[1/5] procfs: get rid of ancient BS in pid_revalidate() uses
diff mbox

Message ID 20180526000410.10804-1-viro@ZenIV.linux.org.uk
State New
Headers show

Commit Message

Al Viro May 26, 2018, 12:04 a.m. UTC
From: Al Viro <viro@zeniv.linux.org.uk>

First of all, calling pid_revalidate() in the end of <pid>/* lookups
is *not* about closing any kind of races; that used to be true once
upon a time, but these days those comments are actively misleading.
Especially since pid_revalidate() doesn't even do d_drop() on
failure anymore.  It doesn't matter, anyway, since once
pid_revalidate() starts returning false, ->d_delete() of those
dentries starts saying "don't keep"; they won't get stuck in
dcache any longer than they are pinned.

These calls cannot be just removed, though - the side effect of
pid_revalidate() (updating i_uid/i_gid/etc.) is what we are calling
it for here.

Let's separate the "update ownership" into a new helper (pid_update_inode())
and use it, both in lookups and in pid_revalidate() itself.

The comments in pid_revalidate() are also out of date - they refer to
the time when pid_revalidate() used to call d_drop() directly...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
---
 fs/proc/base.c       | 55 +++++++++++++++++++++++-----------------------------
 fs/proc/internal.h   |  2 +-
 fs/proc/namespaces.c |  9 +++------
 3 files changed, 28 insertions(+), 38 deletions(-)

Comments

Alexey Dobriyan May 31, 2018, 8:28 a.m. UTC | #1
On Sat, May 26, 2018 at 01:04:06AM +0100, Al Viro wrote:
> +	pid_update_inode(task, inode);
>  	d_set_d_op(dentry, &pid_dentry_operations);
>  	d_add(dentry, inode);
> -	/* Close the race of the process dying before we return the dentry */
> -	if (pid_revalidate(dentry, 0))
> -		return 0;
> -out:
> -	return -ENOENT;
> +	return 0;
>  }

ACK this. I never understood what race is prevented if process can
exit after pid_revalidate().

Patch
diff mbox

diff --git a/fs/proc/base.c b/fs/proc/base.c
index eafa39a3a88c..6e0875505898 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1803,15 +1803,22 @@  int pid_getattr(const struct path *path, struct kstat *stat,
 /* dentry stuff */
 
 /*
- *	Exceptional case: normally we are not allowed to unhash a busy
- * directory. In this case, however, we can do it - no aliasing problems
- * due to the way we treat inodes.
- *
+ * Set <pid>/... inode ownership (can change due to setuid(), etc.)
+ */
+void pid_update_inode(struct task_struct *task, struct inode *inode)
+{
+	task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
+
+	inode->i_mode &= ~(S_ISUID | S_ISGID);
+	security_task_to_inode(task, inode);
+}
+
+/*
  * Rewrite the inode's ownerships here because the owning task may have
  * performed a setuid(), etc.
  *
  */
-int pid_revalidate(struct dentry *dentry, unsigned int flags)
+static int pid_revalidate(struct dentry *dentry, unsigned int flags)
 {
 	struct inode *inode;
 	struct task_struct *task;
@@ -1823,10 +1830,7 @@  int pid_revalidate(struct dentry *dentry, unsigned int flags)
 	task = get_proc_task(inode);
 
 	if (task) {
-		task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
-
-		inode->i_mode &= ~(S_ISUID | S_ISGID);
-		security_task_to_inode(task, inode);
+		pid_update_inode(task, inode);
 		put_task_struct(task);
 		return 1;
 	}
@@ -2438,7 +2442,7 @@  static int proc_pident_instantiate(struct inode *dir,
 
 	inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
 	if (!inode)
-		goto out;
+		return -ENOENT;
 
 	ei = PROC_I(inode);
 	if (S_ISDIR(inode->i_mode))
@@ -2448,13 +2452,10 @@  static int proc_pident_instantiate(struct inode *dir,
 	if (p->fop)
 		inode->i_fop = p->fop;
 	ei->op = p->op;
+	pid_update_inode(task, inode);
 	d_set_d_op(dentry, &pid_dentry_operations);
 	d_add(dentry, inode);
-	/* Close the race of the process dying before we return the dentry */
-	if (pid_revalidate(dentry, 0))
-		return 0;
-out:
-	return -ENOENT;
+	return 0;
 }
 
 static struct dentry *proc_pident_lookup(struct inode *dir, 
@@ -3140,22 +3141,18 @@  static int proc_pid_instantiate(struct inode *dir,
 
 	inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
 	if (!inode)
-		goto out;
+		return -ENOENT;
 
 	inode->i_op = &proc_tgid_base_inode_operations;
 	inode->i_fop = &proc_tgid_base_operations;
 	inode->i_flags|=S_IMMUTABLE;
 
 	set_nlink(inode, nlink_tgid);
+	pid_update_inode(task, inode);
 
 	d_set_d_op(dentry, &pid_dentry_operations);
-
 	d_add(dentry, inode);
-	/* Close the race of the process dying before we return the dentry */
-	if (pid_revalidate(dentry, 0))
-		return 0;
-out:
-	return -ENOENT;
+	return 0;
 }
 
 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
@@ -3434,23 +3431,19 @@  static int proc_task_instantiate(struct inode *dir,
 {
 	struct inode *inode;
 	inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
-
 	if (!inode)
-		goto out;
+		return -ENOENT;
+
 	inode->i_op = &proc_tid_base_inode_operations;
 	inode->i_fop = &proc_tid_base_operations;
-	inode->i_flags|=S_IMMUTABLE;
+	inode->i_flags |= S_IMMUTABLE;
 
 	set_nlink(inode, nlink_tid);
+	pid_update_inode(task, inode);
 
 	d_set_d_op(dentry, &pid_dentry_operations);
-
 	d_add(dentry, inode);
-	/* Close the race of the process dying before we return the dentry */
-	if (pid_revalidate(dentry, 0))
-		return 0;
-out:
-	return -ENOENT;
+	return 0;
 }
 
 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
diff --git a/fs/proc/internal.h b/fs/proc/internal.h
index 0f1692e63cb6..04a455b9ae69 100644
--- a/fs/proc/internal.h
+++ b/fs/proc/internal.h
@@ -147,7 +147,7 @@  extern const struct dentry_operations pid_dentry_operations;
 extern int pid_getattr(const struct path *, struct kstat *, u32, unsigned int);
 extern int proc_setattr(struct dentry *, struct iattr *);
 extern struct inode *proc_pid_make_inode(struct super_block *, struct task_struct *, umode_t);
-extern int pid_revalidate(struct dentry *, unsigned int);
+extern void pid_update_inode(struct task_struct *, struct inode *);
 extern int pid_delete_dentry(const struct dentry *);
 extern int proc_pid_readdir(struct file *, struct dir_context *);
 extern struct dentry *proc_pid_lookup(struct inode *, struct dentry *, unsigned int);
diff --git a/fs/proc/namespaces.c b/fs/proc/namespaces.c
index 59b17e509f46..ad1adce6541d 100644
--- a/fs/proc/namespaces.c
+++ b/fs/proc/namespaces.c
@@ -96,19 +96,16 @@  static int proc_ns_instantiate(struct inode *dir,
 
 	inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK | S_IRWXUGO);
 	if (!inode)
-		goto out;
+		return -ENOENT;
 
 	ei = PROC_I(inode);
 	inode->i_op = &proc_ns_link_inode_operations;
 	ei->ns_ops = ns_ops;
+	pid_update_inode(task, inode);
 
 	d_set_d_op(dentry, &pid_dentry_operations);
 	d_add(dentry, inode);
-	/* Close the race of the process dying before we return the dentry */
-	if (pid_revalidate(dentry, 0))
-		return 0;
-out:
-	return -ENOENT;
+	return 0;
 }
 
 static int proc_ns_dir_readdir(struct file *file, struct dir_context *ctx)