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 inline void mmu_notifier_range_start(struct mmu_notifier *,
71 					    struct mm_struct *,
72 					    unsigned long, unsigned long);
73 static void mmu_notifier_mem_invalidate(struct mmu_notifier *,
74 					struct mm_struct *,
75 					unsigned long, unsigned long);
76 static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *,
77 					   unsigned long, unsigned long);
78 static void do_remove(struct mmu_rb_handler *handler,
79 		      struct list_head *del_list);
80 static void handle_remove(struct work_struct *work);
81 
82 static const struct mmu_notifier_ops mn_opts = {
83 	.invalidate_range_start = mmu_notifier_range_start,
84 };
85 
86 INTERVAL_TREE_DEFINE(struct mmu_rb_node, node, unsigned long, __last,
87 		     mmu_node_start, mmu_node_last, static, __mmu_int_rb);
88 
89 static unsigned long mmu_node_start(struct mmu_rb_node *node)
90 {
91 	return node->addr & PAGE_MASK;
92 }
93 
94 static unsigned long mmu_node_last(struct mmu_rb_node *node)
95 {
96 	return PAGE_ALIGN(node->addr + node->len) - 1;
97 }
98 
99 int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm,
100 			 struct mmu_rb_ops *ops,
101 			 struct workqueue_struct *wq,
102 			 struct mmu_rb_handler **handler)
103 {
104 	struct mmu_rb_handler *handlr;
105 	int ret;
106 
107 	handlr = kmalloc(sizeof(*handlr), GFP_KERNEL);
108 	if (!handlr)
109 		return -ENOMEM;
110 
111 	handlr->root = RB_ROOT_CACHED;
112 	handlr->ops = ops;
113 	handlr->ops_arg = ops_arg;
114 	INIT_HLIST_NODE(&handlr->mn.hlist);
115 	spin_lock_init(&handlr->lock);
116 	handlr->mn.ops = &mn_opts;
117 	handlr->mm = mm;
118 	INIT_WORK(&handlr->del_work, handle_remove);
119 	INIT_LIST_HEAD(&handlr->del_list);
120 	INIT_LIST_HEAD(&handlr->lru_list);
121 	handlr->wq = wq;
122 
123 	ret = mmu_notifier_register(&handlr->mn, handlr->mm);
124 	if (ret) {
125 		kfree(handlr);
126 		return ret;
127 	}
128 
129 	*handler = handlr;
130 	return 0;
131 }
132 
133 void hfi1_mmu_rb_unregister(struct mmu_rb_handler *handler)
134 {
135 	struct mmu_rb_node *rbnode;
136 	struct rb_node *node;
137 	unsigned long flags;
138 	struct list_head del_list;
139 
140 	/* Unregister first so we don't get any more notifications. */
141 	mmu_notifier_unregister(&handler->mn, handler->mm);
142 
143 	/*
144 	 * Make sure the wq delete handler is finished running.  It will not
145 	 * be triggered once the mmu notifiers are unregistered above.
146 	 */
147 	flush_work(&handler->del_work);
148 
149 	INIT_LIST_HEAD(&del_list);
150 
151 	spin_lock_irqsave(&handler->lock, flags);
152 	while ((node = rb_first_cached(&handler->root))) {
153 		rbnode = rb_entry(node, struct mmu_rb_node, node);
154 		rb_erase_cached(node, &handler->root);
155 		/* move from LRU list to delete list */
156 		list_move(&rbnode->list, &del_list);
157 	}
158 	spin_unlock_irqrestore(&handler->lock, flags);
159 
160 	do_remove(handler, &del_list);
161 
162 	kfree(handler);
163 }
164 
165 int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler,
166 		       struct mmu_rb_node *mnode)
167 {
168 	struct mmu_rb_node *node;
169 	unsigned long flags;
170 	int ret = 0;
171 
172 	trace_hfi1_mmu_rb_insert(mnode->addr, mnode->len);
173 	spin_lock_irqsave(&handler->lock, flags);
174 	node = __mmu_rb_search(handler, mnode->addr, mnode->len);
175 	if (node) {
176 		ret = -EINVAL;
177 		goto unlock;
178 	}
179 	__mmu_int_rb_insert(mnode, &handler->root);
180 	list_add(&mnode->list, &handler->lru_list);
181 
182 	ret = handler->ops->insert(handler->ops_arg, mnode);
183 	if (ret) {
184 		__mmu_int_rb_remove(mnode, &handler->root);
185 		list_del(&mnode->list); /* remove from LRU list */
186 	}
187 unlock:
188 	spin_unlock_irqrestore(&handler->lock, flags);
189 	return ret;
190 }
191 
192 /* Caller must hold handler lock */
193 static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *handler,
194 					   unsigned long addr,
195 					   unsigned long len)
196 {
197 	struct mmu_rb_node *node = NULL;
198 
199 	trace_hfi1_mmu_rb_search(addr, len);
200 	if (!handler->ops->filter) {
201 		node = __mmu_int_rb_iter_first(&handler->root, addr,
202 					       (addr + len) - 1);
203 	} else {
204 		for (node = __mmu_int_rb_iter_first(&handler->root, addr,
205 						    (addr + len) - 1);
206 		     node;
207 		     node = __mmu_int_rb_iter_next(node, addr,
208 						   (addr + len) - 1)) {
209 			if (handler->ops->filter(node, addr, len))
210 				return node;
211 		}
212 	}
213 	return node;
214 }
215 
216 bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler,
217 				     unsigned long addr, unsigned long len,
218 				     struct mmu_rb_node **rb_node)
219 {
220 	struct mmu_rb_node *node;
221 	unsigned long flags;
222 	bool ret = false;
223 
224 	spin_lock_irqsave(&handler->lock, flags);
225 	node = __mmu_rb_search(handler, addr, len);
226 	if (node) {
227 		if (node->addr == addr && node->len == len)
228 			goto unlock;
229 		__mmu_int_rb_remove(node, &handler->root);
230 		list_del(&node->list); /* remove from LRU list */
231 		ret = true;
232 	}
233 unlock:
234 	spin_unlock_irqrestore(&handler->lock, flags);
235 	*rb_node = node;
236 	return ret;
237 }
238 
239 void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg)
240 {
241 	struct mmu_rb_node *rbnode, *ptr;
242 	struct list_head del_list;
243 	unsigned long flags;
244 	bool stop = false;
245 
246 	INIT_LIST_HEAD(&del_list);
247 
248 	spin_lock_irqsave(&handler->lock, flags);
249 	list_for_each_entry_safe_reverse(rbnode, ptr, &handler->lru_list,
250 					 list) {
251 		if (handler->ops->evict(handler->ops_arg, rbnode, evict_arg,
252 					&stop)) {
253 			__mmu_int_rb_remove(rbnode, &handler->root);
254 			/* move from LRU list to delete list */
255 			list_move(&rbnode->list, &del_list);
256 		}
257 		if (stop)
258 			break;
259 	}
260 	spin_unlock_irqrestore(&handler->lock, flags);
261 
262 	while (!list_empty(&del_list)) {
263 		rbnode = list_first_entry(&del_list, struct mmu_rb_node, list);
264 		list_del(&rbnode->list);
265 		handler->ops->remove(handler->ops_arg, rbnode);
266 	}
267 }
268 
269 /*
270  * It is up to the caller to ensure that this function does not race with the
271  * mmu invalidate notifier which may be calling the users remove callback on
272  * 'node'.
273  */
274 void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler,
275 			struct mmu_rb_node *node)
276 {
277 	unsigned long flags;
278 
279 	/* Validity of handler and node pointers has been checked by caller. */
280 	trace_hfi1_mmu_rb_remove(node->addr, node->len);
281 	spin_lock_irqsave(&handler->lock, flags);
282 	__mmu_int_rb_remove(node, &handler->root);
283 	list_del(&node->list); /* remove from LRU list */
284 	spin_unlock_irqrestore(&handler->lock, flags);
285 
286 	handler->ops->remove(handler->ops_arg, node);
287 }
288 
289 static inline void mmu_notifier_range_start(struct mmu_notifier *mn,
290 					    struct mm_struct *mm,
291 					    unsigned long start,
292 					    unsigned long end)
293 {
294 	mmu_notifier_mem_invalidate(mn, mm, start, end);
295 }
296 
297 static void mmu_notifier_mem_invalidate(struct mmu_notifier *mn,
298 					struct mm_struct *mm,
299 					unsigned long start, unsigned long end)
300 {
301 	struct mmu_rb_handler *handler =
302 		container_of(mn, struct mmu_rb_handler, mn);
303 	struct rb_root_cached *root = &handler->root;
304 	struct mmu_rb_node *node, *ptr = NULL;
305 	unsigned long flags;
306 	bool added = false;
307 
308 	spin_lock_irqsave(&handler->lock, flags);
309 	for (node = __mmu_int_rb_iter_first(root, start, end - 1);
310 	     node; node = ptr) {
311 		/* Guard against node removal. */
312 		ptr = __mmu_int_rb_iter_next(node, start, end - 1);
313 		trace_hfi1_mmu_mem_invalidate(node->addr, node->len);
314 		if (handler->ops->invalidate(handler->ops_arg, node)) {
315 			__mmu_int_rb_remove(node, root);
316 			/* move from LRU list to delete list */
317 			list_move(&node->list, &handler->del_list);
318 			added = true;
319 		}
320 	}
321 	spin_unlock_irqrestore(&handler->lock, flags);
322 
323 	if (added)
324 		queue_work(handler->wq, &handler->del_work);
325 }
326 
327 /*
328  * Call the remove function for the given handler and the list.  This
329  * is expected to be called with a delete list extracted from handler.
330  * The caller should not be holding the handler lock.
331  */
332 static void do_remove(struct mmu_rb_handler *handler,
333 		      struct list_head *del_list)
334 {
335 	struct mmu_rb_node *node;
336 
337 	while (!list_empty(del_list)) {
338 		node = list_first_entry(del_list, struct mmu_rb_node, list);
339 		list_del(&node->list);
340 		handler->ops->remove(handler->ops_arg, node);
341 	}
342 }
343 
344 /*
345  * Work queue function to remove all nodes that have been queued up to
346  * be removed.  The key feature is that mm->mmap_sem is not being held
347  * and the remove callback can sleep while taking it, if needed.
348  */
349 static void handle_remove(struct work_struct *work)
350 {
351 	struct mmu_rb_handler *handler = container_of(work,
352 						struct mmu_rb_handler,
353 						del_work);
354 	struct list_head del_list;
355 	unsigned long flags;
356 
357 	/* remove anything that is queued to get removed */
358 	spin_lock_irqsave(&handler->lock, flags);
359 	list_replace_init(&handler->del_list, &del_list);
360 	spin_unlock_irqrestore(&handler->lock, flags);
361 
362 	do_remove(handler, &del_list);
363 }
364