sched: latencytop support

LatencyTOP kernel infrastructure; it measures latencies in the scheduler and tracks it system wide and per process. Signed-off-by: Arjan van de Ven <[email protected]> Signed-off-by: Ingo Molnar <[email protected]>
niw · Jan 25, 2008 · 9745512 · 9745512
1 parent 326587b
commit 9745512
Show file tree

Hide file tree

Showing 11 changed files with 429 additions and 1 deletion.
diff --git a/arch/x86/kernel/stacktrace.c b/arch/x86/kernel/stacktrace.c
@@ -33,13 +33,33 @@ static void save_stack_address(void *data, unsigned long addr)
 		trace->entries[trace->nr_entries++] = addr;
 }
 
+static void save_stack_address_nosched(void *data, unsigned long addr)
+{
+	struct stack_trace *trace = (struct stack_trace *)data;
+	if (in_sched_functions(addr))
+		return;
+	if (trace->skip > 0) {
+		trace->skip--;
+		return;
+	}
+	if (trace->nr_entries < trace->max_entries)
+		trace->entries[trace->nr_entries++] = addr;
+}
+
 static const struct stacktrace_ops save_stack_ops = {
 	.warning = save_stack_warning,
 	.warning_symbol = save_stack_warning_symbol,
 	.stack = save_stack_stack,
 	.address = save_stack_address,
 };
 
+static const struct stacktrace_ops save_stack_ops_nosched = {
+	.warning = save_stack_warning,
+	.warning_symbol = save_stack_warning_symbol,
+	.stack = save_stack_stack,
+	.address = save_stack_address_nosched,
+};
+
 /*
  * Save stack-backtrace addresses into a stack_trace buffer.
  */
@@ -50,3 +70,10 @@ void save_stack_trace(struct stack_trace *trace)
 		trace->entries[trace->nr_entries++] = ULONG_MAX;
 }
 EXPORT_SYMBOL(save_stack_trace);
+
+void save_stack_trace_tsk(struct task_struct *tsk, struct stack_trace *trace)
+{
+	dump_trace(tsk, NULL, NULL, &save_stack_ops_nosched, trace);
+	if (trace->nr_entries < trace->max_entries)
+		trace->entries[trace->nr_entries++] = ULONG_MAX;
+}
diff --git a/fs/proc/base.c b/fs/proc/base.c
@@ -310,6 +310,77 @@ static int proc_pid_schedstat(struct task_struct *task, char *buffer)
 }
 #endif
 
+#ifdef CONFIG_LATENCYTOP
+static int lstats_show_proc(struct seq_file *m, void *v)
+{
+	int i;
+	struct task_struct *task = m->private;
+	seq_puts(m, "Latency Top version : v0.1\n");
+
+	for (i = 0; i < 32; i++) {
+		if (task->latency_record[i].backtrace[0]) {
+			int q;
+			seq_printf(m, "%i %li %li ",
+				task->latency_record[i].count,
+				task->latency_record[i].time,
+				task->latency_record[i].max);
+			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
+				char sym[KSYM_NAME_LEN];
+				char *c;
+				if (!task->latency_record[i].backtrace[q])
+					break;
+				if (task->latency_record[i].backtrace[q] == ULONG_MAX)
+					break;
+				sprint_symbol(sym, task->latency_record[i].backtrace[q]);
+				c = strchr(sym, '+');
+				if (c)
+					*c = 0;
+				seq_printf(m, "%s ", sym);
+			}
+			seq_printf(m, "\n");
+		}
+
+	}
+	return 0;
+}
+
+static int lstats_open(struct inode *inode, struct file *file)
+{
+	int ret;
+	struct seq_file *m;
+	struct task_struct *task = get_proc_task(inode);
+
+	ret = single_open(file, lstats_show_proc, NULL);
+	if (!ret) {
+		m = file->private_data;
+		m->private = task;
+	}
+	return ret;
+}
+
+static ssize_t lstats_write(struct file *file, const char __user *buf,
+			    size_t count, loff_t *offs)
+{
+	struct seq_file *m;
+	struct task_struct *task;
+
+	m = file->private_data;
+	task = m->private;
+	clear_all_latency_tracing(task);
+
+	return count;
+}
+
+static const struct file_operations proc_lstats_operations = {
+	.open		= lstats_open,
+	.read		= seq_read,
+	.write		= lstats_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+#endif
+
 /* The badness from the OOM killer */
 unsigned long badness(struct task_struct *p, unsigned long uptime);
 static int proc_oom_score(struct task_struct *task, char *buffer)
@@ -1020,6 +1091,7 @@ static const struct file_operations proc_fault_inject_operations = {
 };
 #endif
 
+
 #ifdef CONFIG_SCHED_DEBUG
 /*
  * Print out various scheduling related per-task fields:
@@ -2230,6 +2302,9 @@ static const struct pid_entry tgid_base_stuff[] = {
 #ifdef CONFIG_SCHEDSTATS
 	INF("schedstat",  S_IRUGO, pid_schedstat),
 #endif
+#ifdef CONFIG_LATENCYTOP
+	REG("latency",  S_IRUGO, lstats),
+#endif
 #ifdef CONFIG_PROC_PID_CPUSET
 	REG("cpuset",     S_IRUGO, cpuset),
 #endif
@@ -2555,6 +2630,9 @@ static const struct pid_entry tid_base_stuff[] = {
 #ifdef CONFIG_SCHEDSTATS
 	INF("schedstat", S_IRUGO, pid_schedstat),
 #endif
+#ifdef CONFIG_LATENCYTOP
+	REG("latency",  S_IRUGO, lstats),
+#endif
 #ifdef CONFIG_PROC_PID_CPUSET
 	REG("cpuset",    S_IRUGO, cpuset),
 #endif

diff --git a/include/linux/latencytop.h b/include/linux/latencytop.h
@@ -0,0 +1,44 @@
+/*
+ * latencytop.h: Infrastructure for displaying latency
+ *
+ * (C) Copyright 2008 Intel Corporation
+ * Author: Arjan van de Ven <[email protected]>
+ *
+ */
+
+#ifndef _INCLUDE_GUARD_LATENCYTOP_H_
+#define _INCLUDE_GUARD_LATENCYTOP_H_
+
+#ifdef CONFIG_LATENCYTOP
+
+#define LT_SAVECOUNT		32
+#define LT_BACKTRACEDEPTH	12
+
+struct latency_record {
+	unsigned long	backtrace[LT_BACKTRACEDEPTH];
+	unsigned int	count;
+	unsigned long	time;
+	unsigned long	max;
+};
+
+
+struct task_struct;
+
+void account_scheduler_latency(struct task_struct *task, int usecs, int inter);
+
+void clear_all_latency_tracing(struct task_struct *p);
+
+#else
+
+static inline void
+account_scheduler_latency(struct task_struct *task, int usecs, int inter)
+{
+}
+
+static inline void clear_all_latency_tracing(struct task_struct *p)
+{
+}
+
+#endif
+
+#endif
diff --git a/include/linux/sched.h b/include/linux/sched.h
@@ -88,6 +88,7 @@ struct sched_param {
 #include <linux/hrtimer.h>
 #include <linux/task_io_accounting.h>
 #include <linux/kobject.h>
+#include <linux/latencytop.h>
 
 #include <asm/processor.h>
 
@@ -1220,6 +1221,10 @@ struct task_struct {
 	int make_it_fail;
 #endif
 	struct prop_local_single dirties;
+#ifdef CONFIG_LATENCYTOP
+	int latency_record_count;
+	struct latency_record latency_record[LT_SAVECOUNT];
+#endif
 };
 
 /*

diff --git a/include/linux/stacktrace.h b/include/linux/stacktrace.h
@@ -9,10 +9,13 @@ struct stack_trace {
 };
 
 extern void save_stack_trace(struct stack_trace *trace);
+extern void save_stack_trace_tsk(struct task_struct *tsk,
+				struct stack_trace *trace);
 
 extern void print_stack_trace(struct stack_trace *trace, int spaces);
 #else
 # define save_stack_trace(trace)			do { } while (0)
+# define save_stack_trace_tsk(tsk, trace)		do { } while (0)
 # define print_stack_trace(trace, spaces)		do { } while (0)
 #endif
 

diff --git a/kernel/Makefile b/kernel/Makefile
@@ -62,6 +62,7 @@ obj-$(CONFIG_SYSCTL) += utsname_sysctl.o
 obj-$(CONFIG_TASK_DELAY_ACCT) += delayacct.o
 obj-$(CONFIG_TASKSTATS) += taskstats.o tsacct.o
 obj-$(CONFIG_MARKERS) += marker.o
+obj-$(CONFIG_LATENCYTOP) += latencytop.o
 
 ifneq ($(CONFIG_SCHED_NO_NO_OMIT_FRAME_POINTER),y)
 # According to Alan Modra <[email protected]>, the -fno-omit-frame-pointer is

diff --git a/kernel/fork.c b/kernel/fork.c
@@ -1205,6 +1205,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 #ifdef TIF_SYSCALL_EMU
 	clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
 #endif
+	clear_all_latency_tracing(p);
 
 	/* Our parent execution domain becomes current domain
 	   These must match for thread signalling to apply */