Re: lib/54135: sparc userland can not deal with 8k page size

[christos%astron.com@localhost (Christos Zoulas)]

In article <20190425142001.D54517A1CF%mollari.NetBSD.org@localhost>,
Martin Husemann  <gnats-bugs%netbsd.org@localhost> wrote:
>The following reply was made to PR lib/54135; it has been noted by GNATS.
>
>From: Martin Husemann <martin%duskware.de@localhost>
>To: Christos Zoulas <christos%zoulas.com@localhost>
>Cc: gnats-bugs%netbsd.org@localhost
>Subject: Re: lib/54135: sparc userland can not deal with 8k page size
>Date: Thu, 25 Apr 2019 16:17:50 +0200
>
> On Thu, Apr 25, 2019 at 07:39:41AM -0400, Christos Zoulas wrote:
> > This is proving a tough nut to crack. Perhaps going at it backwards
>is easier (revert all of it and assign os_page = PAGE after the first
>test.
> 
> I think we are doing it wrong. The comments in jemalloc code all seem to
> suggest that the real page size must be <= the PAGE define, but we are
> doing:
> 
> #define LG_PAGE MIN_PAGE_SHIFT
> 
> instead of MAX_PAGE_SHIFT. Why is it this way around?
> 
> This issue should also break userland on Sun4 systems (i.e. not only on
> sparc64 kernels).

We used to have it MAX_PAGE_SHIFT:

http://mail-index.netbsd.org/port-sandpoint/2019/04/08/msg000457.html

perhaps we can go back to it and relax that constraint...

christos