Re: [RFC PATCH v3 for 4.15 08/24] Provide cpu_opv system call

[Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>]

----- On Nov 20, 2017, at 1:03 PM, Thomas Gleixner tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org wrote:

> On Mon, 20 Nov 2017, Thomas Gleixner wrote:
>> On Mon, 20 Nov 2017, Mathieu Desnoyers wrote:
>> > >> + * The reason why we require all pointer offsets to be calculated by
>> > >> + * user-space beforehand is because we need to use get_user_pages_fast()
>> > >> + * to first pin all pages touched by each operation. This takes care of
>> > > 
>> > > That doesnt explain it either.
>> > 
>> > What kind of explication are you looking for here ? Perhaps being too close
>> > to the implementation prevents me from understanding what is unclear from
>> > your perspective.
>> 
>> What the heck are pointer offsets?
>> 
>> The ops have one or two pointer(s) to a lump of memory. So if a pointer
>> points to the wrong lump of memory then you're screwed, but that's true for
>> all pointers handed to the kernel.
> 
> So I think you mix here the 'user space programmer guide - how to set up
> the magic ops - into the kernel side documentation. The kernel side does
> not care about the pointers, as long as they are valid to access.
> 
> Again. Try to explain things at the conceptual level.

Yes, I took the sentence you suggested for that reason: it removes
usage details that are meant for user-space implementer, which do not
belong in those comments.

Thanks,

Mathieu

> 
> Thanks,
> 
> 	tglx

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com