Re: check for correct flag in uvm_page_unbusy
On Thu, Jun 23, 2011 at 07:05:25PM +0100, Owain Ainsworth wrote: > How about this now? > > I've got some more important uvm diffs on the way I but would like to > get the small ones out of my tree. ok. > On Tue, May 31, 2011 at 12:09:35AM +0100, Owain Ainsworth wrote: > > No functional change since aobjs should never hit this path. However, I > > introduced this diff when I reworked the page releasing stuff for > > objects. > > > > ok? > > > > diff --git uvm/uvm_page.c uvm/uvm_page.c > > index 10ef7d1..27e970a 100644 > > --- uvm/uvm_page.c > > +++ uvm/uvm_page.c > > @@ -1099,7 +1099,7 @@ uvm_page_unbusy(struct vm_page **pgs, int npgs) > > uvm_lock_pageq(); > > pmap_page_protect(pg, VM_PROT_NONE); > > /* XXX won't happen right now */ > > - if (pg->pg_flags & PQ_ANON) > > + if (pg->pg_flags & PQ_AOBJ) > > uao_dropswap(uobj, > > pg->offset >> PAGE_SHIFT); > > uvm_pagefree(pg); > > -- > > 1.7.5 > > > > > > -- > > Blood flows down one leg and up the other. > > -- > Campus sidewalks never exist as the straightest line between two > points. > -- M. M. Johnston > -- Ariane
Re: check for correct flag in uvm_page_unbusy
How about this now? I've got some more important uvm diffs on the way I but would like to get the small ones out of my tree. -0- On Tue, May 31, 2011 at 12:09:35AM +0100, Owain Ainsworth wrote: > No functional change since aobjs should never hit this path. However, I > introduced this diff when I reworked the page releasing stuff for > objects. > > ok? > > diff --git uvm/uvm_page.c uvm/uvm_page.c > index 10ef7d1..27e970a 100644 > --- uvm/uvm_page.c > +++ uvm/uvm_page.c > @@ -1099,7 +1099,7 @@ uvm_page_unbusy(struct vm_page **pgs, int npgs) > uvm_lock_pageq(); > pmap_page_protect(pg, VM_PROT_NONE); > /* XXX won't happen right now */ > - if (pg->pg_flags & PQ_ANON) > + if (pg->pg_flags & PQ_AOBJ) > uao_dropswap(uobj, > pg->offset >> PAGE_SHIFT); > uvm_pagefree(pg); > -- > 1.7.5 > > > -- > Blood flows down one leg and up the other. -- Campus sidewalks never exist as the straightest line between two points. -- M. M. Johnston
check for correct flag in uvm_page_unbusy
No functional change since aobjs should never hit this path. However, I introduced this diff when I reworked the page releasing stuff for objects. ok? diff --git uvm/uvm_page.c uvm/uvm_page.c index 10ef7d1..27e970a 100644 --- uvm/uvm_page.c +++ uvm/uvm_page.c @@ -1099,7 +1099,7 @@ uvm_page_unbusy(struct vm_page **pgs, int npgs) uvm_lock_pageq(); pmap_page_protect(pg, VM_PROT_NONE); /* XXX won't happen right now */ - if (pg->pg_flags & PQ_ANON) + if (pg->pg_flags & PQ_AOBJ) uao_dropswap(uobj, pg->offset >> PAGE_SHIFT); uvm_pagefree(pg); -- 1.7.5 -- Blood flows down one leg and up the other.