Re: [PATCH v2 1/3] bpf: Add bpf_verify_pkcs7_signature() helper

[Daniel Borkmann]

On 6/8/22 1:12 PM, Roberto Sassu wrote:
> Add the bpf_verify_pkcs7_signature() helper, to give the ability to eBPF
> security modules to check the validity of a PKCS#7 signature against
> supplied data.
>
> Use the 'keyring' parameter to select the keyring containing the
> verification key: 0 for the primary keyring, 1 for the primary and
> secondary keyrings, 2 for the platform keyring.
>
> Signed-off-by: Roberto Sassu <roberto.sassu@xxxxxxxxxx>
> ---
>   include/uapi/linux/bpf.h       |  8 ++++++++
>   kernel/bpf/bpf_lsm.c           | 32 ++++++++++++++++++++++++++++++++
>   tools/include/uapi/linux/bpf.h |  8 ++++++++
>   3 files changed, 48 insertions(+)
>
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index f4009dbdf62d..40d0fc0d9493 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -5249,6 +5249,13 @@ union bpf_attr {
>    *		Pointer to the underlying dynptr data, NULL if the dynptr is
>    *		read-only, if the dynptr is invalid, or if the offset and length
>    *		is out of bounds.
> + *
> + * long bpf_verify_pkcs7_signature(u8 *data, u32 datalen, u8 *sig, u32 siglen, u64 keyring)
> + *	Description
> + *		Verify the PKCS#7 *sig* with length *siglen*, on *data* with
> + *		length *datalen*, with key in *keyring*.

Could you also add a description for users about the keyring argument and guidance on when
they should use which in their programs? Above is a bit too terse, imho.

> + *	Return
> + *		0 on success, a negative value on error.
>    */
>   #define __BPF_FUNC_MAPPER(FN)		\
>   	FN(unspec),			\
> @@ -5455,6 +5462,7 @@ union bpf_attr {
>   	FN(dynptr_read),		\
>   	FN(dynptr_write),		\
>   	FN(dynptr_data),		\
> +	FN(verify_pkcs7_signature),	\
>   	/* */
>   
>   /* integer value in 'imm' field of BPF_CALL instruction selects which helper
> diff --git a/kernel/bpf/bpf_lsm.c b/kernel/bpf/bpf_lsm.c
> index c1351df9f7ee..1cda43cb541a 100644
> --- a/kernel/bpf/bpf_lsm.c
> +++ b/kernel/bpf/bpf_lsm.c
> @@ -16,6 +16,7 @@
>   #include <linux/bpf_local_storage.h>
>   #include <linux/btf_ids.h>
>   #include <linux/ima.h>
> +#include <linux/verification.h>
>   
>   /* For every LSM hook that allows attachment of BPF programs, declare a nop
>    * function where a BPF program can be attached.
> @@ -132,6 +133,35 @@ static const struct bpf_func_proto bpf_get_attach_cookie_proto = {
>   	.arg1_type	= ARG_PTR_TO_CTX,
>   };
>   
> +BPF_CALL_5(bpf_verify_pkcs7_signature, u8 *, data, u32, datalen, u8 *, sig,
> +	   u32, siglen, u64, keyring)
> +{
> +	int ret = -EOPNOTSUPP;
> +
> +#ifdef CONFIG_SYSTEM_DATA_VERIFICATION
> +	if (keyring > (unsigned long)VERIFY_USE_PLATFORM_KEYRING)
> +		return -EINVAL;
> +
> +	ret = verify_pkcs7_signature(data, datalen, sig, siglen,
> +				     (struct key *)keyring,
> +				     VERIFYING_UNSPECIFIED_SIGNATURE, NULL,
> +				     NULL);
> +#endif
> +	return ret;
> +}

Looks great! One small nit, I would move all of the BPF_CALL and _proto under the
#ifdef CONFIG_SYSTEM_DATA_VERIFICATION ...

> +static const struct bpf_func_proto bpf_verify_pkcs7_signature_proto = {
> +	.func		= bpf_verify_pkcs7_signature,
> +	.gpl_only	= false,
> +	.ret_type	= RET_INTEGER,
> +	.arg1_type	= ARG_PTR_TO_MEM,
> +	.arg2_type	= ARG_CONST_SIZE_OR_ZERO,
> +	.arg3_type	= ARG_PTR_TO_MEM,
> +	.arg4_type	= ARG_CONST_SIZE_OR_ZERO,
> +	.arg5_type	= ARG_ANYTHING,
> +	.allowed	= bpf_ima_inode_hash_allowed,
> +};
> +
>   static const struct bpf_func_proto *
>   bpf_lsm_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
>   {
> @@ -158,6 +188,8 @@ bpf_lsm_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
>   		return prog->aux->sleepable ? &bpf_ima_file_hash_proto : NULL;
>   	case BPF_FUNC_get_attach_cookie:
>   		return bpf_prog_has_trampoline(prog) ? &bpf_get_attach_cookie_proto : NULL;
> +	case BPF_FUNC_verify_pkcs7_signature:
> +		return prog->aux->sleepable ? &bpf_verify_pkcs7_signature_proto : NULL;

... same here:

#ifdef CONFIG_SYSTEM_DATA_VERIFICATION
	case BPF_FUNC_verify_pkcs7_signature:
		return prog->aux->sleepable ? &bpf_verify_pkcs7_signature_proto : NULL;
#endif

So that bpftool or other feature probes can check for its availability. Otherwise, apps have
a hard time checking whether bpf_verify_pkcs7_signature() helper is available for use or not.

>   	default:
>   		return tracing_prog_func_proto(func_id, prog);
>   	}
> diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
> index f4009dbdf62d..40d0fc0d9493 100644
> --- a/tools/include/uapi/linux/bpf.h
> +++ b/tools/include/uapi/linux/bpf.h
> @@ -5249,6 +5249,13 @@ union bpf_attr {
>    *		Pointer to the underlying dynptr data, NULL if the dynptr is
>    *		read-only, if the dynptr is invalid, or if the offset and length
>    *		is out of bounds.
> + *
> + * long bpf_verify_pkcs7_signature(u8 *data, u32 datalen, u8 *sig, u32 siglen, u64 keyring)
> + *	Description
> + *		Verify the PKCS#7 *sig* with length *siglen*, on *data* with
> + *		length *datalen*, with key in *keyring*.
> + *	Return
> + *		0 on success, a negative value on error.
>    */
>   #define __BPF_FUNC_MAPPER(FN)		\
>   	FN(unspec),			\
> @@ -5455,6 +5462,7 @@ union bpf_attr {
>   	FN(dynptr_read),		\
>   	FN(dynptr_write),		\
>   	FN(dynptr_data),		\
> +	FN(verify_pkcs7_signature),	\
>   	/* */
>   
>   /* integer value in 'imm' field of BPF_CALL instruction selects which helper