Re: [PATCH net-next] ipv4: igmp: annotate data-races around im->users
From: Jiayuan Chen
Date: Thu May 21 2026 - 06:53:47 EST
On 5/21/26 5:36 PM, Yuyang Huang wrote:
/proc/net/igmp walks IPv4 multicast memberships under RCU and
prints im->users without holding RTNL, while multicast join and leave
paths update the field while holding RTNL. Annotate this intentional
lockless snapshot with READ_ONCE() and the matching writers with
WRITE_ONCE().
Signed-off-by: Yuyang Huang <sigefriedhyy@xxxxxxxxx>
---
net/ipv4/igmp.c | 11 +++++++----
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/net/ipv4/igmp.c b/net/ipv4/igmp.c
index 27d120183779..f2aca659b29c 100644
--- a/net/ipv4/igmp.c
+++ b/net/ipv4/igmp.c
@@ -1541,7 +1541,7 @@ static void ____ip_mc_inc_group(struct in_device *in_dev, __be32 addr,
}
if (im) {
- im->users++;
+ WRITE_ONCE(im->users, im->users + 1);
ip_mc_add_src(in_dev, &addr, mode, 0, NULL, 0);
goto out;
}
@@ -1550,7 +1550,7 @@ static void ____ip_mc_inc_group(struct in_device *in_dev, __be32 addr,
if (!im)
goto out;
- im->users = 1;
+ WRITE_ONCE(im->users, 1);
im->interface = in_dev;
in_dev_hold(in_dev);
im->multiaddr = addr;
@@ -1784,7 +1784,10 @@ void __ip_mc_dec_group(struct in_device *in_dev, __be32 addr, gfp_t gfp)
(i = rtnl_dereference(*ip)) != NULL;
ip = &i->next_rcu) {
if (i->multiaddr == addr) {
- if (--i->users == 0) {
+ int new_users = i->users - 1;
+
+ WRITE_ONCE(i->users, new_users);
+ if (new_users == 0) {
Using a local variable here feels slightly awkward, but it does save one extra load compared
to re-reading i->users after the WRITE_ONCE(), so it's acceptable.
One suggestion: please add a Fixes: tag, like other data-race fixes annotated with READ_ONCE()/WRITE_ONCE().