linux-kernel - Re: [PATCH] mm/oom_kill: Ensure MMU notifier range_end() is paired with range

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 10 Mar 2021 23:22:09 -0800
From:   Sean Christopherson <seanjc@...gle.com>
To:     Jason Gunthorpe <jgg@...pe.ca>
Cc:     Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org, David Rientjes <rientjes@...gle.com>,
        Ben Gardon <bgardon@...gle.com>,
        Michal Hocko <mhocko@...e.com>,
        Jérôme Glisse <jglisse@...hat.com>,
        Andrea Arcangeli <aarcange@...hat.com>,
        Johannes Weiner <hannes@...xchg.org>,
        Dimitri Sivanich <dimitri.sivanich@....com>
Subject: Re: [PATCH] mm/oom_kill: Ensure MMU notifier range_end() is paired
 with range_start()

On Wed, Mar 10, 2021, Jason Gunthorpe wrote:
> On Wed, Mar 10, 2021 at 05:20:01PM -0800, Sean Christopherson wrote:
> 
> > > Which I believe is fatal to kvm? These notifiers certainly do not only
> > > happen at process exit.
> > 
> > My point about the process dying is that the existing bug that causes
> > mmu_notifier_count to become imbalanced is benign only because the process is
> > being killed, and thus KVM will stop running its vCPUs.
> 
> Are you saying we only call non-blocking invalidate during a process
> exit event?? 

Yes?  __oom_reap_task_mm() is the only user of _nonblock(), if that's what you're
asking.

> > > So, both of the remaining _end users become corrupted with this patch!
> > 
> > I don't follow.  mn_hlist_invalidate_range_start() iterates over all
> > notifiers, even if a notifier earlier in the chain failed.  How will
> > KVM become imbalanced?
> 
> Er, ok, that got left in a weird way. There is another "bug" where end
> is not supposed to be called if the start failed.

Ha, the best kind of feature.  :-)

> > The existing _end users never fail their _start.  If KVM started failing its
> > start, then yes, it could get corrupted.  
> 
> Well, maybe that is the way out of this now. If we don't permit a
> start to fail if there is an end then we have no problem to unwind it
> as we can continue to call everything. This can't be backported too
> far though, the itree notifier conversions are what made the WARN_ON
> safe today.
> 
> Something very approximately like this is closer to my preference:

Makes sense.  I don't have a strong preference, I'll give this a spin tomorrow.

> diff --git a/mm/mmu_notifier.c b/mm/mmu_notifier.c
> index 61ee40ed804ee5..6d5cd20f81dadc 100644
> --- a/mm/mmu_notifier.c
> +++ b/mm/mmu_notifier.c
> @@ -501,10 +501,25 @@ static int mn_hlist_invalidate_range_start(
>  						"");
>  				WARN_ON(mmu_notifier_range_blockable(range) ||
>  					_ret != -EAGAIN);
> +				/*
> +				 * We call all the notifiers on any EAGAIN,
> +				 * there is no way for a notifier to know if
> +				 * its start method failed, thus a start that
> +				 * does EAGAIN can't also do end.
> +				 */
> +				WARN_ON(ops->invalidate_range_end);
>  				ret = _ret;
>  			}
>  		}
>  	}
> +
> +	if (ret) {
> +		/* Must be non-blocking to get here*/
> +		hlist_for_each_entry_rcu (subscription, &subscriptions->list,
> +					  hlist, srcu_read_lock_held(&srcu))
> +			subscription->ops->invalidate_range_end(subscription,
> +								range);
> +	}
>  	srcu_read_unlock(&srcu, id);
>  
>  	return ret;