Virtualization - Jul 2021 - [PATCH v9 13/17] vdpa: factor out vhost_vdpa_pa_map() and vhost_vdpa_pa_unmap()

? 2021/7/14 ??4:05, Dan Carpenter ??:
> On Wed, Jul 14, 2021 at 10:14:32AM +0800, Jason Wang wrote:
>> ? 2021/7/13 ??7:31, Dan Carpenter ??:
>>> On Tue, Jul 13, 2021 at 04:46:52PM +0800, Xie Yongji wrote:
>>>> @@ -613,37 +618,28 @@ static void vhost_vdpa_unmap(struct
vhost_vdpa *v, u64 iova, u64 size)
>>>>    	}
>>>>    }
>>>> -static int vhost_vdpa_process_iotlb_update(struct vhost_vdpa
*v,
>>>> -					   struct vhost_iotlb_msg *msg)
>>>> +static int vhost_vdpa_pa_map(struct vhost_vdpa *v,
>>>> +			     u64 iova, u64 size, u64 uaddr, u32 perm)
>>>>    {
>>>>    	struct vhost_dev *dev = &v->vdev;
>>>> -	struct vhost_iotlb *iotlb = dev->iotlb;
>>>>    	struct page **page_list;
>>>>    	unsigned long list_size = PAGE_SIZE / sizeof(struct page
*);
>>>>    	unsigned int gup_flags = FOLL_LONGTERM;
>>>>    	unsigned long npages, cur_base, map_pfn, last_pfn = 0;
>>>>    	unsigned long lock_limit, sz2pin, nchunks, i;
>>>> -	u64 iova = msg->iova;
>>>> +	u64 start = iova;
>>>>    	long pinned;
>>>>    	int ret = 0;
>>>> -	if (msg->iova < v->range.first ||
>>>> -	    msg->iova + msg->size - 1 > v->range.last)
>>>> -		return -EINVAL;
>>> This is not related to your patch, but can the "msg->iova +
msg->size"
>>> addition can have an integer overflow.  From looking at the callers
it
>>> seems like it can.  msg comes from:
>>>     vhost_chr_write_iter()
>>>     --> dev->msg_handler(dev, &msg);
>>>         --> vhost_vdpa_process_iotlb_msg()
>>>            --> vhost_vdpa_process_iotlb_update()
>>
>> Yes.
>>
>>
>>> If I'm thinking of the right thing then these are allowed to
overflow to
>>> 0 because of the " - 1" but not further than that.  I
believe the check
>>> needs to be something like:
>>>
>>> 	if (msg->iova < v->range.first ||
>>> 	    msg->iova - 1 > U64_MAX - msg->size ||
>>
>> I guess we don't need - 1 here?
> The - 1 is important.  The highest address is 0xffffffff.  So it goes
> start + size = 0 and then start + size - 1 == 0xffffffff.

Right, so actually

msg->iova = 0xfffffffe, msg->size=2 is valid.

Thanks

>
> I guess we could move the - 1 to the other side?
>
> 	msg->iova > U64_MAX - msg->size + 1 ||
>
> regards,
> dan carpenter
>
>

On Wed, Jul 14, 2021 at 05:41:54PM +0800, Jason Wang
wrote:
> 
> ? 2021/7/14 ??4:05, Dan Carpenter ??:
> > On Wed, Jul 14, 2021 at 10:14:32AM +0800, Jason Wang wrote:
> > > ? 2021/7/13 ??7:31, Dan Carpenter ??:
> > > > On Tue, Jul 13, 2021 at 04:46:52PM +0800, Xie Yongji wrote:
> > > > > @@ -613,37 +618,28 @@ static void
vhost_vdpa_unmap(struct vhost_vdpa *v, u64 iova, u64 size)
> > > > >    	}
> > > > >    }
> > > > > -static int vhost_vdpa_process_iotlb_update(struct
vhost_vdpa *v,
> > > > > -					   struct vhost_iotlb_msg *msg)
> > > > > +static int vhost_vdpa_pa_map(struct vhost_vdpa *v,
> > > > > +			     u64 iova, u64 size, u64 uaddr, u32 perm)
> > > > >    {
> > > > >    	struct vhost_dev *dev = &v->vdev;
> > > > > -	struct vhost_iotlb *iotlb = dev->iotlb;
> > > > >    	struct page **page_list;
> > > > >    	unsigned long list_size = PAGE_SIZE / sizeof(struct
page *);
> > > > >    	unsigned int gup_flags = FOLL_LONGTERM;
> > > > >    	unsigned long npages, cur_base, map_pfn, last_pfn =
0;
> > > > >    	unsigned long lock_limit, sz2pin, nchunks, i;
> > > > > -	u64 iova = msg->iova;
> > > > > +	u64 start = iova;
> > > > >    	long pinned;
> > > > >    	int ret = 0;
> > > > > -	if (msg->iova < v->range.first ||
> > > > > -	    msg->iova + msg->size - 1 >
v->range.last)
> > > > > -		return -EINVAL;
> > > > This is not related to your patch, but can the
"msg->iova + msg->size"
> > > > addition can have an integer overflow.  From looking at the
callers it
> > > > seems like it can.  msg comes from:
> > > >     vhost_chr_write_iter()
> > > >     --> dev->msg_handler(dev, &msg);
> > > >         --> vhost_vdpa_process_iotlb_msg()
> > > >            --> vhost_vdpa_process_iotlb_update()
> > > 
> > > Yes.
> > > 
> > > 
> > > > If I'm thinking of the right thing then these are
allowed to overflow to
> > > > 0 because of the " - 1" but not further than that.
I believe the check
> > > > needs to be something like:
> > > > 
> > > > 	if (msg->iova < v->range.first ||
> > > > 	    msg->iova - 1 > U64_MAX - msg->size ||
> > > 
> > > I guess we don't need - 1 here?
> > The - 1 is important.  The highest address is 0xffffffff.  So it goes
> > start + size = 0 and then start + size - 1 == 0xffffffff.
> 
> 
> Right, so actually
> 
> msg->iova = 0xfffffffe, msg->size=2 is valid.
I believe so, yes.  It's inclusive of 0xfffffffe and 0xffffffff.
(Not an expert).

regards,
dan carpenter