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, 86 insertions, 0 deletions

diff --git a/tools/perf/util/bpf-utils.h b/tools/perf/util/bpf-utils.h
new file mode 100644
index 000000000..a8bc1a232
--- /dev/null
+++ b/tools/perf/util/bpf-utils.h
@@ -0,0 +1,86 @@
+/* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */
+
+#ifndef __PERF_BPF_UTILS_H
+#define __PERF_BPF_UTILS_H
+
+#define ptr_to_u64(ptr)    ((__u64)(unsigned long)(ptr))
+
+#ifdef HAVE_LIBBPF_SUPPORT
+
+#include <bpf/libbpf.h>
+#include <bpf/libbpf_version.h>
+
+#define LIBBPF_CURRENT_VERSION_GEQ(major, minor)                       \
+       (LIBBPF_MAJOR_VERSION > (major) ||                              \
+        (LIBBPF_MAJOR_VERSION == (major) && LIBBPF_MINOR_VERSION >= (minor)))
+
+#if LIBBPF_CURRENT_VERSION_GEQ(1, 7)
+// libbpf 1.7+ support the btf_dump_type_data_opts.emit_strings option.
+#define HAVE_LIBBPF_STRINGS_SUPPORT 1
+#endif
+
+/*
+ * Get bpf_prog_info in continuous memory
+ *
+ * struct bpf_prog_info has multiple arrays. The user has option to choose
+ * arrays to fetch from kernel. The following APIs provide an uniform way to
+ * fetch these data. All arrays in bpf_prog_info are stored in a single
+ * continuous memory region. This makes it easy to store the info in a
+ * file.
+ *
+ * Before writing perf_bpil to files, it is necessary to
+ * translate pointers in bpf_prog_info to offsets. Helper functions
+ * bpil_addr_to_offs() and bpil_offs_to_addr()
+ * are introduced to switch between pointers and offsets.
+ *
+ * Examples:
+ *   # To fetch map_ids and prog_tags:
+ *   __u64 arrays = (1UL << PERF_BPIL_MAP_IDS) |
+ *           (1UL << PERF_BPIL_PROG_TAGS);
+ *   struct perf_bpil *info_linear =
+ *           get_bpf_prog_info_linear(fd, arrays);
+ *
+ *   # To save data in file
+ *   bpil_addr_to_offs(info_linear);
+ *   write(f, info_linear, sizeof(*info_linear) + info_linear->data_len);
+ *
+ *   # To read data from file
+ *   read(f, info_linear, <proper_size>);
+ *   bpil_offs_to_addr(info_linear);
+ */
+enum perf_bpil_array_types {
+	PERF_BPIL_FIRST_ARRAY = 0,
+	PERF_BPIL_JITED_INSNS = 0,
+	PERF_BPIL_XLATED_INSNS,
+	PERF_BPIL_MAP_IDS,
+	PERF_BPIL_JITED_KSYMS,
+	PERF_BPIL_JITED_FUNC_LENS,
+	PERF_BPIL_FUNC_INFO,
+	PERF_BPIL_LINE_INFO,
+	PERF_BPIL_JITED_LINE_INFO,
+	PERF_BPIL_PROG_TAGS,
+	PERF_BPIL_LAST_ARRAY,
+};
+
+struct perf_bpil {
+	/* size of struct bpf_prog_info, when the tool is compiled */
+	__u32			info_len;
+	/* total bytes allocated for data, round up to 8 bytes */
+	__u32			data_len;
+	/* which arrays are included in data */
+	__u64			arrays;
+	struct bpf_prog_info	info;
+	__u8			data[];
+};
+
+struct perf_bpil *
+get_bpf_prog_info_linear(int fd, __u64 arrays);
+
+void
+bpil_addr_to_offs(struct perf_bpil *info_linear);
+
+void
+bpil_offs_to_addr(struct perf_bpil *info_linear);
+
+#endif /* HAVE_LIBBPF_SUPPORT */
+#endif /* __PERF_BPF_UTILS_H */