Linaro-mm-sig April 2012

linaro-mm-sig@lists.linaro.org

43 participants
26 discussions

by Daniel Vetter

Compared to Rob Clark's RFC I've ditched the prepare/finish hooks and corresponding ioctls on the dma_buf file. The major reason for that is that many people seem to be under the impression that this is also for synchronization with outstanding asynchronous processsing. I'm pretty massively opposed to this because: - It boils down reinventing a new rather general-purpose userspace synchronization interface. If we look at things like futexes, this is hard to get right. - Furthermore a lot of kernel code has to interact with this synchronization primitive. This smells a look like the dri1 hw_lock, a horror show I prefer not to reinvent. - Even more fun is that multiple different subsystems would interact here, so we have plenty of opportunities to create funny deadlock scenarios. I think synchronization is a wholesale different problem from data sharing and should be tackled as an orthogonal problem. Now we could demand that prepare/finish may only ensure cache coherency (as Rob intended), but that runs up into the next problem: We not only need mmap support to facilitate sw-only processing nodes in a pipeline (without jumping through hoops by importing the dma_buf into some sw-access only importer), which allows for a nicer ION->dma-buf upgrade path for existing Android userspace. We also need mmap support for existing importing subsystems to support existing userspace libraries. And a loot of these subsystems are expected to export coherent userspace mappings. So prepare/finish can only ever be optional and the exporter /needs/ to support coherent mappings. Given that mmap access is always somewhat fallback-y in nature I've decided to drop this optimization, instead of just making it optional. If we demonstrate a clear need for this, supported by benchmark results, we can always add it in again later as an optional extension. Other differences compared to Rob's RFC is the above mentioned support for mapping a dma-buf through facilities provided by the importer. Which results in mmap support no longer being optional. Note that this dma-buf mmap patch does _not_ support every possible insanity an existing subsystem could pull of with mmap: Because it does not allow to intercept pagefaults and shoot down ptes importing subsystems can't add some magic of their own at these points (e.g. to automatically synchronize with outstanding rendering or set up some special resources). I've done a cursory read through a few mmap implementions of various subsytems and I'm hopeful that we can avoid this (and the complexity it'd bring with it). Additonally I've extended the documentation a bit to explain the hows and whys of this mmap extension. In case we ever want to add support for explicitly cache maneged userspace mmap with a prepare/finish ioctl pair, we could specify that userspace needs to mmap a different part of the dma_buf, e.g. the range starting at dma_buf->size up to dma_buf->size*2. This works because the size of a dma_buf is invariant over it's lifetime. The exporter would obviously need to fall back to coherent mappings for both ranges if a legacy clients maps the coherent range and the architecture cannot suppor conflicting caching policies. Also, this would obviously be optional and userspace needs to be able to fall back to coherent mappings. v2: - Spelling fixes from Rob Clark. - Compile fix for !DMA_BUF from Rob Clark. - Extend commit message to explain how explicitly cache managed mmap support could be added later. - Extend the documentation with implementations notes for exporters that need to manually fake coherency. Cc: Rob Clark <rob.clark(a)linaro.org> Cc: Rebecca Schultz Zavin <rebecca(a)android.com> Signed-Off-by: Daniel Vetter <daniel.vetter(a)ffwll.ch> --- Documentation/dma-buf-sharing.txt | 98 ++++++++++++++++++++++++++++++++++--- drivers/base/dma-buf.c | 64 +++++++++++++++++++++++- include/linux/dma-buf.h | 16 ++++++ 3 files changed, 170 insertions(+), 8 deletions(-) diff --git a/Documentation/dma-buf-sharing.txt b/Documentation/dma-buf-sharing.txt index 3bbd5c5..5ff4d2b 100644 --- a/Documentation/dma-buf-sharing.txt +++ b/Documentation/dma-buf-sharing.txt @@ -29,13 +29,6 @@ The buffer-user in memory, mapped into its own address space, so it can access the same area of memory. -*IMPORTANT*: [see https://lkml.org/lkml/2011/12/20/211 for more details] -For this first version, A buffer shared using the dma_buf sharing API: -- *may* be exported to user space using "mmap" *ONLY* by exporter, outside of - this framework. -- with this new iteration of the dma-buf api cpu access from the kernel has been - enable, see below for the details. - dma-buf operations for device dma only -------------------------------------- @@ -313,6 +306,83 @@ Access to a dma_buf from the kernel context involves three steps: enum dma_data_direction dir); +Direct Userspace Access/mmap Support +------------------------------------ + +Being able to mmap an export dma-buf buffer object has 2 main use-cases: +- CPU fallback processing in a pipeline and +- supporting existing mmap interfaces in importers. + +1. CPU fallback processing in a pipeline + + In many processing pipelines it is sometimes required that the cpu can access + the data in a dma-buf (e.g. for thumbnail creation, snapshots, ...). To avoid + the need to handle this specially in userspace frameworks for buffer sharing + it's ideal if the dma_buf fd itself can be used to access the backing storage + from userspace using mmap. + + Furthermore Android's ION framework already supports this (and is otherwise + rather similar to dma-buf from a userspace consumer side with using fds as + handles, too). So it's beneficial to support this in a similar fashion on + dma-buf to have a good transition path for existing Android userspace. + + No special interfaces, userspace simply calls mmap on the dma-buf fd. + +2. Supporting existing mmap interfaces in exporters + + Similar to the motivation for kernel cpu access it is again important that + the userspace code of a given importing subsystem can use the same interfaces + with a imported dma-buf buffer object as with a native buffer object. This is + especially important for drm where the userspace part of contemporary OpenGL, + X, and other drivers is huge, and reworking them to use a different way to + mmap a buffer rather invasive. + + The assumption in the current dma-buf interfaces is that redirecting the + initial mmap is all that's needed. A survey of some of the existing + subsystems shows that no driver seems to do any nefarious thing like syncing + up with outstanding asynchronous processing on the device or allocating + special resources at fault time. So hopefully this is good enough, since + adding interfaces to intercept pagefaults and allow pte shootdowns would + increase the complexity quite a bit. + + Interface: + int dma_buf_mmap(struct dma_buf *, struct vm_area_struct *, + unsigned long); + + If the importing subsystem simply provides a special-purpose mmap call to set + up a mapping in userspace, calling do_mmap with dma_buf->file will equally + achieve that for a dma-buf object. + +3. Implementation notes for exporters + + Because dma-buf buffers have invariant size over their lifetime, the dma-buf + core checks whether a vma is too large and rejects such mappings. The + exporter hence does not need to duplicate this check. + + Because existing importing subsystems might presume coherent mappings for + userspace, the exporter needs to set up a coherent mapping. If that's not + possible, it needs to fake coherency by manually shooting down ptes when + leaving the cpu domain and flushing caches at fault time. Note that all the + dma_buf files share the same anon inode, hence the exporter needs to replace + the dma_buf file stored in vma->vm_file with it's own if pte shootdown is + requred. This is because the kernel uses the underlying inode's address_space + for vma tracking (and hence pte tracking at shootdown time with + unmap_mapping_range). + + If the above shootdown dance turns out to be too expensive in certain + scenarios, we can extend dma-buf with a more explicit cache tracking scheme + for userspace mappings. But the current assumption is that using mmap is + always a slower path, so some inefficiencies should be acceptable. + + Exporters that shoot down mappings (for any reasons) shall not do any + synchronization at fault time with outstanding device operations. + Synchronization is an orthogonal issue to sharing the backing storage of a + buffer and hence should not be handled by dma-buf itself. This is explictly + mentioned here because many people seem to want something like this, but if + different exporters handle this differently, buffer sharing can fail in + interesting ways depending upong the exporter (if userspace starts depending + upon this implicit synchronization). + Miscellaneous notes ------------------- @@ -336,6 +406,20 @@ Miscellaneous notes the exporting driver to create a dmabuf fd must provide a way to let userspace control setting of O_CLOEXEC flag passed in to dma_buf_fd(). +- If an exporter needs to manually flush caches and hence needs to fake + coherency for mmap support, it needs to be able to zap all the ptes pointing + at the backing storage. Now linux mm needs a struct address_space associated + with the struct file stored in vma->vm_file to do that with the function + unmap_mapping_range. But the dma_buf framework only backs every dma_buf fd + with the anon_file struct file, i.e. all dma_bufs share the same file. + + Hence exporters need to setup their own file (and address_space) association + by setting vma->vm_file and adjusting vma->vm_pgoff in the dma_buf mmap + callback. In the specific case of a gem driver the exporter could use the + shmem file already provided by gem (and set vm_pgoff = 0). Exporters can then + zap ptes by unmapping the corresponding range of the struct address_space + associated with their own file. + References: [1] struct dma_buf_ops in include/linux/dma-buf.h [2] All interfaces mentioned above defined in include/linux/dma-buf.h diff --git a/drivers/base/dma-buf.c b/drivers/base/dma-buf.c index 07cbbc6..8a573b9 100644 --- a/drivers/base/dma-buf.c +++ b/drivers/base/dma-buf.c @@ -44,8 +44,26 @@ static int dma_buf_release(struct inode *inode, struct file *file) return 0; } +static int dma_buf_mmap_internal(struct file *file, struct vm_area_struct *vma) +{ + struct dma_buf *dmabuf; + + if (!is_dma_buf_file(file)) + return -EINVAL; + + /* check for overflowing the buffer's size */ + if (vma->vm_pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) > + dmabuf->size >> PAGE_SHIFT) + return -EINVAL; + + dmabuf = file->private_data; + + return dmabuf->ops->mmap(dmabuf, vma); +} + static const struct file_operations dma_buf_fops = { .release = dma_buf_release, + .mmap = dma_buf_mmap_internal, }; /* @@ -82,7 +100,8 @@ struct dma_buf *dma_buf_export(void *priv, const struct dma_buf_ops *ops, || !ops->unmap_dma_buf || !ops->release || !ops->kmap_atomic - || !ops->kmap)) { + || !ops->kmap + || !ops->mmap)) { return ERR_PTR(-EINVAL); } @@ -406,3 +425,46 @@ void dma_buf_kunmap(struct dma_buf *dmabuf, unsigned long page_num, dmabuf->ops->kunmap(dmabuf, page_num, vaddr); } EXPORT_SYMBOL_GPL(dma_buf_kunmap); + + +/** + * dma_buf_mmap - Setup up a userspace mmap with the given vma + * @dma_buf: [in] buffer that should back the vma + * @vma: [in] vma for the mmap + * @pgoff: [in] offset in pages where this mmap should start within the + * dma-buf buffer. + * + * This function adjusts the passed in vma so that it points at the file of the + * dma_buf operation. It alsog adjusts the starting pgoff and does bounds + * checking on the size of the vma. Then it calls the exporters mmap function to + * set up the mapping. + * + * Can return negative error values, returns 0 on success. + */ +int dma_buf_mmap(struct dma_buf *dmabuf, struct vm_area_struct *vma, + unsigned long pgoff) +{ + if (WARN_ON(!dmabuf || !vma)) + return -EINVAL; + + /* check for offset overflow */ + if (pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) < pgoff) + return -EOVERFLOW; + + /* check for overflowing the buffer's size */ + if (pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) > + dmabuf->size >> PAGE_SHIFT) + return -EINVAL; + + /* readjust the vma */ + if (vma->vm_file) + fput(vma->vm_file); + + vma->vm_file = dmabuf->file; + get_file(vma->vm_file); + + vma->vm_pgoff = pgoff; + + return dmabuf->ops->mmap(dmabuf, vma); +} +EXPORT_SYMBOL_GPL(dma_buf_mmap); diff --git a/include/linux/dma-buf.h b/include/linux/dma-buf.h index 3efbfc2..1f78d15 100644 --- a/include/linux/dma-buf.h +++ b/include/linux/dma-buf.h @@ -61,6 +61,10 @@ struct dma_buf_attachment; * This Callback must not sleep. * @kmap: maps a page from the buffer into kernel address space. * @kunmap: [optional] unmaps a page from the buffer. + * @mmap: used to expose the backing storage to userspace. Note that the + * mapping needs to be coherent - if the exporter doesn't directly + * support this, it needs to fake coherency by shooting down any ptes + * when transitioning away from the cpu domain. */ struct dma_buf_ops { int (*attach)(struct dma_buf *, struct device *, @@ -92,6 +96,8 @@ struct dma_buf_ops { void (*kunmap_atomic)(struct dma_buf *, unsigned long, void *); void *(*kmap)(struct dma_buf *, unsigned long); void (*kunmap)(struct dma_buf *, unsigned long, void *); + + int (*mmap)(struct dma_buf *, struct vm_area_struct *vma); }; /** @@ -167,6 +173,9 @@ void *dma_buf_kmap_atomic(struct dma_buf *, unsigned long); void dma_buf_kunmap_atomic(struct dma_buf *, unsigned long, void *); void *dma_buf_kmap(struct dma_buf *, unsigned long); void dma_buf_kunmap(struct dma_buf *, unsigned long, void *); + +int dma_buf_mmap(struct dma_buf *, struct vm_area_struct *, + unsigned long); #else static inline struct dma_buf_attachment *dma_buf_attach(struct dma_buf *dmabuf, @@ -248,6 +257,13 @@ static inline void dma_buf_kunmap(struct dma_buf *dmabuf, unsigned long pnum, void *vaddr) { } + +static inline int dma_buf_mmap(struct dma_buf *dmabuf, + struct vm_area_struct *vma, + unsigned long pgoff) +{ + return -ENODEV; +} #endif /* CONFIG_DMA_SHARED_BUFFER */ #endif /* __DMA_BUF_H__ */ -- 1.7.10

12 years

[PATCH 0/4] ARM: replace custom consistent dma region with vmalloc

by Marek Szyprowski

Hi! Recent changes to ioremap and unification of vmalloc regions on ARM significantly reduces the possible size of the consistent dma region and limited allowed dma coherent/writecombine allocations. This experimental patch series replaces custom consistent dma regions usage in dma-mapping framework in favour of generic vmalloc areas created on demand for each coherent and writecombine allocations. This patch is based on vanilla v3.4-rc2 release. Best regards Marek Szyprowski Samsung Poland R&D Center Patch summary: Marek Szyprowski (4): mm: vmalloc: use const void * for caller argument mm: vmalloc: export find_vm_area() function mm: vmalloc: add VM_DMA flag to indicate areas used by dma-mapping framework ARM: remove consistent dma region and use common vmalloc range for dma allocations arch/arm/include/asm/dma-mapping.h | 2 +- arch/arm/mm/dma-mapping.c | 220 +++++++----------------------------- include/linux/vmalloc.h | 10 +- mm/vmalloc.c | 31 ++++-- 4 files changed, 67 insertions(+), 196 deletions(-) -- 1.7.1.569.g6f426

12 years

Re: [Linaro-mm-sig] [PATCH 1/1] arm/dts: Tegra30: Add device tree support for SMMU

by Hiroshi Doyu

From: Thierry Reding <thierry.reding(a)avionic-design.de> Subject: Re: [PATCH 1/1] arm/dts: Tegra30: Add device tree support for SMMU Date: Wed, 18 Apr 2012 07:41:40 +0200 Message-ID: <20120418054140.GA17506(a)avionic-0098.adnet.avionic-design.de> > * PGP Signed by an unknown key > > * Hiroshi Doyu wrote: > > On Mon, 16 Apr 2012 17:34:19 +0200 > > Arnd Bergmann <arnd(a)arndb.de> wrote: > > > > > On Monday 16 April 2012, Hiroshi Doyu wrote: > > > > What about using "dma-window" property to specify IOVA range in dtsi as below? > > > > > > > > arch/powerpc/platforms/cell/iommu.c: > > > > > > > > 698 static int __init cell_iommu_get_window(struct device_node *np, > > > > 699 unsigned long *base, > > > > 700 unsigned long *size) > > > > 701 { > > > > 702 const void *dma_window; > > > > 703 unsigned long index; > > > > 704 > > > > 705 /* Use ibm,dma-window if available, else, hard code ! */ > > > > 706 dma_window = of_get_property(np, "ibm,dma-window", NULL); > > > > 707 if (dma_window == NULL) { ^^^^^^^^^^^^^^ > > > > 708 *base = 0; > > > > 709 *size = 0x80000000u; > > > > 710 return -ENODEV; > > > > 711 } > > > > 712 > > > > 713 of_parse_dma_window(np, dma_window, &index, base, size); > > > > 714 return 0; > > > > 715 } > > > > > > > > > > Yes, that's the right way to do it, but I would use the more generic > > > "dma-window" name rather than "ibm,dma-window", which was originally > > > introduced for pseries and for some reason copied into the cell qs2x > > > firmware. > > > > Which name is better, "dma-window" or "iova-window"? > > Considering DMA IOMMU mapping API, "dma-window" may sound appropriate to me. > > But the window isn't necessarily used for DMA transfers. I was going to > propose "aperture", but that doesn't seem quite right since it is a very > GART-specific term. "iova-window" sounds like the best fit for what it > really is. If we can consider IOMMU as one of the implementation of DMA(Direct Memory Access), the prefix "dma-(window)" may make sense here. Then, we don't have to introduce a new concept "IO virtual address(iova)" in addition to the existing "bus address"(?) Anyway, either name would be ok for me;)

12 years

[PATCH v4 00/14] Integration of videobuf2 with dmabuf

by Tomasz Stanislawski

Hello everyone, This patchset adds support for DMABUF [2] importing to V4L2 stack. The support for DMABUF exporting was moved to separate patchset due to dependency on patches for DMA mapping redesign by Marek Szyprowski [4]. v4: - rebased on mainline 3.4-rc2 - included missing importing support for s5p-fimc and s5p-tv - added patch for changing map/unmap for importers - fixes to Documentation part - coding style fixes - pairing {map/unmap}_dmabuf in vb2-core - fixing variable types and semantic of arguments in videobufb2-dma-contig.c v3: - rebased on mainline 3.4-rc1 - split 'code refactor' patch to multiple smaller patches - squashed fixes to Sumit's patches - patchset is no longer dependant on 'DMA mapping redesign' - separated path for handling IO and non-IO mappings - add documentation for DMABUF importing to V4L - removed all DMABUF exporter related code - removed usage of dma_get_pages extension v2: - extended VIDIOC_EXPBUF argument from integer memoffset to struct v4l2_exportbuffer - added patch that breaks DMABUF spec on (un)map_atachment callcacks but allows to work with existing implementation of DMABUF prime in DRM - all dma-contig code refactoring patches were squashed - bugfixes v1: List of changes since [1]. - support for DMA api extension dma_get_pages, the function is used to retrieve pages used to create DMA mapping. - small fixes/code cleanup to videobuf2 - added prepare and finish callbacks to vb2 allocators, it is used keep consistency between dma-cpu acess to the memory (by Marek Szyprowski) - support for exporting of DMABUF buffer in V4L2 and Videobuf2, originated from [3]. - support for dma-buf exporting in vb2-dma-contig allocator - support for DMABUF for s5p-tv and s5p-fimc (capture interface) drivers, originated from [3] - changed handling for userptr buffers (by Marek Szyprowski, Andrzej Pietrasiewicz) - let mmap method to use dma_mmap_writecombine call (by Marek Szyprowski) [1] http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/4296… [2] https://lkml.org/lkml/2011/12/26/29 [3] http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/3635… [4] http://thread.gmane.org/gmane.linux.kernel.cross-arch/12819 Andrzej Pietrasiewicz (1): v4l: vb2-dma-contig: add support for scatterlist in userptr mode Laurent Pinchart (2): v4l: vb2-dma-contig: Shorten vb2_dma_contig prefix to vb2_dc v4l: vb2-dma-contig: Reorder functions Marek Szyprowski (2): v4l: vb2: add prepare/finish callbacks to allocators v4l: vb2-dma-contig: add prepare/finish to dma-contig allocator Sumit Semwal (4): v4l: Add DMABUF as a memory type v4l: vb2: add support for shared buffer (dma_buf) v4l: vb: remove warnings about MEMORY_DMABUF v4l: vb2-dma-contig: add support for dma_buf importing Tomasz Stanislawski (5): Documentation: media: description of DMABUF importing in V4L2 v4l: vb2-dma-contig: Remove unneeded allocation context structure v4l: vb2-dma-contig: change map/unmap behaviour for importers v4l: s5p-tv: mixer: support for dmabuf importing v4l: fimc: support for dmabuf importing Documentation/DocBook/media/v4l/compat.xml | 4 + Documentation/DocBook/media/v4l/io.xml | 179 +++++++ .../DocBook/media/v4l/vidioc-create-bufs.xml | 1 + Documentation/DocBook/media/v4l/vidioc-qbuf.xml | 15 + Documentation/DocBook/media/v4l/vidioc-reqbufs.xml | 47 +- drivers/media/video/Kconfig | 1 + drivers/media/video/s5p-fimc/fimc-capture.c | 2 +- drivers/media/video/s5p-tv/Kconfig | 1 + drivers/media/video/s5p-tv/mixer_video.c | 2 +- drivers/media/video/videobuf-core.c | 4 + drivers/media/video/videobuf2-core.c | 207 +++++++- drivers/media/video/videobuf2-dma-contig.c | 553 +++++++++++++++++--- include/linux/videodev2.h | 7 + include/media/videobuf2-core.h | 34 ++ 14 files changed, 956 insertions(+), 101 deletions(-) -- 1.7.5.4

12 years

[PATCH] WIP: staging: drm/omap: dmabuf/prime mmap support

by Rob Clark

From: Rob Clark <rob(a)ti.com> There are still a few places to cleanup, and possibly things can be made a bit easier for individual drm drivers by some helpers in drm core. But this is enough for a proof of concept. With this, I can allocate cached buffers, fill them w/ software from userspace, and on kernel side with fault handling and PTE shootdown track which pages must be flushed to simulate coherency. Fyi, you can find my kernel branch, with this patch on top of Daniel's dmabuf mmap patch, here: https://github.com/robclark/kernel-omap4/commits/dmabuf/mmap --- drivers/staging/omapdrm/omap_drv.h | 4 + drivers/staging/omapdrm/omap_fb.c | 6 +- drivers/staging/omapdrm/omap_gem.c | 99 ++++++++++++++++++++++++++--- drivers/staging/omapdrm/omap_gem_dmabuf.c | 45 +++++++++++++ 4 files changed, 142 insertions(+), 12 deletions(-) diff --git a/drivers/staging/omapdrm/omap_drv.h b/drivers/staging/omapdrm/omap_drv.h index 2b41152..5812e20 100644 --- a/drivers/staging/omapdrm/omap_drv.h +++ b/drivers/staging/omapdrm/omap_drv.h @@ -138,6 +138,8 @@ int omap_gem_dumb_destroy(struct drm_file *file, struct drm_device *dev, int omap_gem_dumb_create(struct drm_file *file, struct drm_device *dev, struct drm_mode_create_dumb *args); int omap_gem_mmap(struct file *filp, struct vm_area_struct *vma); +int omap_gem_mmap_obj(struct drm_gem_object *obj, + struct vm_area_struct *vma); int omap_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf); int omap_gem_op_start(struct drm_gem_object *obj, enum omap_gem_op op); int omap_gem_op_finish(struct drm_gem_object *obj, enum omap_gem_op op); @@ -145,6 +147,8 @@ int omap_gem_op_sync(struct drm_gem_object *obj, enum omap_gem_op op); int omap_gem_op_async(struct drm_gem_object *obj, enum omap_gem_op op, void (*fxn)(void *arg), void *arg); int omap_gem_roll(struct drm_gem_object *obj, uint32_t roll); +int omap_gem_dma_sync(struct drm_gem_object *obj, + enum dma_data_direction dir); int omap_gem_get_paddr(struct drm_gem_object *obj, dma_addr_t *paddr, bool remap); int omap_gem_put_paddr(struct drm_gem_object *obj); diff --git a/drivers/staging/omapdrm/omap_fb.c b/drivers/staging/omapdrm/omap_fb.c index 04b235b..ac32b3d 100644 --- a/drivers/staging/omapdrm/omap_fb.c +++ b/drivers/staging/omapdrm/omap_fb.c @@ -197,8 +197,10 @@ int omap_framebuffer_replace(struct drm_framebuffer *a, pa->paddr = 0; } - if (pb && !ret) - ret = omap_gem_get_paddr(pb->bo, &pb->paddr, true); + if (pb && !ret) { + ret = omap_gem_get_paddr(pb->bo, &pb->paddr, true) || + omap_gem_dma_sync(pb->bo, DMA_TO_DEVICE); + } } if (ret) { diff --git a/drivers/staging/omapdrm/omap_gem.c b/drivers/staging/omapdrm/omap_gem.c index c5ba334..be7f6c0 100644 --- a/drivers/staging/omapdrm/omap_gem.c +++ b/drivers/staging/omapdrm/omap_gem.c @@ -212,8 +212,11 @@ static DEFINE_SPINLOCK(sync_lock); /** ensure backing pages are allocated */ static int omap_gem_attach_pages(struct drm_gem_object *obj) { + struct drm_device *dev = obj->dev; struct omap_gem_object *omap_obj = to_omap_bo(obj); struct page **pages; + int i, npages = obj->size >> PAGE_SHIFT; + dma_addr_t *addrs; WARN_ON(omap_obj->pages); @@ -231,16 +234,18 @@ static int omap_gem_attach_pages(struct drm_gem_object *obj) * DSS, GPU, etc. are not cache coherent: */ if (omap_obj->flags & (OMAP_BO_WC|OMAP_BO_UNCACHED)) { - int i, npages = obj->size >> PAGE_SHIFT; - dma_addr_t *addrs = kmalloc(npages * sizeof(addrs), GFP_KERNEL); + addrs = kmalloc(npages * sizeof(addrs), GFP_KERNEL); for (i = 0; i < npages; i++) { - addrs[i] = dma_map_page(obj->dev->dev, pages[i], + addrs[i] = dma_map_page(dev->dev, pages[i], 0, PAGE_SIZE, DMA_BIDIRECTIONAL); } - omap_obj->addrs = addrs; + } else { + addrs = kzalloc(npages * sizeof(addrs), GFP_KERNEL); } + omap_obj->addrs = addrs; omap_obj->pages = pages; + return 0; } @@ -258,10 +263,11 @@ static void omap_gem_detach_pages(struct drm_gem_object *obj) dma_unmap_page(obj->dev->dev, omap_obj->addrs[i], PAGE_SIZE, DMA_BIDIRECTIONAL); } - kfree(omap_obj->addrs); - omap_obj->addrs = NULL; } + kfree(omap_obj->addrs); + omap_obj->addrs = NULL; + _drm_gem_put_pages(obj, omap_obj->pages, true, false); omap_obj->pages = NULL; } @@ -336,6 +342,13 @@ static int fault_1d(struct drm_gem_object *obj, vma->vm_start) >> PAGE_SHIFT; if (omap_obj->pages) { + if ((omap_obj->flags & OMAP_BO_CACHE_MASK) == OMAP_BO_CACHED) { + if (omap_obj->addrs[pgoff]) { + dma_unmap_page(obj->dev->dev, omap_obj->addrs[pgoff], + PAGE_SIZE, DMA_BIDIRECTIONAL); + omap_obj->addrs[pgoff] = 0; + } + } pfn = page_to_pfn(omap_obj->pages[pgoff]); } else { BUG_ON(!(omap_obj->flags & OMAP_BO_DMA)); @@ -510,7 +523,6 @@ fail: /** We override mainly to fix up some of the vm mapping flags.. */ int omap_gem_mmap(struct file *filp, struct vm_area_struct *vma) { - struct omap_gem_object *omap_obj; int ret; ret = drm_gem_mmap(filp, vma); @@ -519,8 +531,14 @@ int omap_gem_mmap(struct file *filp, struct vm_area_struct *vma) return ret; } - /* after drm_gem_mmap(), it is safe to access the obj */ - omap_obj = to_omap_bo(vma->vm_private_data); + return omap_gem_mmap_obj(vma->vm_private_data, vma); +} + + +int omap_gem_mmap_obj(struct drm_gem_object *obj, + struct vm_area_struct *vma) +{ + struct omap_gem_object *omap_obj = to_omap_bo(obj); vma->vm_flags &= ~VM_PFNMAP; vma->vm_flags |= VM_MIXEDMAP; @@ -530,12 +548,34 @@ int omap_gem_mmap(struct file *filp, struct vm_area_struct *vma) } else if (omap_obj->flags & OMAP_BO_UNCACHED) { vma->vm_page_prot = pgprot_noncached(vm_get_page_prot(vma->vm_flags)); } else { + /* + * We do have some private objects, at least for scanout buffers + * on hardware without DMM/TILER. But these are allocated write- + * combine + */ + if (WARN_ON(!obj->filp)) + return -EINVAL; + + // TODO quick hack to avoid needing to re-invent my own vmops: + WARN_ON(obj->filp->private_data); + obj->filp->private_data = vma->vm_file->private_data; + fput(vma->vm_file); + get_file(obj->filp); + + /* + * Shunt off cached objs to shmem file so they have their own + * address_space (so unmap_mapping_range does what we want) + */ + vma->vm_pgoff = 0; + vma->vm_file = obj->filp; + vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); } - return ret; + return 0; } + /** * omap_gem_dumb_create - create a dumb buffer * @drm_file: our client file @@ -645,6 +685,44 @@ fail: return ret; } +int omap_gem_dma_sync(struct drm_gem_object *obj, + enum dma_data_direction dir) +{ + struct drm_device *dev = obj->dev; + struct omap_gem_object *omap_obj = to_omap_bo(obj); + int ret = 0; + + // TODO not sure if I want this to be automagic or new flag set + // when allocating buffer (or maybe instead a flag to explictly + // disable? + + if ((omap_obj->flags & OMAP_BO_CACHE_MASK) == OMAP_BO_CACHED) { + int i, npages = obj->size >> PAGE_SHIFT; + struct page **pages = omap_obj->pages; + bool dirty = false; + + if (WARN_ON(!obj->filp)) + return -EINVAL; + + for (i = 0; i < npages; i++) { + if (!omap_obj->addrs[i]) { + // TODO using BIDIRECTIONAL for now because in the + // fault handler we don't really know the direction.. + omap_obj->addrs[i] = dma_map_page(dev->dev, pages[i], 0, + PAGE_SIZE, DMA_BIDIRECTIONAL); + dirty = true; + } + } + + if (dirty) { + unmap_mapping_range(obj->filp->f_mapping, 0, + omap_gem_mmap_size(obj), 1); + } + } + + return ret; +} + /* Get physical address for DMA.. if 'remap' is true, and the buffer is not * already contiguous, remap it to pin in physically contiguous memory.. (ie. * map in TILER) @@ -709,6 +787,7 @@ int omap_gem_get_paddr(struct drm_gem_object *obj, *paddr = omap_obj->paddr; } else { ret = -EINVAL; + goto fail; } fail: diff --git a/drivers/staging/omapdrm/omap_gem_dmabuf.c b/drivers/staging/omapdrm/omap_gem_dmabuf.c index 2fa39e8..85a0b1c 100644 --- a/drivers/staging/omapdrm/omap_gem_dmabuf.c +++ b/drivers/staging/omapdrm/omap_gem_dmabuf.c @@ -34,6 +34,8 @@ static struct sg_table *omap_gem_map_dma_buf( if (!sg) return ERR_PTR(-ENOMEM); + omap_gem_dma_sync(obj, dir); + /* camera, etc, need physically contiguous.. but we need a * better way to know this.. */ @@ -131,6 +133,48 @@ static void omap_gem_dmabuf_kunmap(struct dma_buf *buffer, kunmap(pages[page_num]); } +/* TODO: perhaps this could be a helper: */ +static int omap_gem_dmabuf_mmap(struct dma_buf *buffer, + struct vm_area_struct *vma) +{ + struct drm_gem_object *obj = buffer->priv; + int ret = 0; + + if (WARN_ON(!obj->filp)) + return -EINVAL; + + // TODO maybe we can split up drm_gem_mmap to avoid duplicating + // some here.. or at least have a drm_dmabuf helper.. + + /* Check for valid size. */ + if (omap_gem_mmap_size(obj) < vma->vm_end - vma->vm_start) { + ret = -EINVAL; + goto out_unlock; + } + + if (!obj->dev->driver->gem_vm_ops) { + ret = -EINVAL; + goto out_unlock; + } + + vma->vm_flags |= VM_RESERVED | VM_IO | VM_PFNMAP | VM_DONTEXPAND; + vma->vm_ops = obj->dev->driver->gem_vm_ops; + vma->vm_private_data = obj; + vma->vm_page_prot = pgprot_writecombine(vm_get_page_prot(vma->vm_flags)); + + /* Take a ref for this mapping of the object, so that the fault + * handler can dereference the mmap offset's pointer to the object. + * This reference is cleaned up by the corresponding vm_close + * (which should happen whether the vma was created by this call, or + * by a vm_open due to mremap or partial unmap or whatever). + */ + vma->vm_ops->open(vma); + +out_unlock: + + return omap_gem_mmap_obj(obj, vma); +} + struct dma_buf_ops omap_dmabuf_ops = { .map_dma_buf = omap_gem_map_dma_buf, .unmap_dma_buf = omap_gem_unmap_dma_buf, @@ -141,6 +185,7 @@ struct dma_buf_ops omap_dmabuf_ops = { .kunmap_atomic = omap_gem_dmabuf_kunmap_atomic, .kmap = omap_gem_dmabuf_kmap, .kunmap = omap_gem_dmabuf_kunmap, + .mmap = omap_gem_dmabuf_mmap, }; struct dma_buf * omap_gem_prime_export(struct drm_device *dev, -- 1.7.9.1

12 years

[PATCHv8 00/10] ARM: DMA-mapping framework redesign

by Marek Szyprowski

Hello, Linux v3.4-rc2, which include dma-mapping preparation patches, has been released two days ago, now it's time for the next spin of ARM dma-mapping redesign patches. This version includes various fixes posted separately to v7, mainly related to incorrect io address space bitmap setup and a major issue with broken mmap for memory which comes from dma_declare_coherent(). The patches have been also rebased onto Linux v3.4-rc2 which comes with dma_map_ops related changes. The code has been tested on Samsung Exynos4 'UniversalC210' and NURI boards with IOMMU driver posted by KyongHo Cho, I will put separate branch which shows how to integrate this driver with this patchset. The patches are also available on my git repository at: git://git.linaro.org/people/mszyprowski/linux-dma-mapping.git 3.4-rc2-arm-dma-v8 History of the development: v1: (initial version of the DMA-mapping redesign patches): http://www.spinics.net/lists/linux-mm/msg21241.html v2: http://lists.linaro.org/pipermail/linaro-mm-sig/2011-September/000571.html http://lists.linaro.org/pipermail/linaro-mm-sig/2011-September/000577.html v3: http://www.spinics.net/lists/linux-mm/msg25490.html v4 and v5: http://www.spinics.net/lists/arm-kernel/msg151147.html http://www.spinics.net/lists/arm-kernel/msg154889.html v6: http://www.spinics.net/lists/linux-mm/msg29903.html v7: http://www.spinics.net/lists/arm-kernel/msg162149.html Best regards Marek Szyprowski Samsung Poland R&D Center Patch summary: Marek Szyprowski (10): common: add dma_mmap_from_coherent() function ARM: dma-mapping: use pr_* instread of printk ARM: dma-mapping: introduce ARM_DMA_ERROR constant ARM: dma-mapping: remove offset parameter to prepare for generic dma_ops ARM: dma-mapping: use asm-generic/dma-mapping-common.h ARM: dma-mapping: implement dma sg methods on top of any generic dma ops ARM: dma-mapping: move all dma bounce code to separate dma ops structure ARM: dma-mapping: remove redundant code and cleanup ARM: dma-mapping: use alloc, mmap, free from dma_ops ARM: dma-mapping: add support for IOMMU mapper arch/arm/Kconfig | 9 + arch/arm/common/dmabounce.c | 84 +++- arch/arm/include/asm/device.h | 4 + arch/arm/include/asm/dma-iommu.h | 34 ++ arch/arm/include/asm/dma-mapping.h | 407 ++++----------- arch/arm/mm/dma-mapping.c | 1019 ++++++++++++++++++++++++++++++------ arch/arm/mm/vmregion.h | 2 +- drivers/base/dma-coherent.c | 42 ++ include/asm-generic/dma-coherent.h | 4 +- 9 files changed, 1138 insertions(+), 467 deletions(-) create mode 100644 arch/arm/include/asm/dma-iommu.h -- 1.7.1.569.g6f426

12 years

[PATCH 00/11] Integration of videobuf2 with dmabuf

by Tomasz Stanislawski

Hello everyone, This patchset adds support for DMABUF [2] importing to V4L2 stack. It was updated after Laurent Pinchart's review. The support for DMABUF exporting was moved to separate patchset due to dependency on patches for DMA mapping redesign by Marek Szyprowski [4]. v3: - rebased on mainline 3.4-rc1 - split 'code refactor' patch to multiple smaller patches - squashed fixes to Sumit's patches - patchset is no longer dependant on 'DMA mapping redesign' - separated path for handling IO and non-IO mappings - add documentation for DMABUF importing to V4L - removed all DMABUF exporter related code - removed usage of dma_get_pages extension v2: - extended VIDIOC_EXPBUF argument from integer memoffset to struct v4l2_exportbuffer - added patch that breaks DMABUF spec on (un)map_atachment callcacks but allows to work with existing implementation of DMABUF prime in DRM - all dma-contig code refactoring patches were squashed - bugfixes v1: List of changes since [1]. - support for DMA api extension dma_get_pages, the function is used to retrieve pages used to create DMA mapping. - small fixes/code cleanup to videobuf2 - added prepare and finish callbacks to vb2 allocators, it is used keep consistency between dma-cpu acess to the memory (by Marek Szyprowski) - support for exporting of DMABUF buffer in V4L2 and Videobuf2, originated from [3]. - support for dma-buf exporting in vb2-dma-contig allocator - support for DMABUF for s5p-tv and s5p-fimc (capture interface) drivers, originated from [3] - changed handling for userptr buffers (by Marek Szyprowski, Andrzej Pietrasiewicz) - let mmap method to use dma_mmap_writecombine call (by Marek Szyprowski) [1] http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/4296… [2] https://lkml.org/lkml/2011/12/26/29 [3] http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/3635… [4] http://thread.gmane.org/gmane.linux.kernel.cross-arch/12819 Andrzej Pietrasiewicz (1): v4l: vb2-dma-contig: add support for scatterlist in userptr mode Laurent Pinchart (2): v4l: vb2-dma-contig: Shorten vb2_dma_contig prefix to vb2_dc v4l: vb2-dma-contig: Reorder functions Marek Szyprowski (2): v4l: vb2: add prepare/finish callbacks to allocators v4l: vb2-dma-contig: add prepare/finish to dma-contig allocator Sumit Semwal (4): v4l: Add DMABUF as a memory type v4l: vb2: add support for shared buffer (dma_buf) v4l: vb: remove warnings about MEMORY_DMABUF v4l: vb2: Add dma-contig allocator as dma_buf user Tomasz Stanislawski (2): Documentation: media: description of DMABUF importing in V4L2 v4l: vb2-dma-contig: Remove unneeded allocation context structure Documentation/DocBook/media/v4l/compat.xml | 4 + Documentation/DocBook/media/v4l/io.xml | 177 +++++++ .../DocBook/media/v4l/vidioc-create-bufs.xml | 1 + Documentation/DocBook/media/v4l/vidioc-qbuf.xml | 15 + Documentation/DocBook/media/v4l/vidioc-reqbufs.xml | 8 +- drivers/media/video/videobuf-core.c | 4 + drivers/media/video/videobuf2-core.c | 195 +++++++- drivers/media/video/videobuf2-dma-contig.c | 536 +++++++++++++++++--- include/linux/videodev2.h | 8 + include/media/videobuf2-core.h | 38 ++ 10 files changed, 910 insertions(+), 76 deletions(-) -- 1.7.5.4

12 years

overlay UI layer of ICS supports by preserved memory allocation in virtualized environment

by Kwang Sik Shin

I'm interested in CMA scheme or preserved memory allocation of ION (ION_HEAP_TYPE_CARVEOUT heap type) I knew many SoC vendors (TI, nvidia, qualcom) already used similar approaches for multimedia processing in the past. However, reservation of physical memory should waste memory and limit available system memory, so now they use an IOMMU to support a contiguous virtual address region. In addition to, ICS (Ice Cream Sandwitch) support overlay UI layer requiring lots of contiguous memory allocation area. Now I'm suggesting an idea to use lots of contiguous memory allocation area without IOMMU. So ION_HEAP_TYPE_CARVEOUT heap type of ION is one of my considerations, but you know that it might exhaust much physical memory. However, I think physical memory wasting problem can be solved in the virtualized system. In case of virtualized environment, physical memory is intermediate physical memory instead of real physical memory, so a preserved carveout heap does not exhaust real physical memory any more. In addition to, if virtual address and intermediate physical address are 32 and 40bits respectively, virtual memory is not exhausted neither. How do you think about my suggestion?

12 years

[PATCH 0/4] [RFC] Addition of dma-mapping IOMMU framework in Exynos DRM

by Prathyush

From: Prathyush K <prathyush.k(a)samsung.com> With these set of patches, the Exynos5 DRM module will be able to allocate, export and map non-contiguous memory for GEM objects using the dma-mapping IOMMU framework. These patches are based on the following references: 1> Exynos 5 SYSMMU driver by Kyongho Cho http://lists.infradead.org/pipermail/linux-arm-kernel/2011-December/078636.… 2> DMA-MAPPING IOMMU framework by Marek Szyprowski http://lists.infradead.org/pipermail/linux-arm-kernel/2011-June/053636.html 3> DRM-PRIME by Dave Airlie http://patches.linaro.org/7480/ 4> DRM Exynos DMABUF module by Inki Dae http://git.infradead.org/users/kmpark/linux-2.6-samsung/commit/ d59d3e521d8a9d27e7b329f89f24a8473f32f50c/ 5> Support for dma_get_pages by Tomasz Stanislawski http://patchwork.linuxtv.org/patch/9639/ The updated GEM framework by Inki Dae is not considered for this patch set. IOMMU mapping is created for exynos-drm-device initially. Allocation of GEM objects happens through dma_alloc_writecombine which inturn calls arm_iommu_alloc_attrs. This will allocate a non-contig set of pages, allocate a iova and map all the pages. The following changes are done to support non-contiguous memory for GEM objects [PATCH 1/4]: drm/exynos: DMABUF: Added support for exporting non-contig buffers. This patch is for creating and mapping a SGT in map_dma_buf by retrieving the pages by calling dma_get_pages. [PATCH 2/4]: drm/exynos: Mapping of gem objects uses dma_mmap_writecombine. This patch is for mapping the non-contiguous GEM objects to user space by calling dma_mmap_writecombine. [PATCH 3/4]: drm/Exynos: Added 'disable' function to Exynos drm crtc module. This patch is required for safe release of DRM. If a FB of a CRTC is released, the driver tries to disable the CRTC if that is supported by the CRTC. This patch adds Exynos DRM CRTC disable functionality. [PATCH 4/4]: drm: Releasing FBs before releasing GEM objects during drm_release. This patch is required for safe release of DRM. During drm release, all the FBs and GEM objects are released. A gem object which is used as a FB must not be freed first before releasing the FB. This patch modifies the drm release function to first release the FBs and then release the GEMs. drivers/gpu/drm/drm_fops.c | 6 +- drivers/gpu/drm/exynos/exynos_drm_crtc.c | 11 +++ drivers/gpu/drm/exynos/exynos_drm_dmabuf.c | 98 +++++++++++++++++++++++----- drivers/gpu/drm/exynos/exynos_drm_gem.c | 70 +++++++++----------- 4 files changed, 128 insertions(+), 57 deletions(-)

12 years

[PATCH/RFC] ARM: Exynos4: Integrate IOMMU aware DMA-mapping

by Marek Szyprowski

Hi! This is an example of the IOMMU aware DMA-mapping implementation usage on a Samsung Exynos4 based NURI board. The ARM DMA-mapping IOMMU aware implementation is available in the [1] thread: This patch essentially registers DMA-mmaping/IOMMU support for FIMC and MFC devices and performs some tweaks in clocks hierarchy to get SYSMMU driver working correctly. The drivers have been tested with mainline V4L2 drivers for FIMC and MFC hardware. For easier testing I've created a separate kernel branch with all required prerequisite patches. It is based on lastest kgene/for-next branch and is available on my git repository: git://git.linaro.org/people/mszyprowski/linux-dma-mapping.git 3.4-rc2-arm-dma-v8-samsung This patch requires the following items: 1. ARM DMA-mapping patches [1] 2. Exynos SYSMMU driver v12 [2] 3. Exynos SYSMMU driver runtime pm fixes 4. Exynos4 gen_pd power domain driver fixes Runtime pm and power domain patches are required on Samsung Nuri board, but might be optional on boards where bootloader doesn't disable all devices on boot. [1] http://www.spinics.net/lists/linux-arch/msg17331.html [2] https://lkml.org/lkml/2012/3/15/51 Best regards Marek Szyprowski Samsung Poland R&D Center Patch summary: Marek Szyprowski (1): ARM: Exynos4: integrate SYSMMU driver with DMA-mapping interface arch/arm/mach-exynos/Kconfig | 1 + arch/arm/mach-exynos/clock-exynos4.c | 64 +++++++++++++++------------- arch/arm/mach-exynos/dev-sysmmu.c | 44 +++++++++++++++++++ arch/arm/mach-exynos/include/mach/sysmmu.h | 3 + drivers/iommu/Kconfig | 1 + 5 files changed, 84 insertions(+), 29 deletions(-) -- 1.7.1.569.g6f426

12 years

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

Linaro-mm-sig April 2012