From e99db35508cbe6b5fe4a16b8d50fcfe8fdfdedcc Mon Sep 17 00:00:00 2001 From: Volodymyr Huti Date: Thu, 14 Jul 2022 16:44:39 +0300 Subject: [PATCH] QPPB DEMO v1 (Kernel) --- include/net/dst.h | 1 + include/net/ip_fib.h | 5 ++++- include/net/route.h | 1 + include/uapi/linux/rtnetlink.h | 2 ++ net/ipv4/Makefile | 5 +++++ net/ipv4/fib_frontend.c | 6 +++++- net/ipv4/fib_lookup.h | 1 + net/ipv4/fib_semantics.c | 4 ++++ net/ipv4/fib_trie.c | 6 ++++-- net/ipv4/ip_forward.c | 21 +++++++++++++++++++++ net/ipv4/ip_output.c | 10 ++++++++++ net/ipv4/route.c | 4 +++- 12 files changed, 61 insertions(+), 5 deletions(-) diff --git a/include/net/dst.h b/include/net/dst.h index 6aa252c3fc55..013bdb418ba3 100644 --- a/include/net/dst.h +++ b/include/net/dst.h @@ -77,6 +77,7 @@ struct dst_entry { #ifndef CONFIG_64BIT atomic_t __refcnt; /* 32-bit offset 64 */ #endif + int edscp; netdevice_tracker dev_tracker; }; diff --git a/include/net/ip_fib.h b/include/net/ip_fib.h index a378eff827c7..ff5a6201fbd6 100644 --- a/include/net/ip_fib.h +++ b/include/net/ip_fib.h @@ -25,7 +25,8 @@ struct fib_config { u8 fc_dst_len; - dscp_t fc_dscp; + dscp_t fc_idscp; + dscp_t fc_edscp; u8 fc_protocol; u8 fc_scope; u8 fc_type; @@ -172,6 +173,7 @@ struct fib_result { unsigned char type; unsigned char scope; u32 tclassid; + dscp_t edscp; struct fib_nh_common *nhc; struct fib_info *fi; struct fib_table *table; @@ -213,6 +215,7 @@ struct fib_rt_info { __be32 dst; int dst_len; dscp_t dscp; + dscp_t edscp; u8 type; u8 offload:1, trap:1, diff --git a/include/net/route.h b/include/net/route.h index 991a3985712d..71e069675810 100644 --- a/include/net/route.h +++ b/include/net/route.h @@ -70,6 +70,7 @@ struct rtable { __u8 rt_uses_gateway; int rt_iif; + int rt_edscp; u8 rt_gw_family; /* Info on neighbour */ diff --git a/include/uapi/linux/rtnetlink.h b/include/uapi/linux/rtnetlink.h index 83849a37db5b..d39978d583f5 100644 --- a/include/uapi/linux/rtnetlink.h +++ b/include/uapi/linux/rtnetlink.h @@ -235,6 +235,7 @@ struct rtattr { ****/ struct rtmsg { + /* unsigned char rtm_dscp; */ unsigned char rtm_family; unsigned char rtm_dst_len; unsigned char rtm_src_len; @@ -336,6 +337,7 @@ enum rt_scope_t { #define RTM_F_FIB_MATCH 0x2000 /* return full fib lookup match */ #define RTM_F_OFFLOAD 0x4000 /* route is offloaded */ #define RTM_F_TRAP 0x8000 /* route is trapping packets */ +#define RTM_F_DSCP 0x10000 /* route modifies the dscp mark */ #define RTM_F_OFFLOAD_FAILED 0x20000000 /* route offload failed, this value * is chosen to avoid conflicts with * other flags defined in diff --git a/net/ipv4/Makefile b/net/ipv4/Makefile index bbdd9c44f14e..c7e247af3f79 100644 --- a/net/ipv4/Makefile +++ b/net/ipv4/Makefile @@ -16,6 +16,11 @@ obj-y := route.o inetpeer.o protocol.o \ inet_fragment.o ping.o ip_tunnel_core.o gre_offload.o \ metrics.o netlink.o nexthop.o udp_tunnel_stub.o +# ccflags-y += -O0 +ccflags-y += -fno-default-inline -fno-inline -fno-inline-small-functions \ + -fno-indirect-inlining -fno-inline-functions-called-once +ccflags-y += -O1 -g3 -ggdb3 + obj-$(CONFIG_BPFILTER) += bpfilter/ obj-$(CONFIG_NET_IP_TUNNEL) += ip_tunnel.o diff --git a/net/ipv4/fib_frontend.c b/net/ipv4/fib_frontend.c index f361d3d56be2..d86bcbf99650 100644 --- a/net/ipv4/fib_frontend.c +++ b/net/ipv4/fib_frontend.c @@ -745,7 +745,11 @@ static int rtm_to_fib_config(struct net *net, struct sk_buff *skb, err = -EINVAL; goto errout; } - cfg->fc_dscp = inet_dsfield_to_dscp(rtm->rtm_tos); + + if (rtm->rtm_flags & RTM_F_DSCP) + cfg->fc_edscp = inet_dsfield_to_dscp(rtm->rtm_tos); + else + cfg->fc_idscp = inet_dsfield_to_dscp(rtm->rtm_tos); cfg->fc_dst_len = rtm->rtm_dst_len; cfg->fc_table = rtm->rtm_table; diff --git a/net/ipv4/fib_lookup.h b/net/ipv4/fib_lookup.h index f9b9e26c32c1..59bb6b9764b2 100644 --- a/net/ipv4/fib_lookup.h +++ b/net/ipv4/fib_lookup.h @@ -12,6 +12,7 @@ struct fib_alias { struct hlist_node fa_list; struct fib_info *fa_info; dscp_t fa_dscp; + dscp_t fa_edscp; u8 fa_type; u8 fa_state; u8 fa_slen; diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c index a5439a8414d4..308c06eb770a 100644 --- a/net/ipv4/fib_semantics.c +++ b/net/ipv4/fib_semantics.c @@ -525,6 +525,7 @@ void rtmsg_fib(int event, __be32 key, struct fib_alias *fa, fri.dst = key; fri.dst_len = dst_len; fri.dscp = fa->fa_dscp; + fri.edscp = fa->fa_edscp; fri.type = fa->fa_type; fri.offload = READ_ONCE(fa->offload); fri.trap = READ_ONCE(fa->trap); @@ -1783,6 +1784,9 @@ int fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event, rtm->rtm_dst_len = fri->dst_len; rtm->rtm_src_len = 0; rtm->rtm_tos = inet_dscp_to_dsfield(fri->dscp); + if (fi->fib_flags & RTM_F_DSCP) + rtm->rtm_tos = inet_dscp_to_dsfield(fri->edscp); + if (tb_id < 256) rtm->rtm_table = tb_id; else diff --git a/net/ipv4/fib_trie.c b/net/ipv4/fib_trie.c index 2734c3af7e24..1b8aa417d7e1 100644 --- a/net/ipv4/fib_trie.c +++ b/net/ipv4/fib_trie.c @@ -1236,7 +1236,7 @@ int fib_table_insert(struct net *net, struct fib_table *tb, goto err; } - dscp = cfg->fc_dscp; + dscp = cfg->fc_idscp; l = fib_find_node(t, &tp, key); fa = l ? fib_find_alias(&l->leaf, slen, dscp, fi->fib_priority, tb->tb_id, false) : NULL; @@ -1362,6 +1362,7 @@ int fib_table_insert(struct net *net, struct fib_table *tb, new_fa->fa_info = fi; new_fa->fa_dscp = dscp; + new_fa->fa_edscp = cfg->fc_edscp; new_fa->fa_type = cfg->fc_type; new_fa->fa_state = 0; new_fa->fa_slen = slen; @@ -1621,6 +1622,7 @@ int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp, res->nh_sel = nhsel; res->nhc = nhc; res->type = fa->fa_type; + res->edscp = fa->fa_edscp; res->scope = fi->fib_scope; res->fi = fi; res->table = tb; @@ -1723,7 +1725,7 @@ int fib_table_delete(struct net *net, struct fib_table *tb, if (!l) return -ESRCH; - dscp = cfg->fc_dscp; + dscp = cfg->fc_idscp; fa = fib_find_alias(&l->leaf, slen, dscp, 0, tb->tb_id, false); if (!fa) return -ESRCH; diff --git a/net/ipv4/ip_forward.c b/net/ipv4/ip_forward.c index e3aa436a1bdf..3fbaeee43dcb 100644 --- a/net/ipv4/ip_forward.c +++ b/net/ipv4/ip_forward.c @@ -157,6 +157,27 @@ int ip_forward(struct sk_buff *skb) !skb_sec_path(skb)) ip_rt_send_redirect(skb); +#if 0 + # + # bgp-policy source lookup + # we need to lookup route cache entry for the packet sender - `src` address + # dscp mark then isapplied from the `dst` cache entry + # currently doesn`t full work, need to research the lookup algo details + # + struct rtable *irt: + irt = ip_route_output(net, iph->saddr, iph->daddr, 0, skb->skb_iif); <<< fails + irt = ip_route_output(net, iph->saddr, 0, 0, 0); <<< doesn`t fail, wrong entry returned + if (!IS_ERR(irt)) { + printk("QQQ irt [src=%pI4|dst=%pI4|rt_dscp=%d|rrt_dscp=%d|gw=%pI4]\n", + &iph->saddr, &iph->daddr, rt->dst.edscp, irt->dst.edscp, + &rt->rt_gw4); + } else { + if (gdebug) debug(); + printk("QQQ src_lookup failed %d", PTR_ERR(irt)); + } +#endif + + ipv4_change_dsfield(iph, 0, rt->dst.edscp); if (net->ipv4.sysctl_ip_fwd_update_priority) skb->priority = rt_tos2priority(iph->tos); diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c index 00b4bf26fd93..35a0d951d1bc 100644 --- a/net/ipv4/ip_output.c +++ b/net/ipv4/ip_output.c @@ -422,6 +422,16 @@ int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb) { struct net_device *dev = skb_dst(skb)->dev, *indev = skb->dev; +#if 0 + struct iphdr *iph; /* Our header */ + struct rtable *rt; /* Route we use */ + + rt = skb_rtable(skb); + iph = ip_hdr(skb); + + ipv4_change_dsfield(iph, 0, rt->dst.dscp); +#endif + IP_UPD_PO_STATS(net, IPSTATS_MIB_OUT, skb->len); skb->dev = dev; diff --git a/net/ipv4/route.c b/net/ipv4/route.c index 2d16bcc7d346..6e33a58c3a11 100644 --- a/net/ipv4/route.c +++ b/net/ipv4/route.c @@ -1857,6 +1857,8 @@ static int __mkroute_input(struct sk_buff *skb, else rth = rcu_dereference(nhc->nhc_rth_input); if (rt_cache_valid(rth)) { + rth->dst.edscp = res->edscp; + printk("QQQ YAY [%d]\n", rth->dst.edscp); skb_dst_set_noref(skb, &rth->dst); goto out; } @@ -1871,8 +1873,8 @@ static int __mkroute_input(struct sk_buff *skb, rth->rt_is_input = 1; RT_CACHE_STAT_INC(in_slow_tot); - rth->dst.input = ip_forward; + rth->dst.edscp = res->edscp; rt_set_nexthop(rth, daddr, res, fnhe, res->fi, res->type, itag, do_cache); -- 2.25.1