[PATCH] rbd: add barriers near done flag operations

[Alex Elder]

Somehow, I missed this little item in Documentation/atomic_ops.txt:
    *** WARNING: atomic_read() and atomic_set() DO NOT IMPLY BARRIERS! ***

Create and use some helper functions that include the proper memory
barriers for manipulating the done field.

Signed-off-by: Alex Elder <el...@inktank.com>
---
 drivers/block/rbd.c |   30 ++++++++++++++++++++++++------
 1 file changed, 24 insertions(+), 6 deletions(-)

diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
index 91983a60..982963e 100644
--- a/drivers/block/rbd.c
+++ b/drivers/block/rbd.c
@@ -1216,10 +1216,28 @@ static int rbd_obj_request_wait(struct
rbd_obj_request *obj_request)
        return wait_for_completion_interruptible(&obj_request->completion);
 }

+static void obj_request_done_init(struct rbd_obj_request *obj_request)
+{
+       atomic_set(&obj_request->done, 0);
+       smp_wmb();
+}
+
+static void obj_request_done_set(struct rbd_obj_request *obj_request)
+{
+       atomic_set(&obj_request->done, 1);
+       smp_wmb();
+}
+
+static bool obj_request_done_test(struct rbd_obj_request *obj_request)
+{
+       smp_rmb();
+       return atomic_read(&obj_request->done) != 0;
+}
+
 static void rbd_osd_trivial_callback(struct rbd_obj_request *obj_request,
                                struct ceph_osd_op *op)
 {
-       atomic_set(&obj_request->done, 1);
+       obj_request_done_set(obj_request);
 }

 static void rbd_obj_request_complete(struct rbd_obj_request *obj_request)
@@ -1249,14 +1267,14 @@ static void rbd_osd_read_callback(struct
rbd_obj_request *obj_request,
                xferred = obj_request->length;
        }
        obj_request->xferred = xferred;
-       atomic_set(&obj_request->done, 1);
+       obj_request_done_set(obj_request);
 }

 static void rbd_osd_write_callback(struct rbd_obj_request *obj_request,
                                struct ceph_osd_op *op)
 {
        obj_request->xferred = le64_to_cpu(op->extent.length);
-       atomic_set(&obj_request->done, 1);
+       obj_request_done_set(obj_request);
 }

 static void rbd_osd_req_callback(struct ceph_osd_request *osd_req,
@@ -1300,7 +1318,7 @@ static void rbd_osd_req_callback(struct
ceph_osd_request *osd_req,
                break;
        }

-       if (atomic_read(&obj_request->done))
+       if (obj_request_done_test(obj_request))
                rbd_obj_request_complete(obj_request);
 }

@@ -1407,7 +1425,7 @@ static struct rbd_obj_request
*rbd_obj_request_create(const char *object_name,
        obj_request->which = BAD_WHICH;
        obj_request->type = type;
        INIT_LIST_HEAD(&obj_request->links);
-       atomic_set(&obj_request->done, 0);
+       obj_request_done_init(obj_request);
        init_completion(&obj_request->completion);
        kref_init(&obj_request->kref);

@@ -1611,7 +1629,7 @@ static void rbd_img_obj_callback(struct
rbd_obj_request *obj_request)
                rbd_assert(more);
                rbd_assert(which < img_request->obj_request_count);

-               if (!atomic_read(&obj_request->done))
+               if (!obj_request_done_test(obj_request))
                        break;

                rbd_assert(obj_request->xferred <= (u64) UINT_MAX);
-- 
1.7.9.5

--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html