Re: Proof-of-concept: better(?) page-table manipulation API

From: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
To: Matthew Wilcox <willy@infradead.org>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>,
	Andy Lutomirski <luto@amacapital.net>,
	Andrew Morton <akpm@linux-foundation.org>,
	Michal Hocko <mhocko@kernel.org>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	X86 ML <x86@kernel.org>, Linux-MM <linux-mm@kvack.org>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: Proof-of-concept: better(?) page-table manipulation API
Date: Mon, 7 May 2018 16:10:59 +0300	[thread overview]
Message-ID: <20180507131059.3cdpfd4dcnwznhyw@black.fi.intel.com> (raw)
In-Reply-To: <20180507122346.GE18116@bombadil.infradead.org>

On Mon, May 07, 2018 at 12:23:46PM +0000, Matthew Wilcox wrote:
> On Mon, May 07, 2018 at 02:31:25PM +0300, Kirill A. Shutemov wrote:
> > > Also, what does lvl == 0 mean?  Is it the top or the bottom?  I think a
> > > comment would be helpful.
> > 
> > It is bottom. But it should be up to architecture to decide.
> 
> That's not true because ...
> 
> > > > +static inline void ptp_walk(ptp_t *ptp, unsigned long addr)
> > > > +{
> > > > +       ptp->ptr = (unsigned long *)ptp_page_vaddr(ptp);
> > > > +       ptp->ptr += __pt_index(addr, --ptp->lvl);
> > > > +}
> > > 
> > > Can you add a comment that says what this function does?
> > 
> > Okay, I will.
> > 
> > > Why does it not change the level?
> > 
> > It does. --ptp->lvl.
> 
> ... you've hardcoded that walking down decrements the level by 1.
> 
> I don't see that as a defect; it's just part of the API that needs
> documenting.

You assume that the function is a generic one. This may or may not be
true.

This is subject for refinement anyway.

-- 
 Kirill A. Shutemov