Skip to contentSkip to conversation

i2c: i2c-mxs: Use DMA mode even for small transfers

PATCH Applied Fabio Estevam 8 participants 17 messages Jul 1, 2013 – Aug 5, 2013 6 messages not in the archive

Subsystems (2)
lore
Share & tools
lore

Fabio Estevam

author reviewer
#0 lore raw

PATCH v2 1/1 i2c: i2c-mxs: Use DMA mode even for small transfers

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

Let's use DMA mode even for small transfers.

Without this patch, i2c reads the incorrect sgtl5000 version on a mx28evk when touchscreen is enabled:

[    5.856270] sgtl5000 0-000a: Device with ID register 0 is not a sgtl5000     
[    9.877307] sgtl5000 0-000a: ASoC: failed to probe CODEC -19                 
[    9.883528] mxs-sgtl5000 sound.12: ASoC: failed to instantiate card -19      
[    9.892955] mxs-sgtl5000 sound.12: snd_soc_register_card failed (-19)  

---
Applies against 3.10

Changes since v1:
- Keep the PIO code

 drivers/i2c/busses/i2c-mxs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/i2c/busses/i2c-mxs.c +1 -1
index 2039f23..6d8094d 100644
@@ -494,7 +494,7 @@ static int mxs_i2c_xfer_msg(struct i2c_adapter *adap, struct i2c_msg *msg,
 	 * based on this empirical measurement and a lot of previous frobbing.
 	 */
 	i2c->cmd_err = 0;
-	if (msg->len < 8) {
+	if (0) {	/* disable PIO mode until a proper fix is made */
 		ret = mxs_i2c_pio_setup_xfer(adap, msg, flags);
 		if (ret)
 			mxs_i2c_reset(i2c);
Signature · 1 line
1.8.1.2
Acked-by on v2 1/1
#1 lore raw
Quoted 8 lines · Fabio Estevam

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

It also causes sgtl5000 register write failure [1]. Changing it back to DMA mode fixes the failure. So,

[1] http://thread.gmane.org/gmane.linux.alsa.devel/109715/focus=109747

author reviewer
#2 lore raw

Hi Wolfram,

Quoted 9 lines · Shawn Guo
Quoted 8 lines · Fabio Estevam

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

Quoted 7 lines

It also causes sgtl5000 register write failure [1]. Changing it back to DMA mode fixes the failure. So,

[1] http://thread.gmane.org/gmane.linux.alsa.devel/109715/focus=109747

Any comment about this patch, please?

If you are happy with it, could we please have it applied into stable?

3.10 is currently broken without this patch.

Regards,

Fabio Estevam

A message that is not in the archive (sent off-list or not yet received): CAOMZO5Dyk_5CzcXvG7eJ3h07i=So0DS7MjtMQVHfpfx9Se_Teg-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#4 lore raw

Dear Fabio Estevam,

Quoted 18 lines

Hi Wolfram,

Quoted 15 lines · Shawn Guo
Quoted 8 lines · Fabio Estevam

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

It also causes sgtl5000 register write failure [1]. Changing it back to DMA mode fixes the failure. So,

Quoted 7 lines
Quoted 1 line

[1] http://thread.gmane.org/gmane.linux.alsa.devel/109715/focus=109747

Any comment about this patch, please?

If you are happy with it, could we please have it applied into stable?

3.10 is currently broken without this patch.

Hi, either this or the PIO fix, but that's a RFC and quite a big slab of change which still needs some work. So please apply this and let's try the PIOfix for 3.11 or 3.12.

Best regards,
Marek Vasut

author reviewer
#5 lore raw
In reply to Fabio Estevam

Wolfram,

Quoted 5 lines · Fabio Estevam

Any comment about this patch, please?

If you are happy with it, could we please have it applied into stable?

3.10 is currently broken without this patch.

Ping?

#6 lore raw

Dear Fabio Estevam,

Quoted 8 lines

Wolfram,

Quoted 5 lines · Fabio Estevam

Any comment about this patch, please?

If you are happy with it, could we please have it applied into stable?

3.10 is currently broken without this patch.

Yes, the big question is if we want to use this patch or fix the PIO. I think this one is much less intrusive, so I'd vouch for yours for 3.10 .

Best regards,
Marek Vasut

author reviewer
#7 lore raw

Hi Marek,

Quoted 2 lines · Marek Vasut

Yes, the big question is if we want to use this patch or fix the PIO. I think this one is much less intrusive, so I'd vouch for yours for 3.10 .

Correct. Also according to Documentation/stable_kernel_rules.txt :

"Rules on what kind of patches are accepted, and which ones are not, into the "-stable" tree:

....
- It cannot be bigger than 100 lines, with context."

Your PIO fix is really appreciated, but it surpasses 100 lines, so it would be better if we go with this one first so that it can reach 3.10, and then we can use your PIO fix later.

Regards,

Fabio Estevam

#8 lore raw
Quoted 12 lines · Fabio Estevam

Hi Marek,

Quoted 2 lines · Marek Vasut

Yes, the big question is if we want to use this patch or fix the PIO. I think this one is much less intrusive, so I'd vouch for yours for 3.10 .

Correct. Also according to Documentation/stable_kernel_rules.txt :

"Rules on what kind of patches are accepted, and which ones are not, into the "-stable" tree:

Quoted 6 lines

....
- It cannot be bigger than 100 lines, with context."

Your PIO fix is really appreciated, but it surpasses 100 lines, so it would be better if we go with this one first so that it can reach 3.10, and then we can use your PIO fix later.

That one is not a hard and fast rule, but small localised fixes are preferred if possible.

Ben.

Signature · 2 lines
Ben Hutchings
Once a job is fouled up, anything done to improve it makes it worse.
#9 lore raw

Dear Ben Hutchings,

Quoted 15 lines · Fabio Estevam
Quoted 14 lines

Hi Marek,

Quoted 3 lines · Marek Vasut

Yes, the big question is if we want to use this patch or fix the PIO. I think this one is much less intrusive, so I'd vouch for yours for 3.10 .

Correct. Also according to Documentation/stable_kernel_rules.txt :

"Rules on what kind of patches are accepted, and which ones are not, into the "-stable" tree:

....

Quoted 8 lines
Quoted 5 lines

- It cannot be bigger than 100 lines, with context."

Your PIO fix is really appreciated, but it surpasses 100 lines, so it would be better if we go with this one first so that it can reach 3.10, and then we can use your PIO fix later.

That one is not a hard and fast rule, but small localised fixes are preferred if possible.

Full agreement on this one, let's apply this patch.

Best regards,
Marek Vasut

A message that is not in the archive (sent off-list or not yet received): 1372713261-20551-1-git-send-email-festevam-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org

Acked-by
#11 lore raw

Am Montag, den 01.07.2013, 18:14 -0300 schrieb Fabio Estevam:

Quoted 34 lines

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

Let's use DMA mode even for small transfers.

Without this patch, i2c reads the incorrect sgtl5000 version on a mx28evk when touchscreen is enabled:

[    5.856270] sgtl5000 0-000a: Device with ID register 0 is not a sgtl5000     
[    9.877307] sgtl5000 0-000a: ASoC: failed to probe CODEC -19                 
[    9.883528] mxs-sgtl5000 sound.12: ASoC: failed to instantiate card -19      
[    9.892955] mxs-sgtl5000 sound.12: snd_soc_register_card failed (-19)  

---
Applies against 3.10

Changes since v1:
- Keep the PIO code

 drivers/i2c/busses/i2c-mxs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/i2c/busses/i2c-mxs.c +0 -0
index 2039f23..6d8094d 100644

drivers/i2c/busses/i2c-mxs.c static int mxs_i2c_xfer_msg(struct i2c_adapter *adap, struct i2c_msg *msg, · line 500

Quoted 8 lines
drivers/i2c/busses/i2c-mxs.c +1 -1
@@ -494,7 +494,7 @@ static int mxs_i2c_xfer_msg(struct i2c_adapter *adap, struct i2c_msg *msg,
 	 * based on this empirical measurement and a lot of previous frobbing.
 	 */
 	i2c->cmd_err = 0;
-	if (msg->len < 8) {
+	if (0) {	/* disable PIO mode until a proper fix is made */
 		ret = mxs_i2c_pio_setup_xfer(adap, msg, flags);
 		if (ret)
 			mxs_i2c_reset(i2c);

I still plan to take another look at the PIO mode, but higher priority things keep popping up in front of me. So this patch is:

Signature · 4 lines
Pengutronix e.K.                           | Lucas Stach                 |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-5076 |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
#12 lore raw

Dear Lucas Stach,

Quoted 47 lines

Am Montag, den 01.07.2013, 18:14 -0300 schrieb Fabio Estevam:

Quoted 46 lines

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

Let's use DMA mode even for small transfers.

Without this patch, i2c reads the incorrect sgtl5000 version on a mx28evk when touchscreen is enabled:

[    5.856270] sgtl5000 0-000a: Device with ID register 0 is not a
sgtl5000 [    9.877307] sgtl5000 0-000a: ASoC: failed to probe CODEC -19
[    9.883528] mxs-sgtl5000 sound.12: ASoC: failed to instantiate card

-19 [ 9.892955] mxs-sgtl5000 sound.12: snd_soc_register_card failed
(-19)

---
Applies against 3.10

Changes since v1:
- Keep the PIO code

 drivers/i2c/busses/i2c-mxs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/i2c/busses/i2c-mxs.c +1 -1
index 2039f23..6d8094d 100644
@@ -494,7 +494,7 @@ static int mxs_i2c_xfer_msg(struct i2c_adapter *adap,
 struct i2c_msg *msg,
 
 	 * based on this empirical measurement and a lot of previous frobbing.
 	 */
 
 	i2c->cmd_err = 0;
 
-	if (msg->len < 8) {
+	if (0) {	/* disable PIO mode until a proper fix is made */
Quoted 8 lines

drivers/i2c/busses/i2c-mxs.c static int mxs_i2c_xfer_msg(struct i2c_adapter *adap, struct i2c_msg *msg, · line 500

Quoted 4 lines
drivers/i2c/busses/i2c-mxs.c +0 -0
 		ret = mxs_i2c_pio_setup_xfer(adap, msg, flags);
 		if (ret)
 
 			mxs_i2c_reset(i2c);

I still plan to take another look at the PIO mode, but higher priority things keep popping up in front of me. So this patch is:

Ok, update. So far I arrived at a point where I can avoid the PIO trouble.

If I only do transfers shorter or equal in length to 3 bytes via PIO, it works as expected. If the transfer is longer, the LA shows very long transfer happening actually. Therefore, the pumping of data in loop to/from PIO via the DATA register doesn't work.

I will update you more later, once I figure out something else. Now I need some sleep.

Best regards,
Marek Vasut

A message that is not in the archive (sent off-list or not yet received): [email protected]

#14 lore raw

Hi,

Quoted 60 lines

Dear Lucas Stach,

Quoted 47 lines

Am Montag, den 01.07.2013, 18:14 -0300 schrieb Fabio Estevam:

Quoted 46 lines

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

Let's use DMA mode even for small transfers.

Without this patch, i2c reads the incorrect sgtl5000 version on a mx28evk when touchscreen is enabled:

[    5.856270] sgtl5000 0-000a: Device with ID register 0 is not a

sgtl5000 [ 9.877307] sgtl5000 0-000a: ASoC: failed to probe CODEC
-19 [ 9.883528] mxs-sgtl5000 sound.12: ASoC: failed to instantiate
card -19 [ 9.892955] mxs-sgtl5000 sound.12: snd_soc_register_card
failed (-19)

---
Applies against 3.10

Changes since v1:
- Keep the PIO code

 drivers/i2c/busses/i2c-mxs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
(quoted diff) +0 -0

b/drivers/i2c/busses/i2c-mxs.c index 2039f23..6d8094d 100644

drivers/i2c/busses/i2c-mxs.c +1 -1
@@ -494,7 +494,7 @@ static int mxs_i2c_xfer_msg(struct i2c_adapter
 *adap, struct i2c_msg *msg,
 
 	 * based on this empirical measurement and a lot of previous
 	 frobbing. */
 
 	i2c->cmd_err = 0;
 
-	if (msg->len < 8) {
+	if (0) {	/* disable PIO mode until a proper fix is made */
Quoted 8 lines

drivers/i2c/busses/i2c-mxs.c

Quoted 4 lines
drivers/i2c/busses/i2c-mxs.c +0 -0
 		ret = mxs_i2c_pio_setup_xfer(adap, msg, flags);
 		if (ret)
 
 			mxs_i2c_reset(i2c);

I still plan to take another look at the PIO mode, but higher priority things keep popping up in front of me. So this patch is:

Ok, update. So far I arrived at a point where I can avoid the PIO trouble.

Quoted 7 lines

If I only do transfers shorter or equal in length to 3 bytes via PIO, it works as expected. If the transfer is longer, the LA shows very long transfer happening actually. Therefore, the pumping of data in loop to/from PIO via the DATA register doesn't work.

I will update you more later, once I figure out something else. Now I need some sleep.

I'm attaching a patch. Alex, please give it a go and see if it fixes your issue. It is _VERY_ ugly.

The basic idea behind the the patch is that, as (attempted to be) explained above, subsequent writes to DATA register in PIO mode cause constant generation of clock on the bus and therefore a very long transfer of zero data. This confuses the I2C peripherals of course.

The patch implements clock stretching for PIO writes (maybe we need this for reads too) by making the controller blast out only 4 (or less) bytes of data in each write into the DATA register. To prevent interruption of the transfer between writes into the DATA register, the SCK is held low using the RETAIN_CLOCK bit.

But (!) here comes the caveat. The PIO was introduced to speed up small transfers. Introducing clock stretching into PIO mode operation might completely remove this advantage. This has to be measured again.

I will continue once I sleep a little. Pardon my {language, code}, it's too early in the morning already.

Best regards,
Marek Vasut

drivers/i2c/busses/i2c-mxs.c +75 -15
index df8ff5a..b125183 100644
@@ -35,6 +35,7 @@
 
 #define MXS_I2C_CTRL0		(0x00)
 #define MXS_I2C_CTRL0_SET	(0x04)
+#define MXS_I2C_CTRL0_CLR	(0x08)
 
 #define MXS_I2C_CTRL0_SFTRST			0x80000000
 #define MXS_I2C_CTRL0_RUN			0x20000000
@@ -123,6 +124,32 @@ struct mxs_i2c_dev {
 	bool				dma_read;
 };
 
+static void mxs_i2c_dump(struct mxs_i2c_dev *i2c, int rd)
+{
+	pr_err("=====================================\n");
+
+	pr_err("0x000: %08x %08x %08x %08x\n",
+		readl(i2c->regs + 0x00),
+		readl(i2c->regs + 0x10),
+		readl(i2c->regs + 0x20),
+		readl(i2c->regs + 0x30));
+	pr_err("0x040: %08x %08x %08x %08x\n",
+		readl(i2c->regs + 0x40),
+		readl(i2c->regs + 0x50),
+		readl(i2c->regs + 0x60),
+		readl(i2c->regs + 0x70));
+	pr_err("0x080: %08x %08x %08x %08x\n",
+		readl(i2c->regs + 0x80),
+		rd ? readl(i2c->regs + 0x90) : 0,
+		rd ? readl(i2c->regs + 0xa0) : 0,
+		readl(i2c->regs + 0xb0));
+	pr_err("0x0c0: %08x %08x\n",
+		readl(i2c->regs + 0xc0),
+		readl(i2c->regs + 0xd0));
+
+	pr_err("=====================================\n");
+}
+
 static void mxs_i2c_reset(struct mxs_i2c_dev *i2c)
 {
 	stmp_reset_block(i2c->regs);
@@ -366,6 +393,7 @@ static void mxs_i2c_pio_trigger_cmd(struct mxs_i2c_dev *i2c, u32 cmd)
 	writel(reg, i2c->regs + MXS_I2C_CTRL0);
 }
 
+#include <linux/delay.h>
 static int mxs_i2c_pio_setup_xfer(struct i2c_adapter *adap,
 			struct i2c_msg *msg, uint32_t flags)
 {
@@ -401,6 +429,7 @@ static int mxs_i2c_pio_setup_xfer(struct i2c_adapter *adap,
 		mxs_i2c_pio_trigger_cmd(i2c,
 					MXS_CMD_I2C_READ | flags |
 					MXS_I2C_CTRL0_XFER_COUNT(msg->len));
+//		mxs_i2c_dump(i2c, 0);
 
 		for (i = 0; i < msg->len; i++) {
 			if ((i & 3) == 0) {
@@ -418,38 +447,66 @@ static int mxs_i2c_pio_setup_xfer(struct i2c_adapter *adap,
 		addr_data |= I2C_SMBUS_WRITE;
 
 		/* WRITE command. */
-		mxs_i2c_pio_trigger_cmd(i2c,
-					MXS_CMD_I2C_WRITE | flags |
-					MXS_I2C_CTRL0_XFER_COUNT(msg->len + 1));
+//		mxs_i2c_pio_trigger_cmd(i2c,
+//					MXS_CMD_I2C_WRITE | flags |
+//					MXS_I2C_CTRL0_XFER_COUNT(msg->len + 1));
 
 		/*
 		 * The LSB of data buffer is the first byte blasted across
 		 * the bus. Higher order bytes follow. Thus the following
 		 * filling schematic.
 		 */
+
 		data = addr_data << 24;
 		for (i = 0; i < msg->len; i++) {
 			data >>= 8;
 			data |= (msg->buf[i] << 24);
 			if ((i & 3) == 2) {
-				ret = mxs_i2c_pio_wait_dmareq(i2c);
-				if (ret)
-					return ret;
+				mxs_i2c_pio_trigger_cmd(i2c,
+					MXS_CMD_I2C_WRITE /*| flags*/ | (1 << 21) |
+					MXS_I2C_CTRL0_XFER_COUNT(4));
+
+	//			ret = mxs_i2c_pio_wait_dmareq(i2c);
+	//			if (ret)
+	//				return ret;
+
+//				mxs_i2c_dump(i2c, 0);
+//				pr_err("writing %08x\n", data);
+
 				writel(data, i2c->regs + MXS_I2C_DATA);
-				writel(MXS_I2C_DEBUG0_DMAREQ,
-				       i2c->regs + MXS_I2C_DEBUG0_CLR);
+//				writel(1 << 21, i2c->regs + MXS_I2C_CTRL0_SET);
+
+	//			writel(MXS_I2C_DEBUG0_DMAREQ,
+	//			       i2c->regs + MXS_I2C_DEBUG0_CLR);
+//				mdelay(20);
+//				mxs_i2c_dump(i2c, 0);
+
 			}
 		}
 
 		shifts_left = 24 - (i & 3) * 8;
 		if (shifts_left) {
 			data >>= shifts_left;
-			ret = mxs_i2c_pio_wait_dmareq(i2c);
-			if (ret)
-				return ret;
+			if (msg->len <= 3)
+				flags |= MXS_CMD_I2C_WRITE;
+			mxs_i2c_pio_trigger_cmd(i2c,
+				MXS_I2C_CTRL0_MASTER_MODE | MXS_I2C_CTRL0_DIRECTION | flags |
+				MXS_I2C_CTRL0_XFER_COUNT((i & 3) + 1));
+
+	//		ret = mxs_i2c_pio_wait_dmareq(i2c);
+	//		if (ret)
+	//			return ret;
+
 			writel(data, i2c->regs + MXS_I2C_DATA);
-			writel(MXS_I2C_DEBUG0_DMAREQ,
-			       i2c->regs + MXS_I2C_DEBUG0_CLR);
+
+//			mxs_i2c_dump(i2c, 0);
+//			pr_err("writing %08x %i\n", data, (i & 3) + 1);
+
+	//		writel(MXS_I2C_DEBUG0_DMAREQ,
+	//		       i2c->regs + MXS_I2C_DEBUG0_CLR);
+//			mdelay(20);
+//			mxs_i2c_dump(i2c, 0);
+
 		}
 	}
 
@@ -480,8 +537,10 @@ static int mxs_i2c_xfer_msg(struct i2c_adapter *adap, struct i2c_msg *msg,
 
 	flags = stop ? MXS_I2C_CTRL0_POST_SEND_STOP : 0;
 
-	dev_dbg(i2c->dev, "addr: 0x%04x, len: %d, flags: 0x%x, stop: %d\n",
-		msg->addr, msg->len, msg->flags, stop);
+	dev_dbg(i2c->dev, "addr: 0x%04x, len: %d, flags: 0x%x [%c], stop: %d\n",
+		msg->addr, msg->len, msg->flags, msg->flags & I2C_M_RD ? 'R':'W', stop);
+//	if (msg->len)
+//		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_NONE, 16, 1, msg->buf, msg->len, false);
 
 	if (msg->len == 0)
 		return -EINVAL;
@@ -497,6 +556,7 @@ static int mxs_i2c_xfer_msg(struct i2c_adapter *adap, struct i2c_msg *msg,
 		ret = mxs_i2c_pio_setup_xfer(adap, msg, flags);
 		if (ret)
 			mxs_i2c_reset(i2c);
+		i2c->cmd_err = ret;
 	} else {
 		INIT_COMPLETION(i2c->cmd_complete);
 		ret = mxs_i2c_dma_setup_xfer(adap, msg, flags);
#15 lore raw

Hi Marek,

Quoted 4 lines · Marek Vasut

I'm attaching a patch. Alex, please give it a go and see if it fixes your issue. It is _VERY_ ugly.

Quite ugly ;)

It indeed seems to fix the issue.

Quoted 11 lines

The basic idea behind the the patch is that, as (attempted to be) explained above, subsequent writes to DATA register in PIO mode cause constant generation of clock on the bus and therefore a very long transfer of zero data. This confuses the I2C peripherals of course.

The patch implements clock stretching for PIO writes (maybe we need this for reads too) by making the controller blast out only 4 (or less) bytes of data in each write into the DATA register. To prevent interruption of the transfer between writes into the DATA register, the SCK is held low using the RETAIN_CLOCK bit.

Quoted 4 lines

But (!) here comes the caveat. The PIO was introduced to speed up small transfers. Introducing clock stretching into PIO mode operation might completely remove this advantage. This has to be measured again.

And now, PIO mode is slower than DMA...

Signature · 3 lines
Alexandre Belloni, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

A message that is not in the archive (sent off-list or not yet received): 51D3E1D2.4070706-wi1+55ScJUtKEb57/[email protected]

#17 lore raw

Dear Alexandre Belloni,

Quoted 21 lines

Hi Marek,

Quoted 2 lines · Marek Vasut

I'm attaching a patch. Alex, please give it a go and see if it fixes your issue. It is _VERY_ ugly.

Quite ugly ;)

It indeed seems to fix the issue.

Quoted 11 lines

The basic idea behind the the patch is that, as (attempted to be) explained above, subsequent writes to DATA register in PIO mode cause constant generation of clock on the bus and therefore a very long transfer of zero data. This confuses the I2C peripherals of course.

The patch implements clock stretching for PIO writes (maybe we need this for reads too) by making the controller blast out only 4 (or less) bytes of data in each write into the DATA register. To prevent interruption of the transfer between writes into the DATA register, the SCK is held low using the RETAIN_CLOCK bit.

Quoted 5 lines
Quoted 3 lines

But (!) here comes the caveat. The PIO was introduced to speed up small transfers. Introducing clock stretching into PIO mode operation might completely remove this advantage. This has to be measured again.

And now, PIO mode is slower than DMA...

Further digging needed then. I will update you when I'm done, but I won't be in my office tomorrow, so probably on friday.

Best regards,
Marek Vasut

A message that is not in the archive (sent off-list or not yet received): [email protected]

#19 lore raw

Am Mittwoch, den 03.07.2013, 06:37 +0200 schrieb Marek Vasut:

Quoted 84 lines

Hi,

Quoted 60 lines

Dear Lucas Stach,

Quoted 47 lines

Am Montag, den 01.07.2013, 18:14 -0300 schrieb Fabio Estevam:

Quoted 46 lines

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

Let's use DMA mode even for small transfers.

Without this patch, i2c reads the incorrect sgtl5000 version on a mx28evk when touchscreen is enabled:

[    5.856270] sgtl5000 0-000a: Device with ID register 0 is not a

sgtl5000 [ 9.877307] sgtl5000 0-000a: ASoC: failed to probe CODEC
-19 [ 9.883528] mxs-sgtl5000 sound.12: ASoC: failed to instantiate
card -19 [ 9.892955] mxs-sgtl5000 sound.12: snd_soc_register_card
failed (-19)

---
Applies against 3.10

Changes since v1:
- Keep the PIO code

 drivers/i2c/busses/i2c-mxs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
(quoted diff) +0 -0

b/drivers/i2c/busses/i2c-mxs.c index 2039f23..6d8094d 100644

drivers/i2c/busses/i2c-mxs.c +1 -1
@@ -494,7 +494,7 @@ static int mxs_i2c_xfer_msg(struct i2c_adapter
 *adap, struct i2c_msg *msg,
 
 	 * based on this empirical measurement and a lot of previous
 	 frobbing. */
 
 	i2c->cmd_err = 0;
 
-	if (msg->len < 8) {
+	if (0) {	/* disable PIO mode until a proper fix is made */
Quoted 8 lines

drivers/i2c/busses/i2c-mxs.c static int mxs_i2c_xfer_msg(struct i2c_adapter *adap, struct i2c_msg *msg, · line 558

Quoted 4 lines
drivers/i2c/busses/i2c-mxs.c +0 -0
 		ret = mxs_i2c_pio_setup_xfer(adap, msg, flags);
 		if (ret)
 
 			mxs_i2c_reset(i2c);

I still plan to take another look at the PIO mode, but higher priority things keep popping up in front of me. So this patch is:

Ok, update. So far I arrived at a point where I can avoid the PIO trouble.

Quoted 7 lines

If I only do transfers shorter or equal in length to 3 bytes via PIO, it works as expected. If the transfer is longer, the LA shows very long transfer happening actually. Therefore, the pumping of data in loop to/from PIO via the DATA register doesn't work.

I will update you more later, once I figure out something else. Now I need some sleep.

I'm attaching a patch. Alex, please give it a go and see if it fixes your issue. It is _VERY_ ugly.

The basic idea behind the the patch is that, as (attempted to be) explained above, subsequent writes to DATA register in PIO mode cause constant generation of clock on the bus and therefore a very long transfer of zero data. This confuses the I2C peripherals of course.

The patch implements clock stretching for PIO writes (maybe we need this for reads too) by making the controller blast out only 4 (or less) bytes of data in each write into the DATA register. To prevent interruption of the transfer between writes into the DATA register, the SCK is held low using the RETAIN_CLOCK bit.

Quoted 7 lines

But (!) here comes the caveat. The PIO was introduced to speed up small transfers. Introducing clock stretching into PIO mode operation might completely remove this advantage. This has to be measured again.

I will continue once I sleep a little. Pardon my {language, code}, it's too early in the morning already.

Ok, so if this patch works, we may just want to restrict PIO mode to 4 or 3 Byte transfers, where we don't need to use this ugly clock stretching. This way we may gain the PIO speedup for single register reads/writes, which is a pretty common operation, but fall back to the more reliable DMA mode for bigger transfers.

Regards,
Lucas

Signature · 4 lines
Pengutronix e.K.                           | Lucas Stach                 |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-5076 |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

A message that is not in the archive (sent off-list or not yet received): 1372842695.4440.3.camel-WzVe3FnzCwFR6QfukMTsflXZhhPuCNm+@public.gmane.org

#21 lore raw

Dear Lucas Stach,

Quoted 44 lines

Am Mittwoch, den 03.07.2013, 06:37 +0200 schrieb Marek Vasut:

Quoted 43 lines

Hi,

Quoted 41 lines

Dear Lucas Stach,

Quoted 39 lines

Am Montag, den 01.07.2013, 18:14 -0300 schrieb Fabio Estevam:

Quoted 38 lines

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

Let's use DMA mode even for small transfers.

Without this patch, i2c reads the incorrect sgtl5000 version on a mx28evk when touchscreen is enabled:

[    5.856270] sgtl5000 0-000a: Device with ID register 0 is not a

sgtl5000 [ 9.877307] sgtl5000 0-000a: ASoC: failed to probe
CODEC -19 [ 9.883528] mxs-sgtl5000 sound.12: ASoC: failed to
instantiate card -19 [ 9.892955] mxs-sgtl5000 sound.12:
snd_soc_register_card failed (-19)

---
Applies against 3.10

Changes since v1:
- Keep the PIO code

 drivers/i2c/busses/i2c-mxs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
(quoted diff) +0 -0

b/drivers/i2c/busses/i2c-mxs.c index 2039f23..6d8094d 100644

drivers/i2c/busses/i2c-mxs.c +0 -0
@@ -494,7 +494,7 @@ static int mxs_i2c_xfer_msg(struct i2c_adapter
 *adap, struct i2c_msg *msg,
 
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

drivers/i2c/busses/i2c-mxs.c static int mxs_i2c_xfer_msg(struct i2c_adapter

Quoted 7 lines
drivers/i2c/busses/i2c-mxs.c +1 -1
 	 * based on this empirical measurement and a lot of previous
 	 frobbing. */
 
 	i2c->cmd_err = 0;
 
-	if (msg->len < 8) {
+	if (0) {	/* disable PIO mode until a proper fix is made 

*/

Quoted 39 lines
Quoted 39 lines
Quoted 13 lines
Quoted 9 lines

drivers/i2c/busses/i2c-mxs.c static int mxs_i2c_xfer_msg(struct i2c_adapter *adap, struct i2c_msg *msg, · line 558

Quoted 5 lines
drivers/i2c/busses/i2c-mxs.c +0 -0
 		ret = mxs_i2c_pio_setup_xfer(adap, msg, flags);
 		if (ret)
 
 			mxs_i2c_reset(i2c);

I still plan to take another look at the PIO mode, but higher priority things keep popping up in front of me. So this patch is:

Ok, update. So far I arrived at a point where I can avoid the PIO trouble.

Quoted 7 lines

If I only do transfers shorter or equal in length to 3 bytes via PIO, it works as expected. If the transfer is longer, the LA shows very long transfer happening actually. Therefore, the pumping of data in loop to/from PIO via the DATA register doesn't work.

I will update you more later, once I figure out something else. Now I need some sleep.

I'm attaching a patch. Alex, please give it a go and see if it fixes your issue. It is _VERY_ ugly.

The basic idea behind the the patch is that, as (attempted to be) explained above, subsequent writes to DATA register in PIO mode cause constant generation of clock on the bus and therefore a very long transfer of zero data. This confuses the I2C peripherals of course.

The patch implements clock stretching for PIO writes (maybe we need this for reads too) by making the controller blast out only 4 (or less) bytes of data in each write into the DATA register. To prevent interruption of the transfer between writes into the DATA register, the SCK is held low using the RETAIN_CLOCK bit.

But (!) here comes the caveat. The PIO was introduced to speed up small transfers. Introducing clock stretching into PIO mode operation might completely remove this advantage. This has to be measured again.

Quoted 5 lines
Quoted 2 lines

I will continue once I sleep a little. Pardon my {language, code}, it's too early in the morning already.

Ok, so if this patch works, we may just want to restrict PIO mode to 4 or 3 Byte transfers

But then we will end up with pretty useless PIO implementation, since most of the register IO is around 4-bytes long ;-)

Quoted 4 lines

where we don't need to use this ugly clock
stretching. This way we may gain the PIO speedup for single register reads/writes, which is a pretty common operation, but fall back to the more reliable DMA mode for bigger transfers.

I will need to check some more and especially do some profiling and timing, otherwise further discussion here is moot .

Best regards,
Marek Vasut

Applied
#22 lore raw
In reply to Fabio Estevam
Quoted 13 lines · Fabio Estevam

From: Fabio Estevam <[email protected]>

Recently we have been seing some reports about PIO mode not working properly.

- http://www.spinics.net/lists/linux-i2c/msg11985.html
- http://marc.info/?l=linux-i2c&m=137235593101385&w=2
- https://lkml.org/lkml/2013/6/24/430

Let's use DMA mode even for small transfers.

Without this patch, i2c reads the incorrect sgtl5000 version on a mx28evk when touchscreen is enabled:

Quoted 7 lines
[    5.856270] sgtl5000 0-000a: Device with ID register 0 is not a sgtl5000     
[    9.877307] sgtl5000 0-000a: ASoC: failed to probe CODEC -19                 
[    9.883528] mxs-sgtl5000 sound.12: ASoC: failed to instantiate card -19      
[    9.892955] mxs-sgtl5000 sound.12: snd_soc_register_card failed (-19)  

Applied to for-current, thanks!