2022-09-29 02:13:09

by ChiYuan Huang

Subject: [PATCH v2] mfd: mt6360: add bound check in regmap read/write function

From: ChiYuan Huang <[email protected]>

Fix the potential risk for null pointer if bank index is over the maximum.

Refer to the discussion list for the experiment result on mt6370.
https://lore.kernel.org/all/[email protected]/
If not to check the bound, there is the same issue on mt6360.

Fixes: 3b0850440a06c (mfd: mt6360: Merge different sub-devices I2C read/write)
Cc: [email protected]
Signed-off-by: ChiYuan Huang <[email protected]>
---
Since v2:
- Assign i2c bank variable after bank index is already checked.

---
drivers/mfd/mt6360-core.c | 14 ++++++++++++--
1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/drivers/mfd/mt6360-core.c b/drivers/mfd/mt6360-core.c
index 6eaa677..d3b32eb 100644
--- a/drivers/mfd/mt6360-core.c
+++ b/drivers/mfd/mt6360-core.c
@@ -402,7 +402,7 @@ static int mt6360_regmap_read(void *context, const void *reg, size_t reg_size,
struct mt6360_ddata *ddata = context;
u8 bank = *(u8 *)reg;
u8 reg_addr = *(u8 *)(reg + 1);
- struct i2c_client *i2c = ddata->i2c[bank];
+ struct i2c_client *i2c;
bool crc_needed = false;
u8 *buf;
int buf_len = MT6360_ALLOC_READ_SIZE(val_size);
@@ -410,6 +410,11 @@ static int mt6360_regmap_read(void *context, const void *reg, size_t reg_size,
u8 crc;
int ret;

+ if (bank >= MT6360_SLAVE_MAX)
+ return -EINVAL;
+
+ i2c = ddata->i2c[bank];
+
if (bank == MT6360_SLAVE_PMIC || bank == MT6360_SLAVE_LDO) {
crc_needed = true;
ret = mt6360_xlate_pmicldo_addr(&reg_addr, val_size);
@@ -453,13 +458,18 @@ static int mt6360_regmap_write(void *context, const void *val, size_t val_size)
struct mt6360_ddata *ddata = context;
u8 bank = *(u8 *)val;
u8 reg_addr = *(u8 *)(val + 1);
- struct i2c_client *i2c = ddata->i2c[bank];
+ struct i2c_client *i2c;
bool crc_needed = false;
u8 *buf;
int buf_len = MT6360_ALLOC_WRITE_SIZE(val_size);
int write_size = val_size - MT6360_REGMAP_REG_BYTE_SIZE;
int ret;

+ if (bank >= MT6360_SLAVE_MAX)
+ return -EINVAL;
+
+ i2c = ddata->i2c[bank];
+
if (bank == MT6360_SLAVE_PMIC || bank == MT6360_SLAVE_LDO) {
crc_needed = true;
ret = mt6360_xlate_pmicldo_addr(&reg_addr, val_size - MT6360_REGMAP_REG_BYTE_SIZE);
--
2.7.4

2022-10-20 07:56:36

by ChiYuan Huang

Subject: Re: [PATCH v2] mfd: mt6360: add bound check in regmap read/write function

Hi,

cy_huang <[email protected]> 於 2022年9月29日 週四 上午10:00寫道：
>
> From: ChiYuan Huang <[email protected]>
>
> Fix the potential risk for null pointer if bank index is over the maximum.
>
> Refer to the discussion list for the experiment result on mt6370.
> https://lore.kernel.org/all/[email protected]/
> If not to check the bound, there is the same issue on mt6360.
>
> Fixes: 3b0850440a06c (mfd: mt6360: Merge different sub-devices I2C read/write)
> Cc: [email protected]
> Signed-off-by: ChiYuan Huang <[email protected]>
> ---
> Since v2:
> - Assign i2c bank variable after bank index is already checked.
>
> ---
> drivers/mfd/mt6360-core.c | 14 ++++++++++++--
> 1 file changed, 12 insertions(+), 2 deletions(-)
>
Any comment?
> diff --git a/drivers/mfd/mt6360-core.c b/drivers/mfd/mt6360-core.c
> index 6eaa677..d3b32eb 100644
> --- a/drivers/mfd/mt6360-core.c
> +++ b/drivers/mfd/mt6360-core.c
> @@ -402,7 +402,7 @@ static int mt6360_regmap_read(void *context, const void *reg, size_t reg_size,
> struct mt6360_ddata *ddata = context;
> u8 bank = *(u8 *)reg;
> u8 reg_addr = *(u8 *)(reg + 1);
> - struct i2c_client *i2c = ddata->i2c[bank];
> + struct i2c_client *i2c;
> bool crc_needed = false;
> u8 *buf;
> int buf_len = MT6360_ALLOC_READ_SIZE(val_size);
> @@ -410,6 +410,11 @@ static int mt6360_regmap_read(void *context, const void *reg, size_t reg_size,
> u8 crc;
> int ret;
>
> + if (bank >= MT6360_SLAVE_MAX)
> + return -EINVAL;
> +
> + i2c = ddata->i2c[bank];
> +
> if (bank == MT6360_SLAVE_PMIC || bank == MT6360_SLAVE_LDO) {
> crc_needed = true;
> ret = mt6360_xlate_pmicldo_addr(&reg_addr, val_size);
> @@ -453,13 +458,18 @@ static int mt6360_regmap_write(void *context, const void *val, size_t val_size)
> struct mt6360_ddata *ddata = context;
> u8 bank = *(u8 *)val;
> u8 reg_addr = *(u8 *)(val + 1);
> - struct i2c_client *i2c = ddata->i2c[bank];
> + struct i2c_client *i2c;
> bool crc_needed = false;
> u8 *buf;
> int buf_len = MT6360_ALLOC_WRITE_SIZE(val_size);
> int write_size = val_size - MT6360_REGMAP_REG_BYTE_SIZE;
> int ret;
>
> + if (bank >= MT6360_SLAVE_MAX)
> + return -EINVAL;
> +
> + i2c = ddata->i2c[bank];
> +
> if (bank == MT6360_SLAVE_PMIC || bank == MT6360_SLAVE_LDO) {
> crc_needed = true;
> ret = mt6360_xlate_pmicldo_addr(&reg_addr, val_size - MT6360_REGMAP_REG_BYTE_SIZE);
> --
> 2.7.4
>

2022-10-31 14:44:56

by Lee Jones

Subject: Re: [PATCH v2] mfd: mt6360: add bound check in regmap read/write function

On Thu, 29 Sep 2022, cy_huang wrote:

> From: ChiYuan Huang <[email protected]>
>
> Fix the potential risk for null pointer if bank index is over the maximum.
>
> Refer to the discussion list for the experiment result on mt6370.
> https://lore.kernel.org/all/[email protected]/
> If not to check the bound, there is the same issue on mt6360.
>
> Fixes: 3b0850440a06c (mfd: mt6360: Merge different sub-devices I2C read/write)
> Cc: [email protected]
> Signed-off-by: ChiYuan Huang <[email protected]>
> ---
> Since v2:
> - Assign i2c bank variable after bank index is already checked.
>
> ---
> drivers/mfd/mt6360-core.c | 14 ++++++++++++--
> 1 file changed, 12 insertions(+), 2 deletions(-)

Applied, thanks.

--
Lee Jones [李琼斯]