ANDROID: dma-buf: Add support for partial cache maintenance

In order to improve performance, allow dma-buf clients to
apply cache maintenance to only a subset of a dma-buf.

Kernel clients will be able to use the dma_buf_begin_cpu_access_partial
and dma_buf_end_cpu_access_partial functions to only apply cache
maintenance to a range within the dma-buf.

Bug: 133508579
Test: ion-unit-tests
Change-Id: Icce61fc21b1542f5248daea34f713184449a62c3
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Sandeep Patil <sspatil@google.com>

drivers/dma-buf/dma-buf.c

@@ -957,6 +957,30 @@ int dma_buf_begin_cpu_access(struct dma_buf *dmabuf,
 }
 EXPORT_SYMBOL_GPL(dma_buf_begin_cpu_access);
 
+int dma_buf_begin_cpu_access_partial(struct dma_buf *dmabuf,
+				     enum dma_data_direction direction,
+				     unsigned int offset, unsigned int len)
+{
+	int ret = 0;
+
+	if (WARN_ON(!dmabuf))
+		return -EINVAL;
+
+	if (dmabuf->ops->begin_cpu_access_partial)
+		ret = dmabuf->ops->begin_cpu_access_partial(dmabuf, direction,
+							    offset, len);
+
+	/* Ensure that all fences are waited upon - but we first allow
+	 * the native handler the chance to do so more efficiently if it
+	 * chooses. A double invocation here will be reasonably cheap no-op.
+	 */
+	if (ret == 0)
+		ret = __dma_buf_begin_cpu_access(dmabuf, direction);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dma_buf_begin_cpu_access_partial);
+
 /**
  * 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
@@ -983,6 +1007,22 @@ int dma_buf_end_cpu_access(struct dma_buf *dmabuf,
 }
 EXPORT_SYMBOL_GPL(dma_buf_end_cpu_access);
 
+int dma_buf_end_cpu_access_partial(struct dma_buf *dmabuf,
+				   enum dma_data_direction direction,
+				   unsigned int offset, unsigned int len)
+{
+	int ret = 0;
+
+	WARN_ON(!dmabuf);
+
+	if (dmabuf->ops->end_cpu_access_partial)
+		ret = dmabuf->ops->end_cpu_access_partial(dmabuf, direction,
+							  offset, len);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dma_buf_end_cpu_access_partial);
+
 /**
  * dma_buf_kmap - Map a page of the buffer object into kernel address space. The
  * same restrictions as for kmap and friends apply.

include/linux/dma-buf.h

@@ -178,6 +178,41 @@ struct dma_buf_ops {
 	 */
 	int (*begin_cpu_access)(struct dma_buf *, enum dma_data_direction);
 
+	/**
+	 * @begin_cpu_access_partial:
+	 *
+	 * This is called from dma_buf_begin_cpu_access_partial() and allows the
+	 * exporter to ensure that the memory specified in the range is
+	 * 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 access direction. The direction can be used by the
+	 * exporter to optimize the cache flushing, i.e. access 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 callback is optional.
+	 *
+	 * FIXME: This is both called through the DMA_BUF_IOCTL_SYNC command
+	 * from userspace (where storage shouldn't be pinned to avoid handing
+	 * de-factor mlock rights to userspace) and for the kernel-internal
+	 * users of the various kmap interfaces, where the backing storage must
+	 * be pinned to guarantee that the atomic kmap calls can succeed. Since
+	 * there's no in-kernel users of the kmap interfaces yet this isn't a
+	 * real problem.
+	 *
+	 * Returns:
+	 *
+	 * 0 on success or a negative error code on failure. This can for
+	 * example fail when the backing storage can't be allocated. Can also
+	 * return -ERESTARTSYS or -EINTR when the call has been interrupted and
+	 * needs to be restarted.
+	 */
+	int (*begin_cpu_access_partial)(struct dma_buf *dmabuf,
+					enum dma_data_direction,
+					unsigned int offset, unsigned int len);
+
 	/**
 	 * @end_cpu_access:
 	 *
@@ -197,6 +232,28 @@ struct dma_buf_ops {
 	 */
 	int (*end_cpu_access)(struct dma_buf *, enum dma_data_direction);
 
+	/**
+	 * @end_cpu_access_partial:
+	 *
+	 * This is called from dma_buf_end_cpu_access_partial() when the
+	 * importer is done accessing the CPU. The exporter can use to limit
+	 * cache flushing to only the range specefied and to unpin any
+	 * resources pinned in @begin_cpu_access_umapped.
+	 * The result of any dma_buf kmap calls after end_cpu_access_partial is
+	 * undefined.
+	 *
+	 * This callback is optional.
+	 *
+	 * Returns:
+	 *
+	 * 0 on success or a negative error code on failure. Can return
+	 * -ERESTARTSYS or -EINTR when the call has been interrupted and needs
+	 * to be restarted.
+	 */
+	int (*end_cpu_access_partial)(struct dma_buf *dmabuf,
+				      enum dma_data_direction,
+				      unsigned int offset, unsigned int len);
+
 	/**
 	 * @mmap:
 	 *
@@ -411,8 +468,14 @@ 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,
 			     enum dma_data_direction dir);
+int dma_buf_begin_cpu_access_partial(struct dma_buf *dma_buf,
+				     enum dma_data_direction dir,
+				     unsigned int offset, unsigned int len);
 int dma_buf_end_cpu_access(struct dma_buf *dma_buf,
 			   enum dma_data_direction dir);
+int dma_buf_end_cpu_access_partial(struct dma_buf *dma_buf,
+				     enum dma_data_direction dir,
+				     unsigned int offset, unsigned int len);
 void *dma_buf_kmap(struct dma_buf *, unsigned long);
 void dma_buf_kunmap(struct dma_buf *, unsigned long, void *);