2 * Copyright (c) 2006 Oracle. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
33 #include <linux/kernel.h>
34 #include <linux/slab.h>
35 #include <linux/rculist.h>
36 #include <linux/llist.h>
38 #include "rds_single_path.h"
41 struct workqueue_struct *rds_ib_mr_wq;
43 static DEFINE_PER_CPU(unsigned long, clean_list_grace);
44 #define CLEAN_LIST_BUSY_BIT 0
46 static struct rds_ib_device *rds_ib_get_device(__be32 ipaddr)
48 struct rds_ib_device *rds_ibdev;
49 struct rds_ib_ipaddr *i_ipaddr;
52 list_for_each_entry_rcu(rds_ibdev, &rds_ib_devices, list) {
53 list_for_each_entry_rcu(i_ipaddr, &rds_ibdev->ipaddr_list, list) {
54 if (i_ipaddr->ipaddr == ipaddr) {
55 refcount_inc(&rds_ibdev->refcount);
66 static int rds_ib_add_ipaddr(struct rds_ib_device *rds_ibdev, __be32 ipaddr)
68 struct rds_ib_ipaddr *i_ipaddr;
70 i_ipaddr = kmalloc(sizeof *i_ipaddr, GFP_KERNEL);
74 i_ipaddr->ipaddr = ipaddr;
76 spin_lock_irq(&rds_ibdev->spinlock);
77 list_add_tail_rcu(&i_ipaddr->list, &rds_ibdev->ipaddr_list);
78 spin_unlock_irq(&rds_ibdev->spinlock);
83 static void rds_ib_remove_ipaddr(struct rds_ib_device *rds_ibdev, __be32 ipaddr)
85 struct rds_ib_ipaddr *i_ipaddr;
86 struct rds_ib_ipaddr *to_free = NULL;
89 spin_lock_irq(&rds_ibdev->spinlock);
90 list_for_each_entry_rcu(i_ipaddr, &rds_ibdev->ipaddr_list, list) {
91 if (i_ipaddr->ipaddr == ipaddr) {
92 list_del_rcu(&i_ipaddr->list);
97 spin_unlock_irq(&rds_ibdev->spinlock);
100 kfree_rcu(to_free, rcu);
103 int rds_ib_update_ipaddr(struct rds_ib_device *rds_ibdev, __be32 ipaddr)
105 struct rds_ib_device *rds_ibdev_old;
107 rds_ibdev_old = rds_ib_get_device(ipaddr);
109 return rds_ib_add_ipaddr(rds_ibdev, ipaddr);
111 if (rds_ibdev_old != rds_ibdev) {
112 rds_ib_remove_ipaddr(rds_ibdev_old, ipaddr);
113 rds_ib_dev_put(rds_ibdev_old);
114 return rds_ib_add_ipaddr(rds_ibdev, ipaddr);
116 rds_ib_dev_put(rds_ibdev_old);
121 void rds_ib_add_conn(struct rds_ib_device *rds_ibdev, struct rds_connection *conn)
123 struct rds_ib_connection *ic = conn->c_transport_data;
125 /* conn was previously on the nodev_conns_list */
126 spin_lock_irq(&ib_nodev_conns_lock);
127 BUG_ON(list_empty(&ib_nodev_conns));
128 BUG_ON(list_empty(&ic->ib_node));
129 list_del(&ic->ib_node);
131 spin_lock(&rds_ibdev->spinlock);
132 list_add_tail(&ic->ib_node, &rds_ibdev->conn_list);
133 spin_unlock(&rds_ibdev->spinlock);
134 spin_unlock_irq(&ib_nodev_conns_lock);
136 ic->rds_ibdev = rds_ibdev;
137 refcount_inc(&rds_ibdev->refcount);
140 void rds_ib_remove_conn(struct rds_ib_device *rds_ibdev, struct rds_connection *conn)
142 struct rds_ib_connection *ic = conn->c_transport_data;
144 /* place conn on nodev_conns_list */
145 spin_lock(&ib_nodev_conns_lock);
147 spin_lock_irq(&rds_ibdev->spinlock);
148 BUG_ON(list_empty(&ic->ib_node));
149 list_del(&ic->ib_node);
150 spin_unlock_irq(&rds_ibdev->spinlock);
152 list_add_tail(&ic->ib_node, &ib_nodev_conns);
154 spin_unlock(&ib_nodev_conns_lock);
156 ic->rds_ibdev = NULL;
157 rds_ib_dev_put(rds_ibdev);
160 void rds_ib_destroy_nodev_conns(void)
162 struct rds_ib_connection *ic, *_ic;
165 /* avoid calling conn_destroy with irqs off */
166 spin_lock_irq(&ib_nodev_conns_lock);
167 list_splice(&ib_nodev_conns, &tmp_list);
168 spin_unlock_irq(&ib_nodev_conns_lock);
170 list_for_each_entry_safe(ic, _ic, &tmp_list, ib_node)
171 rds_conn_destroy(ic->conn);
174 void rds_ib_get_mr_info(struct rds_ib_device *rds_ibdev, struct rds_info_rdma_connection *iinfo)
176 struct rds_ib_mr_pool *pool_1m = rds_ibdev->mr_1m_pool;
178 iinfo->rdma_mr_max = pool_1m->max_items;
179 iinfo->rdma_mr_size = pool_1m->fmr_attr.max_pages;
182 struct rds_ib_mr *rds_ib_reuse_mr(struct rds_ib_mr_pool *pool)
184 struct rds_ib_mr *ibmr = NULL;
185 struct llist_node *ret;
189 flag = this_cpu_ptr(&clean_list_grace);
190 set_bit(CLEAN_LIST_BUSY_BIT, flag);
191 ret = llist_del_first(&pool->clean_list);
193 ibmr = llist_entry(ret, struct rds_ib_mr, llnode);
194 if (pool->pool_type == RDS_IB_MR_8K_POOL)
195 rds_ib_stats_inc(s_ib_rdma_mr_8k_reused);
197 rds_ib_stats_inc(s_ib_rdma_mr_1m_reused);
200 clear_bit(CLEAN_LIST_BUSY_BIT, flag);
205 static inline void wait_clean_list_grace(void)
210 for_each_online_cpu(cpu) {
211 flag = &per_cpu(clean_list_grace, cpu);
212 while (test_bit(CLEAN_LIST_BUSY_BIT, flag))
217 void rds_ib_sync_mr(void *trans_private, int direction)
219 struct rds_ib_mr *ibmr = trans_private;
220 struct rds_ib_device *rds_ibdev = ibmr->device;
223 case DMA_FROM_DEVICE:
224 ib_dma_sync_sg_for_cpu(rds_ibdev->dev, ibmr->sg,
225 ibmr->sg_dma_len, DMA_BIDIRECTIONAL);
228 ib_dma_sync_sg_for_device(rds_ibdev->dev, ibmr->sg,
229 ibmr->sg_dma_len, DMA_BIDIRECTIONAL);
234 void __rds_ib_teardown_mr(struct rds_ib_mr *ibmr)
236 struct rds_ib_device *rds_ibdev = ibmr->device;
238 if (ibmr->sg_dma_len) {
239 ib_dma_unmap_sg(rds_ibdev->dev,
240 ibmr->sg, ibmr->sg_len,
242 ibmr->sg_dma_len = 0;
245 /* Release the s/g list */
249 for (i = 0; i < ibmr->sg_len; ++i) {
250 struct page *page = sg_page(&ibmr->sg[i]);
252 /* FIXME we need a way to tell a r/w MR
254 WARN_ON(!page->mapping && irqs_disabled());
255 set_page_dirty(page);
265 void rds_ib_teardown_mr(struct rds_ib_mr *ibmr)
267 unsigned int pinned = ibmr->sg_len;
269 __rds_ib_teardown_mr(ibmr);
271 struct rds_ib_mr_pool *pool = ibmr->pool;
273 atomic_sub(pinned, &pool->free_pinned);
277 static inline unsigned int rds_ib_flush_goal(struct rds_ib_mr_pool *pool, int free_all)
279 unsigned int item_count;
281 item_count = atomic_read(&pool->item_count);
289 * given an llist of mrs, put them all into the list_head for more processing
291 static unsigned int llist_append_to_list(struct llist_head *llist,
292 struct list_head *list)
294 struct rds_ib_mr *ibmr;
295 struct llist_node *node;
296 struct llist_node *next;
297 unsigned int count = 0;
299 node = llist_del_all(llist);
302 ibmr = llist_entry(node, struct rds_ib_mr, llnode);
303 list_add_tail(&ibmr->unmap_list, list);
311 * this takes a list head of mrs and turns it into linked llist nodes
312 * of clusters. Each cluster has linked llist nodes of
313 * MR_CLUSTER_SIZE mrs that are ready for reuse.
315 static void list_to_llist_nodes(struct rds_ib_mr_pool *pool,
316 struct list_head *list,
317 struct llist_node **nodes_head,
318 struct llist_node **nodes_tail)
320 struct rds_ib_mr *ibmr;
321 struct llist_node *cur = NULL;
322 struct llist_node **next = nodes_head;
324 list_for_each_entry(ibmr, list, unmap_list) {
334 * Flush our pool of MRs.
335 * At a minimum, all currently unused MRs are unmapped.
336 * If the number of MRs allocated exceeds the limit, we also try
337 * to free as many MRs as needed to get back to this limit.
339 int rds_ib_flush_mr_pool(struct rds_ib_mr_pool *pool,
340 int free_all, struct rds_ib_mr **ibmr_ret)
342 struct rds_ib_mr *ibmr;
343 struct llist_node *clean_nodes;
344 struct llist_node *clean_tail;
345 LIST_HEAD(unmap_list);
346 unsigned long unpinned = 0;
347 unsigned int nfreed = 0, dirty_to_clean = 0, free_goal;
349 if (pool->pool_type == RDS_IB_MR_8K_POOL)
350 rds_ib_stats_inc(s_ib_rdma_mr_8k_pool_flush);
352 rds_ib_stats_inc(s_ib_rdma_mr_1m_pool_flush);
356 while (!mutex_trylock(&pool->flush_lock)) {
357 ibmr = rds_ib_reuse_mr(pool);
360 finish_wait(&pool->flush_wait, &wait);
364 prepare_to_wait(&pool->flush_wait, &wait,
365 TASK_UNINTERRUPTIBLE);
366 if (llist_empty(&pool->clean_list))
369 ibmr = rds_ib_reuse_mr(pool);
372 finish_wait(&pool->flush_wait, &wait);
376 finish_wait(&pool->flush_wait, &wait);
378 mutex_lock(&pool->flush_lock);
381 ibmr = rds_ib_reuse_mr(pool);
388 /* Get the list of all MRs to be dropped. Ordering matters -
389 * we want to put drop_list ahead of free_list.
391 dirty_to_clean = llist_append_to_list(&pool->drop_list, &unmap_list);
392 dirty_to_clean += llist_append_to_list(&pool->free_list, &unmap_list);
394 llist_append_to_list(&pool->clean_list, &unmap_list);
396 free_goal = rds_ib_flush_goal(pool, free_all);
398 if (list_empty(&unmap_list))
401 if (pool->use_fastreg)
402 rds_ib_unreg_frmr(&unmap_list, &nfreed, &unpinned, free_goal);
404 rds_ib_unreg_fmr(&unmap_list, &nfreed, &unpinned, free_goal);
406 if (!list_empty(&unmap_list)) {
407 /* we have to make sure that none of the things we're about
408 * to put on the clean list would race with other cpus trying
409 * to pull items off. The llist would explode if we managed to
410 * remove something from the clean list and then add it back again
411 * while another CPU was spinning on that same item in llist_del_first.
413 * This is pretty unlikely, but just in case wait for an llist grace period
414 * here before adding anything back into the clean list.
416 wait_clean_list_grace();
418 list_to_llist_nodes(pool, &unmap_list, &clean_nodes, &clean_tail);
420 *ibmr_ret = llist_entry(clean_nodes, struct rds_ib_mr, llnode);
422 /* more than one entry in llist nodes */
423 if (clean_nodes->next)
424 llist_add_batch(clean_nodes->next, clean_tail, &pool->clean_list);
428 atomic_sub(unpinned, &pool->free_pinned);
429 atomic_sub(dirty_to_clean, &pool->dirty_count);
430 atomic_sub(nfreed, &pool->item_count);
433 mutex_unlock(&pool->flush_lock);
434 if (waitqueue_active(&pool->flush_wait))
435 wake_up(&pool->flush_wait);
440 struct rds_ib_mr *rds_ib_try_reuse_ibmr(struct rds_ib_mr_pool *pool)
442 struct rds_ib_mr *ibmr = NULL;
445 if (atomic_read(&pool->dirty_count) >= pool->max_items_soft / 10)
446 queue_delayed_work(rds_ib_mr_wq, &pool->flush_worker, 10);
449 ibmr = rds_ib_reuse_mr(pool);
453 if (atomic_inc_return(&pool->item_count) <= pool->max_items)
456 atomic_dec(&pool->item_count);
459 if (pool->pool_type == RDS_IB_MR_8K_POOL)
460 rds_ib_stats_inc(s_ib_rdma_mr_8k_pool_depleted);
462 rds_ib_stats_inc(s_ib_rdma_mr_1m_pool_depleted);
463 return ERR_PTR(-EAGAIN);
466 /* We do have some empty MRs. Flush them out. */
467 if (pool->pool_type == RDS_IB_MR_8K_POOL)
468 rds_ib_stats_inc(s_ib_rdma_mr_8k_pool_wait);
470 rds_ib_stats_inc(s_ib_rdma_mr_1m_pool_wait);
472 rds_ib_flush_mr_pool(pool, 0, &ibmr);
480 static void rds_ib_mr_pool_flush_worker(struct work_struct *work)
482 struct rds_ib_mr_pool *pool = container_of(work, struct rds_ib_mr_pool, flush_worker.work);
484 rds_ib_flush_mr_pool(pool, 0, NULL);
487 void rds_ib_free_mr(void *trans_private, int invalidate)
489 struct rds_ib_mr *ibmr = trans_private;
490 struct rds_ib_mr_pool *pool = ibmr->pool;
491 struct rds_ib_device *rds_ibdev = ibmr->device;
493 rdsdebug("RDS/IB: free_mr nents %u\n", ibmr->sg_len);
495 /* Return it to the pool's free list */
496 if (rds_ibdev->use_fastreg)
497 rds_ib_free_frmr_list(ibmr);
499 rds_ib_free_fmr_list(ibmr);
501 atomic_add(ibmr->sg_len, &pool->free_pinned);
502 atomic_inc(&pool->dirty_count);
504 /* If we've pinned too many pages, request a flush */
505 if (atomic_read(&pool->free_pinned) >= pool->max_free_pinned ||
506 atomic_read(&pool->dirty_count) >= pool->max_items / 5)
507 queue_delayed_work(rds_ib_mr_wq, &pool->flush_worker, 10);
510 if (likely(!in_interrupt())) {
511 rds_ib_flush_mr_pool(pool, 0, NULL);
513 /* We get here if the user created a MR marked
514 * as use_once and invalidate at the same time.
516 queue_delayed_work(rds_ib_mr_wq,
517 &pool->flush_worker, 10);
521 rds_ib_dev_put(rds_ibdev);
524 void rds_ib_flush_mrs(void)
526 struct rds_ib_device *rds_ibdev;
528 down_read(&rds_ib_devices_lock);
529 list_for_each_entry(rds_ibdev, &rds_ib_devices, list) {
530 if (rds_ibdev->mr_8k_pool)
531 rds_ib_flush_mr_pool(rds_ibdev->mr_8k_pool, 0, NULL);
533 if (rds_ibdev->mr_1m_pool)
534 rds_ib_flush_mr_pool(rds_ibdev->mr_1m_pool, 0, NULL);
536 up_read(&rds_ib_devices_lock);
539 void *rds_ib_get_mr(struct scatterlist *sg, unsigned long nents,
540 struct rds_sock *rs, u32 *key_ret)
542 struct rds_ib_device *rds_ibdev;
543 struct rds_ib_mr *ibmr = NULL;
544 struct rds_ib_connection *ic = rs->rs_conn->c_transport_data;
547 rds_ibdev = rds_ib_get_device(rs->rs_bound_addr);
553 if (!rds_ibdev->mr_8k_pool || !rds_ibdev->mr_1m_pool) {
558 if (rds_ibdev->use_fastreg)
559 ibmr = rds_ib_reg_frmr(rds_ibdev, ic, sg, nents, key_ret);
561 ibmr = rds_ib_reg_fmr(rds_ibdev, sg, nents, key_ret);
567 pr_warn("RDS/IB: rds_ib_get_mr failed (errno=%d)\n", ret);
570 rds_ib_dev_put(rds_ibdev);
575 void rds_ib_destroy_mr_pool(struct rds_ib_mr_pool *pool)
577 cancel_delayed_work_sync(&pool->flush_worker);
578 rds_ib_flush_mr_pool(pool, 1, NULL);
579 WARN_ON(atomic_read(&pool->item_count));
580 WARN_ON(atomic_read(&pool->free_pinned));
584 struct rds_ib_mr_pool *rds_ib_create_mr_pool(struct rds_ib_device *rds_ibdev,
587 struct rds_ib_mr_pool *pool;
589 pool = kzalloc(sizeof(*pool), GFP_KERNEL);
591 return ERR_PTR(-ENOMEM);
593 pool->pool_type = pool_type;
594 init_llist_head(&pool->free_list);
595 init_llist_head(&pool->drop_list);
596 init_llist_head(&pool->clean_list);
597 mutex_init(&pool->flush_lock);
598 init_waitqueue_head(&pool->flush_wait);
599 INIT_DELAYED_WORK(&pool->flush_worker, rds_ib_mr_pool_flush_worker);
601 if (pool_type == RDS_IB_MR_1M_POOL) {
602 /* +1 allows for unaligned MRs */
603 pool->fmr_attr.max_pages = RDS_MR_1M_MSG_SIZE + 1;
604 pool->max_items = rds_ibdev->max_1m_mrs;
606 /* pool_type == RDS_IB_MR_8K_POOL */
607 pool->fmr_attr.max_pages = RDS_MR_8K_MSG_SIZE + 1;
608 pool->max_items = rds_ibdev->max_8k_mrs;
611 pool->max_free_pinned = pool->max_items * pool->fmr_attr.max_pages / 4;
612 pool->fmr_attr.max_maps = rds_ibdev->fmr_max_remaps;
613 pool->fmr_attr.page_shift = PAGE_SHIFT;
614 pool->max_items_soft = rds_ibdev->max_mrs * 3 / 4;
615 pool->use_fastreg = rds_ibdev->use_fastreg;
620 int rds_ib_mr_init(void)
622 rds_ib_mr_wq = alloc_workqueue("rds_mr_flushd", WQ_MEM_RECLAIM, 0);
628 /* By the time this is called all the IB devices should have been torn down and
629 * had their pools freed. As each pool is freed its work struct is waited on,
630 * so the pool flushing work queue should be idle by the time we get here.
632 void rds_ib_mr_exit(void)
634 destroy_workqueue(rds_ib_mr_wq);