From: Jan Kara <jack@suse.cz>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Dave Jones <davej@redhat.com>, Jan Kara <jack@suse.cz>,
Linux Kernel <linux-kernel@vger.kernel.org>,
jkosina@suse.cz
Subject: Re: fanotify use after free.
Date: Thu, 23 Jan 2014 00:36:22 +0100 [thread overview]
Message-ID: <20140122233622.GB27916@quack.suse.cz> (raw)
In-Reply-To: <CA+55aFxnv050RqhhLBbsvV4=dcu-_ZXokbhmuO8Y4zVcvVJ6OQ@mail.gmail.com>
[-- Attachment #1: Type: text/plain, Size: 2479 bytes --]
On Wed 22-01-14 10:20:01, Linus Torvalds wrote:
> On Tue, Jan 21, 2014 at 10:27 PM, Dave Jones <davej@redhat.com> wrote:
> >
> > BUG fanotify_event_info (Not tainted): Poison overwritten
>
> Looking at the poison data, it seems that is is the
>
> u32 response;
>
> field that has been overwritten (with all zero).
>
> That doesn't really help me guess where the bug is, though. That code
> is crazy. For example, looking at one place where it is set, we have:
>
> - process_access_response():
>
> re->event->response = response;
>
> wake_up(&group->fanotify_data.access_waitq);
>
> kmem_cache_free(fanotify_response_event_cache, re);
>
> which looks buggy in *so* many ways. In particular, we're doing a
> kmem_cache_free() on "re", but what happened to "re->event" that we
> just used? There was no release of that anywhere. Wut?
>
> So it seems that the lifetime of these "fanotify_event_info"
> structures is completely buggered. I don't even see any *attempt* to
> maintain reference counts or other lifetime info. People free the
> containers that point to them without doing anything at all about the
> fsnotify_event that contains the fanotify_event_info that they point
> to.
>
> Jan - how is the lifetime of the fanotify_event_info tied to the
> lifetime of the fanotify_response_event structure that contains
> pointers into it? Because I don't see it.
Yeah, I messed that up. They aren't tied in any way - well, in fact they
end up being tied but in a wrong way. fanotify_event_info lives from the
moment event happens to the moment user reads the event. At that moment the
fanotify_response_event gets created (for those special permission events),
pointing to fanotify_event_info which will get freed just several lines
further :-|
But refcounting seems like an overkill for this - there is exactly one
fanotify_response_event structure iff it is a permission event. So
something like the (completely untested) attached patch should fix the
problem. But I agree it's a bit ugly so we might want something different.
I'll try to think about something better tomorrow.
> And considering that it's the response field that gets overwritten, it
> really sounds like *that* is the exact issue at play here - there is
> some fanotify_response_event structure holding a pointer to the
> fanotify_event that is embedded into a fanotify_event_info that has
> been freed.
Honza
--
Jan Kara <jack@suse.cz>
SUSE Labs, CR
[-- Attachment #2: fanotify_corruption.diff --]
[-- Type: text/x-patch, Size: 1098 bytes --]
diff --git a/fs/notify/fanotify/fanotify.c b/fs/notify/fanotify/fanotify.c
index 58772623f02a..756e9b047e27 100644
--- a/fs/notify/fanotify/fanotify.c
+++ b/fs/notify/fanotify/fanotify.c
@@ -201,8 +201,10 @@ static int fanotify_handle_event(struct fsnotify_group *group,
}
#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
- if (fsn_event->mask & FAN_ALL_PERM_EVENTS)
+ if (fsn_event->mask & FAN_ALL_PERM_EVENTS) {
ret = fanotify_get_response_from_access(group, event);
+ fsnotify_destroy_event(group, event);
+ }
#endif
return ret;
}
diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
index 57d7c083cb4b..d493c72c71fd 100644
--- a/fs/notify/fanotify/fanotify_user.c
+++ b/fs/notify/fanotify/fanotify_user.c
@@ -319,7 +319,8 @@ static ssize_t fanotify_read(struct file *file, char __user *buf,
if (IS_ERR(kevent))
break;
ret = copy_event_to_user(group, kevent, buf);
- fsnotify_destroy_event(group, kevent);
+ if (!(kevent->mask & FAN_ALL_PERM_EVENTS))
+ fsnotify_destroy_event(group, kevent);
if (ret < 0)
break;
buf += ret;
next prev parent reply other threads:[~2014-01-22 23:36 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-01-22 6:27 Dave Jones
2014-01-22 16:43 ` Dave Jones
2014-01-22 18:20 ` Linus Torvalds
2014-01-22 23:36 ` Jan Kara [this message]
2014-01-23 0:08 ` Linus Torvalds
2014-01-23 0:32 ` Dave Jones
2014-01-23 15:05 ` Jan Kara
2014-01-23 10:23 ` Jiri Kosina
2014-01-23 15:05 ` Jan Kara
2014-01-23 15:07 ` Jiri Kosina
2014-01-23 23:55 ` Jan Kara
2014-01-24 7:26 ` Jiri Kosina
2014-01-27 23:40 ` Jan Kara
2014-01-28 6:10 ` Dave Jones
2014-01-28 8:02 ` Jan Kara
2014-01-28 11:07 ` Jiri Kosina
2014-01-28 14:53 ` Jan Kara
2014-01-28 15:24 ` Dave Jones
2014-01-28 10:53 ` Jiri Kosina
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20140122233622.GB27916@quack.suse.cz \
--to=jack@suse.cz \
--cc=davej@redhat.com \
--cc=jkosina@suse.cz \
--cc=linux-kernel@vger.kernel.org \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®