Skip to contentSkip to conversation

bpf: Add trampoline helpers

PATCH Superseded Jiri Olsa 4 participants 17 messages Jan 21, 2020 – Jan 23, 2020

Subsystems (9)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Jiri Olsa

author reviewer
#0 lore raw

v3 cover bpf: Add trampoline helpers

hi,
adding helpers for trampolines and 2 other fixes to have kernel support for loading trampoline programs in bcc/bpftrace.

Original rfc post [1].

Speedup output of perf bench while running klockstat.py on kprobes vs trampolines:

    Without:
            $ perf bench sched messaging -l 50000
            ...
                 Total time: 18.571 [sec]

    With current kprobe tracing:
            $ perf bench sched messaging -l 50000
            ...
                 Total time: 183.395 [sec]

    With kfunc tracing:
            $ perf bench sched messaging -l 50000
            ...
                 Total time: 39.773 [sec]

v3 changes:
- added ack from John Fastabend for patch 1
- move out is_bpf_image_address from is_bpf_text_address call [David]

v2 changes:
- make the unwind work for dispatcher as well
- added test for allowed trampolines count
- used raw tp pt_regs nest-arrays for trampoline helpers

thanks,
jirka

[1] https://lore.kernel.org/netdev/[email protected]/
---
Jiri Olsa (6):
bpf: Allow ctx access for pointers to scalar
bpf: Add bpf_perf_event_output_kfunc
bpf: Add bpf_get_stackid_kfunc
bpf: Add bpf_get_stack_kfunc
bpf: Allow to resolve bpf trampoline and dispatcher in unwind selftest/bpf: Add test for allowed trampolines count

 include/linux/bpf.h                                       |  12 +++++++++++-
 kernel/bpf/btf.c                                          |  13 ++++++++++++-
 kernel/bpf/dispatcher.c                                   |   4 ++--
 kernel/bpf/trampoline.c                                   |  82 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-------
 kernel/extable.c                                          |   7 +++++--
 kernel/trace/bpf_trace.c                                  |  96 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/testing/selftests/bpf/prog_tests/trampoline_count.c | 112 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/testing/selftests/bpf/progs/test_trampoline_count.c |  21 +++++++++++++++++++++
 8 files changed, 334 insertions(+), 13 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/trampoline_count.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_trampoline_count.c
author reviewer
#1 lore raw

PATCH v3 1/6 bpf: Allow ctx access for pointers to scalar

When accessing the context we allow access to arguments with scalar type and pointer to struct. But we omit pointer to scalar type, which is the case for many functions and same case as when accessing scalar.

Adding the check if the pointer is to scalar type and allow it.

---
 kernel/bpf/btf.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)
kernel/bpf/btf.c +12 -1
index 832b5d7fd892..207ae554e0ce 100644
@@ -3668,7 +3668,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 		    const struct bpf_prog *prog,
 		    struct bpf_insn_access_aux *info)
 {
-	const struct btf_type *t = prog->aux->attach_func_proto;
+	const struct btf_type *tp, *t = prog->aux->attach_func_proto;
 	struct bpf_prog *tgt_prog = prog->aux->linked_prog;
 	struct btf *btf = bpf_prog_get_target_btf(prog);
 	const char *tname = prog->aux->attach_func_name;
@@ -3730,6 +3730,17 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 		 */
 		return true;
 
+	tp = btf_type_by_id(btf, t->type);
+	/* skip modifiers */
+	while (btf_type_is_modifier(tp))
+		tp = btf_type_by_id(btf, tp->type);
+
+	if (btf_type_is_int(tp) || btf_type_is_enum(tp))
+		/* This is a pointer scalar.
+		 * It is the same as scalar from the verifier safety pov.
+		 */
+		return true;
+
 	/* this is a pointer to another type */
 	info->reg_type = PTR_TO_BTF_ID;
 
Signature · 1 line
2.24.1
#2 lore raw

Re: [PATCH 1/6] bpf: Allow ctx access for pointers to scalar

Quoted 31 lines · Jiri Olsa

When accessing the context we allow access to arguments with scalar type and pointer to struct. But we omit pointer to scalar type, which is the case for many functions and same case as when accessing scalar.

Adding the check if the pointer is to scalar type and allow it.

---
 kernel/bpf/btf.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)
kernel/bpf/btf.c +1 -1
index 832b5d7fd892..207ae554e0ce 100644
@@ -3668,7 +3668,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
                    const struct bpf_prog *prog,
                    struct bpf_insn_access_aux *info)
 {
-       const struct btf_type *t = prog->aux->attach_func_proto;
+       const struct btf_type *tp, *t = prog->aux->attach_func_proto;
        struct bpf_prog *tgt_prog = prog->aux->linked_prog;
        struct btf *btf = bpf_prog_get_target_btf(prog);
        const char *tname = prog->aux->attach_func_name;
@@ -3730,6 +3730,17 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
                 */
                return true;
 

kernel/bpf/btf.c bool btf_ctx_access(int off, int size, enum bpf_access_type type, · line 3742

Quoted 10 lines
kernel/bpf/btf.c +10 -0
+       tp = btf_type_by_id(btf, t->type);
+       /* skip modifiers */
+       while (btf_type_is_modifier(tp))
+               tp = btf_type_by_id(btf, tp->type);
+
+       if (btf_type_is_int(tp) || btf_type_is_enum(tp))
+               /* This is a pointer scalar.
+                * It is the same as scalar from the verifier safety pov.
+                */
+               return true;

The reason I didn't do it earlier is I was thinking to represent it as PTR_TO_BTF_ID as well, so that corresponding u8..u64 access into this memory would still be possible.
I'm trying to analyze the situation that returning a scalar now and converting to PTR_TO_BTF_ID in the future will keep progs passing the verifier. Is it really the case?
Could you give a specific example that needs this support? It will help me understand this backward compatibility concern. What prog is doing with that 'u32 *' that is seen as scalar ? It cannot dereference it. Use it as what?

#3 lore raw

Re: [PATCH 1/6] bpf: Allow ctx access for pointers to scalar

Quoted 45 lines · Alexei Starovoitov
Quoted 31 lines · Jiri Olsa

When accessing the context we allow access to arguments with scalar type and pointer to struct. But we omit pointer to scalar type, which is the case for many functions and same case as when accessing scalar.

Adding the check if the pointer is to scalar type and allow it.

---
  kernel/bpf/btf.c | 13 ++++++++++++-
  1 file changed, 12 insertions(+), 1 deletion(-)
kernel/bpf/btf.c +1 -1
index 832b5d7fd892..207ae554e0ce 100644
@@ -3668,7 +3668,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
                     const struct bpf_prog *prog,
                     struct bpf_insn_access_aux *info)
  {
-       const struct btf_type *t = prog->aux->attach_func_proto;
+       const struct btf_type *tp, *t = prog->aux->attach_func_proto;
         struct bpf_prog *tgt_prog = prog->aux->linked_prog;
         struct btf *btf = bpf_prog_get_target_btf(prog);
         const char *tname = prog->aux->attach_func_name;
@@ -3730,6 +3730,17 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
                  */
                 return true;
 

kernel/bpf/btf.c bool btf_ctx_access(int off, int size, enum bpf_access_type type, · line 3742

Quoted 10 lines
kernel/bpf/btf.c +10 -0
+       tp = btf_type_by_id(btf, t->type);
+       /* skip modifiers */
+       while (btf_type_is_modifier(tp))
+               tp = btf_type_by_id(btf, tp->type);
+
+       if (btf_type_is_int(tp) || btf_type_is_enum(tp))
+               /* This is a pointer scalar.
+                * It is the same as scalar from the verifier safety pov.
+                */
+               return true;

The reason I didn't do it earlier is I was thinking to represent it as PTR_TO_BTF_ID as well, so that corresponding u8..u64

Quoted 8 lines

access into this memory would still be possible.
I'm trying to analyze the situation that returning a scalar now and converting to PTR_TO_BTF_ID in the future will keep progs passing the verifier. Is it really the case?
Could you give a specific example that needs this support? It will help me understand this backward compatibility concern. What prog is doing with that 'u32 *' that is seen as scalar ? It cannot dereference it. Use it as what?

If this is from original bcc code, it will use bpf_probe_read for dereference. This is what I understand when I first reviewed this patch. But it will be good to get Jiri's confirmation.

#4 lore raw

Re: [PATCH 1/6] bpf: Allow ctx access for pointers to scalar

Quoted 57 lines · Yonghong Song
Quoted 45 lines · Alexei Starovoitov
Quoted 31 lines · Jiri Olsa

When accessing the context we allow access to arguments with scalar type and pointer to struct. But we omit pointer to scalar type, which is the case for many functions and same case as when accessing scalar.

Adding the check if the pointer is to scalar type and allow it.

---
  kernel/bpf/btf.c | 13 ++++++++++++-
  1 file changed, 12 insertions(+), 1 deletion(-)
kernel/bpf/btf.c +1 -1
index 832b5d7fd892..207ae554e0ce 100644
@@ -3668,7 +3668,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
                     const struct bpf_prog *prog,
                     struct bpf_insn_access_aux *info)
  {
-       const struct btf_type *t = prog->aux->attach_func_proto;
+       const struct btf_type *tp, *t = prog->aux->attach_func_proto;
         struct bpf_prog *tgt_prog = prog->aux->linked_prog;
         struct btf *btf = bpf_prog_get_target_btf(prog);
         const char *tname = prog->aux->attach_func_name;
@@ -3730,6 +3730,17 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
                  */
                 return true;
 

kernel/bpf/btf.c bool btf_ctx_access(int off, int size, enum bpf_access_type type, · line 3742

Quoted 10 lines
kernel/bpf/btf.c +10 -0
+       tp = btf_type_by_id(btf, t->type);
+       /* skip modifiers */
+       while (btf_type_is_modifier(tp))
+               tp = btf_type_by_id(btf, tp->type);
+
+       if (btf_type_is_int(tp) || btf_type_is_enum(tp))
+               /* This is a pointer scalar.
+                * It is the same as scalar from the verifier safety pov.
+                */
+               return true;

The reason I didn't do it earlier is I was thinking to represent it as PTR_TO_BTF_ID as well, so that corresponding u8..u64

Quoted 8 lines

access into this memory would still be possible.
I'm trying to analyze the situation that returning a scalar now and converting to PTR_TO_BTF_ID in the future will keep progs passing the verifier. Is it really the case?
Could you give a specific example that needs this support? It will help me understand this backward compatibility concern. What prog is doing with that 'u32 *' that is seen as scalar ? It cannot dereference it. Use it as what?

Quoted 3 lines

If this is from original bcc code, it will use bpf_probe_read for dereference. This is what I understand when I first reviewed this patch. But it will be good to get Jiri's confirmation.

it blocked me from accessing 'filename' argument when I probed do_sys_open via trampoline in bcc, like:

	KRETFUNC_PROBE(do_sys_open)
	{
	    const char *filename = (const char *) args[1];

AFAICS the current code does not allow for trampoline arguments being other pointers than to void or struct, the patch should detect that the argument is pointer to scalar type and let it pass

jirka

#5 lore raw

Re: [PATCH 1/6] bpf: Allow ctx access for pointers to scalar

Quoted 65 lines · Jiri Olsa
Quoted 57 lines · Yonghong Song
Quoted 45 lines · Alexei Starovoitov
Quoted 31 lines · Jiri Olsa

When accessing the context we allow access to arguments with scalar type and pointer to struct. But we omit pointer to scalar type, which is the case for many functions and same case as when accessing scalar.

Adding the check if the pointer is to scalar type and allow it.

---
  kernel/bpf/btf.c | 13 ++++++++++++-
  1 file changed, 12 insertions(+), 1 deletion(-)
kernel/bpf/btf.c +1 -1
index 832b5d7fd892..207ae554e0ce 100644
@@ -3668,7 +3668,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
                     const struct bpf_prog *prog,
                     struct bpf_insn_access_aux *info)
  {
-       const struct btf_type *t = prog->aux->attach_func_proto;
+       const struct btf_type *tp, *t = prog->aux->attach_func_proto;
         struct bpf_prog *tgt_prog = prog->aux->linked_prog;
         struct btf *btf = bpf_prog_get_target_btf(prog);
         const char *tname = prog->aux->attach_func_name;
@@ -3730,6 +3730,17 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
                  */
                 return true;
 

kernel/bpf/btf.c bool btf_ctx_access(int off, int size, enum bpf_access_type type, · line 3742

Quoted 10 lines
kernel/bpf/btf.c +10 -0
+       tp = btf_type_by_id(btf, t->type);
+       /* skip modifiers */
+       while (btf_type_is_modifier(tp))
+               tp = btf_type_by_id(btf, tp->type);
+
+       if (btf_type_is_int(tp) || btf_type_is_enum(tp))
+               /* This is a pointer scalar.
+                * It is the same as scalar from the verifier safety pov.
+                */
+               return true;

The reason I didn't do it earlier is I was thinking to represent it as PTR_TO_BTF_ID as well, so that corresponding u8..u64

Quoted 8 lines

access into this memory would still be possible.
I'm trying to analyze the situation that returning a scalar now and converting to PTR_TO_BTF_ID in the future will keep progs passing the verifier. Is it really the case?
Could you give a specific example that needs this support? It will help me understand this backward compatibility concern. What prog is doing with that 'u32 *' that is seen as scalar ? It cannot dereference it. Use it as what?

Quoted 3 lines

If this is from original bcc code, it will use bpf_probe_read for dereference. This is what I understand when I first reviewed this patch. But it will be good to get Jiri's confirmation.

it blocked me from accessing 'filename' argument when I probed do_sys_open via trampoline in bcc, like:

Quoted 8 lines
	KRETFUNC_PROBE(do_sys_open)
	{
	    const char *filename = (const char *) args[1];

AFAICS the current code does not allow for trampoline arguments being other pointers than to void or struct, the patch should detect that the argument is pointer to scalar type and let it pass

Got it. I've looked up your bcc patches and I agree that there is no way to workaround. BTF type argument of that kernel function is 'const char *' and the verifier will enforce that if bpf program tries to cast it the verifier will still see 'const char *'. (It's done this way by design). How about we special case 'char *' in the verifier? Then my concern regarding future extensibility of 'int *' and 'long *' will go away.
Compilers have a long history special casing 'char *'. In particular signed char because it's a pointer to null terminated string. I think it's still a special pointer from pointer aliasing point of view. I think the verifier can treat it as scalar here too. In the future the verifier will get smarter and will recognize it as PTR_TO_NULL_STRING while 'u8 *', 'u32 *' will be PTR_TO_BTF_ID. I think it will solve this particular issue. I like conservative approach to the verifier improvements: start with strict checking and relax it on case-by-case. Instead of accepting wide range of cases and cause potential compatibility issues.

#6 lore raw

Re: [PATCH 1/6] bpf: Allow ctx access for pointers to scalar

On Wed, Jan 22, 2020 at 08:09:59AM -0800, Alexei Starovoitov wrote:

SNIP

Quoted 18 lines
Quoted 9 lines
Quoted 5 lines
Quoted 1 line

It cannot dereference it. Use it as what?

If this is from original bcc code, it will use bpf_probe_read for dereference. This is what I understand when I first reviewed this patch. But it will be good to get Jiri's confirmation.

it blocked me from accessing 'filename' argument when I probed do_sys_open via trampoline in bcc, like:

Quoted 8 lines
	KRETFUNC_PROBE(do_sys_open)
	{
	    const char *filename = (const char *) args[1];

AFAICS the current code does not allow for trampoline arguments being other pointers than to void or struct, the patch should detect that the argument is pointer to scalar type and let it pass

Quoted 15 lines

Got it. I've looked up your bcc patches and I agree that there is no way to workaround. BTF type argument of that kernel function is 'const char *' and the verifier will enforce that if bpf program tries to cast it the verifier will still see 'const char *'. (It's done this way by design). How about we special case 'char *' in the verifier? Then my concern regarding future extensibility of 'int *' and 'long *' will go away.
Compilers have a long history special casing 'char *'. In particular signed char because it's a pointer to null terminated string. I think it's still a special pointer from pointer aliasing point of view. I think the verifier can treat it as scalar here too. In the future the verifier will get smarter and will recognize it as PTR_TO_NULL_STRING while 'u8 *', 'u32 *' will be PTR_TO_BTF_ID. I think it will solve this particular issue. I like conservative approach to the verifier improvements: start with strict checking and relax it on case-by-case. Instead of accepting wide range of cases and cause potential compatibility issues.

ok, so something like below?

jirka

---

kernel/bpf/btf.c +16 -0
index 832b5d7fd892..dd678b8e00b7 100644
@@ -3664,6 +3664,19 @@ struct btf *bpf_prog_get_target_btf(const struct bpf_prog *prog)
 	}
 }
 
+static bool is_string_ptr(struct btf *btf, const struct btf_type *t)
+{
+	/* t comes in already as a pointer */
+	t = btf_type_by_id(btf, t->type);
+
+	/* allow const */
+	if (BTF_INFO_KIND(t->info) == BTF_KIND_CONST)
+		t = btf_type_by_id(btf, t->type);
+
+	/* char, signed char, unsigned char */
+	return btf_type_is_int(t) && t->size == 1;
+}
+
 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 		    const struct bpf_prog *prog,
 		    struct bpf_insn_access_aux *info)
@@ -3730,6 +3743,9 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 		 */
 		return true;
 
+	if (is_string_ptr(btf, t))
+		return true;
+
 	/* this is a pointer to another type */
 	info->reg_type = PTR_TO_BTF_ID;
 
#7 lore raw

Re: [PATCH 1/6] bpf: Allow ctx access for pointers to scalar

Quoted 55 lines · Jiri Olsa

On Wed, Jan 22, 2020 at 08:09:59AM -0800, Alexei Starovoitov wrote:

SNIP

Quoted 18 lines
Quoted 9 lines
Quoted 5 lines
Quoted 1 line

It cannot dereference it. Use it as what?

If this is from original bcc code, it will use bpf_probe_read for dereference. This is what I understand when I first reviewed this patch. But it will be good to get Jiri's confirmation.

it blocked me from accessing 'filename' argument when I probed do_sys_open via trampoline in bcc, like:

Quoted 8 lines
    KRETFUNC_PROBE(do_sys_open)
    {
        const char *filename = (const char *) args[1];

AFAICS the current code does not allow for trampoline arguments being other pointers than to void or struct, the patch should detect that the argument is pointer to scalar type and let it pass

Quoted 15 lines

Got it. I've looked up your bcc patches and I agree that there is no way to workaround. BTF type argument of that kernel function is 'const char *' and the verifier will enforce that if bpf program tries to cast it the verifier will still see 'const char *'. (It's done this way by design). How about we special case 'char *' in the verifier? Then my concern regarding future extensibility of 'int *' and 'long *' will go away.
Compilers have a long history special casing 'char *'. In particular signed char because it's a pointer to null terminated string. I think it's still a special pointer from pointer aliasing point of view. I think the verifier can treat it as scalar here too. In the future the verifier will get smarter and will recognize it as PTR_TO_NULL_STRING while 'u8 *', 'u32 *' will be PTR_TO_BTF_ID. I think it will solve this particular issue. I like conservative approach to the verifier improvements: start with strict checking and relax it on case-by-case. Instead of accepting wide range of cases and cause potential compatibility issues.

ok, so something like below?

jirka

---

kernel/bpf/btf.c +2 -0
index 832b5d7fd892..dd678b8e00b7 100644
@@ -3664,6 +3664,19 @@ struct btf *bpf_prog_get_target_btf(const struct bpf_prog *prog)
        }
 }
 
+static bool is_string_ptr(struct btf *btf, const struct btf_type *t)
+{

kernel/bpf/btf.c struct btf *bpf_prog_get_target_btf(const struct bpf_prog *prog) · line 3678

Quoted 10 lines
kernel/bpf/btf.c +10 -0
+       /* t comes in already as a pointer */
+       t = btf_type_by_id(btf, t->type);
+
+       /* allow const */
+       if (BTF_INFO_KIND(t->info) == BTF_KIND_CONST)
+               t = btf_type_by_id(btf, t->type);
+
+       /* char, signed char, unsigned char */
+       return btf_type_is_int(t) && t->size == 1;
+}

yep. looks like btf doesn't distinguish signedness for chars. So above is good.

author reviewer
#8 lore raw

PATCH v3 2/6 bpf: Add bpf_perf_event_output_kfunc

In reply to Jiri Olsa

Adding support to use perf_event_output in
BPF_TRACE_FENTRY/BPF_TRACE_FEXIT programs.

Using nesting regs array from raw tracepoint helpers.

---
 kernel/trace/bpf_trace.c | 41 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 41 insertions(+)
kernel/trace/bpf_trace.c +41 -0
index 19e793aa441a..6a18e2ae6e30 100644
@@ -1172,6 +1172,43 @@ raw_tp_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	}
 }
 
+BPF_CALL_5(bpf_perf_event_output_kfunc, void *, ctx, struct bpf_map *, map,
+	   u64, flags, void *, data, u64, size)
+{
+	struct pt_regs *regs = get_bpf_raw_tp_regs();
+	int ret;
+
+	if (IS_ERR(regs))
+		return PTR_ERR(regs);
+
+	perf_fetch_caller_regs(regs);
+	ret = ____bpf_perf_event_output(regs, map, flags, data, size);
+	put_bpf_raw_tp_regs();
+	return ret;
+}
+
+static const struct bpf_func_proto bpf_perf_event_output_proto_kfunc = {
+	.func		= bpf_perf_event_output_kfunc,
+	.gpl_only	= true,
+	.ret_type	= RET_INTEGER,
+	.arg1_type	= ARG_PTR_TO_CTX,
+	.arg2_type	= ARG_CONST_MAP_PTR,
+	.arg3_type	= ARG_ANYTHING,
+	.arg4_type	= ARG_PTR_TO_MEM,
+	.arg5_type	= ARG_CONST_SIZE_OR_ZERO,
+};
+
+static const struct bpf_func_proto *
+kfunc_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
+{
+	switch (func_id) {
+	case BPF_FUNC_perf_event_output:
+		return &bpf_perf_event_output_proto_kfunc;
+	default:
+		return tracing_func_proto(func_id, prog);
+	}
+}
+
 static const struct bpf_func_proto *
 tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 {
@@ -1181,6 +1218,10 @@ tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_skb_output_proto;
 #endif
 	default:
+		if (prog->expected_attach_type == BPF_TRACE_FENTRY ||
+		    prog->expected_attach_type == BPF_TRACE_FEXIT)
+			return kfunc_prog_func_proto(func_id, prog);
+
 		return raw_tp_prog_func_proto(func_id, prog);
 	}
 }
Signature · 1 line
2.24.1
#9 lore raw

Re: [PATCH 2/6] bpf: Add bpf_perf_event_output_kfunc

Quoted 22 lines · Jiri Olsa

Adding support to use perf_event_output in
BPF_TRACE_FENTRY/BPF_TRACE_FEXIT programs.

Using nesting regs array from raw tracepoint helpers.

---
 kernel/trace/bpf_trace.c | 41 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 41 insertions(+)
kernel/trace/bpf_trace.c +4 -0
index 19e793aa441a..6a18e2ae6e30 100644
@@ -1172,6 +1172,43 @@ raw_tp_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	}
 }
 
+BPF_CALL_5(bpf_perf_event_output_kfunc, void *, ctx, struct bpf_map *, map,
+	   u64, flags, void *, data, u64, size)
+{
+	struct pt_regs *regs = get_bpf_raw_tp_regs();

kernel/trace/bpf_trace.c raw_tp_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) · line 1188

Quoted 10 lines
kernel/trace/bpf_trace.c +10 -0
+	int ret;
+
+	if (IS_ERR(regs))
+		return PTR_ERR(regs);
+
+	perf_fetch_caller_regs(regs);
+	ret = ____bpf_perf_event_output(regs, map, flags, data, size);
+	put_bpf_raw_tp_regs();
+	return ret;
+}

I'm not sure why copy paste bpf_perf_event_output_raw_tp() into new function.

kernel/trace/bpf_trace.c tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) · line 1225

Quoted 9 lines
kernel/trace/bpf_trace.c +4 -0
@@ -1181,6 +1218,10 @@ tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_skb_output_proto;
 #endif
 	default:
+		if (prog->expected_attach_type == BPF_TRACE_FENTRY ||
+		    prog->expected_attach_type == BPF_TRACE_FEXIT)
+			return kfunc_prog_func_proto(func_id, prog);
+
 		return raw_tp_prog_func_proto(func_id, prog);

Are you saying bpf_perf_event_output_raw_tp() for some reason didn't work for fentry/fexit?
But above is exact copy-paste and it somehow worked?

Ditto for patches 3,4.

#10 lore raw

Re: [PATCH 2/6] bpf: Add bpf_perf_event_output_kfunc

Quoted 50 lines · Alexei Starovoitov
Quoted 22 lines · Jiri Olsa

Adding support to use perf_event_output in
BPF_TRACE_FENTRY/BPF_TRACE_FEXIT programs.

Using nesting regs array from raw tracepoint helpers.

---
 kernel/trace/bpf_trace.c | 41 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 41 insertions(+)
kernel/trace/bpf_trace.c +4 -0
index 19e793aa441a..6a18e2ae6e30 100644
@@ -1172,6 +1172,43 @@ raw_tp_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	}
 }
 
+BPF_CALL_5(bpf_perf_event_output_kfunc, void *, ctx, struct bpf_map *, map,
+	   u64, flags, void *, data, u64, size)
+{
+	struct pt_regs *regs = get_bpf_raw_tp_regs();

kernel/trace/bpf_trace.c raw_tp_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) · line 1188

Quoted 10 lines
kernel/trace/bpf_trace.c +10 -0
+	int ret;
+
+	if (IS_ERR(regs))
+		return PTR_ERR(regs);
+
+	perf_fetch_caller_regs(regs);
+	ret = ____bpf_perf_event_output(regs, map, flags, data, size);
+	put_bpf_raw_tp_regs();
+	return ret;
+}

I'm not sure why copy paste bpf_perf_event_output_raw_tp() into new function.

kernel/trace/bpf_trace.c tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) · line 1225

Quoted 9 lines
kernel/trace/bpf_trace.c +4 -0
@@ -1181,6 +1218,10 @@ tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_skb_output_proto;
 #endif
 	default:
+		if (prog->expected_attach_type == BPF_TRACE_FENTRY ||
+		    prog->expected_attach_type == BPF_TRACE_FEXIT)
+			return kfunc_prog_func_proto(func_id, prog);
+
 		return raw_tp_prog_func_proto(func_id, prog);

Are you saying bpf_perf_event_output_raw_tp() for some reason didn't work for fentry/fexit?
But above is exact copy-paste and it somehow worked?

Quoted 1 line

Ditto for patches 3,4.

ugh right.. did not realize that after switching to the rawtp regs nest arrays it's identical and we don't need that

jirka

author reviewer
#11 lore raw

PATCH v3 3/6 bpf: Add bpf_get_stackid_kfunc

In reply to Jiri Olsa

Adding support to use bpf_get_stackid_kfunc in
BPF_TRACE_FENTRY/BPF_TRACE_FEXIT programs.

---
 kernel/trace/bpf_trace.c | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)
kernel/trace/bpf_trace.c +27 -0
index 6a18e2ae6e30..18d6e96751c4 100644
@@ -1198,12 +1198,39 @@ static const struct bpf_func_proto bpf_perf_event_output_proto_kfunc = {
 	.arg5_type	= ARG_CONST_SIZE_OR_ZERO,
 };
 
+BPF_CALL_3(bpf_get_stackid_kfunc, void*, args,
+	   struct bpf_map *, map, u64, flags)
+{
+	struct pt_regs *regs = get_bpf_raw_tp_regs();
+	int ret;
+
+	if (IS_ERR(regs))
+		return PTR_ERR(regs);
+
+	perf_fetch_caller_regs(regs);
+	ret = bpf_get_stackid((unsigned long) regs, (unsigned long) map,
+			      flags, 0, 0);
+	put_bpf_raw_tp_regs();
+	return ret;
+}
+
+static const struct bpf_func_proto bpf_get_stackid_proto_kfunc = {
+	.func		= bpf_get_stackid_kfunc,
+	.gpl_only	= true,
+	.ret_type	= RET_INTEGER,
+	.arg1_type	= ARG_PTR_TO_CTX,
+	.arg2_type	= ARG_CONST_MAP_PTR,
+	.arg3_type	= ARG_ANYTHING,
+};
+
 static const struct bpf_func_proto *
 kfunc_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 {
 	switch (func_id) {
 	case BPF_FUNC_perf_event_output:
 		return &bpf_perf_event_output_proto_kfunc;
+	case BPF_FUNC_get_stackid:
+		return &bpf_get_stackid_proto_kfunc;
 	default:
 		return tracing_func_proto(func_id, prog);
 	}
Signature · 1 line
2.24.1
author reviewer
#12 lore raw

PATCH v3 4/6 bpf: Add bpf_get_stack_kfunc

In reply to Jiri Olsa

Adding support to use bpf_get_stack_kfunc in
BPF_TRACE_FENTRY/BPF_TRACE_FEXIT programs.

---
 kernel/trace/bpf_trace.c | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)
kernel/trace/bpf_trace.c +28 -0
index 18d6e96751c4..5c8edede3ac4 100644
@@ -1223,6 +1223,32 @@ static const struct bpf_func_proto bpf_get_stackid_proto_kfunc = {
 	.arg3_type	= ARG_ANYTHING,
 };
 
+BPF_CALL_4(bpf_get_stack_kfunc, void*, args,
+	   void *, buf, u32, size, u64, flags)
+{
+	struct pt_regs *regs = get_bpf_raw_tp_regs();
+	int ret;
+
+	if (IS_ERR(regs))
+		return PTR_ERR(regs);
+
+	perf_fetch_caller_regs(regs);
+	ret = bpf_get_stack((unsigned long) regs, (unsigned long) buf,
+			    (unsigned long) size, flags, 0);
+	put_bpf_raw_tp_regs();
+	return ret;
+}
+
+static const struct bpf_func_proto bpf_get_stack_proto_kfunc = {
+	.func		= bpf_get_stack_raw_tp,
+	.gpl_only	= true,
+	.ret_type	= RET_INTEGER,
+	.arg1_type	= ARG_PTR_TO_CTX,
+	.arg2_type	= ARG_PTR_TO_MEM,
+	.arg3_type	= ARG_CONST_SIZE_OR_ZERO,
+	.arg4_type	= ARG_ANYTHING,
+};
+
 static const struct bpf_func_proto *
 kfunc_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 {
@@ -1231,6 +1257,8 @@ kfunc_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_perf_event_output_proto_kfunc;
 	case BPF_FUNC_get_stackid:
 		return &bpf_get_stackid_proto_kfunc;
+	case BPF_FUNC_get_stack:
+		return &bpf_get_stack_proto_kfunc;
 	default:
 		return tracing_func_proto(func_id, prog);
 	}
Signature · 1 line
2.24.1
author reviewer
#13 lore raw

PATCH v3 5/6 bpf: Allow to resolve bpf trampoline and dispatcher in unwind

In reply to Jiri Olsa

When unwinding the stack we need to identify each address to successfully continue. Adding latch tree to keep trampolines for quick lookup during the unwind.

The patch uses first 48 bytes for latch tree node, leaving 4048 bytes from the rest of the page for trampoline or dispatcher generated code.

It's still enough not to affect trampoline and dispatcher progs maximum counts.

---
 include/linux/bpf.h     | 12 +++++-
 kernel/bpf/dispatcher.c |  4 +-
 kernel/bpf/trampoline.c | 82 +++++++++++++++++++++++++++++++++++++----
 kernel/extable.c        |  7 +++-
 4 files changed, 93 insertions(+), 12 deletions(-)
include/linux/bpf.h +11 -1
index 8e3b8f4ad183..46b60ed4387b 100644
@@ -519,7 +519,6 @@ struct bpf_trampoline *bpf_trampoline_lookup(u64 key);
 int bpf_trampoline_link_prog(struct bpf_prog *prog);
 int bpf_trampoline_unlink_prog(struct bpf_prog *prog);
 void bpf_trampoline_put(struct bpf_trampoline *tr);
-void *bpf_jit_alloc_exec_page(void);
 #define BPF_DISPATCHER_INIT(name) {			\
 	.mutex = __MUTEX_INITIALIZER(name.mutex),	\
 	.func = &name##func,				\
@@ -551,6 +550,13 @@ void *bpf_jit_alloc_exec_page(void);
 #define BPF_DISPATCHER_PTR(name) (&name)
 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
 				struct bpf_prog *to);
+struct bpf_image {
+	struct latch_tree_node tnode;
+	unsigned char data[];
+};
+#define BPF_IMAGE_SIZE (PAGE_SIZE - sizeof(struct bpf_image))
+bool is_bpf_image_address(unsigned long address);
+void *bpf_image_alloc(void);
 #else
 static inline struct bpf_trampoline *bpf_trampoline_lookup(u64 key)
 {
@@ -572,6 +578,10 @@ static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
 					      struct bpf_prog *from,
 					      struct bpf_prog *to) {}
+static inline bool is_bpf_image_address(unsigned long address)
+{
+	return false;
+}
 #endif
 
 struct bpf_func_info_aux {
kernel/bpf/dispatcher.c +2 -2
index 204ee61a3904..b3e5b214fed8 100644
@@ -113,7 +113,7 @@ static void bpf_dispatcher_update(struct bpf_dispatcher *d, int prev_num_progs)
 		noff = 0;
 	} else {
 		old = d->image + d->image_off;
-		noff = d->image_off ^ (PAGE_SIZE / 2);
+		noff = d->image_off ^ (BPF_IMAGE_SIZE / 2);
 	}
 
 	new = d->num_progs ? d->image + noff : NULL;
@@ -140,7 +140,7 @@ void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
 
 	mutex_lock(&d->mutex);
 	if (!d->image) {
-		d->image = bpf_jit_alloc_exec_page();
+		d->image = bpf_image_alloc();
 		if (!d->image)
 			goto out;
 	}
kernel/bpf/trampoline.c +75 -7

129 lines not on this page · raw message

kernel/extable.c +5 -2
index f6920a11e28a..a0024f27d3a1 100644
@@ -131,8 +131,9 @@ int kernel_text_address(unsigned long addr)
 	 * triggers a stack trace, or a WARN() that happens during
 	 * coming back from idle, or cpu on or offlining.
 	 *
-	 * is_module_text_address() as well as the kprobe slots
-	 * and is_bpf_text_address() require RCU to be watching.
+	 * is_module_text_address() as well as the kprobe slots,
+	 * is_bpf_text_address() and is_bpf_image_address require
+	 * RCU to be watching.
 	 */
 	no_rcu = !rcu_is_watching();
 
@@ -148,6 +149,8 @@ int kernel_text_address(unsigned long addr)
 		goto out;
 	if (is_bpf_text_address(addr))
 		goto out;
+	if (is_bpf_image_address(addr))
+		goto out;
 	ret = 0;
 out:
 	if (no_rcu)
Signature · 1 line
2.24.1
author reviewer
#14 lore raw

PATCH v3 6/6 selftest/bpf: Add test for allowed trampolines count

In reply to Jiri Olsa

There's limit of 40 programs tht can be attached to trampoline for one function. Adding test that tries to attach that many plus one extra that needs to fail.

---
 .../bpf/prog_tests/trampoline_count.c         | 112 ++++++++++++++++++
 .../bpf/progs/test_trampoline_count.c         |  21 ++++
 2 files changed, 133 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/trampoline_count.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_trampoline_count.c
tools/testing/selftests/bpf/prog_tests/trampoline_count.c new file +112 -0
new file mode 100644
index 000000000000..1235f3d1cc50
@@ -0,0 +1,112 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#define _GNU_SOURCE
+#include <sched.h>
+#include <sys/prctl.h>
+#include <test_progs.h>
+
+#define MAX_TRAMP_PROGS 40
+
+struct inst {
+	struct bpf_object *obj;
+	struct bpf_link   *link_fentry;
+	struct bpf_link   *link_fexit;
+};
+
+static int test_task_rename(void)
+{
+	int fd, duration = 0, err;
+	char buf[] = "test_overhead";
+
+	fd = open("/proc/self/comm", O_WRONLY|O_TRUNC);
+	if (CHECK(fd < 0, "open /proc", "err %d", errno))
+		return -1;
+	err = write(fd, buf, sizeof(buf));
+	if (err < 0) {
+		CHECK(err < 0, "task rename", "err %d", errno);
+		close(fd);
+		return -1;
+	}
+	close(fd);
+	return 0;
+}
+
+static struct bpf_link *load(struct bpf_object *obj, const char *name)
+{
+	struct bpf_program *prog;
+	int duration = 0;
+
+	prog = bpf_object__find_program_by_title(obj, name);
+	if (CHECK(!prog, "find_probe", "prog '%s' not found\n", name))
+		return ERR_PTR(-EINVAL);
+	return bpf_program__attach_trace(prog);
+}
+
+void test_trampoline_count(void)
+{
+	const char *fentry_name = "fentry/__set_task_comm";
+	const char *fexit_name = "fexit/__set_task_comm";
+	const char *object = "test_trampoline_count.o";
+	struct inst inst[MAX_TRAMP_PROGS] = { 0 };
+	int err, i = 0, duration = 0;
+	struct bpf_object *obj;
+	struct bpf_link *link;
+	char comm[16] = {};
+
+	/* attach 'allowed' 40 trampoline programs */
+	for (i = 0; i < MAX_TRAMP_PROGS; i++) {
+		obj = bpf_object__open_file(object, NULL);
+		if (CHECK(IS_ERR(obj), "obj_open_file", "err %ld\n", PTR_ERR(obj)))
+			goto cleanup;
+
+		err = bpf_object__load(obj);
+		if (CHECK(err, "obj_load", "err %d\n", err))
+			goto cleanup;
+		inst[i].obj = obj;
+
+		if (rand() % 2) {
+			link = load(obj, fentry_name);
+			if (CHECK(IS_ERR(link), "attach prog", "err %ld\n", PTR_ERR(link)))
+				goto cleanup;
+			inst[i].link_fentry = link;
+		} else {
+			link = load(obj, fexit_name);
+			if (CHECK(IS_ERR(link), "attach prog", "err %ld\n", PTR_ERR(link)))
+				goto cleanup;
+			inst[i].link_fexit = link;
+		}
+	}
+
+	/* and try 1 extra.. */
+	obj = bpf_object__open_file(object, NULL);
+	if (CHECK(IS_ERR(obj), "obj_open_file", "err %ld\n", PTR_ERR(obj)))
+		goto cleanup;
+
+	err = bpf_object__load(obj);
+	if (CHECK(err, "obj_load", "err %d\n", err))
+		goto cleanup_extra;
+
+	/* ..that needs to fail */
+	link = load(obj, fentry_name);
+	if (CHECK(!IS_ERR(link), "cannot attach over the limit", "err %ld\n", PTR_ERR(link))) {
+		bpf_link__destroy(link);
+		goto cleanup_extra;
+	}
+
+	/* with E2BIG error */
+	CHECK(PTR_ERR(link) != -E2BIG, "proper error check", "err %ld\n", PTR_ERR(link));
+
+	/* and finaly execute the probe */
+	if (CHECK_FAIL(prctl(PR_GET_NAME, comm, 0L, 0L, 0L)))
+		goto cleanup_extra;
+	CHECK_FAIL(test_task_rename());
+	CHECK_FAIL(prctl(PR_SET_NAME, comm, 0L, 0L, 0L));
+
+cleanup_extra:
+	bpf_object__close(obj);
+cleanup:
+	while (--i) {
+		bpf_link__destroy(inst[i].link_fentry);
+		bpf_link__destroy(inst[i].link_fexit);
+		bpf_object__close(inst[i].obj);
+	}
+}
tools/testing/selftests/bpf/progs/test_trampoline_count.c new file +21 -0
new file mode 100644
index 000000000000..e51e6e3a81c2
@@ -0,0 +1,21 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <stdbool.h>
+#include <stddef.h>
+#include <linux/bpf.h>
+#include "bpf_trace_helpers.h"
+
+struct task_struct;
+
+SEC("fentry/__set_task_comm")
+int BPF_PROG(prog1, struct task_struct *tsk, const char *buf, bool exec)
+{
+	return 0;
+}
+
+SEC("fexit/__set_task_comm")
+int BPF_PROG(prog2, struct task_struct *tsk, const char *buf, bool exec)
+{
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";
Signature · 1 line
2.24.1
#15 lore raw

Re: [PATCH 6/6] selftest/bpf: Add test for allowed trampolines count

Quoted 7 lines · Jiri Olsa

There's limit of 40 programs tht can be attached to trampoline for one function. Adding test that tries to attach that many plus one extra that needs to fail.

I don't mind another test. Just pointing out that there is one for this purpose already :) prog_tests/fexit_stress.c
Yours is better. Mine wasn't that sophisticated. :)

#16 lore raw

Re: [PATCH 6/6] selftest/bpf: Add test for allowed trampolines count

Quoted 7 lines · Alexei Starovoitov
Quoted 6 lines · Jiri Olsa

There's limit of 40 programs tht can be attached to trampoline for one function. Adding test that tries to attach that many plus one extra that needs to fail.

Quoted 6 lines
Quoted 1 line

I don't mind another test. Just pointing out that there is one for this purpose already :) prog_tests/fexit_stress.c
Yours is better. Mine wasn't that sophisticated. :)

ok ;-) did not notice that one.. just wanted to be sure the unwind change won't screw that

jirka