From 94bf4b64690b3e3426f4e9758d2948491d79fb68 Mon Sep 17 00:00:00 2001
From: Jamie Liu <jamieliu@google.com>
Date: Mon, 26 Jun 2023 13:41:35 -0700
Subject: [PATCH] Return consistent IDs for PID namespaces via procfs.

PiperOrigin-RevId: 543529000
---
 pkg/sentry/fsimpl/proc/task.go       |  6 +++---
 pkg/sentry/fsimpl/proc/task_files.go | 13 ++++++++++++-
 pkg/sentry/kernel/threads.go         | 16 ++++++++++++++++
 3 files changed, 31 insertions(+), 4 deletions(-)

diff --git a/pkg/sentry/fsimpl/proc/task.go b/pkg/sentry/fsimpl/proc/task.go
index 89ce31d64..4ab06eb1d 100644
--- a/pkg/sentry/fsimpl/proc/task.go
+++ b/pkg/sentry/fsimpl/proc/task.go
@@ -72,9 +72,9 @@ func (fs *filesystem) newTaskInode(ctx context.Context, task *kernel.Task, pidns
 		"mounts":    fs.newTaskOwnedInode(ctx, task, fs.NextIno(), 0444, &mountsData{fs: fs, task: task}),
 		"net":       fs.newTaskNetDir(ctx, task),
 		"ns": fs.newTaskOwnedDir(ctx, task, fs.NextIno(), 0511, map[string]kernfs.Inode{
-			"net":  fs.newNamespaceSymlink(ctx, task, fs.NextIno(), "net"),
-			"pid":  fs.newNamespaceSymlink(ctx, task, fs.NextIno(), "pid"),
-			"user": fs.newNamespaceSymlink(ctx, task, fs.NextIno(), "user"),
+			"net":  fs.newFakeNamespaceSymlink(ctx, task, fs.NextIno(), "net"),
+			"pid":  fs.newPIDNamespaceSymlink(ctx, task, fs.NextIno()),
+			"user": fs.newFakeNamespaceSymlink(ctx, task, fs.NextIno(), "user"),
 		}),
 		"oom_score":     fs.newTaskOwnedInode(ctx, task, fs.NextIno(), 0444, newStaticFile("0\n")),
 		"oom_score_adj": fs.newTaskOwnedInode(ctx, task, fs.NextIno(), 0644, &oomScoreAdj{task: task}),
diff --git a/pkg/sentry/fsimpl/proc/task_files.go b/pkg/sentry/fsimpl/proc/task_files.go
index da24fdc12..a5b1f6691 100644
--- a/pkg/sentry/fsimpl/proc/task_files.go
+++ b/pkg/sentry/fsimpl/proc/task_files.go
@@ -1226,7 +1226,18 @@ type namespaceSymlink struct {
 	task *kernel.Task
 }
 
-func (fs *filesystem) newNamespaceSymlink(ctx context.Context, task *kernel.Task, ino uint64, ns string) kernfs.Inode {
+func (fs *filesystem) newPIDNamespaceSymlink(ctx context.Context, task *kernel.Task, ino uint64) kernfs.Inode {
+	target := fmt.Sprintf("pid:[%d]", task.PIDNamespace().ID())
+
+	inode := &namespaceSymlink{task: task}
+	// Note: credentials are overridden by taskOwnedInode.
+	inode.Init(ctx, task.Credentials(), linux.UNNAMED_MAJOR, fs.devMinor, ino, target)
+
+	taskInode := &taskOwnedInode{Inode: inode, owner: task}
+	return taskInode
+}
+
+func (fs *filesystem) newFakeNamespaceSymlink(ctx context.Context, task *kernel.Task, ino uint64, ns string) kernfs.Inode {
 	// Namespace symlinks should contain the namespace name and the inode number
 	// for the namespace instance, so for example user:[123456]. We currently fake
 	// the inode number by sticking the symlink inode in its place.
diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go
index b90118ee2..791c4d1bd 100644
--- a/pkg/sentry/kernel/threads.go
+++ b/pkg/sentry/kernel/threads.go
@@ -149,6 +149,9 @@ type PIDNamespace struct {
 	// appropriate capabilities in userns. The userns pointer is immutable.
 	userns *auth.UserNamespace
 
+	// id is a unique ID assigned to the PID namespace. id is immutable.
+	id uint64
+
 	// The following fields are protected by owner.mu.
 
 	// last is the last ThreadID to be allocated in this namespace.
@@ -198,6 +201,7 @@ func newPIDNamespace(ts *TaskSet, parent *PIDNamespace, userns *auth.UserNamespa
 		owner:         ts,
 		parent:        parent,
 		userns:        userns,
+		id:            lastPIDNSID.Add(1),
 		tasks:         make(map[ThreadID]*Task),
 		tids:          make(map[*Task]ThreadID),
 		tgids:         make(map[*ThreadGroup]ThreadID),
@@ -209,6 +213,13 @@ func newPIDNamespace(ts *TaskSet, parent *PIDNamespace, userns *auth.UserNamespa
 	}
 }
 
+// lastPIDNSID is the last value of PIDNamespace.ID assigned to a PID
+// namespace.
+//
+// This is global rather than being per-TaskSet or Kernel because
+// NewRootPIDNamespace() is called before the Kernel is initialized.
+var lastPIDNSID atomicbitops.Uint64
+
 // NewRootPIDNamespace creates the root PID namespace. 'owner' is not available
 // yet when root namespace is created and must be set by caller.
 func NewRootPIDNamespace(userns *auth.UserNamespace) *PIDNamespace {
@@ -230,6 +241,11 @@ func (ns *PIDNamespace) TaskWithID(tid ThreadID) *Task {
 	return t
 }
 
+// ID returns a non-zero ID that is unique across PID namespaces.
+func (ns *PIDNamespace) ID() uint64 {
+	return ns.id
+}
+
 // ThreadGroupWithID returns the thread group led by the task with thread ID
 // tid in PID namespace ns. If no task has that TID, or if the task with that
 // TID is not a thread group leader, ThreadGroupWithID returns nil.