Re: [PATCH] mm/mempolicy: add node_empty check in SYSC_migrate_pages
From: Vlastimil Babka
Date: Wed Oct 18 2017 - 03:54:16 EST
+CC linux-api
On 10/18/2017 03:37 AM, Yisheng Xie wrote:
> As Xiaojun reported the ltp of migrate_pages01 will failed on ARCH arm64
> system whoes has 4 nodes[0...3], all have memory and CONFIG_NODES_SHIFT=2:
>
> migrate_pages01 0 TINFO : test_invalid_nodes
> migrate_pages01 14 TFAIL : migrate_pages_common.c:45: unexpected failure - returned value = 0, expected: -1
> migrate_pages01 15 TFAIL : migrate_pages_common.c:55: call succeeded unexpectedly
>
> In this case the test_invalid_nodes of migrate_pages01 will call:
> SYSC_migrate_pages as:
>
> migrate_pages(0, , {0x0000000000000001}, 64, , {0x0000000000000010}, 64) = 0
is 64 here the maxnode parameter of migrate_pages() ?
> For MAX_NUMNODES is 4, so 0x10 nodemask will tread as empty set which makes
> nodes_subset(*new, node_states[N_MEMORY])
According to manpage of migrate_pages:
EINVAL The value specified by maxnode exceeds a kernel-imposed
limit. Or, old_nodes or new_nodes specifies one or more node IDs that
are greater than the maximum supported node ID. 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.
if maxnode parameter is 64, but MAX_NUMNODES ("kernel-imposed limit") is
4, we should get EINVAL just because of that. I don't see such check in
the migrate_pages implementation though. But then at least the
"new_nodes specifies one or more node IDs that are greater than the
maximum supported node ID" part should trigger here, because you have
node number 8 set in the new_nodes nodemask, right?
get_nodes() should be checking this according to comment:
/* When the user specified more nodes than supported just check
if the non supported part is all zero. */
Somehow that doesn't seem to work then? I think we should look into
this. Your patch may still be needed, or not, after that is resolved.
> return true, as empty set is subset of any set.
>
> So this is a common issue which also can happens in X86_64 system eg. 8 nodes[0..7],
> all with memory and CONFIG_NODES_SHIFT=3. Fix it by adding node_empty check in
> SYSC_migrate_pages.
>
> Reported-by: Tan Xiaojun <tanxiaojun@xxxxxxxxxx>
> Signed-off-by: Yisheng Xie <xieyisheng1@xxxxxxxxxx>
> ---
> mm/mempolicy.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/mm/mempolicy.c b/mm/mempolicy.c
> index a2af6d5..1dfd3cc 100644
> --- a/mm/mempolicy.c
> +++ b/mm/mempolicy.c
> @@ -1388,6 +1388,11 @@ static int copy_nodes_to_user(unsigned long __user *mask, unsigned long maxnode,
> if (err)
> goto out;
>
> + if (nodes_empty(*new)) {
> + err = -EINVAL;
> + goto out;
> + }
> +
> /* Find the mm_struct */
> rcu_read_lock();
> task = pid ? find_task_by_vpid(pid) : current;
>