source3/modules/vfs_aio_fork.c

   1 /*
   2  * Simulate the Posix AIO using mmap/fork
   3  *
   4  * Copyright (C) Volker Lendecke 2008
   5  * Copyright (C) Jeremy Allison 2010
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License as published by
   9  * the Free Software Foundation; either version 2 of the License, or
  10  * (at your option) any later version.
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License
  18  * along with this program; if not, write to the Free Software
  19  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  20  */
  21
  22 #include "includes.h"
  23 #include "system/filesys.h"
  24 #include "system/shmem.h"
  25 #include "smbd/smbd.h"
  26 #include "smbd/globals.h"
  27 #include "lib/async_req/async_sock.h"
  28 #include "lib/util/tevent_unix.h"
  29 #include "lib/sys_rw.h"
  30 #include "lib/sys_rw_data.h"
  31
  32 #if !defined(HAVE_STRUCT_MSGHDR_MSG_CONTROL) && !defined(HAVE_STRUCT_MSGHDR_MSG_ACCRIGHTS)
  33 # error Can not pass file descriptors
  34 #endif
  35
  36 #undef recvmsg
  37
  38 #ifndef MAP_FILE
  39 #define MAP_FILE 0
  40 #endif
  41
  42 struct aio_fork_config {
  43         bool erratic_testing_mode;
  44 };
  45
  46 struct mmap_area {
  47         size_t size;
  48         volatile void *ptr;
  49 };
  50
  51 static int mmap_area_destructor(struct mmap_area *area)
  52 {
  53         munmap((void *)area->ptr, area->size);
  54         return 0;
  55 }
  56
  57 static struct mmap_area *mmap_area_init(TALLOC_CTX *mem_ctx, size_t size)
  58 {
  59         struct mmap_area *result;
  60         int fd;
  61
  62         result = talloc(mem_ctx, struct mmap_area);
  63         if (result == NULL) {
  64                 DEBUG(0, ("talloc failed\n"));
  65                 goto fail;
  66         }
  67
  68         fd = open("/dev/zero", O_RDWR);
  69         if (fd == -1) {
  70                 DEBUG(3, ("open(\"/dev/zero\") failed: %s\n",
  71                           strerror(errno)));
  72                 goto fail;
  73         }
  74
  75         result->ptr = mmap(NULL, size, PROT_READ|PROT_WRITE,
  76                            MAP_SHARED|MAP_FILE, fd, 0);
  77         close(fd);
  78         if (result->ptr == MAP_FAILED) {
  79                 DEBUG(1, ("mmap failed: %s\n", strerror(errno)));
  80                 goto fail;
  81         }
  82
  83         result->size = size;
  84         talloc_set_destructor(result, mmap_area_destructor);
  85
  86         return result;
  87
  88 fail:
  89         TALLOC_FREE(result);
  90         return NULL;
  91 }
  92
  93 enum cmd_type {
  94         READ_CMD,
  95         WRITE_CMD,
  96         FSYNC_CMD
  97 };
  98
  99 static const char *cmd_type_str(enum cmd_type cmd)
 100 {
 101         const char *result;
 102
 103         switch (cmd) {
 104         case READ_CMD:
 105                 result = "READ";
 106                 break;
 107         case WRITE_CMD:
 108                 result = "WRITE";
 109                 break;
 110         case FSYNC_CMD:
 111                 result = "FSYNC";
 112                 break;
 113         default:
 114                 result = "<UNKNOWN>";
 115                 break;
 116         }
 117         return result;
 118 }
 119
 120 struct rw_cmd {
 121         size_t n;
 122         off_t offset;
 123         enum cmd_type cmd;
 124         bool erratic_testing_mode;
 125 };
 126
 127 struct rw_ret {
 128         ssize_t size;
 129         int ret_errno;
 130 };
 131
 132 struct aio_child_list;
 133
 134 struct aio_child {
 135         struct aio_child *prev, *next;
 136         struct aio_child_list *list;
 137         pid_t pid;
 138         int sockfd;
 139         struct mmap_area *map;
 140         bool dont_delete;       /* Marked as in use since last cleanup */
 141         bool busy;
 142 };
 143
 144 struct aio_child_list {
 145         struct aio_child *children;
 146         struct tevent_timer *cleanup_event;
 147 };
 148
 149 static void free_aio_children(void **p)
 150 {
 151         TALLOC_FREE(*p);
 152 }
 153
 154 static ssize_t read_fd(int fd, void *ptr, size_t nbytes, int *recvfd)
 155 {
 156         struct msghdr msg;
 157         struct iovec iov[1];
 158         ssize_t n;
 159 #ifndef HAVE_STRUCT_MSGHDR_MSG_CONTROL
 160         int newfd;
 161
 162         ZERO_STRUCT(msg);
 163         msg.msg_accrights = (caddr_t) &newfd;
 164         msg.msg_accrightslen = sizeof(int);
 165 #else
 166
 167         union {
 168           struct cmsghdr        cm;
 169           char                          control[CMSG_SPACE(sizeof(int))];
 170         } control_un;
 171         struct cmsghdr  *cmptr;
 172
 173         ZERO_STRUCT(msg);
 174         ZERO_STRUCT(control_un);
 175
 176         msg.msg_control = control_un.control;
 177         msg.msg_controllen = sizeof(control_un.control);
 178 #endif
 179
 180         msg.msg_name = NULL;
 181         msg.msg_namelen = 0;
 182
 183         iov[0].iov_base = (void *)ptr;
 184         iov[0].iov_len = nbytes;
 185         msg.msg_iov = iov;
 186         msg.msg_iovlen = 1;
 187
 188         if ( (n = recvmsg(fd, &msg, 0)) <= 0) {
 189                 return(n);
 190         }
 191
 192 #ifdef  HAVE_STRUCT_MSGHDR_MSG_CONTROL
 193         if ((cmptr = CMSG_FIRSTHDR(&msg)) != NULL
 194             && cmptr->cmsg_len == CMSG_LEN(sizeof(int))) {
 195                 if (cmptr->cmsg_level != SOL_SOCKET) {
 196                         DEBUG(10, ("control level != SOL_SOCKET"));
 197                         errno = EINVAL;
 198                         return -1;
 199                 }
 200                 if (cmptr->cmsg_type != SCM_RIGHTS) {
 201                         DEBUG(10, ("control type != SCM_RIGHTS"));
 202                         errno = EINVAL;
 203                         return -1;
 204                 }
 205                 memcpy(recvfd, CMSG_DATA(cmptr), sizeof(*recvfd));
 206         } else {
 207                 *recvfd = -1;           /* descriptor was not passed */
 208         }
 209 #else
 210         if (msg.msg_accrightslen == sizeof(int)) {
 211                 *recvfd = newfd;
 212         }
 213         else {
 214                 *recvfd = -1;           /* descriptor was not passed */
 215         }
 216 #endif
 217
 218         return(n);
 219 }
 220
 221 static ssize_t write_fd(int fd, void *ptr, size_t nbytes, int sendfd)
 222 {
 223         struct msghdr   msg;
 224         struct iovec    iov[1];
 225
 226 #ifdef  HAVE_STRUCT_MSGHDR_MSG_CONTROL
 227         union {
 228                 struct cmsghdr  cm;
 229                 char control[CMSG_SPACE(sizeof(int))];
 230         } control_un;
 231         struct cmsghdr  *cmptr;
 232
 233         ZERO_STRUCT(msg);
 234         ZERO_STRUCT(control_un);
 235
 236         msg.msg_control = control_un.control;
 237         msg.msg_controllen = sizeof(control_un.control);
 238
 239         cmptr = CMSG_FIRSTHDR(&msg);
 240         cmptr->cmsg_len = CMSG_LEN(sizeof(int));
 241         cmptr->cmsg_level = SOL_SOCKET;
 242         cmptr->cmsg_type = SCM_RIGHTS;
 243         memcpy(CMSG_DATA(cmptr), &sendfd, sizeof(sendfd));
 244 #else
 245         ZERO_STRUCT(msg);
 246         msg.msg_accrights = (caddr_t) &sendfd;
 247         msg.msg_accrightslen = sizeof(int);
 248 #endif
 249
 250         msg.msg_name = NULL;
 251         msg.msg_namelen = 0;
 252
 253         ZERO_STRUCT(iov);
 254         iov[0].iov_base = (void *)ptr;
 255         iov[0].iov_len = nbytes;
 256         msg.msg_iov = iov;
 257         msg.msg_iovlen = 1;
 258
 259         return (sendmsg(fd, &msg, 0));
 260 }
 261
 262 static void aio_child_cleanup(struct tevent_context *event_ctx,
 263                               struct tevent_timer *te,
 264                               struct timeval now,
 265                               void *private_data)
 266 {
 267         struct aio_child_list *list = talloc_get_type_abort(
 268                 private_data, struct aio_child_list);
 269         struct aio_child *child, *next;
 270
 271         TALLOC_FREE(list->cleanup_event);
 272
 273         for (child = list->children; child != NULL; child = next) {
 274                 next = child->next;
 275
 276                 if (child->busy) {
 277                         DEBUG(10, ("child %d currently active\n",
 278                                    (int)child->pid));
 279                         continue;
 280                 }
 281
 282                 if (child->dont_delete) {
 283                         DEBUG(10, ("Child %d was active since last cleanup\n",
 284                                    (int)child->pid));
 285                         child->dont_delete = false;
 286                         continue;
 287                 }
 288
 289                 DEBUG(10, ("Child %d idle for more than 30 seconds, "
 290                            "deleting\n", (int)child->pid));
 291
 292                 TALLOC_FREE(child);
 293                 child = next;
 294         }
 295
 296         if (list->children != NULL) {
 297                 /*
 298                  * Re-schedule the next cleanup round
 299                  */
 300                 list->cleanup_event = tevent_add_timer(server_event_context(), list,
 301                                                       timeval_add(&now, 30, 0),
 302                                                       aio_child_cleanup, list);
 303
 304         }
 305 }
 306
 307 static struct aio_child_list *init_aio_children(struct vfs_handle_struct *handle)
 308 {
 309         struct aio_child_list *data = NULL;
 310
 311         if (SMB_VFS_HANDLE_TEST_DATA(handle)) {
 312                 SMB_VFS_HANDLE_GET_DATA(handle, data, struct aio_child_list,
 313                                         return NULL);
 314         }
 315
 316         if (data == NULL) {
 317                 data = talloc_zero(NULL, struct aio_child_list);
 318                 if (data == NULL) {
 319                         return NULL;
 320                 }
 321         }
 322
 323         /*
 324          * Regardless of whether the child_list had been around or not, make
 325          * sure that we have a cleanup timed event. This timed event will
 326          * delete itself when it finds that no children are around anymore.
 327          */
 328
 329         if (data->cleanup_event == NULL) {
 330                 data->cleanup_event = tevent_add_timer(server_event_context(), data,
 331                                                       timeval_current_ofs(30, 0),
 332                                                       aio_child_cleanup, data);
 333                 if (data->cleanup_event == NULL) {
 334                         TALLOC_FREE(data);
 335                         return NULL;
 336                 }
 337         }
 338
 339         if (!SMB_VFS_HANDLE_TEST_DATA(handle)) {
 340                 SMB_VFS_HANDLE_SET_DATA(handle, data, free_aio_children,
 341                                         struct aio_child_list, return False);
 342         }
 343
 344         return data;
 345 }
 346
 347 static void aio_child_loop(int sockfd, struct mmap_area *map)
 348 {
 349         while (true) {
 350                 int fd = -1;
 351                 ssize_t ret;
 352                 struct rw_cmd cmd_struct;
 353                 struct rw_ret ret_struct;
 354
 355                 ret = read_fd(sockfd, &cmd_struct, sizeof(cmd_struct), &fd);
 356                 if (ret != sizeof(cmd_struct)) {
 357                         DEBUG(10, ("read_fd returned %d: %s\n", (int)ret,
 358                                    strerror(errno)));
 359                         exit(1);
 360                 }
 361
 362                 DEBUG(10, ("aio_child_loop: %s %d bytes at %d from fd %d\n",
 363                            cmd_type_str(cmd_struct.cmd),
 364                            (int)cmd_struct.n, (int)cmd_struct.offset, fd));
 365
 366                 if (cmd_struct.erratic_testing_mode) {
 367                         /*
 368                          * For developer testing, we want erratic behaviour for
 369                          * async I/O times
 370                          */
 371                         uint8_t randval;
 372                         unsigned msecs;
 373                         /*
 374                          * use generate_random_buffer, we just forked from a
 375                          * common parent state
 376                          */
 377                         generate_random_buffer(&randval, sizeof(randval));
 378                         msecs = randval + 20;
 379                         DEBUG(10, ("delaying for %u msecs\n", msecs));
 380                         smb_msleep(msecs);
 381                 }
 382
 383                 ZERO_STRUCT(ret_struct);
 384
 385                 switch (cmd_struct.cmd) {
 386                 case READ_CMD:
 387                         ret_struct.size = sys_pread(
 388                                 fd, (void *)map->ptr, cmd_struct.n,
 389                                 cmd_struct.offset);
 390 #if 0
 391 /* This breaks "make test" when run with aio_fork module. */
 392 #ifdef DEVELOPER
 393                         ret_struct.size = MAX(1, ret_struct.size * 0.9);
 394 #endif
 395 #endif
 396                         break;
 397                 case WRITE_CMD:
 398                         ret_struct.size = sys_pwrite(
 399                                 fd, (void *)map->ptr, cmd_struct.n,
 400                                 cmd_struct.offset);
 401                         break;
 402                 case FSYNC_CMD:
 403                         ret_struct.size = fsync(fd);
 404                         break;
 405                 default:
 406                         ret_struct.size = -1;
 407                         errno = EINVAL;
 408                 }
 409
 410                 DEBUG(10, ("aio_child_loop: syscall returned %d\n",
 411                            (int)ret_struct.size));
 412
 413                 if (ret_struct.size == -1) {
 414                         ret_struct.ret_errno = errno;
 415                 }
 416
 417                 /*
 418                  * Close the fd before telling our parent we're done. The
 419                  * parent might close and re-open the file very quickly, and
 420                  * with system-level share modes (GPFS) we would get an
 421                  * unjustified SHARING_VIOLATION.
 422                  */
 423                 close(fd);
 424
 425                 ret = write_data(sockfd, (char *)&ret_struct,
 426                                  sizeof(ret_struct));
 427                 if (ret != sizeof(ret_struct)) {
 428                         DEBUG(10, ("could not write ret_struct: %s\n",
 429                                    strerror(errno)));
 430                         exit(2);
 431                 }
 432         }
 433 }
 434
 435 static int aio_child_destructor(struct aio_child *child)
 436 {
 437         char c=0;
 438
 439         SMB_ASSERT(!child->busy);
 440
 441         DEBUG(10, ("aio_child_destructor: removing child %d on fd %d\n",
 442                         child->pid, child->sockfd));
 443
 444         /*
 445          * closing the sockfd makes the child not return from recvmsg() on RHEL
 446          * 5.5 so instead force the child to exit by writing bad data to it
 447          */
 448         write(child->sockfd, &c, sizeof(c));
 449         close(child->sockfd);
 450         DLIST_REMOVE(child->list->children, child);
 451         return 0;
 452 }
 453
 454 /*
 455  * We have to close all fd's in open files, we might incorrectly hold a system
 456  * level share mode on a file.
 457  */
 458
 459 static struct files_struct *close_fsp_fd(struct files_struct *fsp,
 460                                          void *private_data)
 461 {
 462         if ((fsp->fh != NULL) && (fsp->fh->fd != -1)) {
 463                 close(fsp->fh->fd);
 464                 fsp->fh->fd = -1;
 465         }
 466         return NULL;
 467 }
 468
 469 static int create_aio_child(struct smbd_server_connection *sconn,
 470                             struct aio_child_list *children,
 471                             size_t map_size,
 472                             struct aio_child **presult)
 473 {
 474         struct aio_child *result;
 475         int fdpair[2];
 476         int ret;
 477
 478         fdpair[0] = fdpair[1] = -1;
 479
 480         result = talloc_zero(children, struct aio_child);
 481         if (result == NULL) {
 482                 return ENOMEM;
 483         }
 484
 485         if (socketpair(AF_UNIX, SOCK_STREAM, 0, fdpair) == -1) {
 486                 ret = errno;
 487                 DEBUG(10, ("socketpair() failed: %s\n", strerror(errno)));
 488                 goto fail;
 489         }
 490
 491         DEBUG(10, ("fdpair = %d/%d\n", fdpair[0], fdpair[1]));
 492
 493         result->map = mmap_area_init(result, map_size);
 494         if (result->map == NULL) {
 495                 ret = errno;
 496                 DEBUG(0, ("Could not create mmap area\n"));
 497                 goto fail;
 498         }
 499
 500         result->pid = fork();
 501         if (result->pid == -1) {
 502                 ret = errno;
 503                 DEBUG(0, ("fork failed: %s\n", strerror(errno)));
 504                 goto fail;
 505         }
 506
 507         if (result->pid == 0) {
 508                 close(fdpair[0]);
 509                 result->sockfd = fdpair[1];
 510                 files_forall(sconn, close_fsp_fd, NULL);
 511                 aio_child_loop(result->sockfd, result->map);
 512         }
 513
 514         DEBUG(10, ("Child %d created with sockfd %d\n",
 515                         result->pid, fdpair[0]));
 516
 517         result->sockfd = fdpair[0];
 518         close(fdpair[1]);
 519
 520         result->list = children;
 521         DLIST_ADD(children->children, result);
 522
 523         talloc_set_destructor(result, aio_child_destructor);
 524
 525         *presult = result;
 526
 527         return 0;
 528
 529  fail:
 530         if (fdpair[0] != -1) close(fdpair[0]);
 531         if (fdpair[1] != -1) close(fdpair[1]);
 532         TALLOC_FREE(result);
 533
 534         return ret;
 535 }
 536
 537 static int get_idle_child(struct vfs_handle_struct *handle,
 538                           struct aio_child **pchild)
 539 {
 540         struct aio_child_list *children;
 541         struct aio_child *child;
 542
 543         children = init_aio_children(handle);
 544         if (children == NULL) {
 545                 return ENOMEM;
 546         }
 547
 548         for (child = children->children; child != NULL; child = child->next) {
 549                 if (!child->busy) {
 550                         break;
 551                 }
 552         }
 553
 554         if (child == NULL) {
 555                 int ret;
 556
 557                 DEBUG(10, ("no idle child found, creating new one\n"));
 558
 559                 ret = create_aio_child(handle->conn->sconn, children,
 560                                           128*1024, &child);
 561                 if (ret != 0) {
 562                         DEBUG(10, ("create_aio_child failed: %s\n",
 563                                    strerror(errno)));
 564                         return ret;
 565                 }
 566         }
 567
 568         child->dont_delete = true;
 569         child->busy = true;
 570
 571         *pchild = child;
 572         return 0;
 573 }
 574
 575 struct aio_fork_pread_state {
 576         struct aio_child *child;
 577         ssize_t ret;
 578         int err;
 579 };
 580
 581 static void aio_fork_pread_done(struct tevent_req *subreq);
 582
 583 static struct tevent_req *aio_fork_pread_send(struct vfs_handle_struct *handle,
 584                                               TALLOC_CTX *mem_ctx,
 585                                               struct tevent_context *ev,
 586                                               struct files_struct *fsp,
 587                                               void *data,
 588                                               size_t n, off_t offset)
 589 {
 590         struct tevent_req *req, *subreq;
 591         struct aio_fork_pread_state *state;
 592         struct rw_cmd cmd;
 593         ssize_t written;
 594         int err;
 595         struct aio_fork_config *config;
 596
 597         SMB_VFS_HANDLE_GET_DATA(handle, config,
 598                                 struct aio_fork_config,
 599                                 return NULL);
 600
 601         req = tevent_req_create(mem_ctx, &state, struct aio_fork_pread_state);
 602         if (req == NULL) {
 603                 return NULL;
 604         }
 605
 606         if (n > 128*1024) {
 607                 /* TODO: support variable buffers */
 608                 tevent_req_error(req, EINVAL);
 609                 return tevent_req_post(req, ev);
 610         }
 611
 612         err = get_idle_child(handle, &state->child);
 613         if (err != 0) {
 614                 tevent_req_error(req, err);
 615                 return tevent_req_post(req, ev);
 616         }
 617
 618         ZERO_STRUCT(cmd);
 619         cmd.n = n;
 620         cmd.offset = offset;
 621         cmd.cmd = READ_CMD;
 622         cmd.erratic_testing_mode = config->erratic_testing_mode;
 623
 624         DEBUG(10, ("sending fd %d to child %d\n", fsp->fh->fd,
 625                    (int)state->child->pid));
 626
 627         /*
 628          * Not making this async. We're writing into an empty unix
 629          * domain socket. This should never block.
 630          */
 631         written = write_fd(state->child->sockfd, &cmd, sizeof(cmd),
 632                            fsp->fh->fd);
 633         if (written == -1) {
 634                 err = errno;
 635
 636                 TALLOC_FREE(state->child);
 637
 638                 DEBUG(10, ("write_fd failed: %s\n", strerror(err)));
 639                 tevent_req_error(req, err);
 640                 return tevent_req_post(req, ev);
 641         }
 642
 643         subreq = read_packet_send(state, ev, state->child->sockfd,
 644                                   sizeof(struct rw_ret), NULL, NULL);
 645         if (tevent_req_nomem(subreq, req)) {
 646                 TALLOC_FREE(state->child); /* we sent sth down */
 647                 return tevent_req_post(req, ev);
 648         }
 649         tevent_req_set_callback(subreq, aio_fork_pread_done, req);
 650         return req;
 651 }
 652
 653 static void aio_fork_pread_done(struct tevent_req *subreq)
 654 {
 655         struct tevent_req *req = tevent_req_callback_data(
 656                 subreq, struct tevent_req);
 657         struct aio_fork_pread_state *state = tevent_req_data(
 658                 req, struct aio_fork_pread_state);
 659         ssize_t nread;
 660         uint8_t *buf;
 661         int err;
 662         struct rw_ret *retbuf;
 663
 664         nread = read_packet_recv(subreq, talloc_tos(), &buf, &err);
 665         TALLOC_FREE(subreq);
 666         if (nread == -1) {
 667                 TALLOC_FREE(state->child);
 668                 tevent_req_error(req, err);
 669                 return;
 670         }
 671
 672         state->child->busy = false;
 673
 674         retbuf = (struct rw_ret *)buf;
 675         state->ret = retbuf->size;
 676         state->err = retbuf->ret_errno;
 677         tevent_req_done(req);
 678 }
 679
 680 static ssize_t aio_fork_pread_recv(struct tevent_req *req, int *err)
 681 {
 682         struct aio_fork_pread_state *state = tevent_req_data(
 683                 req, struct aio_fork_pread_state);
 684
 685         if (tevent_req_is_unix_error(req, err)) {
 686                 return -1;
 687         }
 688         if (state->ret == -1) {
 689                 *err = state->err;
 690         }
 691         return state->ret;
 692 }
 693
 694 struct aio_fork_pwrite_state {
 695         struct aio_child *child;
 696         ssize_t ret;
 697         int err;
 698 };
 699
 700 static void aio_fork_pwrite_done(struct tevent_req *subreq);
 701
 702 static struct tevent_req *aio_fork_pwrite_send(
 703         struct vfs_handle_struct *handle, TALLOC_CTX *mem_ctx,
 704         struct tevent_context *ev, struct files_struct *fsp,
 705         const void *data, size_t n, off_t offset)
 706 {
 707         struct tevent_req *req, *subreq;
 708         struct aio_fork_pwrite_state *state;
 709         struct rw_cmd cmd;
 710         ssize_t written;
 711         int err;
 712         struct aio_fork_config *config;
 713         SMB_VFS_HANDLE_GET_DATA(handle, config,
 714                                 struct aio_fork_config,
 715                                 return NULL);
 716
 717         req = tevent_req_create(mem_ctx, &state, struct aio_fork_pwrite_state);
 718         if (req == NULL) {
 719                 return NULL;
 720         }
 721
 722         if (n > 128*1024) {
 723                 /* TODO: support variable buffers */
 724                 tevent_req_error(req, EINVAL);
 725                 return tevent_req_post(req, ev);
 726         }
 727
 728         err = get_idle_child(handle, &state->child);
 729         if (err != 0) {
 730                 tevent_req_error(req, err);
 731                 return tevent_req_post(req, ev);
 732         }
 733
 734         ZERO_STRUCT(cmd);
 735         cmd.n = n;
 736         cmd.offset = offset;
 737         cmd.cmd = WRITE_CMD;
 738         cmd.erratic_testing_mode = config->erratic_testing_mode;
 739
 740         DEBUG(10, ("sending fd %d to child %d\n", fsp->fh->fd,
 741                    (int)state->child->pid));
 742
 743         /*
 744          * Not making this async. We're writing into an empty unix
 745          * domain socket. This should never block.
 746          */
 747         written = write_fd(state->child->sockfd, &cmd, sizeof(cmd),
 748                            fsp->fh->fd);
 749         if (written == -1) {
 750                 err = errno;
 751
 752                 TALLOC_FREE(state->child);
 753
 754                 DEBUG(10, ("write_fd failed: %s\n", strerror(err)));
 755                 tevent_req_error(req, err);
 756                 return tevent_req_post(req, ev);
 757         }
 758
 759         subreq = read_packet_send(state, ev, state->child->sockfd,
 760                                   sizeof(struct rw_ret), NULL, NULL);
 761         if (tevent_req_nomem(subreq, req)) {
 762                 TALLOC_FREE(state->child); /* we sent sth down */
 763                 return tevent_req_post(req, ev);
 764         }
 765         tevent_req_set_callback(subreq, aio_fork_pwrite_done, req);
 766         return req;
 767 }
 768
 769 static void aio_fork_pwrite_done(struct tevent_req *subreq)
 770 {
 771         struct tevent_req *req = tevent_req_callback_data(
 772                 subreq, struct tevent_req);
 773         struct aio_fork_pwrite_state *state = tevent_req_data(
 774                 req, struct aio_fork_pwrite_state);
 775         ssize_t nread;
 776         uint8_t *buf;
 777         int err;
 778         struct rw_ret *retbuf;
 779
 780         nread = read_packet_recv(subreq, talloc_tos(), &buf, &err);
 781         TALLOC_FREE(subreq);
 782         if (nread == -1) {
 783                 TALLOC_FREE(state->child);
 784                 tevent_req_error(req, err);
 785                 return;
 786         }
 787
 788         state->child->busy = false;
 789
 790         retbuf = (struct rw_ret *)buf;
 791         state->ret = retbuf->size;
 792         state->err = retbuf->ret_errno;
 793         tevent_req_done(req);
 794 }
 795
 796 static ssize_t aio_fork_pwrite_recv(struct tevent_req *req, int *err)
 797 {
 798         struct aio_fork_pwrite_state *state = tevent_req_data(
 799                 req, struct aio_fork_pwrite_state);
 800
 801         if (tevent_req_is_unix_error(req, err)) {
 802                 return -1;
 803         }
 804         if (state->ret == -1) {
 805                 *err = state->err;
 806         }
 807         return state->ret;
 808 }
 809
 810 struct aio_fork_fsync_state {
 811         struct aio_child *child;
 812         ssize_t ret;
 813         int err;
 814 };
 815
 816 static void aio_fork_fsync_done(struct tevent_req *subreq);
 817
 818 static struct tevent_req *aio_fork_fsync_send(
 819         struct vfs_handle_struct *handle, TALLOC_CTX *mem_ctx,
 820         struct tevent_context *ev, struct files_struct *fsp)
 821 {
 822         struct tevent_req *req, *subreq;
 823         struct aio_fork_fsync_state *state;
 824         struct rw_cmd cmd;
 825         ssize_t written;
 826         int err;
 827         struct aio_fork_config *config;
 828
 829         SMB_VFS_HANDLE_GET_DATA(handle, config,
 830                                 struct aio_fork_config,
 831                                 return NULL);
 832
 833         req = tevent_req_create(mem_ctx, &state, struct aio_fork_fsync_state);
 834         if (req == NULL) {
 835                 return NULL;
 836         }
 837
 838         err = get_idle_child(handle, &state->child);
 839         if (err != 0) {
 840                 tevent_req_error(req, err);
 841                 return tevent_req_post(req, ev);
 842         }
 843
 844         ZERO_STRUCT(cmd);
 845         cmd.cmd = FSYNC_CMD;
 846         cmd.erratic_testing_mode = config->erratic_testing_mode;
 847
 848         DEBUG(10, ("sending fd %d to child %d\n", fsp->fh->fd,
 849                    (int)state->child->pid));
 850
 851         /*
 852          * Not making this async. We're writing into an empty unix
 853          * domain socket. This should never block.
 854          */
 855         written = write_fd(state->child->sockfd, &cmd, sizeof(cmd),
 856                            fsp->fh->fd);
 857         if (written == -1) {
 858                 err = errno;
 859
 860                 TALLOC_FREE(state->child);
 861
 862                 DEBUG(10, ("write_fd failed: %s\n", strerror(err)));
 863                 tevent_req_error(req, err);
 864                 return tevent_req_post(req, ev);
 865         }
 866
 867         subreq = read_packet_send(state, ev, state->child->sockfd,
 868                                   sizeof(struct rw_ret), NULL, NULL);
 869         if (tevent_req_nomem(subreq, req)) {
 870                 TALLOC_FREE(state->child); /* we sent sth down */
 871                 return tevent_req_post(req, ev);
 872         }
 873         tevent_req_set_callback(subreq, aio_fork_fsync_done, req);
 874         return req;
 875 }
 876
 877 static void aio_fork_fsync_done(struct tevent_req *subreq)
 878 {
 879         struct tevent_req *req = tevent_req_callback_data(
 880                 subreq, struct tevent_req);
 881         struct aio_fork_fsync_state *state = tevent_req_data(
 882                 req, struct aio_fork_fsync_state);
 883         ssize_t nread;
 884         uint8_t *buf;
 885         int err;
 886         struct rw_ret *retbuf;
 887
 888         nread = read_packet_recv(subreq, talloc_tos(), &buf, &err);
 889         TALLOC_FREE(subreq);
 890         if (nread == -1) {
 891                 TALLOC_FREE(state->child);
 892                 tevent_req_error(req, err);
 893                 return;
 894         }
 895
 896         state->child->busy = false;
 897
 898         retbuf = (struct rw_ret *)buf;
 899         state->ret = retbuf->size;
 900         state->err = retbuf->ret_errno;
 901         tevent_req_done(req);
 902 }
 903
 904 static int aio_fork_fsync_recv(struct tevent_req *req, int *err)
 905 {
 906         struct aio_fork_fsync_state *state = tevent_req_data(
 907                 req, struct aio_fork_fsync_state);
 908
 909         if (tevent_req_is_unix_error(req, err)) {
 910                 return -1;
 911         }
 912         if (state->ret == -1) {
 913                 *err = state->err;
 914         }
 915         return state->ret;
 916 }
 917
 918 static int aio_fork_connect(vfs_handle_struct *handle, const char *service,
 919                             const char *user)
 920 {
 921         int ret;
 922         struct aio_fork_config *config;
 923         ret = SMB_VFS_NEXT_CONNECT(handle, service, user);
 924
 925         if (ret < 0) {
 926                 return ret;
 927         }
 928
 929         config = talloc_zero(handle->conn, struct aio_fork_config);
 930         if (!config) {
 931                 SMB_VFS_NEXT_DISCONNECT(handle);
 932                 DEBUG(0, ("talloc_zero() failed\n"));
 933                 return -1;
 934         }
 935
 936         config->erratic_testing_mode = lp_parm_bool(SNUM(handle->conn), "vfs_aio_fork",
 937                                                     "erratic_testing_mode", false);
 938
 939         SMB_VFS_HANDLE_SET_DATA(handle, config,
 940                                 NULL, struct aio_fork_config,
 941                                 return -1);
 942
 943         /*********************************************************************
 944          * How many threads to initialize ?
 945          * 100 per process seems insane as a default until you realize that
 946          * (a) Threads terminate after 1 second when idle.
 947          * (b) Throttling is done in SMB2 via the crediting algorithm.
 948          * (c) SMB1 clients are limited to max_mux (50) outstanding
 949          *     requests and Windows clients don't use this anyway.
 950          * Essentially we want this to be unlimited unless smb.conf
 951          * says different.
 952          *********************************************************************/
 953         aio_pending_size = 100;
 954         return 0;
 955 }
 956
 957 static struct vfs_fn_pointers vfs_aio_fork_fns = {
 958         .connect_fn = aio_fork_connect,
 959         .pread_send_fn = aio_fork_pread_send,
 960         .pread_recv_fn = aio_fork_pread_recv,
 961         .pwrite_send_fn = aio_fork_pwrite_send,
 962         .pwrite_recv_fn = aio_fork_pwrite_recv,
 963         .fsync_send_fn = aio_fork_fsync_send,
 964         .fsync_recv_fn = aio_fork_fsync_recv,
 965 };
 966
 967 NTSTATUS vfs_aio_fork_init(void);
 968 NTSTATUS vfs_aio_fork_init(void)
 969 {
 970         return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
 971                                 "aio_fork", &vfs_aio_fork_fns);
 972 }