On Thu, Mar 07, 2024 at 12:23:39AM +0100, Richard Weinberger wrote:
> Bit 58 denotes that a PTE is writable.
> The main use case is detecting CoW mappings.
>
> Signed-off-by: Richard Weinberger <richard@xxxxxx>
> ---
> Documentation/admin-guide/mm/pagemap.rst | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/Documentation/admin-guide/mm/pagemap.rst b/Documentation/admin-guide/mm/pagemap.rst
> index f5f065c67615..81ffe3601b96 100644
> --- a/Documentation/admin-guide/mm/pagemap.rst
> +++ b/Documentation/admin-guide/mm/pagemap.rst
> @@ -21,7 +21,8 @@ There are four components to pagemap:
> * Bit 56 page exclusively mapped (since 4.2)
> * Bit 57 pte is uffd-wp write-protected (since 5.13) (see
> Documentation/admin-guide/mm/userfaultfd.rst)
> - * Bits 58-60 zero
> + * Bit 58 pte is writable (since 6.10)
I really think we need to be careful about talking about 'writable' again
because people are easily confused about the difference between a writable
_mapping_ and a writable _page table entry_.
Of course you mention PTE here, but I think it might be better to say
something like:
* Bit 58 raw pte r/w flag (since 6.10)
> + * Bits 59-60 zero
> * Bit 61 page is file-page or shared-anon (since 3.5)
> * Bit 62 page swapped
> * Bit 63 page present
> @@ -37,6 +38,11 @@ There are four components to pagemap:
> precisely which pages are mapped (or in swap) and comparing mapped
> pages between processes.
>
> + Bit 58 is useful to detect CoW mappings; however, it does not indicate
> + whether the page mapping is writable or not. If an anonymous mapping is
> + writable but the write bit is not set, it means that the next write access
> + will cause a page fault, and copy-on-write will happen.
> +
David has addressed the copy vs. anon exclusive remap issue, but I also
feel this needs some balking out.
I would simply rephrase this in terms of whether a write fault occurs or
not e.g.:
Bit 58 indicates whether the PTE has the write flag set. If this flag is
unset, then write accesses for this mapping will cause a fault for this
page. If the mapping is private (whether anonymous or file-backed), this
can result in a Copy-on-Write (though if anonymous-excusive the flag
will simply be set). If file-backed, this being cleared may simply
indicate that this file page is clean.
> Efficient users of this interface will use ``/proc/pid/maps`` to
> determine which areas of memory are actually mapped and llseek to
> skip over unmapped regions.
> --
> 2.35.3
>
