blob: e8fdb172adbb23940b8d2d698cd72fa216cd6ba1 [file] [log] [blame]
Andy Grover08b48a12009-02-24 15:30:32 +00001/*
2 * Copyright (c) 2006 Oracle. All rights reserved.
3 *
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:
9 *
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
12 * conditions are met:
13 *
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer.
17 *
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.
22 *
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
30 * SOFTWARE.
31 *
32 */
33#include <linux/kernel.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090034#include <linux/slab.h>
Chris Mason764f2dd2010-04-22 21:59:15 -040035#include <linux/rculist.h>
Huang Ying1bc144b2011-08-30 15:21:30 +000036#include <linux/llist.h>
Andy Grover08b48a12009-02-24 15:30:32 +000037
38#include "rds.h"
Andy Grover08b48a12009-02-24 15:30:32 +000039#include "ib.h"
40
Chris Mason6fa70da2010-06-11 11:17:59 -070041static DEFINE_PER_CPU(unsigned long, clean_list_grace);
42#define CLEAN_LIST_BUSY_BIT 0
Andy Grover08b48a12009-02-24 15:30:32 +000043
44/*
45 * This is stored as mr->r_trans_private.
46 */
47struct rds_ib_mr {
48 struct rds_ib_device *device;
49 struct rds_ib_mr_pool *pool;
50 struct ib_fmr *fmr;
Chris Mason6fa70da2010-06-11 11:17:59 -070051
Huang Ying1bc144b2011-08-30 15:21:30 +000052 struct llist_node llnode;
Chris Mason6fa70da2010-06-11 11:17:59 -070053
54 /* unmap_list is for freeing */
55 struct list_head unmap_list;
Andy Grover08b48a12009-02-24 15:30:32 +000056 unsigned int remap_count;
57
58 struct scatterlist *sg;
59 unsigned int sg_len;
60 u64 *dma;
61 int sg_dma_len;
62};
63
64/*
65 * Our own little FMR pool
66 */
67struct rds_ib_mr_pool {
68 struct mutex flush_lock; /* serialize fmr invalidate */
Chris Mason7a0ff5d2010-06-11 11:26:02 -070069 struct delayed_work flush_worker; /* flush worker */
Andy Grover08b48a12009-02-24 15:30:32 +000070
Andy Grover08b48a12009-02-24 15:30:32 +000071 atomic_t item_count; /* total # of MRs */
72 atomic_t dirty_count; /* # dirty of MRs */
Chris Mason6fa70da2010-06-11 11:17:59 -070073
Huang Ying1bc144b2011-08-30 15:21:30 +000074 struct llist_head drop_list; /* MRs that have reached their max_maps limit */
75 struct llist_head free_list; /* unused MRs */
76 struct llist_head clean_list; /* global unused & unamapped MRs */
Chris Mason6fa70da2010-06-11 11:17:59 -070077 wait_queue_head_t flush_wait;
78
Andy Grover08b48a12009-02-24 15:30:32 +000079 atomic_t free_pinned; /* memory pinned by free MRs */
80 unsigned long max_items;
81 unsigned long max_items_soft;
82 unsigned long max_free_pinned;
83 struct ib_fmr_attr fmr_attr;
84};
85
Chris Mason6fa70da2010-06-11 11:17:59 -070086static int rds_ib_flush_mr_pool(struct rds_ib_mr_pool *pool, int free_all, struct rds_ib_mr **);
Andy Grover08b48a12009-02-24 15:30:32 +000087static void rds_ib_teardown_mr(struct rds_ib_mr *ibmr);
88static void rds_ib_mr_pool_flush_worker(struct work_struct *work);
89
90static struct rds_ib_device *rds_ib_get_device(__be32 ipaddr)
91{
92 struct rds_ib_device *rds_ibdev;
93 struct rds_ib_ipaddr *i_ipaddr;
94
Zach Brownea819862010-07-15 12:34:33 -070095 rcu_read_lock();
96 list_for_each_entry_rcu(rds_ibdev, &rds_ib_devices, list) {
Chris Mason764f2dd2010-04-22 21:59:15 -040097 list_for_each_entry_rcu(i_ipaddr, &rds_ibdev->ipaddr_list, list) {
Andy Grover08b48a12009-02-24 15:30:32 +000098 if (i_ipaddr->ipaddr == ipaddr) {
Zach Brown3e0249f2010-05-18 15:48:51 -070099 atomic_inc(&rds_ibdev->refcount);
Chris Mason764f2dd2010-04-22 21:59:15 -0400100 rcu_read_unlock();
Andy Grover08b48a12009-02-24 15:30:32 +0000101 return rds_ibdev;
102 }
103 }
Andy Grover08b48a12009-02-24 15:30:32 +0000104 }
Zach Brownea819862010-07-15 12:34:33 -0700105 rcu_read_unlock();
Andy Grover08b48a12009-02-24 15:30:32 +0000106
107 return NULL;
108}
109
110static int rds_ib_add_ipaddr(struct rds_ib_device *rds_ibdev, __be32 ipaddr)
111{
112 struct rds_ib_ipaddr *i_ipaddr;
113
114 i_ipaddr = kmalloc(sizeof *i_ipaddr, GFP_KERNEL);
115 if (!i_ipaddr)
116 return -ENOMEM;
117
118 i_ipaddr->ipaddr = ipaddr;
119
120 spin_lock_irq(&rds_ibdev->spinlock);
Chris Mason764f2dd2010-04-22 21:59:15 -0400121 list_add_tail_rcu(&i_ipaddr->list, &rds_ibdev->ipaddr_list);
Andy Grover08b48a12009-02-24 15:30:32 +0000122 spin_unlock_irq(&rds_ibdev->spinlock);
123
124 return 0;
125}
126
127static void rds_ib_remove_ipaddr(struct rds_ib_device *rds_ibdev, __be32 ipaddr)
128{
Andy Grover4a818022010-04-23 11:04:21 -0700129 struct rds_ib_ipaddr *i_ipaddr;
Chris Mason764f2dd2010-04-22 21:59:15 -0400130 struct rds_ib_ipaddr *to_free = NULL;
131
Andy Grover08b48a12009-02-24 15:30:32 +0000132
133 spin_lock_irq(&rds_ibdev->spinlock);
Chris Mason764f2dd2010-04-22 21:59:15 -0400134 list_for_each_entry_rcu(i_ipaddr, &rds_ibdev->ipaddr_list, list) {
Andy Grover08b48a12009-02-24 15:30:32 +0000135 if (i_ipaddr->ipaddr == ipaddr) {
Chris Mason764f2dd2010-04-22 21:59:15 -0400136 list_del_rcu(&i_ipaddr->list);
137 to_free = i_ipaddr;
Andy Grover08b48a12009-02-24 15:30:32 +0000138 break;
139 }
140 }
141 spin_unlock_irq(&rds_ibdev->spinlock);
Chris Mason764f2dd2010-04-22 21:59:15 -0400142
143 if (to_free) {
144 synchronize_rcu();
145 kfree(to_free);
146 }
Andy Grover08b48a12009-02-24 15:30:32 +0000147}
148
149int rds_ib_update_ipaddr(struct rds_ib_device *rds_ibdev, __be32 ipaddr)
150{
151 struct rds_ib_device *rds_ibdev_old;
152
153 rds_ibdev_old = rds_ib_get_device(ipaddr);
Zach Brown3e0249f2010-05-18 15:48:51 -0700154 if (rds_ibdev_old) {
Andy Grover08b48a12009-02-24 15:30:32 +0000155 rds_ib_remove_ipaddr(rds_ibdev_old, ipaddr);
Zach Brown3e0249f2010-05-18 15:48:51 -0700156 rds_ib_dev_put(rds_ibdev_old);
157 }
Andy Grover08b48a12009-02-24 15:30:32 +0000158
159 return rds_ib_add_ipaddr(rds_ibdev, ipaddr);
160}
161
Andy Grover745cbcc2009-04-01 08:20:19 +0000162void rds_ib_add_conn(struct rds_ib_device *rds_ibdev, struct rds_connection *conn)
Andy Grover08b48a12009-02-24 15:30:32 +0000163{
164 struct rds_ib_connection *ic = conn->c_transport_data;
165
166 /* conn was previously on the nodev_conns_list */
167 spin_lock_irq(&ib_nodev_conns_lock);
168 BUG_ON(list_empty(&ib_nodev_conns));
169 BUG_ON(list_empty(&ic->ib_node));
170 list_del(&ic->ib_node);
Andy Grover08b48a12009-02-24 15:30:32 +0000171
Dan Carpenteraef3ea32010-09-18 13:44:14 +0000172 spin_lock(&rds_ibdev->spinlock);
Andy Grover08b48a12009-02-24 15:30:32 +0000173 list_add_tail(&ic->ib_node, &rds_ibdev->conn_list);
Dan Carpenteraef3ea32010-09-18 13:44:14 +0000174 spin_unlock(&rds_ibdev->spinlock);
Andy Grover08b48a12009-02-24 15:30:32 +0000175 spin_unlock_irq(&ib_nodev_conns_lock);
176
Andy Grover745cbcc2009-04-01 08:20:19 +0000177 ic->rds_ibdev = rds_ibdev;
Zach Brown3e0249f2010-05-18 15:48:51 -0700178 atomic_inc(&rds_ibdev->refcount);
Andy Grover08b48a12009-02-24 15:30:32 +0000179}
180
Andy Grover745cbcc2009-04-01 08:20:19 +0000181void rds_ib_remove_conn(struct rds_ib_device *rds_ibdev, struct rds_connection *conn)
182{
183 struct rds_ib_connection *ic = conn->c_transport_data;
184
185 /* place conn on nodev_conns_list */
186 spin_lock(&ib_nodev_conns_lock);
187
188 spin_lock_irq(&rds_ibdev->spinlock);
189 BUG_ON(list_empty(&ic->ib_node));
190 list_del(&ic->ib_node);
191 spin_unlock_irq(&rds_ibdev->spinlock);
192
193 list_add_tail(&ic->ib_node, &ib_nodev_conns);
194
195 spin_unlock(&ib_nodev_conns_lock);
196
197 ic->rds_ibdev = NULL;
Zach Brown3e0249f2010-05-18 15:48:51 -0700198 rds_ib_dev_put(rds_ibdev);
Andy Grover745cbcc2009-04-01 08:20:19 +0000199}
200
Zach Brown8aeb1ba2010-06-25 14:58:16 -0700201void rds_ib_destroy_nodev_conns(void)
Andy Grover08b48a12009-02-24 15:30:32 +0000202{
203 struct rds_ib_connection *ic, *_ic;
204 LIST_HEAD(tmp_list);
205
206 /* avoid calling conn_destroy with irqs off */
Zach Brown8aeb1ba2010-06-25 14:58:16 -0700207 spin_lock_irq(&ib_nodev_conns_lock);
208 list_splice(&ib_nodev_conns, &tmp_list);
209 spin_unlock_irq(&ib_nodev_conns_lock);
Andy Grover08b48a12009-02-24 15:30:32 +0000210
Andy Grover433d3082009-10-30 08:51:55 +0000211 list_for_each_entry_safe(ic, _ic, &tmp_list, ib_node)
Andy Grover08b48a12009-02-24 15:30:32 +0000212 rds_conn_destroy(ic->conn);
Andy Grover08b48a12009-02-24 15:30:32 +0000213}
214
215struct rds_ib_mr_pool *rds_ib_create_mr_pool(struct rds_ib_device *rds_ibdev)
216{
217 struct rds_ib_mr_pool *pool;
218
219 pool = kzalloc(sizeof(*pool), GFP_KERNEL);
220 if (!pool)
221 return ERR_PTR(-ENOMEM);
222
Huang Ying1bc144b2011-08-30 15:21:30 +0000223 init_llist_head(&pool->free_list);
224 init_llist_head(&pool->drop_list);
225 init_llist_head(&pool->clean_list);
Andy Grover08b48a12009-02-24 15:30:32 +0000226 mutex_init(&pool->flush_lock);
Chris Mason6fa70da2010-06-11 11:17:59 -0700227 init_waitqueue_head(&pool->flush_wait);
Chris Mason7a0ff5d2010-06-11 11:26:02 -0700228 INIT_DELAYED_WORK(&pool->flush_worker, rds_ib_mr_pool_flush_worker);
Andy Grover08b48a12009-02-24 15:30:32 +0000229
230 pool->fmr_attr.max_pages = fmr_message_size;
231 pool->fmr_attr.max_maps = rds_ibdev->fmr_max_remaps;
Andy Grovera870d622009-07-17 13:13:33 +0000232 pool->fmr_attr.page_shift = PAGE_SHIFT;
Andy Grover08b48a12009-02-24 15:30:32 +0000233 pool->max_free_pinned = rds_ibdev->max_fmrs * fmr_message_size / 4;
234
235 /* We never allow more than max_items MRs to be allocated.
236 * When we exceed more than max_items_soft, we start freeing
237 * items more aggressively.
238 * Make sure that max_items > max_items_soft > max_items / 2
239 */
240 pool->max_items_soft = rds_ibdev->max_fmrs * 3 / 4;
241 pool->max_items = rds_ibdev->max_fmrs;
242
243 return pool;
244}
245
246void rds_ib_get_mr_info(struct rds_ib_device *rds_ibdev, struct rds_info_rdma_connection *iinfo)
247{
248 struct rds_ib_mr_pool *pool = rds_ibdev->mr_pool;
249
250 iinfo->rdma_mr_max = pool->max_items;
251 iinfo->rdma_mr_size = pool->fmr_attr.max_pages;
252}
253
254void rds_ib_destroy_mr_pool(struct rds_ib_mr_pool *pool)
255{
Chris Mason7a0ff5d2010-06-11 11:26:02 -0700256 cancel_delayed_work_sync(&pool->flush_worker);
Chris Mason6fa70da2010-06-11 11:17:59 -0700257 rds_ib_flush_mr_pool(pool, 1, NULL);
Andy Grover571c02f2010-03-11 13:50:01 +0000258 WARN_ON(atomic_read(&pool->item_count));
259 WARN_ON(atomic_read(&pool->free_pinned));
Andy Grover08b48a12009-02-24 15:30:32 +0000260 kfree(pool);
261}
262
263static inline struct rds_ib_mr *rds_ib_reuse_fmr(struct rds_ib_mr_pool *pool)
264{
265 struct rds_ib_mr *ibmr = NULL;
Huang Ying1bc144b2011-08-30 15:21:30 +0000266 struct llist_node *ret;
Chris Mason6fa70da2010-06-11 11:17:59 -0700267 unsigned long *flag;
Andy Grover08b48a12009-02-24 15:30:32 +0000268
Chris Mason6fa70da2010-06-11 11:17:59 -0700269 preempt_disable();
270 flag = &__get_cpu_var(clean_list_grace);
271 set_bit(CLEAN_LIST_BUSY_BIT, flag);
Huang Ying1bc144b2011-08-30 15:21:30 +0000272 ret = llist_del_first(&pool->clean_list);
Chris Mason6fa70da2010-06-11 11:17:59 -0700273 if (ret)
Huang Ying1bc144b2011-08-30 15:21:30 +0000274 ibmr = llist_entry(ret, struct rds_ib_mr, llnode);
Andy Grover08b48a12009-02-24 15:30:32 +0000275
Chris Mason6fa70da2010-06-11 11:17:59 -0700276 clear_bit(CLEAN_LIST_BUSY_BIT, flag);
277 preempt_enable();
Andy Grover08b48a12009-02-24 15:30:32 +0000278 return ibmr;
279}
280
Chris Mason6fa70da2010-06-11 11:17:59 -0700281static inline void wait_clean_list_grace(void)
282{
283 int cpu;
284 unsigned long *flag;
285
286 for_each_online_cpu(cpu) {
287 flag = &per_cpu(clean_list_grace, cpu);
288 while (test_bit(CLEAN_LIST_BUSY_BIT, flag))
289 cpu_relax();
290 }
291}
292
Andy Grover08b48a12009-02-24 15:30:32 +0000293static struct rds_ib_mr *rds_ib_alloc_fmr(struct rds_ib_device *rds_ibdev)
294{
295 struct rds_ib_mr_pool *pool = rds_ibdev->mr_pool;
296 struct rds_ib_mr *ibmr = NULL;
297 int err = 0, iter = 0;
298
Chris Mason8576f372010-07-19 17:06:46 -0700299 if (atomic_read(&pool->dirty_count) >= pool->max_items / 10)
Tejun Heoc534a102011-02-01 11:42:43 +0100300 schedule_delayed_work(&pool->flush_worker, 10);
Chris Mason8576f372010-07-19 17:06:46 -0700301
Andy Grover08b48a12009-02-24 15:30:32 +0000302 while (1) {
303 ibmr = rds_ib_reuse_fmr(pool);
304 if (ibmr)
305 return ibmr;
306
307 /* No clean MRs - now we have the choice of either
308 * allocating a fresh MR up to the limit imposed by the
309 * driver, or flush any dirty unused MRs.
310 * We try to avoid stalling in the send path if possible,
311 * so we allocate as long as we're allowed to.
312 *
313 * We're fussy with enforcing the FMR limit, though. If the driver
314 * tells us we can't use more than N fmrs, we shouldn't start
315 * arguing with it */
316 if (atomic_inc_return(&pool->item_count) <= pool->max_items)
317 break;
318
319 atomic_dec(&pool->item_count);
320
321 if (++iter > 2) {
322 rds_ib_stats_inc(s_ib_rdma_mr_pool_depleted);
323 return ERR_PTR(-EAGAIN);
324 }
325
326 /* We do have some empty MRs. Flush them out. */
327 rds_ib_stats_inc(s_ib_rdma_mr_pool_wait);
Chris Mason6fa70da2010-06-11 11:17:59 -0700328 rds_ib_flush_mr_pool(pool, 0, &ibmr);
329 if (ibmr)
330 return ibmr;
Andy Grover08b48a12009-02-24 15:30:32 +0000331 }
332
Andy Grovere4c52c92010-04-23 10:49:53 -0700333 ibmr = kzalloc_node(sizeof(*ibmr), GFP_KERNEL, rdsibdev_to_node(rds_ibdev));
Andy Grover08b48a12009-02-24 15:30:32 +0000334 if (!ibmr) {
335 err = -ENOMEM;
336 goto out_no_cigar;
337 }
338
Chris Mason38a4e5e2010-05-11 15:09:45 -0700339 memset(ibmr, 0, sizeof(*ibmr));
340
Andy Grover08b48a12009-02-24 15:30:32 +0000341 ibmr->fmr = ib_alloc_fmr(rds_ibdev->pd,
342 (IB_ACCESS_LOCAL_WRITE |
343 IB_ACCESS_REMOTE_READ |
Andy Grover15133f62010-01-12 14:33:38 -0800344 IB_ACCESS_REMOTE_WRITE|
345 IB_ACCESS_REMOTE_ATOMIC),
Andy Grover08b48a12009-02-24 15:30:32 +0000346 &pool->fmr_attr);
347 if (IS_ERR(ibmr->fmr)) {
348 err = PTR_ERR(ibmr->fmr);
349 ibmr->fmr = NULL;
350 printk(KERN_WARNING "RDS/IB: ib_alloc_fmr failed (err=%d)\n", err);
351 goto out_no_cigar;
352 }
353
354 rds_ib_stats_inc(s_ib_rdma_mr_alloc);
355 return ibmr;
356
357out_no_cigar:
358 if (ibmr) {
359 if (ibmr->fmr)
360 ib_dealloc_fmr(ibmr->fmr);
361 kfree(ibmr);
362 }
363 atomic_dec(&pool->item_count);
364 return ERR_PTR(err);
365}
366
367static int rds_ib_map_fmr(struct rds_ib_device *rds_ibdev, struct rds_ib_mr *ibmr,
368 struct scatterlist *sg, unsigned int nents)
369{
370 struct ib_device *dev = rds_ibdev->dev;
371 struct scatterlist *scat = sg;
372 u64 io_addr = 0;
373 u64 *dma_pages;
374 u32 len;
375 int page_cnt, sg_dma_len;
376 int i, j;
377 int ret;
378
379 sg_dma_len = ib_dma_map_sg(dev, sg, nents,
380 DMA_BIDIRECTIONAL);
381 if (unlikely(!sg_dma_len)) {
382 printk(KERN_WARNING "RDS/IB: dma_map_sg failed!\n");
383 return -EBUSY;
384 }
385
386 len = 0;
387 page_cnt = 0;
388
389 for (i = 0; i < sg_dma_len; ++i) {
390 unsigned int dma_len = ib_sg_dma_len(dev, &scat[i]);
391 u64 dma_addr = ib_sg_dma_address(dev, &scat[i]);
392
Andy Grovera870d622009-07-17 13:13:33 +0000393 if (dma_addr & ~PAGE_MASK) {
Andy Grover08b48a12009-02-24 15:30:32 +0000394 if (i > 0)
395 return -EINVAL;
396 else
397 ++page_cnt;
398 }
Andy Grovera870d622009-07-17 13:13:33 +0000399 if ((dma_addr + dma_len) & ~PAGE_MASK) {
Andy Grover08b48a12009-02-24 15:30:32 +0000400 if (i < sg_dma_len - 1)
401 return -EINVAL;
402 else
403 ++page_cnt;
404 }
405
406 len += dma_len;
407 }
408
Andy Grovera870d622009-07-17 13:13:33 +0000409 page_cnt += len >> PAGE_SHIFT;
Andy Grover08b48a12009-02-24 15:30:32 +0000410 if (page_cnt > fmr_message_size)
411 return -EINVAL;
412
Andy Grovere4c52c92010-04-23 10:49:53 -0700413 dma_pages = kmalloc_node(sizeof(u64) * page_cnt, GFP_ATOMIC,
414 rdsibdev_to_node(rds_ibdev));
Andy Grover08b48a12009-02-24 15:30:32 +0000415 if (!dma_pages)
416 return -ENOMEM;
417
418 page_cnt = 0;
419 for (i = 0; i < sg_dma_len; ++i) {
420 unsigned int dma_len = ib_sg_dma_len(dev, &scat[i]);
421 u64 dma_addr = ib_sg_dma_address(dev, &scat[i]);
422
Andy Grovera870d622009-07-17 13:13:33 +0000423 for (j = 0; j < dma_len; j += PAGE_SIZE)
Andy Grover08b48a12009-02-24 15:30:32 +0000424 dma_pages[page_cnt++] =
Andy Grovera870d622009-07-17 13:13:33 +0000425 (dma_addr & PAGE_MASK) + j;
Andy Grover08b48a12009-02-24 15:30:32 +0000426 }
427
428 ret = ib_map_phys_fmr(ibmr->fmr,
429 dma_pages, page_cnt, io_addr);
430 if (ret)
431 goto out;
432
433 /* Success - we successfully remapped the MR, so we can
434 * safely tear down the old mapping. */
435 rds_ib_teardown_mr(ibmr);
436
437 ibmr->sg = scat;
438 ibmr->sg_len = nents;
439 ibmr->sg_dma_len = sg_dma_len;
440 ibmr->remap_count++;
441
442 rds_ib_stats_inc(s_ib_rdma_mr_used);
443 ret = 0;
444
445out:
446 kfree(dma_pages);
447
448 return ret;
449}
450
451void rds_ib_sync_mr(void *trans_private, int direction)
452{
453 struct rds_ib_mr *ibmr = trans_private;
454 struct rds_ib_device *rds_ibdev = ibmr->device;
455
456 switch (direction) {
457 case DMA_FROM_DEVICE:
458 ib_dma_sync_sg_for_cpu(rds_ibdev->dev, ibmr->sg,
459 ibmr->sg_dma_len, DMA_BIDIRECTIONAL);
460 break;
461 case DMA_TO_DEVICE:
462 ib_dma_sync_sg_for_device(rds_ibdev->dev, ibmr->sg,
463 ibmr->sg_dma_len, DMA_BIDIRECTIONAL);
464 break;
465 }
466}
467
468static void __rds_ib_teardown_mr(struct rds_ib_mr *ibmr)
469{
470 struct rds_ib_device *rds_ibdev = ibmr->device;
471
472 if (ibmr->sg_dma_len) {
473 ib_dma_unmap_sg(rds_ibdev->dev,
474 ibmr->sg, ibmr->sg_len,
475 DMA_BIDIRECTIONAL);
476 ibmr->sg_dma_len = 0;
477 }
478
479 /* Release the s/g list */
480 if (ibmr->sg_len) {
481 unsigned int i;
482
483 for (i = 0; i < ibmr->sg_len; ++i) {
484 struct page *page = sg_page(&ibmr->sg[i]);
485
486 /* FIXME we need a way to tell a r/w MR
487 * from a r/o MR */
Andy Grover9e2effb2010-03-12 16:22:32 -0800488 BUG_ON(irqs_disabled());
Andy Grover08b48a12009-02-24 15:30:32 +0000489 set_page_dirty(page);
490 put_page(page);
491 }
492 kfree(ibmr->sg);
493
494 ibmr->sg = NULL;
495 ibmr->sg_len = 0;
496 }
497}
498
499static void rds_ib_teardown_mr(struct rds_ib_mr *ibmr)
500{
501 unsigned int pinned = ibmr->sg_len;
502
503 __rds_ib_teardown_mr(ibmr);
504 if (pinned) {
505 struct rds_ib_device *rds_ibdev = ibmr->device;
506 struct rds_ib_mr_pool *pool = rds_ibdev->mr_pool;
507
508 atomic_sub(pinned, &pool->free_pinned);
509 }
510}
511
512static inline unsigned int rds_ib_flush_goal(struct rds_ib_mr_pool *pool, int free_all)
513{
514 unsigned int item_count;
515
516 item_count = atomic_read(&pool->item_count);
517 if (free_all)
518 return item_count;
519
520 return 0;
521}
522
523/*
Huang Ying1bc144b2011-08-30 15:21:30 +0000524 * given an llist of mrs, put them all into the list_head for more processing
Chris Mason6fa70da2010-06-11 11:17:59 -0700525 */
Huang Ying1bc144b2011-08-30 15:21:30 +0000526static void llist_append_to_list(struct llist_head *llist, struct list_head *list)
Chris Mason6fa70da2010-06-11 11:17:59 -0700527{
528 struct rds_ib_mr *ibmr;
Huang Ying1bc144b2011-08-30 15:21:30 +0000529 struct llist_node *node;
530 struct llist_node *next;
Chris Mason6fa70da2010-06-11 11:17:59 -0700531
Huang Ying1bc144b2011-08-30 15:21:30 +0000532 node = llist_del_all(llist);
533 while (node) {
534 next = node->next;
535 ibmr = llist_entry(node, struct rds_ib_mr, llnode);
Chris Mason6fa70da2010-06-11 11:17:59 -0700536 list_add_tail(&ibmr->unmap_list, list);
Huang Ying1bc144b2011-08-30 15:21:30 +0000537 node = next;
Chris Mason6fa70da2010-06-11 11:17:59 -0700538 }
539}
540
541/*
Huang Ying1bc144b2011-08-30 15:21:30 +0000542 * this takes a list head of mrs and turns it into linked llist nodes
543 * of clusters. Each cluster has linked llist nodes of
544 * MR_CLUSTER_SIZE mrs that are ready for reuse.
Chris Mason6fa70da2010-06-11 11:17:59 -0700545 */
Huang Ying1bc144b2011-08-30 15:21:30 +0000546static void list_to_llist_nodes(struct rds_ib_mr_pool *pool,
547 struct list_head *list,
548 struct llist_node **nodes_head,
549 struct llist_node **nodes_tail)
Chris Mason6fa70da2010-06-11 11:17:59 -0700550{
551 struct rds_ib_mr *ibmr;
Huang Ying1bc144b2011-08-30 15:21:30 +0000552 struct llist_node *cur = NULL;
553 struct llist_node **next = nodes_head;
Chris Mason6fa70da2010-06-11 11:17:59 -0700554
555 list_for_each_entry(ibmr, list, unmap_list) {
Huang Ying1bc144b2011-08-30 15:21:30 +0000556 cur = &ibmr->llnode;
557 *next = cur;
558 next = &cur->next;
Chris Mason6fa70da2010-06-11 11:17:59 -0700559 }
Huang Ying1bc144b2011-08-30 15:21:30 +0000560 *next = NULL;
561 *nodes_tail = cur;
Chris Mason6fa70da2010-06-11 11:17:59 -0700562}
563
564/*
Andy Grover08b48a12009-02-24 15:30:32 +0000565 * Flush our pool of MRs.
566 * At a minimum, all currently unused MRs are unmapped.
567 * If the number of MRs allocated exceeds the limit, we also try
568 * to free as many MRs as needed to get back to this limit.
569 */
Chris Mason6fa70da2010-06-11 11:17:59 -0700570static int rds_ib_flush_mr_pool(struct rds_ib_mr_pool *pool,
571 int free_all, struct rds_ib_mr **ibmr_ret)
Andy Grover08b48a12009-02-24 15:30:32 +0000572{
573 struct rds_ib_mr *ibmr, *next;
Huang Ying1bc144b2011-08-30 15:21:30 +0000574 struct llist_node *clean_nodes;
575 struct llist_node *clean_tail;
Andy Grover08b48a12009-02-24 15:30:32 +0000576 LIST_HEAD(unmap_list);
577 LIST_HEAD(fmr_list);
578 unsigned long unpinned = 0;
Andy Grover08b48a12009-02-24 15:30:32 +0000579 unsigned int nfreed = 0, ncleaned = 0, free_goal;
580 int ret = 0;
581
582 rds_ib_stats_inc(s_ib_rdma_mr_pool_flush);
583
Chris Mason6fa70da2010-06-11 11:17:59 -0700584 if (ibmr_ret) {
585 DEFINE_WAIT(wait);
586 while(!mutex_trylock(&pool->flush_lock)) {
587 ibmr = rds_ib_reuse_fmr(pool);
588 if (ibmr) {
589 *ibmr_ret = ibmr;
590 finish_wait(&pool->flush_wait, &wait);
591 goto out_nolock;
592 }
Andy Grover08b48a12009-02-24 15:30:32 +0000593
Chris Mason6fa70da2010-06-11 11:17:59 -0700594 prepare_to_wait(&pool->flush_wait, &wait,
595 TASK_UNINTERRUPTIBLE);
Huang Ying1bc144b2011-08-30 15:21:30 +0000596 if (llist_empty(&pool->clean_list))
Chris Mason6fa70da2010-06-11 11:17:59 -0700597 schedule();
598
599 ibmr = rds_ib_reuse_fmr(pool);
600 if (ibmr) {
601 *ibmr_ret = ibmr;
602 finish_wait(&pool->flush_wait, &wait);
603 goto out_nolock;
604 }
605 }
606 finish_wait(&pool->flush_wait, &wait);
607 } else
608 mutex_lock(&pool->flush_lock);
609
610 if (ibmr_ret) {
611 ibmr = rds_ib_reuse_fmr(pool);
612 if (ibmr) {
613 *ibmr_ret = ibmr;
614 goto out;
615 }
616 }
617
Andy Grover08b48a12009-02-24 15:30:32 +0000618 /* Get the list of all MRs to be dropped. Ordering matters -
Chris Mason6fa70da2010-06-11 11:17:59 -0700619 * we want to put drop_list ahead of free_list.
620 */
Huang Ying1bc144b2011-08-30 15:21:30 +0000621 llist_append_to_list(&pool->drop_list, &unmap_list);
622 llist_append_to_list(&pool->free_list, &unmap_list);
Andy Grover08b48a12009-02-24 15:30:32 +0000623 if (free_all)
Huang Ying1bc144b2011-08-30 15:21:30 +0000624 llist_append_to_list(&pool->clean_list, &unmap_list);
Andy Grover08b48a12009-02-24 15:30:32 +0000625
626 free_goal = rds_ib_flush_goal(pool, free_all);
627
628 if (list_empty(&unmap_list))
629 goto out;
630
631 /* String all ib_mr's onto one list and hand them to ib_unmap_fmr */
Chris Mason6fa70da2010-06-11 11:17:59 -0700632 list_for_each_entry(ibmr, &unmap_list, unmap_list)
Andy Grover08b48a12009-02-24 15:30:32 +0000633 list_add(&ibmr->fmr->list, &fmr_list);
Chris Mason6fa70da2010-06-11 11:17:59 -0700634
Andy Grover08b48a12009-02-24 15:30:32 +0000635 ret = ib_unmap_fmr(&fmr_list);
636 if (ret)
637 printk(KERN_WARNING "RDS/IB: ib_unmap_fmr failed (err=%d)\n", ret);
638
639 /* Now we can destroy the DMA mapping and unpin any pages */
Chris Mason6fa70da2010-06-11 11:17:59 -0700640 list_for_each_entry_safe(ibmr, next, &unmap_list, unmap_list) {
Andy Grover08b48a12009-02-24 15:30:32 +0000641 unpinned += ibmr->sg_len;
642 __rds_ib_teardown_mr(ibmr);
643 if (nfreed < free_goal || ibmr->remap_count >= pool->fmr_attr.max_maps) {
644 rds_ib_stats_inc(s_ib_rdma_mr_free);
Chris Mason6fa70da2010-06-11 11:17:59 -0700645 list_del(&ibmr->unmap_list);
Andy Grover08b48a12009-02-24 15:30:32 +0000646 ib_dealloc_fmr(ibmr->fmr);
647 kfree(ibmr);
648 nfreed++;
649 }
650 ncleaned++;
651 }
652
Chris Mason6fa70da2010-06-11 11:17:59 -0700653 if (!list_empty(&unmap_list)) {
654 /* we have to make sure that none of the things we're about
655 * to put on the clean list would race with other cpus trying
Huang Ying1bc144b2011-08-30 15:21:30 +0000656 * to pull items off. The llist would explode if we managed to
Chris Mason6fa70da2010-06-11 11:17:59 -0700657 * remove something from the clean list and then add it back again
Huang Ying1bc144b2011-08-30 15:21:30 +0000658 * while another CPU was spinning on that same item in llist_del_first.
Chris Mason6fa70da2010-06-11 11:17:59 -0700659 *
Huang Ying1bc144b2011-08-30 15:21:30 +0000660 * This is pretty unlikely, but just in case wait for an llist grace period
Chris Mason6fa70da2010-06-11 11:17:59 -0700661 * here before adding anything back into the clean list.
662 */
663 wait_clean_list_grace();
664
Huang Ying1bc144b2011-08-30 15:21:30 +0000665 list_to_llist_nodes(pool, &unmap_list, &clean_nodes, &clean_tail);
Chris Mason6fa70da2010-06-11 11:17:59 -0700666 if (ibmr_ret)
Huang Ying1bc144b2011-08-30 15:21:30 +0000667 *ibmr_ret = llist_entry(clean_nodes, struct rds_ib_mr, llnode);
Chris Mason6fa70da2010-06-11 11:17:59 -0700668
Huang Ying1bc144b2011-08-30 15:21:30 +0000669 /* more than one entry in llist nodes */
670 if (clean_nodes->next)
671 llist_add_batch(clean_nodes->next, clean_tail, &pool->clean_list);
Chris Mason6fa70da2010-06-11 11:17:59 -0700672
673 }
Andy Grover08b48a12009-02-24 15:30:32 +0000674
675 atomic_sub(unpinned, &pool->free_pinned);
676 atomic_sub(ncleaned, &pool->dirty_count);
677 atomic_sub(nfreed, &pool->item_count);
678
679out:
680 mutex_unlock(&pool->flush_lock);
Chris Mason6fa70da2010-06-11 11:17:59 -0700681 if (waitqueue_active(&pool->flush_wait))
682 wake_up(&pool->flush_wait);
683out_nolock:
Andy Grover08b48a12009-02-24 15:30:32 +0000684 return ret;
685}
686
687static void rds_ib_mr_pool_flush_worker(struct work_struct *work)
688{
Chris Mason7a0ff5d2010-06-11 11:26:02 -0700689 struct rds_ib_mr_pool *pool = container_of(work, struct rds_ib_mr_pool, flush_worker.work);
Andy Grover08b48a12009-02-24 15:30:32 +0000690
Chris Mason6fa70da2010-06-11 11:17:59 -0700691 rds_ib_flush_mr_pool(pool, 0, NULL);
Andy Grover08b48a12009-02-24 15:30:32 +0000692}
693
694void rds_ib_free_mr(void *trans_private, int invalidate)
695{
696 struct rds_ib_mr *ibmr = trans_private;
697 struct rds_ib_device *rds_ibdev = ibmr->device;
698 struct rds_ib_mr_pool *pool = rds_ibdev->mr_pool;
Andy Grover08b48a12009-02-24 15:30:32 +0000699
700 rdsdebug("RDS/IB: free_mr nents %u\n", ibmr->sg_len);
701
702 /* Return it to the pool's free list */
Andy Grover08b48a12009-02-24 15:30:32 +0000703 if (ibmr->remap_count >= pool->fmr_attr.max_maps)
Huang Ying1bc144b2011-08-30 15:21:30 +0000704 llist_add(&ibmr->llnode, &pool->drop_list);
Andy Grover08b48a12009-02-24 15:30:32 +0000705 else
Huang Ying1bc144b2011-08-30 15:21:30 +0000706 llist_add(&ibmr->llnode, &pool->free_list);
Andy Grover08b48a12009-02-24 15:30:32 +0000707
708 atomic_add(ibmr->sg_len, &pool->free_pinned);
709 atomic_inc(&pool->dirty_count);
Andy Grover08b48a12009-02-24 15:30:32 +0000710
711 /* If we've pinned too many pages, request a flush */
Joe Perchesf64f9e72009-11-29 16:55:45 -0800712 if (atomic_read(&pool->free_pinned) >= pool->max_free_pinned ||
713 atomic_read(&pool->dirty_count) >= pool->max_items / 10)
Tejun Heoc534a102011-02-01 11:42:43 +0100714 schedule_delayed_work(&pool->flush_worker, 10);
Andy Grover08b48a12009-02-24 15:30:32 +0000715
716 if (invalidate) {
717 if (likely(!in_interrupt())) {
Chris Mason6fa70da2010-06-11 11:17:59 -0700718 rds_ib_flush_mr_pool(pool, 0, NULL);
Andy Grover08b48a12009-02-24 15:30:32 +0000719 } else {
720 /* We get here if the user created a MR marked
721 * as use_once and invalidate at the same time. */
Tejun Heoc534a102011-02-01 11:42:43 +0100722 schedule_delayed_work(&pool->flush_worker, 10);
Andy Grover08b48a12009-02-24 15:30:32 +0000723 }
724 }
Zach Brown3e0249f2010-05-18 15:48:51 -0700725
726 rds_ib_dev_put(rds_ibdev);
Andy Grover08b48a12009-02-24 15:30:32 +0000727}
728
729void rds_ib_flush_mrs(void)
730{
731 struct rds_ib_device *rds_ibdev;
732
Zach Brownea819862010-07-15 12:34:33 -0700733 down_read(&rds_ib_devices_lock);
Andy Grover08b48a12009-02-24 15:30:32 +0000734 list_for_each_entry(rds_ibdev, &rds_ib_devices, list) {
735 struct rds_ib_mr_pool *pool = rds_ibdev->mr_pool;
736
737 if (pool)
Chris Mason6fa70da2010-06-11 11:17:59 -0700738 rds_ib_flush_mr_pool(pool, 0, NULL);
Andy Grover08b48a12009-02-24 15:30:32 +0000739 }
Zach Brownea819862010-07-15 12:34:33 -0700740 up_read(&rds_ib_devices_lock);
Andy Grover08b48a12009-02-24 15:30:32 +0000741}
742
743void *rds_ib_get_mr(struct scatterlist *sg, unsigned long nents,
744 struct rds_sock *rs, u32 *key_ret)
745{
746 struct rds_ib_device *rds_ibdev;
747 struct rds_ib_mr *ibmr = NULL;
748 int ret;
749
750 rds_ibdev = rds_ib_get_device(rs->rs_bound_addr);
751 if (!rds_ibdev) {
752 ret = -ENODEV;
753 goto out;
754 }
755
756 if (!rds_ibdev->mr_pool) {
757 ret = -ENODEV;
758 goto out;
759 }
760
761 ibmr = rds_ib_alloc_fmr(rds_ibdev);
762 if (IS_ERR(ibmr))
763 return ibmr;
764
765 ret = rds_ib_map_fmr(rds_ibdev, ibmr, sg, nents);
766 if (ret == 0)
767 *key_ret = ibmr->fmr->rkey;
768 else
769 printk(KERN_WARNING "RDS/IB: map_fmr failed (errno=%d)\n", ret);
770
771 ibmr->device = rds_ibdev;
Zach Brown3e0249f2010-05-18 15:48:51 -0700772 rds_ibdev = NULL;
Andy Grover08b48a12009-02-24 15:30:32 +0000773
774 out:
775 if (ret) {
776 if (ibmr)
777 rds_ib_free_mr(ibmr, 0);
778 ibmr = ERR_PTR(ret);
779 }
Zach Brown3e0249f2010-05-18 15:48:51 -0700780 if (rds_ibdev)
781 rds_ib_dev_put(rds_ibdev);
Andy Grover08b48a12009-02-24 15:30:32 +0000782 return ibmr;
783}
Chris Mason6fa70da2010-06-11 11:17:59 -0700784