From: Sasha Levin <sashal@kernel.org>
To: stable@vger.kernel.org
Cc: Kuniyuki Iwashima <kuniyu@google.com>,
Eric Dumazet <edumazet@google.com>,
Jakub Kicinski <kuba@kernel.org>, Sasha Levin <sashal@kernel.org>
Subject: [PATCH 6.6.y 4/5] ipv6: Move ipv6_fl_list from ipv6_pinfo to inet_sock.
Date: Wed, 16 Sep 2026 23:13:51 -0400 [thread overview]
Message-ID: <20260917031352.3971853-4-sashal@kernel.org> (raw)
In-Reply-To: <20260917031352.3971853-1-sashal@kernel.org>
From: Kuniyuki Iwashima <kuniyu@google.com>
[ Upstream commit 1c17f4373d4db1e1f0ebd3ddcd8e7a642927a826 ]
In {tcp6,udp6,raw6}_sock, struct ipv6_pinfo is always placed at
the beginning of a new cache line because
1. __alignof__(struct tcp_sock) is 64 due to ____cacheline_aligned
of __cacheline_group_begin(tcp_sock_write_tx)
2. __alignof__(struct udp_sock) is 64 due to ____cacheline_aligned
of struct numa_drop_counters
3. in raw6_sock, struct numa_drop_counters is placed before
struct ipv6_pinfo
. struct ipv6_pinfo is 136 bytes, but the last cache line is
only used by ipv6_fl_list:
$ pahole -C ipv6_pinfo vmlinux
struct ipv6_pinfo {
...
/* --- cacheline 2 boundary (128 bytes) --- */
struct ipv6_fl_socklist * ipv6_fl_list; /* 128 8 */
/* size: 136, cachelines: 3, members: 23 */
Let's move ipv6_fl_list from struct ipv6_pinfo to struct inet_sock
to save a full cache line for {tcp6,udp6,raw6}_sock.
Now, struct ipv6_pinfo is 128 bytes, and {tcp6,udp6,raw6}_sock have
64 bytes less, while {tcp,udp,raw}_sock retain the same size.
Before:
# grep -E "^(RAW|UDP[^L\-]|TCP)" /proc/slabinfo | awk '{print $1, "\t", $4}'
RAWv6 1408
UDPv6 1472
TCPv6 2560
RAW 1152
UDP 1280
TCP 2368
After:
# grep -E "^(RAW|UDP[^L\-]|TCP)" /proc/slabinfo | awk '{print $1, "\t", $4}'
RAWv6 1344
UDPv6 1408
TCPv6 2496
RAW 1152
UDP 1280
TCP 2368
Also, ipv6_fl_list and inet_flags (SNDFLOW bit) are placed in the
same cache line.
$ pahole -C inet_sock vmlinux
...
/* --- cacheline 11 boundary (704 bytes) was 56 bytes ago --- */
struct ipv6_pinfo * pinet6; /* 760 8 */
/* --- cacheline 12 boundary (768 bytes) --- */
struct ipv6_fl_socklist * ipv6_fl_list; /* 768 8 */
unsigned long inet_flags; /* 776 8 */
Doc churn is due to the insufficient Type column (only 1 space short).
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20251014224210.2964778-1-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Backport notes for Linux 6.6:
Drop the cacheline documentation update because the file does not exist
in this tree. Remove only ipv6_fl_list from the older ipv6_pinfo layout,
keeping its transmit options, packet options and cork fields intact.
Apply the IPv4-mapped TCP child change in the existing
tcp_v6_mapped_child_init() helper. Also update both IPv6 DCCP child paths,
which still exist in 6.6, so cloned sockets do not inherit the listener's
flow-label list. Retain the ipv6_pinfo pointer in ipv6_flowlabel_get(),
where 6.6 still uses it for the repflow flag. No functions are added.
This preserves the socket-based flow-label iterator needed by
8d6cd1885085 ("ipv6: flowlabel: cap duplicate leases per socket").
The target applies cleanly with a three-way cherry-pick; retain the
stable branch's existing mem_check() policy.
[ sashal: Reduced backport -- upstream 1c17f4373d4db touches 7 file(s), this
backport carries 7. Not backported here:
Documentation/networking/net_cachelines/inet_sock.rst
This note is generated from the file lists only; see the resolution record
for the reasoning. ]
Stable-dep-of: 8d6cd1885085 ("ipv6: flowlabel: cap duplicate leases per socket")
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
.../chelsio/inline_crypto/chtls/chtls_cm.c | 4 +-
include/linux/ipv6.h | 1 -
include/net/inet_sock.h | 1 +
net/dccp/ipv6.c | 4 +-
net/ipv6/ip6_flowlabel.c | 43 +++++++++----------
net/ipv6/tcp_ipv6.c | 10 ++---
net/sctp/ipv6.c | 8 ++--
7 files changed, 36 insertions(+), 35 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_cm.c b/drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_cm.c
index 6f6525983130e..6d69c19ea7c30 100644
--- a/drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_cm.c
+++ b/drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_cm.c
@@ -1197,12 +1197,12 @@ static struct sock *chtls_recv_sock(struct sock *lsk,
struct ipv6_pinfo *newnp = inet6_sk(newsk);
struct ipv6_pinfo *np = inet6_sk(lsk);
- inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
+ newinet->pinet6 = &newtcp6sk->inet6;
+ newinet->ipv6_fl_list = NULL;
memcpy(newnp, np, sizeof(struct ipv6_pinfo));
newsk->sk_v6_daddr = treq->ir_v6_rmt_addr;
newsk->sk_v6_rcv_saddr = treq->ir_v6_loc_addr;
inet6_sk(newsk)->saddr = treq->ir_v6_loc_addr;
- newnp->ipv6_fl_list = NULL;
newnp->pktoptions = NULL;
newsk->sk_bound_dev_if = treq->ir_iif;
newinet->inet_opt = NULL;
diff --git a/include/linux/ipv6.h b/include/linux/ipv6.h
index fc52145ab732e..4c577356d7e9d 100644
--- a/include/linux/ipv6.h
+++ b/include/linux/ipv6.h
@@ -285,7 +285,6 @@ struct ipv6_pinfo {
struct ipv6_mc_socklist __rcu *ipv6_mc_list;
struct ipv6_ac_socklist *ipv6_ac_list;
- struct ipv6_fl_socklist __rcu *ipv6_fl_list;
struct ipv6_txoptions __rcu *opt;
struct sk_buff *pktoptions;
diff --git a/include/net/inet_sock.h b/include/net/inet_sock.h
index 2790ba58ffe5c..ee3456bc1a401 100644
--- a/include/net/inet_sock.h
+++ b/include/net/inet_sock.h
@@ -211,6 +211,7 @@ struct inet_sock {
struct sock sk;
#if IS_ENABLED(CONFIG_IPV6)
struct ipv6_pinfo *pinet6;
+ struct ipv6_fl_socklist __rcu *ipv6_fl_list;
#endif
/* Socket demultiplex comparisons on incoming packets. */
#define inet_daddr sk.__sk_common.skc_daddr
diff --git a/net/dccp/ipv6.c b/net/dccp/ipv6.c
index ae617bd8afab8..a4158fc6a6846 100644
--- a/net/dccp/ipv6.c
+++ b/net/dccp/ipv6.c
@@ -454,7 +454,7 @@ static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
newnp->opt = NULL;
newnp->ipv6_mc_list = NULL;
newnp->ipv6_ac_list = NULL;
- newnp->ipv6_fl_list = NULL;
+ newinet->ipv6_fl_list = NULL;
newnp->mcast_oif = inet_iif(skb);
newnp->mcast_hops = ip_hdr(skb)->ttl;
@@ -521,7 +521,7 @@ static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
newnp->ipv6_mc_list = NULL;
newnp->ipv6_ac_list = NULL;
- newnp->ipv6_fl_list = NULL;
+ newinet->ipv6_fl_list = NULL;
newnp->pktoptions = NULL;
newnp->opt = NULL;
newnp->mcast_oif = inet6_iif(skb);
diff --git a/net/ipv6/ip6_flowlabel.c b/net/ipv6/ip6_flowlabel.c
index bf909fa69a977..04ec8f0bda7d1 100644
--- a/net/ipv6/ip6_flowlabel.c
+++ b/net/ipv6/ip6_flowlabel.c
@@ -66,8 +66,8 @@ EXPORT_SYMBOL(ipv6_flowlabel_exclusive);
fl != NULL; \
fl = rcu_dereference(fl->next))
-#define for_each_sk_fl_rcu(np, sfl) \
- for (sfl = rcu_dereference(np->ipv6_fl_list); \
+#define for_each_sk_fl_rcu(sk, sfl) \
+ for (sfl = rcu_dereference(inet_sk(sk)->ipv6_fl_list); \
sfl != NULL; \
sfl = rcu_dereference(sfl->next))
@@ -257,12 +257,11 @@ static struct ip6_flowlabel *fl_intern(struct net *net,
struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label)
{
struct ipv6_fl_socklist *sfl;
- struct ipv6_pinfo *np = inet6_sk(sk);
label &= IPV6_FLOWLABEL_MASK;
rcu_read_lock();
- for_each_sk_fl_rcu(np, sfl) {
+ for_each_sk_fl_rcu(sk, sfl) {
struct ip6_flowlabel *fl = sfl->fl;
if (fl->label == label && atomic_inc_not_zero(&fl->users)) {
@@ -278,16 +277,16 @@ EXPORT_SYMBOL_GPL(__fl6_sock_lookup);
void fl6_free_socklist(struct sock *sk)
{
- struct ipv6_pinfo *np = inet6_sk(sk);
+ struct inet_sock *inet = inet_sk(sk);
struct ipv6_fl_socklist *sfl;
- if (!rcu_access_pointer(np->ipv6_fl_list))
+ if (!rcu_access_pointer(inet->ipv6_fl_list))
return;
spin_lock_bh(&ip6_sk_fl_lock);
- while ((sfl = rcu_dereference_protected(np->ipv6_fl_list,
+ while ((sfl = rcu_dereference_protected(inet->ipv6_fl_list,
lockdep_is_held(&ip6_sk_fl_lock))) != NULL) {
- np->ipv6_fl_list = sfl->next;
+ inet->ipv6_fl_list = sfl->next;
spin_unlock_bh(&ip6_sk_fl_lock);
fl_release(sfl->fl);
@@ -465,16 +464,15 @@ fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq,
static int mem_check(struct sock *sk)
{
- struct ipv6_pinfo *np = inet6_sk(sk);
- struct ipv6_fl_socklist *sfl;
int room = FL_MAX_SIZE - atomic_read(&fl_size);
+ struct ipv6_fl_socklist *sfl;
int count = 0;
if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
return 0;
rcu_read_lock();
- for_each_sk_fl_rcu(np, sfl)
+ for_each_sk_fl_rcu(sk, sfl)
count++;
rcu_read_unlock();
@@ -487,13 +485,15 @@ static int mem_check(struct sock *sk)
return 0;
}
-static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
- struct ip6_flowlabel *fl)
+static inline void fl_link(struct sock *sk, struct ipv6_fl_socklist *sfl,
+ struct ip6_flowlabel *fl)
{
+ struct inet_sock *inet = inet_sk(sk);
+
spin_lock_bh(&ip6_sk_fl_lock);
sfl->fl = fl;
- sfl->next = np->ipv6_fl_list;
- rcu_assign_pointer(np->ipv6_fl_list, sfl);
+ sfl->next = inet->ipv6_fl_list;
+ rcu_assign_pointer(inet->ipv6_fl_list, sfl);
spin_unlock_bh(&ip6_sk_fl_lock);
}
@@ -515,7 +515,7 @@ int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
rcu_read_lock();
- for_each_sk_fl_rcu(np, sfl) {
+ for_each_sk_fl_rcu(sk, sfl) {
if (sfl->fl->label == (np->flow_label & IPV6_FLOWLABEL_MASK)) {
spin_lock_bh(&ip6_fl_lock);
freq->flr_label = sfl->fl->label;
@@ -554,7 +554,7 @@ static int ipv6_flowlabel_put(struct sock *sk, struct in6_flowlabel_req *freq)
}
spin_lock_bh(&ip6_sk_fl_lock);
- for (sflp = &np->ipv6_fl_list;
+ for (sflp = &inet_sk(sk)->ipv6_fl_list;
(sfl = socklist_dereference(*sflp)) != NULL;
sflp = &sfl->next) {
if (sfl->fl->label == freq->flr_label)
@@ -574,13 +574,12 @@ static int ipv6_flowlabel_put(struct sock *sk, struct in6_flowlabel_req *freq)
static int ipv6_flowlabel_renew(struct sock *sk, struct in6_flowlabel_req *freq)
{
- struct ipv6_pinfo *np = inet6_sk(sk);
struct net *net = sock_net(sk);
struct ipv6_fl_socklist *sfl;
int err;
rcu_read_lock();
- for_each_sk_fl_rcu(np, sfl) {
+ for_each_sk_fl_rcu(sk, sfl) {
if (sfl->fl->label == freq->flr_label) {
err = fl6_renew(sfl->fl, freq->flr_linger,
freq->flr_expires);
@@ -640,7 +639,7 @@ static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq,
if (freq->flr_label) {
err = -EEXIST;
rcu_read_lock();
- for_each_sk_fl_rcu(np, sfl) {
+ for_each_sk_fl_rcu(sk, sfl) {
if (sfl->fl->label == freq->flr_label) {
if (freq->flr_flags & IPV6_FL_F_EXCL) {
rcu_read_unlock();
@@ -677,7 +676,7 @@ static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq,
fl1->linger = fl->linger;
if ((long)(fl->expires - fl1->expires) > 0)
fl1->expires = fl->expires;
- fl_link(np, sfl1, fl1);
+ fl_link(sk, sfl1, fl1);
fl_free(fl);
return 0;
@@ -711,7 +710,7 @@ static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq,
}
}
- fl_link(np, sfl1, fl);
+ fl_link(sk, sfl1, fl);
return 0;
done:
fl_free(fl);
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index aef3c7bc3d80d..458c1d6e9021c 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1187,6 +1187,7 @@ static void tcp_v6_mapped_child_init(struct sock *newsk, const struct sock *sk)
struct ipv6_pinfo *newnp;
newinet->pinet6 = newnp = tcp_inet6_sk(newsk);
+ newinet->ipv6_fl_list = NULL;
memcpy(newnp, tcp_inet6_sk(sk), sizeof(struct ipv6_pinfo));
@@ -1202,7 +1203,6 @@ static void tcp_v6_mapped_child_init(struct sock *newsk, const struct sock *sk)
newnp->ipv6_mc_list = NULL;
newnp->ipv6_ac_list = NULL;
- newnp->ipv6_fl_list = NULL;
newnp->pktoptions = NULL;
newnp->opt = NULL;
@@ -1265,10 +1265,12 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
newsk->sk_gso_type = SKB_GSO_TCPV6;
inet6_sk_rx_dst_set(newsk, skb);
- inet_sk(newsk)->pinet6 = tcp_inet6_sk(newsk);
+ newinet = inet_sk(newsk);
+ newinet->pinet6 = tcp_inet6_sk(newsk);
+ newinet->ipv6_fl_list = NULL;
+ newinet->inet_opt = NULL;
newtp = tcp_sk(newsk);
- newinet = inet_sk(newsk);
newnp = tcp_inet6_sk(newsk);
memcpy(newnp, np, sizeof(struct ipv6_pinfo));
@@ -1284,10 +1286,8 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
First: no IPv4 options.
*/
- newinet->inet_opt = NULL;
newnp->ipv6_mc_list = NULL;
newnp->ipv6_ac_list = NULL;
- newnp->ipv6_fl_list = NULL;
/* Clone RX bits */
newnp->rxopt.all = np->rxopt.all;
diff --git a/net/sctp/ipv6.c b/net/sctp/ipv6.c
index 0673857cb3d8b..5db4220aacf72 100644
--- a/net/sctp/ipv6.c
+++ b/net/sctp/ipv6.c
@@ -782,9 +782,10 @@ static struct sock *sctp_v6_create_accept_sk(struct sock *sk,
struct sctp_association *asoc,
bool kern)
{
- struct sock *newsk;
struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
struct sctp6_sock *newsctp6sk;
+ struct inet_sock *newinet;
+ struct sock *newsk;
newsk = sk_alloc(sock_net(sk), PF_INET6, GFP_KERNEL, sk->sk_prot, kern);
if (!newsk)
@@ -796,7 +797,9 @@ static struct sock *sctp_v6_create_accept_sk(struct sock *sk,
sock_reset_flag(sk, SOCK_ZAPPED);
newsctp6sk = (struct sctp6_sock *)newsk;
- inet_sk(newsk)->pinet6 = &newsctp6sk->inet6;
+ newinet = inet_sk(newsk);
+ newinet->pinet6 = &newsctp6sk->inet6;
+ newinet->ipv6_fl_list = NULL;
sctp_sk(newsk)->v4mapped = sctp_sk(sk)->v4mapped;
@@ -805,7 +808,6 @@ static struct sock *sctp_v6_create_accept_sk(struct sock *sk,
memcpy(newnp, np, sizeof(struct ipv6_pinfo));
newnp->ipv6_mc_list = NULL;
newnp->ipv6_ac_list = NULL;
- newnp->ipv6_fl_list = NULL;
sctp_v6_copy_ip_options(sk, newsk);
--
2.53.0
next prev parent reply other threads:[~2026-09-17 3:13 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-16 15:59 FAILED: patch "[PATCH] ipv6: flowlabel: cap duplicate leases per socket" failed to apply to 6.6-stable tree gregkh
2026-09-17 3:13 ` [PATCH 6.6.y 1/5] ipv4: remove fib_devindex_hashfn() Sasha Levin
2026-09-17 3:13 ` [PATCH 6.6.y 2/5] ipv4: use rcu in ip_fib_check_default() Sasha Levin
2026-09-17 3:13 ` [PATCH 6.6.y 3/5] ipv6: make ipv6_pinfo.saddr_cache a boolean Sasha Levin
2026-09-17 3:13 ` Sasha Levin [this message]
2026-09-17 3:13 ` [PATCH 6.6.y 5/5] ipv6: flowlabel: cap duplicate leases per socket Sasha Levin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260917031352.3971853-4-sashal@kernel.org \
--to=sashal@kernel.org \
--cc=edumazet@google.com \
--cc=kuba@kernel.org \
--cc=kuniyu@google.com \
--cc=stable@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox