lkml.org 
[lkml]   [2019]   [Mar]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: KASAN: use-after-free Read in get_mem_cgroup_from_mm
On Wed, Mar 06, 2019 at 03:41:06PM +0800, zhong jiang wrote:
> On 2019/3/6 14:26, Mike Rapoport wrote:
> > Hi,
> >
> > On Wed, Mar 06, 2019 at 01:53:12PM +0800, zhong jiang wrote:
> >> On 2019/3/6 10:05, Andrea Arcangeli wrote:
> >>> Hello everyone,
> >>>
> >>> [ CC'ed Mike and Peter ]
> >>>
> >>> On Tue, Mar 05, 2019 at 02:42:00PM +0800, zhong jiang wrote:
> >>>> On 2019/3/5 14:26, Dmitry Vyukov wrote:
> >>>>> On Mon, Mar 4, 2019 at 4:32 PM zhong jiang <zhongjiang@huawei.com> wrote:
> >>>>>> On 2019/3/4 22:11, Dmitry Vyukov wrote:
> >>>>>>> On Mon, Mar 4, 2019 at 3:00 PM zhong jiang <zhongjiang@huawei.com> wrote:
> >>>>>>>> On 2019/3/4 15:40, Dmitry Vyukov wrote:
> >>>>>>>>> On Sun, Mar 3, 2019 at 5:19 PM zhong jiang <zhongjiang@huawei.com> wrote:
> >>>>>>>>>> Hi, guys
> >>>>>>>>>>
> >>>>>>>>>> I also hit the following issue. but it fails to reproduce the issue by the log.
> >>>>>>>>>>
> >>>>>>>>>> it seems to the case that we access the mm->owner and deference it will result in the UAF.
> >>>>>>>>>> But it should not be possible that we specify the incomplete process to be the mm->owner.
> >>>>>>>>>>
> >>>>>>>>>> Any thoughts?
> >>>>>>>>> FWIW syzbot was able to reproduce this with this reproducer.
> >>>>>>>>> This looks like a very subtle race (threaded reproducer that runs
> >>>>>>>>> repeatedly in multiple processes), so most likely we are looking for
> >>>>>>>>> something like few instructions inconsistency window.
> >>>>>>>>>
> >>>>>>>> I has a little doubtful about the instrustions inconsistency window.
> >>>>>>>>
> >>>>>>>> I guess that you mean some smb barriers should be taken into account.:-)
> >>>>>>>>
> >>>>>>>> Because IMO, It should not be the lock case to result in the issue.
> >>>>>>> Since the crash was triggered on x86 _most likley_ this is not a
> >>>>>>> missed barrier. What I meant is that one thread needs to executed some
> >>>>>>> code, while another thread is stopped within few instructions.
> >>>>>>>
> >>>>>>>
> >>>>>> It is weird and I can not find any relationship you had said with the issue.:-(
> >>>>>>
> >>>>>> Because It is the cause that mm->owner has been freed, whereas we still deference it.
> >>>>>>
> >>>>>> From the lastest freed task call trace, It fails to create process.
> >>>>>>
> >>>>>> Am I miss something or I misunderstand your meaning. Please correct me.
> >>>>> Your analysis looks correct. I am just saying that the root cause of
> >>>>> this use-after-free seems to be a race condition.
> >>>>>
> >>>>>
> >>>>>
> >>>> Yep, Indeed, I can not figure out how the race works. I will dig up further.
> >>> Yes it's a race condition.
> >>>
> >>> We were aware about the non-cooperative fork userfaultfd feature
> >>> creating userfaultfd file descriptor that gets reported to the parent
> >>> uffd, despite they belong to mm created by failed forks.
> >>>
> >>> https://www.spinics.net/lists/linux-mm/msg136357.html
> >>>
> >> Hi, Andrea
> >>
> >> I still not clear why uffd ioctl can use the incomplete process as the mm->owner.
> >> and how to produce the race.
> > There is a C reproducer in the syzcaller report:
> >
> > https://syzkaller.appspot.com/x/repro.c?x=172fa5a3400000
> >
> >> From your above explainations, My underdtanding is that the process handling do_exexve
> >> will have a temporary mm, which will be used by the UUFD ioctl.
> > The race is between userfaultfd operation and fork() failure:
> >
> > forking thread | userfaultfd monitor thread
> > --------------------------------+-------------------------------
> > fork() |
> > dup_mmap() |
> > dup_userfaultfd() |
> > dup_userfaultfd_complete() |
> > | read(UFFD_EVENT_FORK)
> > | uffdio_copy()
> > | mmget_not_zero()
> > goto bad_fork_something |
> > ... |
> > bad_fork_free: |
> > free_task() |
> > | mem_cgroup_from_task()
> > | /* access stale mm->owner */
> >
> Hi, Mike

Hi, Zhong,

>
> forking thread fails to create the process ,and then free the allocated task struct.
> Other userfaultfd monitor thread should not access the stale mm->owner.
>
> The parent process and child process do not share the mm struct. Userfaultfd monitor thread's
> mm->owner should not point to the freed child task_struct.

IIUC the problem is that above mm (of the mm->owner) is the child
process's mm rather than the uffd monitor's. When
dup_userfaultfd_complete() is called there will be a new userfaultfd
context sent to the uffd monitor thread which linked to the chlid
process's mm, and if the monitor thread do UFFDIO_COPY upon the newly
received userfaultfd it'll operate on that new mm too.

>
> and due to the existence of tasklist_lock, we can not specify the mm->owner to freed task_struct.
>
> I miss something,=-O
>
> Thanks,
> zhong jiang
> >> Thanks,
> >> zhong jiang
>
>

Regards,

--
Peter Xu

\
 
 \ /
  Last update: 2019-03-06 09:13    [W:0.080 / U:64.872 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site