Skip to contentSkip to conversation

fix do_wp_page_mk_pte_writable() in 2.6.10-rc2-mm3

bugfix William Lee Irwin III 1 participant 1 message Nov 22, 2004

Subsystems (2)
lore
Share & tools
lore
#0 lore raw
Quoted 2 lines · Andrew Morton

4level-core-patch.patch
4level core patch

vma->vm_ops->page_mkwrite() is supposed to be able to sleep, but this function doesn't unmap the pte across the call and worse yet, reuses the pte across a drop and reacquisition of ->page_table_lock.

mm/memory.c +4 -1
Index: mm3-2.6.10-rc2/mm/memory.c
===================================================================
@@ -1268,6 +1268,7 @@
 static inline int do_wp_page_mk_pte_writable(struct mm_struct *mm,
 					     struct vm_area_struct *vma,
 					     unsigned long address,
+					     pmd_t *pmd,
 					     pte_t *page_table,
 					     struct page *old_page,
 					     pte_t pte)
@@ -1279,6 +1280,7 @@
 	if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
 		/* Notify the page owner without the lock held so they can
 		 * sleep if they want to */
+		pte_unmap(page_table);
 		spin_unlock(&mm->page_table_lock);
 
 		if (vma->vm_ops->page_mkwrite(vma, old_page) < 0)
@@ -1291,6 +1293,7 @@
 		 * return, as we can count on the MMU to tell us if they didn't
 		 * also make it writable
 		 */
+		page_table = pte_offset_map(pmd, address);
 		if (!pte_same(*page_table, pte))
 			goto minor_fault;
 	}
@@ -1352,7 +1355,7 @@
 		unlock_page(old_page);
 		if (reuse)
 			/* We can just make the PTE writable */
-			return do_wp_page_mk_pte_writable(mm, vma, address,
+			return do_wp_page_mk_pte_writable(mm, vma, address, pmd,
 							  page_table, old_page,
 							  pte);
 	}