Commit efe9f68a authored by Patrick McHardy's avatar Patrick McHardy Committed by David S. Miller

[NETFILTER]: nf_conntrack_sctp: rename "newconntrack" variable

The name is misleading, it holds the new connection state, so rename it
to "newstate". Also rename "oldsctpstate" to "oldstate" for consistency.
Signed-off-by: default avatarPatrick McHardy <kaber@trash.net>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent b37e933a
......@@ -231,9 +231,9 @@ static int do_basic_checks(struct nf_conn *ct,
return count == 0;
}
static int new_state(enum ip_conntrack_dir dir,
enum sctp_conntrack cur_state,
int chunk_type)
static int sctp_new_state(enum ip_conntrack_dir dir,
enum sctp_conntrack cur_state,
int chunk_type)
{
int i;
......@@ -299,7 +299,7 @@ static int sctp_packet(struct nf_conn *ct,
int pf,
unsigned int hooknum)
{
enum sctp_conntrack newconntrack, oldsctpstate;
enum sctp_conntrack new_state, old_state;
enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
sctp_sctphdr_t _sctph, *sh;
sctp_chunkhdr_t _sch, *sch;
......@@ -324,7 +324,7 @@ static int sctp_packet(struct nf_conn *ct,
goto out;
}
oldsctpstate = newconntrack = SCTP_CONNTRACK_MAX;
old_state = new_state = SCTP_CONNTRACK_MAX;
for_each_sctp_chunk (skb, sch, _sch, offset, dataoff, count) {
write_lock_bh(&sctp_lock);
......@@ -350,14 +350,14 @@ static int sctp_packet(struct nf_conn *ct,
goto out_unlock;
}
oldsctpstate = ct->proto.sctp.state;
newconntrack = new_state(dir, oldsctpstate, sch->type);
old_state = ct->proto.sctp.state;
new_state = sctp_new_state(dir, old_state, sch->type);
/* Invalid */
if (newconntrack == SCTP_CONNTRACK_MAX) {
if (new_state == SCTP_CONNTRACK_MAX) {
pr_debug("nf_conntrack_sctp: Invalid dir=%i ctype=%u "
"conntrack=%u\n",
dir, sch->type, oldsctpstate);
dir, sch->type, old_state);
goto out_unlock;
}
......@@ -375,17 +375,17 @@ static int sctp_packet(struct nf_conn *ct,
ct->proto.sctp.vtag[!dir] = ih->init_tag;
}
ct->proto.sctp.state = newconntrack;
if (oldsctpstate != newconntrack)
ct->proto.sctp.state = new_state;
if (old_state != new_state)
nf_conntrack_event_cache(IPCT_PROTOINFO, skb);
write_unlock_bh(&sctp_lock);
}
nf_ct_refresh_acct(ct, ctinfo, skb, *sctp_timeouts[newconntrack]);
nf_ct_refresh_acct(ct, ctinfo, skb, *sctp_timeouts[new_state]);
if (oldsctpstate == SCTP_CONNTRACK_COOKIE_ECHOED &&
if (old_state == SCTP_CONNTRACK_COOKIE_ECHOED &&
dir == IP_CT_DIR_REPLY &&
newconntrack == SCTP_CONNTRACK_ESTABLISHED) {
new_state == SCTP_CONNTRACK_ESTABLISHED) {
pr_debug("Setting assured bit\n");
set_bit(IPS_ASSURED_BIT, &ct->status);
nf_conntrack_event_cache(IPCT_STATUS, skb);
......@@ -403,7 +403,7 @@ static int sctp_packet(struct nf_conn *ct,
static int sctp_new(struct nf_conn *ct, const struct sk_buff *skb,
unsigned int dataoff)
{
enum sctp_conntrack newconntrack;
enum sctp_conntrack new_state;
sctp_sctphdr_t _sctph, *sh;
sctp_chunkhdr_t _sch, *sch;
u_int32_t offset, count;
......@@ -422,15 +422,15 @@ static int sctp_new(struct nf_conn *ct, const struct sk_buff *skb,
test_bit(SCTP_CID_COOKIE_ACK, map))
return 0;
newconntrack = SCTP_CONNTRACK_MAX;
new_state = SCTP_CONNTRACK_MAX;
for_each_sctp_chunk (skb, sch, _sch, offset, dataoff, count) {
/* Don't need lock here: this conntrack not in circulation yet */
newconntrack = new_state(IP_CT_DIR_ORIGINAL,
SCTP_CONNTRACK_NONE, sch->type);
new_state = sctp_new_state(IP_CT_DIR_ORIGINAL,
SCTP_CONNTRACK_NONE, sch->type);
/* Invalid: delete conntrack */
if (newconntrack == SCTP_CONNTRACK_NONE ||
newconntrack == SCTP_CONNTRACK_MAX) {
if (new_state == SCTP_CONNTRACK_NONE ||
new_state == SCTP_CONNTRACK_MAX) {
pr_debug("nf_conntrack_sctp: invalid new deleting.\n");
return 0;
}
......@@ -463,7 +463,7 @@ static int sctp_new(struct nf_conn *ct, const struct sk_buff *skb,
ct->proto.sctp.vtag[IP_CT_DIR_REPLY] = sh->vtag;
}
ct->proto.sctp.state = newconntrack;
ct->proto.sctp.state = new_state;
}
return 1;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment