[PATCH] drivers: of: fix possible overflow

[PATCH] drivers: of: fix possible overflow

[Abdelrahman Youssef]

while parsing the fdt header, the name of the begining node marked by
FDT_BEGIN_NODE that is part of the struct block moves out of the block
that results in heap-overflow.

So this patch checks if the length of name (maxlen) + the offset of
the struct block exceeds the size of the whole block.

Signed-off-by: Abdelrahman Youssef <abdelrahmanyossef12@gmail.com>
---
 drivers/of/fdt.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
index 2c3ea31394..7dc8ee2529 100644
--- a/drivers/of/fdt.c
+++ b/drivers/of/fdt.c
@@ -210,6 +210,11 @@ static struct device_node *__of_unflatten_dtb(const void *infdt, int size,
 			maxlen = (unsigned long)fdt + f.off_dt_struct +
 				f.size_dt_struct - (unsigned long)name;
 
+			if (maxlen + f.off_dt_struct > f.size_dt_struct) {
+				ret = -ESPIPE;
+				goto err;
+			}
+
 			len = strnlen(name, maxlen + 1);
 			if (len > maxlen) {
 				ret = -ESPIPE;
-- 
2.43.0

Re: [PATCH] drivers: of: fix possible overflow

[Ahmad Fatoum]

Hello Abdelrahman,

Thanks for your patch.

On 12.11.24 13:54, Abdelrahman Youssef wrote:
> while parsing the fdt header, the name of the begining node marked by
> FDT_BEGIN_NODE that is part of the struct block moves out of the block
> that results in heap-overflow.
> 
> So this patch checks if the length of name (maxlen) + the offset of
> the struct block exceeds the size of the whole block.
> 
> Signed-off-by: Abdelrahman Youssef <abdelrahmanyossef12@gmail.com>
> ---
>  drivers/of/fdt.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
> index 2c3ea31394..7dc8ee2529 100644
> --- a/drivers/of/fdt.c
> +++ b/drivers/of/fdt.c
> @@ -210,6 +210,11 @@ static struct device_node *__of_unflatten_dtb(const void *infdt, int size,
>  			maxlen = (unsigned long)fdt + f.off_dt_struct +
>  				f.size_dt_struct - (unsigned long)name;
>  
> +			if (maxlen + f.off_dt_struct > f.size_dt_struct) {

This looks wrong:

 - f.off_dt_struct is the offset in bytes of the struct area relative to
   the start of the FDT

 - f.size_dt_struct is the size of the struct area, i.e. the first byte
   after the struct area and it does _not_ include the offset.

It doesn't make sense to subtract the size from the offset.

I have looked into it a little deeper now and I believe the root cause is
an integer overflow; if name _starts_ after the end of the struct area,
the computation for maxlen will result in a negative number and that
negative number would wrap around to a very big positive number.

We should change maxlen to be a (signed) int and return an error if
it becomes negative. Can you verify that fixes the issue and send a patch?

Cheers,
Ahmad

> +				ret = -ESPIPE;
> +				goto err;
> +			}
> +
>  			len = strnlen(name, maxlen + 1);
>  			if (len > maxlen) {
>  				ret = -ESPIPE;


-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |