Re: [PATCH RFC 1/3] vsock: support sockmap

On Thu, Jan 19, 2023 at 11:39:36AM +0100, Stefano Garzarella wrote:

On Wed, Jan 18, 2023 at 12:27:39PM -0800, Bobby Eshleman wrote:

...

This patch adds sockmap support for vsock sockets. It is intended to be usable by all transports, but only the virtio transport is implemented.

Signed-off-by: Bobby Eshleman bobby.eshleman@bytedance.com

drivers/vhost/vsock.c | 1 + include/linux/virtio_vsock.h | 1 + include/net/af_vsock.h | 17 +++ net/vmw_vsock/Makefile | 1 + net/vmw_vsock/af_vsock.c | 59 +++++++++-- net/vmw_vsock/virtio_transport.c | 2 + net/vmw_vsock/virtio_transport_common.c | 22 ++++ net/vmw_vsock/vsock_bpf.c | 180 ++++++++++++++++++++++++++++++++ net/vmw_vsock/vsock_loopback.c | 2 + 9 files changed, 279 insertions(+), 6 deletions(-)

./scripts/checkpatch.pl --strict prints some simple warnings/checks that I suggest to fix :-)

Oops, thanks. New machine, forgot my pre-commit hook. Putting in place now.

...

diff --git a/drivers/vhost/vsock.c b/drivers/vhost/vsock.c index 1f3b89c885cca..3c6dc036b9044 100644 --- a/drivers/vhost/vsock.c +++ b/drivers/vhost/vsock.c @@ -439,6 +439,7 @@ static struct virtio_transport vhost_transport = { .notify_send_post_enqueue = virtio_transport_notify_send_post_enqueue, .notify_buffer_size = virtio_transport_notify_buffer_size,

• .read_skb = virtio_transport_read_skb,
  

  },

  .send_pkt = vhost_transport_send_pkt,

diff --git a/include/linux/virtio_vsock.h b/include/linux/virtio_vsock.h index 3f9c166113063..c58453699ee98 100644 --- a/include/linux/virtio_vsock.h +++ b/include/linux/virtio_vsock.h @@ -245,4 +245,5 @@ u32 virtio_transport_get_credit(struct virtio_vsock_sock *vvs, u32 wanted); void virtio_transport_put_credit(struct virtio_vsock_sock *vvs, u32 credit); void virtio_transport_deliver_tap_pkt(struct sk_buff *skb); int virtio_transport_purge_skbs(void *vsk, struct sk_buff_head *list); +int virtio_transport_read_skb(struct vsock_sock *vsk, skb_read_actor_t read_actor); #endif /* _LINUX_VIRTIO_VSOCK_H */ diff --git a/include/net/af_vsock.h b/include/net/af_vsock.h index 568a87c5e0d0f..a73f5fbd296af 100644 --- a/include/net/af_vsock.h +++ b/include/net/af_vsock.h @@ -75,6 +75,7 @@ struct vsock_sock { void *trans; };

+s64 vsock_connectible_has_data(struct vsock_sock *vsk); s64 vsock_stream_has_data(struct vsock_sock *vsk); s64 vsock_stream_has_space(struct vsock_sock *vsk); struct sock *vsock_create_connected(struct sock *parent); @@ -173,6 +174,9 @@ struct vsock_transport {

/* Addressing. */ u32 (*get_local_cid)(void);

• /* Read a single skb */
• int (*read_skb)(struct vsock_sock *, skb_read_actor_t);

};

/**** CORE ****/ @@ -225,5 +229,18 @@ int vsock_init_tap(void); int vsock_add_tap(struct vsock_tap *vt); int vsock_remove_tap(struct vsock_tap *vt); void vsock_deliver_tap(struct sk_buff *build_skb(void *opaque), void *opaque); +int vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,

• 	      int flags);
  

+int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,

• 	       size_t len, int flags);
  

+#ifdef CONFIG_BPF_SYSCALL +extern struct proto vsock_proto; +int vsock_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore); +void __init vsock_bpf_build_proto(void); +#else +static inline void __init vsock_bpf_build_proto(void) +{} +#endif

#endif /* __AF_VSOCK_H__ */ diff --git a/net/vmw_vsock/Makefile b/net/vmw_vsock/Makefile index 6a943ec95c4a5..5da74c4a9f1d1 100644 --- a/net/vmw_vsock/Makefile +++ b/net/vmw_vsock/Makefile @@ -8,6 +8,7 @@ obj-$(CONFIG_HYPERV_VSOCKETS) += hv_sock.o obj-$(CONFIG_VSOCKETS_LOOPBACK) += vsock_loopback.o

vsock-y += af_vsock.o af_vsock_tap.o vsock_addr.o +vsock-$(CONFIG_BPF_SYSCALL) += vsock_bpf.o

vsock_diag-y += diag.o

diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c index d593d5b6d4b15..7081b3a992c1e 100644 --- a/net/vmw_vsock/af_vsock.c +++ b/net/vmw_vsock/af_vsock.c @@ -116,10 +116,13 @@ static void vsock_sk_destruct(struct sock *sk); static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);

/* Protocol family. */ -static struct proto vsock_proto = { +struct proto vsock_proto = { .name = "AF_VSOCK", .owner = THIS_MODULE, .obj_size = sizeof(struct vsock_sock), +#ifdef CONFIG_BPF_SYSCALL

• .psock_update_sk_prot = vsock_bpf_update_proto,

+#endif };

/* The default peer timeout indicates how long we will wait for a peer response @@ -865,7 +868,7 @@ s64 vsock_stream_has_data(struct vsock_sock *vsk) } EXPORT_SYMBOL_GPL(vsock_stream_has_data);

-static s64 vsock_connectible_has_data(struct vsock_sock *vsk) +s64 vsock_connectible_has_data(struct vsock_sock *vsk) { struct sock *sk = sk_vsock(vsk);

@@ -874,6 +877,7 @@ static s64 vsock_connectible_has_data(struct vsock_sock *vsk) else return vsock_stream_has_data(vsk); } +EXPORT_SYMBOL_GPL(vsock_connectible_has_data);

s64 vsock_stream_has_space(struct vsock_sock *vsk) { @@ -1131,6 +1135,19 @@ static __poll_t vsock_poll(struct file *file, struct socket *sock, return mask; }

+static int vsock_read_skb(struct sock *sk, skb_read_actor_t read_actor) +{

• struct vsock_sock *vsk = vsock_sk(sk);
• if (!vsk->transport)

• return -ENODEV;
  

• if (!vsk->transport->read_skb)

• return -EOPNOTSUPP;
  

• return vsk->transport->read_skb(vsk, read_actor);

+}

static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) { @@ -1241,19 +1258,32 @@ static int vsock_dgram_connect(struct socket *sock,

memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); sock->state = SS_CONNECTED;

• sk->sk_state = TCP_ESTABLISHED;

Why we need this change? If it's a fix, we should put it in another patch.

This is just required by sockmap's function that determines if a socket is a valid one to add to a map. It will refuse to add any socket that is not TCP_ESTABLISHED to a sockmap.

This was the approach that unix dgrams took, so I followed here.

...

out: release_sock(sk); return err; }

-static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,

• 	       size_t len, int flags)
  

+int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,

• 	size_t len, int flags)
  

{

• struct vsock_sock *vsk = vsock_sk(sock->sk);

• const struct proto *prot;

We should use the guard for this statement as in vsock_connectible_recvmsg().

Got it.

...

• struct vsock_sock *vsk;
• struct sock *sk;
• sk = sock->sk;
• vsk = vsock_sk(sk);

+#ifdef CONFIG_BPF_SYSCALL

• prot = READ_ONCE(sk->sk_prot);
• if (prot != &vsock_proto)

• return prot->recvmsg(sk, msg, len, flags, NULL);
  

+#endif

return vsk->transport->dgram_dequeue(vsk, msg, len, flags); } +EXPORT_SYMBOL_GPL(vsock_dgram_recvmsg);

static const struct proto_ops vsock_dgram_ops = { .family = PF_VSOCK, @@ -1272,6 +1302,7 @@ static const struct proto_ops vsock_dgram_ops = { .recvmsg = vsock_dgram_recvmsg, .mmap = sock_no_mmap, .sendpage = sock_no_sendpage,

• .read_skb = vsock_read_skb,

};

static int vsock_transport_cancel_pkt(struct vsock_sock *vsk) @@ -2085,13 +2116,16 @@ static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg, return err; }

-static int +int vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, int flags) { struct sock *sk; struct vsock_sock *vsk; const struct vsock_transport *transport; +#ifdef CONFIG_BPF_SYSCALL

• const struct proto *prot;

+#endif int err;

sk = sock->sk; @@ -2138,6 +2172,14 @@ vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, goto out; }

+#ifdef CONFIG_BPF_SYSCALL

• prot = READ_ONCE(sk->sk_prot);
• if (prot != &vsock_proto) {

• release_sock(sk);
  

• return prot->recvmsg(sk, msg, len, flags, NULL);
  

• }

+#endif

• if (sk->sk_type == SOCK_STREAM) err = __vsock_stream_recvmsg(sk, msg, len, flags); else

@@ -2147,6 +2189,7 @@ vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, release_sock(sk); return err; } +EXPORT_SYMBOL_GPL(vsock_connectible_recvmsg);

static int vsock_set_rcvlowat(struct sock *sk, int val) { @@ -2187,6 +2230,7 @@ static const struct proto_ops vsock_stream_ops = { .mmap = sock_no_mmap, .sendpage = sock_no_sendpage, .set_rcvlowat = vsock_set_rcvlowat,

• .read_skb = vsock_read_skb,

};

static const struct proto_ops vsock_seqpacket_ops = { @@ -2208,6 +2252,7 @@ static const struct proto_ops vsock_seqpacket_ops = { .recvmsg = vsock_connectible_recvmsg, .mmap = sock_no_mmap, .sendpage = sock_no_sendpage,

• .read_skb = vsock_read_skb,

};

static int vsock_create(struct net *net, struct socket *sock, @@ -2347,6 +2392,8 @@ static int __init vsock_init(void) goto err_unregister_proto; }

• vsock_bpf_build_proto();
• return 0;

err_unregister_proto: diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c index 28b5a8e8e0948..e95df847176b6 100644 --- a/net/vmw_vsock/virtio_transport.c +++ b/net/vmw_vsock/virtio_transport.c @@ -457,6 +457,8 @@ static struct virtio_transport virtio_transport = { .notify_send_pre_enqueue = virtio_transport_notify_send_pre_enqueue, .notify_send_post_enqueue = virtio_transport_notify_send_post_enqueue, .notify_buffer_size = virtio_transport_notify_buffer_size,

• .read_skb = virtio_transport_read_skb,
  

  },

  .send_pkt = virtio_transport_send_pkt,

diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c index a1581c77cf84a..9a87ead5b1fc5 100644 --- a/net/vmw_vsock/virtio_transport_common.c +++ b/net/vmw_vsock/virtio_transport_common.c @@ -1388,6 +1388,28 @@ int virtio_transport_purge_skbs(void *vsk, struct sk_buff_head *queue) } EXPORT_SYMBOL_GPL(virtio_transport_purge_skbs);

+int virtio_transport_read_skb(struct vsock_sock *vsk, skb_read_actor_t recv_actor) +{

• struct virtio_vsock_sock *vvs = vsk->trans;
• struct sock *sk = sk_vsock(vsk);
• struct sk_buff *skb;
• int copied = 0;

We could avoid initializing `copied`, since it is overwritten later.

Got it.

...

• int off = 0;
• int err;
• spin_lock_bh(&vvs->rx_lock);
• skb = __skb_recv_datagram(sk, &vvs->rx_queue, MSG_DONTWAIT, &off, &err);

Will this work also for STREAM and SEQPACKET sockets?

Yep, it is used for non-datagram sockets as well because it is free of race conditions, and handles waits/errors sensibly. For example, in unix_accept():

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/net/...

...

• spin_unlock_bh(&vvs->rx_lock);
• if (!skb)

• return err;
  

• copied = recv_actor(sk, skb);
• kfree_skb(skb);

I would have moved these steps to vsock_read_skb() to avoid duplicating these steps in each transport. Perhaps not all transports want to pass skb ownership to the caller, though, so maybe we can leave it that way for now.

That is a good point though. I bet your initial hunch is the right one. If even one other transport duplicates this, then I'd say it is worth pulling up into vsock_read_skb().

...

• return copied;

+} +EXPORT_SYMBOL_GPL(virtio_transport_read_skb);

MODULE_LICENSE("GPL v2"); MODULE_AUTHOR("Asias He"); MODULE_DESCRIPTION("common code for virtio vsock"); diff --git a/net/vmw_vsock/vsock_bpf.c b/net/vmw_vsock/vsock_bpf.c new file mode 100644 index 0000000000000..9e11282d3bc1f --- /dev/null +++ b/net/vmw_vsock/vsock_bpf.c @@ -0,0 +1,180 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2022 Bobby Eshleman bobby.eshleman@bytedance.com

• • Based off of net/unix/unix_bpf.c
• */

+#include <linux/bpf.h> +#include <linux/module.h> +#include <linux/skmsg.h> +#include <linux/socket.h> +#include <net/af_vsock.h> +#include <net/sock.h>

+#define vsock_sk_has_data(__sk, __psock) \

• ({	!skb_queue_empty(&__sk->sk_receive_queue) ||	\
  

• 	!skb_queue_empty(&__psock->ingress_skb) ||	\
  

• 	!list_empty(&__psock->ingress_msg);		\
  

• })
  

+static struct proto *vsock_dgram_prot_saved __read_mostly; +static DEFINE_SPINLOCK(vsock_dgram_prot_lock); +static struct proto vsock_dgram_bpf_prot;

+static bool vsock_has_data(struct vsock_sock *vsk, struct sk_psock *psock) +{

• struct sock *sk = sk_vsock(vsk);
• s64 ret;
• ret = vsock_connectible_has_data(vsk);
• if (ret > 0)

• return true;
  

• return vsock_sk_has_data(sk, psock);

+}

+static int vsock_msg_wait_data(struct sock *sk, struct sk_psock *psock, long timeo) +{

• struct vsock_sock *vsk;
• int err;
• DEFINE_WAIT(wait);
• vsk = vsock_sk(sk);
• err = 0;
• while (vsock_has_data(vsk, psock)) {

• prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
  

• if (sk->sk_err != 0 ||
  

•     (sk->sk_shutdown & RCV_SHUTDOWN) ||
  

•     (vsk->peer_shutdown & SEND_SHUTDOWN)) {
  

• 	break;
  

• }
  

• if (timeo == 0) {
  

• 	err = -EAGAIN;
  

• 	break;
  

• }
  

• release_sock(sk);
  

• timeo = schedule_timeout(timeo);
  

• lock_sock(sk);
  

• if (signal_pending(current)) {
  

• 	err = sock_intr_errno(timeo);
  

• 	break;
  

• } else if (timeo == 0) {
  

• 	err = -EAGAIN;
  

• 	break;
  

• }
  

• }
• finish_wait(sk_sleep(sk), &wait);
• if (err)

• return err;
  

• return 0;

+}

+static int vsock_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int flags) +{

• int err;
• struct socket *sock = sk->sk_socket;
• if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)

• err = vsock_connectible_recvmsg(sock, msg, len, flags);
  

• else

Could it happen that it is not DGRAM and we should return an error in this case?

I'm not sure but for the sake of safety, I'll add that.

Thanks, Stefano

Thanks for the review Stefano.

Best, Bobby