mm: fix races between swapoff and flush dcache

Thanks to commit 4b3ef9d ("mm/swap: split swap cache into 64MB
trunks"), after swapoff the address_space associated with the swap
device will be freed.  So page_mapping() users which may touch the
address_space need some kind of mechanism to prevent the address_space
from being freed during accessing.

The dcache flushing functions (flush_dcache_page(), etc) in architecture
specific code may access the address_space of swap device for anonymous
pages in swap cache via page_mapping() function.  But in some cases
there are no mechanisms to prevent the swap device from being swapoff,
for example,

  CPU1					CPU2
  __get_user_pages()			swapoff()
    flush_dcache_page()
      mapping = page_mapping()
        ...				  exit_swap_address_space()
        ...				    kvfree(spaces)
        mapping_mapped(mapping)

The address space may be accessed after being freed.

But from cachetlb.txt and Russell King, flush_dcache_page() only care
about file cache pages, for anonymous pages, flush_anon_page() should be
used.  The implementation of flush_dcache_page() in all architectures
follows this too.  They will check whether page_mapping() is NULL and
whether mapping_mapped() is true to determine whether to flush the
dcache immediately.  And they will use interval tree (mapping->i_mmap)
to find all user space mappings.  While mapping_mapped() and
mapping->i_mmap isn't used by anonymous pages in swap cache at all.

So, to fix the race between swapoff and flush dcache, __page_mapping()
is add to return the address_space for file cache pages and NULL
otherwise.  All page_mapping() invoking in flush dcache functions are
replaced with page_mapping_file().

[[email protected]: simplify page_mapping_file(), per Mike]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: "Huang, Ying" <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Johannes Weiner <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Chen Liqin <[email protected]>
Cc: Russell King <[email protected]>
Cc: Yoshinori Sato <[email protected]>
Cc: "James E.J. Bottomley" <[email protected]>
Cc: Guan Xuetao <[email protected]>
Cc: "David S. Miller" <[email protected]>
Cc: Chris Zankel <[email protected]>
Cc: Vineet Gupta <[email protected]>
Cc: Ley Foon Tan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Mike Rapoport <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

arch/arc/mm/cache.c

@@ -833,7 +833,7 @@ void flush_dcache_page(struct page *page)
 	}
 
 	/* don't handle anon pages here */
-	mapping = page_mapping(page);
+	mapping = page_mapping_file(page);
 	if (!mapping)
 		return;
 

arch/arm/mm/copypage-v4mc.c

@@ -70,7 +70,7 @@ void v4_mc_copy_user_highpage(struct page *to, struct page *from,
 	void *kto = kmap_atomic(to);
 
 	if (!test_and_set_bit(PG_dcache_clean, &from->flags))
-		__flush_dcache_page(page_mapping(from), from);
+		__flush_dcache_page(page_mapping_file(from), from);
 
 	raw_spin_lock(&minicache_lock);
 

arch/arm/mm/copypage-v6.c

@@ -76,7 +76,7 @@ static void v6_copy_user_highpage_aliasing(struct page *to,
 	unsigned long kfrom, kto;
 
 	if (!test_and_set_bit(PG_dcache_clean, &from->flags))
-		__flush_dcache_page(page_mapping(from), from);
+		__flush_dcache_page(page_mapping_file(from), from);
 
 	/* FIXME: not highmem safe */
 	discard_old_kernel_data(page_address(to));

arch/arm/mm/copypage-xscale.c

@@ -90,7 +90,7 @@ void xscale_mc_copy_user_highpage(struct page *to, struct page *from,
 	void *kto = kmap_atomic(to);
 
 	if (!test_and_set_bit(PG_dcache_clean, &from->flags))
-		__flush_dcache_page(page_mapping(from), from);
+		__flush_dcache_page(page_mapping_file(from), from);
 
 	raw_spin_lock(&minicache_lock);
 

arch/arm/mm/fault-armv.c

@@ -195,7 +195,7 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long addr,
 	if (page == ZERO_PAGE(0))
 		return;
 
-	mapping = page_mapping(page);
+	mapping = page_mapping_file(page);
 	if (!test_and_set_bit(PG_dcache_clean, &page->flags))
 		__flush_dcache_page(mapping, page);
 	if (mapping) {

arch/arm/mm/flush.c

@@ -285,7 +285,7 @@ void __sync_icache_dcache(pte_t pteval)
 
 	page = pfn_to_page(pfn);
 	if (cache_is_vipt_aliasing())
-		mapping = page_mapping(page);
+		mapping = page_mapping_file(page);
 	else
 		mapping = NULL;
 
@@ -333,7 +333,7 @@ void flush_dcache_page(struct page *page)
 		return;
 	}
 
-	mapping = page_mapping(page);
+	mapping = page_mapping_file(page);
 
 	if (!cache_ops_need_broadcast() &&
 	    mapping && !page_mapcount(page))
@@ -363,7 +363,7 @@ void flush_kernel_dcache_page(struct page *page)
 	if (cache_is_vivt() || cache_is_vipt_aliasing()) {
 		struct address_space *mapping;
 
-		mapping = page_mapping(page);
+		mapping = page_mapping_file(page);
 
 		if (!mapping || mapping_mapped(mapping)) {
 			void *addr;

arch/mips/mm/cache.c

@@ -86,7 +86,7 @@ SYSCALL_DEFINE3(cacheflush, unsigned long, addr, unsigned long, bytes,
 
 void __flush_dcache_page(struct page *page)
 {
-	struct address_space *mapping = page_mapping(page);
+	struct address_space *mapping = page_mapping_file(page);
 	unsigned long addr;
 
 	if (mapping && !mapping_mapped(mapping)) {

arch/nios2/mm/cacheflush.c

@@ -180,7 +180,7 @@ void flush_dcache_page(struct page *page)
 	if (page == ZERO_PAGE(0))
 		return;
 
-	mapping = page_mapping(page);
+	mapping = page_mapping_file(page);
 
 	/* Flush this page if there are aliases. */
 	if (mapping && !mapping_mapped(mapping)) {
@@ -215,7 +215,7 @@ void update_mmu_cache(struct vm_area_struct *vma,
 	if (page == ZERO_PAGE(0))
 		return;
 
-	mapping = page_mapping(page);
+	mapping = page_mapping_file(page);
 	if (!test_and_set_bit(PG_dcache_clean, &page->flags))
 		__flush_dcache_page(mapping, page);
 

arch/parisc/kernel/cache.c

@@ -88,7 +88,8 @@ update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t *ptep)
 		return;
 
 	page = pfn_to_page(pfn);
-	if (page_mapping(page) && test_bit(PG_dcache_dirty, &page->flags)) {
+	if (page_mapping_file(page) &&
+	    test_bit(PG_dcache_dirty, &page->flags)) {
 		flush_kernel_dcache_page_addr(pfn_va(pfn));
 		clear_bit(PG_dcache_dirty, &page->flags);
 	} else if (parisc_requires_coherency())
@@ -304,7 +305,7 @@ __flush_cache_page(struct vm_area_struct *vma, unsigned long vmaddr,
 
 void flush_dcache_page(struct page *page)
 {
-	struct address_space *mapping = page_mapping(page);
+	struct address_space *mapping = page_mapping_file(page);
 	struct vm_area_struct *mpnt;
 	unsigned long offset;
 	unsigned long addr, old_addr = 0;

arch/sh/mm/cache-sh4.c

@@ -112,7 +112,7 @@ static void sh4_flush_dcache_page(void *arg)
 	struct page *page = arg;
 	unsigned long addr = (unsigned long)page_address(page);
 #ifndef CONFIG_SMP
-	struct address_space *mapping = page_mapping(page);
+	struct address_space *mapping = page_mapping_file(page);
 
 	if (mapping && !mapping_mapped(mapping))
 		clear_bit(PG_dcache_clean, &page->flags);

arch/sh/mm/cache-sh7705.c

@@ -136,7 +136,7 @@ static void __flush_dcache_page(unsigned long phys)
 static void sh7705_flush_dcache_page(void *arg)
 {
 	struct page *page = arg;
-	struct address_space *mapping = page_mapping(page);
+	struct address_space *mapping = page_mapping_file(page);
 
 	if (mapping && !mapping_mapped(mapping))
 		clear_bit(PG_dcache_clean, &page->flags);

arch/sparc/kernel/smp_64.c

@@ -929,9 +929,9 @@ static inline void __local_flush_dcache_page(struct page *page)
 #ifdef DCACHE_ALIASING_POSSIBLE
 	__flush_dcache_page(page_address(page),
 			    ((tlb_type == spitfire) &&
-			     page_mapping(page) != NULL));
+			     page_mapping_file(page) != NULL));
 #else
-	if (page_mapping(page) != NULL &&
+	if (page_mapping_file(page) != NULL &&
 	    tlb_type == spitfire)
 		__flush_icache_page(__pa(page_address(page)));
 #endif
@@ -958,7 +958,7 @@ void smp_flush_dcache_page_impl(struct page *page, int cpu)
 
 		if (tlb_type == spitfire) {
 			data0 = ((u64)&xcall_flush_dcache_page_spitfire);
-			if (page_mapping(page) != NULL)
+			if (page_mapping_file(page) != NULL)
 				data0 |= ((u64)1 << 32);
 		} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
 #ifdef DCACHE_ALIASING_POSSIBLE
@@ -994,7 +994,7 @@ void flush_dcache_page_all(struct mm_struct *mm, struct page *page)
 	pg_addr = page_address(page);
 	if (tlb_type == spitfire) {
 		data0 = ((u64)&xcall_flush_dcache_page_spitfire);
-		if (page_mapping(page) != NULL)
+		if (page_mapping_file(page) != NULL)
 			data0 |= ((u64)1 << 32);
 	} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
 #ifdef DCACHE_ALIASING_POSSIBLE

arch/sparc/mm/init_64.c

@@ -206,9 +206,9 @@ inline void flush_dcache_page_impl(struct page *page)
 #ifdef DCACHE_ALIASING_POSSIBLE
 	__flush_dcache_page(page_address(page),
 			    ((tlb_type == spitfire) &&
-			     page_mapping(page) != NULL));
+			     page_mapping_file(page) != NULL));
 #else
-	if (page_mapping(page) != NULL &&
+	if (page_mapping_file(page) != NULL &&
 	    tlb_type == spitfire)
 		__flush_icache_page(__pa(page_address(page)));
 #endif
@@ -490,7 +490,7 @@ void flush_dcache_page(struct page *page)
 
 	this_cpu = get_cpu();
 
-	mapping = page_mapping(page);
+	mapping = page_mapping_file(page);
 	if (mapping && !mapping_mapped(mapping)) {
 		int dirty = test_bit(PG_dcache_dirty, &page->flags);
 		if (dirty) {

arch/sparc/mm/tlb.c

@@ -128,7 +128,7 @@ void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
 			goto no_cache_flush;
 
 		/* A real file page? */
-		mapping = page_mapping(page);
+		mapping = page_mapping_file(page);
 		if (!mapping)
 			goto no_cache_flush;
 

arch/unicore32/mm/flush.c

@@ -83,7 +83,7 @@ void flush_dcache_page(struct page *page)
 	if (page == ZERO_PAGE(0))
 		return;
 
-	mapping = page_mapping(page);
+	mapping = page_mapping_file(page);
 
 	if (mapping && !mapping_mapped(mapping))
 		clear_bit(PG_dcache_clean, &page->flags);

arch/unicore32/mm/mmu.c

@@ -503,7 +503,7 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long addr,
 	if (page == ZERO_PAGE(0))
 		return;
 
-	mapping = page_mapping(page);
+	mapping = page_mapping_file(page);
 	if (!test_and_set_bit(PG_dcache_clean, &page->flags))
 		__flush_dcache_page(mapping, page);
 	if (mapping)

arch/xtensa/mm/cache.c

@@ -127,7 +127,7 @@ EXPORT_SYMBOL(copy_user_highpage);
 
 void flush_dcache_page(struct page *page)
 {
-	struct address_space *mapping = page_mapping(page);
+	struct address_space *mapping = page_mapping_file(page);
 
 	/*
 	 * If we have a mapping but the page is not mapped to user-space

include/linux/mm.h

@@ -1155,6 +1155,7 @@ static inline pgoff_t page_index(struct page *page)
 
 bool page_mapped(struct page *page);
 struct address_space *page_mapping(struct page *page);
+struct address_space *page_mapping_file(struct page *page);
 
 /*
  * Return true only if the page has been allocated with

mm/util.c

@@ -515,6 +515,16 @@ struct address_space *page_mapping(struct page *page)
 }
 EXPORT_SYMBOL(page_mapping);
 
+/*
+ * For file cache pages, return the address_space, otherwise return NULL
+ */
+struct address_space *page_mapping_file(struct page *page)
+{
+	if (unlikely(PageSwapCache(page)))
+		return NULL;
+	return page_mapping(page);
+}
+
 /* Slow path of page_mapcount() for compound pages */
 int __page_mapcount(struct page *page)
 {