[PATCH] (3/5) sane procfs/dcache interaction

 - sane dentry retention.  Namely, we don't kill /proc/<pid> dentries at the
   first opportunity (as the current tree does).  Instead we do the following:
	* ->d_delete() kills it only if process is already dead.
	* all ->lookup() in proc/base.c end with checking if process is still
	  alive and unhash if it isn't.
	* proc_pid_lookup() (lookup for /proc/<pid>) caches reference to dentry
	  in task_struct.  It's _not_ counted in ->d_count.
	* ->d_iput() resets said reference to NULL.
	* release_task() (burying a zombie) checks if there is a cached
	  reference and if there is - shrinks the subtree.
	* tasklist_lock is used for exclusion.
   That way we are guaranteed that after release_task() all dentries in
   /proc/<pid> will go away as soon as possible; OTOH, before release_task()
   we have normal retention policy - they go away under memory pressure with
   the same rules as for dentries on any other fs.

4 files changed, 47 insertions, 4 deletions

diff --git a/fs/proc/base.c b/fs/proc/base.c
index 44c1189b6eed..7dfa4f269ab6 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -747,7 +747,7 @@ static int pid_fd_revalidate(struct dentry * dentry, int flags)
  * directory. In this case, however, we can do it - no aliasing problems
  * due to the way we treat inodes.
  */
-static int pid_base_revalidate(struct dentry * dentry, int flags)
+static int pid_revalidate(struct dentry * dentry, int flags)
 {
 	if (proc_task(dentry->d_inode)->pid)
 		return 1;
@@ -755,25 +755,42 @@ static int pid_base_revalidate(struct dentry * dentry, int flags)
 	return 0;
 }
 
-static int pid_delete_dentry(struct dentry * dentry)
+static void pid_base_iput(struct dentry *dentry, struct inode *inode)
+{
+	struct task_struct *task = proc_task(inode);
+	write_lock_irq(&tasklist_lock);
+	if (task->proc_dentry == dentry)
+		task->proc_dentry = NULL;
+	write_unlock_irq(&tasklist_lock);
+	iput(inode);
+}
+
+static int pid_fd_delete_dentry(struct dentry * dentry)
 {
 	return 1;
 }
 
+static int pid_delete_dentry(struct dentry * dentry)
+{
+	return proc_task(dentry->d_inode)->pid == 0;
+}
+
 static struct dentry_operations pid_fd_dentry_operations =
 {
 	d_revalidate:	pid_fd_revalidate,
-	d_delete:	pid_delete_dentry,
+	d_delete:	pid_fd_delete_dentry,
 };
 
 static struct dentry_operations pid_dentry_operations =
 {
+	d_revalidate:	pid_revalidate,
 	d_delete:	pid_delete_dentry,
 };
 
 static struct dentry_operations pid_base_dentry_operations =
 {
-	d_revalidate:	pid_base_revalidate,
+	d_revalidate:	pid_revalidate,
+	d_iput:		pid_base_iput,
 	d_delete:	pid_delete_dentry,
 };
 
@@ -842,6 +859,8 @@ static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry)
 		inode->i_mode |= S_IWUSR | S_IXUSR;
 	dentry->d_op = &pid_fd_dentry_operations;
 	d_add(dentry, inode);
+	if (!proc_task(dentry->d_inode)->pid)
+		d_drop(dentry);
 	return NULL;
 
 out_unlock2:
@@ -959,6 +978,8 @@ static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
 	}
 	dentry->d_op = &pid_dentry_operations;
 	d_add(dentry, inode);
+	if (!proc_task(dentry->d_inode)->pid)
+		d_drop(dentry);
 	return NULL;
 
 out:
@@ -1045,6 +1066,11 @@ struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry)
 
 	dentry->d_op = &pid_base_dentry_operations;
 	d_add(dentry, inode);
+	read_lock(&tasklist_lock);
+	proc_task(dentry->d_inode)->proc_dentry = dentry;
+	read_unlock(&tasklist_lock);
+	if (!proc_task(dentry->d_inode)->pid)
+		d_drop(dentry);
 	return NULL;
 out:
 	return ERR_PTR(-ENOENT);
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 8f0df64bb723..a65b376d99da 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -346,6 +346,7 @@ struct task_struct {
 
 /* journalling filesystem info */
 	void *journal_info;
+	struct dentry *proc_dentry;
 };
 
 extern void __put_task_struct(struct task_struct *tsk);
diff --git a/kernel/exit.c b/kernel/exit.c
index 58b05e3ef9f8..e3fdd078f449 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -29,13 +29,28 @@ int getrusage(struct task_struct *, int, struct rusage *);
 
 static inline void __unhash_process(struct task_struct *p)
 {
+	struct dentry *proc_dentry;
 	write_lock_irq(&tasklist_lock);
 	nr_threads--;
 	unhash_pid(p);
 	REMOVE_LINKS(p);
 	list_del(&p->thread_group);
 	p->pid = 0;
+	proc_dentry = p->proc_dentry;
+	if (unlikely(proc_dentry)) {
+		spin_lock(&dcache_lock);
+		if (!list_empty(&proc_dentry->d_hash)) {
+			dget_locked(proc_dentry);
+			list_del_init(&proc_dentry->d_hash);
+		} else
+			proc_dentry = NULL;
+		spin_unlock(&dcache_lock);
+	}
 	write_unlock_irq(&tasklist_lock);
+	if (unlikely(proc_dentry)) {
+		shrink_dcache_parent(proc_dentry);
+		dput(proc_dentry);
+	}
 }
 
 static void release_task(struct task_struct * p)
diff --git a/kernel/fork.c b/kernel/fork.c
index d7a274cbec95..380c1bafe75c 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -665,6 +665,7 @@ int do_fork(unsigned long clone_flags, unsigned long stack_start,
 
 	copy_flags(clone_flags, p);
 	p->pid = get_pid(clone_flags);
+	p->proc_dentry = NULL;
 
 	INIT_LIST_HEAD(&p->run_list);