Re: [PATCH 08/14] zram: check that backends array has at least one backend

Re: [PATCH 08/14] zram: check that backends array has at least one backend

[Sergey Senozhatsky]

On (24/05/05 14:13), Sergey Senozhatsky wrote:
> On (24/05/05 13:39), Sergey Senozhatsky wrote:
> [..]
> > > I guess just pick one if none were selected.
> 
> How do I pick one if none were selected? Does Kconfig support
> something like that?

This triggers Kconfig error:

config ZRAM_EMPTY_BACKENDS_FIXUP
       bool
       depends on ZRAM && !ZRAM_BACKEND_LZO && !ZRAM_BACKEND_LZ4 && \
               !ZRAM_BACKEND_LZ4HC && !ZRAM_BACKEND_ZSTD && \
               !ZRAM_BACKEND_DEFLATE
       select ZRAM_BACKEND_LZO


drivers/block/zram/Kconfig:17:error: recursive dependency detected!
drivers/block/zram/Kconfig:17:  symbol ZRAM_BACKEND_LZO is selected by ZRAM_EMPTY_BACKENDS_FIXUP
drivers/block/zram/Kconfig:52:  symbol ZRAM_EMPTY_BACKENDS_FIXUP depends on ZRAM_BACKEND_LZO


I'm a little surprised by this - EMPTY_BACKENDS_FIXUP does not depend
on ZRAM_BACKEND_LZO, it depends on NOT ZRAM_BACKEND_LZO.

Let me Cc linux-kbuild. Kbuild folks, how do I workaround this?

Re: [PATCH 08/14] zram: check that backends array has at least one backend

[Sergey Senozhatsky]

On (24/05/05 15:48), Sergey Senozhatsky wrote:
> On (24/05/05 14:13), Sergey Senozhatsky wrote:
> > On (24/05/05 13:39), Sergey Senozhatsky wrote:
> > [..]
> > > > I guess just pick one if none were selected.
> > 
> > How do I pick one if none were selected? Does Kconfig support
> > something like that?
> 
> This triggers Kconfig error:
> 
> config ZRAM_EMPTY_BACKENDS_FIXUP
>        bool
>        depends on ZRAM && !ZRAM_BACKEND_LZO && !ZRAM_BACKEND_LZ4 && \
>                !ZRAM_BACKEND_LZ4HC && !ZRAM_BACKEND_ZSTD && \
>                !ZRAM_BACKEND_DEFLATE
>        select ZRAM_BACKEND_LZO
> 
> 
> drivers/block/zram/Kconfig:17:error: recursive dependency detected!
> drivers/block/zram/Kconfig:17:  symbol ZRAM_BACKEND_LZO is selected by ZRAM_EMPTY_BACKENDS_FIXUP
> drivers/block/zram/Kconfig:52:  symbol ZRAM_EMPTY_BACKENDS_FIXUP depends on ZRAM_BACKEND_LZO
> 
> 
> I'm a little surprised by this - EMPTY_BACKENDS_FIXUP does not depend
> on ZRAM_BACKEND_LZO, it depends on NOT ZRAM_BACKEND_LZO.
> 
> Let me Cc linux-kbuild. Kbuild folks, how do I workaround this?

Is this how one does it?

config ZRAM_BACKEND_LZO
       bool "lzo and lzo-rle compression support"
       depends on ZRAM
       default y if !ZRAM_BACKEND_LZ4 && !ZRAM_BACKEND_LZ4HC && \
               !ZRAM_BACKEND_ZSTD && !ZRAM_BACKEND_DEFLATE
       default n
       select LZO_COMPRESS
       select LZO_DECOMPRESS


User still can select N and then we'll have empty backends, but
at least default is Y if none of the algorithms were selected.
Is it good enough?

Re: [PATCH 08/14] zram: check that backends array has at least one backend

[Masahiro Yamada]

On Mon, May 6, 2024 at 12:22 PM Sergey Senozhatsky
<senozhatsky@chromium.org> wrote:
>
> On (24/05/05 15:48), Sergey Senozhatsky wrote:
> > On (24/05/05 14:13), Sergey Senozhatsky wrote:
> > > On (24/05/05 13:39), Sergey Senozhatsky wrote:
> > > [..]
> > > > > I guess just pick one if none were selected.
> > >
> > > How do I pick one if none were selected? Does Kconfig support
> > > something like that?
> >
> > This triggers Kconfig error:
> >
> > config ZRAM_EMPTY_BACKENDS_FIXUP
> >        bool
> >        depends on ZRAM && !ZRAM_BACKEND_LZO && !ZRAM_BACKEND_LZ4 && \
> >                !ZRAM_BACKEND_LZ4HC && !ZRAM_BACKEND_ZSTD && \
> >                !ZRAM_BACKEND_DEFLATE
> >        select ZRAM_BACKEND_LZO
> >
> >
> > drivers/block/zram/Kconfig:17:error: recursive dependency detected!
> > drivers/block/zram/Kconfig:17:  symbol ZRAM_BACKEND_LZO is selected by ZRAM_EMPTY_BACKENDS_FIXUP
> > drivers/block/zram/Kconfig:52:  symbol ZRAM_EMPTY_BACKENDS_FIXUP depends on ZRAM_BACKEND_LZO
> >
> >
> > I'm a little surprised by this - EMPTY_BACKENDS_FIXUP does not depend
> > on ZRAM_BACKEND_LZO, it depends on NOT ZRAM_BACKEND_LZO.
> >
> > Let me Cc linux-kbuild. Kbuild folks, how do I workaround this?
>
> Is this how one does it?
>
> config ZRAM_BACKEND_LZO
>        bool "lzo and lzo-rle compression support"
>        depends on ZRAM
>        default y if !ZRAM_BACKEND_LZ4 && !ZRAM_BACKEND_LZ4HC && \
>                !ZRAM_BACKEND_ZSTD && !ZRAM_BACKEND_DEFLATE
>        default n
>        select LZO_COMPRESS
>        select LZO_DECOMPRESS
>
>
> User still can select N and then we'll have empty backends, but
> at least default is Y if none of the algorithms were selected.
> Is it good enough?


I interpret this sentence into:

"randconfig will eventually disable all ZRAM_BACKEND_*,
causing the build error again.
Is it good enough, Arnd?"



Some possible solutions:



config ZRAM_BACKEND_FORCE_LZO
        def_bool !ZRAM_BACKEND_LZ4 && !ZRAM_BACKEND_LZ4HC && \
                 !ZRAM_BACKEND_ZSTD && !ZRAM_BACKEND_DEFLATE
        depends on ZRAM
        select ZRAM_BACKEND_LZO

config ZRAM_BACKEND_LZO
        bool "lzo and lzo-rle compression support"
        depends on ZRAM
        select LZO_COMPRESS
        select LZO_DECOMPRESS


OR


config ZRAM_BACKEND_FORCE_LZO
        def_bool !ZRAM_BACKEND_LZ4 && !ZRAM_BACKEND_LZ4HC && \
                 !ZRAM_BACKEND_ZSTD && !ZRAM_BACKEND_DEFLATE

config ZRAM_BACKEND_LZO
        bool "lzo and lzo-rle compression support" if !ZRAM_BACKEND_FORCE_LZO
        depends on ZRAM
        default ZRAM_BACKEND_FORCE_LZO
        select LZO_COMPRESS
        select LZO_DECOMPRESS




BTW, "default n" you are adding are redundant.

-- 
Best Regards
Masahiro Yamada

Re: [PATCH 08/14] zram: check that backends array has at least one backend

[Sergey Senozhatsky]

On (24/05/06 14:20), Masahiro Yamada wrote:
> config ZRAM_BACKEND_FORCE_LZO
>         def_bool !ZRAM_BACKEND_LZ4 && !ZRAM_BACKEND_LZ4HC && \
>                  !ZRAM_BACKEND_ZSTD && !ZRAM_BACKEND_DEFLATE
> 
> config ZRAM_BACKEND_LZO
>         bool "lzo and lzo-rle compression support" if !ZRAM_BACKEND_FORCE_LZO
>         depends on ZRAM
>         default ZRAM_BACKEND_FORCE_LZO
>         select LZO_COMPRESS
>         select LZO_DECOMPRESS

I'll take this one. Thank you.

> BTW, "default n" you are adding are redundant.

OK.