* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2013, Intel Corporation.
+ * Copyright (c) 2011, 2014, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
{
}
+static void osc_read_ahead_release(const struct lu_env *env,
+ void *cbdata)
+{
+ struct ldlm_lock *dlmlock = cbdata;
+ struct lustre_handle lockh;
+
+ ldlm_lock2handle(dlmlock, &lockh);
+ ldlm_lock_decref(&lockh, LCK_PR);
+ LDLM_LOCK_PUT(dlmlock);
+}
+
+static int osc_io_read_ahead(const struct lu_env *env,
+ const struct cl_io_slice *ios,
+ pgoff_t start, struct cl_read_ahead *ra)
+{
+ struct osc_object *osc = cl2osc(ios->cis_obj);
+ struct ldlm_lock *dlmlock;
+ int result = -ENODATA;
+ ENTRY;
+
+ dlmlock = osc_dlmlock_at_pgoff(env, osc, start, 0);
+ if (dlmlock != NULL) {
+ if (dlmlock->l_req_mode != LCK_PR) {
+ struct lustre_handle lockh;
+ ldlm_lock2handle(dlmlock, &lockh);
+ ldlm_lock_addref(&lockh, LCK_PR);
+ ldlm_lock_decref(&lockh, dlmlock->l_req_mode);
+ }
+
+ ra->cra_end = cl_index(osc2cl(osc),
+ dlmlock->l_policy_data.l_extent.end);
+ ra->cra_release = osc_read_ahead_release;
+ ra->cra_cbdata = dlmlock;
+ result = 0;
+ }
+
+ RETURN(result);
+}
+
/**
* An implementation of cl_io_operations::cio_io_submit() method for osc
* layer. Iterates over pages in the in-queue, prepares each for io by calling
const struct cl_io_slice *ios,
enum cl_req_type crt, struct cl_2queue *queue)
{
- struct cl_page *page;
- struct cl_page *tmp;
- struct client_obd *cli = NULL;
- struct osc_object *osc = NULL; /* to keep gcc happy */
- struct osc_page *opg;
- struct cl_io *io;
- CFS_LIST_HEAD (list);
+ struct cl_page *page;
+ struct cl_page *tmp;
+ struct client_obd *cli = NULL;
+ struct osc_object *osc = NULL; /* to keep gcc happy */
+ struct osc_page *opg;
+ struct cl_io *io;
+ struct list_head list = LIST_HEAD_INIT(list);
struct cl_page_list *qin = &queue->c2_qin;
struct cl_page_list *qout = &queue->c2_qout;
- int queued = 0;
+ unsigned int queued = 0;
int result = 0;
int cmd;
int brw_flags;
- int max_pages;
+ unsigned int max_pages;
LASSERT(qin->pl_nr > 0);
oap = &opg->ops_oap;
LASSERT(osc == oap->oap_obj);
- if (!cfs_list_empty(&oap->oap_pending_item) ||
- !cfs_list_empty(&oap->oap_rpc_item)) {
+ if (!list_empty(&oap->oap_pending_item) ||
+ !list_empty(&oap->oap_rpc_item)) {
CDEBUG(D_CACHE, "Busy oap %p page %p for submit.\n",
oap, opg);
result = -EBUSY;
continue;
}
- cl_page_list_move(qout, qin, page);
spin_lock(&oap->oap_lock);
oap->oap_async_flags = ASYNC_URGENT|ASYNC_READY;
oap->oap_async_flags |= ASYNC_COUNT_STABLE;
spin_unlock(&oap->oap_lock);
osc_page_submit(env, opg, crt, brw_flags);
- cfs_list_add_tail(&oap->oap_pending_item, &list);
+ list_add_tail(&oap->oap_pending_item, &list);
+
+ if (page->cp_sync_io != NULL)
+ cl_page_list_move(qout, qin, page);
+ else /* async IO */
+ cl_page_list_del(env, qin, page);
+
if (++queued == max_pages) {
queued = 0;
result = osc_queue_sync_pages(env, osc, &list, cmd,
* Expand stripe KMS if necessary.
*/
static void osc_page_touch_at(const struct lu_env *env,
- struct cl_object *obj, pgoff_t idx, unsigned to)
+ struct cl_object *obj, pgoff_t idx, size_t to)
{
struct lov_oinfo *loi = cl2osc(obj)->oo_oinfo;
struct cl_attr *attr = &osc_env_info(env)->oti_attr;
attr->cat_mtime = attr->cat_ctime = LTIME_S(CFS_CURRENT_TIME);
valid = CAT_MTIME | CAT_CTIME;
- if (kms > loi->loi_kms) {
- attr->cat_kms = kms;
- valid |= CAT_KMS;
- }
- if (kms > loi->loi_lvb.lvb_size) {
- attr->cat_size = kms;
- valid |= CAT_SIZE;
- }
- cl_object_attr_set(env, obj, attr, valid);
- cl_object_attr_unlock(obj);
+ if (kms > loi->loi_kms) {
+ attr->cat_kms = kms;
+ valid |= CAT_KMS;
+ }
+ if (kms > loi->loi_lvb.lvb_size) {
+ attr->cat_size = kms;
+ valid |= CAT_SIZE;
+ }
+ cl_object_attr_update(env, obj, attr, valid);
+ cl_object_attr_unlock(obj);
}
static int osc_io_commit_async(const struct lu_env *env,
opg = osc_cl_page_osc(page, osc);
oap = &opg->ops_oap;
- if (!cfs_list_empty(&oap->oap_rpc_item)) {
+ if (!list_empty(&oap->oap_rpc_item)) {
CDEBUG(D_CACHE, "Busy oap %p page %p for submit.\n",
oap, opg);
result = -EBUSY;
}
/* The page may be already in dirty cache. */
- if (cfs_list_empty(&oap->oap_pending_item)) {
+ if (list_empty(&oap->oap_pending_item)) {
result = osc_page_cache_add(env, &opg->ops_cl, io);
if (result != 0)
break;
struct osc_object *osc = cl2osc(ios->cis_obj);
struct client_obd *cli = osc_cli(osc);
unsigned long c;
- unsigned int npages;
- unsigned int max_pages;
+ unsigned long npages;
+ unsigned long max_pages;
ENTRY;
if (cl_io_is_append(io))
RETURN(0);
npages = io->u.ci_rw.crw_count >> PAGE_CACHE_SHIFT;
- if (io->u.ci_rw.crw_pos & ~CFS_PAGE_MASK)
+ if (io->u.ci_rw.crw_pos & ~PAGE_MASK)
++npages;
max_pages = cli->cl_max_pages_per_rpc * cli->cl_max_rpcs_in_flight;
if (npages > max_pages)
npages = max_pages;
- c = atomic_read(cli->cl_lru_left);
+ c = atomic_long_read(cli->cl_lru_left);
if (c < npages && osc_lru_reclaim(cli) > 0)
- c = atomic_read(cli->cl_lru_left);
+ c = atomic_long_read(cli->cl_lru_left);
while (c >= npages) {
- if (c == atomic_cmpxchg(cli->cl_lru_left, c, c - npages)) {
+ if (c == atomic_long_cmpxchg(cli->cl_lru_left, c, c - npages)) {
oio->oi_lru_reserved = npages;
break;
}
- c = atomic_read(cli->cl_lru_left);
+ c = atomic_long_read(cli->cl_lru_left);
}
RETURN(0);
struct client_obd *cli = osc_cli(osc);
if (oio->oi_lru_reserved > 0) {
- atomic_add(oio->oi_lru_reserved, cli->cl_lru_left);
+ atomic_long_add(oio->oi_lru_reserved, cli->cl_lru_left);
oio->oi_lru_reserved = 0;
}
+ oio->oi_write_osclock = NULL;
}
static int osc_io_fault_start(const struct lu_env *env,
- const struct cl_io_slice *ios)
+ const struct cl_io_slice *ios)
{
- struct cl_io *io;
- struct cl_fault_io *fio;
-
- ENTRY;
+ struct cl_io *io;
+ struct cl_fault_io *fio;
+ ENTRY;
- io = ios->cis_io;
- fio = &io->u.ci_fault;
- CDEBUG(D_INFO, "%lu %d %d\n",
- fio->ft_index, fio->ft_writable, fio->ft_nob);
- /*
- * If mapping is writeable, adjust kms to cover this page,
- * but do not extend kms beyond actual file size.
- * See bug 10919.
- */
- if (fio->ft_writable)
- osc_page_touch_at(env, ios->cis_obj,
- fio->ft_index, fio->ft_nob);
- RETURN(0);
+ io = ios->cis_io;
+ fio = &io->u.ci_fault;
+ CDEBUG(D_INFO, "%lu %d %zu\n",
+ fio->ft_index, fio->ft_writable, fio->ft_nob);
+ /*
+ * If mapping is writeable, adjust kms to cover this page,
+ * but do not extend kms beyond actual file size.
+ * See bug 10919.
+ */
+ if (fio->ft_writable)
+ osc_page_touch_at(env, ios->cis_obj,
+ fio->ft_index, fio->ft_nob);
+ RETURN(0);
}
static int osc_async_upcall(void *a, int rc)
return 0;
}
-#if defined(__KERNEL__)
/**
* Checks that there are no pages being written in the extent being truncated.
*/
oap = &ops->ops_oap;
if (oap->oap_cmd & OBD_BRW_WRITE &&
- !cfs_list_empty(&oap->oap_pending_item))
+ !list_empty(&oap->oap_pending_item))
CL_PAGE_DEBUG(D_ERROR, env, page, "exists " LPU64 "/%s.\n",
start, current->comm);
-#ifdef __linux__
if (PageLocked(page->cp_vmpage))
CDEBUG(D_CACHE, "page %p index %lu locked for %d.\n",
ops, osc_index(ops), oap->oap_cmd & OBD_BRW_RWMASK);
-#endif
return CLP_GANG_OKAY;
}
start + partial, CL_PAGE_EOF,
trunc_check_cb, (void *)&size);
}
-#else /* __KERNEL__ */
-static void osc_trunc_check(const struct lu_env *env, struct cl_io *io,
- struct osc_io *oio, __u64 size)
-{
- return;
-}
-#endif
static int osc_io_setattr_start(const struct lu_env *env,
const struct cl_io_slice *slice)
}
if (ia_valid & ATTR_CTIME_SET) {
attr->cat_ctime = lvb->lvb_ctime;
- cl_valid |= CAT_CTIME;
- }
- result = cl_object_attr_set(env, obj, attr, cl_valid);
- }
- cl_object_attr_unlock(obj);
- }
- memset(oa, 0, sizeof(*oa));
- if (result == 0) {
+ cl_valid |= CAT_CTIME;
+ }
+ result = cl_object_attr_update(env, obj, attr,
+ cl_valid);
+ }
+ cl_object_attr_unlock(obj);
+ }
+ memset(oa, 0, sizeof(*oa));
+ if (result == 0) {
oa->o_oi = loi->loi_oi;
- oa->o_mtime = attr->cat_mtime;
- oa->o_atime = attr->cat_atime;
- oa->o_ctime = attr->cat_ctime;
- oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP | OBD_MD_FLATIME |
- OBD_MD_FLCTIME | OBD_MD_FLMTIME;
+ obdo_set_parent_fid(oa, io->u.ci_setattr.sa_parent_fid);
+ oa->o_stripe_idx = io->u.ci_setattr.sa_stripe_index;
+ oa->o_valid |= OBD_MD_FLID | OBD_MD_FLGROUP;
+ if (ia_valid & ATTR_CTIME) {
+ oa->o_valid |= OBD_MD_FLCTIME;
+ oa->o_ctime = attr->cat_ctime;
+ }
+ if (ia_valid & ATTR_ATIME) {
+ oa->o_valid |= OBD_MD_FLATIME;
+ oa->o_atime = attr->cat_atime;
+ }
+ if (ia_valid & ATTR_MTIME) {
+ oa->o_valid |= OBD_MD_FLMTIME;
+ oa->o_mtime = attr->cat_mtime;
+ }
if (ia_valid & ATTR_SIZE) {
oa->o_size = size;
oa->o_blocks = OBD_OBJECT_EOF;
LASSERT(oio->oi_lockless == 0);
}
+ if (ia_valid & ATTR_ATTR_FLAG) {
+ oa->o_flags = io->u.ci_setattr.sa_attr_flags;
+ oa->o_valid |= OBD_MD_FLFLAGS;
+ }
+
oinfo.oi_oa = oa;
oinfo.oi_capa = io->u.ci_setattr.sa_capa;
init_completion(&cbargs->opc_sync);
&oinfo, osc_async_upcall,
cbargs, PTLRPCD_SET);
else
- result = osc_setattr_async_base(osc_export(cl2osc(obj)),
- &oinfo, NULL,
- osc_async_upcall,
- cbargs, PTLRPCD_SET);
+ result = osc_setattr_async(osc_export(cl2osc(obj)),
+ &oinfo,
+ osc_async_upcall,
+ cbargs, PTLRPCD_SET);
cbargs->opc_rpc_sent = result == 0;
}
return result;
}
}
+struct osc_data_version_args {
+ struct osc_io *dva_oio;
+};
+
+static int
+osc_data_version_interpret(const struct lu_env *env, struct ptlrpc_request *req,
+ void *arg, int rc)
+{
+ struct osc_data_version_args *dva = arg;
+ struct osc_io *oio = dva->dva_oio;
+ const struct ost_body *body;
+
+ ENTRY;
+ if (rc < 0)
+ GOTO(out, rc);
+
+ body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
+ if (body == NULL)
+ GOTO(out, rc = -EPROTO);
+
+ lustre_get_wire_obdo(&req->rq_import->imp_connect_data, &oio->oi_oa,
+ &body->oa);
+ EXIT;
+out:
+ oio->oi_cbarg.opc_rc = rc;
+ complete(&oio->oi_cbarg.opc_sync);
+
+ return 0;
+}
+
+static int osc_io_data_version_start(const struct lu_env *env,
+ const struct cl_io_slice *slice)
+{
+ struct cl_data_version_io *dv = &slice->cis_io->u.ci_data_version;
+ struct osc_io *oio = cl2osc_io(env, slice);
+ struct obdo *oa = &oio->oi_oa;
+ struct osc_async_cbargs *cbargs = &oio->oi_cbarg;
+ struct osc_object *obj = cl2osc(slice->cis_obj);
+ struct lov_oinfo *loi = obj->oo_oinfo;
+ struct obd_export *exp = osc_export(obj);
+ struct ptlrpc_request *req;
+ struct ost_body *body;
+ struct osc_data_version_args *dva;
+ int rc;
+
+ ENTRY;
+ memset(oa, 0, sizeof(*oa));
+ oa->o_oi = loi->loi_oi;
+ oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP;
+
+ if (dv->dv_flags & (LL_DV_RD_FLUSH | LL_DV_WR_FLUSH)) {
+ oa->o_valid |= OBD_MD_FLFLAGS;
+ oa->o_flags |= OBD_FL_SRVLOCK;
+ if (dv->dv_flags & LL_DV_WR_FLUSH)
+ oa->o_flags |= OBD_FL_FLUSH;
+ }
+
+ init_completion(&cbargs->opc_sync);
+
+ req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
+ if (req == NULL)
+ RETURN(-ENOMEM);
+
+ rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
+ if (rc < 0) {
+ ptlrpc_request_free(req);
+ RETURN(rc);
+ }
+
+ body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
+ lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
+
+ ptlrpc_request_set_replen(req);
+ req->rq_interpret_reply = osc_data_version_interpret;
+ CLASSERT(sizeof(*dva) <= sizeof(req->rq_async_args));
+ dva = ptlrpc_req_async_args(req);
+ dva->dva_oio = oio;
+
+ ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
+
+ RETURN(0);
+}
+
+static void osc_io_data_version_end(const struct lu_env *env,
+ const struct cl_io_slice *slice)
+{
+ struct cl_data_version_io *dv = &slice->cis_io->u.ci_data_version;
+ struct osc_io *oio = cl2osc_io(env, slice);
+ struct osc_async_cbargs *cbargs = &oio->oi_cbarg;
+
+ ENTRY;
+ wait_for_completion(&cbargs->opc_sync);
+
+ if (cbargs->opc_rc != 0) {
+ slice->cis_io->ci_result = cbargs->opc_rc;
+ } else if (!(oio->oi_oa.o_valid & OBD_MD_FLDATAVERSION)) {
+ slice->cis_io->ci_result = -EOPNOTSUPP;
+ } else {
+ dv->dv_data_version = oio->oi_oa.o_data_version;
+ slice->cis_io->ci_result = 0;
+ }
+
+ EXIT;
+}
+
static int osc_io_read_start(const struct lu_env *env,
const struct cl_io_slice *slice)
{
- struct osc_io *oio = cl2osc_io(env, slice);
struct cl_object *obj = slice->cis_obj;
struct cl_attr *attr = &osc_env_info(env)->oti_attr;
int rc = 0;
ENTRY;
- if (oio->oi_lockless == 0 && !slice->cis_io->ci_noatime) {
+ if (!slice->cis_io->ci_noatime) {
cl_object_attr_lock(obj);
attr->cat_atime = LTIME_S(CFS_CURRENT_TIME);
- rc = cl_object_attr_set(env, obj, attr, CAT_ATIME);
+ rc = cl_object_attr_update(env, obj, attr, CAT_ATIME);
cl_object_attr_unlock(obj);
}
static int osc_io_write_start(const struct lu_env *env,
const struct cl_io_slice *slice)
{
- struct osc_io *oio = cl2osc_io(env, slice);
- struct cl_object *obj = slice->cis_obj;
- struct cl_attr *attr = &osc_env_info(env)->oti_attr;
- int result = 0;
- ENTRY;
-
- if (oio->oi_lockless == 0) {
- OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_DELAY_SETTIME, 1);
- cl_object_attr_lock(obj);
- result = cl_object_attr_get(env, obj, attr);
- if (result == 0) {
- attr->cat_mtime = attr->cat_ctime =
- LTIME_S(CFS_CURRENT_TIME);
- result = cl_object_attr_set(env, obj, attr,
- CAT_MTIME | CAT_CTIME);
- }
- cl_object_attr_unlock(obj);
- }
- RETURN(result);
+ struct cl_object *obj = slice->cis_obj;
+ struct cl_attr *attr = &osc_env_info(env)->oti_attr;
+ int rc = 0;
+ ENTRY;
+
+ OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_DELAY_SETTIME, 1);
+ cl_object_attr_lock(obj);
+ attr->cat_mtime = attr->cat_ctime = LTIME_S(CFS_CURRENT_TIME);
+ rc = cl_object_attr_update(env, obj, attr, CAT_MTIME | CAT_CTIME);
+ cl_object_attr_unlock(obj);
+
+ RETURN(rc);
}
static int osc_fsync_ost(const struct lu_env *env, struct osc_object *obj,
.cio_start = osc_io_setattr_start,
.cio_end = osc_io_setattr_end
},
+ [CIT_DATA_VERSION] = {
+ .cio_start = osc_io_data_version_start,
+ .cio_end = osc_io_data_version_end,
+ },
[CIT_FAULT] = {
.cio_start = osc_io_fault_start,
.cio_end = osc_io_end,
.cio_fini = osc_io_fini
}
},
+ .cio_read_ahead = osc_io_read_ahead,
.cio_submit = osc_io_submit,
.cio_commit_async = osc_io_commit_async
};
static void osc_req_attr_set(const struct lu_env *env,
const struct cl_req_slice *slice,
const struct cl_object *obj,
- struct cl_req_attr *attr, obd_valid flags)
+ struct cl_req_attr *attr, u64 flags)
{
struct lov_oinfo *oinfo;
struct cl_req *clerq;
struct cl_page *apage; /* _some_ page in @clerq */
- struct cl_lock *lock; /* _some_ lock protecting @apage */
- struct osc_lock *olck;
+ struct ldlm_lock *lock; /* _some_ lock protecting @apage */
struct osc_page *opg;
struct obdo *oa;
struct ost_lvb *lvb;
oa->o_valid |= OBD_MD_FLID;
}
if (flags & OBD_MD_FLHANDLE) {
- struct cl_object *subobj;
-
clerq = slice->crs_req;
- LASSERT(!cfs_list_empty(&clerq->crq_pages));
+ LASSERT(!list_empty(&clerq->crq_pages));
apage = container_of(clerq->crq_pages.next,
struct cl_page, cp_flight);
opg = osc_cl_page_osc(apage, NULL);
- subobj = opg->ops_cl.cpl_obj;
- lock = cl_lock_at_pgoff(env, subobj, osc_index(opg),
- NULL, 1, 1);
- if (lock == NULL) {
- struct cl_object_header *head;
- struct cl_lock *scan;
-
- head = cl_object_header(subobj);
- cfs_list_for_each_entry(scan, &head->coh_locks,
- cll_linkage)
- CL_LOCK_DEBUG(D_ERROR, env, scan,
- "no cover page!\n");
- CL_PAGE_DEBUG(D_ERROR, env, apage,
- "dump uncover page!\n");
- libcfs_debug_dumpstack(NULL);
- LBUG();
- }
+ lock = osc_dlmlock_at_pgoff(env, cl2osc(obj), osc_index(opg),
+ OSC_DAP_FL_TEST_LOCK | OSC_DAP_FL_CANCELING);
+ if (lock == NULL && !opg->ops_srvlock) {
+ struct ldlm_resource *res;
+ struct ldlm_res_id *resname;
+
+ CL_PAGE_DEBUG(D_ERROR, env, apage, "uncovered page!\n");
+
+ resname = &osc_env_info(env)->oti_resname;
+ ostid_build_res_name(&oinfo->loi_oi, resname);
+ res = ldlm_resource_get(
+ osc_export(cl2osc(obj))->exp_obd->obd_namespace,
+ NULL, resname, LDLM_EXTENT, 0);
+ ldlm_resource_dump(D_ERROR, res);
+
+ libcfs_debug_dumpstack(NULL);
+ LBUG();
+ }
- olck = osc_lock_at(lock);
- LASSERT(olck != NULL);
- LASSERT(ergo(opg->ops_srvlock, olck->ols_lock == NULL));
- /* check for lockless io. */
- if (olck->ols_lock != NULL) {
- oa->o_handle = olck->ols_lock->l_remote_handle;
- oa->o_valid |= OBD_MD_FLHANDLE;
- }
- cl_lock_put(env, lock);
- }
+ /* check for lockless io. */
+ if (lock != NULL) {
+ oa->o_handle = lock->l_remote_handle;
+ oa->o_valid |= OBD_MD_FLHANDLE;
+ LDLM_LOCK_PUT(lock);
+ }
+ }
}
static const struct cl_req_operations osc_req_ops = {
}
int osc_req_init(const struct lu_env *env, struct cl_device *dev,
- struct cl_req *req)
+ struct cl_req *req)
{
- struct osc_req *or;
- int result;
-
- OBD_SLAB_ALLOC_PTR_GFP(or, osc_req_kmem, __GFP_IO);
- if (or != NULL) {
- cl_req_slice_add(req, &or->or_cl, dev, &osc_req_ops);
- result = 0;
- } else
- result = -ENOMEM;
- return result;
+ struct osc_req *or;
+ int result;
+
+ OBD_SLAB_ALLOC_PTR_GFP(or, osc_req_kmem, GFP_NOFS);
+ if (or != NULL) {
+ cl_req_slice_add(req, &or->or_cl, dev, &osc_req_ops);
+ result = 0;
+ } else
+ result = -ENOMEM;
+ return result;
}
/** @} osc */