Re: [PATCH v15 08/26] x86/mm: Introduce _PAGE_COW

[Yu, Yu-cheng]

On 12/10/2020 9:41 AM, Borislav Petkov wrote:
> On Tue, Dec 08, 2020 at 11:24:16AM -0800, Yu, Yu-cheng wrote:
> > Case (a) is a normal writable data page that has gone through fork(). So it
>
> Writable >
> > has W=0, D=1.  But here, the software chooses not to use the D bit, and
>
> But it has W=0. So not writable?

Maybe I will change to: A page in a writable vma, has been modified and 
gone through fork().

> > instead, W=0, COW=1.
>
> So the "new" way of denoting that the page is modified is COW=1
> *when* on CET hw. The D=1 bit is still used on the rest thus the two
> _PAGE_DIRTY_BITS.
>
> Am I close?

COW=1 is only used in copy-on-write situation (when CET is enabled).  If 
W=1, D bit is used.

> > Case (b) is a normal read-only data page.  Since it is read-only, fork()
> > won't affect it.  In __get_user_pages(), a copy of the read-only page is
> > needed, and the page is duplicated.  The software sets COW=1 for the new
> > copy.
>
> That makes more sense.
>
> > Thread-A is writing to a writable page, and the page's PTE is becoming W=1,
> > D=1.  In the middle of it, Thread-B is changing the PTE to W=0.
>
> Yah, add that to the explanation pls.

Sure.