When support for bit offsets of more than one byte was added it
unfortunately left the cell->bytes value at the number of bytes read
including the offset. Make sure to update it to the proper number of
meaningful bytes in the returned data.

Fixes: 7a06ef751077 ("nvmem: core: fix bit offsets of more than one byte")
Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com>
---
 drivers/nvmem/core.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/nvmem/core.c b/drivers/nvmem/core.c
index 387c88c55259..5f6d22545e39 100644
--- a/drivers/nvmem/core.c
+++ b/drivers/nvmem/core.c
@@ -1640,6 +1640,8 @@ static void nvmem_shift_read_buffer_in_place(struct nvmem_cell_entry *cell, void
 	/* clear msb bits if any leftover in the last byte */
 	if (cell->nbits % BITS_PER_BYTE)
 		*p &= GENMASK((cell->nbits % BITS_PER_BYTE) - 1, 0);
+
+	cell->bytes = (p - (u8 *)buf) + 1;
 }
 
 static int __nvmem_cell_read(struct nvmem_device *nvmem,
-- 
2.43.0

On 9/30/25 3:22 PM, Emil Renner Berthing wrote:
> When support for bit offsets of more than one byte was added it
> unfortunately left the cell->bytes value at the number of bytes read
> including the offset. Make sure to update it to the proper number of
> meaningful bytes in the returned data.
Sorry somehow I missed this patch.

If the number of bytes are different to the bytes that are part of cell
then the parsing code seems to have missed some cases, which is the
right place to fix.

Can you share more details on the values that you pass as part of the
cell definition.

> Fixes: 7a06ef751077 ("nvmem: core: fix bit offsets of more than one byte")

missing CC stable

> Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com>
> ---
>  drivers/nvmem/core.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/nvmem/core.c b/drivers/nvmem/core.c
> index 387c88c55259..5f6d22545e39 100644
> --- a/drivers/nvmem/core.c
> +++ b/drivers/nvmem/core.c
> @@ -1640,6 +1640,8 @@ static void nvmem_shift_read_buffer_in_place(struct nvmem_cell_entry *cell, void
>  	/* clear msb bits if any leftover in the last byte */
>  	if (cell->nbits % BITS_PER_BYTE)
>  		*p &= GENMASK((cell->nbits % BITS_PER_BYTE) - 1, 0);
> +
> +	cell->bytes = (p - (u8 *)buf) + 1;
cell bytes can only be updated during parsing time not during every
access time.


--srini>  }
>  
>  static int __nvmem_cell_read(struct nvmem_device *nvmem,

Quoting Srinivas Kandagatla (2025-11-10 09:32:08)
>
>
> On 9/30/25 3:22 PM, Emil Renner Berthing wrote:
> > When support for bit offsets of more than one byte was added it
> > unfortunately left the cell->bytes value at the number of bytes read
> > including the offset. Make sure to update it to the proper number of
> > meaningful bytes in the returned data.
> Sorry somehow I missed this patch.
>
> If the number of bytes are different to the bytes that are part of cell
> then the parsing code seems to have missed some cases, which is the
> right place to fix.
>
> Can you share more details on the values that you pass as part of the
> cell definition.

Yes, the node effectively looks like this:

sid: efuse@3006000 {
        compatible = "allwinner,sun20i-d1-sid";
        reg = <0x3006000 0x1000>;
        #address-cells = <1>;
        #size-cells = <1>;

        ths_calib: ths-calib@14 {
                reg = <0x14 0x4>;
        };

        bg_trim: bg-trim@28 {
                reg = <0x28 0x4>;
                bits = <16 8>;
        };
};

/Emil