[PATCH v6 10/15] perf dso: Support BPF programs in dso__read_symbol

[Ian Rogers <irogers@google.com>]

Set the buffer to the code in the BPF linear info. This enables BPF
JIT code disassembly by LLVM and capstone. Move the common but minimal
disassmble_bpf_image call to disassemble_objdump so that it is only
called after falling back to the objdump option. Similarly move the
disassmble_bpf function to disassemble_objdump and rename to
disassmble_bpf_libbfd to make it clearer that this support relies on
libbfd.

Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/util/disasm.c |  12 +++--
 tools/perf/util/dso.c    | 100 ++++++++++++++++++++++++++-------------
 tools/perf/util/libbfd.c |   4 +-
 tools/perf/util/libbfd.h |   6 +--
 4 files changed, 80 insertions(+), 42 deletions(-)

diff --git a/tools/perf/util/disasm.c b/tools/perf/util/disasm.c
index a1240543c89c..e64902e520ab 100644
--- a/tools/perf/util/disasm.c
+++ b/tools/perf/util/disasm.c
@@ -1521,6 +1521,12 @@ static int symbol__disassemble_objdump(const char *filename, struct symbol *sym,
 	struct child_process objdump_process;
 	int err;
 
+	if (dso__binary_type(dso) == DSO_BINARY_TYPE__BPF_PROG_INFO)
+		return symbol__disassemble_bpf_libbfd(sym, args);
+
+	if (dso__binary_type(dso) == DSO_BINARY_TYPE__BPF_IMAGE)
+		return symbol__disassemble_bpf_image(sym, args);
+
 	err = asprintf(&command,
 		 "%s %s%s --start-address=0x%016" PRIx64
 		 " --stop-address=0x%016" PRIx64
@@ -1655,11 +1661,7 @@ int symbol__disassemble(struct symbol *sym, struct annotate_args *args)
 
 	pr_debug("annotating [%p] %30s : [%p] %30s\n", dso, dso__long_name(dso), sym, sym->name);
 
-	if (dso__binary_type(dso) == DSO_BINARY_TYPE__BPF_PROG_INFO) {
-		return symbol__disassemble_bpf(sym, args);
-	} else if (dso__binary_type(dso) == DSO_BINARY_TYPE__BPF_IMAGE) {
-		return symbol__disassemble_bpf_image(sym, args);
-	} else if (dso__binary_type(dso) == DSO_BINARY_TYPE__NOT_FOUND) {
+	if (dso__binary_type(dso) == DSO_BINARY_TYPE__NOT_FOUND) {
 		return SYMBOL_ANNOTATE_ERRNO__COULDNT_DETERMINE_FILE_TYPE;
 	} else if (dso__is_kcore(dso)) {
 		kce.addr = map__rip_2objdump(map, sym->start);
diff --git a/tools/perf/util/dso.c b/tools/perf/util/dso.c
index 87d075942de6..0aed5c8691bd 100644
--- a/tools/perf/util/dso.c
+++ b/tools/perf/util/dso.c
@@ -1816,23 +1816,17 @@ static int find_file_offset(u64 start, u64 len, u64 pgoff, void *arg)
 	return 0;
 }
 
-const u8 *dso__read_symbol(struct dso *dso, const char *symfs_filename,
-			   const struct map *map, const struct symbol *sym,
-			   u8 **out_buf, u64 *out_buf_len, bool *is_64bit)
+static const u8 *__dso__read_symbol(struct dso *dso, const char *symfs_filename,
+				    u64 start, size_t len,
+				    u8 **out_buf, u64 *out_buf_len, bool *is_64bit)
 {
 	struct nscookie nsc;
-	u64 start = map__rip_2objdump(map, sym->start);
-	u64 end = map__rip_2objdump(map, sym->end);
-	int fd, count;
-	u8 *buf = NULL;
-	size_t len;
+	int fd;
+	ssize_t count;
 	struct find_file_offset_data data = {
 		.ip = start,
 	};
-
-	*out_buf = NULL;
-	*out_buf_len = 0;
-	*is_64bit = false;
+	u8 *code_buf = NULL;
 
 	nsinfo__mountns_enter(dso__nsinfo(dso), &nsc);
 	fd = open(symfs_filename, O_RDONLY);
@@ -1840,28 +1834,70 @@ const u8 *dso__read_symbol(struct dso *dso, const char *symfs_filename,
 	if (fd < 0)
 		return NULL;
 
-	if (file__read_maps(fd, /*exe=*/true, find_file_offset, &data, is_64bit) == 0)
-		goto err;
-
-	len = end - start;
-	buf = malloc(len);
-	if (buf == NULL)
-		goto err;
-
-	count = pread(fd, buf, len, data.offset);
+	if (file__read_maps(fd, /*exe=*/true, find_file_offset, &data, is_64bit) == 0) {
+		close(fd);
+		return NULL;
+	}
+	code_buf = malloc(len);
+	if (code_buf == NULL) {
+		close(fd);
+		return NULL;
+	}
+	count = pread(fd, code_buf, len, data.offset);
 	close(fd);
-	fd = -1;
+	if ((u64)count != len) {
+		free(code_buf);
+		return NULL;
+	}
+	*out_buf = code_buf;
+	*out_buf_len = len;
+	return code_buf;
+}
 
-	if ((u64)count != len)
-		goto err;
+/*
+ * Read a symbol into memory for disassembly by a library like capstone of
+ * libLLVM. If memory is allocated out_buf holds it.
+ */
+const u8 *dso__read_symbol(struct dso *dso, const char *symfs_filename,
+			   const struct map *map, const struct symbol *sym,
+			   u8 **out_buf, u64 *out_buf_len, bool *is_64bit)
+{
+	u64 start = map__rip_2objdump(map, sym->start);
+	u64 end = map__rip_2objdump(map, sym->end);
+	size_t len = end - start;
 
-	*out_buf = buf;
-	*out_buf_len = len;
-	return buf;
+	*out_buf = NULL;
+	*out_buf_len = 0;
+	*is_64bit = false;
 
-err:
-	if (fd >= 0)
-		close(fd);
-	free(buf);
-	return NULL;
+	if (dso__binary_type(dso) == DSO_BINARY_TYPE__BPF_IMAGE) {
+		/*
+		 * Note, there is fallback BPF image disassembly in the objdump
+		 * version but it currently does nothing.
+		 */
+		return NULL;
+	}
+	if (dso__binary_type(dso) == DSO_BINARY_TYPE__BPF_PROG_INFO) {
+#ifdef HAVE_LIBBPF_SUPPORT
+		struct bpf_prog_info_node *info_node;
+		struct perf_bpil *info_linear;
+
+		*is_64bit = sizeof(void *) == sizeof(u64);
+		info_node = perf_env__find_bpf_prog_info(dso__bpf_prog(dso)->env,
+							 dso__bpf_prog(dso)->id);
+		if (!info_node) {
+			errno = SYMBOL_ANNOTATE_ERRNO__BPF_MISSING_BTF;
+			return NULL;
+		}
+		info_linear = info_node->info_linear;
+		assert(len <= info_linear->info.jited_prog_len);
+		*out_buf_len = len;
+		return (const u8 *)(uintptr_t)(info_linear->info.jited_prog_insns);
+#else
+		pr_debug("No BPF program disassembly support\n");
+		return NULL;
+#endif
+	}
+	return __dso__read_symbol(dso, symfs_filename, start, len,
+				  out_buf, out_buf_len, is_64bit);
 }
diff --git a/tools/perf/util/libbfd.c b/tools/perf/util/libbfd.c
index 09a0eeb78a1a..01147fbf73b3 100644
--- a/tools/perf/util/libbfd.c
+++ b/tools/perf/util/libbfd.c
@@ -448,8 +448,8 @@ int libbfd_filename__read_debuglink(const char *filename, char *debuglink,
 	return err;
 }
 
-int symbol__disassemble_bpf(struct symbol *sym __maybe_unused,
-			    struct annotate_args *args  __maybe_unused)
+int symbol__disassemble_bpf_libbfd(struct symbol *sym __maybe_unused,
+				   struct annotate_args *args  __maybe_unused)
 {
 #ifdef HAVE_LIBBPF_SUPPORT
 	struct annotation *notes = symbol__annotation(sym);
diff --git a/tools/perf/util/libbfd.h b/tools/perf/util/libbfd.h
index 7441e95f8ec0..e300f171d1bd 100644
--- a/tools/perf/util/libbfd.h
+++ b/tools/perf/util/libbfd.h
@@ -29,7 +29,7 @@ int libbfd__read_build_id(const char *filename, struct build_id *bid, bool block
 
 int libbfd_filename__read_debuglink(const char *filename, char *debuglink, size_t size);
 
-int symbol__disassemble_bpf(struct symbol *sym, struct annotate_args *args);
+int symbol__disassemble_bpf_libbfd(struct symbol *sym, struct annotate_args *args);
 
 #else // !defined(HAVE_LIBBFD_SUPPORT)
 #include "annotate.h"
@@ -72,8 +72,8 @@ static inline int libbfd_filename__read_debuglink(const char *filename __always_
 	return -1;
 }
 
-static inline int symbol__disassemble_bpf(struct symbol *sym __always_unused,
-					  struct annotate_args *args __always_unused)
+static inline int symbol__disassemble_bpf_libbfd(struct symbol *sym __always_unused,
+						 struct annotate_args *args __always_unused)
 {
 	return SYMBOL_ANNOTATE_ERRNO__NO_LIBOPCODES_FOR_BPF;
 }
-- 
2.51.0.570.gb178f27e6d-goog