From adf850bcca4eec45f06c6ad3e297a350a8a319bc Mon Sep 17 00:00:00 2001 From: Peter Ujfalusi Date: Wed, 11 Nov 2015 12:37:55 +0200 Subject: dmaengine: omap-dma: Correct status reporting for memcpy During mem copy both src and dst position moves at the same pace. Check the dst position for progress reporting. Signed-off-by: Peter Ujfalusi Tested-by: Tomi Valkeinen Signed-off-by: Jyri Sarha Signed-off-by: Vinod Koul --- drivers/dma/omap-dma.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/dma/omap-dma.c') diff --git a/drivers/dma/omap-dma.c b/drivers/dma/omap-dma.c index 1dfc71c90123..8ed39ce24d46 100644 --- a/drivers/dma/omap-dma.c +++ b/drivers/dma/omap-dma.c @@ -719,7 +719,7 @@ static enum dma_status omap_dma_tx_status(struct dma_chan *chan, if (d->dir == DMA_MEM_TO_DEV) pos = omap_dma_get_src_pos(c); - else if (d->dir == DMA_DEV_TO_MEM) + else if (d->dir == DMA_DEV_TO_MEM || d->dir == DMA_MEM_TO_MEM) pos = omap_dma_get_dst_pos(c); else pos = 0; -- cgit v1.2.3 From e8a5e79c17f977d1820ffe9b6ac259c953b3b34b Mon Sep 17 00:00:00 2001 From: Peter Ujfalusi Date: Wed, 11 Nov 2015 12:37:56 +0200 Subject: dmaengine: omap-dma: Clean up the prep_slave_sg sg list walk code The for_each_sg() macro's last parameter is inteded to be used as counter. We can use 'i' instead of 'j' within the loop for indexes. Signed-off-by: Peter Ujfalusi Signed-off-by: Vinod Koul --- drivers/dma/omap-dma.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'drivers/dma/omap-dma.c') diff --git a/drivers/dma/omap-dma.c b/drivers/dma/omap-dma.c index 8ed39ce24d46..4afc2c18d451 100644 --- a/drivers/dma/omap-dma.c +++ b/drivers/dma/omap-dma.c @@ -768,7 +768,7 @@ static struct dma_async_tx_descriptor *omap_dma_prep_slave_sg( struct scatterlist *sgent; struct omap_desc *d; dma_addr_t dev_addr; - unsigned i, j = 0, es, en, frame_bytes; + unsigned i, es, en, frame_bytes; u32 burst; if (dir == DMA_DEV_TO_MEM) { @@ -845,13 +845,12 @@ static struct dma_async_tx_descriptor *omap_dma_prep_slave_sg( en = burst; frame_bytes = es_bytes[es] * en; for_each_sg(sgl, sgent, sglen, i) { - d->sg[j].addr = sg_dma_address(sgent); - d->sg[j].en = en; - d->sg[j].fn = sg_dma_len(sgent) / frame_bytes; - j++; + d->sg[i].addr = sg_dma_address(sgent); + d->sg[i].en = en; + d->sg[i].fn = sg_dma_len(sgent) / frame_bytes; } - d->sglen = j; + d->sglen = sglen; return vchan_tx_prep(&c->vc, &d->vd, tx_flags); } -- cgit v1.2.3 From 1c1d25f9f933211b622b0e209716372480051361 Mon Sep 17 00:00:00 2001 From: Peter Ujfalusi Date: Wed, 11 Nov 2015 12:37:57 +0200 Subject: dmaengine: omap-dma: Remove tasklet to start the transfers The use of tasklet to actually start the DMA transfer slightly decreases the DMA throughput since it adds small scheduling delay when the transfer is started. In normal use, even with high I/O load the tasklet would start one transaction at a time, however running the DMAtest for memcpy on all available channels will cause the tasklet to start about 15 transfers. The performance numbers on OMAP4 PandaBoard-es (test_buf_size = 6553): With the tasklet: dmatest: dma0chan30-copy: summary 5000 tests, 0 failures 186 iops 593 KB/s (0) dmatest: dma0chan8-copy0: summary 5000 tests, 0 failures 184 iops 584 KB/s (0) dmatest: dma0chan13-copy: summary 5000 tests, 0 failures 184 iops 585 KB/s (0) dmatest: dma0chan12-copy: summary 5000 tests, 0 failures 184 iops 585 KB/s (0) dmatest: dma0chan7-copy0: summary 5000 tests, 0 failures 183 iops 581 KB/s (0) With this patch (no tasklet): dmatest: dma0chan4-copy0: summary 5000 tests, 0 failures 199 iops 644 KB/s (0) dmatest: dma0chan5-copy0: summary 5000 tests, 0 failures 199 iops 645 KB/s (0) dmatest: dma0chan6-copy0: summary 5000 tests, 0 failures 199 iops 637 KB/s (0) dmatest: dma0chan24-copy: summary 5000 tests, 0 failures 199 iops 638 KB/s (0) dmatest: dma0chan16-copy: summary 5000 tests, 0 failures 199 iops 638 KB/s (0) Signed-off-by: Peter Ujfalusi Signed-off-by: Vinod Koul --- drivers/dma/omap-dma.c | 59 ++------------------------------------------------ 1 file changed, 2 insertions(+), 57 deletions(-) (limited to 'drivers/dma/omap-dma.c') diff --git a/drivers/dma/omap-dma.c b/drivers/dma/omap-dma.c index 4afc2c18d451..4e4642f561f5 100644 --- a/drivers/dma/omap-dma.c +++ b/drivers/dma/omap-dma.c @@ -28,8 +28,6 @@ struct omap_dmadev { struct dma_device ddev; spinlock_t lock; - struct tasklet_struct task; - struct list_head pending; void __iomem *base; const struct omap_dma_reg *reg_map; struct omap_system_dma_plat_info *plat; @@ -42,7 +40,6 @@ struct omap_dmadev { struct omap_chan { struct virt_dma_chan vc; - struct list_head node; void __iomem *channel_base; const struct omap_dma_reg *reg_map; uint32_t ccr; @@ -454,33 +451,6 @@ static void omap_dma_callback(int ch, u16 status, void *data) spin_unlock_irqrestore(&c->vc.lock, flags); } -/* - * This callback schedules all pending channels. We could be more - * clever here by postponing allocation of the real DMA channels to - * this point, and freeing them when our virtual channel becomes idle. - * - * We would then need to deal with 'all channels in-use' - */ -static void omap_dma_sched(unsigned long data) -{ - struct omap_dmadev *d = (struct omap_dmadev *)data; - LIST_HEAD(head); - - spin_lock_irq(&d->lock); - list_splice_tail_init(&d->pending, &head); - spin_unlock_irq(&d->lock); - - while (!list_empty(&head)) { - struct omap_chan *c = list_first_entry(&head, - struct omap_chan, node); - - spin_lock_irq(&c->vc.lock); - list_del_init(&c->node); - omap_dma_start_desc(c); - spin_unlock_irq(&c->vc.lock); - } -} - static irqreturn_t omap_dma_irq(int irq, void *devid) { struct omap_dmadev *od = devid; @@ -739,22 +709,8 @@ static void omap_dma_issue_pending(struct dma_chan *chan) unsigned long flags; spin_lock_irqsave(&c->vc.lock, flags); - if (vchan_issue_pending(&c->vc) && !c->desc) { - /* - * c->cyclic is used only by audio and in this case the DMA need - * to be started without delay. - */ - if (!c->cyclic) { - struct omap_dmadev *d = to_omap_dma_dev(chan->device); - spin_lock(&d->lock); - if (list_empty(&c->node)) - list_add_tail(&c->node, &d->pending); - spin_unlock(&d->lock); - tasklet_schedule(&d->task); - } else { - omap_dma_start_desc(c); - } - } + if (vchan_issue_pending(&c->vc) && !c->desc) + omap_dma_start_desc(c); spin_unlock_irqrestore(&c->vc.lock, flags); } @@ -1017,17 +973,11 @@ static int omap_dma_slave_config(struct dma_chan *chan, struct dma_slave_config static int omap_dma_terminate_all(struct dma_chan *chan) { struct omap_chan *c = to_omap_dma_chan(chan); - struct omap_dmadev *d = to_omap_dma_dev(c->vc.chan.device); unsigned long flags; LIST_HEAD(head); spin_lock_irqsave(&c->vc.lock, flags); - /* Prevent this channel being scheduled */ - spin_lock(&d->lock); - list_del_init(&c->node); - spin_unlock(&d->lock); - /* * Stop DMA activity: we assume the callback will not be called * after omap_dma_stop() returns (even if it does, it will see @@ -1101,14 +1051,12 @@ static int omap_dma_chan_init(struct omap_dmadev *od) c->reg_map = od->reg_map; c->vc.desc_free = omap_dma_desc_free; vchan_init(&c->vc, &od->ddev); - INIT_LIST_HEAD(&c->node); return 0; } static void omap_dma_free(struct omap_dmadev *od) { - tasklet_kill(&od->task); while (!list_empty(&od->ddev.channels)) { struct omap_chan *c = list_first_entry(&od->ddev.channels, struct omap_chan, vc.chan.device_node); @@ -1164,12 +1112,9 @@ static int omap_dma_probe(struct platform_device *pdev) od->ddev.residue_granularity = DMA_RESIDUE_GRANULARITY_BURST; od->ddev.dev = &pdev->dev; INIT_LIST_HEAD(&od->ddev.channels); - INIT_LIST_HEAD(&od->pending); spin_lock_init(&od->lock); spin_lock_init(&od->irq_lock); - tasklet_init(&od->task, omap_dma_sched, (unsigned long)od); - od->dma_requests = OMAP_SDMA_REQUESTS; if (pdev->dev.of_node && of_property_read_u32(pdev->dev.of_node, "dma-requests", -- cgit v1.2.3 From 1a7cf7b26f2594bb1c622f76765f77d3a5140293 Mon Sep 17 00:00:00 2001 From: Peter Ujfalusi Date: Wed, 11 Nov 2015 12:37:58 +0200 Subject: dmaengine: omap-dma: Handle cases when the channel is polled for completion When a DMA client driver decides that it is not providing callback for completion of a transfer (and/or does not set the DMA_PREP_INTERRUPT) but it will poll the status of the transfer (in case of short memcpy for example) we will not get interrupt for the completion of the transfer and will not mark the transaction as done. Check the channel enable bit in the CCR when the status is queried and if the channel is no longer active, we call the omap_dma_callback() to handle the transfer completion. Signed-off-by: Peter Ujfalusi Signed-off-by: Vinod Koul --- drivers/dma/omap-dma.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'drivers/dma/omap-dma.c') diff --git a/drivers/dma/omap-dma.c b/drivers/dma/omap-dma.c index 4e4642f561f5..f86827ac0c8a 100644 --- a/drivers/dma/omap-dma.c +++ b/drivers/dma/omap-dma.c @@ -673,8 +673,14 @@ static enum dma_status omap_dma_tx_status(struct dma_chan *chan, struct omap_chan *c = to_omap_dma_chan(chan); struct virt_dma_desc *vd; enum dma_status ret; + uint32_t ccr; unsigned long flags; + ccr = omap_dma_chan_read(c, CCR); + /* The channel is no longer active, handle the completion right away */ + if (!(ccr & CCR_ENABLE)) + omap_dma_callback(c->dma_ch, 0, c); + ret = dma_cookie_status(chan, cookie, txstate); if (ret == DMA_COMPLETE || !txstate) return ret; -- cgit v1.2.3