Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.kernel > #1264690 > unrolled thread
| Started by | Martin KaFai Lau <kafai@fb.com> |
|---|---|
| First post | 2015-11-07 02:40 +0100 |
| Last post | 2015-11-13 08:00 +0100 |
| Articles | 2 — 2 participants |
Back to article view | Back to linux.kernel
[RFC] kprobes: Use percpu counter to collect nhit statistics Martin KaFai Lau <kafai@fb.com> - 2015-11-07 02:40 +0100
RE: [RFC] kprobes: Use percpu counter to collect nhit statistics 平松雅巳 / HIRAMATU，MASAMI <masami.hiramatsu.pt@hitachi.com> - 2015-11-13 08:00 +0100
| From | Martin KaFai Lau <kafai@fb.com> |
|---|---|
| Date | 2015-11-07 02:40 +0100 |
| Subject | [RFC] kprobes: Use percpu counter to collect nhit statistics |
| Message-ID | <qs0E1-3XU-1@gated-at.bofh.it> |
When doing ebpf+kprobe on some hot TCP functions (e.g.
tcp_rcv_established), the kprobe_dispatcher shows up in 'perf report'.
In kprobe_dispatcher(), there is a lot of cache bouncing
in 'tk->nhit++'. 'tk->nhit' and 'tk->tp.flags' also share
the same cacheline.
perf report (cycles:pp):
8.30% ipv4_dst_check
4.74% copy_user_enhanced_fast_string
3.93% dst_release
2.80% tcp_v4_rcv
2.31% queued_spin_lock_slowpath
2.30% _raw_spin_lock
1.88% mlx4_en_process_rx_cq
1.84% eth_get_headlen
1.81% ip_rcv_finish
~~~~
1.71% kprobe_dispatcher
~~~~
1.55% mlx4_en_xmit
1.09% __probe_kernel_read
perf report after patch:
9.15% ipv4_dst_check
5.00% copy_user_enhanced_fast_string
4.12% dst_release
2.96% tcp_v4_rcv
2.50% _raw_spin_lock
2.39% queued_spin_lock_slowpath
2.11% eth_get_headlen
2.03% mlx4_en_process_rx_cq
1.69% mlx4_en_xmit
1.19% ip_rcv_finish
1.12% __probe_kernel_read
1.02% ehci_hcd_cleanup
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
---
kernel/trace/trace_kprobe.c | 19 +++++++++++++++----
1 file changed, 15 insertions(+), 4 deletions(-)
diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index c995644..91a441b 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -30,7 +30,7 @@
struct trace_kprobe {
struct list_head list;
struct kretprobe rp; /* Use rp.kp for kprobe use */
- unsigned long nhit;
+ unsigned long * __percpu nhit;
const char *symbol; /* symbol name */
struct trace_probe tp;
};
@@ -274,6 +274,10 @@ static struct trace_kprobe *alloc_trace_kprobe(const char *group,
if (!tk)
return ERR_PTR(ret);
+ tk->nhit = alloc_percpu(unsigned long);
+ if (!tk->nhit)
+ goto error;
+
if (symbol) {
tk->symbol = kstrdup(symbol, GFP_KERNEL);
if (!tk->symbol)
@@ -313,6 +317,7 @@ static struct trace_kprobe *alloc_trace_kprobe(const char *group,
error:
kfree(tk->tp.call.name);
kfree(tk->symbol);
+ free_percpu(tk->nhit);
kfree(tk);
return ERR_PTR(ret);
}
@@ -327,6 +332,7 @@ static void free_trace_kprobe(struct trace_kprobe *tk)
kfree(tk->tp.call.class->system);
kfree(tk->tp.call.name);
kfree(tk->symbol);
+ free_percpu(tk->nhit);
kfree(tk);
}
@@ -874,9 +880,14 @@ static const struct file_operations kprobe_events_ops = {
static int probes_profile_seq_show(struct seq_file *m, void *v)
{
struct trace_kprobe *tk = v;
+ unsigned long nhit = 0;
+ int cpu;
+
+ for_each_possible_cpu(cpu)
+ nhit += *per_cpu_ptr(tk->nhit, cpu);
seq_printf(m, " %-44s %15lu %15lu\n",
- trace_event_name(&tk->tp.call), tk->nhit,
+ trace_event_name(&tk->tp.call), nhit,
tk->rp.kp.nmissed);
return 0;
@@ -1225,7 +1236,7 @@ static int kprobe_dispatcher(struct kprobe *kp, struct pt_regs *regs)
{
struct trace_kprobe *tk = container_of(kp, struct trace_kprobe, rp.kp);
- tk->nhit++;
+ raw_cpu_inc(*tk->nhit);
if (tk->tp.flags & TP_FLAG_TRACE)
kprobe_trace_func(tk, regs);
@@ -1242,7 +1253,7 @@ kretprobe_dispatcher(struct kretprobe_instance *ri, struct pt_regs *regs)
{
struct trace_kprobe *tk = container_of(ri->rp, struct trace_kprobe, rp);
- tk->nhit++;
+ raw_cpu_inc(*tk->nhit);
if (tk->tp.flags & TP_FLAG_TRACE)
kretprobe_trace_func(tk, ri, regs);
--
2.5.1
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[toc] | [next] | [standalone]
| From | 平松雅巳 / HIRAMATU，MASAMI <masami.hiramatsu.pt@hitachi.com> |
|---|---|
| Date | 2015-11-13 08:00 +0100 |
| Message-ID | <quguZ-2Yu-5@gated-at.bofh.it> |
| In reply to | #1264690 |
RnJvbTogTWFydGluIEthRmFpIExhdSBbbWFpbHRvOmthZmFpQGZiLmNvbV0NCj4NCj5XaGVuIGRv aW5nIGVicGYra3Byb2JlIG9uIHNvbWUgaG90IFRDUCBmdW5jdGlvbnMgKGUuZy4NCj50Y3BfcmN2 X2VzdGFibGlzaGVkKSwgdGhlIGtwcm9iZV9kaXNwYXRjaGVyIHNob3dzIHVwIGluICdwZXJmIHJl cG9ydCcuDQo+DQo+SW4ga3Byb2JlX2Rpc3BhdGNoZXIoKSwgdGhlcmUgaXMgYSBsb3Qgb2YgY2Fj aGUgYm91bmNpbmcNCj5pbiAndGstPm5oaXQrKycuICAndGstPm5oaXQnIGFuZCAndGstPnRwLmZs YWdzJyBhbHNvIHNoYXJlDQo+dGhlIHNhbWUgY2FjaGVsaW5lLg0KPg0KPnBlcmYgcmVwb3J0IChj eWNsZXM6cHApOg0KPjguMzAlICBpcHY0X2RzdF9jaGVjaw0KPjQuNzQlICBjb3B5X3VzZXJfZW5o YW5jZWRfZmFzdF9zdHJpbmcNCj4zLjkzJSAgZHN0X3JlbGVhc2UNCj4yLjgwJSAgdGNwX3Y0X3Jj dg0KPjIuMzElICBxdWV1ZWRfc3Bpbl9sb2NrX3Nsb3dwYXRoDQo+Mi4zMCUgIF9yYXdfc3Bpbl9s b2NrDQo+MS44OCUgIG1seDRfZW5fcHJvY2Vzc19yeF9jcQ0KPjEuODQlICBldGhfZ2V0X2hlYWRs ZW4NCj4xLjgxJSAgaXBfcmN2X2ZpbmlzaA0KPn5+fn4NCj4xLjcxJSAga3Byb2JlX2Rpc3BhdGNo ZXINCj5+fn5+DQo+MS41NSUgIG1seDRfZW5feG1pdA0KPjEuMDklICBfX3Byb2JlX2tlcm5lbF9y ZWFkDQo+DQo+cGVyZiByZXBvcnQgYWZ0ZXIgcGF0Y2g6DQo+OS4xNSUgIGlwdjRfZHN0X2NoZWNr DQo+NS4wMCUgIGNvcHlfdXNlcl9lbmhhbmNlZF9mYXN0X3N0cmluZw0KPjQuMTIlICBkc3RfcmVs ZWFzZQ0KPjIuOTYlICB0Y3BfdjRfcmN2DQo+Mi41MCUgIF9yYXdfc3Bpbl9sb2NrDQo+Mi4zOSUg IHF1ZXVlZF9zcGluX2xvY2tfc2xvd3BhdGgNCj4yLjExJSAgZXRoX2dldF9oZWFkbGVuDQo+Mi4w MyUgIG1seDRfZW5fcHJvY2Vzc19yeF9jcQ0KPjEuNjklICBtbHg0X2VuX3htaXQNCj4xLjE5JSAg aXBfcmN2X2ZpbmlzaA0KPjEuMTIlICBfX3Byb2JlX2tlcm5lbF9yZWFkDQo+MS4wMiUgIGVoY2lf aGNkX2NsZWFudXANCg0KR29vZCBjYXRjaCEgTG9va3MgZ29vZCB0byBtZSENCg0KQWNrZWQtYnk6 IE1hc2FtaSBIaXJhbWF0c3UgPG1hc2FtaS5oaXJhbWF0c3UucHRAaGl0YWNoaS5jb20+DQoNClRo YW5rcyEhDQoNCj4NCj5TaWduZWQtb2ZmLWJ5OiBNYXJ0aW4gS2FGYWkgTGF1IDxrYWZhaUBmYi5j b20+DQo+LS0tDQo+IGtlcm5lbC90cmFjZS90cmFjZV9rcHJvYmUuYyB8IDE5ICsrKysrKysrKysr KysrKy0tLS0NCj4gMSBmaWxlIGNoYW5nZWQsIDE1IGluc2VydGlvbnMoKyksIDQgZGVsZXRpb25z KC0pDQo+DQo+ZGlmZiAtLWdpdCBhL2tlcm5lbC90cmFjZS90cmFjZV9rcHJvYmUuYyBiL2tlcm5l bC90cmFjZS90cmFjZV9rcHJvYmUuYw0KPmluZGV4IGM5OTU2NDQuLjkxYTQ0MWIgMTAwNjQ0DQo+ LS0tIGEva2VybmVsL3RyYWNlL3RyYWNlX2twcm9iZS5jDQo+KysrIGIva2VybmVsL3RyYWNlL3Ry YWNlX2twcm9iZS5jDQo+QEAgLTMwLDcgKzMwLDcgQEANCj4gc3RydWN0IHRyYWNlX2twcm9iZSB7 DQo+IAlzdHJ1Y3QgbGlzdF9oZWFkCWxpc3Q7DQo+IAlzdHJ1Y3Qga3JldHByb2JlCXJwOwkvKiBV c2UgcnAua3AgZm9yIGtwcm9iZSB1c2UgKi8NCj4tCXVuc2lnbmVkIGxvbmcgCQluaGl0Ow0KPisJ dW5zaWduZWQgbG9uZyAqIF9fcGVyY3B1IG5oaXQ7DQo+IAljb25zdCBjaGFyCQkqc3ltYm9sOwkv KiBzeW1ib2wgbmFtZSAqLw0KPiAJc3RydWN0IHRyYWNlX3Byb2JlCXRwOw0KPiB9Ow0KPkBAIC0y NzQsNiArMjc0LDEwIEBAIHN0YXRpYyBzdHJ1Y3QgdHJhY2Vfa3Byb2JlICphbGxvY190cmFjZV9r cHJvYmUoY29uc3QgY2hhciAqZ3JvdXAsDQo+IAlpZiAoIXRrKQ0KPiAJCXJldHVybiBFUlJfUFRS KHJldCk7DQo+DQo+Kwl0ay0+bmhpdCA9IGFsbG9jX3BlcmNwdSh1bnNpZ25lZCBsb25nKTsNCj4r CWlmICghdGstPm5oaXQpDQo+KwkJZ290byBlcnJvcjsNCj4rDQo+IAlpZiAoc3ltYm9sKSB7DQo+ IAkJdGstPnN5bWJvbCA9IGtzdHJkdXAoc3ltYm9sLCBHRlBfS0VSTkVMKTsNCj4gCQlpZiAoIXRr LT5zeW1ib2wpDQo+QEAgLTMxMyw2ICszMTcsNyBAQCBzdGF0aWMgc3RydWN0IHRyYWNlX2twcm9i ZSAqYWxsb2NfdHJhY2Vfa3Byb2JlKGNvbnN0IGNoYXIgKmdyb3VwLA0KPiBlcnJvcjoNCj4gCWtm cmVlKHRrLT50cC5jYWxsLm5hbWUpOw0KPiAJa2ZyZWUodGstPnN5bWJvbCk7DQo+KwlmcmVlX3Bl cmNwdSh0ay0+bmhpdCk7DQo+IAlrZnJlZSh0ayk7DQo+IAlyZXR1cm4gRVJSX1BUUihyZXQpOw0K PiB9DQo+QEAgLTMyNyw2ICszMzIsNyBAQCBzdGF0aWMgdm9pZCBmcmVlX3RyYWNlX2twcm9iZShz dHJ1Y3QgdHJhY2Vfa3Byb2JlICp0aykNCj4gCWtmcmVlKHRrLT50cC5jYWxsLmNsYXNzLT5zeXN0 ZW0pOw0KPiAJa2ZyZWUodGstPnRwLmNhbGwubmFtZSk7DQo+IAlrZnJlZSh0ay0+c3ltYm9sKTsN Cj4rCWZyZWVfcGVyY3B1KHRrLT5uaGl0KTsNCj4gCWtmcmVlKHRrKTsNCj4gfQ0KPg0KPkBAIC04 NzQsOSArODgwLDE0IEBAIHN0YXRpYyBjb25zdCBzdHJ1Y3QgZmlsZV9vcGVyYXRpb25zIGtwcm9i ZV9ldmVudHNfb3BzID0gew0KPiBzdGF0aWMgaW50IHByb2Jlc19wcm9maWxlX3NlcV9zaG93KHN0 cnVjdCBzZXFfZmlsZSAqbSwgdm9pZCAqdikNCj4gew0KPiAJc3RydWN0IHRyYWNlX2twcm9iZSAq dGsgPSB2Ow0KPisJdW5zaWduZWQgbG9uZyBuaGl0ID0gMDsNCj4rCWludCBjcHU7DQo+Kw0KPisJ Zm9yX2VhY2hfcG9zc2libGVfY3B1KGNwdSkNCj4rCQluaGl0ICs9ICpwZXJfY3B1X3B0cih0ay0+ bmhpdCwgY3B1KTsNCj4NCj4gCXNlcV9wcmludGYobSwgIiAgJS00NHMgJTE1bHUgJTE1bHVcbiIs DQo+LQkJICAgdHJhY2VfZXZlbnRfbmFtZSgmdGstPnRwLmNhbGwpLCB0ay0+bmhpdCwNCj4rCQkg ICB0cmFjZV9ldmVudF9uYW1lKCZ0ay0+dHAuY2FsbCksIG5oaXQsDQo+IAkJICAgdGstPnJwLmtw Lm5taXNzZWQpOw0KPg0KPiAJcmV0dXJuIDA7DQo+QEAgLTEyMjUsNyArMTIzNiw3IEBAIHN0YXRp YyBpbnQga3Byb2JlX2Rpc3BhdGNoZXIoc3RydWN0IGtwcm9iZSAqa3AsIHN0cnVjdCBwdF9yZWdz ICpyZWdzKQ0KPiB7DQo+IAlzdHJ1Y3QgdHJhY2Vfa3Byb2JlICp0ayA9IGNvbnRhaW5lcl9vZihr cCwgc3RydWN0IHRyYWNlX2twcm9iZSwgcnAua3ApOw0KPg0KPi0JdGstPm5oaXQrKzsNCj4rCXJh d19jcHVfaW5jKCp0ay0+bmhpdCk7DQo+DQo+IAlpZiAodGstPnRwLmZsYWdzICYgVFBfRkxBR19U UkFDRSkNCj4gCQlrcHJvYmVfdHJhY2VfZnVuYyh0aywgcmVncyk7DQo+QEAgLTEyNDIsNyArMTI1 Myw3IEBAIGtyZXRwcm9iZV9kaXNwYXRjaGVyKHN0cnVjdCBrcmV0cHJvYmVfaW5zdGFuY2UgKnJp LCBzdHJ1Y3QgcHRfcmVncyAqcmVncykNCj4gew0KPiAJc3RydWN0IHRyYWNlX2twcm9iZSAqdGsg PSBjb250YWluZXJfb2YocmktPnJwLCBzdHJ1Y3QgdHJhY2Vfa3Byb2JlLCBycCk7DQo+DQo+LQl0 ay0+bmhpdCsrOw0KPisJcmF3X2NwdV9pbmMoKnRrLT5uaGl0KTsNCj4NCj4gCWlmICh0ay0+dHAu ZmxhZ3MgJiBUUF9GTEFHX1RSQUNFKQ0KPiAJCWtyZXRwcm9iZV90cmFjZV9mdW5jKHRrLCByaSwg cmVncyk7DQo+LS0NCj4yLjUuMQ0KDQo= -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[toc] | [prev] | [standalone]
Back to top | Article view | linux.kernel
csiph-web