jcs.org
qemu with hax to log dma reads & writes
jcs.org/2018/11/12/vfio
1/*
2 * Block replication tests
3 *
4 * Copyright (c) 2016 FUJITSU LIMITED
5 * Author: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
6 *
7 * This work is licensed under the terms of the GNU GPL, version 2 or
8 * later. See the COPYING file in the top-level directory.
9 */
10
11#include "qemu/osdep.h"
12
13#include "qapi/error.h"
14#include "qapi/qmp/qdict.h"
15#include "qemu/option.h"
16#include "qemu/main-loop.h"
17#include "replication.h"
18#include "block/block_int.h"
19#include "block/qdict.h"
20#include "sysemu/block-backend.h"
21
22#define IMG_SIZE (64 * 1024 * 1024)
23
24/* primary */
25#define P_ID "primary-id"
26static char p_local_disk[] = "/tmp/p_local_disk.XXXXXX";
27
28/* secondary */
29#define S_ID "secondary-id"
30#define S_LOCAL_DISK_ID "secondary-local-disk-id"
31static char s_local_disk[] = "/tmp/s_local_disk.XXXXXX";
32static char s_active_disk[] = "/tmp/s_active_disk.XXXXXX";
33static char s_hidden_disk[] = "/tmp/s_hidden_disk.XXXXXX";
34
35/* FIXME: steal from blockdev.c */
36QemuOptsList qemu_drive_opts = {
37 .name = "drive",
38 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
39 .desc = {
40 { /* end of list */ }
41 },
42};
43
44#define NOT_DONE 0x7fffffff
45
46static void blk_rw_done(void *opaque, int ret)
47{
48 *(int *)opaque = ret;
49}
50
51static void test_blk_read(BlockBackend *blk, long pattern,
52 int64_t pattern_offset, int64_t pattern_count,
53 int64_t offset, int64_t count,
54 bool expect_failed)
55{
56 void *pattern_buf = NULL;
57 QEMUIOVector qiov;
58 void *cmp_buf = NULL;
59 int async_ret = NOT_DONE;
60
61 if (pattern) {
62 cmp_buf = g_malloc(pattern_count);
63 memset(cmp_buf, pattern, pattern_count);
64 }
65
66 pattern_buf = g_malloc(count);
67 if (pattern) {
68 memset(pattern_buf, pattern, count);
69 } else {
70 memset(pattern_buf, 0x00, count);
71 }
72
73 qemu_iovec_init(&qiov, 1);
74 qemu_iovec_add(&qiov, pattern_buf, count);
75
76 blk_aio_preadv(blk, offset, &qiov, 0, blk_rw_done, &async_ret);
77 while (async_ret == NOT_DONE) {
78 main_loop_wait(false);
79 }
80
81 if (expect_failed) {
82 g_assert(async_ret != 0);
83 } else {
84 g_assert(async_ret == 0);
85 if (pattern) {
86 g_assert(memcmp(pattern_buf + pattern_offset,
87 cmp_buf, pattern_count) <= 0);
88 }
89 }
90
91 g_free(pattern_buf);
92 g_free(cmp_buf);
93 qemu_iovec_destroy(&qiov);
94}
95
96static void test_blk_write(BlockBackend *blk, long pattern, int64_t offset,
97 int64_t count, bool expect_failed)
98{
99 void *pattern_buf = NULL;
100 QEMUIOVector qiov;
101 int async_ret = NOT_DONE;
102
103 pattern_buf = g_malloc(count);
104 if (pattern) {
105 memset(pattern_buf, pattern, count);
106 } else {
107 memset(pattern_buf, 0x00, count);
108 }
109
110 qemu_iovec_init(&qiov, 1);
111 qemu_iovec_add(&qiov, pattern_buf, count);
112
113 blk_aio_pwritev(blk, offset, &qiov, 0, blk_rw_done, &async_ret);
114 while (async_ret == NOT_DONE) {
115 main_loop_wait(false);
116 }
117
118 if (expect_failed) {
119 g_assert(async_ret != 0);
120 } else {
121 g_assert(async_ret == 0);
122 }
123
124 g_free(pattern_buf);
125 qemu_iovec_destroy(&qiov);
126}
127
128/*
129 * Create a uniquely-named empty temporary file.
130 */
131static void make_temp(char *template)
132{
133 int fd;
134
135 fd = mkstemp(template);
136 g_assert(fd >= 0);
137 close(fd);
138}
139
140static void prepare_imgs(void)
141{
142 make_temp(p_local_disk);
143 make_temp(s_local_disk);
144 make_temp(s_active_disk);
145 make_temp(s_hidden_disk);
146
147 /* Primary */
148 bdrv_img_create(p_local_disk, "qcow2", NULL, NULL, NULL, IMG_SIZE,
149 BDRV_O_RDWR, true, &error_abort);
150
151 /* Secondary */
152 bdrv_img_create(s_local_disk, "qcow2", NULL, NULL, NULL, IMG_SIZE,
153 BDRV_O_RDWR, true, &error_abort);
154 bdrv_img_create(s_active_disk, "qcow2", NULL, NULL, NULL, IMG_SIZE,
155 BDRV_O_RDWR, true, &error_abort);
156 bdrv_img_create(s_hidden_disk, "qcow2", NULL, NULL, NULL, IMG_SIZE,
157 BDRV_O_RDWR, true, &error_abort);
158}
159
160static void cleanup_imgs(void)
161{
162 /* Primary */
163 unlink(p_local_disk);
164
165 /* Secondary */
166 unlink(s_local_disk);
167 unlink(s_active_disk);
168 unlink(s_hidden_disk);
169}
170
171static BlockBackend *start_primary(void)
172{
173 BlockBackend *blk;
174 QemuOpts *opts;
175 QDict *qdict;
176 char *cmdline;
177
178 cmdline = g_strdup_printf("driver=replication,mode=primary,node-name=xxx,"
179 "file.driver=qcow2,file.file.filename=%s,"
180 "file.file.locking=off"
181 , p_local_disk);
182 opts = qemu_opts_parse_noisily(&qemu_drive_opts, cmdline, false);
183 g_free(cmdline);
184
185 qdict = qemu_opts_to_qdict(opts, NULL);
186 qdict_set_default_str(qdict, BDRV_OPT_CACHE_DIRECT, "off");
187 qdict_set_default_str(qdict, BDRV_OPT_CACHE_NO_FLUSH, "off");
188
189 blk = blk_new_open(NULL, NULL, qdict, BDRV_O_RDWR, &error_abort);
190 g_assert(blk);
191
192 monitor_add_blk(blk, P_ID, &error_abort);
193
194 qemu_opts_del(opts);
195
196 return blk;
197}
198
199static void teardown_primary(void)
200{
201 BlockBackend *blk;
202 AioContext *ctx;
203
204 /* remove P_ID */
205 blk = blk_by_name(P_ID);
206 assert(blk);
207
208 ctx = blk_get_aio_context(blk);
209 aio_context_acquire(ctx);
210 monitor_remove_blk(blk);
211 blk_unref(blk);
212 aio_context_release(ctx);
213}
214
215static void test_primary_read(void)
216{
217 BlockBackend *blk;
218
219 blk = start_primary();
220
221 /* read from 0 to IMG_SIZE */
222 test_blk_read(blk, 0, 0, IMG_SIZE, 0, IMG_SIZE, true);
223
224 teardown_primary();
225}
226
227static void test_primary_write(void)
228{
229 BlockBackend *blk;
230
231 blk = start_primary();
232
233 /* write from 0 to IMG_SIZE */
234 test_blk_write(blk, 0, 0, IMG_SIZE, true);
235
236 teardown_primary();
237}
238
239static void test_primary_start(void)
240{
241 BlockBackend *blk = NULL;
242
243 blk = start_primary();
244
245 replication_start_all(REPLICATION_MODE_PRIMARY, &error_abort);
246
247 /* read from 0 to IMG_SIZE */
248 test_blk_read(blk, 0, 0, IMG_SIZE, 0, IMG_SIZE, true);
249
250 /* write 0x22 from 0 to IMG_SIZE */
251 test_blk_write(blk, 0x22, 0, IMG_SIZE, false);
252
253 teardown_primary();
254}
255
256static void test_primary_stop(void)
257{
258 bool failover = true;
259
260 start_primary();
261
262 replication_start_all(REPLICATION_MODE_PRIMARY, &error_abort);
263
264 replication_stop_all(failover, &error_abort);
265
266 teardown_primary();
267}
268
269static void test_primary_do_checkpoint(void)
270{
271 start_primary();
272
273 replication_start_all(REPLICATION_MODE_PRIMARY, &error_abort);
274
275 replication_do_checkpoint_all(&error_abort);
276
277 teardown_primary();
278}
279
280static void test_primary_get_error_all(void)
281{
282 start_primary();
283
284 replication_start_all(REPLICATION_MODE_PRIMARY, &error_abort);
285
286 replication_get_error_all(&error_abort);
287
288 teardown_primary();
289}
290
291static BlockBackend *start_secondary(void)
292{
293 QemuOpts *opts;
294 QDict *qdict;
295 BlockBackend *blk;
296 char *cmdline;
297
298 /* add s_local_disk and forge S_LOCAL_DISK_ID */
299 cmdline = g_strdup_printf("file.filename=%s,driver=qcow2,"
300 "file.locking=off",
301 s_local_disk);
302 opts = qemu_opts_parse_noisily(&qemu_drive_opts, cmdline, false);
303 g_free(cmdline);
304
305 qdict = qemu_opts_to_qdict(opts, NULL);
306 qdict_set_default_str(qdict, BDRV_OPT_CACHE_DIRECT, "off");
307 qdict_set_default_str(qdict, BDRV_OPT_CACHE_NO_FLUSH, "off");
308
309 blk = blk_new_open(NULL, NULL, qdict, BDRV_O_RDWR, &error_abort);
310 assert(blk);
311 monitor_add_blk(blk, S_LOCAL_DISK_ID, &error_abort);
312
313 /* format s_local_disk with pattern "0x11" */
314 test_blk_write(blk, 0x11, 0, IMG_SIZE, false);
315
316 qemu_opts_del(opts);
317
318 /* add S_(ACTIVE/HIDDEN)_DISK and forge S_ID */
319 cmdline = g_strdup_printf("driver=replication,mode=secondary,top-id=%s,"
320 "file.driver=qcow2,file.file.filename=%s,"
321 "file.file.locking=off,"
322 "file.backing.driver=qcow2,"
323 "file.backing.file.filename=%s,"
324 "file.backing.file.locking=off,"
325 "file.backing.backing=%s"
326 , S_ID, s_active_disk, s_hidden_disk
327 , S_LOCAL_DISK_ID);
328 opts = qemu_opts_parse_noisily(&qemu_drive_opts, cmdline, false);
329 g_free(cmdline);
330
331 qdict = qemu_opts_to_qdict(opts, NULL);
332 qdict_set_default_str(qdict, BDRV_OPT_CACHE_DIRECT, "off");
333 qdict_set_default_str(qdict, BDRV_OPT_CACHE_NO_FLUSH, "off");
334
335 blk = blk_new_open(NULL, NULL, qdict, BDRV_O_RDWR, &error_abort);
336 assert(blk);
337 monitor_add_blk(blk, S_ID, &error_abort);
338
339 qemu_opts_del(opts);
340
341 return blk;
342}
343
344static void teardown_secondary(void)
345{
346 /* only need to destroy two BBs */
347 BlockBackend *blk;
348 AioContext *ctx;
349
350 /* remove S_LOCAL_DISK_ID */
351 blk = blk_by_name(S_LOCAL_DISK_ID);
352 assert(blk);
353
354 ctx = blk_get_aio_context(blk);
355 aio_context_acquire(ctx);
356 monitor_remove_blk(blk);
357 blk_unref(blk);
358 aio_context_release(ctx);
359
360 /* remove S_ID */
361 blk = blk_by_name(S_ID);
362 assert(blk);
363
364 ctx = blk_get_aio_context(blk);
365 aio_context_acquire(ctx);
366 monitor_remove_blk(blk);
367 blk_unref(blk);
368 aio_context_release(ctx);
369}
370
371static void test_secondary_read(void)
372{
373 BlockBackend *blk;
374
375 blk = start_secondary();
376
377 /* read from 0 to IMG_SIZE */
378 test_blk_read(blk, 0, 0, IMG_SIZE, 0, IMG_SIZE, true);
379
380 teardown_secondary();
381}
382
383static void test_secondary_write(void)
384{
385 BlockBackend *blk;
386
387 blk = start_secondary();
388
389 /* write from 0 to IMG_SIZE */
390 test_blk_write(blk, 0, 0, IMG_SIZE, true);
391
392 teardown_secondary();
393}
394
395static void test_secondary_start(void)
396{
397 BlockBackend *top_blk, *local_blk;
398 bool failover = true;
399
400 top_blk = start_secondary();
401 replication_start_all(REPLICATION_MODE_SECONDARY, &error_abort);
402
403 /* read from s_local_disk (0, IMG_SIZE) */
404 test_blk_read(top_blk, 0x11, 0, IMG_SIZE, 0, IMG_SIZE, false);
405
406 /* write 0x22 to s_local_disk (IMG_SIZE / 2, IMG_SIZE) */
407 local_blk = blk_by_name(S_LOCAL_DISK_ID);
408 test_blk_write(local_blk, 0x22, IMG_SIZE / 2, IMG_SIZE / 2, false);
409
410 /* replication will backup s_local_disk to s_hidden_disk */
411 test_blk_read(top_blk, 0x11, IMG_SIZE / 2,
412 IMG_SIZE / 2, 0, IMG_SIZE, false);
413
414 /* write 0x33 to s_active_disk (0, IMG_SIZE / 2) */
415 test_blk_write(top_blk, 0x33, 0, IMG_SIZE / 2, false);
416
417 /* read from s_active_disk (0, IMG_SIZE/2) */
418 test_blk_read(top_blk, 0x33, 0, IMG_SIZE / 2,
419 0, IMG_SIZE / 2, false);
420
421 /* unblock top_bs */
422 replication_stop_all(failover, &error_abort);
423
424 teardown_secondary();
425}
426
427
428static void test_secondary_stop(void)
429{
430 BlockBackend *top_blk, *local_blk;
431 bool failover = true;
432
433 top_blk = start_secondary();
434 replication_start_all(REPLICATION_MODE_SECONDARY, &error_abort);
435
436 /* write 0x22 to s_local_disk (IMG_SIZE / 2, IMG_SIZE) */
437 local_blk = blk_by_name(S_LOCAL_DISK_ID);
438 test_blk_write(local_blk, 0x22, IMG_SIZE / 2, IMG_SIZE / 2, false);
439
440 /* replication will backup s_local_disk to s_hidden_disk */
441 test_blk_read(top_blk, 0x11, IMG_SIZE / 2,
442 IMG_SIZE / 2, 0, IMG_SIZE, false);
443
444 /* write 0x33 to s_active_disk (0, IMG_SIZE / 2) */
445 test_blk_write(top_blk, 0x33, 0, IMG_SIZE / 2, false);
446
447 /* do active commit */
448 replication_stop_all(failover, &error_abort);
449
450 /* read from s_local_disk (0, IMG_SIZE / 2) */
451 test_blk_read(top_blk, 0x33, 0, IMG_SIZE / 2,
452 0, IMG_SIZE / 2, false);
453
454
455 /* read from s_local_disk (IMG_SIZE / 2, IMG_SIZE) */
456 test_blk_read(top_blk, 0x22, IMG_SIZE / 2,
457 IMG_SIZE / 2, 0, IMG_SIZE, false);
458
459 teardown_secondary();
460}
461
462static void test_secondary_continuous_replication(void)
463{
464 BlockBackend *top_blk, *local_blk;
465
466 top_blk = start_secondary();
467 replication_start_all(REPLICATION_MODE_SECONDARY, &error_abort);
468
469 /* write 0x22 to s_local_disk (IMG_SIZE / 2, IMG_SIZE) */
470 local_blk = blk_by_name(S_LOCAL_DISK_ID);
471 test_blk_write(local_blk, 0x22, IMG_SIZE / 2, IMG_SIZE / 2, false);
472
473 /* replication will backup s_local_disk to s_hidden_disk */
474 test_blk_read(top_blk, 0x11, IMG_SIZE / 2,
475 IMG_SIZE / 2, 0, IMG_SIZE, false);
476
477 /* write 0x33 to s_active_disk (0, IMG_SIZE / 2) */
478 test_blk_write(top_blk, 0x33, 0, IMG_SIZE / 2, false);
479
480 /* do failover (active commit) */
481 replication_stop_all(true, &error_abort);
482
483 /* it should ignore all requests from now on */
484
485 /* start after failover */
486 replication_start_all(REPLICATION_MODE_PRIMARY, &error_abort);
487
488 /* checkpoint */
489 replication_do_checkpoint_all(&error_abort);
490
491 /* stop */
492 replication_stop_all(true, &error_abort);
493
494 /* read from s_local_disk (0, IMG_SIZE / 2) */
495 test_blk_read(top_blk, 0x33, 0, IMG_SIZE / 2,
496 0, IMG_SIZE / 2, false);
497
498
499 /* read from s_local_disk (IMG_SIZE / 2, IMG_SIZE) */
500 test_blk_read(top_blk, 0x22, IMG_SIZE / 2,
501 IMG_SIZE / 2, 0, IMG_SIZE, false);
502
503 teardown_secondary();
504}
505
506static void test_secondary_do_checkpoint(void)
507{
508 BlockBackend *top_blk, *local_blk;
509 bool failover = true;
510
511 top_blk = start_secondary();
512 replication_start_all(REPLICATION_MODE_SECONDARY, &error_abort);
513
514 /* write 0x22 to s_local_disk (IMG_SIZE / 2, IMG_SIZE) */
515 local_blk = blk_by_name(S_LOCAL_DISK_ID);
516 test_blk_write(local_blk, 0x22, IMG_SIZE / 2,
517 IMG_SIZE / 2, false);
518
519 /* replication will backup s_local_disk to s_hidden_disk */
520 test_blk_read(top_blk, 0x11, IMG_SIZE / 2,
521 IMG_SIZE / 2, 0, IMG_SIZE, false);
522
523 replication_do_checkpoint_all(&error_abort);
524
525 /* after checkpoint, read pattern 0x22 from s_local_disk */
526 test_blk_read(top_blk, 0x22, IMG_SIZE / 2,
527 IMG_SIZE / 2, 0, IMG_SIZE, false);
528
529 /* unblock top_bs */
530 replication_stop_all(failover, &error_abort);
531
532 teardown_secondary();
533}
534
535static void test_secondary_get_error_all(void)
536{
537 bool failover = true;
538
539 start_secondary();
540 replication_start_all(REPLICATION_MODE_SECONDARY, &error_abort);
541
542 replication_get_error_all(&error_abort);
543
544 /* unblock top_bs */
545 replication_stop_all(failover, &error_abort);
546
547 teardown_secondary();
548}
549
550static void sigabrt_handler(int signo)
551{
552 cleanup_imgs();
553}
554
555static void setup_sigabrt_handler(void)
556{
557 struct sigaction sigact;
558
559 sigact = (struct sigaction) {
560 .sa_handler = sigabrt_handler,
561 .sa_flags = SA_RESETHAND,
562 };
563 sigemptyset(&sigact.sa_mask);
564 sigaction(SIGABRT, &sigact, NULL);
565}
566
567int main(int argc, char **argv)
568{
569 int ret;
570 qemu_init_main_loop(&error_fatal);
571 bdrv_init();
572
573 g_test_init(&argc, &argv, NULL);
574 setup_sigabrt_handler();
575
576 prepare_imgs();
577
578 /* Primary */
579 g_test_add_func("/replication/primary/read", test_primary_read);
580 g_test_add_func("/replication/primary/write", test_primary_write);
581 g_test_add_func("/replication/primary/start", test_primary_start);
582 g_test_add_func("/replication/primary/stop", test_primary_stop);
583 g_test_add_func("/replication/primary/do_checkpoint",
584 test_primary_do_checkpoint);
585 g_test_add_func("/replication/primary/get_error_all",
586 test_primary_get_error_all);
587
588 /* Secondary */
589 g_test_add_func("/replication/secondary/read", test_secondary_read);
590 g_test_add_func("/replication/secondary/write", test_secondary_write);
591 g_test_add_func("/replication/secondary/start", test_secondary_start);
592 g_test_add_func("/replication/secondary/stop", test_secondary_stop);
593 g_test_add_func("/replication/secondary/continuous_replication",
594 test_secondary_continuous_replication);
595 g_test_add_func("/replication/secondary/do_checkpoint",
596 test_secondary_do_checkpoint);
597 g_test_add_func("/replication/secondary/get_error_all",
598 test_secondary_get_error_all);
599
600 ret = g_test_run();
601
602 cleanup_imgs();
603
604 return ret;
605}