[PATCH next] nvdimm: Prevent integer overflow in ramdax_get_config_data()

[PATCH next] nvdimm: Prevent integer overflow in ramdax_get_config_data()

[Dan Carpenter]

The "cmd->in_offset" variable comes from the user via the __nd_ioctl()
function.  The problem is that the "cmd->in_offset + cmd->in_length"
addition could have an integer wrapping issue if cmd->in_offset is close
to UINT_MAX .  The "cmd->in_length" variable has been capped, but the
"cmd->in_offset" variable has not.  Both of these variables are type u32.

Fixes: 43bc0aa19a21 ("nvdimm: allow exposing RAM carveouts as NVDIMM DIMM devices")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
---
 drivers/nvdimm/ramdax.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/nvdimm/ramdax.c b/drivers/nvdimm/ramdax.c
index 63cf05791829..faa6f3101972 100644
--- a/drivers/nvdimm/ramdax.c
+++ b/drivers/nvdimm/ramdax.c
@@ -143,7 +143,7 @@ static int ramdax_get_config_data(struct nvdimm *nvdimm, int buf_len,
 		return -EINVAL;
 	if (struct_size(cmd, out_buf, cmd->in_length) > buf_len)
 		return -EINVAL;
-	if (cmd->in_offset + cmd->in_length > LABEL_AREA_SIZE)
+	if (size_add(cmd->in_offset, cmd->in_length) > LABEL_AREA_SIZE)
 		return -EINVAL;
 
 	memcpy(cmd->out_buf, dimm->label_area + cmd->in_offset, cmd->in_length);
-- 
2.51.0

Re: [PATCH next] nvdimm: Prevent integer overflow in ramdax_get_config_data()

[Ira Weiny]

Dan Carpenter wrote:
> The "cmd->in_offset" variable comes from the user via the __nd_ioctl()
> function.  The problem is that the "cmd->in_offset + cmd->in_length"
> addition could have an integer wrapping issue if cmd->in_offset is close
> to UINT_MAX .  The "cmd->in_length" variable has been capped, but the
> "cmd->in_offset" variable has not.  Both of these variables are type u32.

Does ramdax_set_config_data() also need this?  I'm not quite following
where in_length is capped so I'm inclined to add size_add in both set and
get.

Ira

> 
> Fixes: 43bc0aa19a21 ("nvdimm: allow exposing RAM carveouts as NVDIMM DIMM devices")
> Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
> ---
>  drivers/nvdimm/ramdax.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/nvdimm/ramdax.c b/drivers/nvdimm/ramdax.c
> index 63cf05791829..faa6f3101972 100644
> --- a/drivers/nvdimm/ramdax.c
> +++ b/drivers/nvdimm/ramdax.c
> @@ -143,7 +143,7 @@ static int ramdax_get_config_data(struct nvdimm *nvdimm, int buf_len,
>  		return -EINVAL;
>  	if (struct_size(cmd, out_buf, cmd->in_length) > buf_len)
>  		return -EINVAL;
> -	if (cmd->in_offset + cmd->in_length > LABEL_AREA_SIZE)
> +	if (size_add(cmd->in_offset, cmd->in_length) > LABEL_AREA_SIZE)
>  		return -EINVAL;
>  
>  	memcpy(cmd->out_buf, dimm->label_area + cmd->in_offset, cmd->in_length);
> -- 
> 2.51.0
> 

Re: [PATCH next] nvdimm: Prevent integer overflow in ramdax_get_config_data()

[Dan Carpenter]

On Tue, Nov 25, 2025 at 10:38:29AM -0600, Ira Weiny wrote:
> Dan Carpenter wrote:
> > The "cmd->in_offset" variable comes from the user via the __nd_ioctl()
> > function.  The problem is that the "cmd->in_offset + cmd->in_length"
> > addition could have an integer wrapping issue if cmd->in_offset is close
> > to UINT_MAX .  The "cmd->in_length" variable has been capped, but the
> > "cmd->in_offset" variable has not.  Both of these variables are type u32.
> 
> Does ramdax_set_config_data() also need this?

Yes.  It does.  These are from Smatch warnings, right.  They take a few
rebuilds for the taint information to propagate from the ioctl to the
ramdax_get_config_data() function.  When I rebuilt it, then it propagates
to both so I would have seen the ramdax_set_config_data() tomorrow.

But they're called from the same function so the taint data should
have propagated to both at the same time...  WTF?  I don't know what
happened.  Anyway, I will fix that and resend.

Thanks for noticing.

>  I'm not quite following where in_length is capped so I'm inclined to
> add size_add in both set and get.

I meant that the  if (struct_size(cmd, in_buf, cmd->in_length) > buf_len)
line checks that cmd->in_length is okay.

regards,
dan carpenter