net/ipv6: Add support for path selection using hash of 5-tuple
Subsystems (5)
cover net/ipv6: Add support for path selection using hash of 5-tuple
Hardware supports multipath selection using the standard L4 5-tuple instead of just L3 and the flow label. In addition, some network operators prefer IPv6 path selection to use the 5-tuple. To that end, add support to IPv6 for multipath hash policy similar to bf4e0a3db97eb ("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.
A separate sysctl is added for IPv6, allowing IPv4 and IPv6 to use different algorithms if desired.
The first 2 patches modify the IPv4 variant so that at the end of the patch set the ipv4 and ipv6 implementations are direct parallels.
Patch 3 refactors the existing rt6_multipath_hash in preparation for adding the policy option.
Patch 4 renames the existing netevent to have IPv4 in the name so ipv4 changes can be distinguished from IPv6 if the netevent handler cares.
Patch 5 adds the L4 hash support.
Patch 6 adds the hook for the netevent to the spectrum driver to update the ASIC.
Patch 7 removes no longer used code.
David Ahern (7):
net/ipv4: Pass net to fib_multipath_hash instead of fib_info
net: Align ip_multipath_l3_keys and ip6_multipath_l3_keys net/ipv6: Make rt6_multipath_hash similar to fib_multipath_hash
net: Rename NETEVENT_MULTIPATH_HASH_UPDATE
net/ipv6: Add support for path selection using hash of 5-tuple
mlxsw: spectrum_router: Add support for ipv6 hash policy update
net: Remove unused get_hash_from_flow functions
Documentation/networking/ip-sysctl.txt | 7 +++ .../net/ethernet/mellanox/mlxsw/spectrum_router.c | 13 ++++- include/net/flow.h | 15 ----- include/net/ip6_route.h | 3 +- include/net/ip_fib.h | 2 +- include/net/netevent.h | 3 +- include/net/netns/ipv6.h | 1 + net/core/flow_dissector.c | 16 ------ net/ipv4/fib_semantics.c | 2 +- net/ipv4/route.c | 25 +++++---- net/ipv4/sysctl_net_ipv4.c | 2 +- net/ipv6/icmp.c | 2 +- net/ipv6/route.c | 64 ++++++++++++++++++---- net/ipv6/sysctl_net_ipv6.c | 26 +++++++++ 14 files changed, 118 insertions(+), 63 deletions(-)
Signature · 1 line
2.11.0
RFC PATCH 1/7 net/ipv4: Pass net to fib_multipath_hash instead of fib_info
fib_multipath_hash only needs net struct to check a sysctl. Make it clear by passing net instead of fib_info. In the need this allows alignment between the ipv4 and ipv6 versions.
--- include/net/ip_fib.h | 2 +- net/ipv4/fib_semantics.c | 2 +- net/ipv4/route.c | 5 ++--- 3 files changed, 4 insertions(+), 5 deletions(-)
include/net/ip_fib.h +1 -1
@@ -370,7 +370,7 @@ int fib_sync_down_addr(struct net_device *dev, __be32 local);int fib_sync_up (struct net_device *dev,unsigned int nh_flags);#ifdef CONFIG_IP_ROUTE_MULTIPATH -int fib_multipath_hash (const struct fib_info *fi,const struct flowi4 *fl4, +int fib_multipath_hash (const struct net *net,const struct flowi4 *fl4,const struct sk_buff *skb);#endif void fib_select_multipath (struct fib_result *res,int hash);
net/ipv4/fib_semantics.c +1 -1
@@ -1767,7 +1767,7 @@ void fib_select_path(struct net *net, struct fib_result *res,#ifdef CONFIG_IP_ROUTE_MULTIPATH if (res->fi->fib_nhs >1 && oif_check) { -int h =fib_multipath_hash (res->fi, fl4, skb); +int h =fib_multipath_hash (res->fi->fib_net, fl4, skb);fib_select_multipath (res, h); }
net/ipv4/route.c +2 -3
@@ -1797,10 +1797,9 @@ static void ip_multipath_l3_keys(const struct sk_buff *skb, }/* if skb is set it will be used and fl4 can be NULL */ -int fib_multipath_hash (const struct fib_info *fi,const struct flowi4 *fl4, +int fib_multipath_hash (const struct net *net,const struct flowi4 *fl4,const struct sk_buff *skb) { -struct net *net = fi->fib_net;struct flow_keys hash_keys;u32 mhash;
@@ -1856,7 +1855,7 @@ static int ip_mkroute_input(struct sk_buff *skb, {#ifdef CONFIG_IP_ROUTE_MULTIPATH if (res->fi && res->fi->fib_nhs >1 ) { -int h =fib_multipath_hash (res->fi,NULL , skb); +int h =fib_multipath_hash (res->fi->fib_net,NULL , skb);fib_select_multipath (res, h); }
Signature · 1 line
2.11.0
RFC PATCH 2/7 net: Align ip_multipath_l3_keys and ip6_multipath_l3_keys
In reply to David AhernSymmetry is good and allows easy comparison that ipv4 and ipv6 are doing the same thing. To that end, change ip_multipath_l3_keys to set addresses at the end after the icmp compares, and move the initialization of ipv6 flow keys to rt6_multipath_hash.
--- net/ipv4/route.c | 20 +++++++++++--------- net/ipv6/route.c | 4 ++-- 2 files changed, 13 insertions(+), 11 deletions(-)
net/ipv4/route.c +11 -9
@@ -1763,37 +1763,39 @@ static void ip_multipath_l3_keys(const struct sk_buff *skb, struct flow_keys *hash_keys) {const struct iphdr *outer_iph =ip_hdr (skb); +const struct iphdr *key_iph = outer_iph;const struct iphdr *inner_iph;const struct icmphdr *icmph;struct iphdr _inner_iph;struct icmphdr _icmph; - hash_keys->addrs.v4addrs.src = outer_iph->saddr; - hash_keys->addrs.v4addrs.dst = outer_iph->daddr;if (likely (outer_iph->protocol !=IPPROTO_ICMP )) -return ; + goto out;if (unlikely ((outer_iph->frag_off &htons (IP_OFFSET )) !=0 )) -return ; + goto out; icmph =skb_header_pointer (skb, outer_iph->ihl *4 ,sizeof (_icmph), &_icmph);if (!icmph) -return ; + goto out;if (icmph->type !=ICMP_DEST_UNREACH && icmph->type !=ICMP_REDIRECT && icmph->type !=ICMP_TIME_EXCEEDED && icmph->type !=ICMP_PARAMETERPROB ) -return ; + goto out; inner_iph =skb_header_pointer (skb, outer_iph->ihl *4 +sizeof (_icmph),sizeof (_inner_iph), &_inner_iph);if (!inner_iph) -return ; - hash_keys->addrs.v4addrs.src = inner_iph->saddr; - hash_keys->addrs.v4addrs.dst = inner_iph->daddr; + goto out; + + key_iph = inner_iph; +out: + hash_keys->addrs.v4addrs.src = key_iph->saddr; + hash_keys->addrs.v4addrs.dst = key_iph->daddr; }/* if skb is set it will be used and fl4 can be NULL */
net/ipv6/route.c +2 -2
@@ -1812,8 +1812,6 @@ static void ip6_multipath_l3_keys(const struct sk_buff *skb, key_iph = inner_iph; out: -memset (keys,0 ,sizeof (*keys)); - keys->control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ; keys->addrs.v6addrs.src = key_iph->saddr; keys->addrs.v6addrs.dst = key_iph->daddr; keys->tags.flow_label =ip6_flowinfo (key_iph);
@@ -1826,6 +1824,8 @@ u32 rt6_multipath_hash(const struct flowi6 *fl6, const struct sk_buff *skb)struct flow_keys hash_keys;if (skb) { +memset (&hash_keys,0 ,sizeof (hash_keys)); + hash_keys.control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ;ip6_multipath_l3_keys (skb, &hash_keys);return flow_hash_from_keys (&hash_keys) >>1 ; }
Signature · 1 line
2.11.0
RFC PATCH 3/7 net/ipv6: Make rt6_multipath_hash similar to fib_multipath_hash
In reply to David AhernMake rt6_multipath_hash more of a direct parallel to fib_multipath_hash and reduce stack and overhead in the process: get_hash_from_flowi6 is just a wrapper around __get_hash_from_flowi6 with another stack allocation for flow_keys. Move setting the addresses, protocol and label into rt6_multipath_hash and allow it to make the call to flow_hash_from_keys.
--- net/ipv6/route.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-)
net/ipv6/route.c +10 -4
@@ -1822,15 +1822,21 @@ static void ip6_multipath_l3_keys(const struct sk_buff *skb,u32 rt6_multipath_hash (const struct flowi6 *fl6,const struct sk_buff *skb) {struct flow_keys hash_keys; +u32 mhash; +memset (&hash_keys,0 ,sizeof (hash_keys)); + hash_keys.control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ;if (skb) { -memset (&hash_keys,0 ,sizeof (hash_keys)); - hash_keys.control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ;ip6_multipath_l3_keys (skb, &hash_keys); -return flow_hash_from_keys (&hash_keys) >>1 ; + }else { + hash_keys.addrs.v6addrs.src = fl6->saddr; + hash_keys.addrs.v6addrs.dst = fl6->daddr; + hash_keys.tags.flow_label = (__force u32 )fl6->flowlabel; + hash_keys.basic.ip_proto = fl6->flowi6_proto; } + mhash =flow_hash_from_keys (&hash_keys); -return get_hash_from_flowi6 (fl6) >>1 ; +return mhash >>1 ; }void ip6_route_input (struct sk_buff *skb)
Signature · 1 line
2.11.0
RFC PATCH 4/7 net: Rename NETEVENT_MULTIPATH_HASH_UPDATE
In reply to David AhernRename NETEVENT_MULTIPATH_HASH_UPDATE to NETEVENT_IPV4_MPATH_HASH_UPDATE to denote it relates to a change in the IPv4 hash policy.
--- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 2 +- include/net/netevent.h | 2 +- net/ipv4/sysctl_net_ipv4.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-)
drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +1 -1
@@ -2413,7 +2413,7 @@ static int mlxsw_sp_router_netevent_event(struct notifier_block *nb,mlxsw_core_schedule_work (&net_work->work);mlxsw_sp_port_dev_put (mlxsw_sp_port);break ; -case NETEVENT_MULTIPATH_HASH_UPDATE : +case NETEVENT_IPV4_MPATH_HASH_UPDATE : net = ptr;if (!net_eq (net, &init_net))
include/net/netevent.h +1 -1
@@ -26,7 +26,7 @@ enum netevent_notif_type {NETEVENT_NEIGH_UPDATE =1 ,/* arg is struct neighbour ptr */ NETEVENT_REDIRECT ,/* arg is struct netevent_redirect ptr */ NETEVENT_DELAY_PROBE_TIME_UPDATE ,/* arg is struct neigh_parms ptr */ - NETEVENT_MULTIPATH_HASH_UPDATE, /* arg is struct net ptr */ + NETEVENT_IPV4_MPATH_HASH_UPDATE, /* arg is struct net ptr */ };int register_netevent_notifier (struct notifier_block *nb);
net/ipv4/sysctl_net_ipv4.c +1 -1
@@ -400,7 +400,7 @@ static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write, ret =proc_dointvec_minmax (table, write, buffer, lenp, ppos);if (write && ret ==0 ) -call_netevent_notifiers (NETEVENT_MULTIPATH_HASH_UPDATE , net); +call_netevent_notifiers (NETEVENT_IPV4_MPATH_HASH_UPDATE , net);return ret; }
Signature · 1 line
2.11.0
RFC PATCH 5/7 net/ipv6: Add support for path selection using hash of 5-tuple
In reply to David AhernSome operators prefer IPv6 path selection to use a standard 5-tuple hash rather than just an L3 hash with the flow the label. To that end add support to IPv6 for multipath hash policy similar to bf4e0a3db97eb ("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.
--- Documentation/networking/ip-sysctl.txt | 7 ++++ include/net/ip6_route.h | 3 +- include/net/netevent.h | 1 + include/net/netns/ipv6.h | 1 + net/ipv6/icmp.c | 2 +- net/ipv6/route.c | 64 ++++++++++++++++++++++++++-------- net/ipv6/sysctl_net_ipv6.c | 26 ++++++++++++++ 7 files changed, 87 insertions(+), 17 deletions(-)
Documentation/networking/ip-sysctl.txt +7 -0
@@ -1363,6 +1363,13 @@ flowlabel_reflect - BOOLEAN FALSE: disabled Default: FALSE +fib_multipath_hash_policy - INTEGER + Controls which hash policy to use for multipath routes. + Default: 0 (Layer 3) + Possible values: + 0 - Layer 3 (source and destination addresses plus flow label) + 1 - Layer 4 (standard 5-tuple) + anycast_src_echo_reply - BOOLEAN Controls the use of anycast addresses as source addresses for ICMPv6 echo reply
include/net/ip6_route.h +2 -1
@@ -127,7 +127,8 @@ static inline int ip6_route_get_saddr(struct net *net, struct rt6_info *rt,struct rt6_info *rt6_lookup (struct net *net,const struct in6_addr *daddr,const struct in6_addr *saddr,int oif,int flags); -u32 rt6_multipath_hash (const struct flowi6 *fl6,const struct sk_buff *skb); +u32 rt6_multipath_hash (const struct net *net,const struct flowi6 *fl6, +const struct sk_buff *skb);struct dst_entry *icmp6_dst_alloc (struct net_device *dev,struct flowi6 *fl6);
include/net/netevent.h +1 -0
@@ -27,6 +27,7 @@ enum netevent_notif_type {NETEVENT_REDIRECT ,/* arg is struct netevent_redirect ptr */ NETEVENT_DELAY_PROBE_TIME_UPDATE ,/* arg is struct neigh_parms ptr */ NETEVENT_IPV4_MPATH_HASH_UPDATE ,/* arg is struct net ptr */ + NETEVENT_IPV6_MPATH_HASH_UPDATE, /* arg is struct net ptr */ };int register_netevent_notifier (struct notifier_block *nb);
include/net/netns/ipv6.h +1 -0
@@ -28,6 +28,7 @@ struct netns_sysctl_ipv6 {int ip6_rt_gc_elasticity;int ip6_rt_mtu_expires;int ip6_rt_min_advmss; +int multipath_hash_policy;int flowlabel_consistency;int auto_flowlabels;int icmpv6_time;
net/ipv6/icmp.c +1 -1
@@ -522,7 +522,7 @@ static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info, fl6.fl6_icmp_type = type; fl6.fl6_icmp_code = code; fl6.flowi6_uid =sock_net_uid (net,NULL ); - fl6.mp_hash =rt6_multipath_hash (&fl6, skb); + fl6.mp_hash =rt6_multipath_hash (net, &fl6, skb);security_skb_classify_flow (skb,flowi6_to_flowi (&fl6)); sk =icmpv6_xmit_lock (net);
net/ipv6/route.c +49 -15
104 lines not on this page · raw message
net/ipv6/sysctl_net_ipv6.c +26 -0
@@ -16,14 +16,31 @@#include <net/ipv6.h> #include <net/addrconf.h> #include <net/inet_frag.h> +#include <net/netevent.h> #ifdef CONFIG_NETLABEL #include <net/calipso.h> #endif +static int zero;static int one =1 ;static int auto_flowlabels_min;static int auto_flowlabels_max =IP6_AUTO_FLOW_LABEL_MAX ; +static int proc_rt6_multipath_hash_policy (struct ctl_table *table,int write, +void __user *buffer,size_t *lenp, +loff_t *ppos) +{ +struct net *net; +int ret; + + net = container_of(table->data, struct net, + ipv6.sysctl.multipath_hash_policy); + ret =proc_dointvec_minmax (table, write, buffer, lenp, ppos); +if (write && ret ==0 ) +call_netevent_notifiers (NETEVENT_IPV6_MPATH_HASH_UPDATE , net); + +return ret; +}static struct ctl_table ipv6_table_template[] = { {
@@ -126,6 +143,15 @@ static struct ctl_table ipv6_table_template[] = { .mode = 0644, .proc_handler = proc_dointvec }, + { + .procname = "fib_multipath_hash_policy", + .data = &init_net.ipv6.sysctl.multipath_hash_policy, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_rt6_multipath_hash_policy, + .extra1 = &zero, + .extra2 = &one, + }, { } };
Signature · 1 line
2.11.0
Le 13/02/2018 à 01:06, David Ahern a écrit :
Quoted 9 lines
Some operators prefer IPv6 path selection to use a standard 5-tuple hash rather than just an L3 hash with the flow the label. To that end add support to IPv6 for multipath hash policy similar to
bf4e0a3db97eb("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.Signed-off-by David Ahern <[email protected]>---
Do you plan to add the nl NETCONF support for this new sysctl?
Regards,
Nicolas
Quoted 8 lines · Nicolas Dichtel
Le 13/02/2018 à 01:06, David Ahern a écrit :
Quoted 7 lines
Some operators prefer IPv6 path selection to use a standard 5-tuple hash rather than just an L3 hash with the flow the label. To that end add support to IPv6 for multipath hash policy similar to
bf4e0a3db97eb("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.
Quoted 4 lines
Quoted 2 lines
Signed-off-by David Ahern <[email protected]>---
Do you plan to add the nl NETCONF support for this new sysctl?
hmmm.... doesn't exist for the current IPv4 version, but in-kernel drivers are notified. I wasn't planning on it, but it can be added.
Le 13/02/2018 à 21:35, David Ahern a écrit :
Quoted 9 lines · Nicolas Dichtel
Quoted 8 lines
Le 13/02/2018 à 01:06, David Ahern a écrit :
Quoted 7 lines
Some operators prefer IPv6 path selection to use a standard 5-tuple hash rather than just an L3 hash with the flow the label. To that end add support to IPv6 for multipath hash policy similar to
bf4e0a3db97eb("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines
Signed-off-by David Ahern <[email protected]>---
Do you plan to add the nl NETCONF support for this new sysctl?
hmmm.... doesn't exist for the current IPv4 version, but in-kernel drivers are notified. I wasn't planning on it, but it can be added.
Right for ipv4, it was on my todo list ;-)
Quoted 12 lines · Nicolas Dichtel
Le 13/02/2018 à 21:35, David Ahern a écrit :
Quoted 11 lines · Nicolas Dichtel
Quoted 10 lines
Le 13/02/2018 à 01:06, David Ahern a écrit :
Quoted 9 lines
Some operators prefer IPv6 path selection to use a standard 5-tuple hash rather than just an L3 hash with the flow the label. To that end add support to IPv6 for multipath hash policy similar to
bf4e0a3db97eb("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.Signed-off-by David Ahern <[email protected]>---
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
Do you plan to add the nl NETCONF support for this new sysctl?
hmmm.... doesn't exist for the current IPv4 version, but in-kernel drivers are notified. I wasn't planning on it, but it can be added.
Right for ipv4, it was on my todo list ;-)
Awesome, then ipv6 is a 1-liner after you add ipv4 support. :-)
Le 13/02/2018 à 22:02, David Ahern a écrit :
Quoted 16 lines · Nicolas Dichtel
Quoted 15 lines
Le 13/02/2018 à 21:35, David Ahern a écrit :
Quoted 14 lines · Nicolas Dichtel
Quoted 12 lines
Le 13/02/2018 à 01:06, David Ahern a écrit :
Quoted 9 lines
Some operators prefer IPv6 path selection to use a standard 5-tuple hash rather than just an L3 hash with the flow the label. To that end add support to IPv6 for multipath hash policy similar to
bf4e0a3db97eb("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.Signed-off-by David Ahern <[email protected]>---
Do you plan to add the nl NETCONF support for this new sysctl?
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
hmmm.... doesn't exist for the current IPv4 version, but in-kernel drivers are notified. I wasn't planning on it, but it can be added.
Right for ipv4, it was on my todo list ;-)
Awesome, then ipv6 is a 1-liner after you add ipv4 support. :-)
Sure :)
Quoted 8 lines · David Ahern
Some operators prefer IPv6 path selection to use a standard 5-tuple hash rather than just an L3 hash with the flow the label. To that end add support to IPv6 for multipath hash policy similar to
bf4e0a3db97eb("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.Signed-off-by David Ahern <[email protected]>
[...]
Quoted 38 lines
net/ipv6/route.c +20
-10@@ -1819,20 +1820,53 @@ static void ip6_multipath_l3_keys(const struct sk_buff *skb, }/* if skb is set it will be used and fl6 can be NULL */ -u32 rt6_multipath_hash (const struct flowi6 *fl6,const struct sk_buff *skb) +u32 rt6_multipath_hash (const struct net *net,const struct flowi6 *fl6, +const struct sk_buff *skb) {struct flow_keys hash_keys;u32 mhash; -memset (&hash_keys,0 ,sizeof (hash_keys)); - hash_keys.control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ; -if (skb) { -ip6_multipath_l3_keys (skb, &hash_keys); - }else { - hash_keys.addrs.v6addrs.src = fl6->saddr; - hash_keys.addrs.v6addrs.dst = fl6->daddr; - hash_keys.tags.flow_label = (__force u32 )fl6->flowlabel; - hash_keys.basic.ip_proto = fl6->flowi6_proto; +switch (net->ipv6.sysctl.multipath_hash_policy) { +case 0 : +memset (&hash_keys,0 ,sizeof (hash_keys)); + hash_keys.control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ; +if (skb) { +ip6_multipath_l3_keys (skb, &hash_keys); + }else { + hash_keys.addrs.v6addrs.src = fl6->saddr; + hash_keys.addrs.v6addrs.dst = fl6->daddr; + hash_keys.tags.flow_label = (__force u32 )fl6->flowlabel; + hash_keys.basic.ip_proto = fl6->flowi6_proto; + } +break ; +case 1 : +if (skb) { +unsigned int flag =FLOW_DISSECTOR_F_STOP_AT_ENCAP ; +struct flow_keys keys; +
net/ipv6/route.c static void ip6_multipath_l3_keys(const struct sk_buff *skb, · line 1856
Quoted 10 lines
net/ipv6/route.c +10
-0+/* short-circuit if we already have L4 hash present */ +if (skb->l4_hash) +return skb_get_hash_raw (skb) >>1 ; + +memset (&hash_keys,0 ,sizeof (hash_keys)); + +skb_flow_dissect_flow_keys (skb, &keys, flag); + + hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src; + hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
Shouldn't you add:
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
?
Otherwise flow_hash_from_keys() will not consistentify the ports and the addresses and it will also not use the addresses for the hash computation.
It's missing from fib_multipath_hash() as well, so I might be missing something.
net/ipv6/route.c static void ip6_multipath_l3_keys(const struct sk_buff *skb, · line 1859
Quoted 3 lines
net/ipv6/route.c +3
-0+ hash_keys.ports.src = keys.ports.src; + hash_keys.ports.dst = keys.ports.dst; + hash_keys.tags.flow_label = keys.tags.flow_label;
Why are you using the flow label?
Quoted 22 lines
net/ipv6/route.c +11
-1+ hash_keys.basic.ip_proto = keys.basic.ip_proto; + }else { +memset (&hash_keys,0 ,sizeof (hash_keys)); + hash_keys.control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ; + hash_keys.addrs.v6addrs.src = fl6->saddr; + hash_keys.addrs.v6addrs.dst = fl6->daddr; + hash_keys.ports.src = fl6->fl6_sport; + hash_keys.ports.dst = fl6->fl6_dport; + hash_keys.basic.ip_proto = fl6->flowi6_proto; + } } mhash =flow_hash_from_keys (&hash_keys);@@ -1858,7 +1892,7 @@ void ip6_route_input(struct sk_buff *skb)if (tun_info && !(tun_info->mode &IP_TUNNEL_INFO_TX )) fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;if (unlikely (fl6.flowi6_proto ==IPPROTO_ICMPV6 )) - fl6.mp_hash =rt6_multipath_hash (&fl6, skb); + fl6.mp_hash =rt6_multipath_hash (net, &fl6, skb);skb_dst_drop (skb);skb_dst_set (skb,ip6_route_input_lookup (net, skb->dev, &fl6, flags)); }
Quoted 48 lines · Ido Schimmel
Quoted 33 lines
net/ipv6/route.c +20
-9+u32 rt6_multipath_hash (const struct net *net,const struct flowi6 *fl6, +const struct sk_buff *skb) {struct flow_keys hash_keys;u32 mhash; -memset (&hash_keys,0 ,sizeof (hash_keys)); - hash_keys.control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ; -if (skb) { -ip6_multipath_l3_keys (skb, &hash_keys); - }else { - hash_keys.addrs.v6addrs.src = fl6->saddr; - hash_keys.addrs.v6addrs.dst = fl6->daddr; - hash_keys.tags.flow_label = (__force u32 )fl6->flowlabel; - hash_keys.basic.ip_proto = fl6->flowi6_proto; +switch (net->ipv6.sysctl.multipath_hash_policy) { +case 0 : +memset (&hash_keys,0 ,sizeof (hash_keys)); + hash_keys.control.addr_type =FLOW_DISSECTOR_KEY_IPV6_ADDRS ; +if (skb) { +ip6_multipath_l3_keys (skb, &hash_keys); + }else { + hash_keys.addrs.v6addrs.src = fl6->saddr; + hash_keys.addrs.v6addrs.dst = fl6->daddr; + hash_keys.tags.flow_label = (__force u32 )fl6->flowlabel; + hash_keys.basic.ip_proto = fl6->flowi6_proto; + } +break ; +case 1 : +if (skb) { +unsigned int flag =FLOW_DISSECTOR_F_STOP_AT_ENCAP ; +struct flow_keys keys; +net/ipv6/route.c static void ip6_multipath_l3_keys(const struct sk_buff *skb, · line 1856
Quoted 10 lines
net/ipv6/route.c +10
-0+/* short-circuit if we already have L4 hash present */ +if (skb->l4_hash) +return skb_get_hash_raw (skb) >>1 ; + +memset (&hash_keys,0 ,sizeof (hash_keys)); + +skb_flow_dissect_flow_keys (skb, &keys, flag); + + hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src; + hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;Shouldn't you add:
hash_keys.control.addr_type =
FLOW_DISSECTOR_KEY_IPV6_ADDRS;
Quoted 8 lines
?
Otherwise
flow_hash_from_keys()will not consistentify the ports and the addresses and it will also not use the addresses for the hash computation.It's missing from
fib_multipath_hash()as well, so I might be missing something.
Yes, I think you are correct. The oversight here is based on the missing set in the ipv4 variant.
Quoted 6 lines
net/ipv6/route.c static void ip6_multipath_l3_keys(const struct sk_buff *skb, · line 1859
Quoted 4 lines
net/ipv6/route.c +3
-0+ hash_keys.ports.src = keys.ports.src; + hash_keys.ports.dst = keys.ports.dst; + hash_keys.tags.flow_label = keys.tags.flow_label;Why are you using the flow label?
will remove. It is supposed to be an L4 hash.
RFC PATCH 6/7 mlxsw: spectrum_router: Add support for ipv6 hash policy update
In reply to David AhernSimilar to 28678f07f127d ("mlxsw: spectrum_router: Update multipath hash parameters upon netevents") for IPv4, make sure the kernel and asic are using the same hash algorithm for path selection.
--- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +10 -1
@@ -2414,6 +2414,7 @@ static int mlxsw_sp_router_netevent_event(struct notifier_block *nb,mlxsw_sp_port_dev_put (mlxsw_sp_port);break ;case NETEVENT_IPV4_MPATH_HASH_UPDATE : +case NETEVENT_IPV6_MPATH_HASH_UPDATE : net = ptr;if (!net_eq (net, &init_net))
@@ -7006,13 +7007,21 @@ static void mlxsw_sp_mp4_hash_init(char *recr2_pl)static void mlxsw_sp_mp6_hash_init (char *recr2_pl) { +bool only_l3 = !init_net.ipv6.sysctl.multipath_hash_policy; +mlxsw_sp_mp_hash_header_set (recr2_pl,MLXSW_REG_RECR2_IPV6_EN_NOT_TCP_NOT_UDP );mlxsw_sp_mp_hash_header_set (recr2_pl,MLXSW_REG_RECR2_IPV6_EN_TCP_UDP );mlxsw_reg_recr2_ipv6_sip_enable (recr2_pl);mlxsw_reg_recr2_ipv6_dip_enable (recr2_pl); -mlxsw_sp_mp_hash_field_set (recr2_pl,MLXSW_REG_RECR2_IPV6_FLOW_LABEL );mlxsw_sp_mp_hash_field_set (recr2_pl,MLXSW_REG_RECR2_IPV6_NEXT_HEADER ); +if (only_l3) { +mlxsw_sp_mp_hash_field_set (recr2_pl,MLXSW_REG_RECR2_IPV6_FLOW_LABEL ); + }else { +mlxsw_sp_mp_hash_header_set (recr2_pl,MLXSW_REG_RECR2_TCP_UDP_EN_IPV6 ); +mlxsw_sp_mp_hash_field_set (recr2_pl,MLXSW_REG_RECR2_TCP_UDP_SPORT ); +mlxsw_sp_mp_hash_field_set (recr2_pl,MLXSW_REG_RECR2_TCP_UDP_DPORT ); + } }static int mlxsw_sp_mp_hash_init (struct mlxsw_sp *mlxsw_sp)
Signature · 1 line
2.11.0
RFC PATCH 7/7 net: Remove unused get_hash_from_flow functions
In reply to David Ahern__get_hash_from_flowi6 is still used for flowlabels, but the IPv4 variant and the wrappers to both are not used. Remove them.
--- include/net/flow.h | 15 --------------- net/core/flow_dissector.c | 16 ---------------- 2 files changed, 31 deletions(-)
include/net/flow.h +0 -15
@@ -221,21 +221,6 @@ static inline unsigned int flow_key_size(u16 family) }__u32 __get_hash_from_flowi6 (const struct flowi6 *fl6,struct flow_keys *keys); - -static inline __u32 get_hash_from_flowi6 (const struct flowi6 *fl6) -{ -struct flow_keys keys; - -return __get_hash_from_flowi6 (fl6, &keys); -} -__u32 __get_hash_from_flowi4 (const struct flowi4 *fl4,struct flow_keys *keys); -static inline __u32 get_hash_from_flowi4 (const struct flowi4 *fl4) -{ -struct flow_keys keys; - -return __get_hash_from_flowi4 (fl4, &keys); -} -#endif
net/core/flow_dissector.c +0 -16
@@ -1341,22 +1341,6 @@ __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys) }EXPORT_SYMBOL (__get_hash_from_flowi6); -__u32 __get_hash_from_flowi4 (const struct flowi4 *fl4,struct flow_keys *keys) -{ -memset (keys,0 ,sizeof (*keys)); - - keys->addrs.v4addrs.src = fl4->saddr; - keys->addrs.v4addrs.dst = fl4->daddr; - keys->control.addr_type =FLOW_DISSECTOR_KEY_IPV4_ADDRS ; - keys->ports.src = fl4->fl4_sport; - keys->ports.dst = fl4->fl4_dport; - keys->keyid.keyid = fl4->fl4_gre_key; - keys->basic.ip_proto = fl4->flowi4_proto; - -return flow_hash_from_keys (keys); -} -EXPORT_SYMBOL (__get_hash_from_flowi4); -static const struct flow_dissector_key flow_keys_dissector_keys[] = { { .key_id =FLOW_DISSECTOR_KEY_CONTROL ,
Signature · 1 line
2.11.0
Quoted 3 lines · David Ahern
Hardware supports multipath selection using the standard L4 5-tuple instead of just L3 and the flow label. In addition, some network operators prefer IPv6 path selection to use the 5-tuple.
The HW supports using flow label and AFAIK that is the preferred approach by the community (?)
Quoted 1 line
To that end, add support to IPv6 for multipath hash policy
so a question comes up if/what are the disadvantaged
to support 5-tuple. E.g Tom was commenting that such DPI is problematic when multiple IPv6 header extensions are used.
Quoted 6 lines
similar to
bf4e0a3db97eb("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.A separate sysctl is added for IPv6, allowing IPv4 and IPv6 to use different algorithms if desired.
Quoted 5 lines · Or Gerlitz
Quoted 3 lines · David Ahern
Hardware supports multipath selection using the standard L4 5-tuple instead of just L3 and the flow label. In addition, some network operators prefer IPv6 path selection to use the 5-tuple.
Quoted 8 lines
The HW supports using flow label and AFAIK that is the preferred approach by the community (?)
Quoted 1 line
To that end, add support to IPv6 for multipath hash policy
so a question comes up if/what are the disadvantaged
to support 5-tuple. E.g Tom was commenting that such DPI is problematic when multiple IPv6 header extensions are used.
Tom is much more qualified to answer this, but I think the problem is that the flow label isn't always set. Also, apparently some devices change the flow label mid flow. See:
"At Fastly, this hashing is performed by an Ethernet switch ASIC, and to avoid breakage, the IPv6 hashing function must not include the flow label. As in IPv4, the hash function includes the source and destination information in the L3 and L4 headers."
https://blog.apnic.net/2018/01/11/ipv6-flow-label-misuse-hashing/
https://www.youtube.com/watch?v=b0CRjOpnT7w
https://pc.nanog.org/static/published/meetings/NANOG71/1531/20171003_Jaeggli_Lightning_Talk_Ipv6_v1.pdf
Quoted 11 lines · Ido Schimmel
Quoted 10 lines · Or Gerlitz
Quoted 3 lines · David Ahern
Hardware supports multipath selection using the standard L4 5-tuple instead of just L3 and the flow label. In addition, some network operators prefer IPv6 path selection to use the 5-tuple.
The HW supports using flow label and AFAIK that is the preferred approach by the community (?)
Quoted 1 line
To that end, add support to IPv6 for multipath hash policy
Quoted 7 lines
Quoted 3 lines
so a question comes up if/what are the disadvantaged
to support 5-tuple. E.g Tom was commenting that such DPI is problematic when multiple IPv6 header extensions are used.Tom is much more qualified to answer this, but I think the problem is that the flow label isn't always set. Also, apparently some devices change the flow label mid flow. See:
OK. But note we have two ends to deal with here (1) generation (2) usage
E.g if the kernel generates flow label but uses source port we have inconsistent environment. Problem is that the generation and usage typically don't happen on the same network point.
Quoted 5 lines · Or Gerlitz
[...] note we have two ends to deal with here (1) generation (2) usage
E.g if the kernel generates flow label but uses source port we have inconsistent environment. Problem is that the generation and usage typically don't happen on the same network point.
David, what's your thinking here? e.g for overly networks environments
Quoted 6 lines · Ido Schimmel
Quoted 5 lines · Or Gerlitz
Quoted 3 lines · David Ahern
Hardware supports multipath selection using the standard L4 5-tuple instead of just L3 and the flow label. In addition, some network operators prefer IPv6 path selection to use the 5-tuple.
Quoted 8 lines
Quoted 8 lines
The HW supports using flow label and AFAIK that is the preferred approach by the community (?)
Quoted 1 line
To that end, add support to IPv6 for multipath hash policy
so a question comes up if/what are the disadvantaged
to support 5-tuple. E.g Tom was commenting that such DPI is problematic when multiple IPv6 header extensions are used.
Pros and cons to both approaches (L3 only or L4). We (Cumulus Networks) use L4 5-tuple hash for both IPv4 and IPv6. When I asked around various experts all of them gave me a puzzled look as to why I was asking the question. Basically, the unanimous response was of course it is an L4 hash.
Quoted 13 lines
Tom is much more qualified to answer this, but I think the problem is that the flow label isn't always set. Also, apparently some devices change the flow label mid flow. See:
"At Fastly, this hashing is performed by an Ethernet switch ASIC, and to avoid breakage, the IPv6 hashing function must not include the flow label. As in IPv4, the hash function includes the source and destination information in the L3 and L4 headers."
https://blog.apnic.net/2018/01/11/ipv6-flow-label-misuse-hashing/
https://www.youtube.com/watch?v=b0CRjOpnT7w
https://pc.nanog.org/static/published/meetings/NANOG71/1531/20171003_Jaeggli_Lightning_Talk_Ipv6_v1.pdf
Quoted 12 lines · David Ahern
Quoted 11 lines · Ido Schimmel
Quoted 10 lines · Or Gerlitz
Quoted 3 lines · David Ahern
Hardware supports multipath selection using the standard L4 5-tuple instead of just L3 and the flow label. In addition, some network operators prefer IPv6 path selection to use the 5-tuple.
The HW supports using flow label and AFAIK that is the preferred approach by the community (?)
Quoted 1 line
To that end, add support to IPv6 for multipath hash policy
Quoted 8 lines
Quoted 3 lines
Quoted 3 lines
so a question comes up if/what are the disadvantaged
to support 5-tuple. E.g Tom was commenting that such DPI is problematic when multiple IPv6 header extensions are used.Pros and cons to both approaches (L3 only or L4). We (Cumulus Networks) use L4 5-tuple hash for both IPv4 and IPv6. When I asked around various experts all of them gave me a puzzled look as to why I was asking the question. Basically, the unanimous response was of course it is an L4 hash.
how the various systems you are dealing with do with traffic that involves ipv6 extension headers? what about environments with GRE? in ipv4 GRE fabrics are just broken for ECMP, in ipv6 they can fly with flow label but will crash again with L4 hash.
Quoted 22 lines · Or Gerlitz
Quoted 12 lines · David Ahern
Quoted 11 lines · Ido Schimmel
Quoted 10 lines · Or Gerlitz
Quoted 3 lines · David Ahern
Hardware supports multipath selection using the standard L4 5-tuple instead of just L3 and the flow label. In addition, some network operators prefer IPv6 path selection to use the 5-tuple.
The HW supports using flow label and AFAIK that is the preferred approach by the community (?)
Quoted 1 line
To that end, add support to IPv6 for multipath hash policy
Quoted 8 lines
Quoted 3 lines
Quoted 3 lines
so a question comes up if/what are the disadvantaged
to support 5-tuple. E.g Tom was commenting that such DPI is problematic when multiple IPv6 header extensions are used.Pros and cons to both approaches (L3 only or L4). We (Cumulus Networks) use L4 5-tuple hash for both IPv4 and IPv6. When I asked around various experts all of them gave me a puzzled look as to why I was asking the question. Basically, the unanimous response was of course it is an L4 hash.
Quoted 5 lines
how the various systems you are dealing with do with traffic that involves ipv6 extension headers? what about environments with GRE? in ipv4 GRE fabrics are just broken for ECMP, in ipv6 they can fly with flow label but will crash again with L4 hash.
If you like your ecmp hash algorithm, you can keep your ecmp hash algorithm.
This gives users a choice; it is not a requirement to move from L3 only to L4. Further, this makes IPv6 on par with IPv4 with a choice between L3 and L4 and allows users to decide what works best for them.
Quoted 7 lines · David Ahern
Quoted 4 lines · Or Gerlitz
how the various systems you are dealing with do with traffic that involves ipv6 extension headers? what about environments with GRE? in ipv4 GRE fabrics are just broken for ECMP, in ipv6 they can fly with flow label but will crash again with L4 hash.
Quoted 5 lines
If you like your ecmp hash algorithm, you can keep your ecmp hash algorithm.
This gives users a choice; it is not a requirement to move from L3 only to L4. Further, this makes IPv6 on par with IPv4 with a choice between L3 and L4 and allows users to decide what works best for them.
Looking in the code for tunnels e.g in the vxlan xmit path (but I believe this is the case for
other UDP tunnels as well), a call is made to generate the source udp port as the hash of
the overlay tuple regardless if we are on v4/v6. Next, when it comes to ipv6, the kernel
computes the flow label which effectively takes into account the inner header.
Hi David,
Quoted 20 lines · David Ahern
Hardware supports multipath selection using the standard L4 5-tuple instead of just L3 and the flow label. In addition, some network operators prefer IPv6 path selection to use the 5-tuple. To that end, add support to IPv6 for multipath hash policy similar to
bf4e0a3db97eb("net: ipv4: add support for ECMP hash policy choice"). The default is still L3 which covers source and destination addresses along with flow label and IPv6 protocol.A separate sysctl is added for IPv6, allowing IPv4 and IPv6 to use different algorithms if desired.
The first 2 patches modify the IPv4 variant so that at the end of the patch set the ipv4 and ipv6 implementations are direct parallels.
Patch 3 refactors the existing
rt6_multipath_hashin preparation for adding the policy option.Patch 4 renames the existing netevent to have IPv4 in the name so ipv4 changes can be distinguished from IPv6 if the netevent handler cares.
Quoted 6 lines
Patch 5 adds the L4 hash support.
Patch 6 adds the hook for the netevent to the spectrum driver to update the ASIC.
Patch 7 removes no longer used code.
Thanks for working on this.
I've yet to review the patches, but I did test them. The good news is that the offloaded path works as expected. Different flows (random UDP destination port, zero flowlabel) are hashed to different nexthops. With the default policy all the flows are hashed to the same nexthop.
However, it seems that the kernel isn't spreading the traffic as expected and the same nexthop is always hit. I'll look into it later this week when I have some more time.
Thanks!