fix NULL pointer dereference in __vm_enough_memory()

The new exec code inserts an accounted vma into an mm struct which is not
current->mm.  The existing memory check code has a hard coded assumption
that this does not happen as does the security code.

As the correct mm is known we pass the mm to the security method and the
helper function.  A new security test is added for the case where we need
to pass the mm and the existing one is modified to pass current->mm to
avoid the need to change large amounts of code.

(Thanks to Tobias for fixing rejects and testing)

Signed-off-by: Alan Cox <alan@redhat.com>
Cc: WU Fengguang <wfg@mail.ustc.edu.cn>
Cc: James Morris <jmorris@redhat.com>
Cc: Tobias Diedrich <ranma+kernel@tdiedrich.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 2 insertions, 2 deletions

diff --git a/security/commoncap.c b/security/commoncap.c
index 338606eb7238..7520361663e8 100644
--- a/security/commoncap.c
+++ b/security/commoncap.c
@@ -315,13 +315,13 @@ int cap_syslog (int type)
 	return 0;
 }
 
-int cap_vm_enough_memory(long pages)
+int cap_vm_enough_memory(struct mm_struct *mm, long pages)
 {
 	int cap_sys_admin = 0;
 
 	if (cap_capable(current, CAP_SYS_ADMIN) == 0)
 		cap_sys_admin = 1;
-	return __vm_enough_memory(pages, cap_sys_admin);
+	return __vm_enough_memory(mm, pages, cap_sys_admin);
 }
 
 EXPORT_SYMBOL(cap_capable);