v3 -> v4: - Rebasing - Cast bpf_[per|this]_cpu_ptr's parameter to void __percpu * before passing into per_cpu_ptr.
v2 -> v3: - Rename functions and variables in verifier for better readability. - Stick to logging message convention in libbpf. - Move bpf_per_cpu_ptr and bpf_this_cpu_ptr from trace-specific helper set to base helper set. - More specific test in ksyms_btf. - Fix return type cast in bpf_*_cpu_ptr. - Fix btf leak in ksyms_btf selftest. - Fix return error code for kallsyms_find().
v1 -> v2: - Move check_pseudo_btf_id from check_ld_imm() to replace_map_fd_with_map_ptr() and rename the latter. - Add bpf_this_cpu_ptr(). - Use bpf_core_types_are_compat() in libbpf.c for checking type compatibility. - Rewrite typed ksym extern type in BTF with int to save space. - Minor revision of bpf_per_cpu_ptr()'s comments. - Avoid using long in tests that use skeleton. - Refactored test_ksyms.c by moving kallsyms_find() to trace_helpers.c - Fold the patches that sync include/linux/uapi and tools/include/linux/uapi.
rfc -> v1: - Encode VAR's btf_id for PSEUDO_BTF_ID. - More checks in verifier. Checking the btf_id passed as PSEUDO_BTF_ID is valid VAR, its name and type. - Checks in libbpf on type compatibility of ksyms. - Add bpf_per_cpu_ptr() to access kernel percpu vars. Introduced new ARG and RET types for this helper.
This patch series extends the previously added __ksym externs with btf support.
Right now the __ksym externs are treated as pure 64-bit scalar value. Libbpf replaces ld_imm64 insn of __ksym by its kernel address at load time. This patch series extend those externs with their btf info. Note that btf support for __ksym must come with the kernel btf that has VARs encoded to work properly. The corresponding chagnes in pahole is available at [1] (with a fix at [2] for gcc 4.9+).
The first 3 patches in this series add support for general kernel global variables, which include verifier checking (01/06), libpf support (02/06) and selftests for getting typed ksym extern's kernel address (03/06).
The next 3 patches extends that capability further by introducing helpers bpf_per_cpu_ptr() and bpf_this_cpu_ptr(), which allows accessing kernel percpu variables correctly (04/06 and 05/06).
The tests of this feature were performed against pahole that is extended with [1] and [2]. For kernel BTF that does not have VARs encoded, the selftests will be skipped.
[1] https://git.kernel.org/pub/scm/devel/pahole/pahole.git/commit/?id=f3d9054ba8... [2] https://www.spinics.net/lists/dwarves/msg00451.html
Hao Luo (6): bpf: Introduce pseudo_btf_id bpf/libbpf: BTF support for typed ksyms selftests/bpf: ksyms_btf to test typed ksyms bpf: Introduce bpf_per_cpu_ptr() bpf: Introducte bpf_this_cpu_ptr() bpf/selftests: Test for bpf_per_cpu_ptr() and bpf_this_cpu_ptr()
include/linux/bpf.h | 6 + include/linux/bpf_verifier.h | 7 + include/linux/btf.h | 26 +++ include/uapi/linux/bpf.h | 67 +++++- kernel/bpf/btf.c | 25 --- kernel/bpf/helpers.c | 32 +++ kernel/bpf/verifier.c | 190 ++++++++++++++++-- kernel/trace/bpf_trace.c | 4 + tools/include/uapi/linux/bpf.h | 67 +++++- tools/lib/bpf/libbpf.c | 112 +++++++++-- .../testing/selftests/bpf/prog_tests/ksyms.c | 38 ++-- .../selftests/bpf/prog_tests/ksyms_btf.c | 88 ++++++++ .../selftests/bpf/progs/test_ksyms_btf.c | 55 +++++ tools/testing/selftests/bpf/trace_helpers.c | 27 +++ tools/testing/selftests/bpf/trace_helpers.h | 4 + 15 files changed, 653 insertions(+), 95 deletions(-) create mode 100644 tools/testing/selftests/bpf/prog_tests/ksyms_btf.c create mode 100644 tools/testing/selftests/bpf/progs/test_ksyms_btf.c
Pseudo_btf_id is a type of ld_imm insn that associates a btf_id to a ksym so that further dereferences on the ksym can use the BTF info to validate accesses. Internally, when seeing a pseudo_btf_id ld insn, the verifier reads the btf_id stored in the insn[0]'s imm field and marks the dst_reg as PTR_TO_BTF_ID. The btf_id points to a VAR_KIND, which is encoded in btf_vminux by pahole. If the VAR is not of a struct type, the dst reg will be marked as PTR_TO_MEM instead of PTR_TO_BTF_ID and the mem_size is resolved to the size of the VAR's type.
From the VAR btf_id, the verifier can also read the address of the
ksym's corresponding kernel var from kallsyms and use that to fill dst_reg.
Therefore, the proper functionality of pseudo_btf_id depends on (1) kallsyms and (2) the encoding of kernel global VARs in pahole, which should be available since pahole v1.18.
Acked-by: Andrii Nakryiko andriin@fb.com Signed-off-by: Hao Luo haoluo@google.com --- include/linux/bpf_verifier.h | 7 ++ include/linux/btf.h | 15 ++++ include/uapi/linux/bpf.h | 36 +++++++--- kernel/bpf/btf.c | 15 ---- kernel/bpf/verifier.c | 125 +++++++++++++++++++++++++++++---- tools/include/uapi/linux/bpf.h | 36 +++++++--- 6 files changed, 188 insertions(+), 46 deletions(-)
diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h index 363b4f1c562a..e83ef6f6bf43 100644 --- a/include/linux/bpf_verifier.h +++ b/include/linux/bpf_verifier.h @@ -308,6 +308,13 @@ struct bpf_insn_aux_data { u32 map_index; /* index into used_maps[] */ u32 map_off; /* offset from value base address */ }; + struct { + enum bpf_reg_type reg_type; /* type of pseudo_btf_id */ + union { + u32 btf_id; /* btf_id for struct typed var */ + u32 mem_size; /* mem_size for non-struct typed var */ + }; + } btf_var; }; u64 map_key_state; /* constant (32 bit) key tracking for maps */ int ctx_field_size; /* the ctx field size for load insn, maybe 0 */ diff --git a/include/linux/btf.h b/include/linux/btf.h index 024e16ff7dcc..af1244180588 100644 --- a/include/linux/btf.h +++ b/include/linux/btf.h @@ -145,6 +145,21 @@ static inline bool btf_type_is_func_proto(const struct btf_type *t) return BTF_INFO_KIND(t->info) == BTF_KIND_FUNC_PROTO; }
+static inline bool btf_type_is_var(const struct btf_type *t) +{ + return BTF_INFO_KIND(t->info) == BTF_KIND_VAR; +} + +/* union is only a special case of struct: + * all its offsetof(member) == 0 + */ +static inline bool btf_type_is_struct(const struct btf_type *t) +{ + u8 kind = BTF_INFO_KIND(t->info); + + return kind == BTF_KIND_STRUCT || kind == BTF_KIND_UNION; +} + static inline u16 btf_type_vlen(const struct btf_type *t) { return BTF_INFO_VLEN(t->info); diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 2b1d3f16cbd1..1defa5f23078 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -356,18 +356,36 @@ enum bpf_link_type { #define BPF_F_SLEEPABLE (1U << 4)
/* When BPF ldimm64's insn[0].src_reg != 0 then this can have - * two extensions: - * - * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE - * insn[0].imm: map fd map fd - * insn[1].imm: 0 offset into value - * insn[0].off: 0 0 - * insn[1].off: 0 0 - * ldimm64 rewrite: address of map address of map[0]+offset - * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE + * the following extensions: + * + * insn[0].src_reg: BPF_PSEUDO_MAP_FD + * insn[0].imm: map fd + * insn[1].imm: 0 + * insn[0].off: 0 + * insn[1].off: 0 + * ldimm64 rewrite: address of map + * verifier type: CONST_PTR_TO_MAP */ #define BPF_PSEUDO_MAP_FD 1 +/* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE + * insn[0].imm: map fd + * insn[1].imm: offset into value + * insn[0].off: 0 + * insn[1].off: 0 + * ldimm64 rewrite: address of map[0]+offset + * verifier type: PTR_TO_MAP_VALUE + */ #define BPF_PSEUDO_MAP_VALUE 2 +/* insn[0].src_reg: BPF_PSEUDO_BTF_ID + * insn[0].imm: kernel btd id of VAR + * insn[1].imm: 0 + * insn[0].off: 0 + * insn[1].off: 0 + * ldimm64 rewrite: address of the kernel variable + * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var + * is struct/union. + */ +#define BPF_PSEUDO_BTF_ID 3
/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative * offset to another bpf function diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c index 4d0ee7839fdb..00569afe3d0d 100644 --- a/kernel/bpf/btf.c +++ b/kernel/bpf/btf.c @@ -440,16 +440,6 @@ static bool btf_type_nosize_or_null(const struct btf_type *t) return !t || btf_type_nosize(t); }
-/* union is only a special case of struct: - * all its offsetof(member) == 0 - */ -static bool btf_type_is_struct(const struct btf_type *t) -{ - u8 kind = BTF_INFO_KIND(t->info); - - return kind == BTF_KIND_STRUCT || kind == BTF_KIND_UNION; -} - static bool __btf_type_is_struct(const struct btf_type *t) { return BTF_INFO_KIND(t->info) == BTF_KIND_STRUCT; @@ -460,11 +450,6 @@ static bool btf_type_is_array(const struct btf_type *t) return BTF_INFO_KIND(t->info) == BTF_KIND_ARRAY; }
-static bool btf_type_is_var(const struct btf_type *t) -{ - return BTF_INFO_KIND(t->info) == BTF_KIND_VAR; -} - static bool btf_type_is_datasec(const struct btf_type *t) { return BTF_INFO_KIND(t->info) == BTF_KIND_DATASEC; diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 015a1c074b6b..fe4965079773 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -7488,6 +7488,7 @@ static int check_ld_imm(struct bpf_verifier_env *env, struct bpf_insn *insn) { struct bpf_insn_aux_data *aux = cur_aux(env); struct bpf_reg_state *regs = cur_regs(env); + struct bpf_reg_state *dst_reg; struct bpf_map *map; int err;
@@ -7504,25 +7505,44 @@ static int check_ld_imm(struct bpf_verifier_env *env, struct bpf_insn *insn) if (err) return err;
+ dst_reg = ®s[insn->dst_reg]; if (insn->src_reg == 0) { u64 imm = ((u64)(insn + 1)->imm << 32) | (u32)insn->imm;
- regs[insn->dst_reg].type = SCALAR_VALUE; + dst_reg->type = SCALAR_VALUE; __mark_reg_known(®s[insn->dst_reg], imm); return 0; }
+ if (insn->src_reg == BPF_PSEUDO_BTF_ID) { + mark_reg_known_zero(env, regs, insn->dst_reg); + + dst_reg->type = aux->btf_var.reg_type; + switch (dst_reg->type) { + case PTR_TO_MEM: + dst_reg->mem_size = aux->btf_var.mem_size; + break; + case PTR_TO_BTF_ID: + dst_reg->btf_id = aux->btf_var.btf_id; + break; + default: + verbose(env, "bpf verifier is misconfigured\n"); + return -EFAULT; + } + return 0; + } + map = env->used_maps[aux->map_index]; mark_reg_known_zero(env, regs, insn->dst_reg); - regs[insn->dst_reg].map_ptr = map; + dst_reg->map_ptr = map;
if (insn->src_reg == BPF_PSEUDO_MAP_VALUE) { - regs[insn->dst_reg].type = PTR_TO_MAP_VALUE; - regs[insn->dst_reg].off = aux->map_off; + dst_reg->type = PTR_TO_MAP_VALUE; + dst_reg->off = aux->map_off; if (map_value_has_spin_lock(map)) - regs[insn->dst_reg].id = ++env->id_gen; + dst_reg->id = ++env->id_gen; } else if (insn->src_reg == BPF_PSEUDO_MAP_FD) { - regs[insn->dst_reg].type = CONST_PTR_TO_MAP; + dst_reg->type = CONST_PTR_TO_MAP; } else { verbose(env, "bpf verifier is misconfigured\n"); return -EINVAL; @@ -9424,6 +9444,73 @@ static int do_check(struct bpf_verifier_env *env) return 0; }
+/* replace pseudo btf_id with kernel symbol address */ +static int check_pseudo_btf_id(struct bpf_verifier_env *env, + struct bpf_insn *insn, + struct bpf_insn_aux_data *aux) +{ + u32 type, id = insn->imm; + const struct btf_type *t; + const char *sym_name; + u64 addr; + + if (!btf_vmlinux) { + verbose(env, "kernel is missing BTF, make sure CONFIG_DEBUG_INFO_BTF=y is specified in Kconfig.\n"); + return -EINVAL; + } + + if (insn[1].imm != 0) { + verbose(env, "reserved field (insn[1].imm) is used in pseudo_btf_id ldimm64 insn.\n"); + return -EINVAL; + } + + t = btf_type_by_id(btf_vmlinux, id); + if (!t) { + verbose(env, "ldimm64 insn specifies invalid btf_id %d.\n", id); + return -ENOENT; + } + + if (!btf_type_is_var(t)) { + verbose(env, "pseudo btf_id %d in ldimm64 isn't KIND_VAR.\n", + id); + return -EINVAL; + } + + sym_name = btf_name_by_offset(btf_vmlinux, t->name_off); + addr = kallsyms_lookup_name(sym_name); + if (!addr) { + verbose(env, "ldimm64 failed to find the address for kernel symbol '%s'.\n", + sym_name); + return -ENOENT; + } + + insn[0].imm = (u32)addr; + insn[1].imm = addr >> 32; + + type = t->type; + t = btf_type_skip_modifiers(btf_vmlinux, type, NULL); + if (!btf_type_is_struct(t)) { + const struct btf_type *ret; + const char *tname; + u32 tsize; + + /* resolve the type size of ksym. */ + ret = btf_resolve_size(btf_vmlinux, t, &tsize); + if (IS_ERR(ret)) { + tname = btf_name_by_offset(btf_vmlinux, t->name_off); + verbose(env, "ldimm64 unable to resolve the size of type '%s': %ld\n", + tname, PTR_ERR(ret)); + return -EINVAL; + } + aux->btf_var.reg_type = PTR_TO_MEM; + aux->btf_var.mem_size = tsize; + } else { + aux->btf_var.reg_type = PTR_TO_BTF_ID; + aux->btf_var.btf_id = type; + } + return 0; +} + static int check_map_prealloc(struct bpf_map *map) { return (map->map_type != BPF_MAP_TYPE_HASH && @@ -9534,10 +9621,14 @@ static bool bpf_map_is_cgroup_storage(struct bpf_map *map) map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE); }
-/* look for pseudo eBPF instructions that access map FDs and - * replace them with actual map pointers +/* find and rewrite pseudo imm in ld_imm64 instructions: + * + * 1. if it accesses map FD, replace it with actual map pointer. + * 2. if it accesses btf_id of a VAR, replace it with pointer to the var. + * + * NOTE: btf_vmlinux is required for converting pseudo btf_id. */ -static int replace_map_fd_with_map_ptr(struct bpf_verifier_env *env) +static int resolve_pseudo_ldimm64(struct bpf_verifier_env *env) { struct bpf_insn *insn = env->prog->insnsi; int insn_cnt = env->prog->len; @@ -9578,6 +9669,14 @@ static int replace_map_fd_with_map_ptr(struct bpf_verifier_env *env) /* valid generic load 64-bit imm */ goto next_insn;
+ if (insn[0].src_reg == BPF_PSEUDO_BTF_ID) { + aux = &env->insn_aux_data[i]; + err = check_pseudo_btf_id(env, insn, aux); + if (err) + return err; + goto next_insn; + } + /* In final convert_pseudo_ld_imm64() step, this is * converted into regular 64-bit imm load insn. */ @@ -11633,10 +11732,6 @@ int bpf_check(struct bpf_prog **prog, union bpf_attr *attr, if (is_priv) env->test_state_freq = attr->prog_flags & BPF_F_TEST_STATE_FREQ;
- ret = replace_map_fd_with_map_ptr(env); - if (ret < 0) - goto skip_full_check; - if (bpf_prog_is_dev_bound(env->prog->aux)) { ret = bpf_prog_offload_verifier_prep(env->prog); if (ret) @@ -11662,6 +11757,10 @@ int bpf_check(struct bpf_prog **prog, union bpf_attr *attr, if (ret) goto skip_full_check;
+ ret = resolve_pseudo_ldimm64(env); + if (ret < 0) + goto skip_full_check; + ret = check_cfg(env); if (ret < 0) goto skip_full_check; diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index 2b1d3f16cbd1..1defa5f23078 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -356,18 +356,36 @@ enum bpf_link_type { #define BPF_F_SLEEPABLE (1U << 4)
/* When BPF ldimm64's insn[0].src_reg != 0 then this can have - * two extensions: - * - * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE - * insn[0].imm: map fd map fd - * insn[1].imm: 0 offset into value - * insn[0].off: 0 0 - * insn[1].off: 0 0 - * ldimm64 rewrite: address of map address of map[0]+offset - * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE + * the following extensions: + * + * insn[0].src_reg: BPF_PSEUDO_MAP_FD + * insn[0].imm: map fd + * insn[1].imm: 0 + * insn[0].off: 0 + * insn[1].off: 0 + * ldimm64 rewrite: address of map + * verifier type: CONST_PTR_TO_MAP */ #define BPF_PSEUDO_MAP_FD 1 +/* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE + * insn[0].imm: map fd + * insn[1].imm: offset into value + * insn[0].off: 0 + * insn[1].off: 0 + * ldimm64 rewrite: address of map[0]+offset + * verifier type: PTR_TO_MAP_VALUE + */ #define BPF_PSEUDO_MAP_VALUE 2 +/* insn[0].src_reg: BPF_PSEUDO_BTF_ID + * insn[0].imm: kernel btd id of VAR + * insn[1].imm: 0 + * insn[0].off: 0 + * insn[1].off: 0 + * ldimm64 rewrite: address of the kernel variable + * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var + * is struct/union. + */ +#define BPF_PSEUDO_BTF_ID 3
/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative * offset to another bpf function
On Tue, Sep 29, 2020 at 4:50 PM Hao Luo haoluo@google.com wrote:
ret = replace_map_fd_with_map_ptr(env);if (ret < 0)goto skip_full_check;if (bpf_prog_is_dev_bound(env->prog->aux)) { ret = bpf_prog_offload_verifier_prep(env->prog); if (ret)@@ -11662,6 +11757,10 @@ int bpf_check(struct bpf_prog **prog, union bpf_attr *attr, if (ret) goto skip_full_check;
ret = resolve_pseudo_ldimm64(env);if (ret < 0)goto skip_full_check;
Hao,
this change broke several tests in test_verifier: #21/u empty prog FAIL Unexpected error message! EXP: unknown opcode 00 RES: last insn is not an exit or jmp
#656/u test5 ld_imm64 FAIL Unexpected error message! EXP: invalid bpf_ld_imm64 insn RES: last insn is not an exit or jmp
#656/p test5 ld_imm64 FAIL Unexpected error message! EXP: invalid bpf_ld_imm64 insn RES: last insn is not an exit or jmp
Please send a fix. Thanks
Ack. Will do.
On Tue, Oct 6, 2020 at 12:05 PM Alexei Starovoitov alexei.starovoitov@gmail.com wrote:
On Tue, Sep 29, 2020 at 4:50 PM Hao Luo haoluo@google.com wrote:
ret = replace_map_fd_with_map_ptr(env);if (ret < 0)goto skip_full_check;if (bpf_prog_is_dev_bound(env->prog->aux)) { ret = bpf_prog_offload_verifier_prep(env->prog); if (ret)@@ -11662,6 +11757,10 @@ int bpf_check(struct bpf_prog **prog, union bpf_attr *attr, if (ret) goto skip_full_check;
ret = resolve_pseudo_ldimm64(env);if (ret < 0)goto skip_full_check;Hao,
this change broke several tests in test_verifier: #21/u empty prog FAIL Unexpected error message! EXP: unknown opcode 00 RES: last insn is not an exit or jmp
#656/u test5 ld_imm64 FAIL Unexpected error message! EXP: invalid bpf_ld_imm64 insn RES: last insn is not an exit or jmp
#656/p test5 ld_imm64 FAIL Unexpected error message! EXP: invalid bpf_ld_imm64 insn RES: last insn is not an exit or jmp
Please send a fix. Thanks
If a ksym is defined with a type, libbpf will try to find the ksym's btf information from kernel btf. If a valid btf entry for the ksym is found, libbpf can pass in the found btf id to the verifier, which validates the ksym's type and value.
Typeless ksyms (i.e. those defined as 'void') will not have such btf_id, but it has the symbol's address (read from kallsyms) and its value is treated as a raw pointer.
Acked-by: Andrii Nakryiko andriin@fb.com Signed-off-by: Hao Luo haoluo@google.com --- tools/lib/bpf/libbpf.c | 112 ++++++++++++++++++++++++++++++++++++----- 1 file changed, 99 insertions(+), 13 deletions(-)
diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c index a4f55f8a460d..9b36c52b8511 100644 --- a/tools/lib/bpf/libbpf.c +++ b/tools/lib/bpf/libbpf.c @@ -390,6 +390,12 @@ struct extern_desc { } kcfg; struct { unsigned long long addr; + + /* target btf_id of the corresponding kernel var. */ + int vmlinux_btf_id; + + /* local btf_id of the ksym extern's type. */ + __u32 type_id; } ksym; }; }; @@ -2522,12 +2528,23 @@ static int bpf_object__load_vmlinux_btf(struct bpf_object *obj) { bool need_vmlinux_btf = false; struct bpf_program *prog; - int err; + int i, err;
/* CO-RE relocations need kernel BTF */ if (obj->btf_ext && obj->btf_ext->core_relo_info.len) need_vmlinux_btf = true;
+ /* Support for typed ksyms needs kernel BTF */ + for (i = 0; i < obj->nr_extern; i++) { + const struct extern_desc *ext; + + ext = &obj->externs[i]; + if (ext->type == EXT_KSYM && ext->ksym.type_id) { + need_vmlinux_btf = true; + break; + } + } + bpf_object__for_each_program(prog, obj) { if (!prog->load) continue; @@ -3156,16 +3173,10 @@ static int bpf_object__collect_externs(struct bpf_object *obj) return -ENOTSUP; } } else if (strcmp(sec_name, KSYMS_SEC) == 0) { - const struct btf_type *vt; - ksym_sec = sec; ext->type = EXT_KSYM; - - vt = skip_mods_and_typedefs(obj->btf, t->type, NULL); - if (!btf_is_void(vt)) { - pr_warn("extern (ksym) '%s' is not typeless (void)\n", ext_name); - return -ENOTSUP; - } + skip_mods_and_typedefs(obj->btf, t->type, + &ext->ksym.type_id); } else { pr_warn("unrecognized extern section '%s'\n", sec_name); return -ENOTSUP; @@ -5800,8 +5811,13 @@ bpf_object__relocate_data(struct bpf_object *obj, struct bpf_program *prog) insn[0].imm = obj->maps[obj->kconfig_map_idx].fd; insn[1].imm = ext->kcfg.data_off; } else /* EXT_KSYM */ { - insn[0].imm = (__u32)ext->ksym.addr; - insn[1].imm = ext->ksym.addr >> 32; + if (ext->ksym.type_id) { /* typed ksyms */ + insn[0].src_reg = BPF_PSEUDO_BTF_ID; + insn[0].imm = ext->ksym.vmlinux_btf_id; + } else { /* typeless ksyms */ + insn[0].imm = (__u32)ext->ksym.addr; + insn[1].imm = ext->ksym.addr >> 32; + } } relo->processed = true; break; @@ -6933,10 +6949,72 @@ static int bpf_object__read_kallsyms_file(struct bpf_object *obj) return err; }
+static int bpf_object__resolve_ksyms_btf_id(struct bpf_object *obj) +{ + struct extern_desc *ext; + int i, id; + + for (i = 0; i < obj->nr_extern; i++) { + const struct btf_type *targ_var, *targ_type; + __u32 targ_type_id, local_type_id; + const char *targ_var_name; + int ret; + + ext = &obj->externs[i]; + if (ext->type != EXT_KSYM || !ext->ksym.type_id) + continue; + + id = btf__find_by_name_kind(obj->btf_vmlinux, ext->name, + BTF_KIND_VAR); + if (id <= 0) { + pr_warn("extern (ksym) '%s': failed to find BTF ID in vmlinux BTF.\n", + ext->name); + return -ESRCH; + } + + /* find local type_id */ + local_type_id = ext->ksym.type_id; + + /* find target type_id */ + targ_var = btf__type_by_id(obj->btf_vmlinux, id); + targ_var_name = btf__name_by_offset(obj->btf_vmlinux, + targ_var->name_off); + targ_type = skip_mods_and_typedefs(obj->btf_vmlinux, + targ_var->type, + &targ_type_id); + + ret = bpf_core_types_are_compat(obj->btf, local_type_id, + obj->btf_vmlinux, targ_type_id); + if (ret <= 0) { + const struct btf_type *local_type; + const char *targ_name, *local_name; + + local_type = btf__type_by_id(obj->btf, local_type_id); + local_name = btf__name_by_offset(obj->btf, + local_type->name_off); + targ_name = btf__name_by_offset(obj->btf_vmlinux, + targ_type->name_off); + + pr_warn("extern (ksym) '%s': incompatible types, expected [%d] %s %s, but kernel has [%d] %s %s\n", + ext->name, local_type_id, + btf_kind_str(local_type), local_name, targ_type_id, + btf_kind_str(targ_type), targ_name); + return -EINVAL; + } + + ext->is_set = true; + ext->ksym.vmlinux_btf_id = id; + pr_debug("extern (ksym) '%s': resolved to [%d] %s %s\n", + ext->name, id, btf_kind_str(targ_var), targ_var_name); + } + return 0; +} + static int bpf_object__resolve_externs(struct bpf_object *obj, const char *extra_kconfig) { bool need_config = false, need_kallsyms = false; + bool need_vmlinux_btf = false; struct extern_desc *ext; void *kcfg_data = NULL; int err, i; @@ -6967,7 +7045,10 @@ static int bpf_object__resolve_externs(struct bpf_object *obj, strncmp(ext->name, "CONFIG_", 7) == 0) { need_config = true; } else if (ext->type == EXT_KSYM) { - need_kallsyms = true; + if (ext->ksym.type_id) + need_vmlinux_btf = true; + else + need_kallsyms = true; } else { pr_warn("unrecognized extern '%s'\n", ext->name); return -EINVAL; @@ -6996,6 +7077,11 @@ static int bpf_object__resolve_externs(struct bpf_object *obj, if (err) return -EINVAL; } + if (need_vmlinux_btf) { + err = bpf_object__resolve_ksyms_btf_id(obj); + if (err) + return -EINVAL; + } for (i = 0; i < obj->nr_extern; i++) { ext = &obj->externs[i];
@@ -7028,10 +7114,10 @@ int bpf_object__load_xattr(struct bpf_object_load_attr *attr) }
err = bpf_object__probe_loading(obj); + err = err ? : bpf_object__load_vmlinux_btf(obj); err = err ? : bpf_object__resolve_externs(obj, obj->kconfig); err = err ? : bpf_object__sanitize_and_load_btf(obj); err = err ? : bpf_object__sanitize_maps(obj); - err = err ? : bpf_object__load_vmlinux_btf(obj); err = err ? : bpf_object__init_kern_struct_ops_maps(obj); err = err ? : bpf_object__create_maps(obj); err = err ? : bpf_object__relocate(obj, attr->target_btf_path);
Selftests for typed ksyms. Tests two types of ksyms: one is a struct, the other is a plain int. This tests two paths in the kernel. Struct ksyms will be converted into PTR_TO_BTF_ID by the verifier while int typed ksyms will be converted into PTR_TO_MEM.
Acked-by: Andrii Nakryiko andriin@fb.com Signed-off-by: Hao Luo haoluo@google.com --- .../testing/selftests/bpf/prog_tests/ksyms.c | 38 ++++------ .../selftests/bpf/prog_tests/ksyms_btf.c | 70 +++++++++++++++++++ .../selftests/bpf/progs/test_ksyms_btf.c | 23 ++++++ tools/testing/selftests/bpf/trace_helpers.c | 27 +++++++ tools/testing/selftests/bpf/trace_helpers.h | 4 ++ 5 files changed, 137 insertions(+), 25 deletions(-) create mode 100644 tools/testing/selftests/bpf/prog_tests/ksyms_btf.c create mode 100644 tools/testing/selftests/bpf/progs/test_ksyms_btf.c
diff --git a/tools/testing/selftests/bpf/prog_tests/ksyms.c b/tools/testing/selftests/bpf/prog_tests/ksyms.c index b771804b2342..b295969b263b 100644 --- a/tools/testing/selftests/bpf/prog_tests/ksyms.c +++ b/tools/testing/selftests/bpf/prog_tests/ksyms.c @@ -7,40 +7,28 @@
static int duration;
-static __u64 kallsyms_find(const char *sym) -{ - char type, name[500]; - __u64 addr, res = 0; - FILE *f; - - f = fopen("/proc/kallsyms", "r"); - if (CHECK(!f, "kallsyms_fopen", "failed to open: %d\n", errno)) - return 0; - - while (fscanf(f, "%llx %c %499s%*[^\n]\n", &addr, &type, name) > 0) { - if (strcmp(name, sym) == 0) { - res = addr; - goto out; - } - } - - CHECK(false, "not_found", "symbol %s not found\n", sym); -out: - fclose(f); - return res; -} - void test_ksyms(void) { - __u64 per_cpu_start_addr = kallsyms_find("__per_cpu_start"); - __u64 link_fops_addr = kallsyms_find("bpf_link_fops"); const char *btf_path = "/sys/kernel/btf/vmlinux"; struct test_ksyms *skel; struct test_ksyms__data *data; + __u64 link_fops_addr, per_cpu_start_addr; struct stat st; __u64 btf_size; int err;
+ err = kallsyms_find("bpf_link_fops", &link_fops_addr); + if (CHECK(err == -EINVAL, "kallsyms_fopen", "failed to open: %d\n", errno)) + return; + if (CHECK(err == -ENOENT, "ksym_find", "symbol 'bpf_link_fops' not found\n")) + return; + + err = kallsyms_find("__per_cpu_start", &per_cpu_start_addr); + if (CHECK(err == -EINVAL, "kallsyms_fopen", "failed to open: %d\n", errno)) + return; + if (CHECK(err == -ENOENT, "ksym_find", "symbol 'per_cpu_start' not found\n")) + return; + if (CHECK(stat(btf_path, &st), "stat_btf", "err %d\n", errno)) return; btf_size = st.st_size; diff --git a/tools/testing/selftests/bpf/prog_tests/ksyms_btf.c b/tools/testing/selftests/bpf/prog_tests/ksyms_btf.c new file mode 100644 index 000000000000..c6ef06c0629a --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/ksyms_btf.c @@ -0,0 +1,70 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2020 Google */ + +#include <test_progs.h> +#include <bpf/libbpf.h> +#include <bpf/btf.h> +#include "test_ksyms_btf.skel.h" + +static int duration; + +void test_ksyms_btf(void) +{ + __u64 runqueues_addr, bpf_prog_active_addr; + struct test_ksyms_btf *skel = NULL; + struct test_ksyms_btf__data *data; + struct btf *btf; + int percpu_datasec; + int err; + + err = kallsyms_find("runqueues", &runqueues_addr); + if (CHECK(err == -EINVAL, "kallsyms_fopen", "failed to open: %d\n", errno)) + return; + if (CHECK(err == -ENOENT, "ksym_find", "symbol 'runqueues' not found\n")) + return; + + err = kallsyms_find("bpf_prog_active", &bpf_prog_active_addr); + if (CHECK(err == -EINVAL, "kallsyms_fopen", "failed to open: %d\n", errno)) + return; + if (CHECK(err == -ENOENT, "ksym_find", "symbol 'bpf_prog_active' not found\n")) + return; + + btf = libbpf_find_kernel_btf(); + if (CHECK(IS_ERR(btf), "btf_exists", "failed to load kernel BTF: %ld\n", + PTR_ERR(btf))) + return; + + percpu_datasec = btf__find_by_name_kind(btf, ".data..percpu", + BTF_KIND_DATASEC); + if (percpu_datasec < 0) { + printf("%s:SKIP:no PERCPU DATASEC in kernel btf\n", + __func__); + test__skip(); + goto cleanup; + } + + skel = test_ksyms_btf__open_and_load(); + if (CHECK(!skel, "skel_open", "failed to open and load skeleton\n")) + goto cleanup; + + err = test_ksyms_btf__attach(skel); + if (CHECK(err, "skel_attach", "skeleton attach failed: %d\n", err)) + goto cleanup; + + /* trigger tracepoint */ + usleep(1); + + data = skel->data; + CHECK(data->out__runqueues_addr != runqueues_addr, "runqueues_addr", + "got %llu, exp %llu\n", + (unsigned long long)data->out__runqueues_addr, + (unsigned long long)runqueues_addr); + CHECK(data->out__bpf_prog_active_addr != bpf_prog_active_addr, "bpf_prog_active_addr", + "got %llu, exp %llu\n", + (unsigned long long)data->out__bpf_prog_active_addr, + (unsigned long long)bpf_prog_active_addr); + +cleanup: + btf__free(btf); + test_ksyms_btf__destroy(skel); +} diff --git a/tools/testing/selftests/bpf/progs/test_ksyms_btf.c b/tools/testing/selftests/bpf/progs/test_ksyms_btf.c new file mode 100644 index 000000000000..7dde2082131d --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_ksyms_btf.c @@ -0,0 +1,23 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2020 Google */ + +#include "vmlinux.h" + +#include <bpf/bpf_helpers.h> + +__u64 out__runqueues_addr = -1; +__u64 out__bpf_prog_active_addr = -1; + +extern const struct rq runqueues __ksym; /* struct type global var. */ +extern const int bpf_prog_active __ksym; /* int type global var. */ + +SEC("raw_tp/sys_enter") +int handler(const void *ctx) +{ + out__runqueues_addr = (__u64)&runqueues; + out__bpf_prog_active_addr = (__u64)&bpf_prog_active; + + return 0; +} + +char _license[] SEC("license") = "GPL"; diff --git a/tools/testing/selftests/bpf/trace_helpers.c b/tools/testing/selftests/bpf/trace_helpers.c index 4d0e913bbb22..1bbd1d9830c8 100644 --- a/tools/testing/selftests/bpf/trace_helpers.c +++ b/tools/testing/selftests/bpf/trace_helpers.c @@ -90,6 +90,33 @@ long ksym_get_addr(const char *name) return 0; }
+/* open kallsyms and read symbol addresses on the fly. Without caching all symbols, + * this is faster than load + find. + */ +int kallsyms_find(const char *sym, unsigned long long *addr) +{ + char type, name[500]; + unsigned long long value; + int err = 0; + FILE *f; + + f = fopen("/proc/kallsyms", "r"); + if (!f) + return -EINVAL; + + while (fscanf(f, "%llx %c %499s%*[^\n]\n", &value, &type, name) > 0) { + if (strcmp(name, sym) == 0) { + *addr = value; + goto out; + } + } + err = -ENOENT; + +out: + fclose(f); + return err; +} + void read_trace_pipe(void) { int trace_fd; diff --git a/tools/testing/selftests/bpf/trace_helpers.h b/tools/testing/selftests/bpf/trace_helpers.h index 25ef597dd03f..f62fdef9e589 100644 --- a/tools/testing/selftests/bpf/trace_helpers.h +++ b/tools/testing/selftests/bpf/trace_helpers.h @@ -12,6 +12,10 @@ struct ksym { int load_kallsyms(void); struct ksym *ksym_search(long key); long ksym_get_addr(const char *name); + +/* open kallsyms and find addresses on the fly, faster than load + search. */ +int kallsyms_find(const char *sym, unsigned long long *addr); + void read_trace_pipe(void);
#endif
Add bpf_per_cpu_ptr() to help bpf programs access percpu vars. bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the kernel except that it may return NULL. This happens when the cpu parameter is out of range. So the caller must check the returned value.
Acked-by: Andrii Nakryiko andriin@fb.com Signed-off-by: Hao Luo haoluo@google.com --- include/linux/bpf.h | 4 +++ include/linux/btf.h | 11 ++++++ include/uapi/linux/bpf.h | 18 ++++++++++ kernel/bpf/btf.c | 10 ------ kernel/bpf/helpers.c | 18 ++++++++++ kernel/bpf/verifier.c | 64 ++++++++++++++++++++++++++++++++-- kernel/trace/bpf_trace.c | 2 ++ tools/include/uapi/linux/bpf.h | 18 ++++++++++ 8 files changed, 132 insertions(+), 13 deletions(-)
diff --git a/include/linux/bpf.h b/include/linux/bpf.h index 50e5c4b52bd1..9dde15b2479d 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -293,6 +293,7 @@ enum bpf_arg_type { ARG_PTR_TO_ALLOC_MEM_OR_NULL, /* pointer to dynamically allocated memory or NULL */ ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ + ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ __BPF_ARG_TYPE_MAX, };
@@ -307,6 +308,7 @@ enum bpf_return_type { RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */ RET_PTR_TO_ALLOC_MEM_OR_NULL, /* returns a pointer to dynamically allocated memory or NULL */ RET_PTR_TO_BTF_ID_OR_NULL, /* returns a pointer to a btf_id or NULL */ + RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL, /* returns a pointer to a valid memory or a btf_id or NULL */ };
/* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs @@ -405,6 +407,7 @@ enum bpf_reg_type { PTR_TO_RDONLY_BUF_OR_NULL, /* reg points to a readonly buffer or NULL */ PTR_TO_RDWR_BUF, /* reg points to a read/write buffer */ PTR_TO_RDWR_BUF_OR_NULL, /* reg points to a read/write buffer or NULL */ + PTR_TO_PERCPU_BTF_ID, /* reg points to a percpu kernel variable */ };
/* The information passed from prog-specific *_is_valid_access @@ -1828,6 +1831,7 @@ extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; extern const struct bpf_func_proto bpf_copy_from_user_proto; extern const struct bpf_func_proto bpf_snprintf_btf_proto; +extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
const struct bpf_func_proto *bpf_tracing_func_proto( enum bpf_func_id func_id, const struct bpf_prog *prog); diff --git a/include/linux/btf.h b/include/linux/btf.h index af1244180588..2bf641829664 100644 --- a/include/linux/btf.h +++ b/include/linux/btf.h @@ -110,6 +110,11 @@ btf_resolve_size(const struct btf *btf, const struct btf_type *type, i < btf_type_vlen(struct_type); \ i++, member++)
+#define for_each_vsi(i, datasec_type, member) \ + for (i = 0, member = btf_type_var_secinfo(datasec_type); \ + i < btf_type_vlen(datasec_type); \ + i++, member++) + static inline bool btf_type_is_ptr(const struct btf_type *t) { return BTF_INFO_KIND(t->info) == BTF_KIND_PTR; @@ -194,6 +199,12 @@ static inline const struct btf_member *btf_type_member(const struct btf_type *t) return (const struct btf_member *)(t + 1); }
+static inline const struct btf_var_secinfo *btf_type_var_secinfo( + const struct btf_type *t) +{ + return (const struct btf_var_secinfo *)(t + 1); +} + #ifdef CONFIG_BPF_SYSCALL const struct btf_type *btf_type_by_id(const struct btf *btf, u32 type_id); const char *btf_name_by_offset(const struct btf *btf, u32 offset); diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 1defa5f23078..feae87eaa8c6 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -3661,6 +3661,23 @@ union bpf_attr { * *flags* are identical to those used for bpf_snprintf_btf. * Return * 0 on success or a negative error in case of failure. + * + * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) + * Description + * Take a pointer to a percpu ksym, *percpu_ptr*, and return a + * pointer to the percpu kernel variable on *cpu*. A ksym is an + * extern variable decorated with '__ksym'. For ksym, there is a + * global var (either static or global) defined of the same name + * in the kernel. The ksym is percpu if the global var is percpu. + * The returned pointer points to the global percpu var on *cpu*. + * + * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the + * kernel, except that bpf_per_cpu_ptr() may return NULL. This + * happens if *cpu* is larger than nr_cpu_ids. The caller of + * bpf_per_cpu_ptr() must check the returned value. + * Return + * A pointer pointing to the kernel percpu variable on *cpu*, or + * NULL, if *cpu* is invalid. */ #define __BPF_FUNC_MAPPER(FN) \ FN(unspec), \ @@ -3814,6 +3831,7 @@ union bpf_attr { FN(copy_from_user), \ FN(snprintf_btf), \ FN(seq_printf_btf), \ + FN(bpf_per_cpu_ptr), \ /* */
/* integer value in 'imm' field of BPF_CALL instruction selects which helper diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c index 00569afe3d0d..ed7d02e8bc93 100644 --- a/kernel/bpf/btf.c +++ b/kernel/bpf/btf.c @@ -188,11 +188,6 @@ i < btf_type_vlen(struct_type); \ i++, member++)
-#define for_each_vsi(i, struct_type, member) \ - for (i = 0, member = btf_type_var_secinfo(struct_type); \ - i < btf_type_vlen(struct_type); \ - i++, member++) - #define for_each_vsi_from(i, from, struct_type, member) \ for (i = from, member = btf_type_var_secinfo(struct_type) + from; \ i < btf_type_vlen(struct_type); \ @@ -598,11 +593,6 @@ static const struct btf_var *btf_type_var(const struct btf_type *t) return (const struct btf_var *)(t + 1); }
-static const struct btf_var_secinfo *btf_type_var_secinfo(const struct btf_type *t) -{ - return (const struct btf_var_secinfo *)(t + 1); -} - static const struct btf_kind_operations *btf_type_ops(const struct btf_type *t) { return kind_ops[BTF_INFO_KIND(t->info)]; diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c index e825441781ab..14fe3f64fd82 100644 --- a/kernel/bpf/helpers.c +++ b/kernel/bpf/helpers.c @@ -623,6 +623,22 @@ const struct bpf_func_proto bpf_copy_from_user_proto = { .arg3_type = ARG_ANYTHING, };
+BPF_CALL_2(bpf_per_cpu_ptr, const void *, ptr, u32, cpu) +{ + if (cpu >= nr_cpu_ids) + return (unsigned long)NULL; + + return (unsigned long)per_cpu_ptr((const void __percpu *)ptr, cpu); +} + +const struct bpf_func_proto bpf_per_cpu_ptr_proto = { + .func = bpf_per_cpu_ptr, + .gpl_only = false, + .ret_type = RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL, + .arg1_type = ARG_PTR_TO_PERCPU_BTF_ID, + .arg2_type = ARG_ANYTHING, +}; + const struct bpf_func_proto bpf_get_current_task_proto __weak; const struct bpf_func_proto bpf_probe_read_user_proto __weak; const struct bpf_func_proto bpf_probe_read_user_str_proto __weak; @@ -689,6 +705,8 @@ bpf_base_func_proto(enum bpf_func_id func_id) return &bpf_snprintf_btf_proto; case BPF_FUNC_jiffies64: return &bpf_jiffies64_proto; + case BPF_FUNC_bpf_per_cpu_ptr: + return &bpf_per_cpu_ptr_proto; default: break; } diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index fe4965079773..216b8ece23ce 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -238,6 +238,8 @@ struct bpf_call_arg_meta { u64 msize_max_value; int ref_obj_id; int func_id; + u32 btf_id; + u32 ret_btf_id; };
struct btf *btf_vmlinux; @@ -517,6 +519,7 @@ static const char * const reg_type_str[] = { [PTR_TO_XDP_SOCK] = "xdp_sock", [PTR_TO_BTF_ID] = "ptr_", [PTR_TO_BTF_ID_OR_NULL] = "ptr_or_null_", + [PTR_TO_PERCPU_BTF_ID] = "percpu_ptr_", [PTR_TO_MEM] = "mem", [PTR_TO_MEM_OR_NULL] = "mem_or_null", [PTR_TO_RDONLY_BUF] = "rdonly_buf", @@ -583,7 +586,9 @@ static void print_verifier_state(struct bpf_verifier_env *env, /* reg->off should be 0 for SCALAR_VALUE */ verbose(env, "%lld", reg->var_off.value + reg->off); } else { - if (t == PTR_TO_BTF_ID || t == PTR_TO_BTF_ID_OR_NULL) + if (t == PTR_TO_BTF_ID || + t == PTR_TO_BTF_ID_OR_NULL || + t == PTR_TO_PERCPU_BTF_ID) verbose(env, "%s", kernel_type_name(reg->btf_id)); verbose(env, "(id=%d", reg->id); if (reg_type_may_be_refcounted_or_null(t)) @@ -2204,6 +2209,7 @@ static bool is_spillable_regtype(enum bpf_reg_type type) case PTR_TO_RDONLY_BUF_OR_NULL: case PTR_TO_RDWR_BUF: case PTR_TO_RDWR_BUF_OR_NULL: + case PTR_TO_PERCPU_BTF_ID: return true; default: return false; @@ -4017,6 +4023,7 @@ static const struct bpf_reg_types alloc_mem_types = { .types = { PTR_TO_MEM } }; static const struct bpf_reg_types const_map_ptr_types = { .types = { CONST_PTR_TO_MAP } }; static const struct bpf_reg_types btf_ptr_types = { .types = { PTR_TO_BTF_ID } }; static const struct bpf_reg_types spin_lock_types = { .types = { PTR_TO_MAP_VALUE } }; +static const struct bpf_reg_types percpu_btf_ptr_types = { .types = { PTR_TO_PERCPU_BTF_ID } };
static const struct bpf_reg_types *compatible_reg_types[__BPF_ARG_TYPE_MAX] = { [ARG_PTR_TO_MAP_KEY] = &map_key_value_types, @@ -4042,6 +4049,7 @@ static const struct bpf_reg_types *compatible_reg_types[__BPF_ARG_TYPE_MAX] = { [ARG_PTR_TO_ALLOC_MEM_OR_NULL] = &alloc_mem_types, [ARG_PTR_TO_INT] = &int_ptr_types, [ARG_PTR_TO_LONG] = &int_ptr_types, + [ARG_PTR_TO_PERCPU_BTF_ID] = &percpu_btf_ptr_types, };
static int check_reg_type(struct bpf_verifier_env *env, u32 regno, @@ -4205,6 +4213,12 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg, err = check_helper_mem_access(env, regno, meta->map_ptr->value_size, false, meta); + } else if (arg_type == ARG_PTR_TO_PERCPU_BTF_ID) { + if (!reg->btf_id) { + verbose(env, "Helper has invalid btf_id in R%d\n", regno); + return -EACCES; + } + meta->ret_btf_id = reg->btf_id; } else if (arg_type == ARG_PTR_TO_SPIN_LOCK) { if (meta->func_id == BPF_FUNC_spin_lock) { if (process_spin_lock(env, regno, true)) @@ -5114,6 +5128,30 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn regs[BPF_REG_0].type = PTR_TO_MEM_OR_NULL; regs[BPF_REG_0].id = ++env->id_gen; regs[BPF_REG_0].mem_size = meta.mem_size; + } else if (fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL) { + const struct btf_type *t; + + mark_reg_known_zero(env, regs, BPF_REG_0); + t = btf_type_skip_modifiers(btf_vmlinux, meta.ret_btf_id, NULL); + if (!btf_type_is_struct(t)) { + u32 tsize; + const struct btf_type *ret; + const char *tname; + + /* resolve the type size of ksym. */ + ret = btf_resolve_size(btf_vmlinux, t, &tsize); + if (IS_ERR(ret)) { + tname = btf_name_by_offset(btf_vmlinux, t->name_off); + verbose(env, "unable to resolve the size of type '%s': %ld\n", + tname, PTR_ERR(ret)); + return -EINVAL; + } + regs[BPF_REG_0].type = PTR_TO_MEM_OR_NULL; + regs[BPF_REG_0].mem_size = tsize; + } else { + regs[BPF_REG_0].type = PTR_TO_BTF_ID_OR_NULL; + regs[BPF_REG_0].btf_id = meta.ret_btf_id; + } } else if (fn->ret_type == RET_PTR_TO_BTF_ID_OR_NULL) { int ret_btf_id;
@@ -7523,6 +7561,7 @@ static int check_ld_imm(struct bpf_verifier_env *env, struct bpf_insn *insn) dst_reg->mem_size = aux->btf_var.mem_size; break; case PTR_TO_BTF_ID: + case PTR_TO_PERCPU_BTF_ID: dst_reg->btf_id = aux->btf_var.btf_id; break; default: @@ -9449,10 +9488,14 @@ static int check_pseudo_btf_id(struct bpf_verifier_env *env, struct bpf_insn *insn, struct bpf_insn_aux_data *aux) { - u32 type, id = insn->imm; + u32 datasec_id, type, id = insn->imm; + const struct btf_var_secinfo *vsi; + const struct btf_type *datasec; const struct btf_type *t; const char *sym_name; + bool percpu = false; u64 addr; + int i;
if (!btf_vmlinux) { verbose(env, "kernel is missing BTF, make sure CONFIG_DEBUG_INFO_BTF=y is specified in Kconfig.\n"); @@ -9484,12 +9527,27 @@ static int check_pseudo_btf_id(struct bpf_verifier_env *env, return -ENOENT; }
+ datasec_id = btf_find_by_name_kind(btf_vmlinux, ".data..percpu", + BTF_KIND_DATASEC); + if (datasec_id > 0) { + datasec = btf_type_by_id(btf_vmlinux, datasec_id); + for_each_vsi(i, datasec, vsi) { + if (vsi->type == id) { + percpu = true; + break; + } + } + } + insn[0].imm = (u32)addr; insn[1].imm = addr >> 32;
type = t->type; t = btf_type_skip_modifiers(btf_vmlinux, type, NULL); - if (!btf_type_is_struct(t)) { + if (percpu) { + aux->btf_var.reg_type = PTR_TO_PERCPU_BTF_ID; + aux->btf_var.btf_id = type; + } else if (!btf_type_is_struct(t)) { const struct btf_type *ret; const char *tname; u32 tsize; diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c index e118a83439c3..364a322e2898 100644 --- a/kernel/trace/bpf_trace.c +++ b/kernel/trace/bpf_trace.c @@ -1327,6 +1327,8 @@ bpf_tracing_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) return prog->aux->sleepable ? &bpf_copy_from_user_proto : NULL; case BPF_FUNC_snprintf_btf: return &bpf_snprintf_btf_proto; + case BPF_FUNC_bpf_per_cpu_ptr: + return &bpf_per_cpu_ptr_proto; default: return NULL; } diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index 1defa5f23078..feae87eaa8c6 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -3661,6 +3661,23 @@ union bpf_attr { * *flags* are identical to those used for bpf_snprintf_btf. * Return * 0 on success or a negative error in case of failure. + * + * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) + * Description + * Take a pointer to a percpu ksym, *percpu_ptr*, and return a + * pointer to the percpu kernel variable on *cpu*. A ksym is an + * extern variable decorated with '__ksym'. For ksym, there is a + * global var (either static or global) defined of the same name + * in the kernel. The ksym is percpu if the global var is percpu. + * The returned pointer points to the global percpu var on *cpu*. + * + * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the + * kernel, except that bpf_per_cpu_ptr() may return NULL. This + * happens if *cpu* is larger than nr_cpu_ids. The caller of + * bpf_per_cpu_ptr() must check the returned value. + * Return + * A pointer pointing to the kernel percpu variable on *cpu*, or + * NULL, if *cpu* is invalid. */ #define __BPF_FUNC_MAPPER(FN) \ FN(unspec), \ @@ -3814,6 +3831,7 @@ union bpf_attr { FN(copy_from_user), \ FN(snprintf_btf), \ FN(seq_printf_btf), \ + FN(bpf_per_cpu_ptr), \ /* */
/* integer value in 'imm' field of BPF_CALL instruction selects which helper
Add bpf_this_cpu_ptr() to help access percpu var on this cpu. This helper always returns a valid pointer, therefore no need to check returned value for NULL. Also note that all programs run with preemption disabled, which means that the returned pointer is stable during all the execution of the program.
Acked-by: Andrii Nakryiko andriin@fb.com Signed-off-by: Hao Luo haoluo@google.com --- include/linux/bpf.h | 2 ++ include/uapi/linux/bpf.h | 13 +++++++++++++ kernel/bpf/helpers.c | 14 ++++++++++++++ kernel/bpf/verifier.c | 11 ++++++++--- kernel/trace/bpf_trace.c | 2 ++ tools/include/uapi/linux/bpf.h | 13 +++++++++++++ 6 files changed, 52 insertions(+), 3 deletions(-)
diff --git a/include/linux/bpf.h b/include/linux/bpf.h index 9dde15b2479d..dc63eeed4fd9 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -309,6 +309,7 @@ enum bpf_return_type { RET_PTR_TO_ALLOC_MEM_OR_NULL, /* returns a pointer to dynamically allocated memory or NULL */ RET_PTR_TO_BTF_ID_OR_NULL, /* returns a pointer to a btf_id or NULL */ RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL, /* returns a pointer to a valid memory or a btf_id or NULL */ + RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ };
/* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs @@ -1832,6 +1833,7 @@ extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; extern const struct bpf_func_proto bpf_copy_from_user_proto; extern const struct bpf_func_proto bpf_snprintf_btf_proto; extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; +extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
const struct bpf_func_proto *bpf_tracing_func_proto( enum bpf_func_id func_id, const struct bpf_prog *prog); diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index feae87eaa8c6..8b360fd42094 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -3678,6 +3678,18 @@ union bpf_attr { * Return * A pointer pointing to the kernel percpu variable on *cpu*, or * NULL, if *cpu* is invalid. + * + * void *bpf_this_cpu_ptr(const void *percpu_ptr) + * Description + * Take a pointer to a percpu ksym, *percpu_ptr*, and return a + * pointer to the percpu kernel variable on this cpu. See the + * description of 'ksym' in **bpf_per_cpu_ptr**\ (). + * + * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in + * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would + * never return NULL. + * Return + * A pointer pointing to the kernel percpu variable on this cpu. */ #define __BPF_FUNC_MAPPER(FN) \ FN(unspec), \ @@ -3832,6 +3844,7 @@ union bpf_attr { FN(snprintf_btf), \ FN(seq_printf_btf), \ FN(bpf_per_cpu_ptr), \ + FN(bpf_this_cpu_ptr), \ /* */
/* integer value in 'imm' field of BPF_CALL instruction selects which helper diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c index 14fe3f64fd82..25520f5eeaf6 100644 --- a/kernel/bpf/helpers.c +++ b/kernel/bpf/helpers.c @@ -639,6 +639,18 @@ const struct bpf_func_proto bpf_per_cpu_ptr_proto = { .arg2_type = ARG_ANYTHING, };
+BPF_CALL_1(bpf_this_cpu_ptr, const void *, percpu_ptr) +{ + return (unsigned long)this_cpu_ptr((const void __percpu *)percpu_ptr); +} + +const struct bpf_func_proto bpf_this_cpu_ptr_proto = { + .func = bpf_this_cpu_ptr, + .gpl_only = false, + .ret_type = RET_PTR_TO_MEM_OR_BTF_ID, + .arg1_type = ARG_PTR_TO_PERCPU_BTF_ID, +}; + const struct bpf_func_proto bpf_get_current_task_proto __weak; const struct bpf_func_proto bpf_probe_read_user_proto __weak; const struct bpf_func_proto bpf_probe_read_user_str_proto __weak; @@ -707,6 +719,8 @@ bpf_base_func_proto(enum bpf_func_id func_id) return &bpf_jiffies64_proto; case BPF_FUNC_bpf_per_cpu_ptr: return &bpf_per_cpu_ptr_proto; + case BPF_FUNC_bpf_this_cpu_ptr: + return &bpf_this_cpu_ptr_proto; default: break; } diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 216b8ece23ce..d9dbf271ebab 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -5128,7 +5128,8 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn regs[BPF_REG_0].type = PTR_TO_MEM_OR_NULL; regs[BPF_REG_0].id = ++env->id_gen; regs[BPF_REG_0].mem_size = meta.mem_size; - } else if (fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL) { + } else if (fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL || + fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID) { const struct btf_type *t;
mark_reg_known_zero(env, regs, BPF_REG_0); @@ -5146,10 +5147,14 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn tname, PTR_ERR(ret)); return -EINVAL; } - regs[BPF_REG_0].type = PTR_TO_MEM_OR_NULL; + regs[BPF_REG_0].type = + fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID ? + PTR_TO_MEM : PTR_TO_MEM_OR_NULL; regs[BPF_REG_0].mem_size = tsize; } else { - regs[BPF_REG_0].type = PTR_TO_BTF_ID_OR_NULL; + regs[BPF_REG_0].type = + fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID ? + PTR_TO_BTF_ID : PTR_TO_BTF_ID_OR_NULL; regs[BPF_REG_0].btf_id = meta.ret_btf_id; } } else if (fn->ret_type == RET_PTR_TO_BTF_ID_OR_NULL) { diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c index 364a322e2898..a136a6a63a71 100644 --- a/kernel/trace/bpf_trace.c +++ b/kernel/trace/bpf_trace.c @@ -1329,6 +1329,8 @@ bpf_tracing_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) return &bpf_snprintf_btf_proto; case BPF_FUNC_bpf_per_cpu_ptr: return &bpf_per_cpu_ptr_proto; + case BPF_FUNC_bpf_this_cpu_ptr: + return &bpf_this_cpu_ptr_proto; default: return NULL; } diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index feae87eaa8c6..8b360fd42094 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -3678,6 +3678,18 @@ union bpf_attr { * Return * A pointer pointing to the kernel percpu variable on *cpu*, or * NULL, if *cpu* is invalid. + * + * void *bpf_this_cpu_ptr(const void *percpu_ptr) + * Description + * Take a pointer to a percpu ksym, *percpu_ptr*, and return a + * pointer to the percpu kernel variable on this cpu. See the + * description of 'ksym' in **bpf_per_cpu_ptr**\ (). + * + * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in + * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would + * never return NULL. + * Return + * A pointer pointing to the kernel percpu variable on this cpu. */ #define __BPF_FUNC_MAPPER(FN) \ FN(unspec), \ @@ -3832,6 +3844,7 @@ union bpf_attr { FN(snprintf_btf), \ FN(seq_printf_btf), \ FN(bpf_per_cpu_ptr), \ + FN(bpf_this_cpu_ptr), \ /* */
/* integer value in 'imm' field of BPF_CALL instruction selects which helper
Test bpf_per_cpu_ptr() and bpf_this_cpu_ptr(). Test two paths in the kernel. If the base pointer points to a struct, the returned reg is of type PTR_TO_BTF_ID. Direct pointer dereference can be applied on the returned variable. If the base pointer isn't a struct, the returned reg is of type PTR_TO_MEM, which also supports direct pointer dereference.
Acked-by: Andrii Nakryiko andriin@fb.com Signed-off-by: Hao Luo haoluo@google.com --- .../selftests/bpf/prog_tests/ksyms_btf.c | 18 +++++++++++ .../selftests/bpf/progs/test_ksyms_btf.c | 32 +++++++++++++++++++ 2 files changed, 50 insertions(+)
diff --git a/tools/testing/selftests/bpf/prog_tests/ksyms_btf.c b/tools/testing/selftests/bpf/prog_tests/ksyms_btf.c index c6ef06c0629a..28e26bd3e0ca 100644 --- a/tools/testing/selftests/bpf/prog_tests/ksyms_btf.c +++ b/tools/testing/selftests/bpf/prog_tests/ksyms_btf.c @@ -11,6 +11,8 @@ static int duration; void test_ksyms_btf(void) { __u64 runqueues_addr, bpf_prog_active_addr; + __u32 this_rq_cpu; + int this_bpf_prog_active; struct test_ksyms_btf *skel = NULL; struct test_ksyms_btf__data *data; struct btf *btf; @@ -64,6 +66,22 @@ void test_ksyms_btf(void) (unsigned long long)data->out__bpf_prog_active_addr, (unsigned long long)bpf_prog_active_addr);
+ CHECK(data->out__rq_cpu == -1, "rq_cpu", + "got %u, exp != -1\n", data->out__rq_cpu); + CHECK(data->out__bpf_prog_active < 0, "bpf_prog_active", + "got %d, exp >= 0\n", data->out__bpf_prog_active); + CHECK(data->out__cpu_0_rq_cpu != 0, "cpu_rq(0)->cpu", + "got %u, exp 0\n", data->out__cpu_0_rq_cpu); + + this_rq_cpu = data->out__this_rq_cpu; + CHECK(this_rq_cpu != data->out__rq_cpu, "this_rq_cpu", + "got %u, exp %u\n", this_rq_cpu, data->out__rq_cpu); + + this_bpf_prog_active = data->out__this_bpf_prog_active; + CHECK(this_bpf_prog_active != data->out__bpf_prog_active, "this_bpf_prog_active", + "got %d, exp %d\n", this_bpf_prog_active, + data->out__bpf_prog_active); + cleanup: btf__free(btf); test_ksyms_btf__destroy(skel); diff --git a/tools/testing/selftests/bpf/progs/test_ksyms_btf.c b/tools/testing/selftests/bpf/progs/test_ksyms_btf.c index 7dde2082131d..bb8ea9270f29 100644 --- a/tools/testing/selftests/bpf/progs/test_ksyms_btf.c +++ b/tools/testing/selftests/bpf/progs/test_ksyms_btf.c @@ -8,15 +8,47 @@ __u64 out__runqueues_addr = -1; __u64 out__bpf_prog_active_addr = -1;
+__u32 out__rq_cpu = -1; /* percpu struct fields */ +int out__bpf_prog_active = -1; /* percpu int */ + +__u32 out__this_rq_cpu = -1; +int out__this_bpf_prog_active = -1; + +__u32 out__cpu_0_rq_cpu = -1; /* cpu_rq(0)->cpu */ + extern const struct rq runqueues __ksym; /* struct type global var. */ extern const int bpf_prog_active __ksym; /* int type global var. */
SEC("raw_tp/sys_enter") int handler(const void *ctx) { + struct rq *rq; + int *active; + __u32 cpu; + out__runqueues_addr = (__u64)&runqueues; out__bpf_prog_active_addr = (__u64)&bpf_prog_active;
+ cpu = bpf_get_smp_processor_id(); + + /* test bpf_per_cpu_ptr() */ + rq = (struct rq *)bpf_per_cpu_ptr(&runqueues, cpu); + if (rq) + out__rq_cpu = rq->cpu; + active = (int *)bpf_per_cpu_ptr(&bpf_prog_active, cpu); + if (active) + out__bpf_prog_active = *active; + + rq = (struct rq *)bpf_per_cpu_ptr(&runqueues, 0); + if (rq) /* should always be valid, but we can't spare the check. */ + out__cpu_0_rq_cpu = rq->cpu; + + /* test bpf_this_cpu_ptr */ + rq = (struct rq *)bpf_this_cpu_ptr(&runqueues); + out__this_rq_cpu = rq->cpu; + active = (int *)bpf_this_cpu_ptr(&bpf_prog_active); + out__this_bpf_prog_active = *active; + return 0; }
On Tue, Sep 29, 2020 at 4:50 PM Hao Luo haoluo@google.com wrote:
v3 -> v4:
- Rebasing
- Cast bpf_[per|this]_cpu_ptr's parameter to void __percpu * before passing into per_cpu_ptr.
Looks good, but doesn't work: ./test_progs -t ksyms_btf test_ksyms_btf:PASS:kallsyms_fopen 0 nsec test_ksyms_btf:PASS:ksym_find 0 nsec test_ksyms_btf:PASS:kallsyms_fopen 0 nsec test_ksyms_btf:PASS:ksym_find 0 nsec test_ksyms_btf:PASS:btf_exists 0 nsec libbpf: extern (ksym) 'bpf_prog_active': incompatible types, expected [4] int int, but kernel has [18729] var bpf_user_rnd_state libbpf: failed to load object 'test_ksyms_btf' libbpf: failed to load BPF skeleton 'test_ksyms_btf': -22 test_ksyms_btf:FAIL:skel_open failed to open and load skeleton #43 ksyms_btf:FAIL
I have the latest pahole from master. Any ideas?
Ah, this is the bug in pahole described in https://lkml.org/lkml/2020/8/20/1862. I proposed a fix [1] but it hasn't reached pahole's master branch. Let me ask Arnaldo to see if he is OK merging it.
[1] https://www.spinics.net/lists/dwarves/msg00451.html
On Tue, Sep 29, 2020 at 9:36 PM Alexei Starovoitov alexei.starovoitov@gmail.com wrote:
On Tue, Sep 29, 2020 at 4:50 PM Hao Luo haoluo@google.com wrote:
v3 -> v4:
- Rebasing
- Cast bpf_[per|this]_cpu_ptr's parameter to void __percpu * before passing into per_cpu_ptr.
Looks good, but doesn't work: ./test_progs -t ksyms_btf test_ksyms_btf:PASS:kallsyms_fopen 0 nsec test_ksyms_btf:PASS:ksym_find 0 nsec test_ksyms_btf:PASS:kallsyms_fopen 0 nsec test_ksyms_btf:PASS:ksym_find 0 nsec test_ksyms_btf:PASS:btf_exists 0 nsec libbpf: extern (ksym) 'bpf_prog_active': incompatible types, expected [4] int int, but kernel has [18729] var bpf_user_rnd_state libbpf: failed to load object 'test_ksyms_btf' libbpf: failed to load BPF skeleton 'test_ksyms_btf': -22 test_ksyms_btf:FAIL:skel_open failed to open and load skeleton #43 ksyms_btf:FAIL
I have the latest pahole from master. Any ideas?
On Tue, Sep 29, 2020 at 11:48 PM Hao Luo haoluo@google.com wrote:
Ah, this is the bug in pahole described in https://lkml.org/lkml/2020/8/20/1862. I proposed a fix [1] but it hasn't reached pahole's master branch. Let me ask Arnaldo to see if he is OK merging it.
[1] https://www.spinics.net/lists/dwarves/msg00451.html
On Tue, Sep 29, 2020 at 9:36 PM Alexei Starovoitov alexei.starovoitov@gmail.com wrote:
On Tue, Sep 29, 2020 at 4:50 PM Hao Luo haoluo@google.com wrote:
v3 -> v4:
- Rebasing
- Cast bpf_[per|this]_cpu_ptr's parameter to void __percpu * before passing into per_cpu_ptr.
I've rebased it myself and applied. Thanks Hao.
Thanks, Alexei and Andrii and other reviewers for the comments. It's a pleasure to work with you and contribute to bpf.
Hao
On Fri, Oct 2, 2020 at 3:16 PM Alexei Starovoitov alexei.starovoitov@gmail.com wrote:
On Tue, Sep 29, 2020 at 11:48 PM Hao Luo haoluo@google.com wrote:
Ah, this is the bug in pahole described in https://lkml.org/lkml/2020/8/20/1862. I proposed a fix [1] but it hasn't reached pahole's master branch. Let me ask Arnaldo to see if he is OK merging it.
[1] https://www.spinics.net/lists/dwarves/msg00451.html
On Tue, Sep 29, 2020 at 9:36 PM Alexei Starovoitov alexei.starovoitov@gmail.com wrote:
On Tue, Sep 29, 2020 at 4:50 PM Hao Luo haoluo@google.com wrote:
v3 -> v4:
- Rebasing
- Cast bpf_[per|this]_cpu_ptr's parameter to void __percpu * before passing into per_cpu_ptr.
I've rebased it myself and applied. Thanks Hao.
linux-kselftest-mirror@lists.linaro.org
-
Alexei Starovoitov -
Hao Luo