]> www.infradead.org Git - users/dwmw2/linux.git/commitdiff
xen/gntdev: Ignore failure to unmap INVALID_GRANT_HANDLE
authorDemi Marie Obenour <demi@invisiblethingslab.com>
Sun, 10 Jul 2022 23:05:22 +0000 (19:05 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 29 Jul 2022 15:19:07 +0000 (17:19 +0200)
commit 166d3863231667c4f64dee72b77d1102cdfad11f upstream.

The error paths of gntdev_mmap() can call unmap_grant_pages() even
though not all of the pages have been successfully mapped.  This will
trigger the WARN_ON()s in __unmap_grant_pages_done().  The number of
warnings can be very large; I have observed thousands of lines of
warnings in the systemd journal.

Avoid this problem by only warning on unmapping failure if the handle
being unmapped is not INVALID_GRANT_HANDLE.  The handle field of any
page that was not successfully mapped will be INVALID_GRANT_HANDLE, so
this catches all cases where unmapping can legitimately fail.

Fixes: dbe97cff7dd9 ("xen/gntdev: Avoid blocking in unmap_grant_pages()")
Cc: stable@vger.kernel.org
Suggested-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220710230522.1563-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/xen/gntdev.c

index f415c056ff8ab8d808ee2bacfaa3cad57af28204..54fee4087bf1078803c230ad2081aafa8415cf53 100644 (file)
@@ -401,7 +401,8 @@ static void __unmap_grant_pages_done(int result,
        unsigned int offset = data->unmap_ops - map->unmap_ops;
 
        for (i = 0; i < data->count; i++) {
-               WARN_ON(map->unmap_ops[offset+i].status);
+               WARN_ON(map->unmap_ops[offset+i].status &&
+                       map->unmap_ops[offset+i].handle != -1);
                pr_debug("unmap handle=%d st=%d\n",
                        map->unmap_ops[offset+i].handle,
                        map->unmap_ops[offset+i].status);