OSDN Git Service

cachefiles: Wait rather than BUG'ing on "Unexpected object collision"
authorKiran Kumar Modukuri <kiran.modukuri@gmail.com>
Thu, 21 Jun 2018 20:25:53 +0000 (13:25 -0700)
committerDavid Howells <dhowells@redhat.com>
Wed, 25 Jul 2018 13:49:00 +0000 (14:49 +0100)
If we meet a conflicting object that is marked FSCACHE_OBJECT_IS_LIVE in
the active object tree, we have been emitting a BUG after logging
information about it and the new object.

Instead, we should wait for the CACHEFILES_OBJECT_ACTIVE flag to be cleared
on the old object (or return an error).  The ACTIVE flag should be cleared
after it has been removed from the active object tree.  A timeout of 60s is
used in the wait, so we shouldn't be able to get stuck there.

Fixes: 9ae326a69004 ("CacheFiles: A cache that backs onto a mounted filesystem")
Signed-off-by: Kiran Kumar Modukuri <kiran.modukuri@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
fs/cachefiles/namei.c

index b5d6dd7..af2b17b 100644 (file)
@@ -192,7 +192,6 @@ wait_for_old_object:
                pr_err("\n");
                pr_err("Error: Unexpected object collision\n");
                cachefiles_printk_object(object, xobject);
-               BUG();
        }
        atomic_inc(&xobject->usage);
        write_unlock(&cache->active_lock);