Re: [PATCH v3 bpf-next 1/2] bpf: separate bpf_get_[stack|stackid] for perf events BPF

[Song Liu <songliubraving@xxxxxx>]

> On Jul 21, 2020, at 3:43 PM, Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote:
> 
> On Tue, Jul 21, 2020 at 3:40 PM Song Liu <songliubraving@xxxxxx> wrote:
>> 
>> We only need to block precise_ip >= 2. precise_ip == 1 is OK.
> 
> Are you sure?
> intel_pmu_hw_config() has:
> if (event->attr.precise_ip) {
>    if (event->attr.sample_type & PERF_SAMPLE_CALLCHAIN)
>            event->attr.sample_type |= __PERF_SAMPLE_CALLCHAIN_EARLY;
> }

The bit that breaks the unwinder was in setup_pebs_fixed_sample_data():

                if (x86_pmu.intel_cap.pebs_format >= 2) {
                        set_linear_ip(regs, pebs->real_ip);
                        regs->flags |= PERF_EFLAGS_EXACT;
                } 

"real_ip" causes the issue. 

But on a second thought, it is probably better also blocks precise_ip == 1, 
to match the logic in intel_pmu_hw_config(). 

Thanks,
Song