Thread (9 messages) flat view 9 messages, 3 authors, 2017-11-28

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

From: Vlastimil Babka <hidden>
Date: 2017-11-27 17:27:18
Also in: linux-mm, lkml

On 11/17/2017 02:37 AM, Yisheng Xie wrote:
quoted hunk ↗ jump to hunk
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;
        }
 
-	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. 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?

Note there's still a danger that we are breaking existing code so this
will have to be reverted in any case...
+
+	nodes_and(*new, *new, node_states[N_MEMORY]);
+	if (nodes_empty(*new))
 		goto out_put;
-	}
 
 	err = security_task_movememory(task);
 	if (err)
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help