1/*-
2 * Copyright (c) 2012 NetApp, Inc.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD: releng/11.0/usr.sbin/bhyve/mem.c 269700 2014-08-08 03:49:01Z neel $
27 */
28
29/*
30 * Memory ranges are represented with an RB tree. On insertion, the range
31 * is checked for overlaps. On lookup, the key has the same base and limit
32 * so it can be searched within the range.
33 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: releng/11.0/usr.sbin/bhyve/mem.c 269700 2014-08-08 03:49:01Z neel $");
37
38#include <sys/types.h>
39#include <sys/tree.h>
40#include <sys/errno.h>
41#include <machine/vmm.h>
42#include <machine/vmm_instruction_emul.h>
43
44#include <stdio.h>
45#include <stdlib.h>
46#include <assert.h>
47#include <pthread.h>
48
49#include "mem.h"
50
51struct mmio_rb_range {
52	RB_ENTRY(mmio_rb_range)	mr_link;	/* RB tree links */
53	struct mem_range	mr_param;
54	uint64_t                mr_base;
55	uint64_t                mr_end;
56};
57
58struct mmio_rb_tree;
59RB_PROTOTYPE(mmio_rb_tree, mmio_rb_range, mr_link, mmio_rb_range_compare);
60
61RB_HEAD(mmio_rb_tree, mmio_rb_range) mmio_rb_root, mmio_rb_fallback;
62
63/*
64 * Per-vCPU cache. Since most accesses from a vCPU will be to
65 * consecutive addresses in a range, it makes sense to cache the
66 * result of a lookup.
67 */
68static struct mmio_rb_range	*mmio_hint[VM_MAXCPU];
69
70static pthread_rwlock_t mmio_rwlock;
71
72static int
73mmio_rb_range_compare(struct mmio_rb_range *a, struct mmio_rb_range *b)
74{
75	if (a->mr_end < b->mr_base)
76		return (-1);
77	else if (a->mr_base > b->mr_end)
78		return (1);
79	return (0);
80}
81
82static int
83mmio_rb_lookup(struct mmio_rb_tree *rbt, uint64_t addr,
84    struct mmio_rb_range **entry)
85{
86	struct mmio_rb_range find, *res;
87
88	find.mr_base = find.mr_end = addr;
89
90	res = RB_FIND(mmio_rb_tree, rbt, &find);
91
92	if (res != NULL) {
93		*entry = res;
94		return (0);
95	}
96
97	return (ENOENT);
98}
99
100static int
101mmio_rb_add(struct mmio_rb_tree *rbt, struct mmio_rb_range *new)
102{
103	struct mmio_rb_range *overlap;
104
105	overlap = RB_INSERT(mmio_rb_tree, rbt, new);
106
107	if (overlap != NULL) {
108#ifdef RB_DEBUG
109		printf("overlap detected: new %lx:%lx, tree %lx:%lx\n",
110		       new->mr_base, new->mr_end,
111		       overlap->mr_base, overlap->mr_end);
112#endif
113
114		return (EEXIST);
115	}
116
117	return (0);
118}
119
120#if 0
121static void
122mmio_rb_dump(struct mmio_rb_tree *rbt)
123{
124	struct mmio_rb_range *np;
125
126	pthread_rwlock_rdlock(&mmio_rwlock);
127	RB_FOREACH(np, mmio_rb_tree, rbt) {
128		printf(" %lx:%lx, %s\n", np->mr_base, np->mr_end,
129		       np->mr_param.name);
130	}
131	pthread_rwlock_unlock(&mmio_rwlock);
132}
133#endif
134
135RB_GENERATE(mmio_rb_tree, mmio_rb_range, mr_link, mmio_rb_range_compare);
136
137static int
138mem_read(void *ctx, int vcpu, uint64_t gpa, uint64_t *rval, int size, void *arg)
139{
140	int error;
141	struct mem_range *mr = arg;
142
143	error = (*mr->handler)(ctx, vcpu, MEM_F_READ, gpa, size,
144			       rval, mr->arg1, mr->arg2);
145	return (error);
146}
147
148static int
149mem_write(void *ctx, int vcpu, uint64_t gpa, uint64_t wval, int size, void *arg)
150{
151	int error;
152	struct mem_range *mr = arg;
153
154	error = (*mr->handler)(ctx, vcpu, MEM_F_WRITE, gpa, size,
155			       &wval, mr->arg1, mr->arg2);
156	return (error);
157}
158
159int
160emulate_mem(struct vmctx *ctx, int vcpu, uint64_t paddr, struct vie *vie,
161    struct vm_guest_paging *paging)
162
163{
164	struct mmio_rb_range *entry;
165	int err, immutable;
166
167	pthread_rwlock_rdlock(&mmio_rwlock);
168	/*
169	 * First check the per-vCPU cache
170	 */
171	if (mmio_hint[vcpu] &&
172	    paddr >= mmio_hint[vcpu]->mr_base &&
173	    paddr <= mmio_hint[vcpu]->mr_end) {
174		entry = mmio_hint[vcpu];
175	} else
176		entry = NULL;
177
178	if (entry == NULL) {
179		if (mmio_rb_lookup(&mmio_rb_root, paddr, &entry) == 0) {
180			/* Update the per-vCPU cache */
181			mmio_hint[vcpu] = entry;
182		} else if (mmio_rb_lookup(&mmio_rb_fallback, paddr, &entry)) {
183			pthread_rwlock_unlock(&mmio_rwlock);
184			return (ESRCH);
185		}
186	}
187
188	assert(entry != NULL);
189
190	/*
191	 * An 'immutable' memory range is guaranteed to be never removed
192	 * so there is no need to hold 'mmio_rwlock' while calling the
193	 * handler.
194	 *
195	 * XXX writes to the PCIR_COMMAND register can cause register_mem()
196	 * to be called. If the guest is using PCI extended config space
197	 * to modify the PCIR_COMMAND register then register_mem() can
198	 * deadlock on 'mmio_rwlock'. However by registering the extended
199	 * config space window as 'immutable' the deadlock can be avoided.
200	 */
201	immutable = (entry->mr_param.flags & MEM_F_IMMUTABLE);
202	if (immutable)
203		pthread_rwlock_unlock(&mmio_rwlock);
204
205	err = vmm_emulate_instruction(ctx, vcpu, paddr, vie, paging,
206				      mem_read, mem_write, &entry->mr_param);
207
208	if (!immutable)
209		pthread_rwlock_unlock(&mmio_rwlock);
210
211	return (err);
212}
213
214static int
215register_mem_int(struct mmio_rb_tree *rbt, struct mem_range *memp)
216{
217	struct mmio_rb_range *entry, *mrp;
218	int		err;
219
220	err = 0;
221
222	mrp = malloc(sizeof(struct mmio_rb_range));
223
224	if (mrp != NULL) {
225		mrp->mr_param = *memp;
226		mrp->mr_base = memp->base;
227		mrp->mr_end = memp->base + memp->size - 1;
228		pthread_rwlock_wrlock(&mmio_rwlock);
229		if (mmio_rb_lookup(rbt, memp->base, &entry) != 0)
230			err = mmio_rb_add(rbt, mrp);
231		pthread_rwlock_unlock(&mmio_rwlock);
232		if (err)
233			free(mrp);
234	} else
235		err = ENOMEM;
236
237	return (err);
238}
239
240int
241register_mem(struct mem_range *memp)
242{
243
244	return (register_mem_int(&mmio_rb_root, memp));
245}
246
247int
248register_mem_fallback(struct mem_range *memp)
249{
250
251	return (register_mem_int(&mmio_rb_fallback, memp));
252}
253
254int
255unregister_mem(struct mem_range *memp)
256{
257	struct mem_range *mr;
258	struct mmio_rb_range *entry = NULL;
259	int err, i;
260
261	pthread_rwlock_wrlock(&mmio_rwlock);
262	err = mmio_rb_lookup(&mmio_rb_root, memp->base, &entry);
263	if (err == 0) {
264		mr = &entry->mr_param;
265		assert(mr->name == memp->name);
266		assert(mr->base == memp->base && mr->size == memp->size);
267		assert((mr->flags & MEM_F_IMMUTABLE) == 0);
268		RB_REMOVE(mmio_rb_tree, &mmio_rb_root, entry);
269
270		/* flush Per-vCPU cache */
271		for (i=0; i < VM_MAXCPU; i++) {
272			if (mmio_hint[i] == entry)
273				mmio_hint[i] = NULL;
274		}
275	}
276	pthread_rwlock_unlock(&mmio_rwlock);
277
278	if (entry)
279		free(entry);
280
281	return (err);
282}
283
284void
285init_mem(void)
286{
287
288	RB_INIT(&mmio_rb_root);
289	RB_INIT(&mmio_rb_fallback);
290	pthread_rwlock_init(&mmio_rwlock, NULL);
291}
292