| Message ID | 20230919175254.144417-1-jrife@google.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net,v4,1/3] net: replace calls to sock->ops->connect() with kernel_connect() | expand |
Jordan Rife wrote: > 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 > Signed-off-by: Jordan Rife <jrife@google.com> Reviewed-by: Willem de Bruijn <willemb@google.com>
On Wed, 2023-09-20 at 09:28 -0400, Willem de Bruijn wrote: > Jordan Rife wrote: > > 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 > > Signed-off-by: Jordan Rife <jrife@google.com> > > Reviewed-by: Willem de Bruijn <willemb@google.com> CC Jens and Pavel, as I guess io_uring likely want to use __sock_sendmsg(), in a follow-up patch. Cheers, Paolo
On 9/21/23 09:32, Paolo Abeni wrote: > On Wed, 2023-09-20 at 09:28 -0400, Willem de Bruijn wrote: >> Jordan Rife wrote: >>> 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 >>> Signed-off-by: Jordan Rife <jrife@google.com> >> >> Reviewed-by: Willem de Bruijn <willemb@google.com> > > CC Jens and Pavel, as I guess io_uring likely want to use > __sock_sendmsg(), in a follow-up patch. Yeah, likely so. Thanks Paolo, we'll take a look
diff --git a/net/socket.c b/net/socket.c index c8b08b32f097e..a39ec136f5cff 100644 --- a/net/socket.c +++ b/net/socket.c @@ -737,6 +737,14 @@ static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) return ret; } +static int __sock_sendmsg(struct socket *sock, struct msghdr *msg) +{ + int err = security_socket_sendmsg(sock, msg, + msg_data_left(msg)); + + return err ?: sock_sendmsg_nosec(sock, msg); +} + /** * sock_sendmsg - send a message through @sock * @sock: socket @@ -747,10 +755,19 @@ static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) */ int sock_sendmsg(struct socket *sock, struct msghdr *msg) { - int err = security_socket_sendmsg(sock, msg, - msg_data_left(msg)); + struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name; + struct sockaddr_storage address; + int ret; - return err ?: sock_sendmsg_nosec(sock, msg); + if (msg->msg_name) { + memcpy(&address, msg->msg_name, msg->msg_namelen); + msg->msg_name = &address; + } + + ret = __sock_sendmsg(sock, msg); + msg->msg_name = save_addr; + + return ret; } EXPORT_SYMBOL(sock_sendmsg); @@ -1138,7 +1155,7 @@ static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) if (sock->type == SOCK_SEQPACKET) msg.msg_flags |= MSG_EOR; - res = sock_sendmsg(sock, &msg); + res = __sock_sendmsg(sock, &msg); *from = msg.msg_iter; return res; } @@ -2174,7 +2191,7 @@ int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, if (sock->file->f_flags & O_NONBLOCK) flags |= MSG_DONTWAIT; msg.msg_flags = flags; - err = sock_sendmsg(sock, &msg); + err = __sock_sendmsg(sock, &msg); out_put: fput_light(sock->file, fput_needed); @@ -2538,7 +2555,7 @@ static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, err = sock_sendmsg_nosec(sock, msg_sys); goto out_freectl; } - err = sock_sendmsg(sock, msg_sys); + err = __sock_sendmsg(sock, msg_sys); /* * If this is sendmmsg() and sending to current destination address was * successful, remember it.
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 Signed-off-by: Jordan Rife <jrife@google.com> --- v3->v4: Maintain reverse xmas tree order for variable declarations. Remove precondition check for msg_namelen. v2->v3: Add "Fixes" tag. v1->v2: Split up original patch into patch series. Perform address copy in sock_sendmsg() instead of sock->ops->sendmsg(). net/socket.c | 29 +++++++++++++++++++++++------ 1 file changed, 23 insertions(+), 6 deletions(-)