Re: [malware-list] A few concerns about fanotify implementation.

[Lino Sanfilippo <LinoSanfilippo@gmx.de>]

On Fri, Jun 03, 2011 at 01:43:09PM +0400, Vasily Novikov wrote:

> 3. The fanotify file descriptor is always ready to be written to it. But  

This is not correct. The fanotify file is only writeable if there is
at least one event on the access list (meaning at least one file access event
has been read but not already been confirmed by userspace).
Otherwise you will get -ENOENT.

The applied patch should handle this correctly.

Lino


Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
---
 fs/notify/fanotify/fanotify_user.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
index 9fde1c0..f39bcc4 100644
--- a/fs/notify/fanotify/fanotify_user.c
+++ b/fs/notify/fanotify/fanotify_user.c
@@ -304,10 +304,15 @@ static unsigned int fanotify_poll(struct file *file, poll_table *wait)
 	int ret = 0;
 
 	poll_wait(file, &group->notification_waitq, wait);
+	poll_wait(file, &group->fanotify_data.access_waitq, wait);
 	mutex_lock(&group->notification_mutex);
 	if (!fsnotify_notify_queue_is_empty(group))
-		ret = POLLIN | POLLRDNORM;
+		ret |= POLLIN | POLLRDNORM;
 	mutex_unlock(&group->notification_mutex);
+	mutex_lock(&group->fanotify_data.access_mutex);
+	if (!list_empty(&group->fanotify_data.access_list))
+		ret |= POLLOUT | POLLWRNORM;
+	mutex_unlock(&group->fanotify_data.access_mutex);
 
 	return ret;
 }
-- 
1.5.6.5