Migrated killsnoop to bpf_perf_event

man/man8/killsnoop.8

@@ -11,6 +11,10 @@ is sending signals.
 This works by tracing the kernel sys_kill() function using dynamic tracing, and
 will need updating to match any changes to this function.
 
+This makes use of a Linux 4.5 feature (bpf_perf_event_output());
+for kernels older than 4.5, see the version under tools/old,
+which uses an older mechanism.
+
 Since this uses BPF, only the root user can use this tool.
 .SH REQUIREMENTS
 CONFIG_BPF and bcc.

tools/killsnoop.py

@@ -10,10 +10,12 @@
 # Licensed under the Apache License, Version 2.0 (the "License")
 #
 # 20-Sep-2015   Brendan Gregg   Created this.
+# 19-Feb-2016   Allan McAleavy migrated to BPF_PERF_OUTPUT
 
 from __future__ import print_function
 from bcc import BPF
 import argparse
+import ctypes as ct
 
 # arguments
 examples = """examples:
@@ -38,15 +40,40 @@ debug = 0
 # define BPF program
 bpf_text = """
 #include <uapi/linux/ptrace.h>
+#include <linux/sched.h>
+
+struct val_t {
+   u64 pid;
+   u64 ts;
+   char comm[TASK_COMM_LEN];
+};
+
+struct data_t {
+   u64 pid;
+   u64 tpid;
+   int sig;
+   int ret;
+   u64 ts;
+   u64 delta;
+   char comm[TASK_COMM_LEN];
+};
 
 BPF_HASH(args_pid, u32, int);
 BPF_HASH(args_sig, u32, int);
+BPF_HASH(infotmp, u32, struct val_t);
+BPF_PERF_OUTPUT(events);
 
 int kprobe__sys_kill(struct pt_regs *ctx, int tpid, int sig)
 {
+    struct val_t val = {};
     u32 pid = bpf_get_current_pid_tgid();
 
     FILTER
+    if (bpf_get_current_comm(&val.comm, sizeof(val.comm)) == 0) {
+        val.pid = bpf_get_current_pid_tgid();
+        val.ts = bpf_ktime_get_ns();
+        infotmp.update(&pid, &val);
+    }
     args_pid.update(&pid, &tpid);
     args_sig.update(&pid, &sig);
 
@@ -55,8 +82,11 @@ int kprobe__sys_kill(struct pt_regs *ctx, int tpid, int sig)
 
 int kretprobe__sys_kill(struct pt_regs *ctx)
 {
-    int *tpidp, *sigp, ret = ctx->ax;
+    struct data_t data = {};
+    struct val_t *valp;
+    int *tpidp, *sigp;
     u32 pid = bpf_get_current_pid_tgid();
+    u64 tsp = bpf_ktime_get_ns();
 
     tpidp = args_pid.lookup(&pid);
     sigp = args_sig.lookup(&pid);
@@ -64,7 +94,22 @@ int kretprobe__sys_kill(struct pt_regs *ctx)
         return 0;   // missed entry
     }
 
-    bpf_trace_printk("%d %d %d\\n", *tpidp, *sigp, ret);
+    valp = infotmp.lookup(&pid);
+    if (valp == 0) {
+        // missed entry
+        return 0;
+    }
+
+    bpf_probe_read(&data.comm, sizeof(data.comm), valp->comm);
+    data.pid = pid;
+    data.delta = tsp - valp->ts;
+    data.ts = tsp / 1000;
+    data.tpid = *tpidp;
+    data.ret = ctx->ax;
+    data.sig = *sigp;
+
+    events.perf_submit(ctx, &data, sizeof(data));
+    infotmp.delete(&pid);
     args_pid.delete(&pid);
     args_sig.delete(&pid);
 
@@ -82,25 +127,57 @@ if debug:
 # initialize BPF
 b = BPF(text=bpf_text)
 
+TASK_COMM_LEN = 16    # linux/sched.h
+
+class Data(ct.Structure):
+    _fields_ = [
+        ("pid", ct.c_ulonglong),
+        ("tpid", ct.c_ulonglong),
+        ("sig", ct.c_int),
+        ("ret", ct.c_int),
+        ("ts", ct.c_ulonglong),
+        ("delta", ct.c_ulonglong),
+        ("comm", ct.c_char * TASK_COMM_LEN)
+    ]
+
+start_ts = 0
+prev_ts = 0
+delta = 0
+
 # header
 if args.timestamp:
     print("%-14s" % ("TIME(s)"), end="")
 print("%-6s %-16s %-4s %-6s %s" % ("PID", "COMM", "SIG", "TPID", "RESULT"))
 
-start_ts = 0
+# process event
+def print_event(cpu, data, size):
+    event = ct.cast(data, ct.POINTER(Data)).contents
+    global start_ts
+    global prev_ts
+    global delta
 
-# format output
-while 1:
-    (task, pid, cpu, flags, ts, msg) = b.trace_fields()
-    (tpid_s, sig_s, ret_s) = msg.split(" ")
+    if start_ts == 0:
+        prev_ts = start_ts
+
+    if start_ts == 1:
+        delta = float(delta) + (event.ts - prev_ts)
 
-    ret = int(ret_s)
-    if (args.failed and (ret >= 0)):
-        continue
+    if (args.failed and (event.ret >= 0)):
+        start_ts = 1
+        prev_ts = event.ts
+        return
 
     # print columns
     if args.timestamp:
-        if start_ts == 0:
-            start_ts = ts
-        print("%-14.9f" % (ts - start_ts), end="")
-    print("%-6d %-16s %-4s %-6s %s" % (pid, task, sig_s, tpid_s, ret_s))
+        print("%-14.9f" % (delta / 1000000), end="")
+
+    print("%-6d %-16s %-4d %-6d %d" % (event.pid, event.comm, event.sig,
+        event.tpid, event.ret))
+
+    prev_ts = event.ts
+    start_ts = 1
+
+# loop with callback to print_event
+b["events"].open_perf_buffer(print_event)
+while 1:
+    b.kprobe_poll()

tools/old/killsnoop.py

+#!/usr/bin/python
+# @lint-avoid-python-3-compatibility-imports
+#
+# killsnoop Trace signals issued by the kill() syscall.
+#           For Linux, uses BCC, eBPF. Embedded C.
+#
+# USAGE: killsnoop [-h] [-t] [-x] [-p PID]
+#
+# Copyright (c) 2015 Brendan Gregg.
+# Licensed under the Apache License, Version 2.0 (the "License")
+#
+# 20-Sep-2015   Brendan Gregg   Created this.
+
+from __future__ import print_function
+from bcc import BPF
+import argparse
+
+# arguments
+examples = """examples:
+    ./killsnoop           # trace all kill() signals
+    ./killsnoop -t        # include timestamps
+    ./killsnoop -x        # only show failed kills
+    ./killsnoop -p 181    # only trace PID 181
+"""
+parser = argparse.ArgumentParser(
+    description="Trace signals issued by the kill() syscall",
+    formatter_class=argparse.RawDescriptionHelpFormatter,
+    epilog=examples)
+parser.add_argument("-t", "--timestamp", action="store_true",
+    help="include timestamp on output")
+parser.add_argument("-x", "--failed", action="store_true",
+    help="only show failed opens")
+parser.add_argument("-p", "--pid",
+    help="trace this PID only")
+args = parser.parse_args()
+debug = 0
+
+# define BPF program
+bpf_text = """
+#include <uapi/linux/ptrace.h>
+
+BPF_HASH(args_pid, u32, int);
+BPF_HASH(args_sig, u32, int);
+
+int kprobe__sys_kill(struct pt_regs *ctx, int tpid, int sig)
+{
+    u32 pid = bpf_get_current_pid_tgid();
+
+    FILTER
+    args_pid.update(&pid, &tpid);
+    args_sig.update(&pid, &sig);
+
+    return 0;
+};
+
+int kretprobe__sys_kill(struct pt_regs *ctx)
+{
+    int *tpidp, *sigp, ret = ctx->ax;
+    u32 pid = bpf_get_current_pid_tgid();
+
+    tpidp = args_pid.lookup(&pid);
+    sigp = args_sig.lookup(&pid);
+    if (tpidp == 0 || sigp == 0) {
+        return 0;   // missed entry
+    }
+
+    bpf_trace_printk("%d %d %d\\n", *tpidp, *sigp, ret);
+    args_pid.delete(&pid);
+    args_sig.delete(&pid);
+
+    return 0;
+}
+"""
+if args.pid:
+    bpf_text = bpf_text.replace('FILTER',
+        'if (pid != %s) { return 0; }' % args.pid)
+else:
+    bpf_text = bpf_text.replace('FILTER', '')
+if debug:
+    print(bpf_text)
+
+# initialize BPF
+b = BPF(text=bpf_text)
+
+# header
+if args.timestamp:
+    print("%-14s" % ("TIME(s)"), end="")
+print("%-6s %-16s %-4s %-6s %s" % ("PID", "COMM", "SIG", "TPID", "RESULT"))
+
+start_ts = 0
+
+# format output
+while 1:
+    (task, pid, cpu, flags, ts, msg) = b.trace_fields()
+    (tpid_s, sig_s, ret_s) = msg.split(" ")
+
+    ret = int(ret_s)
+    if (args.failed and (ret >= 0)):
+        continue
+
+    # print columns
+    if args.timestamp:
+        if start_ts == 0:
+            start_ts = ts
+        print("%-14.9f" % (ts - start_ts), end="")
+    print("%-6d %-16s %-4s %-6s %s" % (pid, task, sig_s, tpid_s, ret_s))