Re: [External] Re: [PATCH] mm: rename the oldflags and parameter in memalloc_flags_*()

[Kent Overstreet <kent.overstreet@linux.dev>]

On Thu, Jun 19, 2025 at 11:17:58AM +0800, Zhongkun He wrote:
> On Thu, Jun 19, 2025 at 8:07 AM Kent Overstreet
> <kent.overstreet@linux.dev> wrote:
> >
> > On Wed, Jun 18, 2025 at 04:43:34PM -0700, Andrew Morton wrote:
> > > On Wed, 18 Jun 2025 15:03:28 +0800 Zhongkun He <hezhongkun.hzk@bytedance.com> wrote:
> > >
> > > > The variable name oldflags can indeed be misleading, because
> > > > it does not store the complete original value of flags.
> > > > Instead, it records which flags from the given set are not
> > > > currently set. So rename it.
> > > >
> > >
> > > Your email client is mangling the patches in strange ways.  Please send
> > > yourself a patch, figure out why it didn't apply?
> > >
> > > > --- a/include/linux/sched/mm.h
> > > > +++ b/include/linux/sched/mm.h
> > > > @@ -322,21 +322,21 @@ static inline void might_alloc(gfp_t gfp_mask)
> > > >  }
> > > >
> > > >  /**
> > > > - * memalloc_flags_save - Add a PF_* flag to current->flags, save old value
> > > > + * memalloc_flags_save - Add a PF_* flag to current->flags, return saved flags mask
> > > >   *
> > > >   * This allows PF_* flags to be conveniently added, irrespective of current
> > > >   * value, and then the old version restored with memalloc_flags_restore().
> > > >   */
> > > > -static inline unsigned memalloc_flags_save(unsigned flags)
> > > > +static inline unsigned int memalloc_flags_save(unsigned int flags_mask)
> > > >  {
> > > > -   unsigned oldflags = ~current->flags & flags;
> > > > -   current->flags |= flags;
> > > > -   return oldflags;
> > > > +   unsigned int saved_flags_mask = ~current->flags & flags_mask;
> > > > +
> > > > +   current->flags |= flags_mask;
> > > > +   return saved_flags_mask;
> > > >  }
> > > >
> > > > -static inline void memalloc_flags_restore(unsigned flags)
> > > > +static inline void memalloc_flags_restore(unsigned int flags_mask)
> > > >  {
> > > > -   current->flags &= ~flags;
> > > > +   current->flags &= ~flags_mask;
> > > >  }
> > >
> > > I guess so.  Maybe.  A bit.  Kent, what do you think?
> >
> > Eesh, seems like pointless verbosity to me. Maybe don't change it if it
> > doesn't need to be changed?
> 
> Hi Kent, thanks for your feedback.
> How about this version,  only change the 'old' to 'saved'.
> The function does not return the old current->flags value. Instead,
> it returns the subset of flags that were not previously set in current->flags,
> so they can later be cleared by memalloc_flags_restore(). The name savedflags
> makes this behavior clearer and avoids confusion.

Why change it at all? The returned flags parameter is opaque state that
should only be used by memalloc_flags_restore(), it's not something the
caller should be looking at.

> 
>  /**
> - * memalloc_flags_save - Add a PF_* flag to current->flags, save old value
> + * memalloc_flags_save - Add a PF_* flag to current->flags, return saved flags
>   *
>   * This allows PF_* flags to be conveniently added, irrespective of current
>   * value, and then the old version restored with memalloc_flags_restore().
>   */
>  static inline unsigned memalloc_flags_save(unsigned flags)
>  {
> -       unsigned oldflags = ~current->flags & flags;
> +       unsigned savedflags = ~current->flags & flags;
>         current->flags |= flags;
> -       return oldflags;
> +       return savedflags;
>  }
> 
> Thanks,
> Zhongkun