2
0
mirror of https://github.com/checkpoint-restore/criu synced 2025-08-22 18:07:57 +00:00
criu/sockets.c

781 lines
18 KiB
C
Raw Permalink Normal View History

#include <unistd.h>
#include <sys/socket.h>
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
#include <linux/netlink.h>
#include <linux/rtnetlink.h>
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
#include <netinet/tcp.h>
#include <errno.h>
#include <linux/if.h>
#include <linux/filter.h>
#include <string.h>
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
#include "libnetlink.h"
#include "sockets.h"
#include "unix_diag.h"
#include "inet_diag.h"
#include "packet_diag.h"
#include "netlink_diag.h"
#include "files.h"
#include "util-pie.h"
#include "sk-packet.h"
#include "namespaces.h"
#include "net.h"
#include "fs-magic.h"
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
#ifndef SOCK_DIAG_BY_FAMILY
#define SOCK_DIAG_BY_FAMILY 20
#endif
#define SK_HASH_SIZE 32
#ifndef SO_GET_FILTER
#define SO_GET_FILTER SO_ATTACH_FILTER
#endif
struct sock_diag_greq {
u8 family;
u8 protocol;
};
struct sock_diag_req {
struct nlmsghdr hdr;
union {
struct unix_diag_req u;
struct inet_diag_req_v2 i;
struct packet_diag_req p;
struct netlink_diag_req n;
struct sock_diag_greq g;
} r;
};
enum socket_cl_bits
{
NETLINK_CL_BIT,
INET_TCP_CL_BIT,
INET_UDP_CL_BIT,
INET_UDPLITE_CL_BIT,
INET6_TCP_CL_BIT,
INET6_UDP_CL_BIT,
INET6_UDPLITE_CL_BIT,
UNIX_CL_BIT,
PACKET_CL_BIT,
_MAX_CL_BIT,
};
#define MAX_CL_BIT (_MAX_CL_BIT - 1)
static DECLARE_BITMAP(socket_cl_bits, MAX_CL_BIT);
static inline
enum socket_cl_bits get_collect_bit_nr(unsigned int family, unsigned int proto)
{
if (family == AF_NETLINK)
return NETLINK_CL_BIT;
if (family == AF_UNIX)
return UNIX_CL_BIT;
if (family == AF_PACKET)
return PACKET_CL_BIT;
if (family == AF_INET) {
if (proto == IPPROTO_TCP)
return INET_TCP_CL_BIT;
if (proto == IPPROTO_UDP)
return INET_UDP_CL_BIT;
if (proto == IPPROTO_UDPLITE)
return INET_UDPLITE_CL_BIT;
}
if (family == AF_INET6) {
if (proto == IPPROTO_TCP)
return INET6_TCP_CL_BIT;
if (proto == IPPROTO_UDP)
return INET6_UDP_CL_BIT;
if (proto == IPPROTO_UDPLITE)
return INET6_UDPLITE_CL_BIT;
}
pr_err("Unknown pair family %d proto %d\n", family, proto);
BUG();
return -1;
}
static void set_collect_bit(unsigned int family, unsigned int proto)
{
enum socket_cl_bits nr;
nr = get_collect_bit_nr(family, proto);
set_bit(nr, socket_cl_bits);
}
bool socket_test_collect_bit(unsigned int family, unsigned int proto)
{
enum socket_cl_bits nr;
nr = get_collect_bit_nr(family, proto);
return test_bit(nr, socket_cl_bits) != 0;
}
static int probe_recv_one(struct nlmsghdr *h, void *arg)
{
pr_err("PROBE RECEIVED\n");
return -1;
}
static int probe_err(int err, void *arg)
{
int expected_err = *(int *)arg;
if (err == expected_err)
return 0;
pr_err("Diag module missing (%d)\n", err);
return err;
}
static inline void probe_diag(int nl, struct sock_diag_req *req, int expected_err)
{
do_rtnl_req(nl, req, req->hdr.nlmsg_len, probe_recv_one, probe_err, &expected_err);
}
void preload_socket_modules()
{
int nl;
struct sock_diag_req req;
/*
* If the task to dump (e.g. an LXC container) has any netlink
* KOBJECT_UEVENT socket open and the _diag modules aren't
* loaded is dumped, criu will freeze the task and then the
* kernel will send it messages on the socket, and then we will
* fail to dump because the socket has pending data. The Real
* Solution is to dump this pending data, but we just make sure
* modules are there beforehand for now so that the first dump
* doesn't fail.
*/
nl = socket(PF_NETLINK, SOCK_RAW, NETLINK_SOCK_DIAG);
if (nl < 0)
return;
pr_info("Probing sock diag modules\n");
memset(&req, 0, sizeof(req));
req.hdr.nlmsg_type = SOCK_DIAG_BY_FAMILY;
req.hdr.nlmsg_seq = CR_NLMSG_SEQ;
/*
* Probe UNIX, netlink and packet diag-s by feeding
* to the kernel request that is shorter than they
* expect, byt still containing the family to make
* sure the family handler is there. The family-level
* diag module would report EINVAL in this case.
*/
req.hdr.nlmsg_len = sizeof(req.hdr) + sizeof(req.r.g);
req.hdr.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST;
req.r.g.family = AF_UNIX;
probe_diag(nl, &req, -EINVAL);
req.r.g.family = AF_PACKET;
probe_diag(nl, &req, -EINVAL);
req.r.g.family = AF_NETLINK;
probe_diag(nl, &req, -EINVAL);
/*
* TCP and UDP(LITE) diags do not support such trick, only
* inet_diag module can be probed like that. For the protocol
* level ones it's OK to request for exact non-existing socket
* and check for ENOENT being reported back as error.
*/
req.hdr.nlmsg_len = sizeof(req.hdr) + sizeof(req.r.i);
req.hdr.nlmsg_flags = NLM_F_REQUEST;
req.r.i.sdiag_family = AF_INET;
req.r.i.sdiag_protocol = IPPROTO_TCP;
probe_diag(nl, &req, -ENOENT);
req.r.i.sdiag_protocol = IPPROTO_UDP; /* UDLITE is merged with UDP */
probe_diag(nl, &req, -ENOENT);
close(nl);
pr_info("Done probing\n");
}
static int dump_bound_dev(int sk, SkOptsEntry *soe)
{
int ret;
char dev[IFNAMSIZ];
socklen_t len = sizeof(dev);
ret = getsockopt(sk, SOL_SOCKET, SO_BINDTODEVICE, &dev, &len);
if (ret) {
pr_perror("Can't get bound dev");
return ret;
}
if (len == 0)
return 0;
pr_debug("\tDumping %s bound dev for sk\n", dev);
soe->so_bound_dev = xmalloc(len);
if (soe->so_bound_dev == NULL)
return -1;
strcpy(soe->so_bound_dev, dev);
return 0;
}
static int restore_bound_dev(int sk, SkOptsEntry *soe)
{
char *n = soe->so_bound_dev;
if (!n)
return 0;
pr_debug("\tBinding socket to %s dev\n", n);
return do_restore_opt(sk, SOL_SOCKET, SO_BINDTODEVICE, n, strlen(n));
}
/*
* Protobuf handles le/be himself, but the sock_filter is not just u64,
* it's a structure and we have to preserve the fields order to be able
* to move socket image across architectures.
*/
static void encode_filter(struct sock_filter *f, u64 *img, int n)
{
int i;
BUILD_BUG_ON(sizeof(*f) != sizeof(*img));
for (i = 0; i < n; i++)
img[i] = ((u64)f[i].code << 48) |
((u64)f[i].jt << 40) |
((u64)f[i].jf << 32) |
((u64)f[i].k << 0);
}
static void decode_filter(u64 *img, struct sock_filter *f, int n)
{
int i;
for (i = 0; i < n; i++) {
f[i].code = img[i] >> 48;
f[i].jt = img[i] >> 40;
f[i].jf = img[i] >> 32;
f[i].k = img[i] >> 0;
}
}
static int dump_socket_filter(int sk, SkOptsEntry *soe)
{
socklen_t len = 0;
int ret;
struct sock_filter *flt;
ret = getsockopt(sk, SOL_SOCKET, SO_GET_FILTER, NULL, &len);
if (ret) {
pr_perror("Can't get socket filter len");
return ret;
}
if (!len) {
pr_info("No filter for socket\n");
return 0;
}
flt = xmalloc(len * sizeof(*flt));
if (!flt)
return -1;
ret = getsockopt(sk, SOL_SOCKET, SO_GET_FILTER, flt, &len);
if (ret) {
pr_perror("Can't get socket filter");
xfree(flt);
return ret;
}
soe->so_filter = xmalloc(len * sizeof(*soe->so_filter));
if (!soe->so_filter) {
xfree(flt);
return -1;
}
encode_filter(flt, soe->so_filter, len);
soe->n_so_filter = len;
xfree(flt);
return 0;
}
static int restore_socket_filter(int sk, SkOptsEntry *soe)
{
int ret;
struct sock_fprog sfp;
if (!soe->n_so_filter)
return 0;
pr_info("Restoring socket filter\n");
sfp.len = soe->n_so_filter;
sfp.filter = xmalloc(soe->n_so_filter * sfp.len);
if (!sfp.filter)
return -1;
decode_filter(soe->so_filter, sfp.filter, sfp.len);
ret = restore_opt(sk, SOL_SOCKET, SO_ATTACH_FILTER, &sfp);
xfree(sfp.filter);
return ret;
}
static struct socket_desc *sockets[SK_HASH_SIZE];
struct socket_desc *lookup_socket(int ino, int family, int proto)
{
struct socket_desc *sd;
if (!socket_test_collect_bit(family, proto)) {
pr_err("Sockets (family %d, proto %d) are not collected\n",
family, proto);
return ERR_PTR(-EINVAL);
}
pr_debug("\tSearching for socket %x (family %d)\n", ino, family);
for (sd = sockets[ino % SK_HASH_SIZE]; sd; sd = sd->next)
if (sd->ino == ino) {
BUG_ON(sd->family != family);
return sd;
}
return NULL;
}
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
int sk_collect_one(int ino, int family, struct socket_desc *d)
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
{
struct socket_desc **chain;
d->ino = ino;
d->family = family;
d->already_dumped = 0;
chain = &sockets[ino % SK_HASH_SIZE];
d->next = *chain;
*chain = d;
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
return 0;
}
int do_restore_opt(int sk, int level, int name, void *val, int len)
{
if (setsockopt(sk, level, name, val, len) < 0) {
pr_perror("Can't set %d:%d (len %d)", level, name, len);
return -1;
}
return 0;
}
static int sk_setbufs(void *arg, int fd)
{
u32 *buf = (u32 *)arg;
if (restore_opt(fd, SOL_SOCKET, SO_SNDBUFFORCE, &buf[0]))
return -1;
if (restore_opt(fd, SOL_SOCKET, SO_RCVBUFFORCE, &buf[1]))
return -1;
return 0;
}
/*
* Set sizes of buffers to maximum and prevent blocking
* Caller of this fn should call other socket restoring
* routines to drop the non-blocking and set proper send
* and receive buffers.
*/
int restore_prepare_socket(int sk)
{
int flags;
/* In kernel a bufsize has type int and a value is doubled. */
u32 maxbuf[2] = { INT_MAX / 2, INT_MAX / 2 };
if (userns_call(sk_setbufs, 0, maxbuf, sizeof(maxbuf), sk))
return -1;
/* Prevent blocking on restore */
flags = fcntl(sk, F_GETFL, 0);
if (flags == -1) {
pr_perror("Unable to get flags for %d", sk);
return -1;
}
if (fcntl(sk, F_SETFL, flags | O_NONBLOCK) ) {
pr_perror("Unable to set O_NONBLOCK for %d", sk);
return -1;
}
return 0;
}
int restore_socket_opts(int sk, SkOptsEntry *soe)
{
int ret = 0, val;
struct timeval tv;
u32 bufs[2] = { soe->so_sndbuf, soe->so_rcvbuf };
pr_info("%d restore sndbuf %d rcv buf %d\n", sk, soe->so_sndbuf, soe->so_rcvbuf);
/* setsockopt() multiplies the input values by 2 */
ret |= userns_call(sk_setbufs, UNS_ASYNC, bufs, sizeof(bufs), sk);
if (soe->has_so_priority) {
pr_debug("\trestore priority %d for socket\n", soe->so_priority);
ret |= restore_opt(sk, SOL_SOCKET, SO_PRIORITY, &soe->so_priority);
}
if (soe->has_so_rcvlowat) {
pr_debug("\trestore rcvlowat %d for socket\n", soe->so_rcvlowat);
ret |= restore_opt(sk, SOL_SOCKET, SO_RCVLOWAT, &soe->so_rcvlowat);
}
if (soe->has_so_mark) {
pr_debug("\trestore mark %d for socket\n", soe->so_mark);
ret |= restore_opt(sk, SOL_SOCKET, SO_MARK, &soe->so_mark);
}
if (soe->has_so_passcred && soe->so_passcred) {
val = 1;
pr_debug("\tset passcred for socket\n");
ret |= restore_opt(sk, SOL_SOCKET, SO_PASSCRED, &val);
}
if (soe->has_so_passsec && soe->so_passsec) {
val = 1;
pr_debug("\tset passsec for socket\n");
ret |= restore_opt(sk, SOL_SOCKET, SO_PASSSEC, &val);
}
if (soe->has_so_dontroute && soe->so_dontroute) {
val = 1;
pr_debug("\tset dontroute for socket\n");
ret |= restore_opt(sk, SOL_SOCKET, SO_DONTROUTE, &val);
}
if (soe->has_so_no_check && soe->so_no_check) {
val = 1;
pr_debug("\tset no_check for socket\n");
ret |= restore_opt(sk, SOL_SOCKET, SO_NO_CHECK, &val);
}
tv.tv_sec = soe->so_snd_tmo_sec;
tv.tv_usec = soe->so_snd_tmo_usec;
ret |= restore_opt(sk, SOL_SOCKET, SO_SNDTIMEO, &tv);
tv.tv_sec = soe->so_rcv_tmo_sec;
tv.tv_usec = soe->so_rcv_tmo_usec;
ret |= restore_opt(sk, SOL_SOCKET, SO_RCVTIMEO, &tv);
ret |= restore_bound_dev(sk, soe);
ret |= restore_socket_filter(sk, soe);
/* The restore of SO_REUSEADDR depends on type of socket */
return ret;
}
int do_dump_opt(int sk, int level, int name, void *val, int len)
{
socklen_t aux = len;
if (getsockopt(sk, level, name, val, &aux) < 0) {
pr_perror("Can't get %d:%d opt", level, name);
return -1;
}
if (aux != len) {
pr_err("Len mismatch on %d:%d : %d, want %d\n",
level, name, aux, len);
return -1;
}
return 0;
}
int dump_socket_opts(int sk, SkOptsEntry *soe)
{
int ret = 0, val;
struct timeval tv;
ret |= dump_opt(sk, SOL_SOCKET, SO_SNDBUF, &soe->so_sndbuf);
ret |= dump_opt(sk, SOL_SOCKET, SO_RCVBUF, &soe->so_rcvbuf);
soe->has_so_priority = true;
ret |= dump_opt(sk, SOL_SOCKET, SO_PRIORITY, &soe->so_priority);
soe->has_so_rcvlowat = true;
ret |= dump_opt(sk, SOL_SOCKET, SO_RCVLOWAT, &soe->so_rcvlowat);
soe->has_so_mark = true;
ret |= dump_opt(sk, SOL_SOCKET, SO_MARK, &soe->so_mark);
ret |= dump_opt(sk, SOL_SOCKET, SO_SNDTIMEO, &tv);
soe->so_snd_tmo_sec = tv.tv_sec;
soe->so_snd_tmo_usec = tv.tv_usec;
ret |= dump_opt(sk, SOL_SOCKET, SO_RCVTIMEO, &tv);
soe->so_rcv_tmo_sec = tv.tv_sec;
soe->so_rcv_tmo_usec = tv.tv_usec;
ret |= dump_opt(sk, SOL_SOCKET, SO_REUSEADDR, &val);
soe->reuseaddr = val ? true : false;
soe->has_reuseaddr = true;
ret |= dump_opt(sk, SOL_SOCKET, SO_PASSCRED, &val);
soe->has_so_passcred = true;
soe->so_passcred = val ? true : false;
ret |= dump_opt(sk, SOL_SOCKET, SO_PASSSEC, &val);
soe->has_so_passsec = true;
soe->so_passsec = val ? true : false;
ret |= dump_opt(sk, SOL_SOCKET, SO_DONTROUTE, &val);
soe->has_so_dontroute = true;
soe->so_dontroute = val ? true : false;
ret |= dump_opt(sk, SOL_SOCKET, SO_NO_CHECK, &val);
soe->has_so_no_check = true;
soe->so_no_check = val ? true : false;
ret |= dump_bound_dev(sk, soe);
ret |= dump_socket_filter(sk, soe);
return ret;
}
void release_skopts(SkOptsEntry *soe)
{
xfree(soe->so_filter);
xfree(soe->so_bound_dev);
}
int dump_socket(struct fd_parms *p, int lfd, struct cr_img *img)
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
{
int family;
const struct fdtype_ops *ops;
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
if (dump_opt(lfd, SOL_SOCKET, SO_DOMAIN, &family))
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
return -1;
switch (family) {
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
case AF_UNIX:
ops = &unix_dump_ops;
break;
case AF_INET:
ops = &inet_dump_ops;
break;
case AF_INET6:
ops = &inet6_dump_ops;
break;
case AF_PACKET:
ops = &packet_dump_ops;
break;
case AF_NETLINK:
ops = &netlink_dump_ops;
break;
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
default:
pr_err("BUG! Unknown socket collected (family %d)\n", family);
return -1;
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
}
return do_dump_gen_file(p, lfd, ops, img);
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
}
static int inet_receive_one(struct nlmsghdr *h, void *arg)
{
struct inet_diag_req_v2 *i = arg;
int type;
switch (i->sdiag_protocol) {
case IPPROTO_TCP:
type = SOCK_STREAM;
break;
case IPPROTO_UDP:
case IPPROTO_UDPLITE:
type = SOCK_DGRAM;
break;
default:
BUG_ON(1);
return -1;
}
return inet_collect_one(h, i->sdiag_family, type);
}
static int do_collect_req(int nl, struct sock_diag_req *req, int size,
int (*receive_callback)(struct nlmsghdr *h, void *), void *arg)
{
int tmp;
tmp = do_rtnl_req(nl, req, size, receive_callback, NULL, arg);
if (tmp == 0)
set_collect_bit(req->r.n.sdiag_family, req->r.n.sdiag_protocol);
return tmp;
}
int collect_sockets(struct ns_id *ns)
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
{
int err = 0, tmp;
int nl = ns->net.nlsk;
struct sock_diag_req req;
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
memset(&req, 0, sizeof(req));
req.hdr.nlmsg_len = sizeof(req);
req.hdr.nlmsg_type = SOCK_DIAG_BY_FAMILY;
req.hdr.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST;
req.hdr.nlmsg_seq = CR_NLMSG_SEQ;
/* Collect UNIX sockets */
req.r.u.sdiag_family = AF_UNIX;
req.r.u.udiag_states = -1; /* All */
req.r.u.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_VFS |
UDIAG_SHOW_PEER | UDIAG_SHOW_ICONS |
UDIAG_SHOW_RQLEN;
tmp = do_collect_req(nl, &req, sizeof(req), unix_receive_one, NULL);
if (tmp)
err = tmp;
/* Collect IPv4 TCP sockets */
req.r.i.sdiag_family = AF_INET;
req.r.i.sdiag_protocol = IPPROTO_TCP;
req.r.i.idiag_ext = 0;
/* Only listening and established sockets supported yet */
req.r.i.idiag_states = (1 << TCP_LISTEN) | (1 << TCP_ESTABLISHED);
tmp = do_collect_req(nl, &req, sizeof(req), inet_receive_one, &req.r.i);
if (tmp)
err = tmp;
/* Collect IPv4 UDP sockets */
req.r.i.sdiag_family = AF_INET;
req.r.i.sdiag_protocol = IPPROTO_UDP;
req.r.i.idiag_ext = 0;
req.r.i.idiag_states = -1; /* All */
tmp = do_collect_req(nl, &req, sizeof(req), inet_receive_one, &req.r.i);
if (tmp)
err = tmp;
/* Collect IPv4 UDP-lite sockets */
req.r.i.sdiag_family = AF_INET;
req.r.i.sdiag_protocol = IPPROTO_UDPLITE;
req.r.i.idiag_ext = 0;
req.r.i.idiag_states = -1; /* All */
tmp = do_collect_req(nl, &req, sizeof(req), inet_receive_one, &req.r.i);
if (tmp)
err = tmp;
/* Collect IPv6 TCP sockets */
req.r.i.sdiag_family = AF_INET6;
req.r.i.sdiag_protocol = IPPROTO_TCP;
req.r.i.idiag_ext = 0;
/* Only listening sockets supported yet */
req.r.i.idiag_states = (1 << TCP_LISTEN) | (1 << TCP_ESTABLISHED);
tmp = do_collect_req(nl, &req, sizeof(req), inet_receive_one, &req.r.i);
if (tmp)
err = tmp;
/* Collect IPv6 UDP sockets */
req.r.i.sdiag_family = AF_INET6;
req.r.i.sdiag_protocol = IPPROTO_UDP;
req.r.i.idiag_ext = 0;
req.r.i.idiag_states = -1; /* All */
tmp = do_collect_req(nl, &req, sizeof(req), inet_receive_one, &req.r.i);
if (tmp)
err = tmp;
/* Collect IPv6 UDP-lite sockets */
req.r.i.sdiag_family = AF_INET6;
req.r.i.sdiag_protocol = IPPROTO_UDPLITE;
req.r.i.idiag_ext = 0;
req.r.i.idiag_states = -1; /* All */
tmp = do_collect_req(nl, &req, sizeof(req), inet_receive_one, &req.r.i);
if (tmp)
err = tmp;
req.r.p.sdiag_family = AF_PACKET;
req.r.p.sdiag_protocol = 0;
req.r.p.pdiag_show = PACKET_SHOW_INFO | PACKET_SHOW_MCLIST |
PACKET_SHOW_FANOUT | PACKET_SHOW_RING_CFG;
tmp = do_collect_req(nl, &req, sizeof(req), packet_receive_one, NULL);
if (tmp) {
pr_warn("The current kernel doesn't support packet_diag\n");
if (ns->pid == 0 || tmp != -ENOENT) /* Fedora 19 */
err = tmp;
}
req.r.n.sdiag_family = AF_NETLINK;
req.r.n.sdiag_protocol = NDIAG_PROTO_ALL;
req.r.n.ndiag_show = NDIAG_SHOW_GROUPS;
tmp = do_collect_req(nl, &req, sizeof(req), netlink_receive_one, NULL);
if (tmp) {
pr_warn("The current kernel doesn't support netlink_diag\n");
if (ns->pid == 0 || tmp != -ENOENT) /* Fedora 19 */
err = tmp;
}
/* don't need anymore */
close(nl);
ns->net.nlsk = -1;
if (ns->pid == getpid()) {
/*
* If netns isn't dumped, criu will fail only
* if an unsupported socket will be really dumped.
*/
pr_info("Uncollected sockets! Will probably fail later.\n");
err = 0;
}
Unix sockets initial support Currently it can only work with stream sockets, which have no skbs in queues (listening or established -- both work OK). The cpt part uses the sock_diag engine that was merged to Dave recently to collect sockets. Then it dumps sockets by checking the filesystem ID of a failed-to-open through /proc/pid/fd descriptors (sockets do not allow for such tricks with opens through proc) against SOCKFS_TYPE. The rst part is more tricky. Listen sockets are just restored, this is simple. Connected sockets are restored like this: 1. One end establishes a listening anon socket at the desired descriptor; 2. The other end just creates a socket at the desired descriptor; 3. All sockets, that are to be connect()-ed call connect. Unix sockets do not block connect() till the accept() time and thus we continue with... 4. ... all listening sockets call accept() and ... dup2 the new fd into the accepting end. There's a problem with this approach -- socket names are not preserved, but looking into our OpenVZ implementation I think this is OK for existing apps. What should be done next is: 1. Need to merge the file IDs patches in our tree and make Andrey to support files sharing. This will solve the sk = socket(); fork(); case. Currently it simply doesn't work :( 2. Need to add support for DGRAM sockets -- I wrote comment how to do it in the can_dump_unix_sk() 3. Need to add support for in-flight connections 4. Implement support for UDP sockets (quite simple) 5. Implement support for listening TCP sockets (also not very complex) 6. Implement support for connected TCP scokets (hard one, Tejun's patches are not very good for this from my POV) Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we have the plans page yet?). Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till you do the shared files support for sockets. Signed-off-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
2011-12-26 22:12:03 +04:00
return err;
}
static inline char *unknown(u32 val)
{
static char unk[12];
snprintf(unk, sizeof(unk), "x%d", val);
return unk;
}
char *skfamily2s(u32 f)
{
if (f == AF_INET)
return " inet";
else if (f == AF_INET6)
return "inet6";
else
return unknown(f);
}
char *sktype2s(u32 t)
{
if (t == SOCK_STREAM)
return "stream";
else if (t == SOCK_DGRAM)
return " dgram";
else
return unknown(t);
}
char *skproto2s(u32 p)
{
if (p == IPPROTO_UDP)
return "udp";
else if (p == IPPROTO_UDPLITE)
return "udpl";
else if (p == IPPROTO_TCP)
return "tcp";
else
return unknown(p);
}
char *skstate2s(u32 state)
{
if (state == TCP_ESTABLISHED)
return " estab";
else if (state == TCP_CLOSE)
return "closed";
else if (state == TCP_LISTEN)
return "listen";
else
return unknown(state);
}