d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix a regression in the TCP handling of received segments.

Browse Source 
When receiving TCP segments the stack protects itself by limiting
the resources allocated for a TCP connections. This patch adds
an exception to these limitations for the TCP segement which is the next
expected in-sequence segment. Without this patch, TCP connections
may stall and finally fail in some cases of packet loss.

Reported by:		jhb@
Reviewed by:		jtl@, rrs@
MFC after:		3 days
Sponsored by:		Netflix, Inc.
Differential Revision:	https://reviews.freebsd.org/D18580
This commit is contained in:
Michael Tuexen 2018-12-20 16:05:30 +00:00
parent eacff37a61
commit 09423f72fd
1 changed files with 22 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
sys/netinet/tcp_reass.c
View File

@ -579,7 +579,8 @@ tcp_reass(struct tcpcb *tp, struct tcphdr *th, tcp_seq *seq_start,
*/
lenofoh = tcp_reass_overhead_of_chain(m, &mlast);
sb = &tp->t_inpcb->inp_socket->so_rcv;
if ((sb->sb_mbcnt + tp->t_segqmbuflen + lenofoh) > sb->sb_mbmax) {
if ((th->th_seq != tp->rcv_nxt || !TCPS_HAVEESTABLISHED(tp->t_state)) &&
(sb->sb_mbcnt + tp->t_segqmbuflen + lenofoh) > sb->sb_mbmax) {
/* No room */
TCPSTAT_INC(tcps_rcvreassfull);
#ifdef TCP_REASS_COUNTERS
@ -588,6 +589,11 @@ tcp_reass(struct tcpcb *tp, struct tcphdr *th, tcp_seq *seq_start,
#ifdef TCP_REASS_LOGGING
tcp_log_reassm(tp, NULL, NULL, th->th_seq, lenofoh, TCP_R_LOG_LIMIT_REACHED, 0);
#endif
if ((s = tcp_log_addrs(&tp->t_inpcb->inp_inc, th, NULL, NULL))) {
log(LOG_DEBUG, "%s; %s: mbuf count limit reached, "
"segment dropped\n", s, __func__);
free(s, M_TCPLOG);
}
m_freem(m);
*tlenp = 0;
#ifdef TCP_REASS_LOGGING
@ -936,6 +942,20 @@ tcp_reass(struct tcpcb *tp, struct tcphdr *th, tcp_seq *seq_start,
* is understood.
*/
new_entry:
if (th->th_seq == tp->rcv_nxt && TCPS_HAVEESTABLISHED(tp->t_state)) {
tp->rcv_nxt += *tlenp;
flags = th->th_flags & TH_FIN;
TCPSTAT_INC(tcps_rcvoopack);
TCPSTAT_ADD(tcps_rcvoobyte, *tlenp);
SOCKBUF_LOCK(&so->so_rcv);
if (so->so_rcv.sb_state & SBS_CANTRCVMORE) {
m_freem(m);
} else {
sbappendstream_locked(&so->so_rcv, m, 0);
}
sorwakeup_locked(so);
return (flags);
}
if (tcp_new_limits) {
if ((tp->t_segqlen > tcp_reass_queue_guard) &&
(*tlenp < MSIZE)) {
@ -960,9 +980,7 @@ tcp_reass(struct tcpcb *tp, struct tcphdr *th, tcp_seq *seq_start,
return (0);
}
} else {
if ((th->th_seq != tp->rcv_nxt || !TCPS_HAVEESTABLISHED(tp->t_state)) &&
tp->t_segqlen >= min((so->so_rcv.sb_hiwat / tp->t_maxseg) + 1,
if (tp->t_segqlen >= min((so->so_rcv.sb_hiwat / tp->t_maxseg) + 1,
tcp_reass_maxqueuelen)) {
TCPSTAT_INC(tcps_rcvreassfull);
*tlenp = 0;