fork,memcg: fix crash in free_thread_stack on memcg charge fail

Commit 9b6f7e1 ("mm: rework memcg kernel stack accounting") will
result in fork failing if allocating a kernel stack for a task in
dup_task_struct exceeds the kernel memory allowance for that cgroup.

Unfortunately, it also results in a crash.

This is due to the code jumping to free_stack and calling
free_thread_stack when the memcg kernel stack charge fails, but without
tsk->stack pointing at the freshly allocated stack.

This in turn results in the vfree_atomic in free_thread_stack oopsing
with a backtrace like this:

#5 [ffffc900244efc88] die at ffffffff8101f0ab
 #6 [ffffc900244efcb8] do_general_protection at ffffffff8101cb86
 #7 [ffffc900244efce0] general_protection at ffffffff818ff082
    [exception RIP: llist_add_batch+7]
    RIP: ffffffff8150d487  RSP: ffffc900244efd98  RFLAGS: 00010282
    RAX: 0000000000000000  RBX: ffff88085ef55980  RCX: 0000000000000000
    RDX: ffff88085ef55980  RSI: 343834343531203a  RDI: 343834343531203a
    RBP: ffffc900244efd98   R8: 0000000000000001   R9: ffff8808578c3600
    R10: 0000000000000000  R11: 0000000000000001  R12: ffff88029f6c21c0
    R13: 0000000000000286  R14: ffff880147759b00  R15: 0000000000000000
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 #8 [ffffc900244efda0] vfree_atomic at ffffffff811df2c7
 #9 [ffffc900244efdb8] copy_process at ffffffff81086e37
#10 [ffffc900244efe98] _do_fork at ffffffff810884e0
#11 [ffffc900244eff10] sys_vfork at ffffffff810887ff
#12 [ffffc900244eff20] do_syscall_64 at ffffffff81002a43
    RIP: 000000000049b948  RSP: 00007ffcdb307830  RFLAGS: 00000246
    RAX: ffffffffffffffda  RBX: 0000000000896030  RCX: 000000000049b948
    RDX: 0000000000000000  RSI: 00007ffcdb307790  RDI: 00000000005d7421
    RBP: 000000000067370f   R8: 00007ffcdb3077b0   R9: 000000000001ed00
    R10: 0000000000000008  R11: 0000000000000246  R12: 0000000000000040
    R13: 000000000000000f  R14: 0000000000000000  R15: 000000000088d018
    ORIG_RAX: 000000000000003a  CS: 0033  SS: 002b

The simplest fix is to assign tsk->stack right where it is allocated.

Link: http://lkml.kernel.org/r/[email protected]
Fixes: 9b6f7e1 ("mm: rework memcg kernel stack accounting")
Signed-off-by: Rik van Riel <[email protected]>
Acked-by: Roman Gushchin <[email protected]>
Acked-by: Michal Hocko <[email protected]>
Cc: Shakeel Butt <[email protected]>
Cc: Johannes Weiner <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

kernel/fork.c

@@ -240,8 +240,10 @@ static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
 	 * free_thread_stack() can be called in interrupt context,
 	 * so cache the vm_struct.
 	 */
-	if (stack)
+	if (stack) {
 		tsk->stack_vm_area = find_vm_area(stack);
+		tsk->stack = stack;
+	}
 	return stack;
 #else
 	struct page *page = alloc_pages_node(node, THREADINFO_GFP,
@@ -288,7 +290,10 @@ static struct kmem_cache *thread_stack_cache;
 static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
 						  int node)
 {
-	return kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
+	unsigned long *stack;
+	stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
+	tsk->stack = stack;
+	return stack;
 }
 
 static void free_thread_stack(struct task_struct *tsk)