1/*-
2 * Copyright (c) 2013  Peter Grehan <grehan@freebsd.org>
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 THE AUTHOR ``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 THE AUTHOR 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 * $FreeBSD$
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD$");
31
32#include <sys/param.h>
33#include <sys/queue.h>
34#include <sys/errno.h>
35#include <sys/stat.h>
36#include <sys/ioctl.h>
37#include <sys/disk.h>
38
39#include <assert.h>
40#include <fcntl.h>
41#include <stdio.h>
42#include <stdlib.h>
43#include <string.h>
44#include <pthread.h>
45#include <pthread_np.h>
46#include <unistd.h>
47
48#include "bhyverun.h"
49#include "block_if.h"
50
51#define BLOCKIF_SIG	0xb109b109
52
53#define BLOCKIF_MAXREQ	32
54
55enum blockop {
56	BOP_READ,
57	BOP_WRITE,
58	BOP_FLUSH,
59	BOP_CANCEL
60};
61
62enum blockstat {
63	BST_FREE,
64	BST_INUSE
65};
66
67struct blockif_elem {
68	TAILQ_ENTRY(blockif_elem) be_link;
69	struct blockif_req  *be_req;
70	enum blockop	     be_op;
71	enum blockstat	     be_status;
72};
73
74struct blockif_ctxt {
75	int			bc_magic;
76	int			bc_fd;
77	int			bc_rdonly;
78	off_t			bc_size;
79	int			bc_sectsz;
80	pthread_t		bc_btid;
81        pthread_mutex_t		bc_mtx;
82        pthread_cond_t		bc_cond;
83	int			bc_closing;
84
85	/* Request elements and free/inuse queues */
86	TAILQ_HEAD(, blockif_elem) bc_freeq;
87	TAILQ_HEAD(, blockif_elem) bc_inuseq;
88	u_int			bc_req_count;
89	struct blockif_elem	bc_reqs[BLOCKIF_MAXREQ];
90};
91
92static int
93blockif_enqueue(struct blockif_ctxt *bc, struct blockif_req *breq,
94		enum blockop op)
95{
96	struct blockif_elem *be;
97
98	assert(bc->bc_req_count < BLOCKIF_MAXREQ);
99
100	be = TAILQ_FIRST(&bc->bc_freeq);
101	assert(be != NULL);
102	assert(be->be_status == BST_FREE);
103
104	TAILQ_REMOVE(&bc->bc_freeq, be, be_link);
105	be->be_status = BST_INUSE;
106	be->be_req = breq;
107	be->be_op = op;
108	TAILQ_INSERT_TAIL(&bc->bc_inuseq, be, be_link);
109
110	bc->bc_req_count++;
111
112	return (0);
113}
114
115static int
116blockif_dequeue(struct blockif_ctxt *bc, struct blockif_elem *el)
117{
118	struct blockif_elem *be;
119
120	if (bc->bc_req_count == 0)
121		return (ENOENT);
122
123	be = TAILQ_FIRST(&bc->bc_inuseq);
124	assert(be != NULL);
125	assert(be->be_status == BST_INUSE);
126	*el = *be;
127
128	TAILQ_REMOVE(&bc->bc_inuseq, be, be_link);
129	be->be_status = BST_FREE;
130	be->be_req = NULL;
131	TAILQ_INSERT_TAIL(&bc->bc_freeq, be, be_link);
132
133	bc->bc_req_count--;
134
135	return (0);
136}
137
138static void
139blockif_proc(struct blockif_ctxt *bc, struct blockif_elem *be)
140{
141	struct blockif_req *br;
142	int err;
143
144	br = be->be_req;
145	err = 0;
146
147	switch (be->be_op) {
148	case BOP_READ:
149		if (preadv(bc->bc_fd, br->br_iov, br->br_iovcnt,
150			   br->br_offset) < 0)
151			err = errno;
152		break;
153	case BOP_WRITE:
154		if (bc->bc_rdonly)
155			err = EROFS;
156		else if (pwritev(bc->bc_fd, br->br_iov, br->br_iovcnt,
157			     br->br_offset) < 0)
158			err = errno;
159		break;
160	case BOP_FLUSH:
161		break;
162	case BOP_CANCEL:
163		err = EINTR;
164		break;
165	default:
166		err = EINVAL;
167		break;
168	}
169
170	(*br->br_callback)(br, err);
171}
172
173static void *
174blockif_thr(void *arg)
175{
176	struct blockif_ctxt *bc;
177	struct blockif_elem req;
178
179	bc = arg;
180
181	for (;;) {
182		pthread_mutex_lock(&bc->bc_mtx);
183		while (!blockif_dequeue(bc, &req)) {
184			pthread_mutex_unlock(&bc->bc_mtx);
185			blockif_proc(bc, &req);
186			pthread_mutex_lock(&bc->bc_mtx);
187		}
188		pthread_cond_wait(&bc->bc_cond, &bc->bc_mtx);
189		pthread_mutex_unlock(&bc->bc_mtx);
190
191		/*
192		 * Check ctxt status here to see if exit requested
193		 */
194		if (bc->bc_closing)
195			pthread_exit(NULL);
196	}
197
198	/* Not reached */
199	return (NULL);
200}
201
202struct blockif_ctxt *
203blockif_open(const char *optstr, const char *ident)
204{
205	char tname[MAXCOMLEN + 1];
206	char *nopt, *xopts;
207	struct blockif_ctxt *bc;
208	struct stat sbuf;
209	off_t size;
210	int extra, fd, i, sectsz;
211	int nocache, sync, ro;
212
213	nocache = 0;
214	sync = 0;
215	ro = 0;
216
217	/*
218	 * The first element in the optstring is always a pathname.
219	 * Optional elements follow
220	 */
221	nopt = strdup(optstr);
222	for (xopts = strtok(nopt, ",");
223	     xopts != NULL;
224	     xopts = strtok(NULL, ",")) {
225		if (!strcmp(xopts, "nocache"))
226			nocache = 1;
227		else if (!strcmp(xopts, "sync"))
228			sync = 1;
229		else if (!strcmp(xopts, "ro"))
230			ro = 1;
231	}
232
233	extra = 0;
234	if (nocache)
235		extra |= O_DIRECT;
236	if (sync)
237		extra |= O_SYNC;
238
239	fd = open(nopt, (ro ? O_RDONLY : O_RDWR) | extra);
240	if (fd < 0 && !ro) {
241		/* Attempt a r/w fail with a r/o open */
242		fd = open(nopt, O_RDONLY | extra);
243		ro = 1;
244	}
245
246	if (fd < 0) {
247		perror("Could not open backing file");
248		return (NULL);
249	}
250
251        if (fstat(fd, &sbuf) < 0) {
252                perror("Could not stat backing file");
253                close(fd);
254                return (NULL);
255        }
256
257        /*
258	 * Deal with raw devices
259	 */
260        size = sbuf.st_size;
261	sectsz = DEV_BSIZE;
262	if (S_ISCHR(sbuf.st_mode)) {
263		if (ioctl(fd, DIOCGMEDIASIZE, &size) < 0 ||
264		    ioctl(fd, DIOCGSECTORSIZE, &sectsz)) {
265			perror("Could not fetch dev blk/sector size");
266			close(fd);
267			return (NULL);
268		}
269		assert(size != 0);
270		assert(sectsz != 0);
271	}
272
273	bc = calloc(1, sizeof(struct blockif_ctxt));
274	if (bc == NULL) {
275		close(fd);
276		return (NULL);
277	}
278
279	bc->bc_magic = BLOCKIF_SIG;
280	bc->bc_fd = fd;
281	bc->bc_size = size;
282	bc->bc_sectsz = sectsz;
283	pthread_mutex_init(&bc->bc_mtx, NULL);
284	pthread_cond_init(&bc->bc_cond, NULL);
285	TAILQ_INIT(&bc->bc_freeq);
286	TAILQ_INIT(&bc->bc_inuseq);
287	bc->bc_req_count = 0;
288	for (i = 0; i < BLOCKIF_MAXREQ; i++) {
289		bc->bc_reqs[i].be_status = BST_FREE;
290		TAILQ_INSERT_HEAD(&bc->bc_freeq, &bc->bc_reqs[i], be_link);
291	}
292
293	pthread_create(&bc->bc_btid, NULL, blockif_thr, bc);
294
295	snprintf(tname, sizeof(tname), "blk-%s", ident);
296	pthread_set_name_np(bc->bc_btid, tname);
297
298	return (bc);
299}
300
301static int
302blockif_request(struct blockif_ctxt *bc, struct blockif_req *breq,
303		enum blockop op)
304{
305	int err;
306
307	err = 0;
308
309	pthread_mutex_lock(&bc->bc_mtx);
310	if (bc->bc_req_count < BLOCKIF_MAXREQ) {
311		/*
312		 * Enqueue and inform the block i/o thread
313		 * that there is work available
314		 */
315		blockif_enqueue(bc, breq, op);
316		pthread_cond_signal(&bc->bc_cond);
317	} else {
318		/*
319		 * Callers are not allowed to enqueue more than
320		 * the specified blockif queue limit. Return an
321		 * error to indicate that the queue length has been
322		 * exceeded.
323		 */
324		err = E2BIG;
325	}
326	pthread_mutex_unlock(&bc->bc_mtx);
327
328	return (err);
329}
330
331int
332blockif_read(struct blockif_ctxt *bc, struct blockif_req *breq)
333{
334
335	assert(bc->bc_magic == BLOCKIF_SIG);
336	return (blockif_request(bc, breq, BOP_READ));
337}
338
339int
340blockif_write(struct blockif_ctxt *bc, struct blockif_req *breq)
341{
342
343	assert(bc->bc_magic == BLOCKIF_SIG);
344	return (blockif_request(bc, breq, BOP_WRITE));
345}
346
347int
348blockif_flush(struct blockif_ctxt *bc, struct blockif_req *breq)
349{
350
351	assert(bc->bc_magic == BLOCKIF_SIG);
352	return (blockif_request(bc, breq, BOP_FLUSH));
353}
354
355int
356blockif_cancel(struct blockif_ctxt *bc, struct blockif_req *breq)
357{
358
359	assert(bc->bc_magic == BLOCKIF_SIG);
360	return (blockif_request(bc, breq, BOP_CANCEL));
361}
362
363int
364blockif_close(struct blockif_ctxt *bc)
365{
366	void *jval;
367	int err;
368
369	err = 0;
370
371	assert(bc->bc_magic == BLOCKIF_SIG);
372
373	/*
374	 * Stop the block i/o thread
375	 */
376	bc->bc_closing = 1;
377	pthread_cond_signal(&bc->bc_cond);
378	pthread_join(bc->bc_btid, &jval);
379
380	/* XXX Cancel queued i/o's ??? */
381
382	/*
383	 * Release resources
384	 */
385	bc->bc_magic = 0;
386	close(bc->bc_fd);
387	free(bc);
388
389	return (0);
390}
391
392/*
393 * Return virtual C/H/S values for a given block. Use the algorithm
394 * outlined in the VHD specification to calculate values.
395 */
396void
397blockif_chs(struct blockif_ctxt *bc, uint16_t *c, uint8_t *h, uint8_t *s)
398{
399	off_t sectors;		/* total sectors of the block dev */
400	off_t hcyl;		/* cylinders times heads */
401	uint16_t secpt;		/* sectors per track */
402	uint8_t heads;
403
404	assert(bc->bc_magic == BLOCKIF_SIG);
405
406	sectors = bc->bc_size / bc->bc_sectsz;
407
408	/* Clamp the size to the largest possible with CHS */
409	if (sectors > 65535UL*16*255)
410		sectors = 65535UL*16*255;
411
412	if (sectors >= 65536UL*16*63) {
413		secpt = 255;
414		heads = 16;
415		hcyl = sectors / secpt;
416	} else {
417		secpt = 17;
418		hcyl = sectors / secpt;
419		heads = (hcyl + 1023) / 1024;
420
421		if (heads < 4)
422			heads = 4;
423
424		if (hcyl >= (heads * 1024) || heads > 16) {
425			secpt = 31;
426			heads = 16;
427			hcyl = sectors / secpt;
428		}
429		if (hcyl >= (heads * 1024)) {
430			secpt = 63;
431			heads = 16;
432			hcyl = sectors / secpt;
433		}
434	}
435
436	*c = hcyl / heads;
437	*h = heads;
438	*s = secpt;
439}
440
441/*
442 * Accessors
443 */
444off_t
445blockif_size(struct blockif_ctxt *bc)
446{
447
448	assert(bc->bc_magic == BLOCKIF_SIG);
449	return (bc->bc_size);
450}
451
452int
453blockif_sectsz(struct blockif_ctxt *bc)
454{
455
456	assert(bc->bc_magic == BLOCKIF_SIG);
457	return (bc->bc_sectsz);
458}
459
460int
461blockif_queuesz(struct blockif_ctxt *bc)
462{
463
464	assert(bc->bc_magic == BLOCKIF_SIG);
465	return (BLOCKIF_MAXREQ);
466}
467
468int
469blockif_is_ro(struct blockif_ctxt *bc)
470{
471
472	assert(bc->bc_magic == BLOCKIF_SIG);
473	return (bc->bc_rdonly);
474}
475