1602adf40SYehuda Sadeh /* 2602adf40SYehuda Sadeh rbd.c -- Export ceph rados objects as a Linux block device 3602adf40SYehuda Sadeh 4602adf40SYehuda Sadeh 5602adf40SYehuda Sadeh based on drivers/block/osdblk.c: 6602adf40SYehuda Sadeh 7602adf40SYehuda Sadeh Copyright 2009 Red Hat, Inc. 8602adf40SYehuda Sadeh 9602adf40SYehuda Sadeh This program is free software; you can redistribute it and/or modify 10602adf40SYehuda Sadeh it under the terms of the GNU General Public License as published by 11602adf40SYehuda Sadeh the Free Software Foundation. 12602adf40SYehuda Sadeh 13602adf40SYehuda Sadeh This program is distributed in the hope that it will be useful, 14602adf40SYehuda Sadeh but WITHOUT ANY WARRANTY; without even the implied warranty of 15602adf40SYehuda Sadeh MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 16602adf40SYehuda Sadeh GNU General Public License for more details. 17602adf40SYehuda Sadeh 18602adf40SYehuda Sadeh You should have received a copy of the GNU General Public License 19602adf40SYehuda Sadeh along with this program; see the file COPYING. If not, write to 20602adf40SYehuda Sadeh the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 21602adf40SYehuda Sadeh 22602adf40SYehuda Sadeh 23602adf40SYehuda Sadeh 24dfc5606dSYehuda Sadeh For usage instructions, please refer to: 25602adf40SYehuda Sadeh 26dfc5606dSYehuda Sadeh Documentation/ABI/testing/sysfs-bus-rbd 27602adf40SYehuda Sadeh 28602adf40SYehuda Sadeh */ 29602adf40SYehuda Sadeh 30602adf40SYehuda Sadeh #include <linux/ceph/libceph.h> 31602adf40SYehuda Sadeh #include <linux/ceph/osd_client.h> 32602adf40SYehuda Sadeh #include <linux/ceph/mon_client.h> 33602adf40SYehuda Sadeh #include <linux/ceph/decode.h> 3459c2be1eSYehuda Sadeh #include <linux/parser.h> 35602adf40SYehuda Sadeh 36602adf40SYehuda Sadeh #include <linux/kernel.h> 37602adf40SYehuda Sadeh #include <linux/device.h> 38602adf40SYehuda Sadeh #include <linux/module.h> 39602adf40SYehuda Sadeh #include <linux/fs.h> 40602adf40SYehuda Sadeh #include <linux/blkdev.h> 41602adf40SYehuda Sadeh 42602adf40SYehuda Sadeh #include "rbd_types.h" 43602adf40SYehuda Sadeh 44aafb230eSAlex Elder #define RBD_DEBUG /* Activate rbd_assert() calls */ 45aafb230eSAlex Elder 46593a9e7bSAlex Elder /* 47593a9e7bSAlex Elder * The basic unit of block I/O is a sector. It is interpreted in a 48593a9e7bSAlex Elder * number of contexts in Linux (blk, bio, genhd), but the default is 49593a9e7bSAlex Elder * universally 512 bytes. These symbols are just slightly more 50593a9e7bSAlex Elder * meaningful than the bare numbers they represent. 51593a9e7bSAlex Elder */ 52593a9e7bSAlex Elder #define SECTOR_SHIFT 9 53593a9e7bSAlex Elder #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) 54593a9e7bSAlex Elder 55df111be6SAlex Elder /* It might be useful to have this defined elsewhere too */ 56df111be6SAlex Elder 570ec8ce87SAlex Elder #define U32_MAX ((u32) (~0U)) 58df111be6SAlex Elder #define U64_MAX ((u64) (~0ULL)) 59df111be6SAlex Elder 60f0f8cef5SAlex Elder #define RBD_DRV_NAME "rbd" 61f0f8cef5SAlex Elder #define RBD_DRV_NAME_LONG "rbd (rados block device)" 62602adf40SYehuda Sadeh 63602adf40SYehuda Sadeh #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ 64602adf40SYehuda Sadeh 65d4b125e9SAlex Elder #define RBD_SNAP_DEV_NAME_PREFIX "snap_" 66d4b125e9SAlex Elder #define RBD_MAX_SNAP_NAME_LEN \ 67d4b125e9SAlex Elder (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1)) 68d4b125e9SAlex Elder 6935d489f9SAlex Elder #define RBD_MAX_SNAP_COUNT 510 /* allows max snapc to fit in 4KB */ 70602adf40SYehuda Sadeh #define RBD_MAX_OPT_LEN 1024 71602adf40SYehuda Sadeh 72602adf40SYehuda Sadeh #define RBD_SNAP_HEAD_NAME "-" 73602adf40SYehuda Sadeh 749e15b77dSAlex Elder /* This allows a single page to hold an image name sent by OSD */ 759e15b77dSAlex Elder #define RBD_IMAGE_NAME_LEN_MAX (PAGE_SIZE - sizeof (__le32) - 1) 76589d30e0SAlex Elder #define RBD_IMAGE_ID_LEN_MAX 64 779e15b77dSAlex Elder 781e130199SAlex Elder #define RBD_OBJ_PREFIX_LEN_MAX 64 79589d30e0SAlex Elder 80d889140cSAlex Elder /* Feature bits */ 81d889140cSAlex Elder 82d889140cSAlex Elder #define RBD_FEATURE_LAYERING 1 83d889140cSAlex Elder 84d889140cSAlex Elder /* Features supported by this (client software) implementation. */ 85d889140cSAlex Elder 86d889140cSAlex Elder #define RBD_FEATURES_ALL (0) 87d889140cSAlex Elder 8881a89793SAlex Elder /* 8981a89793SAlex Elder * An RBD device name will be "rbd#", where the "rbd" comes from 9081a89793SAlex Elder * RBD_DRV_NAME above, and # is a unique integer identifier. 9181a89793SAlex Elder * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big 9281a89793SAlex Elder * enough to hold all possible device names. 9381a89793SAlex Elder */ 94602adf40SYehuda Sadeh #define DEV_NAME_LEN 32 9581a89793SAlex Elder #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) 96602adf40SYehuda Sadeh 97cc0538b6SAlex Elder #define RBD_READ_ONLY_DEFAULT false 9859c2be1eSYehuda Sadeh 99602adf40SYehuda Sadeh /* 100602adf40SYehuda Sadeh * block device image metadata (in-memory version) 101602adf40SYehuda Sadeh */ 102602adf40SYehuda Sadeh struct rbd_image_header { 103f84344f3SAlex Elder /* These four fields never change for a given rbd image */ 104849b4260SAlex Elder char *object_prefix; 10534b13184SAlex Elder u64 features; 106602adf40SYehuda Sadeh __u8 obj_order; 107602adf40SYehuda Sadeh __u8 crypt_type; 108602adf40SYehuda Sadeh __u8 comp_type; 109602adf40SYehuda Sadeh 110f84344f3SAlex Elder /* The remaining fields need to be updated occasionally */ 111f84344f3SAlex Elder u64 image_size; 112f84344f3SAlex Elder struct ceph_snap_context *snapc; 113602adf40SYehuda Sadeh char *snap_names; 114602adf40SYehuda Sadeh u64 *snap_sizes; 11559c2be1eSYehuda Sadeh 11659c2be1eSYehuda Sadeh u64 obj_version; 11759c2be1eSYehuda Sadeh }; 11859c2be1eSYehuda Sadeh 1190d7dbfceSAlex Elder /* 1200d7dbfceSAlex Elder * An rbd image specification. 1210d7dbfceSAlex Elder * 1220d7dbfceSAlex Elder * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely 123c66c6e0cSAlex Elder * identify an image. Each rbd_dev structure includes a pointer to 124c66c6e0cSAlex Elder * an rbd_spec structure that encapsulates this identity. 125c66c6e0cSAlex Elder * 126c66c6e0cSAlex Elder * Each of the id's in an rbd_spec has an associated name. For a 127c66c6e0cSAlex Elder * user-mapped image, the names are supplied and the id's associated 128c66c6e0cSAlex Elder * with them are looked up. For a layered image, a parent image is 129c66c6e0cSAlex Elder * defined by the tuple, and the names are looked up. 130c66c6e0cSAlex Elder * 131c66c6e0cSAlex Elder * An rbd_dev structure contains a parent_spec pointer which is 132c66c6e0cSAlex Elder * non-null if the image it represents is a child in a layered 133c66c6e0cSAlex Elder * image. This pointer will refer to the rbd_spec structure used 134c66c6e0cSAlex Elder * by the parent rbd_dev for its own identity (i.e., the structure 135c66c6e0cSAlex Elder * is shared between the parent and child). 136c66c6e0cSAlex Elder * 137c66c6e0cSAlex Elder * Since these structures are populated once, during the discovery 138c66c6e0cSAlex Elder * phase of image construction, they are effectively immutable so 139c66c6e0cSAlex Elder * we make no effort to synchronize access to them. 140c66c6e0cSAlex Elder * 141c66c6e0cSAlex Elder * Note that code herein does not assume the image name is known (it 142c66c6e0cSAlex Elder * could be a null pointer). 1430d7dbfceSAlex Elder */ 1440d7dbfceSAlex Elder struct rbd_spec { 1450d7dbfceSAlex Elder u64 pool_id; 1460d7dbfceSAlex Elder char *pool_name; 1470d7dbfceSAlex Elder 1480d7dbfceSAlex Elder char *image_id; 1490d7dbfceSAlex Elder char *image_name; 1500d7dbfceSAlex Elder 1510d7dbfceSAlex Elder u64 snap_id; 1520d7dbfceSAlex Elder char *snap_name; 1530d7dbfceSAlex Elder 1540d7dbfceSAlex Elder struct kref kref; 1550d7dbfceSAlex Elder }; 1560d7dbfceSAlex Elder 15759c2be1eSYehuda Sadeh struct rbd_options { 158cc0538b6SAlex Elder bool read_only; 159602adf40SYehuda Sadeh }; 160602adf40SYehuda Sadeh 161602adf40SYehuda Sadeh /* 162f0f8cef5SAlex Elder * an instance of the client. multiple devices may share an rbd client. 163602adf40SYehuda Sadeh */ 164602adf40SYehuda Sadeh struct rbd_client { 165602adf40SYehuda Sadeh struct ceph_client *client; 166602adf40SYehuda Sadeh struct kref kref; 167602adf40SYehuda Sadeh struct list_head node; 168602adf40SYehuda Sadeh }; 169602adf40SYehuda Sadeh 170602adf40SYehuda Sadeh /* 171f0f8cef5SAlex Elder * a request completion status 172602adf40SYehuda Sadeh */ 1731fec7093SYehuda Sadeh struct rbd_req_status { 1741fec7093SYehuda Sadeh int done; 1758986cb37SAlex Elder s32 rc; 1761fec7093SYehuda Sadeh u64 bytes; 1771fec7093SYehuda Sadeh }; 1781fec7093SYehuda Sadeh 1791fec7093SYehuda Sadeh /* 1801fec7093SYehuda Sadeh * a collection of requests 1811fec7093SYehuda Sadeh */ 1821fec7093SYehuda Sadeh struct rbd_req_coll { 1831fec7093SYehuda Sadeh int total; 1841fec7093SYehuda Sadeh int num_done; 1851fec7093SYehuda Sadeh struct kref kref; 1861fec7093SYehuda Sadeh struct rbd_req_status status[0]; 187602adf40SYehuda Sadeh }; 188602adf40SYehuda Sadeh 189f0f8cef5SAlex Elder /* 190f0f8cef5SAlex Elder * a single io request 191f0f8cef5SAlex Elder */ 192f0f8cef5SAlex Elder struct rbd_request { 193f0f8cef5SAlex Elder struct request *rq; /* blk layer request */ 194f0f8cef5SAlex Elder struct bio *bio; /* cloned bio */ 195f0f8cef5SAlex Elder struct page **pages; /* list of used pages */ 196f0f8cef5SAlex Elder u64 len; 197f0f8cef5SAlex Elder int coll_index; 198f0f8cef5SAlex Elder struct rbd_req_coll *coll; 199f0f8cef5SAlex Elder }; 200f0f8cef5SAlex Elder 201dfc5606dSYehuda Sadeh struct rbd_snap { 202dfc5606dSYehuda Sadeh struct device dev; 203dfc5606dSYehuda Sadeh const char *name; 2043591538fSJosh Durgin u64 size; 205dfc5606dSYehuda Sadeh struct list_head node; 206dfc5606dSYehuda Sadeh u64 id; 20734b13184SAlex Elder u64 features; 208dfc5606dSYehuda Sadeh }; 209dfc5606dSYehuda Sadeh 210f84344f3SAlex Elder struct rbd_mapping { 21199c1f08fSAlex Elder u64 size; 21234b13184SAlex Elder u64 features; 213f84344f3SAlex Elder bool read_only; 214f84344f3SAlex Elder }; 215f84344f3SAlex Elder 216602adf40SYehuda Sadeh /* 217602adf40SYehuda Sadeh * a single device 218602adf40SYehuda Sadeh */ 219602adf40SYehuda Sadeh struct rbd_device { 220de71a297SAlex Elder int dev_id; /* blkdev unique id */ 221602adf40SYehuda Sadeh 222602adf40SYehuda Sadeh int major; /* blkdev assigned major */ 223602adf40SYehuda Sadeh struct gendisk *disk; /* blkdev's gendisk and rq */ 224602adf40SYehuda Sadeh 225a30b71b9SAlex Elder u32 image_format; /* Either 1 or 2 */ 226602adf40SYehuda Sadeh struct rbd_client *rbd_client; 227602adf40SYehuda Sadeh 228602adf40SYehuda Sadeh char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ 229602adf40SYehuda Sadeh 230602adf40SYehuda Sadeh spinlock_t lock; /* queue lock */ 231602adf40SYehuda Sadeh 232602adf40SYehuda Sadeh struct rbd_image_header header; 233d78b650aSAlex Elder atomic_t exists; 2340d7dbfceSAlex Elder struct rbd_spec *spec; 235602adf40SYehuda Sadeh 2360d7dbfceSAlex Elder char *header_name; 237971f839aSAlex Elder 23859c2be1eSYehuda Sadeh struct ceph_osd_event *watch_event; 23959c2be1eSYehuda Sadeh struct ceph_osd_request *watch_request; 24059c2be1eSYehuda Sadeh 24186b00e0dSAlex Elder struct rbd_spec *parent_spec; 24286b00e0dSAlex Elder u64 parent_overlap; 24386b00e0dSAlex Elder 244c666601aSJosh Durgin /* protects updating the header */ 245c666601aSJosh Durgin struct rw_semaphore header_rwsem; 246f84344f3SAlex Elder 247f84344f3SAlex Elder struct rbd_mapping mapping; 248602adf40SYehuda Sadeh 249602adf40SYehuda Sadeh struct list_head node; 250dfc5606dSYehuda Sadeh 251dfc5606dSYehuda Sadeh /* list of snapshots */ 252dfc5606dSYehuda Sadeh struct list_head snaps; 253dfc5606dSYehuda Sadeh 254dfc5606dSYehuda Sadeh /* sysfs related */ 255dfc5606dSYehuda Sadeh struct device dev; 25642382b70SAlex Elder unsigned long open_count; 257dfc5606dSYehuda Sadeh }; 258dfc5606dSYehuda Sadeh 259602adf40SYehuda Sadeh static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ 260e124a82fSAlex Elder 261602adf40SYehuda Sadeh static LIST_HEAD(rbd_dev_list); /* devices */ 262e124a82fSAlex Elder static DEFINE_SPINLOCK(rbd_dev_list_lock); 263e124a82fSAlex Elder 264602adf40SYehuda Sadeh static LIST_HEAD(rbd_client_list); /* clients */ 265432b8587SAlex Elder static DEFINE_SPINLOCK(rbd_client_list_lock); 266602adf40SYehuda Sadeh 267304f6808SAlex Elder static int rbd_dev_snaps_update(struct rbd_device *rbd_dev); 268304f6808SAlex Elder static int rbd_dev_snaps_register(struct rbd_device *rbd_dev); 269304f6808SAlex Elder 270dfc5606dSYehuda Sadeh static void rbd_dev_release(struct device *dev); 27141f38c2bSAlex Elder static void rbd_remove_snap_dev(struct rbd_snap *snap); 272dfc5606dSYehuda Sadeh 273f0f8cef5SAlex Elder static ssize_t rbd_add(struct bus_type *bus, const char *buf, 274f0f8cef5SAlex Elder size_t count); 275f0f8cef5SAlex Elder static ssize_t rbd_remove(struct bus_type *bus, const char *buf, 276f0f8cef5SAlex Elder size_t count); 277f0f8cef5SAlex Elder 278f0f8cef5SAlex Elder static struct bus_attribute rbd_bus_attrs[] = { 279f0f8cef5SAlex Elder __ATTR(add, S_IWUSR, NULL, rbd_add), 280f0f8cef5SAlex Elder __ATTR(remove, S_IWUSR, NULL, rbd_remove), 281f0f8cef5SAlex Elder __ATTR_NULL 282f0f8cef5SAlex Elder }; 283f0f8cef5SAlex Elder 284f0f8cef5SAlex Elder static struct bus_type rbd_bus_type = { 285f0f8cef5SAlex Elder .name = "rbd", 286f0f8cef5SAlex Elder .bus_attrs = rbd_bus_attrs, 287f0f8cef5SAlex Elder }; 288f0f8cef5SAlex Elder 289f0f8cef5SAlex Elder static void rbd_root_dev_release(struct device *dev) 290f0f8cef5SAlex Elder { 291f0f8cef5SAlex Elder } 292f0f8cef5SAlex Elder 293f0f8cef5SAlex Elder static struct device rbd_root_dev = { 294f0f8cef5SAlex Elder .init_name = "rbd", 295f0f8cef5SAlex Elder .release = rbd_root_dev_release, 296f0f8cef5SAlex Elder }; 297f0f8cef5SAlex Elder 29806ecc6cbSAlex Elder static __printf(2, 3) 29906ecc6cbSAlex Elder void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...) 30006ecc6cbSAlex Elder { 30106ecc6cbSAlex Elder struct va_format vaf; 30206ecc6cbSAlex Elder va_list args; 30306ecc6cbSAlex Elder 30406ecc6cbSAlex Elder va_start(args, fmt); 30506ecc6cbSAlex Elder vaf.fmt = fmt; 30606ecc6cbSAlex Elder vaf.va = &args; 30706ecc6cbSAlex Elder 30806ecc6cbSAlex Elder if (!rbd_dev) 30906ecc6cbSAlex Elder printk(KERN_WARNING "%s: %pV\n", RBD_DRV_NAME, &vaf); 31006ecc6cbSAlex Elder else if (rbd_dev->disk) 31106ecc6cbSAlex Elder printk(KERN_WARNING "%s: %s: %pV\n", 31206ecc6cbSAlex Elder RBD_DRV_NAME, rbd_dev->disk->disk_name, &vaf); 31306ecc6cbSAlex Elder else if (rbd_dev->spec && rbd_dev->spec->image_name) 31406ecc6cbSAlex Elder printk(KERN_WARNING "%s: image %s: %pV\n", 31506ecc6cbSAlex Elder RBD_DRV_NAME, rbd_dev->spec->image_name, &vaf); 31606ecc6cbSAlex Elder else if (rbd_dev->spec && rbd_dev->spec->image_id) 31706ecc6cbSAlex Elder printk(KERN_WARNING "%s: id %s: %pV\n", 31806ecc6cbSAlex Elder RBD_DRV_NAME, rbd_dev->spec->image_id, &vaf); 31906ecc6cbSAlex Elder else /* punt */ 32006ecc6cbSAlex Elder printk(KERN_WARNING "%s: rbd_dev %p: %pV\n", 32106ecc6cbSAlex Elder RBD_DRV_NAME, rbd_dev, &vaf); 32206ecc6cbSAlex Elder va_end(args); 32306ecc6cbSAlex Elder } 32406ecc6cbSAlex Elder 325aafb230eSAlex Elder #ifdef RBD_DEBUG 326aafb230eSAlex Elder #define rbd_assert(expr) \ 327aafb230eSAlex Elder if (unlikely(!(expr))) { \ 328aafb230eSAlex Elder printk(KERN_ERR "\nAssertion failure in %s() " \ 329aafb230eSAlex Elder "at line %d:\n\n" \ 330aafb230eSAlex Elder "\trbd_assert(%s);\n\n", \ 331aafb230eSAlex Elder __func__, __LINE__, #expr); \ 332aafb230eSAlex Elder BUG(); \ 333aafb230eSAlex Elder } 334aafb230eSAlex Elder #else /* !RBD_DEBUG */ 335aafb230eSAlex Elder # define rbd_assert(expr) ((void) 0) 336aafb230eSAlex Elder #endif /* !RBD_DEBUG */ 337dfc5606dSYehuda Sadeh 338117973fbSAlex Elder static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver); 339117973fbSAlex Elder static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver); 34059c2be1eSYehuda Sadeh 341602adf40SYehuda Sadeh static int rbd_open(struct block_device *bdev, fmode_t mode) 342602adf40SYehuda Sadeh { 343f0f8cef5SAlex Elder struct rbd_device *rbd_dev = bdev->bd_disk->private_data; 344602adf40SYehuda Sadeh 345f84344f3SAlex Elder if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only) 346602adf40SYehuda Sadeh return -EROFS; 347602adf40SYehuda Sadeh 34842382b70SAlex Elder mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); 349c3e946ceSAlex Elder (void) get_device(&rbd_dev->dev); 350f84344f3SAlex Elder set_device_ro(bdev, rbd_dev->mapping.read_only); 35142382b70SAlex Elder rbd_dev->open_count++; 35242382b70SAlex Elder mutex_unlock(&ctl_mutex); 353340c7a2bSAlex Elder 354602adf40SYehuda Sadeh return 0; 355602adf40SYehuda Sadeh } 356602adf40SYehuda Sadeh 357dfc5606dSYehuda Sadeh static int rbd_release(struct gendisk *disk, fmode_t mode) 358dfc5606dSYehuda Sadeh { 359dfc5606dSYehuda Sadeh struct rbd_device *rbd_dev = disk->private_data; 360dfc5606dSYehuda Sadeh 36142382b70SAlex Elder mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); 36242382b70SAlex Elder rbd_assert(rbd_dev->open_count > 0); 36342382b70SAlex Elder rbd_dev->open_count--; 364c3e946ceSAlex Elder put_device(&rbd_dev->dev); 36542382b70SAlex Elder mutex_unlock(&ctl_mutex); 366dfc5606dSYehuda Sadeh 367dfc5606dSYehuda Sadeh return 0; 368dfc5606dSYehuda Sadeh } 369dfc5606dSYehuda Sadeh 370602adf40SYehuda Sadeh static const struct block_device_operations rbd_bd_ops = { 371602adf40SYehuda Sadeh .owner = THIS_MODULE, 372602adf40SYehuda Sadeh .open = rbd_open, 373dfc5606dSYehuda Sadeh .release = rbd_release, 374602adf40SYehuda Sadeh }; 375602adf40SYehuda Sadeh 376602adf40SYehuda Sadeh /* 377602adf40SYehuda Sadeh * Initialize an rbd client instance. 37843ae4701SAlex Elder * We own *ceph_opts. 379602adf40SYehuda Sadeh */ 380f8c38929SAlex Elder static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) 381602adf40SYehuda Sadeh { 382602adf40SYehuda Sadeh struct rbd_client *rbdc; 383602adf40SYehuda Sadeh int ret = -ENOMEM; 384602adf40SYehuda Sadeh 385602adf40SYehuda Sadeh dout("rbd_client_create\n"); 386602adf40SYehuda Sadeh rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); 387602adf40SYehuda Sadeh if (!rbdc) 388602adf40SYehuda Sadeh goto out_opt; 389602adf40SYehuda Sadeh 390602adf40SYehuda Sadeh kref_init(&rbdc->kref); 391602adf40SYehuda Sadeh INIT_LIST_HEAD(&rbdc->node); 392602adf40SYehuda Sadeh 393bc534d86SAlex Elder mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); 394bc534d86SAlex Elder 39543ae4701SAlex Elder rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); 396602adf40SYehuda Sadeh if (IS_ERR(rbdc->client)) 397bc534d86SAlex Elder goto out_mutex; 39843ae4701SAlex Elder ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ 399602adf40SYehuda Sadeh 400602adf40SYehuda Sadeh ret = ceph_open_session(rbdc->client); 401602adf40SYehuda Sadeh if (ret < 0) 402602adf40SYehuda Sadeh goto out_err; 403602adf40SYehuda Sadeh 404432b8587SAlex Elder spin_lock(&rbd_client_list_lock); 405602adf40SYehuda Sadeh list_add_tail(&rbdc->node, &rbd_client_list); 406432b8587SAlex Elder spin_unlock(&rbd_client_list_lock); 407602adf40SYehuda Sadeh 408bc534d86SAlex Elder mutex_unlock(&ctl_mutex); 409bc534d86SAlex Elder 410602adf40SYehuda Sadeh dout("rbd_client_create created %p\n", rbdc); 411602adf40SYehuda Sadeh return rbdc; 412602adf40SYehuda Sadeh 413602adf40SYehuda Sadeh out_err: 414602adf40SYehuda Sadeh ceph_destroy_client(rbdc->client); 415bc534d86SAlex Elder out_mutex: 416bc534d86SAlex Elder mutex_unlock(&ctl_mutex); 417602adf40SYehuda Sadeh kfree(rbdc); 418602adf40SYehuda Sadeh out_opt: 41943ae4701SAlex Elder if (ceph_opts) 42043ae4701SAlex Elder ceph_destroy_options(ceph_opts); 42128f259b7SVasiliy Kulikov return ERR_PTR(ret); 422602adf40SYehuda Sadeh } 423602adf40SYehuda Sadeh 424602adf40SYehuda Sadeh /* 4251f7ba331SAlex Elder * Find a ceph client with specific addr and configuration. If 4261f7ba331SAlex Elder * found, bump its reference count. 427602adf40SYehuda Sadeh */ 4281f7ba331SAlex Elder static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) 429602adf40SYehuda Sadeh { 430602adf40SYehuda Sadeh struct rbd_client *client_node; 4311f7ba331SAlex Elder bool found = false; 432602adf40SYehuda Sadeh 43343ae4701SAlex Elder if (ceph_opts->flags & CEPH_OPT_NOSHARE) 434602adf40SYehuda Sadeh return NULL; 435602adf40SYehuda Sadeh 4361f7ba331SAlex Elder spin_lock(&rbd_client_list_lock); 4371f7ba331SAlex Elder list_for_each_entry(client_node, &rbd_client_list, node) { 4381f7ba331SAlex Elder if (!ceph_compare_options(ceph_opts, client_node->client)) { 4391f7ba331SAlex Elder kref_get(&client_node->kref); 4401f7ba331SAlex Elder found = true; 4411f7ba331SAlex Elder break; 4421f7ba331SAlex Elder } 4431f7ba331SAlex Elder } 4441f7ba331SAlex Elder spin_unlock(&rbd_client_list_lock); 4451f7ba331SAlex Elder 4461f7ba331SAlex Elder return found ? client_node : NULL; 447602adf40SYehuda Sadeh } 448602adf40SYehuda Sadeh 449602adf40SYehuda Sadeh /* 45059c2be1eSYehuda Sadeh * mount options 45159c2be1eSYehuda Sadeh */ 45259c2be1eSYehuda Sadeh enum { 45359c2be1eSYehuda Sadeh Opt_last_int, 45459c2be1eSYehuda Sadeh /* int args above */ 45559c2be1eSYehuda Sadeh Opt_last_string, 45659c2be1eSYehuda Sadeh /* string args above */ 457cc0538b6SAlex Elder Opt_read_only, 458cc0538b6SAlex Elder Opt_read_write, 459cc0538b6SAlex Elder /* Boolean args above */ 460cc0538b6SAlex Elder Opt_last_bool, 46159c2be1eSYehuda Sadeh }; 46259c2be1eSYehuda Sadeh 46343ae4701SAlex Elder static match_table_t rbd_opts_tokens = { 46459c2be1eSYehuda Sadeh /* int args above */ 46559c2be1eSYehuda Sadeh /* string args above */ 466be466c1cSAlex Elder {Opt_read_only, "read_only"}, 467cc0538b6SAlex Elder {Opt_read_only, "ro"}, /* Alternate spelling */ 468cc0538b6SAlex Elder {Opt_read_write, "read_write"}, 469cc0538b6SAlex Elder {Opt_read_write, "rw"}, /* Alternate spelling */ 470cc0538b6SAlex Elder /* Boolean args above */ 47159c2be1eSYehuda Sadeh {-1, NULL} 47259c2be1eSYehuda Sadeh }; 47359c2be1eSYehuda Sadeh 47459c2be1eSYehuda Sadeh static int parse_rbd_opts_token(char *c, void *private) 47559c2be1eSYehuda Sadeh { 47643ae4701SAlex Elder struct rbd_options *rbd_opts = private; 47759c2be1eSYehuda Sadeh substring_t argstr[MAX_OPT_ARGS]; 47859c2be1eSYehuda Sadeh int token, intval, ret; 47959c2be1eSYehuda Sadeh 48043ae4701SAlex Elder token = match_token(c, rbd_opts_tokens, argstr); 48159c2be1eSYehuda Sadeh if (token < 0) 48259c2be1eSYehuda Sadeh return -EINVAL; 48359c2be1eSYehuda Sadeh 48459c2be1eSYehuda Sadeh if (token < Opt_last_int) { 48559c2be1eSYehuda Sadeh ret = match_int(&argstr[0], &intval); 48659c2be1eSYehuda Sadeh if (ret < 0) { 48759c2be1eSYehuda Sadeh pr_err("bad mount option arg (not int) " 48859c2be1eSYehuda Sadeh "at '%s'\n", c); 48959c2be1eSYehuda Sadeh return ret; 49059c2be1eSYehuda Sadeh } 49159c2be1eSYehuda Sadeh dout("got int token %d val %d\n", token, intval); 49259c2be1eSYehuda Sadeh } else if (token > Opt_last_int && token < Opt_last_string) { 49359c2be1eSYehuda Sadeh dout("got string token %d val %s\n", token, 49459c2be1eSYehuda Sadeh argstr[0].from); 495cc0538b6SAlex Elder } else if (token > Opt_last_string && token < Opt_last_bool) { 496cc0538b6SAlex Elder dout("got Boolean token %d\n", token); 49759c2be1eSYehuda Sadeh } else { 49859c2be1eSYehuda Sadeh dout("got token %d\n", token); 49959c2be1eSYehuda Sadeh } 50059c2be1eSYehuda Sadeh 50159c2be1eSYehuda Sadeh switch (token) { 502cc0538b6SAlex Elder case Opt_read_only: 503cc0538b6SAlex Elder rbd_opts->read_only = true; 504cc0538b6SAlex Elder break; 505cc0538b6SAlex Elder case Opt_read_write: 506cc0538b6SAlex Elder rbd_opts->read_only = false; 507cc0538b6SAlex Elder break; 50859c2be1eSYehuda Sadeh default: 509aafb230eSAlex Elder rbd_assert(false); 510aafb230eSAlex Elder break; 51159c2be1eSYehuda Sadeh } 51259c2be1eSYehuda Sadeh return 0; 51359c2be1eSYehuda Sadeh } 51459c2be1eSYehuda Sadeh 51559c2be1eSYehuda Sadeh /* 516602adf40SYehuda Sadeh * Get a ceph client with specific addr and configuration, if one does 517602adf40SYehuda Sadeh * not exist create it. 518602adf40SYehuda Sadeh */ 5199d3997fdSAlex Elder static struct rbd_client *rbd_get_client(struct ceph_options *ceph_opts) 520602adf40SYehuda Sadeh { 521f8c38929SAlex Elder struct rbd_client *rbdc; 52259c2be1eSYehuda Sadeh 5231f7ba331SAlex Elder rbdc = rbd_client_find(ceph_opts); 5249d3997fdSAlex Elder if (rbdc) /* using an existing client */ 52543ae4701SAlex Elder ceph_destroy_options(ceph_opts); 5269d3997fdSAlex Elder else 527f8c38929SAlex Elder rbdc = rbd_client_create(ceph_opts); 528d720bcb0SAlex Elder 5299d3997fdSAlex Elder return rbdc; 530602adf40SYehuda Sadeh } 531602adf40SYehuda Sadeh 532602adf40SYehuda Sadeh /* 533602adf40SYehuda Sadeh * Destroy ceph client 534d23a4b3fSAlex Elder * 535432b8587SAlex Elder * Caller must hold rbd_client_list_lock. 536602adf40SYehuda Sadeh */ 537602adf40SYehuda Sadeh static void rbd_client_release(struct kref *kref) 538602adf40SYehuda Sadeh { 539602adf40SYehuda Sadeh struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); 540602adf40SYehuda Sadeh 541602adf40SYehuda Sadeh dout("rbd_release_client %p\n", rbdc); 542cd9d9f5dSAlex Elder spin_lock(&rbd_client_list_lock); 543602adf40SYehuda Sadeh list_del(&rbdc->node); 544cd9d9f5dSAlex Elder spin_unlock(&rbd_client_list_lock); 545602adf40SYehuda Sadeh 546602adf40SYehuda Sadeh ceph_destroy_client(rbdc->client); 547602adf40SYehuda Sadeh kfree(rbdc); 548602adf40SYehuda Sadeh } 549602adf40SYehuda Sadeh 550602adf40SYehuda Sadeh /* 551602adf40SYehuda Sadeh * Drop reference to ceph client node. If it's not referenced anymore, release 552602adf40SYehuda Sadeh * it. 553602adf40SYehuda Sadeh */ 5549d3997fdSAlex Elder static void rbd_put_client(struct rbd_client *rbdc) 555602adf40SYehuda Sadeh { 556c53d5893SAlex Elder if (rbdc) 5579d3997fdSAlex Elder kref_put(&rbdc->kref, rbd_client_release); 558602adf40SYehuda Sadeh } 559602adf40SYehuda Sadeh 5601fec7093SYehuda Sadeh /* 5611fec7093SYehuda Sadeh * Destroy requests collection 5621fec7093SYehuda Sadeh */ 5631fec7093SYehuda Sadeh static void rbd_coll_release(struct kref *kref) 5641fec7093SYehuda Sadeh { 5651fec7093SYehuda Sadeh struct rbd_req_coll *coll = 5661fec7093SYehuda Sadeh container_of(kref, struct rbd_req_coll, kref); 5671fec7093SYehuda Sadeh 5681fec7093SYehuda Sadeh dout("rbd_coll_release %p\n", coll); 5691fec7093SYehuda Sadeh kfree(coll); 5701fec7093SYehuda Sadeh } 571602adf40SYehuda Sadeh 572a30b71b9SAlex Elder static bool rbd_image_format_valid(u32 image_format) 573a30b71b9SAlex Elder { 574a30b71b9SAlex Elder return image_format == 1 || image_format == 2; 575a30b71b9SAlex Elder } 576a30b71b9SAlex Elder 5778e94af8eSAlex Elder static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) 5788e94af8eSAlex Elder { 579103a150fSAlex Elder size_t size; 580103a150fSAlex Elder u32 snap_count; 581103a150fSAlex Elder 582103a150fSAlex Elder /* The header has to start with the magic rbd header text */ 583103a150fSAlex Elder if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) 584103a150fSAlex Elder return false; 585103a150fSAlex Elder 586db2388b6SAlex Elder /* The bio layer requires at least sector-sized I/O */ 587db2388b6SAlex Elder 588db2388b6SAlex Elder if (ondisk->options.order < SECTOR_SHIFT) 589db2388b6SAlex Elder return false; 590db2388b6SAlex Elder 591db2388b6SAlex Elder /* If we use u64 in a few spots we may be able to loosen this */ 592db2388b6SAlex Elder 593db2388b6SAlex Elder if (ondisk->options.order > 8 * sizeof (int) - 1) 594db2388b6SAlex Elder return false; 595db2388b6SAlex Elder 596103a150fSAlex Elder /* 597103a150fSAlex Elder * The size of a snapshot header has to fit in a size_t, and 598103a150fSAlex Elder * that limits the number of snapshots. 599103a150fSAlex Elder */ 600103a150fSAlex Elder snap_count = le32_to_cpu(ondisk->snap_count); 601103a150fSAlex Elder size = SIZE_MAX - sizeof (struct ceph_snap_context); 602103a150fSAlex Elder if (snap_count > size / sizeof (__le64)) 603103a150fSAlex Elder return false; 604103a150fSAlex Elder 605103a150fSAlex Elder /* 606103a150fSAlex Elder * Not only that, but the size of the entire the snapshot 607103a150fSAlex Elder * header must also be representable in a size_t. 608103a150fSAlex Elder */ 609103a150fSAlex Elder size -= snap_count * sizeof (__le64); 610103a150fSAlex Elder if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) 611103a150fSAlex Elder return false; 612103a150fSAlex Elder 613103a150fSAlex Elder return true; 6148e94af8eSAlex Elder } 6158e94af8eSAlex Elder 616602adf40SYehuda Sadeh /* 617602adf40SYehuda Sadeh * Create a new header structure, translate header format from the on-disk 618602adf40SYehuda Sadeh * header. 619602adf40SYehuda Sadeh */ 620602adf40SYehuda Sadeh static int rbd_header_from_disk(struct rbd_image_header *header, 6214156d998SAlex Elder struct rbd_image_header_ondisk *ondisk) 622602adf40SYehuda Sadeh { 623ccece235SAlex Elder u32 snap_count; 62458c17b0eSAlex Elder size_t len; 625d2bb24e5SAlex Elder size_t size; 626621901d6SAlex Elder u32 i; 627602adf40SYehuda Sadeh 6286a52325fSAlex Elder memset(header, 0, sizeof (*header)); 6296a52325fSAlex Elder 630103a150fSAlex Elder snap_count = le32_to_cpu(ondisk->snap_count); 631103a150fSAlex Elder 63258c17b0eSAlex Elder len = strnlen(ondisk->object_prefix, sizeof (ondisk->object_prefix)); 63358c17b0eSAlex Elder header->object_prefix = kmalloc(len + 1, GFP_KERNEL); 6346a52325fSAlex Elder if (!header->object_prefix) 635602adf40SYehuda Sadeh return -ENOMEM; 63658c17b0eSAlex Elder memcpy(header->object_prefix, ondisk->object_prefix, len); 63758c17b0eSAlex Elder header->object_prefix[len] = '\0'; 63800f1f36fSAlex Elder 639602adf40SYehuda Sadeh if (snap_count) { 640f785cc1dSAlex Elder u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); 641f785cc1dSAlex Elder 642621901d6SAlex Elder /* Save a copy of the snapshot names */ 643621901d6SAlex Elder 644f785cc1dSAlex Elder if (snap_names_len > (u64) SIZE_MAX) 645f785cc1dSAlex Elder return -EIO; 646f785cc1dSAlex Elder header->snap_names = kmalloc(snap_names_len, GFP_KERNEL); 647602adf40SYehuda Sadeh if (!header->snap_names) 6486a52325fSAlex Elder goto out_err; 649f785cc1dSAlex Elder /* 650f785cc1dSAlex Elder * Note that rbd_dev_v1_header_read() guarantees 651f785cc1dSAlex Elder * the ondisk buffer we're working with has 652f785cc1dSAlex Elder * snap_names_len bytes beyond the end of the 653f785cc1dSAlex Elder * snapshot id array, this memcpy() is safe. 654f785cc1dSAlex Elder */ 655f785cc1dSAlex Elder memcpy(header->snap_names, &ondisk->snaps[snap_count], 656f785cc1dSAlex Elder snap_names_len); 6576a52325fSAlex Elder 658621901d6SAlex Elder /* Record each snapshot's size */ 659621901d6SAlex Elder 660d2bb24e5SAlex Elder size = snap_count * sizeof (*header->snap_sizes); 661d2bb24e5SAlex Elder header->snap_sizes = kmalloc(size, GFP_KERNEL); 662602adf40SYehuda Sadeh if (!header->snap_sizes) 6636a52325fSAlex Elder goto out_err; 664621901d6SAlex Elder for (i = 0; i < snap_count; i++) 665621901d6SAlex Elder header->snap_sizes[i] = 666621901d6SAlex Elder le64_to_cpu(ondisk->snaps[i].image_size); 667602adf40SYehuda Sadeh } else { 668ccece235SAlex Elder WARN_ON(ondisk->snap_names_len); 669602adf40SYehuda Sadeh header->snap_names = NULL; 670602adf40SYehuda Sadeh header->snap_sizes = NULL; 671602adf40SYehuda Sadeh } 672849b4260SAlex Elder 67334b13184SAlex Elder header->features = 0; /* No features support in v1 images */ 674602adf40SYehuda Sadeh header->obj_order = ondisk->options.order; 675602adf40SYehuda Sadeh header->crypt_type = ondisk->options.crypt_type; 676602adf40SYehuda Sadeh header->comp_type = ondisk->options.comp_type; 6776a52325fSAlex Elder 678621901d6SAlex Elder /* Allocate and fill in the snapshot context */ 679621901d6SAlex Elder 680f84344f3SAlex Elder header->image_size = le64_to_cpu(ondisk->image_size); 6816a52325fSAlex Elder size = sizeof (struct ceph_snap_context); 6826a52325fSAlex Elder size += snap_count * sizeof (header->snapc->snaps[0]); 6836a52325fSAlex Elder header->snapc = kzalloc(size, GFP_KERNEL); 6846a52325fSAlex Elder if (!header->snapc) 6856a52325fSAlex Elder goto out_err; 686602adf40SYehuda Sadeh 687602adf40SYehuda Sadeh atomic_set(&header->snapc->nref, 1); 688505cbb9bSAlex Elder header->snapc->seq = le64_to_cpu(ondisk->snap_seq); 689602adf40SYehuda Sadeh header->snapc->num_snaps = snap_count; 690621901d6SAlex Elder for (i = 0; i < snap_count; i++) 691602adf40SYehuda Sadeh header->snapc->snaps[i] = 692602adf40SYehuda Sadeh le64_to_cpu(ondisk->snaps[i].id); 693602adf40SYehuda Sadeh 694602adf40SYehuda Sadeh return 0; 695602adf40SYehuda Sadeh 6966a52325fSAlex Elder out_err: 697849b4260SAlex Elder kfree(header->snap_sizes); 698ccece235SAlex Elder header->snap_sizes = NULL; 699602adf40SYehuda Sadeh kfree(header->snap_names); 700ccece235SAlex Elder header->snap_names = NULL; 7016a52325fSAlex Elder kfree(header->object_prefix); 7026a52325fSAlex Elder header->object_prefix = NULL; 703ccece235SAlex Elder 70400f1f36fSAlex Elder return -ENOMEM; 705602adf40SYehuda Sadeh } 706602adf40SYehuda Sadeh 7079e15b77dSAlex Elder static const char *rbd_snap_name(struct rbd_device *rbd_dev, u64 snap_id) 7089e15b77dSAlex Elder { 7099e15b77dSAlex Elder struct rbd_snap *snap; 7109e15b77dSAlex Elder 7119e15b77dSAlex Elder if (snap_id == CEPH_NOSNAP) 7129e15b77dSAlex Elder return RBD_SNAP_HEAD_NAME; 7139e15b77dSAlex Elder 7149e15b77dSAlex Elder list_for_each_entry(snap, &rbd_dev->snaps, node) 7159e15b77dSAlex Elder if (snap_id == snap->id) 7169e15b77dSAlex Elder return snap->name; 7179e15b77dSAlex Elder 7189e15b77dSAlex Elder return NULL; 7199e15b77dSAlex Elder } 7209e15b77dSAlex Elder 7218836b995SAlex Elder static int snap_by_name(struct rbd_device *rbd_dev, const char *snap_name) 722602adf40SYehuda Sadeh { 723602adf40SYehuda Sadeh 724e86924a8SAlex Elder struct rbd_snap *snap; 72500f1f36fSAlex Elder 726e86924a8SAlex Elder list_for_each_entry(snap, &rbd_dev->snaps, node) { 727e86924a8SAlex Elder if (!strcmp(snap_name, snap->name)) { 7280d7dbfceSAlex Elder rbd_dev->spec->snap_id = snap->id; 729e86924a8SAlex Elder rbd_dev->mapping.size = snap->size; 73034b13184SAlex Elder rbd_dev->mapping.features = snap->features; 73100f1f36fSAlex Elder 732e86924a8SAlex Elder return 0; 733602adf40SYehuda Sadeh } 73400f1f36fSAlex Elder } 735e86924a8SAlex Elder 73600f1f36fSAlex Elder return -ENOENT; 73700f1f36fSAlex Elder } 738602adf40SYehuda Sadeh 739819d52bfSAlex Elder static int rbd_dev_set_mapping(struct rbd_device *rbd_dev) 740602adf40SYehuda Sadeh { 74178dc447dSAlex Elder int ret; 742602adf40SYehuda Sadeh 7430d7dbfceSAlex Elder if (!memcmp(rbd_dev->spec->snap_name, RBD_SNAP_HEAD_NAME, 744cc9d734cSJosh Durgin sizeof (RBD_SNAP_HEAD_NAME))) { 7450d7dbfceSAlex Elder rbd_dev->spec->snap_id = CEPH_NOSNAP; 74699c1f08fSAlex Elder rbd_dev->mapping.size = rbd_dev->header.image_size; 74734b13184SAlex Elder rbd_dev->mapping.features = rbd_dev->header.features; 748e86924a8SAlex Elder ret = 0; 749602adf40SYehuda Sadeh } else { 7500d7dbfceSAlex Elder ret = snap_by_name(rbd_dev, rbd_dev->spec->snap_name); 751602adf40SYehuda Sadeh if (ret < 0) 752602adf40SYehuda Sadeh goto done; 753f84344f3SAlex Elder rbd_dev->mapping.read_only = true; 754602adf40SYehuda Sadeh } 755d78b650aSAlex Elder atomic_set(&rbd_dev->exists, 1); 756602adf40SYehuda Sadeh done: 757602adf40SYehuda Sadeh return ret; 758602adf40SYehuda Sadeh } 759602adf40SYehuda Sadeh 760602adf40SYehuda Sadeh static void rbd_header_free(struct rbd_image_header *header) 761602adf40SYehuda Sadeh { 762849b4260SAlex Elder kfree(header->object_prefix); 763d78fd7aeSAlex Elder header->object_prefix = NULL; 764602adf40SYehuda Sadeh kfree(header->snap_sizes); 765d78fd7aeSAlex Elder header->snap_sizes = NULL; 766849b4260SAlex Elder kfree(header->snap_names); 767d78fd7aeSAlex Elder header->snap_names = NULL; 768d1d25646SJosh Durgin ceph_put_snap_context(header->snapc); 769d78fd7aeSAlex Elder header->snapc = NULL; 770602adf40SYehuda Sadeh } 771602adf40SYehuda Sadeh 77265ccfe21SAlex Elder static char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset) 773602adf40SYehuda Sadeh { 77465ccfe21SAlex Elder char *name; 77565ccfe21SAlex Elder u64 segment; 77665ccfe21SAlex Elder int ret; 777602adf40SYehuda Sadeh 7782fd82b9eSAlex Elder name = kmalloc(MAX_OBJ_NAME_SIZE + 1, GFP_NOIO); 77965ccfe21SAlex Elder if (!name) 78065ccfe21SAlex Elder return NULL; 78165ccfe21SAlex Elder segment = offset >> rbd_dev->header.obj_order; 7822fd82b9eSAlex Elder ret = snprintf(name, MAX_OBJ_NAME_SIZE + 1, "%s.%012llx", 78365ccfe21SAlex Elder rbd_dev->header.object_prefix, segment); 7842fd82b9eSAlex Elder if (ret < 0 || ret > MAX_OBJ_NAME_SIZE) { 78565ccfe21SAlex Elder pr_err("error formatting segment name for #%llu (%d)\n", 78665ccfe21SAlex Elder segment, ret); 78765ccfe21SAlex Elder kfree(name); 78865ccfe21SAlex Elder name = NULL; 78965ccfe21SAlex Elder } 790602adf40SYehuda Sadeh 79165ccfe21SAlex Elder return name; 79265ccfe21SAlex Elder } 793602adf40SYehuda Sadeh 79465ccfe21SAlex Elder static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) 79565ccfe21SAlex Elder { 79665ccfe21SAlex Elder u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; 797602adf40SYehuda Sadeh 79865ccfe21SAlex Elder return offset & (segment_size - 1); 79965ccfe21SAlex Elder } 80065ccfe21SAlex Elder 80165ccfe21SAlex Elder static u64 rbd_segment_length(struct rbd_device *rbd_dev, 80265ccfe21SAlex Elder u64 offset, u64 length) 80365ccfe21SAlex Elder { 80465ccfe21SAlex Elder u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; 80565ccfe21SAlex Elder 80665ccfe21SAlex Elder offset &= segment_size - 1; 80765ccfe21SAlex Elder 808aafb230eSAlex Elder rbd_assert(length <= U64_MAX - offset); 80965ccfe21SAlex Elder if (offset + length > segment_size) 81065ccfe21SAlex Elder length = segment_size - offset; 81165ccfe21SAlex Elder 81265ccfe21SAlex Elder return length; 813602adf40SYehuda Sadeh } 814602adf40SYehuda Sadeh 8151fec7093SYehuda Sadeh static int rbd_get_num_segments(struct rbd_image_header *header, 8161fec7093SYehuda Sadeh u64 ofs, u64 len) 8171fec7093SYehuda Sadeh { 818df111be6SAlex Elder u64 start_seg; 819df111be6SAlex Elder u64 end_seg; 820df111be6SAlex Elder 821df111be6SAlex Elder if (!len) 822df111be6SAlex Elder return 0; 823df111be6SAlex Elder if (len - 1 > U64_MAX - ofs) 824df111be6SAlex Elder return -ERANGE; 825df111be6SAlex Elder 826df111be6SAlex Elder start_seg = ofs >> header->obj_order; 827df111be6SAlex Elder end_seg = (ofs + len - 1) >> header->obj_order; 828df111be6SAlex Elder 8291fec7093SYehuda Sadeh return end_seg - start_seg + 1; 8301fec7093SYehuda Sadeh } 8311fec7093SYehuda Sadeh 832602adf40SYehuda Sadeh /* 833029bcbd8SJosh Durgin * returns the size of an object in the image 834029bcbd8SJosh Durgin */ 835029bcbd8SJosh Durgin static u64 rbd_obj_bytes(struct rbd_image_header *header) 836029bcbd8SJosh Durgin { 837029bcbd8SJosh Durgin return 1 << header->obj_order; 838029bcbd8SJosh Durgin } 839029bcbd8SJosh Durgin 840029bcbd8SJosh Durgin /* 841602adf40SYehuda Sadeh * bio helpers 842602adf40SYehuda Sadeh */ 843602adf40SYehuda Sadeh 844602adf40SYehuda Sadeh static void bio_chain_put(struct bio *chain) 845602adf40SYehuda Sadeh { 846602adf40SYehuda Sadeh struct bio *tmp; 847602adf40SYehuda Sadeh 848602adf40SYehuda Sadeh while (chain) { 849602adf40SYehuda Sadeh tmp = chain; 850602adf40SYehuda Sadeh chain = chain->bi_next; 851602adf40SYehuda Sadeh bio_put(tmp); 852602adf40SYehuda Sadeh } 853602adf40SYehuda Sadeh } 854602adf40SYehuda Sadeh 855602adf40SYehuda Sadeh /* 856602adf40SYehuda Sadeh * zeros a bio chain, starting at specific offset 857602adf40SYehuda Sadeh */ 858602adf40SYehuda Sadeh static void zero_bio_chain(struct bio *chain, int start_ofs) 859602adf40SYehuda Sadeh { 860602adf40SYehuda Sadeh struct bio_vec *bv; 861602adf40SYehuda Sadeh unsigned long flags; 862602adf40SYehuda Sadeh void *buf; 863602adf40SYehuda Sadeh int i; 864602adf40SYehuda Sadeh int pos = 0; 865602adf40SYehuda Sadeh 866602adf40SYehuda Sadeh while (chain) { 867602adf40SYehuda Sadeh bio_for_each_segment(bv, chain, i) { 868602adf40SYehuda Sadeh if (pos + bv->bv_len > start_ofs) { 869602adf40SYehuda Sadeh int remainder = max(start_ofs - pos, 0); 870602adf40SYehuda Sadeh buf = bvec_kmap_irq(bv, &flags); 871602adf40SYehuda Sadeh memset(buf + remainder, 0, 872602adf40SYehuda Sadeh bv->bv_len - remainder); 87385b5aaa6SDan Carpenter bvec_kunmap_irq(buf, &flags); 874602adf40SYehuda Sadeh } 875602adf40SYehuda Sadeh pos += bv->bv_len; 876602adf40SYehuda Sadeh } 877602adf40SYehuda Sadeh 878602adf40SYehuda Sadeh chain = chain->bi_next; 879602adf40SYehuda Sadeh } 880602adf40SYehuda Sadeh } 881602adf40SYehuda Sadeh 882602adf40SYehuda Sadeh /* 883f7760dadSAlex Elder * Clone a portion of a bio, starting at the given byte offset 884f7760dadSAlex Elder * and continuing for the number of bytes indicated. 885602adf40SYehuda Sadeh */ 886f7760dadSAlex Elder static struct bio *bio_clone_range(struct bio *bio_src, 887f7760dadSAlex Elder unsigned int offset, 888f7760dadSAlex Elder unsigned int len, 889f7760dadSAlex Elder gfp_t gfpmask) 890602adf40SYehuda Sadeh { 891f7760dadSAlex Elder struct bio_vec *bv; 892f7760dadSAlex Elder unsigned int resid; 893f7760dadSAlex Elder unsigned short idx; 894f7760dadSAlex Elder unsigned int voff; 895f7760dadSAlex Elder unsigned short end_idx; 896f7760dadSAlex Elder unsigned short vcnt; 897f7760dadSAlex Elder struct bio *bio; 898602adf40SYehuda Sadeh 899f7760dadSAlex Elder /* Handle the easy case for the caller */ 900f7760dadSAlex Elder 901f7760dadSAlex Elder if (!offset && len == bio_src->bi_size) 902f7760dadSAlex Elder return bio_clone(bio_src, gfpmask); 903f7760dadSAlex Elder 904f7760dadSAlex Elder if (WARN_ON_ONCE(!len)) 905f7760dadSAlex Elder return NULL; 906f7760dadSAlex Elder if (WARN_ON_ONCE(len > bio_src->bi_size)) 907f7760dadSAlex Elder return NULL; 908f7760dadSAlex Elder if (WARN_ON_ONCE(offset > bio_src->bi_size - len)) 909f7760dadSAlex Elder return NULL; 910f7760dadSAlex Elder 911f7760dadSAlex Elder /* Find first affected segment... */ 912f7760dadSAlex Elder 913f7760dadSAlex Elder resid = offset; 914f7760dadSAlex Elder __bio_for_each_segment(bv, bio_src, idx, 0) { 915f7760dadSAlex Elder if (resid < bv->bv_len) 916f7760dadSAlex Elder break; 917f7760dadSAlex Elder resid -= bv->bv_len; 918602adf40SYehuda Sadeh } 919f7760dadSAlex Elder voff = resid; 920602adf40SYehuda Sadeh 921f7760dadSAlex Elder /* ...and the last affected segment */ 922542582fcSAlex Elder 923f7760dadSAlex Elder resid += len; 924f7760dadSAlex Elder __bio_for_each_segment(bv, bio_src, end_idx, idx) { 925f7760dadSAlex Elder if (resid <= bv->bv_len) 926f7760dadSAlex Elder break; 927f7760dadSAlex Elder resid -= bv->bv_len; 928f7760dadSAlex Elder } 929f7760dadSAlex Elder vcnt = end_idx - idx + 1; 930602adf40SYehuda Sadeh 931f7760dadSAlex Elder /* Build the clone */ 932f7760dadSAlex Elder 933f7760dadSAlex Elder bio = bio_alloc(gfpmask, (unsigned int) vcnt); 934f7760dadSAlex Elder if (!bio) 935f7760dadSAlex Elder return NULL; /* ENOMEM */ 936f7760dadSAlex Elder 937f7760dadSAlex Elder bio->bi_bdev = bio_src->bi_bdev; 938f7760dadSAlex Elder bio->bi_sector = bio_src->bi_sector + (offset >> SECTOR_SHIFT); 939f7760dadSAlex Elder bio->bi_rw = bio_src->bi_rw; 940f7760dadSAlex Elder bio->bi_flags |= 1 << BIO_CLONED; 941602adf40SYehuda Sadeh 942602adf40SYehuda Sadeh /* 943f7760dadSAlex Elder * Copy over our part of the bio_vec, then update the first 944f7760dadSAlex Elder * and last (or only) entries. 945602adf40SYehuda Sadeh */ 946f7760dadSAlex Elder memcpy(&bio->bi_io_vec[0], &bio_src->bi_io_vec[idx], 947f7760dadSAlex Elder vcnt * sizeof (struct bio_vec)); 948f7760dadSAlex Elder bio->bi_io_vec[0].bv_offset += voff; 949f7760dadSAlex Elder if (vcnt > 1) { 950f7760dadSAlex Elder bio->bi_io_vec[0].bv_len -= voff; 951f7760dadSAlex Elder bio->bi_io_vec[vcnt - 1].bv_len = resid; 952602adf40SYehuda Sadeh } else { 953f7760dadSAlex Elder bio->bi_io_vec[0].bv_len = len; 954602adf40SYehuda Sadeh } 955602adf40SYehuda Sadeh 956f7760dadSAlex Elder bio->bi_vcnt = vcnt; 957f7760dadSAlex Elder bio->bi_size = len; 958f7760dadSAlex Elder bio->bi_idx = 0; 959602adf40SYehuda Sadeh 960f7760dadSAlex Elder return bio; 961602adf40SYehuda Sadeh } 962602adf40SYehuda Sadeh 963f7760dadSAlex Elder /* 964f7760dadSAlex Elder * Clone a portion of a bio chain, starting at the given byte offset 965f7760dadSAlex Elder * into the first bio in the source chain and continuing for the 966f7760dadSAlex Elder * number of bytes indicated. The result is another bio chain of 967f7760dadSAlex Elder * exactly the given length, or a null pointer on error. 968f7760dadSAlex Elder * 969f7760dadSAlex Elder * The bio_src and offset parameters are both in-out. On entry they 970f7760dadSAlex Elder * refer to the first source bio and the offset into that bio where 971f7760dadSAlex Elder * the start of data to be cloned is located. 972f7760dadSAlex Elder * 973f7760dadSAlex Elder * On return, bio_src is updated to refer to the bio in the source 974f7760dadSAlex Elder * chain that contains first un-cloned byte, and *offset will 975f7760dadSAlex Elder * contain the offset of that byte within that bio. 976f7760dadSAlex Elder */ 977f7760dadSAlex Elder static struct bio *bio_chain_clone_range(struct bio **bio_src, 978f7760dadSAlex Elder unsigned int *offset, 979f7760dadSAlex Elder unsigned int len, 980f7760dadSAlex Elder gfp_t gfpmask) 981f7760dadSAlex Elder { 982f7760dadSAlex Elder struct bio *bi = *bio_src; 983f7760dadSAlex Elder unsigned int off = *offset; 984f7760dadSAlex Elder struct bio *chain = NULL; 985f7760dadSAlex Elder struct bio **end; 986602adf40SYehuda Sadeh 987f7760dadSAlex Elder /* Build up a chain of clone bios up to the limit */ 988602adf40SYehuda Sadeh 989f7760dadSAlex Elder if (!bi || off >= bi->bi_size || !len) 990f7760dadSAlex Elder return NULL; /* Nothing to clone */ 991602adf40SYehuda Sadeh 992f7760dadSAlex Elder end = &chain; 993f7760dadSAlex Elder while (len) { 994f7760dadSAlex Elder unsigned int bi_size; 995f7760dadSAlex Elder struct bio *bio; 996f7760dadSAlex Elder 997f5400b7aSAlex Elder if (!bi) { 998f5400b7aSAlex Elder rbd_warn(NULL, "bio_chain exhausted with %u left", len); 999f7760dadSAlex Elder goto out_err; /* EINVAL; ran out of bio's */ 1000f5400b7aSAlex Elder } 1001f7760dadSAlex Elder bi_size = min_t(unsigned int, bi->bi_size - off, len); 1002f7760dadSAlex Elder bio = bio_clone_range(bi, off, bi_size, gfpmask); 1003f7760dadSAlex Elder if (!bio) 1004f7760dadSAlex Elder goto out_err; /* ENOMEM */ 1005f7760dadSAlex Elder 1006f7760dadSAlex Elder *end = bio; 1007f7760dadSAlex Elder end = &bio->bi_next; 1008f7760dadSAlex Elder 1009f7760dadSAlex Elder off += bi_size; 1010f7760dadSAlex Elder if (off == bi->bi_size) { 1011f7760dadSAlex Elder bi = bi->bi_next; 1012f7760dadSAlex Elder off = 0; 1013f7760dadSAlex Elder } 1014f7760dadSAlex Elder len -= bi_size; 1015f7760dadSAlex Elder } 1016f7760dadSAlex Elder *bio_src = bi; 1017f7760dadSAlex Elder *offset = off; 1018f7760dadSAlex Elder 1019f7760dadSAlex Elder return chain; 1020f7760dadSAlex Elder out_err: 1021f7760dadSAlex Elder bio_chain_put(chain); 1022f7760dadSAlex Elder 1023602adf40SYehuda Sadeh return NULL; 1024602adf40SYehuda Sadeh } 1025602adf40SYehuda Sadeh 1026602adf40SYehuda Sadeh /* 1027602adf40SYehuda Sadeh * helpers for osd request op vectors. 1028602adf40SYehuda Sadeh */ 102957cfc106SAlex Elder static struct ceph_osd_req_op *rbd_create_rw_ops(int num_ops, 103057cfc106SAlex Elder int opcode, u32 payload_len) 1031602adf40SYehuda Sadeh { 103257cfc106SAlex Elder struct ceph_osd_req_op *ops; 103357cfc106SAlex Elder 103457cfc106SAlex Elder ops = kzalloc(sizeof (*ops) * (num_ops + 1), GFP_NOIO); 103557cfc106SAlex Elder if (!ops) 103657cfc106SAlex Elder return NULL; 103757cfc106SAlex Elder 103857cfc106SAlex Elder ops[0].op = opcode; 103957cfc106SAlex Elder 1040602adf40SYehuda Sadeh /* 1041602adf40SYehuda Sadeh * op extent offset and length will be set later on 1042602adf40SYehuda Sadeh * in calc_raw_layout() 1043602adf40SYehuda Sadeh */ 104457cfc106SAlex Elder ops[0].payload_len = payload_len; 104557cfc106SAlex Elder 104657cfc106SAlex Elder return ops; 1047602adf40SYehuda Sadeh } 1048602adf40SYehuda Sadeh 1049602adf40SYehuda Sadeh static void rbd_destroy_ops(struct ceph_osd_req_op *ops) 1050602adf40SYehuda Sadeh { 1051602adf40SYehuda Sadeh kfree(ops); 1052602adf40SYehuda Sadeh } 1053602adf40SYehuda Sadeh 10541fec7093SYehuda Sadeh static void rbd_coll_end_req_index(struct request *rq, 10551fec7093SYehuda Sadeh struct rbd_req_coll *coll, 10561fec7093SYehuda Sadeh int index, 10578986cb37SAlex Elder s32 ret, u64 len) 10581fec7093SYehuda Sadeh { 10591fec7093SYehuda Sadeh struct request_queue *q; 10601fec7093SYehuda Sadeh int min, max, i; 10611fec7093SYehuda Sadeh 1062bd919d45SAlex Elder dout("rbd_coll_end_req_index %p index %d ret %d len %llu\n", 10638986cb37SAlex Elder coll, index, (int)ret, (unsigned long long)len); 10641fec7093SYehuda Sadeh 10651fec7093SYehuda Sadeh if (!rq) 10661fec7093SYehuda Sadeh return; 10671fec7093SYehuda Sadeh 10681fec7093SYehuda Sadeh if (!coll) { 10691fec7093SYehuda Sadeh blk_end_request(rq, ret, len); 10701fec7093SYehuda Sadeh return; 10711fec7093SYehuda Sadeh } 10721fec7093SYehuda Sadeh 10731fec7093SYehuda Sadeh q = rq->q; 10741fec7093SYehuda Sadeh 10751fec7093SYehuda Sadeh spin_lock_irq(q->queue_lock); 10761fec7093SYehuda Sadeh coll->status[index].done = 1; 10771fec7093SYehuda Sadeh coll->status[index].rc = ret; 10781fec7093SYehuda Sadeh coll->status[index].bytes = len; 10791fec7093SYehuda Sadeh max = min = coll->num_done; 10801fec7093SYehuda Sadeh while (max < coll->total && coll->status[max].done) 10811fec7093SYehuda Sadeh max++; 10821fec7093SYehuda Sadeh 10831fec7093SYehuda Sadeh for (i = min; i<max; i++) { 10848986cb37SAlex Elder __blk_end_request(rq, (int)coll->status[i].rc, 10851fec7093SYehuda Sadeh coll->status[i].bytes); 10861fec7093SYehuda Sadeh coll->num_done++; 10871fec7093SYehuda Sadeh kref_put(&coll->kref, rbd_coll_release); 10881fec7093SYehuda Sadeh } 10891fec7093SYehuda Sadeh spin_unlock_irq(q->queue_lock); 10901fec7093SYehuda Sadeh } 10911fec7093SYehuda Sadeh 1092725afc97SAlex Elder static void rbd_coll_end_req(struct rbd_request *rbd_req, 10938986cb37SAlex Elder s32 ret, u64 len) 10941fec7093SYehuda Sadeh { 1095725afc97SAlex Elder rbd_coll_end_req_index(rbd_req->rq, 1096725afc97SAlex Elder rbd_req->coll, rbd_req->coll_index, 1097725afc97SAlex Elder ret, len); 10981fec7093SYehuda Sadeh } 10991fec7093SYehuda Sadeh 11000ec8ce87SAlex Elder static void rbd_layout_init(struct ceph_file_layout *layout, u64 pool_id) 11010ec8ce87SAlex Elder { 11020ec8ce87SAlex Elder memset(layout, 0, sizeof (*layout)); 11030ec8ce87SAlex Elder layout->fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); 11040ec8ce87SAlex Elder layout->fl_stripe_count = cpu_to_le32(1); 11050ec8ce87SAlex Elder layout->fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); 11060ec8ce87SAlex Elder rbd_assert(pool_id <= (u64) U32_MAX); 11070ec8ce87SAlex Elder layout->fl_pg_pool = cpu_to_le32((u32) pool_id); 11080ec8ce87SAlex Elder } 11090ec8ce87SAlex Elder 1110602adf40SYehuda Sadeh /* 1111602adf40SYehuda Sadeh * Send ceph osd request 1112602adf40SYehuda Sadeh */ 1113602adf40SYehuda Sadeh static int rbd_do_request(struct request *rq, 11140ce1a794SAlex Elder struct rbd_device *rbd_dev, 1115602adf40SYehuda Sadeh struct ceph_snap_context *snapc, 1116602adf40SYehuda Sadeh u64 snapid, 1117aded07eaSAlex Elder const char *object_name, u64 ofs, u64 len, 1118602adf40SYehuda Sadeh struct bio *bio, 1119602adf40SYehuda Sadeh struct page **pages, 1120602adf40SYehuda Sadeh int num_pages, 1121602adf40SYehuda Sadeh int flags, 1122602adf40SYehuda Sadeh struct ceph_osd_req_op *ops, 11231fec7093SYehuda Sadeh struct rbd_req_coll *coll, 11241fec7093SYehuda Sadeh int coll_index, 11255f29ddd4SAlex Elder void (*rbd_cb)(struct ceph_osd_request *, 11265f29ddd4SAlex Elder struct ceph_msg *), 112759c2be1eSYehuda Sadeh struct ceph_osd_request **linger_req, 112859c2be1eSYehuda Sadeh u64 *ver) 1129602adf40SYehuda Sadeh { 11305f29ddd4SAlex Elder struct ceph_osd_request *osd_req; 1131602adf40SYehuda Sadeh int ret; 1132602adf40SYehuda Sadeh u64 bno; 1133602adf40SYehuda Sadeh struct timespec mtime = CURRENT_TIME; 1134725afc97SAlex Elder struct rbd_request *rbd_req; 11351dbb4399SAlex Elder struct ceph_osd_client *osdc; 1136602adf40SYehuda Sadeh 1137725afc97SAlex Elder rbd_req = kzalloc(sizeof(*rbd_req), GFP_NOIO); 1138cd323ac0SAlex Elder if (!rbd_req) 11391fec7093SYehuda Sadeh return -ENOMEM; 1140602adf40SYehuda Sadeh 11411fec7093SYehuda Sadeh if (coll) { 1142725afc97SAlex Elder rbd_req->coll = coll; 1143725afc97SAlex Elder rbd_req->coll_index = coll_index; 11441fec7093SYehuda Sadeh } 11451fec7093SYehuda Sadeh 1146f7760dadSAlex Elder dout("rbd_do_request object_name=%s ofs=%llu len=%llu coll=%p[%d]\n", 1147f7760dadSAlex Elder object_name, (unsigned long long) ofs, 1148f7760dadSAlex Elder (unsigned long long) len, coll, coll_index); 1149602adf40SYehuda Sadeh 11500ce1a794SAlex Elder osdc = &rbd_dev->rbd_client->client->osdc; 11515f29ddd4SAlex Elder osd_req = ceph_osdc_alloc_request(osdc, flags, snapc, ops, 11521dbb4399SAlex Elder false, GFP_NOIO, pages, bio); 11535f29ddd4SAlex Elder if (!osd_req) { 11544ad12621SSage Weil ret = -ENOMEM; 1155602adf40SYehuda Sadeh goto done_pages; 1156602adf40SYehuda Sadeh } 1157602adf40SYehuda Sadeh 11585f29ddd4SAlex Elder osd_req->r_callback = rbd_cb; 1159602adf40SYehuda Sadeh 1160725afc97SAlex Elder rbd_req->rq = rq; 1161725afc97SAlex Elder rbd_req->bio = bio; 1162725afc97SAlex Elder rbd_req->pages = pages; 1163725afc97SAlex Elder rbd_req->len = len; 1164602adf40SYehuda Sadeh 11655f29ddd4SAlex Elder osd_req->r_priv = rbd_req; 1166602adf40SYehuda Sadeh 11675f29ddd4SAlex Elder strncpy(osd_req->r_oid, object_name, sizeof(osd_req->r_oid)); 11685f29ddd4SAlex Elder osd_req->r_oid_len = strlen(osd_req->r_oid); 1169602adf40SYehuda Sadeh 11700ec8ce87SAlex Elder rbd_layout_init(&osd_req->r_file_layout, rbd_dev->spec->pool_id); 11710ec8ce87SAlex Elder ret = ceph_calc_raw_layout(osdc, &osd_req->r_file_layout, 1172*4d6b250bSAlex Elder ofs, &len, &bno, osd_req, ops); 11736cae3717SSage Weil rbd_assert(ret == 0); 1174602adf40SYehuda Sadeh 1175*4d6b250bSAlex Elder ceph_osdc_build_request(osd_req, ofs, len, ops, snapc, snapid, &mtime); 1176602adf40SYehuda Sadeh 117759c2be1eSYehuda Sadeh if (linger_req) { 11785f29ddd4SAlex Elder ceph_osdc_set_request_linger(osdc, osd_req); 11795f29ddd4SAlex Elder *linger_req = osd_req; 118059c2be1eSYehuda Sadeh } 118159c2be1eSYehuda Sadeh 11825f29ddd4SAlex Elder ret = ceph_osdc_start_request(osdc, osd_req, false); 1183602adf40SYehuda Sadeh if (ret < 0) 1184602adf40SYehuda Sadeh goto done_err; 1185602adf40SYehuda Sadeh 1186602adf40SYehuda Sadeh if (!rbd_cb) { 11875f29ddd4SAlex Elder u64 version; 11885f29ddd4SAlex Elder 11895f29ddd4SAlex Elder ret = ceph_osdc_wait_request(osdc, osd_req); 11905f29ddd4SAlex Elder version = le64_to_cpu(osd_req->r_reassert_version.version); 119159c2be1eSYehuda Sadeh if (ver) 11925f29ddd4SAlex Elder *ver = version; 11935f29ddd4SAlex Elder dout("reassert_ver=%llu\n", (unsigned long long) version); 11945f29ddd4SAlex Elder ceph_osdc_put_request(osd_req); 1195602adf40SYehuda Sadeh } 1196602adf40SYehuda Sadeh return ret; 1197602adf40SYehuda Sadeh 1198602adf40SYehuda Sadeh done_err: 1199725afc97SAlex Elder bio_chain_put(rbd_req->bio); 12005f29ddd4SAlex Elder ceph_osdc_put_request(osd_req); 1201602adf40SYehuda Sadeh done_pages: 1202725afc97SAlex Elder kfree(rbd_req); 1203602adf40SYehuda Sadeh return ret; 1204602adf40SYehuda Sadeh } 1205602adf40SYehuda Sadeh 1206602adf40SYehuda Sadeh /* 1207602adf40SYehuda Sadeh * Ceph osd op callback 1208602adf40SYehuda Sadeh */ 12095f29ddd4SAlex Elder static void rbd_req_cb(struct ceph_osd_request *osd_req, struct ceph_msg *msg) 1210602adf40SYehuda Sadeh { 12115f29ddd4SAlex Elder struct rbd_request *rbd_req = osd_req->r_priv; 1212602adf40SYehuda Sadeh struct ceph_osd_reply_head *replyhead; 1213602adf40SYehuda Sadeh struct ceph_osd_op *op; 12148986cb37SAlex Elder s32 rc; 1215602adf40SYehuda Sadeh u64 bytes; 1216602adf40SYehuda Sadeh int read_op; 1217602adf40SYehuda Sadeh 1218602adf40SYehuda Sadeh /* parse reply */ 1219602adf40SYehuda Sadeh replyhead = msg->front.iov_base; 1220602adf40SYehuda Sadeh WARN_ON(le32_to_cpu(replyhead->num_ops) == 0); 1221602adf40SYehuda Sadeh op = (void *)(replyhead + 1); 12228986cb37SAlex Elder rc = (s32)le32_to_cpu(replyhead->result); 1223602adf40SYehuda Sadeh bytes = le64_to_cpu(op->extent.length); 1224895cfcc8SDan Carpenter read_op = (le16_to_cpu(op->op) == CEPH_OSD_OP_READ); 1225602adf40SYehuda Sadeh 1226bd919d45SAlex Elder dout("rbd_req_cb bytes=%llu readop=%d rc=%d\n", 1227bd919d45SAlex Elder (unsigned long long) bytes, read_op, (int) rc); 1228602adf40SYehuda Sadeh 12298986cb37SAlex Elder if (rc == (s32)-ENOENT && read_op) { 1230725afc97SAlex Elder zero_bio_chain(rbd_req->bio, 0); 1231602adf40SYehuda Sadeh rc = 0; 1232725afc97SAlex Elder } else if (rc == 0 && read_op && bytes < rbd_req->len) { 1233725afc97SAlex Elder zero_bio_chain(rbd_req->bio, bytes); 1234725afc97SAlex Elder bytes = rbd_req->len; 1235602adf40SYehuda Sadeh } 1236602adf40SYehuda Sadeh 1237725afc97SAlex Elder rbd_coll_end_req(rbd_req, rc, bytes); 1238602adf40SYehuda Sadeh 1239725afc97SAlex Elder if (rbd_req->bio) 1240725afc97SAlex Elder bio_chain_put(rbd_req->bio); 1241602adf40SYehuda Sadeh 12425f29ddd4SAlex Elder ceph_osdc_put_request(osd_req); 1243725afc97SAlex Elder kfree(rbd_req); 1244602adf40SYehuda Sadeh } 1245602adf40SYehuda Sadeh 12465f29ddd4SAlex Elder static void rbd_simple_req_cb(struct ceph_osd_request *osd_req, 12475f29ddd4SAlex Elder struct ceph_msg *msg) 124859c2be1eSYehuda Sadeh { 12495f29ddd4SAlex Elder ceph_osdc_put_request(osd_req); 125059c2be1eSYehuda Sadeh } 125159c2be1eSYehuda Sadeh 1252602adf40SYehuda Sadeh /* 1253602adf40SYehuda Sadeh * Do a synchronous ceph osd operation 1254602adf40SYehuda Sadeh */ 12550ce1a794SAlex Elder static int rbd_req_sync_op(struct rbd_device *rbd_dev, 1256602adf40SYehuda Sadeh int flags, 1257913d2fdcSAlex Elder struct ceph_osd_req_op *ops, 1258aded07eaSAlex Elder const char *object_name, 1259f8d4de6eSAlex Elder u64 ofs, u64 inbound_size, 1260f8d4de6eSAlex Elder char *inbound, 126159c2be1eSYehuda Sadeh struct ceph_osd_request **linger_req, 126259c2be1eSYehuda Sadeh u64 *ver) 1263602adf40SYehuda Sadeh { 1264602adf40SYehuda Sadeh int ret; 1265602adf40SYehuda Sadeh struct page **pages; 1266602adf40SYehuda Sadeh int num_pages; 1267913d2fdcSAlex Elder 1268aafb230eSAlex Elder rbd_assert(ops != NULL); 1269602adf40SYehuda Sadeh 1270f8d4de6eSAlex Elder num_pages = calc_pages_for(ofs, inbound_size); 1271602adf40SYehuda Sadeh pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); 1272b8d0638aSDan Carpenter if (IS_ERR(pages)) 1273b8d0638aSDan Carpenter return PTR_ERR(pages); 1274602adf40SYehuda Sadeh 127525704ac9SAlex Elder ret = rbd_do_request(NULL, rbd_dev, NULL, CEPH_NOSNAP, 1276f8d4de6eSAlex Elder object_name, ofs, inbound_size, NULL, 1277602adf40SYehuda Sadeh pages, num_pages, 1278602adf40SYehuda Sadeh flags, 1279602adf40SYehuda Sadeh ops, 12801fec7093SYehuda Sadeh NULL, 0, 128159c2be1eSYehuda Sadeh NULL, 128259c2be1eSYehuda Sadeh linger_req, ver); 1283602adf40SYehuda Sadeh if (ret < 0) 1284913d2fdcSAlex Elder goto done; 1285602adf40SYehuda Sadeh 1286f8d4de6eSAlex Elder if ((flags & CEPH_OSD_FLAG_READ) && inbound) 1287f8d4de6eSAlex Elder ret = ceph_copy_from_page_vector(pages, inbound, ofs, ret); 1288602adf40SYehuda Sadeh 1289602adf40SYehuda Sadeh done: 1290602adf40SYehuda Sadeh ceph_release_page_vector(pages, num_pages); 1291602adf40SYehuda Sadeh return ret; 1292602adf40SYehuda Sadeh } 1293602adf40SYehuda Sadeh 1294602adf40SYehuda Sadeh /* 1295602adf40SYehuda Sadeh * Do an asynchronous ceph osd operation 1296602adf40SYehuda Sadeh */ 1297602adf40SYehuda Sadeh static int rbd_do_op(struct request *rq, 1298602adf40SYehuda Sadeh struct rbd_device *rbd_dev, 1299602adf40SYehuda Sadeh struct ceph_snap_context *snapc, 1300602adf40SYehuda Sadeh u64 ofs, u64 len, 13011fec7093SYehuda Sadeh struct bio *bio, 13021fec7093SYehuda Sadeh struct rbd_req_coll *coll, 13031fec7093SYehuda Sadeh int coll_index) 1304602adf40SYehuda Sadeh { 1305602adf40SYehuda Sadeh char *seg_name; 1306602adf40SYehuda Sadeh u64 seg_ofs; 1307602adf40SYehuda Sadeh u64 seg_len; 1308602adf40SYehuda Sadeh int ret; 1309602adf40SYehuda Sadeh struct ceph_osd_req_op *ops; 1310602adf40SYehuda Sadeh u32 payload_len; 1311ff2e4bb5SAlex Elder int opcode; 1312ff2e4bb5SAlex Elder int flags; 13134634246dSAlex Elder u64 snapid; 1314602adf40SYehuda Sadeh 131565ccfe21SAlex Elder seg_name = rbd_segment_name(rbd_dev, ofs); 1316602adf40SYehuda Sadeh if (!seg_name) 1317602adf40SYehuda Sadeh return -ENOMEM; 131865ccfe21SAlex Elder seg_len = rbd_segment_length(rbd_dev, ofs, len); 131965ccfe21SAlex Elder seg_ofs = rbd_segment_offset(rbd_dev, ofs); 1320602adf40SYehuda Sadeh 1321ff2e4bb5SAlex Elder if (rq_data_dir(rq) == WRITE) { 1322ff2e4bb5SAlex Elder opcode = CEPH_OSD_OP_WRITE; 1323ff2e4bb5SAlex Elder flags = CEPH_OSD_FLAG_WRITE|CEPH_OSD_FLAG_ONDISK; 13244634246dSAlex Elder snapid = CEPH_NOSNAP; 1325ff2e4bb5SAlex Elder payload_len = seg_len; 1326ff2e4bb5SAlex Elder } else { 1327ff2e4bb5SAlex Elder opcode = CEPH_OSD_OP_READ; 1328ff2e4bb5SAlex Elder flags = CEPH_OSD_FLAG_READ; 1329a7b4c65fSAlex Elder rbd_assert(!snapc); 13300d7dbfceSAlex Elder snapid = rbd_dev->spec->snap_id; 1331ff2e4bb5SAlex Elder payload_len = 0; 1332ff2e4bb5SAlex Elder } 1333602adf40SYehuda Sadeh 133457cfc106SAlex Elder ret = -ENOMEM; 133557cfc106SAlex Elder ops = rbd_create_rw_ops(1, opcode, payload_len); 133657cfc106SAlex Elder if (!ops) 1337602adf40SYehuda Sadeh goto done; 1338602adf40SYehuda Sadeh 1339602adf40SYehuda Sadeh /* we've taken care of segment sizes earlier when we 1340602adf40SYehuda Sadeh cloned the bios. We should never have a segment 1341602adf40SYehuda Sadeh truncated at this point */ 1342aafb230eSAlex Elder rbd_assert(seg_len == len); 1343602adf40SYehuda Sadeh 1344602adf40SYehuda Sadeh ret = rbd_do_request(rq, rbd_dev, snapc, snapid, 1345602adf40SYehuda Sadeh seg_name, seg_ofs, seg_len, 1346602adf40SYehuda Sadeh bio, 1347602adf40SYehuda Sadeh NULL, 0, 1348602adf40SYehuda Sadeh flags, 1349602adf40SYehuda Sadeh ops, 13501fec7093SYehuda Sadeh coll, coll_index, 135159c2be1eSYehuda Sadeh rbd_req_cb, 0, NULL); 1352cd323ac0SAlex Elder if (ret < 0) 1353cd323ac0SAlex Elder rbd_coll_end_req_index(rq, coll, coll_index, 1354cd323ac0SAlex Elder (s32)ret, seg_len); 135511f77002SSage Weil rbd_destroy_ops(ops); 1356602adf40SYehuda Sadeh done: 1357602adf40SYehuda Sadeh kfree(seg_name); 1358602adf40SYehuda Sadeh return ret; 1359602adf40SYehuda Sadeh } 1360602adf40SYehuda Sadeh 1361602adf40SYehuda Sadeh /* 1362602adf40SYehuda Sadeh * Request sync osd read 1363602adf40SYehuda Sadeh */ 13640ce1a794SAlex Elder static int rbd_req_sync_read(struct rbd_device *rbd_dev, 1365aded07eaSAlex Elder const char *object_name, 1366602adf40SYehuda Sadeh u64 ofs, u64 len, 136759c2be1eSYehuda Sadeh char *buf, 136859c2be1eSYehuda Sadeh u64 *ver) 1369602adf40SYehuda Sadeh { 1370913d2fdcSAlex Elder struct ceph_osd_req_op *ops; 1371913d2fdcSAlex Elder int ret; 1372913d2fdcSAlex Elder 1373913d2fdcSAlex Elder ops = rbd_create_rw_ops(1, CEPH_OSD_OP_READ, 0); 1374913d2fdcSAlex Elder if (!ops) 1375913d2fdcSAlex Elder return -ENOMEM; 1376913d2fdcSAlex Elder 137725704ac9SAlex Elder ret = rbd_req_sync_op(rbd_dev, CEPH_OSD_FLAG_READ, 1378913d2fdcSAlex Elder ops, object_name, ofs, len, buf, NULL, ver); 1379913d2fdcSAlex Elder rbd_destroy_ops(ops); 1380913d2fdcSAlex Elder 1381913d2fdcSAlex Elder return ret; 1382602adf40SYehuda Sadeh } 1383602adf40SYehuda Sadeh 1384602adf40SYehuda Sadeh /* 138559c2be1eSYehuda Sadeh * Request sync osd watch 138659c2be1eSYehuda Sadeh */ 13870ce1a794SAlex Elder static int rbd_req_sync_notify_ack(struct rbd_device *rbd_dev, 138859c2be1eSYehuda Sadeh u64 ver, 13897f0a24d8SAlex Elder u64 notify_id) 139059c2be1eSYehuda Sadeh { 139159c2be1eSYehuda Sadeh struct ceph_osd_req_op *ops; 139211f77002SSage Weil int ret; 139311f77002SSage Weil 139457cfc106SAlex Elder ops = rbd_create_rw_ops(1, CEPH_OSD_OP_NOTIFY_ACK, 0); 139557cfc106SAlex Elder if (!ops) 139657cfc106SAlex Elder return -ENOMEM; 139759c2be1eSYehuda Sadeh 1398a71b891bSJosh Durgin ops[0].watch.ver = cpu_to_le64(ver); 139959c2be1eSYehuda Sadeh ops[0].watch.cookie = notify_id; 140059c2be1eSYehuda Sadeh ops[0].watch.flag = 0; 140159c2be1eSYehuda Sadeh 14020ce1a794SAlex Elder ret = rbd_do_request(NULL, rbd_dev, NULL, CEPH_NOSNAP, 14037f0a24d8SAlex Elder rbd_dev->header_name, 0, 0, NULL, 1404ad4f232fSAlex Elder NULL, 0, 140559c2be1eSYehuda Sadeh CEPH_OSD_FLAG_READ, 140659c2be1eSYehuda Sadeh ops, 14071fec7093SYehuda Sadeh NULL, 0, 140859c2be1eSYehuda Sadeh rbd_simple_req_cb, 0, NULL); 140959c2be1eSYehuda Sadeh 141059c2be1eSYehuda Sadeh rbd_destroy_ops(ops); 141159c2be1eSYehuda Sadeh return ret; 141259c2be1eSYehuda Sadeh } 141359c2be1eSYehuda Sadeh 141459c2be1eSYehuda Sadeh static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) 141559c2be1eSYehuda Sadeh { 14160ce1a794SAlex Elder struct rbd_device *rbd_dev = (struct rbd_device *)data; 1417a71b891bSJosh Durgin u64 hver; 141813143d2dSSage Weil int rc; 141913143d2dSSage Weil 14200ce1a794SAlex Elder if (!rbd_dev) 142159c2be1eSYehuda Sadeh return; 142259c2be1eSYehuda Sadeh 1423bd919d45SAlex Elder dout("rbd_watch_cb %s notify_id=%llu opcode=%u\n", 1424bd919d45SAlex Elder rbd_dev->header_name, (unsigned long long) notify_id, 1425bd919d45SAlex Elder (unsigned int) opcode); 1426117973fbSAlex Elder rc = rbd_dev_refresh(rbd_dev, &hver); 142713143d2dSSage Weil if (rc) 142806ecc6cbSAlex Elder rbd_warn(rbd_dev, "got notification but failed to " 142906ecc6cbSAlex Elder " update snaps: %d\n", rc); 143059c2be1eSYehuda Sadeh 14317f0a24d8SAlex Elder rbd_req_sync_notify_ack(rbd_dev, hver, notify_id); 143259c2be1eSYehuda Sadeh } 143359c2be1eSYehuda Sadeh 143459c2be1eSYehuda Sadeh /* 143559c2be1eSYehuda Sadeh * Request sync osd watch 143659c2be1eSYehuda Sadeh */ 14370e6f322dSAlex Elder static int rbd_req_sync_watch(struct rbd_device *rbd_dev) 143859c2be1eSYehuda Sadeh { 143959c2be1eSYehuda Sadeh struct ceph_osd_req_op *ops; 14400ce1a794SAlex Elder struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 144157cfc106SAlex Elder int ret; 144259c2be1eSYehuda Sadeh 144357cfc106SAlex Elder ops = rbd_create_rw_ops(1, CEPH_OSD_OP_WATCH, 0); 144457cfc106SAlex Elder if (!ops) 144557cfc106SAlex Elder return -ENOMEM; 144659c2be1eSYehuda Sadeh 144759c2be1eSYehuda Sadeh ret = ceph_osdc_create_event(osdc, rbd_watch_cb, 0, 14480ce1a794SAlex Elder (void *)rbd_dev, &rbd_dev->watch_event); 144959c2be1eSYehuda Sadeh if (ret < 0) 145059c2be1eSYehuda Sadeh goto fail; 145159c2be1eSYehuda Sadeh 14520e6f322dSAlex Elder ops[0].watch.ver = cpu_to_le64(rbd_dev->header.obj_version); 14530ce1a794SAlex Elder ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); 145459c2be1eSYehuda Sadeh ops[0].watch.flag = 1; 145559c2be1eSYehuda Sadeh 145625704ac9SAlex Elder ret = rbd_req_sync_op(rbd_dev, 145759c2be1eSYehuda Sadeh CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, 145859c2be1eSYehuda Sadeh ops, 14590e6f322dSAlex Elder rbd_dev->header_name, 14600e6f322dSAlex Elder 0, 0, NULL, 14610ce1a794SAlex Elder &rbd_dev->watch_request, NULL); 146259c2be1eSYehuda Sadeh 146359c2be1eSYehuda Sadeh if (ret < 0) 146459c2be1eSYehuda Sadeh goto fail_event; 146559c2be1eSYehuda Sadeh 146659c2be1eSYehuda Sadeh rbd_destroy_ops(ops); 146759c2be1eSYehuda Sadeh return 0; 146859c2be1eSYehuda Sadeh 146959c2be1eSYehuda Sadeh fail_event: 14700ce1a794SAlex Elder ceph_osdc_cancel_event(rbd_dev->watch_event); 14710ce1a794SAlex Elder rbd_dev->watch_event = NULL; 147259c2be1eSYehuda Sadeh fail: 147359c2be1eSYehuda Sadeh rbd_destroy_ops(ops); 147459c2be1eSYehuda Sadeh return ret; 147559c2be1eSYehuda Sadeh } 147659c2be1eSYehuda Sadeh 147779e3057cSYehuda Sadeh /* 147879e3057cSYehuda Sadeh * Request sync osd unwatch 147979e3057cSYehuda Sadeh */ 1480070c633fSAlex Elder static int rbd_req_sync_unwatch(struct rbd_device *rbd_dev) 148179e3057cSYehuda Sadeh { 148279e3057cSYehuda Sadeh struct ceph_osd_req_op *ops; 148357cfc106SAlex Elder int ret; 148479e3057cSYehuda Sadeh 148557cfc106SAlex Elder ops = rbd_create_rw_ops(1, CEPH_OSD_OP_WATCH, 0); 148657cfc106SAlex Elder if (!ops) 148757cfc106SAlex Elder return -ENOMEM; 148879e3057cSYehuda Sadeh 148979e3057cSYehuda Sadeh ops[0].watch.ver = 0; 14900ce1a794SAlex Elder ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); 149179e3057cSYehuda Sadeh ops[0].watch.flag = 0; 149279e3057cSYehuda Sadeh 149325704ac9SAlex Elder ret = rbd_req_sync_op(rbd_dev, 149479e3057cSYehuda Sadeh CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, 149579e3057cSYehuda Sadeh ops, 1496070c633fSAlex Elder rbd_dev->header_name, 1497070c633fSAlex Elder 0, 0, NULL, NULL, NULL); 1498070c633fSAlex Elder 149979e3057cSYehuda Sadeh 150079e3057cSYehuda Sadeh rbd_destroy_ops(ops); 15010ce1a794SAlex Elder ceph_osdc_cancel_event(rbd_dev->watch_event); 15020ce1a794SAlex Elder rbd_dev->watch_event = NULL; 150379e3057cSYehuda Sadeh return ret; 150479e3057cSYehuda Sadeh } 150579e3057cSYehuda Sadeh 150659c2be1eSYehuda Sadeh /* 15073cb4a687SAlex Elder * Synchronous osd object method call 1508602adf40SYehuda Sadeh */ 15090ce1a794SAlex Elder static int rbd_req_sync_exec(struct rbd_device *rbd_dev, 1510aded07eaSAlex Elder const char *object_name, 1511aded07eaSAlex Elder const char *class_name, 1512aded07eaSAlex Elder const char *method_name, 15133cb4a687SAlex Elder const char *outbound, 15143cb4a687SAlex Elder size_t outbound_size, 1515f8d4de6eSAlex Elder char *inbound, 1516f8d4de6eSAlex Elder size_t inbound_size, 151759c2be1eSYehuda Sadeh u64 *ver) 1518602adf40SYehuda Sadeh { 1519602adf40SYehuda Sadeh struct ceph_osd_req_op *ops; 1520aded07eaSAlex Elder int class_name_len = strlen(class_name); 1521aded07eaSAlex Elder int method_name_len = strlen(method_name); 15223cb4a687SAlex Elder int payload_size; 152357cfc106SAlex Elder int ret; 152457cfc106SAlex Elder 15253cb4a687SAlex Elder /* 15263cb4a687SAlex Elder * Any input parameters required by the method we're calling 15273cb4a687SAlex Elder * will be sent along with the class and method names as 15283cb4a687SAlex Elder * part of the message payload. That data and its size are 15293cb4a687SAlex Elder * supplied via the indata and indata_len fields (named from 15303cb4a687SAlex Elder * the perspective of the server side) in the OSD request 15313cb4a687SAlex Elder * operation. 15323cb4a687SAlex Elder */ 15333cb4a687SAlex Elder payload_size = class_name_len + method_name_len + outbound_size; 15343cb4a687SAlex Elder ops = rbd_create_rw_ops(1, CEPH_OSD_OP_CALL, payload_size); 153557cfc106SAlex Elder if (!ops) 153657cfc106SAlex Elder return -ENOMEM; 1537602adf40SYehuda Sadeh 1538aded07eaSAlex Elder ops[0].cls.class_name = class_name; 1539aded07eaSAlex Elder ops[0].cls.class_len = (__u8) class_name_len; 1540aded07eaSAlex Elder ops[0].cls.method_name = method_name; 1541aded07eaSAlex Elder ops[0].cls.method_len = (__u8) method_name_len; 1542602adf40SYehuda Sadeh ops[0].cls.argc = 0; 15433cb4a687SAlex Elder ops[0].cls.indata = outbound; 15443cb4a687SAlex Elder ops[0].cls.indata_len = outbound_size; 1545602adf40SYehuda Sadeh 154625704ac9SAlex Elder ret = rbd_req_sync_op(rbd_dev, CEPH_OSD_FLAG_READ, ops, 1547f8d4de6eSAlex Elder object_name, 0, inbound_size, inbound, 1548f8d4de6eSAlex Elder NULL, ver); 1549602adf40SYehuda Sadeh 1550602adf40SYehuda Sadeh rbd_destroy_ops(ops); 1551602adf40SYehuda Sadeh 1552602adf40SYehuda Sadeh dout("cls_exec returned %d\n", ret); 1553602adf40SYehuda Sadeh return ret; 1554602adf40SYehuda Sadeh } 1555602adf40SYehuda Sadeh 15561fec7093SYehuda Sadeh static struct rbd_req_coll *rbd_alloc_coll(int num_reqs) 15571fec7093SYehuda Sadeh { 15581fec7093SYehuda Sadeh struct rbd_req_coll *coll = 15591fec7093SYehuda Sadeh kzalloc(sizeof(struct rbd_req_coll) + 15601fec7093SYehuda Sadeh sizeof(struct rbd_req_status) * num_reqs, 15611fec7093SYehuda Sadeh GFP_ATOMIC); 15621fec7093SYehuda Sadeh 15631fec7093SYehuda Sadeh if (!coll) 15641fec7093SYehuda Sadeh return NULL; 15651fec7093SYehuda Sadeh coll->total = num_reqs; 15661fec7093SYehuda Sadeh kref_init(&coll->kref); 15671fec7093SYehuda Sadeh return coll; 15681fec7093SYehuda Sadeh } 15691fec7093SYehuda Sadeh 15708295cda7SAlex Elder static int rbd_dev_do_request(struct request *rq, 15718295cda7SAlex Elder struct rbd_device *rbd_dev, 15728295cda7SAlex Elder struct ceph_snap_context *snapc, 15738295cda7SAlex Elder u64 ofs, unsigned int size, 15748295cda7SAlex Elder struct bio *bio_chain) 15758295cda7SAlex Elder { 15768295cda7SAlex Elder int num_segs; 15778295cda7SAlex Elder struct rbd_req_coll *coll; 15788295cda7SAlex Elder unsigned int bio_offset; 15798295cda7SAlex Elder int cur_seg = 0; 15808295cda7SAlex Elder 15818295cda7SAlex Elder dout("%s 0x%x bytes at 0x%llx\n", 15828295cda7SAlex Elder rq_data_dir(rq) == WRITE ? "write" : "read", 15838295cda7SAlex Elder size, (unsigned long long) blk_rq_pos(rq) * SECTOR_SIZE); 15848295cda7SAlex Elder 15858295cda7SAlex Elder num_segs = rbd_get_num_segments(&rbd_dev->header, ofs, size); 15868295cda7SAlex Elder if (num_segs <= 0) 15878295cda7SAlex Elder return num_segs; 15888295cda7SAlex Elder 15898295cda7SAlex Elder coll = rbd_alloc_coll(num_segs); 15908295cda7SAlex Elder if (!coll) 15918295cda7SAlex Elder return -ENOMEM; 15928295cda7SAlex Elder 15938295cda7SAlex Elder bio_offset = 0; 15948295cda7SAlex Elder do { 15958295cda7SAlex Elder u64 limit = rbd_segment_length(rbd_dev, ofs, size); 15968295cda7SAlex Elder unsigned int clone_size; 15978295cda7SAlex Elder struct bio *bio_clone; 15988295cda7SAlex Elder 15998295cda7SAlex Elder BUG_ON(limit > (u64)UINT_MAX); 16008295cda7SAlex Elder clone_size = (unsigned int)limit; 16018295cda7SAlex Elder dout("bio_chain->bi_vcnt=%hu\n", bio_chain->bi_vcnt); 16028295cda7SAlex Elder 16038295cda7SAlex Elder kref_get(&coll->kref); 16048295cda7SAlex Elder 16058295cda7SAlex Elder /* Pass a cloned bio chain via an osd request */ 16068295cda7SAlex Elder 16078295cda7SAlex Elder bio_clone = bio_chain_clone_range(&bio_chain, 16088295cda7SAlex Elder &bio_offset, clone_size, 16098295cda7SAlex Elder GFP_ATOMIC); 16108295cda7SAlex Elder if (bio_clone) 16118295cda7SAlex Elder (void)rbd_do_op(rq, rbd_dev, snapc, 16128295cda7SAlex Elder ofs, clone_size, 16138295cda7SAlex Elder bio_clone, coll, cur_seg); 16148295cda7SAlex Elder else 16158295cda7SAlex Elder rbd_coll_end_req_index(rq, coll, cur_seg, 16168295cda7SAlex Elder (s32)-ENOMEM, 16178295cda7SAlex Elder clone_size); 16188295cda7SAlex Elder size -= clone_size; 16198295cda7SAlex Elder ofs += clone_size; 16208295cda7SAlex Elder 16218295cda7SAlex Elder cur_seg++; 16228295cda7SAlex Elder } while (size > 0); 16238295cda7SAlex Elder kref_put(&coll->kref, rbd_coll_release); 16248295cda7SAlex Elder 16258295cda7SAlex Elder return 0; 16268295cda7SAlex Elder } 16278295cda7SAlex Elder 1628602adf40SYehuda Sadeh /* 1629602adf40SYehuda Sadeh * block device queue callback 1630602adf40SYehuda Sadeh */ 1631602adf40SYehuda Sadeh static void rbd_rq_fn(struct request_queue *q) 1632602adf40SYehuda Sadeh { 1633602adf40SYehuda Sadeh struct rbd_device *rbd_dev = q->queuedata; 1634b395e8b5SAlex Elder bool read_only = rbd_dev->mapping.read_only; 1635602adf40SYehuda Sadeh struct request *rq; 1636602adf40SYehuda Sadeh 163700f1f36fSAlex Elder while ((rq = blk_fetch_request(q))) { 1638b395e8b5SAlex Elder struct ceph_snap_context *snapc = NULL; 1639b395e8b5SAlex Elder unsigned int size = 0; 16408295cda7SAlex Elder int result; 1641602adf40SYehuda Sadeh 1642602adf40SYehuda Sadeh dout("fetched request\n"); 1643602adf40SYehuda Sadeh 1644b395e8b5SAlex Elder /* Filter out block requests we don't understand */ 1645b395e8b5SAlex Elder 1646602adf40SYehuda Sadeh if ((rq->cmd_type != REQ_TYPE_FS)) { 1647602adf40SYehuda Sadeh __blk_end_request_all(rq, 0); 164800f1f36fSAlex Elder continue; 1649602adf40SYehuda Sadeh } 1650602adf40SYehuda Sadeh spin_unlock_irq(q->queue_lock); 1651602adf40SYehuda Sadeh 1652a7b4c65fSAlex Elder /* Write requests need a reference to the snapshot context */ 1653e88a36ecSJosh Durgin 1654a7b4c65fSAlex Elder if (rq_data_dir(rq) == WRITE) { 1655b395e8b5SAlex Elder result = -EROFS; 1656a7b4c65fSAlex Elder if (read_only) /* Can't write to a read-only device */ 1657b395e8b5SAlex Elder goto out_end_request; 1658b395e8b5SAlex Elder 1659a7b4c65fSAlex Elder /* 1660a7b4c65fSAlex Elder * Note that each osd request will take its 1661a7b4c65fSAlex Elder * own reference to the snapshot context 1662a7b4c65fSAlex Elder * supplied. The reference we take here 1663a7b4c65fSAlex Elder * just guarantees the one we provide stays 1664a7b4c65fSAlex Elder * valid. 1665a7b4c65fSAlex Elder */ 1666b395e8b5SAlex Elder down_read(&rbd_dev->header_rwsem); 1667b395e8b5SAlex Elder snapc = ceph_get_snap_context(rbd_dev->header.snapc); 1668d1d25646SJosh Durgin up_read(&rbd_dev->header_rwsem); 1669a7b4c65fSAlex Elder rbd_assert(snapc != NULL); 1670a7b4c65fSAlex Elder } else if (!atomic_read(&rbd_dev->exists)) { 1671b395e8b5SAlex Elder rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); 1672e88a36ecSJosh Durgin dout("request for non-existent snapshot"); 1673b395e8b5SAlex Elder result = -ENXIO; 1674b395e8b5SAlex Elder goto out_end_request; 1675e88a36ecSJosh Durgin } 1676d1d25646SJosh Durgin 1677f7760dadSAlex Elder size = blk_rq_bytes(rq); 1678b395e8b5SAlex Elder result = rbd_dev_do_request(rq, rbd_dev, snapc, 1679b395e8b5SAlex Elder blk_rq_pos(rq) * SECTOR_SIZE, 1680b395e8b5SAlex Elder size, rq->bio); 1681b395e8b5SAlex Elder out_end_request: 1682a7b4c65fSAlex Elder if (snapc) 1683df111be6SAlex Elder ceph_put_snap_context(snapc); 16841fec7093SYehuda Sadeh spin_lock_irq(q->queue_lock); 16858295cda7SAlex Elder if (!size || result < 0) 16868295cda7SAlex Elder __blk_end_request_all(rq, result); 1687602adf40SYehuda Sadeh } 1688602adf40SYehuda Sadeh } 1689602adf40SYehuda Sadeh 1690602adf40SYehuda Sadeh /* 1691602adf40SYehuda Sadeh * a queue callback. Makes sure that we don't create a bio that spans across 1692602adf40SYehuda Sadeh * multiple osd objects. One exception would be with a single page bios, 1693f7760dadSAlex Elder * which we handle later at bio_chain_clone_range() 1694602adf40SYehuda Sadeh */ 1695602adf40SYehuda Sadeh static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, 1696602adf40SYehuda Sadeh struct bio_vec *bvec) 1697602adf40SYehuda Sadeh { 1698602adf40SYehuda Sadeh struct rbd_device *rbd_dev = q->queuedata; 1699e5cfeed2SAlex Elder sector_t sector_offset; 1700e5cfeed2SAlex Elder sector_t sectors_per_obj; 1701e5cfeed2SAlex Elder sector_t obj_sector_offset; 1702e5cfeed2SAlex Elder int ret; 1703602adf40SYehuda Sadeh 1704e5cfeed2SAlex Elder /* 1705e5cfeed2SAlex Elder * Find how far into its rbd object the partition-relative 1706e5cfeed2SAlex Elder * bio start sector is to offset relative to the enclosing 1707e5cfeed2SAlex Elder * device. 1708e5cfeed2SAlex Elder */ 1709e5cfeed2SAlex Elder sector_offset = get_start_sect(bmd->bi_bdev) + bmd->bi_sector; 1710e5cfeed2SAlex Elder sectors_per_obj = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); 1711e5cfeed2SAlex Elder obj_sector_offset = sector_offset & (sectors_per_obj - 1); 1712593a9e7bSAlex Elder 1713e5cfeed2SAlex Elder /* 1714e5cfeed2SAlex Elder * Compute the number of bytes from that offset to the end 1715e5cfeed2SAlex Elder * of the object. Account for what's already used by the bio. 1716e5cfeed2SAlex Elder */ 1717e5cfeed2SAlex Elder ret = (int) (sectors_per_obj - obj_sector_offset) << SECTOR_SHIFT; 1718e5cfeed2SAlex Elder if (ret > bmd->bi_size) 1719e5cfeed2SAlex Elder ret -= bmd->bi_size; 1720e5cfeed2SAlex Elder else 1721e5cfeed2SAlex Elder ret = 0; 1722e5cfeed2SAlex Elder 1723e5cfeed2SAlex Elder /* 1724e5cfeed2SAlex Elder * Don't send back more than was asked for. And if the bio 1725e5cfeed2SAlex Elder * was empty, let the whole thing through because: "Note 1726e5cfeed2SAlex Elder * that a block device *must* allow a single page to be 1727e5cfeed2SAlex Elder * added to an empty bio." 1728e5cfeed2SAlex Elder */ 1729e5cfeed2SAlex Elder rbd_assert(bvec->bv_len <= PAGE_SIZE); 1730e5cfeed2SAlex Elder if (ret > (int) bvec->bv_len || !bmd->bi_size) 1731e5cfeed2SAlex Elder ret = (int) bvec->bv_len; 1732e5cfeed2SAlex Elder 1733e5cfeed2SAlex Elder return ret; 1734602adf40SYehuda Sadeh } 1735602adf40SYehuda Sadeh 1736602adf40SYehuda Sadeh static void rbd_free_disk(struct rbd_device *rbd_dev) 1737602adf40SYehuda Sadeh { 1738602adf40SYehuda Sadeh struct gendisk *disk = rbd_dev->disk; 1739602adf40SYehuda Sadeh 1740602adf40SYehuda Sadeh if (!disk) 1741602adf40SYehuda Sadeh return; 1742602adf40SYehuda Sadeh 1743602adf40SYehuda Sadeh if (disk->flags & GENHD_FL_UP) 1744602adf40SYehuda Sadeh del_gendisk(disk); 1745602adf40SYehuda Sadeh if (disk->queue) 1746602adf40SYehuda Sadeh blk_cleanup_queue(disk->queue); 1747602adf40SYehuda Sadeh put_disk(disk); 1748602adf40SYehuda Sadeh } 1749602adf40SYehuda Sadeh 1750602adf40SYehuda Sadeh /* 17514156d998SAlex Elder * Read the complete header for the given rbd device. 17524156d998SAlex Elder * 17534156d998SAlex Elder * Returns a pointer to a dynamically-allocated buffer containing 17544156d998SAlex Elder * the complete and validated header. Caller can pass the address 17554156d998SAlex Elder * of a variable that will be filled in with the version of the 17564156d998SAlex Elder * header object at the time it was read. 17574156d998SAlex Elder * 17584156d998SAlex Elder * Returns a pointer-coded errno if a failure occurs. 17594156d998SAlex Elder */ 17604156d998SAlex Elder static struct rbd_image_header_ondisk * 17614156d998SAlex Elder rbd_dev_v1_header_read(struct rbd_device *rbd_dev, u64 *version) 17624156d998SAlex Elder { 17634156d998SAlex Elder struct rbd_image_header_ondisk *ondisk = NULL; 17644156d998SAlex Elder u32 snap_count = 0; 17654156d998SAlex Elder u64 names_size = 0; 17664156d998SAlex Elder u32 want_count; 17674156d998SAlex Elder int ret; 17684156d998SAlex Elder 17694156d998SAlex Elder /* 17704156d998SAlex Elder * The complete header will include an array of its 64-bit 17714156d998SAlex Elder * snapshot ids, followed by the names of those snapshots as 17724156d998SAlex Elder * a contiguous block of NUL-terminated strings. Note that 17734156d998SAlex Elder * the number of snapshots could change by the time we read 17744156d998SAlex Elder * it in, in which case we re-read it. 17754156d998SAlex Elder */ 17764156d998SAlex Elder do { 17774156d998SAlex Elder size_t size; 17784156d998SAlex Elder 17794156d998SAlex Elder kfree(ondisk); 17804156d998SAlex Elder 17814156d998SAlex Elder size = sizeof (*ondisk); 17824156d998SAlex Elder size += snap_count * sizeof (struct rbd_image_snap_ondisk); 17834156d998SAlex Elder size += names_size; 17844156d998SAlex Elder ondisk = kmalloc(size, GFP_KERNEL); 17854156d998SAlex Elder if (!ondisk) 17864156d998SAlex Elder return ERR_PTR(-ENOMEM); 17874156d998SAlex Elder 17884775618dSAlex Elder ret = rbd_req_sync_read(rbd_dev, rbd_dev->header_name, 17894156d998SAlex Elder 0, size, 17904156d998SAlex Elder (char *) ondisk, version); 17914156d998SAlex Elder 17924156d998SAlex Elder if (ret < 0) 17934156d998SAlex Elder goto out_err; 17944156d998SAlex Elder if (WARN_ON((size_t) ret < size)) { 17954156d998SAlex Elder ret = -ENXIO; 179606ecc6cbSAlex Elder rbd_warn(rbd_dev, "short header read (want %zd got %d)", 179706ecc6cbSAlex Elder size, ret); 17984156d998SAlex Elder goto out_err; 17994156d998SAlex Elder } 18004156d998SAlex Elder if (!rbd_dev_ondisk_valid(ondisk)) { 18014156d998SAlex Elder ret = -ENXIO; 180206ecc6cbSAlex Elder rbd_warn(rbd_dev, "invalid header"); 18034156d998SAlex Elder goto out_err; 18044156d998SAlex Elder } 18054156d998SAlex Elder 18064156d998SAlex Elder names_size = le64_to_cpu(ondisk->snap_names_len); 18074156d998SAlex Elder want_count = snap_count; 18084156d998SAlex Elder snap_count = le32_to_cpu(ondisk->snap_count); 18094156d998SAlex Elder } while (snap_count != want_count); 18104156d998SAlex Elder 18114156d998SAlex Elder return ondisk; 18124156d998SAlex Elder 18134156d998SAlex Elder out_err: 18144156d998SAlex Elder kfree(ondisk); 18154156d998SAlex Elder 18164156d998SAlex Elder return ERR_PTR(ret); 18174156d998SAlex Elder } 18184156d998SAlex Elder 18194156d998SAlex Elder /* 1820602adf40SYehuda Sadeh * reload the ondisk the header 1821602adf40SYehuda Sadeh */ 1822602adf40SYehuda Sadeh static int rbd_read_header(struct rbd_device *rbd_dev, 1823602adf40SYehuda Sadeh struct rbd_image_header *header) 1824602adf40SYehuda Sadeh { 18254156d998SAlex Elder struct rbd_image_header_ondisk *ondisk; 18264156d998SAlex Elder u64 ver = 0; 18274156d998SAlex Elder int ret; 1828602adf40SYehuda Sadeh 18294156d998SAlex Elder ondisk = rbd_dev_v1_header_read(rbd_dev, &ver); 18304156d998SAlex Elder if (IS_ERR(ondisk)) 18314156d998SAlex Elder return PTR_ERR(ondisk); 18324156d998SAlex Elder ret = rbd_header_from_disk(header, ondisk); 18334156d998SAlex Elder if (ret >= 0) 183459c2be1eSYehuda Sadeh header->obj_version = ver; 18354156d998SAlex Elder kfree(ondisk); 1836602adf40SYehuda Sadeh 18374156d998SAlex Elder return ret; 1838602adf40SYehuda Sadeh } 1839602adf40SYehuda Sadeh 184041f38c2bSAlex Elder static void rbd_remove_all_snaps(struct rbd_device *rbd_dev) 1841dfc5606dSYehuda Sadeh { 1842dfc5606dSYehuda Sadeh struct rbd_snap *snap; 1843a0593290SAlex Elder struct rbd_snap *next; 1844dfc5606dSYehuda Sadeh 1845a0593290SAlex Elder list_for_each_entry_safe(snap, next, &rbd_dev->snaps, node) 184641f38c2bSAlex Elder rbd_remove_snap_dev(snap); 1847dfc5606dSYehuda Sadeh } 1848dfc5606dSYehuda Sadeh 18499478554aSAlex Elder static void rbd_update_mapping_size(struct rbd_device *rbd_dev) 18509478554aSAlex Elder { 18519478554aSAlex Elder sector_t size; 18529478554aSAlex Elder 18530d7dbfceSAlex Elder if (rbd_dev->spec->snap_id != CEPH_NOSNAP) 18549478554aSAlex Elder return; 18559478554aSAlex Elder 18569478554aSAlex Elder size = (sector_t) rbd_dev->header.image_size / SECTOR_SIZE; 18579478554aSAlex Elder dout("setting size to %llu sectors", (unsigned long long) size); 18589478554aSAlex Elder rbd_dev->mapping.size = (u64) size; 18599478554aSAlex Elder set_capacity(rbd_dev->disk, size); 18609478554aSAlex Elder } 18619478554aSAlex Elder 1862602adf40SYehuda Sadeh /* 1863602adf40SYehuda Sadeh * only read the first part of the ondisk header, without the snaps info 1864602adf40SYehuda Sadeh */ 1865117973fbSAlex Elder static int rbd_dev_v1_refresh(struct rbd_device *rbd_dev, u64 *hver) 1866602adf40SYehuda Sadeh { 1867602adf40SYehuda Sadeh int ret; 1868602adf40SYehuda Sadeh struct rbd_image_header h; 1869602adf40SYehuda Sadeh 1870602adf40SYehuda Sadeh ret = rbd_read_header(rbd_dev, &h); 1871602adf40SYehuda Sadeh if (ret < 0) 1872602adf40SYehuda Sadeh return ret; 1873602adf40SYehuda Sadeh 1874a51aa0c0SJosh Durgin down_write(&rbd_dev->header_rwsem); 1875a51aa0c0SJosh Durgin 18769478554aSAlex Elder /* Update image size, and check for resize of mapped image */ 18779478554aSAlex Elder rbd_dev->header.image_size = h.image_size; 18789478554aSAlex Elder rbd_update_mapping_size(rbd_dev); 18799db4b3e3SSage Weil 1880849b4260SAlex Elder /* rbd_dev->header.object_prefix shouldn't change */ 1881602adf40SYehuda Sadeh kfree(rbd_dev->header.snap_sizes); 1882849b4260SAlex Elder kfree(rbd_dev->header.snap_names); 1883d1d25646SJosh Durgin /* osd requests may still refer to snapc */ 1884d1d25646SJosh Durgin ceph_put_snap_context(rbd_dev->header.snapc); 1885602adf40SYehuda Sadeh 1886b813623aSAlex Elder if (hver) 1887b813623aSAlex Elder *hver = h.obj_version; 1888a71b891bSJosh Durgin rbd_dev->header.obj_version = h.obj_version; 188993a24e08SJosh Durgin rbd_dev->header.image_size = h.image_size; 1890602adf40SYehuda Sadeh rbd_dev->header.snapc = h.snapc; 1891602adf40SYehuda Sadeh rbd_dev->header.snap_names = h.snap_names; 1892602adf40SYehuda Sadeh rbd_dev->header.snap_sizes = h.snap_sizes; 1893849b4260SAlex Elder /* Free the extra copy of the object prefix */ 1894849b4260SAlex Elder WARN_ON(strcmp(rbd_dev->header.object_prefix, h.object_prefix)); 1895849b4260SAlex Elder kfree(h.object_prefix); 1896849b4260SAlex Elder 1897304f6808SAlex Elder ret = rbd_dev_snaps_update(rbd_dev); 1898304f6808SAlex Elder if (!ret) 1899304f6808SAlex Elder ret = rbd_dev_snaps_register(rbd_dev); 1900dfc5606dSYehuda Sadeh 1901c666601aSJosh Durgin up_write(&rbd_dev->header_rwsem); 1902602adf40SYehuda Sadeh 1903dfc5606dSYehuda Sadeh return ret; 1904602adf40SYehuda Sadeh } 1905602adf40SYehuda Sadeh 1906117973fbSAlex Elder static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver) 19071fe5e993SAlex Elder { 19081fe5e993SAlex Elder int ret; 19091fe5e993SAlex Elder 1910117973fbSAlex Elder rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); 19111fe5e993SAlex Elder mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); 1912117973fbSAlex Elder if (rbd_dev->image_format == 1) 1913117973fbSAlex Elder ret = rbd_dev_v1_refresh(rbd_dev, hver); 1914117973fbSAlex Elder else 1915117973fbSAlex Elder ret = rbd_dev_v2_refresh(rbd_dev, hver); 19161fe5e993SAlex Elder mutex_unlock(&ctl_mutex); 19171fe5e993SAlex Elder 19181fe5e993SAlex Elder return ret; 19191fe5e993SAlex Elder } 19201fe5e993SAlex Elder 1921602adf40SYehuda Sadeh static int rbd_init_disk(struct rbd_device *rbd_dev) 1922602adf40SYehuda Sadeh { 1923602adf40SYehuda Sadeh struct gendisk *disk; 1924602adf40SYehuda Sadeh struct request_queue *q; 1925593a9e7bSAlex Elder u64 segment_size; 1926602adf40SYehuda Sadeh 1927602adf40SYehuda Sadeh /* create gendisk info */ 1928602adf40SYehuda Sadeh disk = alloc_disk(RBD_MINORS_PER_MAJOR); 1929602adf40SYehuda Sadeh if (!disk) 19301fcdb8aaSAlex Elder return -ENOMEM; 1931602adf40SYehuda Sadeh 1932f0f8cef5SAlex Elder snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", 1933de71a297SAlex Elder rbd_dev->dev_id); 1934602adf40SYehuda Sadeh disk->major = rbd_dev->major; 1935602adf40SYehuda Sadeh disk->first_minor = 0; 1936602adf40SYehuda Sadeh disk->fops = &rbd_bd_ops; 1937602adf40SYehuda Sadeh disk->private_data = rbd_dev; 1938602adf40SYehuda Sadeh 1939602adf40SYehuda Sadeh /* init rq */ 1940602adf40SYehuda Sadeh q = blk_init_queue(rbd_rq_fn, &rbd_dev->lock); 1941602adf40SYehuda Sadeh if (!q) 1942602adf40SYehuda Sadeh goto out_disk; 1943029bcbd8SJosh Durgin 1944593a9e7bSAlex Elder /* We use the default size, but let's be explicit about it. */ 1945593a9e7bSAlex Elder blk_queue_physical_block_size(q, SECTOR_SIZE); 1946593a9e7bSAlex Elder 1947029bcbd8SJosh Durgin /* set io sizes to object size */ 1948593a9e7bSAlex Elder segment_size = rbd_obj_bytes(&rbd_dev->header); 1949593a9e7bSAlex Elder blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); 1950593a9e7bSAlex Elder blk_queue_max_segment_size(q, segment_size); 1951593a9e7bSAlex Elder blk_queue_io_min(q, segment_size); 1952593a9e7bSAlex Elder blk_queue_io_opt(q, segment_size); 1953029bcbd8SJosh Durgin 1954602adf40SYehuda Sadeh blk_queue_merge_bvec(q, rbd_merge_bvec); 1955602adf40SYehuda Sadeh disk->queue = q; 1956602adf40SYehuda Sadeh 1957602adf40SYehuda Sadeh q->queuedata = rbd_dev; 1958602adf40SYehuda Sadeh 1959602adf40SYehuda Sadeh rbd_dev->disk = disk; 1960602adf40SYehuda Sadeh 196112f02944SAlex Elder set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); 196212f02944SAlex Elder 1963602adf40SYehuda Sadeh return 0; 1964602adf40SYehuda Sadeh out_disk: 1965602adf40SYehuda Sadeh put_disk(disk); 19661fcdb8aaSAlex Elder 19671fcdb8aaSAlex Elder return -ENOMEM; 1968602adf40SYehuda Sadeh } 1969602adf40SYehuda Sadeh 1970dfc5606dSYehuda Sadeh /* 1971dfc5606dSYehuda Sadeh sysfs 1972dfc5606dSYehuda Sadeh */ 1973602adf40SYehuda Sadeh 1974593a9e7bSAlex Elder static struct rbd_device *dev_to_rbd_dev(struct device *dev) 1975593a9e7bSAlex Elder { 1976593a9e7bSAlex Elder return container_of(dev, struct rbd_device, dev); 1977593a9e7bSAlex Elder } 1978593a9e7bSAlex Elder 1979dfc5606dSYehuda Sadeh static ssize_t rbd_size_show(struct device *dev, 1980dfc5606dSYehuda Sadeh struct device_attribute *attr, char *buf) 1981602adf40SYehuda Sadeh { 1982593a9e7bSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 1983a51aa0c0SJosh Durgin sector_t size; 1984dfc5606dSYehuda Sadeh 1985a51aa0c0SJosh Durgin down_read(&rbd_dev->header_rwsem); 1986a51aa0c0SJosh Durgin size = get_capacity(rbd_dev->disk); 1987a51aa0c0SJosh Durgin up_read(&rbd_dev->header_rwsem); 1988a51aa0c0SJosh Durgin 1989a51aa0c0SJosh Durgin return sprintf(buf, "%llu\n", (unsigned long long) size * SECTOR_SIZE); 1990602adf40SYehuda Sadeh } 1991602adf40SYehuda Sadeh 199234b13184SAlex Elder /* 199334b13184SAlex Elder * Note this shows the features for whatever's mapped, which is not 199434b13184SAlex Elder * necessarily the base image. 199534b13184SAlex Elder */ 199634b13184SAlex Elder static ssize_t rbd_features_show(struct device *dev, 199734b13184SAlex Elder struct device_attribute *attr, char *buf) 199834b13184SAlex Elder { 199934b13184SAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 200034b13184SAlex Elder 200134b13184SAlex Elder return sprintf(buf, "0x%016llx\n", 200234b13184SAlex Elder (unsigned long long) rbd_dev->mapping.features); 200334b13184SAlex Elder } 200434b13184SAlex Elder 2005dfc5606dSYehuda Sadeh static ssize_t rbd_major_show(struct device *dev, 2006dfc5606dSYehuda Sadeh struct device_attribute *attr, char *buf) 2007602adf40SYehuda Sadeh { 2008593a9e7bSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 2009dfc5606dSYehuda Sadeh 2010dfc5606dSYehuda Sadeh return sprintf(buf, "%d\n", rbd_dev->major); 2011dfc5606dSYehuda Sadeh } 2012dfc5606dSYehuda Sadeh 2013dfc5606dSYehuda Sadeh static ssize_t rbd_client_id_show(struct device *dev, 2014dfc5606dSYehuda Sadeh struct device_attribute *attr, char *buf) 2015dfc5606dSYehuda Sadeh { 2016593a9e7bSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 2017dfc5606dSYehuda Sadeh 20181dbb4399SAlex Elder return sprintf(buf, "client%lld\n", 20191dbb4399SAlex Elder ceph_client_id(rbd_dev->rbd_client->client)); 2020dfc5606dSYehuda Sadeh } 2021dfc5606dSYehuda Sadeh 2022dfc5606dSYehuda Sadeh static ssize_t rbd_pool_show(struct device *dev, 2023dfc5606dSYehuda Sadeh struct device_attribute *attr, char *buf) 2024dfc5606dSYehuda Sadeh { 2025593a9e7bSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 2026dfc5606dSYehuda Sadeh 20270d7dbfceSAlex Elder return sprintf(buf, "%s\n", rbd_dev->spec->pool_name); 2028dfc5606dSYehuda Sadeh } 2029dfc5606dSYehuda Sadeh 20309bb2f334SAlex Elder static ssize_t rbd_pool_id_show(struct device *dev, 20319bb2f334SAlex Elder struct device_attribute *attr, char *buf) 20329bb2f334SAlex Elder { 20339bb2f334SAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 20349bb2f334SAlex Elder 20350d7dbfceSAlex Elder return sprintf(buf, "%llu\n", 20360d7dbfceSAlex Elder (unsigned long long) rbd_dev->spec->pool_id); 20379bb2f334SAlex Elder } 20389bb2f334SAlex Elder 2039dfc5606dSYehuda Sadeh static ssize_t rbd_name_show(struct device *dev, 2040dfc5606dSYehuda Sadeh struct device_attribute *attr, char *buf) 2041dfc5606dSYehuda Sadeh { 2042593a9e7bSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 2043dfc5606dSYehuda Sadeh 2044a92ffdf8SAlex Elder if (rbd_dev->spec->image_name) 20450d7dbfceSAlex Elder return sprintf(buf, "%s\n", rbd_dev->spec->image_name); 2046a92ffdf8SAlex Elder 2047a92ffdf8SAlex Elder return sprintf(buf, "(unknown)\n"); 2048dfc5606dSYehuda Sadeh } 2049dfc5606dSYehuda Sadeh 2050589d30e0SAlex Elder static ssize_t rbd_image_id_show(struct device *dev, 2051589d30e0SAlex Elder struct device_attribute *attr, char *buf) 2052589d30e0SAlex Elder { 2053589d30e0SAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 2054589d30e0SAlex Elder 20550d7dbfceSAlex Elder return sprintf(buf, "%s\n", rbd_dev->spec->image_id); 2056589d30e0SAlex Elder } 2057589d30e0SAlex Elder 205834b13184SAlex Elder /* 205934b13184SAlex Elder * Shows the name of the currently-mapped snapshot (or 206034b13184SAlex Elder * RBD_SNAP_HEAD_NAME for the base image). 206134b13184SAlex Elder */ 2062dfc5606dSYehuda Sadeh static ssize_t rbd_snap_show(struct device *dev, 2063dfc5606dSYehuda Sadeh struct device_attribute *attr, 2064dfc5606dSYehuda Sadeh char *buf) 2065dfc5606dSYehuda Sadeh { 2066593a9e7bSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 2067dfc5606dSYehuda Sadeh 20680d7dbfceSAlex Elder return sprintf(buf, "%s\n", rbd_dev->spec->snap_name); 2069dfc5606dSYehuda Sadeh } 2070dfc5606dSYehuda Sadeh 207186b00e0dSAlex Elder /* 207286b00e0dSAlex Elder * For an rbd v2 image, shows the pool id, image id, and snapshot id 207386b00e0dSAlex Elder * for the parent image. If there is no parent, simply shows 207486b00e0dSAlex Elder * "(no parent image)". 207586b00e0dSAlex Elder */ 207686b00e0dSAlex Elder static ssize_t rbd_parent_show(struct device *dev, 207786b00e0dSAlex Elder struct device_attribute *attr, 207886b00e0dSAlex Elder char *buf) 207986b00e0dSAlex Elder { 208086b00e0dSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 208186b00e0dSAlex Elder struct rbd_spec *spec = rbd_dev->parent_spec; 208286b00e0dSAlex Elder int count; 208386b00e0dSAlex Elder char *bufp = buf; 208486b00e0dSAlex Elder 208586b00e0dSAlex Elder if (!spec) 208686b00e0dSAlex Elder return sprintf(buf, "(no parent image)\n"); 208786b00e0dSAlex Elder 208886b00e0dSAlex Elder count = sprintf(bufp, "pool_id %llu\npool_name %s\n", 208986b00e0dSAlex Elder (unsigned long long) spec->pool_id, spec->pool_name); 209086b00e0dSAlex Elder if (count < 0) 209186b00e0dSAlex Elder return count; 209286b00e0dSAlex Elder bufp += count; 209386b00e0dSAlex Elder 209486b00e0dSAlex Elder count = sprintf(bufp, "image_id %s\nimage_name %s\n", spec->image_id, 209586b00e0dSAlex Elder spec->image_name ? spec->image_name : "(unknown)"); 209686b00e0dSAlex Elder if (count < 0) 209786b00e0dSAlex Elder return count; 209886b00e0dSAlex Elder bufp += count; 209986b00e0dSAlex Elder 210086b00e0dSAlex Elder count = sprintf(bufp, "snap_id %llu\nsnap_name %s\n", 210186b00e0dSAlex Elder (unsigned long long) spec->snap_id, spec->snap_name); 210286b00e0dSAlex Elder if (count < 0) 210386b00e0dSAlex Elder return count; 210486b00e0dSAlex Elder bufp += count; 210586b00e0dSAlex Elder 210686b00e0dSAlex Elder count = sprintf(bufp, "overlap %llu\n", rbd_dev->parent_overlap); 210786b00e0dSAlex Elder if (count < 0) 210886b00e0dSAlex Elder return count; 210986b00e0dSAlex Elder bufp += count; 211086b00e0dSAlex Elder 211186b00e0dSAlex Elder return (ssize_t) (bufp - buf); 211286b00e0dSAlex Elder } 211386b00e0dSAlex Elder 2114dfc5606dSYehuda Sadeh static ssize_t rbd_image_refresh(struct device *dev, 2115dfc5606dSYehuda Sadeh struct device_attribute *attr, 2116dfc5606dSYehuda Sadeh const char *buf, 2117dfc5606dSYehuda Sadeh size_t size) 2118dfc5606dSYehuda Sadeh { 2119593a9e7bSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 2120b813623aSAlex Elder int ret; 2121602adf40SYehuda Sadeh 2122117973fbSAlex Elder ret = rbd_dev_refresh(rbd_dev, NULL); 2123b813623aSAlex Elder 2124b813623aSAlex Elder return ret < 0 ? ret : size; 2125dfc5606dSYehuda Sadeh } 2126602adf40SYehuda Sadeh 2127dfc5606dSYehuda Sadeh static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); 212834b13184SAlex Elder static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL); 2129dfc5606dSYehuda Sadeh static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); 2130dfc5606dSYehuda Sadeh static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); 2131dfc5606dSYehuda Sadeh static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); 21329bb2f334SAlex Elder static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); 2133dfc5606dSYehuda Sadeh static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); 2134589d30e0SAlex Elder static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL); 2135dfc5606dSYehuda Sadeh static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); 2136dfc5606dSYehuda Sadeh static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); 213786b00e0dSAlex Elder static DEVICE_ATTR(parent, S_IRUGO, rbd_parent_show, NULL); 2138dfc5606dSYehuda Sadeh 2139dfc5606dSYehuda Sadeh static struct attribute *rbd_attrs[] = { 2140dfc5606dSYehuda Sadeh &dev_attr_size.attr, 214134b13184SAlex Elder &dev_attr_features.attr, 2142dfc5606dSYehuda Sadeh &dev_attr_major.attr, 2143dfc5606dSYehuda Sadeh &dev_attr_client_id.attr, 2144dfc5606dSYehuda Sadeh &dev_attr_pool.attr, 21459bb2f334SAlex Elder &dev_attr_pool_id.attr, 2146dfc5606dSYehuda Sadeh &dev_attr_name.attr, 2147589d30e0SAlex Elder &dev_attr_image_id.attr, 2148dfc5606dSYehuda Sadeh &dev_attr_current_snap.attr, 214986b00e0dSAlex Elder &dev_attr_parent.attr, 2150dfc5606dSYehuda Sadeh &dev_attr_refresh.attr, 2151dfc5606dSYehuda Sadeh NULL 2152dfc5606dSYehuda Sadeh }; 2153dfc5606dSYehuda Sadeh 2154dfc5606dSYehuda Sadeh static struct attribute_group rbd_attr_group = { 2155dfc5606dSYehuda Sadeh .attrs = rbd_attrs, 2156dfc5606dSYehuda Sadeh }; 2157dfc5606dSYehuda Sadeh 2158dfc5606dSYehuda Sadeh static const struct attribute_group *rbd_attr_groups[] = { 2159dfc5606dSYehuda Sadeh &rbd_attr_group, 2160dfc5606dSYehuda Sadeh NULL 2161dfc5606dSYehuda Sadeh }; 2162dfc5606dSYehuda Sadeh 2163dfc5606dSYehuda Sadeh static void rbd_sysfs_dev_release(struct device *dev) 2164dfc5606dSYehuda Sadeh { 2165dfc5606dSYehuda Sadeh } 2166dfc5606dSYehuda Sadeh 2167dfc5606dSYehuda Sadeh static struct device_type rbd_device_type = { 2168dfc5606dSYehuda Sadeh .name = "rbd", 2169dfc5606dSYehuda Sadeh .groups = rbd_attr_groups, 2170dfc5606dSYehuda Sadeh .release = rbd_sysfs_dev_release, 2171dfc5606dSYehuda Sadeh }; 2172dfc5606dSYehuda Sadeh 2173dfc5606dSYehuda Sadeh 2174dfc5606dSYehuda Sadeh /* 2175dfc5606dSYehuda Sadeh sysfs - snapshots 2176dfc5606dSYehuda Sadeh */ 2177dfc5606dSYehuda Sadeh 2178dfc5606dSYehuda Sadeh static ssize_t rbd_snap_size_show(struct device *dev, 2179dfc5606dSYehuda Sadeh struct device_attribute *attr, 2180dfc5606dSYehuda Sadeh char *buf) 2181dfc5606dSYehuda Sadeh { 2182dfc5606dSYehuda Sadeh struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); 2183dfc5606dSYehuda Sadeh 21843591538fSJosh Durgin return sprintf(buf, "%llu\n", (unsigned long long)snap->size); 2185dfc5606dSYehuda Sadeh } 2186dfc5606dSYehuda Sadeh 2187dfc5606dSYehuda Sadeh static ssize_t rbd_snap_id_show(struct device *dev, 2188dfc5606dSYehuda Sadeh struct device_attribute *attr, 2189dfc5606dSYehuda Sadeh char *buf) 2190dfc5606dSYehuda Sadeh { 2191dfc5606dSYehuda Sadeh struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); 2192dfc5606dSYehuda Sadeh 2193593a9e7bSAlex Elder return sprintf(buf, "%llu\n", (unsigned long long)snap->id); 2194dfc5606dSYehuda Sadeh } 2195dfc5606dSYehuda Sadeh 219634b13184SAlex Elder static ssize_t rbd_snap_features_show(struct device *dev, 219734b13184SAlex Elder struct device_attribute *attr, 219834b13184SAlex Elder char *buf) 219934b13184SAlex Elder { 220034b13184SAlex Elder struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); 220134b13184SAlex Elder 220234b13184SAlex Elder return sprintf(buf, "0x%016llx\n", 220334b13184SAlex Elder (unsigned long long) snap->features); 220434b13184SAlex Elder } 220534b13184SAlex Elder 2206dfc5606dSYehuda Sadeh static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); 2207dfc5606dSYehuda Sadeh static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); 220834b13184SAlex Elder static DEVICE_ATTR(snap_features, S_IRUGO, rbd_snap_features_show, NULL); 2209dfc5606dSYehuda Sadeh 2210dfc5606dSYehuda Sadeh static struct attribute *rbd_snap_attrs[] = { 2211dfc5606dSYehuda Sadeh &dev_attr_snap_size.attr, 2212dfc5606dSYehuda Sadeh &dev_attr_snap_id.attr, 221334b13184SAlex Elder &dev_attr_snap_features.attr, 2214dfc5606dSYehuda Sadeh NULL, 2215dfc5606dSYehuda Sadeh }; 2216dfc5606dSYehuda Sadeh 2217dfc5606dSYehuda Sadeh static struct attribute_group rbd_snap_attr_group = { 2218dfc5606dSYehuda Sadeh .attrs = rbd_snap_attrs, 2219dfc5606dSYehuda Sadeh }; 2220dfc5606dSYehuda Sadeh 2221dfc5606dSYehuda Sadeh static void rbd_snap_dev_release(struct device *dev) 2222dfc5606dSYehuda Sadeh { 2223dfc5606dSYehuda Sadeh struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); 2224dfc5606dSYehuda Sadeh kfree(snap->name); 2225dfc5606dSYehuda Sadeh kfree(snap); 2226dfc5606dSYehuda Sadeh } 2227dfc5606dSYehuda Sadeh 2228dfc5606dSYehuda Sadeh static const struct attribute_group *rbd_snap_attr_groups[] = { 2229dfc5606dSYehuda Sadeh &rbd_snap_attr_group, 2230dfc5606dSYehuda Sadeh NULL 2231dfc5606dSYehuda Sadeh }; 2232dfc5606dSYehuda Sadeh 2233dfc5606dSYehuda Sadeh static struct device_type rbd_snap_device_type = { 2234dfc5606dSYehuda Sadeh .groups = rbd_snap_attr_groups, 2235dfc5606dSYehuda Sadeh .release = rbd_snap_dev_release, 2236dfc5606dSYehuda Sadeh }; 2237dfc5606dSYehuda Sadeh 22388b8fb99cSAlex Elder static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec) 22398b8fb99cSAlex Elder { 22408b8fb99cSAlex Elder kref_get(&spec->kref); 22418b8fb99cSAlex Elder 22428b8fb99cSAlex Elder return spec; 22438b8fb99cSAlex Elder } 22448b8fb99cSAlex Elder 22458b8fb99cSAlex Elder static void rbd_spec_free(struct kref *kref); 22468b8fb99cSAlex Elder static void rbd_spec_put(struct rbd_spec *spec) 22478b8fb99cSAlex Elder { 22488b8fb99cSAlex Elder if (spec) 22498b8fb99cSAlex Elder kref_put(&spec->kref, rbd_spec_free); 22508b8fb99cSAlex Elder } 22518b8fb99cSAlex Elder 22528b8fb99cSAlex Elder static struct rbd_spec *rbd_spec_alloc(void) 22538b8fb99cSAlex Elder { 22548b8fb99cSAlex Elder struct rbd_spec *spec; 22558b8fb99cSAlex Elder 22568b8fb99cSAlex Elder spec = kzalloc(sizeof (*spec), GFP_KERNEL); 22578b8fb99cSAlex Elder if (!spec) 22588b8fb99cSAlex Elder return NULL; 22598b8fb99cSAlex Elder kref_init(&spec->kref); 22608b8fb99cSAlex Elder 22618b8fb99cSAlex Elder rbd_spec_put(rbd_spec_get(spec)); /* TEMPORARY */ 22628b8fb99cSAlex Elder 22638b8fb99cSAlex Elder return spec; 22648b8fb99cSAlex Elder } 22658b8fb99cSAlex Elder 22668b8fb99cSAlex Elder static void rbd_spec_free(struct kref *kref) 22678b8fb99cSAlex Elder { 22688b8fb99cSAlex Elder struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref); 22698b8fb99cSAlex Elder 22708b8fb99cSAlex Elder kfree(spec->pool_name); 22718b8fb99cSAlex Elder kfree(spec->image_id); 22728b8fb99cSAlex Elder kfree(spec->image_name); 22738b8fb99cSAlex Elder kfree(spec->snap_name); 22748b8fb99cSAlex Elder kfree(spec); 22758b8fb99cSAlex Elder } 22768b8fb99cSAlex Elder 2277c53d5893SAlex Elder struct rbd_device *rbd_dev_create(struct rbd_client *rbdc, 2278c53d5893SAlex Elder struct rbd_spec *spec) 2279c53d5893SAlex Elder { 2280c53d5893SAlex Elder struct rbd_device *rbd_dev; 2281c53d5893SAlex Elder 2282c53d5893SAlex Elder rbd_dev = kzalloc(sizeof (*rbd_dev), GFP_KERNEL); 2283c53d5893SAlex Elder if (!rbd_dev) 2284c53d5893SAlex Elder return NULL; 2285c53d5893SAlex Elder 2286c53d5893SAlex Elder spin_lock_init(&rbd_dev->lock); 2287d78b650aSAlex Elder atomic_set(&rbd_dev->exists, 0); 2288c53d5893SAlex Elder INIT_LIST_HEAD(&rbd_dev->node); 2289c53d5893SAlex Elder INIT_LIST_HEAD(&rbd_dev->snaps); 2290c53d5893SAlex Elder init_rwsem(&rbd_dev->header_rwsem); 2291c53d5893SAlex Elder 2292c53d5893SAlex Elder rbd_dev->spec = spec; 2293c53d5893SAlex Elder rbd_dev->rbd_client = rbdc; 2294c53d5893SAlex Elder 2295c53d5893SAlex Elder return rbd_dev; 2296c53d5893SAlex Elder } 2297c53d5893SAlex Elder 2298c53d5893SAlex Elder static void rbd_dev_destroy(struct rbd_device *rbd_dev) 2299c53d5893SAlex Elder { 230086b00e0dSAlex Elder rbd_spec_put(rbd_dev->parent_spec); 2301c53d5893SAlex Elder kfree(rbd_dev->header_name); 2302c53d5893SAlex Elder rbd_put_client(rbd_dev->rbd_client); 2303c53d5893SAlex Elder rbd_spec_put(rbd_dev->spec); 2304c53d5893SAlex Elder kfree(rbd_dev); 2305c53d5893SAlex Elder } 2306c53d5893SAlex Elder 2307304f6808SAlex Elder static bool rbd_snap_registered(struct rbd_snap *snap) 2308304f6808SAlex Elder { 2309304f6808SAlex Elder bool ret = snap->dev.type == &rbd_snap_device_type; 2310304f6808SAlex Elder bool reg = device_is_registered(&snap->dev); 2311304f6808SAlex Elder 2312304f6808SAlex Elder rbd_assert(!ret ^ reg); 2313304f6808SAlex Elder 2314304f6808SAlex Elder return ret; 2315304f6808SAlex Elder } 2316304f6808SAlex Elder 231741f38c2bSAlex Elder static void rbd_remove_snap_dev(struct rbd_snap *snap) 2318dfc5606dSYehuda Sadeh { 2319dfc5606dSYehuda Sadeh list_del(&snap->node); 2320304f6808SAlex Elder if (device_is_registered(&snap->dev)) 2321dfc5606dSYehuda Sadeh device_unregister(&snap->dev); 2322dfc5606dSYehuda Sadeh } 2323dfc5606dSYehuda Sadeh 232414e7085dSAlex Elder static int rbd_register_snap_dev(struct rbd_snap *snap, 2325dfc5606dSYehuda Sadeh struct device *parent) 2326dfc5606dSYehuda Sadeh { 2327dfc5606dSYehuda Sadeh struct device *dev = &snap->dev; 2328dfc5606dSYehuda Sadeh int ret; 2329dfc5606dSYehuda Sadeh 2330dfc5606dSYehuda Sadeh dev->type = &rbd_snap_device_type; 2331dfc5606dSYehuda Sadeh dev->parent = parent; 2332dfc5606dSYehuda Sadeh dev->release = rbd_snap_dev_release; 2333d4b125e9SAlex Elder dev_set_name(dev, "%s%s", RBD_SNAP_DEV_NAME_PREFIX, snap->name); 2334304f6808SAlex Elder dout("%s: registering device for snapshot %s\n", __func__, snap->name); 2335304f6808SAlex Elder 2336dfc5606dSYehuda Sadeh ret = device_register(dev); 2337dfc5606dSYehuda Sadeh 2338dfc5606dSYehuda Sadeh return ret; 2339dfc5606dSYehuda Sadeh } 2340dfc5606dSYehuda Sadeh 23414e891e0aSAlex Elder static struct rbd_snap *__rbd_add_snap_dev(struct rbd_device *rbd_dev, 2342c8d18425SAlex Elder const char *snap_name, 234334b13184SAlex Elder u64 snap_id, u64 snap_size, 234434b13184SAlex Elder u64 snap_features) 2345dfc5606dSYehuda Sadeh { 23464e891e0aSAlex Elder struct rbd_snap *snap; 2347dfc5606dSYehuda Sadeh int ret; 23484e891e0aSAlex Elder 23494e891e0aSAlex Elder snap = kzalloc(sizeof (*snap), GFP_KERNEL); 2350dfc5606dSYehuda Sadeh if (!snap) 23514e891e0aSAlex Elder return ERR_PTR(-ENOMEM); 23524e891e0aSAlex Elder 23534e891e0aSAlex Elder ret = -ENOMEM; 2354c8d18425SAlex Elder snap->name = kstrdup(snap_name, GFP_KERNEL); 23554e891e0aSAlex Elder if (!snap->name) 23564e891e0aSAlex Elder goto err; 23574e891e0aSAlex Elder 2358c8d18425SAlex Elder snap->id = snap_id; 2359c8d18425SAlex Elder snap->size = snap_size; 236034b13184SAlex Elder snap->features = snap_features; 23614e891e0aSAlex Elder 23624e891e0aSAlex Elder return snap; 23634e891e0aSAlex Elder 2364dfc5606dSYehuda Sadeh err: 2365dfc5606dSYehuda Sadeh kfree(snap->name); 2366dfc5606dSYehuda Sadeh kfree(snap); 23674e891e0aSAlex Elder 23684e891e0aSAlex Elder return ERR_PTR(ret); 2369dfc5606dSYehuda Sadeh } 2370dfc5606dSYehuda Sadeh 2371cd892126SAlex Elder static char *rbd_dev_v1_snap_info(struct rbd_device *rbd_dev, u32 which, 2372cd892126SAlex Elder u64 *snap_size, u64 *snap_features) 2373cd892126SAlex Elder { 2374cd892126SAlex Elder char *snap_name; 2375cd892126SAlex Elder 2376cd892126SAlex Elder rbd_assert(which < rbd_dev->header.snapc->num_snaps); 2377cd892126SAlex Elder 2378cd892126SAlex Elder *snap_size = rbd_dev->header.snap_sizes[which]; 2379cd892126SAlex Elder *snap_features = 0; /* No features for v1 */ 2380cd892126SAlex Elder 2381cd892126SAlex Elder /* Skip over names until we find the one we are looking for */ 2382cd892126SAlex Elder 2383cd892126SAlex Elder snap_name = rbd_dev->header.snap_names; 2384cd892126SAlex Elder while (which--) 2385cd892126SAlex Elder snap_name += strlen(snap_name) + 1; 2386cd892126SAlex Elder 2387cd892126SAlex Elder return snap_name; 2388cd892126SAlex Elder } 2389cd892126SAlex Elder 2390dfc5606dSYehuda Sadeh /* 23919d475de5SAlex Elder * Get the size and object order for an image snapshot, or if 23929d475de5SAlex Elder * snap_id is CEPH_NOSNAP, gets this information for the base 23939d475de5SAlex Elder * image. 23949d475de5SAlex Elder */ 23959d475de5SAlex Elder static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, 23969d475de5SAlex Elder u8 *order, u64 *snap_size) 23979d475de5SAlex Elder { 23989d475de5SAlex Elder __le64 snapid = cpu_to_le64(snap_id); 23999d475de5SAlex Elder int ret; 24009d475de5SAlex Elder struct { 24019d475de5SAlex Elder u8 order; 24029d475de5SAlex Elder __le64 size; 24039d475de5SAlex Elder } __attribute__ ((packed)) size_buf = { 0 }; 24049d475de5SAlex Elder 24059d475de5SAlex Elder ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, 24069d475de5SAlex Elder "rbd", "get_size", 24079d475de5SAlex Elder (char *) &snapid, sizeof (snapid), 240807b2391fSAlex Elder (char *) &size_buf, sizeof (size_buf), NULL); 24099d475de5SAlex Elder dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); 24109d475de5SAlex Elder if (ret < 0) 24119d475de5SAlex Elder return ret; 24129d475de5SAlex Elder 24139d475de5SAlex Elder *order = size_buf.order; 24149d475de5SAlex Elder *snap_size = le64_to_cpu(size_buf.size); 24159d475de5SAlex Elder 24169d475de5SAlex Elder dout(" snap_id 0x%016llx order = %u, snap_size = %llu\n", 24179d475de5SAlex Elder (unsigned long long) snap_id, (unsigned int) *order, 24189d475de5SAlex Elder (unsigned long long) *snap_size); 24199d475de5SAlex Elder 24209d475de5SAlex Elder return 0; 24219d475de5SAlex Elder } 24229d475de5SAlex Elder 24239d475de5SAlex Elder static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) 24249d475de5SAlex Elder { 24259d475de5SAlex Elder return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, 24269d475de5SAlex Elder &rbd_dev->header.obj_order, 24279d475de5SAlex Elder &rbd_dev->header.image_size); 24289d475de5SAlex Elder } 24299d475de5SAlex Elder 24301e130199SAlex Elder static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) 24311e130199SAlex Elder { 24321e130199SAlex Elder void *reply_buf; 24331e130199SAlex Elder int ret; 24341e130199SAlex Elder void *p; 24351e130199SAlex Elder 24361e130199SAlex Elder reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL); 24371e130199SAlex Elder if (!reply_buf) 24381e130199SAlex Elder return -ENOMEM; 24391e130199SAlex Elder 24401e130199SAlex Elder ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, 24411e130199SAlex Elder "rbd", "get_object_prefix", 24421e130199SAlex Elder NULL, 0, 244307b2391fSAlex Elder reply_buf, RBD_OBJ_PREFIX_LEN_MAX, NULL); 24441e130199SAlex Elder dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); 24451e130199SAlex Elder if (ret < 0) 24461e130199SAlex Elder goto out; 2447a0ea3a40SAlex Elder ret = 0; /* rbd_req_sync_exec() can return positive */ 24481e130199SAlex Elder 24491e130199SAlex Elder p = reply_buf; 24501e130199SAlex Elder rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, 24511e130199SAlex Elder p + RBD_OBJ_PREFIX_LEN_MAX, 24521e130199SAlex Elder NULL, GFP_NOIO); 24531e130199SAlex Elder 24541e130199SAlex Elder if (IS_ERR(rbd_dev->header.object_prefix)) { 24551e130199SAlex Elder ret = PTR_ERR(rbd_dev->header.object_prefix); 24561e130199SAlex Elder rbd_dev->header.object_prefix = NULL; 24571e130199SAlex Elder } else { 24581e130199SAlex Elder dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); 24591e130199SAlex Elder } 24601e130199SAlex Elder 24611e130199SAlex Elder out: 24621e130199SAlex Elder kfree(reply_buf); 24631e130199SAlex Elder 24641e130199SAlex Elder return ret; 24651e130199SAlex Elder } 24661e130199SAlex Elder 2467b1b5402aSAlex Elder static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, 2468b1b5402aSAlex Elder u64 *snap_features) 2469b1b5402aSAlex Elder { 2470b1b5402aSAlex Elder __le64 snapid = cpu_to_le64(snap_id); 2471b1b5402aSAlex Elder struct { 2472b1b5402aSAlex Elder __le64 features; 2473b1b5402aSAlex Elder __le64 incompat; 2474b1b5402aSAlex Elder } features_buf = { 0 }; 2475d889140cSAlex Elder u64 incompat; 2476b1b5402aSAlex Elder int ret; 2477b1b5402aSAlex Elder 2478b1b5402aSAlex Elder ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, 2479b1b5402aSAlex Elder "rbd", "get_features", 2480b1b5402aSAlex Elder (char *) &snapid, sizeof (snapid), 2481b1b5402aSAlex Elder (char *) &features_buf, sizeof (features_buf), 248207b2391fSAlex Elder NULL); 2483b1b5402aSAlex Elder dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); 2484b1b5402aSAlex Elder if (ret < 0) 2485b1b5402aSAlex Elder return ret; 2486d889140cSAlex Elder 2487d889140cSAlex Elder incompat = le64_to_cpu(features_buf.incompat); 2488d889140cSAlex Elder if (incompat & ~RBD_FEATURES_ALL) 2489b8f5c6edSAlex Elder return -ENXIO; 2490d889140cSAlex Elder 2491b1b5402aSAlex Elder *snap_features = le64_to_cpu(features_buf.features); 2492b1b5402aSAlex Elder 2493b1b5402aSAlex Elder dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", 2494b1b5402aSAlex Elder (unsigned long long) snap_id, 2495b1b5402aSAlex Elder (unsigned long long) *snap_features, 2496b1b5402aSAlex Elder (unsigned long long) le64_to_cpu(features_buf.incompat)); 2497b1b5402aSAlex Elder 2498b1b5402aSAlex Elder return 0; 2499b1b5402aSAlex Elder } 2500b1b5402aSAlex Elder 2501b1b5402aSAlex Elder static int rbd_dev_v2_features(struct rbd_device *rbd_dev) 2502b1b5402aSAlex Elder { 2503b1b5402aSAlex Elder return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, 2504b1b5402aSAlex Elder &rbd_dev->header.features); 2505b1b5402aSAlex Elder } 2506b1b5402aSAlex Elder 250786b00e0dSAlex Elder static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) 250886b00e0dSAlex Elder { 250986b00e0dSAlex Elder struct rbd_spec *parent_spec; 251086b00e0dSAlex Elder size_t size; 251186b00e0dSAlex Elder void *reply_buf = NULL; 251286b00e0dSAlex Elder __le64 snapid; 251386b00e0dSAlex Elder void *p; 251486b00e0dSAlex Elder void *end; 251586b00e0dSAlex Elder char *image_id; 251686b00e0dSAlex Elder u64 overlap; 251786b00e0dSAlex Elder int ret; 251886b00e0dSAlex Elder 251986b00e0dSAlex Elder parent_spec = rbd_spec_alloc(); 252086b00e0dSAlex Elder if (!parent_spec) 252186b00e0dSAlex Elder return -ENOMEM; 252286b00e0dSAlex Elder 252386b00e0dSAlex Elder size = sizeof (__le64) + /* pool_id */ 252486b00e0dSAlex Elder sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX + /* image_id */ 252586b00e0dSAlex Elder sizeof (__le64) + /* snap_id */ 252686b00e0dSAlex Elder sizeof (__le64); /* overlap */ 252786b00e0dSAlex Elder reply_buf = kmalloc(size, GFP_KERNEL); 252886b00e0dSAlex Elder if (!reply_buf) { 252986b00e0dSAlex Elder ret = -ENOMEM; 253086b00e0dSAlex Elder goto out_err; 253186b00e0dSAlex Elder } 253286b00e0dSAlex Elder 253386b00e0dSAlex Elder snapid = cpu_to_le64(CEPH_NOSNAP); 253486b00e0dSAlex Elder ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, 253586b00e0dSAlex Elder "rbd", "get_parent", 253686b00e0dSAlex Elder (char *) &snapid, sizeof (snapid), 253707b2391fSAlex Elder (char *) reply_buf, size, NULL); 253886b00e0dSAlex Elder dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); 253986b00e0dSAlex Elder if (ret < 0) 254086b00e0dSAlex Elder goto out_err; 254186b00e0dSAlex Elder 254286b00e0dSAlex Elder ret = -ERANGE; 254386b00e0dSAlex Elder p = reply_buf; 254486b00e0dSAlex Elder end = (char *) reply_buf + size; 254586b00e0dSAlex Elder ceph_decode_64_safe(&p, end, parent_spec->pool_id, out_err); 254686b00e0dSAlex Elder if (parent_spec->pool_id == CEPH_NOPOOL) 254786b00e0dSAlex Elder goto out; /* No parent? No problem. */ 254886b00e0dSAlex Elder 2549979ed480SAlex Elder image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); 255086b00e0dSAlex Elder if (IS_ERR(image_id)) { 255186b00e0dSAlex Elder ret = PTR_ERR(image_id); 255286b00e0dSAlex Elder goto out_err; 255386b00e0dSAlex Elder } 255486b00e0dSAlex Elder parent_spec->image_id = image_id; 255586b00e0dSAlex Elder ceph_decode_64_safe(&p, end, parent_spec->snap_id, out_err); 255686b00e0dSAlex Elder ceph_decode_64_safe(&p, end, overlap, out_err); 255786b00e0dSAlex Elder 255886b00e0dSAlex Elder rbd_dev->parent_overlap = overlap; 255986b00e0dSAlex Elder rbd_dev->parent_spec = parent_spec; 256086b00e0dSAlex Elder parent_spec = NULL; /* rbd_dev now owns this */ 256186b00e0dSAlex Elder out: 256286b00e0dSAlex Elder ret = 0; 256386b00e0dSAlex Elder out_err: 256486b00e0dSAlex Elder kfree(reply_buf); 256586b00e0dSAlex Elder rbd_spec_put(parent_spec); 256686b00e0dSAlex Elder 256786b00e0dSAlex Elder return ret; 256886b00e0dSAlex Elder } 256986b00e0dSAlex Elder 25709e15b77dSAlex Elder static char *rbd_dev_image_name(struct rbd_device *rbd_dev) 25719e15b77dSAlex Elder { 25729e15b77dSAlex Elder size_t image_id_size; 25739e15b77dSAlex Elder char *image_id; 25749e15b77dSAlex Elder void *p; 25759e15b77dSAlex Elder void *end; 25769e15b77dSAlex Elder size_t size; 25779e15b77dSAlex Elder void *reply_buf = NULL; 25789e15b77dSAlex Elder size_t len = 0; 25799e15b77dSAlex Elder char *image_name = NULL; 25809e15b77dSAlex Elder int ret; 25819e15b77dSAlex Elder 25829e15b77dSAlex Elder rbd_assert(!rbd_dev->spec->image_name); 25839e15b77dSAlex Elder 258469e7a02fSAlex Elder len = strlen(rbd_dev->spec->image_id); 258569e7a02fSAlex Elder image_id_size = sizeof (__le32) + len; 25869e15b77dSAlex Elder image_id = kmalloc(image_id_size, GFP_KERNEL); 25879e15b77dSAlex Elder if (!image_id) 25889e15b77dSAlex Elder return NULL; 25899e15b77dSAlex Elder 25909e15b77dSAlex Elder p = image_id; 25919e15b77dSAlex Elder end = (char *) image_id + image_id_size; 259269e7a02fSAlex Elder ceph_encode_string(&p, end, rbd_dev->spec->image_id, (u32) len); 25939e15b77dSAlex Elder 25949e15b77dSAlex Elder size = sizeof (__le32) + RBD_IMAGE_NAME_LEN_MAX; 25959e15b77dSAlex Elder reply_buf = kmalloc(size, GFP_KERNEL); 25969e15b77dSAlex Elder if (!reply_buf) 25979e15b77dSAlex Elder goto out; 25989e15b77dSAlex Elder 25999e15b77dSAlex Elder ret = rbd_req_sync_exec(rbd_dev, RBD_DIRECTORY, 26009e15b77dSAlex Elder "rbd", "dir_get_name", 26019e15b77dSAlex Elder image_id, image_id_size, 260207b2391fSAlex Elder (char *) reply_buf, size, NULL); 26039e15b77dSAlex Elder if (ret < 0) 26049e15b77dSAlex Elder goto out; 26059e15b77dSAlex Elder p = reply_buf; 26069e15b77dSAlex Elder end = (char *) reply_buf + size; 26079e15b77dSAlex Elder image_name = ceph_extract_encoded_string(&p, end, &len, GFP_KERNEL); 26089e15b77dSAlex Elder if (IS_ERR(image_name)) 26099e15b77dSAlex Elder image_name = NULL; 26109e15b77dSAlex Elder else 26119e15b77dSAlex Elder dout("%s: name is %s len is %zd\n", __func__, image_name, len); 26129e15b77dSAlex Elder out: 26139e15b77dSAlex Elder kfree(reply_buf); 26149e15b77dSAlex Elder kfree(image_id); 26159e15b77dSAlex Elder 26169e15b77dSAlex Elder return image_name; 26179e15b77dSAlex Elder } 26189e15b77dSAlex Elder 26199e15b77dSAlex Elder /* 26209e15b77dSAlex Elder * When a parent image gets probed, we only have the pool, image, 26219e15b77dSAlex Elder * and snapshot ids but not the names of any of them. This call 26229e15b77dSAlex Elder * is made later to fill in those names. It has to be done after 26239e15b77dSAlex Elder * rbd_dev_snaps_update() has completed because some of the 26249e15b77dSAlex Elder * information (in particular, snapshot name) is not available 26259e15b77dSAlex Elder * until then. 26269e15b77dSAlex Elder */ 26279e15b77dSAlex Elder static int rbd_dev_probe_update_spec(struct rbd_device *rbd_dev) 26289e15b77dSAlex Elder { 26299e15b77dSAlex Elder struct ceph_osd_client *osdc; 26309e15b77dSAlex Elder const char *name; 26319e15b77dSAlex Elder void *reply_buf = NULL; 26329e15b77dSAlex Elder int ret; 26339e15b77dSAlex Elder 26349e15b77dSAlex Elder if (rbd_dev->spec->pool_name) 26359e15b77dSAlex Elder return 0; /* Already have the names */ 26369e15b77dSAlex Elder 26379e15b77dSAlex Elder /* Look up the pool name */ 26389e15b77dSAlex Elder 26399e15b77dSAlex Elder osdc = &rbd_dev->rbd_client->client->osdc; 26409e15b77dSAlex Elder name = ceph_pg_pool_name_by_id(osdc->osdmap, rbd_dev->spec->pool_id); 2641935dc89fSAlex Elder if (!name) { 2642935dc89fSAlex Elder rbd_warn(rbd_dev, "there is no pool with id %llu", 2643935dc89fSAlex Elder rbd_dev->spec->pool_id); /* Really a BUG() */ 2644935dc89fSAlex Elder return -EIO; 2645935dc89fSAlex Elder } 26469e15b77dSAlex Elder 26479e15b77dSAlex Elder rbd_dev->spec->pool_name = kstrdup(name, GFP_KERNEL); 26489e15b77dSAlex Elder if (!rbd_dev->spec->pool_name) 26499e15b77dSAlex Elder return -ENOMEM; 26509e15b77dSAlex Elder 26519e15b77dSAlex Elder /* Fetch the image name; tolerate failure here */ 26529e15b77dSAlex Elder 26539e15b77dSAlex Elder name = rbd_dev_image_name(rbd_dev); 265469e7a02fSAlex Elder if (name) 26559e15b77dSAlex Elder rbd_dev->spec->image_name = (char *) name; 265669e7a02fSAlex Elder else 265706ecc6cbSAlex Elder rbd_warn(rbd_dev, "unable to get image name"); 26589e15b77dSAlex Elder 26599e15b77dSAlex Elder /* Look up the snapshot name. */ 26609e15b77dSAlex Elder 26619e15b77dSAlex Elder name = rbd_snap_name(rbd_dev, rbd_dev->spec->snap_id); 26629e15b77dSAlex Elder if (!name) { 2663935dc89fSAlex Elder rbd_warn(rbd_dev, "no snapshot with id %llu", 2664935dc89fSAlex Elder rbd_dev->spec->snap_id); /* Really a BUG() */ 26659e15b77dSAlex Elder ret = -EIO; 26669e15b77dSAlex Elder goto out_err; 26679e15b77dSAlex Elder } 26689e15b77dSAlex Elder rbd_dev->spec->snap_name = kstrdup(name, GFP_KERNEL); 26699e15b77dSAlex Elder if(!rbd_dev->spec->snap_name) 26709e15b77dSAlex Elder goto out_err; 26719e15b77dSAlex Elder 26729e15b77dSAlex Elder return 0; 26739e15b77dSAlex Elder out_err: 26749e15b77dSAlex Elder kfree(reply_buf); 26759e15b77dSAlex Elder kfree(rbd_dev->spec->pool_name); 26769e15b77dSAlex Elder rbd_dev->spec->pool_name = NULL; 26779e15b77dSAlex Elder 26789e15b77dSAlex Elder return ret; 26799e15b77dSAlex Elder } 26809e15b77dSAlex Elder 26816e14b1a6SAlex Elder static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev, u64 *ver) 268235d489f9SAlex Elder { 268335d489f9SAlex Elder size_t size; 268435d489f9SAlex Elder int ret; 268535d489f9SAlex Elder void *reply_buf; 268635d489f9SAlex Elder void *p; 268735d489f9SAlex Elder void *end; 268835d489f9SAlex Elder u64 seq; 268935d489f9SAlex Elder u32 snap_count; 269035d489f9SAlex Elder struct ceph_snap_context *snapc; 269135d489f9SAlex Elder u32 i; 269235d489f9SAlex Elder 269335d489f9SAlex Elder /* 269435d489f9SAlex Elder * We'll need room for the seq value (maximum snapshot id), 269535d489f9SAlex Elder * snapshot count, and array of that many snapshot ids. 269635d489f9SAlex Elder * For now we have a fixed upper limit on the number we're 269735d489f9SAlex Elder * prepared to receive. 269835d489f9SAlex Elder */ 269935d489f9SAlex Elder size = sizeof (__le64) + sizeof (__le32) + 270035d489f9SAlex Elder RBD_MAX_SNAP_COUNT * sizeof (__le64); 270135d489f9SAlex Elder reply_buf = kzalloc(size, GFP_KERNEL); 270235d489f9SAlex Elder if (!reply_buf) 270335d489f9SAlex Elder return -ENOMEM; 270435d489f9SAlex Elder 270535d489f9SAlex Elder ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, 270635d489f9SAlex Elder "rbd", "get_snapcontext", 270735d489f9SAlex Elder NULL, 0, 270807b2391fSAlex Elder reply_buf, size, ver); 270935d489f9SAlex Elder dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); 271035d489f9SAlex Elder if (ret < 0) 271135d489f9SAlex Elder goto out; 271235d489f9SAlex Elder 271335d489f9SAlex Elder ret = -ERANGE; 271435d489f9SAlex Elder p = reply_buf; 271535d489f9SAlex Elder end = (char *) reply_buf + size; 271635d489f9SAlex Elder ceph_decode_64_safe(&p, end, seq, out); 271735d489f9SAlex Elder ceph_decode_32_safe(&p, end, snap_count, out); 271835d489f9SAlex Elder 271935d489f9SAlex Elder /* 272035d489f9SAlex Elder * Make sure the reported number of snapshot ids wouldn't go 272135d489f9SAlex Elder * beyond the end of our buffer. But before checking that, 272235d489f9SAlex Elder * make sure the computed size of the snapshot context we 272335d489f9SAlex Elder * allocate is representable in a size_t. 272435d489f9SAlex Elder */ 272535d489f9SAlex Elder if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) 272635d489f9SAlex Elder / sizeof (u64)) { 272735d489f9SAlex Elder ret = -EINVAL; 272835d489f9SAlex Elder goto out; 272935d489f9SAlex Elder } 273035d489f9SAlex Elder if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) 273135d489f9SAlex Elder goto out; 273235d489f9SAlex Elder 273335d489f9SAlex Elder size = sizeof (struct ceph_snap_context) + 273435d489f9SAlex Elder snap_count * sizeof (snapc->snaps[0]); 273535d489f9SAlex Elder snapc = kmalloc(size, GFP_KERNEL); 273635d489f9SAlex Elder if (!snapc) { 273735d489f9SAlex Elder ret = -ENOMEM; 273835d489f9SAlex Elder goto out; 273935d489f9SAlex Elder } 274035d489f9SAlex Elder 274135d489f9SAlex Elder atomic_set(&snapc->nref, 1); 274235d489f9SAlex Elder snapc->seq = seq; 274335d489f9SAlex Elder snapc->num_snaps = snap_count; 274435d489f9SAlex Elder for (i = 0; i < snap_count; i++) 274535d489f9SAlex Elder snapc->snaps[i] = ceph_decode_64(&p); 274635d489f9SAlex Elder 274735d489f9SAlex Elder rbd_dev->header.snapc = snapc; 274835d489f9SAlex Elder 274935d489f9SAlex Elder dout(" snap context seq = %llu, snap_count = %u\n", 275035d489f9SAlex Elder (unsigned long long) seq, (unsigned int) snap_count); 275135d489f9SAlex Elder 275235d489f9SAlex Elder out: 275335d489f9SAlex Elder kfree(reply_buf); 275435d489f9SAlex Elder 275535d489f9SAlex Elder return 0; 275635d489f9SAlex Elder } 275735d489f9SAlex Elder 2758b8b1e2dbSAlex Elder static char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, u32 which) 2759b8b1e2dbSAlex Elder { 2760b8b1e2dbSAlex Elder size_t size; 2761b8b1e2dbSAlex Elder void *reply_buf; 2762b8b1e2dbSAlex Elder __le64 snap_id; 2763b8b1e2dbSAlex Elder int ret; 2764b8b1e2dbSAlex Elder void *p; 2765b8b1e2dbSAlex Elder void *end; 2766b8b1e2dbSAlex Elder char *snap_name; 2767b8b1e2dbSAlex Elder 2768b8b1e2dbSAlex Elder size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; 2769b8b1e2dbSAlex Elder reply_buf = kmalloc(size, GFP_KERNEL); 2770b8b1e2dbSAlex Elder if (!reply_buf) 2771b8b1e2dbSAlex Elder return ERR_PTR(-ENOMEM); 2772b8b1e2dbSAlex Elder 2773b8b1e2dbSAlex Elder snap_id = cpu_to_le64(rbd_dev->header.snapc->snaps[which]); 2774b8b1e2dbSAlex Elder ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, 2775b8b1e2dbSAlex Elder "rbd", "get_snapshot_name", 2776b8b1e2dbSAlex Elder (char *) &snap_id, sizeof (snap_id), 277707b2391fSAlex Elder reply_buf, size, NULL); 2778b8b1e2dbSAlex Elder dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); 2779b8b1e2dbSAlex Elder if (ret < 0) 2780b8b1e2dbSAlex Elder goto out; 2781b8b1e2dbSAlex Elder 2782b8b1e2dbSAlex Elder p = reply_buf; 2783b8b1e2dbSAlex Elder end = (char *) reply_buf + size; 2784e5c35534SAlex Elder snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); 2785b8b1e2dbSAlex Elder if (IS_ERR(snap_name)) { 2786b8b1e2dbSAlex Elder ret = PTR_ERR(snap_name); 2787b8b1e2dbSAlex Elder goto out; 2788b8b1e2dbSAlex Elder } else { 2789b8b1e2dbSAlex Elder dout(" snap_id 0x%016llx snap_name = %s\n", 2790b8b1e2dbSAlex Elder (unsigned long long) le64_to_cpu(snap_id), snap_name); 2791b8b1e2dbSAlex Elder } 2792b8b1e2dbSAlex Elder kfree(reply_buf); 2793b8b1e2dbSAlex Elder 2794b8b1e2dbSAlex Elder return snap_name; 2795b8b1e2dbSAlex Elder out: 2796b8b1e2dbSAlex Elder kfree(reply_buf); 2797b8b1e2dbSAlex Elder 2798b8b1e2dbSAlex Elder return ERR_PTR(ret); 2799b8b1e2dbSAlex Elder } 2800b8b1e2dbSAlex Elder 2801b8b1e2dbSAlex Elder static char *rbd_dev_v2_snap_info(struct rbd_device *rbd_dev, u32 which, 2802b8b1e2dbSAlex Elder u64 *snap_size, u64 *snap_features) 2803b8b1e2dbSAlex Elder { 2804b8b1e2dbSAlex Elder __le64 snap_id; 2805b8b1e2dbSAlex Elder u8 order; 2806b8b1e2dbSAlex Elder int ret; 2807b8b1e2dbSAlex Elder 2808b8b1e2dbSAlex Elder snap_id = rbd_dev->header.snapc->snaps[which]; 2809b8b1e2dbSAlex Elder ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, &order, snap_size); 2810b8b1e2dbSAlex Elder if (ret) 2811b8b1e2dbSAlex Elder return ERR_PTR(ret); 2812b8b1e2dbSAlex Elder ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, snap_features); 2813b8b1e2dbSAlex Elder if (ret) 2814b8b1e2dbSAlex Elder return ERR_PTR(ret); 2815b8b1e2dbSAlex Elder 2816b8b1e2dbSAlex Elder return rbd_dev_v2_snap_name(rbd_dev, which); 2817b8b1e2dbSAlex Elder } 2818b8b1e2dbSAlex Elder 2819b8b1e2dbSAlex Elder static char *rbd_dev_snap_info(struct rbd_device *rbd_dev, u32 which, 2820b8b1e2dbSAlex Elder u64 *snap_size, u64 *snap_features) 2821b8b1e2dbSAlex Elder { 2822b8b1e2dbSAlex Elder if (rbd_dev->image_format == 1) 2823b8b1e2dbSAlex Elder return rbd_dev_v1_snap_info(rbd_dev, which, 2824b8b1e2dbSAlex Elder snap_size, snap_features); 2825b8b1e2dbSAlex Elder if (rbd_dev->image_format == 2) 2826b8b1e2dbSAlex Elder return rbd_dev_v2_snap_info(rbd_dev, which, 2827b8b1e2dbSAlex Elder snap_size, snap_features); 2828b8b1e2dbSAlex Elder return ERR_PTR(-EINVAL); 2829b8b1e2dbSAlex Elder } 2830b8b1e2dbSAlex Elder 2831117973fbSAlex Elder static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver) 2832117973fbSAlex Elder { 2833117973fbSAlex Elder int ret; 2834117973fbSAlex Elder __u8 obj_order; 2835117973fbSAlex Elder 2836117973fbSAlex Elder down_write(&rbd_dev->header_rwsem); 2837117973fbSAlex Elder 2838117973fbSAlex Elder /* Grab old order first, to see if it changes */ 2839117973fbSAlex Elder 2840117973fbSAlex Elder obj_order = rbd_dev->header.obj_order, 2841117973fbSAlex Elder ret = rbd_dev_v2_image_size(rbd_dev); 2842117973fbSAlex Elder if (ret) 2843117973fbSAlex Elder goto out; 2844117973fbSAlex Elder if (rbd_dev->header.obj_order != obj_order) { 2845117973fbSAlex Elder ret = -EIO; 2846117973fbSAlex Elder goto out; 2847117973fbSAlex Elder } 2848117973fbSAlex Elder rbd_update_mapping_size(rbd_dev); 2849117973fbSAlex Elder 2850117973fbSAlex Elder ret = rbd_dev_v2_snap_context(rbd_dev, hver); 2851117973fbSAlex Elder dout("rbd_dev_v2_snap_context returned %d\n", ret); 2852117973fbSAlex Elder if (ret) 2853117973fbSAlex Elder goto out; 2854117973fbSAlex Elder ret = rbd_dev_snaps_update(rbd_dev); 2855117973fbSAlex Elder dout("rbd_dev_snaps_update returned %d\n", ret); 2856117973fbSAlex Elder if (ret) 2857117973fbSAlex Elder goto out; 2858117973fbSAlex Elder ret = rbd_dev_snaps_register(rbd_dev); 2859117973fbSAlex Elder dout("rbd_dev_snaps_register returned %d\n", ret); 2860117973fbSAlex Elder out: 2861117973fbSAlex Elder up_write(&rbd_dev->header_rwsem); 2862117973fbSAlex Elder 2863117973fbSAlex Elder return ret; 2864117973fbSAlex Elder } 2865117973fbSAlex Elder 28669d475de5SAlex Elder /* 286735938150SAlex Elder * Scan the rbd device's current snapshot list and compare it to the 286835938150SAlex Elder * newly-received snapshot context. Remove any existing snapshots 286935938150SAlex Elder * not present in the new snapshot context. Add a new snapshot for 287035938150SAlex Elder * any snaphots in the snapshot context not in the current list. 287135938150SAlex Elder * And verify there are no changes to snapshots we already know 287235938150SAlex Elder * about. 287335938150SAlex Elder * 287435938150SAlex Elder * Assumes the snapshots in the snapshot context are sorted by 287535938150SAlex Elder * snapshot id, highest id first. (Snapshots in the rbd_dev's list 287635938150SAlex Elder * are also maintained in that order.) 2877dfc5606dSYehuda Sadeh */ 2878304f6808SAlex Elder static int rbd_dev_snaps_update(struct rbd_device *rbd_dev) 2879dfc5606dSYehuda Sadeh { 288035938150SAlex Elder struct ceph_snap_context *snapc = rbd_dev->header.snapc; 288135938150SAlex Elder const u32 snap_count = snapc->num_snaps; 288235938150SAlex Elder struct list_head *head = &rbd_dev->snaps; 288335938150SAlex Elder struct list_head *links = head->next; 288435938150SAlex Elder u32 index = 0; 2885dfc5606dSYehuda Sadeh 28869fcbb800SAlex Elder dout("%s: snap count is %u\n", __func__, (unsigned int) snap_count); 288735938150SAlex Elder while (index < snap_count || links != head) { 288835938150SAlex Elder u64 snap_id; 288935938150SAlex Elder struct rbd_snap *snap; 2890cd892126SAlex Elder char *snap_name; 2891cd892126SAlex Elder u64 snap_size = 0; 2892cd892126SAlex Elder u64 snap_features = 0; 2893dfc5606dSYehuda Sadeh 289435938150SAlex Elder snap_id = index < snap_count ? snapc->snaps[index] 289535938150SAlex Elder : CEPH_NOSNAP; 289635938150SAlex Elder snap = links != head ? list_entry(links, struct rbd_snap, node) 289735938150SAlex Elder : NULL; 2898aafb230eSAlex Elder rbd_assert(!snap || snap->id != CEPH_NOSNAP); 2899dfc5606dSYehuda Sadeh 290035938150SAlex Elder if (snap_id == CEPH_NOSNAP || (snap && snap->id > snap_id)) { 290135938150SAlex Elder struct list_head *next = links->next; 2902dfc5606dSYehuda Sadeh 290335938150SAlex Elder /* Existing snapshot not in the new snap context */ 2904dfc5606dSYehuda Sadeh 29050d7dbfceSAlex Elder if (rbd_dev->spec->snap_id == snap->id) 2906d78b650aSAlex Elder atomic_set(&rbd_dev->exists, 0); 290741f38c2bSAlex Elder rbd_remove_snap_dev(snap); 29089fcbb800SAlex Elder dout("%ssnap id %llu has been removed\n", 29090d7dbfceSAlex Elder rbd_dev->spec->snap_id == snap->id ? 29100d7dbfceSAlex Elder "mapped " : "", 29119fcbb800SAlex Elder (unsigned long long) snap->id); 2912dfc5606dSYehuda Sadeh 291335938150SAlex Elder /* Done with this list entry; advance */ 291435938150SAlex Elder 291535938150SAlex Elder links = next; 291635938150SAlex Elder continue; 2917dfc5606dSYehuda Sadeh } 291835938150SAlex Elder 2919b8b1e2dbSAlex Elder snap_name = rbd_dev_snap_info(rbd_dev, index, 2920cd892126SAlex Elder &snap_size, &snap_features); 2921cd892126SAlex Elder if (IS_ERR(snap_name)) 2922cd892126SAlex Elder return PTR_ERR(snap_name); 2923cd892126SAlex Elder 29249fcbb800SAlex Elder dout("entry %u: snap_id = %llu\n", (unsigned int) snap_count, 29259fcbb800SAlex Elder (unsigned long long) snap_id); 292635938150SAlex Elder if (!snap || (snap_id != CEPH_NOSNAP && snap->id < snap_id)) { 292735938150SAlex Elder struct rbd_snap *new_snap; 292835938150SAlex Elder 292935938150SAlex Elder /* We haven't seen this snapshot before */ 293035938150SAlex Elder 2931c8d18425SAlex Elder new_snap = __rbd_add_snap_dev(rbd_dev, snap_name, 2932cd892126SAlex Elder snap_id, snap_size, snap_features); 29339fcbb800SAlex Elder if (IS_ERR(new_snap)) { 29349fcbb800SAlex Elder int err = PTR_ERR(new_snap); 29359fcbb800SAlex Elder 29369fcbb800SAlex Elder dout(" failed to add dev, error %d\n", err); 29379fcbb800SAlex Elder 29389fcbb800SAlex Elder return err; 29399fcbb800SAlex Elder } 294035938150SAlex Elder 294135938150SAlex Elder /* New goes before existing, or at end of list */ 294235938150SAlex Elder 29439fcbb800SAlex Elder dout(" added dev%s\n", snap ? "" : " at end\n"); 294435938150SAlex Elder if (snap) 294535938150SAlex Elder list_add_tail(&new_snap->node, &snap->node); 294635938150SAlex Elder else 2947523f3258SAlex Elder list_add_tail(&new_snap->node, head); 294835938150SAlex Elder } else { 294935938150SAlex Elder /* Already have this one */ 295035938150SAlex Elder 29519fcbb800SAlex Elder dout(" already present\n"); 29529fcbb800SAlex Elder 2953cd892126SAlex Elder rbd_assert(snap->size == snap_size); 2954aafb230eSAlex Elder rbd_assert(!strcmp(snap->name, snap_name)); 2955cd892126SAlex Elder rbd_assert(snap->features == snap_features); 295635938150SAlex Elder 295735938150SAlex Elder /* Done with this list entry; advance */ 295835938150SAlex Elder 295935938150SAlex Elder links = links->next; 2960dfc5606dSYehuda Sadeh } 296135938150SAlex Elder 296235938150SAlex Elder /* Advance to the next entry in the snapshot context */ 296335938150SAlex Elder 296435938150SAlex Elder index++; 2965dfc5606dSYehuda Sadeh } 29669fcbb800SAlex Elder dout("%s: done\n", __func__); 2967dfc5606dSYehuda Sadeh 2968dfc5606dSYehuda Sadeh return 0; 2969dfc5606dSYehuda Sadeh } 2970dfc5606dSYehuda Sadeh 2971304f6808SAlex Elder /* 2972304f6808SAlex Elder * Scan the list of snapshots and register the devices for any that 2973304f6808SAlex Elder * have not already been registered. 2974304f6808SAlex Elder */ 2975304f6808SAlex Elder static int rbd_dev_snaps_register(struct rbd_device *rbd_dev) 2976304f6808SAlex Elder { 2977304f6808SAlex Elder struct rbd_snap *snap; 2978304f6808SAlex Elder int ret = 0; 2979304f6808SAlex Elder 2980304f6808SAlex Elder dout("%s called\n", __func__); 298186ff77bbSAlex Elder if (WARN_ON(!device_is_registered(&rbd_dev->dev))) 298286ff77bbSAlex Elder return -EIO; 2983304f6808SAlex Elder 2984304f6808SAlex Elder list_for_each_entry(snap, &rbd_dev->snaps, node) { 2985304f6808SAlex Elder if (!rbd_snap_registered(snap)) { 2986304f6808SAlex Elder ret = rbd_register_snap_dev(snap, &rbd_dev->dev); 2987304f6808SAlex Elder if (ret < 0) 2988304f6808SAlex Elder break; 2989304f6808SAlex Elder } 2990304f6808SAlex Elder } 2991304f6808SAlex Elder dout("%s: returning %d\n", __func__, ret); 2992304f6808SAlex Elder 2993304f6808SAlex Elder return ret; 2994304f6808SAlex Elder } 2995304f6808SAlex Elder 2996dfc5606dSYehuda Sadeh static int rbd_bus_add_dev(struct rbd_device *rbd_dev) 2997dfc5606dSYehuda Sadeh { 2998dfc5606dSYehuda Sadeh struct device *dev; 2999cd789ab9SAlex Elder int ret; 3000dfc5606dSYehuda Sadeh 3001dfc5606dSYehuda Sadeh mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); 3002dfc5606dSYehuda Sadeh 3003cd789ab9SAlex Elder dev = &rbd_dev->dev; 3004dfc5606dSYehuda Sadeh dev->bus = &rbd_bus_type; 3005dfc5606dSYehuda Sadeh dev->type = &rbd_device_type; 3006dfc5606dSYehuda Sadeh dev->parent = &rbd_root_dev; 3007dfc5606dSYehuda Sadeh dev->release = rbd_dev_release; 3008de71a297SAlex Elder dev_set_name(dev, "%d", rbd_dev->dev_id); 3009dfc5606dSYehuda Sadeh ret = device_register(dev); 3010dfc5606dSYehuda Sadeh 3011dfc5606dSYehuda Sadeh mutex_unlock(&ctl_mutex); 3012cd789ab9SAlex Elder 3013dfc5606dSYehuda Sadeh return ret; 3014602adf40SYehuda Sadeh } 3015602adf40SYehuda Sadeh 3016dfc5606dSYehuda Sadeh static void rbd_bus_del_dev(struct rbd_device *rbd_dev) 3017dfc5606dSYehuda Sadeh { 3018dfc5606dSYehuda Sadeh device_unregister(&rbd_dev->dev); 3019dfc5606dSYehuda Sadeh } 3020dfc5606dSYehuda Sadeh 302159c2be1eSYehuda Sadeh static int rbd_init_watch_dev(struct rbd_device *rbd_dev) 302259c2be1eSYehuda Sadeh { 302359c2be1eSYehuda Sadeh int ret, rc; 302459c2be1eSYehuda Sadeh 302559c2be1eSYehuda Sadeh do { 30260e6f322dSAlex Elder ret = rbd_req_sync_watch(rbd_dev); 302759c2be1eSYehuda Sadeh if (ret == -ERANGE) { 3028117973fbSAlex Elder rc = rbd_dev_refresh(rbd_dev, NULL); 302959c2be1eSYehuda Sadeh if (rc < 0) 303059c2be1eSYehuda Sadeh return rc; 303159c2be1eSYehuda Sadeh } 303259c2be1eSYehuda Sadeh } while (ret == -ERANGE); 303359c2be1eSYehuda Sadeh 303459c2be1eSYehuda Sadeh return ret; 303559c2be1eSYehuda Sadeh } 303659c2be1eSYehuda Sadeh 3037e2839308SAlex Elder static atomic64_t rbd_dev_id_max = ATOMIC64_INIT(0); 30381ddbe94eSAlex Elder 30391ddbe94eSAlex Elder /* 3040499afd5bSAlex Elder * Get a unique rbd identifier for the given new rbd_dev, and add 3041499afd5bSAlex Elder * the rbd_dev to the global list. The minimum rbd id is 1. 30421ddbe94eSAlex Elder */ 3043e2839308SAlex Elder static void rbd_dev_id_get(struct rbd_device *rbd_dev) 3044b7f23c36SAlex Elder { 3045e2839308SAlex Elder rbd_dev->dev_id = atomic64_inc_return(&rbd_dev_id_max); 3046499afd5bSAlex Elder 3047499afd5bSAlex Elder spin_lock(&rbd_dev_list_lock); 3048499afd5bSAlex Elder list_add_tail(&rbd_dev->node, &rbd_dev_list); 3049499afd5bSAlex Elder spin_unlock(&rbd_dev_list_lock); 3050e2839308SAlex Elder dout("rbd_dev %p given dev id %llu\n", rbd_dev, 3051e2839308SAlex Elder (unsigned long long) rbd_dev->dev_id); 3052b7f23c36SAlex Elder } 3053b7f23c36SAlex Elder 30541ddbe94eSAlex Elder /* 3055499afd5bSAlex Elder * Remove an rbd_dev from the global list, and record that its 3056499afd5bSAlex Elder * identifier is no longer in use. 30571ddbe94eSAlex Elder */ 3058e2839308SAlex Elder static void rbd_dev_id_put(struct rbd_device *rbd_dev) 30591ddbe94eSAlex Elder { 3060d184f6bfSAlex Elder struct list_head *tmp; 3061de71a297SAlex Elder int rbd_id = rbd_dev->dev_id; 3062d184f6bfSAlex Elder int max_id; 3063d184f6bfSAlex Elder 3064aafb230eSAlex Elder rbd_assert(rbd_id > 0); 3065499afd5bSAlex Elder 3066e2839308SAlex Elder dout("rbd_dev %p released dev id %llu\n", rbd_dev, 3067e2839308SAlex Elder (unsigned long long) rbd_dev->dev_id); 3068499afd5bSAlex Elder spin_lock(&rbd_dev_list_lock); 3069499afd5bSAlex Elder list_del_init(&rbd_dev->node); 3070d184f6bfSAlex Elder 3071d184f6bfSAlex Elder /* 3072d184f6bfSAlex Elder * If the id being "put" is not the current maximum, there 3073d184f6bfSAlex Elder * is nothing special we need to do. 3074d184f6bfSAlex Elder */ 3075e2839308SAlex Elder if (rbd_id != atomic64_read(&rbd_dev_id_max)) { 3076d184f6bfSAlex Elder spin_unlock(&rbd_dev_list_lock); 3077d184f6bfSAlex Elder return; 3078d184f6bfSAlex Elder } 3079d184f6bfSAlex Elder 3080d184f6bfSAlex Elder /* 3081d184f6bfSAlex Elder * We need to update the current maximum id. Search the 3082d184f6bfSAlex Elder * list to find out what it is. We're more likely to find 3083d184f6bfSAlex Elder * the maximum at the end, so search the list backward. 3084d184f6bfSAlex Elder */ 3085d184f6bfSAlex Elder max_id = 0; 3086d184f6bfSAlex Elder list_for_each_prev(tmp, &rbd_dev_list) { 3087d184f6bfSAlex Elder struct rbd_device *rbd_dev; 3088d184f6bfSAlex Elder 3089d184f6bfSAlex Elder rbd_dev = list_entry(tmp, struct rbd_device, node); 3090b213e0b1SAlex Elder if (rbd_dev->dev_id > max_id) 3091b213e0b1SAlex Elder max_id = rbd_dev->dev_id; 3092d184f6bfSAlex Elder } 3093499afd5bSAlex Elder spin_unlock(&rbd_dev_list_lock); 30941ddbe94eSAlex Elder 30951ddbe94eSAlex Elder /* 3096e2839308SAlex Elder * The max id could have been updated by rbd_dev_id_get(), in 3097d184f6bfSAlex Elder * which case it now accurately reflects the new maximum. 3098d184f6bfSAlex Elder * Be careful not to overwrite the maximum value in that 3099d184f6bfSAlex Elder * case. 31001ddbe94eSAlex Elder */ 3101e2839308SAlex Elder atomic64_cmpxchg(&rbd_dev_id_max, rbd_id, max_id); 3102e2839308SAlex Elder dout(" max dev id has been reset\n"); 3103b7f23c36SAlex Elder } 3104b7f23c36SAlex Elder 3105a725f65eSAlex Elder /* 3106e28fff26SAlex Elder * Skips over white space at *buf, and updates *buf to point to the 3107e28fff26SAlex Elder * first found non-space character (if any). Returns the length of 3108593a9e7bSAlex Elder * the token (string of non-white space characters) found. Note 3109593a9e7bSAlex Elder * that *buf must be terminated with '\0'. 3110e28fff26SAlex Elder */ 3111e28fff26SAlex Elder static inline size_t next_token(const char **buf) 3112e28fff26SAlex Elder { 3113e28fff26SAlex Elder /* 3114e28fff26SAlex Elder * These are the characters that produce nonzero for 3115e28fff26SAlex Elder * isspace() in the "C" and "POSIX" locales. 3116e28fff26SAlex Elder */ 3117e28fff26SAlex Elder const char *spaces = " \f\n\r\t\v"; 3118e28fff26SAlex Elder 3119e28fff26SAlex Elder *buf += strspn(*buf, spaces); /* Find start of token */ 3120e28fff26SAlex Elder 3121e28fff26SAlex Elder return strcspn(*buf, spaces); /* Return token length */ 3122e28fff26SAlex Elder } 3123e28fff26SAlex Elder 3124e28fff26SAlex Elder /* 3125e28fff26SAlex Elder * Finds the next token in *buf, and if the provided token buffer is 3126e28fff26SAlex Elder * big enough, copies the found token into it. The result, if 3127593a9e7bSAlex Elder * copied, is guaranteed to be terminated with '\0'. Note that *buf 3128593a9e7bSAlex Elder * must be terminated with '\0' on entry. 3129e28fff26SAlex Elder * 3130e28fff26SAlex Elder * Returns the length of the token found (not including the '\0'). 3131e28fff26SAlex Elder * Return value will be 0 if no token is found, and it will be >= 3132e28fff26SAlex Elder * token_size if the token would not fit. 3133e28fff26SAlex Elder * 3134593a9e7bSAlex Elder * The *buf pointer will be updated to point beyond the end of the 3135e28fff26SAlex Elder * found token. Note that this occurs even if the token buffer is 3136e28fff26SAlex Elder * too small to hold it. 3137e28fff26SAlex Elder */ 3138e28fff26SAlex Elder static inline size_t copy_token(const char **buf, 3139e28fff26SAlex Elder char *token, 3140e28fff26SAlex Elder size_t token_size) 3141e28fff26SAlex Elder { 3142e28fff26SAlex Elder size_t len; 3143e28fff26SAlex Elder 3144e28fff26SAlex Elder len = next_token(buf); 3145e28fff26SAlex Elder if (len < token_size) { 3146e28fff26SAlex Elder memcpy(token, *buf, len); 3147e28fff26SAlex Elder *(token + len) = '\0'; 3148e28fff26SAlex Elder } 3149e28fff26SAlex Elder *buf += len; 3150e28fff26SAlex Elder 3151e28fff26SAlex Elder return len; 3152e28fff26SAlex Elder } 3153e28fff26SAlex Elder 3154e28fff26SAlex Elder /* 3155ea3352f4SAlex Elder * Finds the next token in *buf, dynamically allocates a buffer big 3156ea3352f4SAlex Elder * enough to hold a copy of it, and copies the token into the new 3157ea3352f4SAlex Elder * buffer. The copy is guaranteed to be terminated with '\0'. Note 3158ea3352f4SAlex Elder * that a duplicate buffer is created even for a zero-length token. 3159ea3352f4SAlex Elder * 3160ea3352f4SAlex Elder * Returns a pointer to the newly-allocated duplicate, or a null 3161ea3352f4SAlex Elder * pointer if memory for the duplicate was not available. If 3162ea3352f4SAlex Elder * the lenp argument is a non-null pointer, the length of the token 3163ea3352f4SAlex Elder * (not including the '\0') is returned in *lenp. 3164ea3352f4SAlex Elder * 3165ea3352f4SAlex Elder * If successful, the *buf pointer will be updated to point beyond 3166ea3352f4SAlex Elder * the end of the found token. 3167ea3352f4SAlex Elder * 3168ea3352f4SAlex Elder * Note: uses GFP_KERNEL for allocation. 3169ea3352f4SAlex Elder */ 3170ea3352f4SAlex Elder static inline char *dup_token(const char **buf, size_t *lenp) 3171ea3352f4SAlex Elder { 3172ea3352f4SAlex Elder char *dup; 3173ea3352f4SAlex Elder size_t len; 3174ea3352f4SAlex Elder 3175ea3352f4SAlex Elder len = next_token(buf); 31764caf35f9SAlex Elder dup = kmemdup(*buf, len + 1, GFP_KERNEL); 3177ea3352f4SAlex Elder if (!dup) 3178ea3352f4SAlex Elder return NULL; 3179ea3352f4SAlex Elder *(dup + len) = '\0'; 3180ea3352f4SAlex Elder *buf += len; 3181ea3352f4SAlex Elder 3182ea3352f4SAlex Elder if (lenp) 3183ea3352f4SAlex Elder *lenp = len; 3184ea3352f4SAlex Elder 3185ea3352f4SAlex Elder return dup; 3186ea3352f4SAlex Elder } 3187ea3352f4SAlex Elder 3188ea3352f4SAlex Elder /* 3189859c31dfSAlex Elder * Parse the options provided for an "rbd add" (i.e., rbd image 3190859c31dfSAlex Elder * mapping) request. These arrive via a write to /sys/bus/rbd/add, 3191859c31dfSAlex Elder * and the data written is passed here via a NUL-terminated buffer. 3192859c31dfSAlex Elder * Returns 0 if successful or an error code otherwise. 3193d22f76e7SAlex Elder * 3194859c31dfSAlex Elder * The information extracted from these options is recorded in 3195859c31dfSAlex Elder * the other parameters which return dynamically-allocated 3196859c31dfSAlex Elder * structures: 3197859c31dfSAlex Elder * ceph_opts 3198859c31dfSAlex Elder * The address of a pointer that will refer to a ceph options 3199859c31dfSAlex Elder * structure. Caller must release the returned pointer using 3200859c31dfSAlex Elder * ceph_destroy_options() when it is no longer needed. 3201859c31dfSAlex Elder * rbd_opts 3202859c31dfSAlex Elder * Address of an rbd options pointer. Fully initialized by 3203859c31dfSAlex Elder * this function; caller must release with kfree(). 3204859c31dfSAlex Elder * spec 3205859c31dfSAlex Elder * Address of an rbd image specification pointer. Fully 3206859c31dfSAlex Elder * initialized by this function based on parsed options. 3207859c31dfSAlex Elder * Caller must release with rbd_spec_put(). 3208859c31dfSAlex Elder * 3209859c31dfSAlex Elder * The options passed take this form: 3210859c31dfSAlex Elder * <mon_addrs> <options> <pool_name> <image_name> [<snap_id>] 3211859c31dfSAlex Elder * where: 3212859c31dfSAlex Elder * <mon_addrs> 3213859c31dfSAlex Elder * A comma-separated list of one or more monitor addresses. 3214859c31dfSAlex Elder * A monitor address is an ip address, optionally followed 3215859c31dfSAlex Elder * by a port number (separated by a colon). 3216859c31dfSAlex Elder * I.e.: ip1[:port1][,ip2[:port2]...] 3217859c31dfSAlex Elder * <options> 3218859c31dfSAlex Elder * A comma-separated list of ceph and/or rbd options. 3219859c31dfSAlex Elder * <pool_name> 3220859c31dfSAlex Elder * The name of the rados pool containing the rbd image. 3221859c31dfSAlex Elder * <image_name> 3222859c31dfSAlex Elder * The name of the image in that pool to map. 3223859c31dfSAlex Elder * <snap_id> 3224859c31dfSAlex Elder * An optional snapshot id. If provided, the mapping will 3225859c31dfSAlex Elder * present data from the image at the time that snapshot was 3226859c31dfSAlex Elder * created. The image head is used if no snapshot id is 3227859c31dfSAlex Elder * provided. Snapshot mappings are always read-only. 3228a725f65eSAlex Elder */ 3229859c31dfSAlex Elder static int rbd_add_parse_args(const char *buf, 3230dc79b113SAlex Elder struct ceph_options **ceph_opts, 3231859c31dfSAlex Elder struct rbd_options **opts, 3232859c31dfSAlex Elder struct rbd_spec **rbd_spec) 3233a725f65eSAlex Elder { 3234e28fff26SAlex Elder size_t len; 3235859c31dfSAlex Elder char *options; 32360ddebc0cSAlex Elder const char *mon_addrs; 32370ddebc0cSAlex Elder size_t mon_addrs_size; 3238859c31dfSAlex Elder struct rbd_spec *spec = NULL; 32394e9afebaSAlex Elder struct rbd_options *rbd_opts = NULL; 3240859c31dfSAlex Elder struct ceph_options *copts; 3241dc79b113SAlex Elder int ret; 3242e28fff26SAlex Elder 3243e28fff26SAlex Elder /* The first four tokens are required */ 3244e28fff26SAlex Elder 32457ef3214aSAlex Elder len = next_token(&buf); 32464fb5d671SAlex Elder if (!len) { 32474fb5d671SAlex Elder rbd_warn(NULL, "no monitor address(es) provided"); 32484fb5d671SAlex Elder return -EINVAL; 32494fb5d671SAlex Elder } 32500ddebc0cSAlex Elder mon_addrs = buf; 3251f28e565aSAlex Elder mon_addrs_size = len + 1; 32527ef3214aSAlex Elder buf += len; 3253a725f65eSAlex Elder 3254dc79b113SAlex Elder ret = -EINVAL; 3255f28e565aSAlex Elder options = dup_token(&buf, NULL); 3256f28e565aSAlex Elder if (!options) 3257dc79b113SAlex Elder return -ENOMEM; 32584fb5d671SAlex Elder if (!*options) { 32594fb5d671SAlex Elder rbd_warn(NULL, "no options provided"); 32604fb5d671SAlex Elder goto out_err; 32614fb5d671SAlex Elder } 3262a725f65eSAlex Elder 3263859c31dfSAlex Elder spec = rbd_spec_alloc(); 3264859c31dfSAlex Elder if (!spec) 3265f28e565aSAlex Elder goto out_mem; 3266859c31dfSAlex Elder 3267859c31dfSAlex Elder spec->pool_name = dup_token(&buf, NULL); 3268859c31dfSAlex Elder if (!spec->pool_name) 3269859c31dfSAlex Elder goto out_mem; 32704fb5d671SAlex Elder if (!*spec->pool_name) { 32714fb5d671SAlex Elder rbd_warn(NULL, "no pool name provided"); 32724fb5d671SAlex Elder goto out_err; 32734fb5d671SAlex Elder } 3274e28fff26SAlex Elder 327569e7a02fSAlex Elder spec->image_name = dup_token(&buf, NULL); 3276859c31dfSAlex Elder if (!spec->image_name) 3277f28e565aSAlex Elder goto out_mem; 32784fb5d671SAlex Elder if (!*spec->image_name) { 32794fb5d671SAlex Elder rbd_warn(NULL, "no image name provided"); 32804fb5d671SAlex Elder goto out_err; 32814fb5d671SAlex Elder } 3282e28fff26SAlex Elder 3283f28e565aSAlex Elder /* 3284f28e565aSAlex Elder * Snapshot name is optional; default is to use "-" 3285f28e565aSAlex Elder * (indicating the head/no snapshot). 3286f28e565aSAlex Elder */ 32873feeb894SAlex Elder len = next_token(&buf); 3288820a5f3eSAlex Elder if (!len) { 32893feeb894SAlex Elder buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ 32903feeb894SAlex Elder len = sizeof (RBD_SNAP_HEAD_NAME) - 1; 3291f28e565aSAlex Elder } else if (len > RBD_MAX_SNAP_NAME_LEN) { 3292dc79b113SAlex Elder ret = -ENAMETOOLONG; 3293f28e565aSAlex Elder goto out_err; 3294849b4260SAlex Elder } 32954caf35f9SAlex Elder spec->snap_name = kmemdup(buf, len + 1, GFP_KERNEL); 3296859c31dfSAlex Elder if (!spec->snap_name) 3297f28e565aSAlex Elder goto out_mem; 3298859c31dfSAlex Elder *(spec->snap_name + len) = '\0'; 3299e5c35534SAlex Elder 33000ddebc0cSAlex Elder /* Initialize all rbd options to the defaults */ 3301e28fff26SAlex Elder 33024e9afebaSAlex Elder rbd_opts = kzalloc(sizeof (*rbd_opts), GFP_KERNEL); 33034e9afebaSAlex Elder if (!rbd_opts) 33044e9afebaSAlex Elder goto out_mem; 33054e9afebaSAlex Elder 33064e9afebaSAlex Elder rbd_opts->read_only = RBD_READ_ONLY_DEFAULT; 3307d22f76e7SAlex Elder 3308859c31dfSAlex Elder copts = ceph_parse_options(options, mon_addrs, 33090ddebc0cSAlex Elder mon_addrs + mon_addrs_size - 1, 33104e9afebaSAlex Elder parse_rbd_opts_token, rbd_opts); 3311859c31dfSAlex Elder if (IS_ERR(copts)) { 3312859c31dfSAlex Elder ret = PTR_ERR(copts); 3313dc79b113SAlex Elder goto out_err; 3314dc79b113SAlex Elder } 3315859c31dfSAlex Elder kfree(options); 3316859c31dfSAlex Elder 3317859c31dfSAlex Elder *ceph_opts = copts; 33184e9afebaSAlex Elder *opts = rbd_opts; 3319859c31dfSAlex Elder *rbd_spec = spec; 33200ddebc0cSAlex Elder 3321dc79b113SAlex Elder return 0; 3322f28e565aSAlex Elder out_mem: 3323dc79b113SAlex Elder ret = -ENOMEM; 3324d22f76e7SAlex Elder out_err: 3325859c31dfSAlex Elder kfree(rbd_opts); 3326859c31dfSAlex Elder rbd_spec_put(spec); 3327f28e565aSAlex Elder kfree(options); 3328d22f76e7SAlex Elder 3329dc79b113SAlex Elder return ret; 3330a725f65eSAlex Elder } 3331a725f65eSAlex Elder 3332589d30e0SAlex Elder /* 3333589d30e0SAlex Elder * An rbd format 2 image has a unique identifier, distinct from the 3334589d30e0SAlex Elder * name given to it by the user. Internally, that identifier is 3335589d30e0SAlex Elder * what's used to specify the names of objects related to the image. 3336589d30e0SAlex Elder * 3337589d30e0SAlex Elder * A special "rbd id" object is used to map an rbd image name to its 3338589d30e0SAlex Elder * id. If that object doesn't exist, then there is no v2 rbd image 3339589d30e0SAlex Elder * with the supplied name. 3340589d30e0SAlex Elder * 3341589d30e0SAlex Elder * This function will record the given rbd_dev's image_id field if 3342589d30e0SAlex Elder * it can be determined, and in that case will return 0. If any 3343589d30e0SAlex Elder * errors occur a negative errno will be returned and the rbd_dev's 3344589d30e0SAlex Elder * image_id field will be unchanged (and should be NULL). 3345589d30e0SAlex Elder */ 3346589d30e0SAlex Elder static int rbd_dev_image_id(struct rbd_device *rbd_dev) 3347589d30e0SAlex Elder { 3348589d30e0SAlex Elder int ret; 3349589d30e0SAlex Elder size_t size; 3350589d30e0SAlex Elder char *object_name; 3351589d30e0SAlex Elder void *response; 3352589d30e0SAlex Elder void *p; 3353589d30e0SAlex Elder 3354589d30e0SAlex Elder /* 33552c0d0a10SAlex Elder * When probing a parent image, the image id is already 33562c0d0a10SAlex Elder * known (and the image name likely is not). There's no 33572c0d0a10SAlex Elder * need to fetch the image id again in this case. 33582c0d0a10SAlex Elder */ 33592c0d0a10SAlex Elder if (rbd_dev->spec->image_id) 33602c0d0a10SAlex Elder return 0; 33612c0d0a10SAlex Elder 33622c0d0a10SAlex Elder /* 3363589d30e0SAlex Elder * First, see if the format 2 image id file exists, and if 3364589d30e0SAlex Elder * so, get the image's persistent id from it. 3365589d30e0SAlex Elder */ 336669e7a02fSAlex Elder size = sizeof (RBD_ID_PREFIX) + strlen(rbd_dev->spec->image_name); 3367589d30e0SAlex Elder object_name = kmalloc(size, GFP_NOIO); 3368589d30e0SAlex Elder if (!object_name) 3369589d30e0SAlex Elder return -ENOMEM; 33700d7dbfceSAlex Elder sprintf(object_name, "%s%s", RBD_ID_PREFIX, rbd_dev->spec->image_name); 3371589d30e0SAlex Elder dout("rbd id object name is %s\n", object_name); 3372589d30e0SAlex Elder 3373589d30e0SAlex Elder /* Response will be an encoded string, which includes a length */ 3374589d30e0SAlex Elder 3375589d30e0SAlex Elder size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; 3376589d30e0SAlex Elder response = kzalloc(size, GFP_NOIO); 3377589d30e0SAlex Elder if (!response) { 3378589d30e0SAlex Elder ret = -ENOMEM; 3379589d30e0SAlex Elder goto out; 3380589d30e0SAlex Elder } 3381589d30e0SAlex Elder 3382589d30e0SAlex Elder ret = rbd_req_sync_exec(rbd_dev, object_name, 3383589d30e0SAlex Elder "rbd", "get_id", 3384589d30e0SAlex Elder NULL, 0, 338507b2391fSAlex Elder response, RBD_IMAGE_ID_LEN_MAX, NULL); 3386589d30e0SAlex Elder dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); 3387589d30e0SAlex Elder if (ret < 0) 3388589d30e0SAlex Elder goto out; 3389a0ea3a40SAlex Elder ret = 0; /* rbd_req_sync_exec() can return positive */ 3390589d30e0SAlex Elder 3391589d30e0SAlex Elder p = response; 33920d7dbfceSAlex Elder rbd_dev->spec->image_id = ceph_extract_encoded_string(&p, 3393589d30e0SAlex Elder p + RBD_IMAGE_ID_LEN_MAX, 3394979ed480SAlex Elder NULL, GFP_NOIO); 33950d7dbfceSAlex Elder if (IS_ERR(rbd_dev->spec->image_id)) { 33960d7dbfceSAlex Elder ret = PTR_ERR(rbd_dev->spec->image_id); 33970d7dbfceSAlex Elder rbd_dev->spec->image_id = NULL; 3398589d30e0SAlex Elder } else { 33990d7dbfceSAlex Elder dout("image_id is %s\n", rbd_dev->spec->image_id); 3400589d30e0SAlex Elder } 3401589d30e0SAlex Elder out: 3402589d30e0SAlex Elder kfree(response); 3403589d30e0SAlex Elder kfree(object_name); 3404589d30e0SAlex Elder 3405589d30e0SAlex Elder return ret; 3406589d30e0SAlex Elder } 3407589d30e0SAlex Elder 3408a30b71b9SAlex Elder static int rbd_dev_v1_probe(struct rbd_device *rbd_dev) 3409a30b71b9SAlex Elder { 3410a30b71b9SAlex Elder int ret; 3411a30b71b9SAlex Elder size_t size; 3412a30b71b9SAlex Elder 3413a30b71b9SAlex Elder /* Version 1 images have no id; empty string is used */ 3414a30b71b9SAlex Elder 34150d7dbfceSAlex Elder rbd_dev->spec->image_id = kstrdup("", GFP_KERNEL); 34160d7dbfceSAlex Elder if (!rbd_dev->spec->image_id) 3417a30b71b9SAlex Elder return -ENOMEM; 3418a30b71b9SAlex Elder 3419a30b71b9SAlex Elder /* Record the header object name for this rbd image. */ 3420a30b71b9SAlex Elder 342169e7a02fSAlex Elder size = strlen(rbd_dev->spec->image_name) + sizeof (RBD_SUFFIX); 3422a30b71b9SAlex Elder rbd_dev->header_name = kmalloc(size, GFP_KERNEL); 3423a30b71b9SAlex Elder if (!rbd_dev->header_name) { 3424a30b71b9SAlex Elder ret = -ENOMEM; 3425a30b71b9SAlex Elder goto out_err; 3426a30b71b9SAlex Elder } 34270d7dbfceSAlex Elder sprintf(rbd_dev->header_name, "%s%s", 34280d7dbfceSAlex Elder rbd_dev->spec->image_name, RBD_SUFFIX); 3429a30b71b9SAlex Elder 3430a30b71b9SAlex Elder /* Populate rbd image metadata */ 3431a30b71b9SAlex Elder 3432a30b71b9SAlex Elder ret = rbd_read_header(rbd_dev, &rbd_dev->header); 3433a30b71b9SAlex Elder if (ret < 0) 3434a30b71b9SAlex Elder goto out_err; 343586b00e0dSAlex Elder 343686b00e0dSAlex Elder /* Version 1 images have no parent (no layering) */ 343786b00e0dSAlex Elder 343886b00e0dSAlex Elder rbd_dev->parent_spec = NULL; 343986b00e0dSAlex Elder rbd_dev->parent_overlap = 0; 344086b00e0dSAlex Elder 3441a30b71b9SAlex Elder rbd_dev->image_format = 1; 3442a30b71b9SAlex Elder 3443a30b71b9SAlex Elder dout("discovered version 1 image, header name is %s\n", 3444a30b71b9SAlex Elder rbd_dev->header_name); 3445a30b71b9SAlex Elder 3446a30b71b9SAlex Elder return 0; 3447a30b71b9SAlex Elder 3448a30b71b9SAlex Elder out_err: 3449a30b71b9SAlex Elder kfree(rbd_dev->header_name); 3450a30b71b9SAlex Elder rbd_dev->header_name = NULL; 34510d7dbfceSAlex Elder kfree(rbd_dev->spec->image_id); 34520d7dbfceSAlex Elder rbd_dev->spec->image_id = NULL; 3453a30b71b9SAlex Elder 3454a30b71b9SAlex Elder return ret; 3455a30b71b9SAlex Elder } 3456a30b71b9SAlex Elder 3457a30b71b9SAlex Elder static int rbd_dev_v2_probe(struct rbd_device *rbd_dev) 3458a30b71b9SAlex Elder { 3459a30b71b9SAlex Elder size_t size; 34609d475de5SAlex Elder int ret; 34616e14b1a6SAlex Elder u64 ver = 0; 3462a30b71b9SAlex Elder 3463a30b71b9SAlex Elder /* 3464a30b71b9SAlex Elder * Image id was filled in by the caller. Record the header 3465a30b71b9SAlex Elder * object name for this rbd image. 3466a30b71b9SAlex Elder */ 3467979ed480SAlex Elder size = sizeof (RBD_HEADER_PREFIX) + strlen(rbd_dev->spec->image_id); 3468a30b71b9SAlex Elder rbd_dev->header_name = kmalloc(size, GFP_KERNEL); 3469a30b71b9SAlex Elder if (!rbd_dev->header_name) 3470a30b71b9SAlex Elder return -ENOMEM; 3471a30b71b9SAlex Elder sprintf(rbd_dev->header_name, "%s%s", 34720d7dbfceSAlex Elder RBD_HEADER_PREFIX, rbd_dev->spec->image_id); 34739d475de5SAlex Elder 34749d475de5SAlex Elder /* Get the size and object order for the image */ 34759d475de5SAlex Elder 34769d475de5SAlex Elder ret = rbd_dev_v2_image_size(rbd_dev); 34779d475de5SAlex Elder if (ret < 0) 34789d475de5SAlex Elder goto out_err; 34791e130199SAlex Elder 34801e130199SAlex Elder /* Get the object prefix (a.k.a. block_name) for the image */ 34811e130199SAlex Elder 34821e130199SAlex Elder ret = rbd_dev_v2_object_prefix(rbd_dev); 34831e130199SAlex Elder if (ret < 0) 34841e130199SAlex Elder goto out_err; 3485b1b5402aSAlex Elder 3486d889140cSAlex Elder /* Get the and check features for the image */ 3487b1b5402aSAlex Elder 3488b1b5402aSAlex Elder ret = rbd_dev_v2_features(rbd_dev); 3489b1b5402aSAlex Elder if (ret < 0) 3490b1b5402aSAlex Elder goto out_err; 349135d489f9SAlex Elder 349286b00e0dSAlex Elder /* If the image supports layering, get the parent info */ 349386b00e0dSAlex Elder 349486b00e0dSAlex Elder if (rbd_dev->header.features & RBD_FEATURE_LAYERING) { 349586b00e0dSAlex Elder ret = rbd_dev_v2_parent_info(rbd_dev); 349686b00e0dSAlex Elder if (ret < 0) 349786b00e0dSAlex Elder goto out_err; 349886b00e0dSAlex Elder } 349986b00e0dSAlex Elder 35006e14b1a6SAlex Elder /* crypto and compression type aren't (yet) supported for v2 images */ 350135d489f9SAlex Elder 35026e14b1a6SAlex Elder rbd_dev->header.crypt_type = 0; 35036e14b1a6SAlex Elder rbd_dev->header.comp_type = 0; 35046e14b1a6SAlex Elder 35056e14b1a6SAlex Elder /* Get the snapshot context, plus the header version */ 35066e14b1a6SAlex Elder 35076e14b1a6SAlex Elder ret = rbd_dev_v2_snap_context(rbd_dev, &ver); 350835d489f9SAlex Elder if (ret) 350935d489f9SAlex Elder goto out_err; 35106e14b1a6SAlex Elder rbd_dev->header.obj_version = ver; 35116e14b1a6SAlex Elder 3512a30b71b9SAlex Elder rbd_dev->image_format = 2; 3513a30b71b9SAlex Elder 3514a30b71b9SAlex Elder dout("discovered version 2 image, header name is %s\n", 3515a30b71b9SAlex Elder rbd_dev->header_name); 3516a30b71b9SAlex Elder 351735152979SAlex Elder return 0; 35189d475de5SAlex Elder out_err: 351986b00e0dSAlex Elder rbd_dev->parent_overlap = 0; 352086b00e0dSAlex Elder rbd_spec_put(rbd_dev->parent_spec); 352186b00e0dSAlex Elder rbd_dev->parent_spec = NULL; 35229d475de5SAlex Elder kfree(rbd_dev->header_name); 35239d475de5SAlex Elder rbd_dev->header_name = NULL; 35241e130199SAlex Elder kfree(rbd_dev->header.object_prefix); 35251e130199SAlex Elder rbd_dev->header.object_prefix = NULL; 35269d475de5SAlex Elder 35279d475de5SAlex Elder return ret; 3528a30b71b9SAlex Elder } 3529a30b71b9SAlex Elder 353083a06263SAlex Elder static int rbd_dev_probe_finish(struct rbd_device *rbd_dev) 353183a06263SAlex Elder { 353283a06263SAlex Elder int ret; 353383a06263SAlex Elder 353483a06263SAlex Elder /* no need to lock here, as rbd_dev is not registered yet */ 353583a06263SAlex Elder ret = rbd_dev_snaps_update(rbd_dev); 353683a06263SAlex Elder if (ret) 353783a06263SAlex Elder return ret; 353883a06263SAlex Elder 35399e15b77dSAlex Elder ret = rbd_dev_probe_update_spec(rbd_dev); 35409e15b77dSAlex Elder if (ret) 35419e15b77dSAlex Elder goto err_out_snaps; 35429e15b77dSAlex Elder 354383a06263SAlex Elder ret = rbd_dev_set_mapping(rbd_dev); 354483a06263SAlex Elder if (ret) 354583a06263SAlex Elder goto err_out_snaps; 354683a06263SAlex Elder 354783a06263SAlex Elder /* generate unique id: find highest unique id, add one */ 354883a06263SAlex Elder rbd_dev_id_get(rbd_dev); 354983a06263SAlex Elder 355083a06263SAlex Elder /* Fill in the device name, now that we have its id. */ 355183a06263SAlex Elder BUILD_BUG_ON(DEV_NAME_LEN 355283a06263SAlex Elder < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); 355383a06263SAlex Elder sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); 355483a06263SAlex Elder 355583a06263SAlex Elder /* Get our block major device number. */ 355683a06263SAlex Elder 355783a06263SAlex Elder ret = register_blkdev(0, rbd_dev->name); 355883a06263SAlex Elder if (ret < 0) 355983a06263SAlex Elder goto err_out_id; 356083a06263SAlex Elder rbd_dev->major = ret; 356183a06263SAlex Elder 356283a06263SAlex Elder /* Set up the blkdev mapping. */ 356383a06263SAlex Elder 356483a06263SAlex Elder ret = rbd_init_disk(rbd_dev); 356583a06263SAlex Elder if (ret) 356683a06263SAlex Elder goto err_out_blkdev; 356783a06263SAlex Elder 356883a06263SAlex Elder ret = rbd_bus_add_dev(rbd_dev); 356983a06263SAlex Elder if (ret) 357083a06263SAlex Elder goto err_out_disk; 357183a06263SAlex Elder 357283a06263SAlex Elder /* 357383a06263SAlex Elder * At this point cleanup in the event of an error is the job 357483a06263SAlex Elder * of the sysfs code (initiated by rbd_bus_del_dev()). 357583a06263SAlex Elder */ 357683a06263SAlex Elder down_write(&rbd_dev->header_rwsem); 357783a06263SAlex Elder ret = rbd_dev_snaps_register(rbd_dev); 357883a06263SAlex Elder up_write(&rbd_dev->header_rwsem); 357983a06263SAlex Elder if (ret) 358083a06263SAlex Elder goto err_out_bus; 358183a06263SAlex Elder 358283a06263SAlex Elder ret = rbd_init_watch_dev(rbd_dev); 358383a06263SAlex Elder if (ret) 358483a06263SAlex Elder goto err_out_bus; 358583a06263SAlex Elder 358683a06263SAlex Elder /* Everything's ready. Announce the disk to the world. */ 358783a06263SAlex Elder 358883a06263SAlex Elder add_disk(rbd_dev->disk); 358983a06263SAlex Elder 359083a06263SAlex Elder pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name, 359183a06263SAlex Elder (unsigned long long) rbd_dev->mapping.size); 359283a06263SAlex Elder 359383a06263SAlex Elder return ret; 359483a06263SAlex Elder err_out_bus: 359583a06263SAlex Elder /* this will also clean up rest of rbd_dev stuff */ 359683a06263SAlex Elder 359783a06263SAlex Elder rbd_bus_del_dev(rbd_dev); 359883a06263SAlex Elder 359983a06263SAlex Elder return ret; 360083a06263SAlex Elder err_out_disk: 360183a06263SAlex Elder rbd_free_disk(rbd_dev); 360283a06263SAlex Elder err_out_blkdev: 360383a06263SAlex Elder unregister_blkdev(rbd_dev->major, rbd_dev->name); 360483a06263SAlex Elder err_out_id: 360583a06263SAlex Elder rbd_dev_id_put(rbd_dev); 360683a06263SAlex Elder err_out_snaps: 360783a06263SAlex Elder rbd_remove_all_snaps(rbd_dev); 360883a06263SAlex Elder 360983a06263SAlex Elder return ret; 361083a06263SAlex Elder } 361183a06263SAlex Elder 3612a30b71b9SAlex Elder /* 3613a30b71b9SAlex Elder * Probe for the existence of the header object for the given rbd 3614a30b71b9SAlex Elder * device. For format 2 images this includes determining the image 3615a30b71b9SAlex Elder * id. 3616a30b71b9SAlex Elder */ 3617a30b71b9SAlex Elder static int rbd_dev_probe(struct rbd_device *rbd_dev) 3618a30b71b9SAlex Elder { 3619a30b71b9SAlex Elder int ret; 3620a30b71b9SAlex Elder 3621a30b71b9SAlex Elder /* 3622a30b71b9SAlex Elder * Get the id from the image id object. If it's not a 3623a30b71b9SAlex Elder * format 2 image, we'll get ENOENT back, and we'll assume 3624a30b71b9SAlex Elder * it's a format 1 image. 3625a30b71b9SAlex Elder */ 3626a30b71b9SAlex Elder ret = rbd_dev_image_id(rbd_dev); 3627a30b71b9SAlex Elder if (ret) 3628a30b71b9SAlex Elder ret = rbd_dev_v1_probe(rbd_dev); 3629a30b71b9SAlex Elder else 3630a30b71b9SAlex Elder ret = rbd_dev_v2_probe(rbd_dev); 363183a06263SAlex Elder if (ret) { 3632a30b71b9SAlex Elder dout("probe failed, returning %d\n", ret); 3633a30b71b9SAlex Elder 3634a30b71b9SAlex Elder return ret; 3635a30b71b9SAlex Elder } 3636a30b71b9SAlex Elder 363783a06263SAlex Elder ret = rbd_dev_probe_finish(rbd_dev); 363883a06263SAlex Elder if (ret) 363983a06263SAlex Elder rbd_header_free(&rbd_dev->header); 364083a06263SAlex Elder 364183a06263SAlex Elder return ret; 364283a06263SAlex Elder } 364383a06263SAlex Elder 364459c2be1eSYehuda Sadeh static ssize_t rbd_add(struct bus_type *bus, 364559c2be1eSYehuda Sadeh const char *buf, 364659c2be1eSYehuda Sadeh size_t count) 3647602adf40SYehuda Sadeh { 3648cb8627c7SAlex Elder struct rbd_device *rbd_dev = NULL; 3649dc79b113SAlex Elder struct ceph_options *ceph_opts = NULL; 36504e9afebaSAlex Elder struct rbd_options *rbd_opts = NULL; 3651859c31dfSAlex Elder struct rbd_spec *spec = NULL; 36529d3997fdSAlex Elder struct rbd_client *rbdc; 365327cc2594SAlex Elder struct ceph_osd_client *osdc; 365427cc2594SAlex Elder int rc = -ENOMEM; 3655602adf40SYehuda Sadeh 3656602adf40SYehuda Sadeh if (!try_module_get(THIS_MODULE)) 3657602adf40SYehuda Sadeh return -ENODEV; 3658602adf40SYehuda Sadeh 3659a725f65eSAlex Elder /* parse add command */ 3660859c31dfSAlex Elder rc = rbd_add_parse_args(buf, &ceph_opts, &rbd_opts, &spec); 3661dc79b113SAlex Elder if (rc < 0) 3662bd4ba655SAlex Elder goto err_out_module; 3663a725f65eSAlex Elder 36649d3997fdSAlex Elder rbdc = rbd_get_client(ceph_opts); 36659d3997fdSAlex Elder if (IS_ERR(rbdc)) { 36669d3997fdSAlex Elder rc = PTR_ERR(rbdc); 36670ddebc0cSAlex Elder goto err_out_args; 36689d3997fdSAlex Elder } 3669c53d5893SAlex Elder ceph_opts = NULL; /* rbd_dev client now owns this */ 3670602adf40SYehuda Sadeh 3671602adf40SYehuda Sadeh /* pick the pool */ 36729d3997fdSAlex Elder osdc = &rbdc->client->osdc; 3673859c31dfSAlex Elder rc = ceph_pg_poolid_by_name(osdc->osdmap, spec->pool_name); 3674602adf40SYehuda Sadeh if (rc < 0) 3675602adf40SYehuda Sadeh goto err_out_client; 3676859c31dfSAlex Elder spec->pool_id = (u64) rc; 3677859c31dfSAlex Elder 3678c53d5893SAlex Elder rbd_dev = rbd_dev_create(rbdc, spec); 3679bd4ba655SAlex Elder if (!rbd_dev) 3680bd4ba655SAlex Elder goto err_out_client; 3681c53d5893SAlex Elder rbdc = NULL; /* rbd_dev now owns this */ 3682c53d5893SAlex Elder spec = NULL; /* rbd_dev now owns this */ 3683602adf40SYehuda Sadeh 3684bd4ba655SAlex Elder rbd_dev->mapping.read_only = rbd_opts->read_only; 3685c53d5893SAlex Elder kfree(rbd_opts); 3686c53d5893SAlex Elder rbd_opts = NULL; /* done with this */ 3687bd4ba655SAlex Elder 3688a30b71b9SAlex Elder rc = rbd_dev_probe(rbd_dev); 3689a30b71b9SAlex Elder if (rc < 0) 3690c53d5893SAlex Elder goto err_out_rbd_dev; 369105fd6f6fSAlex Elder 3692602adf40SYehuda Sadeh return count; 3693c53d5893SAlex Elder err_out_rbd_dev: 3694c53d5893SAlex Elder rbd_dev_destroy(rbd_dev); 3695bd4ba655SAlex Elder err_out_client: 36969d3997fdSAlex Elder rbd_put_client(rbdc); 36970ddebc0cSAlex Elder err_out_args: 369878cea76eSAlex Elder if (ceph_opts) 369978cea76eSAlex Elder ceph_destroy_options(ceph_opts); 37004e9afebaSAlex Elder kfree(rbd_opts); 3701859c31dfSAlex Elder rbd_spec_put(spec); 3702bd4ba655SAlex Elder err_out_module: 3703bd4ba655SAlex Elder module_put(THIS_MODULE); 370427cc2594SAlex Elder 3705602adf40SYehuda Sadeh dout("Error adding device %s\n", buf); 370627cc2594SAlex Elder 370727cc2594SAlex Elder return (ssize_t) rc; 3708602adf40SYehuda Sadeh } 3709602adf40SYehuda Sadeh 3710de71a297SAlex Elder static struct rbd_device *__rbd_get_dev(unsigned long dev_id) 3711602adf40SYehuda Sadeh { 3712602adf40SYehuda Sadeh struct list_head *tmp; 3713602adf40SYehuda Sadeh struct rbd_device *rbd_dev; 3714602adf40SYehuda Sadeh 3715e124a82fSAlex Elder spin_lock(&rbd_dev_list_lock); 3716602adf40SYehuda Sadeh list_for_each(tmp, &rbd_dev_list) { 3717602adf40SYehuda Sadeh rbd_dev = list_entry(tmp, struct rbd_device, node); 3718de71a297SAlex Elder if (rbd_dev->dev_id == dev_id) { 3719e124a82fSAlex Elder spin_unlock(&rbd_dev_list_lock); 3720602adf40SYehuda Sadeh return rbd_dev; 3721602adf40SYehuda Sadeh } 3722e124a82fSAlex Elder } 3723e124a82fSAlex Elder spin_unlock(&rbd_dev_list_lock); 3724602adf40SYehuda Sadeh return NULL; 3725602adf40SYehuda Sadeh } 3726602adf40SYehuda Sadeh 3727dfc5606dSYehuda Sadeh static void rbd_dev_release(struct device *dev) 3728602adf40SYehuda Sadeh { 3729593a9e7bSAlex Elder struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 3730602adf40SYehuda Sadeh 37311dbb4399SAlex Elder if (rbd_dev->watch_request) { 37321dbb4399SAlex Elder struct ceph_client *client = rbd_dev->rbd_client->client; 37331dbb4399SAlex Elder 37341dbb4399SAlex Elder ceph_osdc_unregister_linger_request(&client->osdc, 373559c2be1eSYehuda Sadeh rbd_dev->watch_request); 37361dbb4399SAlex Elder } 373759c2be1eSYehuda Sadeh if (rbd_dev->watch_event) 3738070c633fSAlex Elder rbd_req_sync_unwatch(rbd_dev); 373959c2be1eSYehuda Sadeh 3740602adf40SYehuda Sadeh 3741602adf40SYehuda Sadeh /* clean up and free blkdev */ 3742602adf40SYehuda Sadeh rbd_free_disk(rbd_dev); 3743602adf40SYehuda Sadeh unregister_blkdev(rbd_dev->major, rbd_dev->name); 374432eec68dSAlex Elder 37452ac4e75dSAlex Elder /* release allocated disk header fields */ 37462ac4e75dSAlex Elder rbd_header_free(&rbd_dev->header); 37472ac4e75dSAlex Elder 374832eec68dSAlex Elder /* done with the id, and with the rbd_dev */ 3749e2839308SAlex Elder rbd_dev_id_put(rbd_dev); 3750c53d5893SAlex Elder rbd_assert(rbd_dev->rbd_client != NULL); 3751c53d5893SAlex Elder rbd_dev_destroy(rbd_dev); 3752602adf40SYehuda Sadeh 3753602adf40SYehuda Sadeh /* release module ref */ 3754602adf40SYehuda Sadeh module_put(THIS_MODULE); 3755602adf40SYehuda Sadeh } 3756602adf40SYehuda Sadeh 3757dfc5606dSYehuda Sadeh static ssize_t rbd_remove(struct bus_type *bus, 3758602adf40SYehuda Sadeh const char *buf, 3759602adf40SYehuda Sadeh size_t count) 3760602adf40SYehuda Sadeh { 3761602adf40SYehuda Sadeh struct rbd_device *rbd_dev = NULL; 3762602adf40SYehuda Sadeh int target_id, rc; 3763602adf40SYehuda Sadeh unsigned long ul; 3764602adf40SYehuda Sadeh int ret = count; 3765602adf40SYehuda Sadeh 3766602adf40SYehuda Sadeh rc = strict_strtoul(buf, 10, &ul); 3767602adf40SYehuda Sadeh if (rc) 3768602adf40SYehuda Sadeh return rc; 3769602adf40SYehuda Sadeh 3770602adf40SYehuda Sadeh /* convert to int; abort if we lost anything in the conversion */ 3771602adf40SYehuda Sadeh target_id = (int) ul; 3772602adf40SYehuda Sadeh if (target_id != ul) 3773602adf40SYehuda Sadeh return -EINVAL; 3774602adf40SYehuda Sadeh 3775602adf40SYehuda Sadeh mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); 3776602adf40SYehuda Sadeh 3777602adf40SYehuda Sadeh rbd_dev = __rbd_get_dev(target_id); 3778602adf40SYehuda Sadeh if (!rbd_dev) { 3779602adf40SYehuda Sadeh ret = -ENOENT; 3780602adf40SYehuda Sadeh goto done; 3781602adf40SYehuda Sadeh } 3782602adf40SYehuda Sadeh 378342382b70SAlex Elder if (rbd_dev->open_count) { 378442382b70SAlex Elder ret = -EBUSY; 378542382b70SAlex Elder goto done; 378642382b70SAlex Elder } 378742382b70SAlex Elder 378841f38c2bSAlex Elder rbd_remove_all_snaps(rbd_dev); 3789dfc5606dSYehuda Sadeh rbd_bus_del_dev(rbd_dev); 3790602adf40SYehuda Sadeh 3791602adf40SYehuda Sadeh done: 3792602adf40SYehuda Sadeh mutex_unlock(&ctl_mutex); 3793aafb230eSAlex Elder 3794602adf40SYehuda Sadeh return ret; 3795602adf40SYehuda Sadeh } 3796602adf40SYehuda Sadeh 3797602adf40SYehuda Sadeh /* 3798602adf40SYehuda Sadeh * create control files in sysfs 3799dfc5606dSYehuda Sadeh * /sys/bus/rbd/... 3800602adf40SYehuda Sadeh */ 3801602adf40SYehuda Sadeh static int rbd_sysfs_init(void) 3802602adf40SYehuda Sadeh { 3803dfc5606dSYehuda Sadeh int ret; 3804602adf40SYehuda Sadeh 3805fed4c143SAlex Elder ret = device_register(&rbd_root_dev); 3806dfc5606dSYehuda Sadeh if (ret < 0) 3807dfc5606dSYehuda Sadeh return ret; 3808602adf40SYehuda Sadeh 3809fed4c143SAlex Elder ret = bus_register(&rbd_bus_type); 3810fed4c143SAlex Elder if (ret < 0) 3811fed4c143SAlex Elder device_unregister(&rbd_root_dev); 3812602adf40SYehuda Sadeh 3813602adf40SYehuda Sadeh return ret; 3814602adf40SYehuda Sadeh } 3815602adf40SYehuda Sadeh 3816602adf40SYehuda Sadeh static void rbd_sysfs_cleanup(void) 3817602adf40SYehuda Sadeh { 3818dfc5606dSYehuda Sadeh bus_unregister(&rbd_bus_type); 3819fed4c143SAlex Elder device_unregister(&rbd_root_dev); 3820602adf40SYehuda Sadeh } 3821602adf40SYehuda Sadeh 3822602adf40SYehuda Sadeh int __init rbd_init(void) 3823602adf40SYehuda Sadeh { 3824602adf40SYehuda Sadeh int rc; 3825602adf40SYehuda Sadeh 3826602adf40SYehuda Sadeh rc = rbd_sysfs_init(); 3827602adf40SYehuda Sadeh if (rc) 3828602adf40SYehuda Sadeh return rc; 3829f0f8cef5SAlex Elder pr_info("loaded " RBD_DRV_NAME_LONG "\n"); 3830602adf40SYehuda Sadeh return 0; 3831602adf40SYehuda Sadeh } 3832602adf40SYehuda Sadeh 3833602adf40SYehuda Sadeh void __exit rbd_exit(void) 3834602adf40SYehuda Sadeh { 3835602adf40SYehuda Sadeh rbd_sysfs_cleanup(); 3836602adf40SYehuda Sadeh } 3837602adf40SYehuda Sadeh 3838602adf40SYehuda Sadeh module_init(rbd_init); 3839602adf40SYehuda Sadeh module_exit(rbd_exit); 3840602adf40SYehuda Sadeh 3841602adf40SYehuda Sadeh MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); 3842602adf40SYehuda Sadeh MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); 3843602adf40SYehuda Sadeh MODULE_DESCRIPTION("rados block device"); 3844602adf40SYehuda Sadeh 3845602adf40SYehuda Sadeh /* following authorship retained from original osdblk.c */ 3846602adf40SYehuda Sadeh MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); 3847602adf40SYehuda Sadeh 3848602adf40SYehuda Sadeh MODULE_LICENSE("GPL"); 3849