leandrof/linux
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge branch 'bpf: expose bpf_{s,g}etsockopt helpers to bind{4,6} hooks'

Browse Source 
Stanislav Fomichev says:

====================

This might be useful for the listener sockets to pre-populate
some options. Since those helpers require locked sockets,
I'm changing bind hooks to lock/unlock the sockets. This
should not cause any performance overhead because at this
point there shouldn't be any socket lock contention and the
locking/unlocking should be cheap.

Also, as part of the series, I convert test_sock_addr bpf
assembly into C (and preserve the narrow load tests) to
make it easier to extend with th bpf_setsockopt later on.

v2:
* remove version from bpf programs (Andrii Nakryiko)
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This commit is contained in:
Alexei Starovoitov 2020-12-02 13:25:11 -08:00
commit 9e83f54f53
7 changed files with 245 additions and 192 deletions
include/linux
bpf-cgroup.h
net
core
filter.c
ipv4
af_inet.c
ipv6
af_inet6.c
tools/testing/selftests/bpf
progs
bind4_prog.cbind6_prog.c
test_sock_addr.c

12
include/linux/bpf-cgroup.h
View File

@ -246,11 +246,11 @@ int bpf_percpu_cgroup_storage_update(struct bpf_map *map, void *key,
__ret; \
})
#define BPF_CGROUP_RUN_PROG_INET4_BIND(sk, uaddr) \
BPF_CGROUP_RUN_SA_PROG(sk, uaddr, BPF_CGROUP_INET4_BIND)
#define BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr) \
BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_INET4_BIND, NULL)
#define BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr) \
BPF_CGROUP_RUN_SA_PROG(sk, uaddr, BPF_CGROUP_INET6_BIND)
#define BPF_CGROUP_RUN_PROG_INET6_BIND_LOCK(sk, uaddr) \
BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_INET6_BIND, NULL)
#define BPF_CGROUP_PRE_CONNECT_ENABLED(sk) (cgroup_bpf_enabled && \
sk->sk_prot->pre_connect)
@ -434,8 +434,8 @@ static inline int bpf_percpu_cgroup_storage_update(struct bpf_map *map,
#define BPF_CGROUP_RUN_PROG_INET_EGRESS(sk,skb) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET_SOCK(sk) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET4_BIND(sk, uaddr) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET6_BIND_LOCK(sk, uaddr) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk) ({ 0; })
#define BPF_CGROUP_RUN_PROG_INET4_CONNECT(sk, uaddr) ({ 0; })

4
net/core/filter.c
View File

@ -6995,6 +6995,8 @@ sock_addr_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
return &bpf_sk_storage_delete_proto;
case BPF_FUNC_setsockopt:
switch (prog->expected_attach_type) {
case BPF_CGROUP_INET4_BIND:
case BPF_CGROUP_INET6_BIND:
case BPF_CGROUP_INET4_CONNECT:
case BPF_CGROUP_INET6_CONNECT:
return &bpf_sock_addr_setsockopt_proto;
@ -7003,6 +7005,8 @@ sock_addr_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
}
case BPF_FUNC_getsockopt:
switch (prog->expected_attach_type) {
case BPF_CGROUP_INET4_BIND:
case BPF_CGROUP_INET6_BIND:
case BPF_CGROUP_INET4_CONNECT:
case BPF_CGROUP_INET6_CONNECT:
return &bpf_sock_addr_getsockopt_proto;

2
net/ipv4/af_inet.c
View File

@ -450,7 +450,7 @@ int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
/* BPF prog is run before any checks are done so that if the prog
* changes context in a wrong way it will be caught.
*/
err = BPF_CGROUP_RUN_PROG_INET4_BIND(sk, uaddr);
err = BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr);
if (err)
return err;

2
net/ipv6/af_inet6.c
View File

@ -451,7 +451,7 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
/* BPF prog is run before any checks are done so that if the prog
* changes context in a wrong way it will be caught.
*/
err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
err = BPF_CGROUP_RUN_PROG_INET6_BIND_LOCK(sk, uaddr);
if (err)
return err;

102
tools/testing/selftests/bpf/progs/bind4_prog.c Normal file
View File

@ -0,0 +1,102 @@
// SPDX-License-Identifier: GPL-2.0
#include <string.h>
#include <linux/stddef.h>
#include <linux/bpf.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <sys/socket.h>
#include <netinet/tcp.h>
#include <linux/if.h>
#include <errno.h>
#include <bpf/bpf_helpers.h>
#include <bpf/bpf_endian.h>
#define SERV4_IP 0xc0a801feU /* 192.168.1.254 */
#define SERV4_PORT 4040
#define SERV4_REWRITE_IP 0x7f000001U /* 127.0.0.1 */
#define SERV4_REWRITE_PORT 4444
#ifndef IFNAMSIZ
#define IFNAMSIZ 16
#endif
static __inline int bind_to_device(struct bpf_sock_addr *ctx)
{
char veth1[IFNAMSIZ] = "test_sock_addr1";
char veth2[IFNAMSIZ] = "test_sock_addr2";
char missing[IFNAMSIZ] = "nonexistent_dev";
char del_bind[IFNAMSIZ] = "";
if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
&veth1, sizeof(veth1)))
return 1;
if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
&veth2, sizeof(veth2)))
return 1;
if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
&missing, sizeof(missing)) != -ENODEV)
return 1;
if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
&del_bind, sizeof(del_bind)))
return 1;
return 0;
}
SEC("cgroup/bind4")
int bind_v4_prog(struct bpf_sock_addr *ctx)
{
struct bpf_sock *sk;
__u32 user_ip4;
__u16 user_port;
sk = ctx->sk;
if (!sk)
return 0;
if (sk->family != AF_INET)
return 0;
if (ctx->type != SOCK_STREAM && ctx->type != SOCK_DGRAM)
return 0;
if (ctx->user_ip4 != bpf_htonl(SERV4_IP) ||
ctx->user_port != bpf_htons(SERV4_PORT))
return 0;
// u8 narrow loads:
user_ip4 = 0;
user_ip4 |= ((volatile __u8 *)&ctx->user_ip4)[0] << 0;
user_ip4 |= ((volatile __u8 *)&ctx->user_ip4)[1] << 8;
user_ip4 |= ((volatile __u8 *)&ctx->user_ip4)[2] << 16;
user_ip4 |= ((volatile __u8 *)&ctx->user_ip4)[3] << 24;
if (ctx->user_ip4 != user_ip4)
return 0;
user_port = 0;
user_port |= ((volatile __u8 *)&ctx->user_port)[0] << 0;
user_port |= ((volatile __u8 *)&ctx->user_port)[1] << 8;
if (ctx->user_port != user_port)
return 0;
// u16 narrow loads:
user_ip4 = 0;
user_ip4 |= ((volatile __u16 *)&ctx->user_ip4)[0] << 0;
user_ip4 |= ((volatile __u16 *)&ctx->user_ip4)[1] << 16;
if (ctx->user_ip4 != user_ip4)
return 0;
/* Bind to device and unbind it. */
if (bind_to_device(ctx))
return 0;
ctx->user_ip4 = bpf_htonl(SERV4_REWRITE_IP);
ctx->user_port = bpf_htons(SERV4_REWRITE_PORT);
return 1;
}
char _license[] SEC("license") = "GPL";

119
tools/testing/selftests/bpf/progs/bind6_prog.c Normal file
View File

@ -0,0 +1,119 @@
// SPDX-License-Identifier: GPL-2.0
#include <string.h>
#include <linux/stddef.h>
#include <linux/bpf.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <sys/socket.h>
#include <netinet/tcp.h>
#include <linux/if.h>
#include <errno.h>
#include <bpf/bpf_helpers.h>
#include <bpf/bpf_endian.h>
#define SERV6_IP_0 0xfaceb00c /* face:b00c:1234:5678::abcd */
#define SERV6_IP_1 0x12345678
#define SERV6_IP_2 0x00000000
#define SERV6_IP_3 0x0000abcd
#define SERV6_PORT 6060
#define SERV6_REWRITE_IP_0 0x00000000
#define SERV6_REWRITE_IP_1 0x00000000
#define SERV6_REWRITE_IP_2 0x00000000
#define SERV6_REWRITE_IP_3 0x00000001
#define SERV6_REWRITE_PORT 6666
#ifndef IFNAMSIZ
#define IFNAMSIZ 16
#endif
static __inline int bind_to_device(struct bpf_sock_addr *ctx)
{
char veth1[IFNAMSIZ] = "test_sock_addr1";
char veth2[IFNAMSIZ] = "test_sock_addr2";
char missing[IFNAMSIZ] = "nonexistent_dev";
char del_bind[IFNAMSIZ] = "";
if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
&veth1, sizeof(veth1)))
return 1;
if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
&veth2, sizeof(veth2)))
return 1;
if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
&missing, sizeof(missing)) != -ENODEV)
return 1;
if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
&del_bind, sizeof(del_bind)))
return 1;
return 0;
}
SEC("cgroup/bind6")
int bind_v6_prog(struct bpf_sock_addr *ctx)
{
struct bpf_sock *sk;
__u32 user_ip6;
__u16 user_port;
int i;
sk = ctx->sk;
if (!sk)
return 0;
if (sk->family != AF_INET6)
return 0;
if (ctx->type != SOCK_STREAM && ctx->type != SOCK_DGRAM)
return 0;
if (ctx->user_ip6[0] != bpf_htonl(SERV6_IP_0) ||
ctx->user_ip6[1] != bpf_htonl(SERV6_IP_1) ||
ctx->user_ip6[2] != bpf_htonl(SERV6_IP_2) ||
ctx->user_ip6[3] != bpf_htonl(SERV6_IP_3) ||
ctx->user_port != bpf_htons(SERV6_PORT))
return 0;
// u8 narrow loads:
for (i = 0; i < 4; i++) {
user_ip6 = 0;
user_ip6 |= ((volatile __u8 *)&ctx->user_ip6[i])[0] << 0;
user_ip6 |= ((volatile __u8 *)&ctx->user_ip6[i])[1] << 8;
user_ip6 |= ((volatile __u8 *)&ctx->user_ip6[i])[2] << 16;
user_ip6 |= ((volatile __u8 *)&ctx->user_ip6[i])[3] << 24;
if (ctx->user_ip6[i] != user_ip6)
return 0;
}
user_port = 0;
user_port |= ((volatile __u8 *)&ctx->user_port)[0] << 0;
user_port |= ((volatile __u8 *)&ctx->user_port)[1] << 8;
if (ctx->user_port != user_port)
return 0;
// u16 narrow loads:
for (i = 0; i < 4; i++) {
user_ip6 = 0;
user_ip6 |= ((volatile __u16 *)&ctx->user_ip6[i])[0] << 0;
user_ip6 |= ((volatile __u16 *)&ctx->user_ip6[i])[1] << 16;
if (ctx->user_ip6[i] != user_ip6)
return 0;
}
/* Bind to device and unbind it. */
if (bind_to_device(ctx))
return 0;
ctx->user_ip6[0] = bpf_htonl(SERV6_REWRITE_IP_0);
ctx->user_ip6[1] = bpf_htonl(SERV6_REWRITE_IP_1);
ctx->user_ip6[2] = bpf_htonl(SERV6_REWRITE_IP_2);
ctx->user_ip6[3] = bpf_htonl(SERV6_REWRITE_IP_3);
ctx->user_port = bpf_htons(SERV6_REWRITE_PORT);
return 1;
}
char _license[] SEC("license") = "GPL";

196
tools/testing/selftests/bpf/test_sock_addr.c
View File

@ -31,6 +31,8 @@
#define CONNECT6_PROG_PATH "./connect6_prog.o"
#define SENDMSG4_PROG_PATH "./sendmsg4_prog.o"
#define SENDMSG6_PROG_PATH "./sendmsg6_prog.o"
#define BIND4_PROG_PATH "./bind4_prog.o"
#define BIND6_PROG_PATH "./bind6_prog.o"
#define SERV4_IP "192.168.1.254"
#define SERV4_REWRITE_IP "127.0.0.1"
@ -660,190 +662,6 @@ static int load_insns(const struct sock_addr_test *test,
return ret;
}
/* [1] These testing programs try to read different context fields, including
* narrow loads of different sizes from user_ip4 and user_ip6, and write to
* those allowed to be overridden.
*
* [2] BPF_LD_IMM64 & BPF_JMP_REG are used below whenever there is a need to
* compare a register with unsigned 32bit integer. BPF_JMP_IMM can't be used
* in such cases since it accepts only _signed_ 32bit integer as IMM
* argument. Also note that BPF_LD_IMM64 contains 2 instructions what matters
* to count jumps properly.
*/
static int bind4_prog_load(const struct sock_addr_test *test)
{
union {
uint8_t u4_addr8[4];
uint16_t u4_addr16[2];
uint32_t u4_addr32;
} ip4, port;
struct sockaddr_in addr4_rw;
if (inet_pton(AF_INET, SERV4_IP, (void *)&ip4) != 1) {
log_err("Invalid IPv4: %s", SERV4_IP);
return -1;
}
port.u4_addr32 = htons(SERV4_PORT);
if (mk_sockaddr(AF_INET, SERV4_REWRITE_IP, SERV4_REWRITE_PORT,
(struct sockaddr *)&addr4_rw, sizeof(addr4_rw)) == -1)
return -1;
/* See [1]. */
struct bpf_insn insns[] = {
BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
/* if (sk.family == AF_INET && */
BPF_LDX_MEM(BPF_W, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, family)),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, AF_INET, 32),
/* (sk.type == SOCK_DGRAM || sk.type == SOCK_STREAM) && */
BPF_LDX_MEM(BPF_W, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, type)),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, SOCK_DGRAM, 1),
BPF_JMP_A(1),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, SOCK_STREAM, 28),
/* 1st_byte_of_user_ip4 == expected && */
BPF_LDX_MEM(BPF_B, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip4)),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, ip4.u4_addr8[0], 26),
/* 2nd_byte_of_user_ip4 == expected && */
BPF_LDX_MEM(BPF_B, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip4) + 1),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, ip4.u4_addr8[1], 24),
/* 3rd_byte_of_user_ip4 == expected && */
BPF_LDX_MEM(BPF_B, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip4) + 2),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, ip4.u4_addr8[2], 22),
/* 4th_byte_of_user_ip4 == expected && */
BPF_LDX_MEM(BPF_B, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip4) + 3),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, ip4.u4_addr8[3], 20),
/* 1st_half_of_user_ip4 == expected && */
BPF_LDX_MEM(BPF_H, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip4)),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, ip4.u4_addr16[0], 18),
/* 2nd_half_of_user_ip4 == expected && */
BPF_LDX_MEM(BPF_H, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip4) + 2),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, ip4.u4_addr16[1], 16),
/* whole_user_ip4 == expected && */
BPF_LDX_MEM(BPF_W, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip4)),
BPF_LD_IMM64(BPF_REG_8, ip4.u4_addr32), /* See [2]. */
BPF_JMP_REG(BPF_JNE, BPF_REG_7, BPF_REG_8, 12),
/* 1st_byte_of_user_port == expected && */
BPF_LDX_MEM(BPF_B, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_port)),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, port.u4_addr8[0], 10),
/* 1st_half_of_user_port == expected && */
BPF_LDX_MEM(BPF_H, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_port)),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, port.u4_addr16[0], 8),
/* user_port == expected) { */
BPF_LDX_MEM(BPF_W, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_port)),
BPF_LD_IMM64(BPF_REG_8, port.u4_addr32), /* See [2]. */
BPF_JMP_REG(BPF_JNE, BPF_REG_7, BPF_REG_8, 4),
/* user_ip4 = addr4_rw.sin_addr */
BPF_MOV32_IMM(BPF_REG_7, addr4_rw.sin_addr.s_addr),
BPF_STX_MEM(BPF_W, BPF_REG_6, BPF_REG_7,
offsetof(struct bpf_sock_addr, user_ip4)),
/* user_port = addr4_rw.sin_port */
BPF_MOV32_IMM(BPF_REG_7, addr4_rw.sin_port),
BPF_STX_MEM(BPF_W, BPF_REG_6, BPF_REG_7,
offsetof(struct bpf_sock_addr, user_port)),
/* } */
/* return 1 */
BPF_MOV64_IMM(BPF_REG_0, 1),
BPF_EXIT_INSN(),
};
return load_insns(test, insns, sizeof(insns) / sizeof(struct bpf_insn));
}
static int bind6_prog_load(const struct sock_addr_test *test)
{
struct sockaddr_in6 addr6_rw;
struct in6_addr ip6;
if (inet_pton(AF_INET6, SERV6_IP, (void *)&ip6) != 1) {
log_err("Invalid IPv6: %s", SERV6_IP);
return -1;
}
if (mk_sockaddr(AF_INET6, SERV6_REWRITE_IP, SERV6_REWRITE_PORT,
(struct sockaddr *)&addr6_rw, sizeof(addr6_rw)) == -1)
return -1;
/* See [1]. */
struct bpf_insn insns[] = {
BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
/* if (sk.family == AF_INET6 && */
BPF_LDX_MEM(BPF_W, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, family)),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, AF_INET6, 18),
/* 5th_byte_of_user_ip6 == expected && */
BPF_LDX_MEM(BPF_B, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip6[1])),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, ip6.s6_addr[4], 16),
/* 3rd_half_of_user_ip6 == expected && */
BPF_LDX_MEM(BPF_H, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip6[1])),
BPF_JMP_IMM(BPF_JNE, BPF_REG_7, ip6.s6_addr16[2], 14),
/* last_word_of_user_ip6 == expected) { */
BPF_LDX_MEM(BPF_W, BPF_REG_7, BPF_REG_6,
offsetof(struct bpf_sock_addr, user_ip6[3])),
BPF_LD_IMM64(BPF_REG_8, ip6.s6_addr32[3]), /* See [2]. */
BPF_JMP_REG(BPF_JNE, BPF_REG_7, BPF_REG_8, 10),
#define STORE_IPV6_WORD(N) \
BPF_MOV32_IMM(BPF_REG_7, addr6_rw.sin6_addr.s6_addr32[N]), \
BPF_STX_MEM(BPF_W, BPF_REG_6, BPF_REG_7, \
offsetof(struct bpf_sock_addr, user_ip6[N]))
/* user_ip6 = addr6_rw.sin6_addr */
STORE_IPV6_WORD(0),
STORE_IPV6_WORD(1),
STORE_IPV6_WORD(2),
STORE_IPV6_WORD(3),
/* user_port = addr6_rw.sin6_port */
BPF_MOV32_IMM(BPF_REG_7, addr6_rw.sin6_port),
BPF_STX_MEM(BPF_W, BPF_REG_6, BPF_REG_7,
offsetof(struct bpf_sock_addr, user_port)),
/* } */
/* return 1 */
BPF_MOV64_IMM(BPF_REG_0, 1),
BPF_EXIT_INSN(),
};
return load_insns(test, insns, sizeof(insns) / sizeof(struct bpf_insn));
}
static int load_path(const struct sock_addr_test *test, const char *path)
{
struct bpf_prog_load_attr attr;
@ -865,6 +683,16 @@ static int load_path(const struct sock_addr_test *test, const char *path)
return prog_fd;
}
static int bind4_prog_load(const struct sock_addr_test *test)
{
return load_path(test, BIND4_PROG_PATH);
}
static int bind6_prog_load(const struct sock_addr_test *test)
{
return load_path(test, BIND6_PROG_PATH);
}
static int connect4_prog_load(const struct sock_addr_test *test)
{
return load_path(test, CONNECT4_PROG_PATH);