Re: [PATCH v2 3/5] zstd: move ZSTD_MAX_CLEVEL to zstd_lib.h

haoqin huang <[email protected]> Wed, 29 Jul 2026 12:32:40 +0800
Newsgroups org.kernel.vger.linux-block,org.kernel.vger.linux-kernel
Message-ID <CAEjiKSmLSrfbtBukdC+fRqDY0tOVnyPtnxBEUyhrvrrwLbaxMA@mail.gmail.com>
On Wed, Jul 29, 2026 at 12:23 PM Sergey Senozhatsky
<[email protected]> wrote:
>
> On (26/07/29 12:16), haoqin huang wrote:
> > > On (26/07/28 17:29), Haoqin Huang wrote:
> > > [..]
> > > >  #define ZSTD_MINMATCH_MAX         7   /* only for ZSTD_fast, other strategies are limited to 6 */
> > > >  #define ZSTD_MINMATCH_MIN         3   /* only for ZSTD_btopt+, faster strategies are limited to 4 */
> > > > +#define ZSTD_MAX_CLEVEL              22
> > > >  #define ZSTD_TARGETLENGTH_MAX    ZSTD_BLOCKSIZE_MAX
> > > >  #define ZSTD_TARGETLENGTH_MIN     0   /* note : comparing this constant to an unsigned results in a tautological test */
> > > >  #define ZSTD_STRATEGY_MIN        ZSTD_fast
> > > > diff --git a/lib/zstd/compress/clevels.h b/lib/zstd/compress/clevels.h
> > > > index 6ab8be6532ef..06565e064456 100644
> > > > --- a/lib/zstd/compress/clevels.h
> > > > +++ b/lib/zstd/compress/clevels.h
> > > > @@ -17,8 +17,6 @@
> > > >
> > > >  /*-=====  Pre-defined compression levels  =====-*/
> > > >
> > > > -#define ZSTD_MAX_CLEVEL     22
> > > > -
> > > >  __attribute__((__unused__))
> > >
> > > Sashiko made a good point.  Can we use zstd_max_clevel() instead?
> >
> > Good point, I'll drop this patch and use zstd_max_clevel() instead
> > in v3.
> >
> > Since it's a runtime function and can't be used for static struct
> > initialization, I plan to set backend_zstd's level_max to -1 as a
> > sentinel value, and query the actual maximum in
> > zcomp_validate_params():
> >
> >     s32 max = backend->level_max;
> >     if (max < 0)
> >         max = zstd_max_clevel();
> >
> > Do you think this approach is feasible?
>
> Hmm, no, that doesn't look good. zcomp should not include
> zstd.h or any other libs directly.  Should params validation
> be a per-backend callback then?

Good point, zcomp.c shouldn't include library headers.  A per-backend
callback would be cleaner.

I'll add an optional validate_params to zcomp_ops: the zstd backend
implements it using zstd_max_clevel() internally, while lzo/deflate
and others just rely on the static caps check (no callback needed).
zcomp.c stays free of any library headers.

I'll send this in v3.