[PATCH v3 5/5] x86/mm: add WARN_ON_ONCE() for wrong large page mapping

[Bin Yang <bin.yang@intel.com>]

If there is a large page which contains an area which requires a
different mapping that the one which the large page provides,
then something went wrong _before_ this code is called.

Here we can catch a case where the existing mapping is wrong
already.

Inspired-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Bin Yang <bin.yang@intel.com>
---
 arch/x86/mm/pageattr.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index fd90c5b..91a250c 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -625,6 +625,7 @@ try_preserve_large_page(pte_t *kpte, unsigned long address,
 
 	psize = page_level_size(level);
 	pmask = page_level_mask(level);
+	addr = address & pmask;
 
 	/*
 	 * Calculate the number of pages, which fit into this large
@@ -636,6 +637,12 @@ try_preserve_large_page(pte_t *kpte, unsigned long address,
 		cpa->numpages = numpages;
 
 	/*
+	 * The old pgprot should not have any protection bit. Otherwise,
+	 * the existing mapping is wrong already.
+	 */
+	WARN_ON_ONCE(needs_static_protections(old_prot, addr, psize, old_pfn));
+
+	/*
 	 * We are safe now. Check whether the new pgprot is the same:
 	 * Convert protection attributes to 4k-format, as cpa->mask* are set
 	 * up accordingly.
@@ -690,7 +697,6 @@ try_preserve_large_page(pte_t *kpte, unsigned long address,
 	 * would anyway result in a split after doing all the check work
 	 * for nothing.
 	 */
-	addr = address & pmask;
 	if (address != addr || cpa->numpages != numpages)
 		goto out_unlock;
 
-- 
2.7.4