On Thu 08-06-17 22:18:22, Michal Hocko wrote:
> On Thu 08-06-17 10:05:57, Matthew Wilcox wrote:
> > On Thu, Jun 08, 2017 at 04:48:31PM +0200, Michal Hocko wrote:
> > > On Wed 07-06-17 13:56:01, David Rientjes wrote:
> > > > I agree it's probably going to bisect to 338a16ba15495 since it's the
> > > > cond_resched() at the line number reported, but I think there must be
> > > > something else going on. I think the list of locks held by khugepaged is
> > > > correct because it matches with the implementation. The preempt_count(),
> > > > as suggested by Andrew, does not. If this is reproducible, I'd like to
> > > > know what preempt_count() is.
> > >
> > > collapse_huge_page
> > > pte_offset_map
> > > kmap_atomic
> > > kmap_atomic_prot
> > > preempt_disable
> > > __collapse_huge_page_copy
> > > pte_unmap
> > > kunmap_atomic
> > > __kunmap_atomic
> > > preempt_enable
> > >
> > > I suspect, so cond_resched seems indeed inappropriate on 32b systems.
> >
> > Then why doesn't it trigger on 64-bit systems too?
> >
> > #ifndef ARCH_HAS_KMAP
> > ...
> > static inline void *kmap_atomic(struct page *page)
> > {
> > preempt_disable();
> > pagefault_disable();
> > return page_address(page);
> > }
> > #define kmap_atomic_prot(page, prot) kmap_atomic(page)
> >
> >
> > ... oh, wait, I see. Because pte_offset_map() doesn't call kmap_atomic()
> > on 64-bit. Indeed, it doesn't necessarily call kmap_atomic() on 32-bit
> > either; only with CONFIG_HIGHPTE enabled. How much of a performance
> > penalty would it be to call kmap_atomic() unconditionally on 64 bit to
> > make sure that this kind of problem doesn't show on 32-bit systems only?
>
> I am not sure I understand why would we map those pages in 64b systems?
> We can access them directly.
But I guess you are primary after syncing the preemptive mode for 64 and
32b systems, right? I agree that having a different model is more than
unfortunate because 32b gets much less testing coverage and so a risk of
introducing a new bug is just a matter of time. Maybe we should make
pte_offset_map disable preemption and currently noop pte_unmap to
preempt_enable. The overhead should be pretty marginal on x86_64 but not
all arches have per-cpu preempt count. So I am not sure we really want
to add this to just for the debugging purposes...
I would just pull the cond_resched out of __collapse_huge_page_copy
right after pte_unmap. But I am not really sure why this cond_resched is
really needed because the changelog of the patch which adds is is quite
terse on details.
--
Michal Hocko
SUSE Labs
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
