*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
* GPL HEADER END
*/
* Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2015, Intel Corporation.
+ * Copyright (c) 2011, 2017, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
* is used to derive a setting for ptlrpcd_per_cpt_max.
*/
static int max_ptlrpcds;
-CFS_MODULE_PARM(max_ptlrpcds, "i", int, 0644,
- "Max ptlrpcd thread count to be started.");
+module_param(max_ptlrpcds, int, 0644);
+MODULE_PARM_DESC(max_ptlrpcds,
+ "Max ptlrpcd thread count to be started (obsolete).");
/*
* ptlrpcd_bind_policy is obsolete, but retained to ensure that
* is used to derive a setting for ptlrpcd_partner_group_size.
*/
static int ptlrpcd_bind_policy;
-CFS_MODULE_PARM(ptlrpcd_bind_policy, "i", int, 0644,
- "Ptlrpcd threads binding mode (obsolete).");
+module_param(ptlrpcd_bind_policy, int, 0644);
+MODULE_PARM_DESC(ptlrpcd_bind_policy,
+ "Ptlrpcd threads binding mode (obsolete).");
/*
* ptlrpcd_per_cpt_max: The maximum number of ptlrpcd threads to run
* in a CPT.
*/
static int ptlrpcd_per_cpt_max;
-CFS_MODULE_PARM(ptlrpcd_per_cpt_max, "i", int, 0644,
- "Max ptlrpcd thread count to be started per cpt.");
+module_param(ptlrpcd_per_cpt_max, int, 0644);
+MODULE_PARM_DESC(ptlrpcd_per_cpt_max,
+ "Max ptlrpcd thread count to be started per CPT.");
/*
* ptlrpcd_partner_group_size: The desired number of threads in each
* a CPT partners of each other.
*/
static int ptlrpcd_partner_group_size;
-CFS_MODULE_PARM(ptlrpcd_partner_group_size, "i", int, 0644,
- "Number of ptlrpcd threads in a partner group.");
+module_param(ptlrpcd_partner_group_size, int, 0644);
+MODULE_PARM_DESC(ptlrpcd_partner_group_size,
+ "Number of ptlrpcd threads in a partner group.");
/*
* ptlrpcd_cpts: A CPT string describing the CPU partitions that
* run ptlrpcd threads on CPTS 0, 1, 2, 3, 5, and 7.
*/
static char *ptlrpcd_cpts;
-CFS_MODULE_PARM(ptlrpcd_cpts, "s", charp, 0644,
- "CPU partitions ptlrpcd threads should run in");
+module_param(ptlrpcd_cpts, charp, 0644);
+MODULE_PARM_DESC(ptlrpcd_cpts,
+ "CPU partitions ptlrpcd threads should run in");
/* ptlrpcds_cpt_idx maps cpt numbers to an index in the ptlrpcds array. */
static int *ptlrpcds_cpt_idx;
LASSERT(req->rq_phase == RQ_PHASE_NEW);
req->rq_set = new;
- req->rq_queued_time = cfs_time_current();
+ req->rq_queued_time = ktime_get_seconds();
}
spin_lock(&new->set_new_req_lock);
if (count == i) {
wake_up(&new->set_waitq);
- /* XXX: It maybe unnecessary to wakeup all the partners. But to
+ /*
+ * XXX: It maybe unnecessary to wakeup all the partners. But to
* guarantee the async RPC can be processed ASAP, we have
- * no other better choice. It maybe fixed in future. */
+ * no other better choice. It maybe fixed in future.
+ */
for (i = 0; i < pc->pc_npartners; i++)
wake_up(&pc->pc_partners[i]->pc_set->set_waitq);
}
* Return transferred RPCs count.
*/
static int ptlrpcd_steal_rqset(struct ptlrpc_request_set *des,
- struct ptlrpc_request_set *src)
+ struct ptlrpc_request_set *src)
{
struct list_head *tmp, *pos;
struct ptlrpc_request *req;
spin_lock(&req->rq_lock);
if (req->rq_invalid_rqset) {
- struct l_wait_info lwi = LWI_TIMEOUT(cfs_time_seconds(5),
- back_to_sleep, NULL);
-
req->rq_invalid_rqset = 0;
spin_unlock(&req->rq_lock);
- l_wait_event(req->rq_set_waitq, (req->rq_set == NULL), &lwi);
+ if (wait_event_idle_timeout(req->rq_set_waitq,
+ req->rq_set == NULL,
+ cfs_time_seconds(5)) == 0)
+ l_wait_event_abortable(req->rq_set_waitq,
+ req->rq_set == NULL);
} else if (req->rq_set) {
- /* If we have a vaid "rq_set", just reuse it to avoid double
- * linked. */
+ /*
+ * If we have a vaid "rq_set", just reuse it to avoid double
+ * linked.
+ */
LASSERT(req->rq_phase == RQ_PHASE_NEW);
LASSERT(req->rq_send_state == LUSTRE_IMP_REPLAY);
pc = ptlrpcd_select_pc(req);
- DEBUG_REQ(D_INFO, req, "add req [%p] to pc [%s:%d]",
+ DEBUG_REQ(D_INFO, req, "add req [%p] to pc [%s+%d]",
req, pc->pc_name, pc->pc_index);
ptlrpc_set_add_new_req(pc, req);
static int ptlrpcd_check(struct lu_env *env, struct ptlrpcd_ctl *pc)
{
struct list_head *tmp, *pos;
- struct ptlrpc_request *req;
- struct ptlrpc_request_set *set = pc->pc_set;
- int rc = 0;
- int rc2;
- ENTRY;
+ struct ptlrpc_request *req;
+ struct ptlrpc_request_set *set = pc->pc_set;
+ int rc = 0;
+ int rc2;
+
+ ENTRY;
if (atomic_read(&set->set_new_count)) {
spin_lock(&set->set_new_req_lock);
spin_unlock(&set->set_new_req_lock);
}
- /* We should call lu_env_refill() before handling new requests to make
+ /*
+ * We should call lu_env_refill() before handling new requests to make
* sure that env key the requests depending on really exists.
*/
rc2 = lu_env_refill(env);
if (atomic_read(&set->set_remaining))
rc |= ptlrpc_check_set(env, set);
- /* NB: ptlrpc_check_set has already moved complted request at the
- * head of seq::set_requests */
+ /*
+ * NB: ptlrpc_check_set has already moved complted request at the
+ * head of seq::set_requests
+ */
list_for_each_safe(pos, tmp, &set->set_requests) {
req = list_entry(pos, struct ptlrpc_request, rq_set_chain);
if (req->rq_phase != RQ_PHASE_COMPLETE)
*/
rc = atomic_read(&set->set_new_count);
- /* If we have nothing to do, check whether we can take some
- * work from our partner threads. */
- if (rc == 0 && pc->pc_npartners > 0) {
- struct ptlrpcd_ctl *partner;
- struct ptlrpc_request_set *ps;
- int first = pc->pc_cursor;
-
- do {
- partner = pc->pc_partners[pc->pc_cursor++];
- if (pc->pc_cursor >= pc->pc_npartners)
- pc->pc_cursor = 0;
- if (partner == NULL)
- continue;
+ /*
+ * If we have nothing to do, check whether we can take some
+ * work from our partner threads.
+ */
+ if (rc == 0 && pc->pc_npartners > 0) {
+ struct ptlrpcd_ctl *partner;
+ struct ptlrpc_request_set *ps;
+ int first = pc->pc_cursor;
+
+ do {
+ partner = pc->pc_partners[pc->pc_cursor++];
+ if (pc->pc_cursor >= pc->pc_npartners)
+ pc->pc_cursor = 0;
+ if (partner == NULL)
+ continue;
spin_lock(&partner->pc_lock);
ps = partner->pc_set;
if (atomic_read(&ps->set_new_count)) {
rc = ptlrpcd_steal_rqset(set, ps);
if (rc > 0)
- CDEBUG(D_RPCTRACE, "transfer %d"
- " async RPCs [%d->%d]\n",
+ CDEBUG(D_RPCTRACE,
+ "transfer %d async RPCs [%d->%d]\n",
rc, partner->pc_index,
pc->pc_index);
}
}
}
- RETURN(rc);
+ RETURN(rc || test_bit(LIOD_STOP, &pc->pc_flags));
}
/**
* Main ptlrpcd thread.
* ptlrpc's code paths like to execute in process context, so we have this
* thread which spins on a set which contains the rpcs and sends them.
- *
*/
static int ptlrpcd(void *arg)
{
struct lu_env env = { .le_ses = &ses };
int rc = 0;
int exit = 0;
+
ENTRY;
unshare_fs_struct();
complete(&pc->pc_starting);
- /*
- * This mainloop strongly resembles ptlrpc_set_wait() except that our
- * set never completes. ptlrpcd_check() calls ptlrpc_check_set() when
- * there are requests in the set. New requests come in on the set's
- * new_req_list and ptlrpcd_check() moves them into the set.
- */
- do {
- struct l_wait_info lwi;
- int timeout;
+ /*
+ * This mainloop strongly resembles ptlrpc_set_wait() except that our
+ * set never completes. ptlrpcd_check() calls ptlrpc_check_set() when
+ * there are requests in the set. New requests come in on the set's
+ * new_req_list and ptlrpcd_check() moves them into the set.
+ */
+ do {
+ struct l_wait_info lwi;
+ time64_t timeout;
- timeout = ptlrpc_set_next_timeout(set);
- lwi = LWI_TIMEOUT(cfs_time_seconds(timeout ? timeout : 1),
- ptlrpc_expired_set, set);
+ timeout = ptlrpc_set_next_timeout(set);
+ lwi = LWI_TIMEOUT(cfs_time_seconds(timeout),
+ ptlrpc_expired_set, set);
lu_context_enter(&env.le_ctx);
lu_context_enter(env.le_ses);
*/
if (test_bit(LIOD_STOP, &pc->pc_flags)) {
if (test_bit(LIOD_FORCE, &pc->pc_flags))
- ptlrpc_abort_set(set);
- exit++;
- }
-
- /*
- * Let's make one more loop to make sure that ptlrpcd_check()
- * copied all raced new rpcs into the set so we can kill them.
- */
- } while (exit < 2);
-
- /*
- * Wait for inflight requests to drain.
- */
+ ptlrpc_abort_set(set);
+ exit++;
+ }
+
+ /*
+ * Let's make one more loop to make sure that ptlrpcd_check()
+ * copied all raced new rpcs into the set so we can kill them.
+ */
+ } while (exit < 2);
+
+ /*
+ * Wait for inflight requests to drain.
+ */
if (!list_empty(&set->set_requests))
- ptlrpc_set_wait(set);
+ ptlrpc_set_wait(&env, set);
lu_context_fini(&env.le_ctx);
lu_context_fini(env.le_ses);
int first;
int i;
int rc = 0;
+
ENTRY;
LASSERT(index >= 0 && index < pd->pd_nthreads);
{
struct task_struct *task;
int rc = 0;
+
ENTRY;
/*
RETURN(0);
}
- /*
- * So far only "client" ptlrpcd uses an environment. In the future,
- * ptlrpcd thread (or a thread-set) has to be given an argument,
- * describing its "scope".
- */
- rc = lu_context_init(&pc->pc_env.le_ctx, LCT_CL_THREAD|LCT_REMEMBER);
- if (rc != 0)
- GOTO(out, rc);
-
task = kthread_run(ptlrpcd, pc, pc->pc_name);
if (IS_ERR(task))
GOTO(out_set, rc = PTR_ERR(task));
spin_unlock(&pc->pc_lock);
ptlrpc_set_destroy(set);
}
- lu_context_fini(&pc->pc_env.le_ctx);
-out:
clear_bit(LIOD_START, &pc->pc_flags);
RETURN(rc);
}
void ptlrpcd_free(struct ptlrpcd_ctl *pc)
{
struct ptlrpc_request_set *set = pc->pc_set;
+
ENTRY;
if (!test_bit(LIOD_START, &pc->pc_flags)) {
}
wait_for_completion(&pc->pc_finishing);
- lu_context_fini(&pc->pc_env.le_ctx);
spin_lock(&pc->pc_lock);
pc->pc_set = NULL;
clear_bit(LIOD_FORCE, &pc->pc_flags);
out:
- if (pc->pc_npartners > 0) {
- LASSERT(pc->pc_partners != NULL);
-
- OBD_FREE(pc->pc_partners,
- sizeof(struct ptlrpcd_ctl *) * pc->pc_npartners);
- pc->pc_partners = NULL;
- }
- pc->pc_npartners = 0;
+ if (pc->pc_npartners > 0) {
+ LASSERT(pc->pc_partners != NULL);
+
+ OBD_FREE(pc->pc_partners,
+ sizeof(struct ptlrpcd_ctl *) * pc->pc_npartners);
+ pc->pc_partners = NULL;
+ }
+ pc->pc_npartners = 0;
pc->pc_error = 0;
- EXIT;
+ EXIT;
}
static void ptlrpcd_fini(void)
int i;
int j;
int ncpts;
+
ENTRY;
if (ptlrpcds != NULL) {
int ncpts;
int cpt;
struct ptlrpcd *pd;
+
ENTRY;
/*
cptable = cfs_cpt_table;
ncpts = cfs_cpt_number(cptable);
if (ptlrpcd_cpts != NULL) {
- struct cfs_expr_list *el;
+ struct cfs_expr_list *el;
size = ncpts * sizeof(ptlrpcds_cpt_idx[0]);
OBD_ALLOC(ptlrpcds_cpt_idx, size);
size = offsetof(struct ptlrpcd, pd_threads[nthreads]);
OBD_CPT_ALLOC(pd, cptable, cpt, size);
+
if (!pd)
GOTO(out, rc = -ENOMEM);
pd->pd_size = size;
int ptlrpcd_addref(void)
{
- int rc = 0;
- ENTRY;
+ int rc = 0;
+
+ ENTRY;
mutex_lock(&ptlrpcd_mutex);
- if (++ptlrpcd_users == 1) {
+ if (++ptlrpcd_users == 1) {
rc = ptlrpcd_init();
if (rc < 0)
ptlrpcd_users--;
}
mutex_unlock(&ptlrpcd_mutex);
- RETURN(rc);
+ RETURN(rc);
}
EXPORT_SYMBOL(ptlrpcd_addref);
void ptlrpcd_decref(void)
{
mutex_lock(&ptlrpcd_mutex);
- if (--ptlrpcd_users == 0)
- ptlrpcd_fini();
+ if (--ptlrpcd_users == 0)
+ ptlrpcd_fini();
mutex_unlock(&ptlrpcd_mutex);
}
EXPORT_SYMBOL(ptlrpcd_decref);