 1);
g2d_userptr->size = size;
start = userptr & PAGE_MASK; offset = userptr & ~PAGE_MASK; end = PAGE_ALIGN(userptr + size); npages = (end - start) >> PAGE_SHIFT;
- g2d_userptr->npages = npages;
- pages = drm_calloc_large(npages, sizeof(struct page *));
- if (!pages) {
DRM_ERROR("failed to allocate pages.\n");ret = -ENOMEM;
- vec = g2d_userptr->vec = frame_vector_create(npages);
- if (!vec) goto err_free;
}
down_read(¤t->mm->mmap_sem);
vma = find_vma(current->mm, userptr);
if (!vma) {
up_read(¤t->mm->mmap_sem);DRM_ERROR("failed to get vm region.\n");
- ret = get_vaddr_frames(start, npages, 1, 1, vec);
- if (ret != npages) {
DRM_ERROR("failed to get user pages from userptr.\n");if (ret < 0)goto err_destroy_framevec;
goto err_free_pages;
goto err_put_framevec;
- if (vma->vm_end < userptr + size) {
up_read(¤t->mm->mmap_sem);DRM_ERROR("vma is too small.\n");
- if (frame_vector_to_pages(vec) < 0) { ret = -EFAULT;
goto err_free_pages;
goto err_put_framevec;
- g2d_userptr->vma = exynos_gem_get_vma(vma);
- if (!g2d_userptr->vma) {
up_read(¤t->mm->mmap_sem);DRM_ERROR("failed to copy vma.\n");ret = -ENOMEM;goto err_free_pages;- }
- g2d_userptr->size = size;
- ret = exynos_gem_get_pages_from_userptr(start & PAGE_MASK,
npages, pages, vma);- if (ret < 0) {
up_read(¤t->mm->mmap_sem);DRM_ERROR("failed to get user pages from userptr.\n");goto err_put_vma;- }
- up_read(¤t->mm->mmap_sem);
- g2d_userptr->pages = pages;
- sgt = kzalloc(sizeof(*sgt), GFP_KERNEL); if (!sgt) { ret = -ENOMEM;
goto err_free_userptr;
goto err_put_framevec;
- ret = sg_alloc_table_from_pages(sgt, pages, npages, offset,
size, GFP_KERNEL);
- ret = sg_alloc_table_from_pages(sgt, frame_vector_pages(vec), npages,
offset, size, GFP_KERNEL);@@ -549,16 +523,11 @@ err_sg_free_table: err_free_sgt: kfree(sgt);
-err_free_userptr:
- exynos_gem_put_pages_to_userptr(g2d_userptr->pages,
g2d_userptr->npages,g2d_userptr->vma);-err_put_vma:
- exynos_gem_put_vma(g2d_userptr->vma);
+err_put_framevec:
- put_vaddr_frames(vec);
-err_free_pages:
- drm_free_large(pages);
+err_destroy_framevec:
- frame_vector_destroy(vec);
err_free: kfree(g2d_userptr); diff --git a/drivers/gpu/drm/exynos/exynos_drm_gem.c b/drivers/gpu/drm/exynos/exynos_drm_gem.c index 0d5b9698d384..47068ae44ced 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_gem.c +++ b/drivers/gpu/drm/exynos/exynos_drm_gem.c @@ -378,103 +378,6 @@ int exynos_drm_gem_get_ioctl(struct drm_device *dev, void *data, return 0; }
-struct vm_area_struct *exynos_gem_get_vma(struct vm_area_struct *vma) -{
- struct vm_area_struct *vma_copy;
- vma_copy = kmalloc(sizeof(*vma_copy), GFP_KERNEL);
- if (!vma_copy)
return NULL;- if (vma->vm_ops && vma->vm_ops->open)
vma->vm_ops->open(vma);- if (vma->vm_file)
get_file(vma->vm_file);- memcpy(vma_copy, vma, sizeof(*vma));
- vma_copy->vm_mm = NULL;
- vma_copy->vm_next = NULL;
- vma_copy->vm_prev = NULL;
- return vma_copy;
-}
-void exynos_gem_put_vma(struct vm_area_struct *vma) -{
- if (!vma)
return;- if (vma->vm_ops && vma->vm_ops->close)
vma->vm_ops->close(vma);- if (vma->vm_file)
fput(vma->vm_file);- kfree(vma);
-}
-int exynos_gem_get_pages_from_userptr(unsigned long start,
unsigned int npages,struct page **pages,struct vm_area_struct *vma)-{
- int get_npages;
- /* the memory region mmaped with VM_PFNMAP. */
- if (vma_is_io(vma)) {
unsigned int i;for (i = 0; i < npages; ++i, start += PAGE_SIZE) {unsigned long pfn;int ret = follow_pfn(vma, start, &pfn);if (ret)return ret;pages[i] = pfn_to_page(pfn);}if (i != npages) {DRM_ERROR("failed to get user_pages.\n");return -EINVAL;}return 0;- }
- get_npages = get_user_pages(current, current->mm, start,
npages, 1, 1, pages, NULL);- get_npages = max(get_npages, 0);
- if (get_npages != npages) {
DRM_ERROR("failed to get user_pages.\n");while (get_npages)put_page(pages[--get_npages]);return -EFAULT;- }
- return 0;
-}
-void exynos_gem_put_pages_to_userptr(struct page **pages,
unsigned int npages,struct vm_area_struct *vma)-{
- if (!vma_is_io(vma)) {
unsigned int i;for (i = 0; i < npages; i++) {set_page_dirty_lock(pages[i]);/** undo the reference we took when populating* the table.*/put_page(pages[i]);}- }
-}
int exynos_gem_map_sgt_with_dma(struct drm_device *drm_dev, struct sg_table *sgt, enum dma_data_direction dir)
Hi Jan,
On 05/13/2015 03:08 PM, Jan Kara wrote:
Hello,
I'm sending the fifth version of my patch series to abstract vma handling from the various media drivers. The patches got some review from mm people and testing from device driver guys so unless someone objects, patches will be queued in media tree for the next merge window.
What is the current status? I saw one comment for patch 9, so I assume it is not quite ready yet.
Let me know when you think it is time to merge.
Regards,
Hans
After this patch set drivers have to know much less details about vmas, their types, and locking. Also quite some code is removed from them. As a bonus drivers get automatically VM_FAULT_RETRY handling. The primary motivation for this series is to remove knowledge about mmap_sem locking from as many places a possible so that we can change it with reasonable effort.
The core of the series is the new helper get_vaddr_frames() which is given a virtual address and it fills in PFNs / struct page pointers (depending on VMA type) into the provided array. If PFNs correspond to normal pages it also grabs references to these pages. The difference from get_user_pages() is that this function can also deal with pfnmap, and io mappings which is what the media drivers need.
I have tested the patches with vivid driver so at least vb2 code got some exposure. Conversion of other drivers was just compile-tested (for x86 so e.g. exynos driver which is only for Samsung platform is completely untested).
HonzaChanges since v4:
- Minor cleanups and fixes pointed out by Mel and Vlasta
- Added Acked-by tags
Changes since v3:
- Added include <linux/vmalloc.h> into mm/gup.c as it's needed for some archs
- Fixed error path for exynos driver
Changes since v2:
- Renamed functions and structures as Mel suggested
- Other minor changes suggested by Mel
- Rebased on top of 4.1-rc2
- Changed functions to get pointer to array of pages / pfns to perform conversion if necessary. This fixes possible issue in the omap I may have introduced in v2 and generally makes the API less errorprone.
-- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon 25-05-15 13:37:56, Hans Verkuil wrote:
Hi Jan,
On 05/13/2015 03:08 PM, Jan Kara wrote:
Hello,
I'm sending the fifth version of my patch series to abstract vma handling from the various media drivers. The patches got some review from mm people and testing from device driver guys so unless someone objects, patches will be queued in media tree for the next merge window.
What is the current status? I saw one comment for patch 9, so I assume it is not quite ready yet.
Let me know when you think it is time to merge.
There was a minor comment to the exynos patch - I've sent updated version of that patch as a reply to the comment. Do you want me to resend the whole series? Otherwise I think you can just pick up the patches and merge them. Thanks!
Honza
After this patch set drivers have to know much less details about vmas, their types, and locking. Also quite some code is removed from them. As a bonus drivers get automatically VM_FAULT_RETRY handling. The primary motivation for this series is to remove knowledge about mmap_sem locking from as many places a possible so that we can change it with reasonable effort.
The core of the series is the new helper get_vaddr_frames() which is given a virtual address and it fills in PFNs / struct page pointers (depending on VMA type) into the provided array. If PFNs correspond to normal pages it also grabs references to these pages. The difference from get_user_pages() is that this function can also deal with pfnmap, and io mappings which is what the media drivers need.
I have tested the patches with vivid driver so at least vb2 code got some exposure. Conversion of other drivers was just compile-tested (for x86 so e.g. exynos driver which is only for Samsung platform is completely untested).
HonzaChanges since v4:
- Minor cleanups and fixes pointed out by Mel and Vlasta
- Added Acked-by tags
Changes since v3:
- Added include <linux/vmalloc.h> into mm/gup.c as it's needed for some archs
- Fixed error path for exynos driver
Changes since v2:
- Renamed functions and structures as Mel suggested
- Other minor changes suggested by Mel
- Rebased on top of 4.1-rc2
- Changed functions to get pointer to array of pages / pfns to perform conversion if necessary. This fixes possible issue in the omap I may have introduced in v2 and generally makes the API less errorprone.
-- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
dri-devel@lists.freedesktop.org
-
Andrew Morton -
Hans Verkuil -
Inki Dae -
Jan Kara