hselasky updated this revision to Diff 5720.
hselasky added a comment.
Herald added a subscriber: imp.
Minor fix for computing correct ip6_plen in ip6_input().
Previously only the first 64K of the payload was consumed.
REPOSITORY
rS FreeBSD src repository
CHANGES SINCE LAST UPDATE
https://reviews.freebsd.org/D1761?vs=3630&id=5720
REVISION DETAIL
https://reviews.freebsd.org/D1761
AFFECTED FILES
sys/conf/options
sys/netinet/ip_input.c
sys/netinet/ip_output.c
sys/netinet/tcp_input.c
sys/netinet/tcp_lro.c
sys/netinet6/ip6_input.c
sys/sys/mbuf.h
EMAIL PREFERENCES
https://reviews.freebsd.org/settings/panel/emailpreferences/
To: hselasky, rrs, glebius, gnn, emaste, lstewart, rwatson, bz, imp, np, jfv,
adrian
Cc: imp, freebsd-net
diff --git a/sys/sys/mbuf.h b/sys/sys/mbuf.h
--- a/sys/sys/mbuf.h
+++ b/sys/sys/mbuf.h
@@ -306,6 +306,7 @@
#define M_HASHTYPE_RSS_UDP_IPV6 9 /* IPv6 UDP 4-tuple */
#define M_HASHTYPE_RSS_UDP_IPV6_EX 10 /* IPv6 UDP 4-tuple + ext hdrs */
+#define M_HASHTYPE_LRO_TCP 254 /* TCP large receive offload */
#define M_HASHTYPE_OPAQUE 255 /* ordering, not affinity */
#define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
diff --git a/sys/netinet6/ip6_input.c b/sys/netinet6/ip6_input.c
--- a/sys/netinet6/ip6_input.c
+++ b/sys/netinet6/ip6_input.c
@@ -694,7 +694,13 @@
* m may be modified in ip6_hopopts_input().
* If a JumboPayload option is included, plen will also be modified.
*/
- plen = (u_int32_t)ntohs(ip6->ip6_plen);
+ if (M_HASHTYPE_GET(m) == M_HASHTYPE_LRO_TCP) {
+ if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
+ plen = 0;
+ else
+ plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
+ } else
+ plen = (u_int32_t)ntohs(ip6->ip6_plen);
if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0)
return;
diff --git a/sys/netinet/tcp_lro.c b/sys/netinet/tcp_lro.c
--- a/sys/netinet/tcp_lro.c
+++ b/sys/netinet/tcp_lro.c
@@ -32,6 +32,7 @@
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
+#include "opt_lro.h"
#include "opt_inet.h"
#include "opt_inet6.h"
@@ -62,6 +63,14 @@
#define LRO_ENTRIES 8 /* # of LRO entries per RX queue. */
#endif
+#ifndef LRO_PAYLOAD_MAX
+#define LRO_PAYLOAD_MAX IP_MAXPACKET
+#endif
+
+#if (LRO_PAYLOAD_MAX < 65535)
+#error "LRO_PAYLOAD_MAX must be at least 65535 bytes"
+#endif
+
#define TCP_LRO_UPDATE_CSUM 1
#ifndef TCP_LRO_UPDATE_CSUM
#define TCP_LRO_INVALID_CSUM 0x0000
@@ -219,8 +228,20 @@
if (le->append_cnt > 0) {
struct tcphdr *th;
uint16_t p_len;
-
- p_len = htons(le->p_len);
+ /*
+ * The TCP/IP stack should use the "m_pkthdr.len"
+ * field instead of the IP-payload length field to
+ * compute the total TCP payload length when it
+ * recognizes the M_HASHTYPE_LRO_TCP hash type. This
+ * allows accumulation of more than 64Kbytes worth of
+ * payload data.
+ */
+ if (le->p_len > IP_MAXPACKET) {
+ M_HASHTYPE_SET(le->m_head, M_HASHTYPE_LRO_TCP);
+ p_len = htons(IP_MAXPACKET);
+ } else {
+ p_len = htons(le->p_len);
+ }
switch (le->eh_type) {
#ifdef INET6
case ETHERTYPE_IPV6:
@@ -501,7 +522,7 @@
}
/* Flush now if appending will result in overflow. */
- if (le->p_len > (65535 - tcp_data_len)) {
+ if (le->p_len > (LRO_PAYLOAD_MAX - tcp_data_len)) {
SLIST_REMOVE(&lc->lro_active, le, lro_entry, next);
tcp_lro_flush(lc, le);
break;
@@ -559,7 +580,7 @@
* If a possible next full length packet would cause an
* overflow, pro-actively flush now.
*/
- if (le->p_len > (65535 - lc->ifp->if_mtu)) {
+ if (le->p_len > (LRO_PAYLOAD_MAX - lc->ifp->if_mtu)) {
SLIST_REMOVE(&lc->lro_active, le, lro_entry, next);
tcp_lro_flush(lc, le);
} else
diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -644,7 +644,10 @@
ip6 = mtod(m, struct ip6_hdr *);
th = (struct tcphdr *)((caddr_t)ip6 + off0);
- tlen = sizeof(*ip6) + ntohs(ip6->ip6_plen) - off0;
+ if (M_HASHTYPE_GET(m) == M_HASHTYPE_LRO_TCP)
+ tlen = m->m_pkthdr.len - off0;
+ else
+ tlen = sizeof(*ip6) + ntohs(ip6->ip6_plen) - off0;
if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
th->th_sum = m->m_pkthdr.csum_data;
@@ -695,8 +698,10 @@
}
ip = mtod(m, struct ip *);
th = (struct tcphdr *)((caddr_t)ip + off0);
- tlen = ntohs(ip->ip_len) - off0;
-
+ if (M_HASHTYPE_GET(m) == M_HASHTYPE_LRO_TCP)
+ tlen = m->m_pkthdr.len - off0;
+ else
+ tlen = ntohs(ip->ip_len) - off0;
if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
th->th_sum = m->m_pkthdr.csum_data;
diff --git a/sys/netinet/ip_output.c b/sys/netinet/ip_output.c
--- a/sys/netinet/ip_output.c
+++ b/sys/netinet/ip_output.c
@@ -124,13 +124,14 @@
struct ifnet *ifp = NULL; /* keep compiler happy */
struct mbuf *m0;
int hlen = sizeof (struct ip);
+ int ip_len;
int mtu;
int error = 0;
struct sockaddr_in *dst;
const struct sockaddr_in *gw;
struct in_ifaddr *ia;
int isbroadcast;
- uint16_t ip_len, ip_off;
+ uint16_t ip_off;
struct route iproute;
struct rtentry *rte; /* cache for ro->ro_rt */
struct in_addr odst;
@@ -169,7 +170,11 @@
hlen = len; /* ip->ip_hl is updated above */
}
ip = mtod(m, struct ip *);
- ip_len = ntohs(ip->ip_len);
+
+ if (M_HASHTYPE_GET(m) == M_HASHTYPE_LRO_TCP)
+ ip_len = m->m_pkthdr.len;
+ else
+ ip_len = ntohs(ip->ip_len);
ip_off = ntohs(ip->ip_off);
if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
@@ -688,9 +693,13 @@
int firstlen;
struct mbuf **mnext;
int nfrags;
- uint16_t ip_len, ip_off;
+ int ip_len;
+ uint16_t ip_off;
- ip_len = ntohs(ip->ip_len);
+ if (M_HASHTYPE_GET(m0) == M_HASHTYPE_LRO_TCP)
+ ip_len = m0->m_pkthdr.len;
+ else
+ ip_len = ntohs(ip->ip_len);
ip_off = ntohs(ip->ip_off);
if (ip_off & IP_DF) { /* Fragmentation not allowed */
diff --git a/sys/netinet/ip_input.c b/sys/netinet/ip_input.c
--- a/sys/netinet/ip_input.c
+++ b/sys/netinet/ip_input.c
@@ -397,7 +397,8 @@
struct ifaddr *ifa;
struct ifnet *ifp;
int checkif, hlen = 0;
- uint16_t sum, ip_len;
+ uint32_t ip_len;
+ uint16_t sum;
int dchg = 0; /* dest changed after fw */
struct in_addr odst; /* original dst address */
@@ -474,7 +475,10 @@
return;
#endif
- ip_len = ntohs(ip->ip_len);
+ if (M_HASHTYPE_GET(m) == M_HASHTYPE_LRO_TCP)
+ ip_len = m->m_pkthdr.len;
+ else
+ ip_len = ntohs(ip->ip_len);
if (ip_len < hlen) {
IPSTAT_INC(ips_badlen);
goto bad;
@@ -900,6 +904,7 @@
struct in_addr dest;
struct route ro;
int error, type = 0, code = 0, mtu = 0;
+ int ip_len;
if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
IPSTAT_INC(ips_cantforward);
@@ -965,8 +970,14 @@
m_free(mcopy);
mcopy = NULL;
}
+
+ if (M_HASHTYPE_GET(m) == M_HASHTYPE_LRO_TCP)
+ ip_len = m->m_pkthdr.len;
+ else
+ ip_len = ntohs(ip->ip_len);
+
if (mcopy != NULL) {
- mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
+ mcopy->m_len = min(ip_len, M_TRAILINGSPACE(mcopy));
mcopy->m_pkthdr.len = mcopy->m_len;
m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
}
@@ -1094,7 +1105,7 @@
if (ia != NULL)
mtu = ia->ia_ifp->if_mtu;
else
- mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
+ mtu = ip_next_mtu(ip_len, 0);
}
IPSTAT_INC(ips_cantfrag);
break;
diff --git a/sys/conf/options b/sys/conf/options
--- a/sys/conf/options
+++ b/sys/conf/options
@@ -402,6 +402,8 @@
DUMMYNET opt_ipdn.h
INET opt_inet.h
INET6 opt_inet6.h
+LRO_ENTRIES opt_lro.h
+LRO_PAYLOAD_MAX opt_lro.h
IPDIVERT
IPFILTER opt_ipfilter.h
IPFILTER_DEFAULT_BLOCK opt_ipfilter.h
_______________________________________________
freebsd-net@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-net
To unsubscribe, send any mail to "freebsd-net-unsubscr...@freebsd.org"
