Skip to contentSkip to conversation

ipv4: fix DO and PROBE pmtu mode regarding local fragmentation with UFO/CORK

PATCH Superseded Hannes Frederic Sowa 2 participants 4 messages Oct 25, 2013 – Oct 29, 2013

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ipv4: fix DO and PROBE pmtu mode regarding local fragmentation with UFO/CORK

UFO as well as UDP_CORK do not respect IP_PMTUDISC_DO and IP_PMTUDISC_PROBE well enough.

UFO enabled packet delivery just appends all frags to the cork and hands it over to the network card. So we just deliver non-DF udp fragments (DF-flag may get overwritten by hardware or virtual UFO enabled interface).

UDP_CORK does enqueue the data until the cork is disengaged. At this point it sets the correct IP_DF and local_df flags and hands it over to ip_fragment which in this case will generate an icmp error which gets appended to the error socket queue. This is not reflected in the syscall error (of course, if UFO is enabled this also won't happen).

Improve this by checking the pmtudisc flags before appending data to the socket and if we still can fit all data in one packet when IP_PMTUDISC_DO or IP_PMTUDISC_PROBE is set, only then proceed.

Maybe, we can relax some other checks around ip_fragment. This needs more research.

---
 net/ipv4/ip_output.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
net/ipv4/ip_output.c +8 -4
index 8fbac7d..27dc1b3 100644
@@ -810,7 +810,7 @@ static int __ip_append_data(struct sock *sk,
 	int copy;
 	int err;
 	int offset = 0;
-	unsigned int maxfraglen, fragheaderlen;
+	unsigned int maxfraglen, fragheaderlen, maxmsgsize;
 	int csummode = CHECKSUM_NONE;
 	struct rtable *rt = (struct rtable *)cork->dst;
 
@@ -823,8 +823,10 @@ static int __ip_append_data(struct sock *sk,
 
 	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
 	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;
+	maxmsgsize = (inet->pmtudisc >= IP_PMTUDISC_DO) ?
+		     maxfraglen : 0xFFFF;
 
-	if (cork->length + length > 0xFFFF - fragheaderlen) {
+	if (cork->length + length > maxmsgsize - fragheaderlen) {
 		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
 			       mtu-exthdrlen);
 		return -EMSGSIZE;
@@ -1122,7 +1124,7 @@ ssize_t	ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
 	int mtu;
 	int len;
 	int err;
-	unsigned int maxfraglen, fragheaderlen, fraggap;
+	unsigned int maxfraglen, fragheaderlen, fraggap, maxmsgsize;
 
 	if (inet->hdrincl)
 		return -EPERM;
@@ -1146,8 +1148,10 @@ ssize_t	ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
 
 	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
 	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;
+	maxmsgsize = (inet->pmtudisc >= IP_PMTUDISC_DO) ?
+		     maxfraglen : 0xFFFF;
 
-	if (cork->length + size > 0xFFFF - fragheaderlen) {
+	if (cork->length + size > maxmsgsize - fragheaderlen) {
 		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, mtu);
 		return -EMSGSIZE;
 	}
Signature · 1 line
1.8.3.1
#1 lore raw

net/ipv4/ip_output.c static int __ip_append_data(struct sock *sk, · line 827

Quoted 4 lines · Hannes Frederic Sowa
net/ipv4/ip_output.c +2 -0
 	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
 	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;
+	maxmsgsize = (inet->pmtudisc >= IP_PMTUDISC_DO) ?
+		     maxfraglen : 0xFFFF;

Oops, sorry. I just realized that the check is too strict. The 64 bit boundary restricton is solely used if we are actually generating fragments. So I should have used mtu directly instead of maxfraglen.

This should only matter in case of some unusual MTUs but I will send a corrected patch as soon as I am at home.

Greetings,

Hannes

#2 lore raw

PATCH v2 1/1 ipv4: fix DO and PROBE pmtu mode regarding local fragmentation with UFO/CORK

In reply to Hannes Frederic Sowa

UFO as well as UDP_CORK do not respect IP_PMTUDISC_DO and IP_PMTUDISC_PROBE well enough.

UFO enabled packet delivery just appends all frags to the cork and hands it over to the network card. So we just deliver non-DF udp fragments (DF-flag may get overwritten by hardware or virtual UFO enabled interface).

UDP_CORK does enqueue the data until the cork is disengaged. At this point it sets the correct IP_DF and local_df flags and hands it over to ip_fragment which in this case will generate an icmp error which gets appended to the error socket queue. This is not reflected in the syscall error (of course, if UFO is enabled this also won't happen).

Improve this by checking the pmtudisc flags before appending data to the socket and if we still can fit all data in one packet when IP_PMTUDISC_DO or IP_PMTUDISC_PROBE is set, only then proceed.

We use (mtu-fragheaderlen) to check for the maximum length because we ensure not to generate a fragment and non-fragmented data does not need to have its length aligned on 64 bit boundaries. Also the passed in ip_options are already aligned correctly.

Maybe, we can relax some other checks around ip_fragment. This needs more research.

---
v2:
Switch from maxfraglen to mtu for length check as outlined in the commit message.

 net/ipv4/ip_output.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
net/ipv4/ip_output.c +8 -4
index 8fbac7d..51be64e 100644
@@ -810,7 +810,7 @@ static int __ip_append_data(struct sock *sk,
 	int copy;
 	int err;
 	int offset = 0;
-	unsigned int maxfraglen, fragheaderlen;
+	unsigned int maxfraglen, fragheaderlen, maxnonfragsize;
 	int csummode = CHECKSUM_NONE;
 	struct rtable *rt = (struct rtable *)cork->dst;
 
@@ -823,8 +823,10 @@ static int __ip_append_data(struct sock *sk,
 
 	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
 	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;
+	maxnonfragsize = (inet->pmtudisc >= IP_PMTUDISC_DO) ?
+			 mtu : 0xFFFF;
 
-	if (cork->length + length > 0xFFFF - fragheaderlen) {
+	if (cork->length + length > maxnonfragsize - fragheaderlen) {
 		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
 			       mtu-exthdrlen);
 		return -EMSGSIZE;
@@ -1122,7 +1124,7 @@ ssize_t	ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
 	int mtu;
 	int len;
 	int err;
-	unsigned int maxfraglen, fragheaderlen, fraggap;
+	unsigned int maxfraglen, fragheaderlen, fraggap, maxnonfragsize;
 
 	if (inet->hdrincl)
 		return -EPERM;
@@ -1146,8 +1148,10 @@ ssize_t	ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
 
 	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
 	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;
+	maxnonfragsize = (inet->pmtudisc >= IP_PMTUDISC_DO) ?
+			 mtu : 0xFFFF;
 
-	if (cork->length + size > 0xFFFF - fragheaderlen) {
+	if (cork->length + size > maxnonfragsize - fragheaderlen) {
 		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, mtu);
 		return -EMSGSIZE;
 	}
Signature · 1 line
1.8.3.1
maintainer
#3 lore raw

From: Hannes Frederic Sowa <[email protected]>
Date: Sun, 27 Oct 2013 17:29:11 +0100

Quoted 23 lines

UFO as well as UDP_CORK do not respect IP_PMTUDISC_DO and IP_PMTUDISC_PROBE well enough.

UFO enabled packet delivery just appends all frags to the cork and hands it over to the network card. So we just deliver non-DF udp fragments (DF-flag may get overwritten by hardware or virtual UFO enabled interface).

UDP_CORK does enqueue the data until the cork is disengaged. At this point it sets the correct IP_DF and local_df flags and hands it over to ip_fragment which in this case will generate an icmp error which gets appended to the error socket queue. This is not reflected in the syscall error (of course, if UFO is enabled this also won't happen).

Improve this by checking the pmtudisc flags before appending data to the socket and if we still can fit all data in one packet when IP_PMTUDISC_DO or IP_PMTUDISC_PROBE is set, only then proceed.

We use (mtu-fragheaderlen) to check for the maximum length because we ensure not to generate a fragment and non-fragmented data does not need to have its length aligned on 64 bit boundaries. Also the passed in ip_options are already aligned correctly.

Quoted 8 lines

Maybe, we can relax some other checks around ip_fragment. This needs more research.

---
v2:
Switch from maxfraglen to mtu for length check as outlined in the commit message.

Looks good, applied.