Skip to contentSkip to conversation

dmaengine: edma: add freescale edma v3 support

RFC PATCH In review Frank Li 2 participants 14 messages May 25, 2023 – May 30, 2023

Subsystems (4)
lore
Share & tools
lore

Frank Li

author maintainer
#0 lore raw

cover dmaengine: edma: add freescale edma v3 support

This patch series introduces support for the eDMA version 3 from Freescale. The eDMA v3 brings alterations in the register layout, particularly, the separation of channel control registers into different channels. The Transfer Control Descriptor (TCD) layout, however, remains identical with only the offset being changed.

The first nine patches aim at tidying up the existing Freescale eDMA code and laying the groundwork for the integration of eDMA v3 support.

Patch 1-10:
These patches primarily focus on cleaning up and refactoring the existing fsl_edma driver code. This is to accommodate the upcoming changes and new features introduced with the eDMA v3.

Patch 11:
This patch introduces support for eDMA v3. In addition, this patch has been designed with an eye towards future upgradability, specifically for transitioning to eDMA v5. The latter involves a significant upgrade where the TCD address would need to support 64 bits.

Patch 12:
This patch focuses on the device tree bindings and their modifications to properly handle and integrate the changes brought about by eDMA v3

----
include v5 TCD64 change at
https://bitbucket.sw.nxp.com/projects/IMX/repos/linux-test-testbuild/pull-requests?create&sourceBranch=refs%2Fheads%2Fedma_upstream_prep_v5

without v5 change
https://bitbucket.sw.nxp.com/projects/IMX/repos/linux-test-testbuild/pull-requests?create&sourceBranch=refs%2Fheads%2Fedma_upstream_prep_v4

These 12 patches were first sets, which try to upstream. After finished then upstream dts part.
v5 need be wait for imx95 tapeout.
----

Frank Li (12):

1   dmaengine: fsl-edma: clean up EXPORT_SYMBOL_GPL in fsl-edma-common.c
2   dmaengine: fsl-edma: clean up fsl_edma_irq_exit()
3   dmaengine: fsl-edma: transition from bool fields to bitmask flags in
     drvdata
4   dmaengine: fsl-edma: remove v3 from enum edma_version
5   dmaengine: fsl-edma: move common IRQ handler to common.c
6   dmaengine: fsl-edma: simply ATTR_DSIZE and ATTR_SSIZE by using ffs()
7   dmaengine: fsl-edma: refactor using devm_clk_get_enabled
8   dmaengine: fsl-edma: move clearing of register interrupt into
     setup_irq function
9   dmaengine: fsl-edma: refactor chan_name setup and safety
10  dmaengine: fsl-edma: move tcd into struct fsl_dma_chan
11  dmaengine: fsl-edma: integrate v3 support
12  dt-bindings: fsl-dma: fsl-edma: add edma3 compatible string
 .../devicetree/bindings/dma/fsl,edma.yaml     |  19 ++
 drivers/dma/Makefile                          |   6 +-
 drivers/dma/fsl-edma-common.c                 | 267 +++++++++++----
 drivers/dma/fsl-edma-common.h                 | 119 ++++++-
 drivers/dma/{fsl-edma.c => fsl-edma-main.c}   | 323 ++++++++++++++----
 drivers/dma/{mcf-edma.c => mcf-edma-main.c}   |  34 +-
 6 files changed, 578 insertions(+), 190 deletions(-)
 rename drivers/dma/{fsl-edma.c => fsl-edma-main.c} (62%)
 rename drivers/dma/{mcf-edma.c => mcf-edma-main.c} (91%)
Signature · 1 line
2.34.1
author maintainer
#1 lore raw

RFC PATCH 1/12 dmaengine: fsl-edma: clean up EXPORT_SYMBOL_GPL in fsl-edma-common.c

Exported functions in fsl-edma-common.c are only used within fsl-edma.c and mcf-edma.c. Global export is unnecessary.

This commit removes all EXPORT_SYMBOL_GPL in fsl-edma-common.c, and renames fsl-edma.c and mcf-edma.c to maintain the same final module names as before, thereby simplifying the codebase.

---
 drivers/dma/Makefile                        |  6 ++++--
 drivers/dma/fsl-edma-common.c               | 18 ------------------
 drivers/dma/{fsl-edma.c => fsl-edma-main.c} |  0
 drivers/dma/{mcf-edma.c => mcf-edma-main.c} |  0
 4 files changed, 4 insertions(+), 20 deletions(-)
 rename drivers/dma/{fsl-edma.c => fsl-edma-main.c} (100%)
 rename drivers/dma/{mcf-edma.c => mcf-edma-main.c} (100%)
drivers/dma/Makefile +4 -2
index a4fd1ce29510..7f3d62c9dee2 100644
@@ -32,8 +32,10 @@ obj-$(CONFIG_DW_DMAC_CORE) += dw/
 obj-$(CONFIG_DW_EDMA) += dw-edma/
 obj-$(CONFIG_EP93XX_DMA) += ep93xx_dma.o
 obj-$(CONFIG_FSL_DMA) += fsldma.o
-obj-$(CONFIG_FSL_EDMA) += fsl-edma.o fsl-edma-common.o
-obj-$(CONFIG_MCF_EDMA) += mcf-edma.o fsl-edma-common.o
+fsl-edma-objs := fsl-edma-main.o fsl-edma-common.o
+obj-$(CONFIG_FSL_EDMA) += fsl-edma.o
+mcf-edma-objs := mcf-edma-main.o fsl-edma-common.o
+obj-$(CONFIG_MCF_EDMA) += mcf-edma.o
 obj-$(CONFIG_FSL_QDMA) += fsl-qdma.o
 obj-$(CONFIG_FSL_RAID) += fsl_raid.o
 obj-$(CONFIG_HISI_DMA) += hisi_dma.o
drivers/dma/fsl-edma-common.c +0 -18
index a06a1575a2a5..ce8d7c9eaf77 100644
@@ -75,7 +75,6 @@ void fsl_edma_disable_request(struct fsl_edma_chan *fsl_chan)
 		iowrite8(EDMA_CEEI_CEEI(ch), regs->ceei);
 	}
 }
-EXPORT_SYMBOL_GPL(fsl_edma_disable_request);
 
 static void mux_configure8(struct fsl_edma_chan *fsl_chan, void __iomem *addr,
 			   u32 off, u32 slot, bool enable)
@@ -126,7 +125,6 @@ void fsl_edma_chan_mux(struct fsl_edma_chan *fsl_chan,
 	else
 		mux_configure8(fsl_chan, muxaddr, ch_off, slot, enable);
 }
-EXPORT_SYMBOL_GPL(fsl_edma_chan_mux);
 
 static unsigned int fsl_edma_get_tcd_attr(enum dma_slave_buswidth addr_width)
 {
@@ -155,7 +153,6 @@ void fsl_edma_free_desc(struct virt_dma_desc *vdesc)
 			      fsl_desc->tcd[i].ptcd);
 	kfree(fsl_desc);
 }
-EXPORT_SYMBOL_GPL(fsl_edma_free_desc);
 
 int fsl_edma_terminate_all(struct dma_chan *chan)
 {
@@ -172,7 +169,6 @@ int fsl_edma_terminate_all(struct dma_chan *chan)
 	vchan_dma_desc_free_list(&fsl_chan->vchan, &head);
 	return 0;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_terminate_all);
 
 int fsl_edma_pause(struct dma_chan *chan)
 {
@@ -188,7 +184,6 @@ int fsl_edma_pause(struct dma_chan *chan)
 	spin_unlock_irqrestore(&fsl_chan->vchan.lock, flags);
 	return 0;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_pause);
 
 int fsl_edma_resume(struct dma_chan *chan)
 {
@@ -204,7 +199,6 @@ int fsl_edma_resume(struct dma_chan *chan)
 	spin_unlock_irqrestore(&fsl_chan->vchan.lock, flags);
 	return 0;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_resume);
 
 static void fsl_edma_unprep_slave_dma(struct fsl_edma_chan *fsl_chan)
 {
@@ -265,7 +259,6 @@ int fsl_edma_slave_config(struct dma_chan *chan,
 
 	return 0;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_slave_config);
 
 static size_t fsl_edma_desc_residue(struct fsl_edma_chan *fsl_chan,
 		struct virt_dma_desc *vdesc, bool in_progress)
@@ -340,7 +333,6 @@ enum dma_status fsl_edma_tx_status(struct dma_chan *chan,
 
 	return fsl_chan->status;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_tx_status);
 
 static void fsl_edma_set_tcd_regs(struct fsl_edma_chan *fsl_chan,
 				  struct fsl_edma_hw_tcd *tcd)
@@ -520,7 +512,6 @@ struct dma_async_tx_descriptor *fsl_edma_prep_dma_cyclic(
 
 	return vchan_tx_prep(&fsl_chan->vchan, &fsl_desc->vdesc, flags);
 }
-EXPORT_SYMBOL_GPL(fsl_edma_prep_dma_cyclic);
 
 struct dma_async_tx_descriptor *fsl_edma_prep_slave_sg(
 		struct dma_chan *chan, struct scatterlist *sgl,
@@ -589,7 +580,6 @@ struct dma_async_tx_descriptor *fsl_edma_prep_slave_sg(
 
 	return vchan_tx_prep(&fsl_chan->vchan, &fsl_desc->vdesc, flags);
 }
-EXPORT_SYMBOL_GPL(fsl_edma_prep_slave_sg);
 
 struct dma_async_tx_descriptor *fsl_edma_prep_memcpy(struct dma_chan *chan,
 						     dma_addr_t dma_dst, dma_addr_t dma_src,
@@ -612,7 +602,6 @@ struct dma_async_tx_descriptor *fsl_edma_prep_memcpy(struct dma_chan *chan,
 
 	return vchan_tx_prep(&fsl_chan->vchan, &fsl_desc->vdesc, flags);
 }
-EXPORT_SYMBOL_GPL(fsl_edma_prep_memcpy);
 
 void fsl_edma_xfer_desc(struct fsl_edma_chan *fsl_chan)
 {
@@ -629,7 +618,6 @@ void fsl_edma_xfer_desc(struct fsl_edma_chan *fsl_chan)
 	fsl_chan->status = DMA_IN_PROGRESS;
 	fsl_chan->idle = false;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_xfer_desc);
 
 void fsl_edma_issue_pending(struct dma_chan *chan)
 {
@@ -649,7 +637,6 @@ void fsl_edma_issue_pending(struct dma_chan *chan)
 
 	spin_unlock_irqrestore(&fsl_chan->vchan.lock, flags);
 }
-EXPORT_SYMBOL_GPL(fsl_edma_issue_pending);
 
 int fsl_edma_alloc_chan_resources(struct dma_chan *chan)
 {
@@ -660,7 +647,6 @@ int fsl_edma_alloc_chan_resources(struct dma_chan *chan)
 				32, 0);
 	return 0;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_alloc_chan_resources);
 
 void fsl_edma_free_chan_resources(struct dma_chan *chan)
 {
@@ -683,7 +669,6 @@ void fsl_edma_free_chan_resources(struct dma_chan *chan)
 	fsl_chan->tcd_pool = NULL;
 	fsl_chan->is_sw = false;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_free_chan_resources);
 
 void fsl_edma_cleanup_vchan(struct dma_device *dmadev)
 {
@@ -695,7 +680,6 @@ void fsl_edma_cleanup_vchan(struct dma_device *dmadev)
 		tasklet_kill(&chan->vchan.task);
 	}
 }
-EXPORT_SYMBOL_GPL(fsl_edma_cleanup_vchan);
 
 /*
  * On the 32 channels Vybrid/mpc577x edma version (here called "v1"),
@@ -743,6 +727,4 @@ void fsl_edma_setup_regs(struct fsl_edma_engine *edma)
 
 	edma->regs.tcd = edma->membase + EDMA_TCD;
 }
-EXPORT_SYMBOL_GPL(fsl_edma_setup_regs);
 
-MODULE_LICENSE("GPL v2");
drivers/dma/fsl-edma-main.c renamed from drivers/dma/fsl-edma.c +0 -0
similarity index 100%
rename from drivers/dma/fsl-edma.c
rename to drivers/dma/fsl-edma-main.c
drivers/dma/mcf-edma-main.c renamed from drivers/dma/mcf-edma.c +0 -0
similarity index 100%
rename from drivers/dma/mcf-edma.c
rename to drivers/dma/mcf-edma-main.c
Signature · 1 line
2.34.1
author maintainer
#2 lore raw

RFC PATCH 2/12 dmaengine: fsl-edma: clean up fsl_edma_irq_exit()

In reply to Frank Li

In the probe function, devm_request_irq is now being used to automatically manage IRQ. It eliminates the need for manual IRQ freeing during removal.

---
 drivers/dma/fsl-edma-main.c | 12 ------------
 1 file changed, 12 deletions(-)
drivers/dma/fsl-edma-main.c +0 -12
index e40769666e39..16305bb9a2ef 100644
@@ -216,17 +216,6 @@ fsl_edma2_irq_init(struct platform_device *pdev,
 	return 0;
 }
 
-static void fsl_edma_irq_exit(
-		struct platform_device *pdev, struct fsl_edma_engine *fsl_edma)
-{
-	if (fsl_edma->txirq == fsl_edma->errirq) {
-		devm_free_irq(&pdev->dev, fsl_edma->txirq, fsl_edma);
-	} else {
-		devm_free_irq(&pdev->dev, fsl_edma->txirq, fsl_edma);
-		devm_free_irq(&pdev->dev, fsl_edma->errirq, fsl_edma);
-	}
-}
-
 static void fsl_disable_clocks(struct fsl_edma_engine *fsl_edma, int nr_clocks)
 {
 	int i;
@@ -427,7 +416,6 @@ static int fsl_edma_remove(struct platform_device *pdev)
 	struct device_node *np = pdev->dev.of_node;
 	struct fsl_edma_engine *fsl_edma = platform_get_drvdata(pdev);
 
-	fsl_edma_irq_exit(pdev, fsl_edma);
 	fsl_edma_cleanup_vchan(&fsl_edma->dma_dev);
 	of_dma_controller_free(np);
 	dma_async_device_unregister(&fsl_edma->dma_dev);
Signature · 1 line
2.34.1
author maintainer
#3 lore raw

RFC PATCH 3/12 dmaengine: fsl-edma: transition from bool fields to bitmask flags in drvdata

In reply to Frank Li

Replace individual bool fields with bitmask flags within drvdata. This will facilitate future extensions, making it easier to add more flags to accommodate new versions of the edma IP.

---
 drivers/dma/fsl-edma-common.c | 2 +-
 drivers/dma/fsl-edma-common.h | 5 +++--
 drivers/dma/fsl-edma-main.c   | 6 +++---
 3 files changed, 7 insertions(+), 6 deletions(-)
drivers/dma/fsl-edma-common.c +1 -1
index ce8d7c9eaf77..10dcc1435d37 100644
@@ -114,7 +114,7 @@ void fsl_edma_chan_mux(struct fsl_edma_chan *fsl_chan,
 	chans_per_mux = fsl_chan->edma->n_chans / dmamux_nr;
 	ch_off = fsl_chan->vchan.chan.chan_id % chans_per_mux;
 
-	if (fsl_chan->edma->drvdata->mux_swap)
+	if (fsl_chan->edma->drvdata->flags & FSL_EDMA_DRV_MUX_SWAP)
 		ch_off += endian_diff[ch_off % 4];
 
 	muxaddr = fsl_chan->edma->muxbase[ch / chans_per_mux];
drivers/dma/fsl-edma-common.h +3 -2
index 004ec4a6bc86..db137a8c558d 100644
@@ -144,11 +144,12 @@ enum edma_version {
 	v3, /* 32ch, i.mx7ulp */
 };
 
+#define FSL_EDMA_DRV_HAS_DMACLK		BIT(0)
+#define FSL_EDMA_DRV_MUX_SWAP		BIT(1)
 struct fsl_edma_drvdata {
 	enum edma_version	version;
 	u32			dmamuxs;
-	bool			has_dmaclk;
-	bool			mux_swap;
+	u32			flags;
 	int			(*setup_irq)(struct platform_device *pdev,
 					     struct fsl_edma_engine *fsl_edma);
 };
drivers/dma/fsl-edma-main.c +3 -3
index 16305bb9a2ef..f5cf95d185f8 100644
@@ -233,14 +233,14 @@ static struct fsl_edma_drvdata vf610_data = {
 static struct fsl_edma_drvdata ls1028a_data = {
 	.version = v1,
 	.dmamuxs = DMAMUX_NR,
-	.mux_swap = true,
+	.flags = FSL_EDMA_DRV_MUX_SWAP,
 	.setup_irq = fsl_edma_irq_init,
 };
 
 static struct fsl_edma_drvdata imx7ulp_data = {
 	.version = v3,
 	.dmamuxs = 1,
-	.has_dmaclk = true,
+	.flags = FSL_EDMA_DRV_HAS_DMACLK,
 	.setup_irq = fsl_edma2_irq_init,
 };
 
@@ -293,7 +293,7 @@ static int fsl_edma_probe(struct platform_device *pdev)
 	fsl_edma_setup_regs(fsl_edma);
 	regs = &fsl_edma->regs;
 
-	if (drvdata->has_dmaclk) {
+	if (drvdata->flags & FSL_EDMA_DRV_HAS_DMACLK) {
 		fsl_edma->dmaclk = devm_clk_get(&pdev->dev, "dma");
 		if (IS_ERR(fsl_edma->dmaclk)) {
 			dev_err(&pdev->dev, "Missing DMA block clock.\n");
Signature · 1 line
2.34.1
author maintainer
#4 lore raw

RFC PATCH 4/12 dmaengine: fsl-edma: remove v3 from enum edma_version

In reply to Frank Li

The new IP of edma refers to version 3 in the hardware documentation. However, the presence of "v3" in the code is merely a software indication to differentiate between certain chips. To prevent confusion in the future, this commit removes "v3" from the enum edma_version.

---
 drivers/dma/fsl-edma-common.c | 2 +-
 drivers/dma/fsl-edma-common.h | 2 +-
 drivers/dma/fsl-edma-main.c   | 4 ++--
 3 files changed, 4 insertions(+), 4 deletions(-)
drivers/dma/fsl-edma-common.c +1 -1
index 10dcc1435d37..e5c7497c1ff3 100644
@@ -120,7 +120,7 @@ void fsl_edma_chan_mux(struct fsl_edma_chan *fsl_chan,
 	muxaddr = fsl_chan->edma->muxbase[ch / chans_per_mux];
 	slot = EDMAMUX_CHCFG_SOURCE(slot);
 
-	if (fsl_chan->edma->drvdata->version == v3)
+	if (fsl_chan->edma->drvdata->flags & FSL_EDMA_DRV_CONFIG32)
 		mux_configure32(fsl_chan, muxaddr, ch_off, slot, enable);
 	else
 		mux_configure8(fsl_chan, muxaddr, ch_off, slot, enable);
drivers/dma/fsl-edma-common.h +1 -1
index db137a8c558d..2f13e687a721 100644
@@ -141,11 +141,11 @@ struct fsl_edma_desc {
 enum edma_version {
 	v1, /* 32ch, Vybrid, mpc57x, etc */
 	v2, /* 64ch Coldfire */
-	v3, /* 32ch, i.mx7ulp */
 };
 
 #define FSL_EDMA_DRV_HAS_DMACLK		BIT(0)
 #define FSL_EDMA_DRV_MUX_SWAP		BIT(1)
+#define FSL_EDMA_DRV_CONFIG32		BIT(2)
 struct fsl_edma_drvdata {
 	enum edma_version	version;
 	u32			dmamuxs;
drivers/dma/fsl-edma-main.c +2 -2
index f5cf95d185f8..31531b8bde78 100644
@@ -238,9 +238,9 @@ static struct fsl_edma_drvdata ls1028a_data = {
 };
 
 static struct fsl_edma_drvdata imx7ulp_data = {
-	.version = v3,
+	.version = v1,
 	.dmamuxs = 1,
-	.flags = FSL_EDMA_DRV_HAS_DMACLK,
+	.flags = FSL_EDMA_DRV_HAS_DMACLK | FSL_EDMA_DRV_CONFIG32,
 	.setup_irq = fsl_edma2_irq_init,
 };
 
Signature · 1 line
2.34.1
author maintainer
#5 lore raw

RFC PATCH 5/12 dmaengine: fsl-edma: move common IRQ handler to common.c

In reply to Frank Li

Move the common part of IRQ handler from fsl-edma-main.c and mcf-edma-main.c to fsl-edma-common.c. This eliminates redundant code, as the both files contains mostly identical code.

---
 drivers/dma/fsl-edma-common.c | 26 ++++++++++++++++++++++++++
 drivers/dma/fsl-edma-common.h |  7 +++++++
 drivers/dma/fsl-edma-main.c   | 30 ++----------------------------
 drivers/dma/mcf-edma-main.c   | 30 ++----------------------------
 4 files changed, 37 insertions(+), 56 deletions(-)
drivers/dma/fsl-edma-common.c +26 -0
index e5c7497c1ff3..839471c42758 100644
@@ -42,6 +42,32 @@
 
 #define EDMA_TCD		0x1000
 
+void fsl_edma_tx_chan_handler(struct fsl_edma_chan *fsl_chan)
+{
+	spin_lock(&fsl_chan->vchan.lock);
+
+	if (!fsl_chan->edesc) {
+		/* terminate_all called before */
+		spin_unlock(&fsl_chan->vchan.lock);
+		return;
+	}
+
+	if (!fsl_chan->edesc->iscyclic) {
+		list_del(&fsl_chan->edesc->vdesc.node);
+		vchan_cookie_complete(&fsl_chan->edesc->vdesc);
+		fsl_chan->edesc = NULL;
+		fsl_chan->status = DMA_COMPLETE;
+		fsl_chan->idle = true;
+	} else {
+		vchan_cyclic_callback(&fsl_chan->edesc->vdesc);
+	}
+
+	if (!fsl_chan->edesc)
+		fsl_edma_xfer_desc(fsl_chan);
+
+	spin_unlock(&fsl_chan->vchan.lock);
+}
+
 static void fsl_edma_enable_request(struct fsl_edma_chan *fsl_chan)
 {
 	struct edma_regs *regs = &fsl_chan->edma->regs;
drivers/dma/fsl-edma-common.h +7 -0
index 2f13e687a721..71e19e20f1cb 100644
@@ -223,6 +223,13 @@ static inline struct fsl_edma_desc *to_fsl_edma_desc(struct virt_dma_desc *vd)
 	return container_of(vd, struct fsl_edma_desc, vdesc);
 }
 
+static inline void fsl_edma_err_chan_handler(struct fsl_edma_chan *fsl_chan)
+{
+	fsl_chan->status = DMA_ERROR;
+	fsl_chan->idle = true;
+}
+
+void fsl_edma_tx_chan_handler(struct fsl_edma_chan *fsl_chan);
 void fsl_edma_disable_request(struct fsl_edma_chan *fsl_chan);
 void fsl_edma_chan_mux(struct fsl_edma_chan *fsl_chan,
 			unsigned int slot, bool enable);
drivers/dma/fsl-edma-main.c +2 -28
index 31531b8bde78..9c0b6fb4cb8f 100644
@@ -33,7 +33,6 @@ static irqreturn_t fsl_edma_tx_handler(int irq, void *dev_id)
 	struct fsl_edma_engine *fsl_edma = dev_id;
 	unsigned int intr, ch;
 	struct edma_regs *regs = &fsl_edma->regs;
-	struct fsl_edma_chan *fsl_chan;
 
 	intr = edma_readl(fsl_edma, regs->intl);
 	if (!intr)
@@ -42,31 +41,7 @@ static irqreturn_t fsl_edma_tx_handler(int irq, void *dev_id)
 	for (ch = 0; ch < fsl_edma->n_chans; ch++) {
 		if (intr & (0x1 << ch)) {
 			edma_writeb(fsl_edma, EDMA_CINT_CINT(ch), regs->cint);
-
-			fsl_chan = &fsl_edma->chans[ch];
-
-			spin_lock(&fsl_chan->vchan.lock);
-
-			if (!fsl_chan->edesc) {
-				/* terminate_all called before */
-				spin_unlock(&fsl_chan->vchan.lock);
-				continue;
-			}
-
-			if (!fsl_chan->edesc->iscyclic) {
-				list_del(&fsl_chan->edesc->vdesc.node);
-				vchan_cookie_complete(&fsl_chan->edesc->vdesc);
-				fsl_chan->edesc = NULL;
-				fsl_chan->status = DMA_COMPLETE;
-				fsl_chan->idle = true;
-			} else {
-				vchan_cyclic_callback(&fsl_chan->edesc->vdesc);
-			}
-
-			if (!fsl_chan->edesc)
-				fsl_edma_xfer_desc(fsl_chan);
-
-			spin_unlock(&fsl_chan->vchan.lock);
+			fsl_edma_tx_chan_handler(&fsl_edma->chans[ch]);
 		}
 	}
 	return IRQ_HANDLED;
@@ -86,8 +61,7 @@ static irqreturn_t fsl_edma_err_handler(int irq, void *dev_id)
 		if (err & (0x1 << ch)) {
 			fsl_edma_disable_request(&fsl_edma->chans[ch]);
 			edma_writeb(fsl_edma, EDMA_CERR_CERR(ch), regs->cerr);
-			fsl_edma->chans[ch].status = DMA_ERROR;
-			fsl_edma->chans[ch].idle = true;
+			fsl_edma_err_chan_handler(&fsl_edma->chans[ch]);
 		}
 	}
 	return IRQ_HANDLED;
drivers/dma/mcf-edma-main.c +2 -28
index ebd8733f72ad..af1b0e8a3021 100644
@@ -19,7 +19,6 @@ static irqreturn_t mcf_edma_tx_handler(int irq, void *dev_id)
 	struct fsl_edma_engine *mcf_edma = dev_id;
 	struct edma_regs *regs = &mcf_edma->regs;
 	unsigned int ch;
-	struct fsl_edma_chan *mcf_chan;
 	u64 intmap;
 
 	intmap = ioread32(regs->inth);
@@ -31,31 +30,7 @@ static irqreturn_t mcf_edma_tx_handler(int irq, void *dev_id)
 	for (ch = 0; ch < mcf_edma->n_chans; ch++) {
 		if (intmap & BIT(ch)) {
 			iowrite8(EDMA_MASK_CH(ch), regs->cint);
-
-			mcf_chan = &mcf_edma->chans[ch];
-
-			spin_lock(&mcf_chan->vchan.lock);
-
-			if (!mcf_chan->edesc) {
-				/* terminate_all called before */
-				spin_unlock(&mcf_chan->vchan.lock);
-				continue;
-			}
-
-			if (!mcf_chan->edesc->iscyclic) {
-				list_del(&mcf_chan->edesc->vdesc.node);
-				vchan_cookie_complete(&mcf_chan->edesc->vdesc);
-				mcf_chan->edesc = NULL;
-				mcf_chan->status = DMA_COMPLETE;
-				mcf_chan->idle = true;
-			} else {
-				vchan_cyclic_callback(&mcf_chan->edesc->vdesc);
-			}
-
-			if (!mcf_chan->edesc)
-				fsl_edma_xfer_desc(mcf_chan);
-
-			spin_unlock(&mcf_chan->vchan.lock);
+			fsl_edma_tx_chan_handler(&mcf_edma->chans[ch]);
 		}
 	}
 
@@ -76,8 +51,7 @@ static irqreturn_t mcf_edma_err_handler(int irq, void *dev_id)
 		if (err & BIT(ch)) {
 			fsl_edma_disable_request(&mcf_edma->chans[ch]);
 			iowrite8(EDMA_CERR_CERR(ch), regs->cerr);
-			mcf_edma->chans[ch].status = DMA_ERROR;
-			mcf_edma->chans[ch].idle = true;
+			fsl_edma_err_chan_handler(&mcf_edma->chans[ch]);
 		}
 	}
 
Signature · 1 line
2.34.1
author maintainer
#6 lore raw

RFC PATCH 6/12 dmaengine: fsl-edma: simply ATTR_DSIZE and ATTR_SSIZE by using ffs()

In reply to Frank Li

Removes all ATTR_DSIZE_*BIT(BYTE) and ATTR_SSIZE_*BIT(BYTE) definitions in edma. Uses ffs() instead, as it gives identical results. This simplifies the code and avoids adding more similar definitions in future V3 version.

---
 drivers/dma/fsl-edma-common.c | 21 ++++++++-------------
 drivers/dma/fsl-edma-common.h | 10 ----------
 2 files changed, 8 insertions(+), 23 deletions(-)
drivers/dma/fsl-edma-common.c +8 -13
index 839471c42758..659bef4a3b23 100644
@@ -154,18 +154,13 @@ void fsl_edma_chan_mux(struct fsl_edma_chan *fsl_chan,
 
 static unsigned int fsl_edma_get_tcd_attr(enum dma_slave_buswidth addr_width)
 {
-	switch (addr_width) {
-	case 1:
-		return EDMA_TCD_ATTR_SSIZE_8BIT | EDMA_TCD_ATTR_DSIZE_8BIT;
-	case 2:
-		return EDMA_TCD_ATTR_SSIZE_16BIT | EDMA_TCD_ATTR_DSIZE_16BIT;
-	case 4:
-		return EDMA_TCD_ATTR_SSIZE_32BIT | EDMA_TCD_ATTR_DSIZE_32BIT;
-	case 8:
-		return EDMA_TCD_ATTR_SSIZE_64BIT | EDMA_TCD_ATTR_DSIZE_64BIT;
-	default:
-		return EDMA_TCD_ATTR_SSIZE_32BIT | EDMA_TCD_ATTR_DSIZE_32BIT;
-	}
+	u32 val;
+
+	if (addr_width == DMA_SLAVE_BUSWIDTH_UNDEFINED)
+		addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+
+	val = ffs(addr_width) - 1;
+	return val | (val << 8);
 }
 
 void fsl_edma_free_desc(struct virt_dma_desc *vdesc)
@@ -623,7 +618,7 @@ struct dma_async_tx_descriptor *fsl_edma_prep_memcpy(struct dma_chan *chan,
 
 	/* To match with copy_align and max_seg_size so 1 tcd is enough */
 	fsl_edma_fill_tcd(fsl_desc->tcd[0].vtcd, dma_src, dma_dst,
-			EDMA_TCD_ATTR_SSIZE_32BYTE | EDMA_TCD_ATTR_DSIZE_32BYTE,
+			fsl_edma_get_tcd_attr(DMA_SLAVE_BUSWIDTH_32_BYTES),
 			32, len, 0, 1, 1, 32, 0, true, true, false);
 
 	return vchan_tx_prep(&fsl_chan->vchan, &fsl_desc->vdesc, flags);
drivers/dma/fsl-edma-common.h +0 -10
index 71e19e20f1cb..3991956dbc11 100644
@@ -29,16 +29,6 @@
 #define EDMA_TCD_ATTR_DMOD(x)		(((x) & GENMASK(4, 0)) << 3)
 #define EDMA_TCD_ATTR_SSIZE(x)		(((x) & GENMASK(2, 0)) << 8)
 #define EDMA_TCD_ATTR_SMOD(x)		(((x) & GENMASK(4, 0)) << 11)
-#define EDMA_TCD_ATTR_DSIZE_8BIT	0
-#define EDMA_TCD_ATTR_DSIZE_16BIT	BIT(0)
-#define EDMA_TCD_ATTR_DSIZE_32BIT	BIT(1)
-#define EDMA_TCD_ATTR_DSIZE_64BIT	(BIT(0) | BIT(1))
-#define EDMA_TCD_ATTR_DSIZE_32BYTE	(BIT(2) | BIT(0))
-#define EDMA_TCD_ATTR_SSIZE_8BIT	0
-#define EDMA_TCD_ATTR_SSIZE_16BIT	(EDMA_TCD_ATTR_DSIZE_16BIT << 8)
-#define EDMA_TCD_ATTR_SSIZE_32BIT	(EDMA_TCD_ATTR_DSIZE_32BIT << 8)
-#define EDMA_TCD_ATTR_SSIZE_64BIT	(EDMA_TCD_ATTR_DSIZE_64BIT << 8)
-#define EDMA_TCD_ATTR_SSIZE_32BYTE	(EDMA_TCD_ATTR_DSIZE_32BYTE << 8)
 
 #define EDMA_TCD_CITER_CITER(x)		((x) & GENMASK(14, 0))
 #define EDMA_TCD_BITER_BITER(x)		((x) & GENMASK(14, 0))
Signature · 1 line
2.34.1
author maintainer
#7 lore raw

RFC PATCH 7/12 dmaengine: fsl-edma: refactor using devm_clk_get_enabled

In reply to Frank Li

Use devm_clk_get_enabled in probe code to reduce error checks, thereby enhancing readability

---
 drivers/dma/fsl-edma-main.c | 19 ++-----------------
 1 file changed, 2 insertions(+), 17 deletions(-)
drivers/dma/fsl-edma-main.c +2 -17
index 9c0b6fb4cb8f..6126717fc87e 100644
@@ -268,17 +268,11 @@ static int fsl_edma_probe(struct platform_device *pdev)
 	regs = &fsl_edma->regs;
 
 	if (drvdata->flags & FSL_EDMA_DRV_HAS_DMACLK) {
-		fsl_edma->dmaclk = devm_clk_get(&pdev->dev, "dma");
+		fsl_edma->dmaclk = devm_clk_get_enabled(&pdev->dev, "dma");
 		if (IS_ERR(fsl_edma->dmaclk)) {
 			dev_err(&pdev->dev, "Missing DMA block clock.\n");
 			return PTR_ERR(fsl_edma->dmaclk);
 		}
-
-		ret = clk_prepare_enable(fsl_edma->dmaclk);
-		if (ret) {
-			dev_err(&pdev->dev, "DMA clk block failed.\n");
-			return ret;
-		}
 	}
 
 	for (i = 0; i < fsl_edma->drvdata->dmamuxs; i++) {
@@ -293,19 +287,12 @@ static int fsl_edma_probe(struct platform_device *pdev)
 		}
 
 		sprintf(clkname, "dmamux%d", i);
-		fsl_edma->muxclk[i] = devm_clk_get(&pdev->dev, clkname);
+		fsl_edma->muxclk[i] = devm_clk_get_enabled(&pdev->dev, clkname);
 		if (IS_ERR(fsl_edma->muxclk[i])) {
 			dev_err(&pdev->dev, "Missing DMAMUX block clock.\n");
 			/* on error: disable all previously enabled clks */
-			fsl_disable_clocks(fsl_edma, i);
 			return PTR_ERR(fsl_edma->muxclk[i]);
 		}
-
-		ret = clk_prepare_enable(fsl_edma->muxclk[i]);
-		if (ret)
-			/* on error: disable all previously enabled clks */
-			fsl_disable_clocks(fsl_edma, i);
-
 	}
 
 	fsl_edma->big_endian = of_property_read_bool(np, "big-endian");
@@ -366,7 +353,6 @@ static int fsl_edma_probe(struct platform_device *pdev)
 	if (ret) {
 		dev_err(&pdev->dev,
 			"Can't register Freescale eDMA engine. (%d)\n", ret);
-		fsl_disable_clocks(fsl_edma, fsl_edma->drvdata->dmamuxs);
 		return ret;
 	}
 
@@ -375,7 +361,6 @@ static int fsl_edma_probe(struct platform_device *pdev)
 		dev_err(&pdev->dev,
 			"Can't register Freescale eDMA of_dma. (%d)\n", ret);
 		dma_async_device_unregister(&fsl_edma->dma_dev);
-		fsl_disable_clocks(fsl_edma, fsl_edma->drvdata->dmamuxs);
 		return ret;
 	}
 
Signature · 1 line
2.34.1
author maintainer
#8 lore raw

RFC PATCH 8/12 dmaengine: fsl-edma: move clearing of register interrupt into setup_irq function

In reply to Frank Li

This accommodates differences in the register layout of EDMA v3 by moving the clearing of register interrupts into the platform-specific set_irq function. This should ensure better compatibility with EDMA v3.

---
 drivers/dma/fsl-edma-main.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/dma/fsl-edma-main.c +4 -1
index 6126717fc87e..cd734779f2eb 100644
@@ -113,6 +113,8 @@ fsl_edma_irq_init(struct platform_device *pdev, struct fsl_edma_engine *fsl_edma
 {
 	int ret;
 
+	edma_writel(fsl_edma, ~0, fsl_edma->regs.intl);
+
 	fsl_edma->txirq = platform_get_irq_byname(pdev, "edma-tx");
 	if (fsl_edma->txirq < 0)
 		return fsl_edma->txirq;
@@ -154,6 +156,8 @@ fsl_edma2_irq_init(struct platform_device *pdev,
 	int i, ret, irq;
 	int count;
 
+	edma_writel(fsl_edma, ~0, fsl_edma->regs.intl);
+
 	count = platform_irq_count(pdev);
 	dev_dbg(&pdev->dev, "%s Found %d interrupts\r\n", __func__, count);
 	if (count <= 2) {
@@ -313,7 +317,6 @@ static int fsl_edma_probe(struct platform_device *pdev)
 		fsl_edma_chan_mux(fsl_chan, 0, false);
 	}
 
-	edma_writel(fsl_edma, ~0, regs->intl);
 	ret = fsl_edma->drvdata->setup_irq(pdev, fsl_edma);
 	if (ret)
 		return ret;
Signature · 1 line
2.34.1
author maintainer
#9 lore raw

RFC PATCH 9/12 dmaengine: fsl-edma: refactor chan_name setup and safety

In reply to Frank Li

Relocated the setup of chan_name from setup_irq() to fsl_chan init. This change anticipates its future use in various locations.

For increased safety, sprintf has been replaced with snprintf. In addition, The size of the fsl_chan->name[] array was expanded from 16 to 32.

---
 drivers/dma/fsl-edma-common.h | 2 +-
 drivers/dma/fsl-edma-main.c   | 5 +++--
 2 files changed, 4 insertions(+), 3 deletions(-)
drivers/dma/fsl-edma-common.h +1 -1
index 3991956dbc11..39b5f903ba87 100644
@@ -116,7 +116,7 @@ struct fsl_edma_chan {
 	dma_addr_t			dma_dev_addr;
 	u32				dma_dev_size;
 	enum dma_data_direction		dma_dir;
-	char				chan_name[16];
+	char				chan_name[32];
 };
 
 struct fsl_edma_desc {
drivers/dma/fsl-edma-main.c +3 -2
index cd734779f2eb..722ebecdc894 100644
@@ -175,8 +175,6 @@ fsl_edma2_irq_init(struct platform_device *pdev,
 		if (irq < 0)
 			return -ENXIO;
 
-		sprintf(fsl_edma->chans[i].chan_name, "eDMA2-CH%02d", i);
-
 		/* The last IRQ is for eDMA err */
 		if (i == count - 1)
 			ret = devm_request_irq(&pdev->dev, irq,
@@ -305,6 +303,9 @@ static int fsl_edma_probe(struct platform_device *pdev)
 	for (i = 0; i < fsl_edma->n_chans; i++) {
 		struct fsl_edma_chan *fsl_chan = &fsl_edma->chans[i];
 
+		snprintf(fsl_chan->chan_name, sizeof(fsl_chan->chan_name), "%s-CH%02d",
+							   dev_name(&pdev->dev), i);
+
 		fsl_chan->edma = fsl_edma;
 		fsl_chan->pm_state = RUNNING;
 		fsl_chan->slave_id = 0;
Signature · 1 line
2.34.1
author maintainer
#10 lore raw

RFC PATCH 10/12 dmaengine: fsl-edma: move tcd into struct fsl_dma_chan

In reply to Frank Li

Relocates the tcd into the fsl_dma_chan structure. This adjustment reduces the need to reference back to fsl_edma_engine, paving the way for EDMA V3 support.

Unified the edma_writel and edma_writew functions for accessing TCD (Transfer Control Descriptor) registers. A new macro is added that can automatically detect whether a 32-bit or 16-bit access should be used based on the structure field definition. This provide better support 64-bit TCD with future v5 version.

---
 drivers/dma/fsl-edma-common.c | 38 +++++++++++++----------------------
 drivers/dma/fsl-edma-common.h | 22 +++++++++++++++++++-
 drivers/dma/fsl-edma-main.c   |  6 ++++--
 drivers/dma/mcf-edma-main.c   |  4 +++-
 4 files changed, 42 insertions(+), 28 deletions(-)
drivers/dma/fsl-edma-common.c +14 -24
index 659bef4a3b23..186b5c41d07f 100644
@@ -40,8 +40,6 @@
 #define EDMA64_ERRH		0x28
 #define EDMA64_ERRL		0x2c
 
-#define EDMA_TCD		0x1000
-
 void fsl_edma_tx_chan_handler(struct fsl_edma_chan *fsl_chan)
 {
 	spin_lock(&fsl_chan->vchan.lock);
@@ -285,8 +283,6 @@ static size_t fsl_edma_desc_residue(struct fsl_edma_chan *fsl_chan,
 		struct virt_dma_desc *vdesc, bool in_progress)
 {
 	struct fsl_edma_desc *edesc = fsl_chan->edesc;
-	struct edma_regs *regs = &fsl_chan->edma->regs;
-	u32 ch = fsl_chan->vchan.chan.chan_id;
 	enum dma_transfer_direction dir = edesc->dirn;
 	dma_addr_t cur_addr, dma_addr;
 	size_t len, size;
@@ -301,9 +297,9 @@ static size_t fsl_edma_desc_residue(struct fsl_edma_chan *fsl_chan,
 		return len;
 
 	if (dir == DMA_MEM_TO_DEV)
-		cur_addr = edma_readl(fsl_chan->edma, &regs->tcd[ch].saddr);
+		cur_addr = edma_read_tcdreg(fsl_chan, saddr);
 	else
-		cur_addr = edma_readl(fsl_chan->edma, &regs->tcd[ch].daddr);
+		cur_addr = edma_read_tcdreg(fsl_chan, daddr);
 
 	/* figure out the finished and calculate the residue */
 	for (i = 0; i < fsl_chan->edesc->n_tcds; i++) {
@@ -358,9 +354,6 @@ enum dma_status fsl_edma_tx_status(struct dma_chan *chan,
 static void fsl_edma_set_tcd_regs(struct fsl_edma_chan *fsl_chan,
 				  struct fsl_edma_hw_tcd *tcd)
 {
-	struct fsl_edma_engine *edma = fsl_chan->edma;
-	struct edma_regs *regs = &fsl_chan->edma->regs;
-	u32 ch = fsl_chan->vchan.chan.chan_id;
 	u16 csr = 0;
 
 	/*
@@ -369,23 +362,22 @@ static void fsl_edma_set_tcd_regs(struct fsl_edma_chan *fsl_chan,
 	 * big- or little-endian obeying the eDMA engine model endian,
 	 * and this is performed from specific edma_write functions
 	 */
-	edma_writew(edma, 0,  &regs->tcd[ch].csr);
+	edma_write_tcdreg(fsl_chan, 0, csr);
 
-	edma_writel(edma, (s32)tcd->saddr, &regs->tcd[ch].saddr);
-	edma_writel(edma, (s32)tcd->daddr, &regs->tcd[ch].daddr);
+	edma_write_tcdreg(fsl_chan, (s32)tcd->saddr, saddr);
+	edma_write_tcdreg(fsl_chan, (s32)tcd->daddr, daddr);
 
-	edma_writew(edma, (s16)tcd->attr, &regs->tcd[ch].attr);
-	edma_writew(edma, tcd->soff, &regs->tcd[ch].soff);
+	edma_write_tcdreg(fsl_chan, (s16)tcd->attr, attr);
+	edma_write_tcdreg(fsl_chan, tcd->soff, soff);
 
-	edma_writel(edma, (s32)tcd->nbytes, &regs->tcd[ch].nbytes);
-	edma_writel(edma, (s32)tcd->slast, &regs->tcd[ch].slast);
+	edma_write_tcdreg(fsl_chan, (s32)tcd->nbytes, nbytes);
+	edma_write_tcdreg(fsl_chan, (s32)tcd->slast, slast);
 
-	edma_writew(edma, (s16)tcd->citer, &regs->tcd[ch].citer);
-	edma_writew(edma, (s16)tcd->biter, &regs->tcd[ch].biter);
-	edma_writew(edma, (s16)tcd->doff, &regs->tcd[ch].doff);
+	edma_write_tcdreg(fsl_chan, (s16)tcd->citer, citer);
+	edma_write_tcdreg(fsl_chan, (s16)tcd->biter, biter);
+	edma_write_tcdreg(fsl_chan, (s16)tcd->doff, doff);
 
-	edma_writel(edma, (s32)tcd->dlast_sga,
-			&regs->tcd[ch].dlast_sga);
+	edma_write_tcdreg(fsl_chan, (s32)tcd->dlast_sga, dlast_sga);
 
 	if (fsl_chan->is_sw) {
 		csr = le16_to_cpu(tcd->csr);
@@ -393,7 +385,7 @@ static void fsl_edma_set_tcd_regs(struct fsl_edma_chan *fsl_chan,
 		tcd->csr = cpu_to_le16(csr);
 	}
 
-	edma_writew(edma, (s16)tcd->csr, &regs->tcd[ch].csr);
+	edma_write_tcdreg(fsl_chan, (s16)tcd->csr, csr);
 }
 
 static inline
@@ -745,7 +737,5 @@ void fsl_edma_setup_regs(struct fsl_edma_engine *edma)
 		edma->regs.errh = edma->membase + EDMA64_ERRH;
 		edma->regs.inth = edma->membase + EDMA64_INTH;
 	}
-
-	edma->regs.tcd = edma->membase + EDMA_TCD;
 }
 
drivers/dma/fsl-edma-common.h +21 -1
index 39b5f903ba87..0a95fd5d74f5 100644
@@ -48,6 +48,8 @@
 
 #define DMAMUX_NR	2
 
+#define EDMA_TCD                0x1000
+
 #define FSL_EDMA_BUSWIDTHS	(BIT(DMA_SLAVE_BUSWIDTH_1_BYTE) | \
 				 BIT(DMA_SLAVE_BUSWIDTH_2_BYTES) | \
 				 BIT(DMA_SLAVE_BUSWIDTH_4_BYTES) | \
@@ -93,7 +95,6 @@ struct edma_regs {
 	void __iomem *intl;
 	void __iomem *errh;
 	void __iomem *errl;
-	struct fsl_edma_hw_tcd __iomem *tcd;
 };
 
 struct fsl_edma_sw_tcd {
@@ -117,6 +118,7 @@ struct fsl_edma_chan {
 	u32				dma_dev_size;
 	enum dma_data_direction		dma_dir;
 	char				chan_name[32];
+	struct fsl_edma_hw_tcd __iomem *tcd;
 };
 
 struct fsl_edma_desc {
@@ -160,6 +162,16 @@ struct fsl_edma_engine {
 	struct fsl_edma_chan	chans[];
 };
 
+#define edma_read_tcdreg(chan, __name)				\
+(sizeof(chan->tcd->__name) == sizeof(u32) ?			\
+	edma_readl(chan->edma, &chan->tcd->__name) :		\
+	edma_readw(chan->edma, &chan->tcd->__name))
+
+#define edma_write_tcdreg(chan, val, __name)			\
+(sizeof(chan->tcd->__name) == sizeof(u32) ?			\
+	edma_writel(chan->edma, val, &chan->tcd->__name) :	\
+	edma_writew(chan->edma, val, &chan->tcd->__name))
+
 /*
  * R/W functions for big- or little-endian registers:
  * The eDMA controller's endian is independent of the CPU core's endian.
@@ -174,6 +186,14 @@ static inline u32 edma_readl(struct fsl_edma_engine *edma, void __iomem *addr)
 		return ioread32(addr);
 }
 
+static inline u16 edma_readw(struct fsl_edma_engine *edma, void __iomem *addr)
+{
+	if (edma->big_endian)
+		return ioread16be(addr);
+	else
+		return ioread16(addr);
+}
+
 static inline void edma_writeb(struct fsl_edma_engine *edma,
 			       u8 val, void __iomem *addr)
 {
drivers/dma/fsl-edma-main.c +4 -2
index 722ebecdc894..8d1d0100f6e6 100644
@@ -312,9 +312,11 @@ static int fsl_edma_probe(struct platform_device *pdev)
 		fsl_chan->idle = true;
 		fsl_chan->dma_dir = DMA_NONE;
 		fsl_chan->vchan.desc_free = fsl_edma_free_desc;
+		fsl_chan->tcd = fsl_edma->membase + EDMA_TCD
+				+ i * sizeof(struct fsl_edma_hw_tcd);
 		vchan_init(&fsl_chan->vchan, &fsl_edma->dma_dev);
 
-		edma_writew(fsl_edma, 0x0, &regs->tcd[i].csr);
+		edma_write_tcdreg(fsl_chan, 0, csr);
 		fsl_edma_chan_mux(fsl_chan, 0, false);
 	}
 
@@ -421,7 +423,7 @@ static int fsl_edma_resume_early(struct device *dev)
 	for (i = 0; i < fsl_edma->n_chans; i++) {
 		fsl_chan = &fsl_edma->chans[i];
 		fsl_chan->pm_state = RUNNING;
-		edma_writew(fsl_edma, 0x0, &regs->tcd[i].csr);
+		edma_write_tcdreg(fsl_chan, 0, csr);
 		if (fsl_chan->slave_id != 0)
 			fsl_edma_chan_mux(fsl_chan, fsl_chan->slave_id, true);
 	}
drivers/dma/mcf-edma-main.c +3 -1
index af1b0e8a3021..d79a04a03fda 100644
@@ -200,7 +200,9 @@ static int mcf_edma_probe(struct platform_device *pdev)
 		mcf_chan->dma_dir = DMA_NONE;
 		mcf_chan->vchan.desc_free = fsl_edma_free_desc;
 		vchan_init(&mcf_chan->vchan, &mcf_edma->dma_dev);
-		iowrite32(0x0, &regs->tcd[i].csr);
+		mcf_chan->tcd = mcf_edma->membase + EDMA_TCD
+				+ i * sizeof(struct fsl_edma_hw_tcd);
+		iowrite32(0x0, &mcf_chan->tcd->csr);
 	}
 
 	iowrite32(~0, regs->inth);
Signature · 1 line
2.34.1
author maintainer
#11 lore raw

RFC PATCH 11/12 dmaengine: fsl-edma: integrate v3 support

In reply to Frank Li

Significant alterations have been made to the EDMA v3's register layout. Now, each channel possesses a separate address space, encapsulating all channel-related controls and statuses, including IRQs. There are changes in bit position definitions as well. However, the fundamental control flow remains analogous to the previous versions.

EDMA v3 was utilized in imx8qm, imx93, and will be in forthcoming chips.

---
 drivers/dma/fsl-edma-common.c | 162 +++++++++++++++++++---
 drivers/dma/fsl-edma-common.h |  71 +++++++++-
 drivers/dma/fsl-edma-main.c   | 252 ++++++++++++++++++++++++++++++++--
 3 files changed, 456 insertions(+), 29 deletions(-)
drivers/dma/fsl-edma-common.c +144 -18
index 186b5c41d07f..946db5733202 100644
@@ -7,6 +7,8 @@
 #include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/dma-mapping.h>
+#include <linux/pm_runtime.h>
+#include <linux/pm_domain.h>
 
 #include "fsl-edma-common.h"
 
@@ -66,12 +68,48 @@ void fsl_edma_tx_chan_handler(struct fsl_edma_chan *fsl_chan)
 	spin_unlock(&fsl_chan->vchan.lock);
 }
 
+static void fsl_edma3_enable_request(struct fsl_edma_chan *fsl_chan)
+{
+	u32 val, flags;
+
+	flags = fsl_edma_drvflags(fsl_chan);
+	val = edma_readl_chreg(fsl_chan, ch_sbr);
+	/* Remote/local swapped wrongly on iMX8 QM Audio edma */
+	if (flags & FSL_EDMA_DRV_QUIRK_SWAPPED) {
+		if (!fsl_chan->is_rxchan)
+			val |= EDMA_V3_CH_SBR_RD;
+		else
+			val |= EDMA_V3_CH_SBR_WR;
+	} else {
+		if (fsl_chan->is_rxchan)
+			val |= EDMA_V3_CH_SBR_RD;
+		else
+			val |= EDMA_V3_CH_SBR_WR;
+	}
+
+	if (fsl_chan->is_remote)
+		val &= ~(EDMA_V3_CH_SBR_RD | EDMA_V3_CH_SBR_WR);
+
+	edma_writel_chreg(fsl_chan, val, ch_sbr);
+
+	if ((flags & (FSL_EDMA_DRV_AXI | FSL_EDMA_DRV_HAS_CHMUX)) &&
+	    fsl_chan->srcid && !edma_readl_chreg(fsl_chan, ch_mux))
+		edma_writel_chreg(fsl_chan, fsl_chan->srcid, ch_mux);
+
+	val = edma_readl_chreg(fsl_chan, ch_csr);
+	val |= EDMA_V3_CH_CSR_ERQ;
+	edma_writel_chreg(fsl_chan, val, ch_csr);
+}
+
 static void fsl_edma_enable_request(struct fsl_edma_chan *fsl_chan)
 {
 	struct edma_regs *regs = &fsl_chan->edma->regs;
 	u32 ch = fsl_chan->vchan.chan.chan_id;
 
-	if (fsl_chan->edma->drvdata->version == v1) {
+	if (fsl_edma_ver(fsl_chan) >= v3)
+		return fsl_edma3_enable_request(fsl_chan);
+
+	if (fsl_edma_ver(fsl_chan) == v1) {
 		edma_writeb(fsl_chan->edma, EDMA_SEEI_SEEI(ch), regs->seei);
 		edma_writeb(fsl_chan->edma, ch, regs->serq);
 	} else {
@@ -83,12 +121,30 @@ static void fsl_edma_enable_request(struct fsl_edma_chan *fsl_chan)
 	}
 }
 
+static void fsl_edma3_disable_request(struct fsl_edma_chan *fsl_chan)
+{
+	u32 val = edma_readl_chreg(fsl_chan, ch_csr);
+	u32 flags;
+
+	flags = fsl_edma_drvflags(fsl_chan);
+
+	if ((flags & (FSL_EDMA_DRV_AXI | FSL_EDMA_DRV_HAS_CHMUX)) &&
+	    fsl_chan->srcid)
+		edma_writel_chreg(fsl_chan, 0, ch_mux);
+
+	val &= ~EDMA_V3_CH_CSR_ERQ;
+	edma_writel_chreg(fsl_chan, val, ch_csr);
+}
+
 void fsl_edma_disable_request(struct fsl_edma_chan *fsl_chan)
 {
 	struct edma_regs *regs = &fsl_chan->edma->regs;
 	u32 ch = fsl_chan->vchan.chan.chan_id;
 
-	if (fsl_chan->edma->drvdata->version == v1) {
+	if (fsl_edma_ver(fsl_chan) >= v3)
+		return fsl_edma3_disable_request(fsl_chan);
+
+	if (fsl_edma_ver(fsl_chan) == v1) {
 		edma_writeb(fsl_chan->edma, ch, regs->cerq);
 		edma_writeb(fsl_chan->edma, EDMA_CEEI_CEEI(ch), regs->ceei);
 	} else {
@@ -186,6 +242,10 @@ int fsl_edma_terminate_all(struct dma_chan *chan)
 	vchan_get_all_descriptors(&fsl_chan->vchan, &head);
 	spin_unlock_irqrestore(&fsl_chan->vchan.lock, flags);
 	vchan_dma_desc_free_list(&fsl_chan->vchan, &head);
+
+	if (fsl_edma_drvflags(fsl_chan) & FSL_EDMA_DRV_HAS_PD)
+		pm_runtime_allow(fsl_chan->pd_dev);
+
 	return 0;
 }
 
@@ -286,12 +346,16 @@ static size_t fsl_edma_desc_residue(struct fsl_edma_chan *fsl_chan,
 	enum dma_transfer_direction dir = edesc->dirn;
 	dma_addr_t cur_addr, dma_addr;
 	size_t len, size;
+	u32 nbytes = 0;
 	int i;
 
 	/* calculate the total size in this desc */
-	for (len = i = 0; i < fsl_chan->edesc->n_tcds; i++)
-		len += le32_to_cpu(edesc->tcd[i].vtcd->nbytes)
-			* le16_to_cpu(edesc->tcd[i].vtcd->biter);
+	for (len = i = 0; i < fsl_chan->edesc->n_tcds; i++) {
+		nbytes = le32_to_cpu(edesc->tcd[i].vtcd->nbytes);
+		if (nbytes & (EDMA_V3_TCD_NBYTES_DMLOE | EDMA_V3_TCD_NBYTES_SMLOE))
+			nbytes = EDMA_V3_TCD_NBYTES_MLOFF_NBYTES(nbytes);
+		len += nbytes * le16_to_cpu(edesc->tcd[i].vtcd->biter);
+	}
 
 	if (!in_progress)
 		return len;
@@ -303,8 +367,12 @@ static size_t fsl_edma_desc_residue(struct fsl_edma_chan *fsl_chan,
 
 	/* figure out the finished and calculate the residue */
 	for (i = 0; i < fsl_chan->edesc->n_tcds; i++) {
-		size = le32_to_cpu(edesc->tcd[i].vtcd->nbytes)
-			* le16_to_cpu(edesc->tcd[i].vtcd->biter);
+		nbytes = le32_to_cpu(edesc->tcd[i].vtcd->nbytes);
+		if (nbytes & (EDMA_V3_TCD_NBYTES_DMLOE | EDMA_V3_TCD_NBYTES_SMLOE))
+			nbytes = EDMA_V3_TCD_NBYTES_MLOFF_NBYTES(nbytes);
+
+		size = nbytes * le16_to_cpu(edesc->tcd[i].vtcd->biter);
+
 		if (dir == DMA_MEM_TO_DEV)
 			dma_addr = le32_to_cpu(edesc->tcd[i].vtcd->saddr);
 		else
@@ -389,13 +457,15 @@ static void fsl_edma_set_tcd_regs(struct fsl_edma_chan *fsl_chan,
 }
 
 static inline
-void fsl_edma_fill_tcd(struct fsl_edma_hw_tcd *tcd, u32 src, u32 dst,
+void fsl_edma_fill_tcd(struct fsl_edma_chan *fsl_chan,
+		       struct fsl_edma_hw_tcd *tcd, u32 src, u32 dst,
 		       u16 attr, u16 soff, u32 nbytes, u32 slast, u16 citer,
 		       u16 biter, u16 doff, u32 dlast_sga, bool major_int,
 		       bool disable_req, bool enable_sg)
 {
+	struct dma_slave_config *cfg = &fsl_chan->cfg;
 	u16 csr = 0;
-
+	u32 burst;
 	/*
 	 * eDMA hardware SGs require the TCDs to be stored in little
 	 * endian format irrespective of the register endian model.
@@ -409,6 +479,21 @@ void fsl_edma_fill_tcd(struct fsl_edma_hw_tcd *tcd, u32 src, u32 dst,
 
 	tcd->soff = cpu_to_le16(soff);
 
+	if (fsl_chan->is_multi_fifo) {
+		/* set mloff to support multiple fifo */
+		burst = cfg->direction == DMA_DEV_TO_MEM ?
+				cfg->src_addr_width : cfg->dst_addr_width;
+		nbytes |= EDMA_V3_TCD_NBYTES_MLOFF(-(burst * 4));
+		/* enable DMLOE/SMLOE */
+		if (cfg->direction == DMA_MEM_TO_DEV) {
+			nbytes |= EDMA_V3_TCD_NBYTES_DMLOE;
+			nbytes &= ~EDMA_V3_TCD_NBYTES_SMLOE;
+		} else {
+			nbytes |= EDMA_V3_TCD_NBYTES_SMLOE;
+			nbytes &= ~EDMA_V3_TCD_NBYTES_DMLOE;
+		}
+	}
+
 	tcd->nbytes = cpu_to_le32(nbytes);
 	tcd->slast = cpu_to_le32(slast);
 
@@ -427,6 +512,12 @@ void fsl_edma_fill_tcd(struct fsl_edma_hw_tcd *tcd, u32 src, u32 dst,
 	if (enable_sg)
 		csr |= EDMA_TCD_CSR_E_SG;
 
+	if (fsl_chan->is_rxchan)
+		csr |= EDMA_TCD_CSR_ACTIVE;
+
+	if (fsl_chan->is_sw)
+		csr |= EDMA_TCD_CSR_START;
+
 	tcd->csr = cpu_to_le16(csr);
 }
 
@@ -466,6 +557,7 @@ struct dma_async_tx_descriptor *fsl_edma_prep_dma_cyclic(
 	struct fsl_edma_chan *fsl_chan = to_fsl_edma_chan(chan);
 	struct fsl_edma_desc *fsl_desc;
 	dma_addr_t dma_buf_next;
+	bool major_int = true;
 	int sg_len, i;
 	u32 src_addr, dst_addr, last_sg, nbytes;
 	u16 soff, doff, iter;
@@ -509,17 +601,23 @@ struct dma_async_tx_descriptor *fsl_edma_prep_dma_cyclic(
 			src_addr = dma_buf_next;
 			dst_addr = fsl_chan->dma_dev_addr;
 			soff = fsl_chan->cfg.dst_addr_width;
-			doff = 0;
-		} else {
+			doff = fsl_chan->is_multi_fifo ? 4 : 0;
+		} else if (direction == DMA_DEV_TO_MEM) {
 			src_addr = fsl_chan->dma_dev_addr;
 			dst_addr = dma_buf_next;
-			soff = 0;
+			soff = fsl_chan->is_multi_fifo ? 4 : 0;
 			doff = fsl_chan->cfg.src_addr_width;
+		} else {
+			/* DMA_DEV_TO_DEV */
+			src_addr = fsl_chan->cfg.src_addr;
+			dst_addr = fsl_chan->cfg.dst_addr;
+			soff = doff = 0;
+			major_int = false;
 		}
 
-		fsl_edma_fill_tcd(fsl_desc->tcd[i].vtcd, src_addr, dst_addr,
+		fsl_edma_fill_tcd(fsl_chan, fsl_desc->tcd[i].vtcd, src_addr, dst_addr,
 				  fsl_chan->attr, soff, nbytes, 0, iter,
-				  iter, doff, last_sg, true, false, true);
+				  iter, doff, last_sg, major_int, false, true);
 		dma_buf_next += period_len;
 	}
 
@@ -568,23 +666,51 @@ struct dma_async_tx_descriptor *fsl_edma_prep_slave_sg(
 			dst_addr = fsl_chan->dma_dev_addr;
 			soff = fsl_chan->cfg.dst_addr_width;
 			doff = 0;
-		} else {
+		} else if (direction == DMA_DEV_TO_MEM) {
 			src_addr = fsl_chan->dma_dev_addr;
 			dst_addr = sg_dma_address(sg);
 			soff = 0;
 			doff = fsl_chan->cfg.src_addr_width;
+		} else {
+			/* DMA_DEV_TO_DEV */
+			src_addr = fsl_chan->cfg.src_addr;
+			dst_addr = fsl_chan->cfg.dst_addr;
+			soff = 0;
+			doff = 0;
 		}
 
+		/*
+		 * Choose the suitable burst length if sg_dma_len is not
+		 * multiple of burst length so that the whole transfer length is
+		 * multiple of minor loop(burst length).
+		 */
+		if (sg_dma_len(sg) % nbytes) {
+			u32 width = (direction == DMA_DEV_TO_MEM) ? doff : soff;
+			u32 burst = (direction == DMA_DEV_TO_MEM) ?
+						fsl_chan->cfg.src_maxburst :
+						fsl_chan->cfg.dst_maxburst;
+			int j;
+
+			for (j = burst; j > 1; j--) {
+				if (!(sg_dma_len(sg) % (j * width))) {
+					nbytes = j * width;
+					break;
+				}
+			}
+			/* Set burst size as 1 if there's no suitable one */
+			if (j == 1)
+				nbytes = width;
+		}
 		iter = sg_dma_len(sg) / nbytes;
 		if (i < sg_len - 1) {
 			last_sg = fsl_desc->tcd[(i + 1)].ptcd;
-			fsl_edma_fill_tcd(fsl_desc->tcd[i].vtcd, src_addr,
+			fsl_edma_fill_tcd(fsl_chan, fsl_desc->tcd[i].vtcd, src_addr,
 					  dst_addr, fsl_chan->attr, soff,
 					  nbytes, 0, iter, iter, doff, last_sg,
 					  false, false, true);
 		} else {
 			last_sg = 0;
-			fsl_edma_fill_tcd(fsl_desc->tcd[i].vtcd, src_addr,
+			fsl_edma_fill_tcd(fsl_chan, fsl_desc->tcd[i].vtcd, src_addr,
 					  dst_addr, fsl_chan->attr, soff,
 					  nbytes, 0, iter, iter, doff, last_sg,
 					  true, true, false);
@@ -609,7 +735,7 @@ struct dma_async_tx_descriptor *fsl_edma_prep_memcpy(struct dma_chan *chan,
 	fsl_chan->is_sw = true;
 
 	/* To match with copy_align and max_seg_size so 1 tcd is enough */
-	fsl_edma_fill_tcd(fsl_desc->tcd[0].vtcd, dma_src, dma_dst,
+	fsl_edma_fill_tcd(fsl_chan, fsl_desc->tcd[0].vtcd, dma_src, dma_dst,
 			fsl_edma_get_tcd_attr(DMA_SLAVE_BUSWIDTH_32_BYTES),
 			32, len, 0, 1, 1, 32, 0, true, true, false);
 
drivers/dma/fsl-edma-common.h +70 -1

134 lines not on this page · raw message

drivers/dma/fsl-edma-main.c +242 -10

359 lines not on this page · raw message

Signature · 1 line
2.34.1
author maintainer
#12 lore raw

RFC PATCH 12/12 dt-bindings: fsl-dma: fsl-edma: add edma3 compatible string

In reply to Frank Li

Extend Freescale eDMA driver bindings to support eDMA3 IP blocks in i.MX8QM and i.MX8QXP SoCs. In i.MX93, both eDMA3 and eDMA4 are now.

---
 .../devicetree/bindings/dma/fsl,edma.yaml     | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
Documentation/devicetree/bindings/dma/fsl,edma.yaml +19 -0
index 5fd8fc604261..4315f854198e 100644
@@ -21,6 +21,10 @@ properties:
       - enum:
           - fsl,vf610-edma
           - fsl,imx7ulp-edma
+          - fsl,imx8qm-edma
+          - fsl,imx8qm-adma
+          - fsl,imx93-edma3
+          - fsl,imx93-edma4
       - items:
           - const: fsl,ls1028a-edma
           - const: fsl,vf610-edma
@@ -101,6 +105,21 @@ allOf:
         reg:
           maxItems: 2
 
+  - if:
+      properties:
+        compatible:
+          contains:
+            const: fsl,imx8qm-edma
+            const: fsl,imx8qm-adma
+            const: fsl,imx93-edma3
+            const: fsl,imx93-edma4
+    then:
+      properties:
+        reg:
+          maxItems: 1
+        interrupts:
+          maxItems: 64
+
 unevaluatedProperties: false
 
 examples:
Signature · 1 line
2.34.1
#13 lore raw

Re: [RFC PATCH 12/12] dt-bindings: fsl-dma: fsl-edma: add edma3 compatible string

Quoted 7 lines · Frank Li

Extend Freescale eDMA driver bindings to support eDMA3 IP blocks in i.MX8QM and i.MX8QXP SoCs. In i.MX93, both eDMA3 and eDMA4 are now.

---
 .../devicetree/bindings/dma/fsl,edma.yaml     | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

Please use scripts/get_maintainers.pl to get a list of necessary people and lists to CC. It might happen, that command when run on an older kernel, gives you outdated entries. Therefore please be sure you base your patches on recent Linux kernel.

You missed at least DT list (maybe more), so this won't be tested. Please resend and include all necessary entries.

Best regards,
Krzysztof