diff mbox series

filemap: replace pte_offset_map() with pte_offset_map_nolock()

Message ID 20240313012913.2395414-1-zhangpeng362@huawei.com (mailing list archive)
State New
Headers show
Series filemap: replace pte_offset_map() with pte_offset_map_nolock() | expand

Commit Message

Peng Zhang March 13, 2024, 1:29 a.m. UTC 
From: ZhangPeng <zhangpeng362@huawei.com>

The vmf->ptl in filemap_fault_recheck_pte_none() is still set from
handle_pte_fault(). But at the same time, we did a pte_unmap(vmf->pte).
After a pte_unmap(vmf->pte) unmap and rcu_read_unlock(), the page table
may be racily changed and vmf->ptl maybe fails to protect the actual
page table.
Fix this by replacing pte_offset_map() with pte_offset_map_nolock().

Fixes: 58f327f2ce80 ("filemap: avoid unnecessary major faults in filemap_fault()")
Suggested-by: David Hildenbrand <david@redhat.com>
Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
 mm/filemap.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Comments

Andrew Morton June 25, 2024, 9:06 p.m. UTC | #1 
On Wed, 13 Mar 2024 09:29:13 +0800 Peng Zhang <zhangpeng362@huawei.com> wrote:

> From: ZhangPeng <zhangpeng362@huawei.com>
> 
> The vmf->ptl in filemap_fault_recheck_pte_none() is still set from
> handle_pte_fault(). But at the same time, we did a pte_unmap(vmf->pte).
> After a pte_unmap(vmf->pte) unmap and rcu_read_unlock(), the page table
> may be racily changed and vmf->ptl maybe fails to protect the actual
> page table.
> Fix this by replacing pte_offset_map() with pte_offset_map_nolock().
> 
> ...
>
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -3207,7 +3207,8 @@ static vm_fault_t filemap_fault_recheck_pte_none(struct vm_fault *vmf)
>  	if (!(vmf->flags & FAULT_FLAG_ORIG_PTE_VALID))
>  		return 0;
>  
> -	ptep = pte_offset_map(vmf->pmd, vmf->address);
> +	ptep = pte_offset_map_nolock(vma->vm_mm, vmf->pmd, vmf->address,
> +				     &vmf->ptl);
>  	if (unlikely(!ptep))
>  		return VM_FAULT_NOPAGE;
>  

whoops, I'm still sitting on this because I didn't know whether we
should backport it.

And...  guess what I say next.  Can we please describe what are the
userspace visible effects of the bug?
Andrew Morton July 4, 2024, 6:40 p.m. UTC | #2 
On Tue, 25 Jun 2024 14:06:43 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:

> On Wed, 13 Mar 2024 09:29:13 +0800 Peng Zhang <zhangpeng362@huawei.com> wrote:
> 
> > From: ZhangPeng <zhangpeng362@huawei.com>
> > 
> > The vmf->ptl in filemap_fault_recheck_pte_none() is still set from
> > handle_pte_fault(). But at the same time, we did a pte_unmap(vmf->pte).
> > After a pte_unmap(vmf->pte) unmap and rcu_read_unlock(), the page table
> > may be racily changed and vmf->ptl maybe fails to protect the actual
> > page table.
> > Fix this by replacing pte_offset_map() with pte_offset_map_nolock().
> > 
> > ...
> >
> > --- a/mm/filemap.c
> > +++ b/mm/filemap.c
> > @@ -3207,7 +3207,8 @@ static vm_fault_t filemap_fault_recheck_pte_none(struct vm_fault *vmf)
> >  	if (!(vmf->flags & FAULT_FLAG_ORIG_PTE_VALID))
> >  		return 0;
> >  
> > -	ptep = pte_offset_map(vmf->pmd, vmf->address);
> > +	ptep = pte_offset_map_nolock(vma->vm_mm, vmf->pmd, vmf->address,
> > +				     &vmf->ptl);
> >  	if (unlikely(!ptep))
> >  		return VM_FAULT_NOPAGE;
> >  
> 
> whoops, I'm still sitting on this because I didn't know whether we
> should backport it.
> 
> And...  guess what I say next.  Can we please describe what are the
> userspace visible effects of the bug?
> 

Nobody?

Oh well, I'll add cc:stable amd move this into mm-hotfixes.


From: ZhangPeng <zhangpeng362@huawei.com>
Subject: filemap: replace pte_offset_map() with pte_offset_map_nolock()
Date: Wed, 13 Mar 2024 09:29:13 +0800

The vmf->ptl in filemap_fault_recheck_pte_none() is still set from
handle_pte_fault().  But at the same time, we did a pte_unmap(vmf->pte). 
After a pte_unmap(vmf->pte) unmap and rcu_read_unlock(), the page table
may be racily changed and vmf->ptl maybe fails to protect the actual page
table.  Fix this by replacing pte_offset_map() with
pte_offset_map_nolock().

Link: https://lkml.kernel.org/r/20240313012913.2395414-1-zhangpeng362@huawei.com
Fixes: 58f327f2ce80 ("filemap: avoid unnecessary major faults in filemap_fault()")
Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
Suggested-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Nanyong Sun <sunnanyong@huawei.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Yin Fengwei <fengwei.yin@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 mm/filemap.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

--- a/mm/filemap.c~filemap-replace-pte_offset_map-with-pte_offset_map_nolock
+++ a/mm/filemap.c
@@ -3231,7 +3231,8 @@ static vm_fault_t filemap_fault_recheck_
 	if (!(vmf->flags & FAULT_FLAG_ORIG_PTE_VALID))
 		return 0;
 
-	ptep = pte_offset_map(vmf->pmd, vmf->address);
+	ptep = pte_offset_map_nolock(vma->vm_mm, vmf->pmd, vmf->address,
+				     &vmf->ptl);
 	if (unlikely(!ptep))
 		return VM_FAULT_NOPAGE;
David Hildenbrand July 4, 2024, 7:05 p.m. UTC | #3 
On 04.07.24 20:40, Andrew Morton wrote:
> On Tue, 25 Jun 2024 14:06:43 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:
> 
>> On Wed, 13 Mar 2024 09:29:13 +0800 Peng Zhang <zhangpeng362@huawei.com> wrote:
>>
>>> From: ZhangPeng <zhangpeng362@huawei.com>
>>>
>>> The vmf->ptl in filemap_fault_recheck_pte_none() is still set from
>>> handle_pte_fault(). But at the same time, we did a pte_unmap(vmf->pte).
>>> After a pte_unmap(vmf->pte) unmap and rcu_read_unlock(), the page table
>>> may be racily changed and vmf->ptl maybe fails to protect the actual
>>> page table.
>>> Fix this by replacing pte_offset_map() with pte_offset_map_nolock().
>>>
>>> ...
>>>
>>> --- a/mm/filemap.c
>>> +++ b/mm/filemap.c
>>> @@ -3207,7 +3207,8 @@ static vm_fault_t filemap_fault_recheck_pte_none(struct vm_fault *vmf)
>>>   	if (!(vmf->flags & FAULT_FLAG_ORIG_PTE_VALID))
>>>   		return 0;
>>>   
>>> -	ptep = pte_offset_map(vmf->pmd, vmf->address);
>>> +	ptep = pte_offset_map_nolock(vma->vm_mm, vmf->pmd, vmf->address,
>>> +				     &vmf->ptl);
>>>   	if (unlikely(!ptep))
>>>   		return VM_FAULT_NOPAGE;
>>>   
>>
>> whoops, I'm still sitting on this because I didn't know whether we
>> should backport it.
>>
>> And...  guess what I say next.  Can we please describe what are the
>> userspace visible effects of the bug?
>>
> 
> Nobody?
> 
> Oh well, I'll add cc:stable amd move this into mm-hotfixes.

Yeah, we should better do that. IIRC, the PTL pointer might be stale 
(use after free?), although I cannot judge how often that should 
actually happen.
Peng Zhang July 6, 2024, 9:15 a.m. UTC | #4 
On 2024/7/5 2:40, Andrew Morton wrote:

> On Tue, 25 Jun 2024 14:06:43 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:
>
>> On Wed, 13 Mar 2024 09:29:13 +0800 Peng Zhang <zhangpeng362@huawei.com> wrote:
>>
>>> From: ZhangPeng <zhangpeng362@huawei.com>
>>>
>>> The vmf->ptl in filemap_fault_recheck_pte_none() is still set from
>>> handle_pte_fault(). But at the same time, we did a pte_unmap(vmf->pte).
>>> After a pte_unmap(vmf->pte) unmap and rcu_read_unlock(), the page table
>>> may be racily changed and vmf->ptl maybe fails to protect the actual
>>> page table.
>>> Fix this by replacing pte_offset_map() with pte_offset_map_nolock().
>>>
>>> ...
>>>
>>> --- a/mm/filemap.c
>>> +++ b/mm/filemap.c
>>> @@ -3207,7 +3207,8 @@ static vm_fault_t filemap_fault_recheck_pte_none(struct vm_fault *vmf)
>>>   	if (!(vmf->flags & FAULT_FLAG_ORIG_PTE_VALID))
>>>   		return 0;
>>>   
>>> -	ptep = pte_offset_map(vmf->pmd, vmf->address);
>>> +	ptep = pte_offset_map_nolock(vma->vm_mm, vmf->pmd, vmf->address,
>>> +				     &vmf->ptl);
>>>   	if (unlikely(!ptep))
>>>   		return VM_FAULT_NOPAGE;
>>>   
>> whoops, I'm still sitting on this because I didn't know whether we
>> should backport it.
>>
>> And...  guess what I say next.  Can we please describe what are the
>> userspace visible effects of the bug?
>>
> Nobody?
>
> Oh well, I'll add cc:stable amd move this into mm-hotfixes.
>
Sorry for the late reply.

As David said, the PTL pointer might be stale so if we continue to use it infilemap_fault_recheck_pte_none(), it might trigger UAF.
Also, if the PTL fails, the issue fixed by commit 58f327f2ce80 ("filemap: avoid unnecessary major faults in filemap_fault()") might reappear.
diff mbox series

Patch

diff --git a/mm/filemap.c b/mm/filemap.c
index 31ab455c4537..222adac7c9c5 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -3207,7 +3207,8 @@  static vm_fault_t filemap_fault_recheck_pte_none(struct vm_fault *vmf)
 	if (!(vmf->flags & FAULT_FLAG_ORIG_PTE_VALID))
 		return 0;
 
-	ptep = pte_offset_map(vmf->pmd, vmf->address);
+	ptep = pte_offset_map_nolock(vma->vm_mm, vmf->pmd, vmf->address,
+				     &vmf->ptl);
 	if (unlikely(!ptep))
 		return VM_FAULT_NOPAGE;