Re: [PATCH 1/6] mm: introduce kv[mz]alloc helpers

[John Hubbard]

On 01/16/2017 01:48 PM, Michal Hocko wrote:
> On Mon 16-01-17 13:15:08, John Hubbard wrote:
> > On 01/16/2017 11:40 AM, Michal Hocko wrote:
> > > On Mon 16-01-17 11:09:37, John Hubbard wrote:
> > > > On 01/16/2017 12:47 AM, Michal Hocko wrote:
> > > > > On Sun 15-01-17 20:34:13, John Hubbard wrote:
> > > [...]
> > > > > > Is that "Reclaim modifiers" line still true, or is it a leftover from an
> > > > > > earlier approach? I am having trouble reconciling it with rest of the
> > > > > > patchset, because:
> > > > > >
> > > > > > a) the flags argument below is effectively passed on to either kmalloc_node
> > > > > > (possibly adding, but not removing flags), or to __vmalloc_node_flags.
> > > > >
> > > > > The above only says thos are _unsupported_ - in other words the behavior
> > > > > is not defined. Even if flags are passed down to kmalloc resp. vmalloc
> > > > > it doesn't mean they are used that way.  Remember that vmalloc uses
> > > > > some hardcoded GFP_KERNEL allocations.  So while I could be really
> > > > > strict about this and mask away these flags I doubt this is worth the
> > > > > additional code.
> > > >
> > > > I do wonder about passing those flags through to kmalloc. Maybe it is worth
> > > > stripping out __GFP_NORETRY and __GFP_NOFAIL, after all. It provides some
> > > > insulation from any future changes to the implementation of kmalloc, and it
> > > > also makes the documentation more believable.
> > >
> > > I am not really convinced that we should take an extra steps for these
> > > flags. There are no existing users for those flags and new users should
> > > follow the documentation.
> >
> > OK, let's just fortify the documentation ever so slightly, then, so that
> > users are more likely to do the right thing. How's this sound:
> >
> > * Reclaim modifiers - __GFP_NORETRY and __GFP_NOFAIL are not supported. (Even
> > * though the current implementation passes the flags on through to kmalloc and
> > * vmalloc, that is done for efficiency and to avoid unnecessary code. The caller
> > * should not pass in these flags.)
> > *
> > * __GFP_REPEAT is supported, but only for large (>64kB) allocations.
> >
> >
> > ? Or is that documentation overkill?
>
> Dunno, it sounds like an overkill to me. It is telling more than
> necessary. If we want to be so vocal about gfp flags then we would have
> to say much more I suspect. E.g. what about __GFP_HIGHMEM? This flag is
> supported for vmalloc while unsupported for kmalloc. I am pretty sure
> there would be other gfp flags to consider and then this would grow
> borringly large and uninteresting to the point when people simply stop
> reading it. Let's just be as simple as possible.

Agreed, on the simplicity point: simple and clear is ideal. But here, it's merely short, and not 
quite simple. :)  People will look at that short bit of documentation, and then notice that the 
flags are, in fact, all passed right on through down to both kmalloc_node and __vmalloc_node_flags.

If you don't want too much documentation, then I'd be inclined to say something higher-level, about 
the intent, rather than mentioning those two flags directly. Because as it stands, the documentation 
contradicts what the code does.

Sorry to go on and on about such a minor point. I'll let it go after this last note.

> --
> Michal Hocko
> SUSE Labs