* [PATCH dwarves 1/2] dwarf_loader: parse dwarf tag DW_TAG_LLVM_annotation
2021-09-20 0:35 [PATCH dwarves 0/2] generate BTF_KIND_TAG types from DW_TAG_LLVM_annotation dwarf tags Yonghong Song
@ 2021-09-20 0:35 ` Yonghong Song
2021-09-21 21:57 ` Andrii Nakryiko
2021-09-20 0:35 ` [PATCH dwarves 2/2] btf_encoder: generate BTF_KIND_TAG from llvm annotations Yonghong Song
1 sibling, 1 reply; 6+ messages in thread
From: Yonghong Song @ 2021-09-20 0:35 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo, dwarves
Cc: Alexei Starovoitov, Andrii Nakryiko, bpf, Daniel Borkmann,
kernel-team, Martin KaFai Lau
Parse dwarf tag DW_TAG_LLVM_annotation. Only record
annotations with btf_tag name which corresponds to
btf_tag attributes in C code. Such information will
be used later by btf_encoder for BTF conversion.
LLVM implementation only supports btf_tag annotations
on struct/union, func, func parameter and variable ([1]).
So we only check existence of corresponding DW tags
in these places. A flag "--skip_encoding_btf_tag"
is introduced if for whatever reason this feature
needs to be disabled.
[1] https://reviews.llvm.org/D106614
Signed-off-by: Yonghong Song <yhs@fb.com>
---
dwarf_loader.c | 85 ++++++++++++++++++++++++++++++++++++++++++++++----
dwarves.h | 10 ++++++
pahole.c | 8 +++++
3 files changed, 97 insertions(+), 6 deletions(-)
diff --git a/dwarf_loader.c b/dwarf_loader.c
index 0213e42..48e1bf0 100644
--- a/dwarf_loader.c
+++ b/dwarf_loader.c
@@ -52,6 +52,10 @@
#define DW_OP_addrx 0xa1
#endif
+#ifndef DW_TAG_LLVM_annotation
+#define DW_TAG_LLVM_annotation 0x6000
+#endif
+
static pthread_mutex_t libdw__lock = PTHREAD_MUTEX_INITIALIZER;
static uint32_t hashtags__bits = 12;
@@ -602,6 +606,7 @@ static void namespace__init(struct namespace *namespace, Dwarf_Die *die,
{
tag__init(&namespace->tag, cu, die);
INIT_LIST_HEAD(&namespace->tags);
+ INIT_LIST_HEAD(&namespace->annots);
namespace->name = attr_string(die, DW_AT_name, conf);
namespace->nr_tags = 0;
namespace->shared_tags = 0;
@@ -719,6 +724,7 @@ static struct variable *variable__new(Dwarf_Die *die, struct cu *cu, struct conf
/* non-defining declaration of an object */
var->declaration = dwarf_hasattr(die, DW_AT_declaration);
var->scope = VSCOPE_UNKNOWN;
+ INIT_LIST_HEAD(&var->annots);
var->ip.addr = 0;
if (!var->declaration && cu->has_addr_info)
var->scope = dwarf__location(die, &var->ip.addr, &var->location);
@@ -854,6 +860,51 @@ static int tag__recode_dwarf_bitfield(struct tag *tag, struct cu *cu, uint16_t b
return -ENOMEM;
}
+static int add_llvm_annotation(Dwarf_Die *die, int component_idx, struct conf_load *conf,
+ struct list_head *head)
+{
+ struct llvm_annotation *annot;
+ const char *name;
+
+ if (conf->skip_encoding_btf_tag)
+ return 0;
+
+ /* Only handle btf_tag annotation for now. */
+ name = attr_string(die, DW_AT_name, conf);
+ if (strcmp(name, "btf_tag") != 0)
+ return 0;
+
+ annot = zalloc(sizeof(*annot));
+ if (!annot)
+ return -ENOMEM;
+
+ annot->value = attr_string(die, DW_AT_const_value, conf);
+ annot->component_idx = component_idx;
+ list_add_tail(&annot->node, head);
+ return 0;
+}
+
+static int add_child_llvm_annotations(Dwarf_Die *die, int component_idx,
+ struct conf_load *conf, struct list_head *head)
+{
+ Dwarf_Die child;
+ int ret;
+
+ if (!dwarf_haschildren(die) || dwarf_child(die, &child) != 0)
+ return 0;
+
+ die = &child;
+ do {
+ if (dwarf_tag(die) == DW_TAG_LLVM_annotation) {
+ ret = add_llvm_annotation(die, component_idx, conf, head);
+ if (ret)
+ return ret;
+ }
+ } while (dwarf_siblingof(die, die) == 0);
+
+ return 0;
+}
+
int class_member__dwarf_recode_bitfield(struct class_member *member,
struct cu *cu)
{
@@ -1092,6 +1143,7 @@ static struct function *function__new(Dwarf_Die *die, struct cu *cu, struct conf
func->accessibility = attr_numeric(die, DW_AT_accessibility);
func->virtuality = attr_numeric(die, DW_AT_virtuality);
INIT_LIST_HEAD(&func->vtable_node);
+ INIT_LIST_HEAD(&func->annots);
INIT_LIST_HEAD(&func->tool_node);
func->vtable_entry = -1;
if (dwarf_hasattr(die, DW_AT_vtable_elem_location))
@@ -1304,16 +1356,21 @@ static struct tag *die__create_new_string_type(Dwarf_Die *die, struct cu *cu)
static struct tag *die__create_new_parameter(Dwarf_Die *die,
struct ftype *ftype,
struct lexblock *lexblock,
- struct cu *cu, struct conf_load *conf)
+ struct cu *cu, struct conf_load *conf,
+ int param_idx)
{
struct parameter *parm = parameter__new(die, cu, conf);
if (parm == NULL)
return NULL;
- if (ftype != NULL)
+ if (ftype != NULL) {
ftype__add_parameter(ftype, parm);
- else {
+ if (param_idx >= 0) {
+ if (add_child_llvm_annotations(die, param_idx, conf, &(tag__function(&ftype->tag)->annots)))
+ return NULL;
+ }
+ } else {
/*
* DW_TAG_formal_parameters on a non DW_TAG_subprogram nor
* DW_TAG_subroutine_type tag happens sometimes, likely due to
@@ -1346,7 +1403,10 @@ static struct tag *die__create_new_variable(Dwarf_Die *die, struct cu *cu, struc
{
struct variable *var = variable__new(die, cu, conf);
- return var ? &var->ip.tag : NULL;
+ if (var == NULL || add_child_llvm_annotations(die, -1, conf, &var->annots))
+ return NULL;
+
+ return &var->ip.tag;
}
static struct tag *die__create_new_subroutine_type(Dwarf_Die *die,
@@ -1371,7 +1431,7 @@ static struct tag *die__create_new_subroutine_type(Dwarf_Die *die,
tag__print_not_supported(dwarf_tag(die));
continue;
case DW_TAG_formal_parameter:
- tag = die__create_new_parameter(die, ftype, NULL, cu, conf);
+ tag = die__create_new_parameter(die, ftype, NULL, cu, conf, -1);
break;
case DW_TAG_unspecified_parameters:
ftype->unspec_parms = 1;
@@ -1455,6 +1515,7 @@ static int die__process_class(Dwarf_Die *die, struct type *class,
struct cu *cu, struct conf_load *conf)
{
const bool is_union = tag__is_union(&class->namespace.tag);
+ int member_idx = 0;
do {
switch (dwarf_tag(die)) {
@@ -1497,8 +1558,15 @@ static int die__process_class(Dwarf_Die *die, struct type *class,
type__add_member(class, member);
cu__hash(cu, &member->tag);
+ if (add_child_llvm_annotations(die, member_idx, conf, &class->namespace.annots))
+ return -ENOMEM;
+ member_idx++;
}
continue;
+ case DW_TAG_LLVM_annotation:
+ if (add_llvm_annotation(die, -1, conf, &class->namespace.annots))
+ return -ENOMEM;
+ continue;
default: {
struct tag *tag = die__process_tag(die, cu, 0, conf);
@@ -1699,6 +1767,7 @@ static struct tag *die__create_new_inline_expansion(Dwarf_Die *die,
static int die__process_function(Dwarf_Die *die, struct ftype *ftype,
struct lexblock *lexblock, struct cu *cu, struct conf_load *conf)
{
+ int param_idx = 0;
Dwarf_Die child;
struct tag *tag;
@@ -1742,7 +1811,7 @@ static int die__process_function(Dwarf_Die *die, struct ftype *ftype,
tag__print_not_supported(dwarf_tag(die));
continue;
case DW_TAG_formal_parameter:
- tag = die__create_new_parameter(die, ftype, lexblock, cu, conf);
+ tag = die__create_new_parameter(die, ftype, lexblock, cu, conf, param_idx++);
break;
case DW_TAG_variable:
tag = die__create_new_variable(die, cu, conf);
@@ -1771,6 +1840,10 @@ static int die__process_function(Dwarf_Die *die, struct ftype *ftype,
if (die__create_new_lexblock(die, cu, lexblock, conf) != 0)
goto out_enomem;
continue;
+ case DW_TAG_LLVM_annotation:
+ if (add_llvm_annotation(die, -1, conf, &(tag__function(&ftype->tag)->annots)))
+ goto out_enomem;
+ continue;
default:
tag = die__process_tag(die, cu, 0, conf);
diff --git a/dwarves.h b/dwarves.h
index 0b69312..30d33fa 100644
--- a/dwarves.h
+++ b/dwarves.h
@@ -58,6 +58,7 @@ struct conf_load {
bool ignore_inline_expansions;
bool ignore_labels;
bool ptr_table_stats;
+ bool skip_encoding_btf_tag;
uint8_t hashtable_bits;
uint8_t max_hashtable_bits;
uint16_t kabi_prefix_len;
@@ -594,6 +595,12 @@ static inline struct ptr_to_member_type *
return (struct ptr_to_member_type *)tag;
}
+struct llvm_annotation {
+ const char *value;
+ int16_t component_idx;
+ struct list_head node;
+};
+
/** struct namespace - base class for enums, structs, unions, typedefs, etc
*
* @tags - class_member, enumerators, etc
@@ -605,6 +612,7 @@ struct namespace {
uint16_t nr_tags;
uint8_t shared_tags;
struct list_head tags;
+ struct list_head annots;
};
static inline struct namespace *tag__namespace(const struct tag *tag)
@@ -686,6 +694,7 @@ struct variable {
enum vscope scope;
struct location location;
struct hlist_node tool_hnode;
+ struct list_head annots;
struct variable *spec;
};
@@ -818,6 +827,7 @@ struct function {
uint8_t btf:1;
int32_t vtable_entry;
struct list_head vtable_node;
+ struct list_head annots;
/* fields used by tools */
union {
struct list_head tool_node;
diff --git a/pahole.c b/pahole.c
index 7f1771b..80271b5 100644
--- a/pahole.c
+++ b/pahole.c
@@ -1124,6 +1124,7 @@ ARGP_PROGRAM_VERSION_HOOK_DEF = dwarves_print_version;
#define ARGP_sort_output 328
#define ARGP_hashbits 329
#define ARGP_devel_stats 330
+#define ARGP_skip_encoding_btf_tag 331
static const struct argp_option pahole__options[] = {
{
@@ -1494,6 +1495,11 @@ static const struct argp_option pahole__options[] = {
.key = ARGP_devel_stats,
.doc = "Print internal data structures stats",
},
+ {
+ .name = "skip_encoding_btf_tag",
+ .key = ARGP_skip_encoding_btf_tag,
+ .doc = "Do not encode TAGs in BTF."
+ },
{
.name = NULL,
}
@@ -1642,6 +1648,8 @@ static error_t pahole__options_parser(int key, char *arg,
conf_load.hashtable_bits = atoi(arg); break;
case ARGP_devel_stats:
conf_load.ptr_table_stats = true; break;
+ case ARGP_skip_encoding_btf_tag:
+ conf_load.skip_encoding_btf_tag = true; break;
default:
return ARGP_ERR_UNKNOWN;
}
--
2.30.2
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH dwarves 2/2] btf_encoder: generate BTF_KIND_TAG from llvm annotations
2021-09-20 0:35 [PATCH dwarves 0/2] generate BTF_KIND_TAG types from DW_TAG_LLVM_annotation dwarf tags Yonghong Song
2021-09-20 0:35 ` [PATCH dwarves 1/2] dwarf_loader: parse dwarf tag DW_TAG_LLVM_annotation Yonghong Song
@ 2021-09-20 0:35 ` Yonghong Song
2021-09-21 22:02 ` Andrii Nakryiko
1 sibling, 1 reply; 6+ messages in thread
From: Yonghong Song @ 2021-09-20 0:35 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo, dwarves
Cc: Alexei Starovoitov, Andrii Nakryiko, bpf, Daniel Borkmann,
kernel-team, Martin KaFai Lau
The following is an example with latest upstream clang:
$ cat t.c
#define __tag1 __attribute__((btf_tag("tag1")))
#define __tag2 __attribute__((btf_tag("tag2")))
struct t {
int a:1 __tag1;
int b __tag2;
} __tag1 __tag2;
int g __tag1 __attribute__((section(".data..percpu")));
int __tag1 foo(struct t *a1, int a2 __tag2) {
return a1->b + a2 + g;
}
$ clang -O2 -g -c t.c
$ pahole -JV t.o
Found per-CPU symbol 'g' at address 0x0
Found 1 per-CPU variables!
Found 1 functions!
File t.o:
[1] INT int size=4 nr_bits=32 encoding=SIGNED
[2] PTR (anon) type_id=3
[3] STRUCT t size=8
a type_id=1 bitfield_size=1 bits_offset=0
b type_id=1 bitfield_size=0 bits_offset=32
[4] TAG tag1 type_id=3 component_idx=0
[5] TAG tag2 type_id=3 component_idx=1
[6] TAG tag1 type_id=3 component_idx=-1
[7] TAG tag2 type_id=3 component_idx=-1
[8] FUNC_PROTO (anon) return=1 args=(2 a1, 1 a2)
[9] FUNC foo type_id=8
[10] TAG tag2 type_id=9 component_idx=1
[11] TAG tag1 type_id=9 component_idx=-1
search cu 't.c' for percpu global variables.
Variable 'g' from CU 't.c' at address 0x0 encoded
[12] VAR g type=1 linkage=1
[13] TAG tag1 type_id=12 component_idx=-1
[14] DATASEC .data..percpu size=4 vlen=1
type=12 offset=0 size=4
$ ...
With additional option --skip_encoding_btf_tag, pahole doesn't
generate BTF_KIND_TAGs any more.
$ pahole -JV --skip_encoding_btf_tag t.o
Found per-CPU symbol 'g' at address 0x0
Found 1 per-CPU variables!
Found 1 functions!
File t.o:
[1] INT int size=4 nr_bits=32 encoding=SIGNED
[2] PTR (anon) type_id=3
[3] STRUCT t size=8
a type_id=1 bitfield_size=1 bits_offset=0
b type_id=1 bitfield_size=0 bits_offset=32
[4] FUNC_PROTO (anon) return=1 args=(2 a1, 1 a2)
[5] FUNC foo type_id=4
search cu 't.c' for percpu global variables.
Variable 'g' from CU 't.c' at address 0x0 encoded
[6] VAR g type=1 linkage=1
[7] DATASEC .data..percpu size=4 vlen=1
type=6 offset=0 size=4
$ ...
Signed-off-by: Yonghong Song <yhs@fb.com>
---
btf_encoder.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 45 insertions(+)
diff --git a/btf_encoder.c b/btf_encoder.c
index 1b4e83d..e983750 100644
--- a/btf_encoder.c
+++ b/btf_encoder.c
@@ -141,6 +141,7 @@ static const char * const btf_kind_str[NR_BTF_KINDS] = {
[BTF_KIND_VAR] = "VAR",
[BTF_KIND_DATASEC] = "DATASEC",
[BTF_KIND_FLOAT] = "FLOAT",
+ [BTF_KIND_TAG] = "TAG",
};
static const char *btf__printable_name(const struct btf *btf, uint32_t offset)
@@ -644,6 +645,26 @@ static int32_t btf_encoder__add_datasec(struct btf_encoder *encoder, const char
return id;
}
+static int32_t btf_encoder__add_tag(struct btf_encoder *encoder, const char *value, uint32_t type,
+ int component_idx)
+{
+ struct btf *btf = encoder->btf;
+ const struct btf_type *t;
+ int32_t id;
+
+ id = btf__add_tag(btf, value, type, component_idx);
+ if (id > 0) {
+ t = btf__type_by_id(btf, id);
+ btf_encoder__log_type(encoder, t, false, true, "type_id=%u component_idx=%d",
+ t->type, component_idx);
+ } else {
+ btf__log_err(btf, BTF_KIND_TAG, value, true, "component_idx=%d Error emitting BTF type",
+ component_idx);
+ }
+
+ return id;
+}
+
/*
* This corresponds to the same macro defined in
* include/linux/kallsyms.h
@@ -1158,6 +1179,7 @@ static int btf_encoder__encode_cu_variables(struct btf_encoder *encoder, struct
struct variable *var = tag__variable(pos);
uint32_t size, type, linkage;
const char *name, *dwarf_name;
+ struct llvm_annotation *annot;
const struct tag *tag;
uint64_t addr;
int id;
@@ -1244,6 +1266,10 @@ static int btf_encoder__encode_cu_variables(struct btf_encoder *encoder, struct
goto out;
}
+ list_for_each_entry(annot, &var->annots, node) {
+ btf_encoder__add_tag(encoder, annot->value, id, annot->component_idx);
+ }
+
/*
* add a BTF_VAR_SECINFO in encoder->percpu_secinfo, which will be added into
* encoder->types later when we add BTF_VAR_DATASEC.
@@ -1359,6 +1385,7 @@ void btf_encoder__delete(struct btf_encoder *encoder)
int btf_encoder__encode_cu(struct btf_encoder *encoder, struct cu *cu)
{
uint32_t type_id_off = btf__get_nr_types(encoder->btf);
+ struct llvm_annotation *annot;
uint32_t core_id;
struct function *fn;
struct tag *pos;
@@ -1396,6 +1423,20 @@ int btf_encoder__encode_cu(struct btf_encoder *encoder, struct cu *cu)
encoder->has_index_type = true;
}
+ cu__for_each_type(cu, core_id, pos) {
+ struct namespace *ns;
+ int btf_type_id;
+
+ if (pos->tag != DW_TAG_structure_type && pos->tag != DW_TAG_union_type)
+ continue;
+
+ btf_type_id = type_id_off + core_id;
+ ns = tag__namespace(pos);
+ list_for_each_entry(annot, &ns->annots, node) {
+ btf_encoder__add_tag(encoder, annot->value, btf_type_id, annot->component_idx);
+ }
+ }
+
cu__for_each_function(cu, core_id, fn) {
int btf_fnproto_id, btf_fn_id;
const char *name;
@@ -1436,6 +1477,10 @@ int btf_encoder__encode_cu(struct btf_encoder *encoder, struct cu *cu)
printf("error: failed to encode function '%s'\n", function__name(fn));
goto out;
}
+
+ list_for_each_entry(annot, &fn->annots, node) {
+ btf_encoder__add_tag(encoder, annot->value, btf_fn_id, annot->component_idx);
+ }
}
if (!encoder->skip_encoding_vars)
--
2.30.2
^ permalink raw reply related [flat|nested] 6+ messages in thread