Merge pull request iovisor#3691 from chenhengqi/add-mdflush

libbpf-tools/.gitignore

@@ -29,6 +29,7 @@
 /llcstat
 /nfsdist
 /nfsslower
+/mdflush
 /mountsnoop
 /numamove
 /offcputime

libbpf-tools/Makefile

@@ -42,6 +42,7 @@ APPS = \
  klockstat \
  ksnoop \
  llcstat \
+ mdflush \
  mountsnoop \
  numamove \
  offcputime \

libbpf-tools/core_fixes.bpf.h

@@ -15,15 +15,34 @@
  */
 struct task_struct___x {
  unsigned int __state;
-};
+} __attribute__((preserve_access_index));
 
-static __s64 get_task_state(void *task)
+/**
+ * commit 309dca309fc3 ("block: store a block_device pointer in struct bio")
+ * adds a new member bi_bdev which is a pointer to struct block_device
+ * see:
+ * https://github.com/torvalds/linux/commit/309dca309fc3
+ */
+struct bio___x {
+ struct block_device *bi_bdev;
+} __attribute__((preserve_access_index));
+
+static __always_inline __s64 get_task_state(void *task)
 {
  struct task_struct___x *t = task;
 
  if (bpf_core_field_exists(t->__state))
- return t->__state;
- return ((struct task_struct *)task)->state;
+ return BPF_CORE_READ(t, __state);
+ return BPF_CORE_READ((struct task_struct *)task, state);
+}
+
+static __always_inline struct gendisk *get_gendisk(void *bio)
+{
+ struct bio___x *b = bio;
+
+ if (bpf_core_field_exists(b->bi_bdev))
+ return BPF_CORE_READ(b, bi_bdev, bd_disk);
+ return BPF_CORE_READ((struct bio *)bio, bi_disk);
 }
 
 #endif /* __CORE_FIXES_BPF_H */

libbpf-tools/mdflush.bpf.c

@@ -0,0 +1,31 @@
+/* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */
+/* Copyright (c) 2021~2022 Hengqi Chen */
+#include <vmlinux.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_core_read.h>
+#include "core_fixes.bpf.h"
+#include "mdflush.h"
+
+struct {
+ __uint(type, BPF_MAP_TYPE_PERF_EVENT_ARRAY);
+ __type(key, __u32);
+ __type(value, __u32);
+} events SEC(".maps");
+
+SEC("fentry/md_flush_request")
+int BPF_PROG(md_flush_request, void *mddev, void *bio)
+{
+ __u64 pid = bpf_get_current_pid_tgid() >> 32;
+ struct event event = {};
+ struct gendisk *gendisk;
+
+ event.pid = pid;
+ gendisk = get_gendisk(bio);
+ BPF_CORE_READ_STR_INTO(event.disk, gendisk, disk_name);
+ bpf_get_current_comm(event.comm, sizeof(event.comm));
+ bpf_perf_event_output(ctx, &events, BPF_F_CURRENT_CPU, &event, sizeof(event));
+ return 0;
+}
+
+char LICENSE[] SEC("license") = "Dual BSD/GPL";

libbpf-tools/mdflush.c

@@ -0,0 +1,152 @@
+/* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */
+
+/*
+ * mdflush Trace md flush events.
+ *
+ * Copyright (c) 2021~2022 Hengqi Chen
+ *
+ * Based on mdflush(8) from BCC by Brendan Gregg.
+ * 08-Nov-2021 Hengqi Chen Created this.
+ */
+#include <argp.h>
+#include <errno.h>
+#include <signal.h>
+#include <string.h>
+#include <time.h>
+
+#include <bpf/libbpf.h>
+#include <bpf/bpf.h>
+#include "mdflush.h"
+#include "mdflush.skel.h"
+#include "trace_helpers.h"
+
+#define PERF_BUFFER_PAGES 16
+#define PERF_POLL_TIMEOUT_MS 100
+#define warn(...) fprintf(stderr, __VA_ARGS__)
+
+static volatile sig_atomic_t exiting = 0;
+static bool verbose = false;
+
+const char *argp_program_version = "mdflush 0.1";
+const char *argp_program_bug_address =
+ "https://github.com/iovisor/bcc/tree/master/libbpf-tools";
+const char argp_program_doc[] =
+"Trace md flush events.\n"
+"\n"
+"USAGE: mdflush\n";
+
+static const struct argp_option opts[] = {
+ { "verbose", 'v', NULL, 0, "Verbose debug output" },
+ { NULL, 'h', NULL, OPTION_HIDDEN, "Show the full help" },
+ {},
+};
+
+static error_t parse_arg(int key, char *arg, struct argp_state *state)
+{
+ switch (key) {
+ case 'h':
+ argp_state_help(state, stderr, ARGP_HELP_STD_HELP);
+ break;
+ case 'v':
+ verbose = true;
+ break;
+ default:
+ return ARGP_ERR_UNKNOWN;
+ }
+ return 0;
+}
+
+static void sig_int(int signo)
+{
+ exiting = 1;
+}
+
+static int libbpf_print_fn(enum libbpf_print_level level, const char *format, va_list args)
+{
+ if (level == LIBBPF_DEBUG && !verbose)
+ return 0;
+ return vfprintf(stderr, format, args);
+}
+
+static void handle_event(void *ctx, int cpu, void *data, __u32 data_sz)
+{
+ struct event *e = data;
+ time_t t;
+ struct tm *tm;
+ char ts[32];
+
+ time(&t);
+ tm = localtime(&t);
+ strftime(ts, sizeof(ts), "%H:%M:%S", tm);
+ printf("%-8s %-7d %-16s %-s\n",
+ ts, e->pid, e->comm, e->disk);
+}
+
+static void handle_lost_events(void *ctx, int cpu, __u64 lost_cnt)
+{
+ warn("lost %llu events on CPU #%d\n", lost_cnt, cpu);
+}
+
+int main(int argc, char **argv)
+{
+ static const struct argp argp = {
+ .options = opts,
+ .parser = parse_arg,
+ .doc = argp_program_doc,
+ };
+ struct perf_buffer *pb = NULL;
+ struct mdflush_bpf *obj;
+ int err;
+
+ err = argp_parse(&argp, argc, argv, 0, NULL, NULL);
+ if (err)
+ return err;
+
+ libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+ libbpf_set_print(libbpf_print_fn);
+
+ obj = mdflush_bpf__open_and_load();
+ if (!obj) {
+ warn("failed to open/load BPF object\n");
+ return 1;
+ }
+
+ err = mdflush_bpf__attach(obj);
+ if (err) {
+ warn("failed to attach BPF programs: %d\n", err);
+ goto cleanup;
+ }
+
+ pb = perf_buffer__new(bpf_map__fd(obj->maps.events), PERF_BUFFER_PAGES,
+ handle_event, handle_lost_events, NULL, NULL);
+ if (!pb) {
+ warn("failed to open perf buffer: %d\n", err);
+ goto cleanup;
+ }
+
+ if (signal(SIGINT, sig_int) == SIG_ERR) {
+ warn("can't set signal handler: %s\n", strerror(errno));
+ err = 1;
+ goto cleanup;
+ }
+
+ printf("Tracing md flush requests... Hit Ctrl-C to end.\n");
+ printf("%-8s %-7s %-16s %-s\n",
+ "TIME", "PID", "COMM", "DEVICE");
+
+ while (!exiting) {
+ err = perf_buffer__poll(pb, PERF_POLL_TIMEOUT_MS);
+ if (err < 0 && err != -EINTR) {
+ warn("error polling perf buffer: %s\n", strerror(-err));
+ goto cleanup;
+ }
+ /* reset err to return 0 if exiting */
+ err = 0;
+ }
+
+cleanup:
+ perf_buffer__free(pb);
+ mdflush_bpf__destroy(obj);
+
+ return err != 0;
+}

libbpf-tools/mdflush.h

@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */
+/* Copyright (c) 2021~2022 Hengqi Chen */
+#ifndef __MDFLUSH_H
+#define __MDFLUSH_H
+
+#define TASK_COMM_LEN 16
+#define DISK_NAME_LEN 32
+
+struct event {
+ __u32 pid;
+ char comm[TASK_COMM_LEN];
+ char disk[DISK_NAME_LEN];
+};
+
+#endif /* __MDFLUSH_H */