Skip to contentSkip to conversation

linux/blktap: don't accesses deallocated data

PATCH In review Jan Beulich 2 participants 2 messages Apr 16, 2009 – Apr 17, 2009 Xen hypervisor interface

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 linux/blktap: don't accesses deallocated data

Short of getting an answer to the respective quesry about a month ago, here is a patch that addresses the described problem (despite me not feeling well without any kind of reaction to the original problem description, as this area of code is certainly not one of those I'd consider myself reasonably knowledgeable about).

Dereferencing filp->private_data->vma in the file_operations.release actor isn't permitted, as the vma generally has been destroyed by that time. The kfree()ing of vma->vm_private_data must be done in the vm_operations.close actor, and the call to zap_page_range() seems redundant with the caller of that actor altogether.

As usual, written and tested on 2.6.27.21 and made apply to the 2.6.18 tree without further testing.

drivers/xen/blktap/blktap.c +32 -35
@@ -296,6 +296,10 @@ static inline int OFFSET_TO_SEG(int offs
 /******************************************************************
  * BLKTAP VM OPS
  */
+struct tap_vma_priv {
+	tap_blkif_t *info;
+	struct page *map[];
+};
 
 static struct page *blktap_nopage(struct vm_area_struct *vma,
 				  unsigned long address,
@@ -316,7 +320,7 @@ static pte_t blktap_clear_pte(struct vm_
 	int offset, seg, usr_idx, pending_idx, mmap_idx;
 	unsigned long uvstart = vma->vm_start + (RING_PAGES << PAGE_SHIFT);
 	unsigned long kvaddr;
-	struct page **map;
+	struct tap_vma_priv *priv;
 	struct page *pg;
 	struct grant_handle_pair *khandle;
 	struct gnttab_unmap_grant_ref unmap[2];
@@ -331,12 +335,12 @@ static pte_t blktap_clear_pte(struct vm_
 					       ptep, is_fullmm);
 
 	info = vma->vm_file->private_data;
-	map = vma->vm_private_data;
+	priv = vma->vm_private_data;
 
 	/* TODO Should these be changed to if statements? */
 	BUG_ON(!info);
 	BUG_ON(!info->idx_map);
-	BUG_ON(!map);
+	BUG_ON(!priv);
 
 	offset = (int) ((uvaddr - uvstart) >> PAGE_SHIFT);
 	usr_idx = OFFSET_TO_USR_IDX(offset);
@@ -348,7 +352,7 @@ static pte_t blktap_clear_pte(struct vm_
 	kvaddr = idx_to_kaddr(mmap_idx, pending_idx, seg);
 	pg = pfn_to_page(__pa(kvaddr) >> PAGE_SHIFT);
 	ClearPageReserved(pg);
-	map[offset + RING_PAGES] = NULL;
+	priv->map[offset + RING_PAGES] = NULL;
 
 	khandle = &pending_handle(mmap_idx, pending_idx, seg);
 
@@ -389,9 +393,20 @@ static pte_t blktap_clear_pte(struct vm_
 	return copy;
 }
 
+static void blktap_vma_close(struct vm_area_struct *vma)
+{
+	struct tap_vma_priv *priv = vma->vm_private_data;
+
+	if (priv) {
+		priv->info->vma = NULL;
+		kfree(priv);
+	}
+}
+
 struct vm_operations_struct blktap_vm_ops = {
 	nopage:   blktap_nopage,
 	zap_pte:  blktap_clear_pte,
+	close:    blktap_vma_close,
 };
 
 /******************************************************************
@@ -609,21 +624,6 @@ static int blktap_release(struct inode *
 	ClearPageReserved(virt_to_page(info->ufe_ring.sring));
 	free_page((unsigned long) info->ufe_ring.sring);
 
-	/* Clear any active mappings and free foreign map table */
-	if (info->vma) {
-		struct mm_struct *mm = info->vma->vm_mm;
-
-		down_write(&mm->mmap_sem);
-		zap_page_range(
-			info->vma, info->vma->vm_start, 
-			info->vma->vm_end - info->vma->vm_start, NULL);
-		up_write(&mm->mmap_sem);
-
-		kfree(info->vma->vm_private_data);
-
-		info->vma = NULL;
-	}
-
 	if (info->idx_map) {
 		kfree(info->idx_map);
 		info->idx_map = NULL;
@@ -662,8 +662,7 @@ static int blktap_release(struct inode *
 static int blktap_mmap(struct file *filp, struct vm_area_struct *vma)
 {
 	int size;
-	struct page **map;
-	int i;
+	struct tap_vma_priv *priv;
 	tap_blkif_t *info = filp->private_data;
 	int ret;
 
@@ -700,18 +699,16 @@ static int blktap_mmap(struct file *filp
 	}
 
 	/* Mark this VM as containing foreign pages, and set up mappings. */
-	map = kzalloc(((vma->vm_end - vma->vm_start) >> PAGE_SHIFT)
-		      * sizeof(struct page *),
-		      GFP_KERNEL);
-	if (map == NULL) {
+	priv = kzalloc(sizeof(*priv) + ((vma->vm_end - vma->vm_start)
+					>> PAGE_SHIFT) * sizeof(*priv->map),
+		       GFP_KERNEL);
+	if (priv == NULL) {
 		WPRINTK("Couldn't alloc VM_FOREIGN map.\n");
 		goto fail;
 	}
+	priv->info = info;
 
-	for (i = 0; i < ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT); i++)
-		map[i] = NULL;
-    
-	vma->vm_private_data = map;
+	vma->vm_private_data = priv;
 	vma->vm_flags |= VM_FOREIGN;
 	vma->vm_flags |= VM_DONTCOPY;
 
@@ -1192,7 +1189,7 @@ static int blktap_read_ufe_ring(tap_blki
 		for (j = 0; j < pending_req->nr_pages; j++) {
 
 			unsigned long kvaddr, uvaddr;
-			struct page **map = info->vma->vm_private_data;
+			struct tap_vma_priv *priv = info->vma->vm_private_data;
 			struct page *pg;
 			int offset;
 
@@ -1203,7 +1200,7 @@ static int blktap_read_ufe_ring(tap_blki
 			ClearPageReserved(pg);
 			offset = (uvaddr - info->vma->vm_start) 
 				>> PAGE_SHIFT;
-			map[offset] = NULL;
+			priv->map[offset] = NULL;
 		}
 		fast_flush_area(pending_req, pending_idx, usr_idx, info->minor);
 		info->idx_map[usr_idx] = INVALID_REQ;
@@ -1369,6 +1366,7 @@ static void dispatch_rw_block_io(blkif_t
 	unsigned int nseg;
 	int ret, i, nr_sects = 0;
 	tap_blkif_t *info;
+	struct tap_vma_priv *priv;
 	blkif_request_t *target;
 	int pending_idx = RTN_PEND_IDX(pending_req,pending_req->mem_idx);
 	int usr_idx;
@@ -1434,6 +1432,7 @@ static void dispatch_rw_block_io(blkif_t
 	pending_req->status    = BLKIF_RSP_OKAY;
 	pending_req->nr_pages  = nseg;
 	op = 0;
+	priv = info->vma->vm_private_data;
 	mm = info->vma->vm_mm;
 	if (!xen_feature(XENFEAT_auto_translated_physmap))
 		down_write(&mm->mmap_sem);
@@ -1517,8 +1516,7 @@ static void dispatch_rw_block_io(blkif_t
 							  >> PAGE_SHIFT));
 			offset = (uvaddr - info->vma->vm_start) >> PAGE_SHIFT;
 			pg = pfn_to_page(__pa(kvaddr) >> PAGE_SHIFT);
-			((struct page **)info->vma->vm_private_data)[offset] =
-				pg;
+			priv->map[offset] = pg;
 		}
 	} else {
 		for (i = 0; i < nseg; i++) {
@@ -1545,8 +1543,7 @@ static void dispatch_rw_block_io(blkif_t
 
 			offset = (uvaddr - info->vma->vm_start) >> PAGE_SHIFT;
 			pg = pfn_to_page(__pa(kvaddr) >> PAGE_SHIFT);
-			((struct page **)info->vma->vm_private_data)[offset] =
-				pg;
+			priv->map[offset] = pg;
 		}
 	}
 
#1 lore raw

Hi.

What about partial munmap? Although blktap doesn't. (NOTE:I'm not objecting about committing this patch. This is surely the first step.)
And I think that gntdev is also broken in the similar way.

vm_private is simply copied to newly created vma's to split vma. So some kind of refcount should be done for partial munmap. There are a few examples in the linux tree.
xenfb.c would be a good example. (xenfb_vm_{open, close}()) And the ia64 xen foreign domain page mapping is also a example which I implemented.
(xen_ia64_privcmd_vma_{open, close}() in arch/ia64/xen/hypervisor.c)

thanks,

Quoted 203 lines · Jan Beulich

Short of getting an answer to the respective quesry about a month ago, here is a patch that addresses the described problem (despite me not feeling well without any kind of reaction to the original problem description, as this area of code is certainly not one of those I'd consider myself reasonably knowledgeable about).

Dereferencing filp->private_data->vma in the file_operations.release actor isn't permitted, as the vma generally has been destroyed by that time. The kfree()ing of vma->vm_private_data must be done in the vm_operations.close actor, and the call to zap_page_range() seems redundant with the caller of that actor altogether.

As usual, written and tested on 2.6.27.21 and made apply to the 2.6.18 tree without further testing.

drivers/xen/blktap/blktap.c +32 -35
@@ -296,6 +296,10 @@ static inline int OFFSET_TO_SEG(int offs
 /******************************************************************
  * BLKTAP VM OPS
  */
+struct tap_vma_priv {
+	tap_blkif_t *info;
+	struct page *map[];
+};
 
 static struct page *blktap_nopage(struct vm_area_struct *vma,
 				  unsigned long address,
@@ -316,7 +320,7 @@ static pte_t blktap_clear_pte(struct vm_
 	int offset, seg, usr_idx, pending_idx, mmap_idx;
 	unsigned long uvstart = vma->vm_start + (RING_PAGES << PAGE_SHIFT);
 	unsigned long kvaddr;
-	struct page **map;
+	struct tap_vma_priv *priv;
 	struct page *pg;
 	struct grant_handle_pair *khandle;
 	struct gnttab_unmap_grant_ref unmap[2];
@@ -331,12 +335,12 @@ static pte_t blktap_clear_pte(struct vm_
 					       ptep, is_fullmm);
 
 	info = vma->vm_file->private_data;
-	map = vma->vm_private_data;
+	priv = vma->vm_private_data;
 
 	/* TODO Should these be changed to if statements? */
 	BUG_ON(!info);
 	BUG_ON(!info->idx_map);
-	BUG_ON(!map);
+	BUG_ON(!priv);
 
 	offset = (int) ((uvaddr - uvstart) >> PAGE_SHIFT);
 	usr_idx = OFFSET_TO_USR_IDX(offset);
@@ -348,7 +352,7 @@ static pte_t blktap_clear_pte(struct vm_
 	kvaddr = idx_to_kaddr(mmap_idx, pending_idx, seg);
 	pg = pfn_to_page(__pa(kvaddr) >> PAGE_SHIFT);
 	ClearPageReserved(pg);
-	map[offset + RING_PAGES] = NULL;
+	priv->map[offset + RING_PAGES] = NULL;
 
 	khandle = &pending_handle(mmap_idx, pending_idx, seg);
 
@@ -389,9 +393,20 @@ static pte_t blktap_clear_pte(struct vm_
 	return copy;
 }
 
+static void blktap_vma_close(struct vm_area_struct *vma)
+{
+	struct tap_vma_priv *priv = vma->vm_private_data;
+
+	if (priv) {
+		priv->info->vma = NULL;
+		kfree(priv);
+	}
+}
+
 struct vm_operations_struct blktap_vm_ops = {
 	nopage:   blktap_nopage,
 	zap_pte:  blktap_clear_pte,
+	close:    blktap_vma_close,
 };
 
 /******************************************************************
@@ -609,21 +624,6 @@ static int blktap_release(struct inode *
 	ClearPageReserved(virt_to_page(info->ufe_ring.sring));
 	free_page((unsigned long) info->ufe_ring.sring);
 
-	/* Clear any active mappings and free foreign map table */
-	if (info->vma) {
-		struct mm_struct *mm = info->vma->vm_mm;
-
-		down_write(&mm->mmap_sem);
-		zap_page_range(
-			info->vma, info->vma->vm_start, 
-			info->vma->vm_end - info->vma->vm_start, NULL);
-		up_write(&mm->mmap_sem);
-
-		kfree(info->vma->vm_private_data);
-
-		info->vma = NULL;
-	}
-
 	if (info->idx_map) {
 		kfree(info->idx_map);
 		info->idx_map = NULL;
@@ -662,8 +662,7 @@ static int blktap_release(struct inode *
 static int blktap_mmap(struct file *filp, struct vm_area_struct *vma)
 {
 	int size;
-	struct page **map;
-	int i;
+	struct tap_vma_priv *priv;
 	tap_blkif_t *info = filp->private_data;
 	int ret;
 
@@ -700,18 +699,16 @@ static int blktap_mmap(struct file *filp
 	}
 
 	/* Mark this VM as containing foreign pages, and set up mappings. */
-	map = kzalloc(((vma->vm_end - vma->vm_start) >> PAGE_SHIFT)
-		      * sizeof(struct page *),
-		      GFP_KERNEL);
-	if (map == NULL) {
+	priv = kzalloc(sizeof(*priv) + ((vma->vm_end - vma->vm_start)
+					>> PAGE_SHIFT) * sizeof(*priv->map),
+		       GFP_KERNEL);
+	if (priv == NULL) {
 		WPRINTK("Couldn't alloc VM_FOREIGN map.\n");
 		goto fail;
 	}
+	priv->info = info;
 
-	for (i = 0; i < ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT); i++)
-		map[i] = NULL;
-    
-	vma->vm_private_data = map;
+	vma->vm_private_data = priv;
 	vma->vm_flags |= VM_FOREIGN;
 	vma->vm_flags |= VM_DONTCOPY;
 
@@ -1192,7 +1189,7 @@ static int blktap_read_ufe_ring(tap_blki
 		for (j = 0; j < pending_req->nr_pages; j++) {
 
 			unsigned long kvaddr, uvaddr;
-			struct page **map = info->vma->vm_private_data;
+			struct tap_vma_priv *priv = info->vma->vm_private_data;
 			struct page *pg;
 			int offset;
 
@@ -1203,7 +1200,7 @@ static int blktap_read_ufe_ring(tap_blki
 			ClearPageReserved(pg);
 			offset = (uvaddr - info->vma->vm_start) 
 				>> PAGE_SHIFT;
-			map[offset] = NULL;
+			priv->map[offset] = NULL;
 		}
 		fast_flush_area(pending_req, pending_idx, usr_idx, info->minor);
 		info->idx_map[usr_idx] = INVALID_REQ;
@@ -1369,6 +1366,7 @@ static void dispatch_rw_block_io(blkif_t
 	unsigned int nseg;
 	int ret, i, nr_sects = 0;
 	tap_blkif_t *info;
+	struct tap_vma_priv *priv;
 	blkif_request_t *target;
 	int pending_idx = RTN_PEND_IDX(pending_req,pending_req->mem_idx);
 	int usr_idx;
@@ -1434,6 +1432,7 @@ static void dispatch_rw_block_io(blkif_t
 	pending_req->status    = BLKIF_RSP_OKAY;
 	pending_req->nr_pages  = nseg;
 	op = 0;
+	priv = info->vma->vm_private_data;
 	mm = info->vma->vm_mm;
 	if (!xen_feature(XENFEAT_auto_translated_physmap))
 		down_write(&mm->mmap_sem);
@@ -1517,8 +1516,7 @@ static void dispatch_rw_block_io(blkif_t
 							  >> PAGE_SHIFT));
 			offset = (uvaddr - info->vma->vm_start) >> PAGE_SHIFT;
 			pg = pfn_to_page(__pa(kvaddr) >> PAGE_SHIFT);
-			((struct page **)info->vma->vm_private_data)[offset] =
-				pg;
+			priv->map[offset] = pg;
 		}
 	} else {
 		for (i = 0; i < nseg; i++) {
@@ -1545,8 +1543,7 @@ static void dispatch_rw_block_io(blkif_t
 
 			offset = (uvaddr - info->vma->vm_start) >> PAGE_SHIFT;
 			pg = pfn_to_page(__pa(kvaddr) >> PAGE_SHIFT);
-			((struct page **)info->vma->vm_private_data)[offset] =
-				pg;
+			priv->map[offset] = pg;
 		}
 	}
 

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

Signature · 1 line
yamahata