[PATCH dwarves 0/2] btf: support typedef DW_TAG_LLVM_annotation

[PATCH dwarves 0/2] btf: support typedef DW_TAG_LLVM_annotation

[Yonghong Song]

Latest llvm is able to generate DW_TAG_LLVM_annotation for typedef
declarations. Latest bpf-next supports BTF_KIND_DECL_TAG for
typedef declarations. This patch implemented dwarf DW_TAG_LLVM_annotation
to btf BTF_KIND_DECL_TAG conversion. Patch 1 is for dwarf_loader
to process DW_TAG_LLVM_annotation tags. Patch 2 is for the
dwarf->btf conversion.

Yonghong Song (2):
  dwarf_loader: support typedef DW_TAG_LLVM_annotation
  btf_encoder: generate BTF_KIND_DECL_TAGs for typedef btf_decl_tag
    attributes

 btf_encoder.c  | 12 +++++++++---
 dwarf_loader.c |  7 ++-----
 2 files changed, 11 insertions(+), 8 deletions(-)

-- 
2.30.2

[PATCH dwarves 1/2] dwarf_loader: support typedef DW_TAG_LLVM_annotation

[Yonghong Song]

llvm commit ([1]) added support for btf_decl_tag attribute
with typedef declaration. Eventually, DW_TAG_LLVM_annotation
tag may appear inside dwarf typedef declaration tag.

kernel support for typedef BTF_KIND_DECL_TAG support
is introduced in [2]. There is no additional libbpf
change needed as the previous libbpf BTF_KIND_DECL_TAG
support is generic enough to cover new typedef use
cases.

This patch added parsing of DW_TAG_LLVM_annotation
for dwarf typedef decl.

  $ cat t.c
  $ clang -O2 -g -c t.c
  $ llvm-dwarfdump --debug-info t.o
    ......
    0x00000033:   DW_TAG_typedef
                    DW_AT_type      (0x00000051 "structure ")
                    DW_AT_name      ("__t")
                    DW_AT_decl_file ("/home/yhs/t.c")
                    DW_AT_decl_line (3)

    0x0000003e:     DW_TAG_LLVM_annotation
                      DW_AT_name    ("btf_decl_tag")
                      DW_AT_const_value     ("tag1")

    0x00000047:     DW_TAG_LLVM_annotation
                      DW_AT_name    ("btf_decl_tag")
                      DW_AT_const_value     ("tag2")

    0x00000050:     NULL

Previously, pahole will issue a warning if typedef tag
contains any child tag. I removed this warning since
it is not true any more. Note that dwarf standard doesn't
prevent typedef decl tag from having nested tags.
In the future if we need to process any tag inside
typedef tag, we can just add code to process it.

  [1] https://reviews.llvm.org/D110127
  [2] https://lore.kernel.org/bpf/20211021195628.4018847-1-yhs@fb.com

Signed-off-by: Yonghong Song <yhs@fb.com>
---
 dwarf_loader.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/dwarf_loader.c b/dwarf_loader.c
index c5bda81..f748bd7 100644
--- a/dwarf_loader.c
+++ b/dwarf_loader.c
@@ -1296,11 +1296,8 @@ static struct tag *die__create_new_typedef(Dwarf_Die *die, struct cu *cu, struct
 	if (tdef == NULL)
 		return NULL;
 
-	if (dwarf_haschildren(die)) {
-		struct dwarf_tag *dtag = tdef->namespace.tag.priv;
-		fprintf(stderr, "%s: DW_TAG_typedef %llx WITH children!\n",
-			__func__, (unsigned long long)dtag->id);
-	}
+	if (add_child_llvm_annotations(die, -1, conf, &tdef->namespace.annots))
+		return NULL;
 
 	return &tdef->namespace.tag;
 }
-- 
2.30.2

[PATCH dwarves 2/2] btf_encoder: generate BTF_KIND_DECL_TAGs for typedef btf_decl_tag attributes

[Yonghong Song]

Emit BTF BTF_KIND_DECL_TAGs for btf_decl_tag attributes attached to
typedef declarations. The following is a simple example:
  $ cat t.c
    #define __tag1 __attribute__((btf_decl_tag("tag1")))
    #define __tag2 __attribute__((btf_decl_tag("tag2")))
    typedef struct { int a; int b; } __t __tag1 __tag2;
    __t g;
  $ clang -O2 -g -c t.c
  $ pahole -JV t.o
    btf_encoder__new: 't.o' doesn't have '.data..percpu' section
    Found 0 per-CPU variables!
    File t.o:
    [1] TYPEDEF __t type_id=2
    [2] STRUCT (anon) size=8
            a type_id=3 bits_offset=0
            b type_id=3 bits_offset=32
    [3] INT int size=4 nr_bits=32 encoding=SIGNED
    [4] DECL_TAG tag1 type_id=1 component_idx=-1
    [5] DECL_TAG tag2 type_id=1 component_idx=-1

Signed-off-by: Yonghong Song <yhs@fb.com>
---
 btf_encoder.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/btf_encoder.c b/btf_encoder.c
index 40f6aa3..2f1f4ae 100644
--- a/btf_encoder.c
+++ b/btf_encoder.c
@@ -1437,19 +1437,25 @@ int btf_encoder__encode_cu(struct btf_encoder *encoder, struct cu *cu)
 	}
 
 	cu__for_each_type(cu, core_id, pos) {
+		const char *tag_name = "typedef";
 		struct namespace *ns;
 
-		if (pos->tag != DW_TAG_structure_type && pos->tag != DW_TAG_union_type)
+		if (pos->tag != DW_TAG_structure_type && pos->tag != DW_TAG_union_type &&
+		    pos->tag != DW_TAG_typedef)
 			continue;
 
+		if (pos->tag == DW_TAG_structure_type)
+			tag_name = "struct";
+		else if (pos->tag == DW_TAG_union_type)
+			tag_name = "union";
+
 		btf_type_id = type_id_off + core_id;
 		ns = tag__namespace(pos);
 		list_for_each_entry(annot, &ns->annots, node) {
 			tag_type_id = btf_encoder__add_decl_tag(encoder, annot->value, btf_type_id, annot->component_idx);
 			if (tag_type_id < 0) {
 				fprintf(stderr, "error: failed to encode tag '%s' to %s '%s' with component_idx %d\n",
-					annot->value, pos->tag == DW_TAG_structure_type ? "struct" : "union",
-					namespace__name(ns), annot->component_idx);
+					annot->value, tag_name, namespace__name(ns), annot->component_idx);
 				goto out;
 			}
 		}
-- 
2.30.2

Re: [PATCH dwarves 2/2] btf_encoder: generate BTF_KIND_DECL_TAGs for typedef btf_decl_tag attributes

[Andrii Nakryiko]

On Wed, Oct 27, 2021 at 4:08 PM Yonghong Song <yhs@fb.com> wrote:
>
> Emit BTF BTF_KIND_DECL_TAGs for btf_decl_tag attributes attached to
> typedef declarations. The following is a simple example:
>   $ cat t.c
>     #define __tag1 __attribute__((btf_decl_tag("tag1")))
>     #define __tag2 __attribute__((btf_decl_tag("tag2")))
>     typedef struct { int a; int b; } __t __tag1 __tag2;
>     __t g;
>   $ clang -O2 -g -c t.c
>   $ pahole -JV t.o
>     btf_encoder__new: 't.o' doesn't have '.data..percpu' section
>     Found 0 per-CPU variables!
>     File t.o:
>     [1] TYPEDEF __t type_id=2
>     [2] STRUCT (anon) size=8
>             a type_id=3 bits_offset=0
>             b type_id=3 bits_offset=32
>     [3] INT int size=4 nr_bits=32 encoding=SIGNED
>     [4] DECL_TAG tag1 type_id=1 component_idx=-1
>     [5] DECL_TAG tag2 type_id=1 component_idx=-1
>
> Signed-off-by: Yonghong Song <yhs@fb.com>
> ---
>  btf_encoder.c | 12 +++++++++---
>  1 file changed, 9 insertions(+), 3 deletions(-)
>
> diff --git a/btf_encoder.c b/btf_encoder.c
> index 40f6aa3..2f1f4ae 100644
> --- a/btf_encoder.c
> +++ b/btf_encoder.c
> @@ -1437,19 +1437,25 @@ int btf_encoder__encode_cu(struct btf_encoder *encoder, struct cu *cu)
>         }
>
>         cu__for_each_type(cu, core_id, pos) {
> +               const char *tag_name = "typedef";
>                 struct namespace *ns;
>
> -               if (pos->tag != DW_TAG_structure_type && pos->tag != DW_TAG_union_type)
> +               if (pos->tag != DW_TAG_structure_type && pos->tag != DW_TAG_union_type &&
> +                   pos->tag != DW_TAG_typedef)
>                         continue;
>
> +               if (pos->tag == DW_TAG_structure_type)
> +                       tag_name = "struct";
> +               else if (pos->tag == DW_TAG_union_type)
> +                       tag_name = "union";

nit: switch instead of these two related sets of if/else blocks would be cleaner


> +
>                 btf_type_id = type_id_off + core_id;
>                 ns = tag__namespace(pos);
>                 list_for_each_entry(annot, &ns->annots, node) {
>                         tag_type_id = btf_encoder__add_decl_tag(encoder, annot->value, btf_type_id, annot->component_idx);
>                         if (tag_type_id < 0) {
>                                 fprintf(stderr, "error: failed to encode tag '%s' to %s '%s' with component_idx %d\n",
> -                                       annot->value, pos->tag == DW_TAG_structure_type ? "struct" : "union",
> -                                       namespace__name(ns), annot->component_idx);
> +                                       annot->value, tag_name, namespace__name(ns), annot->component_idx);
>                                 goto out;
>                         }
>                 }
> --
> 2.30.2
>

Re: [PATCH dwarves 2/2] btf_encoder: generate BTF_KIND_DECL_TAGs for typedef btf_decl_tag attributes

[Yonghong Song]

On 11/1/21 7:58 PM, Andrii Nakryiko wrote:
> On Wed, Oct 27, 2021 at 4:08 PM Yonghong Song <yhs@fb.com> wrote:
>>
>> Emit BTF BTF_KIND_DECL_TAGs for btf_decl_tag attributes attached to
>> typedef declarations. The following is a simple example:
>>    $ cat t.c
>>      #define __tag1 __attribute__((btf_decl_tag("tag1")))
>>      #define __tag2 __attribute__((btf_decl_tag("tag2")))
>>      typedef struct { int a; int b; } __t __tag1 __tag2;
>>      __t g;
>>    $ clang -O2 -g -c t.c
>>    $ pahole -JV t.o
>>      btf_encoder__new: 't.o' doesn't have '.data..percpu' section
>>      Found 0 per-CPU variables!
>>      File t.o:
>>      [1] TYPEDEF __t type_id=2
>>      [2] STRUCT (anon) size=8
>>              a type_id=3 bits_offset=0
>>              b type_id=3 bits_offset=32
>>      [3] INT int size=4 nr_bits=32 encoding=SIGNED
>>      [4] DECL_TAG tag1 type_id=1 component_idx=-1
>>      [5] DECL_TAG tag2 type_id=1 component_idx=-1
>>
>> Signed-off-by: Yonghong Song <yhs@fb.com>
>> ---
>>   btf_encoder.c | 12 +++++++++---
>>   1 file changed, 9 insertions(+), 3 deletions(-)
>>
>> diff --git a/btf_encoder.c b/btf_encoder.c
>> index 40f6aa3..2f1f4ae 100644
>> --- a/btf_encoder.c
>> +++ b/btf_encoder.c
>> @@ -1437,19 +1437,25 @@ int btf_encoder__encode_cu(struct btf_encoder *encoder, struct cu *cu)
>>          }
>>
>>          cu__for_each_type(cu, core_id, pos) {
>> +               const char *tag_name = "typedef";
>>                  struct namespace *ns;
>>
>> -               if (pos->tag != DW_TAG_structure_type && pos->tag != DW_TAG_union_type)
>> +               if (pos->tag != DW_TAG_structure_type && pos->tag != DW_TAG_union_type &&
>> +                   pos->tag != DW_TAG_typedef)
>>                          continue;
>>
>> +               if (pos->tag == DW_TAG_structure_type)
>> +                       tag_name = "struct";
>> +               else if (pos->tag == DW_TAG_union_type)
>> +                       tag_name = "union";
> 
> nit: switch instead of these two related sets of if/else blocks would be cleaner

Sure. Will make the change in v2.

> 
> 
>> +
>>                  btf_type_id = type_id_off + core_id;
>>                  ns = tag__namespace(pos);
>>                  list_for_each_entry(annot, &ns->annots, node) {
>>                          tag_type_id = btf_encoder__add_decl_tag(encoder, annot->value, btf_type_id, annot->component_idx);
>>                          if (tag_type_id < 0) {
>>                                  fprintf(stderr, "error: failed to encode tag '%s' to %s '%s' with component_idx %d\n",
>> -                                       annot->value, pos->tag == DW_TAG_structure_type ? "struct" : "union",
>> -                                       namespace__name(ns), annot->component_idx);
>> +                                       annot->value, tag_name, namespace__name(ns), annot->component_idx);
>>                                  goto out;
>>                          }
>>                  }
>> --
>> 2.30.2
>>