Hi,
On Wed, Jun 03, 2009 at 08:18:25AM -0700, Dan Smith wrote:
> This patch adds basic checkpoint/restart support for AF_UNIX sockets. It
> has been tested with a single and multiple processes, and with data inflight
> at the time of checkpoint. It supports both socketpair()s and path-based
> sockets.
>
> I have an almost-working AF_INET follow-on to this which I can submit after
> this is reviewed and tweaked into acceptance.
>
[...]
> diff --git a/net/socket_cr.c b/net/socket_cr.c
> new file mode 100644
> index 0000000..76759fe
> --- /dev/null
> +++ b/net/socket_cr.c
> @@ -0,0 +1,378 @@
> +/*
> + * Copyright 2009 IBM Corporation
> + *
> + * Author: Dan Smith <danms@xxxxxxxxxx>
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License as
> + * published by the Free Software Foundation, version 2 of the
> + * License.
> + */
> +
> +#include <linux/socket.h>
> +#include <linux/mount.h>
> +#include <linux/file.h>
> +
> +#include <net/af_unix.h>
> +#include <net/tcp_states.h>
> +
> +#include <linux/checkpoint.h>
> +#include <linux/checkpoint_hdr.h>
> +
> +static int sock_copy_buffers(struct sk_buff_head *from, struct sk_buff_head *to)
> +{
> + int count = 0;
> + struct sk_buff *skb;
> +
> + spin_lock(&from->lock);
> +
> + skb_queue_walk(from, skb) {
> + struct sk_buff *tmp;
> +
> + tmp = skb_copy(skb, GFP_KERNEL);
GFP_KERNEL is not allowed here, since from->lock is locked. Not sure that
GFP_ATOMIC is acceptable though. Perhaps it would be better to temporarily move
the queue to a local head, copy it (no spinlock needed), and then push it again.
This would need to block concurrent senders/receivers during this operation,
unless it's guaranteed that they are all frozen.
Thanks,
Louis
> + if (!tmp) {
> + count = -ENOMEM;
> + goto out;
> + }
> + skb_queue_tail(to, tmp);
> + count++;
> + }
> + out:
> + spin_unlock(&from->lock);
> +
> + return count;
> +}
> +
> +static int __sock_write_buffers(struct ckpt_ctx *ctx,
> + struct sk_buff_head *queue)
> +{
> + struct sk_buff *skb;
> + int ret = 0;
> +
> + skb_queue_walk(queue, skb) {
> + ret = ckpt_write_obj_type(ctx, skb->data, skb->len,
> + CKPT_HDR_SOCKET_BUFFER);
> + if (ret)
> + return ret;
> + }
> +
> + return 0;
> +}
> +
> +static int sock_write_buffers(struct ckpt_ctx *ctx, struct sk_buff_head *queue)
> +{
> + struct ckpt_hdr_socket_buffer *h;
> + struct sk_buff_head tmpq;
> + int ret = -ENOMEM;
> +
> + h = ckpt_hdr_get_type(ctx, sizeof(*h), CKPT_HDR_SOCKET_BUFFERS);
> + if (!h)
> + goto out;
> +
> + skb_queue_head_init(&tmpq);
> +
> + h->skb_count = sock_copy_buffers(queue, &tmpq);
> + if (h->skb_count < 0) {
> + ret = h->skb_count;
> + goto out;
> + }
> +
> + ret = ckpt_write_obj(ctx, (struct ckpt_hdr *) h);
> + if (!ret)
> + ret = __sock_write_buffers(ctx, &tmpq);
> +
> + out:
> + ckpt_hdr_put(ctx, h);
> + __skb_queue_purge(&tmpq);
> +
> + return ret;
> +}
> +
> +static int sock_un_checkpoint(struct ckpt_ctx *ctx,
> + struct sock *sock,
> + struct ckpt_hdr_socket *h)
> +{
> + struct unix_sock *sk = unix_sk(sock);
> + struct unix_sock *pr = unix_sk(sk->peer);
> + int new;
> + int ret;
> +
> + h->un.this = ckpt_obj_lookup_add(ctx, sk, CKPT_OBJ_SOCK, &new);
> + if (h->un.this < 0)
> + goto out;
> +
> + if (sk->peer)
> + h->un.peer = ckpt_obj_lookup_add(ctx, pr, CKPT_OBJ_SOCK, &new);
> + else
> + h->un.peer = 0;
> +
> + if (h->un.peer < 0) {
> + ret = h->un.peer;
> + goto out;
> + }
> +
> + ret = ckpt_write_obj(ctx, (struct ckpt_hdr *) h);
> + out:
> + return ret;
> +}
> +
> +static int sock_cptrst(struct ckpt_ctx *ctx,
> + struct sock *sock,
> + struct ckpt_hdr_socket *h,
> + int op)
> +{
> + if (sock->sk_socket) {
> + CKPT_COPY(op, h->socket_flags, sock->sk_socket->flags);
> + CKPT_COPY(op, h->socket_state, sock->sk_socket->state);
> + }
> +
> + CKPT_COPY(op, h->reuse, sock->sk_reuse);
> + CKPT_COPY(op, h->shutdown, sock->sk_shutdown);
> + CKPT_COPY(op, h->userlocks, sock->sk_userlocks);
> + CKPT_COPY(op, h->no_check, sock->sk_no_check);
> + CKPT_COPY(op, h->protocol, sock->sk_protocol);
> + CKPT_COPY(op, h->err, sock->sk_err);
> + CKPT_COPY(op, h->err_soft, sock->sk_err_soft);
> + CKPT_COPY(op, h->priority, sock->sk_priority);
> + CKPT_COPY(op, h->rcvlowat, sock->sk_rcvlowat);
> + CKPT_COPY(op, h->backlog, sock->sk_max_ack_backlog);
> + CKPT_COPY(op, h->rcvtimeo, sock->sk_rcvtimeo);
> + CKPT_COPY(op, h->sndtimeo, sock->sk_sndtimeo);
> + CKPT_COPY(op, h->rcvbuf, sock->sk_rcvbuf);
> + CKPT_COPY(op, h->sndbuf, sock->sk_sndbuf);
> + CKPT_COPY(op, h->bound_dev_if, sock->sk_bound_dev_if);
> + CKPT_COPY(op, h->flags, sock->sk_flags);
> + CKPT_COPY(op, h->lingertime, sock->sk_lingertime);
> +
> + return 0;
> +}
> +
> +int __sock_file_checkpoint(struct ckpt_ctx *ctx, struct file *file)
> +{
> + struct socket *socket = file->private_data;
> + struct sock *sock = socket->sk;
> + struct ckpt_hdr_socket *h;
> + int ret = 0;
> +
> + h = ckpt_hdr_get_type(ctx, sizeof(*h), CKPT_HDR_SOCKET);
> + if (!h)
> + return -ENOMEM;
> +
> + h->family = sock->sk_family;
> + h->state = socket->state;
> + h->sock_state = sock->sk_state;
> + h->reuse = sock->sk_reuse;
> + h->type = sock->sk_type;
> + h->protocol = sock->sk_protocol;
> +
> + h->laddr_len = sizeof(h->laddr);
> + h->raddr_len = sizeof(h->raddr);
> +
> + if (socket->ops->getname(socket, &h->laddr, &h->laddr_len, 0)) {
> + ret = -EINVAL;
> + goto out;
> + }
> +
> + if ((h->sock_state != TCP_LISTEN) &&
> + (h->type != SOCK_DGRAM) &&
> + (socket->ops->getname(socket, &h->raddr, &h->raddr_len, 1))) {
> + ret = -EINVAL;
> + goto out;
> + }
> +
> + sock_cptrst(ctx, sock, h, CKPT_CPT);
> +
> + if (h->family == AF_UNIX) {
> + ret = sock_un_checkpoint(ctx, sock, h);
> + if (ret)
> + goto out;
> + } else {
> + ckpt_debug("unsupported socket type %i\n", h->family);
> + ret = EINVAL;
> + goto out;
> + }
> +
> + ret = sock_write_buffers(ctx, &sock->sk_receive_queue);
> + if (ret)
> + goto out;
> +
> + ret = sock_write_buffers(ctx, &sock->sk_write_queue);
> + if (ret)
> + goto out;
> +
> + /* FIXME: write out-of-order queue for TCP */
> + out:
> + ckpt_hdr_put(ctx, h);
> +
> + return ret;
> +}
> +
> +static int sock_read_buffer(struct ckpt_ctx *ctx,
> + struct sock *sock,
> + struct sk_buff **skb)
> +{
> + struct ckpt_hdr *h;
> + int ret = 0;
> + int len;
> +
> + h = ckpt_read_buf_type(ctx, SKB_MAX_ALLOC, CKPT_HDR_SOCKET_BUFFER);
> + if (IS_ERR(h))
> + return PTR_ERR(h);
> +
> + len = h->len - sizeof(*h);
> +
> + *skb = sock_alloc_send_skb(sock, len, MSG_DONTWAIT, &ret);
> + if (*skb == NULL) {
> + ret = ENOMEM;
> + goto out;
> + }
> +
> + memcpy(skb_put(*skb, len), (char *)(h + 1), len);
> + out:
> + ckpt_hdr_put(ctx, h);
> + return ret;
> +}
> +
> +static int sock_read_buffers(struct ckpt_ctx *ctx,
> + struct sock *sock,
> + struct sk_buff_head *queue)
> +{
> + struct ckpt_hdr_socket_buffer *h;
> + int ret = 0;
> + int i;
> +
> + h = ckpt_read_obj_type(ctx, sizeof(*h), CKPT_HDR_SOCKET_BUFFERS);
> + if (IS_ERR(h)) {
> + ret = PTR_ERR(h);
> + goto out;
> + }
> +
> + for (i = 0; i < h->skb_count; i++) {
> + struct sk_buff *skb = NULL;
> +
> + ret = sock_read_buffer(ctx, sock, &skb);
> + if (ret)
> + break;
> +
> + skb_queue_tail(queue, skb);
> + }
> + out:
> + ckpt_hdr_put(ctx, h);
> +
> + return ret;
> +}
> +
> +static int sock_un_restart(struct ckpt_ctx *ctx,
> + struct ckpt_hdr_socket *h,
> + struct socket *socket)
> +{
> + struct sock *peer;
> + int ret = 0;
> +
> + if (h->sock_state == TCP_ESTABLISHED) {
> + peer = ckpt_obj_fetch(ctx, h->un.peer, CKPT_OBJ_SOCK);
> + if (peer && !IS_ERR(peer)) {
> + /* We're last, so join with peer */
> + struct sock *this = socket->sk;
> +
> + sock_hold(this);
> + sock_hold(peer);
> +
> + unix_sk(this)->peer = peer;
> + unix_sk(peer)->peer = this;
> +
> + this->sk_peercred.pid = task_tgid_vnr(current);
> + current_euid_egid(&this->sk_peercred.uid,
> + &this->sk_peercred.gid);
> +
> + peer->sk_peercred.pid = task_tgid_vnr(current);
> + current_euid_egid(&peer->sk_peercred.uid,
> + &peer->sk_peercred.gid);
> + } else {
> + /* We're first, so add our socket and wait for peer */
> + ckpt_obj_insert(ctx, socket->sk, h->un.this,
> + CKPT_OBJ_SOCK);
> + }
> +
> + } else if (h->sock_state == TCP_LISTEN) {
> + ret = socket->ops->bind(socket,
> + (struct sockaddr *)&h->laddr,
> + h->laddr_len);
> + if (ret < 0)
> + goto out;
> +
> + ret = socket->ops->listen(socket, h->backlog);
> + if (ret < 0)
> + goto out;
> + } else
> + ckpt_debug("unsupported UNIX socket state %i\n", h->state);
> +
> + socket->state = h->state;
> + socket->sk->sk_state = h->sock_state;
> + out:
> + return ret;
> +}
> +
> +struct socket *__sock_file_restore(struct ckpt_ctx *ctx,
> + struct ckpt_hdr_socket *h)
> +{
> + struct socket *socket;
> + int ret;
> +
> + ret = sock_create(h->family, h->type, 0, &socket);
> + if (ret < 0)
> + return ERR_PTR(ret);
> +
> + if (h->family == AF_UNIX) {
> + ret = sock_un_restart(ctx, h, socket);
> + ckpt_debug("sock_un_restart: %i\n", ret);
> + } else {
> + ckpt_debug("unsupported family %i\n", h->family);
> + ret = -EINVAL;
> + }
> +
> + if (ret)
> + goto out;
> +
> + ret = sock_read_buffers(ctx, socket->sk, &socket->sk->sk_receive_queue);
> + if (ret)
> + goto out;
> +
> + ret = sock_read_buffers(ctx, socket->sk, &socket->sk->sk_write_queue);
> + if (ret)
> + goto out;
> + out:
> + if (ret) {
> + sock_release(socket);
> + socket = ERR_PTR(ret);
> + }
> +
> + return socket;
> +}
> +
> +int sock_file_checkpoint(struct ckpt_ctx *ctx, void *ptr)
> +{
> + struct ckpt_hdr_file_socket *h;
> + int ret;
> + struct file *file = ptr;
> +
> + h = ckpt_hdr_get_type(ctx, sizeof(*h), CKPT_HDR_FILE);
> + if (!h)
> + return -ENOMEM;
> +
> + h->common.f_type = CKPT_FILE_SOCKET;
> +
> + ret = checkpoint_file_common(ctx, file, &h->common);
> + if (ret < 0)
> + goto out;
> + ret = ckpt_write_obj(ctx, (struct ckpt_hdr *) h);
> + if (ret < 0)
> + goto out;
> +
> + ret = __sock_file_checkpoint(ctx, file);
> + out:
> + ckpt_hdr_put(ctx, h);
> + return ret;
> +}
> +
> +
> --
> 1.6.0.4
>
> _______________________________________________
> Containers mailing list
> Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx
> https://lists.linux-foundation.org/mailman/listinfo/containers
--
Dr Louis Rilling Kerlabs
Skype: louis.rilling Batiment Germanium
Phone: (+33|0) 6 80 89 08 23 80 avenue des Buttes de Coesmes
http://www.kerlabs.com/ 35700 Rennes
Attachment:
signature.asc
Description: Digital signature
_______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers
