Intel: perf/core: Add a new PERF_EV_CAP_SIBLING event capability

mainline inclusion
from mainline-v5.10-rc1
commit 9f0c4fa1
category: feature
bugzilla: https://gitee.com/openeuler/kernel/issues/I47H3V
CVE: NA

--------------------------------

commit 9f0c4fa1 upstream
Backport summary: backport to kernel 4.19.57 for ICX perf topdown support

Current perf assumes that events in a group are independent. Close an
event doesn't impact the value of the other events in the same group.
If the closed event is a member, after the event closure, other events
are still running like a group. If the closed event is a leader, other
events are running as singleton events.

Add PERF_EV_CAP_SIBLING to allow events to indicate they require being
part of a group, and when the leader dies they cannot exist
independently.
Suggested-by: NPeter Zijlstra <peterz@infradead.org>
Signed-off-by: NKan Liang <kan.liang@linux.intel.com>
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20200723171117.9918-8-kan.liang@linux.intel.comSigned-off-by: NYunying Sun <yunying.sun@intel.com>
Signed-off-by: NJackie Liu <liuyun01@kylinos.cn>
Signed-off-by: NZheng Zengkai <zhengzengkai@huawei.com>
Reviewed-by: NWei Li <liwei391@huawei.com>
Reviewed-by: NXie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: NYang Yingliang <yangyingliang@huawei.com>

include/linux/perf_event.h

@@ -526,9 +526,13 @@ typedef void (*perf_overflow_handler_t)(struct perf_event *,
  * PERF_EV_CAP_SOFTWARE: Is a software event.
  * PERF_EV_CAP_READ_ACTIVE_PKG: A CPU event (or cgroup event) that can be read
  * from any CPU in the package where it is active.
+ * PERF_EV_CAP_SIBLING: An event with this flag must be a group sibling and
+ * cannot be a group leader. If an event with this flag is detached from the
+ * group it is scheduled out and moved into an unrecoverable ERROR state.
  */
 #define PERF_EV_CAP_SOFTWARE		BIT(0)
 #define PERF_EV_CAP_READ_ACTIVE_PKG	BIT(1)
+#define PERF_EV_CAP_SIBLING		BIT(2)
 
 #define SWEVENT_HLIST_BITS		8
 #define SWEVENT_HLIST_SIZE		(1 << SWEVENT_HLIST_BITS)

kernel/events/core.c

@@ -1892,8 +1892,29 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
 	ctx->generation++;
 }
 
+static void put_event(struct perf_event *event);
+static void event_sched_out(struct perf_event *event,
+			    struct perf_cpu_context *cpuctx,
+			    struct perf_event_context *ctx);
+
+/*
+ * Events that have PERF_EV_CAP_SIBLING require being part of a group and
+ * cannot exist on their own, schedule them out and move them into the ERROR
+ * state. Also see _perf_event_enable(), it will not be able to recover
+ * this ERROR state.
+ */
+static inline void perf_remove_sibling_event(struct perf_event *event)
+{
+       struct perf_event_context *ctx = event->ctx;
+       struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
+
+       event_sched_out(event, cpuctx, ctx);
+       perf_event_set_state(event, PERF_EVENT_STATE_ERROR);
+}
+
 static void perf_group_detach(struct perf_event *event)
 {
+	struct perf_event *leader = event->group_leader;
 	struct perf_event *sibling, *tmp;
 	struct perf_event_context *ctx = event->ctx;
 
@@ -1910,7 +1931,7 @@ static void perf_group_detach(struct perf_event *event)
 	/*
 	 * If this is a sibling, remove it from its group.
 	 */
-	if (event->group_leader != event) {
+	if (leader != event) {
 		list_del_init(&event->sibling_list);
 		event->group_leader->nr_siblings--;
 		goto out;
@@ -1923,6 +1944,9 @@ static void perf_group_detach(struct perf_event *event)
 	 */
 	list_for_each_entry_safe(sibling, tmp, &event->sibling_list, sibling_list) {
 
+		if (sibling->event_caps & PERF_EV_CAP_SIBLING)
+			perf_remove_sibling_event(sibling);
+
 		sibling->group_leader = sibling;
 		list_del_init(&sibling->sibling_list);
 
@@ -1944,10 +1968,10 @@ static void perf_group_detach(struct perf_event *event)
 	}
 
 out:
-	perf_event__header_size(event->group_leader);
-
-	for_each_sibling_event(tmp, event->group_leader)
+	for_each_sibling_event(tmp, leader)
 		perf_event__header_size(tmp);
+
+	perf_event__header_size(leader);
 }
 
 static bool is_orphaned_event(struct perf_event *event)
@@ -2727,6 +2751,7 @@ static void _perf_event_enable(struct perf_event *event)
 	raw_spin_lock_irq(&ctx->lock);
 	if (event->state >= PERF_EVENT_STATE_INACTIVE ||
 	    event->state <  PERF_EVENT_STATE_ERROR) {
+out:
 		raw_spin_unlock_irq(&ctx->lock);
 		return;
 	}
@@ -2738,8 +2763,16 @@ static void _perf_event_enable(struct perf_event *event)
 	 * has gone back into error state, as distinct from the task having
 	 * been scheduled away before the cross-call arrived.
 	 */
-	if (event->state == PERF_EVENT_STATE_ERROR)
+	if (event->state == PERF_EVENT_STATE_ERROR) {
+		/*
+		 * Detached SIBLING events cannot leave ERROR state.
+		 */
+		if (event->event_caps & PERF_EV_CAP_SIBLING &&
+		    event->group_leader == event)
+			goto out;
+
 		event->state = PERF_EVENT_STATE_OFF;
+	}
 	raw_spin_unlock_irq(&ctx->lock);
 
 	event_function_call(event, __perf_event_enable, NULL);