On Friday 17 April 2009, Alan Jenkins wrote:
> Linus Torvalds wrote:
> > On Fri, 17 Apr 2009, Alan Jenkins wrote:
> >
> >> As another datapoint: I tried blindly applying the commit to 2.6.29.
> >> The resulting kernel was able to hibernate fine the first time.
> >>
> >
> > Yeah, so it's not that commit per se that causes it. I bet it needs all
> > the IO scheduler changes too - and even when it does that, the end result
> > probably is really just a timing change.
> >
> >
> >> I'm going to be annoying and try something slightly different. In
> >> theory, I should be able to find the "first bad commit" where
> >> cherry-picking 1faa16d22 causes a problem.
> >>
> >
> > Just for fun, try this one first and see if it makes any difference.
> >
> > Maybe the whole "swappiness=0" part was intentional. And maybe it wasn't.
> > This is one trivial patch. Maybe it makes your machine blow up. Who knows?
> >
> > There are other differences in the shrink_all_memory() path wrt the normal
> > memory freeing paths, but they are way more subtle. So I'm suggesting
> > tryign this not becasue I think it's "The Bug(tm)", but because it's an
> > easy test to make, and maybe it makes a difference.
> >
> > Linus
> > ---
> > mm/vmscan.c | 2 ++
> > 1 files changed, 2 insertions(+), 0 deletions(-)
> >
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index 39fdfb1..d3595ed 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -2113,6 +2113,8 @@ unsigned long shrink_all_memory(unsigned long
> > nr_pages)
> > struct scan_control sc = {
> > .gfp_mask = GFP_KERNEL,
> > .may_unmap = 0,
> > + .swap_cluster_max = SWAP_CLUSTER_MAX,
> > + .swappiness = vm_swappiness,
> > .may_writepage = 1,
> > .isolate_pages = isolate_pages_global,
> > };
> >
>
>
> No, that doesn't seem to affect it.
Can you please try to reproduce the problem with the appended debug patch
applied and send the output of dmesg to me?
Rafael
---
mm/vmscan.c | 8 ++++++++
1 file changed, 8 insertions(+)
Index: linux-2.6/mm/vmscan.c
===================================================================
--- linux-2.6.orig/mm/vmscan.c
+++ linux-2.6/mm/vmscan.c
@@ -2115,6 +2115,7 @@ unsigned long shrink_all_memory(unsigned
.may_unmap = 0,
.may_writepage = 1,
.isolate_pages = isolate_pages_global,
+ .nr_reclaimed = 0,
};
current->reclaim_state = &reclaim_state;
@@ -2135,6 +2136,8 @@ unsigned long shrink_all_memory(unsigned
nr_slab -= reclaim_state.reclaimed_slab;
}
+ printk(KERN_INFO "before: sc.nr_reclaimed = %lu\n", sc.nr_reclaimed);
+
/*
* We try to shrink LRUs in 5 passes:
* 0 = Reclaim from inactive_list only
@@ -2168,6 +2171,10 @@ unsigned long shrink_all_memory(unsigned
if (sc.nr_scanned && prio < DEF_PRIORITY - 2)
congestion_wait(WRITE, HZ / 10);
+
+ printk(KERN_INFO "pass = %d, prio = %d, "
+ "sc.nr_reclaimed = %lu\n", pass, prio,
+ sc.nr_reclaimed);
}
}
@@ -2184,6 +2191,7 @@ unsigned long shrink_all_memory(unsigned
reclaim_state.reclaimed_slab > 0);
}
+ printk(KERN_INFO "after: sc.nr_reclaimed = %lu\n", sc.nr_reclaimed);
out:
current->reclaim_state = NULL;
--
To unsubscribe from this list: send the line "unsubscribe kernel-testers" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
