Re: tmpfs status

[Maksym Planeta]

Samuel Thibault <samuel.thibault@gnu.org> writes:

> Maksym Planeta, le Sun 08 Apr 2012 00:48:18 +0300, a écrit :
>> > Existing conditions, agreed.  But the conditions you introduce are about
>> > one-level only.
>> 
>> Really, but I can replace 3 conditions from the beginning of commit with
>> one. Would it fit?
>
> That would be simpler, yes.
>
>> >> Also you will need to allocate memory map for every object is created
>> >> even it would not ever use pageout.
>> >
>> > Well, that's already what we do in pager_alloc, don't we?
>> >
>> 
>> Than I'd rather removed initialization of map from pager_alloc, because,
>> I think, there is no sense to consume space for map if object doesn't
>> use it.
>
> Then let's do it. It's better to make it clear that map can be NULL
> right from the start, and not just cope with some case that seldomly
> happens. It also helps making sure all codepaths are fixed.
>

OK.

> BTW, I guess the #if 0 you added around "XXX we know how kalloc.c works;
> avoid copying." means that the optimization is actually bogus nowadays?
>

No, it isn't bogus and should work, but as it is hack that relies on
implementation and how it works is not obvious (at least from my point
of view), I decided to disable it. Do you think that it should be
returned?

Regards,
Maksym Planeta.