Commit | Line | Data |
---|---|---|
a40e7a59 GB |
1 | /* |
2 | * FIO engines for DDN's Infinite Memory Engine. | |
3 | * This file defines 3 engines: ime_psync, ime_psyncv, and ime_aio | |
4 | * | |
5 | * Copyright (C) 2018 DataDirect Networks. All rights reserved. | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or | |
8 | * modify it under the terms of the GNU General Public License, | |
9 | * version 2 as published by the Free Software Foundation.. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | */ | |
16 | ||
17 | /* | |
18 | * Some details about the new engines are given below: | |
19 | * | |
20 | * | |
21 | * ime_psync: | |
22 | * Most basic engine that issues calls to ime_native whenever an IO is queued. | |
23 | * | |
24 | * ime_psyncv: | |
25 | * This engine tries to queue the IOs (by creating iovecs) if asked by FIO (via | |
26 | * iodepth_batch). It refuses to queue when the iovecs can't be appended, and | |
27 | * waits for FIO to issue a commit. After a call to commit and get_events, new | |
28 | * IOs can be queued. | |
29 | * | |
30 | * ime_aio: | |
31 | * This engine tries to queue the IOs (by creating iovecs) if asked by FIO (via | |
32 | * iodepth_batch). When the iovecs can't be appended to the current request, a | |
33 | * new request for IME is created. These requests will be issued to IME when | |
34 | * commit is called. Contrary to ime_psyncv, there can be several requests at | |
35 | * once. We don't need to wait for a request to terminate before creating a new | |
36 | * one. | |
37 | */ | |
38 | ||
39 | #include <stdio.h> | |
40 | #include <stdlib.h> | |
41 | #include <errno.h> | |
42 | #include <linux/limits.h> | |
43 | #include <ime_native.h> | |
44 | ||
45 | #include "../fio.h" | |
46 | ||
47 | ||
48 | /************************************************************** | |
49 | * Types and constants definitions | |
50 | * | |
51 | **************************************************************/ | |
52 | ||
53 | /* define constants for async IOs */ | |
54 | #define FIO_IME_IN_PROGRESS -1 | |
55 | #define FIO_IME_REQ_ERROR -2 | |
56 | ||
57 | /* This flag is used when some jobs were created using threads. In that | |
58 | case, IME can't be finalized in the engine-specific cleanup function, | |
59 | because other threads might still use IME. Instead, IME is finalized | |
60 | in the destructor (see fio_ime_unregister), only when the flag | |
61 | fio_ime_is_initialized is true (which means at least one thread has | |
62 | initialized IME). */ | |
63 | static bool fio_ime_is_initialized = false; | |
64 | ||
65 | struct imesio_req { | |
66 | int fd; /* File descriptor */ | |
67 | enum fio_ddir ddir; /* Type of IO (read or write) */ | |
68 | off_t offset; /* File offset */ | |
69 | }; | |
70 | struct imeaio_req { | |
71 | struct ime_aiocb iocb; /* IME aio request */ | |
72 | ssize_t status; /* Status of the IME request */ | |
73 | enum fio_ddir ddir; /* Type of IO (read or write) */ | |
74 | pthread_cond_t cond_endio; /* Condition var to notify FIO */ | |
75 | pthread_mutex_t status_mutex; /* Mutex for cond_endio */ | |
76 | }; | |
77 | ||
78 | /* This structure will be used for 2 engines: ime_psyncv and ime_aio */ | |
79 | struct ime_data { | |
80 | union { | |
81 | struct imeaio_req *aioreqs; /* array of aio requests */ | |
82 | struct imesio_req *sioreq; /* pointer to the only syncio request */ | |
83 | }; | |
84 | struct iovec *iovecs; /* array of queued iovecs */ | |
85 | struct io_u **io_us; /* array of queued io_u pointers */ | |
fc002f14 | 86 | struct io_u **event_io_us; /* array of the events retrieved after get_events*/ |
a40e7a59 GB |
87 | unsigned int queued; /* iovecs/io_us in the queue */ |
88 | unsigned int events; /* number of committed iovecs/io_us */ | |
89 | ||
90 | /* variables used to implement a "ring" queue */ | |
91 | unsigned int depth; /* max entries in the queue */ | |
92 | unsigned int head; /* index used to append */ | |
93 | unsigned int tail; /* index used to pop */ | |
94 | unsigned int cur_commit; /* index of the first uncommitted req */ | |
95 | ||
96 | /* offset used by the last iovec (used to check if the iovecs can be appended)*/ | |
97 | unsigned long long last_offset; | |
98 | ||
99 | /* The variables below are used for aio only */ | |
100 | struct imeaio_req *last_req; /* last request awaiting committing */ | |
101 | }; | |
102 | ||
103 | ||
104 | /************************************************************** | |
105 | * Private functions for queueing/unqueueing | |
106 | * | |
107 | **************************************************************/ | |
108 | ||
109 | static void fio_ime_queue_incr (struct ime_data *ime_d) | |
110 | { | |
111 | ime_d->head = (ime_d->head + 1) % ime_d->depth; | |
112 | ime_d->queued++; | |
113 | } | |
38c1392e | 114 | |
a40e7a59 GB |
115 | static void fio_ime_queue_red (struct ime_data *ime_d) |
116 | { | |
117 | ime_d->tail = (ime_d->tail + 1) % ime_d->depth; | |
118 | ime_d->queued--; | |
119 | ime_d->events--; | |
120 | } | |
38c1392e | 121 | |
a40e7a59 GB |
122 | static void fio_ime_queue_commit (struct ime_data *ime_d, int iovcnt) |
123 | { | |
124 | ime_d->cur_commit = (ime_d->cur_commit + iovcnt) % ime_d->depth; | |
125 | ime_d->events += iovcnt; | |
126 | } | |
38c1392e | 127 | |
a40e7a59 GB |
128 | static void fio_ime_queue_reset (struct ime_data *ime_d) |
129 | { | |
130 | ime_d->head = 0; | |
131 | ime_d->tail = 0; | |
132 | ime_d->cur_commit = 0; | |
133 | ime_d->queued = 0; | |
134 | ime_d->events = 0; | |
135 | } | |
136 | ||
a40e7a59 GB |
137 | /************************************************************** |
138 | * General IME functions | |
139 | * (needed for both sync and async IOs) | |
140 | **************************************************************/ | |
141 | ||
142 | static char *fio_set_ime_filename(char* filename) | |
143 | { | |
144 | static __thread char ime_filename[PATH_MAX]; | |
38c1392e JA |
145 | int ret; |
146 | ||
147 | ret = snprintf(ime_filename, PATH_MAX, "%s%s", DEFAULT_IME_FILE_PREFIX, filename); | |
148 | if (ret < PATH_MAX) | |
a40e7a59 | 149 | return ime_filename; |
38c1392e JA |
150 | |
151 | return NULL; | |
a40e7a59 GB |
152 | } |
153 | ||
154 | static int fio_ime_get_file_size(struct thread_data *td, struct fio_file *f) | |
155 | { | |
156 | struct stat buf; | |
157 | int ret; | |
158 | char *ime_filename; | |
159 | ||
160 | dprint(FD_FILE, "get file size %s\n", f->file_name); | |
161 | ||
162 | ime_filename = fio_set_ime_filename(f->file_name); | |
163 | if (ime_filename == NULL) | |
164 | return 1; | |
165 | ret = ime_native_stat(ime_filename, &buf); | |
166 | if (ret == -1) { | |
167 | td_verror(td, errno, "fstat"); | |
168 | return 1; | |
169 | } | |
170 | ||
171 | f->real_file_size = buf.st_size; | |
172 | return 0; | |
173 | } | |
174 | ||
175 | /* This functions mimics the generic_file_open function, but issues | |
176 | IME native calls instead of POSIX calls. */ | |
177 | static int fio_ime_open_file(struct thread_data *td, struct fio_file *f) | |
178 | { | |
179 | int flags = 0; | |
180 | int ret; | |
181 | uint64_t desired_fs; | |
182 | char *ime_filename; | |
183 | ||
184 | dprint(FD_FILE, "fd open %s\n", f->file_name); | |
185 | ||
186 | if (td_trim(td)) { | |
187 | td_verror(td, EINVAL, "IME does not support TRIM operation"); | |
188 | return 1; | |
189 | } | |
190 | ||
a40e7a59 GB |
191 | if (td->o.odirect) |
192 | flags |= O_DIRECT; | |
eb9f8d7f | 193 | flags |= td->o.sync_io; |
a40e7a59 GB |
194 | if (td->o.create_on_open && td->o.allow_create) |
195 | flags |= O_CREAT; | |
196 | ||
197 | if (td_write(td)) { | |
198 | if (!read_only) | |
199 | flags |= O_RDWR; | |
200 | ||
201 | if (td->o.allow_create) | |
202 | flags |= O_CREAT; | |
38c1392e | 203 | } else if (td_read(td)) { |
a40e7a59 | 204 | flags |= O_RDONLY; |
38c1392e | 205 | } else { |
a40e7a59 GB |
206 | /* We should never go here. */ |
207 | td_verror(td, EINVAL, "Unsopported open mode"); | |
208 | return 1; | |
209 | } | |
210 | ||
211 | ime_filename = fio_set_ime_filename(f->file_name); | |
212 | if (ime_filename == NULL) | |
213 | return 1; | |
214 | f->fd = ime_native_open(ime_filename, flags, 0600); | |
215 | if (f->fd == -1) { | |
216 | char buf[FIO_VERROR_SIZE]; | |
217 | int __e = errno; | |
218 | ||
219 | snprintf(buf, sizeof(buf), "open(%s)", f->file_name); | |
220 | td_verror(td, __e, buf); | |
221 | return 1; | |
222 | } | |
223 | ||
224 | /* Now we need to make sure the real file size is sufficient for FIO | |
225 | to do its things. This is normally done before the file open function | |
226 | is called, but because FIO would use POSIX calls, we need to do it | |
227 | ourselves */ | |
228 | ret = fio_ime_get_file_size(td, f); | |
229 | if (ret < 0) { | |
230 | ime_native_close(f->fd); | |
231 | td_verror(td, errno, "ime_get_file_size"); | |
232 | return 1; | |
233 | } | |
234 | ||
235 | desired_fs = f->io_size + f->file_offset; | |
236 | if (td_write(td)) { | |
237 | dprint(FD_FILE, "Laying out file %s%s\n", | |
238 | DEFAULT_IME_FILE_PREFIX, f->file_name); | |
239 | if (!td->o.create_on_open && | |
240 | f->real_file_size < desired_fs && | |
241 | ime_native_ftruncate(f->fd, desired_fs) < 0) { | |
242 | ime_native_close(f->fd); | |
243 | td_verror(td, errno, "ime_native_ftruncate"); | |
244 | return 1; | |
245 | } | |
246 | if (f->real_file_size < desired_fs) | |
247 | f->real_file_size = desired_fs; | |
38c1392e | 248 | } else if (td_read(td) && f->real_file_size < desired_fs) { |
a40e7a59 GB |
249 | ime_native_close(f->fd); |
250 | log_err("error: can't read %lu bytes from file with " | |
251 | "%lu bytes\n", desired_fs, f->real_file_size); | |
252 | return 1; | |
253 | } | |
254 | ||
255 | return 0; | |
256 | } | |
257 | ||
258 | static int fio_ime_close_file(struct thread_data fio_unused *td, struct fio_file *f) | |
259 | { | |
260 | int ret = 0; | |
261 | ||
262 | dprint(FD_FILE, "fd close %s\n", f->file_name); | |
263 | ||
264 | if (ime_native_close(f->fd) < 0) | |
265 | ret = errno; | |
266 | ||
267 | f->fd = -1; | |
268 | return ret; | |
269 | } | |
270 | ||
271 | static int fio_ime_unlink_file(struct thread_data *td, struct fio_file *f) | |
272 | { | |
38c1392e | 273 | char *ime_filename = fio_set_ime_filename(f->file_name); |
a40e7a59 GB |
274 | int ret; |
275 | ||
a40e7a59 GB |
276 | if (ime_filename == NULL) |
277 | return 1; | |
a40e7a59 | 278 | |
38c1392e | 279 | ret = unlink(ime_filename); |
a40e7a59 GB |
280 | return ret < 0 ? errno : 0; |
281 | } | |
282 | ||
283 | static struct io_u *fio_ime_event(struct thread_data *td, int event) | |
284 | { | |
285 | struct ime_data *ime_d = td->io_ops_data; | |
38c1392e | 286 | |
a40e7a59 GB |
287 | return ime_d->event_io_us[event]; |
288 | } | |
289 | ||
290 | /* Setup file used to replace get_file_sizes when settin up the file. | |
291 | Instead we will set real_file_sie to 0 for each file. This way we | |
292 | can avoid calling ime_native_init before the forks are created. */ | |
293 | static int fio_ime_setup(struct thread_data *td) | |
294 | { | |
295 | struct fio_file *f; | |
296 | unsigned int i; | |
297 | ||
298 | for_each_file(td, f, i) { | |
299 | dprint(FD_FILE, "setup: set file size to 0 for %p/%d/%s\n", | |
300 | f, i, f->file_name); | |
301 | f->real_file_size = 0; | |
302 | } | |
303 | ||
304 | return 0; | |
305 | } | |
306 | ||
307 | static int fio_ime_engine_init(struct thread_data *td) | |
308 | { | |
309 | struct fio_file *f; | |
310 | unsigned int i; | |
311 | ||
312 | dprint(FD_IO, "ime engine init\n"); | |
313 | if (fio_ime_is_initialized && !td->o.use_thread) { | |
314 | log_err("Warning: something might go wrong. Not all threads/forks were" | |
315 | " created before the FIO jobs were initialized.\n"); | |
316 | } | |
317 | ||
318 | ime_native_init(); | |
319 | fio_ime_is_initialized = true; | |
320 | ||
321 | /* We have to temporarily set real_file_size so that | |
322 | FIO can initialize properly. It will be corrected | |
323 | on file open. */ | |
324 | for_each_file(td, f, i) | |
325 | f->real_file_size = f->io_size + f->file_offset; | |
326 | ||
327 | return 0; | |
328 | } | |
329 | ||
330 | static void fio_ime_engine_finalize(struct thread_data *td) | |
331 | { | |
332 | /* Only finalize IME when using forks */ | |
333 | if (!td->o.use_thread) { | |
334 | if (ime_native_finalize() < 0) | |
335 | log_err("error in ime_native_finalize\n"); | |
336 | fio_ime_is_initialized = false; | |
337 | } | |
338 | } | |
339 | ||
340 | ||
341 | /************************************************************** | |
342 | * Private functions for blocking IOs | |
343 | * (without iovecs) | |
344 | **************************************************************/ | |
345 | ||
346 | /* Notice: this function comes from the sync engine */ | |
347 | /* It is used by the commit function to return a proper code and fill | |
348 | some attributes in the io_u used for the IO. */ | |
349 | static int fio_ime_psync_end(struct thread_data *td, struct io_u *io_u, ssize_t ret) | |
350 | { | |
351 | if (ret != (ssize_t) io_u->xfer_buflen) { | |
352 | if (ret >= 0) { | |
353 | io_u->resid = io_u->xfer_buflen - ret; | |
354 | io_u->error = 0; | |
355 | return FIO_Q_COMPLETED; | |
356 | } else | |
357 | io_u->error = errno; | |
358 | } | |
359 | ||
360 | if (io_u->error) { | |
361 | io_u_log_error(td, io_u); | |
362 | td_verror(td, io_u->error, "xfer"); | |
363 | } | |
364 | ||
365 | return FIO_Q_COMPLETED; | |
366 | } | |
367 | ||
368 | static enum fio_q_status fio_ime_psync_queue(struct thread_data *td, | |
369 | struct io_u *io_u) | |
370 | { | |
371 | struct fio_file *f = io_u->file; | |
372 | ssize_t ret; | |
373 | ||
374 | fio_ro_check(td, io_u); | |
375 | ||
376 | if (io_u->ddir == DDIR_READ) | |
377 | ret = ime_native_pread(f->fd, io_u->xfer_buf, io_u->xfer_buflen, io_u->offset); | |
378 | else if (io_u->ddir == DDIR_WRITE) | |
379 | ret = ime_native_pwrite(f->fd, io_u->xfer_buf, io_u->xfer_buflen, io_u->offset); | |
380 | else if (io_u->ddir == DDIR_SYNC) | |
381 | ret = ime_native_fsync(f->fd); | |
382 | else { | |
383 | ret = io_u->xfer_buflen; | |
384 | io_u->error = EINVAL; | |
385 | } | |
386 | ||
387 | return fio_ime_psync_end(td, io_u, ret); | |
388 | } | |
389 | ||
390 | ||
391 | /************************************************************** | |
392 | * Private functions for blocking IOs | |
393 | * (with iovecs) | |
394 | **************************************************************/ | |
395 | ||
396 | static bool fio_ime_psyncv_can_queue(struct ime_data *ime_d, struct io_u *io_u) | |
397 | { | |
398 | /* We can only queue if: | |
399 | - There are no queued iovecs | |
400 | - Or if there is at least one: | |
401 | - There must be no event waiting for retrieval | |
402 | - The offsets must be contiguous | |
403 | - The ddir and fd must be the same */ | |
404 | return (ime_d->queued == 0 || ( | |
405 | ime_d->events == 0 && | |
406 | ime_d->last_offset == io_u->offset && | |
407 | ime_d->sioreq->ddir == io_u->ddir && | |
408 | ime_d->sioreq->fd == io_u->file->fd)); | |
409 | } | |
410 | ||
411 | /* Before using this function, we should have already | |
412 | ensured that the queue is not full */ | |
413 | static void fio_ime_psyncv_enqueue(struct ime_data *ime_d, struct io_u *io_u) | |
414 | { | |
415 | struct imesio_req *ioreq = ime_d->sioreq; | |
416 | struct iovec *iov = &ime_d->iovecs[ime_d->head]; | |
417 | ||
418 | iov->iov_base = io_u->xfer_buf; | |
419 | iov->iov_len = io_u->xfer_buflen; | |
420 | ||
421 | if (ime_d->queued == 0) { | |
422 | ioreq->offset = io_u->offset; | |
423 | ioreq->ddir = io_u->ddir; | |
424 | ioreq->fd = io_u->file->fd; | |
425 | } | |
426 | ||
427 | ime_d->io_us[ime_d->head] = io_u; | |
428 | ime_d->last_offset = io_u->offset + io_u->xfer_buflen; | |
429 | fio_ime_queue_incr(ime_d); | |
430 | } | |
431 | ||
432 | /* Tries to queue an IO. It will fail if the IO can't be appended to the | |
433 | current request or if the current request has been committed but not | |
434 | yet retrieved by get_events. */ | |
435 | static enum fio_q_status fio_ime_psyncv_queue(struct thread_data *td, | |
436 | struct io_u *io_u) | |
437 | { | |
438 | struct ime_data *ime_d = td->io_ops_data; | |
439 | ||
440 | fio_ro_check(td, io_u); | |
441 | ||
442 | if (ime_d->queued == ime_d->depth) | |
443 | return FIO_Q_BUSY; | |
444 | ||
445 | if (io_u->ddir == DDIR_READ || io_u->ddir == DDIR_WRITE) { | |
446 | if (!fio_ime_psyncv_can_queue(ime_d, io_u)) | |
447 | return FIO_Q_BUSY; | |
448 | ||
449 | dprint(FD_IO, "queue: ddir=%d at %u commit=%u queued=%u events=%u\n", | |
450 | io_u->ddir, ime_d->head, ime_d->cur_commit, | |
451 | ime_d->queued, ime_d->events); | |
452 | fio_ime_psyncv_enqueue(ime_d, io_u); | |
453 | return FIO_Q_QUEUED; | |
38c1392e | 454 | } else if (io_u->ddir == DDIR_SYNC) { |
a40e7a59 GB |
455 | if (ime_native_fsync(io_u->file->fd) < 0) { |
456 | io_u->error = errno; | |
457 | td_verror(td, io_u->error, "fsync"); | |
458 | } | |
459 | return FIO_Q_COMPLETED; | |
460 | } else { | |
461 | io_u->error = EINVAL; | |
462 | td_verror(td, io_u->error, "wrong ddir"); | |
463 | return FIO_Q_COMPLETED; | |
464 | } | |
465 | } | |
466 | ||
467 | /* Notice: this function comes from the sync engine */ | |
468 | /* It is used by the commit function to return a proper code and fill | |
469 | some attributes in the io_us appended to the current request. */ | |
470 | static int fio_ime_psyncv_end(struct thread_data *td, ssize_t bytes) | |
471 | { | |
472 | struct ime_data *ime_d = td->io_ops_data; | |
473 | struct io_u *io_u; | |
474 | unsigned int i; | |
475 | int err = errno; | |
476 | ||
477 | for (i = 0; i < ime_d->queued; i++) { | |
478 | io_u = ime_d->io_us[i]; | |
479 | ||
480 | if (bytes == -1) | |
481 | io_u->error = err; | |
482 | else { | |
483 | unsigned int this_io; | |
484 | ||
485 | this_io = bytes; | |
486 | if (this_io > io_u->xfer_buflen) | |
487 | this_io = io_u->xfer_buflen; | |
488 | ||
489 | io_u->resid = io_u->xfer_buflen - this_io; | |
490 | io_u->error = 0; | |
491 | bytes -= this_io; | |
492 | } | |
493 | } | |
494 | ||
495 | if (bytes == -1) { | |
496 | td_verror(td, err, "xfer psyncv"); | |
497 | return -err; | |
498 | } | |
499 | ||
500 | return 0; | |
501 | } | |
502 | ||
503 | /* Commits the current request by calling ime_native (with one or several | |
504 | iovecs). After this commit, the corresponding events (one per iovec) | |
505 | can be retrieved by get_events. */ | |
506 | static int fio_ime_psyncv_commit(struct thread_data *td) | |
507 | { | |
508 | struct ime_data *ime_d = td->io_ops_data; | |
509 | struct imesio_req *ioreq; | |
510 | int ret = 0; | |
511 | ||
512 | /* Exit if there are no (new) events to commit | |
513 | or if the previous committed event haven't been retrieved */ | |
514 | if (!ime_d->queued || ime_d->events) | |
515 | return 0; | |
516 | ||
517 | ioreq = ime_d->sioreq; | |
518 | ime_d->events = ime_d->queued; | |
519 | if (ioreq->ddir == DDIR_READ) | |
520 | ret = ime_native_preadv(ioreq->fd, ime_d->iovecs, ime_d->queued, ioreq->offset); | |
521 | else | |
522 | ret = ime_native_pwritev(ioreq->fd, ime_d->iovecs, ime_d->queued, ioreq->offset); | |
523 | ||
524 | dprint(FD_IO, "committed %d iovecs\n", ime_d->queued); | |
525 | ||
526 | return fio_ime_psyncv_end(td, ret); | |
527 | } | |
528 | ||
529 | static int fio_ime_psyncv_getevents(struct thread_data *td, unsigned int min, | |
530 | unsigned int max, const struct timespec *t) | |
531 | { | |
532 | struct ime_data *ime_d = td->io_ops_data; | |
533 | struct io_u *io_u; | |
534 | int events = 0; | |
535 | unsigned int count; | |
536 | ||
537 | if (ime_d->events) { | |
538 | for (count = 0; count < ime_d->events; count++) { | |
539 | io_u = ime_d->io_us[count]; | |
540 | ime_d->event_io_us[events] = io_u; | |
541 | events++; | |
542 | } | |
543 | fio_ime_queue_reset(ime_d); | |
544 | } | |
545 | ||
546 | dprint(FD_IO, "getevents(%u,%u) ret=%d queued=%u events=%u\n", | |
547 | min, max, events, ime_d->queued, ime_d->events); | |
548 | return events; | |
549 | } | |
550 | ||
551 | static int fio_ime_psyncv_init(struct thread_data *td) | |
552 | { | |
553 | struct ime_data *ime_d; | |
554 | ||
555 | if (fio_ime_engine_init(td) < 0) | |
556 | return 1; | |
557 | ||
558 | ime_d = calloc(1, sizeof(*ime_d)); | |
559 | ||
560 | ime_d->sioreq = malloc(sizeof(struct imesio_req)); | |
561 | ime_d->iovecs = malloc(td->o.iodepth * sizeof(struct iovec)); | |
562 | ime_d->io_us = malloc(2 * td->o.iodepth * sizeof(struct io_u *)); | |
563 | ime_d->event_io_us = ime_d->io_us + td->o.iodepth; | |
564 | ||
565 | ime_d->depth = td->o.iodepth; | |
566 | ||
567 | td->io_ops_data = ime_d; | |
568 | return 0; | |
569 | } | |
570 | ||
571 | static void fio_ime_psyncv_clean(struct thread_data *td) | |
572 | { | |
573 | struct ime_data *ime_d = td->io_ops_data; | |
574 | ||
575 | if (ime_d) { | |
576 | free(ime_d->sioreq); | |
577 | free(ime_d->iovecs); | |
578 | free(ime_d->io_us); | |
579 | free(ime_d); | |
580 | td->io_ops_data = NULL; | |
581 | } | |
582 | ||
583 | fio_ime_engine_finalize(td); | |
584 | } | |
585 | ||
586 | ||
587 | /************************************************************** | |
588 | * Private functions for non-blocking IOs | |
589 | * | |
590 | **************************************************************/ | |
591 | ||
592 | void fio_ime_aio_complete_cb (struct ime_aiocb *aiocb, int err, | |
593 | ssize_t bytes) | |
594 | { | |
595 | struct imeaio_req *ioreq = (struct imeaio_req *) aiocb->user_context; | |
596 | ||
597 | pthread_mutex_lock(&ioreq->status_mutex); | |
598 | ioreq->status = err == 0 ? bytes : FIO_IME_REQ_ERROR; | |
599 | pthread_mutex_unlock(&ioreq->status_mutex); | |
600 | ||
601 | pthread_cond_signal(&ioreq->cond_endio); | |
602 | } | |
603 | ||
604 | static bool fio_ime_aio_can_queue (struct ime_data *ime_d, struct io_u *io_u) | |
605 | { | |
606 | /* So far we can queue in any case. */ | |
607 | return true; | |
608 | } | |
609 | static bool fio_ime_aio_can_append (struct ime_data *ime_d, struct io_u *io_u) | |
610 | { | |
611 | /* We can only append if: | |
612 | - The iovecs will be contiguous in the array | |
613 | - There is already a queued iovec | |
614 | - The offsets are contiguous | |
615 | - The ddir and fs are the same */ | |
616 | return (ime_d->head != 0 && | |
617 | ime_d->queued - ime_d->events > 0 && | |
618 | ime_d->last_offset == io_u->offset && | |
619 | ime_d->last_req->ddir == io_u->ddir && | |
620 | ime_d->last_req->iocb.fd == io_u->file->fd); | |
621 | } | |
622 | ||
623 | /* Before using this function, we should have already | |
624 | ensured that the queue is not full */ | |
625 | static void fio_ime_aio_enqueue(struct ime_data *ime_d, struct io_u *io_u) | |
626 | { | |
627 | struct imeaio_req *ioreq = &ime_d->aioreqs[ime_d->head]; | |
628 | struct ime_aiocb *iocb = &ioreq->iocb; | |
629 | struct iovec *iov = &ime_d->iovecs[ime_d->head]; | |
630 | ||
631 | iov->iov_base = io_u->xfer_buf; | |
632 | iov->iov_len = io_u->xfer_buflen; | |
633 | ||
634 | if (fio_ime_aio_can_append(ime_d, io_u)) | |
635 | ime_d->last_req->iocb.iovcnt++; | |
636 | else { | |
637 | ioreq->status = FIO_IME_IN_PROGRESS; | |
638 | ioreq->ddir = io_u->ddir; | |
639 | ime_d->last_req = ioreq; | |
640 | ||
641 | iocb->complete_cb = &fio_ime_aio_complete_cb; | |
642 | iocb->fd = io_u->file->fd; | |
643 | iocb->file_offset = io_u->offset; | |
644 | iocb->iov = iov; | |
645 | iocb->iovcnt = 1; | |
646 | iocb->flags = 0; | |
647 | iocb->user_context = (intptr_t) ioreq; | |
648 | } | |
649 | ||
650 | ime_d->io_us[ime_d->head] = io_u; | |
651 | ime_d->last_offset = io_u->offset + io_u->xfer_buflen; | |
652 | fio_ime_queue_incr(ime_d); | |
653 | } | |
654 | ||
655 | /* Tries to queue an IO. It will create a new request if the IO can't be | |
656 | appended to the current request. It will fail if the queue can't contain | |
657 | any more io_u/iovec. In this case, commit and then get_events need to be | |
658 | called. */ | |
659 | static enum fio_q_status fio_ime_aio_queue(struct thread_data *td, | |
660 | struct io_u *io_u) | |
661 | { | |
662 | struct ime_data *ime_d = td->io_ops_data; | |
663 | ||
664 | fio_ro_check(td, io_u); | |
665 | ||
666 | dprint(FD_IO, "queue: ddir=%d at %u commit=%u queued=%u events=%u\n", | |
667 | io_u->ddir, ime_d->head, ime_d->cur_commit, | |
668 | ime_d->queued, ime_d->events); | |
669 | ||
670 | if (ime_d->queued == ime_d->depth) | |
671 | return FIO_Q_BUSY; | |
672 | ||
673 | if (io_u->ddir == DDIR_READ || io_u->ddir == DDIR_WRITE) { | |
674 | if (!fio_ime_aio_can_queue(ime_d, io_u)) | |
675 | return FIO_Q_BUSY; | |
676 | ||
677 | fio_ime_aio_enqueue(ime_d, io_u); | |
678 | return FIO_Q_QUEUED; | |
38c1392e | 679 | } else if (io_u->ddir == DDIR_SYNC) { |
a40e7a59 GB |
680 | if (ime_native_fsync(io_u->file->fd) < 0) { |
681 | io_u->error = errno; | |
682 | td_verror(td, io_u->error, "fsync"); | |
683 | } | |
684 | return FIO_Q_COMPLETED; | |
685 | } else { | |
686 | io_u->error = EINVAL; | |
687 | td_verror(td, io_u->error, "wrong ddir"); | |
688 | return FIO_Q_COMPLETED; | |
689 | } | |
690 | } | |
691 | ||
692 | static int fio_ime_aio_commit(struct thread_data *td) | |
693 | { | |
694 | struct ime_data *ime_d = td->io_ops_data; | |
695 | struct imeaio_req *ioreq; | |
696 | int ret = 0; | |
697 | ||
698 | /* Loop while there are events to commit */ | |
699 | while (ime_d->queued - ime_d->events) { | |
700 | ioreq = &ime_d->aioreqs[ime_d->cur_commit]; | |
701 | if (ioreq->ddir == DDIR_READ) | |
702 | ret = ime_native_aio_read(&ioreq->iocb); | |
703 | else | |
704 | ret = ime_native_aio_write(&ioreq->iocb); | |
705 | ||
706 | fio_ime_queue_commit(ime_d, ioreq->iocb.iovcnt); | |
707 | ||
708 | /* fio needs a negative error code */ | |
709 | if (ret < 0) { | |
710 | ioreq->status = FIO_IME_REQ_ERROR; | |
711 | return -errno; | |
712 | } | |
713 | ||
714 | io_u_mark_submit(td, ioreq->iocb.iovcnt); | |
715 | dprint(FD_IO, "committed %d iovecs commit=%u queued=%u events=%u\n", | |
716 | ioreq->iocb.iovcnt, ime_d->cur_commit, | |
717 | ime_d->queued, ime_d->events); | |
718 | } | |
719 | ||
720 | return 0; | |
721 | } | |
722 | ||
723 | static int fio_ime_aio_getevents(struct thread_data *td, unsigned int min, | |
724 | unsigned int max, const struct timespec *t) | |
725 | { | |
726 | struct ime_data *ime_d = td->io_ops_data; | |
727 | struct imeaio_req *ioreq; | |
728 | struct io_u *io_u; | |
729 | int events = 0; | |
730 | unsigned int count; | |
731 | ssize_t bytes; | |
732 | ||
733 | while (ime_d->events) { | |
734 | ioreq = &ime_d->aioreqs[ime_d->tail]; | |
735 | ||
736 | /* Break if we already got events, and if we will | |
737 | exceed max if we append the next events */ | |
738 | if (events && events + ioreq->iocb.iovcnt > max) | |
739 | break; | |
740 | ||
741 | if (ioreq->status != FIO_IME_IN_PROGRESS) { | |
742 | ||
743 | bytes = ioreq->status; | |
744 | for (count = 0; count < ioreq->iocb.iovcnt; count++) { | |
745 | io_u = ime_d->io_us[ime_d->tail]; | |
746 | ime_d->event_io_us[events] = io_u; | |
747 | events++; | |
748 | fio_ime_queue_red(ime_d); | |
749 | ||
750 | if (ioreq->status == FIO_IME_REQ_ERROR) | |
751 | io_u->error = EIO; | |
752 | else { | |
753 | io_u->resid = bytes > io_u->xfer_buflen ? | |
754 | 0 : io_u->xfer_buflen - bytes; | |
755 | io_u->error = 0; | |
756 | bytes -= io_u->xfer_buflen - io_u->resid; | |
757 | } | |
758 | } | |
759 | } else { | |
760 | pthread_mutex_lock(&ioreq->status_mutex); | |
38c1392e | 761 | while (ioreq->status == FIO_IME_IN_PROGRESS) |
a40e7a59 | 762 | pthread_cond_wait(&ioreq->cond_endio, &ioreq->status_mutex); |
a40e7a59 GB |
763 | pthread_mutex_unlock(&ioreq->status_mutex); |
764 | } | |
765 | ||
766 | } | |
767 | ||
768 | dprint(FD_IO, "getevents(%u,%u) ret=%d queued=%u events=%u\n", min, max, | |
769 | events, ime_d->queued, ime_d->events); | |
770 | return events; | |
771 | } | |
772 | ||
773 | static int fio_ime_aio_init(struct thread_data *td) | |
774 | { | |
775 | struct ime_data *ime_d; | |
776 | struct imeaio_req *ioreq; | |
777 | unsigned int i; | |
778 | ||
779 | if (fio_ime_engine_init(td) < 0) | |
780 | return 1; | |
781 | ||
782 | ime_d = calloc(1, sizeof(*ime_d)); | |
783 | ||
784 | ime_d->aioreqs = malloc(td->o.iodepth * sizeof(struct imeaio_req)); | |
785 | ime_d->iovecs = malloc(td->o.iodepth * sizeof(struct iovec)); | |
786 | ime_d->io_us = malloc(2 * td->o.iodepth * sizeof(struct io_u *)); | |
787 | ime_d->event_io_us = ime_d->io_us + td->o.iodepth; | |
788 | ||
789 | ime_d->depth = td->o.iodepth; | |
790 | for (i = 0; i < ime_d->depth; i++) { | |
791 | ioreq = &ime_d->aioreqs[i]; | |
792 | pthread_cond_init(&ioreq->cond_endio, NULL); | |
793 | pthread_mutex_init(&ioreq->status_mutex, NULL); | |
794 | } | |
795 | ||
796 | td->io_ops_data = ime_d; | |
797 | return 0; | |
798 | } | |
799 | ||
800 | static void fio_ime_aio_clean(struct thread_data *td) | |
801 | { | |
802 | struct ime_data *ime_d = td->io_ops_data; | |
803 | struct imeaio_req *ioreq; | |
804 | unsigned int i; | |
805 | ||
806 | if (ime_d) { | |
807 | for (i = 0; i < ime_d->depth; i++) { | |
808 | ioreq = &ime_d->aioreqs[i]; | |
809 | pthread_cond_destroy(&ioreq->cond_endio); | |
810 | pthread_mutex_destroy(&ioreq->status_mutex); | |
811 | } | |
812 | free(ime_d->aioreqs); | |
813 | free(ime_d->iovecs); | |
814 | free(ime_d->io_us); | |
815 | free(ime_d); | |
816 | td->io_ops_data = NULL; | |
817 | } | |
818 | ||
819 | fio_ime_engine_finalize(td); | |
820 | } | |
821 | ||
822 | ||
823 | /************************************************************** | |
824 | * IO engines definitions | |
825 | * | |
826 | **************************************************************/ | |
827 | ||
828 | /* The FIO_DISKLESSIO flag used for these engines is necessary to prevent | |
829 | FIO from using POSIX calls. See fio_ime_open_file for more details. */ | |
830 | ||
831 | static struct ioengine_ops ioengine_prw = { | |
832 | .name = "ime_psync", | |
833 | .version = FIO_IOOPS_VERSION, | |
834 | .setup = fio_ime_setup, | |
835 | .init = fio_ime_engine_init, | |
836 | .cleanup = fio_ime_engine_finalize, | |
837 | .queue = fio_ime_psync_queue, | |
838 | .open_file = fio_ime_open_file, | |
839 | .close_file = fio_ime_close_file, | |
840 | .get_file_size = fio_ime_get_file_size, | |
841 | .unlink_file = fio_ime_unlink_file, | |
842 | .flags = FIO_SYNCIO | FIO_DISKLESSIO, | |
843 | }; | |
844 | ||
845 | static struct ioengine_ops ioengine_pvrw = { | |
846 | .name = "ime_psyncv", | |
847 | .version = FIO_IOOPS_VERSION, | |
848 | .setup = fio_ime_setup, | |
849 | .init = fio_ime_psyncv_init, | |
850 | .cleanup = fio_ime_psyncv_clean, | |
851 | .queue = fio_ime_psyncv_queue, | |
852 | .commit = fio_ime_psyncv_commit, | |
853 | .getevents = fio_ime_psyncv_getevents, | |
854 | .event = fio_ime_event, | |
855 | .open_file = fio_ime_open_file, | |
856 | .close_file = fio_ime_close_file, | |
857 | .get_file_size = fio_ime_get_file_size, | |
858 | .unlink_file = fio_ime_unlink_file, | |
859 | .flags = FIO_SYNCIO | FIO_DISKLESSIO, | |
860 | }; | |
861 | ||
862 | static struct ioengine_ops ioengine_aio = { | |
863 | .name = "ime_aio", | |
864 | .version = FIO_IOOPS_VERSION, | |
865 | .setup = fio_ime_setup, | |
866 | .init = fio_ime_aio_init, | |
867 | .cleanup = fio_ime_aio_clean, | |
868 | .queue = fio_ime_aio_queue, | |
869 | .commit = fio_ime_aio_commit, | |
870 | .getevents = fio_ime_aio_getevents, | |
871 | .event = fio_ime_event, | |
872 | .open_file = fio_ime_open_file, | |
873 | .close_file = fio_ime_close_file, | |
874 | .get_file_size = fio_ime_get_file_size, | |
875 | .unlink_file = fio_ime_unlink_file, | |
876 | .flags = FIO_DISKLESSIO, | |
877 | }; | |
878 | ||
879 | static void fio_init fio_ime_register(void) | |
880 | { | |
881 | register_ioengine(&ioengine_prw); | |
882 | register_ioengine(&ioengine_pvrw); | |
883 | register_ioengine(&ioengine_aio); | |
884 | } | |
885 | ||
886 | static void fio_exit fio_ime_unregister(void) | |
887 | { | |
888 | unregister_ioengine(&ioengine_prw); | |
889 | unregister_ioengine(&ioengine_pvrw); | |
890 | unregister_ioengine(&ioengine_aio); | |
38c1392e JA |
891 | |
892 | if (fio_ime_is_initialized && ime_native_finalize() < 0) | |
a40e7a59 | 893 | log_err("Warning: IME did not finalize properly\n"); |
a40e7a59 | 894 | } |