Re: [PATCH RFCv2 02/15] xen/arm: lpae: Use the generic helpers to defined the Xen PT helpers

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Sat, 3 Jul 2021, Julien Grall wrote:
> Hi Stefano,
> 
> Sorry for the late answer.
> 
> On 13/05/2021 23:44, Stefano Stabellini wrote:
> > On Wed, 12 May 2021, Julien Grall wrote:
> > > Hi Stefano,
> > > 
> > > On 12/05/2021 22:30, Stefano Stabellini wrote:
> > > > On Wed, 12 May 2021, Julien Grall wrote:
> > > > > > > +#define LPAE_SHIFT          LPAE_SHIFT_GS(PAGE_SHIFT)
> > > > > > > +#define LPAE_ENTRIES        LPAE_ENTRIES_GS(PAGE_SHIFT)
> > > > > > > +#define LPAE_ENTRY_MASK     LPAE_ENTRY_MASK_GS(PAGE_SHIFT)
> > > > > > > 
> > > > > > > +#define LEVEL_SHIFT(lvl)    LEVEL_SHIFT_GS(PAGE_SHIFT, lvl)
> > > > > > > +#define LEVEL_ORDER(lvl)    LEVEL_ORDER_GS(PAGE_SHIFT, lvl)
> > > > > > > +#define LEVEL_SIZE(lvl)     LEVEL_SIZE_GS(PAGE_SHIFT, lvl)
> > > > > > > +#define LEVEL_MASK(lvl)     (~(LEVEL_SIZE(lvl) - 1))
> > > > > > 
> > > > > > I would avoid adding these 4 macros. It would be OK if they were
> > > > > > just
> > > > > > used within this file but lpae.h is a header: they could end up be
> > > > > > used
> > > > > > anywhere in the xen/ code and they have a very generic name. My
> > > > > > suggestion would be to skip them and just do:
> > > > > 
> > > > > Those macros will be used in follow-up patches. They are pretty useful
> > > > > to
> > > > > avoid introduce static array with the different information for each
> > > > > level.
> > > > > 
> > > > > Would prefix them with XEN_ be better?
> > > > 
> > > > Maybe. The concern I have is that there are multiple page granularities
> > > > (4kb, 16kb, etc) and multiple page sizes (4kb, 2mb, etc). If I just see
> > > > LEVEL_ORDER it is not immediately obvious what granularity and what size
> > > > we are talking about.
> > > 
> > > I am a bit puzzled with your answer. AFAIU, you are happy with the
> > > existing
> > > macros (THIRD_*, SECOND_*) but not with the new macros.
> > > 
> > > In reality, there is no difference because THIRD_* doesn't tell you the
> > > exact
> > > size but only "this is a level 3 mapping".
> > > 
> > > So can you clarify what you are after? IOW is it reworking the current
> > > naming
> > > scheme?
> > 
> > You are right -- there is no real difference between THIRD_*, SECOND_*
> > and LEVEL_*.
> > 
> > The original reason for my comments is that I hadn't read the following
> > patches, and the definition of LEVEL_* macros is simple, they could be
> > open coded. It looked like they were only going to be used to make the
> > definition of THIRD_*, SECOND_* a bit easier. So, at first, I was
> > wondering if they were needed at all.
> > 
> > Secondly, I realized that they were going to be used in *.c files by
> > other patches. That's why they are there. But I started thinking whether
> > we should find a way to make it a bit clearer that they are for Xen
> > pages, currently at 4KB granularity. THIRD_*, SECOND_*, etc. are already
> > generic names which don't convey the granularity or whether they are Xen
> > pages at all. But LEVEL_* seem even more generic.
> > 
> > As I mentioned, I don't have any good suggestions for changes to make
> > here, so unless you can come up with a good idea let's keep it as is.
> 
> I am thinking to use the following naming (diff on top of this patch):
> 
> -#define LPAE_SHIFT          LPAE_SHIFT_GS(PAGE_SHIFT)
> -#define LPAE_ENTRIES        LPAE_ENTRIES_GS(PAGE_SHIFT)
> -#define LPAE_ENTRY_MASK     LPAE_ENTRY_MASK_GS(PAGE_SHIFT)
> +#define XEN_PT_SHIFT          LPAE_SHIFT_GS(PAGE_SHIFT)
> +#define XEN_PT_ENTRIES        LPAE_ENTRIES_GS(PAGE_SHIFT)
> +#define XEN_PT_ENTRY_MASK     LPAE_ENTRY_MASK_GS(PAGE_SHIFT)
> 
> -#define LEVEL_SHIFT(lvl)    LEVEL_SHIFT_GS(PAGE_SHIFT, lvl)
> -#define LEVEL_ORDER(lvl)    LEVEL_ORDER_GS(PAGE_SHIFT, lvl)
> -#define LEVEL_SIZE(lvl)     LEVEL_SIZE_GS(PAGE_SHIFT, lvl)
> -#define LEVEL_MASK(lvl)     (~(LEVEL_SIZE(lvl) - 1))
> +#define XEN_PT_LEVEL_SHIFT(lvl)    LEVEL_SHIFT_GS(PAGE_SHIFT, lvl)
> +#define XEN_PT_LEVEL_ORDER(lvl)    LEVEL_ORDER_GS(PAGE_SHIFT, lvl)
> +#define XEN_PT_LEVEL_SIZE(lvl)     LEVEL_SIZE_GS(PAGE_SHIFT, lvl)
> +#define XEN_PT_LEVEL_MASK(lvl)     (~(LEVEL_SIZE(lvl) - 1))
> 
>  /* Convenience aliases */
> -#define THIRD_SHIFT         LEVEL_SHIFT(3)
> -#define THIRD_ORDER         LEVEL_ORDER(3)
> -#define THIRD_SIZE          LEVEL_SIZE(3)
> -#define THIRD_MASK          LEVEL_MASK(3)
> -
> -#define SECOND_SHIFT        LEVEL_SHIFT(2)
> -#define SECOND_ORDER        LEVEL_ORDER(2)
> -#define SECOND_SIZE         LEVEL_SIZE(2)
> -#define SECOND_MASK         LEVEL_MASK(2)
> -
> -#define FIRST_SHIFT         LEVEL_SHIFT(1)
> -#define FIRST_ORDER         LEVEL_ORDER(1)
> -#define FIRST_SIZE          LEVEL_SIZE(1)
> -#define FIRST_MASK          LEVEL_MASK(1)
> -
> -#define ZEROETH_SHIFT       LEVEL_SHIFT(0)
> -#define ZEROETH_ORDER       LEVEL_ORDER(0)
> -#define ZEROETH_SIZE        LEVEL_SIZE(0)
> -#define ZEROETH_MASK        LEVEL_MASK(0)
> +#define THIRD_SHIFT         XEN_PT_LEVEL_SHIFT(3)
> +#define THIRD_ORDER         XEN_PT_LEVEL_ORDER(3)
> +#define THIRD_SIZE          XEN_PT_LEVEL_SIZE(3)
> +#define THIRD_MASK          XEN_PT_LEVEL_MASK(3)
> +
> +#define SECOND_SHIFT        XEN_PT_LEVEL_SHIFT(2)
> +#define SECOND_ORDER        XEN_PT_LEVEL_ORDER(2)
> +#define SECOND_SIZE         XEN_PT_LEVEL_SIZE(2)
> +#define SECOND_MASK         XEN_PT_LEVEL_MASK(2)
> +
> +#define FIRST_SHIFT         XEN_PT_LEVEL_SHIFT(1)
> +#define FIRST_ORDER         XEN_PT_LEVEL_ORDER(1)
> +#define FIRST_SIZE          XEN_PT_LEVEL_SIZE(1)
> +#define FIRST_MASK          XEN_PT_LEVEL_MASK(1)
> +
> +#define ZEROETH_SHIFT       XEN_PT_LEVEL_SHIFT(0)
> +#define ZEROETH_ORDER       XEN_PT_LEVEL_ORDER(0)
> +#define ZEROETH_SIZE        XEN_PT_LEVEL_SIZE(0)
> +#define ZEROETH_MASK        XEN_PT_LEVEL_MASK(0)
> 
> I don't plan to modify the nameing for ZEROETH*, FIRST*, SECOND*, THIRD*.
> 
> Let me know if you prefer it over the currrent naming.

Yes, I think it is better, thanks!