dw_dmac: allocate dma descriptors from DMA_COHERENT memory
Currently descriptors are allocated from normal cacheable memory and that slows
down filling the descriptors, as we need to call cache_coherency routines
afterwards. It would be better to allocate memory for these descriptors from
DMA_COHERENT memory. This would make code much cleaner too.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
diff --git a/drivers/dma/dw_dmac.c b/drivers/dma/dw_dmac.c
index 28d5f01..76301c4 100644
--- a/drivers/dma/dw_dmac.c
+++ b/drivers/dma/dw_dmac.c
@@ -14,6 +14,7 @@
#include <linux/delay.h>
#include <linux/dmaengine.h>
#include <linux/dma-mapping.h>
+#include <linux/dmapool.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/io.h>
@@ -91,14 +92,6 @@
/*----------------------------------------------------------------------*/
-/*
- * Because we're not relying on writeback from the controller (it may not
- * even be configured into the core!) we don't need to use dma_pool. These
- * descriptors -- and associated data -- are cacheable. We do need to make
- * sure their dcache entries are written back before handing them off to
- * the controller, though.
- */
-
static struct device *chan2dev(struct dma_chan *chan)
{
return &chan->dev->device;
@@ -137,19 +130,6 @@
return ret;
}
-static void dwc_sync_desc_for_cpu(struct dw_dma_chan *dwc, struct dw_desc *desc)
-{
- struct dw_desc *child;
-
- list_for_each_entry(child, &desc->tx_list, desc_node)
- dma_sync_single_for_cpu(chan2parent(&dwc->chan),
- child->txd.phys, sizeof(child->lli),
- DMA_TO_DEVICE);
- dma_sync_single_for_cpu(chan2parent(&dwc->chan),
- desc->txd.phys, sizeof(desc->lli),
- DMA_TO_DEVICE);
-}
-
/*
* Move a descriptor, including any children, to the free list.
* `desc' must not be on any lists.
@@ -161,8 +141,6 @@
if (desc) {
struct dw_desc *child;
- dwc_sync_desc_for_cpu(dwc, desc);
-
spin_lock_irqsave(&dwc->lock, flags);
list_for_each_entry(child, &desc->tx_list, desc_node)
dev_vdbg(chan2dev(&dwc->chan),
@@ -335,8 +313,6 @@
param = txd->callback_param;
}
- dwc_sync_desc_for_cpu(dwc, desc);
-
/* async_tx_ack */
list_for_each_entry(child, &desc->tx_list, desc_node)
async_tx_ack(&child->txd);
@@ -770,25 +746,17 @@
first = desc;
} else {
prev->lli.llp = desc->txd.phys;
- dma_sync_single_for_device(chan2parent(chan),
- prev->txd.phys, sizeof(prev->lli),
- DMA_TO_DEVICE);
list_add_tail(&desc->desc_node,
&first->tx_list);
}
prev = desc;
}
-
if (flags & DMA_PREP_INTERRUPT)
/* Trigger interrupt after last block */
prev->lli.ctllo |= DWC_CTLL_INT_EN;
prev->lli.llp = 0;
- dma_sync_single_for_device(chan2parent(chan),
- prev->txd.phys, sizeof(prev->lli),
- DMA_TO_DEVICE);
-
first->txd.flags = flags;
first->len = len;
@@ -876,10 +844,6 @@
first = desc;
} else {
prev->lli.llp = desc->txd.phys;
- dma_sync_single_for_device(chan2parent(chan),
- prev->txd.phys,
- sizeof(prev->lli),
- DMA_TO_DEVICE);
list_add_tail(&desc->desc_node,
&first->tx_list);
}
@@ -938,10 +902,6 @@
first = desc;
} else {
prev->lli.llp = desc->txd.phys;
- dma_sync_single_for_device(chan2parent(chan),
- prev->txd.phys,
- sizeof(prev->lli),
- DMA_TO_DEVICE);
list_add_tail(&desc->desc_node,
&first->tx_list);
}
@@ -961,10 +921,6 @@
prev->lli.ctllo |= DWC_CTLL_INT_EN;
prev->lli.llp = 0;
- dma_sync_single_for_device(chan2parent(chan),
- prev->txd.phys, sizeof(prev->lli),
- DMA_TO_DEVICE);
-
first->len = total_len;
return &first->txd;
@@ -1118,7 +1074,6 @@
struct dw_desc *desc;
int i;
unsigned long flags;
- int ret;
dev_vdbg(chan2dev(chan), "%s\n", __func__);
@@ -1139,21 +1094,21 @@
spin_lock_irqsave(&dwc->lock, flags);
i = dwc->descs_allocated;
while (dwc->descs_allocated < NR_DESCS_PER_CHANNEL) {
+ dma_addr_t phys;
+
spin_unlock_irqrestore(&dwc->lock, flags);
- desc = kzalloc(sizeof(struct dw_desc), GFP_KERNEL);
+ desc = dma_pool_alloc(dw->desc_pool, GFP_ATOMIC, &phys);
if (!desc)
goto err_desc_alloc;
+ memset(desc, 0, sizeof(struct dw_desc));
+
INIT_LIST_HEAD(&desc->tx_list);
dma_async_tx_descriptor_init(&desc->txd, chan);
desc->txd.tx_submit = dwc_tx_submit;
desc->txd.flags = DMA_CTRL_ACK;
- desc->txd.phys = dma_map_single(chan2parent(chan), &desc->lli,
- sizeof(desc->lli), DMA_TO_DEVICE);
- ret = dma_mapping_error(chan2parent(chan), desc->txd.phys);
- if (ret)
- goto err_desc_alloc;
+ desc->txd.phys = phys;
dwc_desc_put(dwc, desc);
@@ -1168,8 +1123,6 @@
return i;
err_desc_alloc:
- kfree(desc);
-
dev_info(chan2dev(chan), "only allocated %d descriptors\n", i);
return i;
@@ -1204,9 +1157,7 @@
list_for_each_entry_safe(desc, _desc, &list, desc_node) {
dev_vdbg(chan2dev(chan), " freeing descriptor %p\n", desc);
- dma_unmap_single(chan2parent(chan), desc->txd.phys,
- sizeof(desc->lli), DMA_TO_DEVICE);
- kfree(desc);
+ dma_pool_free(dw->desc_pool, desc, desc->txd.phys);
}
dev_vdbg(chan2dev(chan), "%s: done\n", __func__);
@@ -1451,20 +1402,14 @@
desc->lli.ctlhi = (period_len >> reg_width);
cdesc->desc[i] = desc;
- if (last) {
+ if (last)
last->lli.llp = desc->txd.phys;
- dma_sync_single_for_device(chan2parent(chan),
- last->txd.phys, sizeof(last->lli),
- DMA_TO_DEVICE);
- }
last = desc;
}
/* lets make a cyclic list */
last->lli.llp = cdesc->desc[0]->txd.phys;
- dma_sync_single_for_device(chan2parent(chan), last->txd.phys,
- sizeof(last->lli), DMA_TO_DEVICE);
dev_dbg(chan2dev(&dwc->chan), "cyclic prepared buf 0x%llx len %zu "
"period %zu periods %d\n", (unsigned long long)buf_addr,
@@ -1722,6 +1667,14 @@
platform_set_drvdata(pdev, dw);
+ /* create a pool of consistent memory blocks for hardware descriptors */
+ dw->desc_pool = dmam_pool_create("dw_dmac_desc_pool", &pdev->dev,
+ sizeof(struct dw_desc), 4, 0);
+ if (!dw->desc_pool) {
+ dev_err(&pdev->dev, "No memory for descriptors dma pool\n");
+ return -ENOMEM;
+ }
+
tasklet_init(&dw->tasklet, dw_dma_tasklet, (unsigned long)dw);
INIT_LIST_HEAD(&dw->dma.channels);
diff --git a/drivers/dma/dw_dmac_regs.h b/drivers/dma/dw_dmac_regs.h
index 577f2dd..fef296d 100644
--- a/drivers/dma/dw_dmac_regs.h
+++ b/drivers/dma/dw_dmac_regs.h
@@ -235,6 +235,7 @@
struct dw_dma {
struct dma_device dma;
void __iomem *regs;
+ struct dma_pool *desc_pool;
struct tasklet_struct tasklet;
struct clk *clk;