task IO accounting: provide distinct tgid/tid I/O statistics

Report per-thread I/O statistics in /proc/pid/task/tid/io and aggregate
parent I/O statistics in /proc/pid/io.  This approach follows the same
model used to account per-process and per-thread CPU times.

As a practial application, this allows for example to quickly find the top
I/O consumer when a process spawns many child threads that perform the
actual I/O work, because the aggregated I/O statistics can always be found
in /proc/pid/io.

[ Oleg Nesterov points out that we should check that the task is still
  alive before we iterate over the threads, but also says that we can do
  that fixup on top of this later.  - Linus ]

Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrea Righi <righi.andrea@gmail.com>
Cc: Matt Heaton <matt@hostmonster.com>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Acked-by-with-comments: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

4 files changed, 108 insertions, 15 deletions

diff --git a/fs/proc/base.c b/fs/proc/base.c
index 58c3e6a..a891fe4 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -2376,29 +2376,82 @@ static int proc_base_fill_cache(struct file *filp, void *dirent,
 }
 
 #ifdef CONFIG_TASK_IO_ACCOUNTING
-static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
-{
+static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
+{
+	u64 rchar, wchar, syscr, syscw;
+	struct task_io_accounting ioac;
+
+	if (!whole) {
+		rchar = task->rchar;
+		wchar = task->wchar;
+		syscr = task->syscr;
+		syscw = task->syscw;
+		memcpy(&ioac, &task->ioac, sizeof(ioac));
+	} else {
+		unsigned long flags;
+		struct task_struct *t = task;
+		rchar = wchar = syscr = syscw = 0;
+		memset(&ioac, 0, sizeof(ioac));
+
+		rcu_read_lock();
+		do {
+			rchar += t->rchar;
+			wchar += t->wchar;
+			syscr += t->syscr;
+			syscw += t->syscw;
+
+			ioac.read_bytes += t->ioac.read_bytes;
+			ioac.write_bytes += t->ioac.write_bytes;
+			ioac.cancelled_write_bytes +=
+					t->ioac.cancelled_write_bytes;
+			t = next_thread(t);
+		} while (t != task);
+		rcu_read_unlock();
+
+		if (lock_task_sighand(task, &flags)) {
+			struct signal_struct *sig = task->signal;
+
+			rchar += sig->rchar;
+			wchar += sig->wchar;
+			syscr += sig->syscr;
+			syscw += sig->syscw;
+
+			ioac.read_bytes += sig->ioac.read_bytes;
+			ioac.write_bytes += sig->ioac.write_bytes;
+			ioac.cancelled_write_bytes +=
+					sig->ioac.cancelled_write_bytes;
+
+			unlock_task_sighand(task, &flags);
+		}
+	}
+
 	return sprintf(buffer,
-#ifdef CONFIG_TASK_XACCT
 			"rchar: %llu\n"
 			"wchar: %llu\n"
 			"syscr: %llu\n"
 			"syscw: %llu\n"
-#endif
 			"read_bytes: %llu\n"
 			"write_bytes: %llu\n"
 			"cancelled_write_bytes: %llu\n",
-#ifdef CONFIG_TASK_XACCT
-			(unsigned long long)task->rchar,
-			(unsigned long long)task->wchar,
-			(unsigned long long)task->syscr,
-			(unsigned long long)task->syscw,
-#endif
-			(unsigned long long)task->ioac.read_bytes,
-			(unsigned long long)task->ioac.write_bytes,
-			(unsigned long long)task->ioac.cancelled_write_bytes);
+			(unsigned long long)rchar,
+			(unsigned long long)wchar,
+			(unsigned long long)syscr,
+			(unsigned long long)syscw,
+			(unsigned long long)ioac.read_bytes,
+			(unsigned long long)ioac.write_bytes,
+			(unsigned long long)ioac.cancelled_write_bytes);
+}
+
+static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
+{
+	return do_io_accounting(task, buffer, 0);
 }
-#endif
+
+static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
+{
+	return do_io_accounting(task, buffer, 1);
+}
+#endif /* CONFIG_TASK_IO_ACCOUNTING */
 
 /*
  * Thread groups
@@ -2470,7 +2523,7 @@ static const struct pid_entry tgid_base_stuff[] = {
 	REG("coredump_filter", S_IRUGO|S_IWUSR, coredump_filter),
 #endif
 #ifdef CONFIG_TASK_IO_ACCOUNTING
-	INF("io",	S_IRUGO, pid_io_accounting),
+	INF("io",	S_IRUGO, tgid_io_accounting),
 #endif
 };
 
@@ -2797,6 +2850,9 @@ static const struct pid_entry tid_base_stuff[] = {
 #ifdef CONFIG_FAULT_INJECTION
 	REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
 #endif
+#ifdef CONFIG_TASK_IO_ACCOUNTING
+	INF("io",	S_IRUGO, tid_io_accounting),
+#endif
 };
 
 static int proc_tid_base_readdir(struct file * filp,
diff --git a/include/linux/sched.h b/include/linux/sched.h
index af780f2..d22ffe0 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -506,6 +506,10 @@ struct signal_struct {
 	unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
 	unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
 	unsigned long inblock, oublock, cinblock, coublock;
+#ifdef CONFIG_TASK_XACCT
+	u64 rchar, wchar, syscr, syscw;
+#endif
+	struct task_io_accounting ioac;
 
 	/*
 	 * Cumulative ns of scheduled CPU time for dead threads in the
diff --git a/kernel/exit.c b/kernel/exit.c
index 8a4d4d1..ad933bb 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -120,6 +120,18 @@ static void __exit_signal(struct task_struct *tsk)
 		sig->nivcsw += tsk->nivcsw;
 		sig->inblock += task_io_get_inblock(tsk);
 		sig->oublock += task_io_get_oublock(tsk);
+#ifdef CONFIG_TASK_XACCT
+		sig->rchar += tsk->rchar;
+		sig->wchar += tsk->wchar;
+		sig->syscr += tsk->syscr;
+		sig->syscw += tsk->syscw;
+#endif /* CONFIG_TASK_XACCT */
+#ifdef CONFIG_TASK_IO_ACCOUNTING
+		sig->ioac.read_bytes += tsk->ioac.read_bytes;
+		sig->ioac.write_bytes += tsk->ioac.write_bytes;
+		sig->ioac.cancelled_write_bytes +=
+					tsk->ioac.cancelled_write_bytes;
+#endif /* CONFIG_TASK_IO_ACCOUNTING */
 		sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
 		sig = NULL; /* Marker for below. */
 	}
@@ -1366,6 +1378,21 @@ static int wait_task_zombie(struct task_struct *p, int options,
 		psig->coublock +=
 			task_io_get_oublock(p) +
 			sig->oublock + sig->coublock;
+#ifdef CONFIG_TASK_XACCT
+		psig->rchar += p->rchar + sig->rchar;
+		psig->wchar += p->wchar + sig->wchar;
+		psig->syscr += p->syscr + sig->syscr;
+		psig->syscw += p->syscw + sig->syscw;
+#endif /* CONFIG_TASK_XACCT */
+#ifdef CONFIG_TASK_IO_ACCOUNTING
+		psig->ioac.read_bytes +=
+			p->ioac.read_bytes + sig->ioac.read_bytes;
+		psig->ioac.write_bytes +=
+			p->ioac.write_bytes + sig->ioac.write_bytes;
+		psig->ioac.cancelled_write_bytes +=
+				p->ioac.cancelled_write_bytes +
+				sig->ioac.cancelled_write_bytes;
+#endif /* CONFIG_TASK_IO_ACCOUNTING */
 		spin_unlock_irq(&p->parent->sighand->siglock);
 	}
 
diff --git a/kernel/fork.c b/kernel/fork.c
index 813d5c8..b99d73e 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -812,6 +812,12 @@ static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
 	sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
 	sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
 	sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0;
+#ifdef CONFIG_TASK_XACCT
+	sig->rchar = sig->wchar = sig->syscr = sig->syscw = 0;
+#endif
+#ifdef CONFIG_TASK_IO_ACCOUNTING
+	memset(&sig->ioac, 0, sizeof(sig->ioac));
+#endif
 	sig->sum_sched_runtime = 0;
 	INIT_LIST_HEAD(&sig->cpu_timers[0]);
 	INIT_LIST_HEAD(&sig->cpu_timers[1]);