[PATCH v13 3/5] mm/vmalloc: use physical page count in vread_iter()

[Shivam Kalra via B4 Relay]

From: Shivam Kalra <shivamkalra98@xxxxxxxxxxx>

Update vread_iter() to derive the vm area size from vm->nr_pages rather
than get_vm_area_size().

Currently both values are equivalent, but the upcoming vrealloc() shrink
functionality will free pages without reducing the virtual reservation
size. After such a shrink, the old get_vm_area_size() based calculation
would overestimate the mapped range, causing vread_iter() to attempt
reading from unmapped addresses. Switch to vm->nr_pages now so the
reader remains correct once shrink support is added.

Reviewed-by: Uladzislau Rezki (Sony) <urezki@xxxxxxxxx>
Signed-off-by: Shivam Kalra <shivamkalra98@xxxxxxxxxxx>
---
 mm/vmalloc.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 87278cf2046e..d69174afb171 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -4666,7 +4666,14 @@ long vread_iter(struct iov_iter *iter, const char *addr, size_t count)
 		smp_rmb();
 
 		vaddr = (char *) va->va_start;
-		size = vm ? get_vm_area_size(vm) : va_size(va);
+		if (vm)
+			/*
+			 * Cannot use get_vm_area_size() because realloc()
+			 * may shrink the mapping and area->size may be outdated.
+			 */
+			size = vm->nr_pages << PAGE_SHIFT;
+		else
+			size = va_size(va);
 
 		if (addr >= vaddr + size)
 			goto next_va;

-- 
2.43.0