[PATCH net-next 0/3] net: Convert AF_NETLINK and AF_VSOCK to getsockopt_iter API

[PATCH net-next 0/3] net: Convert AF_NETLINK and AF_VSOCK to getsockopt_iter API

[Breno Leitao]

Continue the work to convert protocols to the new getsockopt_iter API.

Convert AF_NETLINK and AF_VSOCK getsockopt implementations to the new
sockopt_t/getsockopt_iter API, and add kselftests that verify the size
and errno semantics are preserved across the conversion.

I chose these two socket families because they are probably one of the
most used  protocols,, ensuring that any potential bugs will be
discovered and reported quickly.

The selftest was added as suggested by Stanislav Fomichev in [1].
Link: https://lore.kernel.org/all/adkSnyihmD1Atfcf@devvm17672.vll0.facebook.com/ [1]

Signed-off-by: Breno Leitao <leitao@debian.org>
---
Breno Leitao (3):
      netlink: convert to getsockopt_iter
      vsock: convert to getsockopt_iter
      net: selftests: add getsockopt_iter regression tests

 net/netlink/af_netlink.c                      |  21 +--
 net/vmw_vsock/af_vsock.c                      |  16 +-
 tools/testing/selftests/net/Makefile          |   1 +
 tools/testing/selftests/net/getsockopt_iter.c | 213 ++++++++++++++++++++++++++
 4 files changed, 232 insertions(+), 19 deletions(-)
---
base-commit: edf4bee4215a173c0534d1851d7523d827149f9e
change-id: 20260501-getsock_one-a62758a9ba25
prerequisite-change-id: 20260501-getsock_iter_first-87f6a74c24e0:v1

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

[PATCH net-next 1/3] netlink: convert to getsockopt_iter

[Breno Leitao]

Convert AF_NETLINK's getsockopt implementation to use the new
getsockopt_iter callback with sockopt_t.

Key changes:
- Replace (char __user *optval, int __user *optlen) with sockopt_t *opt
- Use opt->optlen for buffer length (input) and returned size (output)
- Use copy_to_iter() instead of put_user()/copy_to_user()
- For NETLINK_LIST_MEMBERSHIPS: walk the groups bitmap and emit each
  u32 sequentially via copy_to_iter(), then set opt->optlen to the
  total size required (ALIGN(BITS_TO_BYTES(ngroups), sizeof(u32))).
  The wrapper writes opt->optlen back to userspace even on partial
  failure, preserving the existing API that lets userspace discover
  the needed allocation size.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 net/netlink/af_netlink.c | 21 +++++++++++----------
 1 file changed, 11 insertions(+), 10 deletions(-)

diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
index 2aeb0680807d6..db3be485b4804 100644
--- a/net/netlink/af_netlink.c
+++ b/net/netlink/af_netlink.c
@@ -39,6 +39,7 @@
 #include <linux/fs.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
+#include <linux/uio.h>
 #include <linux/skbuff.h>
 #include <linux/netdevice.h>
 #include <linux/rtnetlink.h>
@@ -1716,18 +1717,18 @@ static int netlink_setsockopt(struct socket *sock, int level, int optname,
 }
 
 static int netlink_getsockopt(struct socket *sock, int level, int optname,
-			      char __user *optval, int __user *optlen)
+			      sockopt_t *opt)
 {
 	struct sock *sk = sock->sk;
 	struct netlink_sock *nlk = nlk_sk(sk);
 	unsigned int flag;
 	int len, val;
+	u32 group;
 
 	if (level != SOL_NETLINK)
 		return -ENOPROTOOPT;
 
-	if (get_user(len, optlen))
-		return -EFAULT;
+	len = opt->optlen;
 	if (len < 0)
 		return -EINVAL;
 
@@ -1751,14 +1752,14 @@ static int netlink_getsockopt(struct socket *sock, int level, int optname,
 
 			idx = pos / sizeof(unsigned long);
 			shift = (pos % sizeof(unsigned long)) * 8;
-			if (put_user((u32)(nlk->groups[idx] >> shift),
-				     (u32 __user *)(optval + pos))) {
+			group = (u32)(nlk->groups[idx] >> shift);
+			if (copy_to_iter(&group, sizeof(u32),
+					 &opt->iter_out) != sizeof(u32)) {
 				err = -EFAULT;
 				break;
 			}
 		}
-		if (put_user(ALIGN(BITS_TO_BYTES(nlk->ngroups), sizeof(u32)), optlen))
-			err = -EFAULT;
+		opt->optlen = ALIGN(BITS_TO_BYTES(nlk->ngroups), sizeof(u32));
 		netlink_unlock_table();
 		return err;
 	}
@@ -1784,8 +1785,8 @@ static int netlink_getsockopt(struct socket *sock, int level, int optname,
 	len = sizeof(int);
 	val = test_bit(flag, &nlk->flags);
 
-	if (put_user(len, optlen) ||
-	    copy_to_user(optval, &val, len))
+	opt->optlen = len;
+	if (copy_to_iter(&val, len, &opt->iter_out) != len)
 		return -EFAULT;
 
 	return 0;
@@ -2813,7 +2814,7 @@ static const struct proto_ops netlink_ops = {
 	.listen =	sock_no_listen,
 	.shutdown =	sock_no_shutdown,
 	.setsockopt =	netlink_setsockopt,
-	.getsockopt =	netlink_getsockopt,
+	.getsockopt_iter = netlink_getsockopt,
 	.sendmsg =	netlink_sendmsg,
 	.recvmsg =	netlink_recvmsg,
 	.mmap =		sock_no_mmap,

-- 
2.52.0

[PATCH net-next 2/3] vsock: convert to getsockopt_iter

[Breno Leitao]

Convert AF_VSOCK's getsockopt implementation to use the new
getsockopt_iter callback with sockopt_t. The single
vsock_connectible_getsockopt() callback is shared by both
vsock_stream_ops and vsock_seqpacket_ops, so both proto_ops are
updated to use .getsockopt_iter.

Key changes:
- Replace (char __user *optval, int __user *optlen) with sockopt_t *opt
- Use opt->optlen for buffer length (input) and returned size (output)
- Use copy_to_iter() instead of put_user()/copy_to_user()

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 net/vmw_vsock/af_vsock.c | 16 +++++++---------
 1 file changed, 7 insertions(+), 9 deletions(-)

diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
index 44037b066a5ff..d4a97eeb596e6 100644
--- a/net/vmw_vsock/af_vsock.c
+++ b/net/vmw_vsock/af_vsock.c
@@ -155,6 +155,7 @@
 #include <linux/random.h>
 #include <linux/skbuff.h>
 #include <linux/smp.h>
+#include <linux/uio.h>
 #include <linux/socket.h>
 #include <linux/stddef.h>
 #include <linux/sysctl.h>
@@ -2091,8 +2092,7 @@ static int vsock_connectible_setsockopt(struct socket *sock,
 
 static int vsock_connectible_getsockopt(struct socket *sock,
 					int level, int optname,
-					char __user *optval,
-					int __user *optlen)
+					sockopt_t *opt)
 {
 	struct sock *sk = sock->sk;
 	struct vsock_sock *vsk = vsock_sk(sk);
@@ -2110,8 +2110,7 @@ static int vsock_connectible_getsockopt(struct socket *sock,
 	if (level != AF_VSOCK)
 		return -ENOPROTOOPT;
 
-	if (get_user(len, optlen))
-		return -EFAULT;
+	len = opt->optlen;
 
 	memset(&v, 0, sizeof(v));
 
@@ -2142,11 +2141,10 @@ static int vsock_connectible_getsockopt(struct socket *sock,
 		return -EINVAL;
 	if (len > lv)
 		len = lv;
-	if (copy_to_user(optval, &v, len))
+	if (copy_to_iter(&v, len, &opt->iter_out) != len)
 		return -EFAULT;
 
-	if (put_user(len, optlen))
-		return -EFAULT;
+	opt->optlen = len;
 
 	return 0;
 }
@@ -2631,7 +2629,7 @@ static const struct proto_ops vsock_stream_ops = {
 	.listen = vsock_listen,
 	.shutdown = vsock_shutdown,
 	.setsockopt = vsock_connectible_setsockopt,
-	.getsockopt = vsock_connectible_getsockopt,
+	.getsockopt_iter = vsock_connectible_getsockopt,
 	.sendmsg = vsock_connectible_sendmsg,
 	.recvmsg = vsock_connectible_recvmsg,
 	.mmap = sock_no_mmap,
@@ -2653,7 +2651,7 @@ static const struct proto_ops vsock_seqpacket_ops = {
 	.listen = vsock_listen,
 	.shutdown = vsock_shutdown,
 	.setsockopt = vsock_connectible_setsockopt,
-	.getsockopt = vsock_connectible_getsockopt,
+	.getsockopt_iter = vsock_connectible_getsockopt,
 	.sendmsg = vsock_connectible_sendmsg,
 	.recvmsg = vsock_connectible_recvmsg,
 	.mmap = sock_no_mmap,

-- 
2.52.0

[PATCH net-next 3/3] net: selftests: add getsockopt_iter regression tests

[Breno Leitao]

Add a single kselftest covering the proto_ops getsockopt_iter
conversions for AF_NETLINK and AF_VSOCK, using one fixture per protocol:

netlink:

NETLINK_PKTINFO covers the flag-style int path (exact size, oversize
clamp, undersize -EINVAL); NETLINK_LIST_MEMBERSHIPS covers the
size-discovery path that always reports the required buffer length back
via optlen, even when the user buffer is too small to receive any group
bits.

vsock:
SO_VM_SOCKETS_BUFFER_SIZE covers the u64 path (exact size, oversize
clamp, undersize -EINVAL).

Each fixture also exercises an unknown optname and a bogus level so
the returned-length / errno semantics preserved by the sockopt_t
conversion are pinned down.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 tools/testing/selftests/net/Makefile          |   1 +
 tools/testing/selftests/net/getsockopt_iter.c | 213 ++++++++++++++++++++++++++
 2 files changed, 214 insertions(+)

diff --git a/tools/testing/selftests/net/Makefile b/tools/testing/selftests/net/Makefile
index a275ed5840265..baa30287cf222 100644
--- a/tools/testing/selftests/net/Makefile
+++ b/tools/testing/selftests/net/Makefile
@@ -176,6 +176,7 @@ TEST_GEN_PROGS := \
 	bind_timewait \
 	bind_wildcard \
 	epoll_busy_poll \
+	getsockopt_iter \
 	icmp_rfc4884 \
 	ipv6_fragmentation \
 	proc_net_pktgen \
diff --git a/tools/testing/selftests/net/getsockopt_iter.c b/tools/testing/selftests/net/getsockopt_iter.c
new file mode 100644
index 0000000000000..179f9e84926fd
--- /dev/null
+++ b/tools/testing/selftests/net/getsockopt_iter.c
@@ -0,0 +1,213 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Quick test for getsockopt{_iter} tests.
+ *
+ * Each fixture targets one converted protocol and pins down the
+ * returned-length / errno semantics across buffer-size variations,
+ * an unknown optname and a bogus level.
+ *
+ * - netlink: NETLINK_PKTINFO covers the flag-style int path; the
+ *   NETLINK_LIST_MEMBERSHIPS cases cover the size-discovery path
+ *   that always reports the required buffer length back via optlen,
+ *   even when the user buffer is too small to receive any group bits.
+ * - vsock:   SO_VM_SOCKETS_BUFFER_SIZE covers the u64 path.
+ *
+ * Author: Breno Leitao <leitao@debian.org>
+ */
+
+#include <errno.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <string.h>
+#include <unistd.h>
+#include <linux/netlink.h>
+#include <linux/rtnetlink.h>
+#include <linux/vm_sockets.h>
+#include <sys/socket.h>
+#include "kselftest_harness.h"
+
+#ifndef AF_VSOCK
+#define AF_VSOCK 40
+#endif
+
+/* ---------- netlink ---------- */
+
+FIXTURE(netlink)
+{
+	int fd;
+};
+
+FIXTURE_SETUP(netlink)
+{
+	int group = RTNLGRP_LINK;
+
+	self->fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
+	if (self->fd < 0)
+		SKIP(return, "AF_NETLINK socket: %s", strerror(errno));
+
+	/* Joining a multicast group grows nlk->ngroups so the
+	 * NETLINK_LIST_MEMBERSHIPS path has a non-zero size to report.
+	 */
+	if (setsockopt(self->fd, SOL_NETLINK, NETLINK_ADD_MEMBERSHIP,
+		       &group, sizeof(group)) < 0)
+		SKIP(return, "NETLINK_ADD_MEMBERSHIP: %s", strerror(errno));
+}
+
+FIXTURE_TEARDOWN(netlink)
+{
+	if (self->fd >= 0)
+		close(self->fd);
+}
+
+TEST_F(netlink, pktinfo_exact)
+{
+	int val = -1;
+	socklen_t optlen = sizeof(val);
+
+	ASSERT_EQ(0, getsockopt(self->fd, SOL_NETLINK, NETLINK_PKTINFO,
+				&val, &optlen));
+	ASSERT_EQ(sizeof(int), optlen);
+	ASSERT_TRUE(val == 0 || val == 1);
+}
+
+TEST_F(netlink, pktinfo_oversize_clamped)
+{
+	char buf[16] = {};
+	socklen_t optlen = sizeof(buf);
+
+	ASSERT_EQ(0, getsockopt(self->fd, SOL_NETLINK, NETLINK_PKTINFO,
+				buf, &optlen));
+	ASSERT_EQ(sizeof(int), optlen);
+}
+
+TEST_F(netlink, pktinfo_undersize)
+{
+	char buf[2] = {};
+	socklen_t optlen = sizeof(buf);
+
+	ASSERT_EQ(-1, getsockopt(self->fd, SOL_NETLINK, NETLINK_PKTINFO,
+				 buf, &optlen));
+	ASSERT_EQ(EINVAL, errno);
+}
+
+TEST_F(netlink, list_memberships_size_discovery)
+{
+	socklen_t optlen = 0;
+	char dummy;
+
+	ASSERT_EQ(0, getsockopt(self->fd, SOL_NETLINK,
+				NETLINK_LIST_MEMBERSHIPS,
+				&dummy, &optlen));
+	ASSERT_GT(optlen, 0);
+	ASSERT_EQ(0, optlen % sizeof(__u32));
+}
+
+TEST_F(netlink, list_memberships_full_read)
+{
+	__u32 buf[64] = {};
+	socklen_t optlen = sizeof(buf);
+
+	ASSERT_EQ(0, getsockopt(self->fd, SOL_NETLINK,
+				NETLINK_LIST_MEMBERSHIPS,
+				buf, &optlen));
+	ASSERT_GT(optlen, 0);
+	ASSERT_LE(optlen, sizeof(buf));
+	ASSERT_EQ(0, optlen % sizeof(__u32));
+}
+
+TEST_F(netlink, bad_level)
+{
+	int val;
+	socklen_t optlen = sizeof(val);
+
+	ASSERT_EQ(-1, getsockopt(self->fd, SOL_SOCKET + 1, NETLINK_PKTINFO,
+				 &val, &optlen));
+	ASSERT_EQ(ENOPROTOOPT, errno);
+}
+
+TEST_F(netlink, bad_optname)
+{
+	int val;
+	socklen_t optlen = sizeof(val);
+
+	ASSERT_EQ(-1, getsockopt(self->fd, SOL_NETLINK, 0x7fff,
+				 &val, &optlen));
+	ASSERT_EQ(ENOPROTOOPT, errno);
+}
+
+/* ---------- vsock ---------- */
+
+FIXTURE(vsock)
+{
+	int fd;
+};
+
+FIXTURE_SETUP(vsock)
+{
+	self->fd = socket(AF_VSOCK, SOCK_STREAM, 0);
+	if (self->fd < 0)
+		SKIP(return, "AF_VSOCK socket: %s", strerror(errno));
+}
+
+FIXTURE_TEARDOWN(vsock)
+{
+	if (self->fd >= 0)
+		close(self->fd);
+}
+
+TEST_F(vsock, buffer_size_exact)
+{
+	uint64_t val = 0;
+	socklen_t optlen = sizeof(val);
+
+	ASSERT_EQ(0, getsockopt(self->fd, AF_VSOCK,
+				SO_VM_SOCKETS_BUFFER_SIZE,
+				&val, &optlen));
+	ASSERT_EQ(sizeof(uint64_t), optlen);
+	ASSERT_GT(val, 0);
+}
+
+TEST_F(vsock, buffer_size_oversize_clamped)
+{
+	char buf[16] = {};
+	socklen_t optlen = sizeof(buf);
+
+	ASSERT_EQ(0, getsockopt(self->fd, AF_VSOCK,
+				SO_VM_SOCKETS_BUFFER_SIZE,
+				buf, &optlen));
+	ASSERT_EQ(sizeof(uint64_t), optlen);
+}
+
+TEST_F(vsock, buffer_size_undersize)
+{
+	char buf[4] = {};
+	socklen_t optlen = sizeof(buf);
+
+	ASSERT_EQ(-1, getsockopt(self->fd, AF_VSOCK,
+				 SO_VM_SOCKETS_BUFFER_SIZE,
+				 buf, &optlen));
+	ASSERT_EQ(EINVAL, errno);
+}
+
+TEST_F(vsock, bad_level)
+{
+	uint64_t val;
+	socklen_t optlen = sizeof(val);
+
+	ASSERT_EQ(-1, getsockopt(self->fd, SOL_SOCKET + 1,
+				 SO_VM_SOCKETS_BUFFER_SIZE,
+				 &val, &optlen));
+	ASSERT_EQ(ENOPROTOOPT, errno);
+}
+
+TEST_F(vsock, bad_optname)
+{
+	uint64_t val;
+	socklen_t optlen = sizeof(val);
+
+	ASSERT_EQ(-1, getsockopt(self->fd, AF_VSOCK, 0x7fff,
+				 &val, &optlen));
+	ASSERT_EQ(ENOPROTOOPT, errno);
+}
+
+TEST_HARNESS_MAIN

-- 
2.52.0

Re: [PATCH net-next 3/3] net: selftests: add getsockopt_iter regression tests

[Bobby Eshleman]

On Fri, May 01, 2026 at 08:52:53AM -0700, Breno Leitao wrote:
> Add a single kselftest covering the proto_ops getsockopt_iter
> conversions for AF_NETLINK and AF_VSOCK, using one fixture per protocol:
> 
> netlink:
> 
> NETLINK_PKTINFO covers the flag-style int path (exact size, oversize
> clamp, undersize -EINVAL); NETLINK_LIST_MEMBERSHIPS covers the
> size-discovery path that always reports the required buffer length back
> via optlen, even when the user buffer is too small to receive any group
> bits.
> 
> vsock:
> SO_VM_SOCKETS_BUFFER_SIZE covers the u64 path (exact size, oversize
> clamp, undersize -EINVAL).
> 
> Each fixture also exercises an unknown optname and a bogus level so
> the returned-length / errno semantics preserved by the sockopt_t
> conversion are pinned down.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

This all looks good to me. My only thought was
SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW/OLD would be a nice add-on for testing
greater than u64, but probably overkill since you already have
NETLINK_LIST_MEMBERSHIPS.

Code all looks reasonable.

Reviewed-by: Bobby Eshleman <bobbyeshleman@meta.com>

> ---
>  tools/testing/selftests/net/Makefile          |   1 +
>  tools/testing/selftests/net/getsockopt_iter.c | 213 ++++++++++++++++++++++++++
>  2 files changed, 214 insertions(+)
> 
> diff --git a/tools/testing/selftests/net/Makefile b/tools/testing/selftests/net/Makefile
> index a275ed5840265..baa30287cf222 100644
> --- a/tools/testing/selftests/net/Makefile
> +++ b/tools/testing/selftests/net/Makefile
> @@ -176,6 +176,7 @@ TEST_GEN_PROGS := \
>  	bind_timewait \
>  	bind_wildcard \
>  	epoll_busy_poll \
> +	getsockopt_iter \
>  	icmp_rfc4884 \
>  	ipv6_fragmentation \
>  	proc_net_pktgen \
> diff --git a/tools/testing/selftests/net/getsockopt_iter.c b/tools/testing/selftests/net/getsockopt_iter.c
> new file mode 100644
> index 0000000000000..179f9e84926fd
> --- /dev/null
> +++ b/tools/testing/selftests/net/getsockopt_iter.c
> @@ -0,0 +1,213 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Quick test for getsockopt{_iter} tests.
> + *
> + * Each fixture targets one converted protocol and pins down the
> + * returned-length / errno semantics across buffer-size variations,
> + * an unknown optname and a bogus level.
> + *
> + * - netlink: NETLINK_PKTINFO covers the flag-style int path; the
> + *   NETLINK_LIST_MEMBERSHIPS cases cover the size-discovery path
> + *   that always reports the required buffer length back via optlen,
> + *   even when the user buffer is too small to receive any group bits.
> + * - vsock:   SO_VM_SOCKETS_BUFFER_SIZE covers the u64 path.
> + *
> + * Author: Breno Leitao <leitao@debian.org>
> + */
> +
> +#include <errno.h>
> +#include <stdint.h>
> +#include <stdio.h>
> +#include <string.h>
> +#include <unistd.h>
> +#include <linux/netlink.h>
> +#include <linux/rtnetlink.h>
> +#include <linux/vm_sockets.h>
> +#include <sys/socket.h>
> +#include "kselftest_harness.h"
> +
> +#ifndef AF_VSOCK
> +#define AF_VSOCK 40
> +#endif
> +
> +/* ---------- netlink ---------- */
> +
> +FIXTURE(netlink)
> +{
> +	int fd;
> +};
> +
> +FIXTURE_SETUP(netlink)
> +{
> +	int group = RTNLGRP_LINK;
> +
> +	self->fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
> +	if (self->fd < 0)
> +		SKIP(return, "AF_NETLINK socket: %s", strerror(errno));
> +
> +	/* Joining a multicast group grows nlk->ngroups so the
> +	 * NETLINK_LIST_MEMBERSHIPS path has a non-zero size to report.
> +	 */
> +	if (setsockopt(self->fd, SOL_NETLINK, NETLINK_ADD_MEMBERSHIP,
> +		       &group, sizeof(group)) < 0)
> +		SKIP(return, "NETLINK_ADD_MEMBERSHIP: %s", strerror(errno));
> +}
> +
> +FIXTURE_TEARDOWN(netlink)
> +{
> +	if (self->fd >= 0)
> +		close(self->fd);
> +}
> +
> +TEST_F(netlink, pktinfo_exact)
> +{
> +	int val = -1;
> +	socklen_t optlen = sizeof(val);
> +
> +	ASSERT_EQ(0, getsockopt(self->fd, SOL_NETLINK, NETLINK_PKTINFO,
> +				&val, &optlen));
> +	ASSERT_EQ(sizeof(int), optlen);
> +	ASSERT_TRUE(val == 0 || val == 1);
> +}
> +
> +TEST_F(netlink, pktinfo_oversize_clamped)
> +{
> +	char buf[16] = {};
> +	socklen_t optlen = sizeof(buf);
> +
> +	ASSERT_EQ(0, getsockopt(self->fd, SOL_NETLINK, NETLINK_PKTINFO,
> +				buf, &optlen));
> +	ASSERT_EQ(sizeof(int), optlen);
> +}
> +
> +TEST_F(netlink, pktinfo_undersize)
> +{
> +	char buf[2] = {};
> +	socklen_t optlen = sizeof(buf);
> +
> +	ASSERT_EQ(-1, getsockopt(self->fd, SOL_NETLINK, NETLINK_PKTINFO,
> +				 buf, &optlen));
> +	ASSERT_EQ(EINVAL, errno);
> +}
> +
> +TEST_F(netlink, list_memberships_size_discovery)
> +{
> +	socklen_t optlen = 0;
> +	char dummy;
> +
> +	ASSERT_EQ(0, getsockopt(self->fd, SOL_NETLINK,
> +				NETLINK_LIST_MEMBERSHIPS,
> +				&dummy, &optlen));
> +	ASSERT_GT(optlen, 0);
> +	ASSERT_EQ(0, optlen % sizeof(__u32));
> +}
> +
> +TEST_F(netlink, list_memberships_full_read)
> +{
> +	__u32 buf[64] = {};
> +	socklen_t optlen = sizeof(buf);
> +
> +	ASSERT_EQ(0, getsockopt(self->fd, SOL_NETLINK,
> +				NETLINK_LIST_MEMBERSHIPS,
> +				buf, &optlen));
> +	ASSERT_GT(optlen, 0);
> +	ASSERT_LE(optlen, sizeof(buf));
> +	ASSERT_EQ(0, optlen % sizeof(__u32));
> +}
> +
> +TEST_F(netlink, bad_level)
> +{
> +	int val;
> +	socklen_t optlen = sizeof(val);
> +
> +	ASSERT_EQ(-1, getsockopt(self->fd, SOL_SOCKET + 1, NETLINK_PKTINFO,
> +				 &val, &optlen));
> +	ASSERT_EQ(ENOPROTOOPT, errno);
> +}
> +
> +TEST_F(netlink, bad_optname)
> +{
> +	int val;
> +	socklen_t optlen = sizeof(val);
> +
> +	ASSERT_EQ(-1, getsockopt(self->fd, SOL_NETLINK, 0x7fff,
> +				 &val, &optlen));
> +	ASSERT_EQ(ENOPROTOOPT, errno);
> +}
> +
> +/* ---------- vsock ---------- */
> +
> +FIXTURE(vsock)
> +{
> +	int fd;
> +};
> +
> +FIXTURE_SETUP(vsock)
> +{
> +	self->fd = socket(AF_VSOCK, SOCK_STREAM, 0);
> +	if (self->fd < 0)
> +		SKIP(return, "AF_VSOCK socket: %s", strerror(errno));
> +}
> +
> +FIXTURE_TEARDOWN(vsock)
> +{
> +	if (self->fd >= 0)
> +		close(self->fd);
> +}
> +
> +TEST_F(vsock, buffer_size_exact)
> +{
> +	uint64_t val = 0;
> +	socklen_t optlen = sizeof(val);
> +
> +	ASSERT_EQ(0, getsockopt(self->fd, AF_VSOCK,
> +				SO_VM_SOCKETS_BUFFER_SIZE,
> +				&val, &optlen));
> +	ASSERT_EQ(sizeof(uint64_t), optlen);
> +	ASSERT_GT(val, 0);
> +}
> +
> +TEST_F(vsock, buffer_size_oversize_clamped)
> +{
> +	char buf[16] = {};
> +	socklen_t optlen = sizeof(buf);
> +
> +	ASSERT_EQ(0, getsockopt(self->fd, AF_VSOCK,
> +				SO_VM_SOCKETS_BUFFER_SIZE,
> +				buf, &optlen));
> +	ASSERT_EQ(sizeof(uint64_t), optlen);
> +}
> +
> +TEST_F(vsock, buffer_size_undersize)
> +{
> +	char buf[4] = {};
> +	socklen_t optlen = sizeof(buf);
> +
> +	ASSERT_EQ(-1, getsockopt(self->fd, AF_VSOCK,
> +				 SO_VM_SOCKETS_BUFFER_SIZE,
> +				 buf, &optlen));
> +	ASSERT_EQ(EINVAL, errno);
> +}
> +
> +TEST_F(vsock, bad_level)
> +{
> +	uint64_t val;
> +	socklen_t optlen = sizeof(val);
> +
> +	ASSERT_EQ(-1, getsockopt(self->fd, SOL_SOCKET + 1,
> +				 SO_VM_SOCKETS_BUFFER_SIZE,
> +				 &val, &optlen));
> +	ASSERT_EQ(ENOPROTOOPT, errno);
> +}
> +
> +TEST_F(vsock, bad_optname)
> +{
> +	uint64_t val;
> +	socklen_t optlen = sizeof(val);
> +
> +	ASSERT_EQ(-1, getsockopt(self->fd, AF_VSOCK, 0x7fff,
> +				 &val, &optlen));
> +	ASSERT_EQ(ENOPROTOOPT, errno);
> +}
> +
> +TEST_HARNESS_MAIN
> 
> -- 
> 2.52.0
> 

Re: [PATCH net-next 2/3] vsock: convert to getsockopt_iter

[Bobby Eshleman]

On Fri, May 01, 2026 at 08:52:52AM -0700, Breno Leitao wrote:
> Convert AF_VSOCK's getsockopt implementation to use the new
> getsockopt_iter callback with sockopt_t. The single
> vsock_connectible_getsockopt() callback is shared by both
> vsock_stream_ops and vsock_seqpacket_ops, so both proto_ops are
> updated to use .getsockopt_iter.
> 
> Key changes:
> - Replace (char __user *optval, int __user *optlen) with sockopt_t *opt
> - Use opt->optlen for buffer length (input) and returned size (output)
> - Use copy_to_iter() instead of put_user()/copy_to_user()
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>
> ---
>  net/vmw_vsock/af_vsock.c | 16 +++++++---------
>  1 file changed, 7 insertions(+), 9 deletions(-)
> 
> diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
> index 44037b066a5ff..d4a97eeb596e6 100644
> --- a/net/vmw_vsock/af_vsock.c
> +++ b/net/vmw_vsock/af_vsock.c
> @@ -155,6 +155,7 @@
>  #include <linux/random.h>
>  #include <linux/skbuff.h>
>  #include <linux/smp.h>
> +#include <linux/uio.h>
>  #include <linux/socket.h>
>  #include <linux/stddef.h>
>  #include <linux/sysctl.h>
> @@ -2091,8 +2092,7 @@ static int vsock_connectible_setsockopt(struct socket *sock,
>  
>  static int vsock_connectible_getsockopt(struct socket *sock,
>  					int level, int optname,
> -					char __user *optval,
> -					int __user *optlen)
> +					sockopt_t *opt)
>  {
>  	struct sock *sk = sock->sk;
>  	struct vsock_sock *vsk = vsock_sk(sk);
> @@ -2110,8 +2110,7 @@ static int vsock_connectible_getsockopt(struct socket *sock,
>  	if (level != AF_VSOCK)
>  		return -ENOPROTOOPT;
>  
> -	if (get_user(len, optlen))
> -		return -EFAULT;
> +	len = opt->optlen;
>  
>  	memset(&v, 0, sizeof(v));
>  
> @@ -2142,11 +2141,10 @@ static int vsock_connectible_getsockopt(struct socket *sock,
>  		return -EINVAL;
>  	if (len > lv)
>  		len = lv;
> -	if (copy_to_user(optval, &v, len))
> +	if (copy_to_iter(&v, len, &opt->iter_out) != len)
>  		return -EFAULT;
>  
> -	if (put_user(len, optlen))
> -		return -EFAULT;
> +	opt->optlen = len;
>  
>  	return 0;
>  }
> @@ -2631,7 +2629,7 @@ static const struct proto_ops vsock_stream_ops = {
>  	.listen = vsock_listen,
>  	.shutdown = vsock_shutdown,
>  	.setsockopt = vsock_connectible_setsockopt,
> -	.getsockopt = vsock_connectible_getsockopt,
> +	.getsockopt_iter = vsock_connectible_getsockopt,
>  	.sendmsg = vsock_connectible_sendmsg,
>  	.recvmsg = vsock_connectible_recvmsg,
>  	.mmap = sock_no_mmap,
> @@ -2653,7 +2651,7 @@ static const struct proto_ops vsock_seqpacket_ops = {
>  	.listen = vsock_listen,
>  	.shutdown = vsock_shutdown,
>  	.setsockopt = vsock_connectible_setsockopt,
> -	.getsockopt = vsock_connectible_getsockopt,
> +	.getsockopt_iter = vsock_connectible_getsockopt,
>  	.sendmsg = vsock_connectible_sendmsg,
>  	.recvmsg = vsock_connectible_recvmsg,
>  	.mmap = sock_no_mmap,
> 
> -- 
> 2.52.0
> 

Reviewed-by: Bobby Eshleman <bobbyeshleman@meta.com>

Re: [PATCH net-next 2/3] vsock: convert to getsockopt_iter

[David Laight]

On Fri, 1 May 2026 10:32:44 -0700
Bobby Eshleman <bobbyeshleman@gmail.com> wrote:

> On Fri, May 01, 2026 at 08:52:52AM -0700, Breno Leitao wrote:
> > Convert AF_VSOCK's getsockopt implementation to use the new
> > getsockopt_iter callback with sockopt_t. The single
> > vsock_connectible_getsockopt() callback is shared by both
> > vsock_stream_ops and vsock_seqpacket_ops, so both proto_ops are
> > updated to use .getsockopt_iter.
> > 
> > Key changes:
> > - Replace (char __user *optval, int __user *optlen) with sockopt_t *opt
> > - Use opt->optlen for buffer length (input) and returned size (output)
> > - Use copy_to_iter() instead of put_user()/copy_to_user()
> > 
> > Signed-off-by: Breno Leitao <leitao@debian.org>
> > ---
> >  net/vmw_vsock/af_vsock.c | 16 +++++++---------
> >  1 file changed, 7 insertions(+), 9 deletions(-)
> > 
> > diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
> > index 44037b066a5ff..d4a97eeb596e6 100644
> > --- a/net/vmw_vsock/af_vsock.c
> > +++ b/net/vmw_vsock/af_vsock.c
> > @@ -155,6 +155,7 @@
> >  #include <linux/random.h>
> >  #include <linux/skbuff.h>
> >  #include <linux/smp.h>
> > +#include <linux/uio.h>
> >  #include <linux/socket.h>
> >  #include <linux/stddef.h>
> >  #include <linux/sysctl.h>
> > @@ -2091,8 +2092,7 @@ static int vsock_connectible_setsockopt(struct socket *sock,
> >  
> >  static int vsock_connectible_getsockopt(struct socket *sock,
> >  					int level, int optname,
> > -					char __user *optval,
> > -					int __user *optlen)
> > +					sockopt_t *opt)
> >  {
> >  	struct sock *sk = sock->sk;
> >  	struct vsock_sock *vsk = vsock_sk(sk);
> > @@ -2110,8 +2110,7 @@ static int vsock_connectible_getsockopt(struct socket *sock,
> >  	if (level != AF_VSOCK)
> >  		return -ENOPROTOOPT;
> >  
> > -	if (get_user(len, optlen))
> > -		return -EFAULT;
> > +	len = opt->optlen;
> >  
> >  	memset(&v, 0, sizeof(v));
> >  
> > @@ -2142,11 +2141,10 @@ static int vsock_connectible_getsockopt(struct socket *sock,
> >  		return -EINVAL;
> >  	if (len > lv)
> >  		len = lv;
> > -	if (copy_to_user(optval, &v, len))
> > +	if (copy_to_iter(&v, len, &opt->iter_out) != len)

I'd wrap that as copy_to_sockopt(&v, len, opt).
or to make the edits easier: copy_to_sockopt(opt, &v, len).
Then if someone decides to change the implementation none of the call
sites need changing.

-- David

> >  		return -EFAULT;
> >  
> > -	if (put_user(len, optlen))
> > -		return -EFAULT;
> > +	opt->optlen = len;
> >  
> >  	return 0;
> >  }
> > @@ -2631,7 +2629,7 @@ static const struct proto_ops vsock_stream_ops = {
> >  	.listen = vsock_listen,
> >  	.shutdown = vsock_shutdown,
> >  	.setsockopt = vsock_connectible_setsockopt,
> > -	.getsockopt = vsock_connectible_getsockopt,
> > +	.getsockopt_iter = vsock_connectible_getsockopt,
> >  	.sendmsg = vsock_connectible_sendmsg,
> >  	.recvmsg = vsock_connectible_recvmsg,
> >  	.mmap = sock_no_mmap,
> > @@ -2653,7 +2651,7 @@ static const struct proto_ops vsock_seqpacket_ops = {
> >  	.listen = vsock_listen,
> >  	.shutdown = vsock_shutdown,
> >  	.setsockopt = vsock_connectible_setsockopt,
> > -	.getsockopt = vsock_connectible_getsockopt,
> > +	.getsockopt_iter = vsock_connectible_getsockopt,
> >  	.sendmsg = vsock_connectible_sendmsg,
> >  	.recvmsg = vsock_connectible_recvmsg,
> >  	.mmap = sock_no_mmap,
> > 
> > -- 
> > 2.52.0
> >   
> 
> Reviewed-by: Bobby Eshleman <bobbyeshleman@meta.com>
> 

Re: [PATCH net-next 1/3] netlink: convert to getsockopt_iter

[Stanislav Fomichev]

On 05/01, Breno Leitao wrote:
> Convert AF_NETLINK's getsockopt implementation to use the new
> getsockopt_iter callback with sockopt_t.
> 
> Key changes:
> - Replace (char __user *optval, int __user *optlen) with sockopt_t *opt
> - Use opt->optlen for buffer length (input) and returned size (output)
> - Use copy_to_iter() instead of put_user()/copy_to_user()
> - For NETLINK_LIST_MEMBERSHIPS: walk the groups bitmap and emit each
>   u32 sequentially via copy_to_iter(), then set opt->optlen to the
>   total size required (ALIGN(BITS_TO_BYTES(ngroups), sizeof(u32))).
>   The wrapper writes opt->optlen back to userspace even on partial
>   failure, preserving the existing API that lets userspace discover
>   the needed allocation size.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

Acked-by: Stanislav Fomichev <sdf@fomichev.me>

Re: [PATCH net-next 2/3] vsock: convert to getsockopt_iter

[Stanislav Fomichev]

On 05/01, Breno Leitao wrote:
> Convert AF_VSOCK's getsockopt implementation to use the new
> getsockopt_iter callback with sockopt_t. The single
> vsock_connectible_getsockopt() callback is shared by both
> vsock_stream_ops and vsock_seqpacket_ops, so both proto_ops are
> updated to use .getsockopt_iter.
> 
> Key changes:
> - Replace (char __user *optval, int __user *optlen) with sockopt_t *opt
> - Use opt->optlen for buffer length (input) and returned size (output)
> - Use copy_to_iter() instead of put_user()/copy_to_user()
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

Acked-by: Stanislav Fomichev <sdf@fomichev.me>

Re: [PATCH net-next 3/3] net: selftests: add getsockopt_iter regression tests

[Stanislav Fomichev]

On 05/01, Breno Leitao wrote:
> Add a single kselftest covering the proto_ops getsockopt_iter
> conversions for AF_NETLINK and AF_VSOCK, using one fixture per protocol:
> 
> netlink:
> 
> NETLINK_PKTINFO covers the flag-style int path (exact size, oversize
> clamp, undersize -EINVAL); NETLINK_LIST_MEMBERSHIPS covers the
> size-discovery path that always reports the required buffer length back
> via optlen, even when the user buffer is too small to receive any group
> bits.
> 
> vsock:
> SO_VM_SOCKETS_BUFFER_SIZE covers the u64 path (exact size, oversize
> clamp, undersize -EINVAL).
> 
> Each fixture also exercises an unknown optname and a bogus level so
> the returned-length / errno semantics preserved by the sockopt_t
> conversion are pinned down.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

A few nits if you happen to respin (or for similar tests in the future):
- add ASSERT_EQ(optlen, xxx), won't hurt
- christmas tree (although not sure how much we care, you'll have to move
   optlen = sizeof() assignment)

Acked-by: Stanislav Fomichev <sdf@fomichev.me>

Re: [PATCH net-next 0/3] net: Convert AF_NETLINK and AF_VSOCK to getsockopt_iter API

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Fri, 01 May 2026 08:52:50 -0700 you wrote:
> Continue the work to convert protocols to the new getsockopt_iter API.
> 
> Convert AF_NETLINK and AF_VSOCK getsockopt implementations to the new
> sockopt_t/getsockopt_iter API, and add kselftests that verify the size
> and errno semantics are preserved across the conversion.
> 
> I chose these two socket families because they are probably one of the
> most used  protocols,, ensuring that any potential bugs will be
> discovered and reported quickly.
> 
> [...]

Here is the summary with links:
  - [net-next,1/3] netlink: convert to getsockopt_iter
    https://git.kernel.org/netdev/net-next/c/390bf43b7788
  - [net-next,2/3] vsock: convert to getsockopt_iter
    https://git.kernel.org/netdev/net-next/c/e21bf72954df
  - [net-next,3/3] net: selftests: add getsockopt_iter regression tests
    https://git.kernel.org/netdev/net-next/c/d39887f55d8e

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html