Skip to contentSkip to conversation

mtd: spi-nor: Rework flash parameter initialization

RFC PATCH Applied Michael Walle 5 participants 22 messages Jun 1 – Jul 7

Subsystems (2)
lore
Share & tools
lore

Michael Walle

author maintainer
#0 lore raw

cover mtd: spi-nor: Rework flash parameter initialization

Try to simplify the flash initialization and get rid of the legacy handling. As default, all the flags of the in-kernel database are taken and amended with the SFDP data.

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

See patch 3 for more information. If feedback is positive, this is intended to be applied to the spi-nor tree after the next merge window, so it will sit around in -next for quite some time and get some testing.

That being said, I've just did a quick test on my boards. Please give it a test on your boards.

Michael Walle (3):
mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag
mtd: spi-nor: don't clear the SNOR_F_4B_OPCODES flag on failure
mtd: spi-nor: rework flash parameter initialization

 drivers/mtd/spi-nor/core.c     | 67 +++++++++++-----------------------
 drivers/mtd/spi-nor/core.h     |  1 -
 drivers/mtd/spi-nor/spansion.c |  2 +-
 3 files changed, 23 insertions(+), 47 deletions(-)
Signature · 1 line
2.47.3
author maintainer
#1 lore raw

RFC PATCH 1/3 mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag

That flag was added in [1] to not "issue any non supported commands". But since we now have the generic SFDP driver, we'll issue the RDSFDP command on most flashes anyway and there was no single complaint. Remove the only user of this flag so we can rework the parameter initialization in the core.

[1] https://lore.kernel.org/linux-mtd/[email protected]/

---
 drivers/mtd/spi-nor/spansion.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/mtd/spi-nor/spansion.c +1 -1
index 65227d989de1..092450375d4e 100644
@@ -868,7 +868,7 @@ static const struct flash_info spansion_nor_parts[] = {
 		.name = "s25fl256s0",
 		.size = SZ_32M,
 		.sector_size = SZ_256K,
-		.no_sfdp_flags = SPI_NOR_SKIP_SFDP | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ,
+		.no_sfdp_flags = SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ,
 		.mfr_flags = USE_CLSR,
 	}, {
 		.id = SNOR_ID(0x01, 0x02, 0x19, 0x4d, 0x00, 0x81),
Signature · 1 line
2.47.3
maintainer
#2 lore raw

Re: [RFC PATCH 1/3] mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag

Hi Michael,

drivers/mtd/spi-nor/spansion.c static const struct flash_info spansion_nor_parts[] = { · line 874

Quoted 13 lines
drivers/mtd/spi-nor/spansion.c +1 -1
index 65227d989de1..092450375d4e 100644
@@ -868,7 +868,7 @@ static const struct flash_info spansion_nor_parts[] = {
 		.name = "s25fl256s0",
 		.size = SZ_32M,
 		.sector_size = SZ_256K,
-		.no_sfdp_flags = SPI_NOR_SKIP_SFDP | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ,
+		.no_sfdp_flags = SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ,
 		.mfr_flags = USE_CLSR,
 	}, {
 		.id = SNOR_ID(0x01, 0x02, 0x19, 0x4d, 0x00, 0x81),

Fine by me. I believe the expected behaviour should just be an empty feedback, leading to the SFDP to be invalid and thus SFDP parsing skipped, right?

author maintainer
#3 lore raw

Re: [RFC PATCH 1/3] mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag

Quoted 11 lines · Miquel Raynal

Hi Michael,

Quoted 9 lines
drivers/mtd/spi-nor/spansion.c +0 -1
index 65227d989de1..092450375d4e 100644
@@ -868,7 +868,7 @@ static const struct flash_info spansion_nor_parts[] = {
 		.name = "s25fl256s0",
 		.size = SZ_32M,
 		.sector_size = SZ_256K,
-		.no_sfdp_flags = SPI_NOR_SKIP_SFDP | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ,
Quoted 8 lines

drivers/mtd/spi-nor/spansion.c static const struct flash_info spansion_nor_parts[] = { · line 874

Quoted 4 lines
drivers/mtd/spi-nor/spansion.c +1 -0
+		.no_sfdp_flags = SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ,
 		.mfr_flags = USE_CLSR,
 	}, {
 		.id = SNOR_ID(0x01, 0x02, 0x19, 0x4d, 0x00, 0x81),

Fine by me. I believe the expected behaviour should just be an empty feedback, leading to the SFDP to be invalid and thus SFDP parsing skipped, right?

Right. Probably it will read all ones or all zeros and the parsing will fail and skipped.

-michael

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

reviewer Reviewed-by on v1 1/3
#4 lore raw

RE: [RFC PATCH 1/3] mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag

In reply to Michael Walle
Quoted 9 lines

That flag was added in [1] to not "issue any non supported commands". But since we now have the generic SFDP driver, we'll issue the RDSFDP command on most flashes anyway and there was no single complaint. Remove the only user of this flag so we can rework the parameter initialization in the core.

[1] https://lore.kernel.org/linux-mtd/[email protected]/

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

author maintainer
#6 lore raw

RFC PATCH 2/3 mtd: spi-nor: don't clear the SNOR_F_4B_OPCODES flag on failure

In reply to Michael Walle

This was introduced in commit 548ed6847f530 ("mtd: spi-nor: Add the SNOR_F_4B_OPCODES flag"). It looks like it was the rollback mechanism if parsing the SFDP failed, but was setting that flag first. Nowadays, that flag can only be set if spi_nor_parse_4bait() was successful. IOW, the flags won't be left in an undefined state if parsing SFDP fails.

Remove the unneeded code to be able to rework spi_nor_init_params().

---
 drivers/mtd/spi-nor/core.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
drivers/mtd/spi-nor/core.c +1 -3
index 4046e137531f..fbf8c2d9c6b5 100644
@@ -3066,10 +3066,8 @@ static void spi_nor_sfdp_init_params_deprecated(struct spi_nor *nor)
 
 	memcpy(&sfdp_params, nor->params, sizeof(sfdp_params));
 
-	if (spi_nor_parse_sfdp(nor)) {
+	if (spi_nor_parse_sfdp(nor))
 		memcpy(nor->params, &sfdp_params, sizeof(*nor->params));
-		nor->flags &= ~SNOR_F_4B_OPCODES;
-	}
 }
 
 /**
Signature · 1 line
2.47.3
maintainer Reviewed-by on v1 2/3
#7 lore raw

Re: [RFC PATCH 2/3] mtd: spi-nor: don't clear the SNOR_F_4B_OPCODES flag on failure

Quoted 9 lines · Michael Walle

This was introduced in commit 548ed6847f530 ("mtd: spi-nor: Add the SNOR_F_4B_OPCODES flag"). It looks like it was the rollback mechanism if parsing the SFDP failed, but was setting that flag first. Nowadays, that flag can only be set if spi_nor_parse_4bait() was successful. IOW, the flags won't be left in an undefined state if parsing SFDP fails.

Remove the unneeded code to be able to rework spi_nor_init_params().

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

reviewer Reviewed-by on v1 2/3
#8 lore raw

RE: [RFC PATCH 2/3] mtd: spi-nor: don't clear the SNOR_F_4B_OPCODES flag on failure

In reply to Michael Walle
Quoted 10 lines

This was introduced in commit 548ed6847f530 ("mtd: spi-nor: Add the SNOR_F_4B_OPCODES flag"). It looks like it was the rollback mechanism if parsing the SFDP failed, but was setting that flag first. Nowadays, that flag can only be set if spi_nor_parse_4bait() was successful. IOW, the flags won't be left in an undefined state if parsing SFDP fails.

Remove the unneeded code to be able to rework spi_nor_init_params().

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

author maintainer
#10 lore raw

RFC PATCH 3/3 mtd: spi-nor: rework flash parameter initialization

In reply to Michael Walle

Rework how the flash parameters are initialized. It used to make a difference whether a flash is in the in-kernel database, whether it supports multi I/O and so on. Recently, that makes more and more problems because flash IDs are reused among legacy flashes (w/o SFDP support) and newer flashes [1,2,3,4].

Simplify the whole parameter initialization, by starting with the parameters we have in the in-kernel database and the try to parse SFDP (and let it override any settings).

[1] https://lore.kernel.org/linux-mtd/CAAyq3SYX9UPwhC_Ume_S2yxhQwimRvB=Y6O_+FFqokhNmw7jQg@mail.gmail.com/
[2] https://lore.kernel.org/linux-mtd/[email protected]/
[3] https://lore.kernel.org/linux-mtd/[email protected]/
[4] https://lore.kernel.org/linux-mtd/[email protected]/

---
 drivers/mtd/spi-nor/core.c | 65 +++++++++++++-------------------------
 drivers/mtd/spi-nor/core.h |  1 -
 2 files changed, 22 insertions(+), 44 deletions(-)
drivers/mtd/spi-nor/core.c +22 -43
index fbf8c2d9c6b5..67e0377b606f 100644
@@ -2868,11 +2868,10 @@ static void spi_nor_manufacturer_init_params(struct spi_nor *nor)
 
 /**
  * spi_nor_no_sfdp_init_params() - Initialize the flash's parameters and
- * settings based on nor->info->sfdp_flags. This method should be called only by
- * flashes that do not define SFDP tables. If the flash supports SFDP but the
- * information is wrong and the settings from this function can not be retrieved
- * by parsing SFDP, one should instead use the fixup hooks and update the wrong
- * bits.
+ * settings based on nor->info->sfdp_flags.
+ * If the flash supports SFDP but the information is wrong and the settings from
+ * this function can not be retrieved by parsing SFDP, one should instead use
+ * the fixup hooks and update the wrong bits.
  * @nor:	pointer to a 'struct spi_nor'.
  */
 static void spi_nor_no_sfdp_init_params(struct spi_nor *nor)
@@ -3053,43 +3052,25 @@ static int spi_nor_late_init_params(struct spi_nor *nor)
 }
 
 /**
- * spi_nor_sfdp_init_params_deprecated() - Deprecated way of initializing flash
- * parameters and settings based on JESD216 SFDP standard.
+ * spi_nor_try_parse_sfdp() - Tries to parse flash parameters based
+ * on JESD216 SFDP standard.
  * @nor:	pointer to a 'struct spi_nor'.
  *
  * The method has a roll-back mechanism: in case the SFDP parsing fails, the
- * legacy flash parameters and settings will be restored.
+ * flash parameters and settings will be restored.
  */
-static void spi_nor_sfdp_init_params_deprecated(struct spi_nor *nor)
+static int spi_nor_try_parse_sfdp(struct spi_nor *nor)
 {
 	struct spi_nor_flash_parameter sfdp_params;
+	int ret;
 
 	memcpy(&sfdp_params, nor->params, sizeof(sfdp_params));
 
-	if (spi_nor_parse_sfdp(nor))
+	ret = spi_nor_parse_sfdp(nor);
+	if (ret)
 		memcpy(nor->params, &sfdp_params, sizeof(*nor->params));
-}
-
-/**
- * spi_nor_init_params_deprecated() - Deprecated way of initializing flash
- * parameters and settings.
- * @nor:	pointer to a 'struct spi_nor'.
- *
- * The method assumes that flash doesn't support SFDP so it initializes flash
- * parameters in spi_nor_no_sfdp_init_params() which later on can be overwritten
- * when parsing SFDP, if supported.
- */
-static void spi_nor_init_params_deprecated(struct spi_nor *nor)
-{
-	spi_nor_no_sfdp_init_params(nor);
 
-	spi_nor_manufacturer_init_params(nor);
-
-	if (nor->info->no_sfdp_flags & (SPI_NOR_DUAL_READ |
-					SPI_NOR_QUAD_READ |
-					SPI_NOR_OCTAL_READ |
-					SPI_NOR_OCTAL_DTR_READ))
-		spi_nor_sfdp_init_params_deprecated(nor);
+	return ret;
 }
 
 /**
@@ -3152,7 +3133,8 @@ static void spi_nor_init_default_params(struct spi_nor *nor)
  *
  * 1/ Default flash parameters initialization. The initializations are done
  *    based on nor->info data:
- *		spi_nor_info_init_params()
+ *		spi_nor_init_default_params()
+ *		spi_nor_no_sfdp_init_params()
  *
  * which can be overwritten by:
  * 2/ Manufacturer flash parameters initialization. The initializations are
@@ -3163,7 +3145,7 @@ static void spi_nor_init_default_params(struct spi_nor *nor)
  * which can be overwritten by:
  * 3/ SFDP flash parameters initialization. JESD216 SFDP is a standard and
  *    should be more accurate that the above.
- *		spi_nor_parse_sfdp() or spi_nor_no_sfdp_init_params()
+ *		spi_nor_try_parse_sfdp()
  *
  *    Please note that there is a ->post_bfpt() fixup hook that can overwrite
  *    the flash parameters and settings immediately after parsing the Basic
@@ -3189,17 +3171,14 @@ static int spi_nor_init_params(struct spi_nor *nor)
 		return -ENOMEM;
 
 	spi_nor_init_default_params(nor);
+	spi_nor_no_sfdp_init_params(nor);
+	spi_nor_manufacturer_init_params(nor);
 
-	if (spi_nor_needs_sfdp(nor)) {
-		ret = spi_nor_parse_sfdp(nor);
-		if (ret) {
-			dev_err(nor->dev, "BFPT parsing failed. Please consider using SPI_NOR_SKIP_SFDP when declaring the flash\n");
-			return ret;
-		}
-	} else if (nor->info->no_sfdp_flags & SPI_NOR_SKIP_SFDP) {
-		spi_nor_no_sfdp_init_params(nor);
-	} else {
-		spi_nor_init_params_deprecated(nor);
+	ret = spi_nor_try_parse_sfdp(nor);
+	if (ret && spi_nor_needs_sfdp(nor)) {
+		dev_err(nor->dev,
+			"SFDP parsing failed. You need to manually declare the flash parameters.\n");
+		return ret;
 	}
 
 	ret = spi_nor_late_init_params(nor);
drivers/mtd/spi-nor/core.h +0 -1
index 2ebc5c3caca1..3d7c31bff5cb 100644
@@ -548,7 +548,6 @@ struct flash_info {
 #define SPI_NOR_HAS_CMP			BIT(10)
 
 	u8 no_sfdp_flags;
-#define SPI_NOR_SKIP_SFDP		BIT(0)
 #define SECT_4K				BIT(1)
 #define SPI_NOR_DUAL_READ		BIT(3)
 #define SPI_NOR_QUAD_READ		BIT(4)
Signature · 1 line
2.47.3
Reviewed-by on v1 3/3
#11 lore raw

Re: [RFC PATCH 3/3] mtd: spi-nor: rework flash parameter initialization

Quoted 9 lines · Michael Walle

Rework how the flash parameters are initialized. It used to make a difference whether a flash is in the in-kernel database, whether it supports multi I/O and so on. Recently, that makes more and more problems because flash IDs are reused among legacy flashes (w/o SFDP support) and newer flashes [1,2,3,4].

Simplify the whole parameter initialization, by starting with the parameters we have in the in-kernel database and the try to parse SFDP (and let it override any settings).

You need to specify here as well about the potential problem of issuing an unsupported command (rdsfdp), although practice showed us it's okay.

I can't remember who raised first the potential of issuing unsupported commands. Was it macronix, sst?

Quoted 111 lines

[1] https://lore.kernel.org/linux-mtd/CAAyq3SYX9UPwhC_Ume_S2yxhQwimRvB=Y6O_+FFqokhNmw7jQg@mail.gmail.com/
[2] https://lore.kernel.org/linux-mtd/[email protected]/
[3] https://lore.kernel.org/linux-mtd/[email protected]/
[4] https://lore.kernel.org/linux-mtd/[email protected]/

---
 drivers/mtd/spi-nor/core.c | 65 +++++++++++++-------------------------
 drivers/mtd/spi-nor/core.h |  1 -
 2 files changed, 22 insertions(+), 44 deletions(-)
drivers/mtd/spi-nor/core.c +17 -34
index fbf8c2d9c6b5..67e0377b606f 100644
@@ -2868,11 +2868,10 @@ static void spi_nor_manufacturer_init_params(struct spi_nor *nor)
 
 /**
  * spi_nor_no_sfdp_init_params() - Initialize the flash's parameters and
- * settings based on nor->info->sfdp_flags. This method should be called only by
- * flashes that do not define SFDP tables. If the flash supports SFDP but the
- * information is wrong and the settings from this function can not be retrieved
- * by parsing SFDP, one should instead use the fixup hooks and update the wrong
- * bits.
+ * settings based on nor->info->sfdp_flags.
+ * If the flash supports SFDP but the information is wrong and the settings from
+ * this function can not be retrieved by parsing SFDP, one should instead use
+ * the fixup hooks and update the wrong bits.
  * @nor:	pointer to a 'struct spi_nor'.
  */
 static void spi_nor_no_sfdp_init_params(struct spi_nor *nor)
@@ -3053,43 +3052,25 @@ static int spi_nor_late_init_params(struct spi_nor *nor)
 }
 
 /**
- * spi_nor_sfdp_init_params_deprecated() - Deprecated way of initializing flash
- * parameters and settings based on JESD216 SFDP standard.
+ * spi_nor_try_parse_sfdp() - Tries to parse flash parameters based
+ * on JESD216 SFDP standard.
  * @nor:	pointer to a 'struct spi_nor'.
  *
  * The method has a roll-back mechanism: in case the SFDP parsing fails, the
- * legacy flash parameters and settings will be restored.
+ * flash parameters and settings will be restored.
  */
-static void spi_nor_sfdp_init_params_deprecated(struct spi_nor *nor)
+static int spi_nor_try_parse_sfdp(struct spi_nor *nor)
 {
 	struct spi_nor_flash_parameter sfdp_params;
+	int ret;
 
 	memcpy(&sfdp_params, nor->params, sizeof(sfdp_params));
 
-	if (spi_nor_parse_sfdp(nor))
+	ret = spi_nor_parse_sfdp(nor);
+	if (ret)
 		memcpy(nor->params, &sfdp_params, sizeof(*nor->params));
-}
-
-/**
- * spi_nor_init_params_deprecated() - Deprecated way of initializing flash
- * parameters and settings.
- * @nor:	pointer to a 'struct spi_nor'.
- *
- * The method assumes that flash doesn't support SFDP so it initializes flash
- * parameters in spi_nor_no_sfdp_init_params() which later on can be overwritten
- * when parsing SFDP, if supported.
- */
-static void spi_nor_init_params_deprecated(struct spi_nor *nor)
-{
-	spi_nor_no_sfdp_init_params(nor);
 
-	spi_nor_manufacturer_init_params(nor);
-
-	if (nor->info->no_sfdp_flags & (SPI_NOR_DUAL_READ |
-					SPI_NOR_QUAD_READ |
-					SPI_NOR_OCTAL_READ |
-					SPI_NOR_OCTAL_DTR_READ))
-		spi_nor_sfdp_init_params_deprecated(nor);
+	return ret;
 }
 
 /**
@@ -3152,7 +3133,8 @@ static void spi_nor_init_default_params(struct spi_nor *nor)
  *
  * 1/ Default flash parameters initialization. The initializations are done
  *    based on nor->info data:
- *		spi_nor_info_init_params()
+ *		spi_nor_init_default_params()
+ *		spi_nor_no_sfdp_init_params()
  *
  * which can be overwritten by:
  * 2/ Manufacturer flash parameters initialization. The initializations are
@@ -3163,7 +3145,7 @@ static void spi_nor_init_default_params(struct spi_nor *nor)
  * which can be overwritten by:
  * 3/ SFDP flash parameters initialization. JESD216 SFDP is a standard and
  *    should be more accurate that the above.
- *		spi_nor_parse_sfdp() or spi_nor_no_sfdp_init_params()
+ *		spi_nor_try_parse_sfdp()
  *
  *    Please note that there is a ->post_bfpt() fixup hook that can overwrite
  *    the flash parameters and settings immediately after parsing the Basic
@@ -3189,17 +3171,14 @@ static int spi_nor_init_params(struct spi_nor *nor)
 		return -ENOMEM;
 
 	spi_nor_init_default_params(nor);
+	spi_nor_no_sfdp_init_params(nor);
+	spi_nor_manufacturer_init_params(nor);
 
-	if (spi_nor_needs_sfdp(nor)) {

drivers/mtd/spi-nor/core.c static int spi_nor_init_params(struct spi_nor *nor) · line 3177

Quoted 10 lines
drivers/mtd/spi-nor/core.c +1 -9
-		ret = spi_nor_parse_sfdp(nor);
-		if (ret) {
-			dev_err(nor->dev, "BFPT parsing failed. Please consider using SPI_NOR_SKIP_SFDP when declaring the flash\n");
-			return ret;
-		}
-	} else if (nor->info->no_sfdp_flags & SPI_NOR_SKIP_SFDP) {
-		spi_nor_no_sfdp_init_params(nor);
-	} else {
-		spi_nor_init_params_deprecated(nor);
+	ret = spi_nor_try_parse_sfdp(nor);

I find it fine. I'm concerned about the inconsistent rollback mechanism, but this is a preexisting problem.

Quoted 12 lines
drivers/mtd/spi-nor/core.c +4 -0
+	if (ret && spi_nor_needs_sfdp(nor)) {
+		dev_err(nor->dev,
+			"SFDP parsing failed. You need to manually declare the flash parameters.\n");
+		return ret;
 	}
 
 	ret = spi_nor_late_init_params(nor);
drivers/mtd/spi-nor/core.h +0 -0
index 2ebc5c3caca1..3d7c31bff5cb 100644

drivers/mtd/spi-nor/core.h struct flash_info { · line 551

Quoted 4 lines
drivers/mtd/spi-nor/core.h +0 -1
@@ -548,7 +548,6 @@ struct flash_info {
 #define SPI_NOR_HAS_CMP			BIT(10)
 
 	u8 no_sfdp_flags;
-#define SPI_NOR_SKIP_SFDP		BIT(0)

sashiko says you need to grep this further and remove all instances.

Quoted 3 lines
drivers/mtd/spi-nor/core.h +0 -0
 #define SECT_4K				BIT(1)
 #define SPI_NOR_DUAL_READ		BIT(3)
 #define SPI_NOR_QUAD_READ		BIT(4)
maintainer
#12 lore raw
In reply to Michael Walle

Hi Michael,

Quoted 9 lines · Michael Walle

Try to simplify the flash initialization and get rid of the legacy handling. As default, all the flags of the in-kernel database are taken and amended with the SFDP data.

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

Quoted 7 lines

See patch 3 for more information. If feedback is positive, this is intended to be applied to the spi-nor tree after the next merge window, so it will sit around in -next for quite some time and get some testing.

That being said, I've just did a quick test on my boards. Please give it a test on your boards.

Interesting cleanup, thanks for pushing it. I'll run some tests with some old and newer SFDP based flashes. I do not have chips without SFDP though.

Thanks,
Miquèl

#13 lore raw

Hi,

Quoted 18 lines

Hi Michael,

Quoted 14 lines · Michael Walle

Try to simplify the flash initialization and get rid of the legacy handling. As default, all the flags of the in-kernel database are taken and amended with the SFDP data.

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

See patch 3 for more information. If feedback is positive, this is intended to be applied to the spi-nor tree after the next merge window, so it will sit around in -next for quite some time and get some testing.

Quoted 7 lines
Quoted 2 lines

That being said, I've just did a quick test on my boards. Please give it a test on your boards.

Interesting cleanup, thanks for pushing it. I'll run some tests with some old and newer SFDP based flashes. I do not have chips without SFDP though.

I do have some non-SFDP chips (S25FL-S). Will get back to you with test result.

Thanks,
Takahiro

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

#14 lore raw
Quoted 23 lines

Hi,

Quoted 21 lines

Hi Michael,

Quoted 16 lines · Michael Walle

Try to simplify the flash initialization and get rid of the legacy handling. As default, all the flags of the in-kernel database are taken and amended with the SFDP data.

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

See patch 3 for more information. If feedback is positive, this is intended to be applied to the spi-nor tree after the next merge window, so it will sit around in -next for quite some time and get some testing.

That being said, I've just did a quick test on my boards. Please give it a test on your boards.

Quoted 6 lines
Quoted 4 lines

Interesting cleanup, thanks for pushing it. I'll run some tests with some old and newer SFDP based flashes. I do not have chips without SFDP though.

I do have some non-SFDP chips (S25FL-S). Will get back to you with test result.

With my AMD Zynq-7000 & Infineon internal SPI controller, s25fl128s0, s25fl128s1, and s25fl256s0 work fine.

Here is the test log of s25fl256s0:

zynq> cat /sys/bus/spi/devices/spi0.0/spi-nor/partname s25fl256s0

zynq> cat /sys/bus/spi/devices/spi0.0/spi-nor/jedec_id 0102194d0080

zynq> cat /sys/bus/spi/devices/spi0.0/spi-nor/manufacturer spansion

zynq> xxd -p /sys/bus/spi/devices/spi0.0/spi-nor/sfdp
xxd: /sys/bus/spi/devices/spi0.0/spi-nor/sfdp: No such file or directory
xxd: /sys/bus/spi/devices/spi0.0/spi-nor/sfdp: Bad file descriptor

zynq> md5sum /sys/bus/spi/devices/spi0.0/spi-nor/sfdp
md5sum: can't open '/sys/bus/spi/devices/spi0.0/spi-nor/sfdp': No such file or directory

zynq> test_spi.sh
2+0 records in
2+0 records out
2097152 bytes (2.0MB) copied, 0.074296 seconds, 26.9MB/s Erased 2097152 bytes from address 0x00000000 in flash
Copied 2097152 bytes from address 0x00000000 in flash to spi_read 0000000 ffff ffff ffff ffff ffff ffff ffff ffff
*
0200000 4bda3a28f4ffe603c0ec1258c0034d65a1a0d35ab7bd523a834608adabf03cc5 spi_read Copied 2097152 bytes from spi_test to address 0x00000000 in flash Copied 2097152 bytes from address 0x00000000 in flash to spi_read 1565a9e729e432053d204d569952d5ee6d669e0ace84a88d681b20e4600df356 spi_read 1565a9e729e432053d204d569952d5ee6d669e0ace84a88d681b20e4600df356 spi_test Erased 2097152 bytes from address 0x00000000 in flash
Copied 2097152 bytes from address 0x00000000 in flash to spi_read 4bda3a28f4ffe603c0ec1258c0034d65a1a0d35ab7bd523a834608adabf03cc5 spi_read 1565a9e729e432053d204d569952d5ee6d669e0ace84a88d681b20e4600df356 spi_test

zynq> cat /sys/kernel/debug/spi-nor/spi0.0/params

name            s25fl256s0
id              01 02 19 4d 00 80
size            32.0 MiB
write size      1
page size       256
address nbytes  4
flags           4B_OPCODES | HAS_16BIT_SR

opcodes

 read           0x6c
  dummy cycles  8
 erase          0xdc
 program        0x12
 8D extension   none

protocols

 read           1S-1S-4S
 write          1S-1S-1S
 register       1S-1S-1S

erase commands
d8 (256 KiB) [0]
c7 (32.0 MiB)

sector map
region (in hex) | erase mask | overlaid
------------------+------------+----------
00000000-01ffffff | [0 ] | no

zynq> cat /sys/kernel/debug/spi-nor/spi0.0/capabilities Supported read modes by the flash
1S-1S-1S

  opcode        0x03
  mode cycles   0
  dummy cycles  0
 1S-1S-1S (fast read)
  opcode        0x0b
  mode cycles   0
  dummy cycles  8
 1S-1S-2S
  opcode        0x3b
  mode cycles   0
  dummy cycles  8
 1S-1S-4S
  opcode        0x6b
  mode cycles   0
  dummy cycles  8

Supported page program modes by the flash
1S-1S-1S
opcode 0x02

maintainer Tested-by on the v1 series
#16 lore raw
In reply to Michael Walle

Hello,

Quoted 9 lines

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

See patch 3 for more information. If feedback is positive, this is intended to be applied to the spi-nor tree after the next merge window, so it will sit around in -next for quite some time and get some testing.

Tested with a Winbond W25Q02JVFIM.

If you can also give a try to my QER rework I would be interested to gather feedback on it and ideally also try to merge this kind of rework early in the cycle. ;)

Thanks,
Miquèl

author maintainer
#17 lore raw
Quoted 3 lines · Miquel Raynal

If you can also give a try to my QER rework I would be interested to gather feedback on it and ideally also try to merge this kind of rework early in the cycle. ;)

Unfortunately, I don't have boards with 4 I/O lanes, because we are mostly using it for the bootloader thus they are pretty small and on top of that, we use hardware write protection. The pin for the latter being shared with on of the I/O lines.

-michael

maintainer Applied
#18 lore raw
In reply to Michael Walle
Quoted 10 lines · Michael Walle

Try to simplify the flash initialization and get rid of the legacy handling. As default, all the flags of the in-kernel database are taken and amended with the SFDP data.

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

[...]

Applied to mtd/fixes, thanks!

[1/3] mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag
(no commit info)
[2/3] mtd: spi-nor: don't clear the SNOR_F_4B_OPCODES flag on failure
(no commit info)
[3/3] mtd: spi-nor: rework flash parameter initialization
(no commit info)

Patche(s) should be available on mtd/linux.git and will be part of the next PR (provided that no robot complains by then).

Kind regards,
Miquèl

author maintainer
#19 lore raw
Quoted 21 lines · Miquel Raynal
Quoted 10 lines · Michael Walle

Try to simplify the flash initialization and get rid of the legacy handling. As default, all the flags of the in-kernel database are taken and amended with the SFDP data.

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

[...]

Applied to mtd/fixes, thanks!

[1/3] mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag
(no commit info)
[2/3] mtd: spi-nor: don't clear the SNOR_F_4B_OPCODES flag on failure
(no commit info)
[3/3] mtd: spi-nor: rework flash parameter initialization
(no commit info)

Quoted 2 lines

Patche(s) should be available on mtd/linux.git and will be part of the next PR (provided that no robot complains by then).

Did you see, that there was a v1 of this series? There wasn't many changes for these three patches. Just what Tudor mentioned that the unsupported RDSFDP opcode should be mentioned in the commit message, too.

Apart from that, the series now includes fixes that Sashiko found, that is, the rollback in case of a failed SFDP parsing is broken. I can also rebase that on top of these three patches.

-michael

maintainer
#20 lore raw

Hello,

Quoted 25 lines · Michael Walle
Quoted 21 lines · Miquel Raynal
Quoted 10 lines · Michael Walle

Try to simplify the flash initialization and get rid of the legacy handling. As default, all the flags of the in-kernel database are taken and amended with the SFDP data.

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

[...]

Applied to mtd/fixes, thanks!

[1/3] mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag
(no commit info)
[2/3] mtd: spi-nor: don't clear the SNOR_F_4B_OPCODES flag on failure
(no commit info)
[3/3] mtd: spi-nor: rework flash parameter initialization
(no commit info)

Quoted 2 lines

Patche(s) should be available on mtd/linux.git and will be part of the next PR (provided that no robot complains by then).

Quoted 8 lines

Did you see, that there was a v1 of this series? There wasn't many changes for these three patches. Just what Tudor mentioned that the unsupported RDSFDP opcode should be mentioned in the commit message, too.

Apart from that, the series now includes fixes that Sashiko found, that is, the rollback in case of a failed SFDP parsing is broken. I can also rebase that on top of these three patches.

Sorry, I actually applied it last Friday and pushed only this morning. I will drop the patches and apply v1.

Thanks,
Miquèl

maintainer
#21 lore raw
Quoted 34 lines · Miquel Raynal

Hello,

Quoted 30 lines · Michael Walle
Quoted 21 lines · Miquel Raynal
Quoted 10 lines · Michael Walle

Try to simplify the flash initialization and get rid of the legacy handling. As default, all the flags of the in-kernel database are taken and amended with the SFDP data.

This might have the consequence that all the flashes now get a RDSFPD opcode which might be an unknown opcode. But that was already the case for any flashes which were unknown to the linux kernel. So far, there was not a single complaint.

[...]

Applied to mtd/fixes, thanks!

[1/3] mtd: spi-nor: spansion: s25fl256s0: remove SKIP_SFDP flag
(no commit info)
[2/3] mtd: spi-nor: don't clear the SNOR_F_4B_OPCODES flag on failure
(no commit info)
[3/3] mtd: spi-nor: rework flash parameter initialization
(no commit info)

Quoted 2 lines

Patche(s) should be available on mtd/linux.git and will be part of the next PR (provided that no robot complains by then).

Did you see, that there was a v1 of this series? There wasn't many changes for these three patches. Just what Tudor mentioned that the unsupported RDSFDP opcode should be mentioned in the commit message, too.

Quoted 6 lines
Quoted 3 lines

Apart from that, the series now includes fixes that Sashiko found, that is, the rollback in case of a failed SFDP parsing is broken. I can also rebase that on top of these three patches.

Sorry, I actually applied it last Friday and pushed only this morning. I will drop the patches and apply v1.

Actually, no.

I did apply it to a local branch for testing, and I forgot about b4 picking it up. So these patches never reached linux-mtd (hence the "no commit" lines below each commit).

There are two ways forward:
- We consider the whole series for mtd/fixes (which I will likely send next week) and I take v1.
- We keep the series for spi-nor/next and I let someone else apply it there.

Thanks,
Miquèl