perf arm64: Inject missing frames when using 'perf record --call-graph=fp'

When unwinding using frame pointers on ARM64, the return address of the
current function may not have been pushed into the stack when a function
was interrupted, which makes perf show an incorrect call graph to the
user.

Consider the following example program:

  void leaf() {
      /* long computation */
  }

  void parent() {
      // (1)
      leaf();
      // (2)
  }

  ... could be compiled into (using gcc -fno-inline -fno-omit-frame-pointer):

  leaf:
      /* long computation */
      nop
      ret
  parent:
      // (1)
      stp     x29, x30, [sp, -16]!
      mov     x29, sp
      bl      parent
      nop
      ldp     x29, x30, [sp], 16
      // (2)
      ret

If the program is interrupted at (1), (2), or any point in "leaf:", the
call graph will skip the callers of the current function. We can unwind
using the dwarf info and check if the return addr is the same as the LR
register, and inject the missing frame into the call graph.

Before this patch, the above example shows the following call-graph when
recording using "--call-graph fp" mode in ARM64:

  # Children      Self  Command   Shared Object     Symbol
  # ........  ........  ........  ................  ......................
  #
      99.86%    99.86%  program3  program3          [.] leaf
  	    |
  	    ---_start
  	       __libc_start_main
  	       main
  	       leaf

As can be seen, the "parent" function is missing. This is specially
problematic in "leaf" because for leaf functions the compiler may always
omit pushing the return addr into the stack. After this patch, it shows
the correct graph:

  # Children      Self  Command   Shared Object     Symbol
  # ........  ........  ........  ................  ......................
  #
      99.86%    99.86%  program3  program3          [.] leaf
  	    |
  	    ---_start
  	       __libc_start_main
  	       main
  	       parent
  	       leaf
Reviewed-by: James Clark <james.clark@arm.com>
Signed-off-by: Alexandre Truong <alexandre.truong@arm.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: John Garry <john.garry@huawei.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: https://lore.kernel.org/r/20211217154521.80603-7-german.gomez@arm.comSigned-off-by: German Gomez <german.gomez@arm.com>
[ Rename machine__normalize_is() to machine__normalized_is(), as suggested by James Clark ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools/perf/util/Build

+perf-y += arm64-frame-pointer-unwind-support.o
 perf-y += annotate.o
 perf-y += block-info.o
 perf-y += block-range.o

tools/perf/util/arm64-frame-pointer-unwind-support.c

+// SPDX-License-Identifier: GPL-2.0
+#include "arm64-frame-pointer-unwind-support.h"
+#include "callchain.h"
+#include "event.h"
+#include "perf_regs.h" // SMPL_REG_MASK
+#include "unwind.h"
+
+#define perf_event_arm_regs perf_event_arm64_regs
+#include "../arch/arm64/include/uapi/asm/perf_regs.h"
+#undef perf_event_arm_regs
+
+struct entries {
+	u64 stack[2];
+	size_t length;
+};
+
+static bool get_leaf_frame_caller_enabled(struct perf_sample *sample)
+{
+	return callchain_param.record_mode == CALLCHAIN_FP && sample->user_regs.regs
+		&& sample->user_regs.mask & SMPL_REG_MASK(PERF_REG_ARM64_LR);
+}
+
+static int add_entry(struct unwind_entry *entry, void *arg)
+{
+	struct entries *entries = arg;
+
+	entries->stack[entries->length++] = entry->ip;
+	return 0;
+}
+
+u64 get_leaf_frame_caller_aarch64(struct perf_sample *sample, struct thread *thread, int usr_idx)
+{
+	int ret;
+	struct entries entries = {};
+	struct regs_dump old_regs = sample->user_regs;
+
+	if (!get_leaf_frame_caller_enabled(sample))
+		return 0;
+
+	/*
+	 * If PC and SP are not recorded, get the value of PC from the stack
+	 * and set its mask. SP is not used when doing the unwinding but it
+	 * still needs to be set to prevent failures.
+	 */
+
+	if (!(sample->user_regs.mask & SMPL_REG_MASK(PERF_REG_ARM64_PC))) {
+		sample->user_regs.cache_mask |= SMPL_REG_MASK(PERF_REG_ARM64_PC);
+		sample->user_regs.cache_regs[PERF_REG_ARM64_PC] = sample->callchain->ips[usr_idx+1];
+	}
+
+	if (!(sample->user_regs.mask & SMPL_REG_MASK(PERF_REG_ARM64_SP))) {
+		sample->user_regs.cache_mask |= SMPL_REG_MASK(PERF_REG_ARM64_SP);
+		sample->user_regs.cache_regs[PERF_REG_ARM64_SP] = 0;
+	}
+
+	ret = unwind__get_entries(add_entry, &entries, thread, sample, 2);
+	sample->user_regs = old_regs;
+
+	if (ret || entries.length != 2)
+		return ret;
+
+	return callchain_param.order == ORDER_CALLER ? entries.stack[0] : entries.stack[1];
+}

tools/perf/util/arm64-frame-pointer-unwind-support.h

+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __PERF_ARM_FRAME_POINTER_UNWIND_SUPPORT_H
+#define __PERF_ARM_FRAME_POINTER_UNWIND_SUPPORT_H
+
+#include "event.h"
+#include "thread.h"
+
+u64 get_leaf_frame_caller_aarch64(struct perf_sample *sample, struct thread *thread, int user_idx);
+
+#endif /* __PERF_ARM_FRAME_POINTER_UNWIND_SUPPORT_H */

tools/perf/util/machine.c

@@ -34,6 +34,7 @@
 #include "bpf-event.h"
 #include <internal/lib.h> // page_size
 #include "cgroup.h"
+#include "arm64-frame-pointer-unwind-support.h"
 
 #include <linux/ctype.h>
 #include <symbol/kallsyms.h>
@@ -2710,10 +2711,13 @@ static int find_prev_cpumode(struct ip_callchain *chain, struct thread *thread,
 	return err;
 }
 
-static u64 get_leaf_frame_caller(struct perf_sample *sample __maybe_unused,
-		struct thread *thread __maybe_unused, int usr_idx __maybe_unused)
+static u64 get_leaf_frame_caller(struct perf_sample *sample,
+		struct thread *thread, int usr_idx)
 {
-	return 0;
+	if (machine__normalized_is(thread->maps->machine, "arm64"))
+		return get_leaf_frame_caller_aarch64(sample, thread, usr_idx);
+	else
+		return 0;
 }
 
 static int thread__resolve_callchain_sample(struct thread *thread,
@@ -3114,14 +3118,19 @@ int machine__set_current_tid(struct machine *machine, int cpu, pid_t pid,
 }
 
 /*
- * Compares the raw arch string. N.B. see instead perf_env__arch() if a
- * normalized arch is needed.
+ * Compares the raw arch string. N.B. see instead perf_env__arch() or
+ * machine__normalized_is() if a normalized arch is needed.
  */
 bool machine__is(struct machine *machine, const char *arch)
 {
 	return machine && !strcmp(perf_env__raw_arch(machine->env), arch);
 }
 
+bool machine__normalized_is(struct machine *machine, const char *arch)
+{
+	return machine && !strcmp(perf_env__arch(machine->env), arch);
+}
+
 int machine__nr_cpus_avail(struct machine *machine)
 {
 	return machine ? perf_env__nr_cpus_avail(machine->env) : 0;

tools/perf/util/machine.h

@@ -208,6 +208,7 @@ static inline bool machine__is_host(struct machine *machine)
 }
 
 bool machine__is(struct machine *machine, const char *arch);
+bool machine__normalized_is(struct machine *machine, const char *arch);
 int machine__nr_cpus_avail(struct machine *machine);
 
 struct thread *__machine__findnew_thread(struct machine *machine, pid_t pid, pid_t tid);