Re: [PATCH V2 14/23] perf/x86/intel/lbr: Support Architectural LBR

[Peter Zijlstra]

On Fri, Jun 26, 2020 at 11:20:11AM -0700, kan.liang@xxxxxxxxxxxxxxx wrote:

> +	if (boot_cpu_has(X86_FEATURE_ARCH_LBR))
> +		intel_pmu_arch_lbr_init();

> +static inline bool is_lbr_call_stack_bit_set(u64 config)
> +{
> +	if (x86_pmu.arch_lbr)
> +		return !!(config & ARCH_LBR_CALL_STACK);
> +
> +	return !!(config & LBR_CALL_STACK);
> +}

> +	if (!x86_pmu.arch_lbr && !pmi && cpuc->lbr_sel)
>  		wrmsrl(MSR_LBR_SELECT, lbr_select);

> +	if (!x86_pmu.arch_lbr)
> +		debugctl |= DEBUGCTLMSR_LBR;

> +	if (x86_pmu.arch_lbr)
> +		wrmsrl(MSR_ARCH_LBR_CTL, lbr_select | ARCH_LBR_CTL_LBREN);
>  }

etc...

> +void __init intel_pmu_arch_lbr_init(void)
> +{
> +	unsigned int unused_edx;
> +	u64 lbr_nr;
> +
> +	/* Arch LBR Capabilities */
> +	cpuid(28, &x86_pmu.lbr_eax.full, &x86_pmu.lbr_ebx.full,
> +		  &x86_pmu.lbr_ecx.full, &unused_edx);
> +
> +	lbr_nr = x86_pmu_get_max_arch_lbr_nr();
> +	if (!lbr_nr)
> +		return;
> +
> +	/* Apply the max depth of Arch LBR */
> +	if (wrmsrl_safe(MSR_ARCH_LBR_DEPTH, lbr_nr))
> +		return;
> +
> +	x86_pmu.lbr_nr = lbr_nr;
> +	x86_get_pmu()->task_ctx_size = sizeof(struct x86_perf_task_context_arch_lbr) +
> +				       lbr_nr * sizeof(struct lbr_entry);
> +
> +	x86_pmu.lbr_from = MSR_ARCH_LBR_FROM_0;
> +	x86_pmu.lbr_to = MSR_ARCH_LBR_TO_0;
> +	x86_pmu.lbr_info = MSR_ARCH_LBR_INFO_0;
> +
> +	/* LBR callstack requires both CPL and Branch Filtering support */
> +	if (!x86_pmu.lbr_ebx.split.lbr_cpl ||
> +	    !x86_pmu.lbr_ebx.split.lbr_filter ||
> +	    !x86_pmu.lbr_ebx.split.lbr_call_stack)
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_CALL_STACK_SHIFT] = LBR_NOT_SUPP;
> +
> +	if (!x86_pmu.lbr_ebx.split.lbr_cpl) {
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_USER_SHIFT] = LBR_NOT_SUPP;
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_KERNEL_SHIFT] = LBR_NOT_SUPP;
> +	} else if (!x86_pmu.lbr_ebx.split.lbr_filter) {
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_ANY_SHIFT] = LBR_NOT_SUPP;
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_ANY_RETURN_SHIFT] = LBR_NOT_SUPP;
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_ANY_CALL_SHIFT] = LBR_NOT_SUPP;
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_IND_CALL_SHIFT] = LBR_NOT_SUPP;
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_COND_SHIFT] = LBR_NOT_SUPP;
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_IND_JUMP_SHIFT] = LBR_NOT_SUPP;
> +		arch_lbr_ctl_map[PERF_SAMPLE_BRANCH_CALL_SHIFT] = LBR_NOT_SUPP;
> +	}
> +
> +	x86_pmu.lbr_ctl_mask = ARCH_LBR_CTL_MASK;
> +	x86_pmu.lbr_ctl_map  = arch_lbr_ctl_map;
> +
> +	if (!x86_pmu.lbr_ebx.split.lbr_cpl && !x86_pmu.lbr_ebx.split.lbr_filter)
> +		x86_pmu.lbr_ctl_map = NULL;
> +
> +	x86_pmu.lbr_reset = intel_pmu_arch_lbr_reset;
> +	x86_pmu.lbr_read = intel_pmu_arch_lbr_read;
> +	x86_pmu.lbr_save = intel_pmu_arch_lbr_save;
> +	x86_pmu.lbr_restore = intel_pmu_arch_lbr_restore;
> +
> +	x86_pmu.arch_lbr = true;
> +	pr_cont("Architectural LBR, ");
> +}

How about we make this here clear FEATURE_ARCH_LBR if it fails and then
do away with x86_pmu.arch_lbr and use
static_cpu_has(X86_FEATURE_ARCH_LBR) a lot more?