[PATCH net-next v2 0/4] net: convert UDP getsockopt to sockopt_t

[PATCH net-next v2 0/4] net: convert UDP getsockopt to sockopt_t

[Breno Leitao]

The leaf proto_ops getsockopt callbacks have been moving to the new
getsockopt_iter()/sockopt_t interface.

I was trying to get SMC into getsockop and retire .getsockopt, but,
I found the best approach is to keep converting other protocols.

This series starts the same conversion one layer down, at the struct proto
getsockopt path, beginning with UDP.

Example of the current code.

	static int udp_getsockopt(struct sock *sk, int level, int optname,
				char __user *optval, int __user *optlen)
	{
		if (level == SOL_UDP)
			return udp_lib_getsockopt(sk, level, optname, optval, optlen);
		return ip_getsockopt(sk, level, optname, optval, optlen);
	}

We want udp_getsockopt to go to .getsockopt_iter, and there are two
approaches in this case:

1) Create a patchset that moves both of them to getsockopt_iter, which
   is will be a huge change (ip_getsockopt() is used in many places)

2) Break this down, and transform from bottoms up. First
   udp_lib_getsockopt() up to the point we can easily convert
   others, such as ip_getsockopt().

I am taking the approach 2), so, the intermediate code will be something
like:

static int udp_getsockopt(struct sock *sk, int level, int optname,
                          char __user *optval, int __user *optlen)
{
        sockopt_t opt;
        int err;

        if (level != SOL_UDP)
                return ip_getsockopt(sk, level, optname, optval, optlen);

	// Convert optlen/optval in sockopt // (first patch)

        err = udp_lib_getsockopt(sk, level, optname, &opt);
}

The work is bottom-up and mergeable in small steps: a protocol's inner
getsockopt helper is switched to sockopt_t behind its existing thin
__user wrapper, one patch at a time.

Once every inner helper speaks sockopt_t, a later series flips the shared
struct proto.getsockopt and inet_connection_sock_af_ops.getsockopt signatures
and drops the transitional wrappers.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
Changes in v2:
- Check for invalid `len` at call site (Stanlislav)
- Do not rewrite `optlen` on error path (Sashiko)
- Converted ipv4 raw
- Expanded the selftest to test this new helper.
- Link to v1: https://lore.kernel.org/r/20260612-getsockopt_phase2-v1-0-7b01f1f5d106@debian.org

---
Breno Leitao (4):
      net: add sockopt_init_user() for getsockopt conversion
      udp: convert udp_lib_getsockopt to sockopt_t
      ipv4: raw: convert do_raw_getsockopt to sockopt_t
      selftests: net: getsockopt_iter: add raw ICMP_FILTER coverage

 include/linux/net.h                           | 23 +++++++
 include/net/udp.h                             |  2 +-
 net/ipv4/raw.c                                | 41 ++++++-----
 net/ipv4/udp.c                                | 39 ++++++++---
 net/ipv6/udp.c                                | 19 +++++-
 tools/testing/selftests/net/getsockopt_iter.c | 97 +++++++++++++++++++++++++++
 6 files changed, 191 insertions(+), 30 deletions(-)
---
base-commit: c8459ee2fef502d6ef6c063751c33d9ac7943eab
change-id: 20260611-getsockopt_phase2-cd495a0115ca

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

[PATCH net-next v2 1/4] net: add sockopt_init_user() for getsockopt conversion

[Breno Leitao]

Add a helper that initializes a user-backed sockopt_t from the (optval,
optlen) __user pair passed to a getsockopt() callback.

It is used by transitional __user getsockopt wrappers while the
proto-layer getsockopt callbacks are converted to take a sockopt_t, and
is removed once the conversion is complete.

The goal is to help to convert leafs. Example:

 sock_common_getsockopt(... char __user *optval, int __user *optlen)
      → udp_getsockopt(sk, level, optname, optval__user, optlen__user)
               → udp_lib_getsockopt(sk, level, optname, &opt)   /* needs a sockopt_t */

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 include/linux/net.h | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/include/linux/net.h b/include/linux/net.h
index f268f395ce473..277188a40c72e 100644
--- a/include/linux/net.h
+++ b/include/linux/net.h
@@ -47,6 +47,29 @@ typedef struct sockopt {
 	int optlen;
 } sockopt_t;
 
+/*
+ * Initialize a user-backed sockopt_t from the (optval, optlen) __user pair of
+ * a getsockopt() callback. Used by transitional __user getsockopt wrappers
+ * while the proto-layer callbacks are converted to take a sockopt_t; the
+ * caller writes opt->optlen back to the user optlen after the callback.
+ */
+static inline int sockopt_init_user(sockopt_t *opt, char __user *optval,
+				    int __user *optlen)
+{
+	int len;
+
+	if (get_user(len, optlen))
+		return -EFAULT;
+	if (len < 0)
+		return -EINVAL;
+
+	iov_iter_ubuf(&opt->iter_out, ITER_DEST, optval, len);
+	iov_iter_ubuf(&opt->iter_in, ITER_SOURCE, optval, len);
+	opt->optlen = len;
+
+	return 0;
+}
+
 struct poll_table_struct;
 struct pipe_inode_info;
 struct inode;

-- 
2.53.0-Meta

Re: [PATCH net-next v2 1/4] net: add sockopt_init_user() for getsockopt conversion

[Stanislav Fomichev]

On 06/30, Breno Leitao wrote:
> Add a helper that initializes a user-backed sockopt_t from the (optval,
> optlen) __user pair passed to a getsockopt() callback.
> 
> It is used by transitional __user getsockopt wrappers while the
> proto-layer getsockopt callbacks are converted to take a sockopt_t, and
> is removed once the conversion is complete.
> 
> The goal is to help to convert leafs. Example:
> 
>  sock_common_getsockopt(... char __user *optval, int __user *optlen)
>       → udp_getsockopt(sk, level, optname, optval__user, optlen__user)
>                → udp_lib_getsockopt(sk, level, optname, &opt)   /* needs a sockopt_t */
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

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

[PATCH net-next v2 2/4] udp: convert udp_lib_getsockopt to sockopt_t

[Breno Leitao]

In preparation for converting the proto-layer getsockopt callbacks to the
sockopt_t interface, switch udp_lib_getsockopt() to take a sockopt_t.

The thin udp_getsockopt()/udpv6_getsockopt() wrappers keep their __user
signature for now: they build a user-backed sockopt_t with
sockopt_init_user(), call the helper, and write the returned length back
to optlen. The helper uses copy_to_iter() instead of copy_to_user().
No functional change.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 include/net/udp.h |  2 +-
 net/ipv4/udp.c    | 39 +++++++++++++++++++++++++++++----------
 net/ipv6/udp.c    | 19 ++++++++++++++++---
 3 files changed, 46 insertions(+), 14 deletions(-)

diff --git a/include/net/udp.h b/include/net/udp.h
index 8262e2b215b4e..1fee17274745f 100644
--- a/include/net/udp.h
+++ b/include/net/udp.h
@@ -430,7 +430,7 @@ struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb,
 				       netdev_features_t features,
 				       bool is_ipv6);
 int udp_lib_getsockopt(struct sock *sk, int level, int optname,
-		       char __user *optval, int __user *optlen);
+		       sockopt_t *opt);
 int udp_lib_setsockopt(struct sock *sk, int level, int optname,
 		       sockptr_t optval, unsigned int optlen,
 		       int (*push_pending_frames)(struct sock *));
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index 70f6cbd4ef73b..59248a59358ca 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -76,6 +76,7 @@
 
 #include <linux/bpf-cgroup.h>
 #include <linux/uaccess.h>
+#include <linux/uio.h>
 #include <asm/ioctls.h>
 #include <linux/memblock.h>
 #include <linux/highmem.h>
@@ -2995,14 +2996,13 @@ static int udp_setsockopt(struct sock *sk, int level, int optname, sockptr_t opt
 }
 
 int udp_lib_getsockopt(struct sock *sk, int level, int optname,
-		       char __user *optval, int __user *optlen)
+		       sockopt_t *opt)
 {
 	struct udp_sock *up = udp_sk(sk);
 	int val, len;
 
-	if (get_user(len, optlen))
-		return -EFAULT;
-
+	len = opt->optlen;
+	/* keep the check so direct sockopt_t callers stay covered. */
 	if (len < 0)
 		return -EINVAL;
 
@@ -3037,9 +3037,8 @@ int udp_lib_getsockopt(struct sock *sk, int level, int optname,
 		return -ENOPROTOOPT;
 	}
 
-	if (put_user(len, optlen))
-		return -EFAULT;
-	if (copy_to_user(optval, &val, len))
+	opt->optlen = len;
+	if (copy_to_iter(&val, len, &opt->iter_out) != len)
 		return -EFAULT;
 	return 0;
 }
@@ -3047,9 +3046,29 @@ int udp_lib_getsockopt(struct sock *sk, int level, int optname,
 static int udp_getsockopt(struct sock *sk, int level, int optname,
 			  char __user *optval, int __user *optlen)
 {
-	if (level == SOL_UDP)
-		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
-	return ip_getsockopt(sk, level, optname, optval, optlen);
+	sockopt_t opt;
+	int err;
+
+	/*
+	 * keep the old __user pointers, until ip_getsockopt() moves
+	 * to sockopt_t
+	 */
+	if (level != SOL_UDP)
+		return ip_getsockopt(sk, level, optname, optval, optlen);
+
+	err = sockopt_init_user(&opt, optval, optlen);
+	if (err)
+		return err;
+
+	err = udp_lib_getsockopt(sk, level, optname, &opt);
+	if (err)
+		return err;
+
+	/* optval was written by copy_to_iter() in udp_lib_getsockopt() */
+	if (put_user(opt.optlen, optlen))
+		return -EFAULT;
+
+	return 0;
 }
 
 /**
diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
index 15e032194eccc..392e18b970454 100644
--- a/net/ipv6/udp.c
+++ b/net/ipv6/udp.c
@@ -1826,9 +1826,22 @@ static int udpv6_setsockopt(struct sock *sk, int level, int optname,
 static int udpv6_getsockopt(struct sock *sk, int level, int optname,
 			    char __user *optval, int __user *optlen)
 {
-	if (level == SOL_UDP)
-		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
-	return ipv6_getsockopt(sk, level, optname, optval, optlen);
+	sockopt_t opt;
+	int err;
+
+	if (level != SOL_UDP)
+		return ipv6_getsockopt(sk, level, optname, optval, optlen);
+
+	err = sockopt_init_user(&opt, optval, optlen);
+	if (err)
+		return err;
+
+	err = udp_lib_getsockopt(sk, level, optname, &opt);
+	if (err)
+		return err;
+	if (put_user(opt.optlen, optlen))
+		return -EFAULT;
+	return 0;
 }
 
 

-- 
2.53.0-Meta

Re: [PATCH net-next v2 2/4] udp: convert udp_lib_getsockopt to sockopt_t

[Stanislav Fomichev]

On 06/30, Breno Leitao wrote:
> In preparation for converting the proto-layer getsockopt callbacks to the
> sockopt_t interface, switch udp_lib_getsockopt() to take a sockopt_t.
> 
> The thin udp_getsockopt()/udpv6_getsockopt() wrappers keep their __user
> signature for now: they build a user-backed sockopt_t with
> sockopt_init_user(), call the helper, and write the returned length back
> to optlen. The helper uses copy_to_iter() instead of copy_to_user().
> No functional change.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

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

[PATCH net-next v2 3/4] ipv4: raw: convert do_raw_getsockopt to sockopt_t

[Breno Leitao]

Continue converting the proto-layer getsockopt callbacks to the sockopt_t
interface, switching do_raw_getsockopt() and its raw_geticmpfilter()
helper to take a sockopt_t.

The thin raw_getsockopt() wrapper keeps its __user signature for now: it
builds a user-backed sockopt_t with sockopt_init_user(), calls the helper,
and writes the returned length back to optlen. The helper uses
copy_to_iter() instead of copy_to_user(). No functional change.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 net/ipv4/raw.c | 41 +++++++++++++++++++++++++----------------
 1 file changed, 25 insertions(+), 16 deletions(-)

diff --git a/net/ipv4/raw.c b/net/ipv4/raw.c
index e9fbab6ad9146..2aebaf8297e04 100644
--- a/net/ipv4/raw.c
+++ b/net/ipv4/raw.c
@@ -809,23 +809,18 @@ static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen)
 	return 0;
 }
 
-static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
+static int raw_geticmpfilter(struct sock *sk, sockopt_t *opt)
 {
-	int len, ret = -EFAULT;
+	int len = opt->optlen;
 
-	if (get_user(len, optlen))
-		goto out;
-	ret = -EINVAL;
 	if (len < 0)
-		goto out;
+		return -EINVAL;
 	if (len > sizeof(struct icmp_filter))
 		len = sizeof(struct icmp_filter);
-	ret = -EFAULT;
-	if (put_user(len, optlen) ||
-	    copy_to_user(optval, &raw_sk(sk)->filter, len))
-		goto out;
-	ret = 0;
-out:	return ret;
+	opt->optlen = len;
+	if (copy_to_iter(&raw_sk(sk)->filter, len, &opt->iter_out) != len)
+		return -EFAULT;
+	return 0;
 }
 
 static int do_raw_setsockopt(struct sock *sk, int optname,
@@ -848,14 +843,13 @@ static int raw_setsockopt(struct sock *sk, int level, int optname,
 	return do_raw_setsockopt(sk, optname, optval, optlen);
 }
 
-static int do_raw_getsockopt(struct sock *sk, int optname,
-			     char __user *optval, int __user *optlen)
+static int do_raw_getsockopt(struct sock *sk, int optname, sockopt_t *opt)
 {
 	if (optname == ICMP_FILTER) {
 		if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
 			return -EOPNOTSUPP;
 		else
-			return raw_geticmpfilter(sk, optval, optlen);
+			return raw_geticmpfilter(sk, opt);
 	}
 	return -ENOPROTOOPT;
 }
@@ -863,9 +857,24 @@ static int do_raw_getsockopt(struct sock *sk, int optname,
 static int raw_getsockopt(struct sock *sk, int level, int optname,
 			  char __user *optval, int __user *optlen)
 {
+	sockopt_t opt;
+	int err;
+
 	if (level != SOL_RAW)
 		return ip_getsockopt(sk, level, optname, optval, optlen);
-	return do_raw_getsockopt(sk, optname, optval, optlen);
+
+	err = sockopt_init_user(&opt, optval, optlen);
+	if (err)
+		return err;
+
+	err = do_raw_getsockopt(sk, optname, &opt);
+	if (err)
+		return err;
+
+	if (put_user(opt.optlen, optlen))
+		return -EFAULT;
+
+	return 0;
 }
 
 static int raw_ioctl(struct sock *sk, int cmd, int *karg)

-- 
2.53.0-Meta

Re: [PATCH net-next v2 3/4] ipv4: raw: convert do_raw_getsockopt to sockopt_t

[Stanislav Fomichev]

On 06/30, Breno Leitao wrote:
> Continue converting the proto-layer getsockopt callbacks to the sockopt_t
> interface, switching do_raw_getsockopt() and its raw_geticmpfilter()
> helper to take a sockopt_t.
> 
> The thin raw_getsockopt() wrapper keeps its __user signature for now: it
> builds a user-backed sockopt_t with sockopt_init_user(), calls the helper,
> and writes the returned length back to optlen. The helper uses
> copy_to_iter() instead of copy_to_user(). No functional change.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

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

[PATCH net-next v2 4/4] selftests: net: getsockopt_iter: add raw ICMP_FILTER coverage

[Breno Leitao]

Exercise the raw getsockopt path now backed by sockopt_t. ICMP_FILTER
returns a fixed-size struct and, unlike the int/u64 options already
covered, clamps the length down to the user buffer on a short read
instead of failing, so check that semantic explicitly along with the
exact and oversized cases, the -EOPNOTSUPP path on a non-ICMP raw
socket, and an unknown optname.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 tools/testing/selftests/net/getsockopt_iter.c | 97 +++++++++++++++++++++++++++
 1 file changed, 97 insertions(+)

diff --git a/tools/testing/selftests/net/getsockopt_iter.c b/tools/testing/selftests/net/getsockopt_iter.c
index 209569354d0e3..fe5a5268bc34e 100644
--- a/tools/testing/selftests/net/getsockopt_iter.c
+++ b/tools/testing/selftests/net/getsockopt_iter.c
@@ -11,6 +11,8 @@
  *   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.
+ * - raw:     ICMP_FILTER covers a fixed-size struct payload that clamps
+ *            the length down on a short buffer instead of failing.
  *
  * Author: Breno Leitao <leitao@debian.org>
  */
@@ -24,12 +26,20 @@
 #include <linux/rtnetlink.h>
 #include <linux/time_types.h>
 #include <linux/vm_sockets.h>
+#include <linux/icmp.h>
+#include <netinet/in.h>
 #include <sys/socket.h>
 #include "kselftest_harness.h"
 
 #ifndef AF_VSOCK
 #define AF_VSOCK 40
 #endif
+#ifndef SOL_RAW
+#define SOL_RAW 255
+#endif
+#ifndef ICMP_FILTER
+#define ICMP_FILTER 1
+#endif
 
 /* ---------- netlink ---------- */
 
@@ -297,4 +307,91 @@ TEST_F(vsock, connect_timeout_old_exact)
 	ASSERT_EQ(sizeof(tv), optlen);
 }
 
+/* ---------- raw (ipv4) ---------- */
+
+FIXTURE(raw)
+{
+	int fd;
+};
+
+FIXTURE_SETUP(raw)
+{
+	struct icmp_filter filt = { .data = 0xdeadbeef };
+
+	self->fd = socket(AF_INET, SOCK_RAW, IPPROTO_ICMP);
+	if (self->fd < 0)
+		SKIP(return, "SOCK_RAW/ICMP socket: %s", strerror(errno));
+
+	if (setsockopt(self->fd, SOL_RAW, ICMP_FILTER, &filt, sizeof(filt)) < 0)
+		SKIP(return, "set ICMP_FILTER: %s", strerror(errno));
+}
+
+FIXTURE_TEARDOWN(raw)
+{
+	if (self->fd >= 0)
+		close(self->fd);
+}
+
+TEST_F(raw, icmpfilter_exact)
+{
+	struct icmp_filter filt = {};
+	socklen_t optlen = sizeof(filt);
+
+	ASSERT_EQ(0, getsockopt(self->fd, SOL_RAW, ICMP_FILTER,
+				&filt, &optlen));
+	ASSERT_EQ(sizeof(filt), optlen);
+	ASSERT_EQ(0xdeadbeef, filt.data);
+}
+
+TEST_F(raw, icmpfilter_oversize_clamped)
+{
+	char buf[16] = {};
+	socklen_t optlen = sizeof(buf);
+
+	ASSERT_EQ(0, getsockopt(self->fd, SOL_RAW, ICMP_FILTER,
+				buf, &optlen));
+	ASSERT_EQ(sizeof(struct icmp_filter), optlen);
+}
+
+/* Unlike the int/u64 options above, ICMP_FILTER clamps the length down
+ * to the user buffer instead of returning EINVAL: a short buffer
+ * succeeds and reports the truncated length back via optlen.
+ */
+TEST_F(raw, icmpfilter_undersize_clamped)
+{
+	char buf[2] = {};
+	socklen_t optlen = sizeof(buf);
+
+	ASSERT_EQ(0, getsockopt(self->fd, SOL_RAW, ICMP_FILTER,
+				buf, &optlen));
+	ASSERT_EQ(sizeof(buf), optlen);
+}
+
+TEST_F(raw, icmpfilter_wrong_proto)
+{
+	struct icmp_filter filt;
+	socklen_t optlen = sizeof(filt);
+	int fd;
+
+	fd = socket(AF_INET, SOCK_RAW, IPPROTO_UDP);
+	if (fd < 0)
+		SKIP(return, "SOCK_RAW/UDP socket: %s", strerror(errno));
+
+	ASSERT_EQ(-1, getsockopt(fd, SOL_RAW, ICMP_FILTER, &filt, &optlen));
+	ASSERT_EQ(EOPNOTSUPP, errno);
+	close(fd);
+}
+
+TEST_F(raw, bad_optname)
+{
+	socklen_t optlen;
+	int val;
+
+	optlen = sizeof(val);
+
+	ASSERT_EQ(-1, getsockopt(self->fd, SOL_RAW, 0x7fff, &val, &optlen));
+	ASSERT_EQ(ENOPROTOOPT, errno);
+	ASSERT_EQ(sizeof(val), optlen);
+}
+
 TEST_HARNESS_MAIN

-- 
2.53.0-Meta

Re: [PATCH net-next v2 4/4] selftests: net: getsockopt_iter: add raw ICMP_FILTER coverage

[Stanislav Fomichev]

On 06/30, Breno Leitao wrote:
> Exercise the raw getsockopt path now backed by sockopt_t. ICMP_FILTER
> returns a fixed-size struct and, unlike the int/u64 options already
> covered, clamps the length down to the user buffer on a short read
> instead of failing, so check that semantic explicitly along with the
> exact and oversized cases, the -EOPNOTSUPP path on a non-ICMP raw
> socket, and an unknown optname.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

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