+
/*
rbd.c -- Export ceph rados objects as a Linux block device
u64 stripe_unit;
u64 stripe_count;
-
- u64 obj_version;
};
/*
struct ceph_osd_request *osd_req;
u64 xferred; /* bytes transferred */
- u64 version;
int result;
rbd_obj_callback_t callback;
static int rbd_dev_snaps_update(struct rbd_device *rbd_dev);
-static void rbd_dev_release(struct device *dev);
+static void rbd_dev_device_release(struct device *dev);
static void rbd_snap_destroy(struct rbd_snap *snap);
static ssize_t rbd_add(struct bus_type *bus, const char *buf,
static void rbd_img_parent_read(struct rbd_obj_request *obj_request);
static void rbd_dev_remove_parent(struct rbd_device *rbd_dev);
-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);
+static int rbd_dev_refresh(struct rbd_device *rbd_dev);
+static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev);
static int rbd_open(struct block_device *bdev, fmode_t mode)
{
kfree(rbdc);
}
-/* Caller has to fill in snapc->seq and snapc->snaps[0..snap_count-1] */
-
-static struct ceph_snap_context *rbd_snap_context_create(u32 snap_count)
-{
- struct ceph_snap_context *snapc;
- size_t size;
-
- size = sizeof (struct ceph_snap_context);
- size += snap_count * sizeof (snapc->snaps[0]);
- snapc = kzalloc(size, GFP_KERNEL);
- if (!snapc)
- return NULL;
-
- atomic_set(&snapc->nref, 1);
- snapc->num_snaps = snap_count;
-
- return snapc;
-}
-
-static inline void rbd_snap_context_get(struct ceph_snap_context *snapc)
-{
- (void)ceph_get_snap_context(snapc);
-}
-
-static inline void rbd_snap_context_put(struct ceph_snap_context *snapc)
-{
- ceph_put_snap_context(snapc);
-}
-
/*
* Drop reference to ceph client node. If it's not referenced anymore, release
* it.
header->image_size = le64_to_cpu(ondisk->image_size);
- header->snapc = rbd_snap_context_create(snap_count);
+ header->snapc = ceph_create_snap_context(snap_count, GFP_KERNEL);
if (!header->snapc)
goto out_err;
header->snapc->seq = le64_to_cpu(ondisk->snap_seq);
rbd_dev->mapping.read_only = true;
}
-static void rbd_header_free(struct rbd_image_header *header)
+static void rbd_dev_clear_mapping(struct rbd_device *rbd_dev)
{
- kfree(header->object_prefix);
- header->object_prefix = NULL;
- kfree(header->snap_sizes);
- header->snap_sizes = NULL;
- kfree(header->snap_names);
- header->snap_names = NULL;
- rbd_snap_context_put(header->snapc);
- header->snapc = NULL;
+ rbd_dev->mapping.size = 0;
+ rbd_dev->mapping.features = 0;
+ rbd_dev->mapping.read_only = true;
}
static const char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset)
if (osd_req->r_result < 0)
obj_request->result = osd_req->r_result;
- obj_request->version = le64_to_cpu(osd_req->r_reassert_version.version);
BUG_ON(osd_req->r_num_ops > 2);
if (write_request) {
down_read(&rbd_dev->header_rwsem);
- rbd_snap_context_get(rbd_dev->header.snapc);
+ ceph_get_snap_context(rbd_dev->header.snapc);
up_read(&rbd_dev->header_rwsem);
}
rbd_assert(img_request->obj_request_count == 0);
if (img_request_write_test(img_request))
- rbd_snap_context_put(img_request->snapc);
+ ceph_put_snap_context(img_request->snapc);
if (img_request_child_test(img_request))
rbd_obj_request_put(img_request->obj_request);
obj_request_done_set(obj_request);
}
-static int rbd_obj_notify_ack(struct rbd_device *rbd_dev,
- u64 ver, u64 notify_id)
+static int rbd_obj_notify_ack(struct rbd_device *rbd_dev, u64 notify_id)
{
struct rbd_obj_request *obj_request;
struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
obj_request->callback = rbd_obj_request_put;
osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_NOTIFY_ACK,
- notify_id, ver, 0);
+ notify_id, 0, 0);
rbd_osd_req_format_read(obj_request);
ret = rbd_obj_request_submit(osdc, obj_request);
static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data)
{
struct rbd_device *rbd_dev = (struct rbd_device *)data;
- u64 hver;
if (!rbd_dev)
return;
dout("%s: \"%s\" notify_id %llu opcode %u\n", __func__,
- rbd_dev->header_name, (unsigned long long) notify_id,
- (unsigned int) opcode);
- (void)rbd_dev_refresh(rbd_dev, &hver);
+ rbd_dev->header_name, (unsigned long long)notify_id,
+ (unsigned int)opcode);
+ (void)rbd_dev_refresh(rbd_dev);
- rbd_obj_notify_ack(rbd_dev, hver, notify_id);
+ rbd_obj_notify_ack(rbd_dev, notify_id);
}
/*
rbd_dev->watch_request->osd_req);
osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_WATCH,
- rbd_dev->watch_event->cookie,
- rbd_dev->header.obj_version, start);
+ rbd_dev->watch_event->cookie, 0, start);
rbd_osd_req_format_write(obj_request);
ret = rbd_obj_request_submit(osdc, obj_request);
const void *outbound,
size_t outbound_size,
void *inbound,
- size_t inbound_size,
- u64 *version)
+ size_t inbound_size)
{
struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
struct rbd_obj_request *obj_request;
rbd_assert(obj_request->xferred < (u64)INT_MAX);
ret = (int)obj_request->xferred;
ceph_copy_from_page_vector(pages, inbound, 0, obj_request->xferred);
- if (version)
- *version = obj_request->version;
out:
if (obj_request)
rbd_obj_request_put(obj_request);
static int rbd_obj_read_sync(struct rbd_device *rbd_dev,
const char *object_name,
- u64 offset, u64 length,
- void *buf, u64 *version)
+ u64 offset, u64 length, void *buf)
{
struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
rbd_assert(obj_request->xferred <= (u64) SIZE_MAX);
size = (size_t) obj_request->xferred;
ceph_copy_from_page_vector(pages, buf, 0, size);
- rbd_assert(size <= (size_t) INT_MAX);
- ret = (int) size;
- if (version)
- *version = obj_request->version;
+ rbd_assert(size <= (size_t)INT_MAX);
+ ret = (int)size;
out:
if (obj_request)
rbd_obj_request_put(obj_request);
* Returns a pointer-coded errno if a failure occurs.
*/
static struct rbd_image_header_ondisk *
-rbd_dev_v1_header_read(struct rbd_device *rbd_dev, u64 *version)
+rbd_dev_v1_header_read(struct rbd_device *rbd_dev)
{
struct rbd_image_header_ondisk *ondisk = NULL;
u32 snap_count = 0;
return ERR_PTR(-ENOMEM);
ret = rbd_obj_read_sync(rbd_dev, rbd_dev->header_name,
- 0, size, ondisk, version);
+ 0, size, ondisk);
if (ret < 0)
goto out_err;
if ((size_t)ret < size) {
struct rbd_image_header *header)
{
struct rbd_image_header_ondisk *ondisk;
- u64 ver = 0;
int ret;
- ondisk = rbd_dev_v1_header_read(rbd_dev, &ver);
+ ondisk = rbd_dev_v1_header_read(rbd_dev);
if (IS_ERR(ondisk))
return PTR_ERR(ondisk);
ret = rbd_header_from_disk(header, ondisk);
- if (ret >= 0)
- header->obj_version = ver;
kfree(ondisk);
return ret;
/*
* only read the first part of the ondisk header, without the snaps info
*/
-static int rbd_dev_v1_refresh(struct rbd_device *rbd_dev, u64 *hver)
+static int rbd_dev_v1_refresh(struct rbd_device *rbd_dev)
{
int ret;
struct rbd_image_header h;
kfree(rbd_dev->header.snap_sizes);
kfree(rbd_dev->header.snap_names);
/* osd requests may still refer to snapc */
- rbd_snap_context_put(rbd_dev->header.snapc);
+ ceph_put_snap_context(rbd_dev->header.snapc);
- if (hver)
- *hver = h.obj_version;
- rbd_dev->header.obj_version = h.obj_version;
rbd_dev->header.image_size = h.image_size;
rbd_dev->header.snapc = h.snapc;
rbd_dev->header.snap_names = h.snap_names;
return ret;
}
-static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver)
+static int rbd_dev_refresh(struct rbd_device *rbd_dev)
{
+ u64 image_size;
int ret;
rbd_assert(rbd_image_format_valid(rbd_dev->image_format));
+ image_size = rbd_dev->header.image_size;
mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING);
if (rbd_dev->image_format == 1)
- ret = rbd_dev_v1_refresh(rbd_dev, hver);
+ ret = rbd_dev_v1_refresh(rbd_dev);
else
- ret = rbd_dev_v2_refresh(rbd_dev, hver);
+ ret = rbd_dev_v2_refresh(rbd_dev);
mutex_unlock(&ctl_mutex);
- revalidate_disk(rbd_dev->disk);
if (ret)
rbd_warn(rbd_dev, "got notification but failed to "
" update snaps: %d\n", ret);
+ if (image_size != rbd_dev->header.image_size)
+ revalidate_disk(rbd_dev->disk);
return ret;
}
struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
int ret;
- ret = rbd_dev_refresh(rbd_dev, NULL);
+ ret = rbd_dev_refresh(rbd_dev);
return ret < 0 ? ret : size;
}
* Returns a dynamically-allocated snapshot name if successful, or a
* pointer-coded error otherwise.
*/
-static char *rbd_dev_v1_snap_info(struct rbd_device *rbd_dev, u32 which,
+static const char *rbd_dev_v1_snap_info(struct rbd_device *rbd_dev, u32 which,
u64 *snap_size, u64 *snap_features)
{
- char *snap_name;
+ const char *snap_name;
int i;
rbd_assert(which < rbd_dev->header.snapc->num_snaps);
ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name,
"rbd", "get_size",
&snapid, sizeof (snapid),
- &size_buf, sizeof (size_buf), NULL);
+ &size_buf, sizeof (size_buf));
dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
if (ret < 0)
return ret;
ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name,
"rbd", "get_object_prefix", NULL, 0,
- reply_buf, RBD_OBJ_PREFIX_LEN_MAX, NULL);
+ reply_buf, RBD_OBJ_PREFIX_LEN_MAX);
dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
if (ret < 0)
goto out;
ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name,
"rbd", "get_features",
&snapid, sizeof (snapid),
- &features_buf, sizeof (features_buf), NULL);
+ &features_buf, sizeof (features_buf));
dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
if (ret < 0)
return ret;
ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name,
"rbd", "get_parent",
&snapid, sizeof (snapid),
- reply_buf, size, NULL);
+ reply_buf, size);
dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
if (ret < 0)
goto out_err;
ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name,
"rbd", "get_stripe_unit_count", NULL, 0,
- (char *)&striping_info_buf, size, NULL);
+ (char *)&striping_info_buf, size);
dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
if (ret < 0)
return ret;
ret = rbd_obj_method_sync(rbd_dev, RBD_DIRECTORY,
"rbd", "dir_get_name",
image_id, image_id_size,
- reply_buf, size, NULL);
+ reply_buf, size);
if (ret < 0)
goto out;
p = reply_buf;
return ret;
}
-static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev, u64 *ver)
+static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev)
{
size_t size;
int ret;
ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name,
"rbd", "get_snapcontext", NULL, 0,
- reply_buf, size, ver);
+ reply_buf, size);
dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
if (ret < 0)
goto out;
goto out;
ret = 0;
- snapc = rbd_snap_context_create(snap_count);
+ snapc = ceph_create_snap_context(snap_count, GFP_KERNEL);
if (!snapc) {
ret = -ENOMEM;
goto out;
return ret;
}
-static char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, u32 which)
+static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, u32 which)
{
size_t size;
void *reply_buf;
ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name,
"rbd", "get_snapshot_name",
&snap_id, sizeof (snap_id),
- reply_buf, size, NULL);
+ reply_buf, size);
dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
if (ret < 0) {
snap_name = ERR_PTR(ret);
return snap_name;
}
-static char *rbd_dev_v2_snap_info(struct rbd_device *rbd_dev, u32 which,
+static const char *rbd_dev_v2_snap_info(struct rbd_device *rbd_dev, u32 which,
u64 *snap_size, u64 *snap_features)
{
u64 snap_id;
u64 size;
u64 features;
- char *snap_name;
+ const char *snap_name;
int ret;
rbd_assert(which < rbd_dev->header.snapc->num_snaps);
return ERR_PTR(ret);
}
-static char *rbd_dev_snap_info(struct rbd_device *rbd_dev, u32 which,
+static const char *rbd_dev_snap_info(struct rbd_device *rbd_dev, u32 which,
u64 *snap_size, u64 *snap_features)
{
if (rbd_dev->image_format == 1)
return ERR_PTR(-EINVAL);
}
-static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver)
+static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev)
{
int ret;
- __u8 obj_order;
down_write(&rbd_dev->header_rwsem);
- /* Grab old order first, to see if it changes */
-
- obj_order = rbd_dev->header.obj_order,
ret = rbd_dev_v2_image_size(rbd_dev);
if (ret)
goto out;
- if (rbd_dev->header.obj_order != obj_order) {
- ret = -EIO;
- goto out;
- }
rbd_update_mapping_size(rbd_dev);
- ret = rbd_dev_v2_snap_context(rbd_dev, hver);
+ ret = rbd_dev_v2_snap_context(rbd_dev);
dout("rbd_dev_v2_snap_context returned %d\n", ret);
if (ret)
goto out;
while (index < snap_count || links != head) {
u64 snap_id;
struct rbd_snap *snap;
- char *snap_name;
+ const char *snap_name;
u64 snap_size = 0;
u64 snap_features = 0;
dev->bus = &rbd_bus_type;
dev->type = &rbd_device_type;
dev->parent = &rbd_root_dev;
- dev->release = rbd_dev_release;
+ dev->release = rbd_dev_device_release;
dev_set_name(dev, "%d", rbd_dev->dev_id);
ret = device_register(dev);
ret = rbd_obj_method_sync(rbd_dev, object_name,
"rbd", "get_id", NULL, 0,
- response, RBD_IMAGE_ID_LEN_MAX, NULL);
+ response, RBD_IMAGE_ID_LEN_MAX);
dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
if (ret == -ENOENT) {
image_id = kstrdup("", GFP_KERNEL);
return ret;
}
+/* Undo whatever state changes are made by v1 or v2 image probe */
+
+static void rbd_dev_unprobe(struct rbd_device *rbd_dev)
+{
+ struct rbd_image_header *header;
+
+ rbd_dev_remove_parent(rbd_dev);
+ rbd_spec_put(rbd_dev->parent_spec);
+ rbd_dev->parent_spec = NULL;
+ rbd_dev->parent_overlap = 0;
+
+ /* Free dynamic fields from the header, then zero it out */
+
+ header = &rbd_dev->header;
+ ceph_put_snap_context(header->snapc);
+ kfree(header->snap_sizes);
+ kfree(header->snap_names);
+ kfree(header->object_prefix);
+ memset(header, 0, sizeof (*header));
+}
+
static int rbd_dev_v1_probe(struct rbd_device *rbd_dev)
{
int ret;
static int rbd_dev_v2_probe(struct rbd_device *rbd_dev)
{
int ret;
- u64 ver = 0;
ret = rbd_dev_v2_image_size(rbd_dev);
if (ret)
ret = rbd_dev_v2_parent_info(rbd_dev);
if (ret)
goto out_err;
- rbd_warn(rbd_dev, "WARNING: kernel support for "
- "layered rbd images is EXPERIMENTAL!");
+
+ /*
+ * Don't print a warning for parent images. We can
+ * tell this point because we won't know its pool
+ * name yet (just its pool id).
+ */
+ if (rbd_dev->spec->pool_name)
+ rbd_warn(rbd_dev, "WARNING: kernel layering "
+ "is EXPERIMENTAL!");
}
/* If the image supports fancy striping, get its parameters */
/* Get the snapshot context, plus the header version */
- ret = rbd_dev_v2_snap_context(rbd_dev, &ver);
+ ret = rbd_dev_v2_snap_context(rbd_dev);
if (ret)
goto out_err;
- rbd_dev->header.obj_version = ver;
dout("discovered version 2 image, header name is %s\n",
rbd_dev->header_name);
return ret;
}
-static int rbd_dev_probe_finish(struct rbd_device *rbd_dev)
+static int rbd_dev_device_setup(struct rbd_device *rbd_dev)
{
int ret;
- int tmp;
-
- ret = rbd_dev_header_watch_sync(rbd_dev, 1);
- if (ret)
- return ret;
ret = rbd_dev_mapping_set(rbd_dev);
if (ret)
unregister_blkdev(rbd_dev->major, rbd_dev->name);
err_out_id:
rbd_dev_id_put(rbd_dev);
- tmp = rbd_dev_header_watch_sync(rbd_dev, 0);
- if (tmp)
- rbd_warn(rbd_dev, "failed to cancel watch event (%d)\n", ret);
rbd_dev_mapping_clear(rbd_dev);
return ret;
return 0;
}
+static void rbd_dev_image_release(struct rbd_device *rbd_dev)
+{
+ int ret;
+
+ rbd_remove_all_snaps(rbd_dev);
+ rbd_dev_unprobe(rbd_dev);
+ ret = rbd_dev_header_watch_sync(rbd_dev, 0);
+ if (ret)
+ rbd_warn(rbd_dev, "failed to cancel watch event (%d)\n", ret);
+ kfree(rbd_dev->header_name);
+ rbd_dev->header_name = NULL;
+ rbd_dev->image_format = 0;
+ kfree(rbd_dev->spec->image_id);
+ rbd_dev->spec->image_id = NULL;
+
+ rbd_dev_destroy(rbd_dev);
+}
+
/*
* Probe for the existence of the header object for the given rbd
* device. For format 2 images this includes determining the image
static int rbd_dev_image_probe(struct rbd_device *rbd_dev)
{
int ret;
+ int tmp;
/*
* Get the id from the image id object. If it's not a
if (ret)
goto err_out_format;
+ ret = rbd_dev_header_watch_sync(rbd_dev, 1);
+ if (ret)
+ goto out_header_name;
+
if (rbd_dev->image_format == 1)
ret = rbd_dev_v1_probe(rbd_dev);
else
ret = rbd_dev_v2_probe(rbd_dev);
if (ret)
- goto out_header_name;
+ goto err_out_watch;
ret = rbd_dev_snaps_update(rbd_dev);
if (ret)
- goto out_header_name;
+ goto err_out_probe;
ret = rbd_dev_spec_update(rbd_dev);
if (ret)
goto err_out_snaps;
ret = rbd_dev_probe_parent(rbd_dev);
- if (ret)
- goto err_out_snaps;
-
- ret = rbd_dev_probe_finish(rbd_dev);
- if (ret)
- goto err_out_parent;
+ if (!ret)
+ return 0;
- return ret;
-err_out_parent:
- rbd_dev_remove_parent(rbd_dev);
- rbd_header_free(&rbd_dev->header);
err_out_snaps:
rbd_remove_all_snaps(rbd_dev);
+err_out_probe:
+ rbd_dev_unprobe(rbd_dev);
+err_out_watch:
+ tmp = rbd_dev_header_watch_sync(rbd_dev, 0);
+ if (tmp)
+ rbd_warn(rbd_dev, "unable to tear down watch request\n");
out_header_name:
kfree(rbd_dev->header_name);
rbd_dev->header_name = NULL;
if (rc < 0)
goto err_out_rbd_dev;
- return count;
+ rc = rbd_dev_device_setup(rbd_dev);
+ if (!rc)
+ return count;
+
+ rbd_dev_image_release(rbd_dev);
err_out_rbd_dev:
- kfree(rbd_dev->header_name);
rbd_dev_destroy(rbd_dev);
err_out_client:
rbd_put_client(rbdc);
return NULL;
}
-static void rbd_dev_release(struct device *dev)
+static void rbd_dev_device_release(struct device *dev)
{
struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
- /* clean up and free blkdev */
rbd_free_disk(rbd_dev);
+ clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
+ rbd_dev_clear_mapping(rbd_dev);
unregister_blkdev(rbd_dev->major, rbd_dev->name);
-
- /* release allocated disk header fields */
- rbd_header_free(&rbd_dev->header);
-
- /* done with the id, and with the rbd_dev */
+ rbd_dev->major = 0;
rbd_dev_id_put(rbd_dev);
rbd_dev_mapping_clear(rbd_dev);
- rbd_assert(rbd_dev->rbd_client != NULL);
- rbd_spec_put(rbd_dev->parent_spec);
- kfree(rbd_dev->header_name);
- rbd_dev_destroy(rbd_dev);
-
- /* release module ref */
- module_put(THIS_MODULE);
}
static void rbd_dev_remove_parent(struct rbd_device *rbd_dev)
struct rbd_device *first = rbd_dev;
struct rbd_device *second = first->parent;
struct rbd_device *third;
- int ret;
/*
* Follow to the parent with no grandparent and
second = third;
}
rbd_assert(second);
- ret = rbd_dev_header_watch_sync(rbd_dev, 0);
- if (ret)
- rbd_warn(rbd_dev,
- "failed to cancel watch event (%d)\n", ret);
- rbd_remove_all_snaps(second);
- rbd_bus_del_dev(second);
+ rbd_dev_image_release(second);
first->parent = NULL;
first->parent_overlap = 0;
spin_unlock_irq(&rbd_dev->lock);
if (ret < 0)
goto done;
-
- ret = rbd_dev_header_watch_sync(rbd_dev, 0);
- if (ret) {
- rbd_warn(rbd_dev, "failed to cancel watch event (%d)\n", ret);
- clear_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags);
- smp_mb();
- return ret;
- }
ret = count;
-
- rbd_dev_remove_parent(rbd_dev);
-
- rbd_remove_all_snaps(rbd_dev);
rbd_bus_del_dev(rbd_dev);
+ rbd_dev_image_release(rbd_dev);
+ module_put(THIS_MODULE);
done:
mutex_unlock(&ctl_mutex);
projects / ~andy / linux / blobdiff