[PATCH bpf-next 0/2] bpf: support proper non-jit func info

[PATCH bpf-next 0/2] bpf: support proper non-jit func info

[Yonghong Song]

Commit 838e96904ff3 ("bpf: Introduce bpf_func_info")
added bpf func info support. The userspace is able
to get better ksym's for bpf programs with jit, and
is able to print out func prototypes.

For a program containing func-to-func calls, the existing
implementation returns user specified number of function
calls and BTF types if jit is enabled. If the jit is not
enabled, it only returns the type for the main function.

This is undesirable. Interpreter may still be used
and we should keep feature identical regardless of
whether jit is enabled or not.
This patch fixed this discrepancy.

The following example shows bpftool output for
the bpf program in selftests test_btf_haskv.o when jit
is disabled:
  $ bpftool prog dump xlated id 1490
  int _dummy_tracepoint(struct dummy_tracepoint_args * arg):
     0: (85) call pc+2#__bpf_prog_run_args32
     1: (b7) r0 = 0
     2: (95) exit
  int test_long_fname_1(struct dummy_tracepoint_args * arg):
     3: (85) call pc+1#__bpf_prog_run_args32
     4: (95) exit
  int test_long_fname_2(struct dummy_tracepoint_args * arg):
     5: (b7) r2 = 0
     6: (63) *(u32 *)(r10 -4) = r2
     7: (79) r1 = *(u64 *)(r1 +8)
     8: (15) if r1 == 0x0 goto pc+9
     9: (bf) r2 = r10
    10: (07) r2 += -4
    11: (18) r1 = map[id:1173]
    13: (85) call bpf_map_lookup_elem#77088
    14: (15) if r0 == 0x0 goto pc+3
    15: (61) r1 = *(u32 *)(r0 +4)
    16: (07) r1 += 1
    17: (63) *(u32 *)(r0 +4) = r1
    18: (95) exit
  $ bpftool prog dump jited id 1490
    no instructions returned

Yonghong Song (2):
  bpf: btf: support proper non-jit func info
  tools/bpf: change selftest test_btf for both jit and non-jit

 include/linux/bpf.h                    |  6 ++-
 include/linux/bpf_verifier.h           |  1 -
 kernel/bpf/core.c                      |  3 +-
 kernel/bpf/syscall.c                   | 33 ++++------------
 kernel/bpf/verifier.c                  | 55 ++++++++++++++++++--------
 tools/testing/selftests/bpf/test_btf.c | 33 ++--------------
 6 files changed, 55 insertions(+), 76 deletions(-)

-- 
2.17.1

[PATCH bpf-next 1/2] bpf: btf: support proper non-jit func info

[Yonghong Song]

Commit 838e96904ff3 ("bpf: Introduce bpf_func_info")
added bpf func info support. The userspace is able
to get better ksym's for bpf programs with jit, and
is able to print out func prototypes.

For a program containing func-to-func calls, the existing
implementation returns user specified number of function
calls and BTF types if jit is enabled. If the jit is not
enabled, it only returns the type for the main function.

This is undesirable. Interpreter may still be used
and we should keep feature identical regardless of
whether jit is enabled or not.
This patch fixed this discrepancy.

Fixes: 838e96904ff3 ("bpf: Introduce bpf_func_info")
Signed-off-by: Yonghong Song <yhs@fb.com>
---
 include/linux/bpf.h          |  6 ++--
 include/linux/bpf_verifier.h |  1 -
 kernel/bpf/core.c            |  3 +-
 kernel/bpf/syscall.c         | 33 ++++++----------------
 kernel/bpf/verifier.c        | 55 +++++++++++++++++++++++++-----------
 5 files changed, 52 insertions(+), 46 deletions(-)

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 7f0e225bf630..e82b7039fc66 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -299,7 +299,8 @@ struct bpf_prog_aux {
 	u32 max_pkt_offset;
 	u32 stack_depth;
 	u32 id;
-	u32 func_cnt;
+	u32 func_cnt; /* used by non-func prog as the number of func progs */
+	u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
 	bool offload_requested;
 	struct bpf_prog **func;
 	void *jit_data; /* JIT specific data. arch dependent */
@@ -317,7 +318,8 @@ struct bpf_prog_aux {
 #endif
 	struct bpf_prog_offload *offload;
 	struct btf *btf;
-	u32 type_id; /* type id for this prog/func */
+	struct bpf_func_info *func_info;
+	u32 func_info_cnt;
 	union {
 		struct work_struct work;
 		struct rcu_head	rcu;
diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h
index 204382f46fd8..11f5df1092d9 100644
--- a/include/linux/bpf_verifier.h
+++ b/include/linux/bpf_verifier.h
@@ -204,7 +204,6 @@ static inline bool bpf_verifier_log_needed(const struct bpf_verifier_log *log)
 struct bpf_subprog_info {
 	u32 start; /* insn idx of function entry point */
 	u16 stack_depth; /* max. stack depth used by this function */
-	u32 type_id; /* btf type_id for this subprog */
 };
 
 /* single container for all structs
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index 16d77012ad3e..002d67c62c8b 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -411,7 +411,8 @@ static void bpf_get_prog_name(const struct bpf_prog *prog, char *sym)
 
 	/* prog->aux->name will be ignored if full btf name is available */
 	if (prog->aux->btf) {
-		type = btf_type_by_id(prog->aux->btf, prog->aux->type_id);
+		type = btf_type_by_id(prog->aux->btf,
+				      prog->aux->func_info[prog->aux->func_idx].type_id);
 		func_name = btf_name_by_offset(prog->aux->btf, type->name_off);
 		snprintf(sym, (size_t)(end - sym), "_%s", func_name);
 		return;
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 998377808102..85cbeec06e50 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -1214,6 +1214,7 @@ static void __bpf_prog_put(struct bpf_prog *prog, bool do_idr_lock)
 		bpf_prog_free_id(prog, do_idr_lock);
 		bpf_prog_kallsyms_del_all(prog);
 		btf_put(prog->aux->btf);
+		kvfree(prog->aux->func_info);
 
 		call_rcu(&prog->aux->rcu, __bpf_prog_put_rcu);
 	}
@@ -2219,46 +2220,28 @@ static int bpf_prog_get_info_by_fd(struct bpf_prog *prog,
 	}
 
 	if (prog->aux->btf) {
+		u32 krec_size = sizeof(struct bpf_func_info);
 		u32 ucnt, urec_size;
 
 		info.btf_id = btf_id(prog->aux->btf);
 
 		ucnt = info.func_info_cnt;
-		info.func_info_cnt = prog->aux->func_cnt ? : 1;
+		info.func_info_cnt = prog->aux->func_info_cnt;
 		urec_size = info.func_info_rec_size;
-		info.func_info_rec_size = sizeof(struct bpf_func_info);
+		info.func_info_rec_size = krec_size;
 		if (ucnt) {
 			/* expect passed-in urec_size is what the kernel expects */
 			if (urec_size != info.func_info_rec_size)
 				return -EINVAL;
 
 			if (bpf_dump_raw_ok()) {
-				struct bpf_func_info kern_finfo;
 				char __user *user_finfo;
-				u32 i, insn_offset;
 
 				user_finfo = u64_to_user_ptr(info.func_info);
-				if (prog->aux->func_cnt) {
-					ucnt = min_t(u32, info.func_info_cnt, ucnt);
-					insn_offset = 0;
-					for (i = 0; i < ucnt; i++) {
-						kern_finfo.insn_offset = insn_offset;
-						kern_finfo.type_id = prog->aux->func[i]->aux->type_id;
-						if (copy_to_user(user_finfo, &kern_finfo,
-								 sizeof(kern_finfo)))
-							return -EFAULT;
-
-						/* func[i]->len holds the prog len */
-						insn_offset += prog->aux->func[i]->len;
-						user_finfo += urec_size;
-					}
-				} else {
-					kern_finfo.insn_offset = 0;
-					kern_finfo.type_id = prog->aux->type_id;
-					if (copy_to_user(user_finfo, &kern_finfo,
-							 sizeof(kern_finfo)))
-						return -EFAULT;
-				}
+				ucnt = min_t(u32, info.func_info_cnt, ucnt);
+				if (copy_to_user(user_finfo, prog->aux->func_info,
+						 krec_size * ucnt))
+					return -EFAULT;
 			} else {
 				info.func_info_cnt = 0;
 			}
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index f102c4fd0c5a..05d95c0e4a26 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -4650,7 +4650,7 @@ static int check_btf_func(struct bpf_prog *prog, struct bpf_verifier_env *env,
 {
 	u32 i, nfuncs, urec_size, min_size, prev_offset;
 	u32 krec_size = sizeof(struct bpf_func_info);
-	struct bpf_func_info krecord = {};
+	struct bpf_func_info *krecord = NULL;
 	const struct btf_type *type;
 	void __user *urecord;
 	struct btf *btf;
@@ -4682,6 +4682,12 @@ static int check_btf_func(struct bpf_prog *prog, struct bpf_verifier_env *env,
 	urecord = u64_to_user_ptr(attr->func_info);
 	min_size = min_t(u32, krec_size, urec_size);
 
+	krecord = kvcalloc(nfuncs, krec_size, GFP_KERNEL | __GFP_NOWARN);
+	if (!krecord) {
+		ret = -ENOMEM;
+		goto free_btf;
+	}
+
 	for (i = 0; i < nfuncs; i++) {
 		ret = bpf_check_uarg_tail_zero(urecord, krec_size, urec_size);
 		if (ret) {
@@ -4696,59 +4702,69 @@ static int check_btf_func(struct bpf_prog *prog, struct bpf_verifier_env *env,
 			goto free_btf;
 		}
 
-		if (copy_from_user(&krecord, urecord, min_size)) {
+		if (copy_from_user(&krecord[i], urecord, min_size)) {
 			ret = -EFAULT;
 			goto free_btf;
 		}
 
 		/* check insn_offset */
 		if (i == 0) {
-			if (krecord.insn_offset) {
+			if (krecord[i].insn_offset) {
 				verbose(env,
 					"nonzero insn_offset %u for the first func info record",
-					krecord.insn_offset);
+					krecord[i].insn_offset);
 				ret = -EINVAL;
 				goto free_btf;
 			}
-		} else if (krecord.insn_offset <= prev_offset) {
+		} else if (krecord[i].insn_offset <= prev_offset) {
 			verbose(env,
 				"same or smaller insn offset (%u) than previous func info record (%u)",
-				krecord.insn_offset, prev_offset);
+				krecord[i].insn_offset, prev_offset);
 			ret = -EINVAL;
 			goto free_btf;
 		}
 
-		if (env->subprog_info[i].start != krecord.insn_offset) {
+		if (env->subprog_info[i].start != krecord[i].insn_offset) {
 			verbose(env, "func_info BTF section doesn't match subprog layout in BPF program\n");
 			ret = -EINVAL;
 			goto free_btf;
 		}
 
 		/* check type_id */
-		type = btf_type_by_id(btf, krecord.type_id);
+		type = btf_type_by_id(btf, krecord[i].type_id);
 		if (!type || BTF_INFO_KIND(type->info) != BTF_KIND_FUNC) {
 			verbose(env, "invalid type id %d in func info",
-				krecord.type_id);
+				krecord[i].type_id);
 			ret = -EINVAL;
 			goto free_btf;
 		}
 
-		if (i == 0)
-			prog->aux->type_id = krecord.type_id;
-		env->subprog_info[i].type_id = krecord.type_id;
-
-		prev_offset = krecord.insn_offset;
+		prev_offset = krecord[i].insn_offset;
 		urecord += urec_size;
 	}
 
 	prog->aux->btf = btf;
+	prog->aux->func_info = krecord;
+	prog->aux->func_info_cnt = nfuncs;
 	return 0;
 
 free_btf:
 	btf_put(btf);
+	kvfree(krecord);
 	return ret;
 }
 
+static void adjust_btf_func(struct bpf_verifier_env *env)
+{
+	int i;
+
+	if (!env->prog->aux->func_info)
+		return;
+
+	for (i = 0; i < env->subprog_cnt; i++)
+		env->prog->aux->func_info[i].insn_offset = env->subprog_info[i].start;
+}
+
 /* check %cur's range satisfies %old's */
 static bool range_within(struct bpf_reg_state *old,
 			 struct bpf_reg_state *cur)
@@ -6043,15 +6059,17 @@ static int jit_subprogs(struct bpf_verifier_env *env)
 		if (bpf_prog_calc_tag(func[i]))
 			goto out_free;
 		func[i]->is_func = 1;
+		func[i]->aux->func_idx = i;
+		/* the btf and func_info will be freed only at prog->aux */
+		func[i]->aux->btf = prog->aux->btf;
+		func[i]->aux->func_info = prog->aux->func_info;
+
 		/* Use bpf_prog_F_tag to indicate functions in stack traces.
 		 * Long term would need debug info to populate names
 		 */
 		func[i]->aux->name[0] = 'F';
 		func[i]->aux->stack_depth = env->subprog_info[i].stack_depth;
 		func[i]->jit_requested = 1;
-		/* the btf will be freed only at prog->aux */
-		func[i]->aux->btf = prog->aux->btf;
-		func[i]->aux->type_id = env->subprog_info[i].type_id;
 		func[i] = bpf_int_jit_compile(func[i]);
 		if (!func[i]->jited) {
 			err = -ENOTSUPP;
@@ -6572,6 +6590,9 @@ int bpf_check(struct bpf_prog **prog, union bpf_attr *attr,
 		convert_pseudo_ld_imm64(env);
 	}
 
+	if (ret == 0)
+		adjust_btf_func(env);
+
 err_release_maps:
 	if (!env->prog->aux->used_maps)
 		/* if we didn't copy map pointers into bpf_prog_info, release
-- 
2.17.1

[PATCH bpf-next 2/2] tools/bpf: change selftest test_btf for both jit and non-jit

[Yonghong Song]

The selftest test_btf is changed to test both jit and non-jit.
The test result should be the same regardless of whether jit
is enabled or not.

Signed-off-by: Yonghong Song <yhs@fb.com>
---
 tools/testing/selftests/bpf/test_btf.c | 33 +++-----------------------
 1 file changed, 3 insertions(+), 30 deletions(-)

diff --git a/tools/testing/selftests/bpf/test_btf.c b/tools/testing/selftests/bpf/test_btf.c
index bcbda7037840..b4c8725b3004 100644
--- a/tools/testing/selftests/bpf/test_btf.c
+++ b/tools/testing/selftests/bpf/test_btf.c
@@ -29,7 +29,6 @@
 static uint32_t pass_cnt;
 static uint32_t error_cnt;
 static uint32_t skip_cnt;
-static bool jit_enabled;
 
 #define CHECK(condition, format...) ({					\
 	int __ret = !!(condition);					\
@@ -65,24 +64,6 @@ static int __base_pr(const char *format, ...)
 	return err;
 }
 
-static bool is_jit_enabled(void)
-{
-	const char *jit_sysctl = "/proc/sys/net/core/bpf_jit_enable";
-	bool enabled = false;
-	int sysctl_fd;
-
-	sysctl_fd = open(jit_sysctl, 0, O_RDONLY);
-	if (sysctl_fd != -1) {
-		char tmpc;
-
-		if (read(sysctl_fd, &tmpc, sizeof(tmpc)) == 1)
-			enabled = (tmpc != '0');
-		close(sysctl_fd);
-	}
-
-	return enabled;
-}
-
 #define BTF_INFO_ENC(kind, root, vlen)			\
 	((!!(root) << 31) | ((kind) << 24) | ((vlen) & BTF_MAX_VLEN))
 
@@ -2547,8 +2528,8 @@ static int do_test_file(unsigned int test_num)
 		  test->btf_kv_notfound))
 		goto done;
 
-	if (!jit_enabled || !has_btf_ext)
-		goto skip_jit;
+	if (!has_btf_ext)
+		goto skip;
 
 	/* get necessary program info */
 	info_len = sizeof(struct bpf_prog_info);
@@ -2636,7 +2617,7 @@ static int do_test_file(unsigned int test_num)
 		finfo = (void *)finfo + rec_size;
 	}
 
-skip_jit:
+skip:
 	fprintf(stderr, "OK");
 
 done:
@@ -3270,12 +3251,6 @@ static int do_test_func_type(int test_num)
 		err = -1;
 		goto done;
 	}
-	if (!jit_enabled) {
-		skip_cnt++;
-		fprintf(stderr, "SKIPPED, please enable sysctl bpf_jit_enable\n");
-		err = 0;
-		goto done;
-	}
 
 	/* get necessary lens */
 	info_len = sizeof(struct bpf_prog_info);
@@ -3452,8 +3427,6 @@ int main(int argc, char **argv)
 	if (args.always_log)
 		libbpf_set_print(__base_pr, __base_pr, __base_pr);
 
-	jit_enabled = is_jit_enabled();
-
 	if (args.raw_test)
 		err |= test_raw();
 
-- 
2.17.1

Re: [PATCH bpf-next 1/2] bpf: btf: support proper non-jit func info

[Martin Lau]

On Sat, Nov 24, 2018 at 11:20:44PM -0800, Yonghong Song wrote:
> Commit 838e96904ff3 ("bpf: Introduce bpf_func_info")
> added bpf func info support. The userspace is able
> to get better ksym's for bpf programs with jit, and
> is able to print out func prototypes.
> 
> For a program containing func-to-func calls, the existing
> implementation returns user specified number of function
> calls and BTF types if jit is enabled. If the jit is not
> enabled, it only returns the type for the main function.
> 
> This is undesirable. Interpreter may still be used
> and we should keep feature identical regardless of
> whether jit is enabled or not.
> This patch fixed this discrepancy.
Acked-by: Martin KaFai Lau <kafai@fb.com>

Re: [PATCH bpf-next 2/2] tools/bpf: change selftest test_btf for both jit and non-jit

[Martin Lau]

On Sat, Nov 24, 2018 at 11:20:45PM -0800, Yonghong Song wrote:
> The selftest test_btf is changed to test both jit and non-jit.
> The test result should be the same regardless of whether jit
> is enabled or not.
Acked-by: Martin KaFai Lau <kafai@fb.com>

Re: [PATCH bpf-next 0/2] bpf: support proper non-jit func info

[Alexei Starovoitov]

On Sat, Nov 24, 2018 at 11:20:43PM -0800, Yonghong Song wrote:
> Commit 838e96904ff3 ("bpf: Introduce bpf_func_info")
> added bpf func info support. The userspace is able
> to get better ksym's for bpf programs with jit, and
> is able to print out func prototypes.
> 
> For a program containing func-to-func calls, the existing
> implementation returns user specified number of function
> calls and BTF types if jit is enabled. If the jit is not
> enabled, it only returns the type for the main function.
> 
> This is undesirable. Interpreter may still be used
> and we should keep feature identical regardless of
> whether jit is enabled or not.
> This patch fixed this discrepancy.
> 
> The following example shows bpftool output for
> the bpf program in selftests test_btf_haskv.o when jit
> is disabled:
>   $ bpftool prog dump xlated id 1490
>   int _dummy_tracepoint(struct dummy_tracepoint_args * arg):
>      0: (85) call pc+2#__bpf_prog_run_args32
>      1: (b7) r0 = 0
>      2: (95) exit
>   int test_long_fname_1(struct dummy_tracepoint_args * arg):
>      3: (85) call pc+1#__bpf_prog_run_args32
>      4: (95) exit
>   int test_long_fname_2(struct dummy_tracepoint_args * arg):
>      5: (b7) r2 = 0
>      6: (63) *(u32 *)(r10 -4) = r2
>      7: (79) r1 = *(u64 *)(r1 +8)
>      8: (15) if r1 == 0x0 goto pc+9
>      9: (bf) r2 = r10
>     10: (07) r2 += -4
>     11: (18) r1 = map[id:1173]
>     13: (85) call bpf_map_lookup_elem#77088
>     14: (15) if r0 == 0x0 goto pc+3
>     15: (61) r1 = *(u32 *)(r0 +4)
>     16: (07) r1 += 1
>     17: (63) *(u32 *)(r0 +4) = r1
>     18: (95) exit
>   $ bpftool prog dump jited id 1490
>     no instructions returned

Applied to bpf-next. Thanks