mtd: spi-nor: macronix: Add post_sfdp fixups for Quad Input Page Program
Subsystems (2)
PATCH 1/1 mtd: spi-nor: macronix: Add post_sfdp fixups for Quad Input Page Program
From: Cheng Ming Lin <[email protected]>
commit 798aafeffb369c5eb36e406b18970ef27baa820d upstream.
Although certain Macronix NOR flash support the Quad Input Page Program feature, the corresponding information in the 4-byte Address Instruction Table of these flash is not properly filled. As a result, this feature cannot be enabled as expected.
To address this issue, a post_sfdp fixups implementation is required to correct the missing information.
--- drivers/mtd/spi-nor/macronix.c | 33 ++++++++++++++++++++++++++++----- 1 file changed, 28 insertions(+), 5 deletions(-)
drivers/mtd/spi-nor/macronix.c +28 -5
@@ -28,8 +28,26 @@ mx25l25635_post_bfpt_fixups(struct spi_nor *nor,return 0 ; } +static int +macronix_qpp4b_post_sfdp_fixups (struct spi_nor *nor) +{ +/* PP_1_1_4_4B is supported but missing in 4BAIT. */ +struct spi_nor_flash_parameter *params = nor->params; + + params->hwcaps.mask |=SNOR_HWCAPS_PP_1_1_4 ; +spi_nor_set_pp_settings (¶ms->page_programs[SNOR_CMD_PP_1_1_4 ], +SPINOR_OP_PP_1_1_4_4B ,SNOR_PROTO_1_1_4 ); + +return 0 ; +} +static const struct spi_nor_fixups mx25l25635_fixups = { .post_bfpt = mx25l25635_post_bfpt_fixups, + .post_sfdp = macronix_qpp4b_post_sfdp_fixups, +}; + +static const struct spi_nor_fixups macronix_qpp4b_fixups = { + .post_sfdp = macronix_qpp4b_post_sfdp_fixups, };static const struct flash_info macronix_nor_parts[] = {
@@ -81,7 +99,8 @@ static const struct flash_info macronix_nor_parts[] = { FIXUP_FLAGS(SPI_NOR_4B_OPCODES) }, { "mx25u51245g", INFO(0xc2253a, 0, 64 * 1024, 1024) NO_SFDP_FLAGS(SECT_4K | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ) - FIXUP_FLAGS(SPI_NOR_4B_OPCODES) }, + FIXUP_FLAGS(SPI_NOR_4B_OPCODES) + .fixups = ¯onix_qpp4b_fixups }, { "mx25uw51245g", INFOB(0xc2813a, 0, 0, 0, 4) PARSE_SFDP FLAGS(SPI_NOR_RWW) },
@@ -91,18 +110,22 @@ static const struct flash_info macronix_nor_parts[] = { { "mx25l25655e", INFO(0xc22619, 0, 64 * 1024, 512) }, { "mx66l51235f", INFO(0xc2201a, 0, 64 * 1024, 1024) NO_SFDP_FLAGS(SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ) - FIXUP_FLAGS(SPI_NOR_4B_OPCODES) }, + FIXUP_FLAGS(SPI_NOR_4B_OPCODES) + .fixups = ¯onix_qpp4b_fixups }, { "mx66u51235f", INFO(0xc2253a, 0, 64 * 1024, 1024) NO_SFDP_FLAGS(SECT_4K | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ) - FIXUP_FLAGS(SPI_NOR_4B_OPCODES) }, + FIXUP_FLAGS(SPI_NOR_4B_OPCODES) + .fixups = ¯onix_qpp4b_fixups }, { "mx66l1g45g", INFO(0xc2201b, 0, 64 * 1024, 2048) NO_SFDP_FLAGS(SECT_4K | SPI_NOR_DUAL_READ | - SPI_NOR_QUAD_READ) }, + SPI_NOR_QUAD_READ) + .fixups = ¯onix_qpp4b_fixups }, { "mx66l1g55g", INFO(0xc2261b, 0, 64 * 1024, 2048) NO_SFDP_FLAGS(SPI_NOR_QUAD_READ) }, { "mx66u2g45g", INFO(0xc2253c, 0, 64 * 1024, 4096) NO_SFDP_FLAGS(SECT_4K | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ) - FIXUP_FLAGS(SPI_NOR_4B_OPCODES) }, + FIXUP_FLAGS(SPI_NOR_4B_OPCODES) + .fixups = ¯onix_qpp4b_fixups }, }; static void macronix_nor_default_init(struct spi_nor *nor)
Signature · 1 line
2.25.1
PATCH 1/1 mtd: spi-nor: macronix: add support for mx66{l2, u1}g45g
From: Cheng Ming Lin <[email protected]>
commit 797bbaa7531f75985b199e484451fa3f954382b3 upstream.
Due to incorrect values in the 4-BAIT table for these two flash IDs, it is necessary to add these two flash IDs with fixups.
--- drivers/mtd/spi-nor/macronix.c | 6 ++++++ 1 file changed, 6 insertions(+)
drivers/mtd/spi-nor/macronix.c +6 -0
@@ -116,10 +116,16 @@ static const struct flash_info macronix_nor_parts[] = { NO_SFDP_FLAGS(SECT_4K | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ)FIXUP_FLAGS (SPI_NOR_4B_OPCODES ) .fixups = ¯onix_qpp4b_fixups }, +/* MX66U1G45G */ + { INFO(0xc2253b, 0, 0, 0) + .fixups = ¯onix_qpp4b_fixups }, { "mx66l1g45g", INFO(0xc2201b, 0, 64 * 1024, 2048)NO_SFDP_FLAGS (SECT_4K |SPI_NOR_DUAL_READ |SPI_NOR_QUAD_READ ) .fixups = ¯onix_qpp4b_fixups }, +/* MX66L2G45G */ + { INFO(0xc2201c, 0, 0, 0) + .fixups = ¯onix_qpp4b_fixups }, { "mx66l1g55g", INFO(0xc2261b, 0, 64 * 1024, 2048) NO_SFDP_FLAGS(SPI_NOR_QUAD_READ) }, {"mx66u2g45g" ,INFO (0xc2253c ,0 ,64 *1024 ,4096 )
Signature · 1 line
2.25.1
Sasha Levin
maintainerI can't take this series for 6.6.y: patch 2 adds flash_info entries with a NULL .name, and 6.6's spi_nor_match_name() has no NULL guard (only added upstream in ac5bfa968b60), so the legacy probe-by-name path can oops at boot.
Please send a v2 that either names the new entries or backports ac5bfa968b60 first.
The 6.12.y series is queued, thanks.
Signature · 2 lines
Thanks, Sasha
Hi Sasha,
Sasha Levin <[email protected]> 於 2026年7月2日週四 上午8:38寫道:
Quoted 5 lines
I can't take this series for 6.6.y: patch 2 adds
flash_infoentries with a NULL .name, and 6.6'sspi_nor_match_name()has no NULL guard (only added upstream inac5bfa968b60), so the legacy probe-by-name path can oops at boot.
Thank you for pointing this out and catching the potential issue.
I have verified this, and you are absolutely right. The issue stems from the strcmp(name, manufacturers[i]->parts[j].name) evaluation within the legacy probe path. Since 6.6.y lacks the null guard, passing a NULL .name will result in a null pointer dereference in strcmp() and cause a kernel oops during boot.
I will add the .name to the new flash entries and submit a v2 series.
Quoted 9 lines
Please send a v2 that either names the new entries or backports
ac5bfa968b60first.The 6.12.y series is queued, thanks.
--
Thanks,
Sasha
Thanks,
Cheng Ming Lin
Quoted 11 lines · Cheng Ming Lin
Hi Sasha,
Sasha Levin <[email protected]> 於 2026年7月2日週四 上午8:38寫道:
Quoted 5 lines
I can't take this series for 6.6.y: patch 2 adds
flash_infoentries with a NULL .name, and 6.6'sspi_nor_match_name()has no NULL guard (only added upstream inac5bfa968b60), so the legacy probe-by-name path can oops at boot.Thank you for pointing this out and catching the potential issue.
Quoted 7 lines
I have verified this, and you are absolutely right. The issue stems from the
strcmp(name, manufacturers[i]->parts[j].name)evaluation within the legacy probe path. Since 6.6.y lacks the null guard, passing a NULL .name will result in a null pointer dereference instrcmp()and cause a kernel oops during boot.I will add the .name to the new flash entries and submit a v2 series.
No, please backport the needed patches. The reason is that the name shouldn't become something an application relies on (it is also exposed via sysfs).
For all people not too involved: we are dropping the name for new flash additions, because it is almost always wrong, due to flash id reuse among almost all flash vendors.
-michael
Quoted 13 lines
Quoted 10 lines
Please send a v2 that either names the new entries or backports
ac5bfa968b60first.The 6.12.y series is queued, thanks.
--
Thanks,
SashaThanks,
Cheng Ming Lin
Hi Michael,
Michael Walle <[email protected]> 於 2026年7月2日週四 下午2:44寫道:
Quoted 21 lines · Cheng Ming Lin
Quoted 11 lines
Hi Sasha,
Sasha Levin <[email protected]> 於 2026年7月2日週四 上午8:38寫道:
Quoted 5 lines
I can't take this series for 6.6.y: patch 2 adds
flash_infoentries with a NULL .name, and 6.6'sspi_nor_match_name()has no NULL guard (only added upstream inac5bfa968b60), so the legacy probe-by-name path can oops at boot.Thank you for pointing this out and catching the potential issue.
Quoted 7 lines
I have verified this, and you are absolutely right. The issue stems from the
strcmp(name, manufacturers[i]->parts[j].name)evaluation within the legacy probe path. Since 6.6.y lacks the null guard, passing a NULL .name will result in a null pointer dereference instrcmp()and cause a kernel oops during boot.I will add the .name to the new flash entries and submit a v2 series.
Quoted 7 lines
No, please backport the needed patches. The reason is that the name shouldn't become something an application relies on (it is also exposed via sysfs).
For all people not too involved: we are dropping the name for new flash additions, because it is almost always wrong, due to flash id reuse among almost all flash vendors.
Thank you for the clarification regarding the upstream policy on dropping flash names. That makes perfect sense given the ID reuse issues.
Unfortunately, our emails crossed paths, and I had already submitted the v2 series (which adds the .name back) just before seeing your message. Please disregard the v2 submission.
To follow your guidance, I will prepare a v3 series. My plan is to:
1. Backport commit ac5bfa968b60 ("mtd: spi-nor: fix flash probing") as the first patch in the v3 series to resolve the NULL pointer dereference issue in 6.6.y.
2. Send my flash addition patches (without the .name field, using comments instead) as the subsequent patches in the series.
Does this structure for the v3 series look good to you?
Quoted 17 lines
-michael
Quoted 13 lines
Quoted 10 lines
Please send a v2 that either names the new entries or backports
ac5bfa968b60first.The 6.12.y series is queued, thanks.
--
Thanks,
SashaThanks,
Cheng Ming Lin
Thanks,
Cheng Ming Lin
Hi,
Quoted 11 lines
To follow your guidance, I will prepare a v3 series. My plan is to:
1. Backport commit
ac5bfa968b60("mtd: spi-nor: fix flash probing") as the first patch in the v3 series to resolve the NULL pointer dereference issue in 6.6.y.2. Send my flash addition patches (without the .name field, using comments instead) as the subsequent patches in the series.
Does this structure for the v3 series look good to you?
Sounds good to me.
-michael