Re: [PATCH] xen/gntdev: fix gntdev_mmap() error exit path

[Luca Fancellu <luca.fancellu@arm.com>]

> On 23 Apr 2021, at 08:00, Juergen Gross <jgross@suse.com> wrote:
> 
> On 23.04.21 08:55, Luca Fancellu wrote:
>>> On 23 Apr 2021, at 06:40, Juergen Gross <jgross@suse.com> wrote:
>>> 
>>> Commit d3eeb1d77c5d0af ("xen/gntdev: use mmu_interval_notifier_insert")
>>> introduced an error in gntdev_mmap(): in case the call of
>>> mmu_interval_notifier_insert_locked() fails the exit path should not
>>> call mmu_interval_notifier_remove(), as this might result in NULL
>>> dereferences.
>>> 
>>> One reason for failure is e.g. a signal pending for the running
>>> process.
>>> 
>>> Fixes: d3eeb1d77c5d0af ("xen/gntdev: use mmu_interval_notifier_insert")
>>> Cc: stable@vger.kernel.org
>>> Reported-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
>>> Tested-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
>>> Signed-off-by: Juergen Gross <jgross@suse.com>
>>> ---
>>> drivers/xen/gntdev.c | 4 +++-
>>> 1 file changed, 3 insertions(+), 1 deletion(-)
>>> 
>>> diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
>>> index f01d58c7a042..a3e7be96527d 100644
>>> --- a/drivers/xen/gntdev.c
>>> +++ b/drivers/xen/gntdev.c
>>> @@ -1017,8 +1017,10 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
>>> 		err = mmu_interval_notifier_insert_locked(
>>> 			&map->notifier, vma->vm_mm, vma->vm_start,
>>> 			vma->vm_end - vma->vm_start, &gntdev_mmu_ops);
>>> -		if (err)
>>> +		if (err) {
>>> +			map->vma = NULL;
>>> 			goto out_unlock_put;
>>> +		}
>>> 	}
>>> 	mutex_unlock(&priv->lock);
>>> 
>>> -- 
>>> 2.26.2
>>> 
>>> 
>> Hi Juergen,
>> I can see from the code that there is another path to out_unlock_put label some lines before:
>>         […]
>>         vma->vm_private_data = map;
>> 	if (map->flags) {
>> 		if ((vma->vm_flags & VM_WRITE) &&
>> 				(map->flags & GNTMAP_readonly))
>> 			goto out_unlock_put;
>> 	} else {
>> 		map->flags = GNTMAP_host_map;
>> 		if (!(vma->vm_flags & VM_WRITE))
>> 			map->flags |= GNTMAP_readonly;
>> 	}
>>         […]
>> Can be the case that use_ptemod is != 0 also for that path?
> 
> map->vma will be NULL in this case, so there will be no problem
> resulting from that path.
> 

Right, thanks, seems good to me.

Reviewed-by: Luca Fancellu <luca.fancellu@arm.com>

Cheers,
Luca

> 
> Juergen
> <OpenPGP_0xB0DE9DD628BF132F.asc>