diff options
Diffstat (limited to 'drivers/block/rbd.c')
| -rw-r--r-- | drivers/block/rbd.c | 1074 |
1 files changed, 638 insertions, 436 deletions
diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index b2c98c1bc037..27b71a0b72d0 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -42,6 +42,7 @@ #include <linux/blkdev.h> #include <linux/slab.h> #include <linux/idr.h> +#include <linux/workqueue.h> #include "rbd_types.h" @@ -209,6 +210,12 @@ enum obj_request_type { OBJ_REQUEST_NODATA, OBJ_REQUEST_BIO, OBJ_REQUEST_PAGES }; +enum obj_operation_type { + OBJ_OP_WRITE, + OBJ_OP_READ, + OBJ_OP_DISCARD, +}; + enum obj_req_flags { OBJ_REQ_DONE, /* completion flag: not done = 0, done = 1 */ OBJ_REQ_IMG_DATA, /* object usage: standalone = 0, image = 1 */ @@ -275,6 +282,7 @@ enum img_req_flags { IMG_REQ_WRITE, /* I/O direction: read = 0, write = 1 */ IMG_REQ_CHILD, /* initiator: block = 0, child image = 1 */ IMG_REQ_LAYERED, /* ENOENT handling: normal = 0, layered = 1 */ + IMG_REQ_DISCARD, /* discard: normal = 0, discard request = 1 */ }; struct rbd_img_request { @@ -332,7 +340,9 @@ struct rbd_device { char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ + struct list_head rq_queue; /* incoming rq queue */ spinlock_t lock; /* queue, flags, open_count */ + struct work_struct rq_work; struct rbd_image_header header; unsigned long flags; /* possibly lock protected */ @@ -391,6 +401,8 @@ static struct kmem_cache *rbd_segment_name_cache; static int rbd_major; static DEFINE_IDA(rbd_dev_id_ida); +static struct workqueue_struct *rbd_wq; + /* * Default to false for now, as single-major requires >= 0.75 version of * userspace rbd utility. @@ -514,7 +526,8 @@ static void rbd_dev_remove_parent(struct rbd_device *rbd_dev); static int rbd_dev_refresh(struct rbd_device *rbd_dev); static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev); -static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev); +static int rbd_dev_header_info(struct rbd_device *rbd_dev); +static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev); static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, u64 snap_id); static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, @@ -780,6 +793,20 @@ static int parse_rbd_opts_token(char *c, void *private) return 0; } +static char* obj_op_name(enum obj_operation_type op_type) +{ + switch (op_type) { + case OBJ_OP_READ: + return "read"; + case OBJ_OP_WRITE: + return "write"; + case OBJ_OP_DISCARD: + return "discard"; + default: + return "???"; + } +} + /* * Get a ceph client with specific addr and configuration, if one does * not exist create it. Either way, ceph_opts is consumed by this @@ -971,12 +998,6 @@ static int rbd_header_from_disk(struct rbd_device *rbd_dev, header->snap_names = snap_names; header->snap_sizes = snap_sizes; - /* Make sure mapping size is consistent with header info */ - - if (rbd_dev->spec->snap_id == CEPH_NOSNAP || first_time) - if (rbd_dev->mapping.size != header->image_size) - rbd_dev->mapping.size = header->image_size; - return 0; out_2big: ret = -EIO; @@ -1139,6 +1160,13 @@ static void rbd_dev_mapping_clear(struct rbd_device *rbd_dev) rbd_dev->mapping.features = 0; } +static void rbd_segment_name_free(const char *name) +{ + /* The explicit cast here is needed to drop the const qualifier */ + + kmem_cache_free(rbd_segment_name_cache, (void *)name); +} + static const char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset) { char *name; @@ -1158,20 +1186,13 @@ static const char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset) if (ret < 0 || ret > CEPH_MAX_OID_NAME_LEN) { pr_err("error formatting segment name for #%llu (%d)\n", segment, ret); - kfree(name); + rbd_segment_name_free(name); name = NULL; } return name; } -static void rbd_segment_name_free(const char *name) -{ - /* The explicit cast here is needed to drop the const qualifier */ - - kmem_cache_free(rbd_segment_name_cache, (void *)name); -} - static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) { u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; @@ -1371,7 +1392,7 @@ static void obj_request_img_data_set(struct rbd_obj_request *obj_request) struct rbd_device *rbd_dev; rbd_dev = obj_request->img_request->rbd_dev; - rbd_warn(rbd_dev, "obj_request %p already marked img_data\n", + rbd_warn(rbd_dev, "obj_request %p already marked img_data", obj_request); } } @@ -1389,7 +1410,7 @@ static void obj_request_done_set(struct rbd_obj_request *obj_request) if (obj_request_img_data_test(obj_request)) rbd_dev = obj_request->img_request->rbd_dev; - rbd_warn(rbd_dev, "obj_request %p already marked done\n", + rbd_warn(rbd_dev, "obj_request %p already marked done", obj_request); } } @@ -1527,11 +1548,37 @@ static bool obj_request_type_valid(enum obj_request_type type) static int rbd_obj_request_submit(struct ceph_osd_client *osdc, struct rbd_obj_request *obj_request) { - dout("%s: osdc %p obj %p\n", __func__, osdc, obj_request); - + dout("%s %p\n", __func__, obj_request); return ceph_osdc_start_request(osdc, obj_request->osd_req, false); } +static void rbd_obj_request_end(struct rbd_obj_request *obj_request) +{ + dout("%s %p\n", __func__, obj_request); + ceph_osdc_cancel_request(obj_request->osd_req); +} + +/* + * Wait for an object request to complete. If interrupted, cancel the + * underlying osd request. + */ +static int rbd_obj_request_wait(struct rbd_obj_request *obj_request) +{ + int ret; + + dout("%s %p\n", __func__, obj_request); + + ret = wait_for_completion_interruptible(&obj_request->completion); + if (ret < 0) { + dout("%s %p interrupted\n", __func__, obj_request); + rbd_obj_request_end(obj_request); + return ret; + } + + dout("%s %p done\n", __func__, obj_request); + return 0; +} + static void rbd_img_request_complete(struct rbd_img_request *img_request) { @@ -1558,15 +1605,6 @@ static void rbd_img_request_complete(struct rbd_img_request *img_request) rbd_img_request_put(img_request); } -/* Caller is responsible for rbd_obj_request_destroy(obj_request) */ - -static int rbd_obj_request_wait(struct rbd_obj_request *obj_request) -{ - dout("%s: obj %p\n", __func__, obj_request); - - return wait_for_completion_interruptible(&obj_request->completion); -} - /* * The default/initial value for all image request flags is 0. Each * is conditionally set to 1 at image request initialization time @@ -1584,6 +1622,21 @@ static bool img_request_write_test(struct rbd_img_request *img_request) return test_bit(IMG_REQ_WRITE, &img_request->flags) != 0; } +/* + * Set the discard flag when the img_request is an discard request + */ +static void img_request_discard_set(struct rbd_img_request *img_request) +{ + set_bit(IMG_REQ_DISCARD, &img_request->flags); + smp_mb(); +} + +static bool img_request_discard_test(struct rbd_img_request *img_request) +{ + smp_mb(); + return test_bit(IMG_REQ_DISCARD, &img_request->flags) != 0; +} + static void img_request_child_set(struct rbd_img_request *img_request) { set_bit(IMG_REQ_CHILD, &img_request->flags); @@ -1620,6 +1673,17 @@ static bool img_request_layered_test(struct rbd_img_request *img_request) return test_bit(IMG_REQ_LAYERED, &img_request->flags) != 0; } +static enum obj_operation_type +rbd_img_request_op_type(struct rbd_img_request *img_request) +{ + if (img_request_write_test(img_request)) + return OBJ_OP_WRITE; + else if (img_request_discard_test(img_request)) + return OBJ_OP_DISCARD; + else + return OBJ_OP_READ; +} + static void rbd_img_obj_request_read_callback(struct rbd_obj_request *obj_request) { @@ -1706,6 +1770,21 @@ static void rbd_osd_write_callback(struct rbd_obj_request *obj_request) obj_request_done_set(obj_request); } +static void rbd_osd_discard_callback(struct rbd_obj_request *obj_request) +{ + dout("%s: obj %p result %d %llu\n", __func__, obj_request, + obj_request->result, obj_request->length); + /* + * There is no such thing as a successful short discard. Set + * it to our originally-requested length. + */ + obj_request->xferred = obj_request->length; + /* discarding a non-existent object is not a problem */ + if (obj_request->result == -ENOENT) + obj_request->result = 0; + obj_request_done_set(obj_request); +} + /* * For a simple stat call there's nothing to do. We'll do more if * this is part of a write sequence for a layered image. @@ -1757,13 +1836,18 @@ static void rbd_osd_req_callback(struct ceph_osd_request *osd_req, case CEPH_OSD_OP_STAT: rbd_osd_stat_callback(obj_request); break; + case CEPH_OSD_OP_DELETE: + case CEPH_OSD_OP_TRUNCATE: + case CEPH_OSD_OP_ZERO: + rbd_osd_discard_callback(obj_request); + break; case CEPH_OSD_OP_CALL: case CEPH_OSD_OP_NOTIFY_ACK: case CEPH_OSD_OP_WATCH: rbd_osd_trivial_callback(obj_request); break; default: - rbd_warn(NULL, "%s: unsupported op %hu\n", + rbd_warn(NULL, "%s: unsupported op %hu", obj_request->object_name, (unsigned short) opcode); break; } @@ -1807,7 +1891,7 @@ static void rbd_osd_req_format_write(struct rbd_obj_request *obj_request) */ static struct ceph_osd_request *rbd_osd_req_create( struct rbd_device *rbd_dev, - bool write_request, + enum obj_operation_type op_type, unsigned int num_ops, struct rbd_obj_request *obj_request) { @@ -1815,16 +1899,18 @@ static struct ceph_osd_request *rbd_osd_req_create( struct ceph_osd_client *osdc; struct ceph_osd_request *osd_req; - if (obj_request_img_data_test(obj_request)) { + if (obj_request_img_data_test(obj_request) && + (op_type == OBJ_OP_DISCARD || op_type == OBJ_OP_WRITE)) { struct rbd_img_request *img_request = obj_request->img_request; - - rbd_assert(write_request == - img_request_write_test(img_request)); - if (write_request) - snapc = img_request->snapc; + if (op_type == OBJ_OP_WRITE) { + rbd_assert(img_request_write_test(img_request)); + } else { + rbd_assert(img_request_discard_test(img_request)); + } + snapc = img_request->snapc; } - rbd_assert(num_ops == 1 || (write_request && num_ops == 2)); + rbd_assert(num_ops == 1 || ((op_type == OBJ_OP_WRITE) && num_ops == 2)); /* Allocate and initialize the request, for the num_ops ops */ @@ -1834,7 +1920,7 @@ static struct ceph_osd_request *rbd_osd_req_create( if (!osd_req) return NULL; /* ENOMEM */ - if (write_request) + if (op_type == OBJ_OP_WRITE || op_type == OBJ_OP_DISCARD) osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; else osd_req->r_flags = CEPH_OSD_FLAG_READ; @@ -1849,9 +1935,10 @@ static struct ceph_osd_request *rbd_osd_req_create( } /* - * Create a copyup osd request based on the information in the - * object request supplied. A copyup request has three osd ops, - * a copyup method call, a hint op, and a write op. + * Create a copyup osd request based on the information in the object + * request supplied. A copyup request has two or three osd ops, a + * copyup method call, potentially a hint op, and a write or truncate + * or zero op. */ static struct ceph_osd_request * rbd_osd_req_create_copyup(struct rbd_obj_request *obj_request) @@ -1861,18 +1948,24 @@ rbd_osd_req_create_copyup(struct rbd_obj_request *obj_request) struct rbd_device *rbd_dev; struct ceph_osd_client *osdc; struct ceph_osd_request *osd_req; + int num_osd_ops = 3; rbd_assert(obj_request_img_data_test(obj_request)); img_request = obj_request->img_request; rbd_assert(img_request); - rbd_assert(img_request_write_test(img_request)); + rbd_assert(img_request_write_test(img_request) || + img_request_discard_test(img_request)); + + if (img_request_discard_test(img_request)) + num_osd_ops = 2; - /* Allocate and initialize the request, for the three ops */ + /* Allocate and initialize the request, for all the ops */ snapc = img_request->snapc; rbd_dev = img_request->rbd_dev; osdc = &rbd_dev->rbd_client->client->osdc; - osd_req = ceph_osdc_alloc_request(osdc, snapc, 3, false, GFP_ATOMIC); + osd_req = ceph_osdc_alloc_request(osdc, snapc, num_osd_ops, + false, GFP_ATOMIC); if (!osd_req) return NULL; /* ENOMEM */ @@ -1998,7 +2091,7 @@ static void rbd_dev_parent_put(struct rbd_device *rbd_dev) if (!counter) rbd_dev_unparent(rbd_dev); else - rbd_warn(rbd_dev, "parent reference underflow\n"); + rbd_warn(rbd_dev, "parent reference underflow"); } /* @@ -2028,7 +2121,7 @@ static bool rbd_dev_parent_get(struct rbd_device *rbd_dev) /* Image was flattened, but parent is not yet torn down */ if (counter < 0) - rbd_warn(rbd_dev, "parent reference overflow\n"); + rbd_warn(rbd_dev, "parent reference overflow"); return false; } @@ -2041,28 +2134,26 @@ static bool rbd_dev_parent_get(struct rbd_device *rbd_dev) static struct rbd_img_request *rbd_img_request_create( struct rbd_device *rbd_dev, u64 offset, u64 length, - bool write_request) + enum obj_operation_type op_type, + struct ceph_snap_context *snapc) { struct rbd_img_request *img_request; - img_request = kmem_cache_alloc(rbd_img_request_cache, GFP_ATOMIC); + img_request = kmem_cache_alloc(rbd_img_request_cache, GFP_NOIO); if (!img_request) return NULL; - if (write_request) { - down_read(&rbd_dev->header_rwsem); - ceph_get_snap_context(rbd_dev->header.snapc); - up_read(&rbd_dev->header_rwsem); - } - img_request->rq = NULL; img_request->rbd_dev = rbd_dev; img_request->offset = offset; img_request->length = length; img_request->flags = 0; - if (write_request) { + if (op_type == OBJ_OP_DISCARD) { + img_request_discard_set(img_request); + img_request->snapc = snapc; + } else if (op_type == OBJ_OP_WRITE) { img_request_write_set(img_request); - img_request->snapc = rbd_dev->header.snapc; + img_request->snapc = snapc; } else { img_request->snap_id = rbd_dev->spec->snap_id; } @@ -2077,8 +2168,7 @@ static struct rbd_img_request *rbd_img_request_create( kref_init(&img_request->kref); dout("%s: rbd_dev %p %s %llu/%llu -> img %p\n", __func__, rbd_dev, - write_request ? "write" : "read", offset, length, - img_request); + obj_op_name(op_type), offset, length, img_request); return img_request; } @@ -2102,7 +2192,8 @@ static void rbd_img_request_destroy(struct kref *kref) rbd_dev_parent_put(img_request->rbd_dev); } - if (img_request_write_test(img_request)) + if (img_request_write_test(img_request) || + img_request_discard_test(img_request)) ceph_put_snap_context(img_request->snapc); kmem_cache_free(rbd_img_request_cache, img_request); @@ -2118,8 +2209,8 @@ static struct rbd_img_request *rbd_parent_request_create( rbd_assert(obj_request->img_request); rbd_dev = obj_request->img_request->rbd_dev; - parent_request = rbd_img_request_create(rbd_dev->parent, - img_offset, length, false); + parent_request = rbd_img_request_create(rbd_dev->parent, img_offset, + length, OBJ_OP_READ, NULL); if (!parent_request) return NULL; @@ -2160,12 +2251,19 @@ static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request) result = obj_request->result; if (result) { struct rbd_device *rbd_dev = img_request->rbd_dev; + enum obj_operation_type op_type; + + if (img_request_discard_test(img_request)) + op_type = OBJ_OP_DISCARD; + else if (img_request_write_test(img_request)) + op_type = OBJ_OP_WRITE; + else + op_type = OBJ_OP_READ; - rbd_warn(rbd_dev, "%s %llx at %llx (%llx)\n", - img_request_write_test(img_request) ? "write" : "read", - obj_request->length, obj_request->img_offset, - obj_request->offset); - rbd_warn(rbd_dev, " result %d xferred %x\n", + rbd_warn(rbd_dev, "%s %llx at %llx (%llx)", + obj_op_name(op_type), obj_request->length, + obj_request->img_offset, obj_request->offset); + rbd_warn(rbd_dev, " result %d xferred %x", result, xferred); if (!img_request->result) img_request->result = result; @@ -2229,6 +2327,67 @@ out: } /* + * Add individual osd ops to the given ceph_osd_request and prepare + * them for submission. num_ops is the current number of + * osd operations already to the object request. + */ +static void rbd_img_obj_request_fill(struct rbd_obj_request *obj_request, + struct ceph_osd_request *osd_request, + enum obj_operation_type op_type, + unsigned int num_ops) +{ + struct rbd_img_request *img_request = obj_request->img_request; + struct rbd_device *rbd_dev = img_request->rbd_dev; + u64 object_size = rbd_obj_bytes(&rbd_dev->header); + u64 offset = obj_request->offset; + u64 length = obj_request->length; + u64 img_end; + u16 opcode; + + if (op_type == OBJ_OP_DISCARD) { + if (!offset && length == object_size && + (!img_request_layered_test(img_request) || + !obj_request_overlaps_parent(obj_request))) { + opcode = CEPH_OSD_OP_DELETE; + } else if ((offset + length == object_size)) { + opcode = CEPH_OSD_OP_TRUNCATE; + } else { + down_read(&rbd_dev->header_rwsem); + img_end = rbd_dev->header.image_size; + up_read(&rbd_dev->header_rwsem); + + if (obj_request->img_offset + length == img_end) + opcode = CEPH_OSD_OP_TRUNCATE; + else + opcode = CEPH_OSD_OP_ZERO; + } + } else if (op_type == OBJ_OP_WRITE) { + opcode = CEPH_OSD_OP_WRITE; + osd_req_op_alloc_hint_init(osd_request, num_ops, + object_size, object_size); + num_ops++; + } else { + opcode = CEPH_OSD_OP_READ; + } + + osd_req_op_extent_init(osd_request, num_ops, opcode, offset, length, + 0, 0); + if (obj_request->type == OBJ_REQUEST_BIO) + osd_req_op_extent_osd_data_bio(osd_request, num_ops, + obj_request->bio_list, length); + else if (obj_request->type == OBJ_REQUEST_PAGES) + osd_req_op_extent_osd_data_pages(osd_request, num_ops, + obj_request->pages, length, + offset & ~PAGE_MASK, false, false); + + /* Discards are also writes */ + if (op_type == OBJ_OP_WRITE || op_type == OBJ_OP_DISCARD) + rbd_osd_req_format_write(obj_request); + else + rbd_osd_req_format_read(obj_request); +} + +/* * Split up an image request into one or more object requests, each * to a different object. The "type" parameter indicates whether * "data_desc" is the pointer to the head of a list of bio @@ -2243,28 +2402,26 @@ static int rbd_img_request_fill(struct rbd_img_request *img_request, struct rbd_device *rbd_dev = img_request->rbd_dev; struct rbd_obj_request *obj_request = NULL; struct rbd_obj_request *next_obj_request; - bool write_request = img_request_write_test(img_request); struct bio *bio_list = NULL; unsigned int bio_offset = 0; struct page **pages = NULL; + enum obj_operation_type op_type; u64 img_offset; u64 resid; - u16 opcode; dout("%s: img %p type %d data_desc %p\n", __func__, img_request, (int)type, data_desc); - opcode = write_request ? CEPH_OSD_OP_WRITE : CEPH_OSD_OP_READ; img_offset = img_request->offset; resid = img_request->length; rbd_assert(resid > 0); + op_type = rbd_img_request_op_type(img_request); if (type == OBJ_REQUEST_BIO) { bio_list = data_desc; rbd_assert(img_offset == bio_list->bi_iter.bi_sector << SECTOR_SHIFT); - } else { - rbd_assert(type == OBJ_REQUEST_PAGES); + } else if (type == OBJ_REQUEST_PAGES) { pages = data_desc; } @@ -2273,7 +2430,6 @@ static int rbd_img_request_fill(struct rbd_img_request *img_request, const char *object_name; u64 offset; u64 length; - unsigned int which = 0; object_name = rbd_segment_name(rbd_dev, img_offset); if (!object_name) @@ -2305,7 +2461,7 @@ static int rbd_img_request_fill(struct rbd_img_request *img_request, GFP_ATOMIC); if (!obj_request->bio_list) goto out_unwind; - } else { + } else if (type == OBJ_REQUEST_PAGES) { unsigned int page_count; obj_request->pages = pages; @@ -2316,38 +2472,19 @@ static int rbd_img_request_fill(struct rbd_img_request *img_request, pages += page_count; } - osd_req = rbd_osd_req_create(rbd_dev, write_request, - (write_request ? 2 : 1), - obj_request); + osd_req = rbd_osd_req_create(rbd_dev, op_type, + (op_type == OBJ_OP_WRITE) ? 2 : 1, + obj_request); if (!osd_req) goto out_unwind; + obj_request->osd_req = osd_req; obj_request->callback = rbd_img_obj_callback; - rbd_img_request_get(img_request); - - if (write_request) { - osd_req_op_alloc_hint_init(osd_req, which, - rbd_obj_bytes(&rbd_dev->header), - rbd_obj_bytes(&rbd_dev->header)); - which++; - } - - osd_req_op_extent_init(osd_req, which, opcode, offset, length, - 0, 0); - if (type == OBJ_REQUEST_BIO) - osd_req_op_extent_osd_data_bio(osd_req, which, - obj_request->bio_list, length); - else - osd_req_op_extent_osd_data_pages(osd_req, which, - obj_request->pages, length, - offset & ~PAGE_MASK, false, false); + obj_request->img_offset = img_offset; - if (write_request) - rbd_osd_req_format_write(obj_request); - else - rbd_osd_req_format_read(obj_request); + rbd_img_obj_request_fill(obj_request, osd_req, op_type, 0); - obj_request->img_offset = img_offset; + rbd_img_request_get(img_request); img_offset += length; resid -= length; @@ -2370,7 +2507,8 @@ rbd_img_obj_copyup_callback(struct rbd_obj_request *obj_request) struct page **pages; u32 page_count; - rbd_assert(obj_request->type == OBJ_REQUEST_BIO); + rbd_assert(obj_request->type == OBJ_REQUEST_BIO || + obj_request->type == OBJ_REQUEST_NODATA); rbd_assert(obj_request_img_data_test(obj_request)); img_request = obj_request->img_request; rbd_assert(img_request); @@ -2408,11 +2546,10 @@ rbd_img_obj_parent_read_full_callback(struct rbd_img_request *img_request) struct ceph_osd_client *osdc; struct rbd_device *rbd_dev; struct page **pages; + enum obj_operation_type op_type; u32 page_count; int img_result; u64 parent_length; - u64 offset; - u64 length; rbd_assert(img_request_child_test(img_request)); @@ -2476,26 +2613,10 @@ rbd_img_obj_parent_read_full_callback(struct rbd_img_request *img_request) osd_req_op_cls_request_data_pages(osd_req, 0, pages, parent_length, 0, false, false); - /* Then the hint op */ - - osd_req_op_alloc_hint_init(osd_req, 1, rbd_obj_bytes(&rbd_dev->header), - rbd_obj_bytes(&rbd_dev->header)); - - /* And the original write request op */ - - offset = orig_request->offset; - length = orig_request->length; - osd_req_op_extent_init(osd_req, 2, CEPH_OSD_OP_WRITE, - offset, length, 0, 0); - if (orig_request->type == OBJ_REQUEST_BIO) - osd_req_op_extent_osd_data_bio(osd_req, 2, - orig_request->bio_list, length); - else - osd_req_op_extent_osd_data_pages(osd_req, 2, - orig_request->pages, length, - offset & ~PAGE_MASK, false, false); + /* Add the other op(s) */ - rbd_osd_req_format_write(orig_request); + op_type = rbd_img_request_op_type(orig_request->img_request); + rbd_img_obj_request_fill(orig_request, osd_req, op_type, 1); /* All set, send it off. */ @@ -2712,7 +2833,7 @@ static int rbd_img_obj_exists_submit(struct rbd_obj_request *obj_request) rbd_assert(obj_request->img_request); rbd_dev = obj_request->img_request->rbd_dev; - stat_request->osd_req = rbd_osd_req_create(rbd_dev, false, 1, + stat_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1, stat_request); if (!stat_request->osd_req) goto out; @@ -2732,11 +2853,10 @@ out: return ret; } -static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) +static bool img_obj_request_simple(struct rbd_obj_request *obj_request) { struct rbd_img_request *img_request; struct rbd_device *rbd_dev; - bool known; rbd_assert(obj_request_img_data_test(obj_request)); @@ -2744,22 +2864,44 @@ static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) rbd_assert(img_request); rbd_dev = img_request->rbd_dev; + /* Reads */ + if (!img_request_write_test(img_request) && + !img_request_discard_test(img_request)) + return true; + + /* Non-layered writes */ + if (!img_request_layered_test(img_request)) + return true; + /* - * Only writes to layered images need special handling. - * Reads and non-layered writes are simple object requests. - * Layered writes that start beyond the end of the overlap - * with the parent have no parent data, so they too are - * simple object requests. Finally, if the target object is - * known to already exist, its parent data has already been - * copied, so a write to the object can also be handled as a - * simple object request. + * Layered writes outside of the parent overlap range don't + * share any data with the parent. */ - if (!img_request_write_test(img_request) || - !img_request_layered_test(img_request) || - !obj_request_overlaps_parent(obj_request) || - ((known = obj_request_known_test(obj_request)) && - obj_request_exists_test(obj_request))) { + if (!obj_request_overlaps_parent(obj_request)) + return true; + + /* + * Entire-object layered writes - we will overwrite whatever + * parent data there is anyway. + */ + if (!obj_request->offset && + obj_request->length == rbd_obj_bytes(&rbd_dev->header)) + return true; + + /* + * If the object is known to already exist, its parent data has + * already been copied. + */ + if (obj_request_known_test(obj_request) && + obj_request_exists_test(obj_request)) + return true; + return false; +} + +static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) +{ + if (img_obj_request_simple(obj_request)) { struct rbd_device *rbd_dev; struct ceph_osd_client *osdc; @@ -2775,7 +2917,7 @@ static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) * start by reading the data for the full target object from * the parent so we can use it for a copyup to the target. */ - if (known) + if (obj_request_known_test(obj_request)) return rbd_img_obj_parent_read_full(obj_request); /* We don't know whether the target exists. Go find out. */ @@ -2916,7 +3058,7 @@ static int rbd_obj_notify_ack_sync(struct rbd_device *rbd_dev, u64 notify_id) return -ENOMEM; ret = -ENOMEM; - obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, 1, + obj_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1, obj_request); if (!obj_request->osd_req) goto out; @@ -2946,154 +3088,135 @@ static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) dout("%s: \"%s\" notify_id %llu opcode %u\n", __func__, rbd_dev->header_name, (unsigned long long)notify_id, (unsigned int)opcode); + + /* + * Until adequate refresh error handling is in place, there is + * not much we can do here, except warn. + * + * See http://tracker.ceph.com/issues/5040 + */ ret = rbd_dev_refresh(rbd_dev); if (ret) - rbd_warn(rbd_dev, "header refresh error (%d)\n", ret); + rbd_warn(rbd_dev, "refresh failed: %d", ret); - rbd_obj_notify_ack_sync(rbd_dev, notify_id); + ret = rbd_obj_notify_ack_sync(rbd_dev, notify_id); + if (ret) + rbd_warn(rbd_dev, "notify_ack ret %d", ret); } /* - * Initiate a watch request, synchronously. + * Send a (un)watch request and wait for the ack. Return a request + * with a ref held on success or error. */ -static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev) +static struct rbd_obj_request *rbd_obj_watch_request_helper( + struct rbd_device *rbd_dev, + bool watch) { struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; struct rbd_obj_request *obj_request; int ret; - rbd_assert(!rbd_dev->watch_event); - rbd_assert(!rbd_dev->watch_request); - - ret = ceph_osdc_create_event(osdc, rbd_watch_cb, rbd_dev, - &rbd_dev->watch_event); - if (ret < 0) - return ret; - - rbd_assert(rbd_dev->watch_event); - obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, OBJ_REQUEST_NODATA); - if (!obj_request) { - ret = -ENOMEM; - goto out_cancel; - } + if (!obj_request) + return ERR_PTR(-ENOMEM); - obj_request->osd_req = rbd_osd_req_create(rbd_dev, true, 1, + obj_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_WRITE, 1, obj_request); if (!obj_request->osd_req) { ret = -ENOMEM; - goto out_put; + goto out; } - ceph_osdc_set_request_linger(osdc, obj_request->osd_req); - osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_WATCH, - rbd_dev->watch_event->cookie, 0, 1); + rbd_dev->watch_event->cookie, 0, watch); rbd_osd_req_format_write(obj_request); + if (watch) + ceph_osdc_set_request_linger(osdc, obj_request->osd_req); + ret = rbd_obj_request_submit(osdc, obj_request); if (ret) - goto out_linger; + goto out; ret = rbd_obj_request_wait(obj_request); if (ret) - goto out_linger; + goto out; ret = obj_request->result; - if (ret) - goto out_linger; - - /* - * A watch request is set to linger, so the underlying osd - * request won't go away until we unregister it. We retain - * a pointer to the object request during that time (in - * rbd_dev->watch_request), so we'll keep a reference to - * it. We'll drop that reference (below) after we've - * unregistered it. - */ - rbd_dev->watch_request = obj_request; + if (ret) { + if (watch) + rbd_obj_request_end(obj_request); + goto out; + } - return 0; + return obj_request; -out_linger: - ceph_osdc_unregister_linger_request(osdc, obj_request->osd_req); -out_put: +out: rbd_obj_request_put(obj_request); -out_cancel: - ceph_osdc_cancel_event(rbd_dev->watch_event); - rbd_dev->watch_event = NULL; - - return ret; + return ERR_PTR(ret); } /* - * Tear down a watch request, synchronously. + * Initiate a watch request, synchronously. */ -static int __rbd_dev_header_unwatch_sync(struct rbd_device *rbd_dev) +static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev) { struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; struct rbd_obj_request *obj_request; int ret; - rbd_assert(rbd_dev->watch_event); - rbd_assert(rbd_dev->watch_request); + rbd_assert(!rbd_dev->watch_event); + rbd_assert(!rbd_dev->watch_request); - obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, - OBJ_REQUEST_NODATA); - if (!obj_request) { - ret = -ENOMEM; - goto out_cancel; - } + ret = ceph_osdc_create_event(osdc, rbd_watch_cb, rbd_dev, + &rbd_dev->watch_event); + if (ret < 0) + return ret; - obj_request->osd_req = rbd_osd_req_create(rbd_dev, true, 1, - obj_request); - if (!obj_request->osd_req) { - ret = -ENOMEM; - goto out_put; + obj_request = rbd_obj_watch_request_helper(rbd_dev, true); + if (IS_ERR(obj_request)) { + ceph_osdc_cancel_event(rbd_dev->watch_event); + rbd_dev->watch_event = NULL; + return PTR_ERR(obj_request); } - osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_WATCH, - rbd_dev->watch_event->cookie, 0, 0); - rbd_osd_req_format_write(obj_request); - - ret = rbd_obj_request_submit(osdc, obj_request); - if (ret) - goto out_put; + /* + * A watch request is set to linger, so the underlying osd + * request won't go away until we unregister it. We retain + * a pointer to the object request during that time (in + * rbd_dev->watch_request), so we'll keep a reference to it. + * We'll drop that reference after we've unregistered it in + * rbd_dev_header_unwatch_sync(). + */ + rbd_dev->watch_request = obj_request; - ret = rbd_obj_request_wait(obj_request); - if (ret) - goto out_put; + return 0; +} - ret = obj_request->result; - if (ret) - goto out_put; +/* + * Tear down a watch request, synchronously. + */ +static void rbd_dev_header_unwatch_sync(struct rbd_device *rbd_dev) +{ + struct rbd_obj_request *obj_request; - /* We have successfully torn down the watch request */ + rbd_assert(rbd_dev->watch_event); + rbd_assert(rbd_dev->watch_request); - ceph_osdc_unregister_linger_request(osdc, - rbd_dev->watch_request->osd_req); + rbd_obj_request_end(rbd_dev->watch_request); rbd_obj_request_put(rbd_dev->watch_request); rbd_dev->watch_request = NULL; -out_put: - rbd_obj_request_put(obj_request); -out_cancel: + obj_request = rbd_obj_watch_request_helper(rbd_dev, false); + if (!IS_ERR(obj_request)) + rbd_obj_request_put(obj_request); + else + rbd_warn(rbd_dev, "unable to tear down watch request (%ld)", + PTR_ERR(obj_request)); + ceph_osdc_cancel_event(rbd_dev->watch_event); rbd_dev->watch_event = NULL; - - return ret; -} - -static void rbd_dev_header_unwatch_sync(struct rbd_device *rbd_dev) -{ - int ret; - - ret = __rbd_dev_header_unwatch_sync(rbd_dev); - if (ret) { - rbd_warn(rbd_dev, "unable to tear down watch request: %d\n", - ret); - } } /* @@ -3136,7 +3259,7 @@ static int rbd_obj_method_sync(struct rbd_device *rbd_dev, obj_request->pages = pages; obj_request->page_count = page_count; - obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, 1, + obj_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1, obj_request); if (!obj_request->osd_req) goto out; @@ -3183,102 +3306,154 @@ out: return ret; } -static void rbd_request_fn(struct request_queue *q) - __releases(q->queue_lock) __acquires(q->queue_lock) +static void rbd_handle_request(struct rbd_device *rbd_dev, struct request *rq) { - struct rbd_device *rbd_dev = q->queuedata; - struct request *rq; + struct rbd_img_request *img_request; + struct ceph_snap_context *snapc = NULL; + u64 offset = (u64)blk_rq_pos(rq) << SECTOR_SHIFT; + u64 length = blk_rq_bytes(rq); + enum obj_operation_type op_type; + u64 mapping_size; int result; - while ((rq = blk_fetch_request(q))) { - bool write_request = rq_data_dir(rq) == WRITE; - struct rbd_img_request *img_request; - u64 offset; - u64 length; - - /* Ignore any non-FS requests that filter through. */ - - if (rq->cmd_type != REQ_TYPE_FS) { - dout("%s: non-fs request type %d\n", __func__, - (int) rq->cmd_type); - __blk_end_request_all(rq, 0); - continue; - } - - /* Ignore/skip any zero-length requests */ + if (rq->cmd_flags & REQ_DISCARD) + op_type = OBJ_OP_DISCARD; + else if (rq->cmd_flags & REQ_WRITE) + op_type = OBJ_OP_WRITE; + else + op_type = OBJ_OP_READ; - offset = (u64) blk_rq_pos(rq) << SECTOR_SHIFT; - length = (u64) blk_rq_bytes(rq); + /* Ignore/skip any zero-length requests */ - if (!length) { - dout("%s: zero-length request\n", __func__); - __blk_end_request_all(rq, 0); - continue; - } - - spin_unlock_irq(q->queue_lock); + if (!length) { + dout("%s: zero-length request\n", __func__); + result = 0; + goto err_rq; + } - /* Disallow writes to a read-only device */ + /* Only reads are allowed to a read-only device */ - if (write_request) { + if (op_type != OBJ_OP_READ) { + if (rbd_dev->mapping.read_only) { result = -EROFS; - if (rbd_dev->mapping.read_only) - goto end_request; - rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP); + goto err_rq; } + rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP); + } - /* - * Quit early if the mapped snapshot no longer - * exists. It's still possible the snapshot will - * have disappeared by the time our request arrives - * at the osd, but there's no sense in sending it if - * we already know. - */ - if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) { - dout("request for non-existent snapshot"); - rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); - result = -ENXIO; - goto end_request; - } + /* + * Quit early if the mapped snapshot no longer exists. It's + * still possible the snapshot will have disappeared by the + * time our request arrives at the osd, but there's no sense in + * sending it if we already know. + */ + if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) { + dout("request for non-existent snapshot"); + rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); + result = -ENXIO; + goto err_rq; + } + if (offset && length > U64_MAX - offset + 1) { + rbd_warn(rbd_dev, "bad request range (%llu~%llu)", offset, + length); result = -EINVAL; - if (offset && length > U64_MAX - offset + 1) { - rbd_warn(rbd_dev, "bad request range (%llu~%llu)\n", - offset, length); - goto end_request; /* Shouldn't happen */ - } + goto err_rq; /* Shouldn't happen */ + } + down_read(&rbd_dev->header_rwsem); + mapping_size = rbd_dev->mapping.size; + if (op_type != OBJ_OP_READ) { + snapc = rbd_dev->header.snapc; + ceph_get_snap_context(snapc); + } + up_read(&rbd_dev->header_rwsem); + + if (offset + length > mapping_size) { + rbd_warn(rbd_dev, "beyond EOD (%llu~%llu > %llu)", offset, + length, mapping_size); result = -EIO; - if (offset + length > rbd_dev->mapping.size) { - rbd_warn(rbd_dev, "beyond EOD (%llu~%llu > %llu)\n", - offset, length, rbd_dev->mapping.size); - goto end_request; - } + goto err_rq; + } + img_request = rbd_img_request_create(rbd_dev, offset, length, op_type, + snapc); + if (!img_request) { result = -ENOMEM; - img_request = rbd_img_request_create(rbd_dev, offset, length, - write_request); - if (!img_request) - goto end_request; - - img_request->rq = rq; + goto err_rq; + } + img_request->rq = rq; + if (op_type == OBJ_OP_DISCARD) + result = rbd_img_request_fill(img_request, OBJ_REQUEST_NODATA, + NULL); + else result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO, - rq->bio); - if (!result) - result = rbd_img_request_submit(img_request); - if (result) - rbd_img_request_put(img_request); -end_request: - spin_lock_irq(q->queue_lock); - if (result < 0) { - rbd_warn(rbd_dev, "%s %llx at %llx result %d\n", - write_request ? "write" : "read", - length, offset, result); - - __blk_end_request_all(rq, result); + rq->bio); + if (result) + goto err_img_request; + + result = rbd_img_request_submit(img_request); + if (result) + goto err_img_request; + + return; + +err_img_request: + rbd_img_request_put(img_request); +err_rq: + if (result) + rbd_warn(rbd_dev, "%s %llx at %llx result %d", + obj_op_name(op_type), length, offset, result); + if (snapc) + ceph_put_snap_context(snapc); + blk_end_request_all(rq, result); +} + +static void rbd_request_workfn(struct work_struct *work) +{ + struct rbd_device *rbd_dev = + container_of(work, struct rbd_device, rq_work); + struct request *rq, *next; + LIST_HEAD(requests); + + spin_lock_irq(&rbd_dev->lock); /* rq->q->queue_lock */ + list_splice_init(&rbd_dev->rq_queue, &requests); + spin_unlock_irq(&rbd_dev->lock); + + list_for_each_entry_safe(rq, next, &requests, queuelist) { + list_del_init(&rq->queuelist); + rbd_handle_request(rbd_dev, rq); + } +} + +/* + * Called with q->queue_lock held and interrupts disabled, possibly on + * the way to schedule(). Do not sleep here! + */ +static void rbd_request_fn(struct request_queue *q) +{ + struct rbd_device *rbd_dev = q->queuedata; + struct request *rq; + int queued = 0; + + rbd_assert(rbd_dev); + + while ((rq = blk_fetch_request(q))) { + /* Ignore any non-FS requests that filter through. */ + if (rq->cmd_type != REQ_TYPE_FS) { + dout("%s: non-fs request type %d\n", __func__, + (int) rq->cmd_type); + __blk_end_request_all(rq, 0); + continue; } + + list_add_tail(&rq->queuelist, &rbd_dev->rq_queue); + queued++; } + + if (queued) + queue_work(rbd_wq, &rbd_dev->rq_work); } /* @@ -3358,7 +3533,7 @@ static int rbd_obj_read_sync(struct rbd_device *rbd_dev, page_count = (u32) calc_pages_for(offset, length); pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); if (IS_ERR(pages)) - ret = PTR_ERR(pages); + return PTR_ERR(pages); ret = -ENOMEM; obj_request = rbd_obj_request_create(object_name, offset, length, @@ -3369,7 +3544,7 @@ static int rbd_obj_read_sync(struct rbd_device *rbd_dev, obj_request->pages = pages; obj_request->page_count = page_count; - obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, 1, + obj_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1, obj_request); if (!obj_request->osd_req) goto out; @@ -3517,24 +3692,37 @@ static int rbd_dev_refresh(struct rbd_device *rbd_dev) u64 mapping_size; int ret; - rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); down_write(&rbd_dev->header_rwsem); mapping_size = rbd_dev->mapping.size; - if (rbd_dev->image_format == 1) - ret = rbd_dev_v1_header_info(rbd_dev); - else - ret = rbd_dev_v2_header_info(rbd_dev); - /* If it's a mapped snapshot, validate its EXISTS flag */ + ret = rbd_dev_header_info(rbd_dev); + if (ret) + return ret; + + /* + * If there is a parent, see if it has disappeared due to the + * mapped image getting flattened. + */ + if (rbd_dev->parent) { + ret = rbd_dev_v2_parent_info(rbd_dev); + if (ret) + return ret; + } + + if (rbd_dev->spec->snap_id == CEPH_NOSNAP) { + if (rbd_dev->mapping.size != rbd_dev->header.image_size) + rbd_dev->mapping.size = rbd_dev->header.image_size; + } else { + /* validate mapped snapshot's EXISTS flag */ + rbd_exists_validate(rbd_dev); + } - rbd_exists_validate(rbd_dev); up_write(&rbd_dev->header_rwsem); - if (mapping_size != rbd_dev->mapping.size) { + if (mapping_size != rbd_dev->mapping.size) rbd_dev_update_size(rbd_dev); - } - return ret; + return 0; } static int rbd_init_disk(struct rbd_device *rbd_dev) @@ -3573,6 +3761,13 @@ static int rbd_init_disk(struct rbd_device *rbd_dev) blk_queue_io_min(q, segment_size); blk_queue_io_opt(q, segment_size); + /* enable the discard support */ + queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); + q->limits.discard_granularity = segment_size; + q->limits.discard_alignment = segment_size; + q->limits.max_discard_sectors = segment_size / SECTOR_SIZE; + q->limits.discard_zeroes_data = 1; + blk_queue_merge_bvec(q, rbd_merge_bvec); disk->queue = q; @@ -3696,46 +3891,36 @@ static ssize_t rbd_snap_show(struct device *dev, } /* - * For an rbd v2 image, shows the pool id, image id, and snapshot id - * for the parent image. If there is no parent, simply shows - * "(no parent image)". + * For a v2 image, shows the chain of parent images, separated by empty + * lines. For v1 images or if there is no parent, shows "(no parent + * image)". */ static ssize_t rbd_parent_show(struct device *dev, - struct device_attribute *attr, - char *buf) + struct device_attribute *attr, + char *buf) { struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); - struct rbd_spec *spec = rbd_dev->parent_spec; - int count; - char *bufp = buf; + ssize_t count = 0; - if (!spec) + if (!rbd_dev->parent) return sprintf(buf, "(no parent image)\n"); - count = sprintf(bufp, "pool_id %llu\npool_name %s\n", - (unsigned long long) spec->pool_id, spec->pool_name); - if (count < 0) - return count; - bufp += count; - - count = sprintf(bufp, "image_id %s\nimage_name %s\n", spec->image_id, - spec->image_name ? spec->image_name : "(unknown)"); - if (count < 0) - return count; - bufp += count; - - count = sprintf(bufp, "snap_id %llu\nsnap_name %s\n", - (unsigned long long) spec->snap_id, spec->snap_name); - if (count < 0) - return count; - bufp += count; - - count = sprintf(bufp, "overlap %llu\n", rbd_dev->parent_overlap); - if (count < 0) - return count; - bufp += count; + for ( ; rbd_dev->parent; rbd_dev = rbd_dev->parent) { + struct rbd_spec *spec = rbd_dev->parent_spec; + + count += sprintf(&buf[count], "%s" + "pool_id %llu\npool_name %s\n" + "image_id %s\nimage_name %s\n" + "snap_id %llu\nsnap_name %s\n" + "overlap %llu\n", + !count ? "" : "\n", /* first? */ + spec->pool_id, spec->pool_name, + spec->image_id, spec->image_name ?: "(unknown)", + spec->snap_id, spec->snap_name, + rbd_dev->parent_overlap); + } - return (ssize_t) (bufp - buf); + return count; } static ssize_t rbd_image_refresh(struct device *dev, @@ -3748,9 +3933,9 @@ static ssize_t rbd_image_refresh(struct device *dev, ret = rbd_dev_refresh(rbd_dev); if (ret) - rbd_warn(rbd_dev, ": manual header refresh error (%d)\n", ret); + return ret; - return ret < 0 ? ret : size; + return size; } static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); @@ -3822,6 +4007,9 @@ static struct rbd_spec *rbd_spec_alloc(void) spec = kzalloc(sizeof (*spec), GFP_KERNEL); if (!spec) return NULL; + + spec->pool_id = CEPH_NOPOOL; + spec->snap_id = CEPH_NOSNAP; kref_init(&spec->kref); return spec; @@ -3848,6 +4036,8 @@ static struct rbd_device *rbd_dev_create(struct rbd_client *rbdc, return NULL; spin_lock_init(&rbd_dev->lock); + INIT_LIST_HEAD(&rbd_dev->rq_queue); + INIT_WORK(&rbd_dev->rq_work, rbd_request_workfn); rbd_dev->flags = 0; atomic_set(&rbd_dev->parent_ref, 0); INIT_LIST_HEAD(&rbd_dev->node); @@ -4021,7 +4211,7 @@ static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) goto out_err; } - snapid = cpu_to_le64(CEPH_NOSNAP); + snapid = cpu_to_le64(rbd_dev->spec->snap_id); ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, "rbd", "get_parent", &snapid, sizeof (snapid), @@ -4059,7 +4249,7 @@ static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) ret = -EIO; if (pool_id > (u64)U32_MAX) { - rbd_warn(NULL, "parent pool id too large (%llu > %u)\n", + rbd_warn(NULL, "parent pool id too large (%llu > %u)", (unsigned long long)pool_id, U32_MAX); goto out_err; } @@ -4083,6 +4273,8 @@ static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) parent_spec->snap_id = snap_id; rbd_dev->parent_spec = parent_spec; parent_spec = NULL; /* rbd_dev now owns this */ + } else { + kfree(image_id); } /* @@ -4110,8 +4302,7 @@ static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) * overlap is zero we just pretend there was * no parent image. */ - rbd_warn(rbd_dev, "ignoring parent of " - "clone with overlap 0\n"); + rbd_warn(rbd_dev, "ignoring parent with overlap 0"); } } out: @@ -4279,18 +4470,38 @@ static u64 rbd_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) } /* - * When an rbd image has a parent image, it is identified by the - * pool, image, and snapshot ids (not names). This function fills - * in the names for those ids. (It's OK if we can't figure out the - * name for an image id, but the pool and snapshot ids should always - * exist and have names.) All names in an rbd spec are dynamically - * allocated. + * An image being mapped will have everything but the snap id. + */ +static int rbd_spec_fill_snap_id(struct rbd_device *rbd_dev) +{ + struct rbd_spec *spec = rbd_dev->spec; + + rbd_assert(spec->pool_id != CEPH_NOPOOL && spec->pool_name); + rbd_assert(spec->image_id && spec->image_name); + rbd_assert(spec->snap_name); + + if (strcmp(spec->snap_name, RBD_SNAP_HEAD_NAME)) { + u64 snap_id; + + snap_id = rbd_snap_id_by_name(rbd_dev, spec->snap_name); + if (snap_id == CEPH_NOSNAP) + return -ENOENT; + + spec->snap_id = snap_id; + } else { + spec->snap_id = CEPH_NOSNAP; + } + + return 0; +} + +/* + * A parent image will have all ids but none of the names. * - * When an image being mapped (not a parent) is probed, we have the - * pool name and pool id, image name and image id, and the snapshot - * name. The only thing we're missing is the snapshot id. + * All names in an rbd spec are dynamically allocated. It's OK if we + * can't figure out the name for an image id. */ -static int rbd_dev_spec_update(struct rbd_device *rbd_dev) +static int rbd_spec_fill_names(struct rbd_device *rbd_dev) { struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; struct rbd_spec *spec = rbd_dev->spec; @@ -4299,24 +4510,9 @@ static int rbd_dev_spec_update(struct rbd_device *rbd_dev) const char *snap_name; int ret; - /* - * An image being mapped will have the pool name (etc.), but - * we need to look up the snapshot id. - */ - if (spec->pool_name) { - if (strcmp(spec->snap_name, RBD_SNAP_HEAD_NAME)) { - u64 snap_id; - - snap_id = rbd_snap_id_by_name(rbd_dev, spec->snap_name); - if (snap_id == CEPH_NOSNAP) - return -ENOENT; - spec->snap_id = snap_id; - } else { - spec->snap_id = CEPH_NOSNAP; - } - - return 0; - } + rbd_assert(spec->pool_id != CEPH_NOPOOL); + rbd_assert(spec->image_id); + rbd_assert(spec->snap_id != CEPH_NOSNAP); /* Get the pool name; we have to make our own copy of this */ @@ -4335,7 +4531,7 @@ static int rbd_dev_spec_update(struct rbd_device *rbd_dev) if (!image_name) rbd_warn(rbd_dev, "unable to get image name"); - /* Look up the snapshot name, and make a copy */ + /* Fetch the snapshot name */ snap_name = rbd_snap_name(rbd_dev, spec->snap_id); if (IS_ERR(snap_name)) { @@ -4348,10 +4544,10 @@ static int rbd_dev_spec_update(struct rbd_device *rbd_dev) spec->snap_name = snap_name; return 0; + out_err: kfree(image_name); kfree(pool_name); - return ret; } @@ -4483,43 +4679,22 @@ static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev) return ret; } - /* - * If the image supports layering, get the parent info. We - * need to probe the first time regardless. Thereafter we - * only need to if there's a parent, to see if it has - * disappeared due to the mapped image getting flattened. - */ - if (rbd_dev->header.features & RBD_FEATURE_LAYERING && - (first_time || rbd_dev->parent_spec)) { - bool warn; - - ret = rbd_dev_v2_parent_info(rbd_dev); - if (ret) - return ret; - - /* - * Print a warning if this is the initial probe and - * the image has a parent. Don't print it if the - * image now being probed is itself a parent. We - * can tell at this point because we won't know its - * pool name yet (just its pool id). - */ - warn = rbd_dev->parent_spec && rbd_dev->spec->pool_name; - if (first_time && warn) - rbd_warn(rbd_dev, "WARNING: kernel layering " - "is EXPERIMENTAL!"); - } - - if (rbd_dev->spec->snap_id == CEPH_NOSNAP) - if (rbd_dev->mapping.size != rbd_dev->header.image_size) - rbd_dev->mapping.size = rbd_dev->header.image_size; - ret = rbd_dev_v2_snap_context(rbd_dev); dout("rbd_dev_v2_snap_context returned %d\n", ret); return ret; } +static int rbd_dev_header_info(struct rbd_device *rbd_dev) +{ + rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); + + if (rbd_dev->image_format == 1) + return rbd_dev_v1_header_info(rbd_dev); + + return rbd_dev_v2_header_info(rbd_dev); +} + static int rbd_bus_add_dev(struct rbd_device *rbd_dev) { struct device *dev; @@ -4907,7 +5082,7 @@ static int rbd_dev_image_id(struct rbd_device *rbd_dev) ret = image_id ? 0 : -ENOMEM; if (!ret) rbd_dev->image_format = 1; - } else if (ret > sizeof (__le32)) { + } else if (ret >= 0) { void *p = response; image_id = ceph_extract_encoded_string(&p, p + ret, @@ -4915,8 +5090,6 @@ static int rbd_dev_image_id(struct rbd_device *rbd_dev) ret = PTR_ERR_OR_ZERO(image_id); if (!ret) rbd_dev->image_format = 2; - } else { - ret = -EINVAL; } if (!ret) { @@ -5066,6 +5239,7 @@ static int rbd_dev_device_setup(struct rbd_device *rbd_dev) ret = rbd_dev_mapping_set(rbd_dev); if (ret) goto err_out_disk; + set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); set_disk_ro(rbd_dev->disk, rbd_dev->mapping.read_only); @@ -5155,8 +5329,6 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, bool mapping) ret = rbd_dev_image_id(rbd_dev); if (ret) return ret; - rbd_assert(rbd_dev->spec->image_id); - rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); ret = rbd_dev_header_name(rbd_dev); if (ret) @@ -5168,25 +5340,45 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, bool mapping) goto out_header_name; } - if (rbd_dev->image_format == 1) - ret = rbd_dev_v1_header_info(rbd_dev); - else - ret = rbd_dev_v2_header_info(rbd_dev); + ret = rbd_dev_header_info(rbd_dev); if (ret) goto err_out_watch; - ret = rbd_dev_spec_update(rbd_dev); + /* + * If this image is the one being mapped, we have pool name and + * id, image name and id, and snap name - need to fill snap id. + * Otherwise this is a parent image, identified by pool, image + * and snap ids - need to fill in names for those ids. + */ + if (mapping) + ret = rbd_spec_fill_snap_id(rbd_dev); + else + ret = rbd_spec_fill_names(rbd_dev); if (ret) goto err_out_probe; + if (rbd_dev->header.features & RBD_FEATURE_LAYERING) { + ret = rbd_dev_v2_parent_info(rbd_dev); + if (ret) + goto err_out_probe; + + /* + * Need to warn users if this image is the one being + * mapped and has a parent. + */ + if (mapping && rbd_dev->parent_spec) + rbd_warn(rbd_dev, + "WARNING: kernel layering is EXPERIMENTAL!"); + } + ret = rbd_dev_probe_parent(rbd_dev); if (ret) goto err_out_probe; dout("discovered format %u image, header name is %s\n", rbd_dev->image_format, rbd_dev->header_name); - return 0; + err_out_probe: rbd_dev_unprobe(rbd_dev); err_out_watch: @@ -5199,9 +5391,6 @@ err_out_format: rbd_dev->image_format = 0; kfree(rbd_dev->spec->image_id); rbd_dev->spec->image_id = NULL; - - dout("probe failed, returning %d\n", ret); - return ret; } @@ -5243,7 +5432,7 @@ static ssize_t do_rbd_add(struct bus_type *bus, /* The ceph file layout needs to fit pool id in 32 bits */ if (spec->pool_id > (u64)U32_MAX) { - rbd_warn(NULL, "pool id too large (%llu > %u)\n", + rbd_warn(NULL, "pool id too large (%llu > %u)", (unsigned long long)spec->pool_id, U32_MAX); rc = -EIO; goto err_out_client; @@ -5517,11 +5706,21 @@ static int __init rbd_init(void) if (rc) return rc; + /* + * The number of active work items is limited by the number of + * rbd devices, so leave @max_active at default. + */ + rbd_wq = alloc_workqueue(RBD_DRV_NAME, WQ_MEM_RECLAIM, 0); + if (!rbd_wq) { + rc = -ENOMEM; + goto err_out_slab; + } + if (single_major) { rbd_major = register_blkdev(0, RBD_DRV_NAME); if (rbd_major < 0) { rc = rbd_major; - goto err_out_slab; + goto err_out_wq; } } @@ -5539,6 +5738,8 @@ static int __init rbd_init(void) err_out_blkdev: if (single_major) unregister_blkdev(rbd_major, RBD_DRV_NAME); +err_out_wq: + destroy_workqueue(rbd_wq); err_out_slab: rbd_slab_exit(); return rc; @@ -5550,6 +5751,7 @@ static void __exit rbd_exit(void) rbd_sysfs_cleanup(); if (single_major) unregister_blkdev(rbd_major, RBD_DRV_NAME); + destroy_workqueue(rbd_wq); rbd_slab_exit(); } |