root / xseg / peers / user / glusterd.c @ feature-glusterd
History | View | Annotate | Download (33.2 kB)
| 1 |
/*
|
|---|---|
| 2 |
* Copyright 2012 GRNET S.A. All rights reserved.
|
| 3 |
*
|
| 4 |
* Redistribution and use in source and binary forms, with or
|
| 5 |
* without modification, are permitted provided that the following
|
| 6 |
* conditions are met:
|
| 7 |
*
|
| 8 |
* 1. Redistributions of source code must retain the above
|
| 9 |
* copyright notice, this list of conditions and the following
|
| 10 |
* disclaimer.
|
| 11 |
* 2. Redistributions in binary form must reproduce the above
|
| 12 |
* copyright notice, this list of conditions and the following
|
| 13 |
* disclaimer in the documentation and/or other materials
|
| 14 |
* provided with the distribution.
|
| 15 |
*
|
| 16 |
* THIS SOFTWARE IS PROVIDED BY GRNET S.A. ``AS IS'' AND ANY EXPRESS
|
| 17 |
* OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
|
| 18 |
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
| 19 |
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL GRNET S.A OR
|
| 20 |
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
| 21 |
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
| 22 |
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
|
| 23 |
* USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
|
| 24 |
* AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
| 25 |
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
|
| 26 |
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
| 27 |
* POSSIBILITY OF SUCH DAMAGE.
|
| 28 |
*
|
| 29 |
* The views and conclusions contained in the software and
|
| 30 |
* documentation are those of the authors and should not be
|
| 31 |
* interpreted as representing official policies, either expressed
|
| 32 |
* or implied, of GRNET S.A.
|
| 33 |
*/
|
| 34 |
|
| 35 |
#include <stdio.h> |
| 36 |
#include <stdlib.h> |
| 37 |
#include <unistd.h> |
| 38 |
#include <xseg/xseg.h> |
| 39 |
#include <peer.h> |
| 40 |
#include <xseg/protocol.h> |
| 41 |
#include <glusterfs/api/glfs.h> |
| 42 |
#include <pthread.h> |
| 43 |
#include <ctype.h> |
| 44 |
#include <errno.h> |
| 45 |
#include <hash.h> |
| 46 |
#include <sys/types.h> |
| 47 |
#include <unistd.h> |
| 48 |
|
| 49 |
|
| 50 |
#define MAX_UNIQUESTR_LEN 128 |
| 51 |
#define LOCK_SUFFIX "_lock" |
| 52 |
#define LOCK_SUFFIX_LEN 5 |
| 53 |
#define HASH_SUFFIX "_hash" |
| 54 |
#define HASH_SUFFIX_LEN 5 |
| 55 |
|
| 56 |
#define MAX_OBJ_NAME (XSEG_MAX_TARGETLEN + LOCK_SUFFIX_LEN + 1) |
| 57 |
#define MAX_GLFS_ARG_LEN 64 |
| 58 |
|
| 59 |
void custom_peer_usage()
|
| 60 |
{
|
| 61 |
fprintf(stderr, "Custom peer options:\n"
|
| 62 |
" Option | Default | \n"
|
| 63 |
" --------------------------------------------\n"
|
| 64 |
" --transport | tcp | Transport type [tcp|rdma|unix]\n"
|
| 65 |
" --server | 127.0.0.1 | A server that is part of the\n"
|
| 66 |
" | | volume\n"
|
| 67 |
" --port | 0(=24007) | Port of server's gluster daemon\n"
|
| 68 |
" --volume | None | Gluster volume to connect to\n"
|
| 69 |
" --uniquestr | pid | Unique string for this instance\n"
|
| 70 |
" --async | 0 | If 1, glusted will use the async\n"
|
| 71 |
" | | equivalent functions\n"
|
| 72 |
"\n"
|
| 73 |
); |
| 74 |
} |
| 75 |
|
| 76 |
#define REQ_UNDEFINED -2 |
| 77 |
#define REQ_FAILED -1 |
| 78 |
#define REQ_SUBMITTED 0 |
| 79 |
#define REQ_COMPLETED 1 |
| 80 |
|
| 81 |
enum gluster_state {
|
| 82 |
ACCEPTED = 0,
|
| 83 |
PENDING = 1,
|
| 84 |
READING = 2,
|
| 85 |
WRITING = 3,
|
| 86 |
STATING = 4,
|
| 87 |
PREHASHING = 5,
|
| 88 |
POSTHASHING = 6
|
| 89 |
}; |
| 90 |
|
| 91 |
struct glusterd {
|
| 92 |
glfs_t *glfs; |
| 93 |
char uniquestr[MAX_UNIQUESTR_LEN + 1]; |
| 94 |
int uniquestr_len;
|
| 95 |
int async;
|
| 96 |
}; |
| 97 |
|
| 98 |
struct gluster_io{
|
| 99 |
char obj_name[MAX_OBJ_NAME + 1]; |
| 100 |
glfs_fd_t *fd; |
| 101 |
enum gluster_state state;
|
| 102 |
uint64_t size; |
| 103 |
char *second_name, *buf;
|
| 104 |
uint64_t read; |
| 105 |
}; |
| 106 |
|
| 107 |
static inline struct glusterd *__get_gluster(struct peerd *peer) |
| 108 |
{
|
| 109 |
return peer->priv;
|
| 110 |
} |
| 111 |
|
| 112 |
static inline int __set_glfs(struct glusterd *gluster, char *volume) |
| 113 |
{
|
| 114 |
gluster->glfs = glfs_new(volume); |
| 115 |
|
| 116 |
if (!gluster->glfs)
|
| 117 |
return -1; |
| 118 |
|
| 119 |
return 0; |
| 120 |
} |
| 121 |
|
| 122 |
static inline glfs_t *__get_glfs(struct glusterd *gluster) |
| 123 |
{
|
| 124 |
return gluster->glfs;
|
| 125 |
} |
| 126 |
|
| 127 |
int handle_read(struct peerd *peer, struct peer_req *pr); |
| 128 |
int handle_write(struct peerd *peer, struct peer_req *pr); |
| 129 |
|
| 130 |
|
| 131 |
static void gluster_complete_read(struct glfs_fd *fd, ssize_t ret, void *data) |
| 132 |
{
|
| 133 |
struct peer_req *pr = (struct peer_req*)data; |
| 134 |
struct peerd *peer = pr->peer;
|
| 135 |
|
| 136 |
pr->retval = ret; |
| 137 |
dispatch(peer, pr, pr->req, dispatch_internal); |
| 138 |
} |
| 139 |
|
| 140 |
static void gluster_complete_write(struct glfs_fd *fd, ssize_t ret, void *data) |
| 141 |
{
|
| 142 |
struct peer_req *pr = (struct peer_req*)data; |
| 143 |
struct peerd *peer = pr->peer;
|
| 144 |
|
| 145 |
pr->retval = ret; |
| 146 |
dispatch(peer, pr, pr->req, dispatch_internal); |
| 147 |
} |
| 148 |
|
| 149 |
static void submit_hook(struct peer_req *pr) |
| 150 |
{
|
| 151 |
struct peerd *peer = pr->peer;
|
| 152 |
struct glusterd *gluster = __get_gluster(peer);
|
| 153 |
|
| 154 |
if (!gluster->async)
|
| 155 |
dispatch(peer, pr, pr->req, dispatch_internal); |
| 156 |
} |
| 157 |
|
| 158 |
static void create_hash_name(struct peer_req *pr, char *hash_name) |
| 159 |
{
|
| 160 |
struct gluster_io *gio = (struct gluster_io *) pr->priv; |
| 161 |
int pos = 0; |
| 162 |
|
| 163 |
strncpy(hash_name, gio->obj_name, strlen(gio->obj_name)); |
| 164 |
pos += strlen(gio->obj_name); |
| 165 |
strncpy(hash_name + pos, HASH_SUFFIX, HASH_SUFFIX_LEN); |
| 166 |
pos += HASH_SUFFIX_LEN; |
| 167 |
hash_name[pos] = 0;
|
| 168 |
} |
| 169 |
|
| 170 |
static int allocate_gio_secname(struct peer_req *pr) |
| 171 |
{
|
| 172 |
struct gluster_io *gio = (struct gluster_io *) pr->priv; |
| 173 |
|
| 174 |
gio->second_name = malloc(MAX_OBJ_NAME + 1);
|
| 175 |
if (!gio->second_name)
|
| 176 |
return -1; |
| 177 |
|
| 178 |
return 0; |
| 179 |
} |
| 180 |
|
| 181 |
static int allocate_gio_buf(struct peer_req *pr) |
| 182 |
{
|
| 183 |
struct gluster_io *gio = (struct gluster_io *) pr->priv; |
| 184 |
struct xseg_request *req = pr->req;
|
| 185 |
|
| 186 |
gio->buf = malloc(req->size); |
| 187 |
if (!gio->buf)
|
| 188 |
return -1; |
| 189 |
|
| 190 |
return 0; |
| 191 |
} |
| 192 |
|
| 193 |
static int prepare_copy(struct peer_req *pr) |
| 194 |
{
|
| 195 |
struct peerd *peer = pr->peer;
|
| 196 |
struct gluster_io *gio = (struct gluster_io *) pr->priv; |
| 197 |
char *data = xseg_get_data(peer->xseg, pr->req);
|
| 198 |
struct xseg_request_copy *xcopy = (struct xseg_request_copy *)data; |
| 199 |
unsigned int end = (xcopy->targetlen > MAX_OBJ_NAME) ? |
| 200 |
MAX_OBJ_NAME : xcopy->targetlen; |
| 201 |
int r;
|
| 202 |
|
| 203 |
r = allocate_gio_secname(pr); |
| 204 |
if (r < 0) |
| 205 |
return r;
|
| 206 |
|
| 207 |
/* FIXME: terminate or fail if targetlen > MAX_OBJ_NAME ? */
|
| 208 |
strncpy(gio->second_name, xcopy->target, end); |
| 209 |
gio->second_name[end] = 0;
|
| 210 |
gio->read = 0;
|
| 211 |
|
| 212 |
r = allocate_gio_buf(pr); |
| 213 |
return r;
|
| 214 |
} |
| 215 |
|
| 216 |
static int prepare_hash(struct peer_req *pr, char *hash_name) |
| 217 |
{
|
| 218 |
int r;
|
| 219 |
|
| 220 |
r = allocate_gio_secname(pr); |
| 221 |
if (r < 0) |
| 222 |
return r;
|
| 223 |
|
| 224 |
create_hash_name(pr, hash_name); |
| 225 |
|
| 226 |
r = allocate_gio_buf(pr); |
| 227 |
return r;
|
| 228 |
} |
| 229 |
|
| 230 |
static int write_lock_owner(struct peer_req *pr, glfs_fd_t *fd) |
| 231 |
{
|
| 232 |
struct peerd *peer = pr->peer;
|
| 233 |
struct glusterd *gluster = __get_gluster(peer);
|
| 234 |
char *me = gluster->uniquestr;
|
| 235 |
int len = strlen(me);
|
| 236 |
int serviced = 0; |
| 237 |
int r = 0; |
| 238 |
|
| 239 |
while (len > serviced) {
|
| 240 |
r = glfs_pwrite(fd, me + serviced, len - serviced, |
| 241 |
0 + serviced, 0); |
| 242 |
if (r > 0) |
| 243 |
serviced += r; |
| 244 |
else if (errno != EINTR) |
| 245 |
return -1; |
| 246 |
} |
| 247 |
|
| 248 |
return 0; |
| 249 |
} |
| 250 |
|
| 251 |
static int read_lock_owner(struct peer_req *pr, glfs_fd_t *fd, char *owner) |
| 252 |
{
|
| 253 |
struct peerd *peer = pr->peer;
|
| 254 |
struct glusterd *gluster = __get_gluster(peer);
|
| 255 |
char *buf = gluster->uniquestr;
|
| 256 |
int len = strlen(buf);
|
| 257 |
int serviced = 0; |
| 258 |
int r = 0; |
| 259 |
|
| 260 |
while (len > serviced) {
|
| 261 |
r = glfs_pread(fd, owner + serviced, len - serviced, |
| 262 |
0 + serviced, 0); |
| 263 |
if (r > 0) |
| 264 |
serviced += r; |
| 265 |
else if (r == 0) |
| 266 |
break;
|
| 267 |
else if (errno != EINTR) |
| 268 |
return -1; |
| 269 |
} |
| 270 |
|
| 271 |
return serviced;
|
| 272 |
} |
| 273 |
|
| 274 |
static int validate_lock_owner(struct peer_req *pr, glfs_fd_t *fd) |
| 275 |
{
|
| 276 |
struct peerd *peer = pr->peer;
|
| 277 |
struct glusterd *gluster = __get_gluster(peer);
|
| 278 |
char lock_owner[MAX_UNIQUESTR_LEN + 1]; |
| 279 |
char *me = gluster->uniquestr;
|
| 280 |
int len = strlen(me);
|
| 281 |
int r = 0; |
| 282 |
|
| 283 |
r = read_lock_owner(pr, fd, lock_owner); |
| 284 |
if (r < 0) { |
| 285 |
XSEGLOG2(&lc, E, "Failed to read lock owner of fd %p", fd);
|
| 286 |
return -1; |
| 287 |
} |
| 288 |
|
| 289 |
if (r != len || strncmp(me, lock_owner, r) != 0) { |
| 290 |
XSEGLOG2(&lc, D, "Lock owner is %s while we are %s",
|
| 291 |
lock_owner, me); |
| 292 |
return -1; |
| 293 |
} |
| 294 |
|
| 295 |
return 0; |
| 296 |
} |
| 297 |
|
| 298 |
static int do_aio_generic(struct peer_req *pr, uint32_t op, |
| 299 |
char *buf, uint64_t size, uint64_t offset)
|
| 300 |
{
|
| 301 |
struct peerd *peer = pr->peer;
|
| 302 |
struct glusterd *gluster = __get_gluster(peer);
|
| 303 |
struct gluster_io *gio = (struct gluster_io *) pr->priv; |
| 304 |
glfs_fd_t *fd = gio->fd; |
| 305 |
int r;
|
| 306 |
|
| 307 |
switch (op) {
|
| 308 |
case X_READ:
|
| 309 |
if (gluster->async) {
|
| 310 |
r = glfs_pread_async(fd, buf, size, offset, 0,
|
| 311 |
gluster_complete_read, pr); |
| 312 |
} else {
|
| 313 |
r = glfs_pread(fd, buf, size, offset, 0);
|
| 314 |
if (r >= 0) |
| 315 |
pr->retval = r; |
| 316 |
} |
| 317 |
break;
|
| 318 |
case X_WRITE:
|
| 319 |
if (gluster->async) {
|
| 320 |
r = glfs_pwrite_async(fd, buf, size, offset, 0,
|
| 321 |
gluster_complete_write, pr); |
| 322 |
} else {
|
| 323 |
r = glfs_pwrite(fd, buf, size, offset, 0);
|
| 324 |
if (r >= 0) |
| 325 |
pr->retval = r; |
| 326 |
} |
| 327 |
break;
|
| 328 |
default:
|
| 329 |
return -1; |
| 330 |
break;
|
| 331 |
} |
| 332 |
return r;
|
| 333 |
} |
| 334 |
|
| 335 |
static int begin_aio_read(struct peer_req *pr, char *buf, |
| 336 |
uint64_t size, uint64_t offset) |
| 337 |
{
|
| 338 |
int r = 0; |
| 339 |
|
| 340 |
r = do_aio_generic(pr, X_READ, buf, size, offset); |
| 341 |
if (r >= 0) |
| 342 |
return REQ_SUBMITTED;
|
| 343 |
else
|
| 344 |
return REQ_FAILED;
|
| 345 |
} |
| 346 |
|
| 347 |
static int begin_aio_write(struct peer_req *pr, char *buf, |
| 348 |
uint64_t size, uint64_t offset) |
| 349 |
{
|
| 350 |
int r = 0; |
| 351 |
|
| 352 |
r = do_aio_generic(pr, X_WRITE, buf, size, offset); |
| 353 |
if (r >= 0) |
| 354 |
return REQ_SUBMITTED;
|
| 355 |
else
|
| 356 |
return REQ_FAILED;
|
| 357 |
} |
| 358 |
|
| 359 |
static int complete_aio_read(struct peer_req *pr, char *buf, |
| 360 |
uint64_t size, uint64_t offset, uint64_t *serviced) |
| 361 |
{
|
| 362 |
int r = 0; |
| 363 |
|
| 364 |
/* Leave on fail or if there are no other data */
|
| 365 |
if (pr->retval < 0) { |
| 366 |
/* TODO: check errors */
|
| 367 |
return REQ_FAILED;
|
| 368 |
} else if (pr->retval == 0) { |
| 369 |
XSEGLOG2(&lc, I, "Zeroing rest of data");
|
| 370 |
memset(buf + *serviced, 0, size - *serviced);
|
| 371 |
*serviced = size; |
| 372 |
return REQ_COMPLETED;
|
| 373 |
} |
| 374 |
|
| 375 |
/*
|
| 376 |
* Else, check if all data have been served and resubmit if necessary
|
| 377 |
*/
|
| 378 |
*serviced += pr->retval; |
| 379 |
if (*serviced == size) {
|
| 380 |
return REQ_COMPLETED;
|
| 381 |
} else {
|
| 382 |
r = do_aio_generic(pr, X_READ, buf + *serviced, |
| 383 |
size - *serviced, offset + *serviced); |
| 384 |
if (r >= 0) |
| 385 |
return REQ_SUBMITTED;
|
| 386 |
else
|
| 387 |
return REQ_FAILED;
|
| 388 |
} |
| 389 |
} |
| 390 |
|
| 391 |
static int complete_aio_write(struct peer_req *pr, char *buf, |
| 392 |
uint64_t size, uint64_t offset, uint64_t *serviced) |
| 393 |
{
|
| 394 |
int r = 0; |
| 395 |
|
| 396 |
/* Leave on fail or if there are no other data */
|
| 397 |
if (pr->retval < 0){ |
| 398 |
/* TODO: check errors */
|
| 399 |
return REQ_FAILED;
|
| 400 |
} |
| 401 |
|
| 402 |
/*
|
| 403 |
* Else, check if all data have been served and resubmit if necessary
|
| 404 |
*/
|
| 405 |
*serviced += pr->retval; |
| 406 |
if (*serviced == size) {
|
| 407 |
return REQ_COMPLETED;
|
| 408 |
} else {
|
| 409 |
r = do_aio_generic(pr, X_WRITE, buf + *serviced, |
| 410 |
size - *serviced, offset + *serviced); |
| 411 |
if (r >= 0) |
| 412 |
return REQ_SUBMITTED;
|
| 413 |
else
|
| 414 |
return REQ_FAILED;
|
| 415 |
} |
| 416 |
|
| 417 |
return r;
|
| 418 |
} |
| 419 |
|
| 420 |
static glfs_fd_t *do_block_create(struct peer_req *pr, char *target, int mode) |
| 421 |
{
|
| 422 |
struct peerd *peer = pr->peer;
|
| 423 |
struct glusterd *gluster = __get_gluster(peer);
|
| 424 |
glfs_t *glfs = __get_glfs(gluster); |
| 425 |
glfs_fd_t *fd = NULL;
|
| 426 |
|
| 427 |
/*
|
| 428 |
* Create the requested file (in O_EXCL mode if requested)
|
| 429 |
* If errno is EINTR, retry. If it is other that EEXIST or EINTR,
|
| 430 |
* leave.
|
| 431 |
*/
|
| 432 |
XSEGLOG2(&lc, D, "Creating target %s", target);
|
| 433 |
do {
|
| 434 |
errno = 0;
|
| 435 |
fd = glfs_creat(glfs, target, |
| 436 |
O_RDWR | O_CREAT | O_TRUNC | mode, |
| 437 |
S_IRUSR | S_IWUSR); |
| 438 |
|
| 439 |
if (fd)
|
| 440 |
return fd;
|
| 441 |
if (errno != EEXIST && errno != EINTR) {
|
| 442 |
XSEGLOG2(&lc, E, "Unexpected error (errno %d) while "
|
| 443 |
"creating %s", errno, target);
|
| 444 |
return fd;
|
| 445 |
} |
| 446 |
} while (errno == EINTR);
|
| 447 |
|
| 448 |
return fd;
|
| 449 |
} |
| 450 |
|
| 451 |
static glfs_fd_t *do_block_open(struct peer_req *pr, char *target, int mode) |
| 452 |
{
|
| 453 |
struct peerd *peer = pr->peer;
|
| 454 |
struct glusterd *gluster = __get_gluster(peer);
|
| 455 |
glfs_t *glfs = __get_glfs(gluster); |
| 456 |
glfs_fd_t *fd; |
| 457 |
|
| 458 |
/*
|
| 459 |
* Open the requested file.
|
| 460 |
* If errno is EINTR, retry. If it is other that ENOENT or EINTR,
|
| 461 |
* leave.
|
| 462 |
*/
|
| 463 |
XSEGLOG2(&lc, D, "Opening target %s", target);
|
| 464 |
do {
|
| 465 |
errno = 0;
|
| 466 |
fd = glfs_open(glfs, target, O_RDWR); |
| 467 |
|
| 468 |
if (fd)
|
| 469 |
return fd;
|
| 470 |
if (errno != ENOENT && errno != EINTR) {
|
| 471 |
XSEGLOG2(&lc, E, "Unexpected error (errno %d) while "
|
| 472 |
"opening %s:", errno, target);
|
| 473 |
return fd;
|
| 474 |
} |
| 475 |
} while (errno == EINTR);
|
| 476 |
|
| 477 |
if (!fd && !(mode & O_CREAT))
|
| 478 |
return NULL; |
| 479 |
|
| 480 |
/*
|
| 481 |
* Create the requested file only if user has demanded so.
|
| 482 |
* If errno is EINTR, retry. Else, leave.
|
| 483 |
*/
|
| 484 |
fd = do_block_create(pr, target, 0);
|
| 485 |
return fd;
|
| 486 |
} |
| 487 |
|
| 488 |
static int do_block_close(struct peer_req *pr) |
| 489 |
{
|
| 490 |
struct gluster_io *gio = (struct gluster_io *)(pr->priv); |
| 491 |
glfs_fd_t *fd = gio->fd; |
| 492 |
int r;
|
| 493 |
|
| 494 |
XSEGLOG2(&lc, D, "Closing fd %p", fd);
|
| 495 |
if (!fd)
|
| 496 |
return 0; |
| 497 |
|
| 498 |
r = glfs_close(fd); |
| 499 |
if (r < 0) |
| 500 |
XSEGLOG2(&lc, E, "Unexpected error (errno %d) while "
|
| 501 |
"closing fd %p:", errno, fd);
|
| 502 |
|
| 503 |
return r;
|
| 504 |
} |
| 505 |
|
| 506 |
/*
|
| 507 |
* do_block_lock() does the following:
|
| 508 |
* a. Checks if there is a lock-file for this object.
|
| 509 |
* i. Checks if the lock-file has been placed by us.
|
| 510 |
* ii. If we own the lock-file or if we are in XF_NOSYNC mode, it returns.
|
| 511 |
* iii. Else, it proceeds to (b)
|
| 512 |
* b. Attempts to create its own lockfile (obj_name + lock suffix). If it
|
| 513 |
* succeeds or is in XF_NOSYNC mode, it returns.
|
| 514 |
* c. If it has not locked the target object and is NOT in XF_NOSYNC mode, it
|
| 515 |
* sleeps for 1 second and it reattempts.
|
| 516 |
*/
|
| 517 |
int do_block_lock(struct peer_req *pr, char *target, int mode) |
| 518 |
{
|
| 519 |
glfs_fd_t *fd = NULL;
|
| 520 |
int r;
|
| 521 |
|
| 522 |
XSEGLOG2(&lc, D, "Locking target %s", target);
|
| 523 |
do {
|
| 524 |
fd = do_block_open(pr, target, 0);
|
| 525 |
if (fd) {
|
| 526 |
r = validate_lock_owner(pr, fd); |
| 527 |
if (r > 0 || mode == XF_NOSYNC) |
| 528 |
break;
|
| 529 |
} |
| 530 |
|
| 531 |
fd = do_block_create(pr, target, O_EXCL); |
| 532 |
if (fd) {
|
| 533 |
r = write_lock_owner(pr, fd); |
| 534 |
break;
|
| 535 |
} else if (mode == XF_NOSYNC) { |
| 536 |
r = -1;
|
| 537 |
break;
|
| 538 |
} |
| 539 |
|
| 540 |
sleep(1);
|
| 541 |
} while (1); |
| 542 |
|
| 543 |
glfs_close(fd); |
| 544 |
|
| 545 |
return r;
|
| 546 |
} |
| 547 |
|
| 548 |
/*
|
| 549 |
* do_block_unlock() does the following:
|
| 550 |
* a. Checks if there is a lock-file for this object. If not, it can safely
|
| 551 |
* return.
|
| 552 |
* b. Checks if the lock-file has been placed by us.
|
| 553 |
* ii. If we own the lock-file or if we are in XF_FORCE mode, we unlink it.
|
| 554 |
* c. If the unlinking was successful or if it failed due to ENOENT, the
|
| 555 |
* unlock operation is considered successful.
|
| 556 |
*/
|
| 557 |
int do_block_unlock(struct peer_req *pr, char *target, int mode) |
| 558 |
{
|
| 559 |
struct peerd *peer = pr->peer;
|
| 560 |
struct glusterd *gluster = __get_gluster(peer);
|
| 561 |
glfs_t *glfs = __get_glfs(gluster); |
| 562 |
glfs_fd_t *fd = NULL;
|
| 563 |
int r = 0; |
| 564 |
|
| 565 |
XSEGLOG2(&lc, D, "Unlocking target %s", target);
|
| 566 |
fd = do_block_open(pr, target, 0);
|
| 567 |
if (!fd) {
|
| 568 |
XSEGLOG2(&lc, D, "Target %s was not locked", target);
|
| 569 |
return -1; |
| 570 |
} |
| 571 |
|
| 572 |
r = validate_lock_owner(pr, fd); |
| 573 |
if (r < 0 && !(mode & XF_FORCE)) |
| 574 |
return -1; |
| 575 |
|
| 576 |
r = glfs_unlink(glfs, target); |
| 577 |
if (r >= 0 || errno == ENOENT) |
| 578 |
return 0; |
| 579 |
else
|
| 580 |
return -1; |
| 581 |
} |
| 582 |
|
| 583 |
int do_block_stat(struct peer_req *pr, char *target, struct stat *buf) |
| 584 |
{
|
| 585 |
struct peerd *peer = pr->peer;
|
| 586 |
struct glusterd *gluster = __get_gluster(peer);
|
| 587 |
glfs_t *glfs = __get_glfs(gluster); |
| 588 |
|
| 589 |
XSEGLOG2(&lc, D, "Stating target %s", target);
|
| 590 |
return glfs_stat(glfs, target, buf);
|
| 591 |
} |
| 592 |
|
| 593 |
int do_block_delete(struct peer_req *pr, char *target) |
| 594 |
{
|
| 595 |
struct peerd *peer = pr->peer;
|
| 596 |
struct glusterd *gluster = __get_gluster(peer);
|
| 597 |
glfs_t *glfs = __get_glfs(gluster); |
| 598 |
|
| 599 |
XSEGLOG2(&lc, D, "Deleting target %s", target);
|
| 600 |
return glfs_unlink(glfs, target);
|
| 601 |
} |
| 602 |
|
| 603 |
int handle_delete(struct peerd *peer, struct peer_req *pr) |
| 604 |
{
|
| 605 |
struct gluster_io *gio = (struct gluster_io *) pr->priv; |
| 606 |
struct xseg_request *req = pr->req;
|
| 607 |
int r;
|
| 608 |
|
| 609 |
XSEGLOG2(&lc, D, "Started for pr %p, req %p, target %s",
|
| 610 |
pr, req, gio->obj_name); |
| 611 |
|
| 612 |
r = do_block_delete(pr, gio->obj_name); |
| 613 |
if (r < 0) { |
| 614 |
XSEGLOG2(&lc, E, "Deletion of %s failed", gio->obj_name);
|
| 615 |
fail(peer, pr); |
| 616 |
} else {
|
| 617 |
XSEGLOG2(&lc, I, "Deletion of %s completed", gio->obj_name);
|
| 618 |
complete(peer, pr); |
| 619 |
} |
| 620 |
return 0; |
| 621 |
} |
| 622 |
|
| 623 |
int handle_info(struct peerd *peer, struct peer_req *pr) |
| 624 |
{
|
| 625 |
struct gluster_io *gio = (struct gluster_io *) pr->priv; |
| 626 |
struct xseg_request *req = pr->req;
|
| 627 |
struct xseg_reply_info *xinfo;
|
| 628 |
struct stat stat;
|
| 629 |
char *req_data;
|
| 630 |
char buf[XSEG_MAX_TARGETLEN + 1]; |
| 631 |
char *target = xseg_get_target(peer->xseg, req);
|
| 632 |
int r;
|
| 633 |
|
| 634 |
XSEGLOG2(&lc, D, "Started for pr %p, req %p, target %s",
|
| 635 |
pr, req, gio->obj_name); |
| 636 |
|
| 637 |
if (req->datalen < sizeof(struct xseg_reply_info)) { |
| 638 |
/* FIXME: Is this normal? */
|
| 639 |
strncpy(buf, target, req->targetlen); |
| 640 |
r = xseg_resize_request(peer->xseg, req, req->targetlen, |
| 641 |
sizeof(struct xseg_reply_info)); |
| 642 |
if (r < 0) { |
| 643 |
XSEGLOG2(&lc, E, "Cannot resize request");
|
| 644 |
fail(peer, pr); |
| 645 |
return -1; |
| 646 |
} |
| 647 |
target = xseg_get_target(peer->xseg, req); |
| 648 |
strncpy(target, buf, req->targetlen); |
| 649 |
} |
| 650 |
|
| 651 |
r = do_block_stat(pr, gio->obj_name, &stat); |
| 652 |
if (r < 0) { |
| 653 |
XSEGLOG2(&lc, E, "Stat failed for %s", gio->obj_name);
|
| 654 |
fail(peer, pr); |
| 655 |
return -1; |
| 656 |
} |
| 657 |
|
| 658 |
req_data = xseg_get_data(peer->xseg, pr->req); |
| 659 |
xinfo = (struct xseg_reply_info *)req_data;
|
| 660 |
xinfo->size = (uint64_t)stat.st_size; |
| 661 |
|
| 662 |
XSEGLOG2(&lc, I, "Getting info of %s completed", gio->obj_name);
|
| 663 |
complete(peer, pr); |
| 664 |
return 0; |
| 665 |
} |
| 666 |
|
| 667 |
void handle_ping(struct peerd *peer, struct peer_req *pr) |
| 668 |
{
|
| 669 |
XSEGLOG2(&lc, D, "Ping accepted. Acknowledging...");
|
| 670 |
|
| 671 |
complete(peer, pr); |
| 672 |
} |
| 673 |
|
| 674 |
int handle_read(struct peerd *peer, struct peer_req *pr) |
| 675 |
{
|
| 676 |
struct gluster_io *gio = (struct gluster_io *) (pr->priv); |
| 677 |
struct xseg_request *req = pr->req;
|
| 678 |
glfs_fd_t *fd; |
| 679 |
char *data = xseg_get_data(peer->xseg, pr->req);
|
| 680 |
char *target = gio->obj_name;
|
| 681 |
int ret = REQ_UNDEFINED;
|
| 682 |
|
| 683 |
XSEGLOG2(&lc, D, "Started for pr %p, req %p, target %s",
|
| 684 |
pr, req, target); |
| 685 |
|
| 686 |
if (gio->state == ACCEPTED) {
|
| 687 |
if (req->datalen < req->size) {
|
| 688 |
XSEGLOG2(&lc, E, "Request datalen is less than "
|
| 689 |
"request size");
|
| 690 |
ret = REQ_FAILED; |
| 691 |
goto out;
|
| 692 |
} |
| 693 |
|
| 694 |
if (!req->size) {
|
| 695 |
ret = REQ_COMPLETED; |
| 696 |
goto out;
|
| 697 |
} |
| 698 |
|
| 699 |
fd = do_block_open(pr, target, 0);
|
| 700 |
if (!fd) {
|
| 701 |
XSEGLOG2(&lc, I, "Object %s does not exist. "
|
| 702 |
"Serving zero data\n", target);
|
| 703 |
/* object not found. return zeros instead */
|
| 704 |
memset(data, 0, req->size);
|
| 705 |
req->serviced = req->size; |
| 706 |
ret = REQ_COMPLETED; |
| 707 |
goto out;
|
| 708 |
} |
| 709 |
gio->fd = fd; |
| 710 |
|
| 711 |
XSEGLOG2(&lc, I, "Reading %s", target);
|
| 712 |
|
| 713 |
gio->state = READING; |
| 714 |
ret = begin_aio_read(pr, data, req->size, req->offset); |
| 715 |
} else if (gio->state == READING) { |
| 716 |
XSEGLOG2(&lc, I, "Reading of %s callback", target);
|
| 717 |
ret = complete_aio_read(pr, data, req->size, |
| 718 |
req->offset, &req->serviced); |
| 719 |
} |
| 720 |
|
| 721 |
out:
|
| 722 |
switch (ret) {
|
| 723 |
case REQ_FAILED:
|
| 724 |
XSEGLOG2(&lc, E, "Reading of %s failed", target);
|
| 725 |
do_block_close(pr); |
| 726 |
fail(peer, pr); |
| 727 |
break;
|
| 728 |
case REQ_SUBMITTED:
|
| 729 |
XSEGLOG2(&lc, I, "Reading of %s submitted", target);
|
| 730 |
submit_hook(pr); |
| 731 |
break;
|
| 732 |
case REQ_COMPLETED:
|
| 733 |
XSEGLOG2(&lc, I, "Reading of %s completed", target);
|
| 734 |
do_block_close(pr); |
| 735 |
complete(peer, pr); |
| 736 |
break;
|
| 737 |
default:
|
| 738 |
XSEGLOG2(&lc, E, "Unknown request state. Failing.");
|
| 739 |
do_block_close(pr); |
| 740 |
fail(peer, pr); |
| 741 |
break;
|
| 742 |
} |
| 743 |
return 0; |
| 744 |
} |
| 745 |
|
| 746 |
int handle_write(struct peerd *peer, struct peer_req *pr) |
| 747 |
{
|
| 748 |
struct gluster_io *gio = (struct gluster_io *) (pr->priv); |
| 749 |
struct xseg_request *req = pr->req;
|
| 750 |
glfs_fd_t *fd; |
| 751 |
char *data = xseg_get_data(peer->xseg, pr->req);
|
| 752 |
char *target = gio->obj_name;
|
| 753 |
int ret = REQ_UNDEFINED;
|
| 754 |
|
| 755 |
XSEGLOG2(&lc, D, "Started for pr %p, req %p, target %s",
|
| 756 |
pr, req, target); |
| 757 |
|
| 758 |
if (gio->state == ACCEPTED) {
|
| 759 |
if (req->datalen < req->size) {
|
| 760 |
XSEGLOG2(&lc, E, "Request datalen is less than "
|
| 761 |
"request size");
|
| 762 |
ret = REQ_FAILED; |
| 763 |
goto out;
|
| 764 |
} |
| 765 |
|
| 766 |
if (!req->size) {
|
| 767 |
/* TODO: Flush data if req->flags & XF_FLUSH */
|
| 768 |
ret = REQ_COMPLETED; |
| 769 |
goto out;
|
| 770 |
} |
| 771 |
|
| 772 |
fd = do_block_open(pr, target, O_CREAT); |
| 773 |
if (!fd) {
|
| 774 |
XSEGLOG2(&lc, E, "Cannot open/create %s", target);
|
| 775 |
ret = REQ_FAILED; |
| 776 |
goto out;
|
| 777 |
} |
| 778 |
gio->fd = fd; |
| 779 |
|
| 780 |
XSEGLOG2(&lc, I, "Writing %s", target);
|
| 781 |
|
| 782 |
gio->state = WRITING; |
| 783 |
ret = begin_aio_write(pr, data, req->size, req->offset); |
| 784 |
} else if (gio->state == WRITING) { |
| 785 |
XSEGLOG2(&lc, I, "Writing of %s callback", target);
|
| 786 |
ret = complete_aio_write(pr, data, req->size, |
| 787 |
req->offset, &req->serviced); |
| 788 |
} |
| 789 |
|
| 790 |
out:
|
| 791 |
switch (ret) {
|
| 792 |
case REQ_FAILED:
|
| 793 |
XSEGLOG2(&lc, E, "Writing of %s failed", target);
|
| 794 |
do_block_close(pr); |
| 795 |
fail(peer, pr); |
| 796 |
break;
|
| 797 |
case REQ_SUBMITTED:
|
| 798 |
XSEGLOG2(&lc, I, "Writing of %s submitted", target);
|
| 799 |
submit_hook(pr); |
| 800 |
break;
|
| 801 |
case REQ_COMPLETED:
|
| 802 |
XSEGLOG2(&lc, I, "Writing of %s completed", target);
|
| 803 |
do_block_close(pr); |
| 804 |
complete(peer, pr); |
| 805 |
break;
|
| 806 |
default:
|
| 807 |
XSEGLOG2(&lc, E, "Unknown request state. Failing.");
|
| 808 |
do_block_close(pr); |
| 809 |
fail(peer, pr); |
| 810 |
break;
|
| 811 |
} |
| 812 |
return 0; |
| 813 |
} |
| 814 |
|
| 815 |
int handle_copy(struct peerd *peer, struct peer_req *pr) |
| 816 |
{
|
| 817 |
struct gluster_io *gio = (struct gluster_io *) (pr->priv); |
| 818 |
struct xseg_request *req = pr->req;
|
| 819 |
glfs_fd_t *fd; |
| 820 |
char *target = gio->obj_name;
|
| 821 |
int r;
|
| 822 |
int ret = REQ_UNDEFINED;
|
| 823 |
|
| 824 |
XSEGLOG2(&lc, D, "Started for pr %p, req %p, target %s",
|
| 825 |
pr, req, target); |
| 826 |
|
| 827 |
if (gio->state == ACCEPTED) {
|
| 828 |
/* Create second name and buf */
|
| 829 |
r = prepare_copy(pr); |
| 830 |
if (r < 0) { |
| 831 |
ret = REQ_FAILED; |
| 832 |
goto out;
|
| 833 |
} |
| 834 |
target = gio->second_name; |
| 835 |
|
| 836 |
XSEGLOG2(&lc, I, "Copy of object %s to object %s started",
|
| 837 |
gio->second_name, gio->obj_name); |
| 838 |
|
| 839 |
if (!req->size) {
|
| 840 |
ret = REQ_COMPLETED; |
| 841 |
goto out;
|
| 842 |
} |
| 843 |
|
| 844 |
/* FIXME: Will we fail here? */
|
| 845 |
fd = do_block_open(pr, target, 0);
|
| 846 |
if (!fd) {
|
| 847 |
XSEGLOG2(&lc, I, "Object %s does not exist. "
|
| 848 |
"Serving zero data\n", target);
|
| 849 |
/* object not found. return zeros instead */
|
| 850 |
memset(gio->buf, 0, req->size);
|
| 851 |
goto write;
|
| 852 |
} |
| 853 |
gio->fd = fd; |
| 854 |
|
| 855 |
XSEGLOG2(&lc, I, "Reading %s", target);
|
| 856 |
|
| 857 |
gio->state = READING; |
| 858 |
gio->read = 0;
|
| 859 |
ret = begin_aio_read(pr, gio->buf, req->size, req->offset); |
| 860 |
} |
| 861 |
else if (gio->state == READING){ |
| 862 |
target = gio->second_name; |
| 863 |
XSEGLOG2(&lc, I, "Reading of %s callback", target);
|
| 864 |
|
| 865 |
ret = complete_aio_read(pr, gio->buf, req->size, |
| 866 |
req->offset, &gio->read); |
| 867 |
|
| 868 |
if (ret == REQ_FAILED || ret == REQ_SUBMITTED)
|
| 869 |
goto out;
|
| 870 |
|
| 871 |
do_block_close(pr); |
| 872 |
write:
|
| 873 |
target = gio->obj_name; |
| 874 |
XSEGLOG2(&lc, I, "Target is %s", target);
|
| 875 |
fd = do_block_open(pr, target, O_CREAT); |
| 876 |
if (!fd) {
|
| 877 |
XSEGLOG2(&lc, E, "Cannot open/create %s", target);
|
| 878 |
ret = REQ_FAILED; |
| 879 |
goto out;
|
| 880 |
} |
| 881 |
gio->fd = fd; |
| 882 |
|
| 883 |
XSEGLOG2(&lc, I, "Writing %s", target);
|
| 884 |
|
| 885 |
gio->state = WRITING; |
| 886 |
ret = begin_aio_write(pr, gio->buf, req->size, req->offset); |
| 887 |
} else if (gio->state == WRITING) { |
| 888 |
XSEGLOG2(&lc, I, "Writing of %s callback", target);
|
| 889 |
ret = complete_aio_write(pr, gio->buf, req->size, |
| 890 |
req->offset, &req->serviced); |
| 891 |
} |
| 892 |
|
| 893 |
out:
|
| 894 |
if (ret != REQ_SUBMITTED) {
|
| 895 |
free(gio->buf); |
| 896 |
free(gio->second_name); |
| 897 |
gio->buf = NULL;
|
| 898 |
gio->second_name = NULL;
|
| 899 |
gio->read = 0;
|
| 900 |
} |
| 901 |
|
| 902 |
switch (ret) {
|
| 903 |
case REQ_FAILED:
|
| 904 |
XSEGLOG2(&lc, E, "Copying of %s failed", target);
|
| 905 |
do_block_close(pr); |
| 906 |
fail(peer, pr); |
| 907 |
break;
|
| 908 |
case REQ_SUBMITTED:
|
| 909 |
submit_hook(pr); |
| 910 |
break;
|
| 911 |
case REQ_COMPLETED:
|
| 912 |
XSEGLOG2(&lc, I, "Copying of %s completed", target);
|
| 913 |
do_block_close(pr); |
| 914 |
complete(peer, pr); |
| 915 |
break;
|
| 916 |
default:
|
| 917 |
XSEGLOG2(&lc, E, "Unknown request state. Failing.");
|
| 918 |
do_block_close(pr); |
| 919 |
fail(peer, pr); |
| 920 |
break;
|
| 921 |
} |
| 922 |
return 0; |
| 923 |
} |
| 924 |
|
| 925 |
int handle_hash(struct peerd *peer, struct peer_req *pr) |
| 926 |
{
|
| 927 |
struct xseg_request *req = pr->req;
|
| 928 |
struct gluster_io *gio = (struct gluster_io *) pr->priv; |
| 929 |
struct xseg_reply_hash *xreply;
|
| 930 |
glfs_fd_t *fd; |
| 931 |
uint64_t trailing_zeros = 0;
|
| 932 |
unsigned char sha[SHA256_DIGEST_SIZE]; |
| 933 |
char hash_name[HEXLIFIED_SHA256_DIGEST_SIZE + 1]; |
| 934 |
uint32_t pos; |
| 935 |
char *target = gio->obj_name;
|
| 936 |
int r;
|
| 937 |
int ret = REQ_UNDEFINED;
|
| 938 |
|
| 939 |
XSEGLOG2(&lc, D, "Started for pr %p, req %p, target %s",
|
| 940 |
pr, req, target); |
| 941 |
|
| 942 |
if (gio->state == ACCEPTED){
|
| 943 |
XSEGLOG2(&lc, I, "Starting hashing of object %s", target);
|
| 944 |
|
| 945 |
if (!req->size) {
|
| 946 |
ret = REQ_COMPLETED; /* or fail? */
|
| 947 |
goto out;
|
| 948 |
} |
| 949 |
|
| 950 |
/* Create hash_name, gio second_name and gio->buf */
|
| 951 |
r = prepare_hash(pr, hash_name); |
| 952 |
if (r < 0) { |
| 953 |
ret = REQ_FAILED; |
| 954 |
goto out;
|
| 955 |
} |
| 956 |
target = hash_name; |
| 957 |
gio->state = PREHASHING; |
| 958 |
|
| 959 |
/* Get correct status */
|
| 960 |
fd = do_block_open(pr, target, 0);
|
| 961 |
if (!fd) {
|
| 962 |
XSEGLOG2(&lc, I, "Hash %s does not exist.", target);
|
| 963 |
goto read;
|
| 964 |
} |
| 965 |
gio->fd = fd; |
| 966 |
|
| 967 |
XSEGLOG2(&lc, I, "Reading %s", target);
|
| 968 |
/* Read contents of hash_name in the gio->second_name buffer */
|
| 969 |
ret = begin_aio_read(pr, gio->second_name, |
| 970 |
HEXLIFIED_SHA256_DIGEST_SIZE, req->offset); |
| 971 |
} else if (gio->state == PREHASHING) { |
| 972 |
target = hash_name; |
| 973 |
XSEGLOG2(&lc, I, "Reading of %s callback", target);
|
| 974 |
ret = complete_aio_read(pr, gio->second_name, |
| 975 |
HEXLIFIED_SHA256_DIGEST_SIZE, |
| 976 |
req->offset, &gio->read); |
| 977 |
|
| 978 |
if (ret == REQ_FAILED || ret == REQ_SUBMITTED)
|
| 979 |
goto out;
|
| 980 |
|
| 981 |
/* Construct answer */
|
| 982 |
XSEGLOG2(&lc, D, "Precalculated hash found");
|
| 983 |
xreply = (struct xseg_reply_hash*)xseg_get_data(peer->xseg, req);
|
| 984 |
r = xseg_resize_request(peer->xseg, pr->req, pr->req->targetlen, |
| 985 |
sizeof(struct xseg_reply_hash)); |
| 986 |
strncpy(xreply->target, gio->second_name, HEXLIFIED_SHA256_DIGEST_SIZE); |
| 987 |
xreply->targetlen = HEXLIFIED_SHA256_DIGEST_SIZE; |
| 988 |
|
| 989 |
XSEGLOG2(&lc, I, "Calculated %s as hash of %s",
|
| 990 |
gio->second_name, gio->obj_name); |
| 991 |
req->serviced = req->size; |
| 992 |
goto out;
|
| 993 |
/* Leave */
|
| 994 |
|
| 995 |
read:
|
| 996 |
/*
|
| 997 |
* We reach this point only if there was no precalculated hash
|
| 998 |
*/
|
| 999 |
gio->state = READING; |
| 1000 |
gio->read = 0;
|
| 1001 |
target = gio->obj_name; |
| 1002 |
|
| 1003 |
fd = do_block_open(pr, target, 0);
|
| 1004 |
if (!fd) {
|
| 1005 |
XSEGLOG2(&lc, I, "Original object %s does not "
|
| 1006 |
"exist.\nServing zeroes.", target);
|
| 1007 |
memset(gio->buf, 0, req->size);
|
| 1008 |
gio->read = req->size; |
| 1009 |
goto hash;
|
| 1010 |
} |
| 1011 |
gio->fd = fd; |
| 1012 |
|
| 1013 |
XSEGLOG2(&lc, I, "Reading %s", target);
|
| 1014 |
ret = begin_aio_read(pr, gio->buf, req->size, req->offset); |
| 1015 |
} else if (gio->state == READING) { |
| 1016 |
target = gio->obj_name; |
| 1017 |
XSEGLOG2(&lc, I, "Reading of %s callback", target);
|
| 1018 |
ret = complete_aio_read(pr, gio->buf, req->size, |
| 1019 |
req->offset, &gio->read); |
| 1020 |
|
| 1021 |
if (ret == REQ_FAILED || ret == REQ_SUBMITTED)
|
| 1022 |
goto out;
|
| 1023 |
|
| 1024 |
do_block_close(pr); |
| 1025 |
|
| 1026 |
hash:
|
| 1027 |
/* Strip here trailing zeroes */
|
| 1028 |
for (; trailing_zeros < gio->read; trailing_zeros++) {
|
| 1029 |
if (gio->buf[gio->read-trailing_zeros -1]) |
| 1030 |
break;
|
| 1031 |
} |
| 1032 |
XSEGLOG2(&lc, D, "Read %llu, Trailing zeros %llu",
|
| 1033 |
gio->read, trailing_zeros); |
| 1034 |
|
| 1035 |
gio->read -= trailing_zeros; |
| 1036 |
SHA256((unsigned char *) gio->buf, gio->read, sha); |
| 1037 |
hexlify(sha, SHA256_DIGEST_SIZE, gio->second_name); |
| 1038 |
gio->second_name[HEXLIFIED_SHA256_DIGEST_SIZE] = 0;
|
| 1039 |
|
| 1040 |
/* Construct reply */
|
| 1041 |
xreply = (struct xseg_reply_hash*)xseg_get_data(peer->xseg, req);
|
| 1042 |
r = xseg_resize_request(peer->xseg, pr->req, pr->req->targetlen, |
| 1043 |
sizeof(struct xseg_reply_hash)); |
| 1044 |
strncpy(xreply->target, gio->second_name, HEXLIFIED_SHA256_DIGEST_SIZE); |
| 1045 |
xreply->targetlen = HEXLIFIED_SHA256_DIGEST_SIZE; |
| 1046 |
|
| 1047 |
XSEGLOG2(&lc, I, "Calculated %s as hash of %s",
|
| 1048 |
gio->second_name, gio->obj_name); |
| 1049 |
/*
|
| 1050 |
* We can't leave since we need to write the
|
| 1051 |
* content-addressable object to the backend.
|
| 1052 |
*/
|
| 1053 |
|
| 1054 |
gio->read = 0;
|
| 1055 |
fd = do_block_create(pr, gio->second_name, O_EXCL); |
| 1056 |
if (!fd) {
|
| 1057 |
XSEGLOG2(&lc, I, "Hash of object %s to object %s completed",
|
| 1058 |
gio->obj_name, gio->second_name); |
| 1059 |
req->serviced = req->size; |
| 1060 |
goto write;
|
| 1061 |
} |
| 1062 |
gio->fd = fd; |
| 1063 |
target = gio->second_name; |
| 1064 |
|
| 1065 |
XSEGLOG2(&lc, I, "Writing %s", target);
|
| 1066 |
|
| 1067 |
gio->state = WRITING; |
| 1068 |
ret = begin_aio_write(pr, gio->buf, req->size, req->offset); |
| 1069 |
} else if (gio->state == WRITING) { |
| 1070 |
target = gio->second_name; |
| 1071 |
XSEGLOG2(&lc, I, "Writing of %s callback", target);
|
| 1072 |
ret = complete_aio_write(pr, gio->buf, req->size, req->offset, |
| 1073 |
&req->serviced); |
| 1074 |
|
| 1075 |
if (ret == REQ_FAILED || ret == REQ_SUBMITTED)
|
| 1076 |
goto out;
|
| 1077 |
|
| 1078 |
XSEGLOG2(&lc, I, "Writing of %s completed", gio->second_name);
|
| 1079 |
|
| 1080 |
/*
|
| 1081 |
* We can't leave since we need to write the precalculated hash
|
| 1082 |
* value to the backend.
|
| 1083 |
*/
|
| 1084 |
do_block_close(pr); |
| 1085 |
write:
|
| 1086 |
pos = 0;
|
| 1087 |
strncpy(hash_name, gio->obj_name, strlen(gio->obj_name)); |
| 1088 |
pos += strlen(gio->obj_name); |
| 1089 |
strncpy(hash_name+pos, HASH_SUFFIX, HASH_SUFFIX_LEN); |
| 1090 |
pos += HASH_SUFFIX_LEN; |
| 1091 |
hash_name[pos] = 0;
|
| 1092 |
|
| 1093 |
gio->state = POSTHASHING; |
| 1094 |
fd = do_block_create(pr, hash_name, O_EXCL); |
| 1095 |
if (!fd) {
|
| 1096 |
XSEGLOG2(&lc, I, "Writing of prehashed value completed");
|
| 1097 |
XSEGLOG2(&lc, I, "Hash of object %s to object %s completed",
|
| 1098 |
gio->obj_name, gio->second_name); |
| 1099 |
req->serviced = req->size; |
| 1100 |
ret = REQ_COMPLETED; |
| 1101 |
/* TODO: Check errors */
|
| 1102 |
goto out;
|
| 1103 |
} |
| 1104 |
gio->fd = fd; |
| 1105 |
target = hash_name; |
| 1106 |
|
| 1107 |
XSEGLOG2(&lc, I, "Writing prehashed value");
|
| 1108 |
ret = begin_aio_write(pr, gio->second_name, |
| 1109 |
HEXLIFIED_SHA256_DIGEST_SIZE, 0);
|
| 1110 |
} else if (gio->state == POSTHASHING) { |
| 1111 |
XSEGLOG2(&lc, I, "Writing of prehashed value callback");
|
| 1112 |
ret = complete_aio_write(pr, gio->second_name, |
| 1113 |
HEXLIFIED_SHA256_DIGEST_SIZE, 0, &gio->read);
|
| 1114 |
|
| 1115 |
if (ret == REQ_FAILED || ret == REQ_SUBMITTED)
|
| 1116 |
goto out;
|
| 1117 |
|
| 1118 |
XSEGLOG2(&lc, I, "Writing of prehashed value completed");
|
| 1119 |
req->serviced = req->size; |
| 1120 |
} |
| 1121 |
|
| 1122 |
out:
|
| 1123 |
target = gio->obj_name; |
| 1124 |
if (ret != REQ_SUBMITTED) {
|
| 1125 |
free(gio->buf); |
| 1126 |
free(gio->second_name); |
| 1127 |
gio->buf = NULL;
|
| 1128 |
gio->second_name = NULL;
|
| 1129 |
gio->read = 0;
|
| 1130 |
} |
| 1131 |
|
| 1132 |
switch (ret) {
|
| 1133 |
case REQ_FAILED:
|
| 1134 |
XSEGLOG2(&lc, E, "Hashing of %s failed", target);
|
| 1135 |
do_block_close(pr); |
| 1136 |
fail(peer, pr); |
| 1137 |
break;
|
| 1138 |
case REQ_SUBMITTED:
|
| 1139 |
submit_hook(pr); |
| 1140 |
break;
|
| 1141 |
case REQ_COMPLETED:
|
| 1142 |
XSEGLOG2(&lc, I, "Hashing of %s completed", target);
|
| 1143 |
do_block_close(pr); |
| 1144 |
complete(peer, pr); |
| 1145 |
break;
|
| 1146 |
default:
|
| 1147 |
XSEGLOG2(&lc, E, "Unknown request state. Failing.");
|
| 1148 |
do_block_close(pr); |
| 1149 |
fail(peer, pr); |
| 1150 |
break;
|
| 1151 |
} |
| 1152 |
return 0; |
| 1153 |
} |
| 1154 |
|
| 1155 |
int handle_acquire(struct peerd *peer, struct peer_req *pr) |
| 1156 |
{
|
| 1157 |
struct gluster_io *gio = (struct gluster_io *)(pr->priv); |
| 1158 |
struct xseg_request *req = pr->req;
|
| 1159 |
uint32_t len = strlen(gio->obj_name); |
| 1160 |
int ret;
|
| 1161 |
|
| 1162 |
strncpy(gio->obj_name + len, LOCK_SUFFIX, LOCK_SUFFIX_LEN); |
| 1163 |
gio->obj_name[len + LOCK_SUFFIX_LEN] = 0;
|
| 1164 |
|
| 1165 |
XSEGLOG2(&lc, I, "Starting lock op for %s", gio->obj_name);
|
| 1166 |
|
| 1167 |
/* TODO: Check error codes and retry if needed */
|
| 1168 |
ret = do_block_lock(pr, gio->obj_name, req->flags & XF_NOSYNC); |
| 1169 |
if (ret < 0) { |
| 1170 |
XSEGLOG2(&lc, E, "Lock op failed for %s", gio->obj_name);
|
| 1171 |
fail(peer, pr); |
| 1172 |
} else {
|
| 1173 |
XSEGLOG2(&lc, I, "Lock op succeeded for %s", gio->obj_name);
|
| 1174 |
complete(peer, pr); |
| 1175 |
} |
| 1176 |
return 0; |
| 1177 |
} |
| 1178 |
|
| 1179 |
|
| 1180 |
int handle_release(struct peerd *peer, struct peer_req *pr) |
| 1181 |
{
|
| 1182 |
struct gluster_io *gio = (struct gluster_io *)(pr->priv); |
| 1183 |
struct xseg_request *req = pr->req;
|
| 1184 |
uint32_t len = strlen(gio->obj_name); |
| 1185 |
int ret;
|
| 1186 |
|
| 1187 |
strncpy(gio->obj_name + len, LOCK_SUFFIX, LOCK_SUFFIX_LEN); |
| 1188 |
gio->obj_name[len + LOCK_SUFFIX_LEN] = 0;
|
| 1189 |
|
| 1190 |
XSEGLOG2(&lc, I, "Starting unlock op for %s", gio->obj_name);
|
| 1191 |
|
| 1192 |
/* TODO: Check error codes and retry if needed */
|
| 1193 |
ret = do_block_unlock(pr, gio->obj_name, req->flags & XF_FORCE); |
| 1194 |
if (ret < 0) { |
| 1195 |
XSEGLOG2(&lc, E, "Unlock op failed for %s", gio->obj_name);
|
| 1196 |
fail(peer, pr); |
| 1197 |
} else {
|
| 1198 |
XSEGLOG2(&lc, I, "Unlock op succeeded for %s", gio->obj_name);
|
| 1199 |
complete(peer, pr); |
| 1200 |
} |
| 1201 |
return 0; |
| 1202 |
} |
| 1203 |
|
| 1204 |
int custom_peer_init(struct peerd *peer, int argc, char *argv[]) |
| 1205 |
{
|
| 1206 |
struct glusterd *gluster = malloc(sizeof(struct glusterd)); |
| 1207 |
struct gluster_io *gio;
|
| 1208 |
glfs_t *glfs = NULL;
|
| 1209 |
char transport[MAX_GLFS_ARG_LEN + 1]; |
| 1210 |
char server[MAX_GLFS_ARG_LEN + 1]; |
| 1211 |
char volume[MAX_GLFS_ARG_LEN + 1]; |
| 1212 |
char uniquestr[MAX_UNIQUESTR_LEN + 1]; |
| 1213 |
int tid;
|
| 1214 |
int port = 0; |
| 1215 |
int ret = 0; |
| 1216 |
int i = 0; |
| 1217 |
int j = 0; |
| 1218 |
|
| 1219 |
if (!gluster) {
|
| 1220 |
perror("malloc");
|
| 1221 |
return -1; |
| 1222 |
} |
| 1223 |
gluster->async = 0;
|
| 1224 |
|
| 1225 |
uniquestr[0] = 0; |
| 1226 |
transport[0] = 0; |
| 1227 |
server[0] = 0; |
| 1228 |
volume[0] = 0; |
| 1229 |
|
| 1230 |
BEGIN_READ_ARGS(argc, argv); |
| 1231 |
READ_ARG_STRING("--transport", transport, MAX_GLFS_ARG_LEN);
|
| 1232 |
READ_ARG_STRING("--server", server, MAX_GLFS_ARG_LEN);
|
| 1233 |
READ_ARG_ULONG("--port", port);
|
| 1234 |
READ_ARG_STRING("--volume", volume, MAX_GLFS_ARG_LEN);
|
| 1235 |
READ_ARG_STRING("--uniquestr", gluster->uniquestr, MAX_UNIQUESTR_LEN);
|
| 1236 |
READ_ARG_BOOL("--async", gluster->async);
|
| 1237 |
END_READ_ARGS(); |
| 1238 |
|
| 1239 |
|
| 1240 |
if (!volume[0]){ |
| 1241 |
XSEGLOG2(&lc, E , "Volume must be provided");
|
| 1242 |
usage(argv[0]);
|
| 1243 |
goto err_arg;
|
| 1244 |
} |
| 1245 |
|
| 1246 |
/* Use defaults if user has not provided his/her own */
|
| 1247 |
if (!transport[0]) |
| 1248 |
strncpy(transport, "tcp", 4); |
| 1249 |
|
| 1250 |
if (!server[0]) |
| 1251 |
strncpy(server, "127.0.0.1", 10); |
| 1252 |
|
| 1253 |
if (!uniquestr[0]) { |
| 1254 |
tid = getpid(); |
| 1255 |
snprintf(gluster->uniquestr, MAX_UNIQUESTR_LEN, "%d", tid);
|
| 1256 |
XSEGLOG2(&lc, W, "Warning, uniquestr not provided.\n"
|
| 1257 |
"\tUsing instead \"%s\" but uniqueness is not "
|
| 1258 |
"guaranteed.", gluster->uniquestr);
|
| 1259 |
} else {
|
| 1260 |
strncpy(gluster->uniquestr, uniquestr, MAX_UNIQUESTR_LEN); |
| 1261 |
} |
| 1262 |
|
| 1263 |
ret = __set_glfs(gluster, volume); |
| 1264 |
if (ret < 0) { |
| 1265 |
XSEGLOG2(&lc, E, "Error at glfs_new");
|
| 1266 |
goto err_glfs;
|
| 1267 |
} |
| 1268 |
glfs = __get_glfs(gluster); |
| 1269 |
|
| 1270 |
ret = glfs_set_volfile_server(glfs, transport, server, port); |
| 1271 |
if (ret < 0) { |
| 1272 |
XSEGLOG2(&lc, E, "Error at glfs_set_volfile_server");
|
| 1273 |
goto err_glfs;
|
| 1274 |
} |
| 1275 |
|
| 1276 |
ret = glfs_init(glfs); |
| 1277 |
if (ret) {
|
| 1278 |
XSEGLOG2(&lc, E, "Error at glfs_init\n");
|
| 1279 |
goto err_glfs;
|
| 1280 |
} |
| 1281 |
|
| 1282 |
peer->priv = (void *)gluster;
|
| 1283 |
for (i = 0; i < peer->nr_ops; i++) { |
| 1284 |
gio = malloc(sizeof(struct gluster_io)); |
| 1285 |
|
| 1286 |
if (!gio)
|
| 1287 |
goto err_gio;
|
| 1288 |
|
| 1289 |
gio->fd = NULL;
|
| 1290 |
gio->buf = 0;
|
| 1291 |
gio->read = 0;
|
| 1292 |
gio->size = 0;
|
| 1293 |
gio->second_name = 0;
|
| 1294 |
peer->peer_reqs[i].priv = (void *) gio;
|
| 1295 |
} |
| 1296 |
return 0; |
| 1297 |
|
| 1298 |
err_arg:
|
| 1299 |
free(gluster); |
| 1300 |
err_gio:
|
| 1301 |
for (j = 0; j < i; j++) |
| 1302 |
free(peer->peer_reqs[j].priv); |
| 1303 |
err_glfs:
|
| 1304 |
glfs_fini(glfs); |
| 1305 |
return -1; |
| 1306 |
} |
| 1307 |
|
| 1308 |
// nothing to do here for now
|
| 1309 |
int custom_arg_parse(int argc, const char *argv[]) |
| 1310 |
{
|
| 1311 |
return 0; |
| 1312 |
} |
| 1313 |
|
| 1314 |
void custom_peer_finalize(struct peerd *peer) |
| 1315 |
{
|
| 1316 |
struct glusterd *gluster = __get_gluster(peer);
|
| 1317 |
glfs_t *glfs = __get_glfs(gluster); |
| 1318 |
|
| 1319 |
glfs_fini(glfs); |
| 1320 |
XSEGLOG2(&lc, I, "Glusterd has closed successfully");
|
| 1321 |
} |
| 1322 |
|
| 1323 |
int dispatch(struct peerd *peer, struct peer_req *pr, struct xseg_request *req, |
| 1324 |
enum dispatch_reason reason)
|
| 1325 |
{
|
| 1326 |
struct gluster_io *gio = (struct gluster_io *) (pr->priv); |
| 1327 |
char *target = xseg_get_target(peer->xseg, pr->req);
|
| 1328 |
unsigned int end = (pr->req->targetlen > MAX_OBJ_NAME) ? |
| 1329 |
MAX_OBJ_NAME : pr->req->targetlen; |
| 1330 |
|
| 1331 |
if (reason == dispatch_accept) {
|
| 1332 |
strncpy(gio->obj_name, target, end); |
| 1333 |
gio->obj_name[end] = 0;
|
| 1334 |
gio->state = ACCEPTED; |
| 1335 |
gio->fd = NULL;
|
| 1336 |
gio->read = 0;
|
| 1337 |
} |
| 1338 |
|
| 1339 |
switch (pr->req->op){
|
| 1340 |
case X_READ:
|
| 1341 |
handle_read(peer, pr); |
| 1342 |
break;
|
| 1343 |
case X_WRITE:
|
| 1344 |
handle_write(peer, pr); |
| 1345 |
break;
|
| 1346 |
case X_DELETE:
|
| 1347 |
if (canDefer(peer))
|
| 1348 |
defer_request(peer, pr); |
| 1349 |
else
|
| 1350 |
handle_delete(peer, pr); |
| 1351 |
break;
|
| 1352 |
case X_INFO:
|
| 1353 |
if (canDefer(peer))
|
| 1354 |
defer_request(peer, pr); |
| 1355 |
else
|
| 1356 |
handle_info(peer, pr); |
| 1357 |
break;
|
| 1358 |
case X_COPY:
|
| 1359 |
if (canDefer(peer))
|
| 1360 |
defer_request(peer, pr); |
| 1361 |
else
|
| 1362 |
handle_copy(peer, pr); |
| 1363 |
break;
|
| 1364 |
case X_ACQUIRE:
|
| 1365 |
handle_acquire(peer, pr); |
| 1366 |
break;
|
| 1367 |
case X_RELEASE:
|
| 1368 |
handle_release(peer, pr); |
| 1369 |
break;
|
| 1370 |
case X_HASH:
|
| 1371 |
handle_hash(peer, pr); |
| 1372 |
break;
|
| 1373 |
case X_PING:
|
| 1374 |
handle_ping(peer, pr); |
| 1375 |
break;
|
| 1376 |
default:
|
| 1377 |
fail(peer, pr); |
| 1378 |
} |
| 1379 |
return 0; |
| 1380 |
} |