ftrace: Synchronize the initialization of ftrace_opsgrafted

There's some internal state that ftrace_ops needs to have set, but since
it can be declared outside of the ftrace.c code, it calls
ftrace_ops_init() on the ops in every global function. The issue is that
if two tasks call it on the same ops at the same time it is possible to
have the initialization of one corrupt the initialization of the other
call.

Create a ops_mutex to use to synchronize every initialization of the
ftrace_ops. The mutex is taken within checking the ftrace_ops flag that
states it was initializied but the flag is checked again after the mutex
has been taken. Checking first outside the mutex allows it to shortcut
having to take the mutex. But then the check needs to be done again after
the mute is taken in case of races.

Cc: stable@vger.kernel.org
Link: https://patch.msgid.link/20260902095501.6b59af20@gandalf.local.home
Fixes: f04f24fb7e48d ("ftrace, kprobes: Fix a deadlock on ftrace_regex_lock")
Reported-by: sashiko-bot@kernel.org
Close: https://lore.kernel.org/all/20260829025528.49A831F000E9@smtp.kernel.org/
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

1 files changed, 242 insertions, 0 deletions

diff --git a/tools/perf/tests/event-times.c b/tools/perf/tests/event-times.c
new file mode 100644
index 000000000..56dd37ca7
--- /dev/null
+++ b/tools/perf/tests/event-times.c
@@ -0,0 +1,242 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/compiler.h>
+#include <linux/string.h>
+#include <errno.h>
+#include <inttypes.h>
+#include <string.h>
+#include <sys/wait.h>
+#include <perf/cpumap.h>
+#include "tests.h"
+#include "evlist.h"
+#include "evsel.h"
+#include "debug.h"
+#include "parse-events.h"
+#include "thread_map.h"
+#include "target.h"
+
+static int attach__enable_on_exec(struct evlist *evlist)
+{
+	struct evsel *evsel = evlist__last(evlist);
+	struct target target = {};
+	const char *argv[] = { "true", NULL, };
+	char sbuf[STRERR_BUFSIZE];
+	int err;
+
+	pr_debug("attaching to spawned child, enable on exec\n");
+
+	err = evlist__create_maps(evlist, &target);
+	if (err < 0) {
+		pr_debug("Not enough memory to create thread/cpu maps\n");
+		return err;
+	}
+
+	err = evlist__prepare_workload(evlist, &target, argv, false, NULL);
+	if (err < 0) {
+		pr_debug("Couldn't run the workload!\n");
+		return err;
+	}
+
+	evsel->core.attr.enable_on_exec = 1;
+
+	err = evlist__open(evlist);
+	if (err < 0) {
+		pr_debug("perf_evlist__open: %s\n",
+			 str_error_r(errno, sbuf, sizeof(sbuf)));
+		return err;
+	}
+
+	return evlist__start_workload(evlist) == 1 ? TEST_OK : TEST_FAIL;
+}
+
+static int detach__enable_on_exec(struct evlist *evlist)
+{
+	waitpid(evlist__workload_pid(evlist), NULL, 0);
+	return 0;
+}
+
+static int attach__current_disabled(struct evlist *evlist)
+{
+	struct evsel *evsel = evlist__last(evlist);
+	struct perf_thread_map *threads;
+	int err;
+
+	pr_debug("attaching to current thread as disabled\n");
+
+	threads = thread_map__new_by_tid(getpid());
+	if (threads == NULL) {
+		pr_debug("thread_map__new\n");
+		return -1;
+	}
+
+	evsel->core.attr.disabled = 1;
+
+	err = evsel__open_per_thread(evsel, threads);
+	if (err) {
+		pr_debug("Failed to open event cpu-clock:u\n");
+		return err;
+	}
+
+	perf_thread_map__put(threads);
+	return evsel__enable(evsel) == 0 ? TEST_OK : TEST_FAIL;
+}
+
+static int attach__current_enabled(struct evlist *evlist)
+{
+	struct evsel *evsel = evlist__last(evlist);
+	struct perf_thread_map *threads;
+	int err;
+
+	pr_debug("attaching to current thread as enabled\n");
+
+	threads = thread_map__new_by_tid(getpid());
+	if (threads == NULL) {
+		pr_debug("failed to call thread_map__new\n");
+		return -1;
+	}
+
+	err = evsel__open_per_thread(evsel, threads);
+
+	perf_thread_map__put(threads);
+	return err == 0 ? TEST_OK : TEST_FAIL;
+}
+
+static int detach__disable(struct evlist *evlist)
+{
+	struct evsel *evsel = evlist__last(evlist);
+
+	return evsel__enable(evsel);
+}
+
+static int attach__cpu_disabled(struct evlist *evlist)
+{
+	struct evsel *evsel = evlist__last(evlist);
+	struct perf_cpu_map *cpus;
+	int err;
+
+	pr_debug("attaching to CPU 0 as enabled\n");
+
+	cpus = perf_cpu_map__new("0");
+	if (cpus == NULL) {
+		pr_debug("failed to call perf_cpu_map__new\n");
+		return -1;
+	}
+
+	evsel->core.attr.disabled = 1;
+
+	err = evsel__open_per_cpu(evsel, cpus, -1);
+	perf_cpu_map__put(cpus);
+	if (err) {
+		if (err == -EACCES)
+			return TEST_SKIP;
+
+		pr_debug("Failed to open event cpu-clock:u\n");
+		return err;
+	}
+
+	return evsel__enable(evsel);
+}
+
+static int attach__cpu_enabled(struct evlist *evlist)
+{
+	struct evsel *evsel = evlist__last(evlist);
+	struct perf_cpu_map *cpus;
+	int err;
+
+	pr_debug("attaching to CPU 0 as enabled\n");
+
+	cpus = perf_cpu_map__new("0");
+	if (cpus == NULL) {
+		pr_debug("failed to call perf_cpu_map__new\n");
+		return -1;
+	}
+
+	err = evsel__open_per_cpu(evsel, cpus, -1);
+	perf_cpu_map__put(cpus);
+	if (err == -EACCES)
+		return TEST_SKIP;
+
+	return err ? TEST_FAIL : TEST_OK;
+}
+
+static int test_times(int (attach)(struct evlist *),
+		      int (detach)(struct evlist *))
+{
+	struct perf_counts_values count;
+	struct evlist *evlist = NULL;
+	struct evsel *evsel;
+	int err = -1, i;
+
+	evlist = evlist__new();
+	if (!evlist) {
+		pr_debug("failed to create event list\n");
+		goto out_err;
+	}
+
+	err = parse_event(evlist, "cpu-clock:u");
+	if (err) {
+		pr_debug("failed to parse event cpu-clock:u\n");
+		goto out_err;
+	}
+
+	evsel = evlist__last(evlist);
+	evsel->core.attr.read_format |=
+		PERF_FORMAT_TOTAL_TIME_ENABLED |
+		PERF_FORMAT_TOTAL_TIME_RUNNING;
+
+	err = attach(evlist);
+	if (err == TEST_SKIP) {
+		pr_debug("  SKIP  : not enough rights\n");
+		evlist__put(evlist);
+		return err;
+	}
+
+	TEST_ASSERT_VAL("failed to attach", !err);
+
+	for (i = 0; i < 100000000; i++) { }
+
+	TEST_ASSERT_VAL("failed to detach", !detach(evlist));
+
+	perf_evsel__read(&evsel->core, 0, 0, &count);
+
+	err = !(count.ena == count.run);
+
+	pr_debug("  %s: ena %" PRIu64", run %" PRIu64"\n",
+		 !err ? "OK    " : "FAILED",
+		 count.ena, count.run);
+
+out_err:
+	evlist__put(evlist);
+	return !err ? TEST_OK : TEST_FAIL;
+}
+
+/*
+ * This test creates software event 'cpu-clock'
+ * attaches it in several ways (explained below)
+ * and checks that enabled and running times
+ * match.
+ */
+static int test__event_times(struct test_suite *test __maybe_unused, int subtest __maybe_unused)
+{
+	int err, ret = 0;
+
+#define _T(attach, detach)			\
+	err = test_times(attach, detach);	\
+	if (err && (ret == TEST_OK || ret == TEST_SKIP))	\
+		ret = err;
+
+	/* attach on newly spawned process after exec */
+	_T(attach__enable_on_exec,   detach__enable_on_exec)
+	/* attach on current process as enabled */
+	_T(attach__current_enabled,  detach__disable)
+	/* attach on current process as disabled */
+	_T(attach__current_disabled, detach__disable)
+	/* attach on cpu as disabled */
+	_T(attach__cpu_disabled,     detach__disable)
+	/* attach on cpu as enabled */
+	_T(attach__cpu_enabled,      detach__disable)
+
+#undef _T
+	return ret;
+}
+
+DEFINE_SUITE("Event times", event_times);