io_uring_cmd_sock() originally supported two ioctl-based cmd_op
operations. Over time, additional operations were added with tail calls
to their helpers.

This approach resulted in the new operations sharing an ioctl check
with the original operations.

io_uring_cmd_sock() now supports 6 operations, so let's move the
implementation of the original two into their own helper, reducing
io_uring_cmd_sock() to a simple dispatcher.

Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
---

Jens, I'm used to net -> net-next taking a week, as it only happens
through Linus' tree.

 io_uring/cmd_net.c | 34 +++++++++++++++++-----------------
 1 file changed, 17 insertions(+), 17 deletions(-)

diff --git a/io_uring/cmd_net.c b/io_uring/cmd_net.c
index 57ddaf874611..56696c4baad1 100644
--- a/io_uring/cmd_net.c
+++ b/io_uring/cmd_net.c
@@ -7,6 +7,21 @@
 #include "uring_cmd.h"
 #include "io_uring.h"
 
+static int io_uring_cmd_get_sock_ioctl(struct socket *sock, int op)
+{
+	struct sock *sk = sock->sk;
+	struct proto *prot = READ_ONCE(sk->sk_prot);
+	int ret, arg = 0;
+
+	if (!prot || !prot->ioctl)
+		return -EOPNOTSUPP;
+
+	ret = prot->ioctl(sk, op, &arg);
+	if (ret)
+		return ret;
+	return arg;
+}
+
 static inline int io_uring_cmd_getsockopt(struct socket *sock,
 					  struct io_uring_cmd *cmd,
 					  unsigned int issue_flags)
@@ -156,27 +171,12 @@ static int io_uring_cmd_getsockname(struct socket *sock,
 int io_uring_cmd_sock(struct io_uring_cmd *cmd, unsigned int issue_flags)
 {
 	struct socket *sock = cmd->file->private_data;
-	struct sock *sk = sock->sk;
-	struct proto *prot = READ_ONCE(sk->sk_prot);
-	int ret, arg = 0;
 
 	switch (cmd->cmd_op) {
 	case SOCKET_URING_OP_SIOCINQ:
-		if (!prot || !prot->ioctl)
-			return -EOPNOTSUPP;
-
-		ret = prot->ioctl(sk, SIOCINQ, &arg);
-		if (ret)
-			return ret;
-		return arg;
+		return io_uring_cmd_get_sock_ioctl(sock, SIOCINQ);
 	case SOCKET_URING_OP_SIOCOUTQ:
-		if (!prot || !prot->ioctl)
-			return -EOPNOTSUPP;
-
-		ret = prot->ioctl(sk, SIOCOUTQ, &arg);
-		if (ret)
-			return ret;
-		return arg;
+		return io_uring_cmd_get_sock_ioctl(sock, SIOCOUTQ);
 	case SOCKET_URING_OP_GETSOCKOPT:
 		return io_uring_cmd_getsockopt(sock, cmd, issue_flags);
 	case SOCKET_URING_OP_SETSOCKOPT:

base-commit: d7861b7cd05a4c02dfa8015048be6821a1af7c5a
-- 
2.51.0

On Mon, 16 Feb 2026 16:03:53 +0000, Asbjørn Sloth Tønnesen wrote:
> io_uring_cmd_sock() originally supported two ioctl-based cmd_op
> operations. Over time, additional operations were added with tail calls
> to their helpers.
> 
> This approach resulted in the new operations sharing an ioctl check
> with the original operations.
> 
> [...]

Applied, thanks!

[1/1] io_uring/cmd_net: split ioctl code out of io_uring_cmd_sock()
      commit: 4148bacb19a84ae2c420d13ad1f4d77981974ba8

Best regards,
-- 
Jens Axboe

On 2/16/26 9:03 AM, Asbjørn Sloth Tønnesen wrote:
> io_uring_cmd_sock() originally supported two ioctl-based cmd_op
> operations. Over time, additional operations were added with tail calls
> to their helpers.
> 
> This approach resulted in the new operations sharing an ioctl check
> with the original operations.
> 
> io_uring_cmd_sock() now supports 6 operations, so let's move the
> implementation of the original two into their own helper, reducing
> io_uring_cmd_sock() to a simple dispatcher.
> 
> Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
> ---
> 
> Jens, I'm used to net -> net-next taking a week, as it only happens
> through Linus' tree.

Looks good to me - since this is just a cleanup, let's defer to 7.1.
I'll kick that off in a week or so, at which point I'll pick this one
up too.

-- 
Jens Axboe

On 2/16/26 5:46 PM, Jens Axboe wrote:
> On 2/16/26 9:03 AM, Asbjørn Sloth Tønnesen wrote:
>> io_uring_cmd_sock() originally supported two ioctl-based cmd_op
>> operations. Over time, additional operations were added with tail calls
>> to their helpers.
>>
>> This approach resulted in the new operations sharing an ioctl check
>> with the original operations.
>>
>> io_uring_cmd_sock() now supports 6 operations, so let's move the
>> implementation of the original two into their own helper, reducing
>> io_uring_cmd_sock() to a simple dispatcher.
>>
>> Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
>> ---
>>
>> Jens, I'm used to net -> net-next taking a week, as it only happens
>> through Linus' tree.
> 
> Looks good to me - since this is just a cleanup, let's defer to 7.1.
> I'll kick that off in a week or so, at which point I'll pick this one
> up too.

Thank you, and sorry for posting during the merge window, I always
intended this for 7.1. I just took it as an invite that you merged into
for-next right after committing my fix to io_uring-7.0, given what I
wrote earlier in the RFC: "I plan to submit v1 once that patch
propagates to for-next.". I wasn't expecting it to happen that quickly.

On 2/16/26 11:31 AM, Asbj?rn Sloth T?nnesen wrote:
> On 2/16/26 5:46 PM, Jens Axboe wrote:
>> On 2/16/26 9:03 AM, Asbj?rn Sloth T?nnesen wrote:
>>> io_uring_cmd_sock() originally supported two ioctl-based cmd_op
>>> operations. Over time, additional operations were added with tail calls
>>> to their helpers.
>>>
>>> This approach resulted in the new operations sharing an ioctl check
>>> with the original operations.
>>>
>>> io_uring_cmd_sock() now supports 6 operations, so let's move the
>>> implementation of the original two into their own helper, reducing
>>> io_uring_cmd_sock() to a simple dispatcher.
>>>
>>> Signed-off-by: Asbj?rn Sloth T?nnesen <ast@fiberby.net>
>>> ---
>>>
>>> Jens, I'm used to net -> net-next taking a week, as it only happens
>>> through Linus' tree.
>>
>> Looks good to me - since this is just a cleanup, let's defer to 7.1.
>> I'll kick that off in a week or so, at which point I'll pick this one
>> up too.
> 
> Thank you, and sorry for posting during the merge window, I always
> intended this for 7.1. I just took it as an invite that you merged into
> for-next right after committing my fix to io_uring-7.0, given what I
> wrote earlier in the RFC: "I plan to submit v1 once that patch
> propagates to for-next.". I wasn't expecting it to happen that quickly.

I do it a bit differently than netdev - my for-next is everything queued
for this release, and the next. You don't need to resend patch headed
for 7.1, unless I for some reason forget to merge it... But I tend to
try and tag these things so I don't forget them. It's a bit easier post
-rc1/2 time as the for-7.x/io_uring branch does exist already and it can
just go straight there.

-- 
Jens Axboe