Skip to contentSkip to conversation

OMAP: DMA: hwmod and DMA as platform device

PATCH Superseded G, Manjunath Kondaiah 4 participants 50 messages Oct 26, 2010 – Nov 11, 2010

Subsystems (5)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v3 cover OMAP: DMA: hwmod and DMA as platform device

This patch series converts DMA library into platform device and hwmod adaptation is done for omap2+ processors.

It also replaces cpu_*is* checks by moving omap1 and omap2+ processor code into repsective mach-omapx directories. The common code in plat-omap will use device attributes for differentiating omap1 and omap2 code.

Build Report:
omap1: Success for omap_h2_1610_defconfig
omap2+: Success for omap2plus_defconfig

Test Report:
omap2+:(multiboot defconfig boots on all the omap2+ boards listed below):
OMAP4BLAZE: http://pastebin.com/HVnim30G
OMAP3630ZOOM3: http://pastebin.com/JJwrtP4F
OMAP243SDP: http://pastebin.com/mz7cVQL3

omap1 : Not tested

Test cases executed:
1. All applicable TI DMA tests which are located at:
http://dev.omapzoom.org/?p=richo/device_driver_test.git;a=tree;f=dma/test_code;h=0d00de3c0fe6933b405c62da63f694883f3e4b8f;hb=2c50a5a58dea0ffc2d31b827935aeef9b9d11253

2. Use case tests : TI MMC tests are executed with differenct types of files systems such as DOS, ext2, ext3 etc on omap4 blaze and omap3630 zoom3. More information can be found at:
http://dev.omapzoom.org/?p=richo/device_driver_test.git;a=tree;f=mmc/test_code;h=d0bc1984eef46ac45719efb02b5c1f4193422a1b;hb=2c50a5a58dea0ffc2d31b827935aeef9b9d11253

Dependencies:
This patch series has dependency on omap_device patch:
https://patchwork.kernel.org/patch/282212/

This patch series applies on:
git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap-2.6.git
Branch: master
commit a83d12a47c9a8c78a184910150797045d69a1570
Merge: f9f47f5 c1fb8df
Author: Tony Lindgren <[email protected]>
Date: Fri Oct 22 11:21:08 2010 -0700

Linux-omap rebuilt: Updated to v2.6.36, add 24xx uart fix

Changelist summary:
v3:
Review comments fixed:
http://www.listware.net/201008/linux-omap/89002-patch-v2-0011-omap-dma-hwmod-and-dma-as-platform-driver.html
- created and tested on latest linux-omap master branch(2.6.36 kernel)

v2:
The review comments which are fixed can be found at:
http://www.spinics.net/lists/linux-omap/msg34291.html
http://www.spinics.net/lists/linux-omap/msg34292.html
http://www.spinics.net/lists/linux-omap/msg34078.html
http://www.spinics.net/lists/linux-omap/msg34083.html

v1:
These changes are based on comments received during internal discussions which has changes(compared to previous version) such as:
- Code optimization
- Patches are rearranged in more meaningful way so that git bisect works at any
intermediate patch in the series.
- Build tested for all omap's(OMAP1 and OMAP2PLUS)
- Boot tested for OMAP3 and OMAP4(appreciate if some one tests on OMAP1/2)
- Applicalbe tests are executed on OMAP3 and OMAP4 boards
- Rebased and added descriptor autoloading feature(only for omap3630 and omap4)

Benoit Cousson (1):
OMAP4: DMA: hwmod: add system DMA

G, Manjunath Kondaiah (12):
OMAP: DMA: Replace read/write macros with functions
OMAP: DMA: Introduce errata handling feature
OMAP: DMA: Introduce DMA device attributes
OMAP2420: DMA: hwmod: add system DMA
OMAP2430: DMA: hwmod: add system DMA
OMAP3: DMA: hwmod: add system DMA
OMAP1: DMA: Introduce DMA driver as platform device OMAP2+: DMA: hwmod: Device registration
OMAP: DMA: Convert DMA library into DMA platform Driver
OMAP: DMA: Use DMA device attributes
sDMA: descriptor autoloading feature
OMAP: PM: DMA: Enable runtime pm

 arch/arm/mach-omap1/Makefile               |    2 +-
 arch/arm/mach-omap1/dma.c                  |  665 ++++++++++
 arch/arm/mach-omap1/include/mach/dma.h     |   51 +
 arch/arm/mach-omap2/Makefile               |    2 +-
 arch/arm/mach-omap2/dma.c                  | 1579 ++++++++++++++++++++++++
 arch/arm/mach-omap2/include/mach/dma.h     |   86 ++
 arch/arm/mach-omap2/omap_hwmod_2420_data.c |   85 ++
 arch/arm/mach-omap2/omap_hwmod_2430_data.c |   85 ++
 arch/arm/mach-omap2/omap_hwmod_3xxx_data.c |   93 ++
 arch/arm/mach-omap2/omap_hwmod_44xx_data.c |   99 ++
 arch/arm/plat-omap/dma.c                   | 1850 ++++------------------------
 arch/arm/plat-omap/include/plat/dma.h      |  387 ++++---
 12 files changed, 3207 insertions(+), 1777 deletions(-)
 create mode 100644 arch/arm/mach-omap1/dma.c
 create mode 100644 arch/arm/mach-omap1/include/mach/dma.h
 create mode 100644 arch/arm/mach-omap2/dma.c
 create mode 100644 arch/arm/mach-omap2/include/mach/dma.h
Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
#1 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

The low level read/write macros are replaced with static inline functions and register offsets are handled through static register offset tables mapped through enumeration constants.

The objective of this patch is to prepare for omap dma driver cleanup and dma hwmod implementation. The code optimization and moving machine specific code to respective mach-omapx dma file will be handled in the rest of this patch series.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/dma.c              |  739 ++++++++++++++++++++++-----------
 arch/arm/plat-omap/include/plat/dma.h |  120 ------
 2 files changed, 497 insertions(+), 362 deletions(-)
arch/arm/plat-omap/dma.c +497 -242
index f5c5b8d..77241e2 100644
@@ -40,6 +40,147 @@
 
 #undef DEBUG
 
+enum {
+	GCR1 = 0,	GSCR,		GRST,		HW_ID,
+	PCH2_ID,	PCH0_ID,	PCH1_ID,	PCHG_ID,
+	PCHD_ID,	CAPS1_0_U,	CAPS1_0_L,	CAPS1_1_U,
+	CAPS1_1_L,	CAPS1_2,	CAPS1_3,	CAPS1_4,
+	PCH2_SR,	PCH0_SR,	PCH1_SR,	PCHD_SR,
+
+	OMAP1_CH_COMMON_START,
+
+	/* Common Registers */
+	CSDP1,		CCR1,		CICR1,		CSR1,
+	CEN1,		CFN1,		CSFI1,		CSEI1,
+	CPC,		CSAC1,		CDAC1,		CDEI1,
+	CDFI1,		CLNK_CTRL1,
+
+	/* Channel specific register offsets */
+	CSSA_L,		CSSA_U,		CDSA_L,		CDSA_U,
+	COLOR_L,	COLOR_U,	CCR1_2,		LCH_CTRL,
+
+	OMAP1_CH_COMMON_END,
+};
+
+static u16 reg_map_omap1[] = {
+	[GCR1]		= 0x400,
+	[GSCR]		= 0x404,
+	[GRST]		= 0x408,
+	[HW_ID]		= 0x442,
+	[PCH2_ID]	= 0x444,
+	[PCH0_ID]	= 0x446,
+	[PCH1_ID]	= 0x448,
+	[PCHG_ID]	= 0x44a,
+	[PCHD_ID]	= 0x44c,
+	[CAPS1_0_U]	= 0x44e,
+	[CAPS1_0_L]	= 0x450,
+	[CAPS1_1_U]	= 0x452,
+	[CAPS1_1_L]	= 0x454,
+	[CAPS1_2]	= 0x456,
+	[CAPS1_3]	= 0x458,
+	[CAPS1_4]	= 0x45a,
+	[PCH2_SR]	= 0x460,
+	[PCH0_SR]	= 0x480,
+	[PCH1_SR]	= 0x482,
+	[PCHD_SR]	= 0x4c0,
+
+	/* Common Registers */
+	[CSDP1]		= 0x00,
+	[CCR1]		= 0x02,
+	[CICR1]		= 0x04,
+	[CSR1]		= 0x06,
+	[CEN1]		= 0x10,
+	[CFN1]		= 0x12,
+	[CSFI1]		= 0x14,
+	[CSEI1]		= 0x16,
+	[CPC]		= 0x18,	/* 15xx only */
+	[CSAC1]		= 0x18,
+	[CDAC1]		= 0x1a,
+	[CDEI1]		= 0x1c,
+	[CDFI1]		= 0x1e,
+	[CLNK_CTRL1]	= 0x28,
+
+	/* Channel specific register offsets */
+	[CSSA_L]	= 0x08,
+	[CSSA_U]	= 0x0a,
+	[CDSA_L]	= 0x0c,
+	[CDSA_U]	= 0x0e,
+	[COLOR_L]	= 0x20,
+	[COLOR_U]	= 0x22,
+	[CCR1_2]	= 0x24,
+	[LCH_CTRL]	= 0x2a,
+};
+
+enum {
+	REVISION = 0,	GCR2,		IRQSTATUS_L0,	IRQSTATUS_L1,
+	IRQSTATUS_L2,	IRQSTATUS_L3,	IRQENABLE_L0,	IRQENABLE_L1,
+	IRQENABLE_L2,	IRQENABLE_L3,	SYSSTATUS,	OCP_SYSCONFIG,
+	CAPS2_0,	CAPS2_2,	CAPS2_3,	CAPS2_4,
+
+	OMAP2_CH_COMMON_START,
+
+	/* Common register offsets */
+	CCR2,		CLNK_CTRL2,	CICR2,		CSR2,
+	CSDP2,		CEN2,		CFN2,		CSEI2,
+	CSFI2,		CDEI2,		CDFI2,		CSAC2,
+	CDAC2,
+
+	/* Channel specific register offsets */
+	CSSA,		CDSA,		CCEN,		CCFN,
+	COLOR,
+
+	/* OMAP4 specific registers */
+	CDP,		CNDP,		CCDN,
+
+	OMAP2_CH_COMMON_END,
+};
+
+static u16 reg_map_omap2[] = {
+	[REVISION]		= 0x00,
+	[GCR2]			= 0x78,
+	[IRQSTATUS_L0]		= 0x08,
+	[IRQSTATUS_L1]		= 0x0c,
+	[IRQSTATUS_L2]		= 0x10,
+	[IRQSTATUS_L3]		= 0x14,
+	[IRQENABLE_L0]		= 0x18,
+	[IRQENABLE_L1]		= 0x1c,
+	[IRQENABLE_L2]		= 0x20,
+	[IRQENABLE_L3]		= 0x24,
+	[SYSSTATUS]		= 0x28,
+	[OCP_SYSCONFIG]		= 0x2c,
+	[CAPS2_0]		= 0x64,
+	[CAPS2_2]		= 0x6c,
+	[CAPS2_3]		= 0x70,
+	[CAPS2_4]		= 0x74,
+
+	/* Common register offsets */
+	[CCR2]			= 0x80,
+	[CLNK_CTRL2]		= 0x84,
+	[CICR2]			= 0x88,
+	[CSR2]			= 0x8c,
+	[CSDP2]			= 0x90,
+	[CEN2]			= 0x94,
+	[CFN2]			= 0x98,
+	[CSEI2]			= 0xa4,
+	[CSFI2]			= 0xa8,
+	[CDEI2]			= 0xac,
+	[CDFI2]			= 0xb0,
+	[CSAC2]			= 0xb4,
+	[CDAC2]			= 0xb8,
+
+	/* Channel specific register offsets */
+	[CSSA]			= 0x9c,
+	[CDSA]			= 0xa0,
+	[CCEN]			= 0xbc,
+	[CCFN]			= 0xc0,
+	[COLOR]			= 0xc4,
+
+	/* OMAP4 specific registers */
+	[CDP]			= 0xd0,
+	[CNDP]			= 0xd4,
+	[CCDN]			= 0xd8,
+};
+
 #ifndef CONFIG_ARCH_OMAP1
 enum { DMA_CH_ALLOC_DONE, DMA_CH_PARAMS_SET_DONE, DMA_CH_STARTED,
 	DMA_CH_QUEUED, DMA_CH_NOTSTARTED, DMA_CH_PAUSED, DMA_CH_LINK_ENABLED
@@ -137,7 +278,7 @@ static int omap_dma_reserve_channels;
 
 static spinlock_t dma_chan_lock;
 static struct omap_dma_lch *dma_chan;
-static void __iomem *omap_dma_base;
+static void __iomem *dma_base;
 
 static const u8 omap1_dma_irq[OMAP1_LOGICAL_DMA_CH_COUNT] = {
 	INT_DMA_CH0_6, INT_DMA_CH1_7, INT_DMA_CH2_8, INT_DMA_CH3,
@@ -154,23 +295,42 @@ static inline void omap_enable_channel_irq(int lch);
 #define REVISIT_24XX()		printk(KERN_ERR "FIXME: no %s on 24xx\n", \
 						__func__);
 
-#define dma_read(reg)							\
-({									\
-	u32 __val;							\
-	if (cpu_class_is_omap1())					\
-		__val = __raw_readw(omap_dma_base + OMAP1_DMA_##reg);	\
-	else								\
-		__val = __raw_readl(omap_dma_base + OMAP_DMA4_##reg);	\
-	__val;								\
-})
-
-#define dma_write(val, reg)						\
-({									\
-	if (cpu_class_is_omap1())					\
-		__raw_writew((u16)(val), omap_dma_base + OMAP1_DMA_##reg); \
-	else								\
-		__raw_writel((val), omap_dma_base + OMAP_DMA4_##reg);	\
-})
+static inline void dma_write(u32 val, int reg, int lch)
+{
+	if (cpu_class_is_omap1()) {
+		if (reg > OMAP1_CH_COMMON_START)
+			__raw_writew(val, dma_base +
+				(reg_map_omap1[reg] + 0x40 * lch));
+		else
+			__raw_writew(val, dma_base + reg_map_omap1[reg]);
+	} else {
+		if (reg > OMAP2_CH_COMMON_START)
+			__raw_writel(val, dma_base +
+				(reg_map_omap2[reg] + 0x60 * lch));
+		else
+			__raw_writel(val, dma_base + reg_map_omap2[reg]);
+	}
+}
+
+static inline u32 dma_read(int reg, int lch)
+{
+	u32 val;
+
+	if (cpu_class_is_omap1()) {
+		if (reg > OMAP1_CH_COMMON_START)
+			val = __raw_readw(dma_base +
+					(reg_map_omap1[reg] + 0x40 * lch));
+		else
+			val = __raw_readw(dma_base + reg_map_omap1[reg]);
+	} else {
+		if (reg > OMAP2_CH_COMMON_START)
+			val = __raw_readl(dma_base +
+					(reg_map_omap2[reg] + 0x60 * lch));
+		else
+			val = __raw_readl(dma_base + reg_map_omap2[reg]);
+	}
+	return val;
+}
 
 #ifdef CONFIG_ARCH_OMAP15XX
 /* Returns 1 if the DMA module is in OMAP1510-compatible mode, 0 otherwise */
@@ -209,11 +369,10 @@ static inline void set_gdma_dev(int req, int dev)
 /* Omap1 only */
 static void clear_lch_regs(int lch)
 {
-	int i;
-	void __iomem *lch_base = omap_dma_base + OMAP1_DMA_CH_BASE(lch);
+	int i = OMAP1_CH_COMMON_START;
 
-	for (i = 0; i < 0x2c; i += 2)
-		__raw_writew(0, lch_base + i);
+	for (; i <= OMAP1_CH_COMMON_END; i += 1)
+		dma_write(0, i, lch);
 }
 
 void omap_set_dma_priority(int lch, int dst_port, int priority)
@@ -248,12 +407,12 @@ void omap_set_dma_priority(int lch, int dst_port, int priority)
 	if (cpu_class_is_omap2()) {
 		u32 ccr;
 
-		ccr = dma_read(CCR(lch));
+		ccr = dma_read(CCR2, lch);
 		if (priority)
 			ccr |= (1 << 6);
 		else
 			ccr &= ~(1 << 6);
-		dma_write(ccr, CCR(lch));
+		dma_write(ccr, CCR2, lch);
 	}
 }
 EXPORT_SYMBOL(omap_set_dma_priority);
@@ -264,31 +423,36 @@ void omap_set_dma_transfer_params(int lch, int data_type, int elem_count,
 {
 	u32 l;
 
-	l = dma_read(CSDP(lch));
-	l &= ~0x03;
-	l |= data_type;
-	dma_write(l, CSDP(lch));
-
 	if (cpu_class_is_omap1()) {
 		u16 ccr;
 
-		ccr = dma_read(CCR(lch));
+		l = dma_read(CSDP1, lch);
+		l &= ~0x03;
+		l |= data_type;
+		dma_write(l, CSDP1, lch);
+
+		ccr = dma_read(CCR1, lch);
 		ccr &= ~(1 << 5);
 		if (sync_mode == OMAP_DMA_SYNC_FRAME)
 			ccr |= 1 << 5;
-		dma_write(ccr, CCR(lch));
+		dma_write(ccr, CCR1, lch);
 
-		ccr = dma_read(CCR2(lch));
+		ccr = dma_read(CCR1_2, lch);
 		ccr &= ~(1 << 2);
 		if (sync_mode == OMAP_DMA_SYNC_BLOCK)
 			ccr |= 1 << 2;
-		dma_write(ccr, CCR2(lch));
+		dma_write(ccr, CCR1_2, lch);
 	}
 
 	if (cpu_class_is_omap2() && dma_trigger) {
 		u32 val;
 
-		val = dma_read(CCR(lch));
+		l = dma_read(CSDP2, lch);
+		l &= ~0x03;
+		l |= data_type;
+		dma_write(l, CSDP2, lch);
+
+		val = dma_read(CCR2, lch);
 
 		/* DMA_SYNCHRO_CONTROL_UPPER depends on the channel number */
 		val &= ~((1 << 23) | (3 << 19) | 0x1f);
@@ -313,11 +477,15 @@ void omap_set_dma_transfer_params(int lch, int data_type, int elem_count,
 		} else {
 			val &= ~(1 << 24);	/* dest synch */
 		}
-		dma_write(val, CCR(lch));
+		dma_write(val, CCR2, lch);
+	}
+	if (cpu_class_is_omap1()) {
+		dma_write(elem_count, CEN1, lch);
+		dma_write(frame_count, CFN1, lch);
+	} else {
+		dma_write(elem_count, CEN2, lch);
+		dma_write(frame_count, CFN2, lch);
 	}
-
-	dma_write(elem_count, CEN(lch));
-	dma_write(frame_count, CFN(lch));
 }
 EXPORT_SYMBOL(omap_set_dma_transfer_params);
 
@@ -328,7 +496,7 @@ void omap_set_dma_color_mode(int lch, enum omap_dma_color_mode mode, u32 color)
 	if (cpu_class_is_omap1()) {
 		u16 w;
 
-		w = dma_read(CCR2(lch));
+		w = dma_read(CCR1_2, lch);
 		w &= ~0x03;
 
 		switch (mode) {
@@ -343,23 +511,23 @@ void omap_set_dma_color_mode(int lch, enum omap_dma_color_mode mode, u32 color)
 		default:
 			BUG();
 		}
-		dma_write(w, CCR2(lch));
+		dma_write(w, CCR1_2, lch);
 
-		w = dma_read(LCH_CTRL(lch));
+		w = dma_read(LCH_CTRL, lch);
 		w &= ~0x0f;
 		/* Default is channel type 2D */
 		if (mode) {
-			dma_write((u16)color, COLOR_L(lch));
-			dma_write((u16)(color >> 16), COLOR_U(lch));
+			dma_write((u16)color, COLOR_L, lch);
+			dma_write((u16)(color >> 16), COLOR_U, lch);
 			w |= 1;		/* Channel type G */
 		}
-		dma_write(w, LCH_CTRL(lch));
+		dma_write(w, LCH_CTRL, lch);
 	}
 
 	if (cpu_class_is_omap2()) {
 		u32 val;
 
-		val = dma_read(CCR(lch));
+		val = dma_read(CCR2, lch);
 		val &= ~((1 << 17) | (1 << 16));
 
 		switch (mode) {
@@ -374,10 +542,10 @@ void omap_set_dma_color_mode(int lch, enum omap_dma_color_mode mode, u32 color)
 		default:
 			BUG();
 		}
-		dma_write(val, CCR(lch));
+		dma_write(val, CCR2, lch);
 
 		color &= 0xffffff;
-		dma_write(color, COLOR(lch));
+		dma_write(color, COLOR, lch);
 	}
 }
 EXPORT_SYMBOL(omap_set_dma_color_mode);
@@ -387,10 +555,10 @@ void omap_set_dma_write_mode(int lch, enum omap_dma_write_mode mode)
 	if (cpu_class_is_omap2()) {
 		u32 csdp;
 
-		csdp = dma_read(CSDP(lch));
+		csdp = dma_read(CSDP2, lch);
 		csdp &= ~(0x3 << 16);
 		csdp |= (mode << 16);
-		dma_write(csdp, CSDP(lch));
+		dma_write(csdp, CSDP2, lch);
 	}
 }
 EXPORT_SYMBOL(omap_set_dma_write_mode);
@@ -400,10 +568,10 @@ void omap_set_dma_channel_mode(int lch, enum omap_dma_channel_mode mode)
 	if (cpu_class_is_omap1() && !cpu_is_omap15xx()) {
 		u32 l;
 
-		l = dma_read(LCH_CTRL(lch));
+		l = dma_read(LCH_CTRL, lch);
 		l &= ~0x7;
 		l |= mode;
-		dma_write(l, LCH_CTRL(lch));
+		dma_write(l, LCH_CTRL, lch);
 	}
 }
 EXPORT_SYMBOL(omap_set_dma_channel_mode);
@@ -418,27 +586,38 @@ void omap_set_dma_src_params(int lch, int src_port, int src_amode,
 	if (cpu_class_is_omap1()) {
 		u16 w;
 
-		w = dma_read(CSDP(lch));
+		w = dma_read(CSDP1, lch);
 		w &= ~(0x1f << 2);
 		w |= src_port << 2;
-		dma_write(w, CSDP(lch));
-	}
+		dma_write(w, CSDP1, lch);
+
+		l = dma_read(CCR1, lch);
+	} else
+		l = dma_read(CCR2, lch);
 
-	l = dma_read(CCR(lch));
 	l &= ~(0x03 << 12);
 	l |= src_amode << 12;
-	dma_write(l, CCR(lch));
+
+	if (cpu_class_is_omap1())
+		dma_write(l, CCR1, lch);
+	else
+		dma_write(l, CCR2, lch);
 
 	if (cpu_class_is_omap1()) {
-		dma_write(src_start >> 16, CSSA_U(lch));
-		dma_write((u16)src_start, CSSA_L(lch));
+		dma_write(src_start >> 16, CSSA_U, lch);
+		dma_write((u16)src_start, CSSA_L, lch);
 	}
 
 	if (cpu_class_is_omap2())
-		dma_write(src_start, CSSA(lch));
+		dma_write(src_start, CSSA, lch);
 
-	dma_write(src_ei, CSEI(lch));
-	dma_write(src_fi, CSFI(lch));
+	if (cpu_class_is_omap1()) {
+		dma_write(src_ei, CSEI1, lch);
+		dma_write(src_fi, CSFI1, lch);
+	} else {
+		dma_write(src_ei, CSEI2, lch);
+		dma_write(src_fi, CSFI2, lch);
+	}
 }
 EXPORT_SYMBOL(omap_set_dma_src_params);
 
@@ -466,8 +645,8 @@ void omap_set_dma_src_index(int lch, int eidx, int fidx)
 	if (cpu_class_is_omap2())
 		return;
 
-	dma_write(eidx, CSEI(lch));
-	dma_write(fidx, CSFI(lch));
+	dma_write(eidx, CSEI1, lch);
+	dma_write(fidx, CSFI1, lch);
 }
 EXPORT_SYMBOL(omap_set_dma_src_index);
 
@@ -475,11 +654,19 @@ void omap_set_dma_src_data_pack(int lch, int enable)
 {
 	u32 l;
 
-	l = dma_read(CSDP(lch));
+	if (cpu_class_is_omap1())
+		l = dma_read(CSDP1, lch);
+	else
+		l = dma_read(CSDP2, lch);
+
 	l &= ~(1 << 6);
 	if (enable)
 		l |= (1 << 6);
-	dma_write(l, CSDP(lch));
+
+	if (cpu_class_is_omap1())
+		dma_write(l, CSDP1, lch);
+	else
+		dma_write(l, CSDP2, lch);
 }
 EXPORT_SYMBOL(omap_set_dma_src_data_pack);
 
@@ -488,7 +675,11 @@ void omap_set_dma_src_burst_mode(int lch, enum omap_dma_burst_mode burst_mode)
 	unsigned int burst = 0;
 	u32 l;
 
-	l = dma_read(CSDP(lch));
+	if (cpu_class_is_omap2())
+		l = dma_read(CSDP2, lch);
+	else
+		l = dma_read(CSDP1, lch);
+
 	l &= ~(0x03 << 7);
 
 	switch (burst_mode) {
@@ -524,7 +715,11 @@ void omap_set_dma_src_burst_mode(int lch, enum omap_dma_burst_mode burst_mode)
 	}
 
 	l |= (burst << 7);
-	dma_write(l, CSDP(lch));
+
+	if (cpu_class_is_omap2())
+		dma_write(l, CSDP2, lch);
+	else
+		dma_write(l, CSDP1, lch);
 }
 EXPORT_SYMBOL(omap_set_dma_src_burst_mode);
 
@@ -536,27 +731,38 @@ void omap_set_dma_dest_params(int lch, int dest_port, int dest_amode,
 	u32 l;
 
 	if (cpu_class_is_omap1()) {
-		l = dma_read(CSDP(lch));
+		l = dma_read(CSDP1, lch);
 		l &= ~(0x1f << 9);
 		l |= dest_port << 9;
-		dma_write(l, CSDP(lch));
-	}
+		dma_write(l, CSDP1, lch);
+
+		l = dma_read(CCR1, lch);
+	} else
+		l = dma_read(CCR2, lch);
 
-	l = dma_read(CCR(lch));
 	l &= ~(0x03 << 14);
 	l |= dest_amode << 14;
-	dma_write(l, CCR(lch));
+
+	if (cpu_class_is_omap1())
+		dma_write(l, CCR1, lch);
+	else
+		dma_write(l, CCR2, lch);
 
 	if (cpu_class_is_omap1()) {
-		dma_write(dest_start >> 16, CDSA_U(lch));
-		dma_write(dest_start, CDSA_L(lch));
+		dma_write(dest_start >> 16, CDSA_U, lch);
+		dma_write(dest_start, CDSA_L, lch);
 	}
 
 	if (cpu_class_is_omap2())
-		dma_write(dest_start, CDSA(lch));
+		dma_write(dest_start, CDSA, lch);
 
-	dma_write(dst_ei, CDEI(lch));
-	dma_write(dst_fi, CDFI(lch));
+	if (cpu_class_is_omap1()) {
+		dma_write(dst_ei, CDEI1, lch);
+		dma_write(dst_fi, CDFI1, lch);
+	} else {
+		dma_write(dst_ei, CDEI2, lch);
+		dma_write(dst_fi, CDFI2, lch);
+	}
 }
 EXPORT_SYMBOL(omap_set_dma_dest_params);
 
@@ -565,8 +771,8 @@ void omap_set_dma_dest_index(int lch, int eidx, int fidx)
 	if (cpu_class_is_omap2())
 		return;
 
-	dma_write(eidx, CDEI(lch));
-	dma_write(fidx, CDFI(lch));
+	dma_write(eidx, CDEI1, lch);
+	dma_write(fidx, CDFI1, lch);
 }
 EXPORT_SYMBOL(omap_set_dma_dest_index);
 
@@ -574,11 +780,19 @@ void omap_set_dma_dest_data_pack(int lch, int enable)
 {
 	u32 l;
 
-	l = dma_read(CSDP(lch));
+	if (cpu_class_is_omap2())
+		l = dma_read(CSDP2, lch);
+	else
+		l = dma_read(CSDP1, lch);
+
 	l &= ~(1 << 13);
 	if (enable)
 		l |= 1 << 13;
-	dma_write(l, CSDP(lch));
+
+	if (cpu_class_is_omap2())
+		dma_write(l, CSDP2, lch);
+	else
+		dma_write(l, CSDP1, lch);
 }
 EXPORT_SYMBOL(omap_set_dma_dest_data_pack);
 
@@ -586,8 +800,11 @@ void omap_set_dma_dest_burst_mode(int lch, enum omap_dma_burst_mode burst_mode)
 {
 	unsigned int burst = 0;
 	u32 l;
+	if (cpu_class_is_omap2())
+		l = dma_read(CSDP2, lch);
+	else
+		l = dma_read(CSDP1, lch);
 
-	l = dma_read(CSDP(lch));
 	l &= ~(0x03 << 14);
 
 	switch (burst_mode) {
@@ -619,8 +836,13 @@ void omap_set_dma_dest_burst_mode(int lch, enum omap_dma_burst_mode burst_mode)
 		BUG();
 		return;
 	}
+
 	l |= (burst << 14);
-	dma_write(l, CSDP(lch));
+
+	if (cpu_class_is_omap2())
+		dma_write(l, CSDP2, lch);
+	else
+		dma_write(l, CSDP1, lch);
 }
 EXPORT_SYMBOL(omap_set_dma_dest_burst_mode);
 
@@ -629,19 +851,21 @@ static inline void omap_enable_channel_irq(int lch)
 	u32 status;
 
 	/* Clear CSR */
-	if (cpu_class_is_omap1())
-		status = dma_read(CSR(lch));
-	else if (cpu_class_is_omap2())
-		dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR(lch));
-
-	/* Enable some nice interrupts. */
-	dma_write(dma_chan[lch].enabled_irqs, CICR(lch));
+	if (cpu_class_is_omap1()) {
+		status = dma_read(CSR1, lch);
+		dma_write(dma_chan[lch].enabled_irqs, CICR1, lch);
+	} else if (cpu_class_is_omap2()) {
+		dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR2, lch);
+		dma_write(dma_chan[lch].enabled_irqs, CICR2, lch);
+	}
 }
 
 static void omap_disable_channel_irq(int lch)
 {
 	if (cpu_class_is_omap2())
-		dma_write(0, CICR(lch));
+		dma_write(0, CICR2, lch);
+	else
+		dma_write(0, CICR1, lch);
 }
 
 void omap_enable_dma_irq(int lch, u16 bits)
@@ -660,44 +884,44 @@ static inline void enable_lnk(int lch)
 {
 	u32 l;
 
-	l = dma_read(CLNK_CTRL(lch));
-
-	if (cpu_class_is_omap1())
+	if (cpu_class_is_omap1()) {
+		l = dma_read(CLNK_CTRL1, lch);
 		l &= ~(1 << 14);
+	} else
+		l = dma_read(CLNK_CTRL2, lch);
 
 	/* Set the ENABLE_LNK bits */
 	if (dma_chan[lch].next_lch != -1)
 		l = dma_chan[lch].next_lch | (1 << 15);
 
-#ifndef CONFIG_ARCH_OMAP1
-	if (cpu_class_is_omap2())
+	if (cpu_class_is_omap2()) {
 		if (dma_chan[lch].next_linked_ch != -1)
 			l = dma_chan[lch].next_linked_ch | (1 << 15);
-#endif
-
-	dma_write(l, CLNK_CTRL(lch));
+		dma_write(l, CLNK_CTRL2, lch);
+	} else
+		dma_write(l, CLNK_CTRL1, lch);
 }
 
 static inline void disable_lnk(int lch)
 {
 	u32 l;
 
-	l = dma_read(CLNK_CTRL(lch));
-
 	/* Disable interrupts */
 	if (cpu_class_is_omap1()) {
-		dma_write(0, CICR(lch));
+		l = dma_read(CLNK_CTRL1, lch);
+		dma_write(0, CICR1, lch);
 		/* Set the STOP_LNK bit */
 		l |= 1 << 14;
+		dma_write(l, CLNK_CTRL1, lch);
 	}
 
 	if (cpu_class_is_omap2()) {
+		l = dma_read(CLNK_CTRL2, lch);
 		omap_disable_channel_irq(lch);
 		/* Clear the ENABLE_LNK bit */
 		l &= ~(1 << 15);
+		dma_write(l, CLNK_CTRL2, lch);
 	}
-
-	dma_write(l, CLNK_CTRL(lch));
 	dma_chan[lch].flags &= ~OMAP_DMA_ACTIVE;
 }
 
@@ -710,9 +934,9 @@ static inline void omap2_enable_irq_lch(int lch)
 		return;
 
 	spin_lock_irqsave(&dma_chan_lock, flags);
-	val = dma_read(IRQENABLE_L0);
+	val = dma_read(IRQENABLE_L0, 0);
 	val |= 1 << lch;
-	dma_write(val, IRQENABLE_L0);
+	dma_write(val, IRQENABLE_L0, 0);
 	spin_unlock_irqrestore(&dma_chan_lock, flags);
 }
 
@@ -725,9 +949,9 @@ static inline void omap2_disable_irq_lch(int lch)
 		return;
 
 	spin_lock_irqsave(&dma_chan_lock, flags);
-	val = dma_read(IRQENABLE_L0);
+	val = dma_read(IRQENABLE_L0, 0);
 	val &= ~(1 << lch);
-	dma_write(val, IRQENABLE_L0);
+	dma_write(val, IRQENABLE_L0, 0);
 	spin_unlock_irqrestore(&dma_chan_lock, flags);
 }
 
@@ -792,17 +1016,17 @@ int omap_request_dma(int dev_id, const char *dev_name,
 		 * Disable the 1510 compatibility mode and set the sync device
 		 * id.
 		 */
-		dma_write(dev_id | (1 << 10), CCR(free_ch));
+		dma_write(dev_id | (1 << 10), CCR1, free_ch);
 	} else if (cpu_is_omap7xx() || cpu_is_omap15xx()) {
-		dma_write(dev_id, CCR(free_ch));
+		dma_write(dev_id, CCR1, free_ch);
 	}
 
 	if (cpu_class_is_omap2()) {
 		omap2_enable_irq_lch(free_ch);
 		omap_enable_channel_irq(free_ch);
 		/* Clear the CSR register and IRQ status register */
-		dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR(free_ch));
-		dma_write(1 << free_ch, IRQSTATUS_L0);
+		dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR2, free_ch);
+		dma_write(1 << free_ch, IRQSTATUS_L0, 0);
 	}
 
 	*dma_ch_out = free_ch;
@@ -823,23 +1047,23 @@ void omap_free_dma(int lch)
 
 	if (cpu_class_is_omap1()) {
 		/* Disable all DMA interrupts for the channel. */
-		dma_write(0, CICR(lch));
+		dma_write(0, CICR1, lch);
 		/* Make sure the DMA transfer is stopped. */
-		dma_write(0, CCR(lch));
+		dma_write(0, CCR1, lch);
 	}
 
 	if (cpu_class_is_omap2()) {
 		omap2_disable_irq_lch(lch);
 
 		/* Clear the CSR register and IRQ status register */
-		dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR(lch));
-		dma_write(1 << lch, IRQSTATUS_L0);
+		dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR2, lch);
+		dma_write(1 << lch, IRQSTATUS_L0, 0);
 
 		/* Disable all DMA interrupts for the channel. */
-		dma_write(0, CICR(lch));
+		dma_write(0, CICR2, lch);
 
 		/* Make sure the DMA transfer is stopped. */
-		dma_write(0, CCR(lch));
+		dma_write(0, CCR2, lch);
 		omap_clear_dma(lch);
 	}
 
@@ -880,7 +1104,7 @@ omap_dma_set_global_params(int arb_rate, int max_fifo_depth, int tparams)
 	reg |= (0x3 & tparams) << 12;
 	reg |= (arb_rate & 0xff) << 16;
 
-	dma_write(reg, GCR);
+	dma_write(reg, GCR2, 0);
 }
 EXPORT_SYMBOL(omap_dma_set_global_params);
 
@@ -903,14 +1127,18 @@ omap_dma_set_prio_lch(int lch, unsigned char read_prio,
 		printk(KERN_ERR "Invalid channel id\n");
 		return -EINVAL;
 	}
-	l = dma_read(CCR(lch));
-	l &= ~((1 << 6) | (1 << 26));
-	if (cpu_is_omap2430() || cpu_is_omap34xx() ||  cpu_is_omap44xx())
+
+	if (cpu_is_omap2430() || cpu_is_omap34xx() ||  cpu_is_omap44xx()) {
+		l = dma_read(CCR2, lch);
+		l &= ~((1 << 6) | (1 << 26));
 		l |= ((read_prio & 0x1) << 6) | ((write_prio & 0x1) << 26);
-	else
+		dma_write(l, CCR2, lch);
+	} else {
+		l = dma_read(CCR1, lch);
+		l &= ~((1 << 6) | (1 << 26));
 		l |= ((read_prio & 0x1) << 6);
-
-	dma_write(l, CCR(lch));
+		dma_write(l, CCR1, lch);
+	}
 
 	return 0;
 }
@@ -929,19 +1157,19 @@ void omap_clear_dma(int lch)
 	if (cpu_class_is_omap1()) {
 		u32 l;
 
-		l = dma_read(CCR(lch));
+		l = dma_read(CCR1, lch);
 		l &= ~OMAP_DMA_CCR_EN;
-		dma_write(l, CCR(lch));
+		dma_write(l, CCR1, lch);
 
 		/* Clear pending interrupts */
-		l = dma_read(CSR(lch));
+		l = dma_read(CSR1, lch);
 	}
 
 	if (cpu_class_is_omap2()) {
-		int i;
-		void __iomem *lch_base = omap_dma_base + OMAP_DMA4_CH_BASE(lch);
-		for (i = 0; i < 0x44; i += 4)
-			__raw_writel(0, lch_base + i);
+		int i = OMAP2_CH_COMMON_START;
+
+		for (; i <= OMAP2_CH_COMMON_END; i++)
+			dma_write(0, i, lch);
 	}
 
 	local_irq_restore(flags);
@@ -957,9 +1185,12 @@ void omap_start_dma(int lch)
 	 * before starting dma transfer.
 	 */
 	if (cpu_is_omap15xx())
-		dma_write(0, CPC(lch));
+		dma_write(0, CPC, lch);
+	else if (cpu_class_is_omap1())
+		dma_write(0, CDAC1, lch);
 	else
-		dma_write(0, CDAC(lch));
+		dma_write(0, CDAC2, lch);
+
 
 	if (!omap_dma_in_1510_mode() && dma_chan[lch].next_lch != -1) {
 		int next_lch, cur_lch;
@@ -989,29 +1220,33 @@ void omap_start_dma(int lch)
 		(cpu_is_omap243x() &&  omap_type() <= OMAP2430_REV_ES1_0)) {
 
 		/* Errata: Need to write lch even if not using chaining */
-		dma_write(lch, CLNK_CTRL(lch));
+		dma_write(lch, CLNK_CTRL2, lch);
 	}
 
 	omap_enable_channel_irq(lch);
 
-	l = dma_read(CCR(lch));
-
-	/*
-	 * Errata: Inter Frame DMA buffering issue (All OMAP2420 and
-	 * OMAP2430ES1.0): DMA will wrongly buffer elements if packing and
-	 * bursting is enabled. This might result in data gets stalled in
-	 * FIFO at the end of the block.
-	 * Workaround: DMA channels must have BUFFERING_DISABLED bit set to
-	 * guarantee no data will stay in the DMA FIFO in case inter frame
-	 * buffering occurs.
-	 */
-	if (cpu_is_omap2420() ||
-	    (cpu_is_omap2430() && (omap_type() == OMAP2430_REV_ES1_0)))
-		l |= OMAP_DMA_CCR_BUFFERING_DISABLE;
-
-	l |= OMAP_DMA_CCR_EN;
-	dma_write(l, CCR(lch));
+	if (cpu_class_is_omap1()) {
+		l = dma_read(CCR1, lch);
+		l |= OMAP_DMA_CCR_EN;
+		dma_write(l, CCR1, lch);
+	} else {
+		l = dma_read(CCR2, lch);
+		/*
+		 * Errata: Inter Frame DMA buffering issue (All OMAP2420 and
+		 * OMAP2430ES1.0): DMA will wrongly buffer elements if packing
+		 * and bursting is enabled. This might result in data gets
+		 * stalled in FIFO at the end of the block.
+		 * Workaround: DMA channels must have BUFFERING_DISABLED bit
+		 * set to guarantee no data will stay in the DMA FIFO in case
+		 * inter frame buffering occurs.
+		 */
+		if (cpu_is_omap2420() || (cpu_is_omap2430() &&
+				(omap_type() == OMAP2430_REV_ES1_0)))
+			l |= OMAP_DMA_CCR_BUFFERING_DISABLE;
 
+		l |= OMAP_DMA_CCR_EN;
+		dma_write(l, CCR2, lch);
+	}
 	dma_chan[lch].flags |= OMAP_DMA_ACTIVE;
 }
 EXPORT_SYMBOL(omap_start_dma);
@@ -1022,41 +1257,45 @@ void omap_stop_dma(int lch)
 
 	/* Disable all interrupts on the channel */
 	if (cpu_class_is_omap1())
-		dma_write(0, CICR(lch));
+		dma_write(0, CICR1, lch);
 
-	l = dma_read(CCR(lch));
 	/* OMAP3 Errata i541: sDMA FIFO draining does not finish */
 	if (cpu_is_omap34xx() && (l & OMAP_DMA_CCR_SEL_SRC_DST_SYNC)) {
 		int i = 0;
 		u32 sys_cf;
 
 		/* Configure No-Standby */
-		l = dma_read(OCP_SYSCONFIG);
+		l = dma_read(OCP_SYSCONFIG, 0);
 		sys_cf = l;
 		l &= ~DMA_SYSCONFIG_MIDLEMODE_MASK;
 		l |= DMA_SYSCONFIG_MIDLEMODE(DMA_IDLEMODE_NO_IDLE);
-		dma_write(l , OCP_SYSCONFIG);
+		dma_write(l , OCP_SYSCONFIG, 0);
 
-		l = dma_read(CCR(lch));
+		l = dma_read(CCR2, lch);
 		l &= ~OMAP_DMA_CCR_EN;
-		dma_write(l, CCR(lch));
+		dma_write(l, CCR2, lch);
 
 		/* Wait for sDMA FIFO drain */
-		l = dma_read(CCR(lch));
+		l = dma_read(CCR2, lch);
 		while (i < 100 && (l & (OMAP_DMA_CCR_RD_ACTIVE |
 					OMAP_DMA_CCR_WR_ACTIVE))) {
 			udelay(5);
 			i++;
-			l = dma_read(CCR(lch));
+			l = dma_read(CCR2, lch);
 		}
 		if (i >= 100)
 			printk(KERN_ERR "DMA drain did not complete on "
 					"lch %d\n", lch);
 		/* Restore OCP_SYSCONFIG */
-		dma_write(sys_cf, OCP_SYSCONFIG);
+		dma_write(sys_cf, OCP_SYSCONFIG, 0);
+	} else if (cpu_class_is_omap1()) {
+		l = dma_read(CCR1, lch);
+		l &= ~OMAP_DMA_CCR_EN;
+		dma_write(l, CCR1, lch);
 	} else {
+		l = dma_read(CCR2, lch);
 		l &= ~OMAP_DMA_CCR_EN;
-		dma_write(l, CCR(lch));
+		dma_write(l, CCR2, lch);
 	}
 
 	if (!omap_dma_in_1510_mode() && dma_chan[lch].next_lch != -1) {
@@ -1122,19 +1361,25 @@ dma_addr_t omap_get_dma_src_pos(int lch)
 	dma_addr_t offset = 0;
 
 	if (cpu_is_omap15xx())
-		offset = dma_read(CPC(lch));
+		offset = dma_read(CPC, lch);
+	else if (cpu_class_is_omap1())
+		offset = dma_read(CSAC1, lch);
 	else
-		offset = dma_read(CSAC(lch));
+		offset = dma_read(CSAC2, lch);
 
 	/*
 	 * omap 3.2/3.3 erratum: sometimes 0 is returned if CSAC/CDAC is
 	 * read before the DMA controller finished disabling the channel.
 	 */
-	if (!cpu_is_omap15xx() && offset == 0)
-		offset = dma_read(CSAC(lch));
+	if (!cpu_is_omap15xx() && offset == 0) {
+		if (cpu_class_is_omap1())
+			offset = dma_read(CSAC1, lch);
+		else
+			offset = dma_read(CSAC2, lch);
+	}
 
 	if (cpu_class_is_omap1())
-		offset |= (dma_read(CSSA_U(lch)) << 16);
+		offset |= (dma_read(CSSA_U, lch) << 16);
 
 	return offset;
 }
@@ -1153,19 +1398,25 @@ dma_addr_t omap_get_dma_dst_pos(int lch)
 	dma_addr_t offset = 0;
 
 	if (cpu_is_omap15xx())
-		offset = dma_read(CPC(lch));
+		offset = dma_read(CPC, lch);
+	else if (cpu_class_is_omap1())
+		offset = dma_read(CDAC1, lch);
 	else
-		offset = dma_read(CDAC(lch));
+		offset = dma_read(CDAC2, lch);
 
 	/*
 	 * omap 3.2/3.3 erratum: sometimes 0 is returned if CSAC/CDAC is
 	 * read before the DMA controller finished disabling the channel.
 	 */
-	if (!cpu_is_omap15xx() && offset == 0)
-		offset = dma_read(CDAC(lch));
+	if (!cpu_is_omap15xx() && offset == 0) {
+		if (cpu_class_is_omap1())
+			offset = dma_read(CDAC1, lch);
+		else
+			offset = dma_read(CDAC2, lch);
+	}
 
 	if (cpu_class_is_omap1())
-		offset |= (dma_read(CDSA_U(lch)) << 16);
+		offset |= (dma_read(CDSA_U, lch) << 16);
 
 	return offset;
 }
@@ -1173,7 +1424,10 @@ EXPORT_SYMBOL(omap_get_dma_dst_pos);
 
 int omap_get_dma_active_status(int lch)
 {
-	return (dma_read(CCR(lch)) & OMAP_DMA_CCR_EN) != 0;
+	if (cpu_class_is_omap1())
+		return (dma_read(CCR1, lch) & OMAP_DMA_CCR_EN) != 0;
+	else
+		return (dma_read(CCR2, lch) & OMAP_DMA_CCR_EN) != 0;
 }
 EXPORT_SYMBOL(omap_get_dma_active_status);
 
@@ -1181,14 +1435,15 @@ int omap_dma_running(void)
 {
 	int lch;
 
-	if (cpu_class_is_omap1())
-		if (omap_lcd_dma_running())
-			return 1;
-
-	for (lch = 0; lch < dma_chan_count; lch++)
-		if (dma_read(CCR(lch)) & OMAP_DMA_CCR_EN)
-			return 1;
-
+	for (lch = 0; lch < dma_chan_count; lch++) {
+		if (cpu_class_is_omap1()) {
+			if (dma_read(CCR1, lch) & OMAP_DMA_CCR_EN)
+				return 1;
+		} else {
+			if (dma_read(CCR2, lch) & OMAP_DMA_CCR_EN)
+				return 1;
+		}
+	}
 	return 0;
 }
 
@@ -1201,8 +1456,8 @@ void omap_dma_link_lch(int lch_head, int lch_queue)
 {
 	if (omap_dma_in_1510_mode()) {
 		if (lch_head == lch_queue) {
-			dma_write(dma_read(CCR(lch_head)) | (3 << 8),
-								CCR(lch_head));
+			dma_write(dma_read(CCR1, lch_head) | (3 << 8),
+							CCR1, lch_head);
 			return;
 		}
 		printk(KERN_ERR "DMA linking is not supported in 1510 mode\n");
@@ -1228,8 +1483,8 @@ void omap_dma_unlink_lch(int lch_head, int lch_queue)
 {
 	if (omap_dma_in_1510_mode()) {
 		if (lch_head == lch_queue) {
-			dma_write(dma_read(CCR(lch_head)) & ~(3 << 8),
-								CCR(lch_head));
+			dma_write(dma_read(CCR1, lch_head) & ~(3 << 8),
+								CCR1, lch_head);
 			return;
 		}
 		printk(KERN_ERR "DMA linking is not supported in 1510 mode\n");
@@ -1281,15 +1536,15 @@ static void create_dma_lch_chain(int lch_head, int lch_queue)
 					lch_queue;
 	}
 
-	l = dma_read(CLNK_CTRL(lch_head));
+	l = dma_read(CLNK_CTRL2, lch_head);
 	l &= ~(0x1f);
 	l |= lch_queue;
-	dma_write(l, CLNK_CTRL(lch_head));
+	dma_write(l, CLNK_CTRL2, lch_head);
 
-	l = dma_read(CLNK_CTRL(lch_queue));
+	l = dma_read(CLNK_CTRL2, lch_queue);
 	l &= ~(0x1f);
 	l |= (dma_chan[lch_queue].next_linked_ch);
-	dma_write(l, CLNK_CTRL(lch_queue));
+	dma_write(l, CLNK_CTRL2, lch_queue);
 }
 
 /**
@@ -1565,13 +1820,13 @@ int omap_dma_chain_a_transfer(int chain_id, int src_start, int dest_start,
 
 	/* Set the params to the free channel */
 	if (src_start != 0)
-		dma_write(src_start, CSSA(lch));
+		dma_write(src_start, CSSA, lch);
 	if (dest_start != 0)
-		dma_write(dest_start, CDSA(lch));
+		dma_write(dest_start, CDSA, lch);
 
 	/* Write the buffer size */
-	dma_write(elem_count, CEN(lch));
-	dma_write(frame_count, CFN(lch));
+	dma_write(elem_count, CEN2, lch);
+	dma_write(frame_count, CFN2, lch);
 
 	/*
 	 * If the chain is dynamically linked,
@@ -1605,7 +1860,7 @@ int omap_dma_chain_a_transfer(int chain_id, int src_start, int dest_start,
 				dma_chan[lch].state = DMA_CH_QUEUED;
 				start_dma = 0;
 				if (0 == ((1 << 7) & dma_read(
-					CCR(dma_chan[lch].prev_linked_ch)))) {
+					CCR2, dma_chan[lch].prev_linked_ch))) {
 					disable_lnk(dma_chan[lch].
 						    prev_linked_ch);
 					pr_debug("\n prev ch is stopped\n");
@@ -1621,7 +1876,7 @@ int omap_dma_chain_a_transfer(int chain_id, int src_start, int dest_start,
 			}
 			omap_enable_channel_irq(lch);
 
-			l = dma_read(CCR(lch));
+			l = dma_read(CCR2, lch);
 
 			if ((0 == (l & (1 << 24))))
 				l &= ~(1 << 25);
@@ -1632,12 +1887,12 @@ int omap_dma_chain_a_transfer(int chain_id, int src_start, int dest_start,
 					l |= (1 << 7);
 					dma_chan[lch].state = DMA_CH_STARTED;
 					pr_debug("starting %d\n", lch);
-					dma_write(l, CCR(lch));
+					dma_write(l, CCR2, lch);
 				} else
 					start_dma = 0;
 			} else {
 				if (0 == (l & (1 << 7)))
-					dma_write(l, CCR(lch));
+					dma_write(l, CCR2, lch);
 			}
 			dma_chan[lch].flags |= OMAP_DMA_ACTIVE;
 		}
@@ -1682,7 +1937,7 @@ int omap_start_dma_chain_transfers(int chain_id)
 		omap_enable_channel_irq(channels[0]);
 	}
 
-	l = dma_read(CCR(channels[0]));
+	l = dma_read(CCR2, channels[0]);
 	l |= (1 << 7);
 	dma_linked_lch[chain_id].chain_state = DMA_CHAIN_STARTED;
 	dma_chan[channels[0]].state = DMA_CH_STARTED;
@@ -1691,7 +1946,7 @@ int omap_start_dma_chain_transfers(int chain_id)
 		l &= ~(1 << 25);
 	else
 		l |= (1 << 25);
-	dma_write(l, CCR(channels[0]));
+	dma_write(l, CCR2, channels[0]);
 
 	dma_chan[channels[0]].flags |= OMAP_DMA_ACTIVE;
 
@@ -1730,18 +1985,18 @@ int omap_stop_dma_chain_transfers(int chain_id)
 	 * DMA Errata:
 	 * Special programming model needed to disable DMA before end of block
 	 */
-	sys_cf = dma_read(OCP_SYSCONFIG);
+	sys_cf = dma_read(OCP_SYSCONFIG, 0);
 	l = sys_cf;
 	/* Middle mode reg set no Standby */
 	l &= ~((1 << 12)|(1 << 13));
-	dma_write(l, OCP_SYSCONFIG);
+	dma_write(l, OCP_SYSCONFIG, 0);
 
 	for (i = 0; i < dma_linked_lch[chain_id].no_of_lchs_linked; i++) {
 
 		/* Stop the Channel transmission */
-		l = dma_read(CCR(channels[i]));
+		l = dma_read(CCR2, channels[i]);
 		l &= ~(1 << 7);
-		dma_write(l, CCR(channels[i]));
+		dma_write(l, CCR2, channels[i]);
 
 		/* Disable the link in all the channels */
 		disable_lnk(channels[i]);
@@ -1754,7 +2009,7 @@ int omap_stop_dma_chain_transfers(int chain_id)
 	OMAP_DMA_CHAIN_QINIT(chain_id);
 
 	/* Errata - put in the old value */
-	dma_write(sys_cf, OCP_SYSCONFIG);
+	dma_write(sys_cf, OCP_SYSCONFIG, 0);
 
 	return 0;
 }
@@ -1796,8 +2051,8 @@ int omap_get_dma_chain_index(int chain_id, int *ei, int *fi)
 	/* Get the current channel */
 	lch = channels[dma_linked_lch[chain_id].q_head];
 
-	*ei = dma_read(CCEN(lch));
-	*fi = dma_read(CCFN(lch));
+	*ei = dma_read(CCEN, lch);
+	*fi = dma_read(CCFN, lch);
 
 	return 0;
 }
@@ -1834,7 +2089,7 @@ int omap_get_dma_chain_dst_pos(int chain_id)
 	/* Get the current channel */
 	lch = channels[dma_linked_lch[chain_id].q_head];
 
-	return dma_read(CDAC(lch));
+	return dma_read(CDAC2, lch);
 }
 EXPORT_SYMBOL(omap_get_dma_chain_dst_pos);
 
@@ -1868,7 +2123,7 @@ int omap_get_dma_chain_src_pos(int chain_id)
 	/* Get the current channel */
 	lch = channels[dma_linked_lch[chain_id].q_head];
 
-	return dma_read(CSAC(lch));
+	return dma_read(CSAC2, lch);
 }
 EXPORT_SYMBOL(omap_get_dma_chain_src_pos);
 #endif	/* ifndef CONFIG_ARCH_OMAP1 */
@@ -1885,7 +2140,7 @@ static int omap1_dma_handle_ch(int ch)
 		csr = dma_chan[ch].saved_csr;
 		dma_chan[ch].saved_csr = 0;
 	} else
-		csr = dma_read(CSR(ch));
+		csr = dma_read(CSR1, ch);
 	if (enable_1510_mode && ch <= 2 && (csr >> 7) != 0) {
 		dma_chan[ch + 6].saved_csr = csr >> 7;
 		csr &= 0x7f;
@@ -1938,13 +2193,13 @@ static irqreturn_t omap1_dma_irq_handler(int irq, void *dev_id)
 
 static int omap2_dma_handle_ch(int ch)
 {
-	u32 status = dma_read(CSR(ch));
+	u32 status = dma_read(CSR2, ch);
 
 	if (!status) {
 		if (printk_ratelimit())
 			printk(KERN_WARNING "Spurious DMA IRQ for lch %d\n",
 				ch);
-		dma_write(1 << ch, IRQSTATUS_L0);
+		dma_write(1 << ch, IRQSTATUS_L0, 0);
 		return 0;
 	}
 	if (unlikely(dma_chan[ch].dev_id == -1)) {
@@ -1968,9 +2223,9 @@ static int omap2_dma_handle_ch(int ch)
 			 */
 			u32 ccr;
 
-			ccr = dma_read(CCR(ch));
+			ccr = dma_read(CCR2, ch);
 			ccr &= ~OMAP_DMA_CCR_EN;
-			dma_write(ccr, CCR(ch));
+			dma_write(ccr, CCR2, ch);
 			dma_chan[ch].flags &= ~OMAP_DMA_ACTIVE;
 		}
 	}
@@ -1981,14 +2236,14 @@ static int omap2_dma_handle_ch(int ch)
 		printk(KERN_INFO "DMA misaligned error with device %d\n",
 		       dma_chan[ch].dev_id);
 
-	dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR(ch));
-	dma_write(1 << ch, IRQSTATUS_L0);
+	dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR2, ch);
+	dma_write(1 << ch, IRQSTATUS_L0, 0);
 
 	/* If the ch is not chained then chain_id will be -1 */
 	if (dma_chan[ch].chain_id != -1) {
 		int chain_id = dma_chan[ch].chain_id;
 		dma_chan[ch].state = DMA_CH_NOTSTARTED;
-		if (dma_read(CLNK_CTRL(ch)) & (1 << 15))
+		if (dma_read(CLNK_CTRL2, ch) & (1 << 15))
 			dma_chan[dma_chan[ch].next_linked_ch].state =
 							DMA_CH_STARTED;
 		if (dma_linked_lch[chain_id].chain_mode ==
@@ -1998,10 +2253,10 @@ static int omap2_dma_handle_ch(int ch)
 		if (!OMAP_DMA_CHAIN_QEMPTY(chain_id))
 			OMAP_DMA_CHAIN_INCQHEAD(chain_id);
 
-		status = dma_read(CSR(ch));
+		status = dma_read(CSR2, ch);
 	}
 
-	dma_write(status, CSR(ch));
+	dma_write(status, CSR2, ch);
 
 	if (likely(dma_chan[ch].callback != NULL))
 		dma_chan[ch].callback(ch, status, dma_chan[ch].data);
@@ -2015,13 +2270,13 @@ static irqreturn_t omap2_dma_irq_handler(int irq, void *dev_id)
 	u32 val, enable_reg;
 	int i;
 
-	val = dma_read(IRQSTATUS_L0);
+	val = dma_read(IRQSTATUS_L0, 0);
 	if (val == 0) {
 		if (printk_ratelimit())
 			printk(KERN_WARNING "Spurious DMA IRQ\n");
 		return IRQ_HANDLED;
 	}
-	enable_reg = dma_read(IRQENABLE_L0);
+	enable_reg = dma_read(IRQENABLE_L0, 0);
 	val &= enable_reg; /* Dispatch only relevant interrupts */
 	for (i = 0; i < dma_lch_count && val != 0; i++) {
 		if (val & 1)
@@ -2047,21 +2302,21 @@ static struct irqaction omap24xx_dma_irq;
 void omap_dma_global_context_save(void)
 {
 	omap_dma_global_context.dma_irqenable_l0 =
-		dma_read(IRQENABLE_L0);
+		dma_read(IRQENABLE_L0, 0);
 	omap_dma_global_context.dma_ocp_sysconfig =
-		dma_read(OCP_SYSCONFIG);
-	omap_dma_global_context.dma_gcr = dma_read(GCR);
+		dma_read(OCP_SYSCONFIG, 0);
+	omap_dma_global_context.dma_gcr = dma_read(GCR2, 0);
 }
 
 void omap_dma_global_context_restore(void)
 {
 	int ch;
 
-	dma_write(omap_dma_global_context.dma_gcr, GCR);
+	dma_write(omap_dma_global_context.dma_gcr, GCR2, 0);
 	dma_write(omap_dma_global_context.dma_ocp_sysconfig,
-		OCP_SYSCONFIG);
+		OCP_SYSCONFIG, 0);
 	dma_write(omap_dma_global_context.dma_irqenable_l0,
-		IRQENABLE_L0);
+		IRQENABLE_L0, 0);
 
 	/*
 	 * A bug in ROM code leaves IRQ status for channels 0 and 1 uncleared
@@ -2070,7 +2325,7 @@ void omap_dma_global_context_restore(void)
 	 * affects only secure devices.
 	 */
 	if (cpu_is_omap34xx() && (omap_type() != OMAP2_DEVICE_TYPE_GP))
-		dma_write(0x3 , IRQSTATUS_L0);
+		dma_write(0x3 , IRQSTATUS_L0, 0);
 
 	for (ch = 0; ch < dma_chan_count; ch++)
 		if (dma_chan[ch].dev_id != -1)
@@ -2101,8 +2356,8 @@ static int __init omap_init_dma(void)
 		return -ENODEV;
 	}
 
-	omap_dma_base = ioremap(base, SZ_4K);
-	BUG_ON(!omap_dma_base);
+	dma_base = ioremap(base, SZ_4K);
+	BUG_ON(!dma_base);
 
 	if (cpu_class_is_omap2() && omap_dma_reserve_channels
 			&& (omap_dma_reserve_channels <= dma_lch_count))
@@ -2130,26 +2385,26 @@ static int __init omap_init_dma(void)
 		enable_1510_mode = 1;
 	} else if (cpu_is_omap16xx() || cpu_is_omap7xx()) {
 		printk(KERN_INFO "OMAP DMA hardware version %d\n",
-		       dma_read(HW_ID));
+		       dma_read(HW_ID, 0));
 		printk(KERN_INFO "DMA capabilities: %08x:%08x:%04x:%04x:%04x\n",
-		       (dma_read(CAPS_0_U) << 16) |
-		       dma_read(CAPS_0_L),
-		       (dma_read(CAPS_1_U) << 16) |
-		       dma_read(CAPS_1_L),
-		       dma_read(CAPS_2), dma_read(CAPS_3),
-		       dma_read(CAPS_4));
+		       (dma_read(CAPS1_0_U, 0) << 16) |
+		       dma_read(CAPS1_0_L, 0),
+		       (dma_read(CAPS1_1_U, 0) << 16) |
+		       dma_read(CAPS1_1_L, 0),
+		       dma_read(CAPS1_2, 0), dma_read(CAPS1_3, 0),
+		       dma_read(CAPS1_4, 0));
 		if (!enable_1510_mode) {
 			u16 w;
 
 			/* Disable OMAP 3.0/3.1 compatibility mode. */
-			w = dma_read(GSCR);
+			w = dma_read(GSCR, 0);
 			w |= 1 << 3;
-			dma_write(w, GSCR);
+			dma_write(w, GSCR, 0);
 			dma_chan_count = 16;
 		} else
 			dma_chan_count = 9;
 	} else if (cpu_class_is_omap2()) {
-		u8 revision = dma_read(REVISION) & 0xff;
+		u8 revision = dma_read(REVISION, 0) & 0xff;
 		printk(KERN_INFO "OMAP DMA hardware revision %d.%d\n",
 		       revision >> 4, revision & 0xf);
 		dma_chan_count = dma_lch_count;
@@ -2208,14 +2463,14 @@ static int __init omap_init_dma(void)
 
 	if (cpu_is_omap34xx() || cpu_is_omap44xx()) {
 		/* Enable smartidle idlemodes and autoidle */
-		u32 v = dma_read(OCP_SYSCONFIG);
+		u32 v = dma_read(OCP_SYSCONFIG, 0);
 		v &= ~(DMA_SYSCONFIG_MIDLEMODE_MASK |
 				DMA_SYSCONFIG_SIDLEMODE_MASK |
 				DMA_SYSCONFIG_AUTOIDLE);
 		v |= (DMA_SYSCONFIG_MIDLEMODE(DMA_IDLEMODE_SMARTIDLE) |
 			DMA_SYSCONFIG_SIDLEMODE(DMA_IDLEMODE_SMARTIDLE) |
 			DMA_SYSCONFIG_AUTOIDLE);
-		dma_write(v , OCP_SYSCONFIG);
+		dma_write(v , OCP_SYSCONFIG, 0);
 		/* reserve dma channels 0 and 1 in high security devices */
 		if (cpu_is_omap34xx() &&
 			(omap_type() != OMAP2_DEVICE_TYPE_GP)) {
@@ -2232,7 +2487,7 @@ out_free:
 	kfree(dma_chan);
 
 out_unmap:
-	iounmap(omap_dma_base);
+	iounmap(dma_base);
 
 	return r;
 }
arch/arm/plat-omap/include/plat/dma.h +0 -120

134 lines not on this page · raw message

Signature · 1 line
1.7.1
#2 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

G, Manjunath Kondaiah had written, on 10/26/2010 08:25 AM, the following:
[...]

Quoted 15 lines
arch/arm/plat-omap/dma.c +6 -0
index f5c5b8d..77241e2 100644
@@ -40,6 +40,147 @@
 
 #undef DEBUG
 
+enum {
+       GCR1 = 0,       GSCR,           GRST,           HW_ID,
+       PCH2_ID,        PCH0_ID,        PCH1_ID,        PCHG_ID,
+       PCHD_ID,        CAPS1_0_U,      CAPS1_0_L,      CAPS1_1_U,
+       CAPS1_1_L,      CAPS1_2,        CAPS1_3,        CAPS1_4,
+       PCH2_SR,        PCH0_SR,        PCH1_SR,        PCHD_SR,

[...]

Quoted 12 lines
arch/arm/plat-omap/dma.c +12 -0
+};
+
+static u16 reg_map_omap1[] = {
+       [GCR1]          = 0x400,
+       [GSCR]          = 0x404,
+       [GRST]          = 0x408,
+       [HW_ID]         = 0x442,
+       [PCH2_ID]       = 0x444,
+       [PCH0_ID]       = 0x446,
+       [PCH1_ID]       = 0x448,
+       [PCHG_ID]       = 0x44a,
+       [PCHD_ID]       = 0x44c,

[...]

Quoted 6 lines
arch/arm/plat-omap/dma.c +6 -0
+       [LCH_CTRL]      = 0x2a,
+};
+
+enum {
+       REVISION = 0,   GCR2,           IRQSTATUS_L0,   IRQSTATUS_L1,
+       IRQSTATUS_L2,   IRQSTATUS_L3,   IRQENABLE_L0,   IRQENABLE_L1,

[...]

Quoted 12 lines
arch/arm/plat-omap/dma.c +12 -0
+
+       /* OMAP4 specific registers */
+       CDP,            CNDP,           CCDN,
+
+       OMAP2_CH_COMMON_END,
+};
+
+static u16 reg_map_omap2[] = {
+       [REVISION]              = 0x00,
+       [GCR2]                  = 0x78,
+       [IRQSTATUS_L0]          = 0x08,
+       [IRQSTATUS_L1]          = 0x0c,

[..]

arch/arm/plat-omap/dma.c line 183

Quoted 6 lines
arch/arm/plat-omap/dma.c +6 -0
+       /* OMAP4 specific registers */
+       [CDP]                   = 0xd0,
+       [CNDP]                  = 0xd4,
+       [CCDN]                  = 0xd8,
+};
+

dumb question: any reason why a struct wont do?
struct reg_map_omap2 {
u16 revision;
...
...
}

[..]

Quoted 44 lines
arch/arm/plat-omap/dma.c +26 -17
-#define dma_read(reg)                                                  \
-({                                                                     \
-       u32 __val;                                                      \
-       if (cpu_class_is_omap1())                                       \
-               __val = __raw_readw(omap_dma_base + OMAP1_DMA_##reg);   \
-       else                                                            \
-               __val = __raw_readl(omap_dma_base + OMAP_DMA4_##reg);   \
-       __val;                                                          \
-})
-
-#define dma_write(val, reg)                                            \
-({                                                                     \
-       if (cpu_class_is_omap1())                                       \
-               __raw_writew((u16)(val), omap_dma_base + OMAP1_DMA_##reg); \
-       else                                                            \
-               __raw_writel((val), omap_dma_base + OMAP_DMA4_##reg);   \
-})
+static inline void dma_write(u32 val, int reg, int lch)
+{
+       if (cpu_class_is_omap1()) {
+               if (reg > OMAP1_CH_COMMON_START)
+                       __raw_writew(val, dma_base +
+                               (reg_map_omap1[reg] + 0x40 * lch));
+               else
+                       __raw_writew(val, dma_base + reg_map_omap1[reg]);
+       } else {
+               if (reg > OMAP2_CH_COMMON_START)
+                       __raw_writel(val, dma_base +
+                               (reg_map_omap2[reg] + 0x60 * lch));
+               else
+                       __raw_writel(val, dma_base + reg_map_omap2[reg]);
+       }
+}
+
+static inline u32 dma_read(int reg, int lch)
+{
+       u32 val;
+
+       if (cpu_class_is_omap1()) {
+               if (reg > OMAP1_CH_COMMON_START)
+                       val = __raw_readw(dma_base +
+                                       (reg_map_omap1[reg] + 0x40 * lch));
+               else

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 333

Quoted 10 lines
arch/arm/plat-omap/dma.c +10 -0
+                       val = __raw_readw(dma_base + reg_map_omap1[reg]);
+       } else {
+               if (reg > OMAP2_CH_COMMON_START)
+                       val = __raw_readl(dma_base +
+                                       (reg_map_omap2[reg] + 0x60 * lch));
+               else
+                       val = __raw_readl(dma_base + reg_map_omap2[reg]);
+       }
+       return val;
+}

What is the benefit of using inline function here? would'nt we increase code size? cant we use a function pointer initialized to class1 or rest? Quote from CodingStyle (15):
"A reasonable rule of thumb is to not put inline at functions that have more than 3 lines of code in them. An exception to this rule are the cases where a parameter is known to be a compiletime constant, and as a result of this constantness you *know* the compiler will be able to optimize most of your function away at compile time. For a good example of this later case, see the kmalloc() inline function.
"
is the expectation that cpu_class_is_omap1 compile time constant hence the actual compiled in code is smaller -candidate for inline?

Quoted 65 lines

#ifdef CONFIG_ARCH_OMAP15XX
/* Returns 1 if the DMA module is in OMAP1510-compatible mode, 0 otherwise */

arch/arm/plat-omap/dma.c +14 -15
@@ -209,11 +369,10 @@ static inline void set_gdma_dev(int req, int dev)
 /* Omap1 only */
 static void clear_lch_regs(int lch)
 {
-       int i;
-       void __iomem *lch_base = omap_dma_base + OMAP1_DMA_CH_BASE(lch);
+       int i = OMAP1_CH_COMMON_START;
 
-       for (i = 0; i < 0x2c; i += 2)
-               __raw_writew(0, lch_base + i);
+       for (; i <= OMAP1_CH_COMMON_END; i += 1)
+               dma_write(0, i, lch);
 }
 
 void omap_set_dma_priority(int lch, int dst_port, int priority)
@@ -248,12 +407,12 @@ void omap_set_dma_priority(int lch, int dst_port, int priority)
        if (cpu_class_is_omap2()) {
                u32 ccr;
 
-               ccr = dma_read(CCR(lch));
+               ccr = dma_read(CCR2, lch);
                if (priority)
                        ccr |= (1 << 6);
                else
                        ccr &= ~(1 << 6);
-               dma_write(ccr, CCR(lch));
+               dma_write(ccr, CCR2, lch);
        }
 }
 EXPORT_SYMBOL(omap_set_dma_priority);
@@ -264,31 +423,36 @@ void omap_set_dma_transfer_params(int lch, int data_type, int elem_count,
 {
        u32 l;
 
-       l = dma_read(CSDP(lch));
-       l &= ~0x03;
-       l |= data_type;
-       dma_write(l, CSDP(lch));
-
        if (cpu_class_is_omap1()) {
                u16 ccr;
 
-               ccr = dma_read(CCR(lch));
+               l = dma_read(CSDP1, lch);
+               l &= ~0x03;
+               l |= data_type;
+               dma_write(l, CSDP1, lch);
+
+               ccr = dma_read(CCR1, lch);
                ccr &= ~(1 << 5);
                if (sync_mode == OMAP_DMA_SYNC_FRAME)
                        ccr |= 1 << 5;
-               dma_write(ccr, CCR(lch));
+               dma_write(ccr, CCR1, lch);
 
-               ccr = dma_read(CCR2(lch));
+               ccr = dma_read(CCR1_2, lch);
                ccr &= ~(1 << 2);
                if (sync_mode == OMAP_DMA_SYNC_BLOCK)
                        ccr |= 1 << 2;
-               dma_write(ccr, CCR2(lch));
+               dma_write(ccr, CCR1_2, lch);

arch/arm/plat-omap/dma.c void omap_set_dma_transfer_params(int lch, int data_type, int elem_count, · line 453

Quoted 10 lines
arch/arm/plat-omap/dma.c +4 -1
        }
 
        if (cpu_class_is_omap2() && dma_trigger) {
                u32 val;
 
-               val = dma_read(CCR(lch));
+               l = dma_read(CSDP2, lch);
+               l &= ~0x03;
+               l |= data_type;
+               dma_write(l, CSDP2, lch);

This seems to me to be a replication mess :( compared to the original CSDP(lch) - which mapped to class1 or class2 as needed - same elsewhere. Code is becoming pretty much needing a refactoring as a result :(

[...]

Signature · 2 lines
Regards,
Nishanth Menon
#3 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

Quoted 23 lines

-----Original Message-----
From: Menon, Nishanth
Sent: Tuesday, October 26, 2010 8:19 PM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org;
linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Kevin Hilman; Shilimkar, Santosh
Subject: Re: [PATCH v3 01/13] OMAP: DMA: Replace read/write macros with functions

G, Manjunath Kondaiah had written, on 10/26/2010 08:25 AM, the following:
[...]

Review

Quoted 2 lines
(quoted diff) +0 -0

b/arch/arm/plat-omap/dma.c index

Quoted 5 lines

f5c5b8d..77241e2 100644

arch/arm/plat-omap/dma.c +0 -0
@@ -40,6 +40,147 @@
 

[...]

Quoted 6 lines
Quoted 5 lines
arch/arm/plat-omap/dma.c +5 -0
+static u16 reg_map_omap2[] = {
+       [REVISION]              = 0x00,
+       [GCR2]                  = 0x78,
+       [IRQSTATUS_L0]          = 0x08,
+       [IRQSTATUS_L1]          = 0x0c,

[..]

Quoted 12 lines

arch/arm/plat-omap/dma.c line 183

Quoted 6 lines
arch/arm/plat-omap/dma.c +6 -0
+       /* OMAP4 specific registers */
+       [CDP]                   = 0xd0,
+       [CNDP]                  = 0xd4,
+       [CCDN]                  = 0xd8,
+};
+

dumb question: any reason why a struct wont do?
struct reg_map_omap2 {
u16 revision;
...
...
}

This approach is as per i2c driver as suggested by kevin.
http://www.spinics.net/lists/linux-omap/msg36446.html

Quoted 74 lines

[..]

Quoted 2 lines

-#define dma_read(reg)

            \
Quoted 1 line

-({

            \
Quoted 1 line

- u32 __val;

            \
Quoted 1 line

- if (cpu_class_is_omap1())

            \
Quoted 1 line

- __val = __raw_readw(omap_dma_base +

OMAP1_DMA_##reg); \

Quoted 1 line

- else

            \
Quoted 1 line

- __val = __raw_readl(omap_dma_base +

OMAP_DMA4_##reg); \

Quoted 1 line

- __val;

            \

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 3 lines
arch/arm/plat-omap/dma.c +0 -3
-})
-
-#define dma_write(val, reg)                                
            \
Quoted 1 line

-({

            \
Quoted 1 line

- if (cpu_class_is_omap1())

            \
Quoted 1 line

- __raw_writew((u16)(val), omap_dma_base +

OMAP1_DMA_##reg); \

Quoted 1 line

- else

            \
Quoted 1 line

- __raw_writel((val), omap_dma_base +

OMAP_DMA4_##reg); \

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 8 lines
arch/arm/plat-omap/dma.c +7 -1
-})
+static inline void dma_write(u32 val, int reg, int lch) {
+       if (cpu_class_is_omap1()) {
+               if (reg > OMAP1_CH_COMMON_START)
+                       __raw_writew(val, dma_base +
+                               (reg_map_omap1[reg] + 0x40 * lch));
+               else
+                       __raw_writew(val, dma_base + 

reg_map_omap1[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 6 lines
arch/arm/plat-omap/dma.c +6 -0
+       } else {
+               if (reg > OMAP2_CH_COMMON_START)
+                       __raw_writel(val, dma_base +
+                               (reg_map_omap2[reg] + 0x60 * lch));
+               else
+                       __raw_writel(val, dma_base + 

reg_map_omap2[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 10 lines
arch/arm/plat-omap/dma.c +10 -0
+       }
+}
+
+static inline u32 dma_read(int reg, int lch) {
+       u32 val;
+
+       if (cpu_class_is_omap1()) {
+               if (reg > OMAP1_CH_COMMON_START)
+                       val = __raw_readw(dma_base +
+                                       (reg_map_omap1[reg] 

+ 0x40 * lch));

arch/arm/plat-omap/dma.c

Quoted 2 lines
arch/arm/plat-omap/dma.c +2 -0
+               else
+                       val = __raw_readw(dma_base + 

reg_map_omap1[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 4 lines
arch/arm/plat-omap/dma.c +4 -0
+       } else {
+               if (reg > OMAP2_CH_COMMON_START)
+                       val = __raw_readl(dma_base +
+                                       (reg_map_omap2[reg] 

+ 0x60 * lch));

arch/arm/plat-omap/dma.c

Quoted 2 lines
arch/arm/plat-omap/dma.c +2 -0
+               else
+                       val = __raw_readl(dma_base + 

reg_map_omap2[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 333

Quoted 3 lines
arch/arm/plat-omap/dma.c +3 -0
+       }
+       return val;
+}
Quoted 15 lines

What is the benefit of using inline function here? would'nt we increase code size? cant we use a function pointer initialized to class1 or rest?
Quote from CodingStyle (15):
"A reasonable rule of thumb is to not put inline at functions that have more than 3 lines of code in them. An exception to this rule are the cases where a parameter is known to be a compiletime constant, and as a result of this constantness you *know* the compiler will be able to optimize most of your function away at compile time. For a good example of this later case, see the kmalloc() inline function.
"
is the expectation that cpu_class_is_omap1 compile time constant hence the actual compiled in code is smaller -candidate for inline?

Detailed discussion and alignment can be found at:
http://www.spinics.net/lists/linux-omap/thrd6.html

Search for:
[PATCH v2 07/11] OMAP2/3/4: DMA: HWMOD: Device registration

Quoted 5 lines
Quoted 5 lines

#ifdef CONFIG_ARCH_OMAP15XX
/* Returns 1 if the DMA module is in OMAP1510-compatible mode, 0

        if (cpu_class_is_omap1()) {

[...]

Quoted 32 lines
Quoted 22 lines
                u16 ccr;
arch/arm/plat-omap/dma.c +9 -4
-               ccr = dma_read(CCR(lch));
+               l = dma_read(CSDP1, lch);
+               l &= ~0x03;
+               l |= data_type;
+               dma_write(l, CSDP1, lch);
+
+               ccr = dma_read(CCR1, lch);
                ccr &= ~(1 << 5);
                if (sync_mode == OMAP_DMA_SYNC_FRAME)
                        ccr |= 1 << 5;
-               dma_write(ccr, CCR(lch));
+               dma_write(ccr, CCR1, lch);
 
-               ccr = dma_read(CCR2(lch));
+               ccr = dma_read(CCR1_2, lch);
                ccr &= ~(1 << 2);
                if (sync_mode == OMAP_DMA_SYNC_BLOCK)
                        ccr |= 1 << 2;
-               dma_write(ccr, CCR2(lch));
+               dma_write(ccr, CCR1_2, lch);

arch/arm/plat-omap/dma.c void omap_set_dma_transfer_params(int lch, int data_type, int elem_count, · line 453

Quoted 10 lines
arch/arm/plat-omap/dma.c +4 -1
        }
 
        if (cpu_class_is_omap2() && dma_trigger) {
                u32 val;
 
-               val = dma_read(CCR(lch));
+               l = dma_read(CSDP2, lch);
+               l &= ~0x03;
+               l |= data_type;
+               dma_write(l, CSDP2, lch);
Quoted 6 lines

This seems to me to be a replication mess :( compared to the original CSDP(lch) - which mapped to class1 or class2 as needed - same elsewhere.
Code is becoming pretty much needing a refactoring as a result :(

[...]

This is interim patch towards hwmod and driver cleanup. As mentioned in the patch description, objective is to replace lowlevel read/write macros with inline functions. For few registers(between omap1 and omap2+), the offset is different. Earlier, this offset difference was handled through macros. Now, it is replaced with inline functions, the code should be seperated for handling between omap1 and omap2+. It will get cleaned up once the code is moved respective mach-omapx directories.

You can have look at: PATCH 11/13

-Manjunath

#4 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

Quoted 3 lines

-----Original Message-----
From: G, Manjunath Kondaiah
Sent: Tuesday, October 26, 2010 10:55 PM

[..]

Quoted 29 lines
Quoted 9 lines

[...]

Review

Quoted 2 lines
(quoted diff) +0 -0

b/arch/arm/plat-omap/dma.c index

Quoted 5 lines

f5c5b8d..77241e2 100644

arch/arm/plat-omap/dma.c +0 -0
@@ -40,6 +40,147 @@
 

[...]

Quoted 6 lines
Quoted 5 lines
arch/arm/plat-omap/dma.c +5 -0
+static u16 reg_map_omap2[] = {
+       [REVISION]              = 0x00,
+       [GCR2]                  = 0x78,
+       [IRQSTATUS_L0]          = 0x08,
+       [IRQSTATUS_L1]          = 0x0c,

[..]

Quoted 12 lines

arch/arm/plat-omap/dma.c line 183

Quoted 6 lines
arch/arm/plat-omap/dma.c +6 -0
+       /* OMAP4 specific registers */
+       [CDP]                   = 0xd0,
+       [CNDP]                  = 0xd4,
+       [CCDN]                  = 0xd8,
+};
+

dumb question: any reason why a struct wont do?
struct reg_map_omap2 {
u16 revision;
...
...
}

Quoted 2 lines

This approach is as per i2c driver as suggested by kevin.
http://www.spinics.net/lists/linux-omap/msg36446.html

Thanks for pointing this out. I2c has what 18 registers while dma has over 40 registers :( patch 11 [1] now I understand this step -> this merges together at later patchset - it starts to make sense now. It becomes reg_map. Thanks - looks like a good change in the eventual code.

[...]

Quoted 14 lines
Quoted 14 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 7 lines
arch/arm/plat-omap/dma.c +7 -0
+static inline void dma_write(u32 val, int reg, int lch) {
+       if (cpu_class_is_omap1()) {
+               if (reg > OMAP1_CH_COMMON_START)
+                       __raw_writew(val, dma_base +
+                               (reg_map_omap1[reg] + 0x40 * lch));
+               else
+                       __raw_writew(val, dma_base +

reg_map_omap1[reg]);

Quoted 6 lines
arch/arm/plat-omap/dma.c +6 -0
+       } else {
+               if (reg > OMAP2_CH_COMMON_START)
+                       __raw_writel(val, dma_base +
+                               (reg_map_omap2[reg] + 0x60 * lch));
+               else
+                       __raw_writel(val, dma_base +
Quoted 3 lines
Quoted 3 lines

reg_map_omap2[reg]);

Quoted 2 lines
arch/arm/plat-omap/dma.c +2 -0
+       }
+}

Eventual code looks like this:

62 static inline void dma_write(u32 val, int reg, int lch) 63 {
64 if (d->dev_caps & IS_WORD_16) {
65 if (reg >= CH_COMMON_START)
66 __raw_writew(val, dma_base +
67 (reg_map[reg] + 0x40 * lch));
68 else
69 __raw_writew(val, dma_base + reg_map[reg]);
70 } else {
71 if (reg > CH_COMMON_START)
72 __raw_writel(val, dma_base +
73 (reg_map[reg] + 0x60 * lch));
74 else
75 __raw_writel(val, dma_base + reg_map[reg]);
76 }
77 }
I don't really see how inline will really help here!

Quoted 39 lines
Quoted 23 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 8 lines
arch/arm/plat-omap/dma.c +8 -0
+
+static inline u32 dma_read(int reg, int lch) {
+       u32 val;
+
+       if (cpu_class_is_omap1()) {
+               if (reg > OMAP1_CH_COMMON_START)
+                       val = __raw_readw(dma_base +
+                                       (reg_map_omap1[reg]

+ 0x40 * lch));

arch/arm/plat-omap/dma.c

Quoted 2 lines
arch/arm/plat-omap/dma.c +2 -0
+               else
+                       val = __raw_readw(dma_base +

reg_map_omap1[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 4 lines
arch/arm/plat-omap/dma.c +4 -0
+       } else {
+               if (reg > OMAP2_CH_COMMON_START)
+                       val = __raw_readl(dma_base +
+                                       (reg_map_omap2[reg]

+ 0x60 * lch));

arch/arm/plat-omap/dma.c

Quoted 2 lines
arch/arm/plat-omap/dma.c +2 -0
+               else
+                       val = __raw_readl(dma_base +

reg_map_omap2[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 333

Quoted 3 lines
arch/arm/plat-omap/dma.c +3 -0
+       }
+       return val;
+}
Quoted 15 lines

What is the benefit of using inline function here? would'nt we increase code size? cant we use a function pointer initialized to class1 or rest?
Quote from CodingStyle (15):
"A reasonable rule of thumb is to not put inline at functions that have more than 3 lines of code in them. An exception to this rule are the cases where a parameter is known to be a compiletime constant, and as a result of this constantness you *know* the compiler will be able to optimize most of your function away at compile time. For a good example of this later case, see the kmalloc() inline function.
"
is the expectation that cpu_class_is_omap1 compile time constant hence the actual compiled in code is smaller -candidate for inline?

Quoted 2 lines

Detailed discussion and alignment can be found at:
http://www.spinics.net/lists/linux-omap/thrd6.html

Better link:
http://marc.info/?t=128264802300006&r=1&w=2

Quoted 3 lines

Search for:
[PATCH v2 07/11] OMAP2/3/4: DMA: HWMOD: Device registration

http://marc.info/?l=linux-omap&m=128464661906497&w=2
my question is slightly different here - debate of suggestion to use inline is based on the size of code involved, the discussion in the thread discussed around 3 lines of code, which made sense, unfortunately, the thread does not answer my question unfortunately for *this* specific patch
- OR do you wish to point me to some specific link which answers this?

Quoted 58 lines
Quoted 7 lines
Quoted 7 lines

#ifdef CONFIG_ARCH_OMAP15XX
/* Returns 1 if the DMA module is in OMAP1510-compatible mode, 0

        if (cpu_class_is_omap1()) {

[...]

Quoted 32 lines
Quoted 22 lines
                u16 ccr;
arch/arm/plat-omap/dma.c +9 -4
-               ccr = dma_read(CCR(lch));
+               l = dma_read(CSDP1, lch);
+               l &= ~0x03;
+               l |= data_type;
+               dma_write(l, CSDP1, lch);
+
+               ccr = dma_read(CCR1, lch);
                ccr &= ~(1 << 5);
                if (sync_mode == OMAP_DMA_SYNC_FRAME)
                        ccr |= 1 << 5;
-               dma_write(ccr, CCR(lch));
+               dma_write(ccr, CCR1, lch);
 
-               ccr = dma_read(CCR2(lch));
+               ccr = dma_read(CCR1_2, lch);
                ccr &= ~(1 << 2);
                if (sync_mode == OMAP_DMA_SYNC_BLOCK)
                        ccr |= 1 << 2;
-               dma_write(ccr, CCR2(lch));
+               dma_write(ccr, CCR1_2, lch);

arch/arm/plat-omap/dma.c void omap_set_dma_transfer_params(int lch, int data_type, int elem_count, · line 453

Quoted 10 lines
arch/arm/plat-omap/dma.c +4 -1
        }
 
        if (cpu_class_is_omap2() && dma_trigger) {
                u32 val;
 
-               val = dma_read(CCR(lch));
+               l = dma_read(CSDP2, lch);
+               l &= ~0x03;
+               l |= data_type;
+               dma_write(l, CSDP2, lch);
Quoted 6 lines

This seems to me to be a replication mess :( compared to the original CSDP(lch) - which mapped to class1 or class2 as needed - same elsewhere.
Code is becoming pretty much needing a refactoring as a result :(

[...]

This is interim patch towards hwmod and driver cleanup. As mentioned in the
patch description, objective is to replace lowlevel read/write macros with inline functions. For few registers(between omap1 and omap2+), the offset is
different. Earlier, this offset difference was handled through macros. Now, it is replaced with inline functions, the code should be seperated for
handling between omap1 and omap2+. It will get cleaned up once the code is moved respective mach-omapx directories.

Quoted 1 line

You can have look at: PATCH 11/13

Ok looks like it never made to ML :( [1] as you refered to in a later patch Makes sense to me now. Thanks.

[1] http://dev.omapzoom.org/?p=manju/kernel-omap3-dev.git;a=commitdiff;h=7457a6b4248772aaa52dfb13a170f891596a512a
Regards,
Nishanth Menon

#5 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

Quoted 12 lines

-----Original Message-----
From: Menon, Nishanth
Sent: Wednesday, October 27, 2010 7:56 PM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org;
linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Kevin Hilman; Shilimkar, Santosh
Subject: RE: [PATCH v3 01/13] OMAP: DMA: Replace read/write macros with functions

[...]

Quoted 98 lines
Quoted 2 lines

This approach is as per i2c driver as suggested by kevin.
http://www.spinics.net/lists/linux-omap/msg36446.html

Thanks for pointing this out. I2c has what 18 registers while dma has over 40 registers :( patch 11 [1] now I understand this step -> this merges
together at later patchset - it starts to make sense now. It becomes reg_map. Thanks - looks like a good change in the eventual code.

[...]

Quoted 5 lines
Quoted 5 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 5 lines
arch/arm/plat-omap/dma.c +5 -0
+static inline void dma_write(u32 val, int reg, int lch) {
+       if (cpu_class_is_omap1()) {
+               if (reg > OMAP1_CH_COMMON_START)
+                       __raw_writew(val, dma_base +
+                               (reg_map_omap1[reg] + 

0x40 * lch));

Quoted 7 lines
Quoted 7 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+               else
+                       __raw_writew(val, dma_base +

reg_map_omap1[reg]);

Quoted 4 lines
arch/arm/plat-omap/dma.c +4 -0
+       } else {
+               if (reg > OMAP2_CH_COMMON_START)
+                       __raw_writel(val, dma_base +
+                               (reg_map_omap2[reg] + 

0x60 * lch));

Quoted 5 lines
Quoted 5 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 308

Quoted 2 lines
arch/arm/plat-omap/dma.c +2 -0
+               else
+                       __raw_writel(val, dma_base +

reg_map_omap2[reg]);

Quoted 2 lines
(quoted diff) +2 -0
+       }
+}

Eventual code looks like this:

62 static inline void dma_write(u32 val, int reg, int lch) 63 {
64 if (d->dev_caps & IS_WORD_16) {
65 if (reg >= CH_COMMON_START)
66 __raw_writew(val, dma_base +
67 (reg_map[reg] + 0x40 * lch));
68 else
69 __raw_writew(val, dma_base + reg_map[reg]);
70 } else {
71 if (reg > CH_COMMON_START)
72 __raw_writel(val, dma_base +
73 (reg_map[reg] + 0x60 * lch));
74 else
75 __raw_writel(val, dma_base + reg_map[reg]);
76 }
77 }
I don't really see how inline will really help here!

Quoted 39 lines
Quoted 23 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 8 lines
arch/arm/plat-omap/dma.c +8 -0
+
+static inline u32 dma_read(int reg, int lch) {
+       u32 val;
+
+       if (cpu_class_is_omap1()) {
+               if (reg > OMAP1_CH_COMMON_START)
+                       val = __raw_readw(dma_base +
+                                       (reg_map_omap1[reg]

+ 0x40 * lch));

Review

Quoted 2 lines
(quoted diff) +2 -0
+               else
+                       val = __raw_readw(dma_base +

reg_map_omap1[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 4 lines
arch/arm/plat-omap/dma.c +4 -0
+       } else {
+               if (reg > OMAP2_CH_COMMON_START)
+                       val = __raw_readl(dma_base +
+                                       (reg_map_omap2[reg]

+ 0x60 * lch));

Review

Quoted 2 lines
(quoted diff) +2 -0
+               else
+                       val = __raw_readl(dma_base +

reg_map_omap2[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 333

Quoted 3 lines
arch/arm/plat-omap/dma.c +3 -0
+       }
+       return val;
+}
Quoted 15 lines

What is the benefit of using inline function here? would'nt we increase code size? cant we use a function pointer initialized to class1 or rest?
Quote from CodingStyle (15):
"A reasonable rule of thumb is to not put inline at functions that have more than 3 lines of code in them. An exception to this rule are the cases where a parameter is known to be a compiletime constant, and as a result of this constantness you *know* the compiler will be able to optimize most of your function away at compile time. For a good example of this later case, see the kmalloc() inline function.
"
is the expectation that cpu_class_is_omap1 compile time constant hence the actual compiled in code is smaller -candidate for inline?

Quoted 2 lines

Detailed discussion and alignment can be found at:
http://www.spinics.net/lists/linux-omap/thrd6.html

Better link:
http://marc.info/?t=128264802300006&r=1&w=2

Quoted 3 lines

Search for:
[PATCH v2 07/11] OMAP2/3/4: DMA: HWMOD: Device registration

Quoted 9 lines

http://marc.info/?l=linux-omap&m=128464661906497&w=2
my question is slightly different here - debate of suggestion to use inline
is based on the size of code involved, the discussion in the thread discussed around 3 lines of code, which made sense, unfortunately, the
thread does not answer my question unfortunately for *this* specific patch
- OR do you wish to point me to some specific link which answers this?

for plat-omap dma code, these read/write api's can be replaced with function pointers.

-Manjunath

#6 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

In reply to G, Manjunath Kondaiah
Quoted 8 lines · Manjunath Kondaiah

The low level read/write macros are replaced with static inline functions and register offsets are handled through static register offset tables mapped through enumeration constants.

The objective of this patch is to prepare for omap dma driver cleanup and dma hwmod implementation. The code optimization and moving machine specific code to respective mach-omapx dma file will be handled in the rest of this patch series.

[...]

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 300

Quoted 10 lines
arch/arm/plat-omap/dma.c +3 -7
-#define dma_write(val, reg)						\
-({									\
-	if (cpu_class_is_omap1())					\
-		__raw_writew((u16)(val), omap_dma_base + OMAP1_DMA_##reg); \
-	else								\
-		__raw_writel((val), omap_dma_base + OMAP_DMA4_##reg);	\
-})
+static inline void dma_write(u32 val, int reg, int lch)
+{
+	if (cpu_class_is_omap1()) {

Reminder: any use of cpu_is_* checks outside of init code will be NAK'd.

cpu_is_* check do not belong at runtime, especially in a crucial path like this.

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 313

Quoted 13 lines
arch/arm/plat-omap/dma.c +13 -0
+		if (reg > OMAP1_CH_COMMON_START)
+			__raw_writew(val, dma_base +
+				(reg_map_omap1[reg] + 0x40 * lch));
+		else
+			__raw_writew(val, dma_base + reg_map_omap1[reg]);
+	} else {
+		if (reg > OMAP2_CH_COMMON_START)
+			__raw_writel(val, dma_base +
+				(reg_map_omap2[reg] + 0x60 * lch));
+		else
+			__raw_writel(val, dma_base + reg_map_omap2[reg]);
+	}
+}

The register base offset, register array and the stride (offset between
instances: 0x40 or 0x60) are detectable at init time, and there's no reason to have conditional code for them. IOW, they should be init time constants. Doing so would greatly simply these functions. In fact the CH_COMMON_START could also be an init time constant as well. So, given the following init_time constants: dma_ch_common_start, dma_stride, reg_map, the above would be reduced to something actually worth inlining, for example (not actually tested):

static inline void dma_write(u32 val, int reg, int lch)

{
        u8 stride = (reg > dma_ch_common_start) ? dma_stride : 0;

        __raw_writel(val, dma_base + (reg_map[reg] + stride * lch));
}

Same applies to dma_read().

Kevin

#7 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

Quoted 50 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 3:08 AM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 01/13] OMAP: DMA: Replace read/write macros with functions

Quoted 1 line · Manjunath Kondaiah

The low level read/write macros are replaced with static

inline functions

Quoted 1 line

and register offsets are handled through static register

offset tables

Quoted 3 lines

mapped through enumeration constants.

The objective of this patch is to prepare for omap dma

driver cleanup

Quoted 1 line

and dma hwmod implementation. The code optimization and

moving machine

Quoted 1 line

specific code to respective mach-omapx dma file will be

handled in the

Quoted 1 line

rest of this patch series.

[...]

Quoted 1 line

-#define dma_write(val, reg)

		\
Quoted 1 line

-({

	\

Review

Quoted 1 line
(quoted diff) +0 -1
-	if (cpu_class_is_omap1())				
	\

Review

Quoted 1 line
(quoted diff) +0 -1
-		__raw_writew((u16)(val), omap_dma_base + 

OMAP1_DMA_##reg); \

arch/arm/plat-omap/dma.c omap_dma_set_prio_lch(int lch, unsigned char read_prio, · line 910

Quoted 1 line
arch/arm/plat-omap/dma.c +0 -1
-	else							
	\

Review

Quoted 1 line
(quoted diff) +0 -1
-		__raw_writel((val), omap_dma_base + 

OMAP_DMA4_##reg); \

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 300

Quoted 4 lines
arch/arm/plat-omap/dma.c +3 -1
-})
+static inline void dma_write(u32 val, int reg, int lch)
+{
+	if (cpu_class_is_omap1()) {

Reminder: any use of cpu_is_* checks outside of init code will be NAK'd.

Quoted 2 lines

cpu_is_* check do not belong at runtime, especially in a crucial path like this.

ok. removed cpu_is_* checks and taken care in init.

Quoted 30 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 6 lines
arch/arm/plat-omap/dma.c +5 -0
+		if (reg > OMAP1_CH_COMMON_START)
+			__raw_writew(val, dma_base +
+				(reg_map_omap1[reg] + 0x40 * lch));
+		else
+			__raw_writew(val, dma_base + 

reg_map_omap1[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 6 lines
arch/arm/plat-omap/dma.c +6 -0
+	} else {
+		if (reg > OMAP2_CH_COMMON_START)
+			__raw_writel(val, dma_base +
+				(reg_map_omap2[reg] + 0x60 * lch));
+		else
+			__raw_writel(val, dma_base + 

reg_map_omap2[reg]);

Review

Quoted 2 lines
(quoted diff) +2 -0
+	}
+}

The register base offset, register array and the stride (offset between
instances: 0x40 or 0x60) are detectable at init time, and there's no reason to have conditional code for them. IOW, they should be init time
constants. Doing so would greatly simply these functions. In fact the
CH_COMMON_START could also be an init time constant as well. So, given
the following init_time constants: dma_ch_common_start, dma_stride, reg_map, the above would be reduced to something actually worth inlining, for example (not actually tested):

Quoted 8 lines

static inline void dma_write(u32 val, int reg, int lch)

{
        u8 stride = (reg > dma_ch_common_start) ? dma_stride : 0;

        __raw_writel(val, dma_base + (reg_map[reg] + stride * lch));
}

Same applies to dma_read().

Thanks for the suggestion. It is taken care along with Tony's comment for handling these read/write's between omap1 and omap2+.

Here is code snippet for handling both omap1(includes 16 bit registers) and omap2+

static inline void dma_write(u32 val, int reg, int lch)

{
        u8  stride;
        u32 offset;

        stride = (reg >= dma_common_ch_start) ? dma_stride : 0;
        offset = reg_map[reg] + (stride * lch);

        if (dma_stride  == 0x40) {
                __raw_writew(val, omap_dma_base + offset);
                if ((reg > CLNK_CTRL && reg < CCR2) ||
                                (reg > PCHD_ID && reg < CAPS_2)) {
                        u32 offset2 = reg_map[reg] + 2 + (stride * lch);
                        __raw_writew(val >> 16, omap_dma_base + offset2);
                }
        } else
                __raw_writel(val, omap_dma_base + offset);
}

static inline u32 dma_read(int reg, int lch)

{
        u8 stride;
        u32 offset, val;

        stride = (reg >= dma_common_ch_start) ? dma_stride : 0;
        offset = reg_map[reg] + (stride * lch);

        if (dma_stride  == 0x40) {
                val = __raw_readw(omap_dma_base + offset);
                if ((reg > CLNK_CTRL && reg < CCR2) ||
                                (reg > PCHD_ID && reg < CAPS_2)) {
                        u16 upper;
                        u32 offset2 = reg_map[reg] + 2 + (stride * lch);
                        upper = __raw_readw(omap_dma_base + offset2);
                        val |= (upper << 16);
                }
        } else
                val = __raw_readl(omap_dma_base + offset);
        return val;
}

-Manjunath

#8 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

"G, Manjunath Kondaiah" <[email protected]> writes:

[...]

Quoted 42 lines
Quoted 26 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 5 lines
arch/arm/plat-omap/dma.c +5 -0
+		if (reg > OMAP1_CH_COMMON_START)
+			__raw_writew(val, dma_base +
+				(reg_map_omap1[reg] + 0x40 * lch));
+		else
+			__raw_writew(val, dma_base + 

reg_map_omap1[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 6 lines
arch/arm/plat-omap/dma.c +6 -0
+	} else {
+		if (reg > OMAP2_CH_COMMON_START)
+			__raw_writel(val, dma_base +
+				(reg_map_omap2[reg] + 0x60 * lch));
+		else
+			__raw_writel(val, dma_base + 

reg_map_omap2[reg]);

Review

Quoted 2 lines
(quoted diff) +2 -0
+	}
+}

The register base offset, register array and the stride (offset between instances: 0x40 or 0x60) are detectable at init time, and there's no reason to have conditional code for them. IOW, they should be init time constants. Doing so would greatly simply these functions. In fact the CH_COMMON_START could also be an init time constant as well. So, given the following init_time constants:
dma_ch_common_start, dma_stride, reg_map, the above would be reduced to something actually worth inlining, for example (not actually tested):

Quoted 8 lines

static inline void dma_write(u32 val, int reg, int lch)

{
        u8 stride = (reg > dma_ch_common_start) ? dma_stride : 0;

        __raw_writel(val, dma_base + (reg_map[reg] + stride * lch));
}

Same applies to dma_read().

Thanks for the suggestion. It is taken care along with Tony's comment for handling these read/write's between omap1 and omap2+.

Here is code snippet for handling both omap1(includes 16 bit registers) and omap2+

static inline void dma_write(u32 val, int reg, int lch)

Quoted 8 lines
{
        u8  stride;
        u32 offset;

        stride = (reg >= dma_common_ch_start) ? dma_stride : 0;
        offset = reg_map[reg] + (stride * lch);

        if (dma_stride  == 0x40) {

The use of hard-coded constants still isn't right here. This is
bascally the same as a cpu_is_omap1 check. After you separate out the device files, I thought you had separate omap1 and omap2+ versions of these, so I don't follow the need for this.

Quoted 1 line
                __raw_writew(val, omap_dma_base + offset);

This could be moved before the 'if' and the 'else' clause removed.

Quoted 8 lines
                if ((reg > CLNK_CTRL && reg < CCR2) ||
                                (reg > PCHD_ID && reg < CAPS_2)) {
                        u32 offset2 = reg_map[reg] + 2 + (stride * lch);
                        __raw_writew(val >> 16, omap_dma_base + offset2);
                }
        } else
                __raw_writel(val, omap_dma_base + offset);
}

Kevin

#9 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

Quoted 70 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 9:33 PM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 01/13] OMAP: DMA: Replace read/write macros with functions

"G, Manjunath Kondaiah" <[email protected]> writes:

[...]

Quoted 3 lines
Quoted 3 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 3 lines
arch/arm/plat-omap/dma.c +3 -0
+		if (reg > OMAP1_CH_COMMON_START)
+			__raw_writew(val, dma_base +
+				(reg_map_omap1[reg] + 

0x40 * lch));

Quoted 7 lines
Quoted 7 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+		else
+			__raw_writew(val, dma_base + 

reg_map_omap1[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 4 lines
arch/arm/plat-omap/dma.c +4 -0
+	} else {
+		if (reg > OMAP2_CH_COMMON_START)
+			__raw_writel(val, dma_base +
+				(reg_map_omap2[reg] + 

0x60 * lch));

Quoted 13 lines
Quoted 13 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 308

Quoted 2 lines
arch/arm/plat-omap/dma.c +2 -0
+		else
+			__raw_writel(val, dma_base + 

reg_map_omap2[reg]);

Review

Quoted 2 lines
(quoted diff) +2 -0
+	}
+}

The register base offset, register array and the stride (offset between instances: 0x40 or 0x60) are detectable at init time, and there's no reason to have conditional code for them. IOW, they should be init time constants. Doing so would greatly simply these functions. In fact the CH_COMMON_START could also be an init time constant as well. So, given the following init_time constants:
dma_ch_common_start, dma_stride, reg_map, the above would

be reduced

Quoted 8 lines
Quoted 8 lines

to something actually worth inlining, for example (not actually tested):

static inline void dma_write(u32 val, int reg, int lch)

{
        u8 stride = (reg > dma_ch_common_start) ? dma_stride : 0;

        __raw_writel(val, dma_base + (reg_map[reg] + 

stride * lch));

Quoted 5 lines
Quoted 3 lines

}

Same applies to dma_read().

Thanks for the suggestion. It is taken care along with

Tony's comment

Quoted 13 lines

for handling these read/write's between omap1 and omap2+.

Here is code snippet for handling both omap1(includes 16 bit registers) and omap2+

static inline void dma_write(u32 val, int reg, int lch)

{
        u8  stride;
        u32 offset;

        stride = (reg >= dma_common_ch_start) ? dma_stride : 0;
        offset = reg_map[reg] + (stride * lch);
Quoted 6 lines
Quoted 1 line
        if (dma_stride  == 0x40) {

The use of hard-coded constants still isn't right here. This is
bascally the same as a cpu_is_omap1 check. After you separate out the device files, I thought you had separate omap1 and omap2+ versions of these, so I don't follow the need for this.

This code will be moved to respective mach-omapx dma files and this conditional check vanishes automatically in PATCH 10/13. Since this patch targets replacing read/write macros with inline functions, no functionality changes(except change in logic for handling 16bit registers for omap1) are done with new patch hence handling omap1 and omap2+ is
done this way.

I hope having the conditional check in this interim patch is ok.

-Manjunath

#10 lore raw

PATCH v3 1/13 OMAP: DMA: Replace read/write macros with functions

Quoted 75 lines · Manjunath Kondaiah
Quoted 68 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 9:33 PM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 01/13] OMAP: DMA: Replace read/write macros with functions

"G, Manjunath Kondaiah" <[email protected]> writes:

[...]

Quoted 3 lines
Quoted 3 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 3 lines
arch/arm/plat-omap/dma.c +3 -0
+		if (reg > OMAP1_CH_COMMON_START)
+			__raw_writew(val, dma_base +
+				(reg_map_omap1[reg] + 

0x40 * lch));

Quoted 7 lines
Quoted 7 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+		else
+			__raw_writew(val, dma_base + 

reg_map_omap1[reg]);

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch);

Quoted 4 lines
arch/arm/plat-omap/dma.c +4 -0
+	} else {
+		if (reg > OMAP2_CH_COMMON_START)
+			__raw_writel(val, dma_base +
+				(reg_map_omap2[reg] + 

0x60 * lch));

Quoted 13 lines
Quoted 13 lines

arch/arm/plat-omap/dma.c static inline void omap_enable_channel_irq(int lch); · line 308

Quoted 2 lines
arch/arm/plat-omap/dma.c +2 -0
+		else
+			__raw_writel(val, dma_base + 

reg_map_omap2[reg]);

Review

Quoted 2 lines
(quoted diff) +2 -0
+	}
+}

The register base offset, register array and the stride (offset between instances: 0x40 or 0x60) are detectable at init time, and there's no reason to have conditional code for them. IOW, they should be init time constants. Doing so would greatly simply these functions. In fact the CH_COMMON_START could also be an init time constant as well. So, given the following init_time constants:
dma_ch_common_start, dma_stride, reg_map, the above would

be reduced

Quoted 8 lines
Quoted 8 lines

to something actually worth inlining, for example (not actually tested):

static inline void dma_write(u32 val, int reg, int lch)

{
        u8 stride = (reg > dma_ch_common_start) ? dma_stride : 0;

        __raw_writel(val, dma_base + (reg_map[reg] + 

stride * lch));

Quoted 5 lines
Quoted 3 lines

}

Same applies to dma_read().

Thanks for the suggestion. It is taken care along with

Tony's comment

Quoted 13 lines

for handling these read/write's between omap1 and omap2+.

Here is code snippet for handling both omap1(includes 16 bit registers) and omap2+

static inline void dma_write(u32 val, int reg, int lch)

{
        u8  stride;
        u32 offset;

        stride = (reg >= dma_common_ch_start) ? dma_stride : 0;
        offset = reg_map[reg] + (stride * lch);
Quoted 6 lines
Quoted 1 line
        if (dma_stride  == 0x40) {

The use of hard-coded constants still isn't right here. This is
bascally the same as a cpu_is_omap1 check. After you separate out the device files, I thought you had separate omap1 and omap2+ versions of these, so I don't follow the need for this.

Quoted 8 lines

This code will be moved to respective mach-omapx dma files and this conditional check vanishes automatically in PATCH 10/13. Since this patch targets replacing read/write macros with inline functions, no functionality changes(except change in logic for handling 16bit registers for omap1) are done with new patch hence handling omap1 and omap2+ is
done this way.

I hope having the conditional check in this interim patch is ok.

Personally, I would rather not have an intermediate step, but if it makes the series smoother, I guess it's OK.

Kevin

#11 lore raw

PATCH v3 2/13 OMAP: DMA: Introduce errata handling feature

In reply to G, Manjunath Kondaiah

Implement errata handling to use flags instead of cpu_is_* and cpu_class_* in the code.

The errata flags are initialized at init time and during runtime we are using the errata variable (via the IS_DMA_ERRATA macro) to execute the required errata workaround.

Reused errata handling patch from Peter Ujfalusi <[email protected]>
https://patchwork.kernel.org/patch/231191/

Cc Peter Ujfalusi <[email protected]>
Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/dma.c              |  134 ++++++++++++++++++++++----------
 arch/arm/plat-omap/include/plat/dma.h |   11 +++
 2 files changed, 103 insertions(+), 42 deletions(-)
arch/arm/plat-omap/dma.c +92 -42
index 77241e2..0ff82d0 100644
@@ -195,6 +195,7 @@ enum { DMA_CHAIN_STARTED, DMA_CHAIN_NOTSTARTED };
 #define OMAP_FUNC_MUX_ARM_BASE		(0xfffe1000 + 0xec)
 
 static int enable_1510_mode;
+static u32 errata;
 
 static struct omap_dma_global_context_registers {
 	u32 dma_irqenable_l0;
@@ -1216,12 +1217,8 @@ void omap_start_dma(int lch)
 
 			cur_lch = next_lch;
 		} while (next_lch != -1);
-	} else if (cpu_is_omap242x() ||
-		(cpu_is_omap243x() &&  omap_type() <= OMAP2430_REV_ES1_0)) {
-
-		/* Errata: Need to write lch even if not using chaining */
+	} else if (IS_DMA_ERRATA(DMA_ERRATA_PARALLEL_CHANNELS))
 		dma_write(lch, CLNK_CTRL2, lch);
-	}
 
 	omap_enable_channel_irq(lch);
 
@@ -1231,17 +1228,8 @@ void omap_start_dma(int lch)
 		dma_write(l, CCR1, lch);
 	} else {
 		l = dma_read(CCR2, lch);
-		/*
-		 * Errata: Inter Frame DMA buffering issue (All OMAP2420 and
-		 * OMAP2430ES1.0): DMA will wrongly buffer elements if packing
-		 * and bursting is enabled. This might result in data gets
-		 * stalled in FIFO at the end of the block.
-		 * Workaround: DMA channels must have BUFFERING_DISABLED bit
-		 * set to guarantee no data will stay in the DMA FIFO in case
-		 * inter frame buffering occurs.
-		 */
-		if (cpu_is_omap2420() || (cpu_is_omap2430() &&
-				(omap_type() == OMAP2430_REV_ES1_0)))
+
+		if (IS_DMA_ERRATA(DMA_ERRATA_IFRAME_BUFFERING))
 			l |= OMAP_DMA_CCR_BUFFERING_DISABLE;
 
 		l |= OMAP_DMA_CCR_EN;
@@ -1253,14 +1241,14 @@ EXPORT_SYMBOL(omap_start_dma);
 
 void omap_stop_dma(int lch)
 {
-	u32 l;
+	u32 l = 0;
 
 	/* Disable all interrupts on the channel */
 	if (cpu_class_is_omap1())
 		dma_write(0, CICR1, lch);
 
-	/* OMAP3 Errata i541: sDMA FIFO draining does not finish */
-	if (cpu_is_omap34xx() && (l & OMAP_DMA_CCR_SEL_SRC_DST_SYNC)) {
+	if (IS_DMA_ERRATA(DMA_ERRATA_i541) &&
+			(l & OMAP_DMA_CCR_SEL_SRC_DST_SYNC)) {
 		int i = 0;
 		u32 sys_cf;
 
@@ -1367,11 +1355,7 @@ dma_addr_t omap_get_dma_src_pos(int lch)
 	else
 		offset = dma_read(CSAC2, lch);
 
-	/*
-	 * omap 3.2/3.3 erratum: sometimes 0 is returned if CSAC/CDAC is
-	 * read before the DMA controller finished disabling the channel.
-	 */
-	if (!cpu_is_omap15xx() && offset == 0) {
+	if (IS_DMA_ERRATA(DMA_ERRATA_3_3) && offset == 0) {
 		if (cpu_class_is_omap1())
 			offset = dma_read(CSAC1, lch);
 		else
@@ -1966,7 +1950,7 @@ int omap_stop_dma_chain_transfers(int chain_id)
 {
 	int *channels;
 	u32 l, i;
-	u32 sys_cf;
+	u32 sys_cf = 0;
 
 	/* Check for input params */
 	if (unlikely((chain_id < 0 || chain_id >= dma_lch_count))) {
@@ -1981,15 +1965,13 @@ int omap_stop_dma_chain_transfers(int chain_id)
 	}
 	channels = dma_linked_lch[chain_id].linked_dmach_q;
 
-	/*
-	 * DMA Errata:
-	 * Special programming model needed to disable DMA before end of block
-	 */
-	sys_cf = dma_read(OCP_SYSCONFIG, 0);
-	l = sys_cf;
-	/* Middle mode reg set no Standby */
-	l &= ~((1 << 12)|(1 << 13));
-	dma_write(l, OCP_SYSCONFIG, 0);
+	if (IS_DMA_ERRATA(DMA_ERRATA_i88)) {
+		sys_cf = dma_read(OCP_SYSCONFIG, 0);
+		l = sys_cf;
+		/* Middle mode reg set no Standby */
+		l &= ~((1 << 12)|(1 << 13));
+		dma_write(l, OCP_SYSCONFIG, 0);
+	}
 
 	for (i = 0; i < dma_linked_lch[chain_id].no_of_lchs_linked; i++) {
 
@@ -2008,8 +1990,8 @@ int omap_stop_dma_chain_transfers(int chain_id)
 	/* Reset the Queue pointers */
 	OMAP_DMA_CHAIN_QINIT(chain_id);
 
-	/* Errata - put in the old value */
-	dma_write(sys_cf, OCP_SYSCONFIG, 0);
+	if (IS_DMA_ERRATA(DMA_ERRATA_i88))
+		dma_write(sys_cf, OCP_SYSCONFIG, 0);
 
 	return 0;
 }
@@ -2215,12 +2197,7 @@ static int omap2_dma_handle_ch(int ch)
 	if (unlikely(status & OMAP2_DMA_TRANS_ERR_IRQ)) {
 		printk(KERN_INFO "DMA transaction error with device %d\n",
 		       dma_chan[ch].dev_id);
-		if (cpu_class_is_omap2()) {
-			/*
-			 * Errata: sDMA Channel is not disabled
-			 * after a transaction error. So we explicitely
-			 * disable the channel
-			 */
+		if (IS_DMA_ERRATA(DMA_ERRATA_i378)) {
 			u32 ccr;
 
 			ccr = dma_read(CCR2, ch);
@@ -2332,6 +2309,76 @@ void omap_dma_global_context_restore(void)
 			omap_clear_dma(ch);
 }
 
+static void configure_dma_errata(void)
+{
+
+	/*
+	 * Erratas applicable for OMAP2430ES1.0 and all omap2420
+	 *
+	 * I.
+	 * Errata ID: XX Inter Frame DMA buffering issue DMA will wrongly
+	 * buffer elements if packing and bursting is enabled. This might
+	 * result in data gets stalled in FIFO at the end of the block.
+	 * Workaround: DMA channels must have BUFFERING_DISABLED bit set to
+	 * guarantee no data will stay in the DMA FIFO in case inter frame
+	 * buffering occurs
+	 *
+	 * II.
+	 * Errata ID: XX DMA may hang when several channels are used in parallel
+	 * In the following configuration, DMA channel hanging can occur:
+	 * a. Channel i, hardware synchronized, is enabled
+	 * b. Another channel (Channel x), software synchronized, is enabled.
+	 * c. Channel i is disabled before end of transfer
+	 * d. Channel i is reenabled.
+	 * e. Steps 1 to 4 are repeated a certain number of times.
+	 * f. A third channel (Channel y), software synchronized, is enabled.
+	 * Channel x and Channel y may hang immediately after step 'f'.
+	 * Workaround:
+	 * For any channel used - make sure NextLCH_ID is set to the value j.
+	 */
+	if (cpu_is_omap2420() || (cpu_is_omap2430() &&
+				(omap_type() == OMAP2430_REV_ES1_0))) {
+
+		SET_DMA_ERRATA(DMA_ERRATA_IFRAME_BUFFERING);
+		SET_DMA_ERRATA(DMA_ERRATA_PARALLEL_CHANNELS);
+	}
+
+	/*
+	 * Errata ID: i378: OMAP2plus: sDMA Channel is not disabled
+	 * after a transaction error.
+	 * Workaround: SW should explicitely disable the channel.
+	 */
+	if (cpu_class_is_omap2())
+		SET_DMA_ERRATA(DMA_ERRATA_i378);
+
+	/*
+	 * Errata ID: i541: sDMA FIFO draining does not finish
+	 * If sDMA channel is disabled on the fly, sDMA enters standby even
+	 * through FIFO Drain is still in progress
+	 * Workaround: Put sDMA in NoStandby more before a logical channel is
+	 * disabled, then put it back to SmartStandby right after the channel
+	 * finishes FIFO draining.
+	 */
+	if (cpu_is_omap34xx())
+		SET_DMA_ERRATA(DMA_ERRATA_i541);
+
+	/*
+	 * Errata ID: i88 : Special programming model needed to disable DMA
+	 * before end of block.
+	 * Workaround: software must ensure that the DMA is configured in No
+	 * Standby mode(DMAx_OCP_SYSCONFIG.MIDLEMODE = "01")
+	 */
+	if (omap_type() == OMAP3430_REV_ES1_0)
+		SET_DMA_ERRATA(DMA_ERRATA_i88);
+
+	/*
+	 * Errata 3.2/3.3: sometimes 0 is returned if CSAC/CDAC is
+	 * read before the DMA controller finished disabling the channel.
+	 */
+	if (!cpu_is_omap15xx())
+		SET_DMA_ERRATA(DMA_ERRATA_3_3);
+}
+
 /*----------------------------------------------------------------------------*/
 
 static int __init omap_init_dma(void)
@@ -2481,6 +2528,9 @@ static int __init omap_init_dma(void)
 		}
 	}
 
+	/* Configure errata handling for all omap's */
+	configure_dma_errata();
+
 	return 0;
 
 out_free:
arch/arm/plat-omap/include/plat/dma.h +11 -0
index 27578f3..5e28d26 100644
@@ -285,6 +285,17 @@
 #define DMA_CH_PRIO_HIGH		0x1
 #define DMA_CH_PRIO_LOW			0x0 /* Def */
 
+/* Errata handling */
+#define IS_DMA_ERRATA(id)		(errata &= (id))
+#define SET_DMA_ERRATA(id)		(errata |= (id))
+
+#define DMA_ERRATA_IFRAME_BUFFERING	(1 << 0)
+#define DMA_ERRATA_PARALLEL_CHANNELS	(1 << 1)
+#define DMA_ERRATA_i378			(1 << 2)
+#define DMA_ERRATA_i541			(1 << 3)
+#define DMA_ERRATA_i88			(1 << 4)
+#define DMA_ERRATA_3_3			(1 << 5)
+
 enum omap_dma_burst_mode {
 	OMAP_DMA_DATA_BURST_DIS = 0,
 	OMAP_DMA_DATA_BURST_4,
Signature · 1 line
1.7.1
#12 lore raw

PATCH v3 2/13 OMAP: DMA: Introduce errata handling feature

Quoted 9 lines · Manjunath Kondaiah

Implement errata handling to use flags instead of cpu_is_* and cpu_class_* in the code.

The errata flags are initialized at init time and during runtime we are using the errata variable (via the IS_DMA_ERRATA macro) to execute the required errata workaround.

Reused errata handling patch from Peter Ujfalusi <[email protected]>
https://patchwork.kernel.org/patch/231191/

When starting from someone else's work, would be very helpful to reviewers (and original authors) if you summarized what you changed/fixed/updated etc.

Quoted 201 lines
Cc Peter Ujfalusi <[email protected]>
Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/dma.c              |  134 ++++++++++++++++++++++----------
 arch/arm/plat-omap/include/plat/dma.h |   11 +++
 2 files changed, 103 insertions(+), 42 deletions(-)
arch/arm/plat-omap/dma.c +73 -42
index 77241e2..0ff82d0 100644
@@ -195,6 +195,7 @@ enum { DMA_CHAIN_STARTED, DMA_CHAIN_NOTSTARTED };
 #define OMAP_FUNC_MUX_ARM_BASE		(0xfffe1000 + 0xec)
 
 static int enable_1510_mode;
+static u32 errata;
 
 static struct omap_dma_global_context_registers {
 	u32 dma_irqenable_l0;
@@ -1216,12 +1217,8 @@ void omap_start_dma(int lch)
 
 			cur_lch = next_lch;
 		} while (next_lch != -1);
-	} else if (cpu_is_omap242x() ||
-		(cpu_is_omap243x() &&  omap_type() <= OMAP2430_REV_ES1_0)) {
-
-		/* Errata: Need to write lch even if not using chaining */
+	} else if (IS_DMA_ERRATA(DMA_ERRATA_PARALLEL_CHANNELS))
 		dma_write(lch, CLNK_CTRL2, lch);
-	}
 
 	omap_enable_channel_irq(lch);
 
@@ -1231,17 +1228,8 @@ void omap_start_dma(int lch)
 		dma_write(l, CCR1, lch);
 	} else {
 		l = dma_read(CCR2, lch);
-		/*
-		 * Errata: Inter Frame DMA buffering issue (All OMAP2420 and
-		 * OMAP2430ES1.0): DMA will wrongly buffer elements if packing
-		 * and bursting is enabled. This might result in data gets
-		 * stalled in FIFO at the end of the block.
-		 * Workaround: DMA channels must have BUFFERING_DISABLED bit
-		 * set to guarantee no data will stay in the DMA FIFO in case
-		 * inter frame buffering occurs.
-		 */
-		if (cpu_is_omap2420() || (cpu_is_omap2430() &&
-				(omap_type() == OMAP2430_REV_ES1_0)))
+
+		if (IS_DMA_ERRATA(DMA_ERRATA_IFRAME_BUFFERING))
 			l |= OMAP_DMA_CCR_BUFFERING_DISABLE;
 
 		l |= OMAP_DMA_CCR_EN;
@@ -1253,14 +1241,14 @@ EXPORT_SYMBOL(omap_start_dma);
 
 void omap_stop_dma(int lch)
 {
-	u32 l;
+	u32 l = 0;
 
 	/* Disable all interrupts on the channel */
 	if (cpu_class_is_omap1())
 		dma_write(0, CICR1, lch);
 
-	/* OMAP3 Errata i541: sDMA FIFO draining does not finish */
-	if (cpu_is_omap34xx() && (l & OMAP_DMA_CCR_SEL_SRC_DST_SYNC)) {
+	if (IS_DMA_ERRATA(DMA_ERRATA_i541) &&
+			(l & OMAP_DMA_CCR_SEL_SRC_DST_SYNC)) {
 		int i = 0;
 		u32 sys_cf;
 
@@ -1367,11 +1355,7 @@ dma_addr_t omap_get_dma_src_pos(int lch)
 	else
 		offset = dma_read(CSAC2, lch);
 
-	/*
-	 * omap 3.2/3.3 erratum: sometimes 0 is returned if CSAC/CDAC is
-	 * read before the DMA controller finished disabling the channel.
-	 */
-	if (!cpu_is_omap15xx() && offset == 0) {
+	if (IS_DMA_ERRATA(DMA_ERRATA_3_3) && offset == 0) {
 		if (cpu_class_is_omap1())
 			offset = dma_read(CSAC1, lch);
 		else
@@ -1966,7 +1950,7 @@ int omap_stop_dma_chain_transfers(int chain_id)
 {
 	int *channels;
 	u32 l, i;
-	u32 sys_cf;
+	u32 sys_cf = 0;
 
 	/* Check for input params */
 	if (unlikely((chain_id < 0 || chain_id >= dma_lch_count))) {
@@ -1981,15 +1965,13 @@ int omap_stop_dma_chain_transfers(int chain_id)
 	}
 	channels = dma_linked_lch[chain_id].linked_dmach_q;
 
-	/*
-	 * DMA Errata:
-	 * Special programming model needed to disable DMA before end of block
-	 */
-	sys_cf = dma_read(OCP_SYSCONFIG, 0);
-	l = sys_cf;
-	/* Middle mode reg set no Standby */
-	l &= ~((1 << 12)|(1 << 13));
-	dma_write(l, OCP_SYSCONFIG, 0);
+	if (IS_DMA_ERRATA(DMA_ERRATA_i88)) {
+		sys_cf = dma_read(OCP_SYSCONFIG, 0);
+		l = sys_cf;
+		/* Middle mode reg set no Standby */
+		l &= ~((1 << 12)|(1 << 13));
+		dma_write(l, OCP_SYSCONFIG, 0);
+	}
 
 	for (i = 0; i < dma_linked_lch[chain_id].no_of_lchs_linked; i++) {
 
@@ -2008,8 +1990,8 @@ int omap_stop_dma_chain_transfers(int chain_id)
 	/* Reset the Queue pointers */
 	OMAP_DMA_CHAIN_QINIT(chain_id);
 
-	/* Errata - put in the old value */
-	dma_write(sys_cf, OCP_SYSCONFIG, 0);
+	if (IS_DMA_ERRATA(DMA_ERRATA_i88))
+		dma_write(sys_cf, OCP_SYSCONFIG, 0);
 	return 0;
 }
@@ -2215,12 +2197,7 @@ static int omap2_dma_handle_ch(int ch)
 	if (unlikely(status & OMAP2_DMA_TRANS_ERR_IRQ)) {
 		printk(KERN_INFO "DMA transaction error with device %d\n",
 		       dma_chan[ch].dev_id);
-		if (cpu_class_is_omap2()) {
-			/*
-			 * Errata: sDMA Channel is not disabled
-			 * after a transaction error. So we explicitely
-			 * disable the channel
-			 */
+		if (IS_DMA_ERRATA(DMA_ERRATA_i378)) {
 			u32 ccr;
 
 			ccr = dma_read(CCR2, ch);
@@ -2332,6 +2309,76 @@ void omap_dma_global_context_restore(void)
 			omap_clear_dma(ch);
 }
 
+static void configure_dma_errata(void)
+{
+
+	/*
+	 * Erratas applicable for OMAP2430ES1.0 and all omap2420
+	 *
+	 * I.
+	 * Errata ID: XX Inter Frame DMA buffering issue DMA will wrongly
+	 * buffer elements if packing and bursting is enabled. This might
+	 * result in data gets stalled in FIFO at the end of the block.
+	 * Workaround: DMA channels must have BUFFERING_DISABLED bit set to
+	 * guarantee no data will stay in the DMA FIFO in case inter frame
+	 * buffering occurs
+	 *
+	 * II.
+	 * Errata ID: XX DMA may hang when several channels are used in parallel
+	 * In the following configuration, DMA channel hanging can occur:
+	 * a. Channel i, hardware synchronized, is enabled
+	 * b. Another channel (Channel x), software synchronized, is enabled.
+	 * c. Channel i is disabled before end of transfer
+	 * d. Channel i is reenabled.
+	 * e. Steps 1 to 4 are repeated a certain number of times.
+	 * f. A third channel (Channel y), software synchronized, is enabled.
+	 * Channel x and Channel y may hang immediately after step 'f'.
+	 * Workaround:
+	 * For any channel used - make sure NextLCH_ID is set to the value j.
+	 */
+	if (cpu_is_omap2420() || (cpu_is_omap2430() &&
+				(omap_type() == OMAP2430_REV_ES1_0))) {
+
+		SET_DMA_ERRATA(DMA_ERRATA_IFRAME_BUFFERING);
+		SET_DMA_ERRATA(DMA_ERRATA_PARALLEL_CHANNELS);
+	}
+
+	/*
+	 * Errata ID: i378: OMAP2plus: sDMA Channel is not disabled
+	 * after a transaction error.
+	 * Workaround: SW should explicitely disable the channel.
+	 */
+	if (cpu_class_is_omap2())
+		SET_DMA_ERRATA(DMA_ERRATA_i378);
+
+	/*
+	 * Errata ID: i541: sDMA FIFO draining does not finish
+	 * If sDMA channel is disabled on the fly, sDMA enters standby even
+	 * through FIFO Drain is still in progress
+	 * Workaround: Put sDMA in NoStandby more before a logical channel is
+	 * disabled, then put it back to SmartStandby right after the channel
+	 * finishes FIFO draining.
+	 */
+	if (cpu_is_omap34xx())
+		SET_DMA_ERRATA(DMA_ERRATA_i541);
+
+	/*

arch/arm/plat-omap/dma.c void omap_dma_global_context_restore(void) · line 2375

Quoted 10 lines
arch/arm/plat-omap/dma.c +10 -0
+	 * Errata ID: i88 : Special programming model needed to disable DMA
+	 * before end of block.
+	 * Workaround: software must ensure that the DMA is configured in No
+	 * Standby mode(DMAx_OCP_SYSCONFIG.MIDLEMODE = "01")
+	 */
+	if (omap_type() == OMAP3430_REV_ES1_0)
+		SET_DMA_ERRATA(DMA_ERRATA_i88);
+
+	/*
+	 * Errata 3.2/3.3: sometimes 0 is returned if CSAC/CDAC is

This doesn't look like a unique errata identifier. Chapter 3 is usually the 'Cautions' section, and I don't listed in either the 34xx (v5.9), 36xx (v1.3) or 44xx (v1.3) errata docs under this number, and don't see
that number in either of the 24xx errata docs either. Please clarify.

Quoted 24 lines
arch/arm/plat-omap/dma.c +9 -0
+	 * read before the DMA controller finished disabling the channel.
+	 */
+	if (!cpu_is_omap15xx())
+		SET_DMA_ERRATA(DMA_ERRATA_3_3);
+}
+
 /*----------------------------------------------------------------------------*/
 
 static int __init omap_init_dma(void)
@@ -2481,6 +2528,9 @@ static int __init omap_init_dma(void)
 		}
 	}
 
+	/* Configure errata handling for all omap's */
+	configure_dma_errata();
+
 	return 0;
 
 out_free:
arch/arm/plat-omap/include/plat/dma.h +0 -0
index 27578f3..5e28d26 100644

arch/arm/plat-omap/include/plat/dma.h line 289

Quoted 5 lines
arch/arm/plat-omap/include/plat/dma.h +2 -0
@@ -285,6 +285,17 @@
 #define DMA_CH_PRIO_HIGH		0x1
 #define DMA_CH_PRIO_LOW			0x0 /* Def */
 
+/* Errata handling */
+#define IS_DMA_ERRATA(id)		(errata &= (id))

Why the '&=' here? I guess it should just be a '&' The '=' was not in Peter's original.

arch/arm/plat-omap/include/plat/dma.h line 297

Quoted 8 lines
arch/arm/plat-omap/include/plat/dma.h +8 -0
+#define SET_DMA_ERRATA(id)		(errata |= (id))
+
+#define DMA_ERRATA_IFRAME_BUFFERING	(1 << 0)
+#define DMA_ERRATA_PARALLEL_CHANNELS	(1 << 1)
+#define DMA_ERRATA_i378			(1 << 2)
+#define DMA_ERRATA_i541			(1 << 3)
+#define DMA_ERRATA_i88			(1 << 4)
+#define DMA_ERRATA_3_3			(1 << 5)

Please use BIT()

arch/arm/plat-omap/include/plat/dma.h line 301

Quoted 3 lines
arch/arm/plat-omap/include/plat/dma.h +0 -0
 enum omap_dma_burst_mode {
 	OMAP_DMA_DATA_BURST_DIS = 0,
 	OMAP_DMA_DATA_BURST_4,

Kevin

#13 lore raw

PATCH v3 2/13 OMAP: DMA: Introduce errata handling feature

Quoted 21 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 3:43 AM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Peter Ujfalusi; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 02/13] OMAP: DMA: Introduce errata handling feature

Quoted 7 lines · Manjunath Kondaiah

Implement errata handling to use flags instead of cpu_is_* and cpu_class_* in the code.

The errata flags are initialized at init time and during runtime we are using the errata variable (via the IS_DMA_ERRATA macro) to execute the required errata workaround.

Quoted 7 lines
Quoted 1 line

Reused errata handling patch from Peter Ujfalusi

<[email protected]>

Quoted 1 line

https://patchwork.kernel.org/patch/231191/

When starting from someone else's work, would be very helpful to reviewers (and original authors) if you summarized what you changed/fixed/updated etc.

ok. I will update patch summary.

Quoted 9 lines
Quoted 8 lines
Cc Peter Ujfalusi <[email protected]>
Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/dma.c              |  134 

++++++++++++++++++++++----------

[...]

Quoted 6 lines

arch/arm/plat-omap/dma.c void omap_dma_global_context_restore(void) · line 2375

Quoted 5 lines
arch/arm/plat-omap/dma.c +5 -0
+	if (omap_type() == OMAP3430_REV_ES1_0)
+		SET_DMA_ERRATA(DMA_ERRATA_i88);
+
+	/*
+	 * Errata 3.2/3.3: sometimes 0 is returned if CSAC/CDAC is
Quoted 7 lines

This doesn't look like a unique errata identifier. Chapter 3 is usually
the 'Cautions' section, and I don't listed in either the 34xx (v5.9), 36xx (v1.3) or 44xx (v1.3) errata docs under this number, and don't see
that number in either of the 24xx errata docs either. Please clarify.

Even I tried to find this info from errata docs prior to migrating all errata's to generic errata handling. This code was exising earlier, hence I retained it "as is", no documentation exists for this errata.

Quoted 32 lines

Review

Quoted 2 lines
(quoted diff) +1 -0
+	 * read before the DMA controller finished disabling 

the channel.

arch/arm/plat-omap/dma.c void omap_dma_global_context_restore(void) · line 2381

Quoted 6 lines
arch/arm/plat-omap/dma.c +5 -0
+	 */
+	if (!cpu_is_omap15xx())
+		SET_DMA_ERRATA(DMA_ERRATA_3_3);
+}
+

/*------------------------------------------------------------ ----------------*/

Review

Quoted 13 lines

static int __init omap_init_dma(void)

arch/arm/plat-omap/dma.c +3 -0
@@ -2481,6 +2528,9 @@ static int __init omap_init_dma(void)
 		}
 	}
 
+	/* Configure errata handling for all omap's */
+	configure_dma_errata();
+
 	return 0;
 
 out_free:
(quoted diff) +0 -0

b/arch/arm/plat-omap/include/plat/dma.h

Quoted 7 lines

index 27578f3..5e28d26 100644

arch/arm/plat-omap/include/plat/dma.h +0 -0
@@ -285,6 +285,17 @@
 #define DMA_CH_PRIO_HIGH		0x1
 #define DMA_CH_PRIO_LOW			0x0 /* Def */
 
Quoted 6 lines

arch/arm/plat-omap/include/plat/dma.h line 289

Quoted 2 lines
arch/arm/plat-omap/include/plat/dma.h +2 -0
+/* Errata handling */
+#define IS_DMA_ERRATA(id)		(errata &= (id))

Why the '&=' here? I guess it should just be a '&' The '=' was not in
Peter's original.

correct. There was bug in peter's original patch for SET_DMA_ERRATA(id) it was | instead |=, while fixing set_dma_errata, accidentally, I have changed is_dma_errata also. Will fix it.

Quoted 11 lines

arch/arm/plat-omap/include/plat/dma.h line 297

Quoted 9 lines
arch/arm/plat-omap/include/plat/dma.h +8 -0
+#define SET_DMA_ERRATA(id)		(errata |= (id))
+
+#define DMA_ERRATA_IFRAME_BUFFERING	(1 << 0)
+#define DMA_ERRATA_PARALLEL_CHANNELS	(1 << 1)
+#define DMA_ERRATA_i378			(1 << 2)
+#define DMA_ERRATA_i541			(1 << 3)
+#define DMA_ERRATA_i88			(1 << 4)
+#define DMA_ERRATA_3_3			(1 << 5)

Please use BIT()

ok.

-Manjunath

#14 lore raw

PATCH v3 2/13 OMAP: DMA: Introduce errata handling feature

"G, Manjunath Kondaiah" <[email protected]> writes:

[...]

Quoted 14 lines
Quoted 6 lines

arch/arm/plat-omap/dma.c void omap_dma_global_context_restore(void) · line 2375

Quoted 5 lines
arch/arm/plat-omap/dma.c +5 -0
+	if (omap_type() == OMAP3430_REV_ES1_0)
+		SET_DMA_ERRATA(DMA_ERRATA_i88);
+
+	/*
+	 * Errata 3.2/3.3: sometimes 0 is returned if CSAC/CDAC is
Quoted 7 lines

This doesn't look like a unique errata identifier. Chapter 3 is usually
the 'Cautions' section, and I don't listed in either the 34xx (v5.9), 36xx (v1.3) or 44xx (v1.3) errata docs under this number, and don't see
that number in either of the 24xx errata docs either. Please clarify.

Quoted 4 lines

Even I tried to find this info from errata docs prior to migrating all errata's to generic errata handling. This code was exising earlier, hence I retained it "as is", no documentation exists for this errata.

Hmm, OK.

According to the code, it seems to affect every IP execpt 1510, yet it's not documented in any of the errata docs.

I wonder what happens if it is removed?

Kevin

#15 lore raw

PATCH v3 2/13 OMAP: DMA: Introduce errata handling feature

Quoted 41 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 9:56 PM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Peter Ujfalusi; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 02/13] OMAP: DMA: Introduce errata handling feature

"G, Manjunath Kondaiah" <[email protected]> writes:

[...]

Quoted 5 lines
Quoted 5 lines

arch/arm/plat-omap/dma.c void omap_dma_global_context_restore(void)

Quoted 5 lines
arch/arm/plat-omap/dma.c +5 -0
+	if (omap_type() == OMAP3430_REV_ES1_0)
+		SET_DMA_ERRATA(DMA_ERRATA_i88);
+
+	/*
+	 * Errata 3.2/3.3: sometimes 0 is returned if 

CSAC/CDAC is

Quoted 4 lines
Quoted 4 lines

This doesn't look like a unique errata identifier. Chapter 3 is usually
the 'Cautions' section, and I don't listed in either the

34xx (v5.9),

Quoted 7 lines
Quoted 4 lines

36xx (v1.3) or 44xx (v1.3) errata docs under this number, and don't see
that number in either of the 24xx errata docs either. Please clarify.

Even I tried to find this info from errata docs prior to migrating all errata's to generic errata handling. This code was

exising earlier,

Quoted 1 line

hence I retained it "as is", no documentation exists for

this errata.

Quoted 1 line
Quoted 7 lines

Hmm, OK.

According to the code, it seems to affect every IP execpt 1510, yet it's
not documented in any of the errata docs.

I wonder what happens if it is removed?

This errata work around is used in two API's:
1. omap_get_dma_src_pos
2. omap_get_dma_dst_pos
which are used in sound and usb drivers.

we can run sound and usb unit tests which are using above two API's and if there no issues, drop this errata work around?

-Manjunath

#16 lore raw

PATCH v3 3/13 OMAP: DMA: Introduce DMA device attributes

In reply to G, Manjunath Kondaiah

This patch introduces OMAP DMA device attributes for using the same in DMA platform driver for all OMAP's and HWMOD database(OMAP2PLUS onwards)

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/include/plat/dma.h |   20 ++++++++++++++++++++
 1 files changed, 20 insertions(+), 0 deletions(-)
arch/arm/plat-omap/include/plat/dma.h +20 -0
index 5e28d26..9757b22 100644
@@ -296,6 +296,22 @@
 #define DMA_ERRATA_i88			(1 << 4)
 #define DMA_ERRATA_3_3			(1 << 5)
 
+/* Attributes for OMAP DMA Contrllers */
+#define ENABLE_1510_MODE		(1 << 0x0)
+#define DMA_LINKED_LCH			(1 << 0x1)
+#define GLOBAL_PRIORITY			(1 << 0x2)
+#define RESERVE_CHANNEL			(1 << 0x3)
+#define SRC_PORT			(1 << 0x4)
+#define DST_PORT			(1 << 0x5)
+#define IS_CSSA_32			(1 << 0x6)
+#define IS_CDSA_32			(1 << 0x7)
+#define SRC_INDEX			(1 << 0x8)
+#define DST_INDEX			(1 << 0x9)
+#define IS_BURST_ONLY4			(1 << 0xA)
+#define CLEAR_CSR_ON_READ		(1 << 0xB)
+#define IS_WORD_16			(1 << 0xC)
+#define IS_RW_PRIORITY			(1 << 0xD)
+
 enum omap_dma_burst_mode {
 	OMAP_DMA_DATA_BURST_DIS = 0,
 	OMAP_DMA_DATA_BURST_4,
@@ -361,6 +377,10 @@ struct omap_dma_channel_params {
 #endif
 };
 
+struct omap_dma_dev_attr {
+	u32 dev_caps;
+	u16 lch_count;
+};
 
 extern void omap_set_dma_priority(int lch, int dst_port, int priority);
 extern int omap_request_dma(int dev_id, const char *dev_name,
Signature · 1 line
1.7.1
#17 lore raw

PATCH v3 3/13 OMAP: DMA: Introduce DMA device attributes

Quoted 24 lines · Manjunath Kondaiah

This patch introduces OMAP DMA device attributes for using the same in DMA platform driver for all OMAP's and HWMOD database(OMAP2PLUS onwards)

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/include/plat/dma.h |   20 ++++++++++++++++++++
 1 files changed, 20 insertions(+), 0 deletions(-)
arch/arm/plat-omap/include/plat/dma.h +5 -0
index 5e28d26..9757b22 100644
@@ -296,6 +296,22 @@
 #define DMA_ERRATA_i88			(1 << 4)
 #define DMA_ERRATA_3_3			(1 << 5)
 
+/* Attributes for OMAP DMA Contrllers */
+#define ENABLE_1510_MODE		(1 << 0x0)
+#define DMA_LINKED_LCH			(1 << 0x1)
+#define GLOBAL_PRIORITY			(1 << 0x2)
+#define RESERVE_CHANNEL			(1 << 0x3)

arch/arm/plat-omap/include/plat/dma.h line 313

Quoted 10 lines
arch/arm/plat-omap/include/plat/dma.h +10 -0
+#define SRC_PORT			(1 << 0x4)
+#define DST_PORT			(1 << 0x5)
+#define IS_CSSA_32			(1 << 0x6)
+#define IS_CDSA_32			(1 << 0x7)
+#define SRC_INDEX			(1 << 0x8)
+#define DST_INDEX			(1 << 0x9)
+#define IS_BURST_ONLY4			(1 << 0xA)
+#define CLEAR_CSR_ON_READ		(1 << 0xB)
+#define IS_WORD_16			(1 << 0xC)
+#define IS_RW_PRIORITY			(1 << 0xD)

Please use BIT() for these.

Quoted 4 lines
arch/arm/plat-omap/include/plat/dma.h +0 -0
 enum omap_dma_burst_mode {
 	OMAP_DMA_DATA_BURST_DIS = 0,
 	OMAP_DMA_DATA_BURST_4,

arch/arm/plat-omap/include/plat/dma.h struct omap_dma_channel_params { · line 386

Quoted 10 lines
arch/arm/plat-omap/include/plat/dma.h +4 -0
@@ -361,6 +377,10 @@ struct omap_dma_channel_params {
 #endif
 };
 
+struct omap_dma_dev_attr {
+	u32 dev_caps;
+	u16 lch_count;
+};
 
 extern void omap_set_dma_priority(int lch, int dst_port, int priority);
 extern int omap_request_dma(int dev_id, const char *dev_name,

Kevin

#18 lore raw

PATCH v3 4/13 OMAP2420: DMA: hwmod: add system DMA

In reply to G, Manjunath Kondaiah

Add OMAP2420 DMA hwmod structures.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap2/omap_hwmod_2420_data.c |   85 ++++++++++++++++++++++++++++
 1 files changed, 85 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_2420_data.c +85 -0
index a1a3dd6..3a51392 100644
@@ -38,6 +38,7 @@ static struct omap_hwmod omap2420_iva_hwmod;
 static struct omap_hwmod omap2420_l3_main_hwmod;
 static struct omap_hwmod omap2420_l4_core_hwmod;
 static struct omap_hwmod omap2420_wd_timer2_hwmod;
+static struct omap_hwmod omap2420_dma_system_hwmod;
 
 /* L3 -> L4_CORE interface */
 static struct omap_hwmod_ocp_if omap2420_l3_main__l4_core = {
@@ -557,6 +558,89 @@ static struct omap_hwmod omap2420_i2c2_hwmod = {
 	.flags		= HWMOD_16BIT_REG,
 };
 
+/* system dma */
+static struct omap_hwmod_class_sysconfig omap2420_dma_sysc = {
+	.rev_offs	= 0x0000,
+	.sysc_offs	= 0x002c,
+	.syss_offs	= 0x0028,
+	.sysc_flags	= (SYSC_HAS_SIDLEMODE | SYSC_HAS_SOFTRESET |
+			   SYSC_HAS_MIDLEMODE | SYSC_HAS_CLOCKACTIVITY |
+			   SYSC_HAS_EMUFREE | SYSC_HAS_AUTOIDLE),
+	.idlemodes	= (SIDLE_FORCE | SIDLE_NO | SIDLE_SMART |
+			   MSTANDBY_FORCE | MSTANDBY_NO | MSTANDBY_SMART),
+	.sysc_fields	= &omap_hwmod_sysc_type1,
+};
+
+static struct omap_hwmod_class omap2420_dma_hwmod_class = {
+	.name = "dma",
+	.sysc = &omap2420_dma_sysc,
+};
+
+/* dma attributes */
+static struct omap_dma_dev_attr dma_dev_attr = {
+	.dev_caps  = DMA_LINKED_LCH | GLOBAL_PRIORITY |
+				IS_CSSA_32 | IS_CDSA_32,
+	.lch_count = OMAP_DMA4_LOGICAL_DMA_CH_COUNT,
+};
+
+static struct omap_hwmod_irq_info omap2420_dma_system_irqs[] = {
+	{ .name = "0", .irq = INT_24XX_SDMA_IRQ0 },
+	{ .name = "1", .irq = INT_24XX_SDMA_IRQ1 },
+	{ .name = "2", .irq = INT_24XX_SDMA_IRQ2 },
+	{ .name = "3", .irq = INT_24XX_SDMA_IRQ3 },
+};
+
+static struct omap_hwmod_addr_space omap2420_dma_system_addrs[] = {
+	{
+		.pa_start	= 0x48056000,
+		.pa_end		= 0x4a0560ff,
+		.flags		= ADDR_TYPE_RT
+	},
+};
+
+/* dma_system -> L3 */
+static struct omap_hwmod_ocp_if omap2420_dma_system__l3 = {
+	.master		= &omap2420_dma_system_hwmod,
+	.slave		= &omap2420_l3_main_hwmod,
+	.clk		= "l3_div_ck",
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma_system master ports */
+static struct omap_hwmod_ocp_if *omap2420_dma_system_masters[] = {
+	&omap2420_dma_system__l3,
+};
+
+/* l4_cfg -> dma_system */
+static struct omap_hwmod_ocp_if omap2420_l4_core__dma_system = {
+	.master		= &omap2420_l4_core_hwmod,
+	.slave		= &omap2420_dma_system_hwmod,
+	.clk		= "l4_div_ck",
+	.addr		= omap2420_dma_system_addrs,
+	.addr_cnt	= ARRAY_SIZE(omap2420_dma_system_addrs),
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma_system slave ports */
+static struct omap_hwmod_ocp_if *omap2420_dma_system_slaves[] = {
+	&omap2420_l4_core__dma_system,
+};
+
+static struct omap_hwmod omap2420_dma_system_hwmod = {
+	.name		= "dma",
+	.class		= &omap2420_dma_hwmod_class,
+	.mpu_irqs	= omap2420_dma_system_irqs,
+	.mpu_irqs_cnt	= ARRAY_SIZE(omap2420_dma_system_irqs),
+	.main_clk	= "l3_div_ck",
+	.slaves		= omap2420_dma_system_slaves,
+	.slaves_cnt	= ARRAY_SIZE(omap2420_dma_system_slaves),
+	.masters	= omap2420_dma_system_masters,
+	.masters_cnt	= ARRAY_SIZE(omap2420_dma_system_masters),
+	.dev_attr	= &dma_dev_attr,
+	.omap_chip	= OMAP_CHIP_INIT(CHIP_IS_OMAP2420),
+	.flags		= HWMOD_NO_IDLEST,
+};
+
 static __initdata struct omap_hwmod *omap2420_hwmods[] = {
 	&omap2420_l3_main_hwmod,
 	&omap2420_l4_core_hwmod,
@@ -569,6 +653,7 @@ static __initdata struct omap_hwmod *omap2420_hwmods[] = {
 	&omap2420_uart3_hwmod,
 	&omap2420_i2c1_hwmod,
 	&omap2420_i2c2_hwmod,
+	&omap2420_dma_system_hwmod,
 	NULL,
 };
 
Signature · 1 line
1.7.1
#19 lore raw

PATCH v3 4/13 OMAP2420: DMA: hwmod: add system DMA

Quoted 1 line · Manjunath Kondaiah

Add OMAP2420 DMA hwmod structures.

s/structures/data/

Also, as pointed out in many other hwmod reviews, when adding hwmod data, subject should be something like:

OMAP2420: hwmod_data: add system DMA

Kevin

Quoted 117 lines
Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap2/omap_hwmod_2420_data.c |   85 ++++++++++++++++++++++++++++
 1 files changed, 85 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_2420_data.c +85 -0
index a1a3dd6..3a51392 100644
@@ -38,6 +38,7 @@ static struct omap_hwmod omap2420_iva_hwmod;
 static struct omap_hwmod omap2420_l3_main_hwmod;
 static struct omap_hwmod omap2420_l4_core_hwmod;
 static struct omap_hwmod omap2420_wd_timer2_hwmod;
+static struct omap_hwmod omap2420_dma_system_hwmod;
 
 /* L3 -> L4_CORE interface */
 static struct omap_hwmod_ocp_if omap2420_l3_main__l4_core = {
@@ -557,6 +558,89 @@ static struct omap_hwmod omap2420_i2c2_hwmod = {
 	.flags		= HWMOD_16BIT_REG,
 };
 
+/* system dma */
+static struct omap_hwmod_class_sysconfig omap2420_dma_sysc = {
+	.rev_offs	= 0x0000,
+	.sysc_offs	= 0x002c,
+	.syss_offs	= 0x0028,
+	.sysc_flags	= (SYSC_HAS_SIDLEMODE | SYSC_HAS_SOFTRESET |
+			   SYSC_HAS_MIDLEMODE | SYSC_HAS_CLOCKACTIVITY |
+			   SYSC_HAS_EMUFREE | SYSC_HAS_AUTOIDLE),
+	.idlemodes	= (SIDLE_FORCE | SIDLE_NO | SIDLE_SMART |
+			   MSTANDBY_FORCE | MSTANDBY_NO | MSTANDBY_SMART),
+	.sysc_fields	= &omap_hwmod_sysc_type1,
+};
+
+static struct omap_hwmod_class omap2420_dma_hwmod_class = {
+	.name = "dma",
+	.sysc = &omap2420_dma_sysc,
+};
+
+/* dma attributes */
+static struct omap_dma_dev_attr dma_dev_attr = {
+	.dev_caps  = DMA_LINKED_LCH | GLOBAL_PRIORITY |
+				IS_CSSA_32 | IS_CDSA_32,
+	.lch_count = OMAP_DMA4_LOGICAL_DMA_CH_COUNT,
+};
+
+static struct omap_hwmod_irq_info omap2420_dma_system_irqs[] = {
+	{ .name = "0", .irq = INT_24XX_SDMA_IRQ0 },
+	{ .name = "1", .irq = INT_24XX_SDMA_IRQ1 },
+	{ .name = "2", .irq = INT_24XX_SDMA_IRQ2 },
+	{ .name = "3", .irq = INT_24XX_SDMA_IRQ3 },
+};
+
+static struct omap_hwmod_addr_space omap2420_dma_system_addrs[] = {
+	{
+		.pa_start	= 0x48056000,
+		.pa_end		= 0x4a0560ff,
+		.flags		= ADDR_TYPE_RT
+	},
+};
+
+/* dma_system -> L3 */
+static struct omap_hwmod_ocp_if omap2420_dma_system__l3 = {
+	.master		= &omap2420_dma_system_hwmod,
+	.slave		= &omap2420_l3_main_hwmod,
+	.clk		= "l3_div_ck",
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma_system master ports */
+static struct omap_hwmod_ocp_if *omap2420_dma_system_masters[] = {
+	&omap2420_dma_system__l3,
+};
+
+/* l4_cfg -> dma_system */
+static struct omap_hwmod_ocp_if omap2420_l4_core__dma_system = {
+	.master		= &omap2420_l4_core_hwmod,
+	.slave		= &omap2420_dma_system_hwmod,
+	.clk		= "l4_div_ck",
+	.addr		= omap2420_dma_system_addrs,
+	.addr_cnt	= ARRAY_SIZE(omap2420_dma_system_addrs),
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma_system slave ports */
+static struct omap_hwmod_ocp_if *omap2420_dma_system_slaves[] = {
+	&omap2420_l4_core__dma_system,
+};
+
+static struct omap_hwmod omap2420_dma_system_hwmod = {
+	.name		= "dma",
+	.class		= &omap2420_dma_hwmod_class,
+	.mpu_irqs	= omap2420_dma_system_irqs,
+	.mpu_irqs_cnt	= ARRAY_SIZE(omap2420_dma_system_irqs),
+	.main_clk	= "l3_div_ck",
+	.slaves		= omap2420_dma_system_slaves,
+	.slaves_cnt	= ARRAY_SIZE(omap2420_dma_system_slaves),
+	.masters	= omap2420_dma_system_masters,
+	.masters_cnt	= ARRAY_SIZE(omap2420_dma_system_masters),
+	.dev_attr	= &dma_dev_attr,
+	.omap_chip	= OMAP_CHIP_INIT(CHIP_IS_OMAP2420),
+	.flags		= HWMOD_NO_IDLEST,
+};
+
 static __initdata struct omap_hwmod *omap2420_hwmods[] = {
 	&omap2420_l3_main_hwmod,
 	&omap2420_l4_core_hwmod,
@@ -569,6 +653,7 @@ static __initdata struct omap_hwmod *omap2420_hwmods[] = {
 	&omap2420_uart3_hwmod,
 	&omap2420_i2c1_hwmod,
 	&omap2420_i2c2_hwmod,
+	&omap2420_dma_system_hwmod,
 	NULL,
 };
#20 lore raw

PATCH v3 4/13 OMAP2420: DMA: hwmod: add system DMA

Quoted 11 lines · Kevin Hilman
Quoted 1 line · Manjunath Kondaiah

Add OMAP2420 DMA hwmod structures.

s/structures/data/

Also, as pointed out in many other hwmod reviews, when adding hwmod data, subject should be something like:

OMAP2420: hwmod_data: add system DMA

Oops... no underscore here please. IOW:

OMAP2420: hwmod data: add system DMA

Sorry for the confusion,

Kevin

#21 lore raw

PATCH v3 5/13 OMAP2430: DMA: hwmod: add system DMA

In reply to G, Manjunath Kondaiah

Add OMAP2430 DMA hwmod structures.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap2/omap_hwmod_2430_data.c |   85 ++++++++++++++++++++++++++++
 1 files changed, 85 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_2430_data.c +85 -0
index 7cf0d3a..d3b7239 100644
@@ -38,6 +38,7 @@ static struct omap_hwmod omap2430_iva_hwmod;
 static struct omap_hwmod omap2430_l3_main_hwmod;
 static struct omap_hwmod omap2430_l4_core_hwmod;
 static struct omap_hwmod omap2430_wd_timer2_hwmod;
+static struct omap_hwmod omap2430_dma_system_hwmod;
 
 /* L3 -> L4_CORE interface */
 static struct omap_hwmod_ocp_if omap2430_l3_main__l4_core = {
@@ -569,6 +570,89 @@ static struct omap_hwmod omap2430_i2c2_hwmod = {
 	.omap_chip	= OMAP_CHIP_INIT(CHIP_IS_OMAP2430),
 };
 
+/* dma_system */
+static struct omap_hwmod_class_sysconfig omap2430_dma_sysc = {
+	.rev_offs	= 0x0000,
+	.sysc_offs	= 0x002c,
+	.syss_offs	= 0x0028,
+	.sysc_flags	= (SYSC_HAS_SIDLEMODE | SYSC_HAS_SOFTRESET |
+			   SYSC_HAS_MIDLEMODE | SYSC_HAS_CLOCKACTIVITY |
+			   SYSC_HAS_EMUFREE | SYSC_HAS_AUTOIDLE),
+	.idlemodes	= (SIDLE_FORCE | SIDLE_NO | SIDLE_SMART |
+			   MSTANDBY_FORCE | MSTANDBY_NO | MSTANDBY_SMART),
+	.sysc_fields	= &omap_hwmod_sysc_type1,
+};
+
+static struct omap_hwmod_class omap2430_dma_hwmod_class = {
+	.name = "dma",
+	.sysc = &omap2430_dma_sysc,
+};
+
+/* dma attributes */
+static struct omap_dma_dev_attr dma_dev_attr = {
+	.dev_caps  = DMA_LINKED_LCH | GLOBAL_PRIORITY |
+				IS_CSSA_32 | IS_CDSA_32 | IS_RW_PRIORITY,
+	.lch_count = OMAP_DMA4_LOGICAL_DMA_CH_COUNT,
+};
+
+static struct omap_hwmod_irq_info omap2430_dma_system_irqs[] = {
+	{ .name = "0", .irq = INT_24XX_SDMA_IRQ0 },
+	{ .name = "1", .irq = INT_24XX_SDMA_IRQ1 },
+	{ .name = "2", .irq = INT_24XX_SDMA_IRQ2 },
+	{ .name = "3", .irq = INT_24XX_SDMA_IRQ3 },
+};
+
+static struct omap_hwmod_addr_space omap2430_dma_system_addrs[] = {
+	{
+		.pa_start	= 0x48056000,
+		.pa_end		= 0x4a0560ff,
+		.flags		= ADDR_TYPE_RT
+	},
+};
+
+/* dma_system -> L3 */
+static struct omap_hwmod_ocp_if omap2430_dma_system__l3 = {
+	.master		= &omap2430_dma_system_hwmod,
+	.slave		= &omap2430_l3_main_hwmod,
+	.clk		= "l3_div_ck",
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma_system master ports */
+static struct omap_hwmod_ocp_if *omap2430_dma_system_masters[] = {
+	&omap2430_dma_system__l3,
+};
+
+/* l4_cfg -> dma_system */
+static struct omap_hwmod_ocp_if omap2430_l4_core__dma_system = {
+	.master		= &omap2430_l4_core_hwmod,
+	.slave		= &omap2430_dma_system_hwmod,
+	.clk		= "l4_div_ck",
+	.addr		= omap2430_dma_system_addrs,
+	.addr_cnt	= ARRAY_SIZE(omap2430_dma_system_addrs),
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma_system slave ports */
+static struct omap_hwmod_ocp_if *omap2430_dma_system_slaves[] = {
+	&omap2430_l4_core__dma_system,
+};
+
+static struct omap_hwmod omap2430_dma_system_hwmod = {
+	.name		= "dma",
+	.class		= &omap2430_dma_hwmod_class,
+	.mpu_irqs	= omap2430_dma_system_irqs,
+	.mpu_irqs_cnt	= ARRAY_SIZE(omap2430_dma_system_irqs),
+	.main_clk	= "l3_div_ck",
+	.slaves		= omap2430_dma_system_slaves,
+	.slaves_cnt	= ARRAY_SIZE(omap2430_dma_system_slaves),
+	.masters	= omap2430_dma_system_masters,
+	.masters_cnt	= ARRAY_SIZE(omap2430_dma_system_masters),
+	.dev_attr	= &dma_dev_attr,
+	.omap_chip	= OMAP_CHIP_INIT(CHIP_IS_OMAP2430),
+	.flags		= HWMOD_NO_IDLEST,
+};
+
 static __initdata struct omap_hwmod *omap2430_hwmods[] = {
 	&omap2430_l3_main_hwmod,
 	&omap2430_l4_core_hwmod,
@@ -581,6 +665,7 @@ static __initdata struct omap_hwmod *omap2430_hwmods[] = {
 	&omap2430_uart3_hwmod,
 	&omap2430_i2c1_hwmod,
 	&omap2430_i2c2_hwmod,
+	&omap2430_dma_system_hwmod,
 	NULL,
 };
 
Signature · 1 line
1.7.1
#22 lore raw

PATCH v3 6/13 OMAP3: DMA: hwmod: add system DMA

In reply to G, Manjunath Kondaiah

Add OMAP3 DMA hwmod structures.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap2/omap_hwmod_3xxx_data.c |   93 ++++++++++++++++++++++++++++
 1 files changed, 93 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +93 -0
index ed6bf4a..4a5231c 100644
@@ -45,6 +45,7 @@ static struct omap_hwmod omap3xxx_wd_timer2_hwmod;
 static struct omap_hwmod omap3xxx_i2c1_hwmod;
 static struct omap_hwmod omap3xxx_i2c2_hwmod;
 static struct omap_hwmod omap3xxx_i2c3_hwmod;
+static struct omap_hwmod omap3xxx_dma_system_hwmod;
 
 /* L3 -> L4_CORE interface */
 static struct omap_hwmod_ocp_if omap3xxx_l3_main__l4_core = {
@@ -736,6 +737,97 @@ static struct omap_hwmod omap3xxx_i2c3_hwmod = {
 	.omap_chip	= OMAP_CHIP_INIT(CHIP_IS_OMAP3430),
 };
 
+/* dma_system -> L3 */
+static struct omap_hwmod_ocp_if omap3xxx_dma_system__l3 = {
+	.master		= &omap3xxx_dma_system_hwmod,
+	.slave		= &omap3xxx_l3_main_hwmod,
+	.clk		= "core_l3_ick",
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma attributes */
+static struct omap_dma_dev_attr dma_dev_attr = {
+	.dev_caps  = DMA_LINKED_LCH | GLOBAL_PRIORITY |
+				IS_CSSA_32 | IS_CDSA_32 | IS_RW_PRIORITY,
+	.lch_count = OMAP_DMA4_LOGICAL_DMA_CH_COUNT,
+};
+
+static struct omap_hwmod_class_sysconfig omap3xxx_dma_sysc = {
+	.rev_offs	= 0x0000,
+	.sysc_offs	= 0x002c,
+	.syss_offs	= 0x0028,
+	.sysc_flags	= (SYSC_HAS_SIDLEMODE | SYSC_HAS_SOFTRESET |
+			   SYSC_HAS_MIDLEMODE | SYSC_HAS_CLOCKACTIVITY |
+			   SYSC_HAS_EMUFREE | SYSC_HAS_AUTOIDLE),
+	.idlemodes	= (SIDLE_FORCE | SIDLE_NO | SIDLE_SMART |
+			   MSTANDBY_FORCE | MSTANDBY_NO | MSTANDBY_SMART),
+	.sysc_fields	= &omap_hwmod_sysc_type1,
+};
+
+static struct omap_hwmod_class omap3xxx_dma_hwmod_class = {
+	.name = "dma",
+	.sysc = &omap3xxx_dma_sysc,
+};
+
+/* dma_system */
+static struct omap_hwmod_irq_info omap3xxx_dma_system_irqs[] = {
+	{ .name = "0", .irq = INT_24XX_SDMA_IRQ0 },
+	{ .name = "1", .irq = INT_24XX_SDMA_IRQ1 },
+	{ .name = "2", .irq = INT_24XX_SDMA_IRQ2 },
+	{ .name = "3", .irq = INT_24XX_SDMA_IRQ3 },
+};
+
+static struct omap_hwmod_addr_space omap3xxx_dma_system_addrs[] = {
+	{
+		.pa_start	= 0x48056000,
+		.pa_end		= 0x4a0560ff,
+		.flags		= ADDR_TYPE_RT
+	},
+};
+
+/* dma_system master ports */
+static struct omap_hwmod_ocp_if *omap3xxx_dma_system_masters[] = {
+	&omap3xxx_dma_system__l3,
+};
+
+/* l4_cfg -> dma_system */
+static struct omap_hwmod_ocp_if omap3xxx_l4_core__dma_system = {
+	.master		= &omap3xxx_l4_core_hwmod,
+	.slave		= &omap3xxx_dma_system_hwmod,
+	.clk		= "core_l4_ick",
+	.addr		= omap3xxx_dma_system_addrs,
+	.addr_cnt	= ARRAY_SIZE(omap3xxx_dma_system_addrs),
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma_system slave ports */
+static struct omap_hwmod_ocp_if *omap3xxx_dma_system_slaves[] = {
+	&omap3xxx_l4_core__dma_system,
+};
+
+static struct omap_hwmod omap3xxx_dma_system_hwmod = {
+	.name		= "dma",
+	.class		= &omap3xxx_dma_hwmod_class,
+	.mpu_irqs	= omap3xxx_dma_system_irqs,
+	.mpu_irqs_cnt	= ARRAY_SIZE(omap3xxx_dma_system_irqs),
+	.main_clk	= "core_l3_ick",
+	.prcm = {
+		.omap2 = {
+			.prcm_reg_id		= 1,
+			.module_bit		= OMAP3430_ST_SDMA_SHIFT,
+			.idlest_reg_id		= 1,
+			.idlest_idle_bit	= OMAP3430_ST_SDMA_SHIFT,
+		},
+	},
+	.slaves		= omap3xxx_dma_system_slaves,
+	.slaves_cnt	= ARRAY_SIZE(omap3xxx_dma_system_slaves),
+	.masters	= omap3xxx_dma_system_masters,
+	.masters_cnt	= ARRAY_SIZE(omap3xxx_dma_system_masters),
+	.dev_attr	= &dma_dev_attr,
+	.omap_chip	= OMAP_CHIP_INIT(CHIP_IS_OMAP3430),
+	.flags = HWMOD_SET_DEFAULT_CLOCKACT,
+};
+
 static __initdata struct omap_hwmod *omap3xxx_hwmods[] = {
 	&omap3xxx_l3_main_hwmod,
 	&omap3xxx_l4_core_hwmod,
@@ -751,6 +843,7 @@ static __initdata struct omap_hwmod *omap3xxx_hwmods[] = {
 	&omap3xxx_i2c1_hwmod,
 	&omap3xxx_i2c2_hwmod,
 	&omap3xxx_i2c3_hwmod,
+	&omap3xxx_dma_system_hwmod,
 	NULL,
 };
 
Signature · 1 line
1.7.1
#23 lore raw

PATCH v3 6/13 OMAP3: DMA: hwmod: add system DMA

Kevin/Benoit,

Quoted 24 lines

-----Original Message-----
From: linux-omap-owner at vger.kernel.org [mailto:linux-omap-owner at vger.kernel.org] On Behalf Of G, Manjunath Kondaiah
Sent: Tuesday, October 26, 2010 6:55 PM
To: linux-omap at vger.kernel.org
Cc: linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Kevin Hilman; Shilimkar, Santosh
Subject: [PATCH v3 06/13] OMAP3: DMA: hwmod: add system DMA

Add OMAP3 DMA hwmod structures.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap2/omap_hwmod_3xxx_data.c |   93 
++++++++++++++++++++++++++++
 1 files changed, 93 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +0 -0
index ed6bf4a..4a5231c 100644

[...]

Quoted 8 lines
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +4 -0
+	.prcm = {
+		.omap2 = {
+			.prcm_reg_id		= 1,
+			.module_bit		= 

OMAP3430_ST_SDMA_SHIFT,

arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +2 -0
+			.idlest_reg_id		= 1,
+			.idlest_idle_bit	= 

OMAP3430_ST_SDMA_SHIFT,

Looks like the .module_offs field is missing here. Since ST_SDMA bit belongs to CM_IDLEST1_CORE register, .module_offs should be CORE_MOD.

But, it is observed that, using .module_offs results in warning dump during bootup. This is due to, ST_SDMA bit in standby mode. Since there is no fclk for DMA, this bit will never change till DMA channel is requested for data transfer.

We might have to bypass calling "omap2_cm_wait_module_ready" in "_wait_target_ready" in omap_hwmod.c

With exising flags, we can use "HWMOD_NO_IDLEST" flag, but this flag has different meaning. How about using new flag like "HWMOD_NO_FCLK_OPCLK" and if it is set, return without calling "omap2_cm_wait_module_ready"

Any better suggestion to handle this issue?

-Manjunath

#24 lore raw

PATCH v3 6/13 OMAP3: DMA: hwmod: add system DMA

Hi Manjunath,

Quoted 42 lines · Manjunath Kondaiah

Kevin/Benoit,

Quoted 18 lines

[mailto:[email protected]] On Behalf Of G, Manjunath Kondaiah
Sent: Tuesday, October 26, 2010 6:55 PM

Add OMAP3 DMA hwmod structures.

Cc Benoit Cousson<[email protected]>
Cc Kevin Hilman<[email protected]>
Cc Santosh Shilimkar<[email protected]>
---
  arch/arm/mach-omap2/omap_hwmod_3xxx_data.c |   93
++++++++++++++++++++++++++++
  1 files changed, 93 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +0 -0
index ed6bf4a..4a5231c 100644

[...]

Quoted 8 lines
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +4 -0
+	.prcm = {
+		.omap2 = {
+			.prcm_reg_id		= 1,
+			.module_bit		=

OMAP3430_ST_SDMA_SHIFT,

arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +2 -0
+			.idlest_reg_id		= 1,
+			.idlest_idle_bit	=

OMAP3430_ST_SDMA_SHIFT,

Looks like the .module_offs field is missing here. Since ST_SDMA bit belongs to CM_IDLEST1_CORE register, .module_offs should be CORE_MOD.

But, it is observed that, using .module_offs results in warning dump during bootup. This is due to, ST_SDMA bit in standby mode. Since there is no fclk for DMA, this bit will never change till DMA channel is requested for data transfer.

We might have to bypass calling "omap2_cm_wait_module_ready" in "_wait_target_ready" in omap_hwmod.c

Quoted 6 lines

With exising flags, we can use "HWMOD_NO_IDLEST" flag, but this flag has different meaning. How about using new flag like "HWMOD_NO_FCLK_OPCLK" and if it is set, return without calling "omap2_cm_wait_module_ready"

Any better suggestion to handle this issue?

Well, yes, use HWMOD_NO_IDLEST and remove idlest_reg_id & idlest_idle_bit entries in the omap2 structure.

ST_SDMA is the standby status of the SDMA, but there is no idle status on that IP, so you cannot do anything.
The standby will just reflect the activity of the master port of the DMA, but it will give you no information about the slave port and thus we do not have any module ready information.

Regards,
Benoit

#25 lore raw

PATCH v3 6/13 OMAP3: DMA: hwmod: add system DMA

Hi Benoit,

Quoted 64 lines

-----Original Message-----
From: Cousson, Benoit
Sent: Thursday, November 04, 2010 9:59 AM
To: G, Manjunath Kondaiah
Cc: Kevin Hilman; linux-arm-kernel at lists.infradead.org; Shilimkar, Santosh; linux-omap at vger.kernel.org
Subject: Re: [PATCH v3 06/13] OMAP3: DMA: hwmod: add system DMA

Hi Manjunath,

Quoted 34 lines · Manjunath Kondaiah

Kevin/Benoit,

Quoted 18 lines

[mailto:linux-omap-owner at vger.kernel.org] On Behalf Of G, Manjunath Kondaiah
Sent: Tuesday, October 26, 2010 6:55 PM

Add OMAP3 DMA hwmod structures.

Cc Benoit Cousson<[email protected]>
Cc Kevin Hilman<[email protected]>
Cc Santosh Shilimkar<[email protected]>
---
  arch/arm/mach-omap2/omap_hwmod_3xxx_data.c |   93
++++++++++++++++++++++++++++
  1 files changed, 93 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +0 -0
index ed6bf4a..4a5231c 100644

[...]

Quoted 8 lines
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +4 -0
+	.prcm = {
+		.omap2 = {
+			.prcm_reg_id		= 1,
+			.module_bit		=

OMAP3430_ST_SDMA_SHIFT,

arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +2 -0
+			.idlest_reg_id		= 1,
+			.idlest_idle_bit	=

OMAP3430_ST_SDMA_SHIFT,

Looks like the .module_offs field is missing here. Since ST_SDMA bit belongs to CM_IDLEST1_CORE register, .module_offs should be CORE_MOD.

Quoted 3 lines

But, it is observed that, using .module_offs results in warning dump during bootup. This is due to, ST_SDMA bit in standby mode. Since there is no fclk for DMA, this bit will never change till

DMA channel

Quoted 11 lines

is requested for data transfer.

We might have to bypass calling "omap2_cm_wait_module_ready" in "_wait_target_ready" in omap_hwmod.c

With exising flags, we can use "HWMOD_NO_IDLEST" flag, but this flag has different meaning. How about using new flag like "HWMOD_NO_FCLK_OPCLK" and if it is set, return without calling "omap2_cm_wait_module_ready"

Any better suggestion to handle this issue?

Well, yes, use HWMOD_NO_IDLEST and remove idlest_reg_id & idlest_idle_bit entries in the omap2 structure.

Quoted 7 lines

ST_SDMA is the standby status of the SDMA, but there is no idle status
on that IP, so you cannot do anything.
The standby will just reflect the activity of the master port of the DMA, but it will give you no information about the slave port and thus
we do not have any module ready information.

Thanks for the feedback.

This change is introduced as per kevin's review comment at:
http://www.mail-archive.com/linux-omap at vger.kernel.org/msg34278.html

I am ok to use "HWMOD_NO_IDLEST" if kevin also agrees

-Manjunath

#26 lore raw

PATCH v3 6/13 OMAP3: DMA: hwmod: add system DMA

Quoted 69 lines · Manjunath Kondaiah

Hi Benoit,

Quoted 59 lines

From: Cousson, Benoit
Sent: Thursday, November 04, 2010 9:59 AM

Hi Manjunath,

Quoted 34 lines · Manjunath Kondaiah

Kevin/Benoit,

Quoted 18 lines

[mailto:[email protected]] On Behalf Of G, Manjunath Kondaiah
Sent: Tuesday, October 26, 2010 6:55 PM

Add OMAP3 DMA hwmod structures.

Cc Benoit Cousson<[email protected]>
Cc Kevin Hilman<[email protected]>
Cc Santosh Shilimkar<[email protected]>
---
   arch/arm/mach-omap2/omap_hwmod_3xxx_data.c |   93
++++++++++++++++++++++++++++
   1 files changed, 93 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +0 -0
index ed6bf4a..4a5231c 100644

[...]

Quoted 8 lines
arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +4 -0
+	.prcm = {
+		.omap2 = {
+			.prcm_reg_id		= 1,
+			.module_bit		=

OMAP3430_ST_SDMA_SHIFT,

arch/arm/mach-omap2/omap_hwmod_3xxx_data.c +2 -0
+			.idlest_reg_id		= 1,
+			.idlest_idle_bit	=

OMAP3430_ST_SDMA_SHIFT,

Looks like the .module_offs field is missing here. Since ST_SDMA bit belongs to CM_IDLEST1_CORE register, .module_offs should be CORE_MOD.

Quoted 3 lines

But, it is observed that, using .module_offs results in warning dump during bootup. This is due to, ST_SDMA bit in standby mode. Since there is no fclk for DMA, this bit will never change till

DMA channel

Quoted 11 lines

is requested for data transfer.

We might have to bypass calling "omap2_cm_wait_module_ready" in "_wait_target_ready" in omap_hwmod.c

With exising flags, we can use "HWMOD_NO_IDLEST" flag, but this flag has different meaning. How about using new flag like "HWMOD_NO_FCLK_OPCLK" and if it is set, return without calling "omap2_cm_wait_module_ready"

Any better suggestion to handle this issue?

Well, yes, use HWMOD_NO_IDLEST and remove idlest_reg_id& idlest_idle_bit entries in the omap2 structure.

Quoted 7 lines

ST_SDMA is the standby status of the SDMA, but there is no idle status
on that IP, so you cannot do anything.
The standby will just reflect the activity of the master port of the DMA, but it will give you no information about the slave port and thus
we do not have any module ready information.

Quoted 4 lines

Thanks for the feedback.

This change is introduced as per kevin's review comment at:
http://www.mail-archive.com/[email protected]/msg34278.html

Well this register is confusing, because he uses the same naming convention but with a difference meaning:

ST_SDMA: System DMA standby status.
- 0x0: System DMA is active.
- 0x1: System DMA is in standby mode.

Quoted 1 line

I am ok to use "HWMOD_NO_IDLEST" if kevin also agrees

He will... Otherwise I will throw him into the New England Aquarium like Linus :-)

Benoit

#27 lore raw

PATCH v3 6/13 OMAP3: DMA: hwmod: add system DMA

On Thu, 2010-11-04 at 08:30 -0400, Cousson, Benoit wrote:

[...]

Quoted 19 lines
Quoted 9 lines
Quoted 9 lines
Quoted 1 line

Any better suggestion to handle this issue?

Well, yes, use HWMOD_NO_IDLEST and remove idlest_reg_id& idlest_idle_bit entries in the omap2 structure.

ST_SDMA is the standby status of the SDMA, but there is no idle status
on that IP, so you cannot do anything.
The standby will just reflect the activity of the master port of

the

Quoted 8 lines
Quoted 3 lines

DMA, but it will give you no information about the slave port and thus
we do not have any module ready information.

Thanks for the feedback.

This change is introduced as per kevin's review comment at:
http://www.mail-archive.com/linux-omap at vger.kernel.org/msg34278.html

Quoted 8 lines

Well this register is confusing, because he uses the same naming convention but with a difference meaning:

ST_SDMA: System DMA standby status.
- 0x0: System DMA is active.
- 0x1: System DMA is in standby mode.

Quoted 1 line

I am ok to use "HWMOD_NO_IDLEST" if kevin also agrees

I agree.

Quoted 2 lines

He will... Otherwise I will throw him into the New England Aquarium like Linus :-)

...if he tried, I think he would end up in the fish tank before me. ;)

Kevin

#28 lore raw

PATCH v3 7/13 OMAP4: DMA: hwmod: add system DMA

In reply to G, Manjunath Kondaiah

From: Benoit Cousson <[email protected]>

Add OMAP4 DMA hwmod structures.

---
 arch/arm/mach-omap2/omap_hwmod_44xx_data.c |   99 ++++++++++++++++++++++++++++
 1 files changed, 99 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/omap_hwmod_44xx_data.c +99 -0
index 0d5c6eb..f749f0d 100644
@@ -22,6 +22,7 @@
 
 #include <plat/omap_hwmod.h>
 #include <plat/cpu.h>
+#include <plat/dma.h>
 
 #include "omap_hwmod_common_data.h"
 
@@ -35,6 +36,7 @@
 #define OMAP44XX_DMA_REQ_START  1
 
 /* Backward references (IPs with Bus Master capability) */
+static struct omap_hwmod omap44xx_dma_system_hwmod;
 static struct omap_hwmod omap44xx_dmm_hwmod;
 static struct omap_hwmod omap44xx_emif_fw_hwmod;
 static struct omap_hwmod omap44xx_l3_instr_hwmod;
@@ -215,6 +217,14 @@ static struct omap_hwmod_ocp_if omap44xx_l3_main_1__l3_main_2 = {
 	.user		= OCP_USER_MPU | OCP_USER_SDMA,
 };
 
+/* dma_system -> l3_main_2 */
+static struct omap_hwmod_ocp_if omap44xx_dma_system__l3_main_2 = {
+	.master		= &omap44xx_dma_system_hwmod,
+	.slave		= &omap44xx_l3_main_2_hwmod,
+	.clk		= "l3_div_ck",
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
 /* l4_cfg -> l3_main_2 */
 static struct omap_hwmod_ocp_if omap44xx_l4_cfg__l3_main_2 = {
 	.master		= &omap44xx_l4_cfg_hwmod,
@@ -226,6 +236,7 @@ static struct omap_hwmod_ocp_if omap44xx_l4_cfg__l3_main_2 = {
 /* l3_main_2 slave ports */
 static struct omap_hwmod_ocp_if *omap44xx_l3_main_2_slaves[] = {
 	&omap44xx_l3_main_1__l3_main_2,
+	&omap44xx_dma_system__l3_main_2,
 	&omap44xx_l4_cfg__l3_main_2,
 };
 
@@ -1043,6 +1054,92 @@ static struct omap_hwmod omap44xx_uart4_hwmod = {
 	.omap_chip	= OMAP_CHIP_INIT(CHIP_IS_OMAP4430),
 };
 
+/*
+ * 'dma' class
+ * dma controller for data exchange between memory to memory (i.e. internal or
+ * external memory) and gp peripherals to memory or memory to gp peripherals
+ */
+
+static struct omap_hwmod_class_sysconfig omap44xx_dma_sysc = {
+	.rev_offs	= 0x0000,
+	.sysc_offs	= 0x002c,
+	.syss_offs	= 0x0028,
+	.sysc_flags	= (SYSC_HAS_SIDLEMODE | SYSC_HAS_SOFTRESET |
+			   SYSC_HAS_MIDLEMODE | SYSC_HAS_CLOCKACTIVITY |
+			   SYSC_HAS_EMUFREE | SYSC_HAS_AUTOIDLE),
+	.idlemodes	= (SIDLE_FORCE | SIDLE_NO | SIDLE_SMART |
+			   MSTANDBY_FORCE | MSTANDBY_NO | MSTANDBY_SMART),
+	.sysc_fields	= &omap_hwmod_sysc_type1,
+};
+
+static struct omap_hwmod_class omap44xx_dma_hwmod_class = {
+	.name = "dma",
+	.sysc = &omap44xx_dma_sysc,
+};
+
+/* dma attributes */
+static struct omap_dma_dev_attr dma_dev_attr = {
+	.dev_caps  = DMA_LINKED_LCH | GLOBAL_PRIORITY |
+				IS_CSSA_32 | IS_CDSA_32 | IS_RW_PRIORITY,
+	.lch_count = OMAP_DMA4_LOGICAL_DMA_CH_COUNT,
+};
+
+/* dma_system */
+static struct omap_hwmod_irq_info omap44xx_dma_system_irqs[] = {
+	{ .name = "0", .irq = 12 + OMAP44XX_IRQ_GIC_START },
+	{ .name = "1", .irq = 13 + OMAP44XX_IRQ_GIC_START },
+	{ .name = "2", .irq = 14 + OMAP44XX_IRQ_GIC_START },
+	{ .name = "3", .irq = 15 + OMAP44XX_IRQ_GIC_START },
+};
+
+static struct omap_hwmod_addr_space omap44xx_dma_system_addrs[] = {
+	{
+		.pa_start	= 0x4a056000,
+		.pa_end		= 0x4a0560ff,
+		.flags		= ADDR_TYPE_RT
+	},
+};
+
+/* dma_system master ports */
+static struct omap_hwmod_ocp_if *omap44xx_dma_system_masters[] = {
+	&omap44xx_dma_system__l3_main_2,
+};
+
+/* l4_cfg -> dma_system */
+static struct omap_hwmod_ocp_if omap44xx_l4_cfg__dma_system = {
+	.master		= &omap44xx_l4_cfg_hwmod,
+	.slave		= &omap44xx_dma_system_hwmod,
+	.clk		= "l4_div_ck",
+	.addr		= omap44xx_dma_system_addrs,
+	.addr_cnt	= ARRAY_SIZE(omap44xx_dma_system_addrs),
+	.user		= OCP_USER_MPU | OCP_USER_SDMA,
+};
+
+/* dma_system slave ports */
+static struct omap_hwmod_ocp_if *omap44xx_dma_system_slaves[] = {
+	&omap44xx_l4_cfg__dma_system,
+};
+
+static struct omap_hwmod omap44xx_dma_system_hwmod = {
+	.name		= "dma_system",
+	.class		= &omap44xx_dma_hwmod_class,
+	.mpu_irqs	= omap44xx_dma_system_irqs,
+	.mpu_irqs_cnt	= ARRAY_SIZE(omap44xx_dma_system_irqs),
+	.main_clk	= "l3_div_ck",
+	.prcm = {
+		.omap4 = {
+			.clkctrl_reg = OMAP4430_CM_SDMA_SDMA_CLKCTRL,
+		},
+	},
+	.slaves		= omap44xx_dma_system_slaves,
+	.slaves_cnt	= ARRAY_SIZE(omap44xx_dma_system_slaves),
+	.masters	= omap44xx_dma_system_masters,
+	.masters_cnt	= ARRAY_SIZE(omap44xx_dma_system_masters),
+	.dev_attr	= &dma_dev_attr,
+	.omap_chip	= OMAP_CHIP_INIT(CHIP_IS_OMAP4430),
+	.flags		= HWMOD_NO_IDLEST,
+};
+
 static __initdata struct omap_hwmod *omap44xx_hwmods[] = {
 	/* dmm class */
 	&omap44xx_dmm_hwmod,
@@ -1077,6 +1174,8 @@ static __initdata struct omap_hwmod *omap44xx_hwmods[] = {
 	&omap44xx_uart2_hwmod,
 	&omap44xx_uart3_hwmod,
 	&omap44xx_uart4_hwmod,
+	/* dma class */
+	&omap44xx_dma_system_hwmod,
 	NULL,
 };
 
Signature · 1 line
1.7.1
#29 lore raw

PATCH v3 8/13 OMAP1: DMA: Introduce DMA driver as platform device

In reply to G, Manjunath Kondaiah

Register OMAP1 DMA driver as platform device and add support for registering through platform device layer using resource structures.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap1/dma.c              |  182 ++++++++++++++++++++++++++++++++
 arch/arm/mach-omap1/include/mach/dma.h |   29 +++++
 2 files changed, 211 insertions(+), 0 deletions(-)
 create mode 100644 arch/arm/mach-omap1/dma.c
 create mode 100644 arch/arm/mach-omap1/include/mach/dma.h
arch/arm/mach-omap1/dma.c new file +182 -0
new file mode 100644
index 0000000..e756069
@@ -0,0 +1,182 @@
+/*
+ * dma.c - OMAP1/OMAP7xx-specific DMA code
+ *
+ * Copyright (C) 2003 - 2008 Nokia Corporation
+ * Author: Juha Yrj??l?? <[email protected]>
+ * DMA channel linking for 1610 by Samuel Ortiz <[email protected]>
+ * Graphics DMA and LCD DMA graphics tranformations
+ * by Imre Deak <[email protected]>
+ * OMAP2/3 support Copyright (C) 2004-2007 Texas Instruments, Inc.
+ * Some functions based on earlier dma-omap.c Copyright (C) 2001 RidgeRun, Inc.
+ *
+ * Copyright (C) 2010 Texas Instruments, Inc.
+ * Converted DMA library into platform driver
+ *                   - G, Manjunath Kondaiah <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/pm_runtime.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/spinlock.h>
+#include <linux/errno.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/device.h>
+
+#include <plat/dma.h>
+#include <plat/tc.h>
+
+#define OMAP1_DMA_BASE			(0xfffed800)
+
+static struct resource res[] __initdata = {
+	[0] = {
+		.start	= OMAP1_DMA_BASE,
+		.end	= OMAP1_DMA_BASE + SZ_2K - 1,
+		.flags	= IORESOURCE_MEM,
+	},
+	[1] = {
+		.name   = "0",
+		.start  = INT_DMA_CH0_6,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[2] = {
+		.name   = "1",
+		.start  = INT_DMA_CH1_7,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[3] = {
+		.name   = "2",
+		.start  = INT_DMA_CH2_8,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[4] = {
+		.name   = "3",
+		.start  = INT_DMA_CH3,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[5] = {
+		.name   = "4",
+		.start  = INT_DMA_CH4,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[6] = {
+		.name   = "5",
+		.start  = INT_DMA_CH5,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[7] = {
+		.name   = "6",
+		.start  = INT_DMA_LCD,
+		.flags  = IORESOURCE_IRQ,
+	},
+	/* irq's for omap16xx and omap7xx */
+	[8] = {
+		.name   = "7",
+		.start  = 53 + IH2_BASE,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[9] = {
+		.name   = "8",
+		.start  = 54 + IH2_BASE,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[10] = {
+		.name  = "9",
+		.start = 55 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[11] = {
+		.name  = "10",
+		.start = 56 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[12] = {
+		.name  = "11",
+		.start = 57 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[13] = {
+		.name  = "12",
+		.start = 58 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[14] = {
+		.name  = "13",
+		.start = 59 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[15] = {
+		.name  = "14",
+		.start = 60 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[16] = {
+		.name  = "15",
+		.start = 61 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[17] = {
+		.name  = "16",
+		.start = 62 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+};
+
+static int __init omap1_system_dma_init(void)
+{
+	struct omap_system_dma_plat_info	*p;
+	struct platform_device			*pdev;
+	int ret;
+
+	pdev = platform_device_alloc("omap_dma_system", 0);
+	if (!pdev) {
+		pr_err("%s: Unable to device alloc for dma\n",
+			__func__);
+		return -ENOMEM;
+	}
+
+	ret = platform_device_add_resources(pdev, res, ARRAY_SIZE(res));
+	if (ret) {
+		pr_err("%s: Unable to add resources for %s%d\n",
+			__func__, pdev->name, pdev->id);
+		goto exit_device_del;
+	}
+
+	p = kzalloc(sizeof(struct omap_system_dma_plat_info), GFP_KERNEL);
+	if (!p) {
+		dev_err(&pdev->dev, "%s: Unable to allocate 'p' for %s\n",
+			__func__, pdev->name);
+		ret = -ENOMEM;
+		goto exit_device_put;
+	}
+
+	ret = platform_device_add_data(pdev, p, sizeof(*p));
+	if (ret) {
+		dev_err(&pdev->dev, "%s: Unable to add resources for %s%d\n",
+			__func__, pdev->name, pdev->id);
+		goto exit_device_put;
+	}
+	ret = platform_device_add(pdev);
+	if (ret) {
+		dev_err(&pdev->dev, "%s: Unable to add resources for %s%d\n",
+			__func__, pdev->name, pdev->id);
+		goto exit_device_put;
+	}
+	return ret;
+
+exit_device_put:
+	platform_device_put(pdev);
+exit_device_del:
+	platform_device_del(pdev);
+
+	return ret;
+}
+arch_initcall(omap1_system_dma_init);
arch/arm/mach-omap1/include/mach/dma.h new file +29 -0
new file mode 100644
index 0000000..7949e3f
@@ -0,0 +1,29 @@
+/*
+ *  OMAP DMA controller register offsets.
+ *
+ *  Copyright (C) 2003 Nokia Corporation
+ *  Author: Juha Yrj??l?? <[email protected]>
+ *
+ *  Copyright (C) 2010 Texas Instruments
+ *  Converted DMA library into platform driver
+ *       by G, Manjunath Kondaiah <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#ifndef __ASM_ARCH_OMAP1_DMA_H
+#define __ASM_ARCH_OMAP1_DMA_H
+
+#endif /* __ASM_ARCH_OMAP1_DMA_H */
Signature · 1 line
1.7.1
#30 lore raw

PATCH v3 8/13 OMAP1: DMA: Introduce DMA driver as platform device

Quoted 34 lines · Manjunath Kondaiah

Register OMAP1 DMA driver as platform device and add support for registering through platform device layer using resource structures.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap1/dma.c              |  182 ++++++++++++++++++++++++++++++++
 arch/arm/mach-omap1/include/mach/dma.h |   29 +++++
 2 files changed, 211 insertions(+), 0 deletions(-)
 create mode 100644 arch/arm/mach-omap1/dma.c
 create mode 100644 arch/arm/mach-omap1/include/mach/dma.h
arch/arm/mach-omap1/dma.c new file +13 -0
new file mode 100644
index 0000000..e756069
@@ -0,0 +1,182 @@
+/*
+ * dma.c - OMAP1/OMAP7xx-specific DMA code
+ *
+ * Copyright (C) 2003 - 2008 Nokia Corporation
+ * Author: Juha Yrj?l? <[email protected]>
+ * DMA channel linking for 1610 by Samuel Ortiz <[email protected]>
+ * Graphics DMA and LCD DMA graphics tranformations
+ * by Imre Deak <[email protected]>
+ * OMAP2/3 support Copyright (C) 2004-2007 Texas Instruments, Inc.
+ * Some functions based on earlier dma-omap.c Copyright (C) 2001 RidgeRun, Inc.
+ *
+ * Copyright (C) 2010 Texas Instruments, Inc.
+ * Converted DMA library into platform driver

arch/arm/mach-omap1/dma.c line 23

Quoted 10 lines
arch/arm/mach-omap1/dma.c new file +10 -0
+ *                   - G, Manjunath Kondaiah <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/pm_runtime.h>

no pm_runtime API usage in this patch. Please add only when needed.

arch/arm/mach-omap1/dma.c line 27

Quoted 4 lines
arch/arm/mach-omap1/dma.c new file +4 -0
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/sched.h>

why is sched.h needed?

arch/arm/mach-omap1/dma.c line 28

Quoted 1 line
arch/arm/mach-omap1/dma.c new file +1 -0
+#include <linux/spinlock.h>

ditto

arch/arm/mach-omap1/dma.c line 32

Quoted 4 lines
arch/arm/mach-omap1/dma.c new file +4 -0
+#include <linux/errno.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/device.h>

When creating a new file like this, it's important to only include headers that are actually used/needed.

Quoted 174 lines
arch/arm/mach-omap1/dma.c new file +149 -0
+#include <plat/dma.h>
+#include <plat/tc.h>
+
+#define OMAP1_DMA_BASE			(0xfffed800)
+
+static struct resource res[] __initdata = {
+	[0] = {
+		.start	= OMAP1_DMA_BASE,
+		.end	= OMAP1_DMA_BASE + SZ_2K - 1,
+		.flags	= IORESOURCE_MEM,
+	},
+	[1] = {
+		.name   = "0",
+		.start  = INT_DMA_CH0_6,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[2] = {
+		.name   = "1",
+		.start  = INT_DMA_CH1_7,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[3] = {
+		.name   = "2",
+		.start  = INT_DMA_CH2_8,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[4] = {
+		.name   = "3",
+		.start  = INT_DMA_CH3,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[5] = {
+		.name   = "4",
+		.start  = INT_DMA_CH4,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[6] = {
+		.name   = "5",
+		.start  = INT_DMA_CH5,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[7] = {
+		.name   = "6",
+		.start  = INT_DMA_LCD,
+		.flags  = IORESOURCE_IRQ,
+	},
+	/* irq's for omap16xx and omap7xx */
+	[8] = {
+		.name   = "7",
+		.start  = 53 + IH2_BASE,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[9] = {
+		.name   = "8",
+		.start  = 54 + IH2_BASE,
+		.flags  = IORESOURCE_IRQ,
+	},
+	[10] = {
+		.name  = "9",
+		.start = 55 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[11] = {
+		.name  = "10",
+		.start = 56 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[12] = {
+		.name  = "11",
+		.start = 57 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[13] = {
+		.name  = "12",
+		.start = 58 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[14] = {
+		.name  = "13",
+		.start = 59 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[15] = {
+		.name  = "14",
+		.start = 60 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[16] = {
+		.name  = "15",
+		.start = 61 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+	[17] = {
+		.name  = "16",
+		.start = 62 + IH2_BASE,
+		.flags = IORESOURCE_IRQ,
+	},
+};
+
+static int __init omap1_system_dma_init(void)
+{
+	struct omap_system_dma_plat_info	*p;
+	struct platform_device			*pdev;
+	int ret;
+
+	pdev = platform_device_alloc("omap_dma_system", 0);
+	if (!pdev) {
+		pr_err("%s: Unable to device alloc for dma\n",
+			__func__);
+		return -ENOMEM;
+	}
+
+	ret = platform_device_add_resources(pdev, res, ARRAY_SIZE(res));
+	if (ret) {
+		pr_err("%s: Unable to add resources for %s%d\n",
+			__func__, pdev->name, pdev->id);
+		goto exit_device_del;
+	}
+
+	p = kzalloc(sizeof(struct omap_system_dma_plat_info), GFP_KERNEL);
+	if (!p) {
+		dev_err(&pdev->dev, "%s: Unable to allocate 'p' for %s\n",
+			__func__, pdev->name);
+		ret = -ENOMEM;
+		goto exit_device_put;
+	}
+
+	ret = platform_device_add_data(pdev, p, sizeof(*p));
+	if (ret) {
+		dev_err(&pdev->dev, "%s: Unable to add resources for %s%d\n",
+			__func__, pdev->name, pdev->id);
+		goto exit_device_put;
+	}
+	ret = platform_device_add(pdev);
+	if (ret) {
+		dev_err(&pdev->dev, "%s: Unable to add resources for %s%d\n",
+			__func__, pdev->name, pdev->id);
+		goto exit_device_put;
+	}
+	return ret;
+
+exit_device_put:
+	platform_device_put(pdev);
+exit_device_del:
+	platform_device_del(pdev);
+
+	return ret;
+}
+arch_initcall(omap1_system_dma_init);
arch/arm/mach-omap1/include/mach/dma.h new file +19 -0
new file mode 100644
index 0000000..7949e3f
@@ -0,0 +1,29 @@
+/*
+ *  OMAP DMA controller register offsets.
+ *
+ *  Copyright (C) 2003 Nokia Corporation
+ *  Author: Juha Yrj?l? <[email protected]>
+ *
+ *  Copyright (C) 2010 Texas Instruments
+ *  Converted DMA library into platform driver
+ *       by G, Manjunath Kondaiah <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.

arch/arm/mach-omap1/include/mach/dma.h line 29

Quoted 10 lines
arch/arm/mach-omap1/include/mach/dma.h new file +10 -0
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#ifndef __ASM_ARCH_OMAP1_DMA_H
+#define __ASM_ARCH_OMAP1_DMA_H
+
+#endif /* __ASM_ARCH_OMAP1_DMA_H */

Please don't create an empty header. Just create it in the series when (if) it's needed. Same goes for next patch.

I see you populate mach/dma.h later in the series, but I don't
understand why those values need to be in the header. More comments on
those patches...

Kevin

#31 lore raw

PATCH v3 8/13 OMAP1: DMA: Introduce DMA driver as platform device

Quoted 77 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 3:54 AM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 08/13] OMAP1: DMA: Introduce DMA driver as platform device

Quoted 10 lines · Manjunath Kondaiah

Register OMAP1 DMA driver as platform device and add support for registering through platform device layer using resource structures.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap1/dma.c              |  182 

++++++++++++++++++++++++++++++++

Quoted 11 lines
 arch/arm/mach-omap1/include/mach/dma.h |   29 +++++
 2 files changed, 211 insertions(+), 0 deletions(-)
 create mode 100644 arch/arm/mach-omap1/dma.c
 create mode 100644 arch/arm/mach-omap1/include/mach/dma.h
arch/arm/mach-omap1/dma.c new file +0 -0
new file mode 100644
index 0000000..e756069

arch/arm/mach-omap1/dma.c line 6

Quoted 6 lines
arch/arm/mach-omap1/dma.c new file +6 -0
@@ -0,0 +1,182 @@
+/*
+ * dma.c - OMAP1/OMAP7xx-specific DMA code
+ *
+ * Copyright (C) 2003 - 2008 Nokia Corporation
+ * Author: Juha Yrj?l? <[email protected]>
+ * DMA channel linking for 1610 by Samuel Ortiz 

<[email protected]>

arch/arm/mach-omap1/dma.c

Quoted 4 lines
arch/arm/mach-omap1/dma.c new file +4 -0
+ * Graphics DMA and LCD DMA graphics tranformations
+ * by Imre Deak <[email protected]>
+ * OMAP2/3 support Copyright (C) 2004-2007 Texas Instruments, Inc.
+ * Some functions based on earlier dma-omap.c Copyright 

(C) 2001 RidgeRun, Inc.

arch/arm/mach-omap1/dma.c

Quoted 6 lines
arch/arm/mach-omap1/dma.c new file +6 -0
+ *
+ * Copyright (C) 2010 Texas Instruments, Inc.
+ * Converted DMA library into platform driver
+ *                   - G, Manjunath Kondaiah <[email protected]>
+ *
+ * This program is free software; you can redistribute it 

and/or modify

arch/arm/mach-omap1/dma.c line 17

Quoted 1 line
arch/arm/mach-omap1/dma.c new file +1 -0
+ * it under the terms of the GNU General Public License 

version 2 as

arch/arm/mach-omap1/dma.c line 23

Quoted 6 lines
arch/arm/mach-omap1/dma.c new file +6 -0
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/pm_runtime.h>

no pm_runtime API usage in this patch. Please add only when needed.

arch/arm/mach-omap1/dma.c line 27

Quoted 4 lines
arch/arm/mach-omap1/dma.c new file +4 -0
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/sched.h>

why is sched.h needed?

arch/arm/mach-omap1/dma.c line 28

Quoted 1 line
arch/arm/mach-omap1/dma.c new file +1 -0
+#include <linux/spinlock.h>

ditto

Quoted 7 lines

arch/arm/mach-omap1/dma.c line 32

Quoted 4 lines
arch/arm/mach-omap1/dma.c new file +4 -0
+#include <linux/errno.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/device.h>

When creating a new file like this, it's important to only include headers that are actually used/needed.

Ok. I will check these headers. It was copy paste from old dma driver.

Quoted 4 lines
Quoted 4 lines
arch/arm/mach-omap1/dma.c new file +3 -0
+#include <plat/dma.h>
+#include <plat/tc.h>
+

[...]

Quoted 35 lines

Review

Quoted 1 line
(quoted diff) +0 -0

b/arch/arm/mach-omap1/include/mach/dma.h

Quoted 6 lines

new file mode 100644
index 0000000..7949e3f

arch/arm/mach-omap1/include/mach/dma.h new file +1 -0
@@ -0,0 +1,29 @@
+/*

arch/arm/mach-omap1/include/mach/dma.h line 11

Quoted 10 lines
arch/arm/mach-omap1/include/mach/dma.h new file +10 -0
+ *  OMAP DMA controller register offsets.
+ *
+ *  Copyright (C) 2003 Nokia Corporation
+ *  Author: Juha Yrj?l? <[email protected]>
+ *
+ *  Copyright (C) 2010 Texas Instruments
+ *  Converted DMA library into platform driver
+ *       by G, Manjunath Kondaiah <[email protected]>
+ *
+ * This program is free software; you can redistribute it 

and/or modify

arch/arm/mach-omap1/include/mach/dma.h line 12

Quoted 1 line
arch/arm/mach-omap1/include/mach/dma.h new file +1 -0
+ * it under the terms of the GNU General Public License as 

published by

arch/arm/mach-omap1/include/mach/dma.h line 13

Quoted 1 line
arch/arm/mach-omap1/include/mach/dma.h new file +1 -0
+ * the Free Software Foundation; either version 2 of the 

License, or

arch/arm/mach-omap1/include/mach/dma.h

Quoted 8 lines
arch/arm/mach-omap1/include/mach/dma.h new file +8 -0
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General 

Public License

arch/arm/mach-omap1/include/mach/dma.h line 23

Quoted 2 lines
arch/arm/mach-omap1/include/mach/dma.h new file +2 -0
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, 

MA 02111-1307 USA

Quoted 10 lines

arch/arm/mach-omap1/include/mach/dma.h line 29

Quoted 6 lines
arch/arm/mach-omap1/include/mach/dma.h new file +6 -0
+ */
+
+#ifndef __ASM_ARCH_OMAP1_DMA_H
+#define __ASM_ARCH_OMAP1_DMA_H
+
+#endif /* __ASM_ARCH_OMAP1_DMA_H */

Please don't create an empty header. Just create it in the series when
(if) it's needed. Same goes for next patch.

ok.

-Manjunath

Quoted 10 lines

I see you populate mach/dma.h later in the series, but I don't
understand why those values need to be in the header. More
comments on
those patches...

Kevin

#32 lore raw

PATCH v3 9/13 OMAP2+: DMA: hwmod: Device registration

In reply to G, Manjunath Kondaiah

Prepare omap2+ dma driver to use hwmod infrastructure so that DMA driver can register as platform device.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap2/dma.c              |   86 ++++++++++++++++++++++++++++++++
 arch/arm/mach-omap2/include/mach/dma.h |   32 ++++++++++++
 2 files changed, 118 insertions(+), 0 deletions(-)
 create mode 100644 arch/arm/mach-omap2/dma.c
 create mode 100644 arch/arm/mach-omap2/include/mach/dma.h
arch/arm/mach-omap2/dma.c new file +86 -0
new file mode 100644
index 0000000..30b20cd
@@ -0,0 +1,86 @@
+/*
+ * dma.c - OMAP2 specific DMA code
+ *
+ * Copyright (C) 2003 - 2008 Nokia Corporation
+ * Author: Juha Yrj??l?? <[email protected]>
+ * DMA channel linking for 1610 by Samuel Ortiz <[email protected]>
+ * Graphics DMA and LCD DMA graphics tranformations
+ * by Imre Deak <[email protected]>
+ * OMAP2/3 support Copyright (C) 2004-2007 Texas Instruments, Inc.
+ * Some functions based on earlier dma-omap.c Copyright (C) 2001 RidgeRun, Inc.
+ *
+ * Copyright (C) 2009 Texas Instruments
+ * Added OMAP4 support - Santosh Shilimkar <[email protected]>
+ *
+ * Copyright (C) 2010 Texas Instruments, Inc.
+ * Converted DMA library into platform driver
+ *	- G, Manjunath Kondaiah <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/pm_runtime.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/spinlock.h>
+#include <linux/errno.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/device.h>
+
+#include <plat/irqs.h>
+#include <plat/omap_hwmod.h>
+#include <plat/omap_device.h>
+#include <plat/dma.h>
+
+static struct omap_device_pm_latency omap2_dma_latency[] = {
+	{
+		.deactivate_func = omap_device_idle_hwmods,
+		.activate_func	 = omap_device_enable_hwmods,
+		.flags		 = OMAP_DEVICE_LATENCY_AUTO_ADJUST,
+	},
+};
+
+/* One time initializations */
+static int __init omap2_system_dma_init_dev(struct omap_hwmod *oh, void *unused)
+{
+	struct omap_device			*od;
+	struct omap_system_dma_plat_info	*p;
+	char					*name = "omap_system_dma";
+
+	p = kzalloc(sizeof(struct omap_system_dma_plat_info), GFP_KERNEL);
+	if (!p) {
+		pr_err("%s: Unable to allocate pdata for %s:%s\n",
+			__func__, name, oh->name);
+		return -ENOMEM;
+	}
+	od = omap_device_build(name, 0, oh, p, sizeof(*p),
+			omap2_dma_latency, ARRAY_SIZE(omap2_dma_latency), 0);
+
+	if (IS_ERR(od)) {
+		pr_err("%s: Cant build omap_device for %s:%s.\n",
+			__func__, name, oh->name);
+		kfree(p);
+		return IS_ERR(od);
+	}
+	kfree(p);
+
+	return 0;
+}
+
+static int __init omap2_system_dma_init(void)
+{
+	int ret;
+
+	ret = omap_hwmod_for_each_by_class("dma",
+			omap2_system_dma_init_dev, NULL);
+
+	return ret;
+}
+arch_initcall(omap2_system_dma_init);
arch/arm/mach-omap2/include/mach/dma.h new file +32 -0
new file mode 100644
index 0000000..d0a7d5b
@@ -0,0 +1,32 @@
+/*
+ *  OMAP DMA controller register offsets.
+ *
+ *  Copyright (C) 2003 Nokia Corporation
+ *  Author: Juha Yrj??l?? <[email protected]>
+ *
+ *  Copyright (C) 2009 Texas Instruments
+ *  Added OMAP4 support - Santosh Shilimkar <[email protected]>
+ *
+ *  Copyright (C) 2010 Texas Instruments
+ *  Converted DMA library into platform driver
+ *          - G, Manjunath Kondaiah <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#ifndef __ASM_ARCH_OMAP2_DMA_H
+#define __ASM_ARCH_OMAP2_DMA_H
+
+#endif /* __ASM_ARCH_OMAP2_DMA_H */
Signature · 1 line
1.7.1
#33 lore raw

PATCH v3 10/13 OMAP: DMA: Convert DMA library into DMA platform Driver

In reply to G, Manjunath Kondaiah

Convert DMA library into DMA platform driver and make use of platform data provided by HWMOD data base for OMAP2PLUS onwards. For OMAP1 processors, the DMA driver in mach-omap uses resource structures for getting platform data.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap1/Makefile           |    2 +-
 arch/arm/mach-omap1/dma.c              |  180 ++++++++++++++++++-
 arch/arm/mach-omap2/Makefile           |    2 +-
 arch/arm/mach-omap2/dma.c              |  209 +++++++++++++++++++++-
 arch/arm/mach-omap2/include/mach/dma.h |    3 +
 arch/arm/plat-omap/dma.c               |  321 +++++++++++---------------------
 arch/arm/plat-omap/include/plat/dma.h  |   50 ++++--
 7 files changed, 538 insertions(+), 229 deletions(-)
arch/arm/mach-omap1/Makefile +1 -1
index 9a304d8..b7dfc54 100644
@@ -3,7 +3,7 @@
 #
 
 # Common support
-obj-y := io.o id.o sram.o irq.o mux.o flash.o serial.o devices.o
+obj-y := io.o id.o sram.o irq.o mux.o flash.o serial.o devices.o dma.o
 obj-y += clock.o clock_data.o opp_data.o
 
 obj-$(CONFIG_OMAP_MCBSP) += mcbsp.o
arch/arm/mach-omap1/dma.c +178 -2

213 lines not on this page · raw message

arch/arm/mach-omap2/Makefile +1 -1
index 7352412..5d250c8 100644
@@ -4,7 +4,7 @@
 
 # Common support
 obj-y := id.o io.o control.o mux.o devices.o serial.o gpmc.o timer-gp.o pm.o \
-	 common.o
+	 common.o dma.o
 
 omap-2-3-common				= irq.o sdrc.o prm2xxx_3xxx.o
 hwmod-common				= omap_hwmod.o \
arch/arm/mach-omap2/dma.c +208 -1

238 lines not on this page · raw message

arch/arm/mach-omap2/include/mach/dma.h +3 -0
index d0a7d5b..d13c5c0 100644
@@ -29,4 +29,7 @@
 #ifndef __ASM_ARCH_OMAP2_DMA_H
 #define __ASM_ARCH_OMAP2_DMA_H
 
+/* Should be part of hwmod data base ? */
+#define OMAP_DMA4_LOGICAL_DMA_CH_COUNT	32	/* REVISIT: Is this 32 + 2? */
+
 #endif /* __ASM_ARCH_OMAP2_DMA_H */
arch/arm/plat-omap/dma.c +110 -211

413 lines not on this page · raw message

arch/arm/plat-omap/include/plat/dma.h +37 -13

67 lines not on this page · raw message

Signature · 1 line
1.7.1
#34 lore raw

PATCH v3 10/13 OMAP: DMA: Convert DMA library into DMA platform Driver

Quoted 9 lines · Manjunath Kondaiah

Convert DMA library into DMA platform driver and make use of platform data provided by HWMOD data base for OMAP2PLUS onwards. For OMAP1 processors, the DMA driver in mach-omap uses resource structures for getting platform data.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>

[...]

arch/arm/mach-omap2/include/mach/dma.h line 34

Quoted 11 lines
arch/arm/mach-omap2/include/mach/dma.h +3 -0
index d0a7d5b..d13c5c0 100644
@@ -29,4 +29,7 @@
 #ifndef __ASM_ARCH_OMAP2_DMA_H
 #define __ASM_ARCH_OMAP2_DMA_H
 
+/* Should be part of hwmod data base ? */
+#define OMAP_DMA4_LOGICAL_DMA_CH_COUNT	32	/* REVISIT: Is this 32 + 2? */
+

There are no users of this in this patch, all users are removed.

Kevin

#35 lore raw

PATCH v3 10/13 OMAP: DMA: Convert DMA library into DMA platform Driver

Quoted 35 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 3:56 AM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 10/13] OMAP: DMA: Convert DMA library into DMA platform Driver

Quoted 9 lines · Manjunath Kondaiah

Convert DMA library into DMA platform driver and make use of platform data provided by HWMOD data base for OMAP2PLUS onwards. For OMAP1 processors, the DMA driver in mach-omap uses resource structures for getting platform data.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>

[...]

Review

Quoted 1 line
(quoted diff) +0 -0

b/arch/arm/mach-omap2/include/mach/dma.h

Quoted 7 lines

index d0a7d5b..d13c5c0 100644

arch/arm/mach-omap2/include/mach/dma.h +0 -0
@@ -29,4 +29,7 @@
 #ifndef __ASM_ARCH_OMAP2_DMA_H
 #define __ASM_ARCH_OMAP2_DMA_H
 
Quoted 6 lines

arch/arm/mach-omap2/include/mach/dma.h line 33

Quoted 2 lines
arch/arm/mach-omap2/include/mach/dma.h +2 -0
+/* Should be part of hwmod data base ? */
+#define OMAP_DMA4_LOGICAL_DMA_CH_COUNT	32	/* 

REVISIT: Is this 32 + 2? */

arch/arm/mach-omap2/include/mach/dma.h line 34

Quoted 1 line
arch/arm/mach-omap2/include/mach/dma.h +1 -0
+

There are no users of this in this patch, all users are removed.

This macro is used in hwmod data base. Do you mean hard coding this value in hwmod db and remove header?

-Manjunath

#36 lore raw

PATCH v3 10/13 OMAP: DMA: Convert DMA library into DMA platform Driver

Quoted 36 lines · Manjunath Kondaiah
Quoted 36 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 3:56 AM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 10/13] OMAP: DMA: Convert DMA library into DMA platform Driver

Quoted 9 lines · Manjunath Kondaiah

Convert DMA library into DMA platform driver and make use of platform data provided by HWMOD data base for OMAP2PLUS onwards. For OMAP1 processors, the DMA driver in mach-omap uses resource structures for getting platform data.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>

[...]

Review

Quoted 1 line
(quoted diff) +0 -0

b/arch/arm/mach-omap2/include/mach/dma.h

arch/arm/mach-omap2/include/mach/dma.h line 33

Quoted 9 lines

index d0a7d5b..d13c5c0 100644

arch/arm/mach-omap2/include/mach/dma.h +2 -0
@@ -29,4 +29,7 @@
 #ifndef __ASM_ARCH_OMAP2_DMA_H
 #define __ASM_ARCH_OMAP2_DMA_H
 
+/* Should be part of hwmod data base ? */
+#define OMAP_DMA4_LOGICAL_DMA_CH_COUNT	32	/* 

REVISIT: Is this 32 + 2? */

Quoted 6 lines
Quoted 3 lines

arch/arm/mach-omap2/include/mach/dma.h line 34

Quoted 1 line
arch/arm/mach-omap2/include/mach/dma.h +1 -0
+

There are no users of this in this patch, all users are removed.

This macro is used in hwmod data base. Do you mean hard coding this value in hwmod db and remove header?

What I mean is that from a reviewers perspective of this patch, you added a new #define that is not used anywhere in this patch. a #define should be added along with any users of it.

That being said, the number of channels of the DMA IP sounds like something that should indeed be associated with the hwmod as it is a fixed HW feature.

Kevin

#37 lore raw

PATCH v3 10/13 OMAP: DMA: Convert DMA library into DMA platform Driver

Quoted 12 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 9:54 PM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 10/13] OMAP: DMA: Convert DMA library into DMA platform Driver

[...]

Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Quoted 1 line

+#define OMAP_DMA4_LOGICAL_DMA_CH_COUNT 32 /*

REVISIT: Is this 32 + 2? */

Quoted 1 line

+

There are no users of this in this patch, all users are removed.

Quoted 7 lines
Quoted 2 lines

This macro is used in hwmod data base. Do you mean hard coding this value in hwmod db and remove header?

What I mean is that from a reviewers perspective of this patch, you added a new #define that is not used anywhere in this patch. a #define
should be added along with any users of it.

No. This macro is used in omap2+ hwmod db through plat/dma.h which includes mach/dma.h

Quoted 4 lines

That being said, the number of channels of the DMA IP sounds like something that should indeed be associated with the hwmod as it is a fixed HW feature.

If so, I can remove this macro and hard code in hwmod db.

-Manjunath

#38 lore raw

PATCH v3 10/13 OMAP: DMA: Convert DMA library into DMA platform Driver

Quoted 9 lines · Manjunath Kondaiah
Quoted 9 lines
Quoted 8 lines
Quoted 5 lines
Quoted 1 line

+#define OMAP_DMA4_LOGICAL_DMA_CH_COUNT 32 /*

REVISIT: Is this 32 + 2? */

Quoted 1 line

+

There are no users of this in this patch, all users are removed.

This macro is used in hwmod data base. Do you mean hard coding this value in hwmod db and remove header?

Quoted 7 lines
Quoted 4 lines

What I mean is that from a reviewers perspective of this patch, you added a new #define that is not used anywhere in this patch. a #define
should be added along with any users of it.

No. This macro is used in omap2+ hwmod db through plat/dma.h which includes mach/dma.h

Which is not part of this patch.

Again, as a general rule, don't introduce code (#defines, headers, functions, anything) in a patch if it is not used.

Quoted 6 lines
Quoted 5 lines

That being said, the number of channels of the DMA IP sounds like something that should indeed be associated with the hwmod as it is a fixed HW feature.

If so, I can remove this macro and hard code in hwmod db.

OK, good.

Kevin

#39 lore raw

PATCH v3 10/13 OMAP: DMA: Convert DMA library into DMA platform Driver

In reply to G, Manjunath Kondaiah
Quoted 56 lines · Manjunath Kondaiah

Convert DMA library into DMA platform driver and make use of platform data provided by HWMOD data base for OMAP2PLUS onwards. For OMAP1 processors, the DMA driver in mach-omap uses resource structures for getting platform data.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap1/Makefile           |    2 +-
 arch/arm/mach-omap1/dma.c              |  180 ++++++++++++++++++-
 arch/arm/mach-omap2/Makefile           |    2 +-
 arch/arm/mach-omap2/dma.c              |  209 +++++++++++++++++++++-
 arch/arm/mach-omap2/include/mach/dma.h |    3 +
 arch/arm/plat-omap/dma.c               |  321 +++++++++++---------------------
 arch/arm/plat-omap/include/plat/dma.h  |   50 ++++--
 7 files changed, 538 insertions(+), 229 deletions(-)
arch/arm/mach-omap1/Makefile +1 -1
index 9a304d8..b7dfc54 100644
@@ -3,7 +3,7 @@
 #
 
 # Common support
-obj-y := io.o id.o sram.o irq.o mux.o flash.o serial.o devices.o
+obj-y := io.o id.o sram.o irq.o mux.o flash.o serial.o devices.o dma.o
 obj-y += clock.o clock_data.o opp_data.o
 
 obj-$(CONFIG_OMAP_MCBSP) += mcbsp.o
arch/arm/mach-omap1/dma.c +16 -0
index e756069..38a7294 100644
@@ -35,6 +35,81 @@
 #include <plat/tc.h>
 
 #define OMAP1_DMA_BASE			(0xfffed800)
+#define OMAP1_LOGICAL_DMA_CH_COUNT	17
+
+static u32 errata;
+static u32 enable_1510_mode;
+
+enum {
+	GCR1 = 0,	GSCR,		GRST,		HW_ID,
+	PCH2_ID,	PCH0_ID,	PCH1_ID,	PCHG_ID,
+	PCHD_ID,	CAPS_0_U,	CAPS_0_L,	CAPS_1_U,
+	CAPS_1_L,	CAPS_2,		CAPS_3,		CAPS_4,
+	PCH2_SR,	PCH0_SR,	PCH1_SR,	PCHD_SR,
+
+	CH_COMMON_START,
+
+	/* Common Registers */
+	CSDP1,		CCR1,		CICR1,		CSR1,

arch/arm/mach-omap1/dma.c line 63

Quoted 10 lines
arch/arm/mach-omap1/dma.c +10 -0
+	CEN1,		CFN1,		CSFI1,		CSEI1,
+	CPC,		CSAC1,		CDAC1,		CDEI1,
+	CDFI1,		CLNK_CTRL1,
+
+	/* Channel specific register offsets */
+	CSSA_L,		CSSA_U,		CDSA_L,		CDSA_U,
+	COLOR_L,	COLOR_U,	CCR1_2,		LCH_CTRL,
+
+	CH_COMMON_END,
+};

These reg_map enums are added here, then promptly moved in the following patch. Please just put them in the right place earlier in the series. Also, at this point, similar versions also still exist in plat-omap/dma.c

This may be a good technique for inflating the number of lines changed in the git history, but it is not review friendly.

Kevin

#40 lore raw

PATCH v3 10/13 OMAP: DMA: Convert DMA library into DMA platform Driver

Quoted 13 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 4:59 AM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 10/13] OMAP: DMA: Convert DMA library into DMA platform Driver

"G, Manjunath Kondaiah" <[email protected]> writes:

[...]

Quoted 43 lines
Quoted 26 lines

obj-$(CONFIG_OMAP_MCBSP) += mcbsp.o

arch/arm/mach-omap1/dma.c +16 -0
index e756069..38a7294 100644
@@ -35,6 +35,81 @@
 #include <plat/tc.h>
 
 #define OMAP1_DMA_BASE			(0xfffed800)
+#define OMAP1_LOGICAL_DMA_CH_COUNT	17
+
+static u32 errata;
+static u32 enable_1510_mode;
+
+enum {
+	GCR1 = 0,	GSCR,		GRST,		HW_ID,
+	PCH2_ID,	PCH0_ID,	PCH1_ID,	PCHG_ID,
+	PCHD_ID,	CAPS_0_U,	CAPS_0_L,	CAPS_1_U,
+	CAPS_1_L,	CAPS_2,		CAPS_3,		CAPS_4,
+	PCH2_SR,	PCH0_SR,	PCH1_SR,	PCHD_SR,
+
+	CH_COMMON_START,
+
+	/* Common Registers */
+	CSDP1,		CCR1,		CICR1,		CSR1,

arch/arm/mach-omap1/dma.c line 63

Quoted 10 lines
arch/arm/mach-omap1/dma.c +10 -0
+	CEN1,		CFN1,		CSFI1,		CSEI1,
+	CPC,		CSAC1,		CDAC1,		CDEI1,
+	CDFI1,		CLNK_CTRL1,
+
+	/* Channel specific register offsets */
+	CSSA_L,		CSSA_U,		CDSA_L,		CDSA_U,
+	COLOR_L,	COLOR_U,	CCR1_2,		LCH_CTRL,
+
+	CH_COMMON_END,
+};

These reg_map enums are added here, then promptly moved in the following
patch. Please just put them in the right place earlier in the series. Also, at this point, similar versions also still exist in plat-omap/dma.c

Quoted 2 lines

This may be a good technique for inflating the number of lines changed in the git history, but it is not review friendly.

Thanks. With Tony's suggestion, these duplicate enums will not exist with next version since there will be common enum in plat-omap from patch-1 of the series. only reg_map tables will be moved to respective mach-omap dma files from plat-omap.

-Manjunath

#41 lore raw

PATCH v3 11/13 OMAP: DMA: Use DMA device attributes

In reply to G, Manjunath Kondaiah

Existing DMA API's are using cpu_*is_* checks for differentiating omap1 and omap2plus code.

Replace cpu_*is_* checks with DMA device attributes and also move API's which are OMAP1 and OMAP2plus specific into respective mach-omap dma driver files.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap1/dma.c              |  385 ++++++-
 arch/arm/mach-omap1/include/mach/dma.h |   22 +
 arch/arm/mach-omap2/dma.c              | 1099 ++++++++++++++++++-
 arch/arm/mach-omap2/include/mach/dma.h |   51 +
 arch/arm/plat-omap/dma.c               | 1872 +++-----------------------------
 arch/arm/plat-omap/include/plat/dma.h  |   77 +-
 6 files changed, 1670 insertions(+), 1836 deletions(-)
arch/arm/mach-omap1/dma.c +346 -39
index 38a7294..d373458 100644
@@ -36,34 +36,16 @@
 
 #define OMAP1_DMA_BASE			(0xfffed800)
 #define OMAP1_LOGICAL_DMA_CH_COUNT	17
+#define OMAP_FUNC_MUX_ARM_BASE		(0xfffe1000 + 0xec)
 
 static u32 errata;
 static u32 enable_1510_mode;
 
-enum {
-	GCR1 = 0,	GSCR,		GRST,		HW_ID,
-	PCH2_ID,	PCH0_ID,	PCH1_ID,	PCHG_ID,
-	PCHD_ID,	CAPS_0_U,	CAPS_0_L,	CAPS_1_U,
-	CAPS_1_L,	CAPS_2,		CAPS_3,		CAPS_4,
-	PCH2_SR,	PCH0_SR,	PCH1_SR,	PCHD_SR,
-
-	CH_COMMON_START,
-
-	/* Common Registers */
-	CSDP1,		CCR1,		CICR1,		CSR1,
-	CEN1,		CFN1,		CSFI1,		CSEI1,
-	CPC,		CSAC1,		CDAC1,		CDEI1,
-	CDFI1,		CLNK_CTRL1,
-
-	/* Channel specific register offsets */
-	CSSA_L,		CSSA_U,		CDSA_L,		CDSA_U,
-	COLOR_L,	COLOR_U,	CCR1_2,		LCH_CTRL,
-
-	CH_COMMON_END,
-};
+static struct omap_dma_lch *dma_chan;
+static struct omap_dma_dev_attr *d;
 
 static u16 reg_map[] = {
-	[GCR1]		= 0x400,
+	[GCR]		= 0x400,
 	[GSCR]		= 0x404,
 	[GRST]		= 0x408,
 	[HW_ID]		= 0x442,
@@ -85,20 +67,20 @@ static u16 reg_map[] = {
 	[PCHD_SR]	= 0x4c0,
 
 	/* Common Registers */
-	[CSDP1]		= 0x00,
-	[CCR1]		= 0x02,
-	[CICR1]		= 0x04,
-	[CSR1]		= 0x06,
-	[CEN1]		= 0x10,
-	[CFN1]		= 0x12,
-	[CSFI1]		= 0x14,
-	[CSEI1]		= 0x16,
+	[CSDP]		= 0x00,
+	[CCR]		= 0x02,
+	[CICR]		= 0x04,
+	[CSR]		= 0x06,
+	[CEN]		= 0x10,
+	[CFN]		= 0x12,
+	[CSFI]		= 0x14,
+	[CSEI]		= 0x16,
 	[CPC]		= 0x18,	/* 15xx only */
-	[CSAC1]		= 0x18,
-	[CDAC1]		= 0x1a,
-	[CDEI1]		= 0x1c,
-	[CDFI1]		= 0x1e,
-	[CLNK_CTRL1]	= 0x28,
+	[CSAC]		= 0x18,
+	[CDAC]		= 0x1a,
+	[CDEI]		= 0x1c,
+	[CDFI]		= 0x1e,
+	[CLNK_CTRL]	= 0x28,
 
 	/* Channel specific register offsets */
 	[CSSA_L]	= 0x08,
@@ -107,7 +89,7 @@ static u16 reg_map[] = {
 	[CDSA_U]	= 0x0e,
 	[COLOR_L]	= 0x20,
 	[COLOR_U]	= 0x22,
-	[CCR1_2]	= 0x24,
+	[CCR2]		= 0x24,
 	[LCH_CTRL]	= 0x2a,
 };
 
@@ -222,6 +204,195 @@ static inline u16 dma_read(int reg, int lch)
 		return __raw_readw(dma_base + reg_map[reg]);
 }
 
+static void
+set_src_params(int lch, unsigned long src_start, int src_ei, int src_fi)
+{
+	dma_write(src_start >> 16, CSSA_U, lch);
+	dma_write((u16)src_start, CSSA_L, lch);
+
+	dma_write(src_ei, CSEI, lch);
+	dma_write(src_fi, CSFI, lch);
+}
+
+static void set_dest_params(int lch, unsigned long dest_start, int dst_ei,
+								int dst_fi)
+{
+	dma_write(dest_start >> 16, CDSA_U, lch);
+	dma_write(dest_start, CDSA_L, lch);
+
+	dma_write(dst_ei, CDEI, lch);
+	dma_write(dst_fi, CDFI, lch);
+}
+
+static inline void omap1_enable_channel_irq(int lch)
+{
+	u32 status;
+
+	/* Clear CSR */
+	status = dma_read(CSR, lch);
+	dma_write(dma_chan[lch].enabled_irqs, CICR, lch);
+}
+
+static void omap1_disable_channel_irq(int lch)
+{
+	dma_write(0, CICR, lch);
+}
+
+static inline void omap1_enable_lnk(int lch)
+{
+	u32 l;
+
+	l = dma_read(CLNK_CTRL, lch);
+	l &= ~(1 << 14);
+
+	/* Set the ENABLE_LNK bits */
+	if (dma_chan[lch].next_lch != -1)
+		l = dma_chan[lch].next_lch | (1 << 15);
+
+	dma_write(l, CLNK_CTRL, lch);
+}
+
+static inline void omap1_disable_lnk(int lch)
+{
+	u32 l;
+
+	/* Disable interrupts */
+	l = dma_read(CLNK_CTRL, lch);
+	dma_write(0, CICR, lch);
+	/* Set the STOP_LNK bit */
+	l |= 1 << 14;
+	dma_write(l, CLNK_CTRL, lch);
+
+	dma_chan[lch].flags &= ~OMAP_DMA_ACTIVE;
+}
+
+static dma_addr_t get_src_pos(int lch)
+{
+	dma_addr_t offset = 0;
+
+	if (enable_1510_mode)
+		offset = dma_read(CPC, lch);
+	else
+		offset = dma_read(CSAC, lch);
+
+	if (IS_DMA_ERRATA(DMA_ERRATA_3_3) && offset == 0)
+		offset = dma_read(CSAC, lch);
+
+	offset |= (dma_read(CSSA_U, lch) << 16);
+
+	return offset;
+}
+
+static dma_addr_t get_dst_pos(int lch)
+{
+	dma_addr_t offset = 0;
+
+	if (enable_1510_mode)
+		offset = dma_read(CPC, lch);
+	else
+		offset = dma_read(CDAC, lch);
+
+	if (IS_DMA_ERRATA(DMA_ERRATA_3_3) && offset == 0)
+		offset = dma_read(CDAC, lch);
+
+	offset |= (dma_read(CDSA_U, lch) << 16);
+
+	return offset;
+}
+
+static void
+set_transfer_params(int lch, int data_type, int sync_mode, int dma_trigger,
+							int src_or_dst_synch)
+{
+	u16 ccr;
+	u32 l;
+
+	l = dma_read(CSDP, lch);
+	l &= ~0x03;
+	l |= data_type;
+	dma_write(l, CSDP, lch);
+
+	ccr = dma_read(CCR, lch);
+	ccr &= ~(1 << 5);
+	if (sync_mode == OMAP_DMA_SYNC_FRAME)
+		ccr |= 1 << 5;
+	dma_write(ccr, CCR, lch);
+
+	ccr = dma_read(CCR2, lch);
+	ccr &= ~(1 << 2);
+	if (sync_mode == OMAP_DMA_SYNC_BLOCK)
+		ccr |= 1 << 2;
+	dma_write(ccr, CCR2, lch);
+}
+
+static void cpc_cdac_init(int lch)
+{
+	if (enable_1510_mode)
+		dma_write(0, CPC, lch);
+	else
+		dma_write(0, CDAC, lch);
+}
+
+static void omap1_clear_lch_regs(int lch)
+{
+	int i = CH_COMMON_START + 1;
+
+	for (; i <= CH_COMMON_END; i += 1)
+		dma_write(0, i, lch);
+}
+
+static inline int omap1_get_gdma_dev(int req)
+{
+	u32 reg = OMAP_FUNC_MUX_ARM_BASE + ((req - 1) / 5) * 4;
+	int shift = ((req - 1) % 5) * 6;
+
+	return ((omap_readl(reg) >> shift) & 0x3f) + 1;
+}
+
+static inline void omap1_set_gdma_dev(int req, int dev)
+{
+	u32 reg = OMAP_FUNC_MUX_ARM_BASE + ((req - 1) / 5) * 4;
+	int shift = ((req - 1) % 5) * 6;
+	u32 l;
+
+	l = omap_readl(reg);
+	l &= ~(0x3f << shift);
+	l |= (dev - 1) << shift;
+	omap_writel(l, reg);
+}
+
+static void set_color_mode(int lch, enum omap_dma_color_mode mode, u32 color)
+{
+	u16 w;
+
+	w = dma_read(CCR2, lch);
+	w &= ~0x03;
+
+	switch (mode) {
+	case OMAP_DMA_CONSTANT_FILL:
+		w |= 0x01;
+		break;
+	case OMAP_DMA_TRANSPARENT_COPY:
+		w |= 0x02;
+		break;
+	case OMAP_DMA_COLOR_DIS:
+		break;
+	default:
+		BUG();
+	}
+	dma_write(w, CCR2, lch);
+
+	w = dma_read(LCH_CTRL, lch);
+	w &= ~0x0f;
+	/* Default is channel type 2D */
+	if (mode) {
+		dma_write((u16)color, COLOR_L, lch);
+		dma_write((u16)(color >> 16), COLOR_U, lch);
+		w |= 1;		/* Channel type G */
+	}
+	dma_write(w, LCH_CTRL, lch);
+}
+
 static void omap1_show_dma_caps(void)
 {
 	if (enable_1510_mode) {
@@ -259,12 +430,93 @@ static u32 configure_dma_errata(void)
 	return errata;
 }
 
+static int omap1_dma_handle_ch(int ch)
+{
+	u32 csr;
+
+	if (enable_1510_mode && ch >= 6) {
+		csr = dma_chan[ch].saved_csr;
+		dma_chan[ch].saved_csr = 0;
+	} else
+		csr = dma_read(CSR, ch);
+	if (enable_1510_mode && ch <= 2 && (csr >> 7) != 0) {
+		dma_chan[ch + 6].saved_csr = csr >> 7;
+		csr &= 0x7f;
+	}
+	if ((csr & 0x3f) == 0)
+		return 0;
+	if (unlikely(dma_chan[ch].dev_id == -1)) {
+		printk(KERN_WARNING "Spurious interrupt from DMA channel "
+		       "%d (CSR %04x)\n", ch, csr);
+		return 0;
+	}
+	if (unlikely(csr & OMAP1_DMA_TOUT_IRQ))
+		printk(KERN_WARNING "DMA timeout with device %d\n",
+		       dma_chan[ch].dev_id);
+	if (unlikely(csr & OMAP_DMA_DROP_IRQ))
+		printk(KERN_WARNING "DMA synchronization event drop occurred "
+		       "with device %d\n", dma_chan[ch].dev_id);
+	if (likely(csr & OMAP_DMA_BLOCK_IRQ))
+		dma_chan[ch].flags &= ~OMAP_DMA_ACTIVE;
+	if (likely(dma_chan[ch].callback != NULL))
+		dma_chan[ch].callback(ch, csr, dma_chan[ch].data);
+
+	return 1;
+}
+
+static irqreturn_t omap1_dma_irq_handler(int irq, void *dev_id)
+{
+	int ch = ((int) dev_id) - 1;
+	int handled = 0;
+
+	for (;;) {
+		int handled_now = 0;
+
+		handled_now += omap1_dma_handle_ch(ch);
+		if (enable_1510_mode && dma_chan[ch + 6].saved_csr)
+			handled_now += omap1_dma_handle_ch(ch + 6);
+		if (!handled_now)
+			break;
+		handled += handled_now;
+	}
+
+	return handled ? IRQ_HANDLED : IRQ_NONE;
+}
+
+void omap_set_dma_channel_mode(int lch, enum omap_dma_channel_mode mode)
+{
+	if (!(d->dev_caps & ENABLE_1510_MODE)) {
+		u32 l;
+
+		l = dma_read(LCH_CTRL, lch);
+		l &= ~0x7;
+		l |= mode;
+		dma_write(l, LCH_CTRL, lch);
+	}
+}
+EXPORT_SYMBOL(omap_set_dma_channel_mode);
+
+void omap_set_dma_src_index(int lch, int eidx, int fidx)
+{
+	dma_write(eidx, CSEI, lch);
+	dma_write(fidx, CSFI, lch);
+}
+EXPORT_SYMBOL(omap_set_dma_src_index);
+
+void omap_set_dma_dest_index(int lch, int eidx, int fidx)
+{
+	dma_write(eidx, CDEI, lch);
+	dma_write(fidx, CDFI, lch);
+}
+EXPORT_SYMBOL(omap_set_dma_dest_index);
+
 static int __init omap1_system_dma_init(void)
 {
 	struct omap_system_dma_plat_info	*p;
-	struct omap_dma_dev_attr		*d;
 	struct platform_device			*pdev;
-	int ret;
+	int ret, ch, irq_rel;
+	char irq_name[4];
+	int dma_irq = 0;
 
 	pdev = platform_device_alloc("omap_dma_system", 0);
 	if (!pdev) {
@@ -287,8 +539,8 @@ static int __init omap1_system_dma_init(void)
 		ret = -ENOMEM;
 		goto exit_device_put;
 	}
-
 	d = p->dma_attr;
+
 	d->chan = kzalloc(sizeof(struct omap_dma_lch) *
 					(d->lch_count), GFP_KERNEL);
 	if (!d->chan) {
@@ -321,10 +573,51 @@ static int __init omap1_system_dma_init(void)
 			d->chan_count = 9;
 	}
 
+	for (ch = 0; ch < d->chan_count; ch++) {
+		if (ch >= 6 && enable_1510_mode)
+			continue;
+
+		sprintf(&irq_name[0], "%d", ch);
+		dma_irq = platform_get_irq_byname(pdev, irq_name);
+		if (dma_irq < 0) {
+			dev_err(&pdev->dev, "%s:unable to get irq\n",
+								__func__);
+			ret = dma_irq;
+			goto exit_release_irq;
+		}
+
+		ret = request_irq(dma_irq, omap1_dma_irq_handler, 0, "DMA",
+							(void *) (ch + 1));
+		if (ret != 0)
+			goto exit_release_irq;
+	}
+
 	p->omap_dma_base	= (void __iomem *)res[0].start;
 	dma_base		= p->omap_dma_base;
 
 	p->show_dma_caps	= omap1_show_dma_caps;
+	p->enable_channel_irq	= omap1_enable_channel_irq;
+	p->disable_channel_irq	= omap1_disable_channel_irq;
+	p->enable_lnk		= omap1_enable_lnk;
+	p->disable_lnk		= omap1_disable_lnk;
+	p->clear_lch_regs	= omap1_clear_lch_regs;
+	p->set_transfer_params	= set_transfer_params;
+	p->set_color_mode	= set_color_mode;
+	p->set_src_params	= set_src_params;
+	p->set_dest_params	= set_dest_params;
+	p->get_src_pos		= get_src_pos;
+	p->get_dst_pos		= get_dst_pos;
+	p->cpc_cdac_init	= cpc_cdac_init;
+
+	if (cpu_is_omap16xx()) {
+		p->get_gdma_dev = omap1_get_gdma_dev;
+		p->set_gdma_dev = omap1_set_gdma_dev;
+	}
+
+	p->enable_irq_lch	= NULL;
+	p->disable_irq_lch	= NULL;
+	p->clear_channel_int	= NULL;
+	p->set_global_params	= NULL;
 	p->disable_irq_lch	= NULL;
 
 	/* Configure errata handling for all omap1+ */
@@ -342,14 +635,28 @@ static int __init omap1_system_dma_init(void)
 			__func__, pdev->name, pdev->id);
 		goto exit_release_chan;
 	}
+	dma_chan = d->chan;
 	return ret;
 
 exit_release_chan:
 	kfree(d->chan);
+
+exit_release_irq:
+	printk(KERN_ERR "unable to request IRQ %d"
+				"for DMA (error %d)\n", dma_irq, ret);
+	if (enable_1510_mode)
+		ch = 6;
+	for (irq_rel = 0; irq_rel < ch; irq_rel++) {
+		dma_irq = platform_get_irq(pdev, irq_rel);
+		free_irq(dma_irq, (void *)(irq_rel + 1));
+	}
+
 exit_release_p:
 	kfree(p);
+
 exit_device_put:
 	platform_device_put(pdev);
+
 exit_device_del:
 	platform_device_del(pdev);
 
arch/arm/mach-omap1/include/mach/dma.h +22 -0
index 7949e3f..2ad11ff 100644
@@ -26,4 +26,26 @@
 #ifndef __ASM_ARCH_OMAP1_DMA_H
 #define __ASM_ARCH_OMAP1_DMA_H
 
+enum {
+	GCR = 0,	GSCR,		GRST,		HW_ID,
+	PCH2_ID,	PCH0_ID,	PCH1_ID,	PCHG_ID,
+	PCHD_ID,	CAPS_0_U,	CAPS_0_L,	CAPS_1_U,
+	CAPS_1_L,	CAPS_2,		CAPS_3,		CAPS_4,
+	PCH2_SR,	PCH0_SR,	PCH1_SR,	PCHD_SR,
+
+	CH_COMMON_START,
+
+	/* Common Registers */
+	CSDP,		CCR,		CICR,		CSR,
+	CEN,		CFN,		CSFI,		CSEI,
+	CPC,		CSAC,		CDAC,		CDEI,
+	CDFI,		CLNK_CTRL,
+
+	/* Channel specific register offsets */
+	CSSA_L,		CSSA_U,		CDSA_L,		CDSA_U,
+	COLOR_L,	COLOR_U,	CCR2,		LCH_CTRL,
+
+	CH_COMMON_END,
+};
+
 #endif /* __ASM_ARCH_OMAP1_DMA_H */
arch/arm/mach-omap2/dma.c +1064 -35

1,171 lines not on this page · raw message

arch/arm/mach-omap2/include/mach/dma.h +51 -0

55 lines not on this page · raw message

arch/arm/plat-omap/dma.c +157 -1715

1,291 lines not on this page · raw message

arch/arm/plat-omap/include/plat/dma.h +30 -47

77 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 1 line
1.7.1

Part of the diff is not shown here. View the raw message.

#42 lore raw

PATCH v3 12/13 OMAP2+: DMA: descriptor autoloading feature

In reply to G, Manjunath Kondaiah

Add sDMA driver support for descriptor autoloading feature. Descriptor autoloading is OMAP sDMA v5 hardware capability that can be exploited for scatter gather scenarios, currently available in OMAP3630 and OMAP4430.

The feature works as described below.
1. A sDMA channel is programmed to be in 'linked list' mode.
2) The client (sDMA user) provides a list of descriptors in a linked list format
3) Each of the 'descriptor' (element in the linked list) contains an updated set of DMA configuration register values.
4) Client starts DMA transfer.
5) sDMA controller loads the first element to its register configuration memory and executes the transfer.
6) After completion, loads the next element (in linked list) to configuration memory and executes the transfer, without MCU intervention.
7) Interrupt is generated after all transfers are completed; this can be configured to be done differently.

Configurations and additional features
1) Fast mode & non-fast mode
Fast mode/non-fast decides on how the first transfer begins. In non-fast mode, the first element in the linked list is loaded only after completing the
transfer according to the configurations already in the sDMA channel registers.
In fast mode, the loading of the first element precedes the transfer.

2) Pause / resume of transfers
A transfer can be paused after a descriptor set has been loaded, provided the pause bit' is set in the linked list element. An ongoing transfer cannot be paused. If the 'pause bit' is set, transfer is not started after loading the register set from memory. Such a transfer can be resumed later.

3) Descriptor types
Three possible configurations of descriptors (initialized as linked list elements) are possible.
Type 1 provides the maximum flexibility, which contains most register definitions of a DMA logical channel.
Fewer options are present in type 2.
Type 3 can just modify source/destinations address of transfers. In all transfers, unmodified registers settings are maintained for the next transfer.

Patch provides options / API for
1) Setting up a descriptor loading for DMA channel for sg type transfers
2) configuration with linked list elements
3) Starting / pause and resume of the said transfers, query state
4) Clearing the sglist mode

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/mach-omap2/dma.c             |  257 +++++++++++++++++++++++++++++++++
 arch/arm/plat-omap/dma.c              |    1 +
 arch/arm/plat-omap/include/plat/dma.h |  111 ++++++++++++++-
 3 files changed, 368 insertions(+), 1 deletions(-)
arch/arm/mach-omap2/dma.c +257 -0
index 6df4835..397e90e 100644
@@ -41,6 +41,7 @@
 
 static u32 errata;
 static int dma_chan_count;
+static int dma_caps0_status;
 
 static struct omap_dma_dev_attr *d;
 static struct dma_link_info *dma_linked_lch;
@@ -240,6 +241,76 @@ static inline void omap2_disable_lnk(int lch)
 	dma_chan[lch].flags &= ~OMAP_DMA_ACTIVE;
 }
 
+static inline void omap_dma_list_set_ntype(struct omap_dma_sglist_node *node,
+					   int value)
+{
+	node->num_of_elem |= ((value) << 29);
+}
+
+static void omap_set_dma_sglist_pausebit(
+		struct omap_dma_list_config_params *lcfg, int nelem, int set)
+{
+	struct omap_dma_sglist_node *sgn = lcfg->sghead;
+
+	if (nelem > 0 && nelem < lcfg->num_elem) {
+		lcfg->pausenode = nelem;
+		sgn += nelem;
+
+		if (set)
+			sgn->next_desc_add_ptr |= DMA_LIST_DESC_PAUSE;
+		else
+			sgn->next_desc_add_ptr &= ~(DMA_LIST_DESC_PAUSE);
+	}
+}
+
+static int dma_sglist_set_phy_params(struct omap_dma_sglist_node *sghead,
+		dma_addr_t phyaddr, int nelem)
+{
+	struct omap_dma_sglist_node *sgcurr, *sgprev;
+	dma_addr_t elem_paddr = phyaddr;
+
+	for (sgprev = sghead;
+		sgprev < sghead + nelem;
+		sgprev++) {
+
+		sgcurr = sgprev + 1;
+		sgprev->next = sgcurr;
+		elem_paddr += (int)sizeof(*sgcurr);
+		sgprev->next_desc_add_ptr = elem_paddr;
+
+		switch (sgcurr->desc_type) {
+		case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE1:
+			omap_dma_list_set_ntype(sgprev, 1);
+			break;
+
+		case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE2a:
+		/* intentional no break */
+		case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE2b:
+			omap_dma_list_set_ntype(sgprev, 2);
+			break;
+
+		case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE3a:
+			/* intentional no break */
+		case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE3b:
+			omap_dma_list_set_ntype(sgprev, 3);
+			break;
+
+		default:
+			return -EINVAL;
+
+		}
+		if (sgcurr->flags & OMAP_DMA_LIST_SRC_VALID)
+			sgprev->num_of_elem |= DMA_LIST_DESC_SRC_VALID;
+		if (sgcurr->flags & OMAP_DMA_LIST_DST_VALID)
+			sgprev->num_of_elem |= DMA_LIST_DESC_DST_VALID;
+		if (sgcurr->flags & OMAP_DMA_LIST_NOTIFY_BLOCK_END)
+			sgprev->num_of_elem |= DMA_LIST_DESC_BLK_END;
+	}
+	sgprev--;
+	sgprev->next_desc_add_ptr = OMAP_DMA_INVALID_DESCRIPTOR_POINTER;
+	return 0;
+}
+
 static void
 omap2_set_color_mode(int lch, enum omap_dma_color_mode mode, u32 color)
 {
@@ -1207,6 +1278,190 @@ int omap_get_dma_chain_src_pos(int chain_id)
 }
 EXPORT_SYMBOL(omap_get_dma_chain_src_pos);
 
+int omap_set_dma_sglist_mode(int lch, struct omap_dma_sglist_node *sgparams,
+	dma_addr_t padd, int nelem, struct omap_dma_channel_params *chparams)
+{
+	struct omap_dma_list_config_params *lcfg;
+	int l = DMA_LIST_CDP_LISTMODE; /* Enable Linked list mode in CDP */
+
+	if ((dma_caps0_status & DMA_CAPS_SGLIST_SUPPORT) == 0) {
+		printk(KERN_ERR "omap DMA: sglist feature not supported\n");
+		return -EPERM;
+	}
+	if (dma_chan[lch].flags & OMAP_DMA_ACTIVE) {
+		printk(KERN_ERR "omap DMA: configuring active DMA channel\n");
+		return -EPERM;
+	}
+
+	if (padd == 0) {
+		printk(KERN_ERR "omap DMA: sglist invalid dma_addr\n");
+		return -EINVAL;
+	}
+	lcfg = &dma_chan[lch].list_config;
+
+	lcfg->sghead = sgparams;
+	lcfg->num_elem = nelem;
+	lcfg->sgheadphy = padd;
+	lcfg->pausenode = -1;
+
+	if (NULL == chparams)
+		l |= DMA_LIST_CDP_FASTMODE;
+	else
+		omap_set_dma_params(lch, chparams);
+
+	dma_write(l, CDP, lch);
+	dma_write(0, CCDN, lch); /* Reset List index numbering */
+	/* Initialize frame and element counters to invalid values */
+	dma_write(OMAP_DMA_INVALID_FRAME_COUNT, CCFN, lch);
+	dma_write(OMAP_DMA_INVALID_ELEM_COUNT, CCEN, lch);
+
+	return dma_sglist_set_phy_params(sgparams, lcfg->sgheadphy, nelem);
+
+}
+EXPORT_SYMBOL(omap_set_dma_sglist_mode);
+
+void omap_clear_dma_sglist_mode(int lch)
+{
+	/* Clear entire CDP which is related to sglist handling */
+	dma_write(0, CDP, lch);
+	dma_write(0, CCDN, lch);
+	/**
+	 * Put back the original enabled irqs, which
+	 * could have been overwritten by type 1 or type 2
+	 * descriptors
+	 */
+	dma_write(dma_chan[lch].enabled_irqs, CICR, lch);
+	return;
+}
+EXPORT_SYMBOL(omap_clear_dma_sglist_mode);
+
+int omap_start_dma_sglist_transfers(int lch, int pauseafter)
+{
+	struct omap_dma_list_config_params *lcfg;
+	struct omap_dma_sglist_node *sgn;
+	unsigned int l, type_id;
+
+	lcfg = &dma_chan[lch].list_config;
+	sgn = lcfg->sghead;
+
+	lcfg->pausenode = 0;
+	omap_set_dma_sglist_pausebit(lcfg, pauseafter, 1);
+
+	/* Program the head descriptor's properties into CDP */
+	switch (lcfg->sghead->desc_type) {
+	case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE1:
+		type_id = DMA_LIST_CDP_TYPE1;
+		break;
+	case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE2a:
+	case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE2b:
+		type_id = DMA_LIST_CDP_TYPE2;
+		break;
+	case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE3a:
+	case OMAP_DMA_SGLIST_DESCRIPTOR_TYPE3b:
+		type_id = DMA_LIST_CDP_TYPE3;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	l = dma_read(CDP, lch);
+	l |= type_id;
+	if (lcfg->sghead->flags & OMAP_DMA_LIST_SRC_VALID)
+		l |= DMA_LIST_CDP_SRC_VALID;
+	if (lcfg->sghead->flags & OMAP_DMA_LIST_DST_VALID)
+		l |= DMA_LIST_CDP_DST_VALID;
+
+	dma_write(l, CDP, lch);
+	dma_write((lcfg->sgheadphy), CNDP, lch);
+	/**
+	 * Barrier needed as writes to the
+	 * descriptor memory needs to be flushed
+	 * before it's used by DMA controller
+	 */
+	wmb();
+	omap_start_dma(lch);
+
+	return 0;
+}
+EXPORT_SYMBOL(omap_start_dma_sglist_transfers);
+
+int omap_resume_dma_sglist_transfers(int lch, int pauseafter)
+{
+	struct omap_dma_list_config_params *lcfg;
+	struct omap_dma_sglist_node *sgn;
+	int l;
+	u32 sys_cf = 0;
+
+	lcfg = &dma_chan[lch].list_config;
+	sgn = lcfg->sghead;
+
+	/* Maintain the pause state in descriptor */
+	omap_set_dma_sglist_pausebit(lcfg, lcfg->pausenode, 0);
+	omap_set_dma_sglist_pausebit(lcfg, pauseafter, 1);
+
+	/**
+	 * Barrier needed as writes to the
+	 * descriptor memory needs to be flushed
+	 * before it's used by DMA controller
+	 */
+	wmb();
+
+	/* Configure No-Standby */
+	if (IS_DMA_ERRATA(DMA_ERRATA_i541))
+		omap_device_mstandby(pd, &sys_cf, true);
+
+	/* Clear pause bit in CDP */
+	l = dma_read(CDP, lch);
+	l &= ~(DMA_LIST_CDP_PAUSEMODE);
+	dma_write(l, CDP, lch);
+
+	omap_start_dma(lch);
+
+	if (IS_DMA_ERRATA(DMA_ERRATA_i541))
+		omap_device_mstandby(pd, &sys_cf, false);
+
+	return 0;
+}
+EXPORT_SYMBOL(omap_resume_dma_sglist_transfers);
+
+void omap_release_dma_sglist(int lch)
+{
+	omap_clear_dma_sglist_mode(lch);
+	omap_free_dma(lch);
+
+	return;
+}
+EXPORT_SYMBOL(omap_release_dma_sglist);
+
+int omap_get_completed_sglist_nodes(int lch)
+{
+	int list_count;
+
+	list_count = dma_read(CCDN, lch);
+	return list_count & 0xffff; /* only 16 LSB bits are valid */
+}
+EXPORT_SYMBOL(omap_get_completed_sglist_nodes);
+
+int omap_dma_sglist_is_paused(int lch)
+{
+	int list_state;
+	list_state = dma_read(CDP, lch);
+	return (list_state & DMA_LIST_CDP_PAUSEMODE) ? 1 : 0;
+}
+EXPORT_SYMBOL(omap_dma_sglist_is_paused);
+
+void omap_dma_set_sglist_fastmode(int lch, int fastmode)
+{
+	int l = dma_read(CDP, lch);
+
+	if (fastmode)
+		l |= DMA_LIST_CDP_FASTMODE;
+	else
+		l &= ~(DMA_LIST_CDP_FASTMODE);
+	dma_write(l, CDP, lch);
+}
+EXPORT_SYMBOL(omap_dma_set_sglist_fastmode);
+
 /* One time initializations */
 static int __init omap2_system_dma_init_dev(struct omap_hwmod *oh, void *unused)
 {
@@ -1245,6 +1500,7 @@ static int __init omap2_system_dma_init_dev(struct omap_hwmod *oh, void *unused)
 	p->disable_lnk			= omap2_disable_lnk;
 
 	p->set_global_params		= omap_dma_set_global_params;
+	p->clear_dma_sglist_mode	= omap_clear_dma_sglist_mode;
 
 	p->clear_lch_regs	= NULL;
 	p->get_gdma_dev		= NULL;
@@ -1306,6 +1562,7 @@ static int __init omap2_system_dma_init_dev(struct omap_hwmod *oh, void *unused)
 		return -ENOMEM;
 	}
 	dma_chan = d->chan;
+	dma_caps0_status = dma_read(CAPS_0, 0);
 
 	return 0;
 }
arch/arm/plat-omap/dma.c +1 -0
index eb00b22..41b14b0 100644
@@ -450,6 +450,7 @@ void omap_free_dma(int lch)
 		/* Make sure the DMA transfer is stopped. */
 		dma_write(0, CCR, lch);
 		omap_clear_dma(lch);
+		p->clear_dma_sglist_mode(lch);
 	}
 	spin_lock_irqsave(&dma_chan_lock, flags);
 	dma_chan[lch].dev_id = -1;
arch/arm/plat-omap/include/plat/dma.h +110 -1

135 lines not on this page · raw message

Signature · 1 line
1.7.1
#43 lore raw

PATCH v3 13/13 OMAP: PM: DMA: Enable runtime pm

In reply to G, Manjunath Kondaiah

Enable runtime pm and use pm_runtime_get and pm_runtime_put for OMAP DMA driver.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/dma.c |   13 +++++++++++++
 1 files changed, 13 insertions(+), 0 deletions(-)
arch/arm/plat-omap/dma.c +13 -0
index 41b14b0..feac7ee 100644
@@ -35,6 +35,7 @@
 #include <linux/io.h>
 #include <linux/slab.h>
 #include <linux/delay.h>
+#include <linux/pm_runtime.h>
 
 #include <asm/system.h>
 #include <mach/hardware.h>
@@ -367,6 +368,8 @@ int omap_request_dma(int dev_id, const char *dev_name,
 	chan = dma_chan + free_ch;
 	chan->dev_id = dev_id;
 
+	pm_runtime_get(&pd->dev);
+
 	if (d->dev_caps & IS_WORD_16)
 		p->clear_lch_regs(free_ch);
 	else
@@ -452,6 +455,7 @@ void omap_free_dma(int lch)
 		omap_clear_dma(lch);
 		p->clear_dma_sglist_mode(lch);
 	}
+	pm_runtime_put(&pd->dev);
 	spin_lock_irqsave(&dma_chan_lock, flags);
 	dma_chan[lch].dev_id = -1;
 	dma_chan[lch].next_lch = -1;
@@ -819,6 +823,11 @@ static int __devinit omap_system_dma_probe(struct platform_device *pdev)
 	dma_chan_count		= d->chan_count;
 	dma_chan		= d->chan;
 
+	/* Enable run time PM */
+	pm_runtime_enable(&pd->dev);
+
+	/* Accessing hw registers, get clock */
+	pm_runtime_get(&pd->dev);
 	for (ch = 0; ch < dma_chan_count; ch++) {
 		unsigned long flags;
 		omap_clear_dma(ch);
@@ -846,6 +855,10 @@ static int __devinit omap_system_dma_probe(struct platform_device *pdev)
 		dma_chan[0].dev_id = 0;
 		dma_chan[1].dev_id = 1;
 	}
+
+	if (!omap_dma_reserve_channels)
+		pm_runtime_put(&pd->dev);
+
 	return 0;
 
 exit_release_region:
Signature · 1 line
1.7.1
#44 lore raw

PATCH v3 13/13 OMAP: PM: DMA: Enable runtime pm

Quoted 24 lines · Manjunath Kondaiah

Enable runtime pm and use pm_runtime_get and pm_runtime_put for OMAP DMA driver.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/dma.c |   13 +++++++++++++
 1 files changed, 13 insertions(+), 0 deletions(-)
arch/arm/plat-omap/dma.c +1 -0
index 41b14b0..feac7ee 100644
@@ -35,6 +35,7 @@
 #include <linux/io.h>
 #include <linux/slab.h>
 #include <linux/delay.h>
+#include <linux/pm_runtime.h>
 
 #include <asm/system.h>
 #include <mach/hardware.h>

arch/arm/plat-omap/dma.c int omap_request_dma(int dev_id, const char *dev_name, · line 371

Quoted 4 lines
arch/arm/plat-omap/dma.c +1 -0
@@ -367,6 +368,8 @@ int omap_request_dma(int dev_id, const char *dev_name,
 	chan = dma_chan + free_ch;
 	chan->dev_id = dev_id;
 
+	pm_runtime_get(&pd->dev);

The _get() call is asynchronous. So if the device was actually idled/disabled, immediately after this call it may still be
idle/disabled. When using the asynchronous versions of the API, the
device should not be touched until the driver's ->runtime_resume() callback is called.

Quoted 12 lines
arch/arm/plat-omap/dma.c +1 -0
 	if (d->dev_caps & IS_WORD_16)
 		p->clear_lch_regs(free_ch);
 	else
@@ -452,6 +455,7 @@ void omap_free_dma(int lch)
 		omap_clear_dma(lch);
 		p->clear_dma_sglist_mode(lch);
 	}
+	pm_runtime_put(&pd->dev);
 	spin_lock_irqsave(&dma_chan_lock, flags);
 	dma_chan[lch].dev_id = -1;
 	dma_chan[lch].next_lch = -1;

arch/arm/plat-omap/dma.c static int __devinit omap_system_dma_probe(struct platform_device *pdev) · line 826

Quoted 4 lines
arch/arm/plat-omap/dma.c +1 -0
@@ -819,6 +823,11 @@ static int __devinit omap_system_dma_probe(struct platform_device *pdev)
 	dma_chan_count		= d->chan_count;
 	dma_chan		= d->chan;
 
+	/* Enable run time PM */

comment is redundant

arch/arm/plat-omap/dma.c static int __devinit omap_system_dma_probe(struct platform_device *pdev) · line 829

Quoted 3 lines
arch/arm/plat-omap/dma.c +3 -0
+	pm_runtime_enable(&pd->dev);
+
+	/* Accessing hw registers, get clock */

comment is wrong, driver doesn't know anything about clocks. Just drop comment as it doesn't tell you anything that the following 'get' doesn't.

arch/arm/plat-omap/dma.c static int __devinit omap_system_dma_probe(struct platform_device *pdev) · line 830

Quoted 1 line
arch/arm/plat-omap/dma.c +1 -0
+	pm_runtime_get(&pd->dev);

see above Re: async calls.

arch/arm/plat-omap/dma.c static int __devinit omap_system_dma_probe(struct platform_device *pdev) · line 861

Quoted 11 lines
arch/arm/plat-omap/dma.c +4 -0
 	for (ch = 0; ch < dma_chan_count; ch++) {
 		unsigned long flags;
 		omap_clear_dma(ch);
@@ -846,6 +855,10 @@ static int __devinit omap_system_dma_probe(struct platform_device *pdev)
 		dma_chan[0].dev_id = 0;
 		dma_chan[1].dev_id = 1;
 	}
+
+	if (!omap_dma_reserve_channels)
+		pm_runtime_put(&pd->dev);
+

So if the reserve channels feature is used, PM is essentially disabled for sDMA. You should add a warning here to that effect as well.

arch/arm/plat-omap/dma.c static int __devinit omap_system_dma_probe(struct platform_device *pdev) · line 864

Quoted 3 lines
arch/arm/plat-omap/dma.c +0 -0
 	return 0;
 
 exit_release_region:

Kevin

#45 lore raw

PATCH v3 13/13 OMAP: PM: DMA: Enable runtime pm

Quoted 42 lines

-----Original Message-----
From: linux-omap-owner at vger.kernel.org [mailto:linux-omap-owner at vger.kernel.org] On Behalf Of Kevin Hilman
Sent: Wednesday, November 10, 2010 4:18 AM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org;
linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 13/13] OMAP: PM: DMA: Enable runtime pm

arch/arm/plat-omap/dma.c int omap_request_dma(int dev_id, const

Quoted 24 lines · Manjunath Kondaiah

Enable runtime pm and use pm_runtime_get and pm_runtime_put for OMAP DMA driver.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/dma.c |   13 +++++++++++++
 1 files changed, 13 insertions(+), 0 deletions(-)
arch/arm/plat-omap/dma.c +1 -0
index 41b14b0..feac7ee 100644
@@ -35,6 +35,7 @@
 #include <linux/io.h>
 #include <linux/slab.h>
 #include <linux/delay.h>
+#include <linux/pm_runtime.h>
 
 #include <asm/system.h>
 #include <mach/hardware.h>
@@ -367,6 +368,8 @@ int omap_request_dma(int dev_id, const 

char *dev_name,

Quoted 3 lines
arch/arm/plat-omap/dma.c +0 -0
 	chan = dma_chan + free_ch;
 	chan->dev_id = dev_id;
 
Quoted 7 lines

arch/arm/plat-omap/dma.c int omap_request_dma(int dev_id, const · line 371

Quoted 1 line
arch/arm/plat-omap/dma.c +1 -0
+	pm_runtime_get(&pd->dev);

The _get() call is asynchronous. So if the device was actually idled/disabled, immediately after this call it may still be
idle/disabled. When using the asynchronous versions of the API, the
device should not be touched until the driver's ->runtime_resume() callback is called.

driver runtime_resume will call omap_device_enable. How do we make we check from driver side, whether omap_device_enable execution is complete or not.

Quoted 14 lines

arch/arm/plat-omap/dma.c static int __devinit

Quoted 13 lines
arch/arm/plat-omap/dma.c +1 -0
 	if (d->dev_caps & IS_WORD_16)
 		p->clear_lch_regs(free_ch);
 	else
@@ -452,6 +455,7 @@ void omap_free_dma(int lch)
 		omap_clear_dma(lch);
 		p->clear_dma_sglist_mode(lch);
 	}
+	pm_runtime_put(&pd->dev);
 	spin_lock_irqsave(&dma_chan_lock, flags);
 	dma_chan[lch].dev_id = -1;
 	dma_chan[lch].next_lch = -1;
@@ -819,6 +823,11 @@ static int __devinit 

omap_system_dma_probe(struct platform_device *pdev)

Quoted 6 lines

arch/arm/plat-omap/dma.c static int __devinit omap_system_dma_probe(struct platform_device *pdev) · line 826

Quoted 4 lines
arch/arm/plat-omap/dma.c +1 -0
 	dma_chan_count		= d->chan_count;
 	dma_chan		= d->chan;
 
+	/* Enable run time PM */

comment is redundant

ok.

Quoted 18 lines

arch/arm/plat-omap/dma.c static int __devinit omap_system_dma_probe(struct platform_device *pdev) · line 829

Quoted 4 lines
arch/arm/plat-omap/dma.c +3 -0
+	pm_runtime_enable(&pd->dev);
+
+	/* Accessing hw registers, get clock */

comment is wrong, driver doesn't know anything about clocks. Just drop comment as it doesn't tell you anything that the following 'get' doesn't.

arch/arm/plat-omap/dma.c static int __devinit · line 830

Quoted 1 line
arch/arm/plat-omap/dma.c +1 -0
+	pm_runtime_get(&pd->dev);

see above Re: async calls.

arch/arm/plat-omap/dma.c static int __devinit

Quoted 4 lines
arch/arm/plat-omap/dma.c +0 -0
 	for (ch = 0; ch < dma_chan_count; ch++) {
 		unsigned long flags;
 		omap_clear_dma(ch);
@@ -846,6 +855,10 @@ static int __devinit 

omap_system_dma_probe(struct platform_device *pdev)

Quoted 10 lines

arch/arm/plat-omap/dma.c static int __devinit omap_system_dma_probe(struct platform_device *pdev) · line 861

Quoted 7 lines
arch/arm/plat-omap/dma.c +4 -0
 		dma_chan[0].dev_id = 0;
 		dma_chan[1].dev_id = 1;
 	}
+
+	if (!omap_dma_reserve_channels)
+		pm_runtime_put(&pd->dev);
+

So if the reserve channels feature is used, PM is essentially disabled for sDMA. You should add a warning here to that effect as well.

I will add warning comment.

-Manjunath

#46 lore raw

PATCH v3 13/13 OMAP: PM: DMA: Enable runtime pm

Quoted 38 lines · Manjunath Kondaiah
Quoted 30 lines

arch/arm/plat-omap/dma.c int omap_request_dma(int dev_id, const

Quoted 24 lines · Manjunath Kondaiah

Enable runtime pm and use pm_runtime_get and pm_runtime_put for OMAP DMA driver.

Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
---
 arch/arm/plat-omap/dma.c |   13 +++++++++++++
 1 files changed, 13 insertions(+), 0 deletions(-)
arch/arm/plat-omap/dma.c +1 -0
index 41b14b0..feac7ee 100644
@@ -35,6 +35,7 @@
 #include <linux/io.h>
 #include <linux/slab.h>
 #include <linux/delay.h>
+#include <linux/pm_runtime.h>
 
 #include <asm/system.h>
 #include <mach/hardware.h>
@@ -367,6 +368,8 @@ int omap_request_dma(int dev_id, const 

char *dev_name,

Quoted 3 lines
arch/arm/plat-omap/dma.c +0 -0
 	chan = dma_chan + free_ch;
 	chan->dev_id = dev_id;
 
Quoted 7 lines

arch/arm/plat-omap/dma.c int omap_request_dma(int dev_id, const · line 371

Quoted 1 line
arch/arm/plat-omap/dma.c +1 -0
+	pm_runtime_get(&pd->dev);

The _get() call is asynchronous. So if the device was actually idled/disabled, immediately after this call it may still be
idle/disabled. When using the asynchronous versions of the API, the
device should not be touched until the driver's ->runtime_resume() callback is called.

Quoted 3 lines

driver runtime_resume will call omap_device_enable. How do we make we check from driver side, whether omap_device_enable execution is complete or not.

It is complete when the driver's runtime_resume hook is called. The OMAP runtime PM core calls omap_device_enable() before calling the driver's callback. See arch/arm/mach-omap2/pm_bus.c for details.

Alternatively, just use _get_sync() which doesn't return until the device is ready.

Kevin

#47 lore raw

v3 cover OMAP: DMA: hwmod and DMA as platform device

In reply to G, Manjunath Kondaiah

Hi,

Quoted 13 lines

-----Original Message-----
From: linux-omap-owner at vger.kernel.org [mailto:linux-omap-owner at vger.kernel.org] On Behalf Of G, Manjunath Kondaiah
Sent: Tuesday, October 26, 2010 6:55 PM
To: linux-omap at vger.kernel.org
Cc: linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Kevin Hilman; Shilimkar, Santosh
Subject: [PATCH v3 00/13] OMAP: DMA: hwmod and DMA as platform device

This patch series converts DMA library into platform device and hwmod adaptation is done for omap2+ processors.

[...]

Quoted 2 lines

G, Manjunath Kondaiah (12):

[...]

Quoted 1 line

OMAP: DMA: Use DMA device attributes

Looks like this patch is not getting posted due to its size(~105 KB) The patch is available for review comments at:
http://dev.omapzoom.org/?p=manju/kernel-omap3-dev.git;a=commitdiff;h=7457a6b4248772aaa52dfb13a170f891596a512a

-Manjunath

#48 lore raw

v3 cover OMAP: DMA: hwmod and DMA as platform device

In reply to G, Manjunath Kondaiah
Quoted 6 lines · Manjunath Kondaiah

This patch series converts DMA library into platform device and hwmod adaptation is done for omap2+ processors.

It also replaces cpu_*is* checks by moving omap1 and omap2+ processor code into repsective mach-omapx directories. The common code in plat-omap will use device attributes for differentiating omap1 and omap2 code.

This summary, and the various changelogs need to be updated to be more descriptive, particularily in the usage of the words 'device' and 'driver'.

The changelog terminology used is confusing, and sometimes wrong. What is actually being done is splitting up the current DMA library into 3
parts: a common platform_driver (plat-omap/dma.c), OMAP1 device code (mach-omap1/dma.c) and OMAP2+ device code (mach-omap2/dma.c)

Quoted 16 lines

Build Report:
omap1: Success for omap_h2_1610_defconfig
omap2+: Success for omap2plus_defconfig

Test Report:
omap2+:(multiboot defconfig boots on all the omap2+ boards listed below):
OMAP4BLAZE: http://pastebin.com/HVnim30G
OMAP3630ZOOM3: http://pastebin.com/JJwrtP4F
OMAP243SDP: http://pastebin.com/mz7cVQL3

omap1 : Not tested

Test cases executed:
1. All applicable TI DMA tests which are located at:
http://dev.omapzoom.org/?p=richo/device_driver_test.git;a=tree;f=dma/test_code;h=0d00de3c0fe6933b405c62da63f694883f3e4b8f;hb=2c50a5a58dea0ffc2d31b827935aeef9b9d11253

Quoted 8 lines

2. Use case tests : TI MMC tests are executed with differenct types of files systems such as DOS, ext2, ext3 etc on omap4 blaze and omap3630 zoom3. More information can be found at:
http://dev.omapzoom.org/?p=richo/device_driver_test.git;a=tree;f=mmc/test_code;h=d0bc1984eef46ac45719efb02b5c1f4193422a1b;hb=2c50a5a58dea0ffc2d31b827935aeef9b9d11253

Dependencies:
This patch series has dependency on omap_device patch:
https://patchwork.kernel.org/patch/282212/

Thanks for the thorough build and test report. Very nice.

Kevin

Quoted 72 lines

This patch series applies on:
git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap-2.6.git
Branch: master
commit a83d12a47c9a8c78a184910150797045d69a1570
Merge: f9f47f5 c1fb8df
Author: Tony Lindgren <[email protected]>
Date: Fri Oct 22 11:21:08 2010 -0700

Linux-omap rebuilt: Updated to v2.6.36, add 24xx uart fix

Changelist summary:
v3:
Review comments fixed:
http://www.listware.net/201008/linux-omap/89002-patch-v2-0011-omap-dma-hwmod-and-dma-as-platform-driver.html
- created and tested on latest linux-omap master branch(2.6.36 kernel)

v2:
The review comments which are fixed can be found at:
http://www.spinics.net/lists/linux-omap/msg34291.html
http://www.spinics.net/lists/linux-omap/msg34292.html
http://www.spinics.net/lists/linux-omap/msg34078.html
http://www.spinics.net/lists/linux-omap/msg34083.html

v1:
These changes are based on comments received during internal discussions which has changes(compared to previous version) such as:
- Code optimization
- Patches are rearranged in more meaningful way so that git bisect works at any
intermediate patch in the series.
- Build tested for all omap's(OMAP1 and OMAP2PLUS)
- Boot tested for OMAP3 and OMAP4(appreciate if some one tests on OMAP1/2)
- Applicalbe tests are executed on OMAP3 and OMAP4 boards
- Rebased and added descriptor autoloading feature(only for omap3630 and omap4)

Benoit Cousson (1):
OMAP4: DMA: hwmod: add system DMA

G, Manjunath Kondaiah (12):
OMAP: DMA: Replace read/write macros with functions
OMAP: DMA: Introduce errata handling feature
OMAP: DMA: Introduce DMA device attributes
OMAP2420: DMA: hwmod: add system DMA
OMAP2430: DMA: hwmod: add system DMA
OMAP3: DMA: hwmod: add system DMA
OMAP1: DMA: Introduce DMA driver as platform device OMAP2+: DMA: hwmod: Device registration
OMAP: DMA: Convert DMA library into DMA platform Driver
OMAP: DMA: Use DMA device attributes
sDMA: descriptor autoloading feature
OMAP: PM: DMA: Enable runtime pm

 arch/arm/mach-omap1/Makefile               |    2 +-
 arch/arm/mach-omap1/dma.c                  |  665 ++++++++++
 arch/arm/mach-omap1/include/mach/dma.h     |   51 +
 arch/arm/mach-omap2/Makefile               |    2 +-
 arch/arm/mach-omap2/dma.c                  | 1579 ++++++++++++++++++++++++
 arch/arm/mach-omap2/include/mach/dma.h     |   86 ++
 arch/arm/mach-omap2/omap_hwmod_2420_data.c |   85 ++
 arch/arm/mach-omap2/omap_hwmod_2430_data.c |   85 ++
 arch/arm/mach-omap2/omap_hwmod_3xxx_data.c |   93 ++
 arch/arm/mach-omap2/omap_hwmod_44xx_data.c |   99 ++
 arch/arm/plat-omap/dma.c                   | 1850 ++++------------------------
 arch/arm/plat-omap/include/plat/dma.h      |  387 ++++---
 12 files changed, 3207 insertions(+), 1777 deletions(-)
 create mode 100644 arch/arm/mach-omap1/dma.c
 create mode 100644 arch/arm/mach-omap1/include/mach/dma.h
 create mode 100644 arch/arm/mach-omap2/dma.c
 create mode 100644 arch/arm/mach-omap2/include/mach/dma.h
Cc Benoit Cousson <[email protected]>
Cc Kevin Hilman <[email protected]>
Cc Santosh Shilimkar <[email protected]>
#49 lore raw

v3 cover OMAP: DMA: hwmod and DMA as platform device

Quoted 28 lines

-----Original Message-----
From: Kevin Hilman [mailto:khilman at deeprootsystems.com]
Sent: Wednesday, November 10, 2010 4:42 AM
To: G, Manjunath Kondaiah
Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Cousson, Benoit; Shilimkar, Santosh
Subject: Re: [PATCH v3 00/13] OMAP: DMA: hwmod and DMA as platform device

Quoted 1 line · Manjunath Kondaiah

This patch series converts DMA library into platform device

and hwmod

Quoted 3 lines

adaptation is done for omap2+ processors.

It also replaces cpu_*is* checks by moving omap1 and omap2+

processor

Quoted 1 line

code into repsective mach-omapx directories. The common

code in plat-omap

Quoted 1 line

will use device attributes for differentiating omap1 and omap2 code.

This summary, and the various changelogs need to be updated to be more descriptive, particularily in the usage of the words 'device' and 'driver'.

Quoted 5 lines

The changelog terminology used is confusing, and sometimes wrong. What
is actually being done is splitting up the current DMA library into 3
parts: a common platform_driver (plat-omap/dma.c), OMAP1 device code (mach-omap1/dma.c) and OMAP2+ device code (mach-omap2/dma.c)

I will incorporate this with next version.

Quoted 12 lines
Quoted 7 lines

Build Report:
omap1: Success for omap_h2_1610_defconfig
omap2+: Success for omap2plus_defconfig

Test Report:
omap2+:(multiboot defconfig boots on all the omap2+ boards

listed below):

Quoted 4 lines

OMAP4BLAZE: http://pastebin.com/HVnim30G
OMAP3630ZOOM3: http://pastebin.com/JJwrtP4F
OMAP243SDP: http://pastebin.com/mz7cVQL3

Quoted 6 lines
Quoted 5 lines

omap1 : Not tested

Test cases executed:
1. All applicable TI DMA tests which are located at:

http://dev.omapzoom.org/?p=richo/device_driver_test.git;a=tree

;f=dma/test_code;h=0d00de3c0fe6933b405c62da63f694883f3e4b8f;hb=> 2c50a5a58dea0ffc2d31b827935aeef9b9d11253

Quoted 8 lines
Quoted 2 lines

2. Use case tests : TI MMC tests are executed with

differenct types of

Quoted 1 line

files systems such as DOS, ext2, ext3 etc on omap4 blaze

and omap3630 zoom3.

Quoted 2 lines

More information can be found at:

http://dev.omapzoom.org/?p=richo/device_driver_test.git;a=tree

;f=mmc/test_code;h=d0bc1984eef46ac45719efb02b5c1f4193422a1b;hb=> 2c50a5a58dea0ffc2d31b827935aeef9b9d11253

Quoted 6 lines
Quoted 4 lines

Dependencies:
This patch series has dependency on omap_device patch:
https://patchwork.kernel.org/patch/282212/

Thanks for the thorough build and test report. Very nice.

Thanks for the feedback.

-Manjunath