Re: [PATCH] staging: vme_user: validate slave window size against buffer size

[Greg KH <gregkh@linuxfoundation.org>]

On Sat, May 09, 2026 at 04:53:18PM +0900, Rion Kiguchi wrote:
> The VME_SET_SLAVE ioctl in drivers/staging/vme_user/vme_user.c accepts
> a user-controlled slave.size and forwards it to vme_slave_set() without
> comparing it against image[minor].size_buf. The slave-image kernel
> buffer is allocated at probe time with a fixed size of PCI_BUF_SIZE
> (0x20000 / 128 KiB), but the configured VME window size can be made
> much larger via the ioctl.
> 
> The subsequent read() / write() handlers (vme_user_read /
> vme_user_write) clamp the I/O range against vme_get_size() (the
> configured window size, attacker-controlled) but never consult
> size_buf. The slave I/O paths buffer_to_user() and buffer_from_user()
> then index image[minor].kern_buf with *ppos values up to
> image_size - 1, well beyond the actual allocation.
> 
> Result: a local user with read/write access to /dev/bus/vme/s* can
> trigger out-of-bounds read and write of the kernel slab adjacent to
> the slave-image buffer.
> 
> Fix: reject slave.size > size_buf in the VME_SET_SLAVE handler. Also
> add defensive bounds checks against size_buf in buffer_to_user() and
> buffer_from_user() so that the I/O paths cannot exceed the
> allocation even if a future ioctl path forgets to validate.
> 
> Cc: stable@vger.kernel.org
> Assisted-by: Claude:claude-opus-4-7
> Signed-off-by: Rion Kiguchi <kiguchi.r.sec@gmail.com>
> ---
>  drivers/staging/vme_user/vme_user.c | 19 ++++++++++++++++++-
>  1 file changed, 18 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/staging/vme_user/vme_user.c b/drivers/staging/vme_user/vme_user.c
> index 11e25c2f6..41b8d5b51 100644
> --- a/drivers/staging/vme_user/vme_user.c
> +++ b/drivers/staging/vme_user/vme_user.c
> @@ -156,6 +156,11 @@ static ssize_t buffer_to_user(unsigned int minor, char __user *buf,
>  {
>  	void *image_ptr;
>  
> +	if (*ppos < 0 || (u64)*ppos >= image[minor].size_buf ||
> +	    count > image[minor].size_buf - (u64)*ppos) {
> +		pr_warn_ratelimited("%s: out-of-bounds access\n", __func__);
> +		return -EINVAL;
> +	}

Why doesn't the check in vme_user_read() already catch this?  You are
duplicating much of the same logic again, are you _SURE_ the
LLM-generated report here is actually correct?

And don't spam the kernel log for when a user sends invalid data, that
would just be a mess.  But if you do want to, use the proper device
information, not just a static function name, which is very generic and
impossible to determine what went wrong (i.e. use the correct logging
functions like dev_err() and the like).

And you need an extra blank line after the check here, your LLM should
know better :)

>  	image_ptr = image[minor].kern_buf + *ppos;
>  	if (copy_to_user(buf, image_ptr, (unsigned long)count))
>  		return -EFAULT;
> @@ -168,6 +173,11 @@ static ssize_t buffer_from_user(unsigned int minor, const char __user *buf,
>  {
>  	void *image_ptr;
>  
> +	if (*ppos < 0 || (u64)*ppos >= image[minor].size_buf ||
> +	    count > image[minor].size_buf - (u64)*ppos) {
> +		pr_warn_ratelimited("%s: out-of-bounds access\n", __func__);
> +		return -EINVAL;
> +	}

Same as above.

thanks,

greg k-h