From 6a19f07018f96a158826db404a393fd0f604dc6d Mon Sep 17 00:00:00 2001 From: Hechao Li Date: Mon, 10 Jun 2019 17:56:50 -0700 Subject: [PATCH 01/33] bpf: add a new API libbpf_num_possible_cpus() ANBZ: #5530 commit 6446b3155521aab37f5a2b99ac264803ea7dced8 upstream. Adding a new API libbpf_num_possible_cpus() that helps user with per-CPU map operations. Signed-off-by: Hechao Li Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/lib/bpf/libbpf.c | 57 ++++++++++++++++++++++++++++++++++++++++ tools/lib/bpf/libbpf.h | 16 +++++++++++ tools/lib/bpf/libbpf.map | 1 + 3 files changed, 74 insertions(+) diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c index 1589c4b24ff3..3c5e878a623b 100644 --- a/tools/lib/bpf/libbpf.c +++ b/tools/lib/bpf/libbpf.c @@ -3872,3 +3872,60 @@ void bpf_program__bpil_offs_to_addr(struct bpf_prog_info_linear *info_linear) desc->array_offset, addr); } } + +int libbpf_num_possible_cpus(void) +{ + static const char *fcpu = "/sys/devices/system/cpu/possible"; + int len = 0, n = 0, il = 0, ir = 0; + unsigned int start = 0, end = 0; + static int cpus; + char buf[128]; + int error = 0; + int fd = -1; + + if (cpus > 0) + return cpus; + + fd = open(fcpu, O_RDONLY); + if (fd < 0) { + error = errno; + pr_warning("Failed to open file %s: %s\n", + fcpu, strerror(error)); + return -error; + } + len = read(fd, buf, sizeof(buf)); + close(fd); + if (len <= 0) { + error = len ? errno : EINVAL; + pr_warning("Failed to read # of possible cpus from %s: %s\n", + fcpu, strerror(error)); + return -error; + } + if (len == sizeof(buf)) { + pr_warning("File %s size overflow\n", fcpu); + return -EOVERFLOW; + } + buf[len] = '\0'; + + for (ir = 0, cpus = 0; ir <= len; ir++) { + /* Each sub string separated by ',' has format \d+-\d+ or \d+ */ + if (buf[ir] == ',' || buf[ir] == '\0') { + buf[ir] = '\0'; + n = sscanf(&buf[il], "%u-%u", &start, &end); + if (n <= 0) { + pr_warning("Failed to get # CPUs from %s\n", + &buf[il]); + return -EINVAL; + } else if (n == 1) { + end = start; + } + cpus += end - start + 1; + il = ir + 1; + } + } + if (cpus <= 0) { + pr_warning("Invalid #CPUs %d from %s\n", cpus, fcpu); + return -EINVAL; + } + return cpus; +} diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h index 1af0d48178c8..2e594a0fa961 100644 --- a/tools/lib/bpf/libbpf.h +++ b/tools/lib/bpf/libbpf.h @@ -454,6 +454,22 @@ bpf_program__bpil_addr_to_offs(struct bpf_prog_info_linear *info_linear); LIBBPF_API void bpf_program__bpil_offs_to_addr(struct bpf_prog_info_linear *info_linear); +/* + * A helper function to get the number of possible CPUs before looking up + * per-CPU maps. Negative errno is returned on failure. + * + * Example usage: + * + * int ncpus = libbpf_num_possible_cpus(); + * if (ncpus < 0) { + * // error handling + * } + * long values[ncpus]; + * bpf_map_lookup_elem(per_cpu_map_fd, key, values); + * + */ +LIBBPF_API int libbpf_num_possible_cpus(void); + #ifdef __cplusplus } /* extern "C" */ #endif diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map index 46dcda89df21..2c6d835620d2 100644 --- a/tools/lib/bpf/libbpf.map +++ b/tools/lib/bpf/libbpf.map @@ -172,4 +172,5 @@ LIBBPF_0.0.4 { btf_dump__new; btf__parse_elf; bpf_object__load_xattr; + libbpf_num_possible_cpus; } LIBBPF_0.0.3; -- Gitee From fabd0ac768dfd83ba315428f36001a6d5ed47a06 Mon Sep 17 00:00:00 2001 From: Hechao Li Date: Mon, 10 Jun 2019 17:56:51 -0700 Subject: [PATCH 02/33] selftests/bpf: remove bpf_util.h from BPF C progs ANBZ: #5530 commit ebb886078bae29bd60eba3157dc6edae4bbed3a1 upstream. Though currently there is no problem including bpf_util.h in kernel space BPF C programs, in next patch in this stack, I will reuse libbpf_num_possible_cpus() in bpf_util.h thus include libbpf.h in it, which will cause BPF C programs compile error. Therefore I will first remove bpf_util.h from all test BPF programs. This can also make it clear that bpf_util.h is a user-space utility while bpf_helpers.h is a kernel space utility. Signed-off-by: Hechao Li Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/bpf_endian.h | 1 + tools/testing/selftests/bpf/progs/sockmap_parse_prog.c | 1 - tools/testing/selftests/bpf/progs/sockmap_tcp_msg_prog.c | 2 +- tools/testing/selftests/bpf/progs/sockmap_verdict_prog.c | 1 - tools/testing/selftests/bpf/progs/test_sysctl_prog.c | 5 ++++- 5 files changed, 6 insertions(+), 4 deletions(-) diff --git a/tools/testing/selftests/bpf/bpf_endian.h b/tools/testing/selftests/bpf/bpf_endian.h index b25595ea4a78..05f036df8a4c 100644 --- a/tools/testing/selftests/bpf/bpf_endian.h +++ b/tools/testing/selftests/bpf/bpf_endian.h @@ -2,6 +2,7 @@ #ifndef __BPF_ENDIAN__ #define __BPF_ENDIAN__ +#include #include /* LLVM's BPF target selects the endianness of the CPU diff --git a/tools/testing/selftests/bpf/progs/sockmap_parse_prog.c b/tools/testing/selftests/bpf/progs/sockmap_parse_prog.c index 0f92858f6226..b061dfb014da 100644 --- a/tools/testing/selftests/bpf/progs/sockmap_parse_prog.c +++ b/tools/testing/selftests/bpf/progs/sockmap_parse_prog.c @@ -1,6 +1,5 @@ #include #include "bpf_helpers.h" -#include "bpf_util.h" #include "bpf_endian.h" int _version SEC("version") = 1; diff --git a/tools/testing/selftests/bpf/progs/sockmap_tcp_msg_prog.c b/tools/testing/selftests/bpf/progs/sockmap_tcp_msg_prog.c index 12a7b5c82ed6..d8663e95b221 100644 --- a/tools/testing/selftests/bpf/progs/sockmap_tcp_msg_prog.c +++ b/tools/testing/selftests/bpf/progs/sockmap_tcp_msg_prog.c @@ -1,6 +1,6 @@ #include + #include "bpf_helpers.h" -#include "bpf_util.h" #include "bpf_endian.h" int _version SEC("version") = 1; diff --git a/tools/testing/selftests/bpf/progs/sockmap_verdict_prog.c b/tools/testing/selftests/bpf/progs/sockmap_verdict_prog.c index 2ce7634a4012..49c9a6cf6825 100644 --- a/tools/testing/selftests/bpf/progs/sockmap_verdict_prog.c +++ b/tools/testing/selftests/bpf/progs/sockmap_verdict_prog.c @@ -1,6 +1,5 @@ #include #include "bpf_helpers.h" -#include "bpf_util.h" #include "bpf_endian.h" int _version SEC("version") = 1; diff --git a/tools/testing/selftests/bpf/progs/test_sysctl_prog.c b/tools/testing/selftests/bpf/progs/test_sysctl_prog.c index a295cad805d7..5cbbff416998 100644 --- a/tools/testing/selftests/bpf/progs/test_sysctl_prog.c +++ b/tools/testing/selftests/bpf/progs/test_sysctl_prog.c @@ -8,7 +8,6 @@ #include #include "bpf_helpers.h" -#include "bpf_util.h" /* Max supported length of a string with unsigned long in base 10 (pow2 - 1). */ #define MAX_ULONG_STR_LEN 0xF @@ -16,6 +15,10 @@ /* Max supported length of sysctl value string (pow2). */ #define MAX_VALUE_STR_LEN 0x40 +#ifndef ARRAY_SIZE +#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) +#endif + static __always_inline int is_tcp_mem(struct bpf_sysctl *ctx) { char tcp_mem_name[] = "net/ipv4/tcp_mem"; -- Gitee From 4831143f9adf382eb1053fd61d85a161b3c372fe Mon Sep 17 00:00:00 2001 From: Hechao Li Date: Mon, 10 Jun 2019 17:56:52 -0700 Subject: [PATCH 03/33] bpf: use libbpf_num_possible_cpus internally ANBZ: #5530 commit 4c587c196d8237169405cbb2195669dbc00fd72b upstream. Use the newly added bpf_num_possible_cpus() in bpftool and selftests and remove duplicate implementations. Signed-off-by: Hechao Li Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/bpf/bpftool/common.c | 53 +++----------------------- tools/testing/selftests/bpf/bpf_util.h | 37 +++--------------- 2 files changed, 10 insertions(+), 80 deletions(-) diff --git a/tools/bpf/bpftool/common.c b/tools/bpf/bpftool/common.c index cc08743dc609..6a71324be628 100644 --- a/tools/bpf/bpftool/common.c +++ b/tools/bpf/bpftool/common.c @@ -21,6 +21,7 @@ #include #include +#include /* libbpf_num_possible_cpus */ #include "main.h" @@ -443,57 +444,13 @@ unsigned int get_page_size(void) unsigned int get_possible_cpus(void) { - static unsigned int result; - char buf[128]; - long int n; - char *ptr; - int fd; - - if (result) - return result; - - fd = open("/sys/devices/system/cpu/possible", O_RDONLY); - if (fd < 0) { - p_err("can't open sysfs possible cpus"); - exit(-1); - } - - n = read(fd, buf, sizeof(buf)); - if (n < 2) { - p_err("can't read sysfs possible cpus"); - exit(-1); - } - close(fd); + int cpus = libbpf_num_possible_cpus(); - if (n == sizeof(buf)) { - p_err("read sysfs possible cpus overflow"); + if (cpus < 0) { + p_err("Can't get # of possible cpus: %s", strerror(-cpus)); exit(-1); } - - ptr = buf; - n = 0; - while (*ptr && *ptr != '\n') { - unsigned int a, b; - - if (sscanf(ptr, "%u-%u", &a, &b) == 2) { - n += b - a + 1; - - ptr = strchr(ptr, '-') + 1; - } else if (sscanf(ptr, "%u", &a) == 1) { - n++; - } else { - assert(0); - } - - while (isdigit(*ptr)) - ptr++; - if (*ptr == ',') - ptr++; - } - - result = n; - - return result; + return cpus; } static char * diff --git a/tools/testing/selftests/bpf/bpf_util.h b/tools/testing/selftests/bpf/bpf_util.h index a29206ebbd13..ec219f84e041 100644 --- a/tools/testing/selftests/bpf/bpf_util.h +++ b/tools/testing/selftests/bpf/bpf_util.h @@ -6,44 +6,17 @@ #include #include #include +#include /* libbpf_num_possible_cpus */ static inline unsigned int bpf_num_possible_cpus(void) { - static const char *fcpu = "/sys/devices/system/cpu/possible"; - unsigned int start, end, possible_cpus = 0; - char buff[128]; - FILE *fp; - int len, n, i, j = 0; + int possible_cpus = libbpf_num_possible_cpus(); - fp = fopen(fcpu, "r"); - if (!fp) { - printf("Failed to open %s: '%s'!\n", fcpu, strerror(errno)); + if (possible_cpus < 0) { + printf("Failed to get # of possible cpus: '%s'!\n", + strerror(-possible_cpus)); exit(1); } - - if (!fgets(buff, sizeof(buff), fp)) { - printf("Failed to read %s!\n", fcpu); - exit(1); - } - - len = strlen(buff); - for (i = 0; i <= len; i++) { - if (buff[i] == ',' || buff[i] == '\0') { - buff[i] = '\0'; - n = sscanf(&buff[j], "%u-%u", &start, &end); - if (n <= 0) { - printf("Failed to retrieve # possible CPUs!\n"); - exit(1); - } else if (n == 1) { - end = start; - } - possible_cpus += end - start + 1; - j = i + 1; - } - } - - fclose(fp); - return possible_cpus; } -- Gitee From 337bd837b82afd7eef03726ecd6cd0678136baee Mon Sep 17 00:00:00 2001 From: Takshak Chahande Date: Wed, 31 Jul 2019 15:10:55 -0700 Subject: [PATCH 04/33] libbpf : make libbpf_num_possible_cpus function thread safe ANBZ: #5530 commit 56fbc24116f458a0ea48f9f37fe770fd791042d9 upstream. Having static variable `cpus` in libbpf_num_possible_cpus function without guarding it with mutex makes this function thread-unsafe. If multiple threads accessing this function, in the current form; it leads to incrementing the static variable value `cpus` in the multiple of total available CPUs. Used local stack variable to calculate the number of possible CPUs and then updated the static variable using WRITE_ONCE(). Changes since v1: * added stack variable to calculate cpus * serialized static variable update using WRITE_ONCE() * fixed Fixes tag Fixes: 6446b3155521 ("bpf: add a new API libbpf_num_possible_cpus()") Signed-off-by: Takshak Chahande Acked-by: Andrey Ignatov Reviewed-by: Jakub Kicinski Signed-off-by: Alexei Starovoitov Signed-off-by: Qiao Ma --- tools/lib/bpf/libbpf.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c index 3c5e878a623b..4c63a5959fa0 100644 --- a/tools/lib/bpf/libbpf.c +++ b/tools/lib/bpf/libbpf.c @@ -3878,13 +3878,15 @@ int libbpf_num_possible_cpus(void) static const char *fcpu = "/sys/devices/system/cpu/possible"; int len = 0, n = 0, il = 0, ir = 0; unsigned int start = 0, end = 0; + int tmp_cpus = 0; static int cpus; char buf[128]; int error = 0; int fd = -1; - if (cpus > 0) - return cpus; + tmp_cpus = READ_ONCE(cpus); + if (tmp_cpus > 0) + return tmp_cpus; fd = open(fcpu, O_RDONLY); if (fd < 0) { @@ -3907,7 +3909,7 @@ int libbpf_num_possible_cpus(void) } buf[len] = '\0'; - for (ir = 0, cpus = 0; ir <= len; ir++) { + for (ir = 0, tmp_cpus = 0; ir <= len; ir++) { /* Each sub string separated by ',' has format \d+-\d+ or \d+ */ if (buf[ir] == ',' || buf[ir] == '\0') { buf[ir] = '\0'; @@ -3919,13 +3921,15 @@ int libbpf_num_possible_cpus(void) } else if (n == 1) { end = start; } - cpus += end - start + 1; + tmp_cpus += end - start + 1; il = ir + 1; } } - if (cpus <= 0) { - pr_warning("Invalid #CPUs %d from %s\n", cpus, fcpu); + if (tmp_cpus <= 0) { + pr_warning("Invalid #CPUs %d from %s\n", tmp_cpus, fcpu); return -EINVAL; } - return cpus; + + WRITE_ONCE(cpus, tmp_cpus); + return tmp_cpus; } -- Gitee From c154102706823f2cb240bf99779de5ababe1cb8e Mon Sep 17 00:00:00 2001 From: Stanislav Fomichev Date: Wed, 12 Jun 2019 10:30:37 -0700 Subject: [PATCH 05/33] bpf: export bpf_sock for BPF_PROG_TYPE_CGROUP_SOCK_ADDR prog type ANBZ: #5530 commit fb85c4a730af221339c1dde1a434b73da0dfc3ed upstream. And let it use bpf_sk_storage_{get,delete} helpers to access socket storage. Kernel context (struct bpf_sock_addr_kern) already has sk member, so I just expose it to the BPF hooks. Using PTR_TO_SOCKET instead of PTR_TO_SOCK_COMMON should be safe because the hook is called on bind/connect. Cc: Martin Lau Signed-off-by: Stanislav Fomichev Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- include/uapi/linux/bpf.h | 1 + net/core/filter.c | 16 ++++++++++++++++ 2 files changed, 17 insertions(+) diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 8e7658d571b8..9231909bd6b4 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -3231,6 +3231,7 @@ struct bpf_sock_addr { __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. * Stored in network byte order. */ + __bpf_md_ptr(struct bpf_sock *, sk); }; /* User bpf_sock_ops struct to access socket values and specify request ops diff --git a/net/core/filter.c b/net/core/filter.c index d1a84712ec0b..2c3b3d3df1ed 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -5896,6 +5896,10 @@ sock_addr_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) case BPF_FUNC_skc_lookup_tcp: return &bpf_sock_addr_skc_lookup_tcp_proto; #endif /* CONFIG_INET */ + case BPF_FUNC_sk_storage_get: + return &bpf_sk_storage_get_proto; + case BPF_FUNC_sk_storage_delete: + return &bpf_sk_storage_delete_proto; default: return bpf_base_func_proto(func_id); } @@ -6801,6 +6805,13 @@ static bool sock_addr_is_valid_access(int off, int size, if (size != size_default) return false; break; + case offsetof(struct bpf_sock_addr, sk): + if (type != BPF_READ) + return false; + if (size != sizeof(__u64)) + return false; + info->reg_type = PTR_TO_SOCKET; + break; default: if (type == BPF_READ) { if (size != size_default) @@ -7735,6 +7746,11 @@ static u32 sock_addr_convert_ctx_access(enum bpf_access_type type, struct bpf_sock_addr_kern, struct in6_addr, t_ctx, s6_addr32[0], BPF_SIZE(si->code), off, tmp_reg); break; + case offsetof(struct bpf_sock_addr, sk): + *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct bpf_sock_addr_kern, sk), + si->dst_reg, si->src_reg, + offsetof(struct bpf_sock_addr_kern, sk)); + break; } return insn - insn_buf; -- Gitee From 51d12dccc5b6fbe311f958f2b7240009d6c02a8d Mon Sep 17 00:00:00 2001 From: Stanislav Fomichev Date: Wed, 12 Jun 2019 10:30:38 -0700 Subject: [PATCH 06/33] bpf: export bpf_sock for BPF_PROG_TYPE_SOCK_OPS prog type ANBZ: #5530 commit 1314ef561102e534e14cb1d37f89f5c1df0b2ea7 upstream. And let it use bpf_sk_storage_{get,delete} helpers to access socket storage. Kernel context (struct bpf_sock_ops_kern) already has sk member, so I just expose it to the BPF hooks. I use PTR_TO_SOCKET_OR_NULL and return NULL in !is_fullsock case. I also export bpf_tcp_sock to make it possible to access tcp socket stats. Cc: Martin Lau Signed-off-by: Stanislav Fomichev Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- include/uapi/linux/bpf.h | 1 + net/core/filter.c | 26 ++++++++++++++++++++++++++ 2 files changed, 27 insertions(+) diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 9231909bd6b4..ed78ecb3c204 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -3283,6 +3283,7 @@ struct bpf_sock_ops { __u32 sk_txhash; __u64 bytes_received; __u64 bytes_acked; + __bpf_md_ptr(struct bpf_sock *, sk); }; /* Definitions for bpf_sock_ops_cb_flags */ diff --git a/net/core/filter.c b/net/core/filter.c index 2c3b3d3df1ed..bd0b252618ea 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -6117,6 +6117,14 @@ sock_ops_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) return &bpf_get_local_storage_proto; case BPF_FUNC_perf_event_output: return &bpf_sockopt_event_output_proto; + case BPF_FUNC_sk_storage_get: + return &bpf_sk_storage_get_proto; + case BPF_FUNC_sk_storage_delete: + return &bpf_sk_storage_delete_proto; +#ifdef CONFIG_INET + case BPF_FUNC_tcp_sock: + return &bpf_tcp_sock_proto; +#endif /* CONFIG_INET */ default: return bpf_base_func_proto(func_id); } @@ -6855,6 +6863,11 @@ static bool sock_ops_is_valid_access(int off, int size, if (size != sizeof(__u64)) return false; break; + case offsetof(struct bpf_sock_ops, sk): + if (size != sizeof(__u64)) + return false; + info->reg_type = PTR_TO_SOCKET_OR_NULL; + break; default: if (size != size_default) return false; @@ -8032,6 +8045,19 @@ static u32 sock_ops_convert_ctx_access(enum bpf_access_type type, SOCK_OPS_GET_OR_SET_FIELD(sk_txhash, sk_txhash, struct sock, type); break; + case offsetof(struct bpf_sock_ops, sk): + *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF( + struct bpf_sock_ops_kern, + is_fullsock), + si->dst_reg, si->src_reg, + offsetof(struct bpf_sock_ops_kern, + is_fullsock)); + *insn++ = BPF_JMP_IMM(BPF_JEQ, si->dst_reg, 0, 1); + *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF( + struct bpf_sock_ops_kern, sk), + si->dst_reg, si->src_reg, + offsetof(struct bpf_sock_ops_kern, sk)); + break; } return insn - insn_buf; } -- Gitee From 638ad213fb90409bc69b9c7f813bc2baa54e96bc Mon Sep 17 00:00:00 2001 From: Stanislav Fomichev Date: Wed, 12 Jun 2019 10:30:39 -0700 Subject: [PATCH 07/33] bpf/tools: sync bpf.h ANBZ: #5530 commit cd17d77705780e2270937fb3cbd2b985adab3edc upstream. Add sk to struct bpf_sock_addr and struct bpf_sock_ops. Cc: Martin Lau Signed-off-by: Stanislav Fomichev Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/include/uapi/linux/bpf.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index c40b9ad94621..76800fa7d57f 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -3228,6 +3228,7 @@ struct bpf_sock_addr { __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. * Stored in network byte order. */ + __bpf_md_ptr(struct bpf_sock *, sk); }; /* User bpf_sock_ops struct to access socket values and specify request ops @@ -3279,6 +3280,7 @@ struct bpf_sock_ops { __u32 sk_txhash; __u64 bytes_received; __u64 bytes_acked; + __bpf_md_ptr(struct bpf_sock *, sk); }; /* Definitions for bpf_sock_ops_cb_flags */ -- Gitee From 4ea228daa171934a8343b7d5256b9f64e461ddf6 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Tue, 11 Aug 2020 15:04:56 -0700 Subject: [PATCH 08/33] bpf: sock_ops sk access may stomp registers when dst_reg = src_reg ANBZ: #5530 commit 84f44df664e9f0e261157e16ee1acd77cc1bb78d upstream. Similar to patch ("bpf: sock_ops ctx access may stomp registers") if the src_reg = dst_reg when reading the sk field of a sock_ops struct we generate xlated code, 53: (61) r9 = *(u32 *)(r9 +28) 54: (15) if r9 == 0x0 goto pc+3 56: (79) r9 = *(u64 *)(r9 +0) This stomps on the r9 reg to do the sk_fullsock check and then when reading the skops->sk field instead of the sk pointer we get the sk_fullsock. To fix use similar pattern noted in the previous fix and use the temp field to save/restore a register used to do sk_fullsock check. After the fix the generated xlated code reads, 52: (7b) *(u64 *)(r9 +32) = r8 53: (61) r8 = *(u32 *)(r9 +28) 54: (15) if r9 == 0x0 goto pc+3 55: (79) r8 = *(u64 *)(r9 +32) 56: (79) r9 = *(u64 *)(r9 +0) 57: (05) goto pc+1 58: (79) r8 = *(u64 *)(r9 +32) Here r9 register was in-use so r8 is chosen as the temporary register. In line 52 r8 is saved in temp variable and at line 54 restored in case fullsock != 0. Finally we handle fullsock == 0 case by restoring at line 58. This adds a new macro SOCK_OPS_GET_SK it is almost possible to merge this with SOCK_OPS_GET_FIELD, but I found the extra branch logic a bit more confusing than just adding a new macro despite a bit of duplicating code. Fixes: 1314ef561102e ("bpf: export bpf_sock for BPF_PROG_TYPE_SOCK_OPS prog type") Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann Acked-by: Song Liu Acked-by: Martin KaFai Lau Link: https://lore.kernel.org/bpf/159718349653.4728.6559437186853473612.stgit@john-Precision-5820-Tower Signed-off-by: Qiao Ma --- net/core/filter.c | 49 ++++++++++++++++++++++++++++++++++++----------- 1 file changed, 38 insertions(+), 11 deletions(-) diff --git a/net/core/filter.c b/net/core/filter.c index bd0b252618ea..782a1cae2167 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -7822,6 +7822,43 @@ static u32 sock_ops_convert_ctx_access(enum bpf_access_type type, } \ } while (0) +#define SOCK_OPS_GET_SK() \ + do { \ + int fullsock_reg = si->dst_reg, reg = BPF_REG_9, jmp = 1; \ + if (si->dst_reg == reg || si->src_reg == reg) \ + reg--; \ + if (si->dst_reg == reg || si->src_reg == reg) \ + reg--; \ + if (si->dst_reg == si->src_reg) { \ + *insn++ = BPF_STX_MEM(BPF_DW, si->src_reg, reg, \ + offsetof(struct bpf_sock_ops_kern, \ + temp)); \ + fullsock_reg = reg; \ + jmp += 2; \ + } \ + *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF( \ + struct bpf_sock_ops_kern, \ + is_fullsock), \ + fullsock_reg, si->src_reg, \ + offsetof(struct bpf_sock_ops_kern, \ + is_fullsock)); \ + *insn++ = BPF_JMP_IMM(BPF_JEQ, fullsock_reg, 0, jmp); \ + if (si->dst_reg == si->src_reg) \ + *insn++ = BPF_LDX_MEM(BPF_DW, reg, si->src_reg, \ + offsetof(struct bpf_sock_ops_kern, \ + temp)); \ + *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF( \ + struct bpf_sock_ops_kern, sk),\ + si->dst_reg, si->src_reg, \ + offsetof(struct bpf_sock_ops_kern, sk));\ + if (si->dst_reg == si->src_reg) { \ + *insn++ = BPF_JMP_A(1); \ + *insn++ = BPF_LDX_MEM(BPF_DW, reg, si->src_reg, \ + offsetof(struct bpf_sock_ops_kern, \ + temp)); \ + } \ + } while (0) + #define SOCK_OPS_GET_TCP_SOCK_FIELD(FIELD) \ SOCK_OPS_GET_FIELD(FIELD, FIELD, struct tcp_sock) @@ -8046,17 +8083,7 @@ static u32 sock_ops_convert_ctx_access(enum bpf_access_type type, struct sock, type); break; case offsetof(struct bpf_sock_ops, sk): - *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF( - struct bpf_sock_ops_kern, - is_fullsock), - si->dst_reg, si->src_reg, - offsetof(struct bpf_sock_ops_kern, - is_fullsock)); - *insn++ = BPF_JMP_IMM(BPF_JEQ, si->dst_reg, 0, 1); - *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF( - struct bpf_sock_ops_kern, sk), - si->dst_reg, si->src_reg, - offsetof(struct bpf_sock_ops_kern, sk)); + SOCK_OPS_GET_SK(); break; } return insn - insn_buf; -- Gitee From d648183cb9db7f399c4da64cd0d3e97a77f7db47 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Tue, 11 Aug 2020 15:05:14 -0700 Subject: [PATCH 09/33] bpf, selftests: Add tests for ctx access in sock_ops with single register ANBZ: #5530 commit 86ed4be68fdee23df4843a59f91c1ac7fc05e860 upstream. To verify fix ("bpf: sock_ops ctx access may stomp registers in corner case") we want to force compiler to generate the following code when accessing a field with BPF_TCP_SOCK_GET_COMMON, r1 = *(u32 *)(r1 + 96) // r1 is skops ptr Rather than depend on clang to do this we add the test with inline asm to the tcpbpf test. This saves us from having to create another runner and ensures that if we break this again test_tcpbpf will crash. With above code we get the xlated code, 11: (7b) *(u64 *)(r1 +32) = r9 12: (61) r9 = *(u32 *)(r1 +28) 13: (15) if r9 == 0x0 goto pc+4 14: (79) r9 = *(u64 *)(r1 +32) 15: (79) r1 = *(u64 *)(r1 +0) 16: (61) r1 = *(u32 *)(r1 +2348) 17: (05) goto pc+1 18: (79) r9 = *(u64 *)(r1 +32) We also add the normal case where src_reg != dst_reg so we can compare code generation easily from llvm-objdump and ensure that case continues to work correctly. The normal code is xlated to, 20: (b7) r1 = 0 21: (61) r1 = *(u32 *)(r3 +28) 22: (15) if r1 == 0x0 goto pc+2 23: (79) r1 = *(u64 *)(r3 +0) 24: (61) r1 = *(u32 *)(r1 +2348) Where the temp variable is not used. Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann Acked-by: Song Liu Acked-by: Martin KaFai Lau Link: https://lore.kernel.org/bpf/159718351457.4728.3295119261717842496.stgit@john-Precision-5820-Tower Signed-off-by: Qiao Ma --- .../testing/selftests/bpf/progs/test_tcpbpf_kern.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c b/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c index 4b7fd540cea9..06bfb769acd2 100644 --- a/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c +++ b/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c @@ -45,12 +45,25 @@ int _version SEC("version") = 1; SEC("sockops") int bpf_testcb(struct bpf_sock_ops *skops) { + struct bpf_sock_ops *reuse = skops; int rv = -1; int bad_call_rv = 0; int good_call_rv = 0; int op; int v = 0; + /* Test reading fields in bpf_sock_ops using single register */ + asm volatile ( + "%[reuse] = *(u32 *)(%[reuse] +96)" + : [reuse] "+r"(reuse) + :); + + asm volatile ( + "%[op] = *(u32 *)(%[skops] +96)" + : [op] "+r"(op) + : [skops] "r"(skops) + :); + op = (int) skops->op; update_event_map(op); -- Gitee From 872b8784a779f8aed6ce82dd347b56214b4d39de Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Tue, 11 Aug 2020 15:05:33 -0700 Subject: [PATCH 10/33] bpf, selftests: Add tests for sock_ops load with r9, r8.r7 registers ANBZ: #5530 commit 8e0c1517565f06027b68caf2875620ddf6914eee upstream. Loads in sock_ops case when using high registers requires extra logic to ensure the correct temporary value is used. We need to ensure the temp register does not use either the src_reg or dst_reg. Lets add an asm test to force the logic is triggered. The xlated code is here, 30: (7b) *(u64 *)(r9 +32) = r7 31: (61) r7 = *(u32 *)(r9 +28) 32: (15) if r7 == 0x0 goto pc+2 33: (79) r7 = *(u64 *)(r9 +0) 34: (63) *(u32 *)(r7 +916) = r8 35: (79) r7 = *(u64 *)(r9 +32) Notice r9 and r8 are not used for temp registers and r7 is chosen. Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann Acked-by: Song Liu Acked-by: Martin KaFai Lau Link: https://lore.kernel.org/bpf/159718353345.4728.8805043614257933227.stgit@john-Precision-5820-Tower Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c b/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c index 06bfb769acd2..86285183059b 100644 --- a/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c +++ b/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c @@ -64,6 +64,13 @@ int bpf_testcb(struct bpf_sock_ops *skops) : [skops] "r"(skops) :); + asm volatile ( + "r9 = %[skops];\n" + "r8 = *(u32 *)(r9 +164);\n" + "*(u32 *)(r9 +164) = r8;\n" + :: [skops] "r"(skops) + : "r9", "r8"); + op = (int) skops->op; update_event_map(op); -- Gitee From 37b9adb025fb16f8baa1305745e70d46bf96405b Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Tue, 11 Aug 2020 15:05:53 -0700 Subject: [PATCH 11/33] bpf, selftests: Add tests to sock_ops for loading sk ANBZ: #5530 commit 9efa9e499799f939968aff1123cc7e8184960e48 upstream. Add tests to directly accesse sock_ops sk field. Then use it to ensure a bad pointer access will fault if something goes wrong. We do three tests: The first test ensures when we read sock_ops sk pointer into the same register that we don't fault as described earlier. Here r9 is chosen as the temp register. The xlated code is, 36: (7b) *(u64 *)(r1 +32) = r9 37: (61) r9 = *(u32 *)(r1 +28) 38: (15) if r9 == 0x0 goto pc+3 39: (79) r9 = *(u64 *)(r1 +32) 40: (79) r1 = *(u64 *)(r1 +0) 41: (05) goto pc+1 42: (79) r9 = *(u64 *)(r1 +32) The second test ensures the temp register selection does not collide with in-use register r9. Shown here r8 is chosen because r9 is the sock_ops pointer. The xlated code is as follows, 46: (7b) *(u64 *)(r9 +32) = r8 47: (61) r8 = *(u32 *)(r9 +28) 48: (15) if r8 == 0x0 goto pc+3 49: (79) r8 = *(u64 *)(r9 +32) 50: (79) r9 = *(u64 *)(r9 +0) 51: (05) goto pc+1 52: (79) r8 = *(u64 *)(r9 +32) And finally, ensure we didn't break the base case where dst_reg does not equal the source register, 56: (61) r2 = *(u32 *)(r1 +28) 57: (15) if r2 == 0x0 goto pc+1 58: (79) r2 = *(u64 *)(r1 +0) Notice it takes us an extra four instructions when src reg is the same as dst reg. One to save the reg, two to restore depending on the branch taken and a goto to jump over the second restore. Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann Acked-by: Song Liu Acked-by: Martin KaFai Lau Link: https://lore.kernel.org/bpf/159718355325.4728.4163036953345999636.stgit@john-Precision-5820-Tower Signed-off-by: Qiao Ma --- .../selftests/bpf/progs/test_tcpbpf_kern.c | 21 +++++++++++++++++++ 1 file changed, 21 insertions(+) diff --git a/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c b/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c index 86285183059b..236f7649bd9c 100644 --- a/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c +++ b/tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c @@ -71,6 +71,27 @@ int bpf_testcb(struct bpf_sock_ops *skops) :: [skops] "r"(skops) : "r9", "r8"); + asm volatile ( + "r1 = %[skops];\n" + "r1 = *(u64 *)(r1 +184);\n" + "if r1 == 0 goto +1;\n" + "r1 = *(u32 *)(r1 +4);\n" + :: [skops] "r"(skops):"r1"); + + asm volatile ( + "r9 = %[skops];\n" + "r9 = *(u64 *)(r9 +184);\n" + "if r9 == 0 goto +1;\n" + "r9 = *(u32 *)(r9 +4);\n" + :: [skops] "r"(skops):"r9"); + + asm volatile ( + "r1 = %[skops];\n" + "r2 = *(u64 *)(r1 +184);\n" + "if r2 == 0 goto +1;\n" + "r2 = *(u32 *)(r2 +4);\n" + :: [skops] "r"(skops):"r1", "r2"); + op = (int) skops->op; update_event_map(op); -- Gitee From e96fb60d52cf071504a4d00c8656d40550aea21b Mon Sep 17 00:00:00 2001 From: Martynas Pumputis Date: Wed, 12 Jun 2019 18:05:41 +0200 Subject: [PATCH 12/33] bpf: sync BPF_FIB_LOOKUP flag changes with BPF uapi ANBZ: #5530 commit 0e265747491cb182429adfbef7d002164566f25c upstream. Sync the changes to the flags made in "bpf: simplify definition of BPF_FIB_LOOKUP related flags" with the BPF UAPI headers. Doing in a separate commit to ease syncing of github/libbpf. Signed-off-by: Martynas Pumputis Acked-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/include/uapi/linux/bpf.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index 76800fa7d57f..5b9627d981cf 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -3396,8 +3396,8 @@ struct bpf_raw_tracepoint_args { /* DIRECT: Skip the FIB rules and go to FIB table associated with device * OUTPUT: Do lookup from egress perspective; default is ingress */ -#define BPF_FIB_LOOKUP_DIRECT BIT(0) -#define BPF_FIB_LOOKUP_OUTPUT BIT(1) +#define BPF_FIB_LOOKUP_DIRECT (1U << 0) +#define BPF_FIB_LOOKUP_OUTPUT (1U << 1) enum { BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ -- Gitee From 1f18219c18ca98bf1e140848625c1613e50eeab4 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Sat, 15 Jun 2019 12:12:17 -0700 Subject: [PATCH 13/33] bpf: track spill/fill of constants ANBZ: #5530 commit f7cf25b2026dc8441e0fa3a202c2aa8a56211e30 upstream. Compilers often spill induction variables into the stack, hence it is necessary for the verifier to track scalar values of the registers through stack slots. Also few bpf programs were incorrectly rejected in the past, since the verifier was not able to track such constants while they were used to compute offsets into packet headers. Tracking constants through the stack significantly decreases the chances of state pruning, since two different constants are considered to be different by state equivalency. End result that cilium tests suffer serious degradation in the number of states processed and corresponding verification time increase. before after bpf_lb-DLB_L3.o 1838 6441 bpf_lb-DLB_L4.o 3218 5908 bpf_lb-DUNKNOWN.o 1064 1064 bpf_lxc-DDROP_ALL.o 26935 93790 bpf_lxc-DUNKNOWN.o 34439 123886 bpf_netdev.o 9721 31413 bpf_overlay.o 6184 18561 bpf_lxc_jit.o 39389 359445 After further debugging turned out that cillium progs are getting hurt by clang due to the same constant tracking issue. Newer clang generates better code by spilling less to the stack. Instead it keeps more constants in the registers which hurts state pruning since the verifier already tracks constants in the registers: old clang new clang (no spill/fill tracking introduced by this patch) bpf_lb-DLB_L3.o 1838 1923 bpf_lb-DLB_L4.o 3218 3077 bpf_lb-DUNKNOWN.o 1064 1062 bpf_lxc-DDROP_ALL.o 26935 166729 bpf_lxc-DUNKNOWN.o 34439 174607 bpf_netdev.o 9721 8407 bpf_overlay.o 6184 5420 bpf_lcx_jit.o 39389 39389 The final table is depressing: old clang old clang new clang new clang const spill/fill const spill/fill bpf_lb-DLB_L3.o 1838 6441 1923 8128 bpf_lb-DLB_L4.o 3218 5908 3077 6707 bpf_lb-DUNKNOWN.o 1064 1064 1062 1062 bpf_lxc-DDROP_ALL.o 26935 93790 166729 380712 bpf_lxc-DUNKNOWN.o 34439 123886 174607 440652 bpf_netdev.o 9721 31413 8407 31904 bpf_overlay.o 6184 18561 5420 23569 bpf_lxc_jit.o 39389 359445 39389 359445 Tracking constants in the registers hurts state pruning already. Adding tracking of constants through stack hurts pruning even more. The later patch address this general constant tracking issue with coarse/precise logic. Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 92 +++++++++++++++++++++++++++++++------------ 1 file changed, 66 insertions(+), 26 deletions(-) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index d8a9cff087b7..73d5623e88bf 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -1377,6 +1377,23 @@ static bool register_is_null(struct bpf_reg_state *reg) return reg->type == SCALAR_VALUE && tnum_equals_const(reg->var_off, 0); } +static bool register_is_const(struct bpf_reg_state *reg) +{ + return reg->type == SCALAR_VALUE && tnum_is_const(reg->var_off); +} + +static void save_register_state(struct bpf_func_state *state, + int spi, struct bpf_reg_state *reg) +{ + int i; + + state->stack[spi].spilled_ptr = *reg; + state->stack[spi].spilled_ptr.live |= REG_LIVE_WRITTEN; + + for (i = 0; i < BPF_REG_SIZE; i++) + state->stack[spi].slot_type[i] = STACK_SPILL; +} + /* check_stack_read/write functions track spill/fill of registers, * stack boundary and alignment are checked in check_mem_access() */ @@ -1386,7 +1403,7 @@ static int check_stack_write(struct bpf_verifier_env *env, { struct bpf_func_state *cur; /* state of the current function */ int i, slot = -off - 1, spi = slot / BPF_REG_SIZE, err; - enum bpf_reg_type type; + struct bpf_reg_state *reg = NULL; err = realloc_func_state(state, round_up(slot + 1, BPF_REG_SIZE), state->acquired_refs, true); @@ -1403,27 +1420,37 @@ static int check_stack_write(struct bpf_verifier_env *env, } cur = env->cur_state->frame[env->cur_state->curframe]; - if (value_regno >= 0 && - is_spillable_regtype((type = cur->regs[value_regno].type))) { + if (value_regno >= 0) + reg = &cur->regs[value_regno]; + if (reg && size == BPF_REG_SIZE && register_is_const(reg) && + !register_is_null(reg) && env->allow_ptr_leaks) { + save_register_state(state, spi, reg); + } else if (reg && is_spillable_regtype(reg->type)) { /* register containing pointer is being spilled into stack */ if (size != BPF_REG_SIZE) { + verbose_linfo(env, insn_idx, "; "); verbose(env, "invalid size of register spill\n"); return -EACCES; } - if (state != cur && type == PTR_TO_STACK) { + if (state != cur && reg->type == PTR_TO_STACK) { verbose(env, "cannot spill pointers to stack into stack frame of the caller\n"); return -EINVAL; } - /* save register state */ - state->stack[spi].spilled_ptr = cur->regs[value_regno]; - state->stack[spi].spilled_ptr.live |= REG_LIVE_WRITTEN; - - for (i = 0; i < BPF_REG_SIZE; i++) { - if (state->stack[spi].slot_type[i] == STACK_MISC && - !env->allow_ptr_leaks) { + if (!env->allow_ptr_leaks) { + bool sanitize = false; + + if (state->stack[spi].slot_type[0] == STACK_SPILL && + register_is_const(&state->stack[spi].spilled_ptr)) + sanitize = true; + for (i = 0; i < BPF_REG_SIZE; i++) + if (state->stack[spi].slot_type[i] == STACK_MISC) { + sanitize = true; + break; + } + if (sanitize) { int *poff = &env->insn_aux_data[insn_idx].sanitize_stack_off; int soff = (-spi - 1) * BPF_REG_SIZE; @@ -1446,8 +1473,8 @@ static int check_stack_write(struct bpf_verifier_env *env, } *poff = soff; } - state->stack[spi].slot_type[i] = STACK_SPILL; } + save_register_state(state, spi, reg); } else { u8 type = STACK_MISC; @@ -1466,8 +1493,7 @@ static int check_stack_write(struct bpf_verifier_env *env, state->stack[spi].spilled_ptr.live |= REG_LIVE_WRITTEN; /* when we zero initialize stack slots mark them as such */ - if (value_regno >= 0 && - register_is_null(&cur->regs[value_regno])) + if (reg && register_is_null(reg)) type = STACK_ZERO; for (i = 0; i < size; i++) @@ -1484,6 +1510,7 @@ static int check_stack_read(struct bpf_verifier_env *env, struct bpf_verifier_state *vstate = env->cur_state; struct bpf_func_state *state = vstate->frame[vstate->curframe]; int i, slot = -off - 1, spi = slot / BPF_REG_SIZE; + struct bpf_reg_state *reg; u8 *stype; if (reg_state->allocated_stack <= slot) { @@ -1492,11 +1519,21 @@ static int check_stack_read(struct bpf_verifier_env *env, return -EACCES; } stype = reg_state->stack[spi].slot_type; + reg = ®_state->stack[spi].spilled_ptr; if (stype[0] == STACK_SPILL) { if (size != BPF_REG_SIZE) { - verbose(env, "invalid size of register spill\n"); - return -EACCES; + if (reg->type != SCALAR_VALUE) { + verbose_linfo(env, env->insn_idx, "; "); + verbose(env, "invalid size of register fill\n"); + return -EACCES; + } + if (value_regno >= 0) { + mark_reg_unknown(env, state->regs, value_regno); + state->regs[value_regno].live |= REG_LIVE_WRITTEN; + } + mark_reg_read(env, reg, reg->parent, REG_LIVE_READ64); + return 0; } for (i = 1; i < BPF_REG_SIZE; i++) { if (stype[(slot - i) % BPF_REG_SIZE] != STACK_SPILL) { @@ -1507,17 +1544,14 @@ static int check_stack_read(struct bpf_verifier_env *env, if (value_regno >= 0) { /* restore register state from stack */ - state->regs[value_regno] = reg_state->stack[spi].spilled_ptr; + state->regs[value_regno] = *reg; /* mark reg as written since spilled pointer state likely * has its liveness marks cleared by is_state_visited() * which resets stack/reg liveness for state transitions */ state->regs[value_regno].live |= REG_LIVE_WRITTEN; } - mark_reg_read(env, ®_state->stack[spi].spilled_ptr, - reg_state->stack[spi].spilled_ptr.parent, - REG_LIVE_READ64); - return 0; + mark_reg_read(env, reg, reg->parent, REG_LIVE_READ64); } else { int zeros = 0; @@ -1532,9 +1566,7 @@ static int check_stack_read(struct bpf_verifier_env *env, off, i, size); return -EACCES; } - mark_reg_read(env, ®_state->stack[spi].spilled_ptr, - reg_state->stack[spi].spilled_ptr.parent, - REG_LIVE_READ64); + mark_reg_read(env, reg, reg->parent, REG_LIVE_READ64); if (value_regno >= 0) { if (zeros == size) { /* any size read into register is zero extended, @@ -1547,8 +1579,8 @@ static int check_stack_read(struct bpf_verifier_env *env, } state->regs[value_regno].live |= REG_LIVE_WRITTEN; } - return 0; } + return 0; } static int check_stack_access(struct bpf_verifier_env *env, @@ -2409,7 +2441,7 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno, { struct bpf_reg_state *reg = reg_state(env, regno); struct bpf_func_state *state = func(env, reg); - int err, min_off, max_off, i, slot, spi; + int err, min_off, max_off, i, j, slot, spi; if (reg->type != PTR_TO_STACK) { /* Allow zero-byte read from NULL, regardless of pointer type */ @@ -2497,6 +2529,14 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno, *stype = STACK_MISC; goto mark; } + if (state->stack[spi].slot_type[0] == STACK_SPILL && + state->stack[spi].spilled_ptr.type == SCALAR_VALUE) { + __mark_reg_unknown(&state->stack[spi].spilled_ptr); + for (j = 0; j < BPF_REG_SIZE; j++) + state->stack[spi].slot_type[j] = STACK_MISC; + goto mark; + } + err: if (tnum_is_const(reg->var_off)) { verbose(env, "invalid indirect read from stack off %d+%d size %d\n", -- Gitee From 6a26b587e6c2e6e5976e16eebc0bbc99dfaa6a17 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Sat, 15 Jun 2019 12:12:18 -0700 Subject: [PATCH 14/33] selftests/bpf: fix tests due to const spill/fill ANBZ: #5530 commit fc559a70d57c6ee5443f7a750858503e94cdc941 upstream. fix tests that incorrectly assumed that the verifier cannot track constants through stack. Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/test_verifier.c | 31 +++++++++++---------- 1 file changed, 17 insertions(+), 14 deletions(-) diff --git a/tools/testing/selftests/bpf/test_verifier.c b/tools/testing/selftests/bpf/test_verifier.c index 44497572ce9a..4a9e92f146c7 100644 --- a/tools/testing/selftests/bpf/test_verifier.c +++ b/tools/testing/selftests/bpf/test_verifier.c @@ -4205,7 +4205,8 @@ static struct bpf_test tests[] = { offsetof(struct __sk_buff, data)), BPF_LDX_MEM(BPF_W, BPF_REG_3, BPF_REG_1, offsetof(struct __sk_buff, data_end)), - BPF_MOV64_IMM(BPF_REG_0, 0xffffffff), + BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_1, + offsetof(struct __sk_buff, mark)), BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_0, -8), BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_10, -8), BPF_ALU64_IMM(BPF_AND, BPF_REG_0, 0xffff), @@ -7451,9 +7452,9 @@ static struct bpf_test tests[] = { { "helper access to variable memory: stack, bitwise AND, zero included", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_1, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, -64), - BPF_MOV64_IMM(BPF_REG_2, 16), BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, -128), BPF_ALU64_IMM(BPF_AND, BPF_REG_2, 64), @@ -7468,9 +7469,9 @@ static struct bpf_test tests[] = { { "helper access to variable memory: stack, bitwise AND + JMP, wrong max", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_1, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, -64), - BPF_MOV64_IMM(BPF_REG_2, 16), BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, -128), BPF_ALU64_IMM(BPF_AND, BPF_REG_2, 65), @@ -7544,9 +7545,9 @@ static struct bpf_test tests[] = { { "helper access to variable memory: stack, JMP, bounds + offset", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_1, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, -64), - BPF_MOV64_IMM(BPF_REG_2, 16), BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, -128), BPF_JMP_IMM(BPF_JGT, BPF_REG_2, 64, 5), @@ -7565,9 +7566,9 @@ static struct bpf_test tests[] = { { "helper access to variable memory: stack, JMP, wrong max", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_1, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, -64), - BPF_MOV64_IMM(BPF_REG_2, 16), BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, -128), BPF_JMP_IMM(BPF_JGT, BPF_REG_2, 65, 4), @@ -7585,9 +7586,9 @@ static struct bpf_test tests[] = { { "helper access to variable memory: stack, JMP, no max check", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_1, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, -64), - BPF_MOV64_IMM(BPF_REG_2, 16), BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, -128), BPF_MOV64_IMM(BPF_REG_4, 0), @@ -7605,9 +7606,9 @@ static struct bpf_test tests[] = { { "helper access to variable memory: stack, JMP, no min check", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_1, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, -64), - BPF_MOV64_IMM(BPF_REG_2, 16), BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, -128), BPF_JMP_IMM(BPF_JGT, BPF_REG_2, 64, 3), @@ -7623,9 +7624,9 @@ static struct bpf_test tests[] = { { "helper access to variable memory: stack, JMP (signed), no min check", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_1, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, -64), - BPF_MOV64_IMM(BPF_REG_2, 16), BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, -128), BPF_JMP_IMM(BPF_JSGT, BPF_REG_2, 64, 3), @@ -7667,6 +7668,7 @@ static struct bpf_test tests[] = { { "helper access to variable memory: map, JMP, wrong max", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_6, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_2, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8), BPF_ST_MEM(BPF_DW, BPF_REG_2, 0, 0), @@ -7674,7 +7676,7 @@ static struct bpf_test tests[] = { BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem), BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 10), BPF_MOV64_REG(BPF_REG_1, BPF_REG_0), - BPF_MOV64_IMM(BPF_REG_2, sizeof(struct test_val)), + BPF_MOV64_REG(BPF_REG_2, BPF_REG_6), BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_10, -128), BPF_JMP_IMM(BPF_JSGT, BPF_REG_2, @@ -7686,7 +7688,7 @@ static struct bpf_test tests[] = { BPF_MOV64_IMM(BPF_REG_0, 0), BPF_EXIT_INSN(), }, - .fixup_map_hash_48b = { 3 }, + .fixup_map_hash_48b = { 4 }, .errstr = "invalid access to map value, value_size=48 off=0 size=49", .result = REJECT, .prog_type = BPF_PROG_TYPE_TRACEPOINT, @@ -7721,6 +7723,7 @@ static struct bpf_test tests[] = { { "helper access to variable memory: map adjusted, JMP, wrong max", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_6, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_2, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8), BPF_ST_MEM(BPF_DW, BPF_REG_2, 0, 0), @@ -7729,7 +7732,7 @@ static struct bpf_test tests[] = { BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 11), BPF_MOV64_REG(BPF_REG_1, BPF_REG_0), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, 20), - BPF_MOV64_IMM(BPF_REG_2, sizeof(struct test_val)), + BPF_MOV64_REG(BPF_REG_2, BPF_REG_6), BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_10, -128), BPF_JMP_IMM(BPF_JSGT, BPF_REG_2, @@ -7741,7 +7744,7 @@ static struct bpf_test tests[] = { BPF_MOV64_IMM(BPF_REG_0, 0), BPF_EXIT_INSN(), }, - .fixup_map_hash_48b = { 3 }, + .fixup_map_hash_48b = { 4 }, .errstr = "R1 min value is outside of the array range", .result = REJECT, .prog_type = BPF_PROG_TYPE_TRACEPOINT, @@ -7763,8 +7766,8 @@ static struct bpf_test tests[] = { { "helper access to variable memory: size > 0 not allowed on NULL (ARG_PTR_TO_MEM_OR_NULL)", .insns = { + BPF_LDX_MEM(BPF_W, BPF_REG_2, BPF_REG_1, 0), BPF_MOV64_IMM(BPF_REG_1, 0), - BPF_MOV64_IMM(BPF_REG_2, 1), BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_10, -128), BPF_ALU64_IMM(BPF_AND, BPF_REG_2, 64), @@ -7991,6 +7994,7 @@ static struct bpf_test tests[] = { { "helper access to variable memory: 8 bytes leak", .insns = { + BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_1, 8), BPF_MOV64_REG(BPF_REG_1, BPF_REG_10), BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, -64), BPF_MOV64_IMM(BPF_REG_0, 0), @@ -8001,7 +8005,6 @@ static struct bpf_test tests[] = { BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_0, -24), BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_0, -16), BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_0, -8), - BPF_MOV64_IMM(BPF_REG_2, 1), BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_2, -128), BPF_LDX_MEM(BPF_DW, BPF_REG_2, BPF_REG_10, -128), BPF_ALU64_IMM(BPF_AND, BPF_REG_2, 63), -- Gitee From c273abf5e7fbdc25cdebb8c873eec82d52e564e0 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Sat, 15 Jun 2019 12:12:20 -0700 Subject: [PATCH 15/33] bpf: introduce bounded loops ANBZ: #5530 commit 2589726d12a1b12eaaa93c7f1ea64287e383c7a5 upstream. Allow the verifier to validate the loops by simulating their execution. Exisiting programs have used '#pragma unroll' to unroll the loops by the compiler. Instead let the verifier simulate all iterations of the loop. In order to do that introduce parentage chain of bpf_verifier_state and 'branches' counter for the number of branches left to explore. See more detailed algorithm description in bpf_verifier.h This algorithm borrows the key idea from Edward Cree approach: https://patchwork.ozlabs.org/patch/877222/ Additional state pruning heuristics make such brute force loop walk practical even for large loops. Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- include/linux/bpf_verifier.h | 51 ++++++++++++- kernel/bpf/verifier.c | 143 ++++++++++++++++++++++++++++++++--- 2 files changed, 181 insertions(+), 13 deletions(-) diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h index 0b96b1077c6d..11a862bad2fb 100644 --- a/include/linux/bpf_verifier.h +++ b/include/linux/bpf_verifier.h @@ -194,6 +194,53 @@ struct bpf_func_state { struct bpf_verifier_state { /* call stack tracking */ struct bpf_func_state *frame[MAX_CALL_FRAMES]; + struct bpf_verifier_state *parent; + /* + * 'branches' field is the number of branches left to explore: + * 0 - all possible paths from this state reached bpf_exit or + * were safely pruned + * 1 - at least one path is being explored. + * This state hasn't reached bpf_exit + * 2 - at least two paths are being explored. + * This state is an immediate parent of two children. + * One is fallthrough branch with branches==1 and another + * state is pushed into stack (to be explored later) also with + * branches==1. The parent of this state has branches==1. + * The verifier state tree connected via 'parent' pointer looks like: + * 1 + * 1 + * 2 -> 1 (first 'if' pushed into stack) + * 1 + * 2 -> 1 (second 'if' pushed into stack) + * 1 + * 1 + * 1 bpf_exit. + * + * Once do_check() reaches bpf_exit, it calls update_branch_counts() + * and the verifier state tree will look: + * 1 + * 1 + * 2 -> 1 (first 'if' pushed into stack) + * 1 + * 1 -> 1 (second 'if' pushed into stack) + * 0 + * 0 + * 0 bpf_exit. + * After pop_stack() the do_check() will resume at second 'if'. + * + * If is_state_visited() sees a state with branches > 0 it means + * there is a loop. If such state is exactly equal to the current state + * it's an infinite loop. Note states_equal() checks for states + * equvalency, so two states being 'states_equal' does not mean + * infinite loop. The exact comparison is provided by + * states_maybe_looping() function. It's a stronger pre-check and + * much faster than states_equal(). + * + * This algorithm may not find all possible infinite loops or + * loop iteration count may be too high. + * In such cases BPF_COMPLEXITY_LIMIT_INSNS limit kicks in. + */ + u32 branches; u32 insn_idx; u32 curframe; u32 active_spin_lock; @@ -313,7 +360,9 @@ struct bpf_verifier_env { } cfg; u32 subprog_cnt; /* number of instructions analyzed by the verifier */ - u32 insn_processed; + u32 prev_insn_processed, insn_processed; + /* number of jmps, calls, exits analyzed so far */ + u32 prev_jmps_processed, jmps_processed; /* total verification time */ u64 verification_time; /* maximum number of verifier states kept in 'branching' instructions */ diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 73d5623e88bf..c742156c8274 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -720,6 +720,8 @@ static int copy_verifier_state(struct bpf_verifier_state *dst_state, dst_state->speculative = src->speculative; dst_state->curframe = src->curframe; dst_state->active_spin_lock = src->active_spin_lock; + dst_state->branches = src->branches; + dst_state->parent = src->parent; for (i = 0; i <= src->curframe; i++) { dst = dst_state->frame[i]; if (!dst) { @@ -735,6 +737,23 @@ static int copy_verifier_state(struct bpf_verifier_state *dst_state, return 0; } +static void update_branch_counts(struct bpf_verifier_env *env, struct bpf_verifier_state *st) +{ + while (st) { + u32 br = --st->branches; + + /* WARN_ON(br > 1) technically makes sense here, + * but see comment in push_stack(), hence: + */ + WARN_ONCE((int)br < 0, + "BUG update_branch_counts:branches_to_explore=%d\n", + br); + if (br) + break; + st = st->parent; + } +} + static int pop_stack(struct bpf_verifier_env *env, int *prev_insn_idx, int *insn_idx) { @@ -788,6 +807,18 @@ static struct bpf_verifier_state *push_stack(struct bpf_verifier_env *env, env->stack_size); goto err; } + if (elem->st.parent) { + ++elem->st.parent->branches; + /* WARN_ON(branches > 2) technically makes sense here, + * but + * 1. speculative states will bump 'branches' for non-branch + * instructions + * 2. is_state_visited() heuristics may decide not to create + * a new state for a sequence of branches and all such current + * and cloned states will be pointing to a single parent state + * which might have large 'branches' count. + */ + } return &elem->st; err: free_verifier_state(env->cur_state, true); @@ -5858,7 +5889,8 @@ static void init_explored_state(struct bpf_verifier_env *env, int idx) * w - next instruction * e - edge */ -static int push_insn(int t, int w, int e, struct bpf_verifier_env *env) +static int push_insn(int t, int w, int e, struct bpf_verifier_env *env, + bool loop_ok) { int *insn_stack = env->cfg.insn_stack; int *insn_state = env->cfg.insn_state; @@ -5888,6 +5920,8 @@ static int push_insn(int t, int w, int e, struct bpf_verifier_env *env) insn_stack[env->cfg.cur_stack++] = w; return 1; } else if ((insn_state[w] & 0xF0) == DISCOVERED) { + if (loop_ok && env->allow_ptr_leaks) + return 0; verbose_linfo(env, t, "%d: ", t); verbose_linfo(env, w, "%d: ", w); verbose(env, "back-edge from insn %d to %d\n", t, w); @@ -5939,7 +5973,7 @@ static int check_cfg(struct bpf_verifier_env *env) if (opcode == BPF_EXIT) { goto mark_explored; } else if (opcode == BPF_CALL) { - ret = push_insn(t, t + 1, FALLTHROUGH, env); + ret = push_insn(t, t + 1, FALLTHROUGH, env, false); if (ret == 1) goto peek_stack; else if (ret < 0) @@ -5948,7 +5982,8 @@ static int check_cfg(struct bpf_verifier_env *env) init_explored_state(env, t + 1); if (insns[t].src_reg == BPF_PSEUDO_CALL) { init_explored_state(env, t); - ret = push_insn(t, t + insns[t].imm + 1, BRANCH, env); + ret = push_insn(t, t + insns[t].imm + 1, BRANCH, + env, false); if (ret == 1) goto peek_stack; else if (ret < 0) @@ -5961,7 +5996,7 @@ static int check_cfg(struct bpf_verifier_env *env) } /* unconditional jump with single edge */ ret = push_insn(t, t + insns[t].off + 1, - FALLTHROUGH, env); + FALLTHROUGH, env, true); if (ret == 1) goto peek_stack; else if (ret < 0) @@ -5974,13 +6009,13 @@ static int check_cfg(struct bpf_verifier_env *env) } else { /* conditional jump with two edges */ init_explored_state(env, t); - ret = push_insn(t, t + 1, FALLTHROUGH, env); + ret = push_insn(t, t + 1, FALLTHROUGH, env, true); if (ret == 1) goto peek_stack; else if (ret < 0) goto err_free; - ret = push_insn(t, t + insns[t].off + 1, BRANCH, env); + ret = push_insn(t, t + insns[t].off + 1, BRANCH, env, true); if (ret == 1) goto peek_stack; else if (ret < 0) @@ -5990,7 +6025,7 @@ static int check_cfg(struct bpf_verifier_env *env) /* all other non-branch instructions with single * fall-through edge */ - ret = push_insn(t, t + 1, FALLTHROUGH, env); + ret = push_insn(t, t + 1, FALLTHROUGH, env, false); if (ret == 1) goto peek_stack; else if (ret < 0) @@ -6423,6 +6458,8 @@ static void clean_live_states(struct bpf_verifier_env *env, int insn, sl = *explored_state(env, insn); while (sl) { + if (sl->state.branches) + goto next; if (sl->state.insn_idx != insn || sl->state.curframe != cur->curframe) goto next; @@ -6787,12 +6824,32 @@ static int propagate_liveness(struct bpf_verifier_env *env, return 0; } +static bool states_maybe_looping(struct bpf_verifier_state *old, + struct bpf_verifier_state *cur) +{ + struct bpf_func_state *fold, *fcur; + int i, fr = cur->curframe; + + if (old->curframe != fr) + return false; + + fold = old->frame[fr]; + fcur = cur->frame[fr]; + for (i = 0; i < MAX_BPF_REG; i++) + if (memcmp(&fold->regs[i], &fcur->regs[i], + offsetof(struct bpf_reg_state, parent))) + return false; + return true; +} + + static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) { struct bpf_verifier_state_list *new_sl; struct bpf_verifier_state_list *sl, **pprev; struct bpf_verifier_state *cur = env->cur_state, *new; int i, j, err, states_cnt = 0; + bool add_new_state = false; if (!env->insn_aux_data[insn_idx].prune_point) /* this 'insn_idx' instruction wasn't marked, so we will not @@ -6800,6 +6857,18 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) */ return 0; + /* bpf progs typically have pruning point every 4 instructions + * http://vger.kernel.org/bpfconf2019.html#session-1 + * Do not add new state for future pruning if the verifier hasn't seen + * at least 2 jumps and at least 8 instructions. + * This heuristics helps decrease 'total_states' and 'peak_states' metric. + * In tests that amounts to up to 50% reduction into total verifier + * memory consumption and 20% verifier time speedup. + */ + if (env->jmps_processed - env->prev_jmps_processed >= 2 && + env->insn_processed - env->prev_insn_processed >= 8) + add_new_state = true; + pprev = explored_state(env, insn_idx); sl = *pprev; @@ -6809,6 +6878,30 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) states_cnt++; if (sl->state.insn_idx != insn_idx) goto next; + if (sl->state.branches) { + if (states_maybe_looping(&sl->state, cur) && + states_equal(env, &sl->state, cur)) { + verbose_linfo(env, insn_idx, "; "); + verbose(env, "infinite loop detected at insn %d\n", insn_idx); + return -EINVAL; + } + /* if the verifier is processing a loop, avoid adding new state + * too often, since different loop iterations have distinct + * states and may not help future pruning. + * This threshold shouldn't be too low to make sure that + * a loop with large bound will be rejected quickly. + * The most abusive loop will be: + * r1 += 1 + * if r1 < 1000000 goto pc-2 + * 1M insn_procssed limit / 100 == 10k peak states. + * This threshold shouldn't be too high either, since states + * at the end of the loop are likely to be useful in pruning. + */ + if (env->jmps_processed - env->prev_jmps_processed < 20 && + env->insn_processed - env->prev_insn_processed < 100) + add_new_state = false; + goto miss; + } if (states_equal(env, &sl->state, cur)) { sl->hit_cnt++; /* reached equivalent register/stack state, @@ -6826,7 +6919,15 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) return err; return 1; } - sl->miss_cnt++; +miss: + /* when new state is not going to be added do not increase miss count. + * Otherwise several loop iterations will remove the state + * recorded earlier. The goal of these heuristics is to have + * states from some iterations of the loop (some in the beginning + * and some at the end) to help pruning. + */ + if (add_new_state) + sl->miss_cnt++; /* heuristic to determine whether this state is beneficial * to keep checking from state equivalence point of view. * Higher numbers increase max_states_per_insn and verification time, @@ -6838,6 +6939,11 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) */ *pprev = sl->next; if (sl->state.frame[0]->regs[0].live & REG_LIVE_DONE) { + u32 br = sl->state.branches; + + WARN_ONCE(br, + "BUG live_done but branches_to_explore %d\n", + br); free_verifier_state(&sl->state, false); kfree(sl); env->peak_states--; @@ -6863,18 +6969,25 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) if (!env->allow_ptr_leaks && states_cnt > BPF_COMPLEXITY_LIMIT_STATES) return 0; - /* there were no equivalent states, remember current one. - * technically the current state is not proven to be safe yet, + if (!add_new_state) + return 0; + + /* There were no equivalent states, remember the current one. + * Technically the current state is not proven to be safe yet, * but it will either reach outer most bpf_exit (which means it's safe) - * or it will be rejected. Since there are no loops, we won't be + * or it will be rejected. When there are no loops the verifier won't be * seeing this tuple (frame[0].callsite, frame[1].callsite, .. insn_idx) - * again on the way to bpf_exit + * again on the way to bpf_exit. + * When looping the sl->state.branches will be > 0 and this state + * will not be considered for equivalence until branches == 0. */ new_sl = kzalloc(sizeof(struct bpf_verifier_state_list), GFP_KERNEL); if (!new_sl) return -ENOMEM; env->total_states++; env->peak_states++; + env->prev_jmps_processed = env->jmps_processed; + env->prev_insn_processed = env->insn_processed; /* add new state to the head of linked list */ new = &new_sl->state; @@ -6885,6 +6998,9 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) return err; } new->insn_idx = insn_idx; + WARN_ONCE(new->branches != 1, + "BUG is_state_visited:branches_to_explore=%d insn %d\n", new->branches, insn_idx); + cur->parent = new; new_sl->next = *explored_state(env, insn_idx); *explored_state(env, insn_idx) = new_sl; /* connect new state to parentage chain. Current frame needs all @@ -6971,6 +7087,7 @@ static int do_check(struct bpf_verifier_env *env) return -ENOMEM; state->curframe = 0; state->speculative = false; + state->branches = 1; state->frame[0] = kzalloc(sizeof(struct bpf_func_state), GFP_KERNEL); if (!state->frame[0]) { kfree(state); @@ -7177,6 +7294,7 @@ static int do_check(struct bpf_verifier_env *env) } else if (class == BPF_JMP || class == BPF_JMP32) { u8 opcode = BPF_OP(insn->code); + env->jmps_processed++; if (opcode == BPF_CALL) { if (BPF_SRC(insn->code) != BPF_K || insn->off != 0 || @@ -7262,6 +7380,7 @@ static int do_check(struct bpf_verifier_env *env) if (err) return err; process_bpf_exit: + update_branch_counts(env, env->cur_state); err = pop_stack(env, &env->prev_insn_idx, &env->insn_idx); if (err < 0) { -- Gitee From b2aea72237635784bceccb46db4d810d2a43446e Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Sat, 15 Jun 2019 12:12:21 -0700 Subject: [PATCH 16/33] bpf: fix callees pruning callers ANBZ: #5530 commit eea1c227b9e9bad295e8ef984004a9acf12bb68c upstream. The commit 7640ead93924 partially resolved the issue of callees incorrectly pruning the callers. With introduction of bounded loops and jmps_processed heuristic single verifier state may contain multiple branches and calls. It's possible that new verifier state (for future pruning) will be allocated inside callee. Then callee will exit (still within the same verifier state). It will go back to the caller and there R6-R9 registers will be read and will trigger mark_reg_read. But the reg->live for all frames but the top frame is not set to LIVE_NONE. Hence mark_reg_read will fail to propagate liveness into parent and future walking will incorrectly conclude that the states are equivalent because LIVE_READ is not set. In other words the rule for parent/live should be: whenever register parentage chain is set the reg->live should be set to LIVE_NONE. is_state_visited logic already follows this rule for spilled registers. Fixes: 7640ead93924 ("bpf: verifier: make sure callees don't prune with caller differences") Fixes: f4d7e40a5b71 ("bpf: introduce function calls (verification)") Signed-off-by: Alexei Starovoitov Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index c742156c8274..338bc0598971 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -7010,17 +7010,18 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) * the state of the call instruction (with WRITTEN set), and r0 comes * from callee with its full parentage chain, anyway. */ - for (j = 0; j <= cur->curframe; j++) - for (i = j < cur->curframe ? BPF_REG_6 : 0; i < BPF_REG_FP; i++) - cur->frame[j]->regs[i].parent = &new->frame[j]->regs[i]; /* clear write marks in current state: the writes we did are not writes * our child did, so they don't screen off its reads from us. * (There are no read marks in current state, because reads always mark * their parent and current state never has children yet. Only * explored_states can get read marks.) */ - for (i = 0; i < BPF_REG_FP; i++) - cur->frame[cur->curframe]->regs[i].live = REG_LIVE_NONE; + for (j = 0; j <= cur->curframe; j++) { + for (i = j < cur->curframe ? BPF_REG_6 : 0; i < BPF_REG_FP; i++) + cur->frame[j]->regs[i].parent = &new->frame[j]->regs[i]; + for (i = 0; i < BPF_REG_FP; i++) + cur->frame[j]->regs[i].live = REG_LIVE_NONE; + } /* all stack frames are accessible from callee, clear them all */ for (j = 0; j <= cur->curframe; j++) { -- Gitee From bc55ff54c5e073e708a49c9797098d98fd703051 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Sat, 15 Jun 2019 12:12:22 -0700 Subject: [PATCH 17/33] selftests/bpf: fix tests ANBZ: #5530 commit aeee380ccfc59c05a4c6e8e762fd645c6bdc773a upstream. Fix tests that assumed no loops. Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/test_verifier.c | 44 +++++++++++---------- 1 file changed, 24 insertions(+), 20 deletions(-) diff --git a/tools/testing/selftests/bpf/test_verifier.c b/tools/testing/selftests/bpf/test_verifier.c index 4a9e92f146c7..71a7b2062e7a 100644 --- a/tools/testing/selftests/bpf/test_verifier.c +++ b/tools/testing/selftests/bpf/test_verifier.c @@ -237,10 +237,10 @@ static void bpf_fill_scale1(struct bpf_test *self) insn[i++] = BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_6, -8 * (k % 64 + 1)); } - /* every jump adds 1 step to insn_processed, so to stay exactly - * within 1m limit add MAX_TEST_INSNS - MAX_JMP_SEQ - 1 MOVs and 1 EXIT + /* is_state_visited() doesn't allocate state for pruning for every jump. + * Hence multiply jmps by 4 to accommodate that heuristic */ - while (i < MAX_TEST_INSNS - MAX_JMP_SEQ - 1) + while (i < MAX_TEST_INSNS - MAX_JMP_SEQ * 4) insn[i++] = BPF_ALU64_IMM(BPF_MOV, BPF_REG_0, 42); insn[i] = BPF_EXIT_INSN(); self->prog_len = i + 1; @@ -269,10 +269,7 @@ static void bpf_fill_scale2(struct bpf_test *self) insn[i++] = BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_6, -8 * (k % (64 - 4 * FUNC_NEST) + 1)); } - /* every jump adds 1 step to insn_processed, so to stay exactly - * within 1m limit add MAX_TEST_INSNS - MAX_JMP_SEQ - 1 MOVs and 1 EXIT - */ - while (i < MAX_TEST_INSNS - MAX_JMP_SEQ - 1) + while (i < MAX_TEST_INSNS - MAX_JMP_SEQ * 4) insn[i++] = BPF_ALU64_IMM(BPF_MOV, BPF_REG_0, 42); insn[i] = BPF_EXIT_INSN(); self->prog_len = i + 1; @@ -947,7 +944,8 @@ static struct bpf_test tests[] = { BPF_JMP_IMM(BPF_JA, 0, 0, -1), BPF_EXIT_INSN(), }, - .errstr = "back-edge", + .errstr = "unreachable insn 1", + .errstr_unpriv = "back-edge", .result = REJECT, }, { @@ -959,19 +957,21 @@ static struct bpf_test tests[] = { BPF_JMP_IMM(BPF_JA, 0, 0, -4), BPF_EXIT_INSN(), }, - .errstr = "back-edge", + .errstr = "unreachable insn 4", + .errstr_unpriv = "back-edge", .result = REJECT, }, { "conditional loop", .insns = { - BPF_MOV64_REG(BPF_REG_1, BPF_REG_0), + BPF_MOV64_REG(BPF_REG_0, BPF_REG_1), BPF_MOV64_REG(BPF_REG_2, BPF_REG_0), BPF_MOV64_REG(BPF_REG_3, BPF_REG_0), BPF_JMP_IMM(BPF_JEQ, BPF_REG_1, 0, -3), BPF_EXIT_INSN(), }, - .errstr = "back-edge", + .errstr = "infinite loop detected", + .errstr_unpriv = "back-edge", .result = REJECT, }, { @@ -11361,9 +11361,11 @@ static struct bpf_test tests[] = { BPF_MOV64_IMM(BPF_REG_0, 3), BPF_JMP_IMM(BPF_JA, 0, 0, -6), }, - .prog_type = BPF_PROG_TYPE_TRACEPOINT, - .errstr = "back-edge from insn", - .result = REJECT, + .prog_type = BPF_PROG_TYPE_SOCKET_FILTER, + .errstr_unpriv = "back-edge from insn", + .result_unpriv = REJECT, + .result = ACCEPT, + .retval = 1, }, { "calls: conditional call 4", @@ -11396,22 +11398,24 @@ static struct bpf_test tests[] = { BPF_MOV64_IMM(BPF_REG_0, 3), BPF_EXIT_INSN(), }, - .prog_type = BPF_PROG_TYPE_TRACEPOINT, - .errstr = "back-edge from insn", - .result = REJECT, + .prog_type = BPF_PROG_TYPE_SCHED_CLS, + .result = ACCEPT, + .retval = 1, }, { "calls: conditional call 6", .insns = { + BPF_MOV64_REG(BPF_REG_6, BPF_REG_1), + BPF_MOV64_REG(BPF_REG_1, BPF_REG_6), BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 1, 0, 2), - BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, -2), + BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, -3), BPF_EXIT_INSN(), BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_1, offsetof(struct __sk_buff, mark)), BPF_EXIT_INSN(), }, - .prog_type = BPF_PROG_TYPE_TRACEPOINT, - .errstr = "back-edge from insn", + .prog_type = BPF_PROG_TYPE_SCHED_CLS, + .errstr = "infinite loop detected", .result = REJECT, }, { -- Gitee From 6d1d7c64bde62ba4c8eaf79613fdd9fff85bb910 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Sat, 15 Jun 2019 12:12:23 -0700 Subject: [PATCH 18/33] selftests/bpf: add basic verifier tests for loops ANBZ: #5530 commit 0d3679e99ae4b7868da22e3b8540fd597df501f5 upstream. This set of tests is a rewrite of Edward's earlier tests: https://patchwork.ozlabs.org/patch/877221/ Signed-off-by: Alexei Starovoitov Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/test_verifier.c | 161 ++++++++++++++++++++ 1 file changed, 161 insertions(+) diff --git a/tools/testing/selftests/bpf/test_verifier.c b/tools/testing/selftests/bpf/test_verifier.c index 71a7b2062e7a..35ffb64c9813 100644 --- a/tools/testing/selftests/bpf/test_verifier.c +++ b/tools/testing/selftests/bpf/test_verifier.c @@ -974,6 +974,167 @@ static struct bpf_test tests[] = { .errstr_unpriv = "back-edge", .result = REJECT, }, + { + "bounded loop, count to 4", + .insns = { + BPF_MOV64_IMM(BPF_REG_0, 0), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 4, -2), + BPF_EXIT_INSN(), + }, + .result = ACCEPT, + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + .retval = 4, + }, + { + "bounded loop, count to 20", + .insns = { + BPF_MOV64_IMM(BPF_REG_0, 0), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 3), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 20, -2), + BPF_EXIT_INSN(), + }, + .result = ACCEPT, + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + }, + { + "bounded loop, count from positive unknown to 4", + .insns = { + BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_get_prandom_u32), + BPF_JMP_IMM(BPF_JSLT, BPF_REG_0, 0, 2), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 4, -2), + BPF_EXIT_INSN(), + }, + .result = ACCEPT, + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + .retval = 4, + }, + { + "bounded loop, count from totally unknown to 4", + .insns = { + BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_get_prandom_u32), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 4, -2), + BPF_EXIT_INSN(), + }, + .result = ACCEPT, + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + }, + { + "bounded loop, count to 4 with equality", + .insns = { + BPF_MOV64_IMM(BPF_REG_0, 0), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 4, -2), + BPF_EXIT_INSN(), + }, + .result = ACCEPT, + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + }, + { + "bounded loop, start in the middle", + .insns = { + BPF_MOV64_IMM(BPF_REG_0, 0), + BPF_JMP_A(1), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 4, -2), + BPF_EXIT_INSN(), + }, + .result = REJECT, + .errstr = "back-edge", + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + .retval = 4, + }, + { + "bounded loop containing a forward jump", + .insns = { + BPF_MOV64_IMM(BPF_REG_0, 0), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_JMP_REG(BPF_JEQ, BPF_REG_0, BPF_REG_0, 0), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 4, -3), + BPF_EXIT_INSN(), + }, + .result = ACCEPT, + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + .retval = 4, + }, + { + "bounded loop that jumps out rather than in", + .insns = { + BPF_MOV64_IMM(BPF_REG_6, 0), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_6, 1), + BPF_JMP_IMM(BPF_JGT, BPF_REG_6, 10000, 2), + BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_get_prandom_u32), + BPF_JMP_A(-4), + BPF_EXIT_INSN(), + }, + .result = ACCEPT, + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + }, + { + "infinite loop after a conditional jump", + .insns = { + BPF_MOV64_IMM(BPF_REG_0, 5), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 4, 2), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_JMP_A(-2), + BPF_EXIT_INSN(), + }, + .result = REJECT, + .errstr = "program is too large", + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + }, + { + "bounded recursion", + .insns = { + BPF_MOV64_IMM(BPF_REG_1, 0), + BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 1, 0, 1), + BPF_EXIT_INSN(), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_1, 1), + BPF_MOV64_REG(BPF_REG_0, BPF_REG_1), + BPF_JMP_IMM(BPF_JLT, BPF_REG_1, 4, 1), + BPF_EXIT_INSN(), + BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 1, 0, -5), + BPF_EXIT_INSN(), + }, + .result = REJECT, + .errstr = "back-edge", + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + }, + { + "infinite loop in two jumps", + .insns = { + BPF_MOV64_IMM(BPF_REG_0, 0), + BPF_JMP_A(0), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 4, -2), + BPF_EXIT_INSN(), + }, + .result = REJECT, + .errstr = "loop detected", + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + }, + { + "infinite loop: three-jump trick", + .insns = { + BPF_MOV64_IMM(BPF_REG_0, 0), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_ALU64_IMM(BPF_AND, BPF_REG_0, 1), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 2, 1), + BPF_EXIT_INSN(), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_ALU64_IMM(BPF_AND, BPF_REG_0, 1), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 2, 1), + BPF_EXIT_INSN(), + BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 1), + BPF_ALU64_IMM(BPF_AND, BPF_REG_0, 1), + BPF_JMP_IMM(BPF_JLT, BPF_REG_0, 2, -11), + BPF_EXIT_INSN(), + }, + .result = REJECT, + .errstr = "loop detected", + .prog_type = BPF_PROG_TYPE_TRACEPOINT, + }, { "read uninitialized register", .insns = { -- Gitee From 26f1dab50053c9db3c436450f3e965447afd5769 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Sat, 15 Jun 2019 12:12:24 -0700 Subject: [PATCH 19/33] selftests/bpf: add realistic loop tests ANBZ: #5530 commit b061017f8b4d0e05d4c11486581a702fb2a975b2 upstream. Add a bunch of loop tests. Most of them are created by replacing '#pragma unroll' with '#pragma clang loop unroll(disable)' Several tests are artificially large: /* partial unroll. llvm will unroll loop ~150 times. * C loop count -> 600. * Asm loop count -> 4. * 16k insns in loop body. * Total of 5 such loops. Total program size ~82k insns. */ "./pyperf600.o", /* no unroll at all. * C loop count -> 600. * ASM loop count -> 600. * ~110 insns in loop body. * Total of 5 such loops. Total program size ~1500 insns. */ "./pyperf600_nounroll.o", /* partial unroll. 19k insn in a loop. * Total program size 20.8k insn. * ~350k processed_insns */ "./strobemeta.o", Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/progs/loop1.c | 28 + tools/testing/selftests/bpf/progs/loop2.c | 28 + tools/testing/selftests/bpf/progs/loop3.c | 22 + tools/testing/selftests/bpf/progs/pyperf.h | 6 +- tools/testing/selftests/bpf/progs/pyperf600.c | 9 + .../selftests/bpf/progs/pyperf600_nounroll.c | 8 + .../testing/selftests/bpf/progs/strobemeta.c | 9 + .../testing/selftests/bpf/progs/strobemeta.h | 528 ++++++++++++++++++ .../bpf/progs/strobemeta_nounroll1.c | 9 + .../bpf/progs/strobemeta_nounroll2.c | 9 + .../selftests/bpf/progs/test_seg6_loop.c | 261 +++++++++ .../selftests/bpf/progs/test_sysctl_loop1.c | 71 +++ .../selftests/bpf/progs/test_sysctl_loop2.c | 72 +++ .../selftests/bpf/progs/test_xdp_loop.c | 231 ++++++++ tools/testing/selftests/bpf/test_progs.c | 67 ++- 15 files changed, 1346 insertions(+), 12 deletions(-) create mode 100644 tools/testing/selftests/bpf/progs/loop1.c create mode 100644 tools/testing/selftests/bpf/progs/loop2.c create mode 100644 tools/testing/selftests/bpf/progs/loop3.c create mode 100644 tools/testing/selftests/bpf/progs/pyperf600.c create mode 100644 tools/testing/selftests/bpf/progs/pyperf600_nounroll.c create mode 100644 tools/testing/selftests/bpf/progs/strobemeta.c create mode 100644 tools/testing/selftests/bpf/progs/strobemeta.h create mode 100644 tools/testing/selftests/bpf/progs/strobemeta_nounroll1.c create mode 100644 tools/testing/selftests/bpf/progs/strobemeta_nounroll2.c create mode 100644 tools/testing/selftests/bpf/progs/test_seg6_loop.c create mode 100644 tools/testing/selftests/bpf/progs/test_sysctl_loop1.c create mode 100644 tools/testing/selftests/bpf/progs/test_sysctl_loop2.c create mode 100644 tools/testing/selftests/bpf/progs/test_xdp_loop.c diff --git a/tools/testing/selftests/bpf/progs/loop1.c b/tools/testing/selftests/bpf/progs/loop1.c new file mode 100644 index 000000000000..dea395af9ea9 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/loop1.c @@ -0,0 +1,28 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook +#include +#include +#include +#include +#include +#include +#include "bpf_helpers.h" + +char _license[] SEC("license") = "GPL"; + +SEC("raw_tracepoint/kfree_skb") +int nested_loops(volatile struct pt_regs* ctx) +{ + int i, j, sum = 0, m; + + for (j = 0; j < 300; j++) + for (i = 0; i < j; i++) { + if (j & 1) + m = ctx->rax; + else + m = j; + sum += i * m; + } + + return sum; +} diff --git a/tools/testing/selftests/bpf/progs/loop2.c b/tools/testing/selftests/bpf/progs/loop2.c new file mode 100644 index 000000000000..0637bd8e8bcf --- /dev/null +++ b/tools/testing/selftests/bpf/progs/loop2.c @@ -0,0 +1,28 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook +#include +#include +#include +#include +#include +#include +#include "bpf_helpers.h" + +char _license[] SEC("license") = "GPL"; + +SEC("raw_tracepoint/consume_skb") +int while_true(volatile struct pt_regs* ctx) +{ + int i = 0; + + while (true) { + if (ctx->rax & 1) + i += 3; + else + i += 7; + if (i > 40) + break; + } + + return i; +} diff --git a/tools/testing/selftests/bpf/progs/loop3.c b/tools/testing/selftests/bpf/progs/loop3.c new file mode 100644 index 000000000000..30a0f6cba080 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/loop3.c @@ -0,0 +1,22 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook +#include +#include +#include +#include +#include +#include +#include "bpf_helpers.h" + +char _license[] SEC("license") = "GPL"; + +SEC("raw_tracepoint/consume_skb") +int while_true(volatile struct pt_regs* ctx) +{ + __u64 i = 0, sum = 0; + do { + i++; + sum += ctx->rax; + } while (i < 0x100000000ULL); + return sum; +} diff --git a/tools/testing/selftests/bpf/progs/pyperf.h b/tools/testing/selftests/bpf/progs/pyperf.h index 0cc5e4ee90bd..6b0781391be5 100644 --- a/tools/testing/selftests/bpf/progs/pyperf.h +++ b/tools/testing/selftests/bpf/progs/pyperf.h @@ -220,7 +220,11 @@ static inline __attribute__((__always_inline__)) int __on_event(struct pt_regs * int32_t* symbol_counter = bpf_map_lookup_elem(&symbolmap, &sym); if (symbol_counter == NULL) return 0; -#pragma unroll +#ifdef NO_UNROLL +#pragma clang loop unroll(disable) +#else +#pragma clang loop unroll(full) +#endif /* Unwind python stack */ for (int i = 0; i < STACK_MAX_LEN; ++i) { if (frame_ptr && get_frame_data(frame_ptr, pidData, &frame, &sym)) { diff --git a/tools/testing/selftests/bpf/progs/pyperf600.c b/tools/testing/selftests/bpf/progs/pyperf600.c new file mode 100644 index 000000000000..cb49b89e37cd --- /dev/null +++ b/tools/testing/selftests/bpf/progs/pyperf600.c @@ -0,0 +1,9 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook +#define STACK_MAX_LEN 600 +/* clang will not unroll the loop 600 times. + * Instead it will unroll it to the amount it deemed + * appropriate, but the loop will still execute 600 times. + * Total program size is around 90k insns + */ +#include "pyperf.h" diff --git a/tools/testing/selftests/bpf/progs/pyperf600_nounroll.c b/tools/testing/selftests/bpf/progs/pyperf600_nounroll.c new file mode 100644 index 000000000000..6beff7502f4d --- /dev/null +++ b/tools/testing/selftests/bpf/progs/pyperf600_nounroll.c @@ -0,0 +1,8 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook +#define STACK_MAX_LEN 600 +#define NO_UNROLL +/* clang will not unroll at all. + * Total program size is around 2k insns + */ +#include "pyperf.h" diff --git a/tools/testing/selftests/bpf/progs/strobemeta.c b/tools/testing/selftests/bpf/progs/strobemeta.c new file mode 100644 index 000000000000..77b50b704184 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/strobemeta.c @@ -0,0 +1,9 @@ +// SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) +// Copyright (c) 2019 Facebook + +#define STROBE_MAX_INTS 2 +#define STROBE_MAX_STRS 25 +#define STROBE_MAX_MAPS 100 +#define STROBE_MAX_MAP_ENTRIES 20 +/* full unroll by llvm #undef NO_UNROLL */ +#include "strobemeta.h" diff --git a/tools/testing/selftests/bpf/progs/strobemeta.h b/tools/testing/selftests/bpf/progs/strobemeta.h new file mode 100644 index 000000000000..1ff73f60a3e4 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/strobemeta.h @@ -0,0 +1,528 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook + +#include +#include +#include +#include +#include +#include +#include +#include "bpf_helpers.h" + +typedef uint32_t pid_t; +struct task_struct {}; + +#define TASK_COMM_LEN 16 +#define PERF_MAX_STACK_DEPTH 127 + +#define STROBE_TYPE_INVALID 0 +#define STROBE_TYPE_INT 1 +#define STROBE_TYPE_STR 2 +#define STROBE_TYPE_MAP 3 + +#define STACK_TABLE_EPOCH_SHIFT 20 +#define STROBE_MAX_STR_LEN 1 +#define STROBE_MAX_CFGS 32 +#define STROBE_MAX_PAYLOAD \ + (STROBE_MAX_STRS * STROBE_MAX_STR_LEN + \ + STROBE_MAX_MAPS * (1 + STROBE_MAX_MAP_ENTRIES * 2) * STROBE_MAX_STR_LEN) + +struct strobe_value_header { + /* + * meaning depends on type: + * 1. int: 0, if value not set, 1 otherwise + * 2. str: 1 always, whether value is set or not is determined by ptr + * 3. map: 1 always, pointer points to additional struct with number + * of entries (up to STROBE_MAX_MAP_ENTRIES) + */ + uint16_t len; + /* + * _reserved might be used for some future fields/flags, but we always + * want to keep strobe_value_header to be 8 bytes, so BPF can read 16 + * bytes in one go and get both header and value + */ + uint8_t _reserved[6]; +}; + +/* + * strobe_value_generic is used from BPF probe only, but needs to be a union + * of strobe_value_int/strobe_value_str/strobe_value_map + */ +struct strobe_value_generic { + struct strobe_value_header header; + union { + int64_t val; + void *ptr; + }; +}; + +struct strobe_value_int { + struct strobe_value_header header; + int64_t value; +}; + +struct strobe_value_str { + struct strobe_value_header header; + const char* value; +}; + +struct strobe_value_map { + struct strobe_value_header header; + const struct strobe_map_raw* value; +}; + +struct strobe_map_entry { + const char* key; + const char* val; +}; + +/* + * Map of C-string key/value pairs with fixed maximum capacity. Each map has + * corresponding int64 ID, which application can use (or ignore) in whatever + * way appropriate. Map is "write-only", there is no way to get data out of + * map. Map is intended to be used to provide metadata for profilers and is + * not to be used for internal in-app communication. All methods are + * thread-safe. + */ +struct strobe_map_raw { + /* + * general purpose unique ID that's up to application to decide + * whether and how to use; for request metadata use case id is unique + * request ID that's used to match metadata with stack traces on + * Strobelight backend side + */ + int64_t id; + /* number of used entries in map */ + int64_t cnt; + /* + * having volatile doesn't change anything on BPF side, but clang + * emits warnings for passing `volatile const char *` into + * bpf_probe_read_str that expects just `const char *` + */ + const char* tag; + /* + * key/value entries, each consisting of 2 pointers to key and value + * C strings + */ + struct strobe_map_entry entries[STROBE_MAX_MAP_ENTRIES]; +}; + +/* Following values define supported values of TLS mode */ +#define TLS_NOT_SET -1 +#define TLS_LOCAL_EXEC 0 +#define TLS_IMM_EXEC 1 +#define TLS_GENERAL_DYN 2 + +/* + * structure that universally represents TLS location (both for static + * executables and shared libraries) + */ +struct strobe_value_loc { + /* + * tls_mode defines what TLS mode was used for particular metavariable: + * - -1 (TLS_NOT_SET) - no metavariable; + * - 0 (TLS_LOCAL_EXEC) - Local Executable mode; + * - 1 (TLS_IMM_EXEC) - Immediate Executable mode; + * - 2 (TLS_GENERAL_DYN) - General Dynamic mode; + * Local Dynamic mode is not yet supported, because never seen in + * practice. Mode defines how offset field is interpreted. See + * calc_location() in below for details. + */ + int64_t tls_mode; + /* + * TLS_LOCAL_EXEC: offset from thread pointer (fs:0 for x86-64, + * tpidr_el0 for aarch64). + * TLS_IMM_EXEC: absolute address of GOT entry containing offset + * from thread pointer; + * TLS_GENERAL_DYN: absolute addres of double GOT entry + * containing tls_index_t struct; + */ + int64_t offset; +}; + +struct strobemeta_cfg { + int64_t req_meta_idx; + struct strobe_value_loc int_locs[STROBE_MAX_INTS]; + struct strobe_value_loc str_locs[STROBE_MAX_STRS]; + struct strobe_value_loc map_locs[STROBE_MAX_MAPS]; +}; + +struct strobe_map_descr { + uint64_t id; + int16_t tag_len; + /* + * cnt <0 - map value isn't set; + * 0 - map has id set, but no key/value entries + */ + int16_t cnt; + /* + * both key_lens[i] and val_lens[i] should be >0 for present key/value + * entry + */ + uint16_t key_lens[STROBE_MAX_MAP_ENTRIES]; + uint16_t val_lens[STROBE_MAX_MAP_ENTRIES]; +}; + +struct strobemeta_payload { + /* req_id has valid request ID, if req_meta_valid == 1 */ + int64_t req_id; + uint8_t req_meta_valid; + /* + * mask has Nth bit set to 1, if Nth metavar was present and + * successfully read + */ + uint64_t int_vals_set_mask; + int64_t int_vals[STROBE_MAX_INTS]; + /* len is >0 for present values */ + uint16_t str_lens[STROBE_MAX_STRS]; + /* if map_descrs[i].cnt == -1, metavar is not present/set */ + struct strobe_map_descr map_descrs[STROBE_MAX_MAPS]; + /* + * payload has compactly packed values of str and map variables in the + * form: strval1\0strval2\0map1key1\0map1val1\0map2key1\0map2val1\0 + * (and so on); str_lens[i], key_lens[i] and val_lens[i] determines + * value length + */ + char payload[STROBE_MAX_PAYLOAD]; +}; + +struct strobelight_bpf_sample { + uint64_t ktime; + char comm[TASK_COMM_LEN]; + pid_t pid; + int user_stack_id; + int kernel_stack_id; + int has_meta; + struct strobemeta_payload metadata; + /* + * makes it possible to pass ( + 1) as data size to + * perf_submit() to avoid perf_submit's paranoia about passing zero as + * size, as it deduces that might be + * **theoretically** zero + */ + char dummy_safeguard; +}; + +struct bpf_map_def SEC("maps") samples = { + .type = BPF_MAP_TYPE_PERF_EVENT_ARRAY, + .key_size = sizeof(int), + .value_size = sizeof(int), + .max_entries = 32, +}; + +struct bpf_map_def SEC("maps") stacks_0 = { + .type = BPF_MAP_TYPE_STACK_TRACE, + .key_size = sizeof(uint32_t), + .value_size = sizeof(uint64_t) * PERF_MAX_STACK_DEPTH, + .max_entries = 16, +}; + +struct bpf_map_def SEC("maps") stacks_1 = { + .type = BPF_MAP_TYPE_STACK_TRACE, + .key_size = sizeof(uint32_t), + .value_size = sizeof(uint64_t) * PERF_MAX_STACK_DEPTH, + .max_entries = 16, +}; + +struct bpf_map_def SEC("maps") sample_heap = { + .type = BPF_MAP_TYPE_PERCPU_ARRAY, + .key_size = sizeof(uint32_t), + .value_size = sizeof(struct strobelight_bpf_sample), + .max_entries = 1, +}; + +struct bpf_map_def SEC("maps") strobemeta_cfgs = { + .type = BPF_MAP_TYPE_PERCPU_ARRAY, + .key_size = sizeof(pid_t), + .value_size = sizeof(struct strobemeta_cfg), + .max_entries = STROBE_MAX_CFGS, +}; + +/* Type for the dtv. */ +/* https://github.com/lattera/glibc/blob/master/nptl/sysdeps/x86_64/tls.h#L34 */ +typedef union dtv { + size_t counter; + struct { + void* val; + bool is_static; + } pointer; +} dtv_t; + +/* Partial definition for tcbhead_t */ +/* https://github.com/bminor/glibc/blob/master/sysdeps/x86_64/nptl/tls.h#L42 */ +struct tcbhead { + void* tcb; + dtv_t* dtv; +}; + +/* + * TLS module/offset information for shared library case. + * For x86-64, this is mapped onto two entries in GOT. + * For aarch64, this is pointed to by second GOT entry. + */ +struct tls_index { + uint64_t module; + uint64_t offset; +}; + +static inline __attribute__((always_inline)) +void *calc_location(struct strobe_value_loc *loc, void *tls_base) +{ + /* + * tls_mode value is: + * - -1 (TLS_NOT_SET), if no metavar is present; + * - 0 (TLS_LOCAL_EXEC), if metavar uses Local Executable mode of TLS + * (offset from fs:0 for x86-64 or tpidr_el0 for aarch64); + * - 1 (TLS_IMM_EXEC), if metavar uses Immediate Executable mode of TLS; + * - 2 (TLS_GENERAL_DYN), if metavar uses General Dynamic mode of TLS; + * This schema allows to use something like: + * (tls_mode + 1) * (tls_base + offset) + * to get NULL for "no metavar" location, or correct pointer for local + * executable mode without doing extra ifs. + */ + if (loc->tls_mode <= TLS_LOCAL_EXEC) { + /* static executable is simple, we just have offset from + * tls_base */ + void *addr = tls_base + loc->offset; + /* multiply by (tls_mode + 1) to get NULL, if we have no + * metavar in this slot */ + return (void *)((loc->tls_mode + 1) * (int64_t)addr); + } + /* + * Other modes are more complicated, we need to jump through few hoops. + * + * For immediate executable mode (currently supported only for aarch64): + * - loc->offset is pointing to a GOT entry containing fixed offset + * relative to tls_base; + * + * For general dynamic mode: + * - loc->offset is pointing to a beginning of double GOT entries; + * - (for aarch64 only) second entry points to tls_index_t struct; + * - (for x86-64 only) two GOT entries are already tls_index_t; + * - tls_index_t->module is used to find start of TLS section in + * which variable resides; + * - tls_index_t->offset provides offset within that TLS section, + * pointing to value of variable. + */ + struct tls_index tls_index; + dtv_t *dtv; + void *tls_ptr; + + bpf_probe_read(&tls_index, sizeof(struct tls_index), + (void *)loc->offset); + /* valid module index is always positive */ + if (tls_index.module > 0) { + /* dtv = ((struct tcbhead *)tls_base)->dtv[tls_index.module] */ + bpf_probe_read(&dtv, sizeof(dtv), + &((struct tcbhead *)tls_base)->dtv); + dtv += tls_index.module; + } else { + dtv = NULL; + } + bpf_probe_read(&tls_ptr, sizeof(void *), dtv); + /* if pointer has (void *)-1 value, then TLS wasn't initialized yet */ + return tls_ptr && tls_ptr != (void *)-1 + ? tls_ptr + tls_index.offset + : NULL; +} + +static inline __attribute__((always_inline)) +void read_int_var(struct strobemeta_cfg *cfg, size_t idx, void *tls_base, + struct strobe_value_generic *value, + struct strobemeta_payload *data) +{ + void *location = calc_location(&cfg->int_locs[idx], tls_base); + if (!location) + return; + + bpf_probe_read(value, sizeof(struct strobe_value_generic), location); + data->int_vals[idx] = value->val; + if (value->header.len) + data->int_vals_set_mask |= (1 << idx); +} + +static inline __attribute__((always_inline)) +uint64_t read_str_var(struct strobemeta_cfg* cfg, size_t idx, void *tls_base, + struct strobe_value_generic *value, + struct strobemeta_payload *data, void *payload) +{ + void *location; + uint32_t len; + + data->str_lens[idx] = 0; + location = calc_location(&cfg->str_locs[idx], tls_base); + if (!location) + return 0; + + bpf_probe_read(value, sizeof(struct strobe_value_generic), location); + len = bpf_probe_read_str(payload, STROBE_MAX_STR_LEN, value->ptr); + /* + * if bpf_probe_read_str returns error (<0), due to casting to + * unsinged int, it will become big number, so next check is + * sufficient to check for errors AND prove to BPF verifier, that + * bpf_probe_read_str won't return anything bigger than + * STROBE_MAX_STR_LEN + */ + if (len > STROBE_MAX_STR_LEN) + return 0; + + data->str_lens[idx] = len; + return len; +} + +static inline __attribute__((always_inline)) +void *read_map_var(struct strobemeta_cfg *cfg, size_t idx, void *tls_base, + struct strobe_value_generic *value, + struct strobemeta_payload* data, void *payload) +{ + struct strobe_map_descr* descr = &data->map_descrs[idx]; + struct strobe_map_raw map; + void *location; + uint32_t len; + int i; + + descr->tag_len = 0; /* presume no tag is set */ + descr->cnt = -1; /* presume no value is set */ + + location = calc_location(&cfg->map_locs[idx], tls_base); + if (!location) + return payload; + + bpf_probe_read(value, sizeof(struct strobe_value_generic), location); + if (bpf_probe_read(&map, sizeof(struct strobe_map_raw), value->ptr)) + return payload; + + descr->id = map.id; + descr->cnt = map.cnt; + if (cfg->req_meta_idx == idx) { + data->req_id = map.id; + data->req_meta_valid = 1; + } + + len = bpf_probe_read_str(payload, STROBE_MAX_STR_LEN, map.tag); + if (len <= STROBE_MAX_STR_LEN) { + descr->tag_len = len; + payload += len; + } + +#ifdef NO_UNROLL +#pragma clang loop unroll(disable) +#else +#pragma unroll +#endif + for (int i = 0; i < STROBE_MAX_MAP_ENTRIES && i < map.cnt; ++i) { + descr->key_lens[i] = 0; + len = bpf_probe_read_str(payload, STROBE_MAX_STR_LEN, + map.entries[i].key); + if (len <= STROBE_MAX_STR_LEN) { + descr->key_lens[i] = len; + payload += len; + } + descr->val_lens[i] = 0; + len = bpf_probe_read_str(payload, STROBE_MAX_STR_LEN, + map.entries[i].val); + if (len <= STROBE_MAX_STR_LEN) { + descr->val_lens[i] = len; + payload += len; + } + } + + return payload; +} + +/* + * read_strobe_meta returns NULL, if no metadata was read; otherwise returns + * pointer to *right after* payload ends + */ +static inline __attribute__((always_inline)) +void *read_strobe_meta(struct task_struct* task, + struct strobemeta_payload* data) { + pid_t pid = bpf_get_current_pid_tgid() >> 32; + struct strobe_value_generic value = {0}; + struct strobemeta_cfg *cfg; + void *tls_base, *payload; + + cfg = bpf_map_lookup_elem(&strobemeta_cfgs, &pid); + if (!cfg) + return NULL; + + data->int_vals_set_mask = 0; + data->req_meta_valid = 0; + payload = data->payload; + /* + * we don't have struct task_struct definition, it should be: + * tls_base = (void *)task->thread.fsbase; + */ + tls_base = (void *)task; + +#ifdef NO_UNROLL +#pragma clang loop unroll(disable) +#else +#pragma unroll +#endif + for (int i = 0; i < STROBE_MAX_INTS; ++i) { + read_int_var(cfg, i, tls_base, &value, data); + } +#ifdef NO_UNROLL +#pragma clang loop unroll(disable) +#else +#pragma unroll +#endif + for (int i = 0; i < STROBE_MAX_STRS; ++i) { + payload += read_str_var(cfg, i, tls_base, &value, data, payload); + } +#ifdef NO_UNROLL +#pragma clang loop unroll(disable) +#else +#pragma unroll +#endif + for (int i = 0; i < STROBE_MAX_MAPS; ++i) { + payload = read_map_var(cfg, i, tls_base, &value, data, payload); + } + /* + * return pointer right after end of payload, so it's possible to + * calculate exact amount of useful data that needs to be sent + */ + return payload; +} + +SEC("raw_tracepoint/kfree_skb") +int on_event(struct pt_regs *ctx) { + pid_t pid = bpf_get_current_pid_tgid() >> 32; + struct strobelight_bpf_sample* sample; + struct task_struct *task; + uint32_t zero = 0; + uint64_t ktime_ns; + void *sample_end; + + sample = bpf_map_lookup_elem(&sample_heap, &zero); + if (!sample) + return 0; /* this will never happen */ + + sample->pid = pid; + bpf_get_current_comm(&sample->comm, TASK_COMM_LEN); + ktime_ns = bpf_ktime_get_ns(); + sample->ktime = ktime_ns; + + task = (struct task_struct *)bpf_get_current_task(); + sample_end = read_strobe_meta(task, &sample->metadata); + sample->has_meta = sample_end != NULL; + sample_end = sample_end ? : &sample->metadata; + + if ((ktime_ns >> STACK_TABLE_EPOCH_SHIFT) & 1) { + sample->kernel_stack_id = bpf_get_stackid(ctx, &stacks_1, 0); + sample->user_stack_id = bpf_get_stackid(ctx, &stacks_1, BPF_F_USER_STACK); + } else { + sample->kernel_stack_id = bpf_get_stackid(ctx, &stacks_0, 0); + sample->user_stack_id = bpf_get_stackid(ctx, &stacks_0, BPF_F_USER_STACK); + } + + uint64_t sample_size = sample_end - (void *)sample; + /* should always be true */ + if (sample_size < sizeof(struct strobelight_bpf_sample)) + bpf_perf_event_output(ctx, &samples, 0, sample, 1 + sample_size); + return 0; +} + +char _license[] SEC("license") = "GPL"; diff --git a/tools/testing/selftests/bpf/progs/strobemeta_nounroll1.c b/tools/testing/selftests/bpf/progs/strobemeta_nounroll1.c new file mode 100644 index 000000000000..f0a1669e11d6 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/strobemeta_nounroll1.c @@ -0,0 +1,9 @@ +// SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) +// Copyright (c) 2019 Facebook + +#define STROBE_MAX_INTS 2 +#define STROBE_MAX_STRS 25 +#define STROBE_MAX_MAPS 13 +#define STROBE_MAX_MAP_ENTRIES 20 +#define NO_UNROLL +#include "strobemeta.h" diff --git a/tools/testing/selftests/bpf/progs/strobemeta_nounroll2.c b/tools/testing/selftests/bpf/progs/strobemeta_nounroll2.c new file mode 100644 index 000000000000..4291a7d642e7 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/strobemeta_nounroll2.c @@ -0,0 +1,9 @@ +// SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) +// Copyright (c) 2019 Facebook + +#define STROBE_MAX_INTS 2 +#define STROBE_MAX_STRS 25 +#define STROBE_MAX_MAPS 30 +#define STROBE_MAX_MAP_ENTRIES 20 +#define NO_UNROLL +#include "strobemeta.h" diff --git a/tools/testing/selftests/bpf/progs/test_seg6_loop.c b/tools/testing/selftests/bpf/progs/test_seg6_loop.c new file mode 100644 index 000000000000..463964d79f73 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_seg6_loop.c @@ -0,0 +1,261 @@ +#include +#include +#include +#include +#include +#include "bpf_helpers.h" +#include "bpf_endian.h" + +/* Packet parsing state machine helpers. */ +#define cursor_advance(_cursor, _len) \ + ({ void *_tmp = _cursor; _cursor += _len; _tmp; }) + +#define SR6_FLAG_ALERT (1 << 4) + +#define htonll(x) ((bpf_htonl(1)) == 1 ? (x) : ((uint64_t)bpf_htonl((x) & \ + 0xFFFFFFFF) << 32) | bpf_htonl((x) >> 32)) +#define ntohll(x) ((bpf_ntohl(1)) == 1 ? (x) : ((uint64_t)bpf_ntohl((x) & \ + 0xFFFFFFFF) << 32) | bpf_ntohl((x) >> 32)) +#define BPF_PACKET_HEADER __attribute__((packed)) + +struct ip6_t { + unsigned int ver:4; + unsigned int priority:8; + unsigned int flow_label:20; + unsigned short payload_len; + unsigned char next_header; + unsigned char hop_limit; + unsigned long long src_hi; + unsigned long long src_lo; + unsigned long long dst_hi; + unsigned long long dst_lo; +} BPF_PACKET_HEADER; + +struct ip6_addr_t { + unsigned long long hi; + unsigned long long lo; +} BPF_PACKET_HEADER; + +struct ip6_srh_t { + unsigned char nexthdr; + unsigned char hdrlen; + unsigned char type; + unsigned char segments_left; + unsigned char first_segment; + unsigned char flags; + unsigned short tag; + + struct ip6_addr_t segments[0]; +} BPF_PACKET_HEADER; + +struct sr6_tlv_t { + unsigned char type; + unsigned char len; + unsigned char value[0]; +} BPF_PACKET_HEADER; + +static __attribute__((always_inline)) struct ip6_srh_t *get_srh(struct __sk_buff *skb) +{ + void *cursor, *data_end; + struct ip6_srh_t *srh; + struct ip6_t *ip; + uint8_t *ipver; + + data_end = (void *)(long)skb->data_end; + cursor = (void *)(long)skb->data; + ipver = (uint8_t *)cursor; + + if ((void *)ipver + sizeof(*ipver) > data_end) + return NULL; + + if ((*ipver >> 4) != 6) + return NULL; + + ip = cursor_advance(cursor, sizeof(*ip)); + if ((void *)ip + sizeof(*ip) > data_end) + return NULL; + + if (ip->next_header != 43) + return NULL; + + srh = cursor_advance(cursor, sizeof(*srh)); + if ((void *)srh + sizeof(*srh) > data_end) + return NULL; + + if (srh->type != 4) + return NULL; + + return srh; +} + +static __attribute__((always_inline)) +int update_tlv_pad(struct __sk_buff *skb, uint32_t new_pad, + uint32_t old_pad, uint32_t pad_off) +{ + int err; + + if (new_pad != old_pad) { + err = bpf_lwt_seg6_adjust_srh(skb, pad_off, + (int) new_pad - (int) old_pad); + if (err) + return err; + } + + if (new_pad > 0) { + char pad_tlv_buf[16] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0}; + struct sr6_tlv_t *pad_tlv = (struct sr6_tlv_t *) pad_tlv_buf; + + pad_tlv->type = SR6_TLV_PADDING; + pad_tlv->len = new_pad - 2; + + err = bpf_lwt_seg6_store_bytes(skb, pad_off, + (void *)pad_tlv_buf, new_pad); + if (err) + return err; + } + + return 0; +} + +static __attribute__((always_inline)) +int is_valid_tlv_boundary(struct __sk_buff *skb, struct ip6_srh_t *srh, + uint32_t *tlv_off, uint32_t *pad_size, + uint32_t *pad_off) +{ + uint32_t srh_off, cur_off; + int offset_valid = 0; + int err; + + srh_off = (char *)srh - (char *)(long)skb->data; + // cur_off = end of segments, start of possible TLVs + cur_off = srh_off + sizeof(*srh) + + sizeof(struct ip6_addr_t) * (srh->first_segment + 1); + + *pad_off = 0; + + // we can only go as far as ~10 TLVs due to the BPF max stack size + #pragma clang loop unroll(disable) + for (int i = 0; i < 100; i++) { + struct sr6_tlv_t tlv; + + if (cur_off == *tlv_off) + offset_valid = 1; + + if (cur_off >= srh_off + ((srh->hdrlen + 1) << 3)) + break; + + err = bpf_skb_load_bytes(skb, cur_off, &tlv, sizeof(tlv)); + if (err) + return err; + + if (tlv.type == SR6_TLV_PADDING) { + *pad_size = tlv.len + sizeof(tlv); + *pad_off = cur_off; + + if (*tlv_off == srh_off) { + *tlv_off = cur_off; + offset_valid = 1; + } + break; + + } else if (tlv.type == SR6_TLV_HMAC) { + break; + } + + cur_off += sizeof(tlv) + tlv.len; + } // we reached the padding or HMAC TLVs, or the end of the SRH + + if (*pad_off == 0) + *pad_off = cur_off; + + if (*tlv_off == -1) + *tlv_off = cur_off; + else if (!offset_valid) + return -EINVAL; + + return 0; +} + +static __attribute__((always_inline)) +int add_tlv(struct __sk_buff *skb, struct ip6_srh_t *srh, uint32_t tlv_off, + struct sr6_tlv_t *itlv, uint8_t tlv_size) +{ + uint32_t srh_off = (char *)srh - (char *)(long)skb->data; + uint8_t len_remaining, new_pad; + uint32_t pad_off = 0; + uint32_t pad_size = 0; + uint32_t partial_srh_len; + int err; + + if (tlv_off != -1) + tlv_off += srh_off; + + if (itlv->type == SR6_TLV_PADDING || itlv->type == SR6_TLV_HMAC) + return -EINVAL; + + err = is_valid_tlv_boundary(skb, srh, &tlv_off, &pad_size, &pad_off); + if (err) + return err; + + err = bpf_lwt_seg6_adjust_srh(skb, tlv_off, sizeof(*itlv) + itlv->len); + if (err) + return err; + + err = bpf_lwt_seg6_store_bytes(skb, tlv_off, (void *)itlv, tlv_size); + if (err) + return err; + + // the following can't be moved inside update_tlv_pad because the + // bpf verifier has some issues with it + pad_off += sizeof(*itlv) + itlv->len; + partial_srh_len = pad_off - srh_off; + len_remaining = partial_srh_len % 8; + new_pad = 8 - len_remaining; + + if (new_pad == 1) // cannot pad for 1 byte only + new_pad = 9; + else if (new_pad == 8) + new_pad = 0; + + return update_tlv_pad(skb, new_pad, pad_size, pad_off); +} + +// Add an Egress TLV fc00::4, add the flag A, +// and apply End.X action to fc42::1 +SEC("lwt_seg6local") +int __add_egr_x(struct __sk_buff *skb) +{ + unsigned long long hi = 0xfc42000000000000; + unsigned long long lo = 0x1; + struct ip6_srh_t *srh = get_srh(skb); + uint8_t new_flags = SR6_FLAG_ALERT; + struct ip6_addr_t addr; + int err, offset; + + if (srh == NULL) + return BPF_DROP; + + uint8_t tlv[20] = {2, 18, 0, 0, 0xfd, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, + 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x4}; + + err = add_tlv(skb, srh, (srh->hdrlen+1) << 3, + (struct sr6_tlv_t *)&tlv, 20); + if (err) + return BPF_DROP; + + offset = sizeof(struct ip6_t) + offsetof(struct ip6_srh_t, flags); + err = bpf_lwt_seg6_store_bytes(skb, offset, + (void *)&new_flags, sizeof(new_flags)); + if (err) + return BPF_DROP; + + addr.lo = htonll(lo); + addr.hi = htonll(hi); + err = bpf_lwt_seg6_action(skb, SEG6_LOCAL_ACTION_END_X, + (void *)&addr, sizeof(addr)); + if (err) + return BPF_DROP; + return BPF_REDIRECT; +} +char __license[] SEC("license") = "GPL"; diff --git a/tools/testing/selftests/bpf/progs/test_sysctl_loop1.c b/tools/testing/selftests/bpf/progs/test_sysctl_loop1.c new file mode 100644 index 000000000000..608a06871572 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_sysctl_loop1.c @@ -0,0 +1,71 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook + +#include +#include + +#include +#include + +#include "bpf_helpers.h" + +#ifndef ARRAY_SIZE +#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) +#endif + +/* tcp_mem sysctl has only 3 ints, but this test is doing TCP_MEM_LOOPS */ +#define TCP_MEM_LOOPS 28 /* because 30 doesn't fit into 512 bytes of stack */ +#define MAX_ULONG_STR_LEN 7 +#define MAX_VALUE_STR_LEN (TCP_MEM_LOOPS * MAX_ULONG_STR_LEN) + +static __always_inline int is_tcp_mem(struct bpf_sysctl *ctx) +{ + volatile char tcp_mem_name[] = "net/ipv4/tcp_mem/very_very_very_very_long_pointless_string"; + unsigned char i; + char name[64]; + int ret; + + memset(name, 0, sizeof(name)); + ret = bpf_sysctl_get_name(ctx, name, sizeof(name), 0); + if (ret < 0 || ret != sizeof(tcp_mem_name) - 1) + return 0; + +#pragma clang loop unroll(disable) + for (i = 0; i < sizeof(tcp_mem_name); ++i) + if (name[i] != tcp_mem_name[i]) + return 0; + + return 1; +} + +SEC("cgroup/sysctl") +int sysctl_tcp_mem(struct bpf_sysctl *ctx) +{ + unsigned long tcp_mem[TCP_MEM_LOOPS] = {}; + char value[MAX_VALUE_STR_LEN]; + unsigned char i, off = 0; + int ret; + + if (ctx->write) + return 0; + + if (!is_tcp_mem(ctx)) + return 0; + + ret = bpf_sysctl_get_current_value(ctx, value, MAX_VALUE_STR_LEN); + if (ret < 0 || ret >= MAX_VALUE_STR_LEN) + return 0; + +#pragma clang loop unroll(disable) + for (i = 0; i < ARRAY_SIZE(tcp_mem); ++i) { + ret = bpf_strtoul(value + off, MAX_ULONG_STR_LEN, 0, + tcp_mem + i); + if (ret <= 0 || ret > MAX_ULONG_STR_LEN) + return 0; + off += ret & MAX_ULONG_STR_LEN; + } + + return tcp_mem[0] < tcp_mem[1] && tcp_mem[1] < tcp_mem[2]; +} + +char _license[] SEC("license") = "GPL"; diff --git a/tools/testing/selftests/bpf/progs/test_sysctl_loop2.c b/tools/testing/selftests/bpf/progs/test_sysctl_loop2.c new file mode 100644 index 000000000000..cb201cbe11e7 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_sysctl_loop2.c @@ -0,0 +1,72 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook + +#include +#include + +#include +#include + +#include "bpf_helpers.h" + +#ifndef ARRAY_SIZE +#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) +#endif + +/* tcp_mem sysctl has only 3 ints, but this test is doing TCP_MEM_LOOPS */ +#define TCP_MEM_LOOPS 20 /* because 30 doesn't fit into 512 bytes of stack */ +#define MAX_ULONG_STR_LEN 7 +#define MAX_VALUE_STR_LEN (TCP_MEM_LOOPS * MAX_ULONG_STR_LEN) + +static __attribute__((noinline)) int is_tcp_mem(struct bpf_sysctl *ctx) +{ + volatile char tcp_mem_name[] = "net/ipv4/tcp_mem/very_very_very_very_long_pointless_string_to_stress_byte_loop"; + unsigned char i; + char name[64]; + int ret; + + memset(name, 0, sizeof(name)); + ret = bpf_sysctl_get_name(ctx, name, sizeof(name), 0); + if (ret < 0 || ret != sizeof(tcp_mem_name) - 1) + return 0; + +#pragma clang loop unroll(disable) + for (i = 0; i < sizeof(tcp_mem_name); ++i) + if (name[i] != tcp_mem_name[i]) + return 0; + + return 1; +} + + +SEC("cgroup/sysctl") +int sysctl_tcp_mem(struct bpf_sysctl *ctx) +{ + unsigned long tcp_mem[TCP_MEM_LOOPS] = {}; + char value[MAX_VALUE_STR_LEN]; + unsigned char i, off = 0; + int ret; + + if (ctx->write) + return 0; + + if (!is_tcp_mem(ctx)) + return 0; + + ret = bpf_sysctl_get_current_value(ctx, value, MAX_VALUE_STR_LEN); + if (ret < 0 || ret >= MAX_VALUE_STR_LEN) + return 0; + +#pragma clang loop unroll(disable) + for (i = 0; i < ARRAY_SIZE(tcp_mem); ++i) { + ret = bpf_strtoul(value + off, MAX_ULONG_STR_LEN, 0, + tcp_mem + i); + if (ret <= 0 || ret > MAX_ULONG_STR_LEN) + return 0; + off += ret & MAX_ULONG_STR_LEN; + } + + return tcp_mem[0] < tcp_mem[1] && tcp_mem[1] < tcp_mem[2]; +} + +char _license[] SEC("license") = "GPL"; diff --git a/tools/testing/selftests/bpf/progs/test_xdp_loop.c b/tools/testing/selftests/bpf/progs/test_xdp_loop.c new file mode 100644 index 000000000000..7fa4677df22e --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_xdp_loop.c @@ -0,0 +1,231 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2019 Facebook +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "bpf_helpers.h" +#include "bpf_endian.h" +#include "test_iptunnel_common.h" + +int _version SEC("version") = 1; + +struct bpf_map_def SEC("maps") rxcnt = { + .type = BPF_MAP_TYPE_PERCPU_ARRAY, + .key_size = sizeof(__u32), + .value_size = sizeof(__u64), + .max_entries = 256, +}; + +struct bpf_map_def SEC("maps") vip2tnl = { + .type = BPF_MAP_TYPE_HASH, + .key_size = sizeof(struct vip), + .value_size = sizeof(struct iptnl_info), + .max_entries = MAX_IPTNL_ENTRIES, +}; + +static __always_inline void count_tx(__u32 protocol) +{ + __u64 *rxcnt_count; + + rxcnt_count = bpf_map_lookup_elem(&rxcnt, &protocol); + if (rxcnt_count) + *rxcnt_count += 1; +} + +static __always_inline int get_dport(void *trans_data, void *data_end, + __u8 protocol) +{ + struct tcphdr *th; + struct udphdr *uh; + + switch (protocol) { + case IPPROTO_TCP: + th = (struct tcphdr *)trans_data; + if (th + 1 > data_end) + return -1; + return th->dest; + case IPPROTO_UDP: + uh = (struct udphdr *)trans_data; + if (uh + 1 > data_end) + return -1; + return uh->dest; + default: + return 0; + } +} + +static __always_inline void set_ethhdr(struct ethhdr *new_eth, + const struct ethhdr *old_eth, + const struct iptnl_info *tnl, + __be16 h_proto) +{ + memcpy(new_eth->h_source, old_eth->h_dest, sizeof(new_eth->h_source)); + memcpy(new_eth->h_dest, tnl->dmac, sizeof(new_eth->h_dest)); + new_eth->h_proto = h_proto; +} + +static __always_inline int handle_ipv4(struct xdp_md *xdp) +{ + void *data_end = (void *)(long)xdp->data_end; + void *data = (void *)(long)xdp->data; + struct iptnl_info *tnl; + struct ethhdr *new_eth; + struct ethhdr *old_eth; + struct iphdr *iph = data + sizeof(struct ethhdr); + __u16 *next_iph; + __u16 payload_len; + struct vip vip = {}; + int dport; + __u32 csum = 0; + int i; + + if (iph + 1 > data_end) + return XDP_DROP; + + dport = get_dport(iph + 1, data_end, iph->protocol); + if (dport == -1) + return XDP_DROP; + + vip.protocol = iph->protocol; + vip.family = AF_INET; + vip.daddr.v4 = iph->daddr; + vip.dport = dport; + payload_len = bpf_ntohs(iph->tot_len); + + tnl = bpf_map_lookup_elem(&vip2tnl, &vip); + /* It only does v4-in-v4 */ + if (!tnl || tnl->family != AF_INET) + return XDP_PASS; + + if (bpf_xdp_adjust_head(xdp, 0 - (int)sizeof(struct iphdr))) + return XDP_DROP; + + data = (void *)(long)xdp->data; + data_end = (void *)(long)xdp->data_end; + + new_eth = data; + iph = data + sizeof(*new_eth); + old_eth = data + sizeof(*iph); + + if (new_eth + 1 > data_end || + old_eth + 1 > data_end || + iph + 1 > data_end) + return XDP_DROP; + + set_ethhdr(new_eth, old_eth, tnl, bpf_htons(ETH_P_IP)); + + iph->version = 4; + iph->ihl = sizeof(*iph) >> 2; + iph->frag_off = 0; + iph->protocol = IPPROTO_IPIP; + iph->check = 0; + iph->tos = 0; + iph->tot_len = bpf_htons(payload_len + sizeof(*iph)); + iph->daddr = tnl->daddr.v4; + iph->saddr = tnl->saddr.v4; + iph->ttl = 8; + + next_iph = (__u16 *)iph; +#pragma clang loop unroll(disable) + for (i = 0; i < sizeof(*iph) >> 1; i++) + csum += *next_iph++; + + iph->check = ~((csum & 0xffff) + (csum >> 16)); + + count_tx(vip.protocol); + + return XDP_TX; +} + +static __always_inline int handle_ipv6(struct xdp_md *xdp) +{ + void *data_end = (void *)(long)xdp->data_end; + void *data = (void *)(long)xdp->data; + struct iptnl_info *tnl; + struct ethhdr *new_eth; + struct ethhdr *old_eth; + struct ipv6hdr *ip6h = data + sizeof(struct ethhdr); + __u16 payload_len; + struct vip vip = {}; + int dport; + + if (ip6h + 1 > data_end) + return XDP_DROP; + + dport = get_dport(ip6h + 1, data_end, ip6h->nexthdr); + if (dport == -1) + return XDP_DROP; + + vip.protocol = ip6h->nexthdr; + vip.family = AF_INET6; + memcpy(vip.daddr.v6, ip6h->daddr.s6_addr32, sizeof(vip.daddr)); + vip.dport = dport; + payload_len = ip6h->payload_len; + + tnl = bpf_map_lookup_elem(&vip2tnl, &vip); + /* It only does v6-in-v6 */ + if (!tnl || tnl->family != AF_INET6) + return XDP_PASS; + + if (bpf_xdp_adjust_head(xdp, 0 - (int)sizeof(struct ipv6hdr))) + return XDP_DROP; + + data = (void *)(long)xdp->data; + data_end = (void *)(long)xdp->data_end; + + new_eth = data; + ip6h = data + sizeof(*new_eth); + old_eth = data + sizeof(*ip6h); + + if (new_eth + 1 > data_end || old_eth + 1 > data_end || + ip6h + 1 > data_end) + return XDP_DROP; + + set_ethhdr(new_eth, old_eth, tnl, bpf_htons(ETH_P_IPV6)); + + ip6h->version = 6; + ip6h->priority = 0; + memset(ip6h->flow_lbl, 0, sizeof(ip6h->flow_lbl)); + ip6h->payload_len = bpf_htons(bpf_ntohs(payload_len) + sizeof(*ip6h)); + ip6h->nexthdr = IPPROTO_IPV6; + ip6h->hop_limit = 8; + memcpy(ip6h->saddr.s6_addr32, tnl->saddr.v6, sizeof(tnl->saddr.v6)); + memcpy(ip6h->daddr.s6_addr32, tnl->daddr.v6, sizeof(tnl->daddr.v6)); + + count_tx(vip.protocol); + + return XDP_TX; +} + +SEC("xdp_tx_iptunnel") +int _xdp_tx_iptunnel(struct xdp_md *xdp) +{ + void *data_end = (void *)(long)xdp->data_end; + void *data = (void *)(long)xdp->data; + struct ethhdr *eth = data; + __u16 h_proto; + + if (eth + 1 > data_end) + return XDP_DROP; + + h_proto = eth->h_proto; + + if (h_proto == bpf_htons(ETH_P_IP)) + return handle_ipv4(xdp); + else if (h_proto == bpf_htons(ETH_P_IPV6)) + + return handle_ipv6(xdp); + else + return XDP_DROP; +} + +char _license[] SEC("license") = "GPL"; diff --git a/tools/testing/selftests/bpf/test_progs.c b/tools/testing/selftests/bpf/test_progs.c index 02d9e9eab033..c2b81f93895d 100644 --- a/tools/testing/selftests/bpf/test_progs.c +++ b/tools/testing/selftests/bpf/test_progs.c @@ -2374,7 +2374,7 @@ static int libbpf_debug_print_verifier_scale(enum libbpf_print_level level, const char *format, va_list args) { if (level != LIBBPF_DEBUG) - return 0; + return vfprintf(stderr, format, args); if (!strstr(format, "verifier log")) return 0; @@ -2400,26 +2400,71 @@ static int check_load(const char *file, enum bpf_prog_type type) void test_bpf_verif_scale(void) { - const char *scale[] = { - "./test_verif_scale1.o", "./test_verif_scale2.o", "./test_verif_scale3.o" + const char *sched_cls[] = { + "./test_verif_scale1.o", "./test_verif_scale2.o", "./test_verif_scale3.o", }; - const char *pyperf[] = { - "./pyperf50.o", "./pyperf100.o", "./pyperf180.o" + const char *raw_tp[] = { + /* full unroll by llvm */ + "./pyperf50.o", "./pyperf100.o", "./pyperf180.o", + + /* partial unroll. llvm will unroll loop ~150 times. + * C loop count -> 600. + * Asm loop count -> 4. + * 16k insns in loop body. + * Total of 5 such loops. Total program size ~82k insns. + */ + "./pyperf600.o", + + /* no unroll at all. + * C loop count -> 600. + * ASM loop count -> 600. + * ~110 insns in loop body. + * Total of 5 such loops. Total program size ~1500 insns. + */ + "./pyperf600_nounroll.o", + + "./loop1.o", "./loop2.o", + + /* partial unroll. 19k insn in a loop. + * Total program size 20.8k insn. + * ~350k processed_insns + */ + "./strobemeta.o", + + /* no unroll, tiny loops */ + "./strobemeta_nounroll1.o", + "./strobemeta_nounroll2.o", + }; + const char *cg_sysctl[] = { + "./test_sysctl_loop1.o", "./test_sysctl_loop2.o", }; int err, i; if (verifier_stats) libbpf_set_print(libbpf_debug_print_verifier_scale); - for (i = 0; i < ARRAY_SIZE(scale); i++) { - err = check_load(scale[i], BPF_PROG_TYPE_SCHED_CLS); - printf("test_scale:%s:%s\n", scale[i], err ? "FAIL" : "OK"); + err = check_load("./loop3.o", BPF_PROG_TYPE_RAW_TRACEPOINT); + printf("test_scale:loop3:%s\n", err ? (error_cnt--, "OK") : "FAIL"); + + for (i = 0; i < ARRAY_SIZE(sched_cls); i++) { + err = check_load(sched_cls[i], BPF_PROG_TYPE_SCHED_CLS); + printf("test_scale:%s:%s\n", sched_cls[i], err ? "FAIL" : "OK"); } - for (i = 0; i < ARRAY_SIZE(pyperf); i++) { - err = check_load(pyperf[i], BPF_PROG_TYPE_RAW_TRACEPOINT); - printf("test_scale:%s:%s\n", pyperf[i], err ? "FAIL" : "OK"); + for (i = 0; i < ARRAY_SIZE(raw_tp); i++) { + err = check_load(raw_tp[i], BPF_PROG_TYPE_RAW_TRACEPOINT); + printf("test_scale:%s:%s\n", raw_tp[i], err ? "FAIL" : "OK"); } + + for (i = 0; i < ARRAY_SIZE(cg_sysctl); i++) { + err = check_load(cg_sysctl[i], BPF_PROG_TYPE_CGROUP_SYSCTL); + printf("test_scale:%s:%s\n", cg_sysctl[i], err ? "FAIL" : "OK"); + } + err = check_load("./test_xdp_loop.o", BPF_PROG_TYPE_XDP); + printf("test_scale:test_xdp_loop:%s\n", err ? "FAIL" : "OK"); + + err = check_load("./test_seg6_loop.o", BPF_PROG_TYPE_LWT_SEG6LOCAL); + printf("test_scale:test_seg6_loop:%s\n", err ? "FAIL" : "OK"); } static void test_global_data_number(struct bpf_object *obj, __u32 duration) -- Gitee From 78098c5073a6f92d9375da3f74040f4bf7998f8f Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Sat, 15 Jun 2019 12:12:25 -0700 Subject: [PATCH 20/33] bpf: precise scalar_value tracking ANBZ: #5530 commit b5dc0163d8fd78e64a7e21f309cf932fda34353e upstream. Introduce precision tracking logic that helps cilium programs the most: old clang old clang new clang new clang with all patches with all patches bpf_lb-DLB_L3.o 1838 2283 1923 1863 bpf_lb-DLB_L4.o 3218 2657 3077 2468 bpf_lb-DUNKNOWN.o 1064 545 1062 544 bpf_lxc-DDROP_ALL.o 26935 23045 166729 22629 bpf_lxc-DUNKNOWN.o 34439 35240 174607 28805 bpf_netdev.o 9721 8753 8407 6801 bpf_overlay.o 6184 7901 5420 4754 bpf_lxc_jit.o 39389 50925 39389 50925 Consider code: 654: (85) call bpf_get_hash_recalc#34 655: (bf) r7 = r0 656: (15) if r8 == 0x0 goto pc+29 657: (bf) r2 = r10 658: (07) r2 += -48 659: (18) r1 = 0xffff8881e41e1b00 661: (85) call bpf_map_lookup_elem#1 662: (15) if r0 == 0x0 goto pc+23 663: (69) r1 = *(u16 *)(r0 +0) 664: (15) if r1 == 0x0 goto pc+21 665: (bf) r8 = r7 666: (57) r8 &= 65535 667: (bf) r2 = r8 668: (3f) r2 /= r1 669: (2f) r2 *= r1 670: (bf) r1 = r8 671: (1f) r1 -= r2 672: (57) r1 &= 255 673: (25) if r1 > 0x1e goto pc+12 R0=map_value(id=0,off=0,ks=20,vs=64,imm=0) R1_w=inv(id=0,umax_value=30,var_off=(0x0; 0x1f)) 674: (67) r1 <<= 1 675: (0f) r0 += r1 At this point the verifier will notice that scalar R1 is used in map pointer adjustment. R1 has to be precise for later operations on R0 to be validated properly. The verifier will backtrack the above code in the following way: last_idx 675 first_idx 664 regs=2 stack=0 before 675: (0f) r0 += r1 // started backtracking R1 regs=2 is a bitmask regs=2 stack=0 before 674: (67) r1 <<= 1 regs=2 stack=0 before 673: (25) if r1 > 0x1e goto pc+12 regs=2 stack=0 before 672: (57) r1 &= 255 regs=2 stack=0 before 671: (1f) r1 -= r2 // now both R1 and R2 has to be precise -> regs=6 mask regs=6 stack=0 before 670: (bf) r1 = r8 // after this insn R8 and R2 has to be precise regs=104 stack=0 before 669: (2f) r2 *= r1 // after this one R8, R2, and R1 regs=106 stack=0 before 668: (3f) r2 /= r1 regs=106 stack=0 before 667: (bf) r2 = r8 regs=102 stack=0 before 666: (57) r8 &= 65535 regs=102 stack=0 before 665: (bf) r8 = r7 regs=82 stack=0 before 664: (15) if r1 == 0x0 goto pc+21 // this is the end of verifier state. The following regs will be marked precised: R1_rw=invP(id=0,umax_value=65535,var_off=(0x0; 0xffff)) R7_rw=invP(id=0) parent didn't have regs=82 stack=0 marks // so backtracking continues into parent state last_idx 663 first_idx 655 regs=82 stack=0 before 663: (69) r1 = *(u16 *)(r0 +0) // R1 was assigned no need to track it further regs=80 stack=0 before 662: (15) if r0 == 0x0 goto pc+23 // keep tracking R7 regs=80 stack=0 before 661: (85) call bpf_map_lookup_elem#1 // keep tracking R7 regs=80 stack=0 before 659: (18) r1 = 0xffff8881e41e1b00 regs=80 stack=0 before 658: (07) r2 += -48 regs=80 stack=0 before 657: (bf) r2 = r10 regs=80 stack=0 before 656: (15) if r8 == 0x0 goto pc+29 regs=80 stack=0 before 655: (bf) r7 = r0 // here the assignment into R7 // mark R0 to be precise: R0_rw=invP(id=0) parent didn't have regs=1 stack=0 marks // regs=1 -> tracking R0 last_idx 654 first_idx 644 regs=1 stack=0 before 654: (85) call bpf_get_hash_recalc#34 // and in the parent frame it was a return value // nothing further to backtrack Two scalar registers not marked precise are equivalent from state pruning point of view. More details in the patch comments. It doesn't support bpf2bpf calls yet and enabled for root only. Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- include/linux/bpf_verifier.h | 18 ++ kernel/bpf/verifier.c | 491 ++++++++++++++++++++++++++++++++++- 2 files changed, 498 insertions(+), 11 deletions(-) diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h index 11a862bad2fb..16f957d08cc5 100644 --- a/include/linux/bpf_verifier.h +++ b/include/linux/bpf_verifier.h @@ -139,6 +139,8 @@ struct bpf_reg_state { */ s32 subreg_def; enum bpf_reg_liveness live; + /* if (!precise && SCALAR_VALUE) min/max/tnum don't affect safety */ + bool precise; }; enum bpf_stack_slot_type { @@ -190,6 +192,11 @@ struct bpf_func_state { struct bpf_stack_state *stack; }; +struct bpf_idx_pair { + u32 prev_idx; + u32 idx; +}; + #define MAX_CALL_FRAMES 8 struct bpf_verifier_state { /* call stack tracking */ @@ -245,6 +252,17 @@ struct bpf_verifier_state { u32 curframe; u32 active_spin_lock; bool speculative; + + /* first and last insn idx of this verifier state */ + u32 first_insn_idx; + u32 last_insn_idx; + /* jmp history recorded from first to last. + * backtracking is using it to go from last to first. + * For most states jmp_history_cnt is [0-3]. + * For loops can go up to ~40. + */ + struct bpf_idx_pair *jmp_history; + u32 jmp_history_cnt; }; #define bpf_get_spilled_reg(slot, frame) \ diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 338bc0598971..4a4683135cf4 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -454,12 +454,12 @@ static void print_verifier_state(struct bpf_verifier_env *env, verbose(env, " R%d", i); print_liveness(env, reg->live); verbose(env, "=%s", reg_type_str[t]); + if (t == SCALAR_VALUE && reg->precise) + verbose(env, "P"); if ((t == SCALAR_VALUE || t == PTR_TO_STACK) && tnum_is_const(reg->var_off)) { /* reg->off should be 0 for SCALAR_VALUE */ verbose(env, "%lld", reg->var_off.value + reg->off); - if (t == PTR_TO_STACK) - verbose(env, ",call_%d", func(env, reg)->callsite); } else { verbose(env, "(id=%d", reg->id); if (reg_type_may_be_refcounted_or_null(t)) @@ -521,11 +521,17 @@ static void print_verifier_state(struct bpf_verifier_env *env, continue; verbose(env, " fp%d", (-i - 1) * BPF_REG_SIZE); print_liveness(env, state->stack[i].spilled_ptr.live); - if (state->stack[i].slot_type[0] == STACK_SPILL) - verbose(env, "=%s", - reg_type_str[state->stack[i].spilled_ptr.type]); - else + if (state->stack[i].slot_type[0] == STACK_SPILL) { + reg = &state->stack[i].spilled_ptr; + t = reg->type; + verbose(env, "=%s", reg_type_str[t]); + if (t == SCALAR_VALUE && reg->precise) + verbose(env, "P"); + if (t == SCALAR_VALUE && tnum_is_const(reg->var_off)) + verbose(env, "%lld", reg->var_off.value + reg->off); + } else { verbose(env, "=%s", types_buf); + } } if (state->acquired_refs && state->refs[0].id) { verbose(env, " refs=%d", state->refs[0].id); @@ -674,6 +680,13 @@ static void free_func_state(struct bpf_func_state *state) kfree(state); } +static void clear_jmp_history(struct bpf_verifier_state *state) +{ + kfree(state->jmp_history); + state->jmp_history = NULL; + state->jmp_history_cnt = 0; +} + static void free_verifier_state(struct bpf_verifier_state *state, bool free_self) { @@ -683,6 +696,7 @@ static void free_verifier_state(struct bpf_verifier_state *state, free_func_state(state->frame[i]); state->frame[i] = NULL; } + clear_jmp_history(state); if (free_self) kfree(state); } @@ -710,8 +724,18 @@ static int copy_verifier_state(struct bpf_verifier_state *dst_state, const struct bpf_verifier_state *src) { struct bpf_func_state *dst; + u32 jmp_sz = sizeof(struct bpf_idx_pair) * src->jmp_history_cnt; int i, err; + if (dst_state->jmp_history_cnt < src->jmp_history_cnt) { + kfree(dst_state->jmp_history); + dst_state->jmp_history = kmalloc(jmp_sz, GFP_USER); + if (!dst_state->jmp_history) + return -ENOMEM; + } + memcpy(dst_state->jmp_history, src->jmp_history, jmp_sz); + dst_state->jmp_history_cnt = src->jmp_history_cnt; + /* if dst has more stack frames then src frame, free them */ for (i = src->curframe + 1; i <= dst_state->curframe; i++) { free_func_state(dst_state->frame[i]); @@ -722,6 +746,8 @@ static int copy_verifier_state(struct bpf_verifier_state *dst_state, dst_state->active_spin_lock = src->active_spin_lock; dst_state->branches = src->branches; dst_state->parent = src->parent; + dst_state->first_insn_idx = src->first_insn_idx; + dst_state->last_insn_idx = src->last_insn_idx; for (i = 0; i <= src->curframe; i++) { dst = dst_state->frame[i]; if (!dst) { @@ -966,6 +992,9 @@ static void __mark_reg_unbounded(struct bpf_reg_state *reg) reg->smax_value = S64_MAX; reg->umin_value = 0; reg->umax_value = U64_MAX; + + /* constant backtracking is enabled for root only for now */ + reg->precise = capable(CAP_SYS_ADMIN) ? false : true; } /* Mark a register as having a completely unknown (scalar) value. */ @@ -1377,6 +1406,389 @@ static int check_reg_arg(struct bpf_verifier_env *env, u32 regno, return 0; } +/* for any branch, call, exit record the history of jmps in the given state */ +static int push_jmp_history(struct bpf_verifier_env *env, + struct bpf_verifier_state *cur) +{ + u32 cnt = cur->jmp_history_cnt; + struct bpf_idx_pair *p; + + cnt++; + p = krealloc(cur->jmp_history, cnt * sizeof(*p), GFP_USER); + if (!p) + return -ENOMEM; + p[cnt - 1].idx = env->insn_idx; + p[cnt - 1].prev_idx = env->prev_insn_idx; + cur->jmp_history = p; + cur->jmp_history_cnt = cnt; + return 0; +} + +/* Backtrack one insn at a time. If idx is not at the top of recorded + * history then previous instruction came from straight line execution. + */ +static int get_prev_insn_idx(struct bpf_verifier_state *st, int i, + u32 *history) +{ + u32 cnt = *history; + + if (cnt && st->jmp_history[cnt - 1].idx == i) { + i = st->jmp_history[cnt - 1].prev_idx; + (*history)--; + } else { + i--; + } + return i; +} + +/* For given verifier state backtrack_insn() is called from the last insn to + * the first insn. Its purpose is to compute a bitmask of registers and + * stack slots that needs precision in the parent verifier state. + */ +static int backtrack_insn(struct bpf_verifier_env *env, int idx, + u32 *reg_mask, u64 *stack_mask) +{ + const struct bpf_insn_cbs cbs = { + .cb_print = verbose, + .private_data = env, + }; + struct bpf_insn *insn = env->prog->insnsi + idx; + u8 class = BPF_CLASS(insn->code); + u8 opcode = BPF_OP(insn->code); + u8 mode = BPF_MODE(insn->code); + u32 dreg = 1u << insn->dst_reg; + u32 sreg = 1u << insn->src_reg; + u32 spi; + + if (insn->code == 0) + return 0; + if (env->log.level & BPF_LOG_LEVEL) { + verbose(env, "regs=%x stack=%llx before ", *reg_mask, *stack_mask); + verbose(env, "%d: ", idx); + print_bpf_insn(&cbs, insn, env->allow_ptr_leaks); + } + + if (class == BPF_ALU || class == BPF_ALU64) { + if (!(*reg_mask & dreg)) + return 0; + if (opcode == BPF_MOV) { + if (BPF_SRC(insn->code) == BPF_X) { + /* dreg = sreg + * dreg needs precision after this insn + * sreg needs precision before this insn + */ + *reg_mask &= ~dreg; + *reg_mask |= sreg; + } else { + /* dreg = K + * dreg needs precision after this insn. + * Corresponding register is already marked + * as precise=true in this verifier state. + * No further markings in parent are necessary + */ + *reg_mask &= ~dreg; + } + } else { + if (BPF_SRC(insn->code) == BPF_X) { + /* dreg += sreg + * both dreg and sreg need precision + * before this insn + */ + *reg_mask |= sreg; + } /* else dreg += K + * dreg still needs precision before this insn + */ + } + } else if (class == BPF_LDX) { + if (!(*reg_mask & dreg)) + return 0; + *reg_mask &= ~dreg; + + /* scalars can only be spilled into stack w/o losing precision. + * Load from any other memory can be zero extended. + * The desire to keep that precision is already indicated + * by 'precise' mark in corresponding register of this state. + * No further tracking necessary. + */ + if (insn->src_reg != BPF_REG_FP) + return 0; + if (BPF_SIZE(insn->code) != BPF_DW) + return 0; + + /* dreg = *(u64 *)[fp - off] was a fill from the stack. + * that [fp - off] slot contains scalar that needs to be + * tracked with precision + */ + spi = (-insn->off - 1) / BPF_REG_SIZE; + if (spi >= 64) { + verbose(env, "BUG spi %d\n", spi); + WARN_ONCE(1, "verifier backtracking bug"); + return -EFAULT; + } + *stack_mask |= 1ull << spi; + } else if (class == BPF_STX) { + if (*reg_mask & dreg) + /* stx shouldn't be using _scalar_ dst_reg + * to access memory. It means backtracking + * encountered a case of pointer subtraction. + */ + return -ENOTSUPP; + /* scalars can only be spilled into stack */ + if (insn->dst_reg != BPF_REG_FP) + return 0; + if (BPF_SIZE(insn->code) != BPF_DW) + return 0; + spi = (-insn->off - 1) / BPF_REG_SIZE; + if (spi >= 64) { + verbose(env, "BUG spi %d\n", spi); + WARN_ONCE(1, "verifier backtracking bug"); + return -EFAULT; + } + if (!(*stack_mask & (1ull << spi))) + return 0; + *stack_mask &= ~(1ull << spi); + *reg_mask |= sreg; + } else if (class == BPF_JMP || class == BPF_JMP32) { + if (opcode == BPF_CALL) { + if (insn->src_reg == BPF_PSEUDO_CALL) + return -ENOTSUPP; + /* regular helper call sets R0 */ + *reg_mask &= ~1; + if (*reg_mask & 0x3f) { + /* if backtracing was looking for registers R1-R5 + * they should have been found already. + */ + verbose(env, "BUG regs %x\n", *reg_mask); + WARN_ONCE(1, "verifier backtracking bug"); + return -EFAULT; + } + } else if (opcode == BPF_EXIT) { + return -ENOTSUPP; + } + } else if (class == BPF_LD) { + if (!(*reg_mask & dreg)) + return 0; + *reg_mask &= ~dreg; + /* It's ld_imm64 or ld_abs or ld_ind. + * For ld_imm64 no further tracking of precision + * into parent is necessary + */ + if (mode == BPF_IND || mode == BPF_ABS) + /* to be analyzed */ + return -ENOTSUPP; + } else if (class == BPF_ST) { + if (*reg_mask & dreg) + /* likely pointer subtraction */ + return -ENOTSUPP; + } + return 0; +} + +/* the scalar precision tracking algorithm: + * . at the start all registers have precise=false. + * . scalar ranges are tracked as normal through alu and jmp insns. + * . once precise value of the scalar register is used in: + * . ptr + scalar alu + * . if (scalar cond K|scalar) + * . helper_call(.., scalar, ...) where ARG_CONST is expected + * backtrack through the verifier states and mark all registers and + * stack slots with spilled constants that these scalar regisers + * should be precise. + * . during state pruning two registers (or spilled stack slots) + * are equivalent if both are not precise. + * + * Note the verifier cannot simply walk register parentage chain, + * since many different registers and stack slots could have been + * used to compute single precise scalar. + * + * The approach of starting with precise=true for all registers and then + * backtrack to mark a register as not precise when the verifier detects + * that program doesn't care about specific value (e.g., when helper + * takes register as ARG_ANYTHING parameter) is not safe. + * + * It's ok to walk single parentage chain of the verifier states. + * It's possible that this backtracking will go all the way till 1st insn. + * All other branches will be explored for needing precision later. + * + * The backtracking needs to deal with cases like: + * R8=map_value(id=0,off=0,ks=4,vs=1952,imm=0) R9_w=map_value(id=0,off=40,ks=4,vs=1952,imm=0) + * r9 -= r8 + * r5 = r9 + * if r5 > 0x79f goto pc+7 + * R5_w=inv(id=0,umax_value=1951,var_off=(0x0; 0x7ff)) + * r5 += 1 + * ... + * call bpf_perf_event_output#25 + * where .arg5_type = ARG_CONST_SIZE_OR_ZERO + * + * and this case: + * r6 = 1 + * call foo // uses callee's r6 inside to compute r0 + * r0 += r6 + * if r0 == 0 goto + * + * to track above reg_mask/stack_mask needs to be independent for each frame. + * + * Also if parent's curframe > frame where backtracking started, + * the verifier need to mark registers in both frames, otherwise callees + * may incorrectly prune callers. This is similar to + * commit 7640ead93924 ("bpf: verifier: make sure callees don't prune with caller differences") + * + * For now backtracking falls back into conservative marking. + */ +static void mark_all_scalars_precise(struct bpf_verifier_env *env, + struct bpf_verifier_state *st) +{ + struct bpf_func_state *func; + struct bpf_reg_state *reg; + int i, j; + + /* big hammer: mark all scalars precise in this path. + * pop_stack may still get !precise scalars. + */ + for (; st; st = st->parent) + for (i = 0; i <= st->curframe; i++) { + func = st->frame[i]; + for (j = 0; j < BPF_REG_FP; j++) { + reg = &func->regs[j]; + if (reg->type != SCALAR_VALUE) + continue; + reg->precise = true; + } + for (j = 0; j < func->allocated_stack / BPF_REG_SIZE; j++) { + if (func->stack[j].slot_type[0] != STACK_SPILL) + continue; + reg = &func->stack[j].spilled_ptr; + if (reg->type != SCALAR_VALUE) + continue; + reg->precise = true; + } + } +} + +static int mark_chain_precision(struct bpf_verifier_env *env, int regno) +{ + struct bpf_verifier_state *st = env->cur_state; + int first_idx = st->first_insn_idx; + int last_idx = env->insn_idx; + struct bpf_func_state *func; + struct bpf_reg_state *reg; + u32 reg_mask = 1u << regno; + u64 stack_mask = 0; + bool skip_first = true; + int i, err; + + if (!env->allow_ptr_leaks) + /* backtracking is root only for now */ + return 0; + + func = st->frame[st->curframe]; + reg = &func->regs[regno]; + if (reg->type != SCALAR_VALUE) { + WARN_ONCE(1, "backtracing misuse"); + return -EFAULT; + } + if (reg->precise) + return 0; + func->regs[regno].precise = true; + + for (;;) { + DECLARE_BITMAP(mask, 64); + bool new_marks = false; + u32 history = st->jmp_history_cnt; + + if (env->log.level & BPF_LOG_LEVEL) + verbose(env, "last_idx %d first_idx %d\n", last_idx, first_idx); + for (i = last_idx;;) { + if (skip_first) { + err = 0; + skip_first = false; + } else { + err = backtrack_insn(env, i, ®_mask, &stack_mask); + } + if (err == -ENOTSUPP) { + mark_all_scalars_precise(env, st); + return 0; + } else if (err) { + return err; + } + if (!reg_mask && !stack_mask) + /* Found assignment(s) into tracked register in this state. + * Since this state is already marked, just return. + * Nothing to be tracked further in the parent state. + */ + return 0; + if (i == first_idx) + break; + i = get_prev_insn_idx(st, i, &history); + if (i >= env->prog->len) { + /* This can happen if backtracking reached insn 0 + * and there are still reg_mask or stack_mask + * to backtrack. + * It means the backtracking missed the spot where + * particular register was initialized with a constant. + */ + verbose(env, "BUG backtracking idx %d\n", i); + WARN_ONCE(1, "verifier backtracking bug"); + return -EFAULT; + } + } + st = st->parent; + if (!st) + break; + + func = st->frame[st->curframe]; + bitmap_from_u64(mask, reg_mask); + for_each_set_bit(i, mask, 32) { + reg = &func->regs[i]; + if (reg->type != SCALAR_VALUE) + continue; + if (!reg->precise) + new_marks = true; + reg->precise = true; + } + + bitmap_from_u64(mask, stack_mask); + for_each_set_bit(i, mask, 64) { + if (i >= func->allocated_stack / BPF_REG_SIZE) { + /* This can happen if backtracking + * is propagating stack precision where + * caller has larger stack frame + * than callee, but backtrack_insn() should + * have returned -ENOTSUPP. + */ + verbose(env, "BUG spi %d stack_size %d\n", + i, func->allocated_stack); + WARN_ONCE(1, "verifier backtracking bug"); + return -EFAULT; + } + + if (func->stack[i].slot_type[0] != STACK_SPILL) + continue; + reg = &func->stack[i].spilled_ptr; + if (reg->type != SCALAR_VALUE) + continue; + if (!reg->precise) + new_marks = true; + reg->precise = true; + } + if (env->log.level & BPF_LOG_LEVEL) { + print_verifier_state(env, func); + verbose(env, "parent %s regs=%x stack=%llx marks\n", + new_marks ? "didn't have" : "already had", + reg_mask, stack_mask); + } + + if (!new_marks) + break; + + last_idx = st->last_insn_idx; + first_idx = st->first_insn_idx; + } + return 0; +} + + static bool is_spillable_regtype(enum bpf_reg_type type) { switch (type) { @@ -1434,6 +1846,7 @@ static int check_stack_write(struct bpf_verifier_env *env, { struct bpf_func_state *cur; /* state of the current function */ int i, slot = -off - 1, spi = slot / BPF_REG_SIZE, err; + u32 dst_reg = env->prog->insnsi[insn_idx].dst_reg; struct bpf_reg_state *reg = NULL; err = realloc_func_state(state, round_up(slot + 1, BPF_REG_SIZE), @@ -1456,6 +1869,17 @@ static int check_stack_write(struct bpf_verifier_env *env, if (reg && size == BPF_REG_SIZE && register_is_const(reg) && !register_is_null(reg) && env->allow_ptr_leaks) { + if (dst_reg != BPF_REG_FP) { + /* The backtracking logic can only recognize explicit + * stack slot address like [fp - 8]. Other spill of + * scalar via different register has to be conervative. + * Backtrack from here and mark all registers as precise + * that contributed into 'reg' being a constant. + */ + err = mark_chain_precision(env, value_regno); + if (err) + return err; + } save_register_state(state, spi, reg); } else if (reg && is_spillable_regtype(reg->type)) { /* register containing pointer is being spilled into stack */ @@ -1524,8 +1948,13 @@ static int check_stack_write(struct bpf_verifier_env *env, state->stack[spi].spilled_ptr.live |= REG_LIVE_WRITTEN; /* when we zero initialize stack slots mark them as such */ - if (reg && register_is_null(reg)) + if (reg && register_is_null(reg)) { + /* backtracking doesn't work for STACK_ZERO yet. */ + err = mark_chain_precision(env, value_regno); + if (err) + return err; type = STACK_ZERO; + } for (i = 0; i < size; i++) state->stack[spi].slot_type[(slot - i) % BPF_REG_SIZE] = @@ -1604,6 +2033,17 @@ static int check_stack_read(struct bpf_verifier_env *env, * so the whole register == const_zero */ __mark_reg_const_zero(&state->regs[value_regno]); + /* backtracking doesn't support STACK_ZERO yet, + * so mark it precise here, so that later + * backtracking can stop here. + * Backtracking may not need this if this register + * doesn't participate in pointer adjustment. + * Forward propagation of precise flag is not + * necessary either. This mark is only to stop + * backtracking. Any register that contributed + * to const 0 was marked precise before spill. + */ + state->regs[value_regno].precise = true; } else { /* have read misc data from the stack */ mark_reg_unknown(env, state->regs, value_regno); @@ -2918,6 +3358,8 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 regno, err = check_helper_mem_access(env, regno - 1, reg->umax_value, zero_size_allowed, meta); + if (!err) + err = mark_chain_precision(env, regno); } else if (arg_type_is_int_ptr(arg_type)) { int size = int_ptr_type_to_size(arg_type); @@ -4521,6 +4963,7 @@ static int adjust_reg_min_max_vals(struct bpf_verifier_env *env, struct bpf_reg_state *regs = state->regs, *dst_reg, *src_reg; struct bpf_reg_state *ptr_reg = NULL, off_reg = {0}; u8 opcode = BPF_OP(insn->code); + int err; dst_reg = ®s[insn->dst_reg]; src_reg = NULL; @@ -4547,11 +4990,17 @@ static int adjust_reg_min_max_vals(struct bpf_verifier_env *env, * This is legal, but we have to reverse our * src/dest handling in computing the range */ + err = mark_chain_precision(env, insn->dst_reg); + if (err) + return err; return adjust_ptr_min_max_vals(env, insn, src_reg, dst_reg); } } else if (ptr_reg) { /* pointer += scalar */ + err = mark_chain_precision(env, insn->src_reg); + if (err) + return err; return adjust_ptr_min_max_vals(env, insn, dst_reg, src_reg); } @@ -5508,6 +5957,13 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, pred = is_branch_taken(dst_reg, src_reg->var_off.value, opcode, is_jmp32); + if (pred >= 0) { + err = mark_chain_precision(env, insn->dst_reg); + if (BPF_SRC(insn->code) == BPF_X && !err) + err = mark_chain_precision(env, insn->src_reg); + if (err) + return err; + } if (pred == 1) { /* Only follow the goto, ignore fall-through. If needed, push * the fall-through branch for simulation under speculative @@ -6001,6 +6457,11 @@ static int check_cfg(struct bpf_verifier_env *env) goto peek_stack; else if (ret < 0) goto err_free; + /* unconditional jmp is not a good pruning point, + * but it's marked, since backtracking needs + * to record jmp history in is_state_visited(). + */ + init_explored_state(env, t + insns[t].off + 1); /* tell verifier to check for equivalent states * after every call and jump */ @@ -6501,6 +6962,8 @@ static bool regsafe(struct bpf_reg_state *rold, struct bpf_reg_state *rcur, switch (rold->type) { case SCALAR_VALUE: if (rcur->type == SCALAR_VALUE) { + if (!rold->precise && !rcur->precise) + return true; /* new val must satisfy old val knowledge */ return range_within(rold, rcur) && tnum_in(rold->var_off, rcur->var_off); @@ -6851,6 +7314,7 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) int i, j, err, states_cnt = 0; bool add_new_state = false; + cur->last_insn_idx = env->prev_insn_idx; if (!env->insn_aux_data[insn_idx].prune_point) /* this 'insn_idx' instruction wasn't marked, so we will not * be doing state search here @@ -6967,10 +7431,10 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) env->max_states_per_insn = states_cnt; if (!env->allow_ptr_leaks && states_cnt > BPF_COMPLEXITY_LIMIT_STATES) - return 0; + return push_jmp_history(env, cur); if (!add_new_state) - return 0; + return push_jmp_history(env, cur); /* There were no equivalent states, remember the current one. * Technically the current state is not proven to be safe yet, @@ -7000,7 +7464,10 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) new->insn_idx = insn_idx; WARN_ONCE(new->branches != 1, "BUG is_state_visited:branches_to_explore=%d insn %d\n", new->branches, insn_idx); + cur->parent = new; + cur->first_insn_idx = insn_idx; + clear_jmp_history(cur); new_sl->next = *explored_state(env, insn_idx); *explored_state(env, insn_idx) = new_sl; /* connect new state to parentage chain. Current frame needs all @@ -7080,6 +7547,7 @@ static int do_check(struct bpf_verifier_env *env) struct bpf_reg_state *regs; int insn_cnt = env->prog->len; bool do_print_state = false; + int prev_insn_idx = -1; env->prev_linfo = NULL; @@ -7105,6 +7573,7 @@ static int do_check(struct bpf_verifier_env *env) u8 class; int err; + env->prev_insn_idx = prev_insn_idx; if (env->insn_idx >= insn_cnt) { verbose(env, "invalid insn idx %d insn_cnt %d\n", env->insn_idx, insn_cnt); @@ -7177,6 +7646,7 @@ static int do_check(struct bpf_verifier_env *env) regs = cur_regs(env); sanitize_mark_insn_seen(env); + prev_insn_idx = env->insn_idx; if (class == BPF_ALU || class == BPF_ALU64) { err = check_alu_op(env, insn); @@ -7350,7 +7820,6 @@ static int do_check(struct bpf_verifier_env *env) if (state->curframe) { /* exit from nested function */ - env->prev_insn_idx = env->insn_idx; err = prepare_func_exit(env, &env->insn_idx); if (err) return err; @@ -7382,7 +7851,7 @@ static int do_check(struct bpf_verifier_env *env) return err; process_bpf_exit: update_branch_counts(env, env->cur_state); - err = pop_stack(env, &env->prev_insn_idx, + err = pop_stack(env, &prev_insn_idx, &env->insn_idx); if (err < 0) { if (err != -ENOENT) -- Gitee From 3eb55eacbd83be970e8ac90a29f8c2d4e5997d48 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Fri, 28 Jun 2019 09:24:09 -0700 Subject: [PATCH 21/33] bpf: fix precision tracking ANBZ: #5530 commit a3ce685dd01a786fa5bc388e47d0066a4f842591 upstream. When equivalent state is found the current state needs to propagate precision marks. Otherwise the verifier will prune the search incorrectly. There is a price for correctness: before before broken fixed cnst spill precise precise bpf_lb-DLB_L3.o 1923 8128 1863 1898 bpf_lb-DLB_L4.o 3077 6707 2468 2666 bpf_lb-DUNKNOWN.o 1062 1062 544 544 bpf_lxc-DDROP_ALL.o 166729 380712 22629 36823 bpf_lxc-DUNKNOWN.o 174607 440652 28805 45325 bpf_netdev.o 8407 31904 6801 7002 bpf_overlay.o 5420 23569 4754 4858 bpf_lxc_jit.o 39389 359445 50925 69631 Overall precision tracking is still very effective. Fixes: b5dc0163d8fd ("bpf: precise scalar_value tracking") Reported-by: Lawrence Brakmo Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Tested-by: Lawrence Brakmo Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 121 +++++++++++++++++++++++++++++++++++++----- 1 file changed, 107 insertions(+), 14 deletions(-) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 4a4683135cf4..fad2ac5a561e 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -1666,16 +1666,18 @@ static void mark_all_scalars_precise(struct bpf_verifier_env *env, } } -static int mark_chain_precision(struct bpf_verifier_env *env, int regno) +static int __mark_chain_precision(struct bpf_verifier_env *env, int regno, + int spi) { struct bpf_verifier_state *st = env->cur_state; int first_idx = st->first_insn_idx; int last_idx = env->insn_idx; struct bpf_func_state *func; struct bpf_reg_state *reg; - u32 reg_mask = 1u << regno; - u64 stack_mask = 0; + u32 reg_mask = regno >= 0 ? 1u << regno : 0; + u64 stack_mask = spi >= 0 ? 1ull << spi : 0; bool skip_first = true; + bool new_marks = false; int i, err; if (!env->allow_ptr_leaks) @@ -1683,18 +1685,43 @@ static int mark_chain_precision(struct bpf_verifier_env *env, int regno) return 0; func = st->frame[st->curframe]; - reg = &func->regs[regno]; - if (reg->type != SCALAR_VALUE) { - WARN_ONCE(1, "backtracing misuse"); - return -EFAULT; + if (regno >= 0) { + reg = &func->regs[regno]; + if (reg->type != SCALAR_VALUE) { + WARN_ONCE(1, "backtracing misuse"); + return -EFAULT; + } + if (!reg->precise) + new_marks = true; + else + reg_mask = 0; + reg->precise = true; } - if (reg->precise) - return 0; - func->regs[regno].precise = true; + while (spi >= 0) { + if (func->stack[spi].slot_type[0] != STACK_SPILL) { + stack_mask = 0; + break; + } + reg = &func->stack[spi].spilled_ptr; + if (reg->type != SCALAR_VALUE) { + stack_mask = 0; + break; + } + if (!reg->precise) + new_marks = true; + else + stack_mask = 0; + reg->precise = true; + break; + } + + if (!new_marks) + return 0; + if (!reg_mask && !stack_mask) + return 0; for (;;) { DECLARE_BITMAP(mask, 64); - bool new_marks = false; u32 history = st->jmp_history_cnt; if (env->log.level & BPF_LOG_LEVEL) @@ -1737,12 +1764,15 @@ static int mark_chain_precision(struct bpf_verifier_env *env, int regno) if (!st) break; + new_marks = false; func = st->frame[st->curframe]; bitmap_from_u64(mask, reg_mask); for_each_set_bit(i, mask, 32) { reg = &func->regs[i]; - if (reg->type != SCALAR_VALUE) + if (reg->type != SCALAR_VALUE) { + reg_mask &= ~(1u << i); continue; + } if (!reg->precise) new_marks = true; reg->precise = true; @@ -1763,11 +1793,15 @@ static int mark_chain_precision(struct bpf_verifier_env *env, int regno) return -EFAULT; } - if (func->stack[i].slot_type[0] != STACK_SPILL) + if (func->stack[i].slot_type[0] != STACK_SPILL) { + stack_mask &= ~(1ull << i); continue; + } reg = &func->stack[i].spilled_ptr; - if (reg->type != SCALAR_VALUE) + if (reg->type != SCALAR_VALUE) { + stack_mask &= ~(1ull << i); continue; + } if (!reg->precise) new_marks = true; reg->precise = true; @@ -1779,6 +1813,8 @@ static int mark_chain_precision(struct bpf_verifier_env *env, int regno) reg_mask, stack_mask); } + if (!reg_mask && !stack_mask) + break; if (!new_marks) break; @@ -1788,6 +1824,15 @@ static int mark_chain_precision(struct bpf_verifier_env *env, int regno) return 0; } +static int mark_chain_precision(struct bpf_verifier_env *env, int regno) +{ + return __mark_chain_precision(env, regno, -1); +} + +static int mark_chain_precision_stack(struct bpf_verifier_env *env, int spi) +{ + return __mark_chain_precision(env, -1, spi); +} static bool is_spillable_regtype(enum bpf_reg_type type) { @@ -7287,6 +7332,46 @@ static int propagate_liveness(struct bpf_verifier_env *env, return 0; } +/* find precise scalars in the previous equivalent state and + * propagate them into the current state + */ +static int propagate_precision(struct bpf_verifier_env *env, + const struct bpf_verifier_state *old) +{ + struct bpf_reg_state *state_reg; + struct bpf_func_state *state; + int i, err = 0; + + state = old->frame[old->curframe]; + state_reg = state->regs; + for (i = 0; i < BPF_REG_FP; i++, state_reg++) { + if (state_reg->type != SCALAR_VALUE || + !state_reg->precise) + continue; + if (env->log.level & BPF_LOG_LEVEL2) + verbose(env, "propagating r%d\n", i); + err = mark_chain_precision(env, i); + if (err < 0) + return err; + } + + for (i = 0; i < state->allocated_stack / BPF_REG_SIZE; i++) { + if (state->stack[i].slot_type[0] != STACK_SPILL) + continue; + state_reg = &state->stack[i].spilled_ptr; + if (state_reg->type != SCALAR_VALUE || + !state_reg->precise) + continue; + if (env->log.level & BPF_LOG_LEVEL2) + verbose(env, "propagating fp%d\n", + (-i - 1) * BPF_REG_SIZE); + err = mark_chain_precision_stack(env, i); + if (err < 0) + return err; + } + return 0; +} + static bool states_maybe_looping(struct bpf_verifier_state *old, struct bpf_verifier_state *cur) { @@ -7379,6 +7464,14 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) * this state and will pop a new one. */ err = propagate_liveness(env, &sl->state, cur); + + /* if previous state reached the exit with precision and + * current state is equivalent to it (except precsion marks) + * the precision needs to be propagated back in + * the current state. + */ + err = err ? : push_jmp_history(env, cur); + err = err ? : propagate_precision(env, &sl->state); if (err) return err; return 1; -- Gitee From 9223eb6652897df67d48f80961f0c8035f005959 Mon Sep 17 00:00:00 2001 From: Andrii Nakryiko Date: Mon, 8 Jul 2019 20:32:44 -0700 Subject: [PATCH 22/33] bpf: fix precision bit propagation for BPF_ST instructions ANBZ: #5530 commit b3b50f05dc501cc2cd90349a7bbfd932af0ceb31 upstream. When backtracking instructions to propagate precision bit for registers and stack slots, one class of instructions (BPF_ST) weren't handled causing extra stack slots to be propagated into parent state. Parent state might not have that much stack allocated, though, which causes warning on invalid stack slot usage. This patch adds handling of BPF_ST instructions: BPF_MEM | | BPF_ST: *(size *) (dst_reg + off) = imm32 Reported-by: syzbot+4da3ff23081bafe74fc2@syzkaller.appspotmail.com Fixes: b5dc0163d8fd ("bpf: precise scalar_value tracking") Cc: Alexei Starovoitov Signed-off-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index fad2ac5a561e..0587935bdeaa 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -1526,9 +1526,9 @@ static int backtrack_insn(struct bpf_verifier_env *env, int idx, return -EFAULT; } *stack_mask |= 1ull << spi; - } else if (class == BPF_STX) { + } else if (class == BPF_STX || class == BPF_ST) { if (*reg_mask & dreg) - /* stx shouldn't be using _scalar_ dst_reg + /* stx & st shouldn't be using _scalar_ dst_reg * to access memory. It means backtracking * encountered a case of pointer subtraction. */ @@ -1547,7 +1547,8 @@ static int backtrack_insn(struct bpf_verifier_env *env, int idx, if (!(*stack_mask & (1ull << spi))) return 0; *stack_mask &= ~(1ull << spi); - *reg_mask |= sreg; + if (class == BPF_STX) + *reg_mask |= sreg; } else if (class == BPF_JMP || class == BPF_JMP32) { if (opcode == BPF_CALL) { if (insn->src_reg == BPF_PSEUDO_CALL) @@ -1576,10 +1577,6 @@ static int backtrack_insn(struct bpf_verifier_env *env, int idx, if (mode == BPF_IND || mode == BPF_ABS) /* to be analyzed */ return -ENOTSUPP; - } else if (class == BPF_ST) { - if (*reg_mask & dreg) - /* likely pointer subtraction */ - return -ENOTSUPP; } return 0; } -- Gitee From 318dcdbadb812704b593c00ed2588c225eac4f7b Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Tue, 30 Jul 2019 18:38:26 -0700 Subject: [PATCH 23/33] bpf: fix x64 JIT code generation for jmp to 1st insn ANBZ: #5530 commit 7c2e988f400e83501e0a3568250780609b7c8263 upstream. Introduction of bounded loops exposed old bug in x64 JIT. JIT maintains the array of offsets to the end of all instructions to compute jmp offsets. addrs[0] - offset of the end of the 1st insn (that includes prologue). addrs[1] - offset of the end of the 2nd insn. JIT didn't keep the offset of the beginning of the 1st insn, since classic BPF didn't have backward jumps and valid extended BPF couldn't have a branch to 1st insn, because it didn't allow loops. With bounded loops it's possible to construct a valid program that jumps backwards to the 1st insn. Fix JIT by computing: addrs[0] - offset of the end of prologue == start of the 1st insn. addrs[1] - offset of the end of 1st insn. v1->v2: - Yonghong noticed a bug in jit linfo. Fix it by passing 'addrs + 1' to bpf_prog_fill_jited_linfo(), since it expects insn_to_jit_off array to be offsets to last byte. Reported-by: syzbot+35101610ff3e83119b1b@syzkaller.appspotmail.com Fixes: 2589726d12a1 ("bpf: introduce bounded loops") Fixes: 0a14842f5a3c ("net: filter: Just In Time compiler for x86-64") Signed-off-by: Alexei Starovoitov Acked-by: Song Liu Signed-off-by: Qiao Ma --- arch/x86/net/bpf_jit_comp.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c index 21faf6d3cc68..0f3125fc4188 100644 --- a/arch/x86/net/bpf_jit_comp.c +++ b/arch/x86/net/bpf_jit_comp.c @@ -394,8 +394,9 @@ static int do_jit(struct bpf_prog *bpf_prog, int *addrs, u8 *image, emit_prologue(&prog, bpf_prog->aux->stack_depth, bpf_prog_was_classic(bpf_prog)); + addrs[0] = prog - temp; - for (i = 0; i < insn_cnt; i++, insn++) { + for (i = 1; i <= insn_cnt; i++, insn++) { const s32 imm32 = insn->imm; u32 dst_reg = insn->dst_reg; u32 src_reg = insn->src_reg; @@ -1130,7 +1131,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) extra_pass = true; goto skip_init_addrs; } - addrs = kmalloc_array(prog->len, sizeof(*addrs), GFP_KERNEL); + addrs = kmalloc_array(prog->len + 1, sizeof(*addrs), GFP_KERNEL); if (!addrs) { prog = orig_prog; goto out_addrs; @@ -1140,7 +1141,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) * Before first pass, make a rough estimation of addrs[] * each BPF instruction is translated to less than 64 bytes */ - for (proglen = 0, i = 0; i < prog->len; i++) { + for (proglen = 0, i = 0; i <= prog->len; i++) { proglen += 64; addrs[i] = proglen; } @@ -1205,7 +1206,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) if (!image || !prog->is_func || extra_pass) { if (image) - bpf_prog_fill_jited_linfo(prog, addrs); + bpf_prog_fill_jited_linfo(prog, addrs + 1); out_addrs: kfree(addrs); kfree(jit_data); -- Gitee From 38b3d64e911b8cf3db5ed91050c4e0c0000d90c5 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Wed, 21 Aug 2019 14:07:10 -0700 Subject: [PATCH 24/33] bpf: fix precision tracking in presence of bpf2bpf calls ANBZ: #5530 commit 6754172c208d9d3dae208c6494611ac167d56688 upstream. While adding extra tests for precision tracking and extra infra to adjust verifier heuristics the existing test "calls: cross frame pruning - liveness propagation" started to fail. The root cause is the same as described in verifer.c comment: * Also if parent's curframe > frame where backtracking started, * the verifier need to mark registers in both frames, otherwise callees * may incorrectly prune callers. This is similar to * commit 7640ead93924 ("bpf: verifier: make sure callees don't prune with caller differences") * For now backtracking falls back into conservative marking. Turned out though that returning -ENOTSUPP from backtrack_insn() and doing mark_all_scalars_precise() in the current parentage chain is not enough. Depending on how is_state_visited() heuristic is creating parentage chain it's possible that callee will incorrectly prune caller. Fix the issue by setting precise=true earlier and more aggressively. Before this fix the precision tracking _within_ functions that don't do bpf2bpf calls would still work. Whereas now precision tracking is completely disabled when bpf2bpf calls are present anywhere in the program. No difference in cilium tests (they don't have bpf2bpf calls). No difference in test_progs though some of them have bpf2bpf calls, but precision tracking wasn't effective there. Fixes: b5dc0163d8fd ("bpf: precise scalar_value tracking") Signed-off-by: Alexei Starovoitov Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 0587935bdeaa..732b041869d6 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -992,9 +992,6 @@ static void __mark_reg_unbounded(struct bpf_reg_state *reg) reg->smax_value = S64_MAX; reg->umin_value = 0; reg->umax_value = U64_MAX; - - /* constant backtracking is enabled for root only for now */ - reg->precise = capable(CAP_SYS_ADMIN) ? false : true; } /* Mark a register as having a completely unknown (scalar) value. */ @@ -1021,7 +1018,11 @@ static void mark_reg_unknown(struct bpf_verifier_env *env, __mark_reg_not_init(regs + regno); return; } - __mark_reg_unknown(regs + regno); + regs += regno; + __mark_reg_unknown(regs); + /* constant backtracking is enabled for root without bpf2bpf calls */ + regs->precise = env->subprog_cnt > 1 || !env->allow_ptr_leaks ? + true : false; } static void __mark_reg_not_init(struct bpf_reg_state *reg) -- Gitee From abebb5c84cdfca47daabb44f1128bb61a4e1757d Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Tue, 3 Sep 2019 15:16:17 -0700 Subject: [PATCH 25/33] bpf: fix precision tracking of stack slots ANBZ: #5530 commit 2339cd6cd0b5401fa3fe886bf1c0cb8822041957 upstream. The problem can be seen in the following two tests: 0: (bf) r3 = r10 1: (55) if r3 != 0x7b goto pc+0 2: (7a) *(u64 *)(r3 -8) = 0 3: (79) r4 = *(u64 *)(r10 -8) .. 0: (85) call bpf_get_prandom_u32#7 1: (bf) r3 = r10 2: (55) if r3 != 0x7b goto pc+0 3: (7b) *(u64 *)(r3 -8) = r0 4: (79) r4 = *(u64 *)(r10 -8) When backtracking need to mark R4 it will mark slot fp-8. But ST or STX into fp-8 could belong to the same block of instructions. When backtracing is done the parent state may have fp-8 slot as "unallocated stack". Which will cause verifier to warn and incorrectly reject such programs. Writes into stack via non-R10 register are rare. llvm always generates canonical stack spill/fill. For such pathological case fall back to conservative precision tracking instead of rejecting. Reported-by: syzbot+c8d66267fd2b5955287e@syzkaller.appspotmail.com Fixes: b5dc0163d8fd ("bpf: precise scalar_value tracking") Signed-off-by: Alexei Starovoitov Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 23 ++++++++++++++--------- 1 file changed, 14 insertions(+), 9 deletions(-) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 732b041869d6..a2f5289c345d 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -1779,16 +1779,21 @@ static int __mark_chain_precision(struct bpf_verifier_env *env, int regno, bitmap_from_u64(mask, stack_mask); for_each_set_bit(i, mask, 64) { if (i >= func->allocated_stack / BPF_REG_SIZE) { - /* This can happen if backtracking - * is propagating stack precision where - * caller has larger stack frame - * than callee, but backtrack_insn() should - * have returned -ENOTSUPP. + /* the sequence of instructions: + * 2: (bf) r3 = r10 + * 3: (7b) *(u64 *)(r3 -8) = r0 + * 4: (79) r4 = *(u64 *)(r10 -8) + * doesn't contain jmps. It's backtracked + * as a single block. + * During backtracking insn 3 is not recognized as + * stack access, so at the end of backtracking + * stack slot fp-8 is still marked in stack_mask. + * However the parent state may not have accessed + * fp-8 and it's "unallocated" stack space. + * In such case fallback to conservative. */ - verbose(env, "BUG spi %d stack_size %d\n", - i, func->allocated_stack); - WARN_ONCE(1, "verifier backtracking bug"); - return -EFAULT; + mark_all_scalars_precise(env, st); + return 0; } if (func->stack[i].slot_type[0] != STACK_SPILL) { -- Gitee From 23b6eeb76ef57370d56bb09feaa31c713b17169b Mon Sep 17 00:00:00 2001 From: Daniel Borkmann Date: Sun, 22 Dec 2019 23:37:40 +0100 Subject: [PATCH 26/33] bpf: Fix precision tracking for unbounded scalars ANBZ: #5530 commit f54c7898ed1c3c9331376c0337a5049c38f66497 upstream. Anatoly has been fuzzing with kBdysch harness and reported a hang in one of the outcomes. Upon closer analysis, it turns out that precise scalar value tracking is missing a few precision markings for unknown scalars: 0: R1=ctx(id=0,off=0,imm=0) R10=fp0 0: (b7) r0 = 0 1: R0_w=invP0 R1=ctx(id=0,off=0,imm=0) R10=fp0 1: (35) if r0 >= 0xf72e goto pc+0 --> only follow fallthrough 2: R0_w=invP0 R1=ctx(id=0,off=0,imm=0) R10=fp0 2: (35) if r0 >= 0x80fe0000 goto pc+0 --> only follow fallthrough 3: R0_w=invP0 R1=ctx(id=0,off=0,imm=0) R10=fp0 3: (14) w0 -= -536870912 4: R0_w=invP536870912 R1=ctx(id=0,off=0,imm=0) R10=fp0 4: (0f) r1 += r0 5: R0_w=invP536870912 R1_w=inv(id=0) R10=fp0 5: (55) if r1 != 0x104c1500 goto pc+0 --> push other branch for later analysis R0_w=invP536870912 R1_w=inv273421568 R10=fp0 6: R0_w=invP536870912 R1_w=inv273421568 R10=fp0 6: (b7) r0 = 0 7: R0=invP0 R1=inv273421568 R10=fp0 7: (76) if w1 s>= 0xffffff00 goto pc+3 --> only follow goto 11: R0=invP0 R1=inv273421568 R10=fp0 11: (95) exit 6: R0_w=invP536870912 R1_w=inv(id=0) R10=fp0 6: (b7) r0 = 0 propagating r0 7: safe processed 11 insns [...] In the analysis of the second path coming after the successful exit above, the path is being pruned at line 7. Pruning analysis found that both r0 are precise P0 and both R1 are non-precise scalars and given prior path with R1 as non-precise scalar succeeded, this one is therefore safe as well. However, problem is that given condition at insn 7 in the first run, we only followed goto and didn't push the other branch for later analysis, we've never walked the few insns in there and therefore dead-code sanitation rewrites it as goto pc-1, causing the hang depending on the skb address hitting these conditions. The issue is that R1 should have been marked as precise as well such that pruning enforces range check and conluded that new R1 is not in range of old R1. In insn 4, we mark R1 (skb) as unknown scalar via __mark_reg_unbounded() but not mark_reg_unbounded() and therefore regs->precise remains as false. Back in b5dc0163d8fd ("bpf: precise scalar_value tracking"), this was not the case since marking out of __mark_reg_unbounded() had this covered as well. Once in both are set as precise in 4 as they should have been, we conclude that given R1 was in prior fall-through path 0x104c1500 and now is completely unknown, the check at insn 7 concludes that we need to continue walking. Analysis after the fix: 0: R1=ctx(id=0,off=0,imm=0) R10=fp0 0: (b7) r0 = 0 1: R0_w=invP0 R1=ctx(id=0,off=0,imm=0) R10=fp0 1: (35) if r0 >= 0xf72e goto pc+0 2: R0_w=invP0 R1=ctx(id=0,off=0,imm=0) R10=fp0 2: (35) if r0 >= 0x80fe0000 goto pc+0 3: R0_w=invP0 R1=ctx(id=0,off=0,imm=0) R10=fp0 3: (14) w0 -= -536870912 4: R0_w=invP536870912 R1=ctx(id=0,off=0,imm=0) R10=fp0 4: (0f) r1 += r0 5: R0_w=invP536870912 R1_w=invP(id=0) R10=fp0 5: (55) if r1 != 0x104c1500 goto pc+0 R0_w=invP536870912 R1_w=invP273421568 R10=fp0 6: R0_w=invP536870912 R1_w=invP273421568 R10=fp0 6: (b7) r0 = 0 7: R0=invP0 R1=invP273421568 R10=fp0 7: (76) if w1 s>= 0xffffff00 goto pc+3 11: R0=invP0 R1=invP273421568 R10=fp0 11: (95) exit 6: R0_w=invP536870912 R1_w=invP(id=0) R10=fp0 6: (b7) r0 = 0 7: R0_w=invP0 R1_w=invP(id=0) R10=fp0 7: (76) if w1 s>= 0xffffff00 goto pc+3 R0_w=invP0 R1_w=invP(id=0) R10=fp0 8: R0_w=invP0 R1_w=invP(id=0) R10=fp0 8: (a5) if r0 < 0x2007002a goto pc+0 9: R0_w=invP0 R1_w=invP(id=0) R10=fp0 9: (57) r0 &= -16316416 10: R0_w=invP0 R1_w=invP(id=0) R10=fp0 10: (a6) if w0 < 0x1201 goto pc+0 11: R0_w=invP0 R1_w=invP(id=0) R10=fp0 11: (95) exit 11: R0=invP0 R1=invP(id=0) R10=fp0 11: (95) exit processed 16 insns [...] Fixes: 6754172c208d ("bpf: fix precision tracking in presence of bpf2bpf calls") Reported-by: Anatoly Trosinenko Signed-off-by: Daniel Borkmann Signed-off-by: Alexei Starovoitov Link: https://lore.kernel.org/bpf/20191222223740.25297-1-daniel@iogearbox.net Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 43 ++++++++++++++++++++++--------------------- 1 file changed, 22 insertions(+), 21 deletions(-) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index a2f5289c345d..3aecee2fcd1c 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -859,7 +859,8 @@ static const int caller_saved[CALLER_SAVED_REGS] = { BPF_REG_0, BPF_REG_1, BPF_REG_2, BPF_REG_3, BPF_REG_4, BPF_REG_5 }; -static void __mark_reg_not_init(struct bpf_reg_state *reg); +static void __mark_reg_not_init(const struct bpf_verifier_env *env, + struct bpf_reg_state *reg); /* Mark the unknown part of a register (variable offset or scalar value) as * known to have the value @imm. @@ -897,7 +898,7 @@ static void mark_reg_known_zero(struct bpf_verifier_env *env, verbose(env, "mark_reg_known_zero(regs, %u)\n", regno); /* Something bad happened, let's kill all regs */ for (regno = 0; regno < MAX_BPF_REG; regno++) - __mark_reg_not_init(regs + regno); + __mark_reg_not_init(env, regs + regno); return; } __mark_reg_known_zero(regs + regno); @@ -995,7 +996,8 @@ static void __mark_reg_unbounded(struct bpf_reg_state *reg) } /* Mark a register as having a completely unknown (scalar) value. */ -static void __mark_reg_unknown(struct bpf_reg_state *reg) +static void __mark_reg_unknown(const struct bpf_verifier_env *env, + struct bpf_reg_state *reg) { /* * Clear type, id, off, and union(map_ptr, range) and @@ -1005,6 +1007,8 @@ static void __mark_reg_unknown(struct bpf_reg_state *reg) reg->type = SCALAR_VALUE; reg->var_off = tnum_unknown; reg->frameno = 0; + reg->precise = env->subprog_cnt > 1 || !env->allow_ptr_leaks ? + true : false; __mark_reg_unbounded(reg); } @@ -1015,19 +1019,16 @@ static void mark_reg_unknown(struct bpf_verifier_env *env, verbose(env, "mark_reg_unknown(regs, %u)\n", regno); /* Something bad happened, let's kill all regs except FP */ for (regno = 0; regno < BPF_REG_FP; regno++) - __mark_reg_not_init(regs + regno); + __mark_reg_not_init(env, regs + regno); return; } - regs += regno; - __mark_reg_unknown(regs); - /* constant backtracking is enabled for root without bpf2bpf calls */ - regs->precise = env->subprog_cnt > 1 || !env->allow_ptr_leaks ? - true : false; + __mark_reg_unknown(env, regs + regno); } -static void __mark_reg_not_init(struct bpf_reg_state *reg) +static void __mark_reg_not_init(const struct bpf_verifier_env *env, + struct bpf_reg_state *reg) { - __mark_reg_unknown(reg); + __mark_reg_unknown(env, reg); reg->type = NOT_INIT; } @@ -1038,10 +1039,10 @@ static void mark_reg_not_init(struct bpf_verifier_env *env, verbose(env, "mark_reg_not_init(regs, %u)\n", regno); /* Something bad happened, let's kill all regs except FP */ for (regno = 0; regno < BPF_REG_FP; regno++) - __mark_reg_not_init(regs + regno); + __mark_reg_not_init(env, regs + regno); return; } - __mark_reg_not_init(regs + regno); + __mark_reg_not_init(env, regs + regno); } #define DEF_NOT_SUBREG (0) @@ -3050,7 +3051,7 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno, } if (state->stack[spi].slot_type[0] == STACK_SPILL && state->stack[spi].spilled_ptr.type == SCALAR_VALUE) { - __mark_reg_unknown(&state->stack[spi].spilled_ptr); + __mark_reg_unknown(env, &state->stack[spi].spilled_ptr); for (j = 0; j < BPF_REG_SIZE; j++) state->stack[spi].slot_type[j] = STACK_MISC; goto mark; @@ -3690,7 +3691,7 @@ static void __clear_all_pkt_pointers(struct bpf_verifier_env *env, if (!reg) continue; if (reg_is_pkt_pointer_any(reg)) - __mark_reg_unknown(reg); + __mark_reg_unknown(env, reg); } } @@ -3718,7 +3719,7 @@ static void release_reg_references(struct bpf_verifier_env *env, if (!reg) continue; if (reg->ref_obj_id == ref_obj_id) - __mark_reg_unknown(reg); + __mark_reg_unknown(env, reg); } } @@ -4531,7 +4532,7 @@ static int adjust_ptr_min_max_vals(struct bpf_verifier_env *env, /* Taint dst register if offset had invalid bounds derived from * e.g. dead branches. */ - __mark_reg_unknown(dst_reg); + __mark_reg_unknown(env, dst_reg); return 0; } @@ -4763,13 +4764,13 @@ static int adjust_scalar_min_max_vals(struct bpf_verifier_env *env, /* Taint dst register if offset had invalid bounds derived from * e.g. dead branches. */ - __mark_reg_unknown(dst_reg); + __mark_reg_unknown(env, dst_reg); return 0; } if (!src_known && opcode != BPF_ADD && opcode != BPF_SUB && opcode != BPF_AND) { - __mark_reg_unknown(dst_reg); + __mark_reg_unknown(env, dst_reg); return 0; } @@ -6899,7 +6900,7 @@ static void clean_func_state(struct bpf_verifier_env *env, /* since the register is unused, clear its state * to make further comparison simpler */ - __mark_reg_not_init(&st->regs[i]); + __mark_reg_not_init(env, &st->regs[i]); } for (i = 0; i < st->allocated_stack / BPF_REG_SIZE; i++) { @@ -6907,7 +6908,7 @@ static void clean_func_state(struct bpf_verifier_env *env, /* liveness must not touch this stack slot anymore */ st->stack[i].spilled_ptr.live |= REG_LIVE_DONE; if (!(live & REG_LIVE_READ)) { - __mark_reg_not_init(&st->stack[i].spilled_ptr); + __mark_reg_not_init(env, &st->stack[i].spilled_ptr); for (j = 0; j < BPF_REG_SIZE; j++) st->stack[i].slot_type[j] = STACK_INVALID; } -- Gitee From ff276ae3bacd47897d0d42ce1a1a8d2febb4865e Mon Sep 17 00:00:00 2001 From: Ilias Apalodimas Date: Thu, 17 Sep 2020 11:49:25 +0300 Subject: [PATCH 27/33] arm64: bpf: Fix branch offset in JIT ANBZ: #5530 commit 32f6865c7aa3c422f710903baa6eb81abc6f559b upstream. Running the eBPF test_verifier leads to random errors looking like this: [ 6525.735488] Unexpected kernel BRK exception at EL1 [ 6525.735502] Internal error: ptrace BRK handler: f2000100 [#1] SMP [ 6525.741609] Modules linked in: nls_utf8 cifs libdes libarc4 dns_resolver fscache binfmt_misc nls_ascii nls_cp437 vfat fat aes_ce_blk crypto_simd cryptd aes_ce_cipher ghash_ce gf128mul efi_pstore sha2_ce sha256_arm64 sha1_ce evdev efivars efivarfs ip_tables x_tables autofs4 btrfs blake2b_generic xor xor_neon zstd_compress raid6_pq libcrc32c crc32c_generic ahci xhci_pci libahci xhci_hcd igb libata i2c_algo_bit nvme realtek usbcore nvme_core scsi_mod t10_pi netsec mdio_devres of_mdio gpio_keys fixed_phy libphy gpio_mb86s7x [ 6525.787760] CPU: 3 PID: 7881 Comm: test_verifier Tainted: G W 5.9.0-rc1+ #47 [ 6525.796111] Hardware name: Socionext SynQuacer E-series DeveloperBox, BIOS build #1 Jun 6 2020 [ 6525.804812] pstate: 20000005 (nzCv daif -PAN -UAO BTYPE=--) [ 6525.810390] pc : bpf_prog_c3d01833289b6311_F+0xc8/0x9f4 [ 6525.815613] lr : bpf_prog_d53bb52e3f4483f9_F+0x38/0xc8c [ 6525.820832] sp : ffff8000130cbb80 [ 6525.824141] x29: ffff8000130cbbb0 x28: 0000000000000000 [ 6525.829451] x27: 000005ef6fcbf39b x26: 0000000000000000 [ 6525.834759] x25: ffff8000130cbb80 x24: ffff800011dc7038 [ 6525.840067] x23: ffff8000130cbd00 x22: ffff0008f624d080 [ 6525.845375] x21: 0000000000000001 x20: ffff800011dc7000 [ 6525.850682] x19: 0000000000000000 x18: 0000000000000000 [ 6525.855990] x17: 0000000000000000 x16: 0000000000000000 [ 6525.861298] x15: 0000000000000000 x14: 0000000000000000 [ 6525.866606] x13: 0000000000000000 x12: 0000000000000000 [ 6525.871913] x11: 0000000000000001 x10: ffff8000000a660c [ 6525.877220] x9 : ffff800010951810 x8 : ffff8000130cbc38 [ 6525.882528] x7 : 0000000000000000 x6 : 0000009864cfa881 [ 6525.887836] x5 : 00ffffffffffffff x4 : 002880ba1a0b3e9f [ 6525.893144] x3 : 0000000000000018 x2 : ffff8000000a4374 [ 6525.898452] x1 : 000000000000000a x0 : 0000000000000009 [ 6525.903760] Call trace: [ 6525.906202] bpf_prog_c3d01833289b6311_F+0xc8/0x9f4 [ 6525.911076] bpf_prog_d53bb52e3f4483f9_F+0x38/0xc8c [ 6525.915957] bpf_dispatcher_xdp_func+0x14/0x20 [ 6525.920398] bpf_test_run+0x70/0x1b0 [ 6525.923969] bpf_prog_test_run_xdp+0xec/0x190 [ 6525.928326] __do_sys_bpf+0xc88/0x1b28 [ 6525.932072] __arm64_sys_bpf+0x24/0x30 [ 6525.935820] el0_svc_common.constprop.0+0x70/0x168 [ 6525.940607] do_el0_svc+0x28/0x88 [ 6525.943920] el0_sync_handler+0x88/0x190 [ 6525.947838] el0_sync+0x140/0x180 [ 6525.951154] Code: d4202000 d4202000 d4202000 d4202000 (d4202000) [ 6525.957249] ---[ end trace cecc3f93b14927e2 ]--- The reason is the offset[] creation and later usage, while building the eBPF body. The code currently omits the first instruction, since build_insn() will increase our ctx->idx before saving it. That was fine up until bounded eBPF loops were introduced. After that introduction, offset[0] must be the offset of the end of prologue which is the start of the 1st insn while, offset[n] holds the offset of the end of n-th insn. When "taken loop with back jump to 1st insn" test runs, it will eventually call bpf2a64_offset(-1, 2, ctx). Since negative indexing is permitted, the current outcome depends on the value stored in ctx->offset[-1], which has nothing to do with our array. If the value happens to be 0 the tests will work. If not this error triggers. commit 7c2e988f400e ("bpf: fix x64 JIT code generation for jmp to 1st insn") fixed an indentical bug on x86 when eBPF bounded loops were introduced. So let's fix it by creating the ctx->offset[] differently. Track the beginning of instruction and account for the extra instruction while calculating the arm instruction offsets. Fixes: 2589726d12a1 ("bpf: introduce bounded loops") Reported-by: Naresh Kamboju Reported-by: Jiri Olsa Co-developed-by: Jean-Philippe Brucker Co-developed-by: Yauheni Kaliuta Signed-off-by: Jean-Philippe Brucker Signed-off-by: Yauheni Kaliuta Signed-off-by: Ilias Apalodimas Acked-by: Will Deacon Link: https://lore.kernel.org/r/20200917084925.177348-1-ilias.apalodimas@linaro.org Signed-off-by: Catalin Marinas Signed-off-by: Qiao Ma --- arch/arm64/net/bpf_jit_comp.c | 43 +++++++++++++++++++++++++---------- 1 file changed, 31 insertions(+), 12 deletions(-) diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c index 9eb0ea9a644f..6e63b6e0688c 100644 --- a/arch/arm64/net/bpf_jit_comp.c +++ b/arch/arm64/net/bpf_jit_comp.c @@ -153,14 +153,17 @@ static inline void emit_addr_mov_i64(const int reg, const u64 val, } } -static inline int bpf2a64_offset(int bpf_to, int bpf_from, +static inline int bpf2a64_offset(int bpf_insn, int off, const struct jit_ctx *ctx) { - int to = ctx->offset[bpf_to]; - /* -1 to account for the Branch instruction */ - int from = ctx->offset[bpf_from] - 1; - - return to - from; + /* BPF JMP offset is relative to the next instruction */ + bpf_insn++; + /* + * Whereas arm64 branch instructions encode the offset + * from the branch itself, so we must subtract 1 from the + * instruction offset. + */ + return ctx->offset[bpf_insn + off] - (ctx->offset[bpf_insn] - 1); } static void jit_fill_hole(void *area, unsigned int size) @@ -546,7 +549,7 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx, /* JUMP off */ case BPF_JMP | BPF_JA: - jmp_offset = bpf2a64_offset(i + off, i, ctx); + jmp_offset = bpf2a64_offset(i, off, ctx); check_imm26(jmp_offset); emit(A64_B(jmp_offset), ctx); break; @@ -573,7 +576,7 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx, case BPF_JMP32 | BPF_JSLE | BPF_X: emit(A64_CMP(is64, dst, src), ctx); emit_cond_jmp: - jmp_offset = bpf2a64_offset(i + off, i, ctx); + jmp_offset = bpf2a64_offset(i, off, ctx); check_imm19(jmp_offset); switch (BPF_OP(code)) { case BPF_JEQ: @@ -798,10 +801,21 @@ static int build_body(struct jit_ctx *ctx, bool extra_pass) const struct bpf_prog *prog = ctx->prog; int i; + /* + * - offset[0] offset of the end of prologue, + * start of the 1st instruction. + * - offset[1] - offset of the end of 1st instruction, + * start of the 2nd instruction + * [....] + * - offset[3] - offset of the end of 3rd instruction, + * start of 4th instruction + */ for (i = 0; i < prog->len; i++) { const struct bpf_insn *insn = &prog->insnsi[i]; int ret; + if (ctx->image == NULL) + ctx->offset[i] = ctx->idx; ret = build_insn(insn, ctx, extra_pass); if (ret > 0) { i++; @@ -809,11 +823,16 @@ static int build_body(struct jit_ctx *ctx, bool extra_pass) ctx->offset[i] = ctx->idx; continue; } - if (ctx->image == NULL) - ctx->offset[i] = ctx->idx; if (ret) return ret; } + /* + * offset is allocated with prog->len + 1 so fill in + * the last element with the offset after the last + * instruction (end of program) + */ + if (ctx->image == NULL) + ctx->offset[i] = ctx->idx; return 0; } @@ -889,7 +908,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) memset(&ctx, 0, sizeof(ctx)); ctx.prog = prog; - ctx.offset = kcalloc(prog->len, sizeof(int), GFP_KERNEL); + ctx.offset = kcalloc(prog->len + 1, sizeof(int), GFP_KERNEL); if (ctx.offset == NULL) { prog = orig_prog; goto out_off; @@ -972,7 +991,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) prog->jited_len = image_size; if (!prog->is_func || extra_pass) { - bpf_prog_fill_jited_linfo(prog, ctx.offset); + bpf_prog_fill_jited_linfo(prog, ctx.offset + 1); out_off: kfree(ctx.offset); kfree(jit_data); -- Gitee From f79e9b2c9c4e802e590a382d7aa5c91bdbbc4837 Mon Sep 17 00:00:00 2001 From: Hou Tao Date: Sat, 26 Feb 2022 20:19:06 +0800 Subject: [PATCH 28/33] bpf, arm64: Feed byte-offset into bpf line info ANBZ: #5530 commit dda7596c109fc382876118627e29db7607cde35d upstream. insn_to_jit_off passed to bpf_prog_fill_jited_linfo() is calculated in instruction granularity instead of bytes granularity, but BPF line info requires byte offset. bpf_prog_fill_jited_linfo() will be the last user of ctx.offset before it is freed, so convert the offset into byte-offset before calling into bpf_prog_fill_jited_linfo() in order to fix the line info dump on arm64. Fixes: 37ab566c178d ("bpf: arm64: Enable arm64 jit to provide bpf_line_info") Suggested-by: Daniel Borkmann Signed-off-by: Hou Tao Signed-off-by: Daniel Borkmann Link: https://lore.kernel.org/bpf/20220226121906.5709-3-houtao1@huawei.com Signed-off-by: Qiao Ma --- arch/arm64/net/bpf_jit_comp.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c index 6e63b6e0688c..6bcce848a720 100644 --- a/arch/arm64/net/bpf_jit_comp.c +++ b/arch/arm64/net/bpf_jit_comp.c @@ -991,6 +991,11 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) prog->jited_len = image_size; if (!prog->is_func || extra_pass) { + int i; + + /* offset[prog->len] is the size of program */ + for (i = 0; i <= prog->len; i++) + ctx.offset[i] *= AARCH64_INSN_SIZE; bpf_prog_fill_jited_linfo(prog, ctx.offset + 1); out_off: kfree(ctx.offset); -- Gitee From e962583a221cd9ea5681d3af0b592345770b3faa Mon Sep 17 00:00:00 2001 From: Andrii Nakryiko Date: Fri, 4 Nov 2022 09:36:44 -0700 Subject: [PATCH 29/33] bpf: propagate precision in ALU/ALU64 operations ANBZ: #5530 commit a3b666bfa9c9edc05bca62a87abafe0936bd7f97 upstream. When processing ALU/ALU64 operations (apart from BPF_MOV, which is handled correctly already; and BPF_NEG and BPF_END are special and don't have source register), if destination register is already marked precise, this causes problem with potentially missing precision tracking for the source register. E.g., when we have r1 >>= r5 and r1 is marked precise, but r5 isn't, this will lead to r5 staying as imprecise. This is due to the precision backtracking logic stopping early when it sees r1 is already marked precise. If r1 wasn't precise, we'd keep backtracking and would add r5 to the set of registers that need to be marked precise. So there is a discrepancy here which can lead to invalid and incompatible states matched due to lack of precision marking on r5. If r1 wasn't precise, precision backtracking would correctly mark both r1 and r5 as precise. This is simple to fix, though. During the forward instruction simulation pass, for arithmetic operations of `scalar = scalar` form (where is ALU or ALU64 operations), if destination register is already precise, mark source register as precise. This applies only when both involved registers are SCALARs. `ptr += scalar` and `scalar += ptr` cases are already handled correctly. This does have (negative) effect on some selftest programs and few Cilium programs. ~/baseline-tmp-results.csv are veristat results with this patch, while ~/baseline-results.csv is without it. See post scriptum for instructions on how to make Cilium programs testable with veristat. Correctness has a price. $ ./veristat -C -e file,prog,insns,states ~/baseline-results.csv ~/baseline-tmp-results.csv | grep -v '+0' File Program Total insns (A) Total insns (B) Total insns (DIFF) Total states (A) Total states (B) Total states (DIFF) ----------------------- -------------------- --------------- --------------- ------------------ ---------------- ---------------- ------------------- bpf_cubic.bpf.linked1.o bpf_cubic_cong_avoid 997 1700 +703 (+70.51%) 62 90 +28 (+45.16%) test_l4lb.bpf.linked1.o balancer_ingress 4559 5469 +910 (+19.96%) 118 126 +8 (+6.78%) ----------------------- -------------------- --------------- --------------- ------------------ ---------------- ---------------- ------------------- $ ./veristat -C -e file,prog,verdict,insns,states ~/baseline-results-cilium.csv ~/baseline-tmp-results-cilium.csv | grep -v '+0' File Program Total insns (A) Total insns (B) Total insns (DIFF) Total states (A) Total states (B) Total states (DIFF) ------------- ------------------------------ --------------- --------------- ------------------ ---------------- ---------------- ------------------- bpf_host.o tail_nodeport_nat_ingress_ipv6 4448 5261 +813 (+18.28%) 234 247 +13 (+5.56%) bpf_host.o tail_nodeport_nat_ipv6_egress 3396 3446 +50 (+1.47%) 201 203 +2 (+1.00%) bpf_lxc.o tail_nodeport_nat_ingress_ipv6 4448 5261 +813 (+18.28%) 234 247 +13 (+5.56%) bpf_overlay.o tail_nodeport_nat_ingress_ipv6 4448 5261 +813 (+18.28%) 234 247 +13 (+5.56%) bpf_xdp.o tail_lb_ipv4 71736 73442 +1706 (+2.38%) 4295 4370 +75 (+1.75%) ------------- ------------------------------ --------------- --------------- ------------------ ---------------- ---------------- ------------------- P.S. To make Cilium ([0]) programs libbpf-compatible and thus veristat-loadable, apply changes from topmost commit in [1], which does minimal changes to Cilium source code, mostly around SEC() annotations and BPF map definitions. [0] https://github.com/cilium/cilium/ [1] https://github.com/anakryiko/cilium/commits/libbpf-friendliness Fixes: b5dc0163d8fd ("bpf: precise scalar_value tracking") Signed-off-by: Andrii Nakryiko Link: https://lore.kernel.org/r/20221104163649.121784-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 3aecee2fcd1c..fe02ec01ab79 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -5052,6 +5052,11 @@ static int adjust_reg_min_max_vals(struct bpf_verifier_env *env, return err; return adjust_ptr_min_max_vals(env, insn, dst_reg, src_reg); + } else if (dst_reg->precise) { + /* if dst_reg is precise, src_reg should be precise as well */ + err = mark_chain_precision(env, insn->src_reg); + if (err) + return err; } } else { /* Pretend the src is a reg with a known value, since we only -- Gitee From 1f518b565d5fe26380c914474aad109d6261962a Mon Sep 17 00:00:00 2001 From: Andrii Nakryiko Date: Fri, 4 Nov 2022 09:36:45 -0700 Subject: [PATCH 30/33] bpf: propagate precision across all frames, not just the last one ANBZ: #5530 commit 529409ea92d590659be487ba0839710329bd8074 upstream. When equivalent completed state is found and it has additional precision restrictions, BPF verifier propagates precision to currently-being-verified state chain (i.e., including parent states) so that if some of the states in the chain are not yet completed, necessary precision restrictions are enforced. Unfortunately, right now this happens only for the last frame (deepest active subprogram's frame), not all the frames. This can lead to incorrect matching of states due to missing precision marker. Currently this doesn't seem possible as BPF verifier forces everything to precise when validated BPF program has any subprograms. But with the next patch lifting this restriction, this becomes problematic. In fact, without this fix, we'll start getting failure in one of the existing test_verifier test cases: #906/p precise: cross frame pruning FAIL Unexpected success to load! verification time 48 usec stack depth 0+0 processed 26 insns (limit 1000000) max_states_per_insn 3 total_states 17 peak_states 17 mark_read 8 This patch adds precision propagation across all frames. Fixes: a3ce685dd01a ("bpf: fix precision tracking") Signed-off-by: Andrii Nakryiko Link: https://lore.kernel.org/r/20221104163649.121784-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov Signed-off-by: Qiao Ma --- kernel/bpf/verifier.c | 74 ++++++++++++++++++++++++------------------- 1 file changed, 41 insertions(+), 33 deletions(-) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index fe02ec01ab79..c667d9e3b571 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -1665,7 +1665,7 @@ static void mark_all_scalars_precise(struct bpf_verifier_env *env, } } -static int __mark_chain_precision(struct bpf_verifier_env *env, int regno, +static int __mark_chain_precision(struct bpf_verifier_env *env, int frame, int regno, int spi) { struct bpf_verifier_state *st = env->cur_state; @@ -1683,7 +1683,7 @@ static int __mark_chain_precision(struct bpf_verifier_env *env, int regno, /* backtracking is root only for now */ return 0; - func = st->frame[st->curframe]; + func = st->frame[frame]; if (regno >= 0) { reg = &func->regs[regno]; if (reg->type != SCALAR_VALUE) { @@ -1764,7 +1764,7 @@ static int __mark_chain_precision(struct bpf_verifier_env *env, int regno, break; new_marks = false; - func = st->frame[st->curframe]; + func = st->frame[frame]; bitmap_from_u64(mask, reg_mask); for_each_set_bit(i, mask, 32) { reg = &func->regs[i]; @@ -1828,14 +1828,19 @@ static int __mark_chain_precision(struct bpf_verifier_env *env, int regno, return 0; } -static int mark_chain_precision(struct bpf_verifier_env *env, int regno) +int mark_chain_precision(struct bpf_verifier_env *env, int regno) { - return __mark_chain_precision(env, regno, -1); + return __mark_chain_precision(env, env->cur_state->curframe, regno, -1); } -static int mark_chain_precision_stack(struct bpf_verifier_env *env, int spi) +static int mark_chain_precision_frame(struct bpf_verifier_env *env, int frame, int regno) { - return __mark_chain_precision(env, -1, spi); + return __mark_chain_precision(env, frame, regno, -1); +} + +static int mark_chain_precision_stack_frame(struct bpf_verifier_env *env, int frame, int spi) +{ + return __mark_chain_precision(env, frame, -1, spi); } static bool is_spillable_regtype(enum bpf_reg_type type) @@ -7349,34 +7354,37 @@ static int propagate_precision(struct bpf_verifier_env *env, { struct bpf_reg_state *state_reg; struct bpf_func_state *state; - int i, err = 0; + int i, err = 0, fr; - state = old->frame[old->curframe]; - state_reg = state->regs; - for (i = 0; i < BPF_REG_FP; i++, state_reg++) { - if (state_reg->type != SCALAR_VALUE || - !state_reg->precise) - continue; - if (env->log.level & BPF_LOG_LEVEL2) - verbose(env, "propagating r%d\n", i); - err = mark_chain_precision(env, i); - if (err < 0) - return err; - } + for (fr = old->curframe; fr >= 0; fr--) { + state = old->frame[fr]; + state_reg = state->regs; + for (i = 0; i < BPF_REG_FP; i++, state_reg++) { + if (state_reg->type != SCALAR_VALUE || + !state_reg->precise) + continue; + if (env->log.level & BPF_LOG_LEVEL2) + verbose(env, "frame %d: propagating r%d\n", i, fr); + err = mark_chain_precision_frame(env, fr, i); + if (err < 0) + return err; + } + + for (i = 0; i < state->allocated_stack / BPF_REG_SIZE; i++) { + if (state->stack[i].slot_type[0] != STACK_SPILL) + continue; + state_reg = &state->stack[i].spilled_ptr; + if (state_reg->type != SCALAR_VALUE || + !state_reg->precise) + continue; + if (env->log.level & BPF_LOG_LEVEL2) + verbose(env, "frame %d: propagating fp%d\n", + (-i - 1) * BPF_REG_SIZE, fr); + err = mark_chain_precision_stack_frame(env, fr, i); + if (err < 0) + return err; + } - for (i = 0; i < state->allocated_stack / BPF_REG_SIZE; i++) { - if (state->stack[i].slot_type[0] != STACK_SPILL) - continue; - state_reg = &state->stack[i].spilled_ptr; - if (state_reg->type != SCALAR_VALUE || - !state_reg->precise) - continue; - if (env->log.level & BPF_LOG_LEVEL2) - verbose(env, "propagating fp%d\n", - (-i - 1) * BPF_REG_SIZE); - err = mark_chain_precision_stack(env, i); - if (err < 0) - return err; } return 0; } -- Gitee From a1404d5dfd6b81a068a6c3fff9b2831e19f17524 Mon Sep 17 00:00:00 2001 From: Yonghong Song Date: Wed, 9 Sep 2020 10:15:42 -0700 Subject: [PATCH 31/33] selftests/bpf: Fix test_sysctl_loop{1, 2} failure due to clang change ANBZ: #5530 commit 7fb5eefd76394cfefb380724a87ca40b47d44405 upstream. Andrii reported that with latest clang, when building selftests, we have error likes: error: progs/test_sysctl_loop1.c:23:16: in function sysctl_tcp_mem i32 (%struct.bpf_sysctl*): Looks like the BPF stack limit of 512 bytes is exceeded. Please move large on stack variables into BPF per-cpu array map. The error is triggered by the following LLVM patch: https://reviews.llvm.org/D87134 For example, the following code is from test_sysctl_loop1.c: static __always_inline int is_tcp_mem(struct bpf_sysctl *ctx) { volatile char tcp_mem_name[] = "net/ipv4/tcp_mem/very_very_very_very_long_pointless_string"; ... } Without the above LLVM patch, the compiler did optimization to load the string (59 bytes long) with 7 64bit loads, 1 8bit load and 1 16bit load, occupying 64 byte stack size. With the above LLVM patch, the compiler only uses 8bit loads, but subregister is 32bit. So stack requirements become 4 * 59 = 236 bytes. Together with other stuff on the stack, total stack size exceeds 512 bytes, hence compiler complains and quits. To fix the issue, removing "volatile" key word or changing "volatile" to "const"/"static const" does not work, the string is put in .rodata.str1.1 section, which libbpf did not process it and errors out with libbpf: elf: skipping unrecognized data section(6) .rodata.str1.1 libbpf: prog 'sysctl_tcp_mem': bad map relo against '.L__const.is_tcp_mem.tcp_mem_name' in section '.rodata.str1.1' Defining the string const as global variable can fix the issue as it puts the string constant in '.rodata' section which is recognized by libbpf. In the future, when libbpf can process '.rodata.str*.*' properly, the global definition can be changed back to local definition. Defining tcp_mem_name as a global, however, triggered a verifier failure. ./test_progs -n 7/21 libbpf: load bpf program failed: Permission denied libbpf: -- BEGIN DUMP LOG --- libbpf: invalid stack off=0 size=1 verification time 6975 usec stack depth 160+64 processed 889 insns (limit 1000000) max_states_per_insn 4 total_states 14 peak_states 14 mark_read 10 libbpf: -- END LOG -- libbpf: failed to load program 'sysctl_tcp_mem' libbpf: failed to load object 'test_sysctl_loop2.o' test_bpf_verif_scale:FAIL:114 #7/21 test_sysctl_loop2.o:FAIL This actually exposed a bpf program bug. In test_sysctl_loop{1,2}, we have code like const char tcp_mem_name[] = "<...long string...>"; ... char name[64]; ... for (i = 0; i < sizeof(tcp_mem_name); ++i) if (name[i] != tcp_mem_name[i]) return 0; In the above code, if sizeof(tcp_mem_name) > 64, name[i] access may be out of bound. The sizeof(tcp_mem_name) is 59 for test_sysctl_loop1.c and 79 for test_sysctl_loop2.c. Without promotion-to-global change, old compiler generates code where the overflowed stack access is actually filled with valid value, so hiding the bpf program bug. With promotion-to-global change, the code is different, more specifically, the previous loading constants to stack is gone, and "name" occupies stack[-64:0] and overflow access triggers a verifier error. To fix the issue, adjust "name" buffer size properly. Reported-by: Andrii Nakryiko Signed-off-by: Yonghong Song Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Link: https://lore.kernel.org/bpf/20200909171542.3673449-1-yhs@fb.com Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/progs/test_sysctl_loop1.c | 4 ++-- tools/testing/selftests/bpf/progs/test_sysctl_loop2.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/tools/testing/selftests/bpf/progs/test_sysctl_loop1.c b/tools/testing/selftests/bpf/progs/test_sysctl_loop1.c index 608a06871572..ac1c0ef3c46c 100644 --- a/tools/testing/selftests/bpf/progs/test_sysctl_loop1.c +++ b/tools/testing/selftests/bpf/progs/test_sysctl_loop1.c @@ -18,11 +18,11 @@ #define MAX_ULONG_STR_LEN 7 #define MAX_VALUE_STR_LEN (TCP_MEM_LOOPS * MAX_ULONG_STR_LEN) +const char tcp_mem_name[] = "net/ipv4/tcp_mem/very_very_very_very_long_pointless_string"; static __always_inline int is_tcp_mem(struct bpf_sysctl *ctx) { - volatile char tcp_mem_name[] = "net/ipv4/tcp_mem/very_very_very_very_long_pointless_string"; unsigned char i; - char name[64]; + char name[sizeof(tcp_mem_name)]; int ret; memset(name, 0, sizeof(name)); diff --git a/tools/testing/selftests/bpf/progs/test_sysctl_loop2.c b/tools/testing/selftests/bpf/progs/test_sysctl_loop2.c index cb201cbe11e7..55251046c9b7 100644 --- a/tools/testing/selftests/bpf/progs/test_sysctl_loop2.c +++ b/tools/testing/selftests/bpf/progs/test_sysctl_loop2.c @@ -18,11 +18,11 @@ #define MAX_ULONG_STR_LEN 7 #define MAX_VALUE_STR_LEN (TCP_MEM_LOOPS * MAX_ULONG_STR_LEN) +const char tcp_mem_name[] = "net/ipv4/tcp_mem/very_very_very_very_long_pointless_string_to_stress_byte_loop"; static __attribute__((noinline)) int is_tcp_mem(struct bpf_sysctl *ctx) { - volatile char tcp_mem_name[] = "net/ipv4/tcp_mem/very_very_very_very_long_pointless_string_to_stress_byte_loop"; unsigned char i; - char name[64]; + char name[sizeof(tcp_mem_name)]; int ret; memset(name, 0, sizeof(name)); -- Gitee From e66c0e7f0180ebbc7429dc43219c54a69e802924 Mon Sep 17 00:00:00 2001 From: Qiao Ma Date: Thu, 18 Jan 2024 20:40:58 +0800 Subject: [PATCH 32/33] anolis: bpf/selftest: do not crash when load progs fail ANBZ: #5530 The patch comes from follow commit: 07b619919d3d ("selftests: bpf: initialize bpf_object pointers where needed") In function check_load(), if prog loading fails, the pointer obj is undefined, and bpf_object__close() will cause a segmentfault. To avoid this problem, setting obj to NULL by default. Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/test_progs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/testing/selftests/bpf/test_progs.c b/tools/testing/selftests/bpf/test_progs.c index c2b81f93895d..fb5566eb0680 100644 --- a/tools/testing/selftests/bpf/test_progs.c +++ b/tools/testing/selftests/bpf/test_progs.c @@ -2384,7 +2384,7 @@ static int libbpf_debug_print_verifier_scale(enum libbpf_print_level level, static int check_load(const char *file, enum bpf_prog_type type) { struct bpf_prog_load_attr attr; - struct bpf_object *obj; + struct bpf_object *obj = NULL; int err, prog_fd; memset(&attr, 0, sizeof(struct bpf_prog_load_attr)); -- Gitee From d58ac133d964f5a85ea0e35eec4f851e6fb01b31 Mon Sep 17 00:00:00 2001 From: Andrii Nakryiko Date: Wed, 25 Sep 2019 11:52:05 -0700 Subject: [PATCH 33/33] selftests/bpf: adjust strobemeta loop to satisfy latest clang ANBZ: #5530 commit 4670d68b9254710fdeaf794cad54d8b2c9929e0a upstream. Some recent changes in latest Clang started causing the following warning when unrolling strobemeta test case main loop: progs/strobemeta.h:416:2: warning: loop not unrolled: the optimizer was unable to perform the requested transformation; the transformation might be disabled or specified as part of an unsupported transformation ordering [-Wpass-failed=transform-warning] This patch simplifies loop's exit condition to depend only on constant max iteration number (STROBE_MAX_MAP_ENTRIES), while moving early termination logic inside the loop body. The changes are equivalent from program logic standpoint, but fixes the warning. It also appears to improve generated BPF code, as it fixes previously failing non-unrolled strobemeta test cases. Cc: Alexei Starovoitov Signed-off-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Signed-off-by: Qiao Ma --- tools/testing/selftests/bpf/progs/strobemeta.h | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/tools/testing/selftests/bpf/progs/strobemeta.h b/tools/testing/selftests/bpf/progs/strobemeta.h index 1ff73f60a3e4..82ee72bc53db 100644 --- a/tools/testing/selftests/bpf/progs/strobemeta.h +++ b/tools/testing/selftests/bpf/progs/strobemeta.h @@ -411,7 +411,10 @@ void *read_map_var(struct strobemeta_cfg *cfg, size_t idx, void *tls_base, #else #pragma unroll #endif - for (int i = 0; i < STROBE_MAX_MAP_ENTRIES && i < map.cnt; ++i) { + for (int i = 0; i < STROBE_MAX_MAP_ENTRIES; ++i) { + if (i >= map.cnt) + break; + descr->key_lens[i] = 0; len = bpf_probe_read_str(payload, STROBE_MAX_STR_LEN, map.entries[i].key); -- Gitee