Another alpha panic

[Alan Cox]

On Fri, Apr 16, 2004 at 03:24:35PM -0400, Andrew Gallatin wrote:
Content-Description: message body text
> 
> Alan Cox writes:
> 
>  > Fixing this should be mechanical in nature: Search for all uses
>  > of PG_UNMANAGED in amd64's or i386's and change alpha's pmap accordingly.
>  > Volunteer?  I can review the patch but not test it.
> 
> How does the attached patch look?  I"ll try a buildworld with
> it before commiting, naturally..
> 
> Its sort of funny that pmap_enter_quick() talks about
> "Enter on the PV list if part of our managed memory"
> and then puts it there anyway..
> 

The patch looks correct.  In regards to the funny comment in
pmap_enter_quick(), I can say the following: once upon a time, all pages
passed to pmap_enter_quick() would have been managed pages.  So, the
check could have been safely omitted as a micro-optimization.

Alan

P.S. A thought regarding style, in pmap_enter_quick(), why not follow
the same idiom used in pmap_enter() for implementing the "management"
bits?  Specifically, something like

	managed = 0;
	if ((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0) {
		pmap_insert_entry(pmap, va, mpte, m);
		managed = PG_MANAGED | PG_FOR | PG_FOW | PG_FOE;
	}
	...
	*pte = ... | managed;