[PATCH 1/3] mmap: protect from stack expantion into low vm addresses

[PATCH 1/3] mmap: protect from stack expantion into low vm addresses

[Eric Paris]

When adding the new security hooks into mmap to enhance protection
against NULL pointers in the kernel I overlooked that a user may be able
to expand the stack all the way down to low addresses.  This flaw was
pointed out by a PaX/grsecurity developer and this patch should provide
protection against this missed code path.

Signed-off-by: Eric Paris <eparis@redhat.com>

---

** Be very careful applying/rediffing this patch.  Standard 3 lines of
context from git diff will misapply the first hunk to expand_upwards
instead of properly in expand_downwards.  This patch was generated using
-U 4 to make sure it applies in the correct place.  **

 mm/mmap.c |    8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

--- kernel-1/mm/mmap.c
+++ kernel-2/mm/mmap.c
@@ -1614,17 +1614,21 @@ static inline int expand_downwards(struc
 	 * so that the anon_vma locking is not a noop.
 	 */
 	if (unlikely(anon_vma_prepare(vma)))
 		return -ENOMEM;
+
+	address &= PAGE_MASK;
+	error = security_file_mmap(0, 0, 0, 0, address, 1);
+	if (error)
+		return error;
+
 	anon_vma_lock(vma);
 
 	/*
 	 * vma->vm_start/vm_end cannot change under us because the caller
 	 * is required to hold the mmap_sem in read mode.  We need the
 	 * anon_vma lock to serialize against concurrent expand_stacks.
 	 */
-	address &= PAGE_MASK;
-	error = 0;
 
 	/* Somebody else might have raced and expanded it already */
 	if (address < vma->vm_start) {
 		unsigned long size, grow;

[PATCH 1/3] mmap: protect from stack expantion into low vm addresses

[Eric Paris]

Add security checks to make sure we are not attempting to expand the
stack into memory protected by mmap_min_addr

Signed-off-by: Eric Paris <eparis@redhat.com>

---

** Be very careful applying/rediffing this patch.  Standard 3 lines of
context from git diff will misapply the first hunk to expand_upwards
instead of properly in expand_downwards.  This patch was generated using
-U 4 to make sure it applies in the correct place.  **

 mm/mmap.c |    8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

--- kernel-1/mm/mmap.c
+++ kernel-2/mm/mmap.c
@@ -1614,17 +1614,21 @@ static inline int expand_downwards(struc
 	 * so that the anon_vma locking is not a noop.
 	 */
 	if (unlikely(anon_vma_prepare(vma)))
 		return -ENOMEM;
+
+	address &= PAGE_MASK;
+	error = security_file_mmap(0, 0, 0, 0, address, 1);
+	if (error)
+		return error;
+
 	anon_vma_lock(vma);
 
 	/*
 	 * vma->vm_start/vm_end cannot change under us because the caller
 	 * is required to hold the mmap_sem in read mode.  We need the
 	 * anon_vma lock to serialize against concurrent expand_stacks.
 	 */
-	address &= PAGE_MASK;
-	error = 0;
 
 	/* Somebody else might have raced and expanded it already */
 	if (address < vma->vm_start) {
 		unsigned long size, grow;

Re: [PATCH 1/3] mmap: protect from stack expantion into low vm addresses

[James Morris]

On Mon, 26 Nov 2007, Eric Paris wrote:

> Add security checks to make sure we are not attempting to expand the
> stack into memory protected by mmap_min_addr
> 
> Signed-off-by: Eric Paris <eparis@redhat.com>

Please include the LSMs list in the CC line (added again) for posts 
relating to security.

Applied to 
git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/selinux-2.6.git#for-akpm

> ---
> 
> ** Be very careful applying/rediffing this patch.  Standard 3 lines of
> context from git diff will misapply the first hunk to expand_upwards
> instead of properly in expand_downwards.  This patch was generated using
> -U 4 to make sure it applies in the correct place.  **

Seems to have applied correctly for me.


> 
>  mm/mmap.c |    8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> --- kernel-1/mm/mmap.c
> +++ kernel-2/mm/mmap.c
> @@ -1614,17 +1614,21 @@ static inline int expand_downwards(struc
>  	 * so that the anon_vma locking is not a noop.
>  	 */
>  	if (unlikely(anon_vma_prepare(vma)))
>  		return -ENOMEM;
> +
> +	address &= PAGE_MASK;
> +	error = security_file_mmap(0, 0, 0, 0, address, 1);
> +	if (error)
> +		return error;
> +
>  	anon_vma_lock(vma);
>  
>  	/*
>  	 * vma->vm_start/vm_end cannot change under us because the caller
>  	 * is required to hold the mmap_sem in read mode.  We need the
>  	 * anon_vma lock to serialize against concurrent expand_stacks.
>  	 */
> -	address &= PAGE_MASK;
> -	error = 0;
>  
>  	/* Somebody else might have raced and expanded it already */
>  	if (address < vma->vm_start) {
>  		unsigned long size, grow;
> 
> 

-- 
James Morris
<jmorris@namei.org>