Re: [PATCH] i2c: dev: Fix memory leak when underlying adapter does not support I2C

From: Andi Shyti
Date: Thu Nov 07 2024 - 12:14:42 EST


Hi Igor,

that's a good catch, but...

On Thu, Nov 07, 2024 at 01:48:27AM +0000, Igor Pylypiv wrote:
> i2cdev_ioctl_rdwr() receives a buffer which is allocated by the caller.

This needs to be a bit re-written. In the commit log you should
describe what the patch does. You are telling where the buffer is
allocated.

> Fixes: 97ca843f6ad3 ("i2c: dev: Check for I2C_FUNC_I2C before calling i2c_transfer")
> Signed-off-by: Igor Pylypiv <ipylypiv@xxxxxxxxxx>
> ---
> drivers/i2c/i2c-dev.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/i2c/i2c-dev.c b/drivers/i2c/i2c-dev.c
> index 61f7c4003d2f..5d15519ef737 100644
> --- a/drivers/i2c/i2c-dev.c
> +++ b/drivers/i2c/i2c-dev.c
> @@ -247,8 +247,10 @@ static noinline int i2cdev_ioctl_rdwr(struct i2c_client *client,
> int i, res;
>
> /* Adapter must support I2C transfers */
> - if (!i2c_check_functionality(client->adapter, I2C_FUNC_I2C))
> + if (!i2c_check_functionality(client->adapter, I2C_FUNC_I2C)) {
> + kfree(msgs);
> return -EOPNOTSUPP;

Please, don't free it here, free it where it has been allocated,
i.e. in compat_i2cdev_ioctl().

Andi

> + }
>
> data_ptrs = kmalloc_array(nmsgs, sizeof(u8 __user *), GFP_KERNEL);
> if (data_ptrs == NULL) {
> --
> 2.47.0.277.g8800431eea-goog
>