fanotify: fix -EOVERFLOW with large files on 64-bit
authorWill Woods <wwoods@redhat.com>
Tue, 6 May 2014 19:50:10 +0000 (12:50 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 6 May 2014 20:04:59 +0000 (13:04 -0700)
On 64-bit systems, O_LARGEFILE is automatically added to flags inside
the open() syscall (also openat(), blkdev_open(), etc).  Userspace
therefore defines O_LARGEFILE to be 0 - you can use it, but it's a
no-op.  Everything should be O_LARGEFILE by default.

But: when fanotify does create_fd() it uses dentry_open(), which skips
all that.  And userspace can't set O_LARGEFILE in fanotify_init()
because it's defined to 0.  So if fanotify gets an event regarding a
large file, the read() will just fail with -EOVERFLOW.

This patch adds O_LARGEFILE to fanotify_init()'s event_f_flags on 64-bit
systems, using the same test as open()/openat()/etc.

Addresses https://bugzilla.redhat.com/show_bug.cgi?id=696821

Signed-off-by: Will Woods <wwoods@redhat.com>
Acked-by: Eric Paris <eparis@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/notify/fanotify/fanotify_user.c

index 4e565c814309d2637c7d9bb8081f03a4d19216d0..732648b270dc1072f679b550fb386147e65b5db0 100644 (file)
@@ -698,6 +698,8 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
        }
        group->overflow_event = &oevent->fse;
 
+       if (force_o_largefile())
+               event_f_flags |= O_LARGEFILE;
        group->fanotify_data.f_flags = event_f_flags;
 #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
        spin_lock_init(&group->fanotify_data.access_lock);