mshv_chunk_stride() rejected huge-page stride whenever pfn_count was
not a multiple of PTRS_PER_PMD, even though the same-stride scan in
mshv_region_process_pfns() already handles the transition between
huge-page and 4K stride for the tail of a run. As a result, a region
backed by a 2 MiB folio with a length that wasn't a multiple of 512
PFNs was processed entirely at 4K stride, issuing one hypercall per
PFN instead of one per 2 MiB.
Reject huge-page stride only when fewer than PTRS_PER_PMD PFNs are
available from this point. The starting GFN alignment requirement is
unchanged. The same-stride scan continues to handle the tail
correctly.
Fixes: 259add0d982cb ("mshv: Align huge page stride with guest mapping")
Signed-off-by: Stanislav Kinsburskii <[email protected]>
---
drivers/hv/mshv_regions.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/hv/mshv_regions.c b/drivers/hv/mshv_regions.c
index fdffd4f002f6f..81e57f727be35 100644
--- a/drivers/hv/mshv_regions.c
+++ b/drivers/hv/mshv_regions.c
@@ -43,7 +43,7 @@ static int mshv_chunk_stride(struct page *page,
*/
if (!PageCompound(page) || !PageHead(page) ||
!IS_ALIGNED(gfn, PTRS_PER_PMD) ||
- !IS_ALIGNED(page_count, PTRS_PER_PMD))
+ page_count < PTRS_PER_PMD)
return 1;
page_order = folio_order(page_folio(page));
