On Fri, 2017-07-14 at 18:31 +1000, Benjamin Herrenschmidt wrote:
> There are a number of conditions in the DSISR that represent
> conditions for which there is no point looking for a VMA or
> trying to update a PTE. In fact, POWER9 adds a few with bad
> AMOs, bad "paste" instruction etc...
> 
> This improve our existing code to use symbolic constants for
> all DSISR bits, properly avoid hash_page() for the new bits,
> and handle both the old and new bits early in do_page_fault()
> 
> Signed-off-by: Benjamin Herrenschmidt <[email protected]>
> --
>

Looks reasonable

Acked-by: Balbir Singh <[email protected]> 

Reply via email to