[v3,net-next,02/15] socket: Pass hold_net flag to __sock_create().

Commit Message

Kuniyuki Iwashima Dec. 13, 2024, 9:21 a.m. UTC

We will introduce a new API to create a kernel socket with netns
refcnt held.

As a prep, let's add a new hold_net argument to __sock_create().

Note that we still do not pass it down to pf->create() for ease
of review; otherwise, this change will be buried in the huge diff.

Another option would be to override the kern parameter, which is
int, but I chose to change parameters for the following two reasons:

  1) Compilers allow us to efficiently make sure that all paths pass
     the parameters down to sk_alloc() as is.

  2) The parameter change breaks out-of-tree drivers, allowing the
     owners to choose an appropriate API.

Regarding 1), there actually was a weird path in smc_ulp_init()
that will be fixed up in the following patch.

While at it, the kernel-doc is fixed up to render the DESCRIPTION
part correctly.

  scripts/kernel-doc -man net/socket.c | scripts/split-man.pl /tmp/man
  man /tmp/man/__sock_create.9

Signed-off-by: Kuniyuki Iwashima <[e-mail]>
---
 net/socket.c | 38 +++++++++++++++++++++++---------------
 1 file changed, 23 insertions(+), 15 deletions(-)

Patch

diff --git a/net/socket.c b/net/socket.c
index 433f346ffc64..e5b4e0d34132 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -1470,22 +1470,28 @@  int sock_wake_async(struct socket_wq *wq, int how, int band)
 EXPORT_SYMBOL(sock_wake_async);
 
 /**
- *	__sock_create - creates a socket
- *	@net: net namespace
- *	@family: protocol family (AF_INET, ...)
- *	@type: communication type (SOCK_STREAM, ...)
- *	@protocol: protocol (0, ...)
- *	@res: new socket
- *	@kern: boolean for kernel space sockets
+ * __sock_create - creates a socket
  *
- *	Creates a new socket and assigns it to @res, passing through LSM.
- *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
- *	be set to true if the socket resides in kernel space.
- *	This function internally uses GFP_KERNEL.
+ * @net: net namespace
+ * @family: protocol family (AF_INET, ...)
+ * @type: communication type (SOCK_STREAM, ...)
+ * @protocol: protocol (0, ...)
+ * @res: new socket
+ * @kern: boolean for kernel space sockets
+ * @hold_net: boolean for netns refcnt
+ *
+ * Creates a new socket and assigns it to @res, passing through LSM.
+ *
+ * @kern must be set to true if userspace cannot touch it via a file
+ * descriptor nor BPF hooks except for LSM.  If @hold_net is false,
+ * the caller must ensure that the socket is always freed before @net.
+ *
+ * Context: Process context. This function internally uses GFP_KERNEL.
+ * Return: 0 or an error. On failure @res is set to %NULL.
  */
 
 static int __sock_create(struct net *net, int family, int type, int protocol,
-			 struct socket **res, int kern)
+			 struct socket **res, bool kern, bool hold_net)
 {
 	int err;
 	struct socket *sock;
@@ -1612,7 +1618,8 @@  static int __sock_create(struct net *net, int family, int type, int protocol,
 
 int sock_create(int family, int type, int protocol, struct socket **res)
 {
-	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
+	return __sock_create(current->nsproxy->net_ns, family, type, protocol,
+			     res, false, true);
 }
 EXPORT_SYMBOL(sock_create);
 
@@ -1628,9 +1635,10 @@  EXPORT_SYMBOL(sock_create);
  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
  */
 
-int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
+int sock_create_kern(struct net *net, int family, int type, int protocol,
+		     struct socket **res)
 {
-	return __sock_create(net, family, type, protocol, res, 1);
+	return __sock_create(net, family, type, protocol, res, true, false);
 }
 EXPORT_SYMBOL(sock_create_kern);