[PATCH] mtd: dc21285: fix bytewise memcpy()

[PATCH] mtd: dc21285: fix bytewise memcpy()

[Arnd Bergmann]

From: Arnd Bergmann <arnd@arndb.de>

The commit that split up the 8/16/32-bit operations in 2004 seems to have
broken the 8-bit case, as clang-21 now points out:

drivers/mtd/maps/dc21285.c:129:97: error: parameter 'len' set but not used [-Werror,-Wunused-but-set-parameter]
  129 | static void dc21285_copy_to_8(struct map_info *map, unsigned long to, const void *from, ssize_t len)

Put back the loop that was in linux-2.6.8 and earlier for this case.

Fixes: 67d4878e4e61 ("NOR flash drivers update")
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 drivers/mtd/maps/dc21285.c | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/drivers/mtd/maps/dc21285.c b/drivers/mtd/maps/dc21285.c
index 70a3db3ab856..1d70bf62e91f 100644
--- a/drivers/mtd/maps/dc21285.c
+++ b/drivers/mtd/maps/dc21285.c
@@ -128,12 +128,14 @@ static void dc21285_copy_to_16(struct map_info *map, unsigned long to, const voi
 
 static void dc21285_copy_to_8(struct map_info *map, unsigned long to, const void *from, ssize_t len)
 {
-	map_word d;
-	d.x[0] = *((uint8_t*)from);
-	dc21285_write8(map, d, to);
-	from++;
-	to++;
-	len--;
+	while (len > 0) {
+		map_word d;
+		d.x[0] = *((uint8_t*)from);
+		dc21285_write8(map, d, to);
+		from++;
+		to++;
+		len--;
+	}
 }
 
 static struct map_info dc21285_map = {
-- 
2.39.5

Re: [PATCH] mtd: dc21285: fix bytewise memcpy()

[Miquel Raynal]

Hello Arnd,

On 07/08/2025 at 09:20:34 +02, Arnd Bergmann <arnd@kernel.org> wrote:

> From: Arnd Bergmann <arnd@arndb.de>
>
> The commit that split up the 8/16/32-bit operations in 2004 seems to have
> broken the 8-bit case, as clang-21 now points out:
>
> drivers/mtd/maps/dc21285.c:129:97: error: parameter 'len' set but not used [-Werror,-Wunused-but-set-parameter]
>   129 | static void dc21285_copy_to_8(struct map_info *map, unsigned long to, const void *from, ssize_t len)
>
> Put back the loop that was in linux-2.6.8 and earlier for this case.
>
> Fixes: 67d4878e4e61 ("NOR flash drivers update")
> Cc: David Woodhouse <dwmw2@infradead.org>
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Thanks for the patch, I'll take it, but that probably means few to no
people still use it, if broken for more than 20 years...

Cheers,
Miquèl

Re: [PATCH] mtd: dc21285: fix bytewise memcpy()

[Arnd Bergmann]

On Thu, Aug 7, 2025, at 09:54, Miquel Raynal wrote:
>
> On 07/08/2025 at 09:20:34 +02, Arnd Bergmann <arnd@kernel.org> wrote:
>
>> From: Arnd Bergmann <arnd@arndb.de>
>>
>> The commit that split up the 8/16/32-bit operations in 2004 seems to have
>> broken the 8-bit case, as clang-21 now points out:
>>
>> drivers/mtd/maps/dc21285.c:129:97: error: parameter 'len' set but not used [-Werror,-Wunused-but-set-parameter]
>>   129 | static void dc21285_copy_to_8(struct map_info *map, unsigned long to, const void *from, ssize_t len)
>>
>> Put back the loop that was in linux-2.6.8 and earlier for this case.
>>
>> Fixes: 67d4878e4e61 ("NOR flash drivers update")
>> Cc: David Woodhouse <dwmw2@infradead.org>
>> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
>
> Thanks for the patch, I'll take it, but that probably means few to no
> people still use it, if broken for more than 20 years...

Yes, certainly. I also have a patch to mark the mach-footbridge
platform as deprecated and schedule it for removal, I need to
resend that patch.

What I suspect happened here is that the bug was harmless
because the few boards that use this driver all have
16-bit or 32-bit flashes, while the 8-bit codepath was indeed
unused.

    Arnd