On Mon, Aug 14, 2017 at 09:47:18PM -0600, Tycho Andersen wrote:
> I'll do that for the next version
Actually looking closer, I think we just need to mirror the
debug_pagealloc_enabled() checks in set_kpte() from
split_large_page(),
diff --git a/arch/x86/mm/xpfo.c b/arch/x86/mm/xpfo.c
index a1344f27406c..c962bd7f34cc 100644
--- a/arch/x86/mm/xpfo.c
+++ b/arch/x86/mm/xpfo.c
@@ -54,9 +54,11 @@ inline void set_kpte(void *kaddr, struct page *page,
pgprot_t prot)
do_split = try_preserve_large_page(pte, (unsigned long)kaddr,
&cpa);
if (do_split) {
- spin_lock(&cpa_lock);
+ if (!debug_pagealloc_enabled())
+ spin_lock(&cpa_lock);
BUG_ON(split_large_page(&cpa, pte, (unsigned
long)kaddr));
- spin_unlock(&cpa_lock);
+ if (!debug_pagealloc_enabled())
+ spin_unlock(&cpa_lock);
}
break;
Tycho