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 int mmu_notifier_range_start(struct mmu_notifier *,
71 		const struct mmu_notifier_range *);
72 static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *,
73 					   unsigned long, unsigned long);
74 static void do_remove(struct mmu_rb_handler *handler,
75 		      struct list_head *del_list);
76 static void handle_remove(struct work_struct *work);
77 
78 static const struct mmu_notifier_ops mn_opts = {
79 	.invalidate_range_start = mmu_notifier_range_start,
80 };
81 
82 INTERVAL_TREE_DEFINE(struct mmu_rb_node, node, unsigned long, __last,
83 		     mmu_node_start, mmu_node_last, static, __mmu_int_rb);
84 
85 static unsigned long mmu_node_start(struct mmu_rb_node *node)
86 {
87 	return node->addr & PAGE_MASK;
88 }
89 
90 static unsigned long mmu_node_last(struct mmu_rb_node *node)
91 {
92 	return PAGE_ALIGN(node->addr + node->len) - 1;
93 }
94 
95 int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm,
96 			 struct mmu_rb_ops *ops,
97 			 struct workqueue_struct *wq,
98 			 struct mmu_rb_handler **handler)
99 {
100 	struct mmu_rb_handler *handlr;
101 	int ret;
102 
103 	handlr = kmalloc(sizeof(*handlr), GFP_KERNEL);
104 	if (!handlr)
105 		return -ENOMEM;
106 
107 	handlr->root = RB_ROOT_CACHED;
108 	handlr->ops = ops;
109 	handlr->ops_arg = ops_arg;
110 	INIT_HLIST_NODE(&handlr->mn.hlist);
111 	spin_lock_init(&handlr->lock);
112 	handlr->mn.ops = &mn_opts;
113 	handlr->mm = mm;
114 	INIT_WORK(&handlr->del_work, handle_remove);
115 	INIT_LIST_HEAD(&handlr->del_list);
116 	INIT_LIST_HEAD(&handlr->lru_list);
117 	handlr->wq = wq;
118 
119 	ret = mmu_notifier_register(&handlr->mn, handlr->mm);
120 	if (ret) {
121 		kfree(handlr);
122 		return ret;
123 	}
124 
125 	*handler = handlr;
126 	return 0;
127 }
128 
129 void hfi1_mmu_rb_unregister(struct mmu_rb_handler *handler)
130 {
131 	struct mmu_rb_node *rbnode;
132 	struct rb_node *node;
133 	unsigned long flags;
134 	struct list_head del_list;
135 
136 	/* Unregister first so we don't get any more notifications. */
137 	mmu_notifier_unregister(&handler->mn, handler->mm);
138 
139 	/*
140 	 * Make sure the wq delete handler is finished running.  It will not
141 	 * be triggered once the mmu notifiers are unregistered above.
142 	 */
143 	flush_work(&handler->del_work);
144 
145 	INIT_LIST_HEAD(&del_list);
146 
147 	spin_lock_irqsave(&handler->lock, flags);
148 	while ((node = rb_first_cached(&handler->root))) {
149 		rbnode = rb_entry(node, struct mmu_rb_node, node);
150 		rb_erase_cached(node, &handler->root);
151 		/* move from LRU list to delete list */
152 		list_move(&rbnode->list, &del_list);
153 	}
154 	spin_unlock_irqrestore(&handler->lock, flags);
155 
156 	do_remove(handler, &del_list);
157 
158 	kfree(handler);
159 }
160 
161 int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler,
162 		       struct mmu_rb_node *mnode)
163 {
164 	struct mmu_rb_node *node;
165 	unsigned long flags;
166 	int ret = 0;
167 
168 	trace_hfi1_mmu_rb_insert(mnode->addr, mnode->len);
169 	spin_lock_irqsave(&handler->lock, flags);
170 	node = __mmu_rb_search(handler, mnode->addr, mnode->len);
171 	if (node) {
172 		ret = -EINVAL;
173 		goto unlock;
174 	}
175 	__mmu_int_rb_insert(mnode, &handler->root);
176 	list_add(&mnode->list, &handler->lru_list);
177 
178 	ret = handler->ops->insert(handler->ops_arg, mnode);
179 	if (ret) {
180 		__mmu_int_rb_remove(mnode, &handler->root);
181 		list_del(&mnode->list); /* remove from LRU list */
182 	}
183 unlock:
184 	spin_unlock_irqrestore(&handler->lock, flags);
185 	return ret;
186 }
187 
188 /* Caller must hold handler lock */
189 static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *handler,
190 					   unsigned long addr,
191 					   unsigned long len)
192 {
193 	struct mmu_rb_node *node = NULL;
194 
195 	trace_hfi1_mmu_rb_search(addr, len);
196 	if (!handler->ops->filter) {
197 		node = __mmu_int_rb_iter_first(&handler->root, addr,
198 					       (addr + len) - 1);
199 	} else {
200 		for (node = __mmu_int_rb_iter_first(&handler->root, addr,
201 						    (addr + len) - 1);
202 		     node;
203 		     node = __mmu_int_rb_iter_next(node, addr,
204 						   (addr + len) - 1)) {
205 			if (handler->ops->filter(node, addr, len))
206 				return node;
207 		}
208 	}
209 	return node;
210 }
211 
212 bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler,
213 				     unsigned long addr, unsigned long len,
214 				     struct mmu_rb_node **rb_node)
215 {
216 	struct mmu_rb_node *node;
217 	unsigned long flags;
218 	bool ret = false;
219 
220 	spin_lock_irqsave(&handler->lock, flags);
221 	node = __mmu_rb_search(handler, addr, len);
222 	if (node) {
223 		if (node->addr == addr && node->len == len)
224 			goto unlock;
225 		__mmu_int_rb_remove(node, &handler->root);
226 		list_del(&node->list); /* remove from LRU list */
227 		ret = true;
228 	}
229 unlock:
230 	spin_unlock_irqrestore(&handler->lock, flags);
231 	*rb_node = node;
232 	return ret;
233 }
234 
235 void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg)
236 {
237 	struct mmu_rb_node *rbnode, *ptr;
238 	struct list_head del_list;
239 	unsigned long flags;
240 	bool stop = false;
241 
242 	INIT_LIST_HEAD(&del_list);
243 
244 	spin_lock_irqsave(&handler->lock, flags);
245 	list_for_each_entry_safe_reverse(rbnode, ptr, &handler->lru_list,
246 					 list) {
247 		if (handler->ops->evict(handler->ops_arg, rbnode, evict_arg,
248 					&stop)) {
249 			__mmu_int_rb_remove(rbnode, &handler->root);
250 			/* move from LRU list to delete list */
251 			list_move(&rbnode->list, &del_list);
252 		}
253 		if (stop)
254 			break;
255 	}
256 	spin_unlock_irqrestore(&handler->lock, flags);
257 
258 	while (!list_empty(&del_list)) {
259 		rbnode = list_first_entry(&del_list, struct mmu_rb_node, list);
260 		list_del(&rbnode->list);
261 		handler->ops->remove(handler->ops_arg, rbnode);
262 	}
263 }
264 
265 /*
266  * It is up to the caller to ensure that this function does not race with the
267  * mmu invalidate notifier which may be calling the users remove callback on
268  * 'node'.
269  */
270 void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler,
271 			struct mmu_rb_node *node)
272 {
273 	unsigned long flags;
274 
275 	/* Validity of handler and node pointers has been checked by caller. */
276 	trace_hfi1_mmu_rb_remove(node->addr, node->len);
277 	spin_lock_irqsave(&handler->lock, flags);
278 	__mmu_int_rb_remove(node, &handler->root);
279 	list_del(&node->list); /* remove from LRU list */
280 	spin_unlock_irqrestore(&handler->lock, flags);
281 
282 	handler->ops->remove(handler->ops_arg, node);
283 }
284 
285 static int mmu_notifier_range_start(struct mmu_notifier *mn,
286 		const struct mmu_notifier_range *range)
287 {
288 	struct mmu_rb_handler *handler =
289 		container_of(mn, struct mmu_rb_handler, mn);
290 	struct rb_root_cached *root = &handler->root;
291 	struct mmu_rb_node *node, *ptr = NULL;
292 	unsigned long flags;
293 	bool added = false;
294 
295 	spin_lock_irqsave(&handler->lock, flags);
296 	for (node = __mmu_int_rb_iter_first(root, range->start, range->end-1);
297 	     node; node = ptr) {
298 		/* Guard against node removal. */
299 		ptr = __mmu_int_rb_iter_next(node, range->start,
300 					     range->end - 1);
301 		trace_hfi1_mmu_mem_invalidate(node->addr, node->len);
302 		if (handler->ops->invalidate(handler->ops_arg, node)) {
303 			__mmu_int_rb_remove(node, root);
304 			/* move from LRU list to delete list */
305 			list_move(&node->list, &handler->del_list);
306 			added = true;
307 		}
308 	}
309 	spin_unlock_irqrestore(&handler->lock, flags);
310 
311 	if (added)
312 		queue_work(handler->wq, &handler->del_work);
313 
314 	return 0;
315 }
316 
317 /*
318  * Call the remove function for the given handler and the list.  This
319  * is expected to be called with a delete list extracted from handler.
320  * The caller should not be holding the handler lock.
321  */
322 static void do_remove(struct mmu_rb_handler *handler,
323 		      struct list_head *del_list)
324 {
325 	struct mmu_rb_node *node;
326 
327 	while (!list_empty(del_list)) {
328 		node = list_first_entry(del_list, struct mmu_rb_node, list);
329 		list_del(&node->list);
330 		handler->ops->remove(handler->ops_arg, node);
331 	}
332 }
333 
334 /*
335  * Work queue function to remove all nodes that have been queued up to
336  * be removed.  The key feature is that mm->mmap_lock is not being held
337  * and the remove callback can sleep while taking it, if needed.
338  */
339 static void handle_remove(struct work_struct *work)
340 {
341 	struct mmu_rb_handler *handler = container_of(work,
342 						struct mmu_rb_handler,
343 						del_work);
344 	struct list_head del_list;
345 	unsigned long flags;
346 
347 	/* remove anything that is queued to get removed */
348 	spin_lock_irqsave(&handler->lock, flags);
349 	list_replace_init(&handler->del_list, &del_list);
350 	spin_unlock_irqrestore(&handler->lock, flags);
351 
352 	do_remove(handler, &del_list);
353 }
354