Skip to content

Commit

Permalink
rbd: implement layered reads
Browse files Browse the repository at this point in the history
Implement layered read requests for format 2 rbd images.

If an rbd image is a clone of a snapshot, the snapshot will be the
clone's "parent" image.  When an object read request on a clone
comes back with ENOENT it indicates that the clone is not yet
populated with that portion of the image's data, and the parent
image should be consulted to satisfy the read.

When this occurs, a new image request is created, directed to the
parent image.  The offset and length of the image are the same as
the image-relative offset and length of the object request that
produced ENOENT.  Data from the parent image therefore satisfies the
object read request for the original image request.

While this code works, it will not be active until we enable the
layering feature (by adding RBD_FEATURE_LAYERING to the value of
RBD_FEATURES_SUPPORTED).

Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
  • Loading branch information
Alex Elder authored and Sage Weil committed May 2, 2013
1 parent 2f82ee5 commit 8b3e1a5
Showing 1 changed file with 85 additions and 12 deletions.
97 changes: 85 additions & 12 deletions drivers/block/rbd.c
Original file line number Diff line number Diff line change
Expand Up @@ -398,6 +398,8 @@ void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...)
# define rbd_assert(expr) ((void) 0)
#endif /* !RBD_DEBUG */

static void rbd_img_parent_read(struct rbd_obj_request *obj_request);

static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver);
static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver);

Expand Down Expand Up @@ -1336,9 +1338,15 @@ static void rbd_osd_trivial_callback(struct rbd_obj_request *obj_request)

static void rbd_osd_read_callback(struct rbd_obj_request *obj_request)
{
dout("%s: obj %p result %d %llu/%llu\n", __func__, obj_request,
obj_request->result, obj_request->xferred, obj_request->length);
if (obj_request->img_request)
struct rbd_img_request *img_request = obj_request->img_request;
bool layered = img_request && img_request_layered_test(img_request);

dout("%s: obj %p img %p result %d %llu/%llu\n", __func__,
obj_request, img_request, obj_request->result,
obj_request->xferred, obj_request->length);
if (layered && obj_request->result == -ENOENT)
rbd_img_parent_read(obj_request);
else if (img_request)
rbd_img_obj_request_read_callback(obj_request);
else
obj_request_done_set(obj_request);
Expand All @@ -1349,9 +1357,8 @@ static void rbd_osd_write_callback(struct rbd_obj_request *obj_request)
dout("%s: obj %p result %d %llu\n", __func__, obj_request,
obj_request->result, obj_request->length);
/*
* There is no such thing as a successful short write.
* Our xferred value is the number of bytes transferred
* back. Set it to our originally-requested length.
* There is no such thing as a successful short write. Set
* it to our originally-requested length.
*/
obj_request->xferred = obj_request->length;
obj_request_done_set(obj_request);
Expand Down Expand Up @@ -1391,7 +1398,7 @@ static void rbd_osd_req_callback(struct ceph_osd_request *osd_req,
* passed to blk_end_request(), which takes an unsigned int.
*/
obj_request->xferred = osd_req->r_reply_op_len[0];
rbd_assert(obj_request->xferred < (u64) UINT_MAX);
rbd_assert(obj_request->xferred < (u64)UINT_MAX);
opcode = osd_req->r_ops[0].op;
switch (opcode) {
case CEPH_OSD_OP_READ:
Expand Down Expand Up @@ -1607,7 +1614,6 @@ static struct rbd_img_request *rbd_img_request_create(
INIT_LIST_HEAD(&img_request->obj_requests);
kref_init(&img_request->kref);

(void) img_request_layered_test(img_request); /* Avoid a warning */
rbd_img_request_get(img_request); /* Avoid a warning */
rbd_img_request_put(img_request); /* TEMPORARY */

Expand Down Expand Up @@ -1635,6 +1641,9 @@ static void rbd_img_request_destroy(struct kref *kref)
if (img_request_write_test(img_request))
ceph_put_snap_context(img_request->snapc);

if (img_request_child_test(img_request))
rbd_obj_request_put(img_request->obj_request);

kfree(img_request);
}

Expand All @@ -1643,13 +1652,11 @@ static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request)
struct rbd_img_request *img_request;
unsigned int xferred;
int result;
bool more;

rbd_assert(obj_request_img_data_test(obj_request));
img_request = obj_request->img_request;

rbd_assert(!img_request_child_test(img_request));
rbd_assert(img_request->rq != NULL);

rbd_assert(obj_request->xferred <= (u64)UINT_MAX);
xferred = (unsigned int)obj_request->xferred;
result = obj_request->result;
Expand All @@ -1666,7 +1673,15 @@ static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request)
img_request->result = result;
}

return blk_end_request(img_request->rq, result, xferred);
if (img_request_child_test(img_request)) {
rbd_assert(img_request->obj_request != NULL);
more = obj_request->which < img_request->obj_request_count - 1;
} else {
rbd_assert(img_request->rq != NULL);
more = blk_end_request(img_request->rq, result, xferred);
}

return more;
}

static void rbd_img_obj_callback(struct rbd_obj_request *obj_request)
Expand Down Expand Up @@ -1811,6 +1826,64 @@ static int rbd_img_request_submit(struct rbd_img_request *img_request)
return 0;
}

static void rbd_img_parent_read_callback(struct rbd_img_request *img_request)
{
struct rbd_obj_request *obj_request;

rbd_assert(img_request_child_test(img_request));

obj_request = img_request->obj_request;
rbd_assert(obj_request != NULL);
obj_request->result = img_request->result;
obj_request->xferred = img_request->xferred;

rbd_img_obj_request_read_callback(obj_request);
rbd_obj_request_complete(obj_request);
}

static void rbd_img_parent_read(struct rbd_obj_request *obj_request)
{
struct rbd_device *rbd_dev;
struct rbd_img_request *img_request;
int result;

rbd_assert(obj_request_img_data_test(obj_request));
rbd_assert(obj_request->img_request != NULL);
rbd_assert(obj_request->result == (s32) -ENOENT);
rbd_assert(obj_request->type == OBJ_REQUEST_BIO);

rbd_dev = obj_request->img_request->rbd_dev;
rbd_assert(rbd_dev->parent != NULL);
/* rbd_read_finish(obj_request, obj_request->length); */
img_request = rbd_img_request_create(rbd_dev->parent,
obj_request->img_offset,
obj_request->length,
false, true);
result = -ENOMEM;
if (!img_request)
goto out_err;

rbd_obj_request_get(obj_request);
img_request->obj_request = obj_request;

result = rbd_img_request_fill_bio(img_request, obj_request->bio_list);
if (result)
goto out_err;

img_request->callback = rbd_img_parent_read_callback;
result = rbd_img_request_submit(img_request);
if (result)
goto out_err;

return;
out_err:
if (img_request)
rbd_img_request_put(img_request);
obj_request->result = result;
obj_request->xferred = 0;
obj_request_done_set(obj_request);
}

static int rbd_obj_notify_ack(struct rbd_device *rbd_dev,
u64 ver, u64 notify_id)
{
Expand Down

0 comments on commit 8b3e1a5

Please sign in to comment.