2023-02-03 07:08:29

by Tudor Ambarus

[permalink] [raw]
Subject: [PATCH v3] mtd: spi-nor: Fix shift-out-of-bounds in spi_nor_set_erase_type

From: Louis Rannou <[email protected]>

spi_nor_set_erase_type() was used either to set or to mask out an erase
type. When we used it to mask out an erase type a shift-out-of-bounds
was hit:
UBSAN: shift-out-of-bounds in drivers/mtd/spi-nor/core.c:2237:24
shift exponent 4294967295 is too large for 32-bit type 'int'

The setting of the size_{shift, mask} and of the opcode are unnecessary
when the erase size is zero, as throughout the code just the erase size
is considered to determine whether an erase type is supported or not.
Setting the opcode to 0xFF was wrong too as nobody guarantees that 0xFF
is an unused opcode. Thus when masking out an erase type, just set the
erase size to zero. This will fix the shift-out-of-bounds.

Fixes: 5390a8df769e ("mtd: spi-nor: add support to non-uniform SFDP SPI NOR flash memories")
Cc: [email protected]
Reported-by: Alexander Stein <[email protected]>
Signed-off-by: Louis Rannou <[email protected]>
[ta: refine changes, new commit message, fix compilation error]
Signed-off-by: Tudor Ambarus <[email protected]>
---
drivers/mtd/spi-nor/core.c | 9 +++++++++
drivers/mtd/spi-nor/core.h | 1 +
drivers/mtd/spi-nor/sfdp.c | 4 ++--
3 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/drivers/mtd/spi-nor/core.c b/drivers/mtd/spi-nor/core.c
index 247d1014879a..22cb18b6c941 100644
--- a/drivers/mtd/spi-nor/core.c
+++ b/drivers/mtd/spi-nor/core.c
@@ -2025,6 +2025,15 @@ void spi_nor_set_erase_type(struct spi_nor_erase_type *erase, u32 size,
erase->size_mask = (1 << erase->size_shift) - 1;
}

+/**
+ * spi_nor_mask_erase_type() - mask out an SPI NOR erase type
+ * @erase: pointer to a structure that describes an SPI NOR erase type
+ */
+void spi_nor_mask_erase_type(struct spi_nor_erase_type *erase)
+{
+ erase->size = 0;
+}
+
/**
* spi_nor_init_uniform_erase_map() - Initialize uniform erase map
* @map: the erase map of the SPI NOR
diff --git a/drivers/mtd/spi-nor/core.h b/drivers/mtd/spi-nor/core.h
index f6d012e1f681..25423225c29d 100644
--- a/drivers/mtd/spi-nor/core.h
+++ b/drivers/mtd/spi-nor/core.h
@@ -681,6 +681,7 @@ void spi_nor_set_pp_settings(struct spi_nor_pp_command *pp, u8 opcode,

void spi_nor_set_erase_type(struct spi_nor_erase_type *erase, u32 size,
u8 opcode);
+void spi_nor_mask_erase_type(struct spi_nor_erase_type *erase);
struct spi_nor_erase_region *
spi_nor_region_next(struct spi_nor_erase_region *region);
void spi_nor_init_uniform_erase_map(struct spi_nor_erase_map *map,
diff --git a/drivers/mtd/spi-nor/sfdp.c b/drivers/mtd/spi-nor/sfdp.c
index fd4daf8fa5df..298ab5e53a8c 100644
--- a/drivers/mtd/spi-nor/sfdp.c
+++ b/drivers/mtd/spi-nor/sfdp.c
@@ -875,7 +875,7 @@ static int spi_nor_init_non_uniform_erase_map(struct spi_nor *nor,
*/
for (i = 0; i < SNOR_ERASE_TYPE_MAX; i++)
if (!(regions_erase_type & BIT(erase[i].idx)))
- spi_nor_set_erase_type(&erase[i], 0, 0xFF);
+ spi_nor_mask_erase_type(&erase[i]);

return 0;
}
@@ -1089,7 +1089,7 @@ static int spi_nor_parse_4bait(struct spi_nor *nor,
erase_type[i].opcode = (dwords[SFDP_DWORD(2)] >>
erase_type[i].idx * 8) & 0xFF;
else
- spi_nor_set_erase_type(&erase_type[i], 0u, 0xFF);
+ spi_nor_mask_erase_type(&erase_type[i]);
}

/*
--
2.34.1



2023-02-03 10:32:22

by Stein, Alexander

[permalink] [raw]
Subject: Re: [PATCH v3] mtd: spi-nor: Fix shift-out-of-bounds in spi_nor_set_erase_type

Hi Tudor,

Am Freitag, dem 03.02.2023 um 09:07 +0200 schrieb Tudor Ambarus:
> From: Louis Rannou <[email protected]>
>
> spi_nor_set_erase_type() was used either to set or to mask out an
> erase
> type. When we used it to mask out an erase type a shift-out-of-bounds
> was hit:
> UBSAN: shift-out-of-bounds in drivers/mtd/spi-nor/core.c:2237:24
> shift exponent 4294967295 is too large for 32-bit type 'int'
>
> The setting of the size_{shift, mask} and of the opcode are
> unnecessary
> when the erase size is zero, as throughout the code just the erase
> size
> is considered to determine whether an erase type is supported or not.
> Setting the opcode to 0xFF was wrong too as nobody guarantees that
> 0xFF
> is an unused opcode. Thus when masking out an erase type, just set
> the
> erase size to zero. This will fix the shift-out-of-bounds.
>
> Fixes: 5390a8df769e ("mtd: spi-nor: add support to non-uniform SFDP
> SPI NOR flash memories")
> Cc: [email protected]
> Reported-by: Alexander Stein <[email protected]>
> Signed-off-by: Louis Rannou <[email protected]>
> [ta: refine changes, new commit message, fix compilation error]
> Signed-off-by: Tudor Ambarus <[email protected]>

Thanks.
Tested-by: Alexander Stein <[email protected]>

> ---
>  drivers/mtd/spi-nor/core.c | 9 +++++++++
>  drivers/mtd/spi-nor/core.h | 1 +
>  drivers/mtd/spi-nor/sfdp.c | 4 ++--
>  3 files changed, 12 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/mtd/spi-nor/core.c b/drivers/mtd/spi-nor/core.c
> index 247d1014879a..22cb18b6c941 100644
> --- a/drivers/mtd/spi-nor/core.c
> +++ b/drivers/mtd/spi-nor/core.c
> @@ -2025,6 +2025,15 @@ void spi_nor_set_erase_type(struct
> spi_nor_erase_type *erase, u32 size,
>         erase->size_mask = (1 << erase->size_shift) - 1;
>  }
>  
> +/**
> + * spi_nor_mask_erase_type() - mask out an SPI NOR erase type
> + * @erase:     pointer to a structure that describes an SPI NOR
> erase type
> + */
> +void spi_nor_mask_erase_type(struct spi_nor_erase_type *erase)
> +{
> +       erase->size = 0;
> +}
> +
>  /**
>   * spi_nor_init_uniform_erase_map() - Initialize uniform erase map
>   * @map:               the erase map of the SPI NOR
> diff --git a/drivers/mtd/spi-nor/core.h b/drivers/mtd/spi-nor/core.h
> index f6d012e1f681..25423225c29d 100644
> --- a/drivers/mtd/spi-nor/core.h
> +++ b/drivers/mtd/spi-nor/core.h
> @@ -681,6 +681,7 @@ void spi_nor_set_pp_settings(struct
> spi_nor_pp_command *pp, u8 opcode,
>  
>  void spi_nor_set_erase_type(struct spi_nor_erase_type *erase, u32
> size,
>                             u8 opcode);
> +void spi_nor_mask_erase_type(struct spi_nor_erase_type *erase);
>  struct spi_nor_erase_region *
>  spi_nor_region_next(struct spi_nor_erase_region *region);
>  void spi_nor_init_uniform_erase_map(struct spi_nor_erase_map *map,
> diff --git a/drivers/mtd/spi-nor/sfdp.c b/drivers/mtd/spi-nor/sfdp.c
> index fd4daf8fa5df..298ab5e53a8c 100644
> --- a/drivers/mtd/spi-nor/sfdp.c
> +++ b/drivers/mtd/spi-nor/sfdp.c
> @@ -875,7 +875,7 @@ static int
> spi_nor_init_non_uniform_erase_map(struct spi_nor *nor,
>          */
>         for (i = 0; i < SNOR_ERASE_TYPE_MAX; i++)
>                 if (!(regions_erase_type & BIT(erase[i].idx)))
> -                       spi_nor_set_erase_type(&erase[i], 0, 0xFF);
> +                       spi_nor_mask_erase_type(&erase[i]);
>  
>         return 0;
>  }
> @@ -1089,7 +1089,7 @@ static int spi_nor_parse_4bait(struct spi_nor
> *nor,
>                         erase_type[i].opcode = (dwords[SFDP_DWORD(2)]
> >>
>                                                 erase_type[i].idx *
> 8) & 0xFF;
>                 else
> -                       spi_nor_set_erase_type(&erase_type[i], 0u,
> 0xFF);
> +                       spi_nor_mask_erase_type(&erase_type[i]);
>         }
>  
>         /*

2023-02-06 08:43:53

by Tudor Ambarus

[permalink] [raw]
Subject: Re: [PATCH v3] mtd: spi-nor: Fix shift-out-of-bounds in spi_nor_set_erase_type

On Fri, 03 Feb 2023 09:07:54 +0200, Tudor Ambarus wrote:
> spi_nor_set_erase_type() was used either to set or to mask out an erase
> type. When we used it to mask out an erase type a shift-out-of-bounds
> was hit:
> UBSAN: shift-out-of-bounds in drivers/mtd/spi-nor/core.c:2237:24
> shift exponent 4294967295 is too large for 32-bit type 'int'
>
> The setting of the size_{shift, mask} and of the opcode are unnecessary
> when the erase size is zero, as throughout the code just the erase size
> is considered to determine whether an erase type is supported or not.
> Setting the opcode to 0xFF was wrong too as nobody guarantees that 0xFF
> is an unused opcode. Thus when masking out an erase type, just set the
> erase size to zero. This will fix the shift-out-of-bounds.
>
> [...]

Applied to spi-nor/next, thanks!

[1/1] mtd: spi-nor: Fix shift-out-of-bounds in spi_nor_set_erase_type
https://git.kernel.org/mtd/c/f0f0cfdc3a02

Best regards,
--
Tudor Ambarus <[email protected]>