rbd: end I/O the entire obj_request on error
authorIlya Dryomov <idryomov@gmail.com>
Sat, 25 Apr 2015 12:56:15 +0000 (15:56 +0300)
committerSage Weil <sage@redhat.com>
Fri, 1 May 2015 23:44:30 +0000 (16:44 -0700)
When we end I/O struct request with error, we need to pass
obj_request->length as @nr_bytes so that the entire obj_request worth
of bytes is completed.  Otherwise block layer ends up confused and we
trip on

    rbd_assert(more ^ (which == img_request->obj_request_count));

in rbd_img_obj_callback() due to more being true no matter what.  We
already do it in most cases but we are missing some, in particular
those where we don't even get a chance to submit any obj_requests, due
to an early -ENOMEM for example.

A number of obj_request->xferred assignments seem to be redundant but
I haven't touched any of obj_request->xferred stuff to keep this small
and isolated.

Cc: Alex Elder <elder@linaro.org>
Cc: stable@vger.kernel.org # 3.10+
Reported-by: Shawn Edwards <lesser.evil@gmail.com>
Reviewed-by: Sage Weil <sage@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
drivers/block/rbd.c

index 812523330a78d438e1397b29803f309b8e360211..ec6c5c6e1ac94b2bcbe0619a7fe62b9e7d0ce4a5 100644 (file)
@@ -2264,6 +2264,11 @@ static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request)
                        result, xferred);
                if (!img_request->result)
                        img_request->result = result;
+               /*
+                * Need to end I/O on the entire obj_request worth of
+                * bytes in case of error.
+                */
+               xferred = obj_request->length;
        }
 
        /* Image object requests don't own their page array */