perf tools: Handle relocatable kernels

DSOs don't have this problem because the kernel emits a
PERF_MMAP for each new executable mapping it performs on
monitored threads.

To fix the kernel case we simulate the same behaviour, by having
'perf record' to synthesize a PERF_MMAP for the kernel, encoded
like this:

[root@doppio ~]# perf record -a -f sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ]
[root@doppio ~]# perf report -D | head -10

0xd0 [0x40]: event: 1
.
. ... raw event: size 64 bytes
.  0000:  01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........
.  0010:  00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ...............
.  0020:  00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........  [kernel
.  0030:  6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00  kallsyms._text]
.  0xd0
[0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text]

I.e. we identify such event as having:

 .pid      = 0
 .filename = [kernel.kallsyms.REFNAME]
 .start    = REFNAME addr in /proc/kallsyms at 'perf record' time

and use now a hardcoded value of '.text' for REFNAME.

Then, later, in 'perf report', if there are any kernel hits and
thus we need to resolve kernel symbols, we search for REFNAME
and if its address changed, relocation happened and we thus must
change the kernel mapping routines to one that uses .pgoff as
the relocation to apply.

This way we use the same mechanism used for the other DSOs and
don't have to do a two pass in all the kernel symbols.

Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

tools/perf/builtin-record.c

@@ -551,6 +551,13 @@ static int __cmd_record(int argc, const char **argv)
 			return err;
 	}
 
+	err = event__synthesize_kernel_mmap(process_synthesized_event,
+					    session, "_text");
+	if (err < 0) {
+		pr_err("Couldn't record kernel reference relocation symbol.\n");
+		return err;
+	}
+
 	if (!system_wide && profile_cpu == -1)
 		event__synthesize_thread(pid, process_synthesized_event,
 					 session);

tools/perf/util/event.c

@@ -189,6 +189,50 @@ void event__synthesize_threads(int (*process)(event_t *event,
 	closedir(proc);
 }
 
+struct process_symbol_args {
+	const char *name;
+	u64	   start;
+};
+
+static int find_symbol_cb(void *arg, const char *name, char type, u64 start)
+{
+	struct process_symbol_args *args = arg;
+
+	if (!symbol_type__is_a(type, MAP__FUNCTION) || strcmp(name, args->name))
+		return 0;
+
+	args->start = start;
+	return 1;
+}
+
+int event__synthesize_kernel_mmap(int (*process)(event_t *event,
+						 struct perf_session *session),
+				  struct perf_session *session,
+				  const char *symbol_name)
+{
+	size_t size;
+	event_t ev = {
+		.header = { .type = PERF_RECORD_MMAP },
+	};
+	/*
+	 * We should get this from /sys/kernel/sections/.text, but till that is
+	 * available use this, and after it is use this as a fallback for older
+	 * kernels.
+	 */
+	struct process_symbol_args args = { .name = symbol_name, };
+
+	if (kallsyms__parse(&args, find_symbol_cb) <= 0)
+		return -ENOENT;
+
+	size = snprintf(ev.mmap.filename, sizeof(ev.mmap.filename),
+			"[kernel.kallsyms.%s]", symbol_name) + 1;
+	size = ALIGN(size, sizeof(u64));
+	ev.mmap.header.size = (sizeof(ev.mmap) - (sizeof(ev.mmap.filename) - size));
+	ev.mmap.start = args.start;
+
+	return process(&ev, session);
+}
+
 static void thread__comm_adjust(struct thread *self)
 {
 	char *comm = self->comm;
@@ -240,9 +284,9 @@ int event__process_lost(event_t *self, struct perf_session *session)
 
 int event__process_mmap(event_t *self, struct perf_session *session)
 {
-	struct thread *thread = perf_session__findnew(session, self->mmap.pid);
+	struct thread *thread;
-	struct map *map = map__new(&self->mmap, MAP__FUNCTION,
+	struct map *map;
-				   session->cwd, session->cwdlen);
+	static const char kmmap_prefix[] = "[kernel.kallsyms.";
 
 	dump_printf(" %d/%d: [%p(%p) @ %p]: %s\n",
 		    self->mmap.pid, self->mmap.tid,
@@ -251,6 +295,20 @@ int event__process_mmap(event_t *self, struct perf_session *session)
 		    (void *)(long)self->mmap.pgoff,
 		    self->mmap.filename);
 
+	if (self->mmap.pid == 0 &&
+	    memcmp(self->mmap.filename, kmmap_prefix,
+		   sizeof(kmmap_prefix) - 1) == 0) {
+		const char *symbol_name = (self->mmap.filename +
+					   sizeof(kmmap_prefix) - 1);
+		perf_session__set_kallsyms_ref_reloc_sym(session, symbol_name,
+							 self->mmap.start);
+		return 0;
+	}
+
+	thread = perf_session__findnew(session, self->mmap.pid);
+	map = map__new(&self->mmap, MAP__FUNCTION,
+		       session->cwd, session->cwdlen);
+
 	if (thread == NULL || map == NULL)
 		dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n");
 	else

tools/perf/util/event.h

@@ -110,6 +110,10 @@ int event__synthesize_thread(pid_t pid,
 void event__synthesize_threads(int (*process)(event_t *event,
 					      struct perf_session *session),
 			       struct perf_session *session);
+int event__synthesize_kernel_mmap(int (*process)(event_t *event,
+						 struct perf_session *session),
+				  struct perf_session *session,
+				  const char *symbol_name);
 
 int event__process_comm(event_t *self, struct perf_session *session);
 int event__process_lost(event_t *self, struct perf_session *session);

tools/perf/util/session.c

@@ -401,3 +401,49 @@ bool perf_session__has_traces(struct perf_session *self, const char *msg)
 
 	return true;
 }
+
+int perf_session__set_kallsyms_ref_reloc_sym(struct perf_session *self,
+					     const char *symbol_name,
+					     u64 addr)
+{
+	char *bracket;
+
+	self->ref_reloc_sym.name = strdup(symbol_name);
+	if (self->ref_reloc_sym.name == NULL)
+		return -ENOMEM;
+
+	bracket = strchr(self->ref_reloc_sym.name, ']');
+	if (bracket)
+		*bracket = '\0';
+
+	self->ref_reloc_sym.addr = addr;
+	return 0;
+}
+
+static u64 map__reloc_map_ip(struct map *map, u64 ip)
+{
+	return ip + (s64)map->pgoff;
+}
+
+static u64 map__reloc_unmap_ip(struct map *map, u64 ip)
+{
+	return ip - (s64)map->pgoff;
+}
+
+void perf_session__reloc_vmlinux_maps(struct perf_session *self,
+				      u64 unrelocated_addr)
+{
+	enum map_type type;
+	s64 reloc = unrelocated_addr - self->ref_reloc_sym.addr;
+
+	if (!reloc)
+		return;
+
+	for (type = 0; type < MAP__NR_TYPES; ++type) {
+		struct map *map = self->vmlinux_maps[type];
+
+		map->map_ip = map__reloc_map_ip;
+		map->unmap_ip = map__reloc_unmap_ip;
+		map->pgoff = reloc;
+	}
+}

tools/perf/util/session.h

@@ -24,6 +24,10 @@ struct perf_session {
 	unsigned long		unknown_events;
 	struct rb_root		hists;
 	u64			sample_type;
+	struct {
+		const char	*name;
+		u64		addr;
+	}			ref_reloc_sym;
 	int			fd;
 	int			cwdlen;
 	char			*cwd;
@@ -59,4 +63,10 @@ bool perf_session__has_traces(struct perf_session *self, const char *msg);
 
 int perf_header__read_build_ids(int input, u64 offset, u64 file_size);
 
+int perf_session__set_kallsyms_ref_reloc_sym(struct perf_session *self,
+					     const char *symbol_name,
+					     u64 addr);
+void perf_session__reloc_vmlinux_maps(struct perf_session *self,
+				      u64 unrelocated_addr);
+
 #endif /* __PERF_SESSION_H */

tools/perf/util/symbol.c

@@ -956,11 +956,15 @@ static int dso__load_sym(struct dso *self, struct map *map,
 
 	elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) {
 		struct symbol *f;
-		const char *elf_name;
+		const char *elf_name = elf_sym__name(&sym, symstrs);
 		char *demangled = NULL;
 		int is_label = elf_sym__is_label(&sym);
 		const char *section_name;
 
+		if (kernel && session->ref_reloc_sym.name != NULL &&
+		    strcmp(elf_name, session->ref_reloc_sym.name) == 0)
+			perf_session__reloc_vmlinux_maps(session, sym.st_value);
+
 		if (!is_label && !elf_sym__is_a(&sym, map->type))
 			continue;
 
@@ -973,7 +977,6 @@ static int dso__load_sym(struct dso *self, struct map *map,
 		if (is_label && !elf_sec__is_a(&shdr, secstrs, map->type))
 			continue;
 
-		elf_name = elf_sym__name(&sym, symstrs);
 		section_name = elf_sec__name(&shdr, secstrs);
 
 		if (kernel || kmodule) {