[3/4] dma: document dma_flags_set_dmabarrier()

[akepner@sgi.com]

Document dma_flags_set_dmabarrier().

Signed-off-by: Arthur Kepner <akepner@sgi.com>

---
 DMA-API.txt |   26 ++++++++++++++++++++++++++
 1 files changed, 26 insertions(+)

diff --git a/Documentation/DMA-API.txt b/Documentation/DMA-API.txt
index cc7a8c3..5fc0bba 100644
--- a/Documentation/DMA-API.txt
+++ b/Documentation/DMA-API.txt
@@ -544,3 +544,29 @@ size is the size (and should be a page-sized multiple).
 The return value will be either a pointer to the processor virtual
 address of the memory, or an error (via PTR_ERR()) if any part of the
 region is occupied.
+
+int 
+dma_flags_set_dmabarrier(int dir)
+
+Amend dir (one of the enum dma_data_direction values), with a 
+platform-specific "dmabarrier" attribute.  The dmabarrier attribute 
+forces a flush of all in-flight DMA when the associated memory 
+region is written to (see example below.)
+
+This provides a mechanism to enforce ordering of DMA on platforms that 
+permit DMA to be reordered between device and host memory (within a 
+NUMA interconnect).  On other platforms this is a nop.
+
+The dmabarrier would be set when the memory region is mapped for DMA, 
+e.g.:
+
+	int count, flags = dma_flags_set_dmabarrier(DMA_BIDIRECTIONAL);
+	....
+	count = dma_map_sg(dev, sglist, nents, flags);
+
+As an example of a situation where this would be useful, suppose that 
+the device does a DMA write to indicate that data is ready and 
+available in memory.  The DMA of the "completion indication" could 
+race with data DMA.  Using a dmabarrier on the memory used for 
+completion indications would prevent the race.
+