Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 11 May 2023 01:21:06 -0700
From:   Lorenzo Stoakes <lstoakes@gmail.com>
To:     Mark Rutland <mark.rutland@arm.com>
Cc:     linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>,
        "Liam R . Howlett" <Liam.Howlett@oracle.com>,
        Vlastimil Babka <vbabka@suse.cz>
Subject: Re: [PATCH] mm/mmap/vma_merge: always check invariants
Message-ID: <ZFylcp2pmUVgxE7g@murray>
References: <df548a6ae3fa135eec3b446eb3dae8eb4227da97.1682885809.git.lstoakes@gmail.com>
 <ZFunF7DmMdK05MoF@FVFF77S0Q05N.cambridge.arm.com>
 <ZFvAnF0DzEUN7F9r@murray>
 <ZFvDrZRV8RnoPR69@FVFF77S0Q05N.cambridge.arm.com>
 <ZFvFoj--H21sxzCQ@murray>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ZFvFoj--H21sxzCQ@murray>
Precedence: bulk

On Wed, May 10, 2023 at 09:26:10AM -0700, Lorenzo Stoakes wrote:
> On Wed, May 10, 2023 at 05:17:49PM +0100, Mark Rutland wrote:
> > On Wed, May 10, 2023 at 09:04:44AM -0700, Lorenzo Stoakes wrote:
> > > On Wed, May 10, 2023 at 03:15:51PM +0100, Mark Rutland wrote:
> > > > Hi,
> > > >
> > > > On Sun, Apr 30, 2023 at 09:19:17PM +0100, Lorenzo Stoakes wrote:
> > > > > We may still have inconsistent input parameters even if we choose not to
> > > > > merge and the vma_merge() invariant checks are useful for checking this
> > > > > with no production runtime cost (these are only relevant when
> > > > > CONFIG_DEBUG_VM is specified).
> > > > >
> > > > > Therefore, perform these checks regardless of whether we merge.
> > > > >
> > > > > This is relevant, as a recent issue (addressed in commit "mm/mempolicy:
> > > > > Correctly update prev when policy is equal on mbind") in the mbind logic
> > > > > was only picked up in the 6.2.y stable branch where these assertions are
> > > > > performed prior to determining mergeability.
> > > > >
> > > > > Had this remained the same in mainline this issue may have been picked up
> > > > > faster, so moving forward let's always check them.
> > > > >
> > > > > Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com>
> > > > > ---
> > > > >  mm/mmap.c | 10 +++++-----
> > > > >  1 file changed, 5 insertions(+), 5 deletions(-)
> > > > >
> > > > > diff --git a/mm/mmap.c b/mm/mmap.c
> > > > > index 5522130ae606..13678edaa22c 100644
> > > > > --- a/mm/mmap.c
> > > > > +++ b/mm/mmap.c
> > > > > @@ -960,17 +960,17 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> > > > >  		merge_next = true;
> > > > >  	}
> > > > >
> > > > > +	/* Verify some invariant that must be enforced by the caller. */
> > > > > +	VM_WARN_ON(prev && addr <= prev->vm_start);
> > > > > +	VM_WARN_ON(curr && (addr != curr->vm_start || end > curr->vm_end));
> > > > > +	VM_WARN_ON(addr >= end);
> > > > > +
> > > >
> > > > I'm seeing this fire a lot when fuzzing v6.4-rc1 on arm64 using Syzkaller.
> > > >
> > >
> > > Thanks, from the line I suspect addr != curr->vm_start, but need to look
> > > into the repro, at lsf/mm so a bit time lagged :)
> >
> > No problem; FWIW I can confirm your theory, the reproducer is causing:
> >
> > 	addr > curr->vm_start
> >
> > ... confirmed the the following hack, log below.
>
> Awesome thanks for that! Just been firing up qemu to do this.
>
> Cases 5-8 should really have addr == curr->vm_start, I wonder if it's
> another case but curr is being set incorrectly, it should in theory not be
> the case.
>
> (See [1] for a visualisation of merge cases as a handy reference)
>
> Of course userfaultfd might be the offender here and might be relying on no
> merge case arising but passing dodgy parameters.
>
> [1]:https://ljs.io/merge_cases.png
>
> >
> > | diff --git a/mm/mmap.c b/mm/mmap.c
> > | index 13678edaa22c..2cdebba15719 100644
> > | --- a/mm/mmap.c
> > | +++ b/mm/mmap.c
> > | @@ -961,9 +961,21 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> > |         }
> > |
> > |         /* Verify some invariant that must be enforced by the caller. */
> > | -       VM_WARN_ON(prev && addr <= prev->vm_start);
> > | -       VM_WARN_ON(curr && (addr != curr->vm_start || end > curr->vm_end));
> > | -       VM_WARN_ON(addr >= end);
> > | +       VM_WARN(prev && addr <= prev->vm_start,
> > | +               "addr = 0x%016lx, prev->vm_start = 0x%016lx\n",
> > | +               addr, prev->vm_start);
> > | +
> > | +       VM_WARN(curr && addr != curr->vm_start,
> > | +               "addr = 0x%016lx, curr->vm_start = 0x%016lx\n",
> > | +               addr, curr->vm_start);
> > | +
> > | +       VM_WARN(curr && addr > curr->vm_end,
> > | +               "addr = 0x%016lx, curr->vm_end = 0x%016lx\n",
> > | +               addr, curr->vm_end);
> > | +
> > | +       VM_WARN(addr >= end,
> > | +               "addr = 0x%016lx, end = 0x%016lx\n",
> > | +               addr, end);
> > |
> > |         if (!merge_prev && !merge_next)
> > |                 return NULL; /* Not mergeable. */
> >
> > ... with that applied, running the reproducer results in:
> >
> > | addr = 0x0000ffff99dc2000, curr->vm_start = 0x0000ffff99db2000
> > | WARNING: CPU: 0 PID: 163 at mm/mmap.c:968 vma_merge+0x3d4/0x1260
> >
> > ... i.e. addr > curr->vm_start
> >
> > Thanks,
> > Mark.

It looks like userfaultfd_register() is indeed using vma_merge() to
determine whether potentially broken input is mergeable, e.g.:-

		if (vma->vm_start < start) {
			ret = split_vma(&vmi, vma, start, 1);
			if (ret)
				break;
		}


So it is proactively using this to determine whether the VMA needs
splitting which is... interesting :)

I need to look at this in detail, but I do suspect uffd could do this check
_before_ the vma_merge() (I'd need to double-check this though), or perhaps
vma_merge could un-warn it.

What's worrying here is that having addr > curr->vm_start but end ==
next->vm_start could result in a broken, overlapped merged.

In any case this needs some post-jet lag analysis (happy for you or Liam or
whoever to step in too if you fancy ;)

THe most proximal solution is to just drop the addr != curr->vm_start check
or at least return NULL if that check is failed, but I think it'd be better
to take a step back and examing the uffd code in detail because the
approach seems like it might have a broken edge case.

I think this check has more so brought out an issue rather than caused one,
but until I can give a post-lsf/mm, post-jet lag analysis I can't be
conclusively coherent on this.

Instinct says we should probably change uffd but might be wrong!