Hello,
On Sat, 26 Apr 2008, Andrew Morton wrote:
Below is 2nd version that handles templates properly, which
should be the actual problem. Large timeouts in backup look like a
problem in incorrect ipvsadm binary, timeouts are not converted from
jiffies. Patch follows:
Result from ip_vs_proto_get() should be checked because
protocol value can be invalid or unsupported in backup. But
for valid message we should not fail for templates which use
IPPROTO_IP. Also, add checks to validate message limits and
connection state. Show NONE for templates.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
---
diff -urp v2.6.25/linux/include/net/ip_vs.h linux/include/net/ip_vs.h
--- v2.6.25/linux/include/net/ip_vs.h 2008-04-17 09:58:08.000000000 +0300
+++ linux/include/net/ip_vs.h 2008-04-26 20:10:46.000000000 +0300
@@ -405,7 +405,8 @@ struct sk_buff;
struct ip_vs_protocol {
struct ip_vs_protocol *next;
char *name;
- __u16 protocol;
+ u16 protocol;
+ u16 num_states;
int dont_defrag;
atomic_t appcnt; /* counter of proto app incs */
int *timeout_table; /* protocol timeout table */
diff -urp v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto.c linux/net/ipv4/ipvs/ip_vs_proto.c
--- v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto.c 2008-04-17 09:58:09.000000000 +0300
+++ linux/net/ipv4/ipvs/ip_vs_proto.c 2008-04-27 10:31:24.000000000 +0300
@@ -148,7 +148,7 @@ const char * ip_vs_state_name(__u16 prot
struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
if (pp == NULL || pp->state_name == NULL)
- return "ERR!";
+ return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
return pp->state_name(state);
}
diff -urp v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto_ah.c linux/net/ipv4/ipvs/ip_vs_proto_ah.c
--- v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto_ah.c 2007-07-10 09:18:43.000000000 +0300
+++ linux/net/ipv4/ipvs/ip_vs_proto_ah.c 2008-04-26 20:14:17.000000000 +0300
@@ -160,6 +160,7 @@ static void ah_exit(struct ip_vs_protoco
struct ip_vs_protocol ip_vs_protocol_ah = {
.name = "AH",
.protocol = IPPROTO_AH,
+ .num_states = 1,
.dont_defrag = 1,
.init = ah_init,
.exit = ah_exit,
diff -urp v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto_esp.c linux/net/ipv4/ipvs/ip_vs_proto_esp.c
--- v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto_esp.c 2008-04-17 09:58:09.000000000 +0300
+++ linux/net/ipv4/ipvs/ip_vs_proto_esp.c 2008-04-26 20:14:39.000000000 +0300
@@ -159,6 +159,7 @@ static void esp_exit(struct ip_vs_protoc
struct ip_vs_protocol ip_vs_protocol_esp = {
.name = "ESP",
.protocol = IPPROTO_ESP,
+ .num_states = 1,
.dont_defrag = 1,
.init = esp_init,
.exit = esp_exit,
diff -urp v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto_tcp.c linux/net/ipv4/ipvs/ip_vs_proto_tcp.c
--- v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto_tcp.c 2008-01-25 10:45:06.000000000 +0200
+++ linux/net/ipv4/ipvs/ip_vs_proto_tcp.c 2008-04-26 20:15:45.000000000 +0300
@@ -594,6 +594,7 @@ static void ip_vs_tcp_exit(struct ip_vs_
struct ip_vs_protocol ip_vs_protocol_tcp = {
.name = "TCP",
.protocol = IPPROTO_TCP,
+ .num_states = IP_VS_TCP_S_LAST,
.dont_defrag = 0,
.appcnt = ATOMIC_INIT(0),
.init = ip_vs_tcp_init,
diff -urp v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto_udp.c linux/net/ipv4/ipvs/ip_vs_proto_udp.c
--- v2.6.25/linux/net/ipv4/ipvs/ip_vs_proto_udp.c 2008-01-25 10:45:06.000000000 +0200
+++ linux/net/ipv4/ipvs/ip_vs_proto_udp.c 2008-04-26 20:15:07.000000000 +0300
@@ -409,6 +409,7 @@ static void udp_exit(struct ip_vs_protoc
struct ip_vs_protocol ip_vs_protocol_udp = {
.name = "UDP",
.protocol = IPPROTO_UDP,
+ .num_states = IP_VS_UDP_S_LAST,
.dont_defrag = 0,
.init = udp_init,
.exit = udp_exit,
diff -urp v2.6.25/linux/net/ipv4/ipvs/ip_vs_sync.c linux/net/ipv4/ipvs/ip_vs_sync.c
--- v2.6.25/linux/net/ipv4/ipvs/ip_vs_sync.c 2008-04-17 09:58:09.000000000 +0300
+++ linux/net/ipv4/ipvs/ip_vs_sync.c 2008-04-27 11:05:48.000000000 +0300
@@ -288,11 +288,16 @@ static void ip_vs_process_message(const
char *p;
int i;
+ if (buflen < sizeof(struct ip_vs_sync_mesg)) {
+ IP_VS_ERR_RL("sync message header too short\n");
+ return;
+ }
+
/* Convert size back to host byte order */
m->size = ntohs(m->size);
if (buflen != m->size) {
- IP_VS_ERR("bogus message\n");
+ IP_VS_ERR_RL("bogus sync message size\n");
return;
}
@@ -307,9 +312,47 @@ static void ip_vs_process_message(const
for (i=0; i<m->nr_conns; i++) {
unsigned flags, state;
- s = (struct ip_vs_sync_conn *)p;
+ if (p + SIMPLE_CONN_SIZE > buffer+buflen) {
+ IP_VS_ERR_RL("bogus conn in sync message\n");
+ return;
+ }
+ s = (struct ip_vs_sync_conn *) p;
flags = ntohs(s->flags) | IP_VS_CONN_F_SYNC;
+ if (flags & IP_VS_CONN_F_SEQ_MASK) {
+ opt = (struct ip_vs_sync_conn_options *)&s[1];
+ p += FULL_CONN_SIZE;
+ if (p > buffer+buflen) {
+ IP_VS_ERR_RL("bogus conn options in sync message\n");
+ return;
+ }
+ } else {
+ opt = NULL;
+ p += SIMPLE_CONN_SIZE;
+ }
+
state = ntohs(s->state);
+ if (!(flags & IP_VS_CONN_F_TEMPLATE)) {
+ pp = ip_vs_proto_get(s->protocol);
+ if (!pp) {
+ IP_VS_ERR_RL("Unsupported protocol %u in sync msg\n",
+ s->protocol);
+ continue;
+ }
+ if (state >= pp->num_states) {
+ IP_VS_DBG(2, "Invalid %s state %u in sync msg\n",
+ pp->name, state);
+ continue;
+ }
+ } else {
+ /* protocol in templates is not used for state/timeout */
+ pp = NULL;
+ if (state > 0) {
+ IP_VS_DBG(2, "Invalid template state %u in sync msg\n",
+ state);
+ state = 0;
+ }
+ }
+
if (!(flags & IP_VS_CONN_F_TEMPLATE))
cp = ip_vs_conn_in_get(s->protocol,
s->caddr, s->cport,
@@ -345,7 +388,6 @@ static void ip_vs_process_message(const
IP_VS_ERR("ip_vs_conn_new failed\n");
return;
}
- cp->state = state;
} else if (!cp->dest) {
dest = ip_vs_try_bind_dest(cp);
if (!dest) {
@@ -371,23 +413,21 @@ static void ip_vs_process_message(const
}
}
- if (flags & IP_VS_CONN_F_SEQ_MASK) {
- opt = (struct ip_vs_sync_conn_options *)&s[1];
+ if (opt)
memcpy(&cp->in_seq, opt, sizeof(*opt));
- p += FULL_CONN_SIZE;
- } else
- p += SIMPLE_CONN_SIZE;
-
atomic_set(&cp->in_pkts, sysctl_ip_vs_sync_threshold[0]);
cp->state = state;
- pp = ip_vs_proto_get(s->protocol);
- cp->timeout = pp->timeout_table[cp->state];
+ /*
+ * We can not recover the right timeout for templates
+ * in all cases, we can not find the right fwmark
+ * virtual service. If needed, we can do it for
+ * non-fwmark persistent services.
+ */
+ if (!(flags & IP_VS_CONN_F_TEMPLATE) && pp->timeout_table)
+ cp->timeout = pp->timeout_table[state];
+ else
+ cp->timeout = (3*60*HZ);
ip_vs_conn_put(cp);
-
- if (p > buffer+buflen) {
- IP_VS_ERR("bogus message\n");
- return;
- }
}
}
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to
majordomo@vger.kernel.org
More majordomo info at
http://vger.kernel.org/majordomo-info.html
