2023-10-06 18:47:32

by Deepak R Varma

[permalink] [raw]
Subject: kmap() transformation: Question about copy_user_highpage

Hello,
I am attempting to work on replacing the kmap[_atomic]() calls by
kmap_local_page() function call. A detail on this change can be found here [0].

I would like to know if this function:
arch/loongarch/mm/init.c::copy_user_highpage()
is currently in use or is it a dead code?

If this code is not in use, can this be removed instead? However, if it is in
use, can you comment why kmap_atomic() was preferred over kmap()
function call?

[0] https://lore.kernel.org/all/[email protected]/

Thank you,
Deepak.



2023-10-07 04:56:43

by Huacai Chen

[permalink] [raw]
Subject: Re: kmap() transformation: Question about copy_user_highpage

Hi, Deepak,

On Sat, Oct 7, 2023 at 2:46 AM Deepak R Varma <[email protected]> wrote:
>
> Hello,
> I am attempting to work on replacing the kmap[_atomic]() calls by
> kmap_local_page() function call. A detail on this change can be found here [0].
>
> I would like to know if this function:
> arch/loongarch/mm/init.c::copy_user_highpage()
> is currently in use or is it a dead code?
>
> If this code is not in use, can this be removed instead? However, if it is in
> use, can you comment why kmap_atomic() was preferred over kmap()
> function call?
I think this function will be used for the 32bit kernel, and we will
add loongarch32 support in the near future.

But on the other hand, you can replace kmap_atomic() with
kmap_local_page(), there is no special reason to use kmap_atomic().

Huacai

>
> [0] https://lore.kernel.org/all/[email protected]/
>
> Thank you,
> Deepak.
>
>
>

2023-10-07 13:02:18

by Deepak R Varma

[permalink] [raw]
Subject: Re: kmap() transformation: Question about copy_user_highpage

On Sat, Oct 07, 2023 at 12:55:14PM +0800, Huacai Chen wrote:
> Hi, Deepak,

Hello Huacai,

>
> On Sat, Oct 7, 2023 at 2:46 AM Deepak R Varma <[email protected]> wrote:
> >
> > Hello,
> > I am attempting to work on replacing the kmap[_atomic]() calls by
> > kmap_local_page() function call. A detail on this change can be found here [0].
> >
> > I would like to know if this function:
> > arch/loongarch/mm/init.c::copy_user_highpage()
> > is currently in use or is it a dead code?
> >
> > If this code is not in use, can this be removed instead? However, if it is in
> > use, can you comment why kmap_atomic() was preferred over kmap()
> > function call?
> I think this function will be used for the 32bit kernel, and we will
> add loongarch32 support in the near future.

Okay. Sounds good.

>
> But on the other hand, you can replace kmap_atomic() with
> kmap_local_page(), there is no special reason to use kmap_atomic().

Looks like you sent in a patch. Thank you. I was working on trying to cross
compile my patch for the arch. I still would like to attempt that. Will request
your help if I get stuck.

regards,
deepak.


>
> Huacai
>
> >
> > [0] https://lore.kernel.org/all/[email protected]/
> >
> > Thank you,
> > Deepak.
> >
> >
> >