close_range(): fix the logics in descriptor table trimming

-----BEGIN PGP SIGNATURE-----
 
 iHUEABYIAB0WIQQqUNBr3gm4hGXdBJlZ7Krx/gZQ6wUCZv90qwAKCRBZ7Krx/gZQ
 629DAP9NT9/ndByazKrP03nQ/ITNhYVd0cby2iPcXZiHGj/fsQD7BF4yUUiBy7EO
 seIY8rGxc8S5TDJAG4HYa7t1Nuksnw4=
 =oji5
 -----END PGP SIGNATURE-----

Merge tag 'pull-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull close_range() fix from Al Viro:
 "Fix the logic in descriptor table trimming"

* tag 'pull-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  close_range(): fix the logics in descriptor table trimming

fs/file.c

@@ -272,59 +272,45 @@ static inline bool fd_is_open(unsigned int fd, const struct fdtable *fdt)
 	return test_bit(fd, fdt->open_fds);
 }
 
-static unsigned int count_open_files(struct fdtable *fdt)
-{
-	unsigned int size = fdt->max_fds;
-	unsigned int i;
-
-	/* Find the last open fd */
-	for (i = size / BITS_PER_LONG; i > 0; ) {
-		if (fdt->open_fds[--i])
-			break;
-	}
-	i = (i + 1) * BITS_PER_LONG;
-	return i;
-}
-
 /*
  * Note that a sane fdtable size always has to be a multiple of
  * BITS_PER_LONG, since we have bitmaps that are sized by this.
  *
- * 'max_fds' will normally already be properly aligned, but it
+ * punch_hole is optional - when close_range() is asked to unshare
- * turns out that in the close_range() -> __close_range() ->
+ * and close, we don't need to copy descriptors in that range, so
- * unshare_fd() -> dup_fd() -> sane_fdtable_size() we can end
+ * a smaller cloned descriptor table might suffice if the last
- * up having a 'max_fds' value that isn't already aligned.
+ * currently opened descriptor falls into that range.
- *
- * Rather than make close_range() have to worry about this,
- * just make that BITS_PER_LONG alignment be part of a sane
- * fdtable size. Becuase that's really what it is.
  */
-static unsigned int sane_fdtable_size(struct fdtable *fdt, unsigned int max_fds)
+static unsigned int sane_fdtable_size(struct fdtable *fdt, struct fd_range *punch_hole)
 {
-	unsigned int count;
+	unsigned int last = find_last_bit(fdt->open_fds, fdt->max_fds);
 
-	count = count_open_files(fdt);
+	if (last == fdt->max_fds)
-	if (max_fds < NR_OPEN_DEFAULT)
+		return NR_OPEN_DEFAULT;
-		max_fds = NR_OPEN_DEFAULT;
+	if (punch_hole && punch_hole->to >= last && punch_hole->from <= last) {
-	return ALIGN(min(count, max_fds), BITS_PER_LONG);
+		last = find_last_bit(fdt->open_fds, punch_hole->from);
+		if (last == punch_hole->from)
+			return NR_OPEN_DEFAULT;
+	}
+	return ALIGN(last + 1, BITS_PER_LONG);
 }
 
 /*
- * Allocate a new files structure and copy contents from the
+ * Allocate a new descriptor table and copy contents from the passed in
- * passed in files structure.
+ * instance.  Returns a pointer to cloned table on success, ERR_PTR()
- * errorp will be valid only when the returned files_struct is NULL.
+ * on failure.  For 'punch_hole' see sane_fdtable_size().
  */
-struct files_struct *dup_fd(struct files_struct *oldf, unsigned int max_fds, int *errorp)
+struct files_struct *dup_fd(struct files_struct *oldf, struct fd_range *punch_hole)
 {
 	struct files_struct *newf;
 	struct file **old_fds, **new_fds;
 	unsigned int open_files, i;
 	struct fdtable *old_fdt, *new_fdt;
+	int error;
 
-	*errorp = -ENOMEM;
 	newf = kmem_cache_alloc(files_cachep, GFP_KERNEL);
 	if (!newf)
-		goto out;
+		return ERR_PTR(-ENOMEM);
 
 	atomic_set(&newf->count, 1);
 
@@ -341,7 +327,7 @@ struct files_struct *dup_fd(struct files_struct *oldf, unsigned int max_fds, int
 
 	spin_lock(&oldf->file_lock);
 	old_fdt = files_fdtable(oldf);
-	open_files = sane_fdtable_size(old_fdt, max_fds);
+	open_files = sane_fdtable_size(old_fdt, punch_hole);
 
 	/*
 	 * Check whether we need to allocate a larger fd array and fd set.
@@ -354,14 +340,14 @@ struct files_struct *dup_fd(struct files_struct *oldf, unsigned int max_fds, int
 
 		new_fdt = alloc_fdtable(open_files - 1);
 		if (!new_fdt) {
-			*errorp = -ENOMEM;
+			error = -ENOMEM;
 			goto out_release;
 		}
 
 		/* beyond sysctl_nr_open; nothing to do */
 		if (unlikely(new_fdt->max_fds < open_files)) {
 			__free_fdtable(new_fdt);
-			*errorp = -EMFILE;
+			error = -EMFILE;
 			goto out_release;
 		}
 
@@ -372,7 +358,7 @@ struct files_struct *dup_fd(struct files_struct *oldf, unsigned int max_fds, int
 		 */
 		spin_lock(&oldf->file_lock);
 		old_fdt = files_fdtable(oldf);
-		open_files = sane_fdtable_size(old_fdt, max_fds);
+		open_files = sane_fdtable_size(old_fdt, punch_hole);
 	}
 
 	copy_fd_bitmaps(new_fdt, old_fdt, open_files / BITS_PER_LONG);
@@ -406,8 +392,7 @@ struct files_struct *dup_fd(struct files_struct *oldf, unsigned int max_fds, int
 
 out_release:
 	kmem_cache_free(files_cachep, newf);
-out:
+	return ERR_PTR(error);
-	return NULL;
 }
 
 static struct fdtable *close_files(struct files_struct * files)
@@ -748,37 +733,25 @@ int __close_range(unsigned fd, unsigned max_fd, unsigned int flags)
 	if (fd > max_fd)
 		return -EINVAL;
 
-	if (flags & CLOSE_RANGE_UNSHARE) {
+	if ((flags & CLOSE_RANGE_UNSHARE) && atomic_read(&cur_fds->count) > 1) {
-		int ret;
+		struct fd_range range = {fd, max_fd}, *punch_hole = &range;
-		unsigned int max_unshare_fds = NR_OPEN_MAX;
 
 		/*
 		 * If the caller requested all fds to be made cloexec we always
 		 * copy all of the file descriptors since they still want to
 		 * use them.
 		 */
-		if (!(flags & CLOSE_RANGE_CLOEXEC)) {
+		if (flags & CLOSE_RANGE_CLOEXEC)
-			/*
+			punch_hole = NULL;
-			 * If the requested range is greater than the current
-			 * maximum, we're closing everything so only copy all
-			 * file descriptors beneath the lowest file descriptor.
-			 */
-			rcu_read_lock();
-			if (max_fd >= last_fd(files_fdtable(cur_fds)))
-				max_unshare_fds = fd;
-			rcu_read_unlock();
-		}
-
-		ret = unshare_fd(CLONE_FILES, max_unshare_fds, &fds);
-		if (ret)
-			return ret;
 
+		fds = dup_fd(cur_fds, punch_hole);
+		if (IS_ERR(fds))
+			return PTR_ERR(fds);
 		/*
 		 * We used to share our file descriptor table, and have now
 		 * created a private one, make sure we're using it below.
 		 */
-		if (fds)
+		swap(cur_fds, fds);
-			swap(cur_fds, fds);
 	}
 
 	if (flags & CLOSE_RANGE_CLOEXEC)

include/linux/fdtable.h

@@ -22,7 +22,6 @@
  * as this is the granularity returned by copy_fdset().
  */
 #define NR_OPEN_DEFAULT BITS_PER_LONG
-#define NR_OPEN_MAX ~0U
 
 struct fdtable {
 	unsigned int max_fds;
@@ -106,7 +105,10 @@ struct task_struct;
 
 void put_files_struct(struct files_struct *fs);
 int unshare_files(void);
-struct files_struct *dup_fd(struct files_struct *, unsigned, int *) __latent_entropy;
+struct fd_range {
+	unsigned int from, to;
+};
+struct files_struct *dup_fd(struct files_struct *, struct fd_range *) __latent_entropy;
 void do_close_on_exec(struct files_struct *);
 int iterate_fd(struct files_struct *, unsigned,
 		int (*)(const void *, struct file *, unsigned),
@@ -115,8 +117,6 @@ int iterate_fd(struct files_struct *, unsigned,
 extern int close_fd(unsigned int fd);
 extern int __close_range(unsigned int fd, unsigned int max_fd, unsigned int flags);
 extern struct file *file_close_fd(unsigned int fd);
-extern int unshare_fd(unsigned long unshare_flags, unsigned int max_fds,
-		      struct files_struct **new_fdp);
 
 extern struct kmem_cache *files_cachep;
 

kernel/fork.c

@@ -1756,33 +1756,30 @@ static int copy_files(unsigned long clone_flags, struct task_struct *tsk,
 		      int no_files)
 {
 	struct files_struct *oldf, *newf;
-	int error = 0;
 
 	/*
 	 * A background process may not have any files ...
 	 */
 	oldf = current->files;
 	if (!oldf)
-		goto out;
+		return 0;
 
 	if (no_files) {
 		tsk->files = NULL;
-		goto out;
+		return 0;
 	}
 
 	if (clone_flags & CLONE_FILES) {
 		atomic_inc(&oldf->count);
-		goto out;
+		return 0;
 	}
 
-	newf = dup_fd(oldf, NR_OPEN_MAX, &error);
+	newf = dup_fd(oldf, NULL);
-	if (!newf)
+	if (IS_ERR(newf))
-		goto out;
+		return PTR_ERR(newf);
 
 	tsk->files = newf;
-	error = 0;
+	return 0;
-out:
-	return error;
 }
 
 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
@@ -3238,17 +3235,16 @@ static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
 /*
  * Unshare file descriptor table if it is being shared
  */
-int unshare_fd(unsigned long unshare_flags, unsigned int max_fds,
+static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
-	       struct files_struct **new_fdp)
 {
 	struct files_struct *fd = current->files;
-	int error = 0;
 
 	if ((unshare_flags & CLONE_FILES) &&
 	    (fd && atomic_read(&fd->count) > 1)) {
-		*new_fdp = dup_fd(fd, max_fds, &error);
+		fd = dup_fd(fd, NULL);
-		if (!*new_fdp)
+		if (IS_ERR(fd))
-			return error;
+			return PTR_ERR(fd);
+		*new_fdp = fd;
 	}
 
 	return 0;
@@ -3306,7 +3302,7 @@ int ksys_unshare(unsigned long unshare_flags)
 	err = unshare_fs(unshare_flags, &new_fs);
 	if (err)
 		goto bad_unshare_out;
-	err = unshare_fd(unshare_flags, NR_OPEN_MAX, &new_fd);
+	err = unshare_fd(unshare_flags, &new_fd);
 	if (err)
 		goto bad_unshare_cleanup_fs;
 	err = unshare_userns(unshare_flags, &new_cred);
@@ -3398,7 +3394,7 @@ int unshare_files(void)
 	struct files_struct *old, *copy = NULL;
 	int error;
 
-	error = unshare_fd(CLONE_FILES, NR_OPEN_MAX, &copy);
+	error = unshare_fd(CLONE_FILES, &copy);
 	if (error || !copy)
 		return error;