diff mbox series

[net-next,v3,1/3] udp: msg_zerocopy

Message ID 20181129202620.233237-2-willemdebruijn.kernel@gmail.com
State Changes Requested, archived
Delegated to: David Miller
Headers show
Series udp msg_zerocopy | expand

Commit Message

Willem de Bruijn Nov. 29, 2018, 8:26 p.m. UTC 
From: Willem de Bruijn <willemb@google.com>

Extend zerocopy to udp sockets. Allow setting sockopt SO_ZEROCOPY and
interpret flag MSG_ZEROCOPY.

This patch was previously part of the zerocopy RFC patchsets. Zerocopy
is not effective at small MTU. With segmentation offload building
larger datagrams, the benefit of page flipping outweights the cost of
generating a completion notification.

tools/testing/selftests/net/msg_zerocopy.sh after applying follow-on
test patch and making skb_orphan_frags_rx same as skb_orphan_frags:

    ipv4 udp -t 1
    tx=191312 (11938 MB) txc=0 zc=n
    rx=191312 (11938 MB)
    ipv4 udp -z -t 1
    tx=304507 (19002 MB) txc=304507 zc=y
    rx=304507 (19002 MB)
    ok
    ipv6 udp -t 1
    tx=174485 (10888 MB) txc=0 zc=n
    rx=174485 (10888 MB)
    ipv6 udp -z -t 1
    tx=294801 (18396 MB) txc=294801 zc=y
    rx=294801 (18396 MB)
    ok

Changes
  v1 -> v2
    - Fixup reverse christmas tree violation
  v2 -> v3
    - Split refcount avoidance optimization into separate patch
      - Fix refcount leak on error in fragmented case
        (thanks to Paolo Abeni for pointing this one out!)
      - Fix refcount inc on zero

Signed-off-by: Willem de Bruijn <willemb@google.com>
---
 include/linux/skbuff.h |  1 +
 net/core/skbuff.c      |  6 ++++++
 net/core/sock.c        |  5 ++++-
 net/ipv4/ip_output.c   | 23 ++++++++++++++++++++++-
 net/ipv6/ip6_output.c  | 23 ++++++++++++++++++++++-
 5 files changed, 55 insertions(+), 3 deletions(-)

Comments

Willem de Bruijn Nov. 29, 2018, 8:29 p.m. UTC | #1 
On Thu, Nov 29, 2018 at 3:26 PM Willem de Bruijn
<willemdebruijn.kernel@gmail.com> wrote:
>
> From: Willem de Bruijn <willemb@google.com>
>
> Extend zerocopy to udp sockets. Allow setting sockopt SO_ZEROCOPY and
> interpret flag MSG_ZEROCOPY.
>
> This patch was previously part of the zerocopy RFC patchsets. Zerocopy
> is not effective at small MTU. With segmentation offload building
> larger datagrams, the benefit of page flipping outweights the cost of
> generating a completion notification.
>
> tools/testing/selftests/net/msg_zerocopy.sh after applying follow-on
> test patch and making skb_orphan_frags_rx same as skb_orphan_frags:
>
>     ipv4 udp -t 1
>     tx=191312 (11938 MB) txc=0 zc=n
>     rx=191312 (11938 MB)
>     ipv4 udp -z -t 1
>     tx=304507 (19002 MB) txc=304507 zc=y
>     rx=304507 (19002 MB)
>     ok
>     ipv6 udp -t 1
>     tx=174485 (10888 MB) txc=0 zc=n
>     rx=174485 (10888 MB)
>     ipv6 udp -z -t 1
>     tx=294801 (18396 MB) txc=294801 zc=y
>     rx=294801 (18396 MB)
>     ok
>
> Changes
>   v1 -> v2
>     - Fixup reverse christmas tree violation
>   v2 -> v3
>     - Split refcount avoidance optimization into separate patch
>       - Fix refcount leak on error in fragmented case
>         (thanks to Paolo Abeni for pointing this one out!)
>       - Fix refcount inc on zero

one additional change that I forgot to add to the changelog

  - test sock_flag SOCK_ZEROCOPY directly in __ip_append_data.
    this is needed since commit 5cf4a8532c99 ("tcp: really ignore
    MSG_ZEROCOPY if no SO_ZEROCOPY") did the same for tcp.
diff mbox series

Patch

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 73902acf2b71..04f52e719571 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -485,6 +485,7 @@  void sock_zerocopy_put_abort(struct ubuf_info *uarg);
 
 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
 
+int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
 			     struct msghdr *msg, int len,
 			     struct ubuf_info *uarg);
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 02cd7ae3d0fb..435bac91d293 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -1105,6 +1105,12 @@  EXPORT_SYMBOL_GPL(sock_zerocopy_put_abort);
 extern int __zerocopy_sg_from_iter(struct sock *sk, struct sk_buff *skb,
 				   struct iov_iter *from, size_t length);
 
+int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len)
+{
+	return __zerocopy_sg_from_iter(skb->sk, skb, &msg->msg_iter, len);
+}
+EXPORT_SYMBOL_GPL(skb_zerocopy_iter_dgram);
+
 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
 			     struct msghdr *msg, int len,
 			     struct ubuf_info *uarg)
diff --git a/net/core/sock.c b/net/core/sock.c
index 6d7e189e3cd9..f5bb89785e47 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -1018,7 +1018,10 @@  int sock_setsockopt(struct socket *sock, int level, int optname,
 
 	case SO_ZEROCOPY:
 		if (sk->sk_family == PF_INET || sk->sk_family == PF_INET6) {
-			if (sk->sk_protocol != IPPROTO_TCP)
+			if (!((sk->sk_type == SOCK_STREAM &&
+			       sk->sk_protocol == IPPROTO_TCP) ||
+			      (sk->sk_type == SOCK_DGRAM &&
+			       sk->sk_protocol == IPPROTO_UDP)))
 				ret = -ENOTSUPP;
 		} else if (sk->sk_family != PF_RDS) {
 			ret = -ENOTSUPP;
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index c09219e7f230..9602746d7175 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -867,6 +867,7 @@  static int __ip_append_data(struct sock *sk,
 			    unsigned int flags)
 {
 	struct inet_sock *inet = inet_sk(sk);
+	struct ubuf_info *uarg = NULL;
 	struct sk_buff *skb;
 
 	struct ip_options *opt = cork->opt;
@@ -916,6 +917,19 @@  static int __ip_append_data(struct sock *sk,
 	    (!exthdrlen || (rt->dst.dev->features & NETIF_F_HW_ESP_TX_CSUM)))
 		csummode = CHECKSUM_PARTIAL;
 
+	if (flags & MSG_ZEROCOPY && length && sock_flag(sk, SOCK_ZEROCOPY)) {
+		uarg = sock_zerocopy_realloc(sk, length, skb_zcopy(skb));
+		if (!uarg)
+			return -ENOBUFS;
+		if (rt->dst.dev->features & NETIF_F_SG &&
+		    csummode == CHECKSUM_PARTIAL) {
+			paged = true;
+		} else {
+			uarg->zerocopy = 0;
+			skb_zcopy_set(skb, uarg);
+		}
+	}
+
 	cork->length += length;
 
 	/* So, what's going on in the loop below?
@@ -1005,6 +1019,7 @@  static int __ip_append_data(struct sock *sk,
 			cork->tx_flags = 0;
 			skb_shinfo(skb)->tskey = tskey;
 			tskey = 0;
+			skb_zcopy_set(skb, uarg);
 
 			/*
 			 *	Find where to start putting bytes.
@@ -1067,7 +1082,7 @@  static int __ip_append_data(struct sock *sk,
 				err = -EFAULT;
 				goto error;
 			}
-		} else {
+		} else if (!uarg || !uarg->zerocopy) {
 			int i = skb_shinfo(skb)->nr_frags;
 
 			err = -ENOMEM;
@@ -1097,6 +1112,10 @@  static int __ip_append_data(struct sock *sk,
 			skb->data_len += copy;
 			skb->truesize += copy;
 			wmem_alloc_delta += copy;
+		} else {
+			err = skb_zerocopy_iter_dgram(skb, from, copy);
+			if (err < 0)
+				goto error;
 		}
 		offset += copy;
 		length -= copy;
@@ -1104,11 +1123,13 @@  static int __ip_append_data(struct sock *sk,
 
 	if (wmem_alloc_delta)
 		refcount_add(wmem_alloc_delta, &sk->sk_wmem_alloc);
+	sock_zerocopy_put(uarg);
 	return 0;
 
 error_efault:
 	err = -EFAULT;
 error:
+	sock_zerocopy_put_abort(uarg);
 	cork->length -= length;
 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTDISCARDS);
 	refcount_add(wmem_alloc_delta, &sk->sk_wmem_alloc);
diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
index 89e0d5118afe..56f09c4c40dc 100644
--- a/net/ipv6/ip6_output.c
+++ b/net/ipv6/ip6_output.c
@@ -1245,6 +1245,7 @@  static int __ip6_append_data(struct sock *sk,
 {
 	struct sk_buff *skb, *skb_prev = NULL;
 	unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu, pmtu;
+	struct ubuf_info *uarg = NULL;
 	int exthdrlen = 0;
 	int dst_exthdrlen = 0;
 	int hh_len;
@@ -1322,6 +1323,19 @@  static int __ip6_append_data(struct sock *sk,
 	    rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM))
 		csummode = CHECKSUM_PARTIAL;
 
+	if (flags & MSG_ZEROCOPY && length && sock_flag(sk, SOCK_ZEROCOPY)) {
+		uarg = sock_zerocopy_realloc(sk, length, skb_zcopy(skb));
+		if (!uarg)
+			return -ENOBUFS;
+		if (rt->dst.dev->features & NETIF_F_SG &&
+		    csummode == CHECKSUM_PARTIAL) {
+			paged = true;
+		} else {
+			uarg->zerocopy = 0;
+			skb_zcopy_set(skb, uarg);
+		}
+	}
+
 	/*
 	 * Let's try using as much space as possible.
 	 * Use MTU if total length of the message fits into the MTU.
@@ -1444,6 +1458,7 @@  static int __ip6_append_data(struct sock *sk,
 			cork->tx_flags = 0;
 			skb_shinfo(skb)->tskey = tskey;
 			tskey = 0;
+			skb_zcopy_set(skb, uarg);
 
 			/*
 			 *	Find where to start putting bytes
@@ -1505,7 +1520,7 @@  static int __ip6_append_data(struct sock *sk,
 				err = -EFAULT;
 				goto error;
 			}
-		} else {
+		} else if (!uarg || !uarg->zerocopy) {
 			int i = skb_shinfo(skb)->nr_frags;
 
 			err = -ENOMEM;
@@ -1535,6 +1550,10 @@  static int __ip6_append_data(struct sock *sk,
 			skb->data_len += copy;
 			skb->truesize += copy;
 			wmem_alloc_delta += copy;
+		} else {
+			err = skb_zerocopy_iter_dgram(skb, from, copy);
+			if (err < 0)
+				goto error;
 		}
 		offset += copy;
 		length -= copy;
@@ -1542,11 +1561,13 @@  static int __ip6_append_data(struct sock *sk,
 
 	if (wmem_alloc_delta)
 		refcount_add(wmem_alloc_delta, &sk->sk_wmem_alloc);
+	sock_zerocopy_put(uarg);
 	return 0;
 
 error_efault:
 	err = -EFAULT;
 error:
+	sock_zerocopy_put_abort(uarg);
 	cork->length -= length;
 	IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
 	refcount_add(wmem_alloc_delta, &sk->sk_wmem_alloc);