[RFC PATCH v6 24/32] perf record: Compile scriptlets if pass '.c' to --event

[Wang Nan <wangnan0@huawei.com>]

This patch enables passing source files to --event directly using:

 # perf record --event bpf-file.c command

This patch does following works:

 1) 'struct bpf_param' is introduced to store compiler related options;

 2) bpf__compile() is introduced for compiling source code by calling
    clang and llc.

 3) Allow passing '.c' file to '--event'. parse_events_load_bpf() is
    expanded to allow caller tell it whether the passed file is source
    file or object.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
---
 tools/perf/util/bpf-loader.c   | 177 ++++++++++++++++++++++++++++++++++++++++-
 tools/perf/util/bpf-loader.h   |  10 ++-
 tools/perf/util/parse-events.c |   4 +-
 tools/perf/util/parse-events.h |   2 +-
 tools/perf/util/parse-events.l |   3 +
 tools/perf/util/parse-events.y |  15 +++-
 6 files changed, 203 insertions(+), 8 deletions(-)

diff --git a/tools/perf/util/bpf-loader.c b/tools/perf/util/bpf-loader.c
index e950baa..e5c03e9 100644
--- a/tools/perf/util/bpf-loader.c
+++ b/tools/perf/util/bpf-loader.c
@@ -28,7 +28,168 @@ DEFINE_PRINT_FN(debug, 1)
 
 static bool libbpf_initialized = false;
 
-int bpf__prepare_load(const char *filename)
+/*
+ * Group all bpf related programs together. Futher we can config them
+ * using perf_default_config().
+ */
+struct bpf_param bpf_param = {
+	.clang_path = "clang",
+	.llc_path = "llc",
+	.clang_opt = "",
+	.llc_opt = "",
+};
+
+static int search_program(const char *def, const char *name,
+			  char *output)
+{
+	char *env, *path, *tmp;
+	char buf[PATH_MAX];
+	int ret;
+
+	if (def && def[0] == '/')
+		if (access(def, F_OK) == 0) {
+			strncpy(output, def, PATH_MAX);
+			return 0;
+		}
+
+	env = getenv("PATH");
+	if (!env)
+		return -1;
+
+	env = strdup(env);
+	if (!env)
+		return -1;
+
+	ret = -1;
+	path = strtok_r(env, ":", &tmp);
+	while (path) {
+		scnprintf(buf, sizeof(buf), "%s/%s", path, name);
+		if (access(buf, F_OK) == 0) {
+			strncpy(output, buf, PATH_MAX);
+			ret = 0;
+			goto out;
+		}
+		path = strtok_r(NULL, ":", &tmp);
+	}
+
+out:
+	free(env);
+	return ret;
+}
+
+#define READ_SIZE	4096
+/*
+ * A tricky way to ensure pclose() always get failure exit status.
+ * If simple passing 'clang | llc' to popen(), and clang fails, its
+ * exit status is !0 but stdout is empty, which is valid for llc, so
+ * pclose() gets 0 as final exit status.
+ *
+ * In CMD_FMT, use 'if !clang ; then echo ERROR ; fi | llc' to ensure
+ * llc get bad input and return !0 exit status.
+ */
+#define CMD_FMT "if ! %s %s -Wno-unused-value -Wno-pointer-sign " \
+		"-emit-llvm -c \"%s\" -O2 -o - ; then echo ERROR ; fi " \
+		" | %s %s -march=bpf -filetype=obj -o %s"
+
+static int bpf__compile(const char *filename, void **p_obj_buf,
+			size_t *p_obj_buf_sz)
+{
+	char clang_path[PATH_MAX], llc_path[PATH_MAX], *cmd = NULL;
+	int err, cmd_sz;
+	FILE *file;
+	void *buf = NULL;
+	size_t buf_sz = 0, read_sz = 0;
+
+	*p_obj_buf = NULL;
+	*p_obj_buf_sz = 0;
+
+	err = search_program(bpf_param.clang_path,
+			     "clang", clang_path);
+	if (!err)
+		err = search_program(bpf_param.llc_path,
+				     "llc", llc_path);
+	if (err) {
+		err = -ENOENT;
+		pr_err("Error:\tunable to find clang or llc in $PATH\n");
+		pr_err("Hint:\tTry install LLVM with BPF backend.\n");
+		goto out;
+	}
+
+	cmd_sz = snprintf(NULL, 0, CMD_FMT,
+			  clang_path, bpf_param.clang_opt, filename,
+			  llc_path, bpf_param.llc_opt, "-");
+
+	if (cmd_sz <= 0) {
+		err = -EINVAL;
+		pr_err("Internal error: sprintf returns %d\n", cmd_sz);
+		goto out;
+	}
+
+	cmd = malloc(cmd_sz + 1);
+	if (!cmd) {
+		err = -ENOMEM;
+		pr_err("Unable to alloc compiling command\n");
+		goto out;
+	}
+
+	err = snprintf(cmd, cmd_sz + 1, CMD_FMT,
+		       clang_path, bpf_param.clang_opt, filename,
+		       llc_path, bpf_param.llc_opt, "-");
+	if (err != cmd_sz) {
+		err = -EINVAL;
+		pr_err("Internal error: snprintf returns %d\n", err);
+		goto out;
+	}
+	cmd[cmd_sz] = '\0';
+
+	file = popen(cmd, "r");
+	if (!file) {
+		err = -EINVAL;
+		pr_err("Error:\tfailed to run command:\n%s\nReason:\t%s\n",
+		       cmd, strerror(errno));
+		goto out;
+	}
+
+	while (!feof(file) && !ferror(file)) {
+		void *new_buf;
+
+		if (buf_sz - read_sz < READ_SIZE) {
+			buf_sz += READ_SIZE;
+			new_buf = realloc(buf, buf_sz);
+			if (!new_buf) {
+				err = -ENOMEM;
+				pr_err("readlloc() failed\n");
+				goto out;
+			}
+			buf = new_buf;
+		}
+
+		read_sz += fread(buf + read_sz, 1, READ_SIZE, file);
+	}
+
+	if (ferror(file))
+		read_sz = -EINVAL;
+
+	err = WEXITSTATUS(pclose(file));
+	if (err || read_sz <= 0) {
+		err = -EINVAL;
+		pr_err("Error:\tsomething went wrong when compiling %s\n",
+		       filename);
+		pr_err("Hint:\tTry manually run following command and check:\n");
+		pr_err("  # " CMD_FMT "\n\n",
+		       clang_path, bpf_param.clang_opt, filename,
+		       llc_path, bpf_param.llc_opt, "/dev/null");
+		goto out;
+	}
+
+	*p_obj_buf_sz = read_sz;
+	*p_obj_buf = buf;
+out:
+	free(cmd);
+	return err;
+}
+
+int bpf__prepare_load(const char *filename, bool source)
 {
 	struct bpf_object *obj;
 
@@ -37,7 +198,19 @@ int bpf__prepare_load(const char *filename)
 				 libbpf_info,
 				 libbpf_debug);
 
-	obj = bpf_object__open(filename);
+	if (source) {
+		void *obj_buf;
+		size_t obj_buf_sz;
+		int err;
+
+		err = bpf__compile(filename, &obj_buf, &obj_buf_sz);
+		if (err)
+			return err;
+		obj = bpf_object__open_buffer(obj_buf, obj_buf_sz);
+		free(obj_buf);
+	} else
+		obj = bpf_object__open(filename);
+
 	if (!obj) {
 		pr_err("bpf: failed to load %s\n", filename);
 		return -EINVAL;
diff --git a/tools/perf/util/bpf-loader.h b/tools/perf/util/bpf-loader.h
index 39d8d1a..5fd015c 100644
--- a/tools/perf/util/bpf-loader.h
+++ b/tools/perf/util/bpf-loader.h
@@ -8,8 +8,16 @@
 #include <linux/compiler.h>
 #include "debug.h"
 
+struct bpf_param {
+	const char *clang_path;
+	const char *llc_path;
+	const char *clang_opt;
+	const char *llc_opt;
+};
+extern struct bpf_param bpf_param;
+
 #ifdef HAVE_LIBBPF_SUPPORT
-int bpf__prepare_load(const char *filename);
+int bpf__prepare_load(const char *filename, bool source);
 
 void bpf__clear(void);
 #else
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 6f7a7b1..aaee24c 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -475,7 +475,7 @@ int parse_events_add_tracepoint(struct list_head *list, int *idx,
 
 int parse_events_load_bpf(struct list_head *list __maybe_unused,
 			  int *idx __maybe_unused,
-			  char *bpf_file_name)
+			  char *bpf_file_name, bool source)
 {
 	/*
 	 * Currently don't link any event to list. BPF object files
@@ -485,7 +485,7 @@ int parse_events_load_bpf(struct list_head *list __maybe_unused,
 	 * problem. After that probe events file by file is possible.
 	 * However, probing cost is still need to be considered.
 	 */
-	return bpf__prepare_load(bpf_file_name);
+	return bpf__prepare_load(bpf_file_name, source);
 }
 
 static int
diff --git a/tools/perf/util/parse-events.h b/tools/perf/util/parse-events.h
index 41b962a..5841d4f 100644
--- a/tools/perf/util/parse-events.h
+++ b/tools/perf/util/parse-events.h
@@ -115,7 +115,7 @@ int parse_events_name(struct list_head *list, char *name);
 int parse_events_add_tracepoint(struct list_head *list, int *idx,
 				char *sys, char *event);
 int parse_events_load_bpf(struct list_head *list, int *idx,
-			  char *bpf_file_name);
+			  char *bpf_file_name, bool source);
 int parse_events_add_numeric(struct parse_events_evlist *data,
 			     struct list_head *list,
 			     u32 type, u64 config,
diff --git a/tools/perf/util/parse-events.l b/tools/perf/util/parse-events.l
index 4acea38..eb821d4 100644
--- a/tools/perf/util/parse-events.l
+++ b/tools/perf/util/parse-events.l
@@ -116,6 +116,7 @@ group		[^,{}/]*[{][^}]*[}][^,{}/]*
 event_pmu	[^,{}/]+[/][^/]*[/][^,{}/]*
 event		[^,{}/]+
 bpf_object	.*\.(o|bpf)
+bpf_source	.*\.c
 
 num_dec		[0-9]+
 num_hex		0x[a-fA-F0-9]+
@@ -161,6 +162,7 @@ modifier_bp	[rwx]{1,3}
 
 {event_pmu}	|
 {bpf_object}	|
+{bpf_source}	|
 {event}		{
 			BEGIN(INITIAL);
 			REWIND(1);
@@ -265,6 +267,7 @@ r{num_raw_hex}		{ return raw(yyscanner); }
 {modifier_event}	{ return str(yyscanner, PE_MODIFIER_EVENT); }
 {name}			{ return pmu_str_check(yyscanner); }
 {bpf_object}		{ return str(yyscanner, PE_BPF_OBJECT); }
+{bpf_source}		{ return str(yyscanner, PE_BPF_SOURCE); }
 "/"			{ BEGIN(config); return '/'; }
 -			{ return '-'; }
 ,			{ BEGIN(event); return ','; }
diff --git a/tools/perf/util/parse-events.y b/tools/perf/util/parse-events.y
index 481f3cd..eeb9768 100644
--- a/tools/perf/util/parse-events.y
+++ b/tools/perf/util/parse-events.y
@@ -42,7 +42,7 @@ static inc_group_count(struct list_head *list,
 %token PE_VALUE PE_VALUE_SYM_HW PE_VALUE_SYM_SW PE_RAW PE_TERM
 %token PE_EVENT_NAME
 %token PE_NAME
-%token PE_BPF_OBJECT
+%token PE_BPF_OBJECT PE_BPF_SOURCE
 %token PE_MODIFIER_EVENT PE_MODIFIER_BP
 %token PE_NAME_CACHE_TYPE PE_NAME_CACHE_OP_RESULT
 %token PE_PREFIX_MEM PE_PREFIX_RAW PE_PREFIX_GROUP
@@ -55,6 +55,7 @@ static inc_group_count(struct list_head *list,
 %type <num> PE_TERM
 %type <str> PE_NAME
 %type <str> PE_BPF_OBJECT
+%type <str> PE_BPF_SOURCE
 %type <str> PE_NAME_CACHE_TYPE
 %type <str> PE_NAME_CACHE_OP_RESULT
 %type <str> PE_MODIFIER_EVENT
@@ -431,7 +432,17 @@ PE_BPF_OBJECT
 	struct list_head *list;
 
 	ALLOC_LIST(list);
-	ABORT_ON(parse_events_load_bpf(list, &data->idx, $1));
+	ABORT_ON(parse_events_load_bpf(list, &data->idx, $1, false));
+	$$ = list;
+}
+|
+PE_BPF_SOURCE
+{
+	struct parse_events_evlist *data = _data;
+	struct list_head *list;
+
+	ALLOC_LIST(list);
+	ABORT_ON(parse_events_load_bpf(list, &data->idx, $1, true));
 	$$ = list;
 }
 
-- 
1.8.3.4