linux.git/net/socket.c, branch v6.1.168

net: Remove RTNL dance for SIOCBRADDIF and SIOCBRDELIF.

2026-01-11T14:19:23+00:00

commit ed3ba9b6e280e14cc3148c1b226ba453f02fa76c upstream.

SIOCBRDELIF is passed to dev_ioctl() first and later forwarded to
br_ioctl_call(), which causes unnecessary RTNL dance and the splat
below [0] under RTNL pressure.

Let's say Thread A is trying to detach a device from a bridge and
Thread B is trying to remove the bridge.

In dev_ioctl(), Thread A bumps the bridge device's refcnt by
netdev_hold() and releases RTNL because the following br_ioctl_call()
also re-acquires RTNL.

In the race window, Thread B could acquire RTNL and try to remove
the bridge device.  Then, rtnl_unlock() by Thread B will release RTNL
and wait for netdev_put() by Thread A.

Thread A, however, must hold RTNL after the unlock in dev_ifsioc(),
which may take long under RTNL pressure, resulting in the splat by
Thread B.

  Thread A (SIOCBRDELIF)           Thread B (SIOCBRDELBR)
  ----------------------           ----------------------
  sock_ioctl                       sock_ioctl
  `- sock_do_ioctl                 `- br_ioctl_call
     `- dev_ioctl                     `- br_ioctl_stub
        |- rtnl_lock                     |
        |- dev_ifsioc                    '
        '  |- dev = __dev_get_by_name(...)
           |- netdev_hold(dev, ...)      .
       /   |- rtnl_unlock  ------.       |
       |   |- br_ioctl_call       `--->  |- rtnl_lock
  Race |   |  `- br_ioctl_stub           |- br_del_bridge
  Window   |     |                       |  |- dev = __dev_get_by_name(...)
       |   |     |  May take long        |  `- br_dev_delete(dev, ...)
       |   |     |  under RTNL pressure  |     `- unregister_netdevice_queue(dev, ...)
       |   |     |               |       `- rtnl_unlock
       \   |     |- rtnl_lock  <-'          `- netdev_run_todo
           |     |- ...                        `- netdev_run_todo
           |     `- rtnl_unlock                   |- __rtnl_unlock
           |                                      |- netdev_wait_allrefs_any
           |- netdev_put(dev, ...)  <----------------'
                                                Wait refcnt decrement
                                                and log splat below

To avoid blocking SIOCBRDELBR unnecessarily, let's not call
dev_ioctl() for SIOCBRADDIF and SIOCBRDELIF.

In the dev_ioctl() path, we do the following:

  1. Copy struct ifreq by get_user_ifreq in sock_do_ioctl()
  2. Check CAP_NET_ADMIN in dev_ioctl()
  3. Call dev_load() in dev_ioctl()
  4. Fetch the master dev from ifr.ifr_name in dev_ifsioc()

3. can be done by request_module() in br_ioctl_call(), so we move
1., 2., and 4. to br_ioctl_stub().

Note that 2. is also checked later in add_del_if(), but it's better
performed before RTNL.

SIOCBRADDIF and SIOCBRDELIF have been processed in dev_ioctl() since
the pre-git era, and there seems to be no specific reason to process
them there.

[0]:
unregister_netdevice: waiting for wpan3 to become free. Usage count = 2
ref_tracker: wpan3@ffff8880662d8608 has 1/1 users at
     __netdev_tracker_alloc include/linux/netdevice.h:4282 [inline]
     netdev_hold include/linux/netdevice.h:4311 [inline]
     dev_ifsioc+0xc6a/0x1160 net/core/dev_ioctl.c:624
     dev_ioctl+0x255/0x10c0 net/core/dev_ioctl.c:826
     sock_do_ioctl+0x1ca/0x260 net/socket.c:1213
     sock_ioctl+0x23a/0x6c0 net/socket.c:1318
     vfs_ioctl fs/ioctl.c:51 [inline]
     __do_sys_ioctl fs/ioctl.c:906 [inline]
     __se_sys_ioctl fs/ioctl.c:892 [inline]
     __x64_sys_ioctl+0x1a4/0x210 fs/ioctl.c:892
     do_syscall_x64 arch/x86/entry/common.c:52 [inline]
     do_syscall_64+0xcb/0x250 arch/x86/entry/common.c:83
     entry_SYSCALL_64_after_hwframe+0x77/0x7f

Fixes: 893b19587534 ("net: bridge: fix ioctl locking")
Reported-by: syzkaller 
Reported-by: yan kang 
Reported-by: yue sun 
Closes: https://lore.kernel.org/netdev/SY8P300MB0421225D54EB92762AE8F0F2A1D32@SY8P300MB0421.AUSP300.PROD.OUTLOOK.COM/
Signed-off-by: Kuniyuki Iwashima 
Acked-by: Stanislav Fomichev 
Reviewed-by: Ido Schimmel 
Acked-by: Nikolay Aleksandrov 
Link: https://patch.msgid.link/20250316192851.19781-1-kuniyu@amazon.com
Signed-off-by: Paolo Abeni 
[cascardo: fixed conflict at dev_ifsioc]
Signed-off-by: Thadeu Lima de Souza Cascardo 
Signed-off-by: Greg Kroah-Hartman

net: explicitly clear the sk pointer, when pf->create fails

2024-10-17T13:22:27+00:00

commit 631083143315d1b192bd7d915b967b37819e88ea upstream.

We have recently noticed the exact same KASAN splat as in commit
6cd4a78d962b ("net: do not leave a dangling sk pointer, when socket
creation fails"). The problem is that commit did not fully address the
problem, as some pf->create implementations do not use sk_common_release
in their error paths.

For example, we can use the same reproducer as in the above commit, but
changing ping to arping. arping uses AF_PACKET socket and if packet_create
fails, it will just sk_free the allocated sk object.

While we could chase all the pf->create implementations and make sure they
NULL the freed sk object on error from the socket, we can't guarantee
future protocols will not make the same mistake.

So it is easier to just explicitly NULL the sk pointer upon return from
pf->create in __sock_create. We do know that pf->create always releases the
allocated sk object on error, so if the pointer is not NULL, it is
definitely dangling.

Fixes: 6cd4a78d962b ("net: do not leave a dangling sk pointer, when socket creation fails")
Signed-off-by: Ignat Korchagin 
Cc: stable@vger.kernel.org
Reviewed-by: Eric Dumazet 
Link: https://patch.msgid.link/20241003170151.69445-1-ignat@cloudflare.com
Signed-off-by: Jakub Kicinski 
Signed-off-by: Greg Kroah-Hartman

splice, net: Add a splice_eof op to file-ops and socket-ops

2024-01-10T16:10:27+00:00

[ Upstream commit 2bfc66850952b6921b2033b09729ec59eabbc81d ]

Add an optional method, ->splice_eof(), to allow splice to indicate the
premature termination of a splice to struct file_operations and struct
proto_ops.

This is called if sendfile() or splice() encounters all of the following
conditions inside splice_direct_to_actor():

 (1) the user did not set SPLICE_F_MORE (splice only), and

 (2) an EOF condition occurred (->splice_read() returned 0), and

 (3) we haven't read enough to fulfill the request (ie. len > 0 still), and

 (4) we have already spliced at least one byte.

A further patch will modify the behaviour of SPLICE_F_MORE to always be
passed to the actor if either the user set it or we haven't yet read
sufficient data to fulfill the request.

Suggested-by: Linus Torvalds 
Link: https://lore.kernel.org/r/CAHk-=wh=V579PDYvkpnTobCLGczbgxpMgGmmhqiTyE34Cpi5Gg@mail.gmail.com/
Signed-off-by: David Howells 
Reviewed-by: Jakub Kicinski 
cc: Jens Axboe 
cc: Christoph Hellwig 
cc: Al Viro 
cc: Matthew Wilcox 
cc: Jan Kara 
cc: Jeff Layton 
cc: David Hildenbrand 
cc: Christian Brauner 
cc: Chuck Lever 
cc: Boris Pismenny 
cc: John Fastabend 
cc: linux-mm@kvack.org
Signed-off-by: Jakub Kicinski 
Stable-dep-of: a0002127cd74 ("udp: move udp->no_check6_tx to udp->udp_flags")
Signed-off-by: Sasha Levin

net: Declare MSG_SPLICE_PAGES internal sendmsg() flag

2024-01-10T16:10:27+00:00

[ Upstream commit b841b901c452d92610f739a36e54978453528876 ]

Declare MSG_SPLICE_PAGES, an internal sendmsg() flag, that hints to a
network protocol that it should splice pages from the source iterator
rather than copying the data if it can.  This flag is added to a list that
is cleared by sendmsg syscalls on entry.

This is intended as a replacement for the ->sendpage() op, allowing a way
to splice in several multipage folios in one go.

Signed-off-by: David Howells 
Reviewed-by: Willem de Bruijn 
cc: Jens Axboe 
cc: Matthew Wilcox 
Signed-off-by: Jakub Kicinski 
Stable-dep-of: a0002127cd74 ("udp: move udp->no_check6_tx to udp->udp_flags")
Signed-off-by: Sasha Levin

net: Save and restore msg_namelen in sock_sendmsg

2024-01-10T16:10:24+00:00

[ Upstream commit 01b2885d9415152bcb12ff1f7788f500a74ea0ed ]

Commit 86a7e0b69bd5 ("net: prevent rewrite of msg_name in
sock_sendmsg()") made sock_sendmsg save the incoming msg_name pointer
and restore it before returning, to insulate the caller against
msg_name being changed by the called code.  If the address length
was also changed however, we may return with an inconsistent structure
where the length doesn't match the address, and attempts to reuse it may
lead to lost packets.

For example, a kernel that doesn't have commit 1c5950fc6fe9 ("udp6: fix
potential access to stale information") will replace a v4 mapped address
with its ipv4 equivalent, and shorten namelen accordingly from 28 to 16.
If the caller attempts to reuse the resulting msg structure, it will have
the original ipv6 (v4 mapped) address but an incorrect v4 length.

Fixes: 86a7e0b69bd5 ("net: prevent rewrite of msg_name in sock_sendmsg()")
Signed-off-by: Marc Dionne 
Reviewed-by: Willem de Bruijn 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

net: annotate data-races around sk->sk_bind_phc

2024-01-10T16:10:23+00:00

[ Upstream commit 251cd405a9e6e70b92fe5afbdd17fd5caf9d3266 ]

sk->sk_bind_phc is read locklessly. Add corresponding annotations.

Fixes: d463126e23f1 ("net: sock: extend SO_TIMESTAMPING for PHC binding")
Signed-off-by: Eric Dumazet 
Cc: Yangbo Lu 
Signed-off-by: David S. Miller 
Stable-dep-of: 7f6ca95d16b9 ("net: Implement missing getsockopt(SO_TIMESTAMPING_NEW)")
Signed-off-by: Sasha Levin

net: annotate data-races around sk->sk_tsflags

2024-01-10T16:10:23+00:00

[ Upstream commit e3390b30a5dfb112e8e802a59c0f68f947b638b2 ]

sk->sk_tsflags can be read locklessly, add corresponding annotations.

Fixes: b9f40e21ef42 ("net-timestamp: move timestamp flags out of sk_flags")
Signed-off-by: Eric Dumazet 
Cc: Willem de Bruijn 
Signed-off-by: David S. Miller 
Stable-dep-of: 7f6ca95d16b9 ("net: Implement missing getsockopt(SO_TIMESTAMPING_NEW)")
Signed-off-by: Sasha Levin

net: prevent address rewrite in kernel_bind()

2023-10-19T21:08:50+00:00

commit c889a99a21bf124c3db08d09df919f0eccc5ea4c upstream.

Similar to the change in commit 0bdf399342c5("net: Avoid address
overwrite in kernel_connect"), BPF hooks run on bind may rewrite the
address passed to kernel_bind(). This change

1) Makes a copy of the bind address in kernel_bind() to insulate
   callers.
2) Replaces direct calls to sock->ops->bind() in net with kernel_bind()

Link: https://lore.kernel.org/netdev/20230912013332.2048422-1-jrife@google.com/
Fixes: 4fbac77d2d09 ("bpf: Hooks for sys_bind")
Cc: stable@vger.kernel.org
Reviewed-by: Willem de Bruijn 
Signed-off-by: Jordan Rife 
Reviewed-by: Simon Horman 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

net: prevent rewrite of msg_name in sock_sendmsg()

2023-10-10T20:00:39+00:00

commit 86a7e0b69bd5b812e48a20c66c2161744f3caa16 upstream.

Callers of sock_sendmsg(), and similarly kernel_sendmsg(), in kernel
space may observe their value of msg_name change in cases where BPF
sendmsg hooks rewrite the send address. This has been confirmed to break
NFS mounts running in UDP mode and has the potential to break other
systems.

This patch:

1) Creates a new function called __sock_sendmsg() with same logic as the
   old sock_sendmsg() function.
2) Replaces calls to sock_sendmsg() made by __sys_sendto() and
   __sys_sendmsg() with __sock_sendmsg() to avoid an unnecessary copy,
   as these system calls are already protected.
3) Modifies sock_sendmsg() so that it makes a copy of msg_name if
   present before passing it down the stack to insulate callers from
   changes to the send address.

Link: https://lore.kernel.org/netdev/20230912013332.2048422-1-jrife@google.com/
Fixes: 1cedee13d25a ("bpf: Hooks for sys_sendmsg")
Cc: stable@vger.kernel.org
Reviewed-by: Willem de Bruijn 
Signed-off-by: Jordan Rife 
Reviewed-by: Simon Horman 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

net: Avoid address overwrite in kernel_connect

2023-09-13T07:42:26+00:00

commit 0bdf399342c5acbd817c9098b6c7ed21f1974312 upstream.

BPF programs that run on connect can rewrite the connect address. For
the connect system call this isn't a problem, because a copy of the address
is made when it is moved into kernel space. However, kernel_connect
simply passes through the address it is given, so the caller may observe
its address value unexpectedly change.

A practical example where this is problematic is where NFS is combined
with a system such as Cilium which implements BPF-based load balancing.
A common pattern in software-defined storage systems is to have an NFS
mount that connects to a persistent virtual IP which in turn maps to an
ephemeral server IP. This is usually done to achieve high availability:
if your server goes down you can quickly spin up a replacement and remap
the virtual IP to that endpoint. With BPF-based load balancing, mounts
will forget the virtual IP address when the address rewrite occurs
because a pointer to the only copy of that address is passed down the
stack. Server failover then breaks, because clients have forgotten the
virtual IP address. Reconnects fail and mounts remain broken. This patch
was tested by setting up a scenario like this and ensuring that NFS
reconnects worked after applying the patch.

Signed-off-by: Jordan Rife 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman