Connectivity/bpf_progs/clatd.c

/*
 * Copyright (C) 2019 The Android Open Source Project
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include <linux/bpf.h>
#include <linux/if.h>
#include <linux/if_ether.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <linux/ip.h>
#include <linux/ipv6.h>
#include <linux/pkt_cls.h>
#include <linux/swab.h>
#include <stdbool.h>
#include <stdint.h>

// bionic kernel uapi linux/udp.h header is munged...
#define __kernel_udphdr udphdr
#include <linux/udp.h>

// The resulting .o needs to load on the Android T beta 3 bpfloader
#define BPFLOADER_MIN_VER BPFLOADER_T_BETA3_VERSION

#include "bpf_helpers.h"
#include "bpf_net_helpers.h"
#include "clatd.h"
#include "clat_mark.h"

// IP flags. (from kernel's include/net/ip.h)
#define IP_CE      0x8000  // Flag: "Congestion" (really reserved 'evil bit')
#define IP_DF      0x4000  // Flag: "Don't Fragment"
#define IP_MF      0x2000  // Flag: "More Fragments"
#define IP_OFFSET  0x1FFF  // "Fragment Offset" part

// from kernel's include/net/ipv6.h
struct frag_hdr {
    __u8   nexthdr;
    __u8   reserved;        // always zero
    __be16 frag_off;        // 13 bit offset, 2 bits zero, 1 bit "More Fragments"
    __be32 identification;
};

DEFINE_BPF_MAP_GRW(clat_ingress6_map, HASH, ClatIngress6Key, ClatIngress6Value, 16, AID_SYSTEM)

static inline __always_inline int nat64(struct __sk_buff* skb,
                                        const bool is_ethernet,
                                        const unsigned kver) {
    // Require ethernet dst mac address to be our unicast address.
    if (is_ethernet && (skb->pkt_type != PACKET_HOST)) return TC_ACT_PIPE;

    // Must be meta-ethernet IPv6 frame
    if (skb->protocol != htons(ETH_P_IPV6)) return TC_ACT_PIPE;

    const int l2_header_size = is_ethernet ? sizeof(struct ethhdr) : 0;

    // Not clear if this is actually necessary considering we use DPA (Direct Packet Access),
    // but we need to make sure we can read the IPv6 header reliably so that we can set
    // skb->mark = 0xDeadC1a7 for packets we fail to offload.
    try_make_writable(skb, l2_header_size + sizeof(struct ipv6hdr));

    void* data = (void*)(long)skb->data;
    const void* data_end = (void*)(long)skb->data_end;
    const struct ethhdr* const eth = is_ethernet ? data : NULL;  // used iff is_ethernet
    const struct ipv6hdr* const ip6 = is_ethernet ? (void*)(eth + 1) : data;

    // Must have (ethernet and) ipv6 header
    if (data + l2_header_size + sizeof(*ip6) > data_end) return TC_ACT_PIPE;

    // Ethertype - if present - must be IPv6
    if (is_ethernet && (eth->h_proto != htons(ETH_P_IPV6))) return TC_ACT_PIPE;

    // IP version must be 6
    if (ip6->version != 6) return TC_ACT_PIPE;

    // Maximum IPv6 payload length that can be translated to IPv4
    // Note: technically this check is too strict for an IPv6 fragment,
    // which by virtue of stripping the extra 8 byte fragment extension header,
    // could thus be 8 bytes larger and still fit in an ipv4 packet post
    // translation.  However... who ever heard of receiving ~64KB frags...
    // fragments are kind of by definition smaller than ingress device mtu,
    // and thus, on the internet, very very unlikely to exceed 1500 bytes.
    if (ntohs(ip6->payload_len) > 0xFFFF - sizeof(struct iphdr)) return TC_ACT_PIPE;

    ClatIngress6Key k = {
            .iif = skb->ifindex,
            .pfx96.in6_u.u6_addr32 =
                    {
                            ip6->saddr.in6_u.u6_addr32[0],
                            ip6->saddr.in6_u.u6_addr32[1],
                            ip6->saddr.in6_u.u6_addr32[2],
                    },
            .local6 = ip6->daddr,
    };

    ClatIngress6Value* v = bpf_clat_ingress6_map_lookup_elem(&k);

    if (!v) return TC_ACT_PIPE;

    __u8 proto = ip6->nexthdr;
    __be16 ip_id = 0;
    __be16 frag_off = htons(IP_DF);
    __u16 tot_len = ntohs(ip6->payload_len) + sizeof(struct iphdr);  // cannot overflow, see above

    if (proto == IPPROTO_FRAGMENT) {
        // Fragment handling requires bpf_skb_adjust_room which is 4.14+
        if (kver < KVER_4_14) return TC_ACT_PIPE;

        // Must have (ethernet and) ipv6 header and ipv6 fragment extension header
        if (data + l2_header_size + sizeof(*ip6) + sizeof(struct frag_hdr) > data_end)
            return TC_ACT_PIPE;
        const struct frag_hdr *frag = (const struct frag_hdr *)(ip6 + 1);
        proto = frag->nexthdr;
        // RFC6145: use bottom 16-bits of network endian 32-bit IPv6 ID field for 16-bit IPv4 field.
        // this is equivalent to: ip_id = htons(ntohl(frag->identification));
        ip_id = frag->identification >> 16;
        // Conversion of 16-bit IPv6 frag offset to 16-bit IPv4 frag offset field.
        // IPv6 is '13 bits of offset in multiples of 8' + 2 zero bits + more fragment bit
        // IPv4 is zero bit + don't frag bit + more frag bit + '13 bits of offset in multiples of 8'
        frag_off = ntohs(frag->frag_off);
        frag_off = ((frag_off & 1) << 13) | (frag_off >> 3);
        frag_off = htons(frag_off);
        // Note that by construction tot_len is guaranteed to not underflow here
        tot_len -= sizeof(struct frag_hdr);
        // This is a badly formed IPv6 packet with less payload than the size of an IPv6 Frag EH
        if (tot_len < sizeof(struct iphdr)) return TC_ACT_PIPE;
    }

    switch (proto) {
        case IPPROTO_TCP:  // For TCP & UDP the checksum neutrality of the chosen IPv6
        case IPPROTO_UDP:  // address means there is no need to update their checksums.
        case IPPROTO_GRE:  // We do not need to bother looking at GRE/ESP headers,
        case IPPROTO_ESP:  // since there is never a checksum to update.
            break;

        default:  // do not know how to handle anything else
            // Mark ingress non-offloaded clat packet for dropping in ip6tables bw_raw_PREROUTING.
            // Non-offloaded clat packet is going to be handled by clat daemon and ip6tables. The
            // duplicate one in ip6tables is not necessary.
            skb->mark = CLAT_MARK;
            return TC_ACT_PIPE;
    }

    struct ethhdr eth2;  // used iff is_ethernet
    if (is_ethernet) {
        eth2 = *eth;                     // Copy over the ethernet header (src/dst mac)
        eth2.h_proto = htons(ETH_P_IP);  // But replace the ethertype
    }

    struct iphdr ip = {
            .version = 4,                                                      // u4
            .ihl = sizeof(struct iphdr) / sizeof(__u32),                       // u4
            .tos = (ip6->priority << 4) + (ip6->flow_lbl[0] >> 4),             // u8
            .tot_len = htons(tot_len),                                         // be16
            .id = ip_id,                                                       // be16
            .frag_off = frag_off,                                              // be16
            .ttl = ip6->hop_limit,                                             // u8
            .protocol = proto,                                                 // u8
            .check = 0,                                                        // u16
            .saddr = ip6->saddr.in6_u.u6_addr32[3],                            // be32
            .daddr = v->local4.s_addr,                                         // be32
    };

    // Calculate the IPv4 one's complement checksum of the IPv4 header.
    __wsum sum4 = 0;
    for (int i = 0; i < sizeof(ip) / sizeof(__u16); ++i) {
        sum4 += ((__u16*)&ip)[i];
    }
    // Note that sum4 is guaranteed to be non-zero by virtue of ip.version == 4
    sum4 = (sum4 & 0xFFFF) + (sum4 >> 16);  // collapse u32 into range 1 .. 0x1FFFE
    sum4 = (sum4 & 0xFFFF) + (sum4 >> 16);  // collapse any potential carry into u16
    ip.check = (__u16)~sum4;                // sum4 cannot be zero, so this is never 0xFFFF

    // Calculate the *negative* IPv6 16-bit one's complement checksum of the IPv6 header.
    __wsum sum6 = 0;
    // We'll end up with a non-zero sum due to ip6->version == 6 (which has '0' bits)
    for (int i = 0; i < sizeof(*ip6) / sizeof(__u16); ++i) {
        sum6 += ~((__u16*)ip6)[i];  // note the bitwise negation
    }

    // Note that there is no L4 checksum update: we are relying on the checksum neutrality
    // of the ipv6 address chosen by netd's ClatdController.

    // Packet mutations begin - point of no return, but if this first modification fails
    // the packet is probably still pristine, so let clatd handle it.
    if (bpf_skb_change_proto(skb, htons(ETH_P_IP), 0)) {
        // Mark ingress non-offloaded clat packet for dropping in ip6tables bw_raw_PREROUTING.
        // Non-offloaded clat packet is going to be handled by clat daemon and ip6tables. The
        // duplicate one in ip6tables is not necessary.
        skb->mark = CLAT_MARK;
        return TC_ACT_PIPE;
    }

    // This takes care of updating the skb->csum field for a CHECKSUM_COMPLETE packet.
    //
    // In such a case, skb->csum is a 16-bit one's complement sum of the entire payload,
    // thus we need to subtract out the ipv6 header's sum, and add in the ipv4 header's sum.
    // However, by construction of ip.check above the checksum of an ipv4 header is zero.
    // Thus we only need to subtract the ipv6 header's sum, which is the same as adding
    // in the sum of the bitwise negation of the ipv6 header.
    //
    // bpf_csum_update() always succeeds if the skb is CHECKSUM_COMPLETE and returns an error
    // (-ENOTSUPP) if it isn't.  So we just ignore the return code.
    //
    // if (skb->ip_summed == CHECKSUM_COMPLETE)
    //   return (skb->csum = csum_add(skb->csum, csum));
    // else
    //   return -ENOTSUPP;
    bpf_csum_update(skb, sum6);

    // Technically 'kver < KVER_4_14' already implies 'frag_off == htons(IP_DF)' due to logic above,
    // thus the initial 'kver >= KVER_4_14' check here is entirely superfluous.
    //
    // However, we *need* the compiler (when compiling the program for 4.9) to entirely
    // optimize out the call to bpf_skb_adjust_room() bpf helper: it's not enough for it to emit
    // an unreachable call to it, it must *not* emit it at all (otherwise the 4.9 kernel's
    // bpf verifier will refuse to load a program with an unknown bpf helper call)
    //
    // This is easiest to achieve by being very explicit in the if clause,
    // better safe than sorry...
    //
    // Note: we currently have no TreeHugger coverage for 4.9-T devices (there are no such
    // Pixel or cuttlefish devices), so likely you won't notice for months if this breaks...
    if (kver >= KVER_4_14 && frag_off != htons(IP_DF)) {
        // If we're converting an IPv6 Fragment, we need to trim off 8 more bytes
        // We're beyond recovery on error here... but hard to imagine how this could fail.
        if (bpf_skb_adjust_room(skb, -(__s32)sizeof(struct frag_hdr), BPF_ADJ_ROOM_NET, /*flags*/0))
            return TC_ACT_SHOT;
    }

    // bpf_skb_change_proto() invalidates all pointers - reload them.
    data = (void*)(long)skb->data;
    data_end = (void*)(long)skb->data_end;

    // I cannot think of any valid way for this error condition to trigger, however I do
    // believe the explicit check is required to keep the in kernel ebpf verifier happy.
    if (data + l2_header_size + sizeof(struct iphdr) > data_end) return TC_ACT_SHOT;

    if (is_ethernet) {
        struct ethhdr* new_eth = data;

        // Copy over the updated ethernet header
        *new_eth = eth2;

        // Copy over the new ipv4 header.
        *(struct iphdr*)(new_eth + 1) = ip;
    } else {
        // Copy over the new ipv4 header without an ethernet header.
        *(struct iphdr*)data = ip;
    }

    // Redirect, possibly back to same interface, so tcpdump sees packet twice.
    if (v->oif) return bpf_redirect(v->oif, BPF_F_INGRESS);

    // Just let it through, tcpdump will not see IPv4 packet.
    return TC_ACT_PIPE;
}

DEFINE_BPF_PROG_KVER("schedcls/ingress6/clat_ether$4_14", AID_ROOT, AID_SYSTEM, sched_cls_ingress6_clat_ether_4_14, KVER_4_14)
(struct __sk_buff* skb) {
    return nat64(skb, ETHER, KVER_4_14);
}

DEFINE_BPF_PROG_KVER_RANGE("schedcls/ingress6/clat_ether$4_9", AID_ROOT, AID_SYSTEM, sched_cls_ingress6_clat_ether_4_9, KVER_NONE, KVER_4_14)
(struct __sk_buff* skb) {
    return nat64(skb, ETHER, KVER_NONE);
}

DEFINE_BPF_PROG_KVER("schedcls/ingress6/clat_rawip$4_14", AID_ROOT, AID_SYSTEM, sched_cls_ingress6_clat_rawip_4_14, KVER_4_14)
(struct __sk_buff* skb) {
    return nat64(skb, RAWIP, KVER_4_14);
}

DEFINE_BPF_PROG_KVER_RANGE("schedcls/ingress6/clat_rawip$4_9", AID_ROOT, AID_SYSTEM, sched_cls_ingress6_clat_rawip_4_9, KVER_NONE, KVER_4_14)
(struct __sk_buff* skb) {
    return nat64(skb, RAWIP, KVER_NONE);
}

DEFINE_BPF_MAP_GRW(clat_egress4_map, HASH, ClatEgress4Key, ClatEgress4Value, 16, AID_SYSTEM)

DEFINE_BPF_PROG("schedcls/egress4/clat_rawip", AID_ROOT, AID_SYSTEM, sched_cls_egress4_clat_rawip)
(struct __sk_buff* skb) {
    // Must be meta-ethernet IPv4 frame
    if (skb->protocol != htons(ETH_P_IP)) return TC_ACT_PIPE;

    // Possibly not needed, but for consistency with nat64 up above
    try_make_writable(skb, sizeof(struct iphdr));

    void* data = (void*)(long)skb->data;
    const void* data_end = (void*)(long)skb->data_end;
    const struct iphdr* const ip4 = data;

    // Must have ipv4 header
    if (data + sizeof(*ip4) > data_end) return TC_ACT_PIPE;

    // IP version must be 4
    if (ip4->version != 4) return TC_ACT_PIPE;

    // We cannot handle IP options, just standard 20 byte == 5 dword minimal IPv4 header
    if (ip4->ihl != 5) return TC_ACT_PIPE;

    // Calculate the IPv4 one's complement checksum of the IPv4 header.
    __wsum sum4 = 0;
    for (int i = 0; i < sizeof(*ip4) / sizeof(__u16); ++i) {
        sum4 += ((__u16*)ip4)[i];
    }
    // Note that sum4 is guaranteed to be non-zero by virtue of ip4->version == 4
    sum4 = (sum4 & 0xFFFF) + (sum4 >> 16);  // collapse u32 into range 1 .. 0x1FFFE
    sum4 = (sum4 & 0xFFFF) + (sum4 >> 16);  // collapse any potential carry into u16
    // for a correct checksum we should get *a* zero, but sum4 must be positive, ie 0xFFFF
    if (sum4 != 0xFFFF) return TC_ACT_PIPE;

    // Minimum IPv4 total length is the size of the header
    if (ntohs(ip4->tot_len) < sizeof(*ip4)) return TC_ACT_PIPE;

    // We are incapable of dealing with IPv4 fragments
    if (ip4->frag_off & ~htons(IP_DF)) return TC_ACT_PIPE;

    switch (ip4->protocol) {
        case IPPROTO_TCP:  // For TCP & UDP the checksum neutrality of the chosen IPv6
        case IPPROTO_GRE:  // address means there is no need to update their checksums.
        case IPPROTO_ESP:  // We do not need to bother looking at GRE/ESP headers,
            break;         // since there is never a checksum to update.

        case IPPROTO_UDP:  // See above comment, but must also have UDP header...
            if (data + sizeof(*ip4) + sizeof(struct udphdr) > data_end) return TC_ACT_PIPE;
            const struct udphdr* uh = (const struct udphdr*)(ip4 + 1);
            // If IPv4/UDP checksum is 0 then fallback to clatd so it can calculate the
            // checksum.  Otherwise the network or more likely the NAT64 gateway might
            // drop the packet because in most cases IPv6/UDP packets with a zero checksum
            // are invalid. See RFC 6935.  TODO: calculate checksum via bpf_csum_diff()
            if (!uh->check) return TC_ACT_PIPE;
            break;

        default:  // do not know how to handle anything else
            return TC_ACT_PIPE;
    }

    ClatEgress4Key k = {
            .iif = skb->ifindex,
            .local4.s_addr = ip4->saddr,
    };

    ClatEgress4Value* v = bpf_clat_egress4_map_lookup_elem(&k);

    if (!v) return TC_ACT_PIPE;

    // Translating without redirecting doesn't make sense.
    if (!v->oif) return TC_ACT_PIPE;

    // This implementation is currently limited to rawip.
    if (v->oifIsEthernet) return TC_ACT_PIPE;

    struct ipv6hdr ip6 = {
            .version = 6,                                    // __u8:4
            .priority = ip4->tos >> 4,                       // __u8:4
            .flow_lbl = {(ip4->tos & 0xF) << 4, 0, 0},       // __u8[3]
            .payload_len = htons(ntohs(ip4->tot_len) - 20),  // __be16
            .nexthdr = ip4->protocol,                        // __u8
            .hop_limit = ip4->ttl,                           // __u8
            .saddr = v->local6,                              // struct in6_addr
            .daddr = v->pfx96,                               // struct in6_addr
    };
    ip6.daddr.in6_u.u6_addr32[3] = ip4->daddr;

    // Calculate the IPv6 16-bit one's complement checksum of the IPv6 header.
    __wsum sum6 = 0;
    // We'll end up with a non-zero sum due to ip6.version == 6
    for (int i = 0; i < sizeof(ip6) / sizeof(__u16); ++i) {
        sum6 += ((__u16*)&ip6)[i];
    }

    // Note that there is no L4 checksum update: we are relying on the checksum neutrality
    // of the ipv6 address chosen by netd's ClatdController.

    // Packet mutations begin - point of no return, but if this first modification fails
    // the packet is probably still pristine, so let clatd handle it.
    if (bpf_skb_change_proto(skb, htons(ETH_P_IPV6), 0)) return TC_ACT_PIPE;

    // This takes care of updating the skb->csum field for a CHECKSUM_COMPLETE packet.
    //
    // In such a case, skb->csum is a 16-bit one's complement sum of the entire payload,
    // thus we need to subtract out the ipv4 header's sum, and add in the ipv6 header's sum.
    // However, we've already verified the ipv4 checksum is correct and thus 0.
    // Thus we only need to add the ipv6 header's sum.
    //
    // bpf_csum_update() always succeeds if the skb is CHECKSUM_COMPLETE and returns an error
    // (-ENOTSUPP) if it isn't.  So we just ignore the return code (see above for more details).
    bpf_csum_update(skb, sum6);

    // bpf_skb_change_proto() invalidates all pointers - reload them.
    data = (void*)(long)skb->data;
    data_end = (void*)(long)skb->data_end;

    // I cannot think of any valid way for this error condition to trigger, however I do
    // believe the explicit check is required to keep the in kernel ebpf verifier happy.
    if (data + sizeof(ip6) > data_end) return TC_ACT_SHOT;

    // Copy over the new ipv6 header without an ethernet header.
    *(struct ipv6hdr*)data = ip6;

    // Redirect to non v4-* interface.  Tcpdump only sees packet after this redirect.
    return bpf_redirect(v->oif, 0 /* this is effectively BPF_F_EGRESS */);
}

LICENSE("Apache 2.0");
CRITICAL("Connectivity");