This is an automated email from the ASF dual-hosted git repository.
wusheng pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/skywalking-rover.git
The following commit(s) were added to refs/heads/main by this push:
new 97857df Improve the stability of Off CPU Profiling (#103)
97857df is described below
commit 97857dfe8f9a677336d2307c820b51c69c5a7a74
Author: mrproliu <[email protected]>
AuthorDate: Sat Oct 21 09:55:04 2023 +0800
Improve the stability of Off CPU Profiling (#103)
---
CHANGES.md | 1 +
pkg/profiling/task/offcpu/runner.go | 24 ++++++++++++++++--------
pkg/tools/profiling/api.go | 19 ++++++++++++++++---
3 files changed, 33 insertions(+), 11 deletions(-)
diff --git a/CHANGES.md b/CHANGES.md
index 97baef8..e392441 100644
--- a/CHANGES.md
+++ b/CHANGES.md
@@ -8,6 +8,7 @@ Release Notes.
* Enhance compatibility when profiling with SSL.
* Update `LabelValue` obtain pod information function to add default value
parameter.
* Add `HasOwnerName` to judgement pod has owner name.
+* Improve the stability of Off CPU Profiling.
#### Bug Fixes
diff --git a/pkg/profiling/task/offcpu/runner.go
b/pkg/profiling/task/offcpu/runner.go
index 83d25ab..492d996 100644
--- a/pkg/profiling/task/offcpu/runner.go
+++ b/pkg/profiling/task/offcpu/runner.go
@@ -26,6 +26,7 @@ import (
"github.com/hashicorp/go-multierror"
+ "github.com/cilium/ebpf"
"github.com/cilium/ebpf/link"
"github.com/apache/skywalking-rover/pkg/logger"
@@ -65,7 +66,7 @@ type Runner struct {
// runtime
previousStacks map[ProcessStack]StackCounter
bpf *bpfObjects
- kprobe link.Link
+ kprobe *btf.Linker
stopChan chan bool
flushDataNotify context.CancelFunc
}
@@ -117,25 +118,32 @@ func (r *Runner) Run(ctx context.Context, notify
base.ProfilingRunningSuccessNot
}
r.bpf = &objs
- kprobe, err := link.Kprobe(r.findMatchesSymbol(),
objs.DoFinishTaskSwitch, nil)
- if err != nil {
+ symbols := r.findMatchesSymbol()
+ linker := btf.NewLinker()
+ switchers := make(map[string]*ebpf.Program)
+ for _, symbol := range symbols {
+ switchers[symbol] = objs.DoFinishTaskSwitch
+ }
+
+ linker.AddLink(link.Kprobe, switchers)
+ if err := linker.HasError(); err != nil {
return fmt.Errorf("link to finish task swtich failure: %v", err)
}
- r.kprobe = kprobe
+ r.kprobe = linker
notify()
<-r.stopChan
return nil
}
-func (r *Runner) findMatchesSymbol() string {
+func (r *Runner) findMatchesSymbol() []string {
if r.kernelProfiling == nil {
- return defaultKernelSymbol
+ return []string{defaultKernelSymbol}
}
- res, err :=
r.kernelProfiling.FindSymbolByRegex(`finish_task_switch(\.\w+\.\d+)?`)
+ res, err :=
r.kernelProfiling.FindMultipleSymbolByRegex(`finish_task_switch(\.\w+\.\d+)?`)
if err != nil {
log.Warnf("found symbol error: %v", err)
- return defaultKernelSymbol
+ return []string{defaultKernelSymbol}
}
return res
}
diff --git a/pkg/tools/profiling/api.go b/pkg/tools/profiling/api.go
index cf57adb..518da05 100644
--- a/pkg/tools/profiling/api.go
+++ b/pkg/tools/profiling/api.go
@@ -142,18 +142,31 @@ func (i *Info) FindSymbolAddress(name string) uint64 {
}
func (i *Info) FindSymbolByRegex(rep string) (string, error) {
- compile, err := regexp.Compile(rep)
+ vals, err := i.FindMultipleSymbolByRegex(rep)
if err != nil {
return "", err
}
+ return vals[0], nil
+}
+
+func (i *Info) FindMultipleSymbolByRegex(rep string) ([]string, error) {
+ compile, err := regexp.Compile(rep)
+ if err != nil {
+ return nil, err
+ }
+ result := make([]string, 0)
for _, m := range i.Modules {
for _, sym := range m.Symbols {
if compile.MatchString(sym.Name) {
- return sym.Name, nil
+ result = append(result, sym.Name)
}
}
}
- return "", fmt.Errorf("cannot found any matches symbol: %s", rep)
+
+ if len(result) == 0 {
+ return nil, fmt.Errorf("cannot found any matches symbol: %s",
rep)
+ }
+ return result, nil
}
func (m *Module) contains(addr uint64) (uint64, bool) {
