Re: [PATCH v28 09/32] x86/mm: Introduce _PAGE_COW

[Yu, Yu-cheng]

On 8/17/2021 2:01 PM, Borislav Petkov wrote:
> On Tue, Aug 17, 2021 at 01:51:52PM -0700, Andy Lutomirski wrote:
> > WRSS can be used from user mode depending on the configuration.
>
> My point being, if you're going to do shadow stack management
> operations, you should check whether the target you're writing to is a
> shadow stack page. Clearly userspace can't do that but userspace will
> get notified of that pretty timely.
>
> > Double-you shmouble-you. You can't write it with MOV, but you can
> > write it from user code and from kernel code. As far as the mm is
> > concerned, I think it should be considered writable.
>
> Because?
>
> > Although... anyone who tries to copy_to_user() it is going to be a bit
> > surprised. Hmm.
>
> Ok, so you see the confusion.

copy_to_user() can run into normal read-only areas too.  The caller can 
handle that just fine.

> In any case, I don't think you can simply look at a shadow stack page as
> simple writable page. There are cases where it is going to be fun.
>
> So why are we even saying that a shadow stack page is writable? Why
> can't we simply say that a shadow stack page is, well, something
> special?

We can visualize the type of a mm area by looking at vma->vm_flags, e.g. 
maybe_mkwrite(), and PTE macros as lower-level operatives.  These two 
have some relations but not one-to-one.  Note that a PTE in a writable 
area is not always pte_write().

I have considered and implemented a shadow stack PTE either pte_write() 
or not.  Making shadow stack as pte_write() results in less arch_* 
macros and less confusion in copy-on-write code.  That is one more thing 
to consider.

Thanks,
Yu-cheng