Re: [PATCH 6/6] Revert "arch: introduce set_direct_map_valid_noflush()"

On Sun Aug 16, 2026 at 12:59 PM CEST, Mike Rapoport (Microsoft) wrote:
> Commit 0c6378a71574 ("arch: introduce set_direct_map_valid_noflush()")
> added set_direct_map_valid_noflush() to allow updating the direct map
> for a physically contiguous range in execmem.
>
> As Brendan recently pointed out [1], this API is confusing because on
> arm64 it means that is sets VALID bit in ptes, while on other
> architectures it is an analog of set_direct_map_default_noflush().
>
> The only user of set_direct_map_valid_noflush() was execmem's ROX cache
> freeing path and it was switched to utilize VM_FLUSH_RESET_PERMS for
> resetting permissions of the direct map alias.
>
> With the last user gone and with set_direct_map_{invalid,default}_noflush()
> accepting number of pages as a parameter, set_direct_map_valid_noflush()
> become a copy of set_memory_valid() on arm64 and a duplicate of
> set_direct_map_{invalid,default}_noflush() on other architecture, it is
> safe to remove set_direct_map_valid_noflush().
>
> Also drop a stale comment in arm64::__kernel_map_pages() that Linus
> bothered to add when merging changes containing set_direct_map_valid_noflush()
> to his tree.
>
> This reverts commit 0c6378a71574daa6cd1534ad42a956e3262756c7.
>
> [1] https://lore.kernel.org/all/[email protected]
>
> Signed-off-by: Mike Rapoport (Microsoft) <[email protected]>

Quick dump of my understanding (ignoring the NG bit on arm64)

- set_direct_map_invalid_noflush():

    x86: clear P and RW (and DIRTY)
    arm64: clear VALID

  So these look out of sync to me

- set_direct_map_default_noflush():

    x86: set P and RW
    arm64: set VALID and WRITE, clear RDONLY

- set_direct_map_valid_noflush(..., true):

    x86: exactly the same as set_direct_map_default_noflush()
         (but with a `nr` arg)
    arm64: set VALID

- set_direct_map_valid_noflush(..., false):

    both archs: exactly the same as set_direct_map_invalid_noflush()
         (but with a `nr` arg)

So basically the big issue here is specifically that
set_direct_map_valid_noflush(..., true) is special on arm64 and not x86.

And we fix that by jut deleting the API. SGTM!

The other issue I can see here is that set_direct_map_invalid_noflush()
clears RW on x86 but doesn't set RDONLY on arm64. So if you unmap
something using set_direct_map_invalid_noflush(), then map it again
using something other than set_direct_map_default_noflush(), you get
different behaviour between the archs.

I think the answer to that is probably: doing that is a bug, i.e.
_invalid_noflush() and _default_noflush() are a pair that you have to
use together. But I haven't checked if this is currently the case. Maybe
it would still make sense to just align these fully?

Anyway, aside from all this yapping, getting rid of _valid_noflush()
seems like an unambiguous win here so thanks for the cleanup!

Reviewed-by: Brendan Jackman <[email protected]>