drivers/infiniband/hw/hfi1/mmu_rb.c

   1 /*
   2  * Copyright(c) 2016 - 2017 Intel Corporation.
   3  *
   4  * This file is provided under a dual BSD/GPLv2 license.  When using or
   5  * redistributing this file, you may do so under either license.
   6  *
   7  * GPL LICENSE SUMMARY
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of version 2 of the GNU General Public License as
  11  * published by the Free Software Foundation.
  12  *
  13  * This program is distributed in the hope that it will be useful, but
  14  * WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * General Public License for more details.
  17  *
  18  * BSD LICENSE
  19  *
  20  * Redistribution and use in source and binary forms, with or without
  21  * modification, are permitted provided that the following conditions
  22  * are met:
  23  *
  24  *  - Redistributions of source code must retain the above copyright
  25  *    notice, this list of conditions and the following disclaimer.
  26  *  - Redistributions in binary form must reproduce the above copyright
  27  *    notice, this list of conditions and the following disclaimer in
  28  *    the documentation and/or other materials provided with the
  29  *    distribution.
  30  *  - Neither the name of Intel Corporation nor the names of its
  31  *    contributors may be used to endorse or promote products derived
  32  *    from this software without specific prior written permission.
  33  *
  34  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  35  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  36  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  37  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  38  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  39  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  40  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  41  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  42  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  43  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  44  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  45  *
  46  */
  47 #include <linux/list.h>
  48 #include <linux/rculist.h>
  49 #include <linux/mmu_notifier.h>
  50 #include <linux/interval_tree_generic.h>
  51
  52 #include "mmu_rb.h"
  53 #include "trace.h"
  54
  55 struct mmu_rb_handler {
  56         struct mmu_notifier mn;
  57         struct rb_root_cached root;
  58         void *ops_arg;
  59         spinlock_t lock;        /* protect the RB tree */
  60         struct mmu_rb_ops *ops;
  61         struct mm_struct *mm;
  62         struct list_head lru_list;
  63         struct work_struct del_work;
  64         struct list_head del_list;
  65         struct workqueue_struct *wq;
  66 };
  67
  68 static unsigned long mmu_node_start(struct mmu_rb_node *);
  69 static unsigned long mmu_node_last(struct mmu_rb_node *);
  70 static void mmu_notifier_range_start(struct mmu_notifier *,
  71                                      struct mm_struct *,
  72                                      unsigned long, unsigned long);
  73 static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *,
  74                                            unsigned long, unsigned long);
  75 static void do_remove(struct mmu_rb_handler *handler,
  76                       struct list_head *del_list);
  77 static void handle_remove(struct work_struct *work);
  78
  79 static const struct mmu_notifier_ops mn_opts = {
  80         .flags = MMU_INVALIDATE_DOES_NOT_BLOCK,
  81         .invalidate_range_start = mmu_notifier_range_start,
  82 };
  83
  84 INTERVAL_TREE_DEFINE(struct mmu_rb_node, node, unsigned long, __last,
  85                      mmu_node_start, mmu_node_last, static, __mmu_int_rb);
  86
  87 static unsigned long mmu_node_start(struct mmu_rb_node *node)
  88 {
  89         return node->addr & PAGE_MASK;
  90 }
  91
  92 static unsigned long mmu_node_last(struct mmu_rb_node *node)
  93 {
  94         return PAGE_ALIGN(node->addr + node->len) - 1;
  95 }
  96
  97 int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm,
  98                          struct mmu_rb_ops *ops,
  99                          struct workqueue_struct *wq,
 100                          struct mmu_rb_handler **handler)
 101 {
 102         struct mmu_rb_handler *handlr;
 103         int ret;
 104
 105         handlr = kmalloc(sizeof(*handlr), GFP_KERNEL);
 106         if (!handlr)
 107                 return -ENOMEM;
 108
 109         handlr->root = RB_ROOT_CACHED;
 110         handlr->ops = ops;
 111         handlr->ops_arg = ops_arg;
 112         INIT_HLIST_NODE(&handlr->mn.hlist);
 113         spin_lock_init(&handlr->lock);
 114         handlr->mn.ops = &mn_opts;
 115         handlr->mm = mm;
 116         INIT_WORK(&handlr->del_work, handle_remove);
 117         INIT_LIST_HEAD(&handlr->del_list);
 118         INIT_LIST_HEAD(&handlr->lru_list);
 119         handlr->wq = wq;
 120
 121         ret = mmu_notifier_register(&handlr->mn, handlr->mm);
 122         if (ret) {
 123                 kfree(handlr);
 124                 return ret;
 125         }
 126
 127         *handler = handlr;
 128         return 0;
 129 }
 130
 131 void hfi1_mmu_rb_unregister(struct mmu_rb_handler *handler)
 132 {
 133         struct mmu_rb_node *rbnode;
 134         struct rb_node *node;
 135         unsigned long flags;
 136         struct list_head del_list;
 137
 138         /* Unregister first so we don't get any more notifications. */
 139         mmu_notifier_unregister(&handler->mn, handler->mm);
 140
 141         /*
 142          * Make sure the wq delete handler is finished running.  It will not
 143          * be triggered once the mmu notifiers are unregistered above.
 144          */
 145         flush_work(&handler->del_work);
 146
 147         INIT_LIST_HEAD(&del_list);
 148
 149         spin_lock_irqsave(&handler->lock, flags);
 150         while ((node = rb_first_cached(&handler->root))) {
 151                 rbnode = rb_entry(node, struct mmu_rb_node, node);
 152                 rb_erase_cached(node, &handler->root);
 153                 /* move from LRU list to delete list */
 154                 list_move(&rbnode->list, &del_list);
 155         }
 156         spin_unlock_irqrestore(&handler->lock, flags);
 157
 158         do_remove(handler, &del_list);
 159
 160         kfree(handler);
 161 }
 162
 163 int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler,
 164                        struct mmu_rb_node *mnode)
 165 {
 166         struct mmu_rb_node *node;
 167         unsigned long flags;
 168         int ret = 0;
 169
 170         trace_hfi1_mmu_rb_insert(mnode->addr, mnode->len);
 171         spin_lock_irqsave(&handler->lock, flags);
 172         node = __mmu_rb_search(handler, mnode->addr, mnode->len);
 173         if (node) {
 174                 ret = -EINVAL;
 175                 goto unlock;
 176         }
 177         __mmu_int_rb_insert(mnode, &handler->root);
 178         list_add(&mnode->list, &handler->lru_list);
 179
 180         ret = handler->ops->insert(handler->ops_arg, mnode);
 181         if (ret) {
 182                 __mmu_int_rb_remove(mnode, &handler->root);
 183                 list_del(&mnode->list); /* remove from LRU list */
 184         }
 185 unlock:
 186         spin_unlock_irqrestore(&handler->lock, flags);
 187         return ret;
 188 }
 189
 190 /* Caller must hold handler lock */
 191 static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *handler,
 192                                            unsigned long addr,
 193                                            unsigned long len)
 194 {
 195         struct mmu_rb_node *node = NULL;
 196
 197         trace_hfi1_mmu_rb_search(addr, len);
 198         if (!handler->ops->filter) {
 199                 node = __mmu_int_rb_iter_first(&handler->root, addr,
 200                                                (addr + len) - 1);
 201         } else {
 202                 for (node = __mmu_int_rb_iter_first(&handler->root, addr,
 203                                                     (addr + len) - 1);
 204                      node;
 205                      node = __mmu_int_rb_iter_next(node, addr,
 206                                                    (addr + len) - 1)) {
 207                         if (handler->ops->filter(node, addr, len))
 208                                 return node;
 209                 }
 210         }
 211         return node;
 212 }
 213
 214 bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler,
 215                                      unsigned long addr, unsigned long len,
 216                                      struct mmu_rb_node **rb_node)
 217 {
 218         struct mmu_rb_node *node;
 219         unsigned long flags;
 220         bool ret = false;
 221
 222         spin_lock_irqsave(&handler->lock, flags);
 223         node = __mmu_rb_search(handler, addr, len);
 224         if (node) {
 225                 if (node->addr == addr && node->len == len)
 226                         goto unlock;
 227                 __mmu_int_rb_remove(node, &handler->root);
 228                 list_del(&node->list); /* remove from LRU list */
 229                 ret = true;
 230         }
 231 unlock:
 232         spin_unlock_irqrestore(&handler->lock, flags);
 233         *rb_node = node;
 234         return ret;
 235 }
 236
 237 void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg)
 238 {
 239         struct mmu_rb_node *rbnode, *ptr;
 240         struct list_head del_list;
 241         unsigned long flags;
 242         bool stop = false;
 243
 244         INIT_LIST_HEAD(&del_list);
 245
 246         spin_lock_irqsave(&handler->lock, flags);
 247         list_for_each_entry_safe_reverse(rbnode, ptr, &handler->lru_list,
 248                                          list) {
 249                 if (handler->ops->evict(handler->ops_arg, rbnode, evict_arg,
 250                                         &stop)) {
 251                         __mmu_int_rb_remove(rbnode, &handler->root);
 252                         /* move from LRU list to delete list */
 253                         list_move(&rbnode->list, &del_list);
 254                 }
 255                 if (stop)
 256                         break;
 257         }
 258         spin_unlock_irqrestore(&handler->lock, flags);
 259
 260         while (!list_empty(&del_list)) {
 261                 rbnode = list_first_entry(&del_list, struct mmu_rb_node, list);
 262                 list_del(&rbnode->list);
 263                 handler->ops->remove(handler->ops_arg, rbnode);
 264         }
 265 }
 266
 267 /*
 268  * It is up to the caller to ensure that this function does not race with the
 269  * mmu invalidate notifier which may be calling the users remove callback on
 270  * 'node'.
 271  */
 272 void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler,
 273                         struct mmu_rb_node *node)
 274 {
 275         unsigned long flags;
 276
 277         /* Validity of handler and node pointers has been checked by caller. */
 278         trace_hfi1_mmu_rb_remove(node->addr, node->len);
 279         spin_lock_irqsave(&handler->lock, flags);
 280         __mmu_int_rb_remove(node, &handler->root);
 281         list_del(&node->list); /* remove from LRU list */
 282         spin_unlock_irqrestore(&handler->lock, flags);
 283
 284         handler->ops->remove(handler->ops_arg, node);
 285 }
 286
 287 static void mmu_notifier_range_start(struct mmu_notifier *mn,
 288                                      struct mm_struct *mm,
 289                                      unsigned long start,
 290                                      unsigned long end)
 291 {
 292         struct mmu_rb_handler *handler =
 293                 container_of(mn, struct mmu_rb_handler, mn);
 294         struct rb_root_cached *root = &handler->root;
 295         struct mmu_rb_node *node, *ptr = NULL;
 296         unsigned long flags;
 297         bool added = false;
 298
 299         spin_lock_irqsave(&handler->lock, flags);
 300         for (node = __mmu_int_rb_iter_first(root, start, end - 1);
 301              node; node = ptr) {
 302                 /* Guard against node removal. */
 303                 ptr = __mmu_int_rb_iter_next(node, start, end - 1);
 304                 trace_hfi1_mmu_mem_invalidate(node->addr, node->len);
 305                 if (handler->ops->invalidate(handler->ops_arg, node)) {
 306                         __mmu_int_rb_remove(node, root);
 307                         /* move from LRU list to delete list */
 308                         list_move(&node->list, &handler->del_list);
 309                         added = true;
 310                 }
 311         }
 312         spin_unlock_irqrestore(&handler->lock, flags);
 313
 314         if (added)
 315                 queue_work(handler->wq, &handler->del_work);
 316 }
 317
 318 /*
 319  * Call the remove function for the given handler and the list.  This
 320  * is expected to be called with a delete list extracted from handler.
 321  * The caller should not be holding the handler lock.
 322  */
 323 static void do_remove(struct mmu_rb_handler *handler,
 324                       struct list_head *del_list)
 325 {
 326         struct mmu_rb_node *node;
 327
 328         while (!list_empty(del_list)) {
 329                 node = list_first_entry(del_list, struct mmu_rb_node, list);
 330                 list_del(&node->list);
 331                 handler->ops->remove(handler->ops_arg, node);
 332         }
 333 }
 334
 335 /*
 336  * Work queue function to remove all nodes that have been queued up to
 337  * be removed.  The key feature is that mm->mmap_sem is not being held
 338  * and the remove callback can sleep while taking it, if needed.
 339  */
 340 static void handle_remove(struct work_struct *work)
 341 {
 342         struct mmu_rb_handler *handler = container_of(work,
 343                                                 struct mmu_rb_handler,
 344                                                 del_work);
 345         struct list_head del_list;
 346         unsigned long flags;
 347
 348         /* remove anything that is queued to get removed */
 349         spin_lock_irqsave(&handler->lock, flags);
 350         list_replace_init(&handler->del_list, &del_list);
 351         spin_unlock_irqrestore(&handler->lock, flags);
 352
 353         do_remove(handler, &del_list);
 354 }