xen: privcmd: do not return pages which we have failed to unmap

This failure represents a hypervisor issue, but if it does occur then nothing
good can come of returning pages which still refer to a foreign owned page
into the general allocation pool.

Instead we are forced to leak them. Log that we have done so.

The potential for failure only exists for autotranslated guest (e.g. ARM and
x86 PVH).

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Reviewed-by: David Vrabel <david.vrabel@citrix.com>

diff --git a/drivers/xen/privcmd.c b/drivers/xen/privcmd.c
index 8e74590fa1bb5d9149e17550c3f4ec7b74142983..569a13b9e856de5c3900050d583844f401243e96 100644 (file)
--- a/drivers/xen/privcmd.c
+++ b/drivers/xen/privcmd.c
@@ -533,12 +533,17 @@ static void privcmd_close(struct vm_area_struct *vma)
 {
        struct page **pages = vma->vm_private_data;
        int numpgs = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
+       int rc;
 
        if (!xen_feature(XENFEAT_auto_translated_physmap) || !numpgs || !pages)
                return;
 
-       xen_unmap_domain_mfn_range(vma, numpgs, pages);
-       free_xenballooned_pages(numpgs, pages);
+       rc = xen_unmap_domain_mfn_range(vma, numpgs, pages);
+       if (rc == 0)
+               free_xenballooned_pages(numpgs, pages);
+       else
+               pr_crit("unable to unmap MFN range: leaking %d pages. rc=%d\n",
+                       numpgs, rc);
        kfree(pages);
 }