Re: [PATCH v3 3/3] mm/mempolicy: add nodes_empty check in SYSC_migrate_pages

Hi Vlastimil,

Thanks for your comment!
On 2017/11/28 1:25, Vlastimil Babka wrote:

On 11/17/2017 02:37 AM, Yisheng Xie wrote:

quoted

As manpage of migrate_pages, the errno should be set to EINVAL when
none of the node IDs specified by new_nodes are on-line and allowed
by the process's current cpuset context, or none of the specified
nodes contain memory. However, when test by following case:

	new_nodes = 0;
	old_nodes = 0xf;
	ret = migrate_pages(pid, old_nodes, new_nodes, MAX);

The ret will be 0 and no errno is set. As the new_nodes is empty,
we should expect EINVAL as documented.

To fix the case like above, this patch check whether target nodes
AND current task_nodes is empty, and then check whether AND
node_states[N_MEMORY] is empty.

Signed-off-by: Yisheng Xie <redacted>
---
 mm/mempolicy.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 65df28d..f604b22 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c

@@ -1433,10 +1433,14 @@ static int copy_nodes_to_user(unsigned long __user *mask, unsigned long maxnode,
 		goto out_put;
 	}

Let me add the whole preceding that ends on the lines above:

        task_nodes = cpuset_mems_allowed(task);
        /* Is the user allowed to access the target nodes? */
        if (!nodes_subset(*new, task_nodes) && !capable(CAP_SYS_NICE)) {
                err = -EPERM;
                goto out_put;
        }

quoted

 
-	if (!nodes_subset(*new, node_states[N_MEMORY])) {
-		err = -EINVAL;
+	task_nodes = cpuset_mems_allowed(current);
+	nodes_and(*new, *new, task_nodes);
+	if (nodes_empty(*new))
+		goto out_put;

So if we have CAP_SYS_NICE, we pass (or rather skip) the EPERM check
above, but the current cpuset restriction still applies regardless. This
doesn't make sense to me? If I get Christoph right in the v2 discussion,
then CAP_SYS_NICE should not allow current cpuset escape. 

hmm, maybe I do not get what you mean, the patch seems do not *escape* the
current cpuset?  if CAP_SYS_NICE it also check current cpuset, right?

In that case,
we should remove the CAP_SYS_NICE check from the EPERM check? Also
should it be a subset check, or a non-empty-intersection check?

So you mean:
1. we should remove the EPERM check above?
2. Not sure we should use subset check, or a non-empty-intersection for current cpuset?
(Please let me know, if have other points.)

For 1: I have checked the manpage of capabilities[1]:
CAP_SYS_NICE
	[...]
	*apply migrate_pages(2) to arbitrary processes* and allow
        processes to be migrated to arbitrary nodes;

	apply move_pages(2) to arbitrary processes;
	[...]

Therefore, IMO, EPERM check should be something like:
	if (currtent->mm != task->mm && !capable(CAP_SYS_NICE)) { // or if (currtent != task && !capable(CAP_SYS_NICE)) ?
		err = -EPERM;
		goto out_put;
	}
And I kept it as unchanged to follow the original code's meaning.(For move_pages
also use the the logical to check EPERM). I also did not want to break the existing code. :)

For 2: we should follow the manpage of migrate_pages about EINVAL, as your listed in
the former discussion:
	  EINVAL... Or, _none_ of the node IDs specified by new_nodes are
 	 	on-line and allowed by the process's current cpuset context, or none of
  		the specified nodes contain memory.

So a non-empty-intersection check for current cpuset should be enough, right?
And Christoph seems do _not oppose_ this point. (I not sure whether he is *agree* or not).

[1] http://man7.org/linux/man-pages/man7/capabilities.7.html

Note there's still a danger that we are breaking existing code so this
will have to be reverted in any case...

I am not oppose if you want to revert this patch, but we should find a
correct way to fix the case above, right? Maybe anther version or a fix to fold?

Thanks
Yisheng Xie

quoted

+
+	nodes_and(*new, *new, node_states[N_MEMORY]);
+	if (nodes_empty(*new))
 		goto out_put;
-	}
 
 	err = security_task_movememory(task);
 	if (err)

.