[PATCH] fix RLIM_NOFILE handling

* dup2() should return -EBADF on exceeded sysctl_nr_open
* dup() should *not* return -EINVAL even if you have rlimit set to 0;
  it should get -EMFILE instead.

Check for orig_start exceeding rlimit taken to sys_fcntl().
Failing expand_files() in dup{2,3}() now gets -EMFILE remapped to -EBADF.
Consequently, remaining checks for rlimit are taken to expand_files().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

3 files changed, 15 insertions, 21 deletions

diff --git a/fs/fcntl.c b/fs/fcntl.c
index 3deec98..61d6251 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -64,11 +64,6 @@ static int locate_fd(unsigned int orig_start, int cloexec)
 	struct fdtable *fdt;
 
 	spin_lock(&files->file_lock);
-
-	error = -EINVAL;
-	if (orig_start >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
-		goto out;
-
 repeat:
 	fdt = files_fdtable(files);
 	/*
@@ -83,10 +78,6 @@ repeat:
 	if (start < fdt->max_fds)
 		newfd = find_next_zero_bit(fdt->open_fds->fds_bits,
 					   fdt->max_fds, start);
-	
-	error = -EMFILE;
-	if (newfd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
-		goto out;
 
 	error = expand_files(files, newfd);
 	if (error < 0)
@@ -141,13 +132,14 @@ asmlinkage long sys_dup3(unsigned int oldfd, unsigned int newfd, int flags)
 	spin_lock(&files->file_lock);
 	if (!(file = fcheck(oldfd)))
 		goto out_unlock;
-	if (newfd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
-		goto out_unlock;
 	get_file(file);			/* We are now finished with oldfd */
 
 	err = expand_files(files, newfd);
-	if (err < 0)
+	if (unlikely(err < 0)) {
+		if (err == -EMFILE)
+			err = -EBADF;
 		goto out_fput;
+	}
 
 	/* To avoid races with open() and dup(), we will mark the fd as
 	 * in-use in the open-file bitmap throughout the entire dup2()
@@ -328,6 +320,8 @@ static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
 	switch (cmd) {
 	case F_DUPFD:
 	case F_DUPFD_CLOEXEC:
+		if (arg >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
+			break;
 		get_file(filp);
 		err = dupfd(filp, arg, cmd == F_DUPFD_CLOEXEC);
 		break;
diff --git a/fs/file.c b/fs/file.c
index 7b3887e..d8773b1 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -250,9 +250,18 @@ int expand_files(struct files_struct *files, int nr)
 	struct fdtable *fdt;
 
 	fdt = files_fdtable(files);
+
+	/*
+	 * N.B. For clone tasks sharing a files structure, this test
+	 * will limit the total number of files that can be opened.
+	 */
+	if (nr >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
+		return -EMFILE;
+
 	/* Do we need to expand? */
 	if (nr < fdt->max_fds)
 		return 0;
+
 	/* Can we expand? */
 	if (nr >= sysctl_nr_open)
 		return -EMFILE;
diff --git a/fs/open.c b/fs/open.c
index 3fe1a685..52647be 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -972,7 +972,6 @@ int get_unused_fd_flags(int flags)
 	int fd, error;
 	struct fdtable *fdt;
 
-  	error = -EMFILE;
 	spin_lock(&files->file_lock);
 
 repeat:
@@ -980,13 +979,6 @@ repeat:
 	fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
 				files->next_fd);
 
-	/*
-	 * N.B. For clone tasks sharing a files structure, this test
-	 * will limit the total number of files that can be opened.
-	 */
-	if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
-		goto out;
-
 	/* Do we need to expand the fd array or fd set?  */
 	error = expand_files(files, fd);
 	if (error < 0)
@@ -997,7 +989,6 @@ repeat:
 	 	 * If we needed to expand the fs array we
 		 * might have blocked - try again.
 		 */
-		error = -EMFILE;
 		goto repeat;
 	}