char *snap_names;
u64 *snap_sizes;
+ u64 stripe_unit;
+ u64 stripe_count;
+
u64 obj_version;
};
*/
struct rbd_spec {
u64 pool_id;
- char *pool_name;
+ const char *pool_name;
- char *image_id;
- char *image_name;
+ const char *image_id;
+ const char *image_name;
u64 snap_id;
- char *snap_name;
+ const char *snap_name;
struct kref kref;
};
u64 parent_overlap;
struct rbd_device *parent;
- u64 stripe_unit;
- u64 stripe_count;
-
/* protects updating the header */
struct rw_semaphore header_rwsem;
size_t count);
static ssize_t rbd_remove(struct bus_type *bus, const char *buf,
size_t count);
-static int rbd_dev_probe(struct rbd_device *rbd_dev);
+static int rbd_dev_image_probe(struct rbd_device *rbd_dev);
static struct bus_attribute rbd_bus_attrs[] = {
__ATTR(add, S_IWUSR, NULL, rbd_add),
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->snap_sizes[i] =
le64_to_cpu(ondisk->snaps[i].image_size);
} else {
- WARN_ON(ondisk->snap_names_len);
header->snap_names = NULL;
header->snap_sizes = NULL;
}
/* Allocate and fill in the snapshot context */
header->image_size = le64_to_cpu(ondisk->image_size);
- size = sizeof (struct ceph_snap_context);
- size += snap_count * sizeof (header->snapc->snaps[0]);
- header->snapc = kzalloc(size, GFP_KERNEL);
+
+ header->snapc = rbd_snap_context_create(snap_count);
if (!header->snapc)
goto out_err;
-
- atomic_set(&header->snapc->nref, 1);
header->snapc->seq = le64_to_cpu(ondisk->snap_seq);
- header->snapc->num_snaps = snap_count;
for (i = 0; i < snap_count; i++)
- header->snapc->snaps[i] =
- le64_to_cpu(ondisk->snaps[i].id);
+ header->snapc->snaps[i] = le64_to_cpu(ondisk->snaps[i].id);
return 0;
return NULL;
}
-static int snap_by_name(struct rbd_device *rbd_dev, const char *snap_name)
+static struct rbd_snap *snap_by_name(struct rbd_device *rbd_dev,
+ const char *snap_name)
{
-
struct rbd_snap *snap;
- list_for_each_entry(snap, &rbd_dev->snaps, node) {
- if (!strcmp(snap_name, snap->name)) {
- rbd_dev->spec->snap_id = snap->id;
- rbd_dev->mapping.size = snap->size;
- rbd_dev->mapping.features = snap->features;
-
- return 0;
- }
- }
+ list_for_each_entry(snap, &rbd_dev->snaps, node)
+ if (!strcmp(snap_name, snap->name))
+ return snap;
- return -ENOENT;
+ return NULL;
}
static int rbd_dev_set_mapping(struct rbd_device *rbd_dev)
{
- int ret;
-
if (!memcmp(rbd_dev->spec->snap_name, RBD_SNAP_HEAD_NAME,
sizeof (RBD_SNAP_HEAD_NAME))) {
- rbd_dev->spec->snap_id = CEPH_NOSNAP;
rbd_dev->mapping.size = rbd_dev->header.image_size;
rbd_dev->mapping.features = rbd_dev->header.features;
- ret = 0;
} else {
- ret = snap_by_name(rbd_dev, rbd_dev->spec->snap_name);
- if (ret < 0)
- goto done;
+ struct rbd_snap *snap;
+
+ snap = snap_by_name(rbd_dev, rbd_dev->spec->snap_name);
+ if (!snap)
+ return -ENOENT;
+ rbd_dev->mapping.size = snap->size;
+ rbd_dev->mapping.features = snap->features;
rbd_dev->mapping.read_only = true;
}
- set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
-done:
- return ret;
+ return 0;
}
static void rbd_header_free(struct rbd_image_header *header)
header->snap_sizes = NULL;
kfree(header->snap_names);
header->snap_names = NULL;
- ceph_put_snap_context(header->snapc);
+ rbd_snap_context_put(header->snapc);
header->snapc = NULL;
}
bool child_request)
{
struct rbd_img_request *img_request;
- struct ceph_snap_context *snapc = NULL;
img_request = kmalloc(sizeof (*img_request), GFP_ATOMIC);
if (!img_request)
if (write_request) {
down_read(&rbd_dev->header_rwsem);
- snapc = ceph_get_snap_context(rbd_dev->header.snapc);
+ rbd_snap_context_get(rbd_dev->header.snapc);
up_read(&rbd_dev->header_rwsem);
- if (WARN_ON(!snapc)) {
- kfree(img_request);
- return NULL; /* Shouldn't happen */
- }
-
}
img_request->rq = NULL;
img_request->flags = 0;
if (write_request) {
img_request_write_set(img_request);
- img_request->snapc = snapc;
+ img_request->snapc = rbd_dev->header.snapc;
} else {
img_request->snap_id = rbd_dev->spec->snap_id;
}
rbd_assert(img_request->obj_request_count == 0);
if (img_request_write_test(img_request))
- ceph_put_snap_context(img_request->snapc);
+ rbd_snap_context_put(img_request->snapc);
if (img_request_child_test(img_request))
rbd_obj_request_put(img_request->obj_request);
}
result = -EINVAL;
- if (WARN_ON(offset && length > U64_MAX - offset + 1))
+ if (offset && length > U64_MAX - offset + 1) {
+ rbd_warn(rbd_dev, "bad request range (%llu~%llu)\n",
+ offset, length);
goto end_request; /* Shouldn't happen */
+ }
result = -ENOMEM;
img_request = rbd_img_request_create(rbd_dev, offset, length,
0, size, ondisk, version);
if (ret < 0)
goto out_err;
- if (WARN_ON((size_t) ret < size)) {
+ if ((size_t)ret < size) {
ret = -ENXIO;
rbd_warn(rbd_dev, "short header read (want %zd got %d)",
size, ret);
static void rbd_update_mapping_size(struct rbd_device *rbd_dev)
{
- sector_t size;
-
if (rbd_dev->spec->snap_id != CEPH_NOSNAP)
return;
- size = (sector_t) rbd_dev->header.image_size / SECTOR_SIZE;
- dout("setting size to %llu sectors", (unsigned long long) size);
- rbd_dev->mapping.size = (u64) size;
- set_capacity(rbd_dev->disk, size);
+ if (rbd_dev->mapping.size != rbd_dev->header.image_size) {
+ sector_t size;
+
+ rbd_dev->mapping.size = rbd_dev->header.image_size;
+ size = (sector_t)rbd_dev->mapping.size / SECTOR_SIZE;
+ dout("setting size to %llu sectors", (unsigned long long)size);
+ set_capacity(rbd_dev->disk, size);
+ }
}
/*
kfree(rbd_dev->header.snap_sizes);
kfree(rbd_dev->header.snap_names);
/* osd requests may still refer to snapc */
- ceph_put_snap_context(rbd_dev->header.snapc);
+ rbd_snap_context_put(rbd_dev->header.snapc);
if (hver)
*hver = h.obj_version;
rbd_dev->header.snap_names = h.snap_names;
rbd_dev->header.snap_sizes = h.snap_sizes;
/* Free the extra copy of the object prefix */
- WARN_ON(strcmp(rbd_dev->header.object_prefix, h.object_prefix));
+ if (strcmp(rbd_dev->header.object_prefix, h.object_prefix))
+ rbd_warn(rbd_dev, "object prefix changed (ignoring)");
kfree(h.object_prefix);
ret = rbd_dev_snaps_update(rbd_dev);
rbd_dev->disk = disk;
- set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE);
-
return 0;
out_disk:
put_disk(disk);
struct device_attribute *attr, char *buf)
{
struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
- sector_t size;
-
- down_read(&rbd_dev->header_rwsem);
- size = get_capacity(rbd_dev->disk);
- up_read(&rbd_dev->header_rwsem);
- return sprintf(buf, "%llu\n", (unsigned long long) size * SECTOR_SIZE);
+ return sprintf(buf, "%llu\n",
+ (unsigned long long)rbd_dev->mapping.size);
}
/*
struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
return sprintf(buf, "0x%016llx\n",
- (unsigned long long) rbd_dev->mapping.features);
+ (unsigned long long)rbd_dev->mapping.features);
}
static ssize_t rbd_major_show(struct device *dev,
{
struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
- return sprintf(buf, "%d\n", rbd_dev->major);
+ if (rbd_dev->major)
+ return sprintf(buf, "%d\n", rbd_dev->major);
+
+ return sprintf(buf, "(none)\n");
+
}
static ssize_t rbd_client_id_show(struct device *dev,
struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
return sprintf(buf, "%llu\n",
- (unsigned long long) rbd_dev->spec->pool_id);
+ (unsigned long long) rbd_dev->spec->pool_id);
}
static ssize_t rbd_name_show(struct device *dev,
static void rbd_dev_destroy(struct rbd_device *rbd_dev)
{
- rbd_spec_put(rbd_dev->parent_spec);
- kfree(rbd_dev->header_name);
rbd_put_client(rbd_dev->rbd_client);
rbd_spec_put(rbd_dev->spec);
kfree(rbd_dev);
/* The ceph file layout needs to fit pool id in 32 bits */
ret = -EIO;
- if (WARN_ON(parent_spec->pool_id > (u64)U32_MAX))
+ if (parent_spec->pool_id > (u64)U32_MAX) {
+ rbd_warn(NULL, "parent pool id too large (%llu > %u)\n",
+ (unsigned long long)parent_spec->pool_id, U32_MAX);
goto out_err;
+ }
image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL);
if (IS_ERR(image_id)) {
"(got %llu want 1)", stripe_count);
return -EINVAL;
}
- rbd_dev->stripe_unit = stripe_unit;
- rbd_dev->stripe_count = stripe_count;
+ rbd_dev->header.stripe_unit = stripe_unit;
+ rbd_dev->header.stripe_count = stripe_count;
return 0;
}
}
/*
- * When a parent image gets probed, we only have the pool, image,
- * and snapshot ids but not the names of any of them. This call
- * is made later to fill in those names. It has to be done after
- * rbd_dev_snaps_update() has completed because some of the
- * information (in particular, snapshot name) is not available
- * until then.
+ * When an rbd image has a parent image, it is identified by the
+ * pool, image, and snapshot ids (not names). This function fills
+ * in the names for those ids. (It's OK if we can't figure out the
+ * name for an image id, but the pool and snapshot ids should always
+ * exist and have names.) All names in an rbd spec are dynamically
+ * allocated.
+ *
+ * When an image being mapped (not a parent) is probed, we have the
+ * pool name and pool id, image name and image id, and the snapshot
+ * name. The only thing we're missing is the snapshot id.
+ *
+ * The set of snapshots for an image is not known until they have
+ * been read by rbd_dev_snaps_update(), so we can't completely fill
+ * in this information until after that has been called.
*/
-static int rbd_dev_probe_update_spec(struct rbd_device *rbd_dev)
+static int rbd_dev_spec_update(struct rbd_device *rbd_dev)
{
- struct ceph_osd_client *osdc;
- const char *name;
- void *reply_buf = NULL;
+ struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
+ struct rbd_spec *spec = rbd_dev->spec;
+ const char *pool_name;
+ const char *image_name;
+ const char *snap_name;
int ret;
- if (rbd_dev->spec->pool_name)
- return 0; /* Already have the names */
+ /*
+ * An image being mapped will have the pool name (etc.), but
+ * we need to look up the snapshot id.
+ */
+ if (spec->pool_name) {
+ if (strcmp(spec->snap_name, RBD_SNAP_HEAD_NAME)) {
+ struct rbd_snap *snap;
+
+ snap = snap_by_name(rbd_dev, spec->snap_name);
+ if (!snap)
+ return -ENOENT;
+ spec->snap_id = snap->id;
+ } else {
+ spec->snap_id = CEPH_NOSNAP;
+ }
- /* Look up the pool name */
+ return 0;
+ }
- osdc = &rbd_dev->rbd_client->client->osdc;
- name = ceph_pg_pool_name_by_id(osdc->osdmap, rbd_dev->spec->pool_id);
- if (!name) {
- rbd_warn(rbd_dev, "there is no pool with id %llu",
- rbd_dev->spec->pool_id); /* Really a BUG() */
+ /* Get the pool name; we have to make our own copy of this */
+
+ pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, spec->pool_id);
+ if (!pool_name) {
+ rbd_warn(rbd_dev, "no pool with id %llu", spec->pool_id);
return -EIO;
}
-
- rbd_dev->spec->pool_name = kstrdup(name, GFP_KERNEL);
- if (!rbd_dev->spec->pool_name)
+ pool_name = kstrdup(pool_name, GFP_KERNEL);
+ if (!pool_name)
return -ENOMEM;
/* Fetch the image name; tolerate failure here */
- name = rbd_dev_image_name(rbd_dev);
- if (name)
- rbd_dev->spec->image_name = (char *)name;
- else
+ image_name = rbd_dev_image_name(rbd_dev);
+ if (!image_name)
rbd_warn(rbd_dev, "unable to get image name");
- /* Look up the snapshot name. */
+ /* Look up the snapshot name, and make a copy */
- name = rbd_snap_name(rbd_dev, rbd_dev->spec->snap_id);
- if (!name) {
- rbd_warn(rbd_dev, "no snapshot with id %llu",
- rbd_dev->spec->snap_id); /* Really a BUG() */
+ snap_name = rbd_snap_name(rbd_dev, spec->snap_id);
+ if (!snap_name) {
+ rbd_warn(rbd_dev, "no snapshot with id %llu", spec->snap_id);
ret = -EIO;
goto out_err;
}
- rbd_dev->spec->snap_name = kstrdup(name, GFP_KERNEL);
- if(!rbd_dev->spec->snap_name)
+ snap_name = kstrdup(snap_name, GFP_KERNEL);
+ if (!snap_name) {
+ ret = -ENOMEM;
goto out_err;
+ }
+
+ spec->pool_name = pool_name;
+ spec->image_name = image_name;
+ spec->snap_name = snap_name;
return 0;
out_err:
- kfree(reply_buf);
- kfree(rbd_dev->spec->pool_name);
- rbd_dev->spec->pool_name = NULL;
+ kfree(image_name);
+ kfree(pool_name);
return ret;
}
}
if (!ceph_has_room(&p, end, snap_count * sizeof (__le64)))
goto out;
+ ret = 0;
- size = sizeof (struct ceph_snap_context) +
- snap_count * sizeof (snapc->snaps[0]);
- snapc = kmalloc(size, GFP_KERNEL);
+ snapc = rbd_snap_context_create(snap_count);
if (!snapc) {
ret = -ENOMEM;
goto out;
}
- ret = 0;
-
- atomic_set(&snapc->nref, 1);
snapc->seq = seq;
- snapc->num_snaps = snap_count;
for (i = 0; i < snap_count; i++)
snapc->snaps[i] = ceph_decode_64(&p);
size_t len;
char *options;
const char *mon_addrs;
+ char *snap_name;
size_t mon_addrs_size;
struct rbd_spec *spec = NULL;
struct rbd_options *rbd_opts = NULL;
ret = -ENAMETOOLONG;
goto out_err;
}
- spec->snap_name = kmemdup(buf, len + 1, GFP_KERNEL);
- if (!spec->snap_name)
+ snap_name = kmemdup(buf, len + 1, GFP_KERNEL);
+ if (!snap_name)
goto out_mem;
- *(spec->snap_name + len) = '\0';
+ *(snap_name + len) = '\0';
+ spec->snap_name = snap_name;
/* Initialize all rbd options to the defaults */
if (ret)
return ret;
- ret = rbd_dev_probe_update_spec(rbd_dev);
+ ret = rbd_dev_spec_update(rbd_dev);
if (ret)
goto err_out_snaps;
}
rbdc = NULL; /* parent now owns reference */
parent_spec = NULL; /* parent now owns reference */
- ret = rbd_dev_probe(parent);
+ ret = rbd_dev_image_probe(parent);
if (ret < 0)
goto err_out_parent;
rbd_dev->parent = parent;
/* Everything's ready. Announce the disk to the world. */
+ set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE);
+ set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
add_disk(rbd_dev->disk);
pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name,
return ret;
err_out_parent:
+ rbd_spec_put(rbd_dev->parent_spec);
+ kfree(rbd_dev->header_name);
rbd_dev_destroy(parent);
err_out_spec:
rbd_spec_put(parent_spec);
* device. For format 2 images this includes determining the image
* id.
*/
-static int rbd_dev_probe(struct rbd_device *rbd_dev)
+static int rbd_dev_image_probe(struct rbd_device *rbd_dev)
{
int ret;
rc = ceph_pg_poolid_by_name(osdc->osdmap, spec->pool_name);
if (rc < 0)
goto err_out_client;
- spec->pool_id = (u64) rc;
+ spec->pool_id = (u64)rc;
/* The ceph file layout needs to fit pool id in 32 bits */
- if (WARN_ON(spec->pool_id > (u64) U32_MAX)) {
+ if (spec->pool_id > (u64)U32_MAX) {
+ rbd_warn(NULL, "pool id too large (%llu > %u)\n",
+ (unsigned long long)spec->pool_id, U32_MAX);
rc = -EIO;
goto err_out_client;
}
kfree(rbd_opts);
rbd_opts = NULL; /* done with this */
- rc = rbd_dev_probe(rbd_dev);
+ rc = rbd_dev_image_probe(rbd_dev);
if (rc < 0)
goto err_out_rbd_dev;
return count;
err_out_rbd_dev:
+ rbd_spec_put(rbd_dev->parent_spec);
+ kfree(rbd_dev->header_name);
rbd_dev_destroy(rbd_dev);
err_out_client:
rbd_put_client(rbdc);
dout("Error adding device %s\n", buf);
- return (ssize_t) rc;
+ return (ssize_t)rc;
}
static struct rbd_device *__rbd_get_dev(unsigned long dev_id)
/* done with the id, and with the rbd_dev */
rbd_dev_id_put(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 */
projects / ~andy / linux / blobdiff