diff mbox series

[net-next,03/14] ipv6: lockless IPV6_MULTICAST_HOPS implementation

Message ID 20230912160212.3467976-4-edumazet@google.com (mailing list archive)
State Accepted
Commit 2da23eb07c91241d962f3ff05565065484cd8929
Delegated to: Netdev Maintainers
Headers show
Series ipv6: round of data-races fixes | expand

Checks

Context Check Description
netdev/series_format success Posting correctly formatted
netdev/tree_selection success Clearly marked for net-next, async
netdev/fixes_present success Fixes tag not required for -next series
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 3184 this patch: 3184
netdev/cc_maintainers warning 12 maintainers not CCed: kadlec@netfilter.org jannh@google.com dccp@vger.kernel.org fw@strlen.de coreteam@netfilter.org kuniyu@amazon.com netfilter-devel@vger.kernel.org pablo@netfilter.org horms@verge.net.au ja@ssi.bg joannelkoong@gmail.com lvs-devel@vger.kernel.org
netdev/build_clang success Errors and warnings before: 1550 this patch: 1550
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/deprecated_api success None detected
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 3432 this patch: 3432
netdev/checkpatch warning WARNING: line length of 81 exceeds 80 columns
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

Eric Dumazet Sept. 12, 2023, 4:02 p.m. UTC 
This fixes data-races around np->mcast_hops,
and make IPV6_MULTICAST_HOPS lockless.

Note that np->mcast_hops is never negative,
thus can fit an u8 field instead of s16.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/linux/ipv6.h            |  9 +--------
 include/net/ipv6.h              |  2 +-
 net/dccp/ipv6.c                 |  2 +-
 net/ipv6/ipv6_sockglue.c        | 28 +++++++++++++++-------------
 net/ipv6/tcp_ipv6.c             |  3 ++-
 net/netfilter/ipvs/ip_vs_sync.c |  2 +-
 6 files changed, 21 insertions(+), 25 deletions(-)

Comments

David Ahern Sept. 14, 2023, 2:55 p.m. UTC | #1 
On 9/12/23 10:02 AM, Eric Dumazet wrote:
> This fixes data-races around np->mcast_hops,
> and make IPV6_MULTICAST_HOPS lockless.
> 
> Note that np->mcast_hops is never negative,
> thus can fit an u8 field instead of s16.
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> ---
>  include/linux/ipv6.h            |  9 +--------
>  include/net/ipv6.h              |  2 +-
>  net/dccp/ipv6.c                 |  2 +-
>  net/ipv6/ipv6_sockglue.c        | 28 +++++++++++++++-------------
>  net/ipv6/tcp_ipv6.c             |  3 ++-
>  net/netfilter/ipvs/ip_vs_sync.c |  2 +-
>  6 files changed, 21 insertions(+), 25 deletions(-)
> 

Reviewed-by: David Ahern <dsahern@kernel.org>
diff mbox series

Patch

diff --git a/include/linux/ipv6.h b/include/linux/ipv6.h
index 68cf1ca949141e419abf2031db2b42105b821ab0..9cc278b5e4f42ce097e57ecd95a50479a947fd82 100644
--- a/include/linux/ipv6.h
+++ b/include/linux/ipv6.h
@@ -214,15 +214,8 @@  struct ipv6_pinfo {
 	__u32			frag_size;
 
 	s16			hop_limit;
+	u8			mcast_hops;
 
-#if defined(__BIG_ENDIAN_BITFIELD)
-	/* Packed in 16bits. */
-	__s16			mcast_hops:9;
-	__u16			__unused_2:7,
-#else
-	__u16			__unused_2:7;
-	__s16			mcast_hops:9;
-#endif
 	int			ucast_oif;
 	int			mcast_oif;
 
diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 61007db0036482e27121747add0eec77f912b54a..0af1a7565a3602e4deb68762267cba454750341e 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -909,7 +909,7 @@  static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
 	int hlimit;
 
 	if (ipv6_addr_is_multicast(&fl6->daddr))
-		hlimit = np->mcast_hops;
+		hlimit = READ_ONCE(np->mcast_hops);
 	else
 		hlimit = READ_ONCE(np->hop_limit);
 	if (hlimit < 0)
diff --git a/net/dccp/ipv6.c b/net/dccp/ipv6.c
index 33f6ccf6ba77b9bcc24054b09857aaee4bb71acf..83617a16b98e70aa577c08a394df63e006e53e9e 100644
--- a/net/dccp/ipv6.c
+++ b/net/dccp/ipv6.c
@@ -676,7 +676,7 @@  static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
 			np->mcast_oif = inet6_iif(opt_skb);
 		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
-			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
+			WRITE_ONCE(np->mcast_hops, ipv6_hdr(opt_skb)->hop_limit);
 		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
 			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
 		if (np->repflow)
diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
index 755fac85a120de44272f685529b579e7118d306b..5fff19a87c75518358bae067dfeb227d6738bb03 100644
--- a/net/ipv6/ipv6_sockglue.c
+++ b/net/ipv6/ipv6_sockglue.c
@@ -431,6 +431,16 @@  int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 			return -EINVAL;
 		inet6_assign_bit(MC6_LOOP, sk, valbool);
 		return 0;
+	case IPV6_MULTICAST_HOPS:
+		if (sk->sk_type == SOCK_STREAM)
+			return retv;
+		if (optlen < sizeof(int))
+			return -EINVAL;
+		if (val > 255 || val < -1)
+			return -EINVAL;
+		WRITE_ONCE(np->mcast_hops,
+			   val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
+		return 0;
 	}
 	if (needs_rtnl)
 		rtnl_lock();
@@ -751,16 +761,6 @@  int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		break;
 	}
 
-	case IPV6_MULTICAST_HOPS:
-		if (sk->sk_type == SOCK_STREAM)
-			break;
-		if (optlen < sizeof(int))
-			goto e_inval;
-		if (val > 255 || val < -1)
-			goto e_inval;
-		np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
-		retv = 0;
-		break;
 
 	case IPV6_UNICAST_IF:
 	{
@@ -1180,7 +1180,8 @@  int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 			}
 			if (np->rxopt.bits.rxhlim) {
-				int hlim = np->mcast_hops;
+				int hlim = READ_ONCE(np->mcast_hops);
+
 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 			}
 			if (np->rxopt.bits.rxtclass) {
@@ -1197,7 +1198,8 @@  int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 			}
 			if (np->rxopt.bits.rxohlim) {
-				int hlim = np->mcast_hops;
+				int hlim = READ_ONCE(np->mcast_hops);
+
 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 			}
 			if (np->rxopt.bits.rxflow) {
@@ -1349,7 +1351,7 @@  int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 		if (optname == IPV6_UNICAST_HOPS)
 			val = READ_ONCE(np->hop_limit);
 		else
-			val = np->mcast_hops;
+			val = READ_ONCE(np->mcast_hops);
 
 		if (val < 0) {
 			rcu_read_lock();
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 3a88545a265d6bd064ecc41d33c9541a75fe0f4d..54db5fab318bc68cf9efbe6f26dacba614fa8562 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1542,7 +1542,8 @@  int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
 			np->mcast_oif = tcp_v6_iif(opt_skb);
 		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
-			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
+			WRITE_ONCE(np->mcast_hops,
+				   ipv6_hdr(opt_skb)->hop_limit);
 		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
 			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
 		if (np->repflow)
diff --git a/net/netfilter/ipvs/ip_vs_sync.c b/net/netfilter/ipvs/ip_vs_sync.c
index 3c2251cabd0439834ca0fc2b8bbf0ecc6cfe9266..df1b33b61059eef1e86baefc63e138108a50a081 100644
--- a/net/netfilter/ipvs/ip_vs_sync.c
+++ b/net/netfilter/ipvs/ip_vs_sync.c
@@ -1322,7 +1322,7 @@  static void set_mcast_ttl(struct sock *sk, u_char ttl)
 		struct ipv6_pinfo *np = inet6_sk(sk);
 
 		/* IPV6_MULTICAST_HOPS */
-		np->mcast_hops = ttl;
+		WRITE_ONCE(np->mcast_hops, ttl);
 	}
 #endif
 	release_sock(sk);