On Sat, May 2, 2020 at 4:29 AM Lese Doru Calin
<lesedorucali...@gmail.com> wrote:
>
> In this year's edition of GSoC, there is a project idea for CRIU to add
> support
> for checkpoint/restore of cork-ed UDP sockets. But to add it, the kernel API
> needs
> to be extended.
> This is what this patch does. It adds a new command, called SIOUDPPENDGET, to
> the
> ioctl syscall regarding UDP sockets, which stores the pending data from the
> write
> queue and the destination address in a struct msghdr. The arg for ioctl needs
> to
> be a pointer to a user space struct msghdr. The syscall returns the number of
> writed
> bytes, if successful, or error. To retrive the data requires the CAP_NET_ADMIN
> capability.
>
> Signed-off-by: Lese Doru Calin <lesedorucali...@gmail.com>
A few concerns:
- why call the BPF recvmsg hook from this ioctl function?
- The patch saves msg_name, but not other relevant state stored in
inet_cork, such as state passed through cmsg. Without that, this might
introduce more subtle bugs than not checkpointing at all.
- Duplicating usercopy code from net/socket.c is fragile and adds
maintenance burden. If anything such refactoring should stay inside
that file. To some extent the same applies to udp_peek_sndq and
net/core/datagram.c.
Less important
- a getsockopt is generally preferred over extending ioctl.
Overall, I'm not sure that this is the right approach. It is quite a
bit of code, for a mostly hypothetical omission to CRIU?
Since these are unreliable datagrams, it is arguably sufficient to
just drop a datagram if checkpoint/restore happened on a corked
socket. That might be simpler.
As long as CRIU is behind a static branch and thus adds no cycles in
the common hot path, I personally don't mind having it in the normal
send/recv path as much -- if that allows it to reuse existing code,
e.g., for copy to user. The MSG_ERRQUEUE path in particular is already
a slow path to loop packets from the send patch back to the sending
process.
> ---
> include/linux/socket.h | 2 +
> include/uapi/linux/sockios.h | 3 +
> net/ipv4/udp.c | 145 +++++++++++++++++++++++++++++++----
> net/socket.c | 4 +-
> 4 files changed, 139 insertions(+), 15 deletions(-)
>
> diff --git a/include/linux/socket.h b/include/linux/socket.h
> index 54338fac45cb..632ba0ea6709 100644
> --- a/include/linux/socket.h
> +++ b/include/linux/socket.h
> @@ -351,6 +351,8 @@ struct ucred {
> #define IPX_TYPE 1
>
> extern int move_addr_to_kernel(void __user *uaddr, int ulen, struct
> sockaddr_storage *kaddr);
> +extern int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
> + void __user *uaddr, int __user *ulen);
> extern int put_cmsg(struct msghdr*, int level, int type, int len, void
> *data);
>
> struct timespec64;
> diff --git a/include/uapi/linux/sockios.h b/include/uapi/linux/sockios.h
> index 7d1bccbbef78..3639fa906604 100644
> --- a/include/uapi/linux/sockios.h
> +++ b/include/uapi/linux/sockios.h
> @@ -153,6 +153,9 @@
> #define SIOCSHWTSTAMP 0x89b0 /* set and get config */
> #define SIOCGHWTSTAMP 0x89b1 /* get config */
>
> +/* UDP socket calls*/
> +#define SIOUDPPENDGET 0x89C0 /* get the pending data from write queue */
> +
> /* Device private ioctl calls */
>
> /*
> diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
> index 32564b350823..f729a5e7f90b 100644
> --- a/net/ipv4/udp.c
> +++ b/net/ipv4/udp.c
> @@ -1620,6 +1620,133 @@ static int first_packet_length(struct sock *sk)
> return res;
> }
>
> +static void udp_set_source_addr(struct sock *sk, struct msghdr *msg,
> + int *addr_len, u32 addr, u16 port)
> +{
> + DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
> +
> + if (sin) {
> + sin->sin_family = AF_INET;
> + sin->sin_port = port;
> + sin->sin_addr.s_addr = addr;
> + memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
> + *addr_len = sizeof(*sin);
> +
> + if (cgroup_bpf_enabled)
> + BPF_CGROUP_RUN_PROG_UDP4_RECVMSG_LOCK(sk,
> + (struct sockaddr *)sin);
> + }
> +}
> +
> +static int udp_peek_sndq(struct sock *sk, struct msghdr *msg, int off, int
> len)
> +{
> + int copy, copied = 0, err = 0;
> + struct sk_buff *skb;
> +
> + skb_queue_walk(&sk->sk_write_queue, skb) {
> + copy = len - copied;
> + if (copy > skb->len - off)
> + copy = skb->len - off;
> +
> + err = skb_copy_datagram_msg(skb, off, msg, copy);
> + if (err)
> + break;
> +
> + copied += copy;
> + if (len <= copied)
> + break;
> + }
> + return err ?: copied;
> +}
> +
> +static int udp_get_pending_write_queue(struct sock *sk, struct msghdr *msg,
> + int *addr_len)
> +{
> + int err = 0, off = sizeof(struct udphdr);
> + struct inet_sock *inet = inet_sk(sk);
> + struct udp_sock *up = udp_sk(sk);
> + struct flowi4 *fl4;
> + struct flowi6 *fl6;
> +
> + switch (up->pending) {
> + case 0:
> + return -ENODATA;
> + case AF_INET:
> + off += sizeof(struct iphdr);
> + fl4 = &inet->cork.fl.u.ip4;
> + udp_set_source_addr(sk, msg, addr_len,
> + fl4->daddr, fl4->fl4_dport);
> + break;
> + case AF_INET6:
> + off += sizeof(struct ipv6hdr);
> + if (msg->msg_name) {
> + DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6,
> + msg->msg_name);
> +
> + fl6 = &inet->cork.fl.u.ip6;
> + sin6->sin6_family = AF_INET6;
> + sin6->sin6_port = fl6->fl6_dport;
> + sin6->sin6_flowinfo = 0;
> + sin6->sin6_addr = fl6->daddr;
> + sin6->sin6_scope_id = fl6->flowi6_oif;
> + *addr_len = sizeof(*sin6);
> +
> + if (cgroup_bpf_enabled)
> + BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk,
> + (struct sockaddr *)sin6);
> + }
> + break;
> + default:
> + return -EINVAL;
> + }
> +
> + lock_sock(sk);
> + if (unlikely(!up->pending)) {
> + release_sock(sk);
> + return -EINVAL;
> + }
> + err = udp_peek_sndq(sk, msg, off, msg_data_left(msg));
> + release_sock(sk);
> + return err;
> +}
> +
> +static int prep_msghdr_recv_pending(struct sock *sk, void __user *argp)
> +{
> + struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
> + struct user_msghdr __user *msg;
> + struct sockaddr __user *uaddr;
> + struct sockaddr_storage addr;
> + struct msghdr msg_sys;
> + int __user *uaddr_len;
> + int err = 0, len = 0;
> +
> + if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
> + return -EPERM;
> +
> + if (!argp)
> + return -EINVAL;
> +
> + msg = (struct user_msghdr __user *)argp;
> + err = recvmsg_copy_msghdr(&msg_sys, msg, 0, &uaddr, &iov);
> + if (err < 0)
> + return err;
> +
> + uaddr_len = &msg->msg_namelen;
> + msg_sys.msg_name = &addr;
> + msg_sys.msg_flags = 0;
> +
> + err = udp_get_pending_write_queue(sk, &msg_sys, &len);
> + msg_sys.msg_namelen = len;
> + len = err;
> +
> + if (uaddr && err >= 0)
> + err = move_addr_to_user(&addr, msg_sys.msg_namelen,
> + uaddr, uaddr_len);
> +
> + kfree(iov);
> + return err < 0 ? err : len;
> +}
> +
> /*
> * IOCTL requests applicable to the UDP protocol
> */
> @@ -1641,6 +1768,9 @@ int udp_ioctl(struct sock *sk, int cmd, unsigned long
> arg)
> return put_user(amount, (int __user *)arg);
> }
>
> + case SIOUDPPENDGET:
> + return prep_msghdr_recv_pending(sk, (void __user *)arg);
> +
> default:
> return -ENOIOCTLCMD;
> }
> @@ -1729,7 +1859,6 @@ int udp_recvmsg(struct sock *sk, struct msghdr *msg,
> size_t len, int noblock,
> int flags, int *addr_len)
> {
> struct inet_sock *inet = inet_sk(sk);
> - DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
> struct sk_buff *skb;
> unsigned int ulen, copied;
> int off, err, peeking = flags & MSG_PEEK;
> @@ -1794,18 +1923,8 @@ int udp_recvmsg(struct sock *sk, struct msghdr *msg,
> size_t len, int noblock,
>
> sock_recv_ts_and_drops(msg, sk, skb);
>
> - /* Copy the address. */
> - if (sin) {
> - sin->sin_family = AF_INET;
> - sin->sin_port = udp_hdr(skb)->source;
> - sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
> - memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
> - *addr_len = sizeof(*sin);
> -
> - if (cgroup_bpf_enabled)
> - BPF_CGROUP_RUN_PROG_UDP4_RECVMSG_LOCK(sk,
> - (struct sockaddr
> *)sin);
> - }
> + udp_set_source_addr(sk, msg, addr_len, ip_hdr(skb)->saddr,
> + udp_hdr(skb)->source);
>
> if (udp_sk(sk)->gro_enabled)
> udp_cmsg_recv(msg, sk, skb);
> diff --git a/net/socket.c b/net/socket.c
> index 2dd739fba866..bd25d528c9a0 100644
> --- a/net/socket.c
> +++ b/net/socket.c
> @@ -217,8 +217,8 @@ int move_addr_to_kernel(void __user *uaddr, int ulen,
> struct sockaddr_storage *k
> * specified. Zero is returned for a success.
> */
>
> -static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
> - void __user *uaddr, int __user *ulen)
> +int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
> + void __user *uaddr, int __user *ulen)
> {
> int err;
> int len;
> --
> 2.17.1
>
