Re: [PATCH] mm: Speed up mremap on large regions

[Joel Fernandes <joelaf@google.com>]

On Wed, Oct 10, 2018 at 5:46 PM, Joel Fernandes <joel@joelfernandes.org> wrote:
> On Wed, Oct 10, 2018 at 01:00:11PM +0300, Kirill A. Shutemov wrote:
[...]
>>
>> My worry is that some architecture has to allocate page table differently
>> depending on virtual address (due to aliasing or something). Original page
>> table was allocated for one virtual address and moving the page table to
>> different spot in virtual address space may break the invariant.
>>
>> > Also the clean up of the argument that you're proposing is a bit out of scope
>> > of this patch but yeah we could clean it up in a separate patch if needed. I
>> > don't feel too strongly about that. It seems cosmetic and in the future if
>> > the address that's passed in is needed, then the architecture can use it.
>>
>> Please, do. This should be pretty mechanical change, but it will help to
>> make sure that none of obscure architecture will be broken by the change.
>>
>
> The thing is its quite a lot of change, I wrote a coccinelle script to do it
> tree wide, following is the diffstat:
>  48 files changed, 91 insertions(+), 124 deletions(-)
>
> Imagine then having to add the address argument back in the future in case
> its ever needed. Is it really worth doing it? Anyway I confirmed that the
> address is NOT used for anything at the moment so your fears of the
> optimization doing anything wonky really don't exist at the moment. I really
> feel this is unnecessary but I am Ok with others agree the second arg to
> pte_alloc should be removed in light of this change. Andrew, what do you
> think?

I meant to say here, "I am Ok if others agree the second arg to
pte_alloc should be removed", but I would really like some input from
the others as well on what they think.