checkpoint: IPv4 listening sockets dumping support

crtools.c

@@ -85,6 +85,13 @@ struct cr_fd_desc_tmpl fdset_template[CR_FD_MAX] = {
 		.fmt	= FMT_FNAME_UNIXSK,
 		.magic	= UNIXSK_MAGIC,
 	},
+
+	/* info about inet sockets */
+	[CR_FD_INETSK] = {
+		.fmt	= FMT_FNAME_INETSK,
+		.magic	= INETSK_MAGIC,
+	},
+
 };
 
 static struct cr_fdset *alloc_cr_fdset(void)

include/crtools.h

@@ -24,6 +24,7 @@ enum {
 	CR_FD_SHMEM,
 	CR_FD_SIGACT,
 	CR_FD_UNIXSK,
+	CR_FD_INETSK,
 
 	CR_FD_MAX
 };
@@ -60,6 +61,7 @@ extern struct cr_fd_desc_tmpl fdset_template[CR_FD_MAX];
 #define FMT_FNAME_VMAS		"vmas-%d.img"
 #define FMT_FNAME_SIGACTS	"sigacts-%d.img"
 #define FMT_FNAME_UNIXSK	"unixsk-%d.img"
+#define FMT_FNAME_INETSK	"inetsk-%d.img"
 
 extern int get_image_path(char *path, int size, const char *fmt, int pid);
 

include/image.h

@@ -13,6 +13,7 @@
 #define PIPES_MAGIC	0x05055050
 #define SIGACT_MAGIC	0x60606060
 #define UNIXSK_MAGIC	0x07070707
+#define INETSK_MAGIC	0x08080808
 
 #define FDINFO_FD	1
 #define FDINFO_MAP	2
@@ -69,6 +70,19 @@ struct unix_sk_entry {
 	u8	name[0];
 } __packed;
 
+struct inet_sk_entry {
+	u32	fd;
+	u32	id;
+	u8	family;
+	u8	type;
+	u8	proto;
+	u8	state;
+	u16	src_port;
+	u8	pad[2];
+	u32	backlog;
+	u32	src_addr[4];
+} __packed;
+
 struct vma_entry {
 	u64	start;
 	u64	end;

include/inet_diag.h

+#ifndef _INET_DIAG_H_
+#define _INET_DIAG_H_ 1
+
+#include <linux/types.h>
+
+/* Just some random number */
+#define TCPDIAG_GETSOCK 18
+#define DCCPDIAG_GETSOCK 19
+
+#define INET_DIAG_GETSOCK_MAX 24
+
+/* Socket identity */
+struct inet_diag_sockid {
+	__be16	idiag_sport;
+	__be16	idiag_dport;
+	__be32	idiag_src[4];
+	__be32	idiag_dst[4];
+	__u32	idiag_if;
+	__u32	idiag_cookie[2];
+#define INET_DIAG_NOCOOKIE (~0U)
+};
+
+/* Request structure */
+
+struct inet_diag_req_compat {
+	__u8	idiag_family;		/* Family of addresses. */
+	__u8	idiag_src_len;
+	__u8	idiag_dst_len;
+	__u8	idiag_ext;		/* Query extended information */
+
+	struct inet_diag_sockid id;
+
+	__u32	idiag_states;		/* States to dump */
+	__u32	idiag_dbs;		/* Tables to dump (NI) */
+};
+
+struct inet_diag_req {
+	__u8	sdiag_family;
+	__u8	sdiag_protocol;
+	__u8	idiag_ext;
+	__u8	pad;
+	__u32	idiag_states;
+	struct inet_diag_sockid id;
+};
+
+enum {
+	INET_DIAG_REQ_NONE,
+	INET_DIAG_REQ_BYTECODE,
+};
+
+#define INET_DIAG_REQ_MAX INET_DIAG_REQ_BYTECODE
+
+/* Bytecode is sequence of 4 byte commands followed by variable arguments.
+ * All the commands identified by "code" are conditional jumps forward:
+ * to offset cc+"yes" or to offset cc+"no". "yes" is supposed to be
+ * length of the command and its arguments.
+ */
+ 
+struct inet_diag_bc_op {
+	unsigned char	code;
+	unsigned char	yes;
+	unsigned short	no;
+};
+
+enum {
+	INET_DIAG_BC_NOP,
+	INET_DIAG_BC_JMP,
+	INET_DIAG_BC_S_GE,
+	INET_DIAG_BC_S_LE,
+	INET_DIAG_BC_D_GE,
+	INET_DIAG_BC_D_LE,
+	INET_DIAG_BC_AUTO,
+	INET_DIAG_BC_S_COND,
+	INET_DIAG_BC_D_COND,
+};
+
+struct inet_diag_hostcond {
+	__u8	family;
+	__u8	prefix_len;
+	int	port;
+	__be32	addr[0];
+};
+
+/* Base info structure. It contains socket identity (addrs/ports/cookie)
+ * and, alas, the information shown by netstat. */
+struct inet_diag_msg {
+	__u8	idiag_family;
+	__u8	idiag_state;
+	__u8	idiag_timer;
+	__u8	idiag_retrans;
+
+	struct inet_diag_sockid id;
+
+	__u32	idiag_expires;
+	__u32	idiag_rqueue;
+	__u32	idiag_wqueue;
+	__u32	idiag_uid;
+	__u32	idiag_inode;
+};
+
+/* Extensions */
+
+enum {
+	INET_DIAG_NONE,
+	INET_DIAG_MEMINFO,
+	INET_DIAG_INFO,
+	INET_DIAG_VEGASINFO,
+	INET_DIAG_CONG,
+	INET_DIAG_TOS,
+	INET_DIAG_TCLASS,
+	INET_DIAG_SKMEMINFO,
+};
+
+#define INET_DIAG_MAX INET_DIAG_SKMEMINFO
+
+
+/* INET_DIAG_MEM */
+
+struct inet_diag_meminfo {
+	__u32	idiag_rmem;
+	__u32	idiag_wmem;
+	__u32	idiag_fmem;
+	__u32	idiag_tmem;
+};
+
+/* INET_DIAG_VEGASINFO */
+
+struct tcpvegas_info {
+	__u32	tcpv_enabled;
+	__u32	tcpv_rttcnt;
+	__u32	tcpv_rtt;
+	__u32	tcpv_minrtt;
+};
+
+#endif /* _INET_DIAG_H_ */

sockets.c

@@ -11,6 +11,7 @@
 #include <errno.h>
 #include <unistd.h>
 #include <limits.h>
+#include <arpa/inet.h>
 
 #include "types.h"
 #include "libnetlink.h"
@@ -19,6 +20,7 @@
 #include "image.h"
 #include "crtools.h"
 #include "util.h"
+#include "inet_diag.h"
 
 static char buf[4096];
 
@@ -52,6 +54,19 @@ struct unix_sk_desc {
 	unsigned int		*icons;
 };
 
+#define INET_ADDR_LEN		40
+
+struct inet_sk_desc {
+	struct socket_desc	sd;
+	unsigned int		type;
+	unsigned int		proto;
+	unsigned int		src_port;
+	unsigned int		state;
+	unsigned int		rqlen;
+	unsigned int		wqlen;
+	unsigned int		src_addr[4];
+};
+
 #define SK_HASH_SIZE	32
 static struct socket_desc *sockets[SK_HASH_SIZE];
 
@@ -78,6 +93,36 @@ static int sk_collect_one(int ino, int family, struct socket_desc *d)
 	return 0;
 }
 
+static void show_one_inet(char *act, struct inet_sk_desc *sk)
+{
+	char src_addr[INET_ADDR_LEN] = "<unknown>";
+
+	if (inet_ntop(AF_INET, (void *)sk->src_addr, src_addr,
+		      INET_ADDR_LEN) == NULL) {
+		pr_err("Failed to translate address: %d\n", errno);
+	}
+
+	dprintk("\t%s: ino %d family %d type %d port %d "
+		"state %d src_addr %s\n",
+		act, sk->sd.ino, sk->sd.family, sk->type, sk->src_port,
+		sk->state, src_addr);
+}
+
+static void show_one_inet_img(char *act, struct inet_sk_entry *e)
+{
+	char src_addr[INET_ADDR_LEN] = "<unknown>";
+
+	if (inet_ntop(AF_INET, (void *)e->src_addr, src_addr,
+		      INET_ADDR_LEN) == NULL) {
+		pr_err("Failed to translate address: %d\n", errno);
+	}
+
+	dprintk("\t%s: fd %d family %d type %d proto %d port %d "
+		"state %d src_addr %d bytes\n",
+		act, e->fd, e->family, e->type, e->proto, e->src_port, e->state,
+		src_addr);
+}
+
 static void show_one_unix(char *act, struct unix_sk_desc *sk)
 {
 	dprintk("\t%s: ino %d type %d state %d name %s\n",
@@ -90,6 +135,69 @@ static void show_one_unix_img(char *act, struct unix_sk_entry *e)
 		act, e->fd, e->type, e->state, e->namelen);
 }
 
+static int can_dump_inet_sk(struct inet_sk_desc *sk)
+{
+	if (sk->sd.family != AF_INET) {
+		pr_err("Only IPv4 sockets for now\n");
+		return 0;
+	}
+
+	if (sk->type != SOCK_STREAM) {
+		pr_err("Only stream inet sockets for now\n");
+		return 0;
+	}
+
+	switch (sk->state) {
+	case TCP_LISTEN:
+		if (sk->rqlen != 0) {
+			/*
+			 * Currently the ICONS nla reports the conn
+			 * requests for listen sockets. Need to pick
+			 * those up and fix the connect job respectively
+			 */
+			pr_err("In-flight connection (l)\n");
+			return 0;
+		}
+		break;
+	default:
+		pr_err("Unknown state %d\n", sk->state);
+		return 0;
+	}
+
+	return 1;
+}
+
+static int dump_one_inet(struct socket_desc *_sk, int fd, struct cr_fdset *cr_fdset)
+{
+	struct inet_sk_desc *sk = (struct inet_sk_desc *)_sk;
+	struct inet_sk_entry ie;
+
+	if (!can_dump_inet_sk(sk))
+		goto err;
+
+	memset(&ie, 0, sizeof(ie));
+
+	ie.fd		= fd;
+	ie.id		= sk->sd.ino;
+	ie.family	= sk->sd.family;
+	ie.type		= sk->type;
+	ie.proto	= sk->proto;
+	ie.state	= sk->state;
+	ie.src_port	= sk->src_port;
+	ie.backlog	= sk->wqlen;
+	memcpy(ie.src_addr, sk->src_addr, sizeof(u32) * 4);
+
+	write_ptr_safe(cr_fdset->fds[CR_FD_INETSK], &ie, err);
+
+	pr_info("Dumping inet socket at %d\n", fd);
+	show_one_inet("Dumping", sk);
+	show_one_inet_img("Dumped", &ie);
+	return 0;
+
+err:
+	return -1;
+}
+
 static int can_dump_unix_sk(struct unix_sk_desc *sk)
 {
 	if (sk->type != SOCK_STREAM &&
@@ -209,6 +317,8 @@ int try_dump_socket(pid_t pid, int fd, struct cr_fdset *cr_fdset)
 	switch (sk->family) {
 	case AF_UNIX:
 		return dump_one_unix(sk, fd, cr_fdset);
+	case AF_INET:
+		return dump_one_inet(sk, fd, cr_fdset);
 	default:
 		pr_err("BUG! Unknown socket collected\n");
 		break;
@@ -217,6 +327,36 @@ int try_dump_socket(pid_t pid, int fd, struct cr_fdset *cr_fdset)
 	return -1;
 }
 
+static int inet_tcp_collect_one(struct inet_diag_msg *m, struct rtattr **tb)
+{
+	struct inet_sk_desc *d;
+
+	d = xzalloc(sizeof(*d));
+	if (!d)
+		return -1;
+
+	d->type = SOCK_STREAM;
+	d->proto = IPPROTO_TCP;
+	d->src_port = ntohs(m->id.idiag_sport);
+	d->state = m->idiag_state;
+	d->rqlen = m->idiag_rqueue;
+	d->wqlen = m->idiag_wqueue;
+	memcpy(d->src_addr, m->id.idiag_src, sizeof(u32) * 4);
+
+	return sk_collect_one(m->idiag_inode, AF_INET, &d->sd);
+}
+
+static int inet_tcp_receive_one(struct nlmsghdr *h)
+{
+	struct inet_diag_msg *m = NLMSG_DATA(h);
+	struct rtattr *tb[INET_DIAG_MAX+1];
+
+	parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(m + 1),
+		     h->nlmsg_len - NLMSG_LENGTH(sizeof(*m)));
+
+	return inet_tcp_collect_one(m, tb);
+}
+
 static int unix_collect_one(struct unix_diag_msg *m, struct rtattr **tb)
 {
 	struct unix_sk_desc *d, **h;
@@ -391,6 +531,7 @@ int collect_sockets(void)
 		struct nlmsghdr hdr;
 		union {
 			struct unix_diag_req u;
+			struct inet_diag_req i;
 		} r;
 	} req;
 
@@ -416,6 +557,14 @@ int collect_sockets(void)
 	if (err)
 		goto out;
 
+	/* Collect IPv4 TCP sockets */
+	req.r.i.sdiag_family	= AF_INET;
+	req.r.i.sdiag_protocol	= IPPROTO_TCP;
+	req.r.i.idiag_ext	= 1 << (INET_DIAG_INFO - 1);
+	/* Only listening sockets supported yet */
+	req.r.i.idiag_states	= 1 << TCP_LISTEN;
+	err = collect_sockets_nl(nl, &req, sizeof(req), inet_tcp_receive_one);
+
 out:
 	close(nl);
 	return err;