Alexei Starovoitov says: ==================== pull-request: bpf-next 2020-07-21 The following pull-request contains BPF updates for your *net-next* tree. We've added 46 non-merge commits during the last 6 day(s) which contain a total of 68 files changed, 4929 insertions(+), 526 deletions(-). The main changes are: 1) Run BPF program on socket lookup, from Jakub. 2) Introduce cpumap, from Lorenzo. 3) s390 JIT fixes, from Ilya. 4) teach riscv JIT to emit compressed insns, from Luke. 5) use build time computed BTF ids in bpf iter, from Yonghong. ==================== Purely independent overlapping changes in both filter.h and xdp.h
Showing the changes to include/uapi/linux (1 of 68 files). Show all
Unified · Split
| @@ -189,6 +189,7 @@ enum bpf_prog_type { · 188 unmodified lines | |||
| 189 | BPF_PROG_TYPE_STRUCT_OPS, | 189 | BPF_PROG_TYPE_STRUCT_OPS, |
| 190 | BPF_PROG_TYPE_EXT, | 190 | BPF_PROG_TYPE_EXT, |
| 191 | BPF_PROG_TYPE_LSM, | 191 | BPF_PROG_TYPE_LSM, |
| 192 | BPF_PROG_TYPE_SK_LOOKUP, | ||
| 192 | }; | 193 | }; |
| 193 | 194 | ||
| 194 | enum bpf_attach_type { | 195 | enum bpf_attach_type { |
| @@ -227,6 +228,8 @@ enum bpf_attach_type { · 32 unmodified lines | |||
| 227 | BPF_CGROUP_INET6_GETSOCKNAME, | 228 | BPF_CGROUP_INET6_GETSOCKNAME, |
| 228 | BPF_XDP_DEVMAP, | 229 | BPF_XDP_DEVMAP, |
| 229 | BPF_CGROUP_INET_SOCK_RELEASE, | 230 | BPF_CGROUP_INET_SOCK_RELEASE, |
| 231 | BPF_XDP_CPUMAP, | ||
| 232 | BPF_SK_LOOKUP, | ||
| 230 | __MAX_BPF_ATTACH_TYPE | 233 | __MAX_BPF_ATTACH_TYPE |
| 231 | }; | 234 | }; |
| 232 | 235 | ||
| @@ -2419,7 +2422,7 @@ union bpf_attr { · 2186 unmodified lines | |||
| 2419 | * Look for an IPv6 socket. | 2422 | * Look for an IPv6 socket. |
| 2420 | * | 2423 | * |
| 2421 | * If the *netns* is a negative signed 32-bit integer, then the | 2424 | * If the *netns* is a negative signed 32-bit integer, then the |
| 2422 | * socket lookup table in the netns associated with the *ctx* will | 2425 | * socket lookup table in the netns associated with the *ctx* |
| 2423 | * will be used. For the TC hooks, this is the netns of the device | 2426 | * will be used. For the TC hooks, this is the netns of the device |
| 2424 | * in the skb. For socket hooks, this is the netns of the socket. | 2427 | * in the skb. For socket hooks, this is the netns of the socket. |
| 2425 | * If *netns* is any other signed 32-bit value greater than or | 2428 | * If *netns* is any other signed 32-bit value greater than or |
| @@ -2456,7 +2459,7 @@ union bpf_attr { · 30 unmodified lines | |||
| 2456 | * Look for an IPv6 socket. | 2459 | * Look for an IPv6 socket. |
| 2457 | * | 2460 | * |
| 2458 | * If the *netns* is a negative signed 32-bit integer, then the | 2461 | * If the *netns* is a negative signed 32-bit integer, then the |
| 2459 | * socket lookup table in the netns associated with the *ctx* will | 2462 | * socket lookup table in the netns associated with the *ctx* |
| 2460 | * will be used. For the TC hooks, this is the netns of the device | 2463 | * will be used. For the TC hooks, this is the netns of the device |
| 2461 | * in the skb. For socket hooks, this is the netns of the socket. | 2464 | * in the skb. For socket hooks, this is the netns of the socket. |
| 2462 | * If *netns* is any other signed 32-bit value greater than or | 2465 | * If *netns* is any other signed 32-bit value greater than or |
| @@ -3068,6 +3071,10 @@ union bpf_attr { · 605 unmodified lines | |||
| 3068 | * | 3071 | * |
| 3069 | * long bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags) | 3072 | * long bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags) |
| 3070 | * Description | 3073 | * Description |
| 3074 | * Helper is overloaded depending on BPF program type. This | ||
| 3075 | * description applies to **BPF_PROG_TYPE_SCHED_CLS** and | ||
| 3076 | * **BPF_PROG_TYPE_SCHED_ACT** programs. | ||
| 3077 | * | ||
| 3071 | * Assign the *sk* to the *skb*. When combined with appropriate | 3078 | * Assign the *sk* to the *skb*. When combined with appropriate |
| 3072 | * routing configuration to receive the packet towards the socket, | 3079 | * routing configuration to receive the packet towards the socket, |
| 3073 | * will cause *skb* to be delivered to the specified socket. | 3080 | * will cause *skb* to be delivered to the specified socket. |
| @@ -3093,6 +3100,56 @@ union bpf_attr { · 19 unmodified lines | |||
| 3093 | * **-ESOCKTNOSUPPORT** if the socket type is not supported | 3100 | * **-ESOCKTNOSUPPORT** if the socket type is not supported |
| 3094 | * (reuseport). | 3101 | * (reuseport). |
| 3095 | * | 3102 | * |
| 3103 | * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) | ||
| 3104 | * Description | ||
| 3105 | * Helper is overloaded depending on BPF program type. This | ||
| 3106 | * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. | ||
| 3107 | * | ||
| 3108 | * Select the *sk* as a result of a socket lookup. | ||
| 3109 | * | ||
| 3110 | * For the operation to succeed passed socket must be compatible | ||
| 3111 | * with the packet description provided by the *ctx* object. | ||
| 3112 | * | ||
| 3113 | * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must | ||
| 3114 | * be an exact match. While IP family (**AF_INET** or | ||
| 3115 | * **AF_INET6**) must be compatible, that is IPv6 sockets | ||
| 3116 | * that are not v6-only can be selected for IPv4 packets. | ||
| 3117 | * | ||
| 3118 | * Only TCP listeners and UDP unconnected sockets can be | ||
| 3119 | * selected. *sk* can also be NULL to reset any previous | ||
| 3120 | * selection. | ||
| 3121 | * | ||
| 3122 | * *flags* argument can combination of following values: | ||
| 3123 | * | ||
| 3124 | * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous | ||
| 3125 | * socket selection, potentially done by a BPF program | ||
| 3126 | * that ran before us. | ||
| 3127 | * | ||
| 3128 | * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip | ||
| 3129 | * load-balancing within reuseport group for the socket | ||
| 3130 | * being selected. | ||
| 3131 | * | ||
| 3132 | * On success *ctx->sk* will point to the selected socket. | ||
| 3133 | * | ||
| 3134 | * Return | ||
| 3135 | * 0 on success, or a negative errno in case of failure. | ||
| 3136 | * | ||
| 3137 | * * **-EAFNOSUPPORT** if socket family (*sk->family*) is | ||
| 3138 | * not compatible with packet family (*ctx->family*). | ||
| 3139 | * | ||
| 3140 | * * **-EEXIST** if socket has been already selected, | ||
| 3141 | * potentially by another program, and | ||
| 3142 | * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. | ||
| 3143 | * | ||
| 3144 | * * **-EINVAL** if unsupported flags were specified. | ||
| 3145 | * | ||
| 3146 | * * **-EPROTOTYPE** if socket L4 protocol | ||
| 3147 | * (*sk->protocol*) doesn't match packet protocol | ||
| 3148 | * (*ctx->protocol*). | ||
| 3149 | * | ||
| 3150 | * * **-ESOCKTNOSUPPORT** if socket is not in allowed | ||
| 3151 | * state (TCP listening or UDP unconnected). | ||
| 3152 | * | ||
| 3096 | * u64 bpf_ktime_get_boot_ns(void) | 3153 | * u64 bpf_ktime_get_boot_ns(void) |
| 3097 | * Description | 3154 | * Description |
| 3098 | * Return the time elapsed since system boot, in nanoseconds. | 3155 | * Return the time elapsed since system boot, in nanoseconds. |
| @@ -3606,6 +3663,12 @@ enum { · 507 unmodified lines | |||
| 3606 | BPF_RINGBUF_HDR_SZ = 8, | 3663 | BPF_RINGBUF_HDR_SZ = 8, |
| 3607 | }; | 3664 | }; |
| 3608 | 3665 | ||
| 3666 | /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ | ||
| 3667 | enum { | ||
| 3668 | BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), | ||
| 3669 | BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), | ||
| 3670 | }; | ||
| 3671 | |||
| 3609 | /* Mode for BPF_FUNC_skb_adjust_room helper. */ | 3672 | /* Mode for BPF_FUNC_skb_adjust_room helper. */ |
| 3610 | enum bpf_adj_room_mode { | 3673 | enum bpf_adj_room_mode { |
| 3611 | BPF_ADJ_ROOM_NET, | 3674 | BPF_ADJ_ROOM_NET, |
| @@ -3849,6 +3912,19 @@ struct bpf_devmap_val { · 237 unmodified lines | |||
| 3849 | } bpf_prog; | 3912 | } bpf_prog; |
| 3850 | }; | 3913 | }; |
| 3851 | 3914 | ||
| 3915 | /* CPUMAP map-value layout | ||
| 3916 | * | ||
| 3917 | * The struct data-layout of map-value is a configuration interface. | ||
| 3918 | * New members can only be added to the end of this structure. | ||
| 3919 | */ | ||
| 3920 | struct bpf_cpumap_val { | ||
| 3921 | __u32 qsize; /* queue size to remote target CPU */ | ||
| 3922 | union { | ||
| 3923 | int fd; /* prog fd on map write */ | ||
| 3924 | __u32 id; /* prog id on map read */ | ||
| 3925 | } bpf_prog; | ||
| 3926 | }; | ||
| 3927 | |||
| 3852 | enum sk_action { | 3928 | enum sk_action { |
| 3853 | SK_DROP = 0, | 3929 | SK_DROP = 0, |
| 3854 | SK_PASS, | 3930 | SK_PASS, |
| @@ -3986,7 +4062,7 @@ struct bpf_link_info { · 131 unmodified lines | |||
| 3986 | 4062 | ||
| 3987 | /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed | 4063 | /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed |
| 3988 | * by user and intended to be used by socket (e.g. to bind to, depends on | 4064 | * by user and intended to be used by socket (e.g. to bind to, depends on |
| 3989 | * attach attach type). | 4065 | * attach type). |
| 3990 | */ | 4066 | */ |
| 3991 | struct bpf_sock_addr { | 4067 | struct bpf_sock_addr { |
| 3992 | __u32 user_family; /* Allows 4-byte read, but no write. */ | 4068 | __u32 user_family; /* Allows 4-byte read, but no write. */ |
| @@ -4335,4 +4411,19 @@ struct bpf_pidns_info { · 342 unmodified lines | |||
| 4335 | __u32 pid; | 4411 | __u32 pid; |
| 4336 | __u32 tgid; | 4412 | __u32 tgid; |
| 4337 | }; | 4413 | }; |
| 4414 | |||
| 4415 | /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ | ||
| 4416 | struct bpf_sk_lookup { | ||
| 4417 | __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ | ||
| 4418 | |||
| 4419 | __u32 family; /* Protocol family (AF_INET, AF_INET6) */ | ||
| 4420 | __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ | ||
| 4421 | __u32 remote_ip4; /* Network byte order */ | ||
| 4422 | __u32 remote_ip6[4]; /* Network byte order */ | ||
| 4423 | __u32 remote_port; /* Network byte order */ | ||
| 4424 | __u32 local_ip4; /* Network byte order */ | ||
| 4425 | __u32 local_ip6[4]; /* Network byte order */ | ||
| 4426 | __u32 local_port; /* Host byte order */ | ||
| 4427 | }; | ||
| 4428 | |||
| 4338 | #endif /* _UAPI__LINUX_BPF_H__ */ | 4429 | #endif /* _UAPI__LINUX_BPF_H__ */ |