Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
[qemu.git] / block / raw-win32.c
1 /*
2 * Block driver for RAW files (win32)
3 *
4 * Copyright (c) 2006 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24 #include "qemu-common.h"
25 #include "qemu/timer.h"
26 #include "block/block_int.h"
27 #include "qemu/module.h"
28 #include "raw-aio.h"
29 #include "trace.h"
30 #include "block/thread-pool.h"
31 #include "qemu/iov.h"
32 #include <windows.h>
33 #include <winioctl.h>
34
35 #define FTYPE_FILE 0
36 #define FTYPE_CD 1
37 #define FTYPE_HARDDISK 2
38
39 static QEMUWin32AIOState *aio;
40
41 typedef struct RawWin32AIOData {
42 BlockDriverState *bs;
43 HANDLE hfile;
44 struct iovec *aio_iov;
45 int aio_niov;
46 size_t aio_nbytes;
47 off64_t aio_offset;
48 int aio_type;
49 } RawWin32AIOData;
50
51 typedef struct BDRVRawState {
52 HANDLE hfile;
53 int type;
54 char drive_path[16]; /* format: "d:\" */
55 QEMUWin32AIOState *aio;
56 } BDRVRawState;
57
58 /*
59 * Read/writes the data to/from a given linear buffer.
60 *
61 * Returns the number of bytes handles or -errno in case of an error. Short
62 * reads are only returned if the end of the file is reached.
63 */
64 static size_t handle_aiocb_rw(RawWin32AIOData *aiocb)
65 {
66 size_t offset = 0;
67 int i;
68
69 for (i = 0; i < aiocb->aio_niov; i++) {
70 OVERLAPPED ov;
71 DWORD ret, ret_count, len;
72
73 memset(&ov, 0, sizeof(ov));
74 ov.Offset = (aiocb->aio_offset + offset);
75 ov.OffsetHigh = (aiocb->aio_offset + offset) >> 32;
76 len = aiocb->aio_iov[i].iov_len;
77 if (aiocb->aio_type & QEMU_AIO_WRITE) {
78 ret = WriteFile(aiocb->hfile, aiocb->aio_iov[i].iov_base,
79 len, &ret_count, &ov);
80 } else {
81 ret = ReadFile(aiocb->hfile, aiocb->aio_iov[i].iov_base,
82 len, &ret_count, &ov);
83 }
84 if (!ret) {
85 ret_count = 0;
86 }
87 if (ret_count != len) {
88 offset += ret_count;
89 break;
90 }
91 offset += len;
92 }
93
94 return offset;
95 }
96
97 static int aio_worker(void *arg)
98 {
99 RawWin32AIOData *aiocb = arg;
100 ssize_t ret = 0;
101 size_t count;
102
103 switch (aiocb->aio_type & QEMU_AIO_TYPE_MASK) {
104 case QEMU_AIO_READ:
105 count = handle_aiocb_rw(aiocb);
106 if (count < aiocb->aio_nbytes && aiocb->bs->growable) {
107 /* A short read means that we have reached EOF. Pad the buffer
108 * with zeros for bytes after EOF. */
109 iov_memset(aiocb->aio_iov, aiocb->aio_niov, count,
110 0, aiocb->aio_nbytes - count);
111
112 count = aiocb->aio_nbytes;
113 }
114 if (count == aiocb->aio_nbytes) {
115 ret = 0;
116 } else {
117 ret = -EINVAL;
118 }
119 break;
120 case QEMU_AIO_WRITE:
121 count = handle_aiocb_rw(aiocb);
122 if (count == aiocb->aio_nbytes) {
123 count = 0;
124 } else {
125 count = -EINVAL;
126 }
127 break;
128 case QEMU_AIO_FLUSH:
129 if (!FlushFileBuffers(aiocb->hfile)) {
130 return -EIO;
131 }
132 break;
133 default:
134 fprintf(stderr, "invalid aio request (0x%x)\n", aiocb->aio_type);
135 ret = -EINVAL;
136 break;
137 }
138
139 g_slice_free(RawWin32AIOData, aiocb);
140 return ret;
141 }
142
143 static BlockDriverAIOCB *paio_submit(BlockDriverState *bs, HANDLE hfile,
144 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
145 BlockDriverCompletionFunc *cb, void *opaque, int type)
146 {
147 RawWin32AIOData *acb = g_slice_new(RawWin32AIOData);
148 ThreadPool *pool;
149
150 acb->bs = bs;
151 acb->hfile = hfile;
152 acb->aio_type = type;
153
154 if (qiov) {
155 acb->aio_iov = qiov->iov;
156 acb->aio_niov = qiov->niov;
157 }
158 acb->aio_nbytes = nb_sectors * 512;
159 acb->aio_offset = sector_num * 512;
160
161 trace_paio_submit(acb, opaque, sector_num, nb_sectors, type);
162 pool = aio_get_thread_pool(bdrv_get_aio_context(bs));
163 return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque);
164 }
165
166 int qemu_ftruncate64(int fd, int64_t length)
167 {
168 LARGE_INTEGER li;
169 DWORD dw;
170 LONG high;
171 HANDLE h;
172 BOOL res;
173
174 if ((GetVersion() & 0x80000000UL) && (length >> 32) != 0)
175 return -1;
176
177 h = (HANDLE)_get_osfhandle(fd);
178
179 /* get current position, ftruncate do not change position */
180 li.HighPart = 0;
181 li.LowPart = SetFilePointer (h, 0, &li.HighPart, FILE_CURRENT);
182 if (li.LowPart == INVALID_SET_FILE_POINTER && GetLastError() != NO_ERROR) {
183 return -1;
184 }
185
186 high = length >> 32;
187 dw = SetFilePointer(h, (DWORD) length, &high, FILE_BEGIN);
188 if (dw == INVALID_SET_FILE_POINTER && GetLastError() != NO_ERROR) {
189 return -1;
190 }
191 res = SetEndOfFile(h);
192
193 /* back to old position */
194 SetFilePointer(h, li.LowPart, &li.HighPart, FILE_BEGIN);
195 return res ? 0 : -1;
196 }
197
198 static int set_sparse(int fd)
199 {
200 DWORD returned;
201 return (int) DeviceIoControl((HANDLE)_get_osfhandle(fd), FSCTL_SET_SPARSE,
202 NULL, 0, NULL, 0, &returned, NULL);
203 }
204
205 static void raw_probe_alignment(BlockDriverState *bs)
206 {
207 BDRVRawState *s = bs->opaque;
208 DWORD sectorsPerCluster, freeClusters, totalClusters, count;
209 DISK_GEOMETRY_EX dg;
210 BOOL status;
211
212 if (s->type == FTYPE_CD) {
213 bs->request_alignment = 2048;
214 return;
215 }
216 if (s->type == FTYPE_HARDDISK) {
217 status = DeviceIoControl(s->hfile, IOCTL_DISK_GET_DRIVE_GEOMETRY_EX,
218 NULL, 0, &dg, sizeof(dg), &count, NULL);
219 if (status != 0) {
220 bs->request_alignment = dg.Geometry.BytesPerSector;
221 return;
222 }
223 /* try GetDiskFreeSpace too */
224 }
225
226 if (s->drive_path[0]) {
227 GetDiskFreeSpace(s->drive_path, &sectorsPerCluster,
228 &dg.Geometry.BytesPerSector,
229 &freeClusters, &totalClusters);
230 bs->request_alignment = dg.Geometry.BytesPerSector;
231 }
232 }
233
234 static void raw_parse_flags(int flags, int *access_flags, DWORD *overlapped)
235 {
236 assert(access_flags != NULL);
237 assert(overlapped != NULL);
238
239 if (flags & BDRV_O_RDWR) {
240 *access_flags = GENERIC_READ | GENERIC_WRITE;
241 } else {
242 *access_flags = GENERIC_READ;
243 }
244
245 *overlapped = FILE_ATTRIBUTE_NORMAL;
246 if (flags & BDRV_O_NATIVE_AIO) {
247 *overlapped |= FILE_FLAG_OVERLAPPED;
248 }
249 if (flags & BDRV_O_NOCACHE) {
250 *overlapped |= FILE_FLAG_NO_BUFFERING;
251 }
252 }
253
254 static void raw_parse_filename(const char *filename, QDict *options,
255 Error **errp)
256 {
257 /* The filename does not have to be prefixed by the protocol name, since
258 * "file" is the default protocol; therefore, the return value of this
259 * function call can be ignored. */
260 strstart(filename, "file:", &filename);
261
262 qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename)));
263 }
264
265 static QemuOptsList raw_runtime_opts = {
266 .name = "raw",
267 .head = QTAILQ_HEAD_INITIALIZER(raw_runtime_opts.head),
268 .desc = {
269 {
270 .name = "filename",
271 .type = QEMU_OPT_STRING,
272 .help = "File name of the image",
273 },
274 { /* end of list */ }
275 },
276 };
277
278 static int raw_open(BlockDriverState *bs, QDict *options, int flags,
279 Error **errp)
280 {
281 BDRVRawState *s = bs->opaque;
282 int access_flags;
283 DWORD overlapped;
284 QemuOpts *opts;
285 Error *local_err = NULL;
286 const char *filename;
287 int ret;
288
289 s->type = FTYPE_FILE;
290
291 opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort);
292 qemu_opts_absorb_qdict(opts, options, &local_err);
293 if (local_err) {
294 error_propagate(errp, local_err);
295 ret = -EINVAL;
296 goto fail;
297 }
298
299 filename = qemu_opt_get(opts, "filename");
300
301 raw_parse_flags(flags, &access_flags, &overlapped);
302
303 if ((flags & BDRV_O_NATIVE_AIO) && aio == NULL) {
304 aio = win32_aio_init();
305 if (aio == NULL) {
306 error_setg(errp, "Could not initialize AIO");
307 ret = -EINVAL;
308 goto fail;
309 }
310 }
311
312 if (filename[0] && filename[1] == ':') {
313 snprintf(s->drive_path, sizeof(s->drive_path), "%c:\\", filename[0]);
314 } else if (filename[0] == '\\' && filename[1] == '\\') {
315 s->drive_path[0] = 0;
316 } else {
317 /* Relative path. */
318 char buf[MAX_PATH];
319 GetCurrentDirectory(MAX_PATH, buf);
320 snprintf(s->drive_path, sizeof(s->drive_path), "%c:\\", buf[0]);
321 }
322
323 s->hfile = CreateFile(filename, access_flags,
324 FILE_SHARE_READ, NULL,
325 OPEN_EXISTING, overlapped, NULL);
326 if (s->hfile == INVALID_HANDLE_VALUE) {
327 int err = GetLastError();
328
329 if (err == ERROR_ACCESS_DENIED) {
330 ret = -EACCES;
331 } else {
332 ret = -EINVAL;
333 }
334 goto fail;
335 }
336
337 if (flags & BDRV_O_NATIVE_AIO) {
338 ret = win32_aio_attach(aio, s->hfile);
339 if (ret < 0) {
340 CloseHandle(s->hfile);
341 error_setg_errno(errp, -ret, "Could not enable AIO");
342 goto fail;
343 }
344 s->aio = aio;
345 }
346
347 raw_probe_alignment(bs);
348 ret = 0;
349 fail:
350 qemu_opts_del(opts);
351 return ret;
352 }
353
354 static BlockDriverAIOCB *raw_aio_readv(BlockDriverState *bs,
355 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
356 BlockDriverCompletionFunc *cb, void *opaque)
357 {
358 BDRVRawState *s = bs->opaque;
359 if (s->aio) {
360 return win32_aio_submit(bs, s->aio, s->hfile, sector_num, qiov,
361 nb_sectors, cb, opaque, QEMU_AIO_READ);
362 } else {
363 return paio_submit(bs, s->hfile, sector_num, qiov, nb_sectors,
364 cb, opaque, QEMU_AIO_READ);
365 }
366 }
367
368 static BlockDriverAIOCB *raw_aio_writev(BlockDriverState *bs,
369 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
370 BlockDriverCompletionFunc *cb, void *opaque)
371 {
372 BDRVRawState *s = bs->opaque;
373 if (s->aio) {
374 return win32_aio_submit(bs, s->aio, s->hfile, sector_num, qiov,
375 nb_sectors, cb, opaque, QEMU_AIO_WRITE);
376 } else {
377 return paio_submit(bs, s->hfile, sector_num, qiov, nb_sectors,
378 cb, opaque, QEMU_AIO_WRITE);
379 }
380 }
381
382 static BlockDriverAIOCB *raw_aio_flush(BlockDriverState *bs,
383 BlockDriverCompletionFunc *cb, void *opaque)
384 {
385 BDRVRawState *s = bs->opaque;
386 return paio_submit(bs, s->hfile, 0, NULL, 0, cb, opaque, QEMU_AIO_FLUSH);
387 }
388
389 static void raw_close(BlockDriverState *bs)
390 {
391 BDRVRawState *s = bs->opaque;
392 CloseHandle(s->hfile);
393 }
394
395 static int raw_truncate(BlockDriverState *bs, int64_t offset)
396 {
397 BDRVRawState *s = bs->opaque;
398 LONG low, high;
399 DWORD dwPtrLow;
400
401 low = offset;
402 high = offset >> 32;
403
404 /*
405 * An error has occurred if the return value is INVALID_SET_FILE_POINTER
406 * and GetLastError doesn't return NO_ERROR.
407 */
408 dwPtrLow = SetFilePointer(s->hfile, low, &high, FILE_BEGIN);
409 if (dwPtrLow == INVALID_SET_FILE_POINTER && GetLastError() != NO_ERROR) {
410 fprintf(stderr, "SetFilePointer error: %lu\n", GetLastError());
411 return -EIO;
412 }
413 if (SetEndOfFile(s->hfile) == 0) {
414 fprintf(stderr, "SetEndOfFile error: %lu\n", GetLastError());
415 return -EIO;
416 }
417 return 0;
418 }
419
420 static int64_t raw_getlength(BlockDriverState *bs)
421 {
422 BDRVRawState *s = bs->opaque;
423 LARGE_INTEGER l;
424 ULARGE_INTEGER available, total, total_free;
425 DISK_GEOMETRY_EX dg;
426 DWORD count;
427 BOOL status;
428
429 switch(s->type) {
430 case FTYPE_FILE:
431 l.LowPart = GetFileSize(s->hfile, (PDWORD)&l.HighPart);
432 if (l.LowPart == 0xffffffffUL && GetLastError() != NO_ERROR)
433 return -EIO;
434 break;
435 case FTYPE_CD:
436 if (!GetDiskFreeSpaceEx(s->drive_path, &available, &total, &total_free))
437 return -EIO;
438 l.QuadPart = total.QuadPart;
439 break;
440 case FTYPE_HARDDISK:
441 status = DeviceIoControl(s->hfile, IOCTL_DISK_GET_DRIVE_GEOMETRY_EX,
442 NULL, 0, &dg, sizeof(dg), &count, NULL);
443 if (status != 0) {
444 l = dg.DiskSize;
445 }
446 break;
447 default:
448 return -EIO;
449 }
450 return l.QuadPart;
451 }
452
453 static int64_t raw_get_allocated_file_size(BlockDriverState *bs)
454 {
455 typedef DWORD (WINAPI * get_compressed_t)(const char *filename,
456 DWORD * high);
457 get_compressed_t get_compressed;
458 struct _stati64 st;
459 const char *filename = bs->filename;
460 /* WinNT support GetCompressedFileSize to determine allocate size */
461 get_compressed =
462 (get_compressed_t) GetProcAddress(GetModuleHandle("kernel32"),
463 "GetCompressedFileSizeA");
464 if (get_compressed) {
465 DWORD high, low;
466 low = get_compressed(filename, &high);
467 if (low != 0xFFFFFFFFlu || GetLastError() == NO_ERROR) {
468 return (((int64_t) high) << 32) + low;
469 }
470 }
471
472 if (_stati64(filename, &st) < 0) {
473 return -1;
474 }
475 return st.st_size;
476 }
477
478 static int raw_create(const char *filename, QEMUOptionParameter *options,
479 Error **errp)
480 {
481 int fd;
482 int64_t total_size = 0;
483
484 strstart(filename, "file:", &filename);
485
486 /* Read out options */
487 while (options && options->name) {
488 if (!strcmp(options->name, BLOCK_OPT_SIZE)) {
489 total_size = options->value.n / 512;
490 }
491 options++;
492 }
493
494 fd = qemu_open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY,
495 0644);
496 if (fd < 0) {
497 error_setg_errno(errp, errno, "Could not create file");
498 return -EIO;
499 }
500 set_sparse(fd);
501 ftruncate(fd, total_size * 512);
502 qemu_close(fd);
503 return 0;
504 }
505
506 static QEMUOptionParameter raw_create_options[] = {
507 {
508 .name = BLOCK_OPT_SIZE,
509 .type = OPT_SIZE,
510 .help = "Virtual disk size"
511 },
512 { NULL }
513 };
514
515 static BlockDriver bdrv_file = {
516 .format_name = "file",
517 .protocol_name = "file",
518 .instance_size = sizeof(BDRVRawState),
519 .bdrv_needs_filename = true,
520 .bdrv_parse_filename = raw_parse_filename,
521 .bdrv_file_open = raw_open,
522 .bdrv_close = raw_close,
523 .bdrv_create = raw_create,
524 .bdrv_has_zero_init = bdrv_has_zero_init_1,
525
526 .bdrv_aio_readv = raw_aio_readv,
527 .bdrv_aio_writev = raw_aio_writev,
528 .bdrv_aio_flush = raw_aio_flush,
529
530 .bdrv_truncate = raw_truncate,
531 .bdrv_getlength = raw_getlength,
532 .bdrv_get_allocated_file_size
533 = raw_get_allocated_file_size,
534
535 .create_options = raw_create_options,
536 };
537
538 /***********************************************/
539 /* host device */
540
541 static int find_cdrom(char *cdrom_name, int cdrom_name_size)
542 {
543 char drives[256], *pdrv = drives;
544 UINT type;
545
546 memset(drives, 0, sizeof(drives));
547 GetLogicalDriveStrings(sizeof(drives), drives);
548 while(pdrv[0] != '\0') {
549 type = GetDriveType(pdrv);
550 switch(type) {
551 case DRIVE_CDROM:
552 snprintf(cdrom_name, cdrom_name_size, "\\\\.\\%c:", pdrv[0]);
553 return 0;
554 break;
555 }
556 pdrv += lstrlen(pdrv) + 1;
557 }
558 return -1;
559 }
560
561 static int find_device_type(BlockDriverState *bs, const char *filename)
562 {
563 BDRVRawState *s = bs->opaque;
564 UINT type;
565 const char *p;
566
567 if (strstart(filename, "\\\\.\\", &p) ||
568 strstart(filename, "//./", &p)) {
569 if (stristart(p, "PhysicalDrive", NULL))
570 return FTYPE_HARDDISK;
571 snprintf(s->drive_path, sizeof(s->drive_path), "%c:\\", p[0]);
572 type = GetDriveType(s->drive_path);
573 switch (type) {
574 case DRIVE_REMOVABLE:
575 case DRIVE_FIXED:
576 return FTYPE_HARDDISK;
577 case DRIVE_CDROM:
578 return FTYPE_CD;
579 default:
580 return FTYPE_FILE;
581 }
582 } else {
583 return FTYPE_FILE;
584 }
585 }
586
587 static int hdev_probe_device(const char *filename)
588 {
589 if (strstart(filename, "/dev/cdrom", NULL))
590 return 100;
591 if (is_windows_drive(filename))
592 return 100;
593 return 0;
594 }
595
596 static void hdev_parse_filename(const char *filename, QDict *options,
597 Error **errp)
598 {
599 /* The prefix is optional, just as for "file". */
600 strstart(filename, "host_device:", &filename);
601
602 qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename)));
603 }
604
605 static int hdev_open(BlockDriverState *bs, QDict *options, int flags,
606 Error **errp)
607 {
608 BDRVRawState *s = bs->opaque;
609 int access_flags, create_flags;
610 int ret = 0;
611 DWORD overlapped;
612 char device_name[64];
613
614 Error *local_err = NULL;
615 const char *filename;
616
617 QemuOpts *opts = qemu_opts_create(&raw_runtime_opts, NULL, 0,
618 &error_abort);
619 qemu_opts_absorb_qdict(opts, options, &local_err);
620 if (local_err) {
621 error_propagate(errp, local_err);
622 ret = -EINVAL;
623 goto done;
624 }
625
626 filename = qemu_opt_get(opts, "filename");
627
628 if (strstart(filename, "/dev/cdrom", NULL)) {
629 if (find_cdrom(device_name, sizeof(device_name)) < 0) {
630 error_setg(errp, "Could not open CD-ROM drive");
631 ret = -ENOENT;
632 goto done;
633 }
634 filename = device_name;
635 } else {
636 /* transform drive letters into device name */
637 if (((filename[0] >= 'a' && filename[0] <= 'z') ||
638 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
639 filename[1] == ':' && filename[2] == '\0') {
640 snprintf(device_name, sizeof(device_name), "\\\\.\\%c:", filename[0]);
641 filename = device_name;
642 }
643 }
644 s->type = find_device_type(bs, filename);
645
646 raw_parse_flags(flags, &access_flags, &overlapped);
647
648 create_flags = OPEN_EXISTING;
649
650 s->hfile = CreateFile(filename, access_flags,
651 FILE_SHARE_READ, NULL,
652 create_flags, overlapped, NULL);
653 if (s->hfile == INVALID_HANDLE_VALUE) {
654 int err = GetLastError();
655
656 if (err == ERROR_ACCESS_DENIED) {
657 ret = -EACCES;
658 } else {
659 ret = -EINVAL;
660 }
661 error_setg_errno(errp, -ret, "Could not open device");
662 goto done;
663 }
664
665 done:
666 qemu_opts_del(opts);
667 return ret;
668 }
669
670 static BlockDriver bdrv_host_device = {
671 .format_name = "host_device",
672 .protocol_name = "host_device",
673 .instance_size = sizeof(BDRVRawState),
674 .bdrv_needs_filename = true,
675 .bdrv_parse_filename = hdev_parse_filename,
676 .bdrv_probe_device = hdev_probe_device,
677 .bdrv_file_open = hdev_open,
678 .bdrv_close = raw_close,
679
680 .bdrv_aio_readv = raw_aio_readv,
681 .bdrv_aio_writev = raw_aio_writev,
682 .bdrv_aio_flush = raw_aio_flush,
683
684 .bdrv_getlength = raw_getlength,
685 .has_variable_length = true,
686
687 .bdrv_get_allocated_file_size
688 = raw_get_allocated_file_size,
689 };
690
691 static void bdrv_file_init(void)
692 {
693 bdrv_register(&bdrv_file);
694 bdrv_register(&bdrv_host_device);
695 }
696
697 block_init(bdrv_file_init);