mempolicy: reduce stack size of migrate_pages()
authorKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Tue, 10 Aug 2010 00:19:01 +0000 (17:19 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 10 Aug 2010 03:44:58 +0000 (20:44 -0700)
migrate_pages() is using >500 bytes stack. Reduce it.

   mm/mempolicy.c: In function 'sys_migrate_pages':
   mm/mempolicy.c:1344: warning: the frame size of 528 bytes is larger than 512 bytes

[akpm@linux-foundation.org: don't play with a might-be-NULL pointer]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/mempolicy.c

index 8a73708d59bb24dc37a8e43ee9ce329035aa9383..f969da5dd8a2136b92aab7e7c1a6ecc9083d6cb9 100644 (file)
@@ -1275,33 +1275,42 @@ SYSCALL_DEFINE4(migrate_pages, pid_t, pid, unsigned long, maxnode,
                const unsigned long __user *, new_nodes)
 {
        const struct cred *cred = current_cred(), *tcred;
-       struct mm_struct *mm;
+       struct mm_struct *mm = NULL;
        struct task_struct *task;
-       nodemask_t old;
-       nodemask_t new;
        nodemask_t task_nodes;
        int err;
+       nodemask_t *old;
+       nodemask_t *new;
+       NODEMASK_SCRATCH(scratch);
+
+       if (!scratch)
+               return -ENOMEM;
 
-       err = get_nodes(&old, old_nodes, maxnode);
+       old = &scratch->mask1;
+       new = &scratch->mask2;
+
+       err = get_nodes(old, old_nodes, maxnode);
        if (err)
-               return err;
+               goto out;
 
-       err = get_nodes(&new, new_nodes, maxnode);
+       err = get_nodes(new, new_nodes, maxnode);
        if (err)
-               return err;
+               goto out;
 
        /* Find the mm_struct */
        read_lock(&tasklist_lock);
        task = pid ? find_task_by_vpid(pid) : current;
        if (!task) {
                read_unlock(&tasklist_lock);
-               return -ESRCH;
+               err = -ESRCH;
+               goto out;
        }
        mm = get_task_mm(task);
        read_unlock(&tasklist_lock);
 
+       err = -EINVAL;
        if (!mm)
-               return -EINVAL;
+               goto out;
 
        /*
         * Check if this process has the right to modify the specified
@@ -1322,12 +1331,12 @@ SYSCALL_DEFINE4(migrate_pages, pid_t, pid, unsigned long, maxnode,
 
        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)) {
+       if (!nodes_subset(*new, task_nodes) && !capable(CAP_SYS_NICE)) {
                err = -EPERM;
                goto out;
        }
 
-       if (!nodes_subset(new, node_states[N_HIGH_MEMORY])) {
+       if (!nodes_subset(*new, node_states[N_HIGH_MEMORY])) {
                err = -EINVAL;
                goto out;
        }
@@ -1336,10 +1345,13 @@ SYSCALL_DEFINE4(migrate_pages, pid_t, pid, unsigned long, maxnode,
        if (err)
                goto out;
 
-       err = do_migrate_pages(mm, &old, &new,
+       err = do_migrate_pages(mm, old, new,
                capable(CAP_SYS_NICE) ? MPOL_MF_MOVE_ALL : MPOL_MF_MOVE);
 out:
-       mmput(mm);
+       if (mm)
+               mmput(mm);
+       NODEMASK_SCRATCH_FREE(scratch);
+
        return err;
 }