On 01.03.23 08:03, Christophe Leroy wrote:
Le 27/02/2023 à 23:29, Rick Edgecombe a écrit :
The x86 Control-flow Enforcement Technology (CET) feature includes a new
type of memory called shadow stack. This shadow stack memory has some
unusual properties, which requires some core mm changes to function
properly.
One of these unusual properties is that shadow stack memory is writable,
but only in limited ways. These limits are applied via a specific PTE
bit combination. Nevertheless, the memory is writable, and core mm code
will need to apply the writable permissions in the typical paths that
call pte_mkwrite().
In addition to VM_WRITE, the shadow stack VMA's will have a flag denoting
that they are special shadow stack flavor of writable memory. So make
pte_mkwrite() take a VMA, so that the x86 implementation of it can know to
create regular writable memory or shadow stack memory.
Apply the same changes for pmd_mkwrite() and huge_pte_mkwrite().
I'm not sure it is a good idea to add a second argument to
pte_mkwrite(). All pte_mkxxxx() only take a pte and nothing else.
We touched on this in previous revisions and so far there was no strong
push back. This turned out to be cleaner and easier than the
alternatives we evaluated.
pte_modify(), for example, takes another argument. Sure, we could try
thinking about passing something else than a VMA to identify the
writability type, but I am not convinced that will look particularly better.
I think you should do the same as commit d9ed9faac283 ("mm: add new
arch_make_huge_pte() method for tile support")
We already have 3 architectures intending to support shadow stacks in
one way or the other. Replacing all pte_mkwrite() with
arch_pte_mkwrite() doesn't sound particularly appealing to me.
--
Thanks,
David / dhildenb
