[PATCH] bpf: core: fix shift-out-of-bounds in ___bpf_prog_run

Greg KH gregkh at linuxfoundation.org
Thu May 27 04:29:39 UTC 2021


On Wed, May 26, 2021 at 11:40:27PM +0200, Kurt Manucredo wrote:
> Fix shift-out-of-bounds in ___bpf_prog_run().
> 
> UBSAN: shift-out-of-bounds in kernel/bpf/core.c:1414:2
> shift exponent 248 is too large for 32-bit type 'unsigned int'
> https://syzkaller.appspot.com/bug?id=edb51be4c9a320186328893287bb30d5eed09231
> 
> Reported-by: syzbot+bed360704c521841c85d at syzkaller.appspotmail.com
> Signed-off-by: Kurt Manucredo <fuzzybritches0 at gmail.com>
> ---
>  kernel/bpf/core.c | 31 +++++++++++++++++++++++++++++--
>  1 file changed, 29 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
> index 5e31ee9f7512..35cba247c531 100644
> --- a/kernel/bpf/core.c
> +++ b/kernel/bpf/core.c
> @@ -1411,11 +1411,38 @@ static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn)
>  	ALU(SUB,  -)
>  	ALU(AND,  &)
>  	ALU(OR,   |)
> -	ALU(LSH, <<)
> -	ALU(RSH, >>)
>  	ALU(XOR,  ^)
>  	ALU(MUL,  *)
>  #undef ALU
> +#define ALU(OPCODE, OP)		\
> +	ALU64_##OPCODE##_X:		\
> +		if (SRC >= 64)		\
> +			DST = 0;	\
> +		else			\
> +			DST = DST OP SRC;	\
> +		CONT;			\
> +	ALU_##OPCODE##_X:		\
> +		if (SRC >= 32)		\
> +			DST = 0;	\
> +		else			\
> +			DST = (u32) DST OP (u32) SRC;	\
> +		CONT;			\
> +	ALU64_##OPCODE##_K:		\
> +		if (IMM >= 64)		\
> +			DST = 0;	\
> +		else			\
> +			DST = DST OP IMM;	\
> +		CONT;			\
> +	ALU_##OPCODE##_K:		\
> +		if (IMM >= 32)		\
> +			DST = 0;	\
> +		else			\
> +			DST = (u32) DST OP (u32) IMM;	\
> +		CONT;
> +
> +	ALU(LSH, <<)
> +	ALU(RSH, >>)
> +#undef ALU
>  	ALU_NEG:
>  		DST = (u32) -DST;
>  		CONT;
> -- 
> 2.30.2
> 

This looks odd, shouldn't the caller be checked before it gets to this
part of the code?

did syzbot pass its tests with this change?

thanks,

greg k-h


More information about the Linux-kernel-mentees mailing list