[PATCH v4] net/smc: release the internal TCP sock on IPPROTO_SMC socket creation failure

IPPROTO_SMC sockets create an internal TCP sock ("clcsock") from the
proto->init hook. When socket creation fails after proto->init has
run - e.g. a cgroup BPF program attached to BPF_CGROUP_INET_SOCK_CREATE
denies the socket - sk_common_release() only invokes sk_prot->destroy
if it is set, but neither smc_inet_prot nor smc_inet6_prot defines it,
and smc_destruct() returns early unless sk_state is SMC_CLOSED. As a
result, every failing socket(AF_INET, SOCK_STREAM, IPPROTO_SMC) call
leaks one tcp_sock, so an unprivileged task able to attach a deny-all
BPF_CGROUP_INET_SOCK_CREATE program to its own cgroup can grow kernel
memory unboundedly.

Add a .destroy hook to both protos that releases the clcsock via
smc_clcsock_release(). smc_sk_init() hashes the sock into the smc
hashinfo before the clcsock is created, and smc_diag dumps walk that
hash dereferencing smc->clcsock without taking clcsock_release_lock,
while sk_common_release() calls .destroy before .unhash. Unhash the
sock before releasing the clcsock, as __smc_release() does, so a
concurrent dump cannot observe the release; the second unhash in
sk_common_release() is a no-op.

Fixes: d25a92ccae6b ("net/smc: Introduce IPPROTO_SMC")
Reported-by: Abaci <[email protected]>
Assisted-by: abaci:qwen3.8-max
Signed-off-by: Yifei Chu <[email protected]>
---
Changes since v3:
- no code change: use full name in Author and Signed-off-by, resend
  as a new thread instead of a reply, as requested. Previous thread:
  https://lore.kernel.org/all/20260813060456.b48HqF9m6NTfFXvq_PL9w_Ih493OogqAqxfAHAfU74A@z/

Changes since v2 (answering the review):
- dropped the smc->clcsock = NULL store: sk_alloc() zeroes the sock
  (__GFP_ZERO / sk_prot_clear_nulls()), so it is already NULL. The
  mutex concern does not apply either: .destroy only runs after
  smc_sk_init() initialized clcsock_release_lock.
- unhash the sock before releasing the clcsock in .destroy so a
  concurrent smc_diag dump cannot observe the release
  (sk_common_release() calls .destroy before .unhash); the second
  unhash is a no-op. Tested on a KASAN+SMC_DIAG kernel: the v2
  version hits a KASAN report in __smc_diag_dump() when a diag dump
  races with failing IPPROTO_SMC socket creation, v3 stays clean
  under the same stress.

 net/smc/smc_inet.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/net/smc/smc_inet.c b/net/smc/smc_inet.c
index a94084b4a498..520b666fdd8f 100644
--- a/net/smc/smc_inet.c
+++ b/net/smc/smc_inet.c
@@ -15,13 +15,16 @@
 
 #include "smc_inet.h"
 #include "smc.h"
+#include "smc_close.h"
 
 static int smc_inet_init_sock(struct sock *sk);
+static void smc_inet_destroy_sock(struct sock *sk);
 
 static struct proto smc_inet_prot = {
 	.name		= "INET_SMC",
 	.owner		= THIS_MODULE,
 	.init		= smc_inet_init_sock,
+	.destroy	= smc_inet_destroy_sock,
 	.hash		= smc_hash_sk,
 	.unhash		= smc_unhash_sk,
 	.release_cb	= smc_release_cb,
@@ -68,6 +71,7 @@ static struct proto smc_inet6_prot = {
 	.name		= "INET6_SMC",
 	.owner		= THIS_MODULE,
 	.init		= smc_inet_init_sock,
+	.destroy	= smc_inet_destroy_sock,
 	.hash		= smc_hash_sk,
 	.unhash		= smc_unhash_sk,
 	.release_cb	= smc_release_cb,
@@ -116,6 +120,18 @@ static int smc_inet_init_sock(struct sock *sk)
 	return smc_create_clcsk(net, sk, sk->sk_family);
 }
 
+static void smc_inet_destroy_sock(struct sock *sk)
+{
+	/* The sock is hashed and smc_diag dumps dereference smc->clcsock
+	 * without clcsock_release_lock, while sk_common_release() calls
+	 * .destroy before .unhash. Unhash first, as __smc_release() does,
+	 * so no dump can observe the clcsock being released; the second
+	 * unhash is a no-op.
+	 */
+	sk->sk_prot->unhash(sk);
+	smc_clcsock_release(smc_sk(sk));
+}
+
 int __init smc_inet_init(void)
 {
 	int rc;
-- 
2.43.5