linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* Re: [PATCH] mtd: spi-nor: allow force unlocking via DT property
       [not found] <2026-08-05171214.2934-1-ptpt52@gmail.com>
@ 2026-08-06 11:24 ` Chen Minqiang
  2026-08-06 11:37 ` [PATCH v2] " Chen Minqiang
  1 sibling, 0 replies; 2+ messages in thread
From: Chen Minqiang @ 2026-08-06 11:24 UTC (permalink / raw)
  To: Pratyush Yadav, Michael Walle, Miquel Raynal, Richard Weinberger,
	Vignesh Raghavendra
  Cc: Takahiro Kuwano, linux-mtd, linux-kernel

Hi,

Thank you for your valuable insight!

You are completely right that for unlisted/generic chips that feature a 4-bit
BP layout (BP3 at bit 5 or bit 6) or a CMP (Complement Protect) bit in SR2,
spi_nor_unlock() won't clear those extra bits without the corresponding flags
(SNOR_F_HAS_4BIT_BP / SNOR_F_HAS_SR2_CMP_BIT6) set by the ID database or SFDP.

However, in practice:
1. The vast majority of 3.3V/1.8V generic SPI NOR flashes (e.g. 4MB-16MB chips
   commonly found in vendor devices like Tenda AX12L Pro) use the standard
   3-bit BP (BP0-BP2, SR1 bits 2..4).
2. The current main issue is that even for these standard 3-bit BP chips, the
   kernel currently skips spi_nor_try_unlock_all() completely at boot time if
   CONFIG_MTD_SPI_NOR_SWP_DISABLE_ON_VOLATILE is set (for non-volatile chips)
   or if SNOR_F_HAS_LOCK is not set in chip flags. As a result, status registers
   locked by factory bootloaders are never cleared.

`linux,force-sr-unlock` serves as a pragmatic DT override to force the unlock
attempt at probe time.

To address your point regarding 4-bit BP and CMP bits for unlisted chips, we
have two potential options:

Option A (Current Best-Effort):
Keep the patch as-is, treating `linux,force-sr-unlock` as a best-effort DT
trigger to invoke standard spi_nor_unlock(). It successfully unlocks the vast
majority of standard 3-bit BP generic chips. For rare unlisted chips with 4-bit
BP or CMP bits, explicit entries can still be added to the ID database when
discovered.

Option B (Aggressive Force-Clear):
When `linux,force-sr-unlock` is present in DT, enhance spi_nor_try_unlock_all()
to perform a broader clear operation on SR1 (masking bits 2..6 to clear BP0-BP3/TB)
and SR2 (clearing CMP bit if SR2 is readable).

Which approach would you prefer? I'd be happy to revise the patch based on your
guidance.

Best regards,
Chen Minqiang

^ permalink raw reply	[flat|nested] 2+ messages in thread

* [PATCH v2] mtd: spi-nor: allow force unlocking via DT property
       [not found] <2026-08-05171214.2934-1-ptpt52@gmail.com>
  2026-08-06 11:24 ` [PATCH] mtd: spi-nor: allow force unlocking via DT property Chen Minqiang
@ 2026-08-06 11:37 ` Chen Minqiang
  1 sibling, 0 replies; 2+ messages in thread
From: Chen Minqiang @ 2026-08-06 11:37 UTC (permalink / raw)
  To: Pratyush Yadav, Michael Walle, Miquel Raynal, Richard Weinberger,
	Vignesh Raghavendra
  Cc: Takahiro Kuwano, linux-mtd, linux-kernel, Chen Minqiang

Some SPI NOR flash chips (such as generic or unlisted chips used in vendor
devices like Tenda AX12L Pro) have Block Protection (BP) bits set in the
Status Register by bootloaders or factory settings, locking flash blocks.

Because vendors frequently switch between various generic SPI NOR flash
chips ("Flash Lottery"), it is impractical to upstream explicit chip ID
flags (SNOR_F_HAS_LOCK) for every possible generic chip variant.

This patch introduces support for the "linux,force-sr-unlock" Device Tree
property:
1. In spi_nor_init_flags(), set the SNOR_F_HAS_LOCK flag on the flash
   instance if "linux,force-sr-unlock" is present in the flash DT node. This
   allows spi_nor_late_init_params() to automatically populate default
   locking_ops without altering swp.c.
2. In spi_nor_init(), trigger spi_nor_try_unlock_all() if "linux,force-sr-unlock"
   is set, invoking Linux kernel's native spi_nor_unlock() mechanism.

Signed-off-by: Chen Minqiang <ptpt52@gmail.com>
---
v1 -> v2:
- Set SNOR_F_HAS_LOCK in spi_nor_init_flags() when "linux,force-sr-unlock"
  is present in DT, allowing spi_nor_late_init_params() to set default
  locking_ops automatically without modifying swp.c.
---
 drivers/mtd/spi-nor/core.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/mtd/spi-nor/core.c b/drivers/mtd/spi-nor/core.c
index ccf4396cdcd0..9aa2aaa10c38 100644
--- a/drivers/mtd/spi-nor/core.c
+++ b/drivers/mtd/spi-nor/core.c
@@ -2956,6 +2956,9 @@ static void spi_nor_init_flags(struct spi_nor *nor)
 	if (of_property_read_bool(np, "no-wp"))
 		nor->flags |= SNOR_F_NO_WP;
 
+	if (of_property_read_bool(np, "linux,force-sr-unlock"))
+		nor->flags |= SNOR_F_HAS_LOCK;
+
 	if (flags & SPI_NOR_SWP_IS_VOLATILE)
 		nor->flags |= SNOR_F_SWP_IS_VOLATILE;
 
@@ -3332,7 +3335,8 @@ static int spi_nor_init(struct spi_nor *nor)
 	spi_nor_cache_sr_lock_bits(nor, NULL);
 	if (IS_ENABLED(CONFIG_MTD_SPI_NOR_SWP_DISABLE) ||
 	    (IS_ENABLED(CONFIG_MTD_SPI_NOR_SWP_DISABLE_ON_VOLATILE) &&
-	     nor->flags & SNOR_F_SWP_IS_VOLATILE)) {
+	     nor->flags & SNOR_F_SWP_IS_VOLATILE) ||
+	    of_property_read_bool(spi_nor_get_flash_node(nor), "linux,force-sr-unlock")) {
 		spi_nor_try_unlock_all(nor);
 	}
 
-- 
2.17.1


^ permalink raw reply related	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2026-08-06 11:37 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
     [not found] <2026-08-05171214.2934-1-ptpt52@gmail.com>
2026-08-06 11:24 ` [PATCH] mtd: spi-nor: allow force unlocking via DT property Chen Minqiang
2026-08-06 11:37 ` [PATCH v2] " Chen Minqiang

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).