[PATCH] unpaged: anon in VM_UNPAGED
copy_one_pte needs to copy the anonymous COWed pages in a VM_UNPAGED area,
zap_pte_range needs to free them, do_wp_page needs to COW them: just like
ordinary pages, not like the unpaged.
But recognizing them is a little subtle: because PageReserved is no longer a
condition for remap_pfn_range, we can now mmap all of /dev/mem (whether the
distro permits, and whether it's advisable on this or that architecture, is
another matter). So if we can see a PageAnon, it may not be ours to mess with
(or may be ours from elsewhere in the address space). I suspect there's an
entertaining insoluble self-referential problem here, but the page_is_anon
function does a good practical job, and MAP_PRIVATE PROT_WRITE VM_UNPAGED will
always be an odd choice.
In updating the comment on page_address_in_vma, noticed a potential NULL
dereference, in a path we don't actually take, but fixed it.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
diff --git a/mm/memory.c b/mm/memory.c
index 107b619..3666a4c 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -350,6 +350,22 @@
}
/*
+ * page_is_anon applies strict checks for an anonymous page belonging to
+ * this vma at this address. It is used on VM_UNPAGED vmas, which are
+ * usually populated with shared originals (which must not be counted),
+ * but occasionally contain private COWed copies (when !VM_SHARED, or
+ * perhaps via ptrace when VM_SHARED). An mmap of /dev/mem might window
+ * free pages, pages from other processes, or from other parts of this:
+ * it's tricky, but try not to be deceived by foreign anonymous pages.
+ */
+static inline int page_is_anon(struct page *page,
+ struct vm_area_struct *vma, unsigned long addr)
+{
+ return page && PageAnon(page) && page_mapped(page) &&
+ page_address_in_vma(page, vma) == addr;
+}
+
+/*
* copy one vm_area from one task to the other. Assumes the page tables
* already present in the new task to be cleared in the whole range
* covered by this vma.
@@ -381,23 +397,22 @@
goto out_set_pte;
}
- /* If the region is VM_UNPAGED, the mapping is not
- * mapped via rmap - duplicate the pte as is.
- */
- if (vm_flags & VM_UNPAGED)
- goto out_set_pte;
-
pfn = pte_pfn(pte);
- /* If the pte points outside of valid memory but
+ page = pfn_valid(pfn)? pfn_to_page(pfn): NULL;
+
+ if (unlikely(vm_flags & VM_UNPAGED))
+ if (!page_is_anon(page, vma, addr))
+ goto out_set_pte;
+
+ /*
+ * If the pte points outside of valid memory but
* the region is not VM_UNPAGED, we have a problem.
*/
- if (unlikely(!pfn_valid(pfn))) {
+ if (unlikely(!page)) {
print_bad_pte(vma, pte, addr);
goto out_set_pte; /* try to do something sane */
}
- page = pfn_to_page(pfn);
-
/*
* If it's a COW mapping, write protect it both
* in the parent and the child
@@ -568,17 +583,20 @@
continue;
}
if (pte_present(ptent)) {
- struct page *page = NULL;
+ struct page *page;
+ unsigned long pfn;
(*zap_work) -= PAGE_SIZE;
- if (!(vma->vm_flags & VM_UNPAGED)) {
- unsigned long pfn = pte_pfn(ptent);
- if (unlikely(!pfn_valid(pfn)))
- print_bad_pte(vma, ptent, addr);
- else
- page = pfn_to_page(pfn);
- }
+ pfn = pte_pfn(ptent);
+ page = pfn_valid(pfn)? pfn_to_page(pfn): NULL;
+
+ if (unlikely(vma->vm_flags & VM_UNPAGED)) {
+ if (!page_is_anon(page, vma, addr))
+ page = NULL;
+ } else if (unlikely(!page))
+ print_bad_pte(vma, ptent, addr);
+
if (unlikely(details) && page) {
/*
* unmap_shared_mapping_pages() wants to
@@ -1295,10 +1313,11 @@
old_page = pfn_to_page(pfn);
src_page = old_page;
- if (unlikely(vma->vm_flags & VM_UNPAGED)) {
- old_page = NULL;
- goto gotten;
- }
+ if (unlikely(vma->vm_flags & VM_UNPAGED))
+ if (!page_is_anon(old_page, vma, address)) {
+ old_page = NULL;
+ goto gotten;
+ }
if (PageAnon(old_page) && !TestSetPageLocked(old_page)) {
int reuse = can_share_swap_page(old_page);