[PATCH v2 1/5] gpu: nova-core: use checked arithmetic in FWSEC firmware parsing

Alexandre Courbot acourbot at nvidia.com
Tue Jan 27 23:58:57 PST 2026


On Tue Jan 27, 2026 at 5:23 AM JST, Joel Fernandes wrote:
> Use checked_add() and checked_mul() when computing offsets from
> firmware-provided values in new_fwsec().
>
> Without checked arithmetic, corrupt firmware could cause integer overflow. The
> danger is not just wrapping to a huge value, but potentially wrapping to a
> small plausible offset that passes validation yet accesses entirely wrong data,
> causing silent corruption or security issues.
>
> Reviewed-by: Zhi Wang <zhiw at nvidia.com>
> Signed-off-by: Joel Fernandes <joelagnelf at nvidia.com>
> ---
>  drivers/gpu/nova-core/firmware/fwsec.rs | 60 ++++++++++++++-----------
>  1 file changed, 35 insertions(+), 25 deletions(-)
>
> diff --git a/drivers/gpu/nova-core/firmware/fwsec.rs b/drivers/gpu/nova-core/firmware/fwsec.rs
> index a8ec08a500ac..71541b1f07d7 100644
> --- a/drivers/gpu/nova-core/firmware/fwsec.rs
> +++ b/drivers/gpu/nova-core/firmware/fwsec.rs
> @@ -46,10 +46,7 @@
>          Signed,
>          Unsigned, //
>      },
> -    num::{
> -        FromSafeCast,
> -        IntoSafeCast, //
> -    },
> +    num::FromSafeCast,
>      vbios::Vbios,
>  };
>  
> @@ -267,7 +264,12 @@ fn new_fwsec(dev: &Device<device::Bound>, bios: &Vbios, cmd: FwsecCommand) -> Re
>          let ucode = bios.fwsec_image().ucode(&desc)?;
>          let mut dma_object = DmaObject::from_data(dev, ucode)?;
>  
> -        let hdr_offset = usize::from_safe_cast(desc.imem_load_size() + desc.interface_offset());
> +        // Compute hdr_offset = imem_load_size + interface_offset.
> +        let hdr_offset = desc
> +            .imem_load_size()
> +            .checked_add(desc.interface_offset())
> +            .map(usize::from_safe_cast)
> +            .ok_or(EINVAL)?;
>          // SAFETY: we have exclusive access to `dma_object`.

Missing empty line before the SAFETY comment (also in other places).

I will fix when applying, no need to resend.



More information about the linux-riscv mailing list