[PATCH RESEND v2 0/2] nfc: llcp: Move to .getsockopt_iter

[PATCH RESEND v2 0/2] nfc: llcp: Move to .getsockopt_iter

[Breno Leitao]

Continue to convert the protocols to getsockopt_iter callback.

Proto llcp is one of the last to be converted, and it was waiting for
the following fixes to land first, avoiding merge conflicts.

 * commit 36812527052c ("nfc: llcp: read llcp_sock->local under the
   socket lock in getsockopt")
 * commit 99985bfa8336 ("nfc: llcp: avoid userspace overflow on invalid
   optlen")

Now that they have landed, let's move llcp to .getsockopt_iter callback

Regarding the whole project migration, this is where we are:

Missing proto_ops migration (as of net-next)

net/can/isotp.c:2064                          isotp_ops                  (.getsockopt = isotp_getsockopt)  [1]
net/can/j1939/socket.c:1375                   j1939_ops                  (.getsockopt = j1939_sk_getsockopt) [1]
net/ieee802154/socket.c:411                   ieee802154_raw_ops         (.getsockopt = sock_common_getsockopt)
net/ieee802154/socket.c:987                   ieee802154_dgram_ops       (.getsockopt = sock_common_getsockopt)
net/ipv4/af_inet.c:1058                       inet_stream_ops            (.getsockopt = sock_common_getsockopt)
net/ipv4/af_inet.c:1094                       inet_dgram_ops             (.getsockopt = sock_common_getsockopt)
net/ipv4/af_inet.c:1126                       inet_sockraw_ops           (.getsockopt = sock_common_getsockopt)
net/ipv6/af_inet6.c:656                       inet6_stream_ops           (.getsockopt = sock_common_getsockopt)
net/ipv6/af_inet6.c:692                       inet6_dgram_ops            (.getsockopt = sock_common_getsockopt)
net/ipv6/raw.c:1290                           inet6_sockraw_ops          (.getsockopt = sock_common_getsockopt)
net/l2tp/l2tp_ip.c:628                        l2tp_ip_ops                (.getsockopt = sock_common_getsockopt)
net/l2tp/l2tp_ip6.c:757                       l2tp_ip6_ops               (.getsockopt = sock_common_getsockopt)
net/mptcp/protocol.c:4753                     mptcp_stream_ops           (.getsockopt = sock_common_getsockopt)
net/mptcp/protocol.c:4862                     mptcp_v6_stream_ops        (.getsockopt = sock_common_getsockopt)
net/phonet/socket.c:451                       phonet_stream_ops          (.getsockopt = sock_common_getsockopt)
net/sctp/ipv6.c:1047                          inet6_seqpacket_ops        (.getsockopt = sock_common_getsockopt)
net/sctp/protocol.c:1106                      inet_seqpacket_ops         (.getsockopt = sock_common_getsockopt)
net/smc/af_smc.c:3345                         smc_sock_ops               (.getsockopt = smc_getsockopt)
net/smc/smc_inet.c:36                         smc_inet_stream_ops        (.getsockopt = smc_getsockopt)
net/smc/smc_inet.c:84                         smc_inet6_stream_ops       (.getsockopt = smc_getsockopt)

Link: https://lore.kernel.org/all/20260507-getsock_two_can-v1-0-3c2ae9edfadc@debian.org/ [1]

Signed-off-by: Breno Leitao <leitao@debian.org>
---
Changes in v2:
- Rebase on net-next, no code change
- Link to v1: https://patch.msgid.link/20260708-getsockopt_phase3-v1-0-d8394f2fc554@debian.org

---
Breno Leitao (2):
      nfc: llcp: widen getsockopt value locals to u32
      nfc: llcp: convert to getsockopt_iter

 net/nfc/llcp_sock.c | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)
---
base-commit: fa5196392691051963726787944db872420dfa75
base-tree: git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git main
change-id: 20260706-getsockopt_phase3-34f7b5eb3e0b

Best regards,
--  
Breno Leitao <leitao@debian.org>

[PATCH RESEND v2 1/2] nfc: llcp: widen getsockopt value locals to u32

[Breno Leitao]

nfc_llcp_getsockopt() returns each option as a u32 (written through a
(u32 __user *) cast), but the temporary variables are smaller than that,
and they are promoted at write side, example:

  u8 rw;
  put_user(rw, (u32 __user *) optval)

That is fine, but I want to widen the rw/miux/remote_miu locals from
u8/u16 to u32 so the following getsockopt_iter conversion can pass their
address to copy_to_iter(), which expects an object of the exact width.

No functional change.

PS: I am sending this as a separated patch to make sure this conversion
explicitly reviewable.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 net/nfc/llcp_sock.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/net/nfc/llcp_sock.c b/net/nfc/llcp_sock.c
index 1e5ee4bcde684f..adb6e06229b4af 100644
--- a/net/nfc/llcp_sock.c
+++ b/net/nfc/llcp_sock.c
@@ -308,8 +308,7 @@ static int nfc_llcp_getsockopt(struct socket *sock, int level, int optname,
 	struct sock *sk = sock->sk;
 	struct nfc_llcp_sock *llcp_sock = nfc_llcp_sock(sk);
 	int len, err = 0;
-	u16 miux, remote_miu;
-	u8 rw;
+	u32 miux, remote_miu, rw;
 
 	pr_debug("%p optname %d\n", sk, optname);
 

-- 
2.53.0-Meta

Re: [PATCH RESEND v2 1/2] nfc: llcp: widen getsockopt value locals to u32

[Simon Horman]

On Mon, Nov 09, 2026 at 05:50:37AM -0800, Breno Leitao wrote:
> nfc_llcp_getsockopt() returns each option as a u32 (written through a
> (u32 __user *) cast), but the temporary variables are smaller than that,
> and they are promoted at write side, example:
> 
>   u8 rw;
>   put_user(rw, (u32 __user *) optval)
> 
> That is fine, but I want to widen the rw/miux/remote_miu locals from
> u8/u16 to u32 so the following getsockopt_iter conversion can pass their
> address to copy_to_iter(), which expects an object of the exact width.
> 
> No functional change.
> 
> PS: I am sending this as a separated patch to make sure this conversion
> explicitly reviewable.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

Reviewed-by: Simon Horman <horms@kernel.org>

[PATCH RESEND v2 2/2] nfc: llcp: convert to getsockopt_iter

[Breno Leitao]

Convert nfc_llcp_getsockopt() to the getsockopt_iter proto_ops callback.
It takes a sockopt_t and writes each value with copy_to_iter() instead of
a put_user() into optval; the socket layer builds the sockopt_t and copies
the length back. The socket lock and optlen validation are unchanged.

No functional change.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 net/nfc/llcp_sock.c | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/net/nfc/llcp_sock.c b/net/nfc/llcp_sock.c
index adb6e06229b4af..8323a0e6a02ea3 100644
--- a/net/nfc/llcp_sock.c
+++ b/net/nfc/llcp_sock.c
@@ -302,21 +302,20 @@ static int nfc_llcp_setsockopt(struct socket *sock, int level, int optname,
 }
 
 static int nfc_llcp_getsockopt(struct socket *sock, int level, int optname,
-			       char __user *optval, int __user *optlen)
+			       sockopt_t *opt)
 {
 	struct nfc_llcp_local *local;
 	struct sock *sk = sock->sk;
 	struct nfc_llcp_sock *llcp_sock = nfc_llcp_sock(sk);
+	u32 miux, remote_miu, remote_lto, remote_rw, rw;
 	int len, err = 0;
-	u32 miux, remote_miu, rw;
 
 	pr_debug("%p optname %d\n", sk, optname);
 
 	if (level != SOL_NFC)
 		return -ENOPROTOOPT;
 
-	if (get_user(len, optlen))
-		return -EFAULT;
+	len = opt->optlen;
 
 	if (len < 0)
 		return -EINVAL;
@@ -337,7 +336,7 @@ static int nfc_llcp_getsockopt(struct socket *sock, int level, int optname,
 	switch (optname) {
 	case NFC_LLCP_RW:
 		rw = llcp_sock->rw > LLCP_MAX_RW ? local->rw : llcp_sock->rw;
-		if (put_user(rw, (u32 __user *) optval))
+		if (copy_to_iter(&rw, len, &opt->iter_out) != len)
 			err = -EFAULT;
 
 		break;
@@ -346,7 +345,7 @@ static int nfc_llcp_getsockopt(struct socket *sock, int level, int optname,
 		miux = be16_to_cpu(llcp_sock->miux) > LLCP_MAX_MIUX ?
 			be16_to_cpu(local->miux) : be16_to_cpu(llcp_sock->miux);
 
-		if (put_user(miux, (u32 __user *) optval))
+		if (copy_to_iter(&miux, len, &opt->iter_out) != len)
 			err = -EFAULT;
 
 		break;
@@ -355,19 +354,21 @@ static int nfc_llcp_getsockopt(struct socket *sock, int level, int optname,
 		remote_miu = llcp_sock->remote_miu > LLCP_MAX_MIU ?
 				local->remote_miu : llcp_sock->remote_miu;
 
-		if (put_user(remote_miu, (u32 __user *) optval))
+		if (copy_to_iter(&remote_miu, len, &opt->iter_out) != len)
 			err = -EFAULT;
 
 		break;
 
 	case NFC_LLCP_REMOTE_LTO:
-		if (put_user(local->remote_lto / 10, (u32 __user *) optval))
+		remote_lto = local->remote_lto / 10;
+		if (copy_to_iter(&remote_lto, len, &opt->iter_out) != len)
 			err = -EFAULT;
 
 		break;
 
 	case NFC_LLCP_REMOTE_RW:
-		if (put_user(llcp_sock->remote_rw, (u32 __user *) optval))
+		remote_rw = llcp_sock->remote_rw;
+		if (copy_to_iter(&remote_rw, len, &opt->iter_out) != len)
 			err = -EFAULT;
 
 		break;
@@ -379,8 +380,7 @@ static int nfc_llcp_getsockopt(struct socket *sock, int level, int optname,
 
 	release_sock(sk);
 
-	if (put_user(len, optlen))
-		return -EFAULT;
+	opt->optlen = len;
 
 	return err;
 }
@@ -966,7 +966,7 @@ static const struct proto_ops llcp_sock_ops = {
 	.listen         = llcp_sock_listen,
 	.shutdown       = sock_no_shutdown,
 	.setsockopt     = nfc_llcp_setsockopt,
-	.getsockopt     = nfc_llcp_getsockopt,
+	.getsockopt_iter = nfc_llcp_getsockopt,
 	.sendmsg        = llcp_sock_sendmsg,
 	.recvmsg        = llcp_sock_recvmsg,
 	.mmap           = sock_no_mmap,

-- 
2.53.0-Meta

Re: [PATCH RESEND v2 2/2] nfc: llcp: convert to getsockopt_iter

[Simon Horman]

On Mon, Nov 09, 2026 at 05:50:38AM -0800, Breno Leitao wrote:
> Convert nfc_llcp_getsockopt() to the getsockopt_iter proto_ops callback.
> It takes a sockopt_t and writes each value with copy_to_iter() instead of
> a put_user() into optval; the socket layer builds the sockopt_t and copies
> the length back. The socket lock and optlen validation are unchanged.
> 
> No functional change.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

Reviewed-by: Simon Horman <horms@kernel.org>