On Sun, Mar 18, 2012 at 6:34 PM, Daniel Vetter daniel.vetter@ffwll.ch wrote:

Big differences to other contenders in the field (like ion) is that this also supports highmem, so we have to split up the cpu access from the kernel side into a prepare and a kmap step.

Prepare is allowed to fail and should do everything required so that the kmap calls can succeed (like swapin/backing storage allocation, flushing, ...).

More in-depth explanations will follow in the follow-up documentation patch.

Changes in v2:

• Clear up begin_cpu_access confusion noticed by Sumit Semwal.
• Don't automatically fallback from the _atomic variants to the

non-atomic variants. The _atomic callbacks are not allowed to  sleep, so we want exporters to make this decision explicit. The  function signatures are explicit, so simpler exporters can still  use the same function for both.

• Make the unmap functions optional. Simpler exporters with permanent

mappings don't need to do anything at unmap time.

Changes in v3:

• Adjust the WARN_ON checks for the new ->ops functions as suggested

by Rob Clark and Sumit Semwal.

• Rebased on top of latest dma-buf-next git.

Signed-Off-by: Daniel Vetter daniel.vetter@ffwll.ch

Signed-off-by: Rob Clark rob@ti.com

note: we should start updating the individual driver support for drm drivers for this (since Dave has prime working now), although this should not block the drm-core support for prime/dmabuf

BR, -R

---

drivers/base/dma-buf.c  |  124 ++++++++++++++++++++++++++++++++++++++++++++++-  include/linux/dma-buf.h |   59 ++++++++++++++++++++++  2 files changed, 182 insertions(+), 1 deletions(-)

diff --git a/drivers/base/dma-buf.c b/drivers/base/dma-buf.c index 5641b9c..2226511 100644 --- a/drivers/base/dma-buf.c +++ b/drivers/base/dma-buf.c @@ -80,7 +80,9 @@ struct dma_buf *dma_buf_export(void *priv, const struct dma_buf_ops *ops,        if (WARN_ON(!priv || !ops                          || !ops->map_dma_buf                          || !ops->unmap_dma_buf

• || !ops->release)) {

• || !ops->release
• || !ops->kmap_atomic
• || !ops->kmap)) {

return ERR_PTR(-EINVAL);        }

@@ -284,3 +286,123 @@ void dma_buf_unmap_attachment(struct dma_buf_attachment *attach,                                                direction);  }  EXPORT_SYMBOL_GPL(dma_buf_unmap_attachment);

+/**

• • dma_buf_begin_cpu_access - Must be called before accessing a dma_buf from the
• • cpu in the kernel context. Calls begin_cpu_access to allow exporter-specific
• • preparations. Coherency is only guaranteed in the specified range for the
• • specified access direction.
• • @dma_buf:   [in]    buffer to prepare cpu access for.
• • @start:     [in]    start of range for cpu access.
• • @len:       [in]    length of range for cpu access.
• • @direction: [in]    length of range for cpu access.
• • Can return negative error values, returns 0 on success.
• */

+int dma_buf_begin_cpu_access(struct dma_buf *dmabuf, size_t start, size_t len,

• enum dma_data_direction direction)

+{

• int ret = 0;
• if (WARN_ON(!dmabuf))
• return EINVAL;
• if (dmabuf->ops->begin_cpu_access)
• ret = dmabuf->ops->begin_cpu_access(dmabuf, start, len, direction);
• return ret;

+} +EXPORT_SYMBOL_GPL(dma_buf_begin_cpu_access);

+/**

• • dma_buf_end_cpu_access - Must be called after accessing a dma_buf from the
• • cpu in the kernel context. Calls end_cpu_access to allow exporter-specific
• • actions. Coherency is only guaranteed in the specified range for the
• • specified access direction.
• • @dma_buf:   [in]    buffer to complete cpu access for.
• • @start:     [in]    start of range for cpu access.
• • @len:       [in]    length of range for cpu access.
• • @direction: [in]    length of range for cpu access.
• • This call must always succeed.
• */

+void dma_buf_end_cpu_access(struct dma_buf *dmabuf, size_t start, size_t len,

• enum dma_data_direction direction)

+{

• WARN_ON(!dmabuf);
• if (dmabuf->ops->end_cpu_access)
• dmabuf->ops->end_cpu_access(dmabuf, start, len, direction);

+} +EXPORT_SYMBOL_GPL(dma_buf_end_cpu_access);

+/**

• • dma_buf_kmap_atomic - Map a page of the buffer object into kernel address
• • space. The same restrictions as for kmap_atomic and friends apply.
• • @dma_buf:   [in]    buffer to map page from.
• • @page_num:  [in]    page in PAGE_SIZE units to map.
• • This call must always succeed, any necessary preparations that might fail
• • need to be done in begin_cpu_access.
• */

+void *dma_buf_kmap_atomic(struct dma_buf *dmabuf, unsigned long page_num) +{

• WARN_ON(!dmabuf);
• return dmabuf->ops->kmap_atomic(dmabuf, page_num);

+} +EXPORT_SYMBOL_GPL(dma_buf_kmap_atomic);

+/**

• • dma_buf_kunmap_atomic - Unmap a page obtained by dma_buf_kmap_atomic.
• • @dma_buf:   [in]    buffer to unmap page from.
• • @page_num:  [in]    page in PAGE_SIZE units to unmap.
• • @vaddr:     [in]    kernel space pointer obtained from dma_buf_kmap_atomic.
• • This call must always succeed.
• */

+void dma_buf_kunmap_atomic(struct dma_buf *dmabuf, unsigned long page_num,

• void *vaddr)

+{

• WARN_ON(!dmabuf);
• if (dmabuf->ops->kunmap_atomic)
• dmabuf->ops->kunmap_atomic(dmabuf, page_num, vaddr);

+} +EXPORT_SYMBOL_GPL(dma_buf_kunmap_atomic);

+/**

• • dma_buf_kmap - Map a page of the buffer object into kernel address space. The
• • same restrictions as for kmap and friends apply.
• • @dma_buf:   [in]    buffer to map page from.
• • @page_num:  [in]    page in PAGE_SIZE units to map.
• • This call must always succeed, any necessary preparations that might fail
• • need to be done in begin_cpu_access.
• */

+void *dma_buf_kmap(struct dma_buf *dmabuf, unsigned long page_num) +{

• WARN_ON(!dmabuf);
• return dmabuf->ops->kmap(dmabuf, page_num);

+} +EXPORT_SYMBOL_GPL(dma_buf_kmap);

+/**

• • dma_buf_kunmap - Unmap a page obtained by dma_buf_kmap.
• • @dma_buf:   [in]    buffer to unmap page from.
• • @page_num:  [in]    page in PAGE_SIZE units to unmap.
• • @vaddr:     [in]    kernel space pointer obtained from dma_buf_kmap.
• • This call must always succeed.
• */

+void dma_buf_kunmap(struct dma_buf *dmabuf, unsigned long page_num,

• void *vaddr)

+{

• WARN_ON(!dmabuf);
• if (dmabuf->ops->kunmap)
• dmabuf->ops->kunmap(dmabuf, page_num, vaddr);

+} +EXPORT_SYMBOL_GPL(dma_buf_kunmap); diff --git a/include/linux/dma-buf.h b/include/linux/dma-buf.h index 24e0f48..ee7ef99 100644 --- a/include/linux/dma-buf.h +++ b/include/linux/dma-buf.h @@ -50,6 +50,17 @@ struct dma_buf_attachment;  * @unmap_dma_buf: decreases usecount of buffer, might deallocate scatter  *                pages.  * @release: release this buffer; to be called after the last dma_buf_put.

• • @begin_cpu_access: [optional] called before cpu access to invalidate cpu
• • caches and allocate backing storage (if not yet done)
• • respectively pin the objet into memory.
• • @end_cpu_access: [optional] called after cpu access to flush cashes.
• • @kmap_atomic: maps a page from the buffer into kernel address
• • space, users may not block until the subsequent unmap call.
• • This callback must not sleep.
• • @kunmap_atomic: [optional] unmaps a atomically mapped page from the buffer.
• • This Callback must not sleep.
• • @kmap: maps a page from the buffer into kernel address space.
• • @kunmap: [optional] unmaps a page from the buffer.

*/  struct dma_buf_ops {        int (*attach)(struct dma_buf *, struct device *, @@ -73,6 +84,14 @@ struct dma_buf_ops {        /* after final dma_buf_put() */        void (*release)(struct dma_buf *);

• int (*begin_cpu_access)(struct dma_buf *, size_t, size_t,
• enum dma_data_direction);
• void (*end_cpu_access)(struct dma_buf *, size_t, size_t,
• enum dma_data_direction);
• void *(*kmap_atomic)(struct dma_buf *, unsigned long);
• void (*kunmap_atomic)(struct dma_buf *, unsigned long, void *);
• void *(*kmap)(struct dma_buf *, unsigned long);
• void (*kunmap)(struct dma_buf *, unsigned long, void *);

};

/** @@ -140,6 +159,14 @@ struct sg_table *dma_buf_map_attachment(struct dma_buf_attachment *,                                        enum dma_data_direction);  void dma_buf_unmap_attachment(struct dma_buf_attachment *, struct sg_table *,                                enum dma_data_direction); +int dma_buf_begin_cpu_access(struct dma_buf *dma_buf, size_t start, size_t len,

• enum dma_data_direction dir);

+void dma_buf_end_cpu_access(struct dma_buf *dma_buf, size_t start, size_t len,

• enum dma_data_direction dir);

+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 *);  #else

static inline struct dma_buf_attachment *dma_buf_attach(struct dma_buf *dmabuf, @@ -188,6 +215,38 @@ static inline void dma_buf_unmap_attachment(struct dma_buf_attachment *attach,        return;  }

+static inline int dma_buf_begin_cpu_access(struct dma_buf *,

• size_t, size_t,
• enum dma_data_direction)

+{

• return -ENODEV;

+}

+static inline void dma_buf_end_cpu_access(struct dma_buf *,

• size_t, size_t,
• enum dma_data_direction)

+{ +}

+static inline void *dma_buf_kmap_atomic(struct dma_buf *, unsigned long) +{

• return NULL;

+}

+static inline void dma_buf_kunmap_atomic(struct dma_buf *, unsigned long,

• void *)

+{ +}

+static inline void *dma_buf_kmap(struct dma_buf *, unsigned long) +{

• return NULL;

+}

+static inline void dma_buf_kunmap(struct dma_buf *, unsigned long,

• void *)

+{ +}  #endif /* CONFIG_DMA_SHARED_BUFFER */

#endif /* __DMA_BUF_H__ */

1.7.7.5

---

dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel

On Sun, Mar 18, 2012 at 6:34 PM, Daniel Vetter daniel.vetter@ffwll.ch wrote:

v2: Fix spelling issues noticed by Rob Clark.

Signed-off-by: Daniel Vetter daniel.vetter@ffwll.ch

Signed-off-by: Rob Clark rob@ti.com

---

Documentation/dma-buf-sharing.txt |  102 +++++++++++++++++++++++++++++++++++-  1 files changed, 99 insertions(+), 3 deletions(-)

diff --git a/Documentation/dma-buf-sharing.txt b/Documentation/dma-buf-sharing.txt index 225f96d..9f3aeef 100644 --- a/Documentation/dma-buf-sharing.txt +++ b/Documentation/dma-buf-sharing.txt @@ -32,8 +32,12 @@ The buffer-user  *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.

-- may be used *ONLY* by importers that do not need CPU access to the buffer.

• 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 +--------------------------------------

The dma_buf buffer sharing API usage contains the following steps:

@@ -219,7 +223,99 @@ NOTES:    If the exporter chooses not to allow an attach() operation once a    map_dma_buf() API has been called, it simply returns an error.

-Miscellaneous notes: +Kernel cpu access to a dma-buf buffer object +--------------------------------------------

+The motivation to allow cpu access from the kernel to a dma-buf object from the +importers side are: +- fallback operations, e.g. if the devices is connected to a usb bus and the

• kernel needs to shuffle the data around first before sending it away.

+- full transparency for existing users on the importer side, i.e. userspace

• should not notice the difference between a normal object from that subsystem
• and an imported one backed by a dma-buf. This is really important for drm
• opengl drivers that expect to still use all the existing upload/download
• paths.

+Access to a dma_buf from the kernel context involves three steps:

+1. Prepare access, which invalidate any necessary caches and make the object

• available for cpu access.

+2. Access the object page-by-page with the dma_buf map apis +3. Finish access, which will flush any necessary cpu caches and free reserved

• resources.

+1. Prepare access

• Before an importer can access a dma_buf object with the cpu from the kernel
• context, it needs to notify the exporter of the access that is about to
• happen.
• Interface:
• int dma_buf_begin_cpu_access(struct dma_buf *dmabuf,
• size_t start, size_t len,
• enum dma_data_direction direction)
• This allows the exporter to ensure that the memory is actually available for
• cpu access - the exporter might need to allocate or swap-in and pin the
• backing storage. The exporter also needs to ensure that cpu access is
• coherent for the given range and access direction. The range and access
• direction can be used by the exporter to optimize the cache flushing, i.e.
• access outside of the range or with a different direction (read instead of
• write) might return stale or even bogus data (e.g. when the exporter needs to
• copy the data to temporary storage).
• This step might fail, e.g. in oom conditions.

+2. Accessing the buffer

• To support dma_buf objects residing in highmem cpu access is page-based using
• an api similar to kmap. Accessing a dma_buf is done in aligned chunks of
• PAGE_SIZE size. Before accessing a chunk it needs to be mapped, which returns
• a pointer in kernel virtual address space. Afterwards the chunk needs to be
• unmapped again. There is no limit on how often a given chunk can be mapped
• and unmapped, i.e. the importer does not need to call begin_cpu_access again
• before mapping the same chunk again.
• Interfaces:
• void *dma_buf_kmap(struct dma_buf *, unsigned long);
• void dma_buf_kunmap(struct dma_buf *, unsigned long, void *);
• There are also atomic variants of these interfaces. Like for kmap they
• facilitate non-blocking fast-paths. Neither the importer nor the exporter (in
• the callback) is allowed to block when using these.
• Interfaces:
• void *dma_buf_kmap_atomic(struct dma_buf *, unsigned long);
• void dma_buf_kunmap_atomic(struct dma_buf *, unsigned long, void *);
• For importers all the restrictions of using kmap apply, like the limited
• supply of kmap_atomic slots. Hence an importer shall only hold onto at most 2
• atomic dma_buf kmaps at the same time (in any given process context).
• dma_buf kmap calls outside of the range specified in begin_cpu_access are
• undefined. If the range is not PAGE_SIZE aligned, kmap needs to succeed on
• the partial chunks at the beginning and end but may return stale or bogus
• data outside of the range (in these partial chunks).
• Note that these calls need to always succeed. The exporter needs to complete
• any preparations that might fail in begin_cpu_access.

+3. Finish access

• When the importer is done accessing the range specified in begin_cpu_access,
• it needs to announce this to the exporter (to facilitate cache flushing and
• unpinning of any pinned resources). The result of of any dma_buf kmap calls
• after end_cpu_access is undefined.
• Interface:
• void dma_buf_end_cpu_access(struct dma_buf *dma_buf,
• size_t start, size_t len,
• enum dma_data_direction dir);

+Miscellaneous notes +-------------------

- Any exporters or users of the dma-buf buffer sharing framework must have   a 'select DMA_SHARED_BUFFER' in their respective Kconfigs.

-- 1.7.7.5

-- 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

Otherwise subsystems will get this wrong and end up with a second export ioctl with the flag and O_CLOEXEC support added.

v2: Fixup the function name and caution exporters to limit the flags to only O_CLOEXEC. Noted by Dave Airlie.

Cc: Dave Airlie airlied@gmail.com Signed-Off-by: Daniel Vetter daniel.vetter@ffwll.ch --- Documentation/dma-buf-sharing.txt | 6 ++++++ 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/Documentation/dma-buf-sharing.txt b/Documentation/dma-buf-sharing.txt index 9f3aeef..a6d4c37 100644 --- a/Documentation/dma-buf-sharing.txt +++ b/Documentation/dma-buf-sharing.txt @@ -319,6 +319,12 @@ Miscellaneous notes - Any exporters or users of the dma-buf buffer sharing framework must have a 'select DMA_SHARED_BUFFER' in their respective Kconfigs.

+- To avoid the wrath of userspace library writers exporting subsystems must have + a flag parameter in the ioctl that creates the dma-buf fd which needs to + support at least the O_CLOEXEC fd flag. This needs to be passed in the flag + parameter of dma_buf_fd. Without any other reasons applying it is recommended + that exporters limit the flags passed to dma_buf_fd to only O_CLOEXEC. + References: [1] struct dma_buf_ops in include/linux/dma-buf.h [2] All interfaces mentioned above defined in include/linux/dma-buf.h