1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
|
package internal
import "C"
import (
"context"
"fmt"
"os"
"os/signal"
"runtime/pprof"
"syscall"
"time"
"ior/internal/flags"
"ior/internal/flamegraph"
"ior/internal/tracepoints"
bpf "github.com/aquasecurity/libbpfgo"
)
// TODO: Generally, write unit tests
// TODO: Integration tests, write C or Cgo code to simulate I/O?
func attachTracepoints(bpfModule *bpf.Module) error {
for _, name := range tracepoints.List {
if !flags.Get().ShouldIAttachTracepoint(name) {
continue
}
fmt.Println("Attaching tracepoint", name)
prog, err := bpfModule.GetProgram(fmt.Sprintf("handle_%s", name))
if err != nil {
return fmt.Errorf("Failed to get BPF program handle_%s: %v", name, err)
}
fmt.Println("Attached prog handle_", name)
if _, err = prog.AttachTracepoint("syscalls", name); err != nil {
// OK, older Kernel versions may not have this tracepoint!
fmt.Printf("Failed to attach to %s tracepoint: %v, kernel version may be too old, skipping", name, err)
continue
}
fmt.Println("Attached tracepoint ", name)
}
return nil
}
func Run() error {
flags.PrintVersion()
iorFile := flags.Get().IorDataFile
var noTraceRun bool
if iorFile != "" {
noTraceRun = true
collapsed := flamegraph.NewCollapsed(iorFile, flags.Get().CollapsedFields, flags.Get().CountField)
collapsedFile, err := collapsed.Write(iorFile)
if err != nil {
return err
}
tool, err := flamegraph.NewTool(collapsedFile)
if err != nil {
return err
}
if err := tool.WriteSVG(); err != nil {
return err
}
}
if noTraceRun {
return nil
}
return runTrace()
}
func runTrace() error {
bpfModule, err := bpf.NewModuleFromFile("ior.bpf.o")
if err != nil {
return err
}
defer bpfModule.Close()
if err := flags.Get().ResizeBPFMaps(bpfModule); err != nil {
return err
}
if err := flags.Get().SetBPF(bpfModule); err != nil {
return err
}
if err := bpfModule.BPFLoadObject(); err != nil {
return err
}
if err := attachTracepoints(bpfModule); err != nil {
return err
}
// 4096 channel size, minimises event drops
ch := make(chan []byte, 4096)
rb, err := bpfModule.InitRingBuf("event_map", ch)
if err != nil {
return err
}
rb.Poll(300)
pprofDone := make(chan struct{})
var cpuProfile, memProfile *os.File
if flags.Get().PprofEnable {
if cpuProfile, err = os.Create("ior.cpuprofile"); err != nil {
return err
}
if memProfile, err = os.Create("ior.memprofile"); err != nil {
return err
}
pprof.StartCPUProfile(cpuProfile)
} else {
close(pprofDone)
}
loop := newEventLoop()
duration := time.Duration(flags.Get().Duration) * time.Second
fmt.Println("Probing for", duration)
ctx, cancel := context.WithTimeout(context.Background(), duration)
signalCh := make(chan os.Signal, 1)
signal.Notify(signalCh, os.Interrupt, syscall.SIGTERM)
go func() {
<-signalCh
fmt.Println("Received signal, shutting down...")
cancel()
}()
go func() {
<-ctx.Done()
fmt.Println(loop.stats())
if flags.Get().PprofEnable {
fmt.Println("Stoppig profiling, writing ior.cpuprofile and ior.memprofile")
pprof.StopCPUProfile()
pprof.WriteHeapProfile(memProfile)
close(pprofDone)
}
}()
startTime := time.Now()
loop.run(ctx, ch)
totalDuration := time.Since(startTime)
<-pprofDone
fmt.Println("Good bye... (unloading BPF tracepoints will take a few seconds...) after", totalDuration)
return nil
}
|
