Skip to contentSkip to conversation

net: ipv6: Add a sysctl to make optimistic addresses useful candidates

RFC PATCH Superseded Erik Kline 3 participants 8 messages Oct 21, 2014 – Oct 22, 2014

Subsystems (4)
Networking Network drivers Documentation IPv4/IPv6

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Erik Kline

author
#0 lore raw

RFC PATCH v2 1/1 net: ipv6: Add a sysctl to make optimistic addresses useful candidates

Add a sysctl that causes an interface's optimistic addresses to be considered equivalent to other non-deprecated addresses for source address selection purposes. Preferred addresses will still take precedence over optimistic addresses, subject to other ranking in the source address selection algorithm.

This is useful where different interfaces are connected to different networks from different ISPs (e.g., a cell network and a home wifi network).

The current behaviour complies with RFC 3484/6724, and it makes sense if the host has only one interface, or has multiple interfaces on the same network (same or cooperating administrative domain(s), but not in the multiple distinct networks case.

For example, if a mobile device has an IPv6 address on an LTE network and then connects to IPv6-enabled wifi, while the wifi IPv6 address is undergoing DAD, IPv6 connections will try use the wifi default route with the LTE IPv6 address, and will get stuck until they time out.

Also, because optimistic addresses can actually be used, issue an RTM_NEWADDR as soon as DAD starts. If DAD fails an separate RTM_DELADDR will be sent.

Also: add an entry in ip-sysctl.txt for optimistic_dad.

Signed-off-by Erik Kline <[email protected]>
---
 Documentation/networking/ip-sysctl.txt | 13 ++++++++++++
 include/linux/ipv6.h                   |  1 +
 include/uapi/linux/ipv6.h              |  1 +
 net/ipv6/addrconf.c                    | 36 ++++++++++++++++++++++++++++++++--
 4 files changed, 49 insertions(+), 2 deletions(-)
Documentation/networking/ip-sysctl.txt +13 -0
index 0307e28..e03cf49 100644
@@ -1452,6 +1452,19 @@ suppress_frag_ndisc - INTEGER
 	1 - (default) discard fragmented neighbor discovery packets
 	0 - allow fragmented neighbor discovery packets
 
+optimistic_dad - BOOLEAN
+	Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
+		0: disabled (default)
+		1: enabled
+
+use_optimistic - BOOLEAN
+	If enabled, do not classify optimistic addresses as deprecated during
+	source address selection.  Preferred addresses will still be chosen
+	before optimistic addresses, subject to other ranking in the source
+	address selection algorithm.
+		0: disabled (default)
+		1: enabled
+
 icmp/*:
 ratelimit - INTEGER
 	Limit the maximal rates for sending ICMPv6 packets.
include/linux/ipv6.h +1 -0
index ff56053..7121a2e 100644
@@ -42,6 +42,7 @@ struct ipv6_devconf {
 	__s32		accept_ra_from_local;
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
 	__s32		optimistic_dad;
+	__s32		use_optimistic;
 #endif
 #ifdef CONFIG_IPV6_MROUTE
 	__s32		mc_forwarding;
include/uapi/linux/ipv6.h +1 -0
index efa2666..e863d08 100644
@@ -164,6 +164,7 @@ enum {
 	DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL,
 	DEVCONF_SUPPRESS_FRAG_NDISC,
 	DEVCONF_ACCEPT_RA_FROM_LOCAL,
+	DEVCONF_USE_OPTIMISTIC,
 	DEVCONF_MAX
 };
net/ipv6/addrconf.c +34 -2
index 725c763..c2fddb7 100644
@@ -1169,6 +1169,9 @@ enum {
 	IPV6_SADDR_RULE_LABEL,
 	IPV6_SADDR_RULE_PRIVACY,
 	IPV6_SADDR_RULE_ORCHID,
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
+#endif
 	IPV6_SADDR_RULE_PREFIX,
 	IPV6_SADDR_RULE_MAX
 };

@@ -1257,10 +1260,17 @@ static int ipv6_get_saddr_eval(struct net *net,
 		score->scopedist = ret;
 		break;
 	case IPV6_SADDR_RULE_PREFERRED:
+	    {
 		/* Rule 3: Avoid deprecated and optimistic addresses */
+		u8 avoid = IFA_F_DEPRECATED;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+		if (!score->ifa->idev->cnf.use_optimistic)
+			avoid |= IFA_F_OPTIMISTIC;
+#endif
 		ret = ipv6_saddr_preferred(score->addr_type) ||
-		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
+		      !(score->ifa->flags & avoid);
 		break;
+	    }
 #ifdef CONFIG_IPV6_MIP6
 	case IPV6_SADDR_RULE_HOA:
 	    {

@@ -1299,6 +1309,14 @@ static int ipv6_get_saddr_eval(struct net *net,
 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
 			ipv6_addr_orchid(dst->addr));
 		break;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
+		/* Optimistic addresses still have lower precedence than other
+		 * preferred addresses.
+		 */
+		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
+		break;
+#endif
 	case IPV6_SADDR_RULE_PREFIX:
 		/* Rule 8: Use longest matching prefix */
 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);

@@ -3222,8 +3240,13 @@ static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
 	 * Optimistic nodes can start receiving
 	 * Frames right away
 	 */
-	if (ifp->flags & IFA_F_OPTIMISTIC)
+	if (ifp->flags & IFA_F_OPTIMISTIC) {
 		ip6_ins_rt(ifp->rt);
+		/* Because optimistic nodes can receive frames, notify
+		 * listeners. If DAD fails, RTM_DELADDR is sent.
+		 */
+		ipv6_ifa_notify(RTM_NEWADDR, ifp);
+	}
 
 	addrconf_dad_kick(ifp);
 out:

@@ -4330,6 +4353,7 @@ static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
+	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
 #endif
 #ifdef CONFIG_IPV6_MROUTE
 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;

@@ -5155,6 +5179,14 @@ static struct addrconf_sysctl_table
 			.proc_handler   = proc_dointvec,
 
 		},
+		{
+			.procname       = "use_optimistic",
+			.data           = &ipv6_devconf.use_optimistic,
+			.maxlen         = sizeof(int),
+			.mode           = 0644,
+			.proc_handler   = proc_dointvec,
+
+		},
 #endif
 #ifdef CONFIG_IPV6_MROUTE
 		{
Signature · 1 line
2.1.0.rc2.206.gedb03e5
#1 lore raw

Hi,

Quoted 119 lines · Erik Kline

Add a sysctl that causes an interface's optimistic addresses to be considered equivalent to other non-deprecated addresses for source address selection purposes. Preferred addresses will still take precedence over optimistic addresses, subject to other ranking in the source address selection algorithm.

This is useful where different interfaces are connected to different networks from different ISPs (e.g., a cell network and a home wifi network).

The current behaviour complies with RFC 3484/6724, and it makes sense if the host has only one interface, or has multiple interfaces on the same network (same or cooperating administrative domain(s), but not in the multiple distinct networks case.

For example, if a mobile device has an IPv6 address on an LTE network and then connects to IPv6-enabled wifi, while the wifi IPv6 address is undergoing DAD, IPv6 connections will try use the wifi default route with the LTE IPv6 address, and will get stuck until they time out.

Also, because optimistic addresses can actually be used, issue an RTM_NEWADDR as soon as DAD starts. If DAD fails an separate RTM_DELADDR will be sent.

Also: add an entry in ip-sysctl.txt for optimistic_dad.

Signed-off-by Erik Kline <[email protected]>
---
 Documentation/networking/ip-sysctl.txt | 13 ++++++++++++
 include/linux/ipv6.h                   |  1 +
 include/uapi/linux/ipv6.h              |  1 +
 net/ipv6/addrconf.c                    | 36 ++++++++++++++++++++++++++++++++--
 4 files changed, 49 insertions(+), 2 deletions(-)
Documentation/networking/ip-sysctl.txt +13 -0
index 0307e28..e03cf49 100644
@@ -1452,6 +1452,19 @@ suppress_frag_ndisc - INTEGER
 	1 - (default) discard fragmented neighbor discovery packets
 	0 - allow fragmented neighbor discovery packets
 
+optimistic_dad - BOOLEAN
+	Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
+		0: disabled (default)
+		1: enabled
+
+use_optimistic - BOOLEAN
+	If enabled, do not classify optimistic addresses as deprecated during
+	source address selection.  Preferred addresses will still be chosen
+	before optimistic addresses, subject to other ranking in the source
+	address selection algorithm.
+		0: disabled (default)
+		1: enabled
+
 icmp/*:
 ratelimit - INTEGER
 	Limit the maximal rates for sending ICMPv6 packets.
include/linux/ipv6.h +1 -0
index ff56053..7121a2e 100644
@@ -42,6 +42,7 @@ struct ipv6_devconf {
 	__s32		accept_ra_from_local;
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
 	__s32		optimistic_dad;
+	__s32		use_optimistic;
 #endif
 #ifdef CONFIG_IPV6_MROUTE
 	__s32		mc_forwarding;
include/uapi/linux/ipv6.h +1 -0
index efa2666..e863d08 100644
@@ -164,6 +164,7 @@ enum {
 	DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL,
 	DEVCONF_SUPPRESS_FRAG_NDISC,
 	DEVCONF_ACCEPT_RA_FROM_LOCAL,
+	DEVCONF_USE_OPTIMISTIC,
 	DEVCONF_MAX
 };
net/ipv6/addrconf.c +11 -1
index 725c763..c2fddb7 100644
@@ -1169,6 +1169,9 @@ enum {
 	IPV6_SADDR_RULE_LABEL,
 	IPV6_SADDR_RULE_PRIVACY,
 	IPV6_SADDR_RULE_ORCHID,
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
+#endif
 	IPV6_SADDR_RULE_PREFIX,
 	IPV6_SADDR_RULE_MAX
 };

@@ -1257,10 +1260,17 @@ static int ipv6_get_saddr_eval(struct net *net,
 		score->scopedist = ret;
 		break;
 	case IPV6_SADDR_RULE_PREFERRED:
+	    {
 		/* Rule 3: Avoid deprecated and optimistic addresses */
+		u8 avoid = IFA_F_DEPRECATED;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+		if (!score->ifa->idev->cnf.use_optimistic)
+			avoid |= IFA_F_OPTIMISTIC;
+#endif
 		ret = ipv6_saddr_preferred(score->addr_type) ||
-		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
+		      !(score->ifa->flags & avoid);
 		break;
+	    }
 #ifdef CONFIG_IPV6_MIP6
 	case IPV6_SADDR_RULE_HOA:
 	    {

@@ -1299,6 +1309,14 @@ static int ipv6_get_saddr_eval(struct net *net,
 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^

net/ipv6/addrconf.c static int ipv6_get_saddr_eval(struct net *net, · line 1319

Quoted 10 lines
net/ipv6/addrconf.c +8 -0
 			ipv6_addr_orchid(dst->addr));
 		break;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
+		/* Optimistic addresses still have lower precedence than other
+		 * preferred addresses.
+		 */
+		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
+		break;
+#endif

I wonder a bit why this rule is not directly ordered after IPV6_SADDR_RULE_PREFERRED? This would e.g. matter for privacy addresses.

Quoted 5 lines
net/ipv6/addrconf.c +0 -0
 	case IPV6_SADDR_RULE_PREFIX:
 		/* Rule 8: Use longest matching prefix */
 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);

@@ -3222,8 +3240,13 @@ static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
 	 * Optimistic nodes can start receiving

net/ipv6/addrconf.c static void addrconf_dad_begin(struct inet6_ifaddr *ifp) · line 3249

Quoted 10 lines
net/ipv6/addrconf.c +6 -1
 	 * Frames right away
 	 */
-	if (ifp->flags & IFA_F_OPTIMISTIC)
+	if (ifp->flags & IFA_F_OPTIMISTIC) {
 		ip6_ins_rt(ifp->rt);
+		/* Because optimistic nodes can receive frames, notify
+		 * listeners. If DAD fails, RTM_DELADDR is sent.
+		 */
+		ipv6_ifa_notify(RTM_NEWADDR, ifp);
+	}

I wonder if we can now delete the ipv6_ifa_notify(RTM_NEWADDR, ifp) in addrconf_dad_completed.

Quoted 15 lines
net/ipv6/addrconf.c +2 -0
 	addrconf_dad_kick(ifp);
 out:

@@ -4330,6 +4353,7 @@ static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
+	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
 #endif
 #ifdef CONFIG_IPV6_MROUTE
 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;

@@ -5155,6 +5179,14 @@ static struct addrconf_sysctl_table
 			.proc_handler   = proc_dointvec,
 
 		},
+		{

net/ipv6/addrconf.c static struct addrconf_sysctl_table · line 5192

Quoted 10 lines
net/ipv6/addrconf.c +7 -0
+			.procname       = "use_optimistic",
+			.data           = &ipv6_devconf.use_optimistic,
+			.maxlen         = sizeof(int),
+			.mode           = 0644,
+			.proc_handler   = proc_dointvec,
+
+		},
 #endif
 #ifdef CONFIG_IPV6_MROUTE
 		{

Otherwise looks good.

Thanks,
Hannes

author
#2 lore raw

Hello.

Quoted 127 lines · Hannes Frederic Sowa

Hi,

Quoted 124 lines · Erik Kline

Add a sysctl that causes an interface's optimistic addresses to be considered equivalent to other non-deprecated addresses for source address selection purposes. Preferred addresses will still take precedence over optimistic addresses, subject to other ranking in the source address selection algorithm.

This is useful where different interfaces are connected to different networks from different ISPs (e.g., a cell network and a home wifi network).

The current behaviour complies with RFC 3484/6724, and it makes sense if the host has only one interface, or has multiple interfaces on the same network (same or cooperating administrative domain(s), but not in the multiple distinct networks case.

For example, if a mobile device has an IPv6 address on an LTE network and then connects to IPv6-enabled wifi, while the wifi IPv6 address is undergoing DAD, IPv6 connections will try use the wifi default route with the LTE IPv6 address, and will get stuck until they time out.

Also, because optimistic addresses can actually be used, issue an RTM_NEWADDR as soon as DAD starts. If DAD fails an separate RTM_DELADDR will be sent.

Also: add an entry in ip-sysctl.txt for optimistic_dad.

Signed-off-by Erik Kline <[email protected]>
---
 Documentation/networking/ip-sysctl.txt | 13 ++++++++++++
 include/linux/ipv6.h                   |  1 +
 include/uapi/linux/ipv6.h              |  1 +
 net/ipv6/addrconf.c                    | 36 ++++++++++++++++++++++++++++++++--
 4 files changed, 49 insertions(+), 2 deletions(-)
Documentation/networking/ip-sysctl.txt +13 -0
index 0307e28..e03cf49 100644
@@ -1452,6 +1452,19 @@ suppress_frag_ndisc - INTEGER
      1 - (default) discard fragmented neighbor discovery packets
      0 - allow fragmented neighbor discovery packets
 
+optimistic_dad - BOOLEAN
+     Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
+             0: disabled (default)
+             1: enabled
+
+use_optimistic - BOOLEAN
+     If enabled, do not classify optimistic addresses as deprecated during
+     source address selection.  Preferred addresses will still be chosen
+     before optimistic addresses, subject to other ranking in the source
+     address selection algorithm.
+             0: disabled (default)
+             1: enabled
+
 icmp/*:
 ratelimit - INTEGER
      Limit the maximal rates for sending ICMPv6 packets.
include/linux/ipv6.h +1 -0
index ff56053..7121a2e 100644
@@ -42,6 +42,7 @@ struct ipv6_devconf {
      __s32           accept_ra_from_local;
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
      __s32           optimistic_dad;
+     __s32           use_optimistic;
 #endif
 #ifdef CONFIG_IPV6_MROUTE
      __s32           mc_forwarding;
include/uapi/linux/ipv6.h +1 -0
index efa2666..e863d08 100644
@@ -164,6 +164,7 @@ enum {
      DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL,
      DEVCONF_SUPPRESS_FRAG_NDISC,
      DEVCONF_ACCEPT_RA_FROM_LOCAL,
+     DEVCONF_USE_OPTIMISTIC,
      DEVCONF_MAX
 };
net/ipv6/addrconf.c +14 -1
index 725c763..c2fddb7 100644
@@ -1169,6 +1169,9 @@ enum {
      IPV6_SADDR_RULE_LABEL,
      IPV6_SADDR_RULE_PRIVACY,
      IPV6_SADDR_RULE_ORCHID,
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+     IPV6_SADDR_RULE_NOT_OPTIMISTIC,
+#endif
      IPV6_SADDR_RULE_PREFIX,
      IPV6_SADDR_RULE_MAX
 };

@@ -1257,10 +1260,17 @@ static int ipv6_get_saddr_eval(struct net *net,
              score->scopedist = ret;
              break;
      case IPV6_SADDR_RULE_PREFERRED:
+         {
              /* Rule 3: Avoid deprecated and optimistic addresses */
+             u8 avoid = IFA_F_DEPRECATED;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+             if (!score->ifa->idev->cnf.use_optimistic)
+                     avoid |= IFA_F_OPTIMISTIC;
+#endif
              ret = ipv6_saddr_preferred(score->addr_type) ||
-                   !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
+                   !(score->ifa->flags & avoid);
              break;
+         }
 #ifdef CONFIG_IPV6_MIP6
      case IPV6_SADDR_RULE_HOA:
          {

@@ -1299,6 +1309,14 @@ static int ipv6_get_saddr_eval(struct net *net,
              ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
                      ipv6_addr_orchid(dst->addr));
              break;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+     case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
+             /* Optimistic addresses still have lower precedence than other
Quoted 8 lines

net/ipv6/addrconf.c static int ipv6_get_saddr_eval(struct net *net, · line 1319

Quoted 5 lines
net/ipv6/addrconf.c +5 -0
+              * preferred addresses.
+              */
+             ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
+             break;
+#endif

I wonder a bit why this rule is not directly ordered after IPV6_SADDR_RULE_PREFERRED? This would e.g. matter for privacy addresses.

Privacy addresses ("tempaddrs") that win in earlier checks are preferred before optimistic in this code (i.e. a tempaddr on the same outgoing interface is preferred before an optimistic address).

Similarly, a non-tentative non-privacy address (same outgoing interface, same label, ...) will match before an optimistic address, but only until DAD completes and the address is no longer optimistic. I think this is in keeping with the spirit of the RFC 3484/6724 rules.

After there's an RFC 7217 implementation, EUI-64-based SLAAC could be disabled by folks.

Quoted 10 lines
Quoted 10 lines
net/ipv6/addrconf.c +1 -1
      case IPV6_SADDR_RULE_PREFIX:
              /* Rule 8: Use longest matching prefix */
              ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);

@@ -3222,8 +3240,13 @@ static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
       * Optimistic nodes can start receiving
       * Frames right away
       */
-     if (ifp->flags & IFA_F_OPTIMISTIC)
+     if (ifp->flags & IFA_F_OPTIMISTIC) {
              ip6_ins_rt(ifp->rt);
Quoted 8 lines

net/ipv6/addrconf.c static void addrconf_dad_begin(struct inet6_ifaddr *ifp) · line 3249

Quoted 5 lines
net/ipv6/addrconf.c +5 -0
+             /* Because optimistic nodes can receive frames, notify
+              * listeners. If DAD fails, RTM_DELADDR is sent.
+              */
+             ipv6_ifa_notify(RTM_NEWADDR, ifp);
+     }

I wonder if we can now delete the ipv6_ifa_notify(RTM_NEWADDR, ifp) in addrconf_dad_completed.

I don't know what everyone's general preference would be, but mine would be to err on the side of notifying on state changes. It seems harmless to me to keep it in, and something in userspace might want to know if/when DAD completes.

Quoted 28 lines
Quoted 15 lines
net/ipv6/addrconf.c +2 -0
      addrconf_dad_kick(ifp);
 out:

@@ -4330,6 +4353,7 @@ static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
      array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
      array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
+     array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
 #endif
 #ifdef CONFIG_IPV6_MROUTE
      array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;

@@ -5155,6 +5179,14 @@ static struct addrconf_sysctl_table
                      .proc_handler   = proc_dointvec,
 
              },
+             {

net/ipv6/addrconf.c static struct addrconf_sysctl_table · line 5192

Quoted 10 lines
net/ipv6/addrconf.c +7 -0
+                     .procname       = "use_optimistic",
+                     .data           = &ipv6_devconf.use_optimistic,
+                     .maxlen         = sizeof(int),
+                     .mode           = 0644,
+                     .proc_handler   = proc_dointvec,
+
+             },
 #endif
 #ifdef CONFIG_IPV6_MROUTE
              {

Otherwise looks good.

Quoted 4 lines

Thanks,
Hannes

Thank /you/.

#3 lore raw
Quoted 140 lines · Erik Kline

Hello.

Quoted 127 lines · Hannes Frederic Sowa

Hi,

Quoted 124 lines · Erik Kline

Add a sysctl that causes an interface's optimistic addresses to be considered equivalent to other non-deprecated addresses for source address selection purposes. Preferred addresses will still take precedence over optimistic addresses, subject to other ranking in the source address selection algorithm.

This is useful where different interfaces are connected to different networks from different ISPs (e.g., a cell network and a home wifi network).

The current behaviour complies with RFC 3484/6724, and it makes sense if the host has only one interface, or has multiple interfaces on the same network (same or cooperating administrative domain(s), but not in the multiple distinct networks case.

For example, if a mobile device has an IPv6 address on an LTE network and then connects to IPv6-enabled wifi, while the wifi IPv6 address is undergoing DAD, IPv6 connections will try use the wifi default route with the LTE IPv6 address, and will get stuck until they time out.

Also, because optimistic addresses can actually be used, issue an RTM_NEWADDR as soon as DAD starts. If DAD fails an separate RTM_DELADDR will be sent.

Also: add an entry in ip-sysctl.txt for optimistic_dad.

Signed-off-by Erik Kline <[email protected]>
---
 Documentation/networking/ip-sysctl.txt | 13 ++++++++++++
 include/linux/ipv6.h                   |  1 +
 include/uapi/linux/ipv6.h              |  1 +
 net/ipv6/addrconf.c                    | 36 ++++++++++++++++++++++++++++++++--
 4 files changed, 49 insertions(+), 2 deletions(-)
Documentation/networking/ip-sysctl.txt +13 -0
index 0307e28..e03cf49 100644
@@ -1452,6 +1452,19 @@ suppress_frag_ndisc - INTEGER
      1 - (default) discard fragmented neighbor discovery packets
      0 - allow fragmented neighbor discovery packets
 
+optimistic_dad - BOOLEAN
+     Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
+             0: disabled (default)
+             1: enabled
+
+use_optimistic - BOOLEAN
+     If enabled, do not classify optimistic addresses as deprecated during
+     source address selection.  Preferred addresses will still be chosen
+     before optimistic addresses, subject to other ranking in the source
+     address selection algorithm.
+             0: disabled (default)
+             1: enabled
+
 icmp/*:
 ratelimit - INTEGER
      Limit the maximal rates for sending ICMPv6 packets.
include/linux/ipv6.h +1 -0
index ff56053..7121a2e 100644
@@ -42,6 +42,7 @@ struct ipv6_devconf {
      __s32           accept_ra_from_local;
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
      __s32           optimistic_dad;
+     __s32           use_optimistic;
 #endif
 #ifdef CONFIG_IPV6_MROUTE
      __s32           mc_forwarding;
include/uapi/linux/ipv6.h +1 -0
index efa2666..e863d08 100644
@@ -164,6 +164,7 @@ enum {
      DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL,
      DEVCONF_SUPPRESS_FRAG_NDISC,
      DEVCONF_ACCEPT_RA_FROM_LOCAL,
+     DEVCONF_USE_OPTIMISTIC,
      DEVCONF_MAX
 };
net/ipv6/addrconf.c +14 -1
index 725c763..c2fddb7 100644
@@ -1169,6 +1169,9 @@ enum {
      IPV6_SADDR_RULE_LABEL,
      IPV6_SADDR_RULE_PRIVACY,
      IPV6_SADDR_RULE_ORCHID,
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+     IPV6_SADDR_RULE_NOT_OPTIMISTIC,
+#endif
      IPV6_SADDR_RULE_PREFIX,
      IPV6_SADDR_RULE_MAX
 };

@@ -1257,10 +1260,17 @@ static int ipv6_get_saddr_eval(struct net *net,
              score->scopedist = ret;
              break;
      case IPV6_SADDR_RULE_PREFERRED:
+         {
              /* Rule 3: Avoid deprecated and optimistic addresses */
+             u8 avoid = IFA_F_DEPRECATED;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+             if (!score->ifa->idev->cnf.use_optimistic)
+                     avoid |= IFA_F_OPTIMISTIC;
+#endif
              ret = ipv6_saddr_preferred(score->addr_type) ||
-                   !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
+                   !(score->ifa->flags & avoid);
              break;
+         }
 #ifdef CONFIG_IPV6_MIP6
      case IPV6_SADDR_RULE_HOA:
          {

@@ -1299,6 +1309,14 @@ static int ipv6_get_saddr_eval(struct net *net,
              ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
                      ipv6_addr_orchid(dst->addr));
              break;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+     case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
+             /* Optimistic addresses still have lower precedence than other
Quoted 8 lines

net/ipv6/addrconf.c static int ipv6_get_saddr_eval(struct net *net, · line 1319

Quoted 5 lines
net/ipv6/addrconf.c +5 -0
+              * preferred addresses.
+              */
+             ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
+             break;
+#endif

I wonder a bit why this rule is not directly ordered after IPV6_SADDR_RULE_PREFERRED? This would e.g. matter for privacy addresses.

Quoted 8 lines

Privacy addresses ("tempaddrs") that win in earlier checks are preferred before optimistic in this code (i.e. a tempaddr on the same outgoing interface is preferred before an optimistic address).

Similarly, a non-tentative non-privacy address (same outgoing interface, same label, ...) will match before an optimistic address, but only until DAD completes and the address is no longer optimistic. I think this is in keeping with the spirit of the RFC 3484/6724 rules.

Oh yes, I see. I had the evaluation order messed up. ;)

So the question I should be asking would be. AFAIR optimistic addresses should be handled like deprecated ones, so I am a bit concerned adding a non-conditional rule before the RULE_PREFIX check.

Shouldn't we only break the tie *after* longest prefix match then? If you do that before longest prefix match I would prefer ret being masked by use_optimisitic flag.

Quoted 2 lines

After there's an RFC 7217 implementation, EUI-64-based SLAAC could be disabled by folks.

Ack.

Quoted 20 lines
Quoted 11 lines
Quoted 11 lines
net/ipv6/addrconf.c +1 -1
      case IPV6_SADDR_RULE_PREFIX:
              /* Rule 8: Use longest matching prefix */
              ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);

@@ -3222,8 +3240,13 @@ static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
       * Optimistic nodes can start receiving
       * Frames right away
       */
-     if (ifp->flags & IFA_F_OPTIMISTIC)
+     if (ifp->flags & IFA_F_OPTIMISTIC) {
              ip6_ins_rt(ifp->rt);
Quoted 8 lines

net/ipv6/addrconf.c static void addrconf_dad_begin(struct inet6_ifaddr *ifp) · line 3249

Quoted 5 lines
net/ipv6/addrconf.c +5 -0
+             /* Because optimistic nodes can receive frames, notify
+              * listeners. If DAD fails, RTM_DELADDR is sent.
+              */
+             ipv6_ifa_notify(RTM_NEWADDR, ifp);
+     }

I wonder if we can now delete the ipv6_ifa_notify(RTM_NEWADDR, ifp) in addrconf_dad_completed.

Quoted 4 lines

I don't know what everyone's general preference would be, but mine would be to err on the side of notifying on state changes. It seems harmless to me to keep it in, and something in userspace might want to know if/when DAD completes.

Userspace expects to communicate with an address which gets announced via RTM_NEWADDR, so I would make the RTM_NEWADDR notifications conditional on use_optimistic flag in both, the completed and the dad_begin function. This sounds like the best option to me, no?

Bye && thanks,
Hannes

author
#4 lore raw
Quoted 152 lines · Hannes Frederic Sowa
Quoted 140 lines · Erik Kline

Hello.

Quoted 127 lines · Hannes Frederic Sowa

Hi,

Quoted 124 lines · Erik Kline

Add a sysctl that causes an interface's optimistic addresses to be considered equivalent to other non-deprecated addresses for source address selection purposes. Preferred addresses will still take precedence over optimistic addresses, subject to other ranking in the source address selection algorithm.

This is useful where different interfaces are connected to different networks from different ISPs (e.g., a cell network and a home wifi network).

The current behaviour complies with RFC 3484/6724, and it makes sense if the host has only one interface, or has multiple interfaces on the same network (same or cooperating administrative domain(s), but not in the multiple distinct networks case.

For example, if a mobile device has an IPv6 address on an LTE network and then connects to IPv6-enabled wifi, while the wifi IPv6 address is undergoing DAD, IPv6 connections will try use the wifi default route with the LTE IPv6 address, and will get stuck until they time out.

Also, because optimistic addresses can actually be used, issue an RTM_NEWADDR as soon as DAD starts. If DAD fails an separate RTM_DELADDR will be sent.

Also: add an entry in ip-sysctl.txt for optimistic_dad.

Signed-off-by Erik Kline <[email protected]>
---
 Documentation/networking/ip-sysctl.txt | 13 ++++++++++++
 include/linux/ipv6.h                   |  1 +
 include/uapi/linux/ipv6.h              |  1 +
 net/ipv6/addrconf.c                    | 36 ++++++++++++++++++++++++++++++++--
 4 files changed, 49 insertions(+), 2 deletions(-)
Documentation/networking/ip-sysctl.txt +13 -0
index 0307e28..e03cf49 100644
@@ -1452,6 +1452,19 @@ suppress_frag_ndisc - INTEGER
      1 - (default) discard fragmented neighbor discovery packets
      0 - allow fragmented neighbor discovery packets
 
+optimistic_dad - BOOLEAN
+     Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
+             0: disabled (default)
+             1: enabled
+
+use_optimistic - BOOLEAN
+     If enabled, do not classify optimistic addresses as deprecated during
+     source address selection.  Preferred addresses will still be chosen
+     before optimistic addresses, subject to other ranking in the source
+     address selection algorithm.
+             0: disabled (default)
+             1: enabled
+
 icmp/*:
 ratelimit - INTEGER
      Limit the maximal rates for sending ICMPv6 packets.
include/linux/ipv6.h +1 -0
index ff56053..7121a2e 100644
@@ -42,6 +42,7 @@ struct ipv6_devconf {
      __s32           accept_ra_from_local;
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
      __s32           optimistic_dad;
+     __s32           use_optimistic;
 #endif
 #ifdef CONFIG_IPV6_MROUTE
      __s32           mc_forwarding;
include/uapi/linux/ipv6.h +1 -0
index efa2666..e863d08 100644
@@ -164,6 +164,7 @@ enum {
      DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL,
      DEVCONF_SUPPRESS_FRAG_NDISC,
      DEVCONF_ACCEPT_RA_FROM_LOCAL,
+     DEVCONF_USE_OPTIMISTIC,
      DEVCONF_MAX
 };
net/ipv6/addrconf.c +14 -1
index 725c763..c2fddb7 100644
@@ -1169,6 +1169,9 @@ enum {
      IPV6_SADDR_RULE_LABEL,
      IPV6_SADDR_RULE_PRIVACY,
      IPV6_SADDR_RULE_ORCHID,
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+     IPV6_SADDR_RULE_NOT_OPTIMISTIC,
+#endif
      IPV6_SADDR_RULE_PREFIX,
      IPV6_SADDR_RULE_MAX
 };

@@ -1257,10 +1260,17 @@ static int ipv6_get_saddr_eval(struct net *net,
              score->scopedist = ret;
              break;
      case IPV6_SADDR_RULE_PREFERRED:
+         {
              /* Rule 3: Avoid deprecated and optimistic addresses */
+             u8 avoid = IFA_F_DEPRECATED;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+             if (!score->ifa->idev->cnf.use_optimistic)
+                     avoid |= IFA_F_OPTIMISTIC;
+#endif
              ret = ipv6_saddr_preferred(score->addr_type) ||
-                   !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
+                   !(score->ifa->flags & avoid);
              break;
+         }
 #ifdef CONFIG_IPV6_MIP6
      case IPV6_SADDR_RULE_HOA:
          {

@@ -1299,6 +1309,14 @@ static int ipv6_get_saddr_eval(struct net *net,
              ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
                      ipv6_addr_orchid(dst->addr));
              break;
+#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
+     case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
+             /* Optimistic addresses still have lower precedence than other
Quoted 8 lines

net/ipv6/addrconf.c static int ipv6_get_saddr_eval(struct net *net, · line 1319

Quoted 5 lines
net/ipv6/addrconf.c +5 -0
+              * preferred addresses.
+              */
+             ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
+             break;
+#endif

I wonder a bit why this rule is not directly ordered after IPV6_SADDR_RULE_PREFERRED? This would e.g. matter for privacy addresses.

Quoted 8 lines

Privacy addresses ("tempaddrs") that win in earlier checks are preferred before optimistic in this code (i.e. a tempaddr on the same outgoing interface is preferred before an optimistic address).

Similarly, a non-tentative non-privacy address (same outgoing interface, same label, ...) will match before an optimistic address, but only until DAD completes and the address is no longer optimistic. I think this is in keeping with the spirit of the RFC 3484/6724 rules.

Oh yes, I see. I had the evaluation order messed up. ;)

Quoted 7 lines

So the question I should be asking would be. AFAIR optimistic addresses should be handled like deprecated ones, so I am a bit concerned adding a non-conditional rule before the RULE_PREFIX check.

Shouldn't we only break the tie *after* longest prefix match then? If you do that before longest prefix match I would prefer ret being masked by use_optimisitic flag.

There was some text suggesting that the longest prefix could be ignored (like for DNS load-balancing reasons). But I think in this particular case it doesn't matter, so I'll move it after. Ack.

Quoted 30 lines
Quoted 2 lines

After there's an RFC 7217 implementation, EUI-64-based SLAAC could be disabled by folks.

Ack.

Quoted 20 lines
Quoted 11 lines
Quoted 11 lines
net/ipv6/addrconf.c +1 -1
      case IPV6_SADDR_RULE_PREFIX:
              /* Rule 8: Use longest matching prefix */
              ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);

@@ -3222,8 +3240,13 @@ static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
       * Optimistic nodes can start receiving
       * Frames right away
       */
-     if (ifp->flags & IFA_F_OPTIMISTIC)
+     if (ifp->flags & IFA_F_OPTIMISTIC) {
              ip6_ins_rt(ifp->rt);
Quoted 8 lines

net/ipv6/addrconf.c static void addrconf_dad_begin(struct inet6_ifaddr *ifp) · line 3249

Quoted 5 lines
net/ipv6/addrconf.c +5 -0
+             /* Because optimistic nodes can receive frames, notify
+              * listeners. If DAD fails, RTM_DELADDR is sent.
+              */
+             ipv6_ifa_notify(RTM_NEWADDR, ifp);
+     }

I wonder if we can now delete the ipv6_ifa_notify(RTM_NEWADDR, ifp) in addrconf_dad_completed.

Quoted 4 lines

I don't know what everyone's general preference would be, but mine would be to err on the side of notifying on state changes. It seems harmless to me to keep it in, and something in userspace might want to know if/when DAD completes.

Quoted 4 lines

Userspace expects to communicate with an address which gets announced via RTM_NEWADDR, so I would make the RTM_NEWADDR notifications conditional on use_optimistic flag in both, the completed and the dad_begin function. This sounds like the best option to me, no?

That's easy enough to do in addrconf_dad_begin(). Unfortunately, by the time we get to addrconf_dad_completed() the IFA_F_OPTIMISTIC flag has already been cleared.

I have a version that attempts to take its best guess as to whether an RTM_NEWADDR _should_ have already been sent--something like:

#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
        // We probably already sent a notification in addrconf_dad_begin().
        if (!ifp->idev->cnf.optimistic_dad || !ifp->idev->cnf.use_optimistic)
#endif
        ipv6_ifa_notify(RTM_NEWADDR, ifp);

but that doesn't seem that clean to me (apart from the ifdef-y messiness of it). Do you think this "best guess" approach is reasonable?

With just the "use_optimistic" check in addrconf_dad_begin(), userspace can still communicate with addresses it gets via RTM_NEWADDR, it will just get /two/ such notifications: one when it can probably use it and one when it definitely can.

Thoughts?

#5 lore raw

Hi,

Quoted 42 lines · Erik Kline
Quoted 24 lines
Quoted 19 lines
Quoted 10 lines
Quoted 10 lines
      case IPV6_SADDR_RULE_PREFIX:
              /* Rule 8: Use longest matching prefix */
              ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
net/ipv6/addrconf.c +1 -1
@@ -3222,8 +3240,13 @@ static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
       * Optimistic nodes can start receiving
       * Frames right away
       */
-     if (ifp->flags & IFA_F_OPTIMISTIC)
+     if (ifp->flags & IFA_F_OPTIMISTIC) {
              ip6_ins_rt(ifp->rt);
Quoted 8 lines

net/ipv6/addrconf.c static void addrconf_dad_begin(struct inet6_ifaddr *ifp) · line 3249

Quoted 5 lines
net/ipv6/addrconf.c +5 -0
+             /* Because optimistic nodes can receive frames, notify
+              * listeners. If DAD fails, RTM_DELADDR is sent.
+              */
+             ipv6_ifa_notify(RTM_NEWADDR, ifp);
+     }

I wonder if we can now delete the ipv6_ifa_notify(RTM_NEWADDR, ifp) in addrconf_dad_completed.

Quoted 4 lines

I don't know what everyone's general preference would be, but mine would be to err on the side of notifying on state changes. It seems harmless to me to keep it in, and something in userspace might want to know if/when DAD completes.

Quoted 4 lines

Userspace expects to communicate with an address which gets announced via RTM_NEWADDR, so I would make the RTM_NEWADDR notifications conditional on use_optimistic flag in both, the completed and the dad_begin function. This sounds like the best option to me, no?

That's easy enough to do in addrconf_dad_begin(). Unfortunately, by the time we get to addrconf_dad_completed() the IFA_F_OPTIMISTIC flag has already been cleared.

I have a version that attempts to take its best guess as to whether an RTM_NEWADDR _should_ have already been sent--something like:

#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
        // We probably already sent a notification in addrconf_dad_begin().
        if (!ifp->idev->cnf.optimistic_dad || !ifp->idev->cnf.use_optimistic)
#endif
        ipv6_ifa_notify(RTM_NEWADDR, ifp);
Quoted 3 lines

but that doesn't seem that clean to me (apart from the ifdef-y messiness of it). Do you think this "best guess" approach is reasonable?

I would definitely ack a patch which removes the macros and the config option CONFIG_IPV6_OPTIMISTIC_DAD completely.

You also can split the ifdefed part into a small helper function:

static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev) {

#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
        return idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
#else
        return false;

#endif
}

...

and then in both locations update the RTM_NEWADDR like this:

addrconf_dad_begin:

if (ipv6_use_optimisitic_addr(idev))
notify...

and the dad_completed with a negated check. I think this is the easiest option.

Quoted 6 lines

With just the "use_optimistic" check in addrconf_dad_begin(), userspace can still communicate with addresses it gets via RTM_NEWADDR, it will just get /two/ such notifications: one when it can probably use it and one when it definitely can.

Thoughts?

In the near future we also must introduce specific DAD events needed for RFC7217 addresses (to count DAD progress and safe the information per prefix in user space). I would prefer to keep the logic for RTM_NEWADDR that the kernel will definitely allow the use of the address but RTM_NEWADDR should be handled idempotent by user space.

Thanks,
Hannes

#6 lore raw
Quoted 5 lines · Hannes Frederic Sowa

In the near future we also must introduce specific DAD events needed for RFC7217 addresses (to count DAD progress and safe the information per prefix in user space). I would prefer to keep the logic for RTM_NEWADDR that the kernel will definitely allow the use of the address but RTM_NEWADDR should be handled idempotent by user space.

Duplicate RTM_NEWADDR notifications aren't a problem though, right? The current code send RTM_NEWADDR every time an RA increases the lifetime of an address, so I think it's reasonable to assume that apps have to be able to deal with it, because it's pretty frequent.

#7 lore raw
Quoted 8 lines · Lorenzo Colitti
Quoted 5 lines · Hannes Frederic Sowa

In the near future we also must introduce specific DAD events needed for RFC7217 addresses (to count DAD progress and safe the information per prefix in user space). I would prefer to keep the logic for RTM_NEWADDR that the kernel will definitely allow the use of the address but RTM_NEWADDR should be handled idempotent by user space.

Quoted 4 lines

Duplicate RTM_NEWADDR notifications aren't a problem though, right? The current code send RTM_NEWADDR every time an RA increases the lifetime of an address, so I think it's reasonable to assume that apps have to be able to deal with it, because it's pretty frequent.

I don't think it is a serious problem (or a problem at all). I would just try to reduce the number of notifications, that's all.

Bye,
Hannes