Viewing: osp_md_object.c
// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (c) 2013, 2017, Intel Corporation.
*/
/*
* OST/MDT proxy device (OSP) Metadata methods
*
* This file implements methods for remote MD object, which include
* dt_object_operations, dt_index_operations and dt_body_operations.
*
* If there are multiple MDTs in one filesystem, one operation might
* include modifications in several MDTs. In such cases, clients
* send the RPC to the master MDT, then the operation is decomposed into
* object updates which will be dispatched to OSD or OSP. The local updates
* go to local OSD and the remote updates go to OSP. In OSP, these remote
* object updates will be packed into an update RPC, sent to the remote MDT
* and handled by Object Update Target (OUT).
*
* In DNE phase I, because of missing complete recovery solution, updates
* will be executed in order and synchronously.
* 1. The transaction is created.
* 2. In transaction declare, it collects and packs remote
* updates (in osp_md_declare_xxx()).
* 3. In transaction start, it sends these remote updates
* to remote MDTs, which will execute these updates synchronously.
* 4. In transaction execute phase, the local updates will be executed
* synchronously.
*
* Author: Di Wang <di.wang@intel.com>
*/
#define DEBUG_SUBSYSTEM S_MDS
#include <llog_swab.h>
#include <lustre_log.h>
#include "osp_internal.h"
#define OUT_UPDATE_BUFFER_SIZE_ADD 4096
#define OUT_UPDATE_BUFFER_SIZE_MAX (256 * 4096) /* 1M update size now */
/**
* osp_create_interpreter() - Interpreter call for object creation
* @env: execution environment
* @reply: update reply
* @req: ptlrpc update request for creating object
* @obj: object to be created
* @data: data used in this function.
* @index: index(position) of create update in the whole updates
* @rc: update result on the remote MDT.
*
* Object creation interpreter, which will be called after creating
* the remote object to set flags and status.
*
* Return only return 0 for now
*/
static int osp_create_interpreter(const struct lu_env *env,
struct object_update_reply *reply,
struct ptlrpc_request *req,
struct osp_object *obj,
void *data, int index, int rc)
{
struct osp_device *osp = lu2osp_dev(obj->opo_obj.do_lu.lo_dev);
spin_lock(&obj->opo_lock);
if (rc != 0 && rc != -EEXIST) {
obj->opo_obj.do_lu.lo_header->loh_attr &= ~LOHA_EXISTS;
obj->opo_non_exist = 1;
}
obj->opo_creating = 0;
spin_unlock(&obj->opo_lock);
/*
* invalidate opo cache for the object after the object is created, so
* attr_get will try to get attr from remote object.
*/
osp_obj_invalidate_cache(obj);
/*
* currently reads from objects being created
* are exceptional - during recovery only, when
* remote llog update fetching can race with
* orphan cleanup. so don't waste memory adding
* a wait queue to every osp object
*/
wake_up_all(&osp->opd_out_waitq);
return 0;
}
/**
* osp_md_declare_create() - Implementation of do_declare_create
* @env: execution environment
* @dt: remote object to be created
* @attr: attribute of the created object
* @hint: creation hint
* @dof: creation format information
* @th: the transaction handle
*
* Implementation of dt_object_operations::do_declare_create
* Create the osp_update_request to track the update for this OSP
* in the transaction.
*
* Return:
* * %0 if preparation succeeds.
* * %negative errno if preparation fails.
*/
int osp_md_declare_create(const struct lu_env *env, struct dt_object *dt,
struct lu_attr *attr, struct dt_allocation_hint *hint,
struct dt_object_format *dof, struct thandle *th)
{
return osp_trans_update_request_create(th);
}
struct object_update *
update_buffer_get_update(struct object_update_request *request,
unsigned int index)
{
void *ptr;
int i;
if (index > request->ourq_count)
return NULL;
ptr = &request->ourq_updates[0];
for (i = 0; i < index; i++)
ptr += object_update_size(ptr);
return ptr;
}
/**
* osp_md_create() - Implementation of dt_object_operations::do_create
* @env: execution environment
* @dt: object to be created
* @attr: attribute of the created object
* @hint: creation hint
* @dof: creation format information
* @th: the transaction handle
*
* It adds an OUT_CREATE sub-request into the OUT RPC that will be flushed
* when the transaction stop, and sets necessary flags for created object.
*
* Return:
* * %0 if packing creation succeeds.
* * %negative errno if packing creation fails.
*/
int osp_md_create(const struct lu_env *env, struct dt_object *dt,
struct lu_attr *attr, struct dt_allocation_hint *hint,
struct dt_object_format *dof, struct thandle *th)
{
struct osp_update_request *update;
struct osp_object *obj = dt2osp_obj(dt);
int rc;
update = thandle_to_osp_update_request(th);
LASSERT(update != NULL);
if (!(attr->la_valid & LA_TYPE)) {
CERROR("%s: create type not specified: valid %llx\n",
dt->do_lu.lo_dev->ld_obd->obd_name, attr->la_valid);
GOTO(out, rc = -EINVAL);
}
rc = OSP_UPDATE_RPC_PACK(env, out_create_pack, update,
lu_object_fid(&dt->do_lu), attr, hint, dof);
if (rc != 0)
GOTO(out, rc);
rc = osp_insert_update_callback(env, update, dt2osp_obj(dt), NULL,
osp_create_interpreter);
if (rc < 0)
GOTO(out, rc);
spin_lock(&obj->opo_lock);
obj->opo_creating = 1;
dt->do_lu.lo_header->loh_attr |= LOHA_EXISTS | (attr->la_mode & S_IFMT);
dt2osp_obj(dt)->opo_non_exist = 0;
obj->opo_stale = 0;
spin_unlock(&obj->opo_lock);
obj->opo_attr = *attr;
out:
return rc;
}
/**
* osp_md_declare_ref_del() - Implementation of do_declare_ref_del
* @env: execution environment
* @dt: object to decrease the reference count.
* @th: the transaction handle of refcount decrease.
*
* Implementation of dt_object_operations::do_declare_ref_del
* Create the osp_update_request to track the update for this OSP
* in the transaction.
*
* Return:
* * %0 if preparation succeeds.
* * %negative errno if preparation fails.
*/
static int osp_md_declare_ref_del(const struct lu_env *env,
struct dt_object *dt, struct thandle *th)
{
return osp_trans_update_request_create(th);
}
/**
* osp_md_ref_del() - Implementation of dt_object_operations::do_ref_del
* @env: execution environment
* @dt: object to decrease the reference count
* @th: the transaction handle
*
* Add an OUT_REF_DEL sub-request into the OUT RPC that will be
* flushed when the transaction stop.
*
* Return:
* * %0 if packing ref_del succeeds.
* * %negative errno if packing fails.
*/
static int osp_md_ref_del(const struct lu_env *env, struct dt_object *dt,
struct thandle *th)
{
struct osp_update_request *update;
int rc;
update = thandle_to_osp_update_request(th);
LASSERT(update != NULL);
rc = OSP_UPDATE_RPC_PACK(env, out_ref_del_pack, update,
lu_object_fid(&dt->do_lu));
return rc;
}
/**
* osp_md_declare_ref_add() - Implementation of do_declare_ref_del
* @env: execution environment
* @dt: object on which to increase the reference count.
* @th: the transaction handle.
*
* Implementation of dt_object_operations::do_declare_ref_del
* Create the osp_update_request to track the update for this OSP
* in the transaction.
*
* Return:
* * %0 if preparation succeeds.
* * %negative errno if preparation fails.
*/
static int osp_md_declare_ref_add(const struct lu_env *env,
struct dt_object *dt, struct thandle *th)
{
return osp_trans_update_request_create(th);
}
/**
* osp_md_ref_add() - Implementation of dt_object_operations::do_ref_add
* @env: execution environment
* @dt: object on which to increase the reference count
* @th: the transaction handle
*
* Add an OUT_REF_ADD sub-request into the OUT RPC that will be flushed
* when the transaction stop.
*
* Return:
* * %0 if packing ref_add succeeds.
* * %negative errno if packing fails.
*/
static int osp_md_ref_add(const struct lu_env *env, struct dt_object *dt,
struct thandle *th)
{
struct osp_update_request *update;
int rc;
update = thandle_to_osp_update_request(th);
LASSERT(update != NULL);
rc = OSP_UPDATE_RPC_PACK(env, out_ref_add_pack, update,
lu_object_fid(&dt->do_lu));
return rc;
}
/**
* osp_md_ah_init() - Implementation of dt_object_operations::do_ah_init
* @env: execution environment
* @ah: the hint to be initialized
* @parent: the parent of the object
* @child: the object to be created
* @child_mode: the mode of the created object
*
* Initialize the allocation hint for object creation, which is usually called
* before the creation, and these hints (parent and child mode) will be sent to
* the remote Object Update Target (OUT) and used in the object create process,
* same as OSD object creation.
*/
static void osp_md_ah_init(const struct lu_env *env,
struct dt_allocation_hint *ah,
struct dt_object *parent,
struct dt_object *child,
umode_t child_mode)
{
LASSERT(ah);
ah->dah_parent = parent;
}
/**
* osp_md_declare_attr_set() - Implementation of do_declare_attr_get
* @env: execution environment
* @dt: object on which to set attributes
* @attr: attributes to be set
* @th: the transaction handle
*
* Implementation of dt_object_operations::do_declare_attr_get
* Create the osp_update_request to track the update for this OSP
* in the transaction.
*
* Return:
* * %0 if preparation succeeds.
* * %negative errno if preparation fails.
*/
int osp_md_declare_attr_set(const struct lu_env *env, struct dt_object *dt,
const struct lu_attr *attr, struct thandle *th)
{
return osp_trans_update_request_create(th);
}
/**
* osp_md_attr_set() - Implementation of dt_object_operations::do_attr_set
* @env: execution environment
* @dt: object to set attributes
* @attr: attributes to be set
* @th: the transaction handle
*
* Set attributes to the specified remote object.
*
* Add the OUT_ATTR_SET sub-request into the OUT RPC that will be flushed
* when the transaction stop.
*
* Return:
* * %0 if packing attr_set succeeds.
* * %negative errno if packing fails.
*/
int osp_md_attr_set(const struct lu_env *env, struct dt_object *dt,
const struct lu_attr *attr, struct thandle *th)
{
struct osp_update_request *update;
int rc;
update = thandle_to_osp_update_request(th);
LASSERT(update != NULL);
rc = OSP_UPDATE_RPC_PACK(env, out_attr_set_pack, update,
lu_object_fid(&dt->do_lu), attr);
return rc;
}
/**
* osp_md_read_lock() - Implementation of dt_object_operations::do_read_lock
* @env: execution environment
* @dt: object to be locked
* @role: lock role from MDD layer, see dt_object_role().
*
* osp_md_{read,write}_lock() will only lock the remote object in the
* local cache, which uses the semaphore (opo_sem) inside the osp_object to
* lock the object. Note: it will not lock the object in the whole cluster,
* which relies on the LDLM lock.
*/
static void osp_md_read_lock(const struct lu_env *env, struct dt_object *dt,
unsigned int role)
{
struct osp_object *obj = dt2osp_obj(dt);
LASSERT(obj->opo_owner != env);
down_read_nested(&obj->opo_sem, role);
LASSERT(obj->opo_owner == NULL);
}
/**
* osp_md_write_lock() - Implementation of dt_object_operations::do_write_lock
* @env: execution environment
* @dt: object to be locked
* @role: lock role from MDD layer, see dt_object_role().
*
* Lock the remote object in write mode.
*/
static void osp_md_write_lock(const struct lu_env *env, struct dt_object *dt,
unsigned int role)
{
struct osp_object *obj = dt2osp_obj(dt);
down_write_nested(&obj->opo_sem, role);
LASSERT(obj->opo_owner == NULL);
obj->opo_owner = env;
}
/**
* osp_md_read_unlock() - Implementation of dt_object_operations::do_read_unlock
* @env: execution environment
* @dt: object to be unlocked
*
* Unlock the read lock of remote object.
*/
static void osp_md_read_unlock(const struct lu_env *env, struct dt_object *dt)
{
struct osp_object *obj = dt2osp_obj(dt);
up_read(&obj->opo_sem);
}
/**
* osp_md_write_unlock() - Implementation of do_write_unlock
* @env: execution environment
* @dt: object to be unlocked
*
* Implementation of dt_object_operations::do_write_unlock
* Unlock the write lock of remote object.
*/
static void osp_md_write_unlock(const struct lu_env *env, struct dt_object *dt)
{
struct osp_object *obj = dt2osp_obj(dt);
LASSERT(obj->opo_owner == env);
obj->opo_owner = NULL;
up_write(&obj->opo_sem);
}
/**
* osp_md_write_locked() - Implementation of do_write_locked
* @env: execution environment
* @dt: object to be tested
*
* Implementation of dt_object_operations::do_write_locked
* Test if the object is locked in write mode.
*
* Return %1 is object is locked else %0
*/
static int osp_md_write_locked(const struct lu_env *env, struct dt_object *dt)
{
struct osp_object *obj = dt2osp_obj(dt);
return obj->opo_owner == env;
}
/**
* osp_md_index_lookup() - Implementation of dt_index_operations::dio_lookup
* @env: execution environment
* @dt: index object to lookup
* @rec: record in which to return lookup result [out]
* @key: key of index which will be looked up
*
* Look up record by key under a remote index object. It packs lookup update
* into RPC, sends to the remote OUT and waits for the lookup result.
*
* Return:
* * %1 if the lookup succeeds.
* * %negative errno if the lookup fails.
*/
static int osp_md_index_lookup(const struct lu_env *env, struct dt_object *dt,
struct dt_rec *rec, const struct dt_key *key)
{
struct lu_buf *lbuf = &osp_env_info(env)->osi_lb2;
struct osp_device *osp = lu2osp_dev(dt->do_lu.lo_dev);
struct dt_device *dt_dev = &osp->opd_dt_dev;
struct osp_update_request *update;
struct object_update_reply *reply;
struct ptlrpc_request *req = NULL;
struct lu_fid *fid;
int rc;
ENTRY;
/* Because it needs send the update buffer right away,
* just create an update buffer, instead of attaching the
* update_remote list of the thandle.
*/
update = osp_update_request_create(dt_dev);
if (IS_ERR(update))
RETURN(PTR_ERR(update));
rc = OSP_UPDATE_RPC_PACK(env, out_index_lookup_pack, update,
lu_object_fid(&dt->do_lu), rec, key);
if (rc != 0) {
CERROR("%s: Insert update error: rc = %d\n",
dt_dev->dd_lu_dev.ld_obd->obd_name, rc);
GOTO(out, rc);
}
rc = osp_remote_sync(env, osp, update, &req);
if (rc < 0)
GOTO(out, rc);
reply = req_capsule_server_sized_get(&req->rq_pill,
&RMF_OUT_UPDATE_REPLY,
OUT_UPDATE_REPLY_SIZE);
if (reply->ourp_magic != UPDATE_REPLY_MAGIC) {
CERROR("%s: Wrong version %x expected %x: rc = %d\n",
dt_dev->dd_lu_dev.ld_obd->obd_name,
reply->ourp_magic, UPDATE_REPLY_MAGIC, -EPROTO);
GOTO(out, rc = -EPROTO);
}
rc = object_update_result_data_get(reply, lbuf, 0);
if (rc < 0)
GOTO(out, rc);
if (lbuf->lb_len != sizeof(*fid)) {
CERROR("%s: lookup "DFID" %s wrong size %d\n",
dt_dev->dd_lu_dev.ld_obd->obd_name,
PFID(lu_object_fid(&dt->do_lu)), (char *)key,
(int)lbuf->lb_len);
GOTO(out, rc = -EINVAL);
}
fid = lbuf->lb_buf;
if (req_capsule_rep_need_swab(&req->rq_pill))
lustre_swab_lu_fid(fid);
if (!fid_is_sane(fid)) {
CERROR("%s: lookup "DFID" %s invalid fid "DFID"\n",
dt_dev->dd_lu_dev.ld_obd->obd_name,
PFID(lu_object_fid(&dt->do_lu)), (char *)key, PFID(fid));
GOTO(out, rc = -EINVAL);
}
memcpy(rec, fid, sizeof(*fid));
GOTO(out, rc = 1);
out:
if (req != NULL)
ptlrpc_req_put(req);
osp_update_request_destroy(env, update);
return rc;
}
/**
* osp_md_declare_index_insert() - Implementation of dio_declare_insert
* @env: execution environment
* @dt: object for which to insert index
* @rec: record of the index which will be inserted
* @key: key of the index which will be inserted
* @th: the transaction handle
*
* Implementation of dt_index_operations::dio_declare_insert
* Create the osp_update_request to track the update for this OSP
* in the transaction.
*
* Return:
* * %0 if preparation succeeds.
* * %negative errno if preparation fails.
*/
static int osp_md_declare_index_insert(const struct lu_env *env,
struct dt_object *dt,
const struct dt_rec *rec,
const struct dt_key *key,
struct thandle *th)
{
return osp_trans_update_request_create(th);
}
/**
* osp_md_index_insert() - Implementation of dt_index_operations::dio_insert
* @env: execution environment
* @dt: object for which to insert index
* @rec: record of the index to be inserted
* @key: key of the index to be inserted
* @th: the transaction handle
*
* Add an OUT_INDEX_INSERT sub-request into the OUT RPC that will
* be flushed when the transaction stop.
*
* Return:
* * %0 if packing index insert succeeds.
* * %negative errno if packing fails.
*/
static int osp_md_index_insert(const struct lu_env *env, struct dt_object *dt,
const struct dt_rec *rec,
const struct dt_key *key, struct thandle *th)
{
struct osp_update_request *update;
int rc;
update = thandle_to_osp_update_request(th);
LASSERT(update != NULL);
rc = OSP_UPDATE_RPC_PACK(env, out_index_insert_pack, update,
lu_object_fid(&dt->do_lu), rec, key);
return rc;
}
/**
* osp_md_declare_index_delete() - Implementation of dio_declare_delete
* @env: execution environment
* @dt: object for which to delete index
* @key: key of the index
* @th: the transaction handle
*
* Implementation of dt_index_operations::dio_declare_delete
* Create the osp_update_request to track the update for this OSP
* in the transaction.
*
* Return:
* * %0 if preparation succeeds.
* * %negative errno if preparation fails.
*/
static int osp_md_declare_index_delete(const struct lu_env *env,
struct dt_object *dt,
const struct dt_key *key,
struct thandle *th)
{
return osp_trans_update_request_create(th);
}
/**
* osp_md_index_delete() - Implementation of dt_index_operations::dio_delete
* @env: execution environment
* @dt: object for which to delete index
* @key: key of the index which will be deleted
* @th: the transaction handle
*
* Add an OUT_INDEX_DELETE sub-request into the OUT RPC that will
* be flushed when the transaction stop.
*
* Return:
* * %0 if packing index delete succeeds.
* * %negative errno if packing fails.
*/
static int osp_md_index_delete(const struct lu_env *env,
struct dt_object *dt,
const struct dt_key *key,
struct thandle *th)
{
struct osp_update_request *update;
int rc;
update = thandle_to_osp_update_request(th);
LASSERT(update != NULL);
rc = OSP_UPDATE_RPC_PACK(env, out_index_delete_pack, update,
lu_object_fid(&dt->do_lu), key);
return rc;
}
/**
* osp_md_index_it_next() - Implementation of dt_index_operations::dio_it.next
* @env: execution environment
* @di: iterator of this iteration
*
* Advance the pointer of the iterator to the next entry. It shares a similar
* internal implementation with osp_orphan_it_next(), which is being used for
* remote orphan index object. This method will be used for remote directory.
*
* Return:
* * %0 if the pointer is advanced successfully.
* * %1 if it reaches to the end of the index object.
* * %negative errno if the pointer cannot be advanced.
*/
static int osp_md_index_it_next(const struct lu_env *env, struct dt_it *di)
{
struct osp_it *it = (struct osp_it *)di;
struct lu_idxpage *idxpage;
struct lu_dirent *ent;
int rc;
ENTRY;
again:
idxpage = it->ooi_cur_idxpage;
if (idxpage) {
if (idxpage->lip_nr == 0)
RETURN(1);
it->ooi_pos_ent++;
ent = (struct lu_dirent *)it->ooi_ent;
if (!ent) {
it->ooi_ent = (struct lu_dirent *)idxpage->lip_entries;
RETURN(0);
}
if (le16_to_cpu(ent->lde_reclen) != 0 &&
it->ooi_pos_ent < idxpage->lip_nr) {
ent = (struct lu_dirent *)((char *)ent +
le16_to_cpu(ent->lde_reclen));
it->ooi_ent = ent;
RETURN(0);
}
it->ooi_ent = NULL;
}
rc = osp_it_next_page(env, di);
if (rc == 0)
GOTO(again, rc);
RETURN(rc);
}
/**
* osp_it_key() - Implementation of dt_index_operations::dio_it.key
* @env: execution environment
* @di: iterator of this iteration
*
* Get the key at current iterator poisiton. These iteration methods
* (dio_it) will only be used for iterating the remote directory, so
* the key is the name of the directory entry.
*
* Return:
* * %name of the current entry
*/
static struct dt_key *osp_it_key(const struct lu_env *env,
const struct dt_it *di)
{
struct osp_it *it = (struct osp_it *)di;
struct lu_dirent *ent = (struct lu_dirent *)it->ooi_ent;
return (struct dt_key *)ent->lde_name;
}
/**
* osp_it_key_size() - Implementation of dt_index_operations::dio_it.key_size
* @env: execution environment
* @di: iterator of this iteration
*
* Get the key size at current iterator poisiton. These iteration methods
* (dio_it) will only be used for iterating the remote directory, so the key
* size is the name size of the directory entry.
*
* Return:
* * %size of name of the current entry
*/
static int osp_it_key_size(const struct lu_env *env, const struct dt_it *di)
{
struct osp_it *it = (struct osp_it *)di;
struct lu_dirent *ent = (struct lu_dirent *)it->ooi_ent;
return (int)le16_to_cpu(ent->lde_namelen);
}
/**
* osp_md_index_it_rec() - Implementation of dt_index_operations::dio_it.rec
* @env: execution environment
* @di: iterator of this iteration
* @rec: the record to be returned [out]
* @attr: attributes of the index object, so it knows how to pack the entry.
*
* Get the record at current iterator position. These iteration methods
* (dio_it) will only be used for iterating the remote directory, so it
* uses lu_dirent_calc_size() to calculate the record size.
*
* Returns 0 always for now
*/
static int osp_md_index_it_rec(const struct lu_env *env, const struct dt_it *di,
struct dt_rec *rec, __u32 attr)
{
struct osp_it *it = (struct osp_it *)di;
struct lu_dirent *ent = (struct lu_dirent *)it->ooi_ent;
size_t reclen;
reclen = lu_dirent_calc_size(le16_to_cpu(ent->lde_namelen), attr);
memcpy(rec, ent, reclen);
return 0;
}
/**
* osp_it_load() - Implementation of dt_index_operations::dio_it.load
* @env: pointer to the thread context
* @di: pointer to the iteration structure
* @hash: the specified position
*
* Locate the iteration cursor to the specified position (cookie).
*
* Return:
* * %positive number for locating to the exactly position or the next
* * %0 for arriving at the end of the iteration
* * %negative error number on failure
*/
static int osp_it_load(const struct lu_env *env, const struct dt_it *di,
__u64 hash)
{
struct osp_it *it = (struct osp_it *)di;
int rc;
it->ooi_next = hash;
rc = osp_md_index_it_next(env, (struct dt_it *)di);
if (rc == 1)
return 0;
if (rc == 0)
return 1;
return rc;
}
const struct dt_index_operations osp_md_index_ops = {
.dio_lookup = osp_md_index_lookup,
.dio_declare_insert = osp_md_declare_index_insert,
.dio_insert = osp_md_index_insert,
.dio_declare_delete = osp_md_declare_index_delete,
.dio_delete = osp_md_index_delete,
.dio_it = {
.init = osp_it_init,
.fini = osp_it_fini,
.get = osp_it_get,
.put = osp_it_put,
.next = osp_md_index_it_next,
.key = osp_it_key,
.key_size = osp_it_key_size,
.rec = osp_md_index_it_rec,
.store = osp_it_store,
.load = osp_it_load,
}
};
/**
* osp_md_xattr_list() - Implement OSP layer do_xattr_list() interface.
* @env: pointer to the thread context
* @dt: pointer to the OSP layer dt_object
* @buf: pointer to the lu_buf to hold the extended attribute [out]
*
* Implement OSP layer dt_object_operations::do_xattr_list() interface.
* List extended attribute from the specified MDT/OST object, result is not
* cached because this is called by directory migration only.
*
* Return:
* * %positive bytes used/required in the buffer
* * %negative error number on failure
*/
static int osp_md_xattr_list(const struct lu_env *env, struct dt_object *dt,
const struct lu_buf *buf)
{
struct osp_device *osp = lu2osp_dev(dt->do_lu.lo_dev);
struct osp_object *obj = dt2osp_obj(dt);
struct dt_device *dev = &osp->opd_dt_dev;
struct lu_buf *rbuf = &osp_env_info(env)->osi_lb2;
struct osp_update_request *update = NULL;
struct ptlrpc_request *req = NULL;
struct object_update_reply *reply;
const char *dname = dt->do_lu.lo_dev->ld_obd->obd_name;
int rc = 0;
ENTRY;
LASSERT(buf);
if (unlikely(obj->opo_non_exist))
RETURN(-ENOENT);
update = osp_update_request_create(dev);
if (IS_ERR(update))
RETURN(PTR_ERR(update));
rc = OSP_UPDATE_RPC_PACK(env, out_xattr_list_pack, update,
lu_object_fid(&dt->do_lu), buf->lb_len);
if (rc) {
CERROR("%s: Insert update error "DFID": rc = %d\n",
dname, PFID(lu_object_fid(&dt->do_lu)), rc);
GOTO(out, rc);
}
rc = osp_remote_sync(env, osp, update, &req);
if (rc < 0) {
if (rc == -ENOENT) {
dt->do_lu.lo_header->loh_attr &= ~LOHA_EXISTS;
obj->opo_non_exist = 1;
}
GOTO(out, rc);
}
reply = req_capsule_server_sized_get(&req->rq_pill,
&RMF_OUT_UPDATE_REPLY,
OUT_UPDATE_REPLY_SIZE);
if (reply->ourp_magic != UPDATE_REPLY_MAGIC) {
DEBUG_REQ(D_ERROR, req,
"%s: Wrong version %x expected %x "DFID": rc = %d",
dname, reply->ourp_magic, UPDATE_REPLY_MAGIC,
PFID(lu_object_fid(&dt->do_lu)), -EPROTO);
GOTO(out, rc = -EPROTO);
}
rc = object_update_result_data_get(reply, rbuf, 0);
if (rc < 0)
GOTO(out, rc);
if (!buf->lb_buf)
GOTO(out, rc);
if (unlikely(buf->lb_len < rbuf->lb_len))
GOTO(out, rc = -ERANGE);
memcpy(buf->lb_buf, rbuf->lb_buf, rbuf->lb_len);
EXIT;
out:
if (req)
ptlrpc_req_put(req);
if (update && !IS_ERR(update))
osp_update_request_destroy(env, update);
return rc;
}
/**
* osp_md_index_try() - Implementation of dt_object_operations::do_index_try
* @env: execution environment
* @dt: index object to be initialized
* @feat: the index feature of the object
*
* Try to initialize the index API pointer for the given object. This
* is the entry point of the index API, i.e. we must call this method
* to initialize the index object before calling other index methods.
*
* Return:
* * %0 if the initialization succeeds.
* * %negative errno if the initialization fails.
*/
static int osp_md_index_try(const struct lu_env *env,
struct dt_object *dt,
const struct dt_index_features *feat)
{
dt->do_index_ops = &osp_md_index_ops;
return 0;
}
/**
* osp_md_object_lock() - Implementation of dt_object_operations::do_object_lock
* @env: execution environment
* @dt: object to be locked
* @lh: lock handle [out]
* @einfo: enqueue information
* @policy: lock policy
*
* Enqueue a lock (by ldlm_cli_enqueue()) of remote object on the remote MDT,
* which will lock the object in the global namespace. And because the
* cross-MDT locks are relatively rare compared with normal local MDT operation,
* let's release it right away, instead of putting it into the LRU list.
*
* Return:
* * %ELDLM_OK if locking the object succeeds.
* * %negative errno if locking fails.
*/
static int osp_md_object_lock(const struct lu_env *env,
struct dt_object *dt,
struct lustre_handle *lh,
struct ldlm_enqueue_info *einfo,
union ldlm_policy_data *policy)
{
struct ldlm_res_id *res_id;
struct osp_device *osp = dt2osp_dev(lu2dt_dev(dt->do_lu.lo_dev));
struct ptlrpc_request *req;
int rc = 0;
__u64 flags = LDLM_FL_NO_LRU;
ENTRY;
res_id = einfo->ei_res_id;
LASSERT(res_id != NULL);
req = ldlm_enqueue_pack(osp->opd_exp, 0);
if (IS_ERR(req))
RETURN(PTR_ERR(req));
osp_set_req_replay(osp, req);
rc = ldlm_cli_enqueue(osp->opd_exp, &req, einfo, res_id,
(const union ldlm_policy_data *)policy, &flags,
NULL, 0, LVB_T_NONE, lh, 0);
ptlrpc_req_put(req);
RETURN(rc == ELDLM_OK ? 0 : -EIO);
}
/**
* osp_md_object_unlock() - Implementation of do_object_unlock
* @env: execution environment
* @dt: object to be unlocked
* @einfo: lock enqueue information
* @policy: lock policy
*
* Implementation of dt_object_operations::do_object_unlock
* Cancel a lock of a remote object.
*
* Return:
* * %Only return 0 for now.
*/
static int osp_md_object_unlock(const struct lu_env *env,
struct dt_object *dt,
struct ldlm_enqueue_info *einfo,
union ldlm_policy_data *policy)
{
struct lustre_handle *lockh = einfo->ei_cbdata;
/* unlock finally */
ldlm_lock_decref(lockh, einfo->ei_mode);
return 0;
}
/**
* osp_md_declare_destroy() - Implement OSP layer do_declare_destroy() interface
* @env: pointer to the thread context
* @dt: pointer to the OSP layer dt_object to be destroyed
* @th: pointer to the transaction handler
*
* Implement OSP layer dt_object_operations::do_declare_destroy() interface.
* Create the dt_update_request to track the update for this OSP
* in the transaction.
*
* Return:
* * %0 for success
* * %negative error number on failure
*/
static int osp_md_declare_destroy(const struct lu_env *env,
struct dt_object *dt,
struct thandle *th)
{
return osp_trans_update_request_create(th);
}
static int osp_destroy_interpreter(const struct lu_env *env,
struct object_update_reply *reply,
struct ptlrpc_request *req,
struct osp_object *obj,
void *data, int index, int rc)
{
return 0;
}
/**
* osp_md_destroy() - Implement OSP layer do_destroy() interface.
* @env: pointer to the thread context
* @dt: pointer to the OSP layer dt_object to be destroyed
* @th: pointer to the transaction handler
*
* Implement OSP layer dt_object_operations::do_destroy() interface.
* Pack the destroy update into the RPC buffer, which will be sent
* to the remote MDT during transaction stop.
*
* It also marks the object as non-cached.
*
* Return:
* * %0 for success
* * %negative error number on failure
*/
static int osp_md_destroy(const struct lu_env *env, struct dt_object *dt,
struct thandle *th)
{
struct osp_object *o = dt2osp_obj(dt);
struct osp_device *osp = lu2osp_dev(dt->do_lu.lo_dev);
struct osp_update_request *update;
struct osp_thandle *oth = thandle_to_osp_thandle(th);
int rc = 0;
ENTRY;
o->opo_non_exist = 1;
o->opo_destroyed = 1;
LASSERT(osp->opd_connect_mdt);
update = thandle_to_osp_update_request(th);
LASSERT(update != NULL);
rc = OSP_UPDATE_RPC_PACK(env, out_destroy_pack, update,
lu_object_fid(&dt->do_lu));
if (rc != 0)
RETURN(rc);
/*
* the object can be stale (due to lost LDLM lock), but
* we still want to destroy it
*/
osp_check_and_set_rpc_version(oth, o);
/* retain the object and it's status until it's destroyed on remote */
rc = osp_insert_update_callback(env, update, o, NULL,
osp_destroy_interpreter);
if (rc != 0)
RETURN(rc);
set_bit(LU_OBJECT_HEARD_BANSHEE, &dt->do_lu.lo_header->loh_flags);
rc = osp_insert_update_callback(env, update, dt2osp_obj(dt), NULL,
NULL);
RETURN(rc);
}
const struct dt_object_operations osp_md_obj_ops = {
.do_read_lock = osp_md_read_lock,
.do_write_lock = osp_md_write_lock,
.do_read_unlock = osp_md_read_unlock,
.do_write_unlock = osp_md_write_unlock,
.do_write_locked = osp_md_write_locked,
.do_declare_create = osp_md_declare_create,
.do_create = osp_md_create,
.do_declare_ref_add = osp_md_declare_ref_add,
.do_ref_add = osp_md_ref_add,
.do_declare_ref_del = osp_md_declare_ref_del,
.do_ref_del = osp_md_ref_del,
.do_declare_destroy = osp_md_declare_destroy,
.do_destroy = osp_md_destroy,
.do_ah_init = osp_md_ah_init,
.do_attr_get = osp_attr_get,
.do_declare_attr_set = osp_md_declare_attr_set,
.do_attr_set = osp_md_attr_set,
.do_xattr_get = osp_xattr_get,
.do_xattr_list = osp_md_xattr_list,
.do_declare_xattr_set = osp_declare_xattr_set,
.do_xattr_set = osp_xattr_set,
.do_declare_xattr_del = osp_declare_xattr_del,
.do_xattr_del = osp_xattr_del,
.do_index_try = osp_md_index_try,
.do_object_lock = osp_md_object_lock,
.do_object_unlock = osp_md_object_unlock,
.do_invalidate = osp_invalidate,
.do_check_stale = osp_check_stale,
};
/**
* osp_md_declare_write() - Implementation of dbo_declare_write
* @env: execution environment
* @dt: object to be written
* @buf: buffer to write which includes an embedded size field
* @pos: offet in the object to start writing at
* @th: transaction handle
*
* Implementation of dt_body_operations::dbo_declare_write
* Create the osp_update_request to track the update for this OSP
* in the transaction.
*
* Return:
* * %0 if preparation succeeds.
* * %negative errno if preparation fails.
*/
static ssize_t osp_md_declare_write(const struct lu_env *env,
struct dt_object *dt,
const struct lu_buf *buf,
loff_t pos, struct thandle *th)
{
struct osp_device *osp = dt2osp_dev(th->th_dev);
int rc;
if (dt2osp_obj(dt)->opo_destroyed)
return -ENOENT;
rc = osp_trans_update_request_create(th);
if (rc != 0)
return rc;
if (osp->opd_update == NULL)
return 0;
if (dt2osp_obj(dt)->opo_stale)
return -ESTALE;
return 0;
}
static int osp_write_interpreter(const struct lu_env *env,
struct object_update_reply *reply,
struct ptlrpc_request *req,
struct osp_object *obj,
void *data, int index, int rc)
{
struct osp_device *osp = lu2osp_dev(obj->opo_obj.do_lu.lo_dev);
if (rc) {
CDEBUG(D_HA, "error "DFID": rc = %d\n",
PFID(lu_object_fid(&obj->opo_obj.do_lu)), rc);
CFS_RACE(OBD_FAIL_OUT_OBJECT_MISS);
spin_lock(&obj->opo_lock);
obj->opo_attr.la_valid = 0;
obj->opo_stale = 1;
spin_unlock(&obj->opo_lock);
}
if (atomic_dec_and_test(&obj->opo_writes_in_flight))
wake_up_all(&osp->opd_out_waitq);
return 0;
}
/**
* osp_md_write() - Implementation of dt_body_operations::dbo_write
* @env: execution environment
* @dt: object to be written
* @buf: buffer to write which includes an embedded size field
* @pos: offet in the object to start writing at
* @th: transaction handle
*
* Pack the write object update into the RPC buffer, which will be sent
* to the remote MDT during transaction stop.
*
* Return:
* * %size of buffer in bytes if packing succeeds.
* * %negative errno if packing fails.
*/
static ssize_t osp_md_write(const struct lu_env *env, struct dt_object *dt,
const struct lu_buf *buf, loff_t *pos,
struct thandle *th)
{
struct osp_object *obj = dt2osp_obj(dt);
struct osp_update_request *update;
struct osp_thandle *oth = thandle_to_osp_thandle(th);
ssize_t rc;
ENTRY;
if (obj->opo_destroyed)
RETURN(-ENOENT);
update = thandle_to_osp_update_request(th);
LASSERT(update != NULL);
CDEBUG(D_INFO, "write "DFID" offset = %llu length = %zu\n",
PFID(lu_object_fid(&dt->do_lu)), *pos, buf->lb_len);
rc = OSP_UPDATE_RPC_PACK(env, out_write_pack, update,
lu_object_fid(&dt->do_lu), buf, *pos);
if (rc < 0)
RETURN(rc);
rc = osp_check_and_set_rpc_version(oth, obj);
if (rc < 0)
RETURN(rc);
/* to be able to invalidate object's state in case of an error */
rc = osp_insert_update_callback(env, update, obj, NULL,
osp_write_interpreter);
if (rc < 0)
RETURN(rc);
/* XXX: how about the write error happened later? */
*pos += buf->lb_len;
if (obj->opo_attr.la_valid & LA_SIZE && obj->opo_attr.la_size < *pos)
obj->opo_attr.la_size = *pos;
spin_lock(&obj->opo_lock);
if (list_empty(&obj->opo_invalidate_cb_list)) {
lu_object_get(&obj->opo_obj.do_lu);
list_add_tail(&obj->opo_invalidate_cb_list,
&update->our_invalidate_cb_list);
}
spin_unlock(&obj->opo_lock);
atomic_inc(&obj->opo_writes_in_flight);
RETURN(buf->lb_len);
}
static inline void orr_le_to_cpu(struct out_read_reply *orr_dst,
const struct out_read_reply *orr_src)
{
orr_dst->orr_size = le32_to_cpu(orr_src->orr_size);
orr_dst->orr_padding = le32_to_cpu(orr_src->orr_padding);
orr_dst->orr_offset = le64_to_cpu(orr_dst->orr_offset);
}
static int osp_md_check_creating(struct osp_object *obj)
{
int rc;
spin_lock(&obj->opo_lock);
rc = obj->opo_creating;
spin_unlock(&obj->opo_lock);
return rc;
}
static ssize_t osp_md_read(const struct lu_env *env, struct dt_object *dt,
struct lu_buf *rbuf, loff_t *pos)
{
struct osp_device *osp = lu2osp_dev(dt->do_lu.lo_dev);
struct dt_device *dt_dev = &osp->opd_dt_dev;
struct lu_buf *lbuf = &osp_env_info(env)->osi_lb2;
char *ptr = rbuf->lb_buf;
struct osp_update_request *update;
struct ptlrpc_request *req = NULL;
struct out_read_reply *orr;
struct ptlrpc_bulk_desc *desc;
struct object_update_reply *reply;
int pages;
int rc;
ENTRY;
if (dt2osp_obj(dt)->opo_destroyed)
RETURN(-ENOENT);
wait_event_idle(osp->opd_out_waitq,
!atomic_read(&dt2osp_obj(dt)->opo_writes_in_flight) &&
osp_md_check_creating(dt2osp_obj(dt)) == 0);
/* Because it needs send the update buffer right away,
* just create an update buffer, instead of attaching the
* update_remote list of the thandle. */
update = osp_update_request_create(dt_dev);
if (IS_ERR(update))
RETURN(PTR_ERR(update));
rc = OSP_UPDATE_RPC_PACK(env, out_read_pack, update,
lu_object_fid(&dt->do_lu),
rbuf->lb_len, *pos);
if (rc != 0) {
CERROR("%s: cannot insert update: rc = %d\n",
dt_dev->dd_lu_dev.ld_obd->obd_name, rc);
GOTO(out_update, rc);
}
CDEBUG(D_INFO, "%s "DFID" read offset %llu size %zu\n",
dt_dev->dd_lu_dev.ld_obd->obd_name,
PFID(lu_object_fid(&dt->do_lu)), *pos, rbuf->lb_len);
rc = osp_prep_update_req(env, osp->opd_obd->u.cli.cl_import, update,
&req);
if (rc != 0)
GOTO(out_update, rc);
/* First *and* last might be partial pages, hence +1 */
pages = DIV_ROUND_UP(rbuf->lb_len, PAGE_SIZE) + 1;
/* allocate bulk descriptor */
desc = ptlrpc_prep_bulk_imp(req, pages, 1,
PTLRPC_BULK_PUT_SINK,
MDS_BULK_PORTAL,
&ptlrpc_bulk_kiov_nopin_ops);
if (desc == NULL)
GOTO(out, rc = -ENOMEM);
desc->bd_frag_ops->add_iov_frag(desc, ptr, rbuf->lb_len);
osp_set_req_replay(osp, req);
req->rq_bulk_read = 1;
/* send request to master and wait for RPC to complete */
rc = ptlrpc_queue_wait(req);
if (rc != 0)
GOTO(out, rc);
rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk,
req->rq_bulk->bd_nob_transferred);
if (rc < 0)
GOTO(out, rc);
reply = req_capsule_server_sized_get(&req->rq_pill,
&RMF_OUT_UPDATE_REPLY,
OUT_UPDATE_REPLY_SIZE);
if (reply->ourp_magic != UPDATE_REPLY_MAGIC) {
CERROR("%s: invalid update reply magic %x expected %x: rc = %d\n",
dt_dev->dd_lu_dev.ld_obd->obd_name,
reply->ourp_magic, UPDATE_REPLY_MAGIC, -EPROTO);
GOTO(out, rc = -EPROTO);
}
rc = object_update_result_data_get(reply, lbuf, 0);
if (rc < 0)
GOTO(out, rc);
if (lbuf->lb_len < sizeof(*orr))
GOTO(out, rc = -EPROTO);
orr = lbuf->lb_buf;
orr_le_to_cpu(orr, orr);
rc = orr->orr_size;
*pos = orr->orr_offset;
out:
ptlrpc_req_put(req);
out_update:
osp_update_request_destroy(env, update);
RETURN(rc);
}
/* These body operation will be used to write symlinks during migration etc */
const struct dt_body_operations osp_md_body_ops = {
.dbo_declare_write = osp_md_declare_write,
.dbo_write = osp_md_write,
.dbo_read = osp_md_read,
};
