lkml.org 
[lkml]   [2014]   [Oct]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH v2 05/12] perf/x86: add cross-HT counter exclusion infrastructure
Date
From: Maria Dimakopoulou <maria.n.dimakopoulou@gmail.com>

This patch adds a new shared_regs style structure to the
per-cpu x86 state (cpuc). It is used to coordinate access
between counters which must be used with exclusion across
HyperThreads on Intel processors. This new struct is not
needed on each PMU, thus is is allocated on demand.

Reviewed-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Maria Dimakopoulou <maria.n.dimakopoulou@gmail.com>
---
arch/x86/kernel/cpu/perf_event.h | 32 +++++++++++++++
arch/x86/kernel/cpu/perf_event_intel.c | 71 +++++++++++++++++++++++++++++++---
2 files changed, 98 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index 667a185..c985270 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -71,6 +71,7 @@ struct event_constraint {
#define PERF_X86_EVENT_COMMITTED 0x8 /* event passed commit_txn */
#define PERF_X86_EVENT_PEBS_LD_HSW 0x10 /* haswell style datala, load */
#define PERF_X86_EVENT_PEBS_NA_HSW 0x20 /* haswell style datala, unknown */
+#define PERF_X86_EVENT_EXCL 0x40 /* HT exclusivity on counter */

struct amd_nb {
int nb_id; /* NorthBridge id */
@@ -121,6 +122,26 @@ struct intel_shared_regs {
unsigned core_id; /* per-core: core id */
};

+enum intel_excl_state_type {
+ INTEL_EXCL_UNUSED = 0, /* counter is unused */
+ INTEL_EXCL_SHARED = 1, /* counter can be used by both threads */
+ INTEL_EXCL_EXCLUSIVE = 2, /* counter can be used by one thread only */
+};
+
+struct intel_excl_states {
+ enum intel_excl_state_type init_state[X86_PMC_IDX_MAX];
+ enum intel_excl_state_type state[X86_PMC_IDX_MAX];
+};
+
+struct intel_excl_cntrs {
+ spinlock_t lock;
+
+ struct intel_excl_states states[2];
+
+ int refcnt; /* per-core: #HT threads */
+ unsigned core_id; /* per-core: core id */
+};
+
#define MAX_LBR_ENTRIES 16

enum {
@@ -183,6 +204,12 @@ struct cpu_hw_events {
* used on Intel NHM/WSM/SNB
*/
struct intel_shared_regs *shared_regs;
+ /*
+ * manage exclusive counter access between hyperthread
+ */
+ struct event_constraint *constraint_list; /* in enable order */
+ struct intel_excl_cntrs *excl_cntrs;
+ int excl_thread_id; /* 0 or 1 */

/*
* AMD specific bits
@@ -206,6 +233,10 @@ struct cpu_hw_events {
#define EVENT_CONSTRAINT(c, n, m) \
__EVENT_CONSTRAINT(c, n, m, HWEIGHT(n), 0, 0)

+#define INTEL_EXCLEVT_CONSTRAINT(c, n) \
+ __EVENT_CONSTRAINT(c, n, ARCH_PERFMON_EVENTSEL_EVENT, HWEIGHT(n),\
+ 0, PERF_X86_EVENT_EXCL)
+
/*
* The overlap flag marks event constraints with overlapping counter
* masks. This is the case if the counter mask of such an event is not
@@ -540,6 +571,7 @@ do { \
*/
#define PMU_FL_NO_HT_SHARING 0x1 /* no hyper-threading resource sharing */
#define PMU_FL_HAS_RSP_1 0x2 /* has 2 equivalent offcore_rsp regs */
+#define PMU_FL_EXCL_CNTRS 0x4 /* has exclusive counter requirements */

#define EVENT_VAR(_id) event_attr_##_id
#define EVENT_PTR(_id) &event_attr_##_id.attr.attr
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 64bbb57..0eece24 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -2131,16 +2131,52 @@ struct intel_shared_regs *allocate_shared_regs(int cpu)
return regs;
}

+struct intel_excl_cntrs *allocate_excl_cntrs(int cpu)
+{
+ struct intel_excl_cntrs *c;
+ int i;
+
+ c = kzalloc_node(sizeof(struct intel_excl_cntrs),
+ GFP_KERNEL, cpu_to_node(cpu));
+ if (c) {
+ spin_lock_init(&c->lock);
+ for (i = 0; i < X86_PMC_IDX_MAX; i++) {
+ c->states[0].state[i] = INTEL_EXCL_UNUSED;
+ c->states[0].init_state[i] = INTEL_EXCL_UNUSED;
+
+ c->states[1].state[i] = INTEL_EXCL_UNUSED;
+ c->states[1].init_state[i] = INTEL_EXCL_UNUSED;
+ }
+ c->core_id = -1;
+ }
+ return c;
+}
+
static int intel_pmu_cpu_prepare(int cpu)
{
struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);

- if (!(x86_pmu.extra_regs || x86_pmu.lbr_sel_map))
- return NOTIFY_OK;
+ if (x86_pmu.extra_regs || x86_pmu.lbr_sel_map) {
+ cpuc->shared_regs = allocate_shared_regs(cpu);
+ if (!cpuc->shared_regs)
+ return NOTIFY_BAD;
+ }

- cpuc->shared_regs = allocate_shared_regs(cpu);
- if (!cpuc->shared_regs)
- return NOTIFY_BAD;
+ if (x86_pmu.flags & PMU_FL_EXCL_CNTRS) {
+ size_t sz = X86_PMC_IDX_MAX * sizeof(struct event_constraint);
+
+ cpuc->constraint_list = kzalloc(sz, GFP_KERNEL);
+ if (!cpuc->constraint_list)
+ return NOTIFY_BAD;
+
+ cpuc->excl_cntrs = allocate_excl_cntrs(cpu);
+ if (!cpuc->excl_cntrs) {
+ kfree(cpuc->constraint_list);
+ kfree(cpuc->shared_regs);
+ return NOTIFY_BAD;
+ }
+ cpuc->excl_thread_id = 0;
+ }

return NOTIFY_OK;
}
@@ -2181,12 +2217,29 @@ static void intel_pmu_cpu_starting(int cpu)

if (x86_pmu.lbr_sel_map)
cpuc->lbr_sel = &cpuc->shared_regs->regs[EXTRA_REG_LBR];
+
+ if (x86_pmu.flags & PMU_FL_EXCL_CNTRS) {
+ for_each_cpu(i, topology_thread_cpumask(cpu)) {
+ struct intel_excl_cntrs *c;
+
+ c = per_cpu(cpu_hw_events, i).excl_cntrs;
+ if (c && c->core_id == core_id) {
+ cpuc->kfree_on_online[1] = cpuc->excl_cntrs;
+ cpuc->excl_cntrs = c;
+ cpuc->excl_thread_id = 1;
+ break;
+ }
+ }
+ cpuc->excl_cntrs->core_id = core_id;
+ cpuc->excl_cntrs->refcnt++;
+ }
}

static void intel_pmu_cpu_dying(int cpu)
{
struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
struct intel_shared_regs *pc;
+ struct intel_excl_cntrs *c;

pc = cpuc->shared_regs;
if (pc) {
@@ -2194,6 +2247,14 @@ static void intel_pmu_cpu_dying(int cpu)
kfree(pc);
cpuc->shared_regs = NULL;
}
+ c = cpuc->excl_cntrs;
+ if (c) {
+ if (c->core_id == -1 || --c->refcnt == 0)
+ kfree(c);
+ cpuc->excl_cntrs = NULL;
+ kfree(cpuc->constraint_list);
+ cpuc->constraint_list = NULL;
+ }

fini_debug_store_on_cpu(cpu);
}
--
1.9.1


\
 
 \ /
  Last update: 2014-10-09 19:01    [W:1.662 / U:0.020 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site