Re: [PATCH] nilfs2: do not use unexported cpu_to_le32()/le32_to_cpu() in uapi header

[Masahiro Yamada]

Konishi-san,


On Sun, Jul 7, 2019 at 3:11 PM Ryusuke Konishi
<konishi.ryusuke@xxxxxxxxxxxxx> wrote:
>
> Yamada-san,
>
> I'm sorry for not being able to follow.
>
> Could you tell me how did you find the build error?
> Still I cannot reproduce the error in my environment.


"You can compile UAPI headers in kernel-space" does not necessarily mean
"you can compile them in user-space".

To make sure UAPI headers can be included from libc etc.
you must compile them after 'make headers_install',
but people often miss to test that.

To automate this testing, I will merge the following for 5.3-rc1:
https://patchwork.kernel.org/patch/11024715/

Once it lands in upstream, all the new breakage
will be caught by 0-day bot.

Currently, nilfs2_ondisk.h is excluded from the test coverage
since it is broken.

Thanks.
Masahiro Yamada



> Or, if it's right that we should not use
> cpu_to_le{16,32,64}/le{16,32,64}_to_cpu() in UAPI  header files,
> I will acknowledge the change and modify nilfs utilities to comply
> with the guideline.
>
> Thanks,
> Ryusuke Konishi
>
> 2019å7æ7æ(æ) 13:31 Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx>:
> >
> > Hi Andrew,
> >
> > On Sun, Jul 7, 2019 at 12:38 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > >
> > > Seems fairly serious.  I'm thinking this needs a cc:stable?
> >
> >
> > Ah, yes.
> >
> >
> > Cc: linux-stable <stable@xxxxxxxxxxxxxxx> # 4.9
> > Fixes: e63e88bc53ba ("nilfs2: move ioctl interface and disk layout to
> > uapi separately")
> >
> >
> > Thanks.
> >
> > --
> > Best Regards
> > Masahiro Yamada



-- 
Best Regards
Masahiro Yamada