Adding upstream version 6.7.7.upstream/6.7.7

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

3 files changed, 467 insertions, 19 deletions

diff --git a/tools/perf/util/bpf_skel/kwork_top.bpf.c b/tools/perf/util/bpf_skel/kwork_top.bpf.c
new file mode 100644
index 0000000000..84c15ccbab
--- /dev/null
+++ b/tools/perf/util/bpf_skel/kwork_top.bpf.c
@@ -0,0 +1,338 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+// Copyright (c) 2022, Huawei
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_core_read.h>
+
+/*
+ * This should be in sync with "util/kwork.h"
+ */
+enum kwork_class_type {
+	KWORK_CLASS_IRQ,
+	KWORK_CLASS_SOFTIRQ,
+	KWORK_CLASS_WORKQUEUE,
+	KWORK_CLASS_SCHED,
+	KWORK_CLASS_MAX,
+};
+
+#define MAX_ENTRIES     102400
+#define MAX_NR_CPUS     2048
+#define PF_KTHREAD      0x00200000
+#define MAX_COMMAND_LEN 16
+
+struct time_data {
+	__u64 timestamp;
+};
+
+struct work_data {
+	__u64 runtime;
+};
+
+struct task_data {
+	__u32 tgid;
+	__u32 is_kthread;
+	char comm[MAX_COMMAND_LEN];
+};
+
+struct work_key {
+	__u32 type;
+	__u32 pid;
+	__u64 task_p;
+};
+
+struct task_key {
+	__u32 pid;
+	__u32 cpu;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_TASK_STORAGE);
+	__uint(map_flags, BPF_F_NO_PREALLOC);
+	__type(key, int);
+	__type(value, struct time_data);
+} kwork_top_task_time SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_HASH);
+	__uint(key_size, sizeof(struct work_key));
+	__uint(value_size, sizeof(struct time_data));
+	__uint(max_entries, MAX_ENTRIES);
+} kwork_top_irq_time SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(struct task_key));
+	__uint(value_size, sizeof(struct task_data));
+	__uint(max_entries, MAX_ENTRIES);
+} kwork_top_tasks SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_HASH);
+	__uint(key_size, sizeof(struct work_key));
+	__uint(value_size, sizeof(struct work_data));
+	__uint(max_entries, MAX_ENTRIES);
+} kwork_top_works SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(u32));
+	__uint(value_size, sizeof(u8));
+	__uint(max_entries, MAX_NR_CPUS);
+} kwork_top_cpu_filter SEC(".maps");
+
+int enabled = 0;
+
+int has_cpu_filter = 0;
+
+__u64 from_timestamp = 0;
+__u64 to_timestamp = 0;
+
+static __always_inline int cpu_is_filtered(__u32 cpu)
+{
+	__u8 *cpu_val;
+
+	if (has_cpu_filter) {
+		cpu_val = bpf_map_lookup_elem(&kwork_top_cpu_filter, &cpu);
+		if (!cpu_val)
+			return 1;
+	}
+
+	return 0;
+}
+
+static __always_inline void update_task_info(struct task_struct *task, __u32 cpu)
+{
+	struct task_key key = {
+		.pid = task->pid,
+		.cpu = cpu,
+	};
+
+	if (!bpf_map_lookup_elem(&kwork_top_tasks, &key)) {
+		struct task_data data = {
+			.tgid = task->tgid,
+			.is_kthread = task->flags & PF_KTHREAD ? 1 : 0,
+		};
+		BPF_CORE_READ_STR_INTO(&data.comm, task, comm);
+
+		bpf_map_update_elem(&kwork_top_tasks, &key, &data, BPF_ANY);
+	}
+}
+
+static __always_inline void update_work(struct work_key *key, __u64 delta)
+{
+	struct work_data *data;
+
+	data = bpf_map_lookup_elem(&kwork_top_works, key);
+	if (data) {
+		data->runtime += delta;
+	} else {
+		struct work_data new_data = {
+			.runtime = delta,
+		};
+
+		bpf_map_update_elem(&kwork_top_works, key, &new_data, BPF_ANY);
+	}
+}
+
+static void on_sched_out(struct task_struct *task, __u64 ts, __u32 cpu)
+{
+	__u64 delta;
+	struct time_data *pelem;
+
+	pelem = bpf_task_storage_get(&kwork_top_task_time, task, NULL, 0);
+	if (pelem)
+		delta = ts - pelem->timestamp;
+	else
+		delta = ts - from_timestamp;
+
+	struct work_key key = {
+		.type = KWORK_CLASS_SCHED,
+		.pid = task->pid,
+		.task_p = (__u64)task,
+	};
+
+	update_work(&key, delta);
+	update_task_info(task, cpu);
+}
+
+static void on_sched_in(struct task_struct *task, __u64 ts)
+{
+	struct time_data *pelem;
+
+	pelem = bpf_task_storage_get(&kwork_top_task_time, task, NULL,
+				     BPF_LOCAL_STORAGE_GET_F_CREATE);
+	if (pelem)
+		pelem->timestamp = ts;
+}
+
+SEC("tp_btf/sched_switch")
+int on_switch(u64 *ctx)
+{
+	struct task_struct *prev, *next;
+
+	prev = (struct task_struct *)ctx[1];
+	next = (struct task_struct *)ctx[2];
+
+	if (!enabled)
+		return 0;
+
+	__u32 cpu = bpf_get_smp_processor_id();
+
+	if (cpu_is_filtered(cpu))
+		return 0;
+
+	__u64 ts = bpf_ktime_get_ns();
+
+	on_sched_out(prev, ts, cpu);
+	on_sched_in(next, ts);
+
+	return 0;
+}
+
+SEC("tp_btf/irq_handler_entry")
+int on_irq_handler_entry(u64 *cxt)
+{
+	struct task_struct *task;
+
+	if (!enabled)
+		return 0;
+
+	__u32 cpu = bpf_get_smp_processor_id();
+
+	if (cpu_is_filtered(cpu))
+		return 0;
+
+	__u64 ts = bpf_ktime_get_ns();
+
+	task = (struct task_struct *)bpf_get_current_task();
+	if (!task)
+		return 0;
+
+	struct work_key key = {
+		.type = KWORK_CLASS_IRQ,
+		.pid = BPF_CORE_READ(task, pid),
+		.task_p = (__u64)task,
+	};
+
+	struct time_data data = {
+		.timestamp = ts,
+	};
+
+	bpf_map_update_elem(&kwork_top_irq_time, &key, &data, BPF_ANY);
+
+	return 0;
+}
+
+SEC("tp_btf/irq_handler_exit")
+int on_irq_handler_exit(u64 *cxt)
+{
+	__u64 delta;
+	struct task_struct *task;
+	struct time_data *pelem;
+
+	if (!enabled)
+		return 0;
+
+	__u32 cpu = bpf_get_smp_processor_id();
+
+	if (cpu_is_filtered(cpu))
+		return 0;
+
+	__u64 ts = bpf_ktime_get_ns();
+
+	task = (struct task_struct *)bpf_get_current_task();
+	if (!task)
+		return 0;
+
+	struct work_key key = {
+		.type = KWORK_CLASS_IRQ,
+		.pid = BPF_CORE_READ(task, pid),
+		.task_p = (__u64)task,
+	};
+
+	pelem = bpf_map_lookup_elem(&kwork_top_irq_time, &key);
+	if (pelem && pelem->timestamp != 0)
+		delta = ts - pelem->timestamp;
+	else
+		delta = ts - from_timestamp;
+
+	update_work(&key, delta);
+
+	return 0;
+}
+
+SEC("tp_btf/softirq_entry")
+int on_softirq_entry(u64 *cxt)
+{
+	struct task_struct *task;
+
+	if (!enabled)
+		return 0;
+
+	__u32 cpu = bpf_get_smp_processor_id();
+
+	if (cpu_is_filtered(cpu))
+		return 0;
+
+	__u64 ts = bpf_ktime_get_ns();
+
+	task = (struct task_struct *)bpf_get_current_task();
+	if (!task)
+		return 0;
+
+	struct work_key key = {
+		.type = KWORK_CLASS_SOFTIRQ,
+		.pid = BPF_CORE_READ(task, pid),
+		.task_p = (__u64)task,
+	};
+
+	struct time_data data = {
+		.timestamp = ts,
+	};
+
+	bpf_map_update_elem(&kwork_top_irq_time, &key, &data, BPF_ANY);
+
+	return 0;
+}
+
+SEC("tp_btf/softirq_exit")
+int on_softirq_exit(u64 *cxt)
+{
+	__u64 delta;
+	struct task_struct *task;
+	struct time_data *pelem;
+
+	if (!enabled)
+		return 0;
+
+	__u32 cpu = bpf_get_smp_processor_id();
+
+	if (cpu_is_filtered(cpu))
+		return 0;
+
+	__u64 ts = bpf_ktime_get_ns();
+
+	task = (struct task_struct *)bpf_get_current_task();
+	if (!task)
+		return 0;
+
+	struct work_key key = {
+		.type = KWORK_CLASS_SOFTIRQ,
+		.pid = BPF_CORE_READ(task, pid),
+		.task_p = (__u64)task,
+	};
+
+	pelem = bpf_map_lookup_elem(&kwork_top_irq_time, &key);
+	if (pelem)
+		delta = ts - pelem->timestamp;
+	else
+		delta = ts - from_timestamp;
+
+	update_work(&key, delta);
+
+	return 0;
+}
+
+char LICENSE[] SEC("license") = "Dual BSD/GPL";
diff --git a/tools/perf/util/bpf_skel/lock_contention.bpf.c b/tools/perf/util/bpf_skel/lock_contention.bpf.c
index 8d3cfbb3cc..95cd8414f6 100644
--- a/tools/perf/util/bpf_skel/lock_contention.bpf.c
+++ b/tools/perf/util/bpf_skel/lock_contention.bpf.c
@@ -42,6 +42,14 @@ struct {
 	__uint(max_entries, MAX_ENTRIES);
 } tstamp SEC(".maps");
 
+/* maintain per-CPU timestamp at the beginning of contention */
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__uint(key_size, sizeof(__u32));
+	__uint(value_size, sizeof(struct tstamp_data));
+	__uint(max_entries, 1);
+} tstamp_cpu SEC(".maps");
+
 /* actual lock contention statistics */
 struct {
 	__uint(type, BPF_MAP_TYPE_HASH);
@@ -92,6 +100,13 @@ struct {
 	__uint(max_entries, 1);
 } addr_filter SEC(".maps");
 
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(__u64));
+	__uint(value_size, sizeof(__u8));
+	__uint(max_entries, 1);
+} cgroup_filter SEC(".maps");
+
 struct rw_semaphore___old {
 	struct task_struct *owner;
 } __attribute__((preserve_access_index));
@@ -114,10 +129,14 @@ int has_cpu;
 int has_task;
 int has_type;
 int has_addr;
+int has_cgroup;
 int needs_callstack;
 int stack_skip;
 int lock_owner;
 
+int use_cgroup_v2;
+int perf_subsys_id = -1;
+
 /* determine the key of lock stat */
 int aggr_mode;
 
@@ -130,6 +149,29 @@ int data_fail;
 int task_map_full;
 int data_map_full;
 
+static inline __u64 get_current_cgroup_id(void)
+{
+	struct task_struct *task;
+	struct cgroup *cgrp;
+
+	if (use_cgroup_v2)
+		return bpf_get_current_cgroup_id();
+
+	task = bpf_get_current_task_btf();
+
+	if (perf_subsys_id == -1) {
+#if __has_builtin(__builtin_preserve_enum_value)
+		perf_subsys_id = bpf_core_enum_value(enum cgroup_subsys_id,
+						     perf_event_cgrp_id);
+#else
+		perf_subsys_id = perf_event_cgrp_id;
+#endif
+	}
+
+	cgrp = BPF_CORE_READ(task, cgroups, subsys[perf_subsys_id], cgroup);
+	return BPF_CORE_READ(cgrp, kn, id);
+}
+
 static inline int can_record(u64 *ctx)
 {
 	if (has_cpu) {
@@ -168,6 +210,15 @@ static inline int can_record(u64 *ctx)
 			return 0;
 	}
 
+	if (has_cgroup) {
+		__u8 *ok;
+		__u64 cgrp = get_current_cgroup_id();
+
+		ok = bpf_map_lookup_elem(&cgroup_filter, &cgrp);
+		if (!ok)
+			return 0;
+	}
+
 	return 1;
 }
 
@@ -268,30 +319,57 @@ static inline __u32 check_lock_type(__u64 lock, __u32 flags)
 	return 0;
 }
 
-SEC("tp_btf/contention_begin")
-int contention_begin(u64 *ctx)
+static inline struct tstamp_data *get_tstamp_elem(__u32 flags)
 {
 	__u32 pid;
 	struct tstamp_data *pelem;
 
-	if (!enabled || !can_record(ctx))
-		return 0;
+	/* Use per-cpu array map for spinlock and rwlock */
+	if (flags == (LCB_F_SPIN | LCB_F_READ) || flags == LCB_F_SPIN ||
+	    flags == (LCB_F_SPIN | LCB_F_WRITE)) {
+		__u32 idx = 0;
+
+		pelem = bpf_map_lookup_elem(&tstamp_cpu, &idx);
+		/* Do not update the element for nested locks */
+		if (pelem && pelem->lock)
+			pelem = NULL;
+		return pelem;
+	}
 
 	pid = bpf_get_current_pid_tgid();
 	pelem = bpf_map_lookup_elem(&tstamp, &pid);
+	/* Do not update the element for nested locks */
 	if (pelem && pelem->lock)
-		return 0;
+		return NULL;
 
 	if (pelem == NULL) {
 		struct tstamp_data zero = {};
 
-		bpf_map_update_elem(&tstamp, &pid, &zero, BPF_ANY);
+		if (bpf_map_update_elem(&tstamp, &pid, &zero, BPF_NOEXIST) < 0) {
+			__sync_fetch_and_add(&task_fail, 1);
+			return NULL;
+		}
+
 		pelem = bpf_map_lookup_elem(&tstamp, &pid);
 		if (pelem == NULL) {
 			__sync_fetch_and_add(&task_fail, 1);
-			return 0;
+			return NULL;
 		}
 	}
+	return pelem;
+}
+
+SEC("tp_btf/contention_begin")
+int contention_begin(u64 *ctx)
+{
+	struct tstamp_data *pelem;
+
+	if (!enabled || !can_record(ctx))
+		return 0;
+
+	pelem = get_tstamp_elem(ctx[1]);
+	if (pelem == NULL)
+		return 0;
 
 	pelem->timestamp = bpf_ktime_get_ns();
 	pelem->lock = (__u64)ctx[0];
@@ -330,23 +408,42 @@ int contention_begin(u64 *ctx)
 SEC("tp_btf/contention_end")
 int contention_end(u64 *ctx)
 {
-	__u32 pid;
+	__u32 pid = 0, idx = 0;
 	struct tstamp_data *pelem;
 	struct contention_key key = {};
 	struct contention_data *data;
 	__u64 duration;
+	bool need_delete = false;
 
 	if (!enabled)
 		return 0;
 
-	pid = bpf_get_current_pid_tgid();
-	pelem = bpf_map_lookup_elem(&tstamp, &pid);
-	if (!pelem || pelem->lock != ctx[0])
-		return 0;
+	/*
+	 * For spinlock and rwlock, it needs to get the timestamp for the
+	 * per-cpu map.  However, contention_end does not have the flags
+	 * so it cannot know whether it reads percpu or hash map.
+	 *
+	 * Try per-cpu map first and check if there's active contention.
+	 * If it is, do not read hash map because it cannot go to sleeping
+	 * locks before releasing the spinning locks.
+	 */
+	pelem = bpf_map_lookup_elem(&tstamp_cpu, &idx);
+	if (pelem && pelem->lock) {
+		if (pelem->lock != ctx[0])
+			return 0;
+	} else {
+		pid = bpf_get_current_pid_tgid();
+		pelem = bpf_map_lookup_elem(&tstamp, &pid);
+		if (!pelem || pelem->lock != ctx[0])
+			return 0;
+		need_delete = true;
+	}
 
 	duration = bpf_ktime_get_ns() - pelem->timestamp;
 	if ((__s64)duration < 0) {
-		bpf_map_delete_elem(&tstamp, &pid);
+		pelem->lock = 0;
+		if (need_delete)
+			bpf_map_delete_elem(&tstamp, &pid);
 		__sync_fetch_and_add(&time_fail, 1);
 		return 0;
 	}
@@ -358,16 +455,22 @@ int contention_end(u64 *ctx)
 	case LOCK_AGGR_TASK:
 		if (lock_owner)
 			key.pid = pelem->flags;
-		else
+		else {
+			if (!need_delete)
+				pid = bpf_get_current_pid_tgid();
 			key.pid = pid;
+		}
 		if (needs_callstack)
 			key.stack_id = pelem->stack_id;
 		break;
 	case LOCK_AGGR_ADDR:
-		key.lock_addr = pelem->lock;
+		key.lock_addr_or_cgroup = pelem->lock;
 		if (needs_callstack)
 			key.stack_id = pelem->stack_id;
 		break;
+	case LOCK_AGGR_CGROUP:
+		key.lock_addr_or_cgroup = get_current_cgroup_id();
+		break;
 	default:
 		/* should not happen */
 		return 0;
@@ -376,7 +479,9 @@ int contention_end(u64 *ctx)
 	data = bpf_map_lookup_elem(&lock_stat, &key);
 	if (!data) {
 		if (data_map_full) {
-			bpf_map_delete_elem(&tstamp, &pid);
+			pelem->lock = 0;
+			if (need_delete)
+				bpf_map_delete_elem(&tstamp, &pid);
 			__sync_fetch_and_add(&data_fail, 1);
 			return 0;
 		}
@@ -399,7 +504,9 @@ int contention_end(u64 *ctx)
 				data_map_full = 1;
 			__sync_fetch_and_add(&data_fail, 1);
 		}
-		bpf_map_delete_elem(&tstamp, &pid);
+		pelem->lock = 0;
+		if (need_delete)
+			bpf_map_delete_elem(&tstamp, &pid);
 		return 0;
 	}
 
@@ -412,7 +519,9 @@ int contention_end(u64 *ctx)
 	if (data->min_time > duration)
 		data->min_time = duration;
 
-	bpf_map_delete_elem(&tstamp, &pid);
+	pelem->lock = 0;
+	if (need_delete)
+		bpf_map_delete_elem(&tstamp, &pid);
 	return 0;
 }
 
diff --git a/tools/perf/util/bpf_skel/lock_data.h b/tools/perf/util/bpf_skel/lock_data.h
index 260062a9f2..08482daf61 100644
--- a/tools/perf/util/bpf_skel/lock_data.h
+++ b/tools/perf/util/bpf_skel/lock_data.h
@@ -6,7 +6,7 @@
 struct contention_key {
 	u32 stack_id;
 	u32 pid;
-	u64 lock_addr;
+	u64 lock_addr_or_cgroup;
 };
 
 #define TASK_COMM_LEN  16
@@ -39,6 +39,7 @@ enum lock_aggr_mode {
 	LOCK_AGGR_ADDR = 0,
 	LOCK_AGGR_TASK,
 	LOCK_AGGR_CALLER,
+	LOCK_AGGR_CGROUP,
 };
 
 enum lock_class_sym {