Skip to contentSkip to conversation

nand_base: Add timeout for NAND reset command

PATCH New Peter Tyser 3 participants 12 messages Feb 4, 2009 – Feb 6, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Without the timeout present an infinite loop can occur if the NAND device is broken or not present.

---
 drivers/mtd/nand/nand_base.c |   19 +++++++++++++++++--
 1 files changed, 17 insertions(+), 2 deletions(-)
drivers/mtd/nand/nand_base.c +17 -2
index bfa7874..a2656eb 100644
@@ -75,6 +75,17 @@
 #include <jffs2/jffs2.h>
 #endif
 
+/*
+ * CONFIG_SYS_NAND_RESET_CNT is used as a timeout mechanism when resetting
+ * a flash.  NAND flash is initialized prior to interrupts so standard timers
+ * can't be used.  CONFIG_SYS_NAND_RESET_CNT should be set to a value
+ * which is greater than (max NAND reset time / NAND status read time).
+ * A conservative default of 200000 (500 us / 25 ns) is used as a default.
+ */
+#ifndef CONFIG_SYS_NAND_RESET_CNT
+#define CONFIG_SYS_NAND_RESET_CNT 200000
+#endif
+
 /* Define default oob placement schemes for large and small page devices */
 static struct nand_ecclayout nand_oob_8 = {
 	.eccbytes = 3,
@@ -524,6 +535,7 @@ static void nand_command(struct mtd_info *mtd, unsigned int command,
 {
 	register struct nand_chip *chip = mtd->priv;
 	int ctrl = NAND_CTRL_CLE | NAND_CTRL_CHANGE;
+	uint32_t rst_sts_cnt = CONFIG_SYS_NAND_RESET_CNT;
 
 	/*
 	 * Write out the command to the device.
@@ -590,7 +602,8 @@ static void nand_command(struct mtd_info *mtd, unsigned int command,
 			       NAND_CTRL_CLE | NAND_CTRL_CHANGE);
 		chip->cmd_ctrl(mtd,
 			       NAND_CMD_NONE, NAND_NCE | NAND_CTRL_CHANGE);
-		while (!(chip->read_byte(mtd) & NAND_STATUS_READY)) ;
+		while (!(chip->read_byte(mtd) & NAND_STATUS_READY) &&
+			(rst_sts_cnt--));
 		return;
 
 		/* This applies to read commands */
@@ -626,6 +639,7 @@ static void nand_command_lp(struct mtd_info *mtd, unsigned int command,
 			    int column, int page_addr)
 {
 	register struct nand_chip *chip = mtd->priv;
+	uint32_t rst_sts_cnt = CONFIG_SYS_NAND_RESET_CNT;
 
 	/* Emulate NAND_CMD_READOOB */
 	if (command == NAND_CMD_READOOB) {
@@ -696,7 +710,8 @@ static void nand_command_lp(struct mtd_info *mtd, unsigned int command,
 			       NAND_NCE | NAND_CLE | NAND_CTRL_CHANGE);
 		chip->cmd_ctrl(mtd, NAND_CMD_NONE,
 			       NAND_NCE | NAND_CTRL_CHANGE);
-		while (!(chip->read_byte(mtd) & NAND_STATUS_READY)) ;
+		while (!(chip->read_byte(mtd) & NAND_STATUS_READY) &&
+			(rst_sts_cnt--));
 		return;
 
 	case NAND_CMD_RNDOUT:
Signature · 1 line
1.6.0.2.GIT
#1 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

drivers/mtd/nand/nand_base.c line 87

Quoted 10 lines · Peter Tyser
drivers/mtd/nand/nand_base.c +10 -0
+/*
+ * CONFIG_SYS_NAND_RESET_CNT is used as a timeout mechanism when resetting
+ * a flash.  NAND flash is initialized prior to interrupts so standard timers
+ * can't be used.  CONFIG_SYS_NAND_RESET_CNT should be set to a value
+ * which is greater than (max NAND reset time / NAND status read time).
+ * A conservative default of 200000 (500 us / 25 ns) is used as a default.
+ */
+#ifndef CONFIG_SYS_NAND_RESET_CNT
+#define CONFIG_SYS_NAND_RESET_CNT 200000
+#endif

Where does 25 ns come from? Should the timeout be in terms of real time rather than iterations (we use get_ticks() for this purpose in fsl_elbc_nand.c)?

-Scott

#2 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Quoted 15 lines · Scott Wood

drivers/mtd/nand/nand_base.c line 87

Quoted 10 lines · Peter Tyser
drivers/mtd/nand/nand_base.c +10 -0
+/*
+ * CONFIG_SYS_NAND_RESET_CNT is used as a timeout mechanism when resetting
+ * a flash.  NAND flash is initialized prior to interrupts so standard timers
+ * can't be used.  CONFIG_SYS_NAND_RESET_CNT should be set to a value
+ * which is greater than (max NAND reset time / NAND status read time).
+ * A conservative default of 200000 (500 us / 25 ns) is used as a default.
+ */
+#ifndef CONFIG_SYS_NAND_RESET_CNT
+#define CONFIG_SYS_NAND_RESET_CNT 200000
+#endif

Where does 25 ns come from? Should the timeout be in terms of real time rather than iterations (we use get_ticks() for this purpose in fsl_elbc_nand.c)?

The 25ns was calculated based on the addition of trp and trhoh from the Micron MT29F8G08 datasheet. Based on the timing diagram for a "Read Status" cycle I thought this would be the minimum cycle time needed to read the chip's status. Other chips (ST, Samsung) I glanced at had > 25 ns read status times as well.

I had tried using get_timer() (I believe nand_wait() would have been perfect to use), but that didn't work due to interrupts being disabled when NAND is probed. I didn't consider using get_ticks()... That seems much better. Is get_ticks() available for all platforms when NAND is initialized?

Assuming get_ticks() is available for all platforms, would you prefer I:
1. re-do the patch using get_ticks()
2. update nand_wait() to use get_ticks instead of get_timer() and use it

Thanks,
Peter

#3 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Quoted 11 lines · Peter Tyser

The 25ns was calculated based on the addition of trp and trhoh from the Micron MT29F8G08 datasheet. Based on the timing diagram for a "Read Status" cycle I thought this would be the minimum cycle time needed to read the chip's status. Other chips (ST, Samsung) I glanced at had > 25 ns read status times as well.

I had tried using get_timer() (I believe nand_wait() would have been perfect to use), but that didn't work due to interrupts being disabled when NAND is probed. I didn't consider using get_ticks()... That seems much better. Is get_ticks() available for all platforms when NAND is initialized?

Probably, but who knows what weirdness is out there.

Quoted 3 lines

Assuming get_ticks() is available for all platforms, would you prefer I:
1. re-do the patch using get_ticks()
2. update nand_wait() to use get_ticks instead of get_timer() and use it

#2 looks better.

-Scott

#4 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Dear Scott Wood,

Quoted 7 lines · In message <[email protected]> you wrote:
Quoted 6 lines · Peter Tyser

The 25ns was calculated based on the addition of trp and trhoh from the Micron MT29F8G08 datasheet. Based on the timing diagram for a "Read Status" cycle I thought this would be the minimum cycle time needed to read the chip's status. Other chips (ST, Samsung) I glanced at had > 25 ns read status times as well.

Quoted 5 lines
Quoted 5 lines

I had tried using get_timer() (I believe nand_wait() would have been perfect to use), but that didn't work due to interrupts being disabled when NAND is probed. I didn't consider using get_ticks()... That seems much better. Is get_ticks() available for all platforms when NAND is initialized?

get_ticks() not a public interface. It should not be used in any common code.

Please use get_timer().

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
You can do this in a number of ways.     IBM chose to do all of them.
Why do you find that funny?        -- D. Taylor, Computer Science 350
#5 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Quoted 16 lines · Wolfgang Denk

Dear Scott Wood,

Quoted 7 lines · In message <[email protected]> you wrote:
Quoted 6 lines · Peter Tyser

The 25ns was calculated based on the addition of trp and trhoh from the Micron MT29F8G08 datasheet. Based on the timing diagram for a "Read Status" cycle I thought this would be the minimum cycle time needed to read the chip's status. Other chips (ST, Samsung) I glanced at had > 25 ns read status times as well.

Quoted 5 lines
Quoted 5 lines

I had tried using get_timer() (I believe nand_wait() would have been perfect to use), but that didn't work due to interrupts being disabled when NAND is probed. I didn't consider using get_ticks()... That seems much better. Is get_ticks() available for all platforms when NAND is initialized?

Quoted 4 lines

get_ticks() not a public interface. It should not be used in any common code.

Please use get_timer().

The problem is that the NAND code is used prior to interrupts being enabled, thus we can't use get_timer(). I used a hokey delay based on (read times * number of iterations). Whats worse, my hokey loop or get_ticks()?

Thanks,
Peter

#6 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Dear Peter Tyser,

Quoted 5 lines · In message <[email protected]> you wrote:
Quoted 5 lines

get_ticks() not a public interface. It should not be used in any common code.

Please use get_timer().

Ah, I see.

Quoted 4 lines

The problem is that the NAND code is used prior to interrupts being enabled, thus we can't use get_timer(). I used a hokey delay based on (read times * number of iterations). Whats worse, my hokey loop or get_ticks()?

I can't tell. Thing is that get_ticks() may or may not exist. I don't think all architectures have it. I don't see it for ixp, leon*, mcf*, microblaze, mips, nios*, ...

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"There is nothing new under the sun, but there are lots of old things
we don't know yet."                                  - Ambrose Bierce
#7 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

In reply to Wolfgang Denk
Quoted 2 lines · Wolfgang Denk

get_ticks() not a public interface. It should not be used in any common code.

Then what is it doing in include/common.h?

Quoted 1 line

Please use get_timer().

But as was pointed out, that depends on interrupts, and thus does not work until very late in the boot process.

-Scott

#8 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Dear Scott,

Quoted 10 lines · In message <[email protected]> you wrote:
Quoted 3 lines

get_ticks() not a public interface. It should not be used in any common code.

Then what is it doing in include/common.h?

Quoted 1 line

Please use get_timer().

But as was pointed out, that depends on interrupts, and thus does not work until very late in the boot process.

I see.

But get_ticks() is an internal interface that may, or may not be preset. There are architectures where it does not exist.

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
If I ever needed a brain transplant, I'd choose a teenager's  because
I'd want a brain that had never been used.
#9 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Hi Scott,

Quoted 11 lines · Wolfgang Denk

Dear Scott,

Quoted 8 lines · In message <[email protected]> you wrote:
Quoted 3 lines

get_ticks() not a public interface. It should not be used in any common code.

Then what is it doing in include/common.h?

Quoted 1 line

Please use get_timer().

Quoted 7 lines
Quoted 2 lines

But as was pointed out, that depends on interrupts, and thus does not work until very late in the boot process.

I see.

But get_ticks() is an internal interface that may, or may not be preset. There are architectures where it does not exist.

Since get_ticks() and get_timer() are not available, is the patch (hokey as it is) acceptable?

Thanks,
Peter

#10 lore raw

PATCH 1/1 nand_base: Add timeout for NAND reset command

Quoted 13 lines · Peter Tyser

Hi Scott,

Quoted 10 lines · Wolfgang Denk

Dear Scott,

Quoted 7 lines · In message <[email protected]> you wrote:
Quoted 2 lines

get_ticks() not a public interface. It should not be used in any common code.

Then what is it doing in include/common.h?

Quoted 1 line

Please use get_timer().

But as was pointed out, that depends on interrupts, and thus does not work until very late in the boot process.

Quoted 7 lines
Quoted 4 lines

I see.

But get_ticks() is an internal interface that may, or may not be preset. There are architectures where it does not exist.

Since get_ticks() and get_timer() are not available, is the patch (hokey as it is) acceptable?

Yes, I'll apply it.

-Scott