fs, proc: Introduce /proc/<pid>/task/<tid>/children entry v7

If it's needed I can wrap all this with CONFIG_CHECKPOINT_RESTORE, should I?

---

Quoted 3 lines · Cyrill Gorcunov

If it's needed I can wrap all this with CONFIG_CHECKPOINT_RESTORE, should I?

---

Oleg, if only I'm not missing something obvious you meant handling like below?

Quoted 7 lines · Cyrill Gorcunov

Quoted 3 lines · Cyrill Gorcunov

If it's needed I can wrap all this with CONFIG_CHECKPOINT_RESTORE, should I?

---

Oleg, if only I'm not missing something obvious you meant handling like below?

Yes, but...

fs/proc/array.c int proc_pid_statm(struct seq_file *m, s · line 554

Quoted 4 lines

fs/proc/array.c +4 -0

+struct proc_pid_children_iter {
+	struct pid_namespace *pid_ns;
+	struct pid *parent_pid;
+};

you forgot to remove this definition.

fs/proc/array.c int proc_pid_statm(struct seq_file *m, s · line 619

Quoted 8 lines

fs/proc/array.c +8 -0

+static int children_seq_show(struct seq_file *seq, void *v)
+{
+	struct inode *inode = seq->private;
+	unsigned long pid;
+
+	pid = (unsigned long)pid_nr_ns(v, inode->i_sb->s_fs_info);
+	return seq_printf(seq, " %lu", pid);
+}

just noticed... why unsigned long and %lu? afaics pid_t/%d should work without any typecasts.

No, I rather forgot to quilt refresh :)

Quoted 12 lines

Review

Quoted 8 lines

(quoted diff) +8 -0

+static int children_seq_show(struct seq_file *seq, void *v)
+{
+	struct inode *inode = seq->private;
+	unsigned long pid;
+
+	pid = (unsigned long)pid_nr_ns(v, inode->i_sb->s_fs_info);
+	return seq_printf(seq, " %lu", pid);
+}

just noticed... why unsigned long and %lu? afaics pid_t/%d should work without any typecasts.

I'm not sure how important it is, but Andrew mentioned in one of email that we might be moving from pid_t from int to long one day (which of course will require extreme huge work on checking code where int->long transition might cause problems). So I thought why should I wait then?

[ Andrew, am I correct? ]

But since I'll be refreshig patch anyway, I drop this.

Here is refreshed one for sure.

From: Cyrill Gorcunov <[email protected]>
Subject: fs, proc: Introduce /proc/<pid>/task/<tid>/children entry v8

When we do checkpoint of a task we need to know the list of children the task, has but there is no easy and fast way to generate reverse parent->children chain from arbitrary <pid> (while a parent pid is provided in "PPid" field of /proc/<pid>/status).

So instead of walking over all pids in the system (creating one big process tree in memory, just to figure out which children a task has) -- we add explicit /proc/<pid>/task/<tid>/children entry, because the kernel already has this kind of information but it is not yet exported.

This is a first level children, not the whole process tree.

v2:
- Kame suggested to use a separated /proc/<pid>/children entry instead of poking /proc/<pid>/status
- Andew suggested to use rcu facility instead of locking tasklist_lock
- Tejun pointed that non-seekable seq file might not be enough for tasks with large number of children

v3:
- To be on a safe side use %lu format for pid_t printing

v4:
- New line get printed when sequence ends not at seq->stop, a nit pointed by Tejun
- Documentation update
- tasklist_lock is back, Oleg pointed that ->children list is actually not rcu-safe

v5:
- Oleg suggested to make /proc/<pid>/task/<tid>/children instead of global /proc/<pid>/children, which eliminates hardness related to threads and children migration, and allows patch to be a way simplier.

v6:
- Drop ptrace_may_access tests, pids are can be found anyway so nothing to protect here.
- Update comments and docs, pointed by Oleg.

v7:
- Use get_pid over proc-pid directly, to simplify code, pointed by Oleg.

v8:
- Obtain a starting pid from the proc's inode directly.

Signed-off-by Cyrill Gorcunov <[email protected]>

Cc Andrew Morton <[email protected]>

Cc Pavel Emelyanov <[email protected]>

Cc Serge Hallyn <[email protected]>

Cc KAMEZAWA Hiroyuki <[email protected]>

Cc Oleg Nesterov <[email protected]>

---
 Documentation/filesystems/proc.txt |   18 +++++
 fs/proc/array.c                    |  123 +++++++++++++++++++++++++++++++++++++
 fs/proc/base.c                     |    1 
 fs/proc/internal.h                 |    1 
 4 files changed, 143 insertions(+)

Index: linux-2.6.git/Documentation/filesystems/proc.txt
===================================================================

Documentation/filesystems/proc.txt +18 -0

@@ -40,6 +40,7 @@ Table of Contents
   3.4	/proc/<pid>/coredump_filter - Core dump filtering settings
   3.5	/proc/<pid>/mountinfo - Information about mounts
   3.6	/proc/<pid>/comm  & /proc/<pid>/task/<tid>/comm
+  3.7   /proc/<pid>/task/<tid>/children - Information about task children
 
   4	Configuring procfs
   4.1	Mount options

@@ -1549,6 +1550,23 @@ then the kernel's TASK_COMM_LEN (current
 comm value.
 
 
+3.7	/proc/<pid>/task/<tid>/children - Information about task children
+-------------------------------------------------------------------------
+This file provides a fast way to retrieve first level children pids
+of a task pointed by <pid>/<tid> pair. The format is a space separated
+stream of pids.
+
+Note the "first level" here -- if a child has own children they will
+not be listed here, one needs to read /proc/<children-pid>/task/<tid>/children
+to obtain the descendants.
+
+Since this interface is intended to be fast and cheap it doesn't
+guarantee to provide precise results and some children might be
+skipped, especially if they've exited right after we printed their
+pids, so one need to either stop or freeze processes being inspected
+if precise results are needed.
+
+
 ------------------------------------------------------------------------------
 Configuring procfs
 ------------------------------------------------------------------------------

Index: linux-2.6.git/fs/proc/array.c
===================================================================

fs/proc/array.c +123 -0

@@ -547,3 +547,126 @@ int proc_pid_statm(struct seq_file *m, s
 
 	return 0;
 }
+
+static struct pid *
+get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
+{
+	struct task_struct *start, *task;
+	struct pid *pid = NULL;
+
+	read_lock(&tasklist_lock);
+
+	start = pid_task(proc_pid(inode), PIDTYPE_PID);
+	if (!start)
+		goto out;
+
+	/*
+	 * Lets try to continue searching first, this gives
+	 * us significant speedup on children-rich processes.
+	 */
+	if (pid_prev) {
+		task = pid_task(pid_prev, PIDTYPE_PID);
+		if (task && task->real_parent == start &&
+		    !(list_empty(&task->sibling))) {
+			if (list_is_last(&task->sibling, &start->children))
+				goto out;
+			task = list_first_entry(&task->sibling,
+						struct task_struct, sibling);
+			pid = get_pid(task_pid(task));
+			goto out;
+		}
+	}
+
+	/*
+	 * Slow search case.
+	 *
+	 * We might miss some children here if children
+	 * are exited while we were not holding the lock,
+	 * but it was never promised to be accurate that
+	 * much.
+	 *
+	 * "Just suppose that the parent sleeps, but N children
+	 *  exit after we printed their tids. Now the slow paths
+	 *  skips N extra children, we miss N tasks." (c)
+	 *
+	 * So one need to stop or freeze the leader and all
+	 * its children to get a precise result.
+	 */
+	list_for_each_entry(task, &start->children, sibling) {
+		if (pos-- == 0) {
+			pid = get_pid(task_pid(task));
+			break;
+		}
+	}
+
+out:
+	read_unlock(&tasklist_lock);
+	return pid;
+}
+
+static int children_seq_show(struct seq_file *seq, void *v)
+{
+	struct inode *inode = seq->private;
+	pid_t pid;
+
+	pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
+	return seq_printf(seq, " %d", pid);
+}
+
+static void *children_seq_start(struct seq_file *seq, loff_t *pos)
+{
+	return get_children_pid(seq->private, NULL, *pos);
+}
+
+static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+	struct pid *pid = NULL;
+
+	pid = get_children_pid(seq->private, v, *pos + 1);
+	if (!pid)
+		seq_printf(seq, "\n");
+	put_pid(v);
+
+	++*pos;
+	return pid;
+}
+
+static void children_seq_stop(struct seq_file *seq, void *v)
+{
+	put_pid(v);
+}
+
+static const struct seq_operations children_seq_ops = {
+	.start	= children_seq_start,
+	.next	= children_seq_next,
+	.stop	= children_seq_stop,
+	.show	= children_seq_show,
+};
+
+static int children_seq_open(struct inode *inode, struct file *file)
+{
+	struct seq_file *m;
+	int ret;
+
+	ret = seq_open(file, &children_seq_ops);
+	if (ret)
+		return ret;
+
+	m = file->private_data;
+	m->private = inode;
+
+	return ret;
+}
+
+int children_seq_release(struct inode *inode, struct file *file)
+{
+	seq_release(inode, file);
+	return 0;
+}
+
+const struct file_operations proc_tid_children_operations = {
+	.open    = children_seq_open,
+	.read    = seq_read,
+	.llseek  = seq_lseek,
+	.release = children_seq_release,
+};

Index: linux-2.6.git/fs/proc/base.c
===================================================================

fs/proc/base.c +1 -0

@@ -3454,6 +3454,7 @@ static const struct pid_entry tid_base_s
 	ONE("stat",      S_IRUGO, proc_tid_stat),
 	ONE("statm",     S_IRUGO, proc_pid_statm),
 	REG("maps",      S_IRUGO, proc_maps_operations),
+	REG("children",  S_IRUGO, proc_tid_children_operations),
 #ifdef CONFIG_NUMA
 	REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
 #endif

Index: linux-2.6.git/fs/proc/internal.h
===================================================================

fs/proc/internal.h +1 -0

@@ -53,6 +53,7 @@ extern int proc_pid_statm(struct seq_fil
 				struct pid *pid, struct task_struct *task);
 extern loff_t mem_lseek(struct file *file, loff_t offset, int orig);
 
+extern const struct file_operations proc_tid_children_operations;
 extern const struct file_operations proc_maps_operations;
 extern const struct file_operations proc_numa_maps_operations;
 extern const struct file_operations proc_smaps_operations;

Quoted 9 lines · Oleg Nesterov

Quoted 7 lines · Cyrill Gorcunov

Quoted 3 lines · Cyrill Gorcunov

If it's needed I can wrap all this with CONFIG_CHECKPOINT_RESTORE, should I?

---

Oleg, if only I'm not missing something obvious you meant handling like below?

Quoted 9 lines

Yes, but...

fs/proc/array.c int proc_pid_statm(struct seq_file *m, s · line 554

Quoted 4 lines

fs/proc/array.c +4 -0

+struct proc_pid_children_iter {
+	struct pid_namespace *pid_ns;
+	struct pid *parent_pid;
+};

you forgot to remove this definition.

No, I rather forgot to quilt refresh :)

Quoted 12 lines

fs/proc/array.c int proc_pid_statm(struct seq_file *m, s · line 619

Quoted 8 lines

fs/proc/array.c +8 -0

+static int children_seq_show(struct seq_file *seq, void *v)
+{
+	struct inode *inode = seq->private;
+	unsigned long pid;
+
+	pid = (unsigned long)pid_nr_ns(v, inode->i_sb->s_fs_info);
+	return seq_printf(seq, " %lu", pid);
+}

just noticed... why unsigned long and %lu? afaics pid_t/%d should work without any typecasts.

I'm not sure how important it is, but Andrew mentioned in one of email that we might be moving from pid_t from int to long one day (which of course will require extreme huge work on checking code where int->long transition might cause problems). So I thought why should I wait then?

[ Andrew, am I correct? ]

Quoted 19 lines

Quoted 13 lines

fs/proc/array.c int proc_pid_statm(struct seq_file *m, s · line 619

Quoted 9 lines

fs/proc/array.c +8 -0

+static int children_seq_show(struct seq_file *seq, void *v)
+{
+	struct inode *inode = seq->private;
+	unsigned long pid;
+
+	pid = (unsigned long)pid_nr_ns(v, inode->i_sb->s_fs_info);
+	return seq_printf(seq, " %lu", pid);
+}

just noticed... why unsigned long and %lu? afaics pid_t/%d should work without any typecasts.

I'm not sure how important it is, but Andrew mentioned in one of email that we might be moving from pid_t from int to long one day (which of course will require extreme huge work on checking code where int->long transition might cause problems). So I thought why should I wait then?

Quoted 1 line

[ Andrew, am I correct? ]

It is going to take a lot to get to a > 32bit pid value. I would not worry about it today unless there is some 64bit arch somewhere that defines pid_t as long.