[PATCH net-next] net: relax SO_TXTIME CAP_NET_ADMIN check

[PATCH net-next] net: relax SO_TXTIME CAP_NET_ADMIN check

[Eric Dumazet]

Now sch_fq has horizon feature, we want to allow QUIC/UDP applications
to use EDT model so that pacing can be offloaded to the kernel (sch_fq)
or the NIC.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
---
 net/core/sock.c | 28 ++++++++++++++++++----------
 1 file changed, 18 insertions(+), 10 deletions(-)

diff --git a/net/core/sock.c b/net/core/sock.c
index b714162213aeae98bfee24d8b457547fe7abab4f..fd85e651ce284b6987f0e8fae94f76ec2c432899 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -1152,23 +1152,31 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_TXTIME:
-		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
-			ret = -EPERM;
-		} else if (optlen != sizeof(struct sock_txtime)) {
+		if (optlen != sizeof(struct sock_txtime)) {
 			ret = -EINVAL;
+			break;
 		} else if (copy_from_user(&sk_txtime, optval,
 			   sizeof(struct sock_txtime))) {
 			ret = -EFAULT;
+			break;
 		} else if (sk_txtime.flags & ~SOF_TXTIME_FLAGS_MASK) {
 			ret = -EINVAL;
-		} else {
-			sock_valbool_flag(sk, SOCK_TXTIME, true);
-			sk->sk_clockid = sk_txtime.clockid;
-			sk->sk_txtime_deadline_mode =
-				!!(sk_txtime.flags & SOF_TXTIME_DEADLINE_MODE);
-			sk->sk_txtime_report_errors =
-				!!(sk_txtime.flags & SOF_TXTIME_REPORT_ERRORS);
+			break;
 		}
+		/* CLOCK_MONOTONIC is only used by sch_fq, and this packet
+		 * scheduler has enough safe guards.
+		 */
+		if (sk_txtime.clockid != CLOCK_MONOTONIC &&
+		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
+			ret = -EPERM;
+			break;
+		}
+		sock_valbool_flag(sk, SOCK_TXTIME, true);
+		sk->sk_clockid = sk_txtime.clockid;
+		sk->sk_txtime_deadline_mode =
+			!!(sk_txtime.flags & SOF_TXTIME_DEADLINE_MODE);
+		sk->sk_txtime_report_errors =
+			!!(sk_txtime.flags & SOF_TXTIME_REPORT_ERRORS);
 		break;
 
 	case SO_BINDTOIFINDEX:
-- 
2.26.2.526.g744177e7f7-goog

Re: [PATCH net-next] net: relax SO_TXTIME CAP_NET_ADMIN check

[Willem de Bruijn]

On Thu, May 7, 2020 at 1:05 PM Eric Dumazet <edumazet@google.com> wrote:
>
> Now sch_fq has horizon feature, we want to allow QUIC/UDP applications
> to use EDT model so that pacing can be offloaded to the kernel (sch_fq)
> or the NIC.
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Willem de Bruijn <willemb@google.com>

Acked-by: Willem de Bruijn <willemb@google.com>

Thanks Eric! This will really help with enabling pacing offload in
real workloads.

Re: [PATCH net-next] net: relax SO_TXTIME CAP_NET_ADMIN check

[Vinicius Costa Gomes]

Hi,

Eric Dumazet <edumazet@google.com> writes:

> Now sch_fq has horizon feature, we want to allow QUIC/UDP applications
> to use EDT model so that pacing can be offloaded to the kernel (sch_fq)
> or the NIC.
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Willem de Bruijn <willemb@google.com>
> ---
>  net/core/sock.c | 28 ++++++++++++++++++----------
>  1 file changed, 18 insertions(+), 10 deletions(-)
>
> diff --git a/net/core/sock.c b/net/core/sock.c
> index b714162213aeae98bfee24d8b457547fe7abab4f..fd85e651ce284b6987f0e8fae94f76ec2c432899 100644
> --- a/net/core/sock.c
> +++ b/net/core/sock.c
> @@ -1152,23 +1152,31 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
>  		break;
>  
>  	case SO_TXTIME:
> -		if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
> -			ret = -EPERM;
> -		} else if (optlen != sizeof(struct sock_txtime)) {
> +		if (optlen != sizeof(struct sock_txtime)) {
>  			ret = -EINVAL;
> +			break;
>  		} else if (copy_from_user(&sk_txtime, optval,
>  			   sizeof(struct sock_txtime))) {
>  			ret = -EFAULT;
> +			break;
>  		} else if (sk_txtime.flags & ~SOF_TXTIME_FLAGS_MASK) {
>  			ret = -EINVAL;
> -		} else {
> -			sock_valbool_flag(sk, SOCK_TXTIME, true);
> -			sk->sk_clockid = sk_txtime.clockid;
> -			sk->sk_txtime_deadline_mode =
> -				!!(sk_txtime.flags & SOF_TXTIME_DEADLINE_MODE);
> -			sk->sk_txtime_report_errors =
> -				!!(sk_txtime.flags & SOF_TXTIME_REPORT_ERRORS);
> +			break;
>  		}
> +		/* CLOCK_MONOTONIC is only used by sch_fq, and this packet
> +		 * scheduler has enough safe guards.
> +		 */
> +		if (sk_txtime.clockid != CLOCK_MONOTONIC &&
> +		    !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
> +			ret = -EPERM;
> +			break;
> +		}

I was a bit worried until I saw the check above.

Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>


> +		sock_valbool_flag(sk, SOCK_TXTIME, true);
> +		sk->sk_clockid = sk_txtime.clockid;
> +		sk->sk_txtime_deadline_mode =
> +			!!(sk_txtime.flags & SOF_TXTIME_DEADLINE_MODE);
> +		sk->sk_txtime_report_errors =
> +			!!(sk_txtime.flags & SOF_TXTIME_REPORT_ERRORS);
>  		break;
>  
>  	case SO_BINDTOIFINDEX:
> -- 
> 2.26.2.526.g744177e7f7-goog
>

-- 
Vinicius

Re: [PATCH net-next] net: relax SO_TXTIME CAP_NET_ADMIN check

[David Miller]

From: Eric Dumazet <edumazet@google.com>
Date: Thu,  7 May 2020 10:05:39 -0700

> Now sch_fq has horizon feature, we want to allow QUIC/UDP applications
> to use EDT model so that pacing can be offloaded to the kernel (sch_fq)
> or the NIC.
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Willem de Bruijn <willemb@google.com>

Applied, thanks Eric.