From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Ujfalusi Subject: [PATCH 01/13] dmaengine: edma: Remove alignment constraint for memcpy Date: Wed, 14 Oct 2015 16:12:12 +0300 Message-ID: <1444828344-21378-2-git-send-email-peter.ujfalusi@ti.com> References: <1444828344-21378-1-git-send-email-peter.ujfalusi@ti.com> Mime-Version: 1.0 Content-Type: text/plain Return-path: In-Reply-To: <1444828344-21378-1-git-send-email-peter.ujfalusi@ti.com> Sender: linux-kernel-owner@vger.kernel.org To: vinod.koul@intel.com, nsekhar@ti.com, linux@arm.linux.org.uk Cc: olof@lixom.net, arnd@arndb.de, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, linux-omap@vger.kernel.org, dmaengine@vger.kernel.org, devicetree@vger.kernel.org, tony@atomide.com, r.schwebel@pengutronix.de List-Id: devicetree@vger.kernel.org Despite the claim by the original commit adding the memcpy support, eDMA does not have constraint on the alignment of src, dst or length in increment mode. Signed-off-by: Peter Ujfalusi --- drivers/dma/edma.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) diff --git a/drivers/dma/edma.c b/drivers/dma/edma.c index 7eefbf1e1c94..b36dfa5458cb 100644 --- a/drivers/dma/edma.c +++ b/drivers/dma/edma.c @@ -1324,6 +1324,7 @@ static struct dma_async_tx_descriptor *edma_prep_dma_memcpy( struct edma_desc *edesc; struct device *dev = chan->device->dev; struct edma_chan *echan = to_edma_chan(chan); + unsigned int width; if (unlikely(!echan || !len)) return NULL; @@ -1336,8 +1337,12 @@ static struct dma_async_tx_descriptor *edma_prep_dma_memcpy( edesc->pset_nr = 1; + width = 1 << __ffs((src | dest | len)); + if (width > DMA_SLAVE_BUSWIDTH_64_BYTES) + width = DMA_SLAVE_BUSWIDTH_64_BYTES; + ret = edma_config_pset(chan, &edesc->pset[0], src, dest, 1, - DMA_SLAVE_BUSWIDTH_4_BYTES, len, DMA_MEM_TO_MEM); + width, len, DMA_MEM_TO_MEM); if (ret < 0) return NULL; @@ -1903,12 +1908,6 @@ static void edma_dma_init(struct edma_cc *ecc, struct dma_device *dma, dma->dev = dev; - /* - * code using dma memcpy must make sure alignment of - * length is at dma->copy_align boundary. - */ - dma->copy_align = DMAENGINE_ALIGN_4_BYTES; - INIT_LIST_HEAD(&dma->channels); } -- 2.6.1