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