1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
30 * Use is subject to license terms.
33 * Copyright (c) 2011 Whamcloud, Inc.
36 * This file is part of Lustre, http://www.lustre.org/
37 * Lustre is a trademark of Sun Microsystems, Inc.
39 * lustre/include/lustre_lib.h
41 * Basic Lustre library routines.
52 #include <libcfs/libcfs.h>
53 #include <lustre/lustre_idl.h>
54 #include <lustre_ver.h>
55 #include <lustre_cfg.h>
56 #if defined(__linux__)
57 #include <linux/lustre_lib.h>
58 #elif defined(__APPLE__)
59 #include <darwin/lustre_lib.h>
60 #elif defined(__WINNT__)
61 #include <winnt/lustre_lib.h>
63 #error Unsupported operating system.
67 struct ptlrpc_request;
70 #include <lustre_ha.h>
71 #include <lustre_net.h>
74 void target_client_add_cb(struct obd_device *obd, __u64 transno, void *cb_data,
76 int target_handle_connect(struct ptlrpc_request *req);
77 int target_handle_disconnect(struct ptlrpc_request *req);
78 void target_destroy_export(struct obd_export *exp);
79 int target_pack_pool_reply(struct ptlrpc_request *req);
80 int target_handle_ping(struct ptlrpc_request *req);
81 void target_committed_to_req(struct ptlrpc_request *req);
82 int do_set_info_async(struct obd_import *imp,
83 int opcode, int version,
84 obd_count keylen, void *key,
85 obd_count vallen, void *val,
86 struct ptlrpc_request_set *set);
88 /* quotacheck callback, dqacq/dqrel callback handler */
89 int target_handle_qc_callback(struct ptlrpc_request *req);
90 #ifdef HAVE_QUOTA_SUPPORT
91 int target_handle_dqacq_callback(struct ptlrpc_request *req);
93 #define target_handle_dqacq_callback(req) ldlm_callback_reply(req, -ENOTSUPP)
96 #define OBD_RECOVERY_MAX_TIME (obd_timeout * 18) /* b13079 */
100 void target_cancel_recovery_timer(struct obd_device *obd);
101 void target_stop_recovery_thread(struct obd_device *obd);
102 void target_cleanup_recovery(struct obd_device *obd);
103 int target_queue_recovery_request(struct ptlrpc_request *req,
104 struct obd_device *obd);
105 void target_send_reply(struct ptlrpc_request *req, int rc, int fail_id);
106 int target_bulk_io(struct obd_export *exp, struct ptlrpc_bulk_desc *desc,
107 struct l_wait_info *lwi);
111 int client_sanobd_setup(struct obd_device *obddev, struct lustre_cfg* lcfg);
112 struct client_obd *client_conn2cli(struct lustre_handle *conn);
115 struct obd_client_handle {
116 struct lustre_handle och_fh;
117 struct lu_fid och_fid;
118 struct md_open_data *och_mod;
122 #define OBD_CLIENT_HANDLE_MAGIC 0xd15ea5ed
125 void statfs_pack(struct obd_statfs *osfs, cfs_kstatfs_t *sfs);
126 void statfs_unpack(cfs_kstatfs_t *sfs, struct obd_statfs *osfs);
132 cfs_semaphore_t l_sem;
133 cfs_spinlock_t l_spin;
136 void l_lock_init(struct lustre_lock *);
137 void l_lock(struct lustre_lock *);
138 void l_unlock(struct lustre_lock *);
139 int l_has_lock(struct lustre_lock *);
145 ECHO_MD_CREATE = 1, /* Open/Create file on MDT */
146 ECHO_MD_MKDIR = 2, /* Mkdir on MDT */
147 ECHO_MD_DESTROY = 3, /* Unlink file on MDT */
148 ECHO_MD_RMDIR = 4, /* Rmdir on MDT */
149 ECHO_MD_LOOKUP = 5, /* Lookup on MDT */
150 ECHO_MD_GETATTR = 6, /* Getattr on MDT */
151 ECHO_MD_SETATTR = 7, /* Setattr on MDT */
152 ECHO_MD_ALLOC_FID = 8, /* Get FIDs from MDT */
158 #define OBD_IOCTL_VERSION 0x00010004
160 struct obd_ioctl_data {
177 struct obdo ioc_obdo1;
178 struct obdo ioc_obdo2;
189 /* buffers the kernel will treat as user pointers */
195 /* inline buffers for various arguments */
208 struct obd_ioctl_hdr {
213 static inline int obd_ioctl_packlen(struct obd_ioctl_data *data)
215 int len = cfs_size_round(sizeof(struct obd_ioctl_data));
216 len += cfs_size_round(data->ioc_inllen1);
217 len += cfs_size_round(data->ioc_inllen2);
218 len += cfs_size_round(data->ioc_inllen3);
219 len += cfs_size_round(data->ioc_inllen4);
224 static inline int obd_ioctl_is_invalid(struct obd_ioctl_data *data)
226 if (data->ioc_len > (1<<30)) {
227 CERROR("OBD ioctl: ioc_len larger than 1<<30\n");
230 if (data->ioc_inllen1 > (1<<30)) {
231 CERROR("OBD ioctl: ioc_inllen1 larger than 1<<30\n");
234 if (data->ioc_inllen2 > (1<<30)) {
235 CERROR("OBD ioctl: ioc_inllen2 larger than 1<<30\n");
238 if (data->ioc_inllen3 > (1<<30)) {
239 CERROR("OBD ioctl: ioc_inllen3 larger than 1<<30\n");
242 if (data->ioc_inllen4 > (1<<30)) {
243 CERROR("OBD ioctl: ioc_inllen4 larger than 1<<30\n");
246 if (data->ioc_inlbuf1 && !data->ioc_inllen1) {
247 CERROR("OBD ioctl: inlbuf1 pointer but 0 length\n");
250 if (data->ioc_inlbuf2 && !data->ioc_inllen2) {
251 CERROR("OBD ioctl: inlbuf2 pointer but 0 length\n");
254 if (data->ioc_inlbuf3 && !data->ioc_inllen3) {
255 CERROR("OBD ioctl: inlbuf3 pointer but 0 length\n");
258 if (data->ioc_inlbuf4 && !data->ioc_inllen4) {
259 CERROR("OBD ioctl: inlbuf4 pointer but 0 length\n");
262 if (data->ioc_pbuf1 && !data->ioc_plen1) {
263 CERROR("OBD ioctl: pbuf1 pointer but 0 length\n");
266 if (data->ioc_pbuf2 && !data->ioc_plen2) {
267 CERROR("OBD ioctl: pbuf2 pointer but 0 length\n");
270 if (data->ioc_plen1 && !data->ioc_pbuf1) {
271 CERROR("OBD ioctl: plen1 set but NULL pointer\n");
274 if (data->ioc_plen2 && !data->ioc_pbuf2) {
275 CERROR("OBD ioctl: plen2 set but NULL pointer\n");
278 if (obd_ioctl_packlen(data) > data->ioc_len) {
279 CERROR("OBD ioctl: packlen exceeds ioc_len (%d > %d)\n",
280 obd_ioctl_packlen(data), data->ioc_len);
287 static inline int obd_ioctl_pack(struct obd_ioctl_data *data, char **pbuf,
291 struct obd_ioctl_data *overlay;
292 data->ioc_len = obd_ioctl_packlen(data);
293 data->ioc_version = OBD_IOCTL_VERSION;
295 if (*pbuf && data->ioc_len > max) {
296 fprintf(stderr, "pbuf %p ioc_len %u max %d\n", *pbuf,
301 *pbuf = malloc(data->ioc_len);
305 overlay = (struct obd_ioctl_data *)*pbuf;
306 memcpy(*pbuf, data, sizeof(*data));
308 ptr = overlay->ioc_bulk;
309 if (data->ioc_inlbuf1)
310 LOGL(data->ioc_inlbuf1, data->ioc_inllen1, ptr);
311 if (data->ioc_inlbuf2)
312 LOGL(data->ioc_inlbuf2, data->ioc_inllen2, ptr);
313 if (data->ioc_inlbuf3)
314 LOGL(data->ioc_inlbuf3, data->ioc_inllen3, ptr);
315 if (data->ioc_inlbuf4)
316 LOGL(data->ioc_inlbuf4, data->ioc_inllen4, ptr);
317 if (obd_ioctl_is_invalid(overlay)) {
318 fprintf(stderr, "ioc_len %u max %d\n",
326 static inline int obd_ioctl_unpack(struct obd_ioctl_data *data, char *pbuf,
330 struct obd_ioctl_data *overlay;
334 overlay = (struct obd_ioctl_data *)pbuf;
336 /* Preserve the caller's buffer pointers */
337 overlay->ioc_inlbuf1 = data->ioc_inlbuf1;
338 overlay->ioc_inlbuf2 = data->ioc_inlbuf2;
339 overlay->ioc_inlbuf3 = data->ioc_inlbuf3;
340 overlay->ioc_inlbuf4 = data->ioc_inlbuf4;
342 memcpy(data, pbuf, sizeof(*data));
344 ptr = overlay->ioc_bulk;
345 if (data->ioc_inlbuf1)
346 LOGU(data->ioc_inlbuf1, data->ioc_inllen1, ptr);
347 if (data->ioc_inlbuf2)
348 LOGU(data->ioc_inlbuf2, data->ioc_inllen2, ptr);
349 if (data->ioc_inlbuf3)
350 LOGU(data->ioc_inlbuf3, data->ioc_inllen3, ptr);
351 if (data->ioc_inlbuf4)
352 LOGU(data->ioc_inlbuf4, data->ioc_inllen4, ptr);
358 #include <obd_support.h>
361 /* function defined in lustre/obdclass/<platform>/<platform>-module.c */
362 int obd_ioctl_getdata(char **buf, int *len, void *arg);
363 int obd_ioctl_popdata(void *arg, void *data, int len);
365 /* buffer MUST be at least the size of obd_ioctl_hdr */
366 static inline int obd_ioctl_getdata(char **buf, int *len, void *arg)
368 struct obd_ioctl_hdr hdr;
369 struct obd_ioctl_data *data;
374 err = cfs_copy_from_user(&hdr, (void *)arg, sizeof(hdr));
378 if (hdr.ioc_version != OBD_IOCTL_VERSION) {
379 CERROR("Version mismatch kernel vs application\n");
383 if (hdr.ioc_len > OBD_MAX_IOCTL_BUFFER) {
384 CERROR("User buffer len %d exceeds %d max buffer\n",
385 hdr.ioc_len, OBD_MAX_IOCTL_BUFFER);
389 if (hdr.ioc_len < sizeof(struct obd_ioctl_data)) {
390 CERROR("User buffer too small for ioctl (%d)\n", hdr.ioc_len);
394 OBD_ALLOC_LARGE(*buf, hdr.ioc_len);
396 CERROR("Cannot allocate control buffer of len %d\n",
401 data = (struct obd_ioctl_data *)*buf;
403 err = cfs_copy_from_user(*buf, (void *)arg, hdr.ioc_len);
405 OBD_FREE_LARGE(*buf, hdr.ioc_len);
409 if (obd_ioctl_is_invalid(data)) {
410 CERROR("ioctl not correctly formatted\n");
411 OBD_FREE_LARGE(*buf, hdr.ioc_len);
415 if (data->ioc_inllen1) {
416 data->ioc_inlbuf1 = &data->ioc_bulk[0];
417 offset += cfs_size_round(data->ioc_inllen1);
420 if (data->ioc_inllen2) {
421 data->ioc_inlbuf2 = &data->ioc_bulk[0] + offset;
422 offset += cfs_size_round(data->ioc_inllen2);
425 if (data->ioc_inllen3) {
426 data->ioc_inlbuf3 = &data->ioc_bulk[0] + offset;
427 offset += cfs_size_round(data->ioc_inllen3);
430 if (data->ioc_inllen4) {
431 data->ioc_inlbuf4 = &data->ioc_bulk[0] + offset;
437 static inline int obd_ioctl_popdata(void *arg, void *data, int len)
439 int err = cfs_copy_to_user(arg, data, len);
446 static inline void obd_ioctl_freedata(char *buf, int len)
450 OBD_FREE_LARGE(buf, len);
456 * BSD ioctl description:
457 * #define IOC_V1 _IOR(g, n1, long)
458 * #define IOC_V2 _IOW(g, n2, long)
460 * ioctl(f, IOC_V1, arg);
461 * arg will be treated as a long value,
463 * ioctl(f, IOC_V2, arg)
464 * arg will be treated as a pointer, bsd will call
465 * copyin(buf, arg, sizeof(long))
467 * To make BSD ioctl handles argument correctly and simplely,
468 * we change _IOR to _IOWR so BSD will copyin obd_ioctl_data
469 * for us. Does this change affect Linux? (XXX Liang)
471 #define OBD_IOC_CREATE _IOWR('f', 101, OBD_IOC_DATA_TYPE)
472 #define OBD_IOC_DESTROY _IOW ('f', 104, OBD_IOC_DATA_TYPE)
473 #define OBD_IOC_PREALLOCATE _IOWR('f', 105, OBD_IOC_DATA_TYPE)
475 #define OBD_IOC_SETATTR _IOW ('f', 107, OBD_IOC_DATA_TYPE)
476 #define OBD_IOC_GETATTR _IOWR ('f', 108, OBD_IOC_DATA_TYPE)
477 #define OBD_IOC_READ _IOWR('f', 109, OBD_IOC_DATA_TYPE)
478 #define OBD_IOC_WRITE _IOWR('f', 110, OBD_IOC_DATA_TYPE)
481 #define OBD_IOC_STATFS _IOWR('f', 113, OBD_IOC_DATA_TYPE)
482 #define OBD_IOC_SYNC _IOW ('f', 114, OBD_IOC_DATA_TYPE)
483 #define OBD_IOC_READ2 _IOWR('f', 115, OBD_IOC_DATA_TYPE)
484 #define OBD_IOC_FORMAT _IOWR('f', 116, OBD_IOC_DATA_TYPE)
485 #define OBD_IOC_PARTITION _IOWR('f', 117, OBD_IOC_DATA_TYPE)
486 #define OBD_IOC_COPY _IOWR('f', 120, OBD_IOC_DATA_TYPE)
487 #define OBD_IOC_MIGR _IOWR('f', 121, OBD_IOC_DATA_TYPE)
488 #define OBD_IOC_PUNCH _IOWR('f', 122, OBD_IOC_DATA_TYPE)
490 #define OBD_IOC_MODULE_DEBUG _IOWR('f', 124, OBD_IOC_DATA_TYPE)
491 #define OBD_IOC_BRW_READ _IOWR('f', 125, OBD_IOC_DATA_TYPE)
492 #define OBD_IOC_BRW_WRITE _IOWR('f', 126, OBD_IOC_DATA_TYPE)
493 #define OBD_IOC_NAME2DEV _IOWR('f', 127, OBD_IOC_DATA_TYPE)
494 #define OBD_IOC_UUID2DEV _IOWR('f', 130, OBD_IOC_DATA_TYPE)
495 #define OBD_IOC_GETNAME _IOWR('f', 131, OBD_IOC_DATA_TYPE)
497 #define OBD_IOC_LOV_GET_CONFIG _IOWR('f', 132, OBD_IOC_DATA_TYPE)
498 #define OBD_IOC_CLIENT_RECOVER _IOW ('f', 133, OBD_IOC_DATA_TYPE)
499 #define OBD_IOC_PING_TARGET _IOW ('f', 136, OBD_IOC_DATA_TYPE)
501 #define OBD_IOC_DEC_FS_USE_COUNT _IO ('f', 139 )
502 #define OBD_IOC_NO_TRANSNO _IOW ('f', 140, OBD_IOC_DATA_TYPE)
503 #define OBD_IOC_SET_READONLY _IOW ('f', 141, OBD_IOC_DATA_TYPE)
504 #define OBD_IOC_ABORT_RECOVERY _IOR ('f', 142, OBD_IOC_DATA_TYPE)
506 #define OBD_IOC_ROOT_SQUASH _IOWR('f', 143, OBD_IOC_DATA_TYPE)
508 #define OBD_GET_VERSION _IOWR ('f', 144, OBD_IOC_DATA_TYPE)
510 #define OBD_IOC_GSS_SUPPORT _IOWR('f', 145, OBD_IOC_DATA_TYPE)
512 #define OBD_IOC_CLOSE_UUID _IOWR ('f', 147, OBD_IOC_DATA_TYPE)
514 #define OBD_IOC_CHANGELOG_SEND _IOW ('f', 148, OBD_IOC_DATA_TYPE)
515 #define OBD_IOC_GETDEVICE _IOWR ('f', 149, OBD_IOC_DATA_TYPE)
516 #define OBD_IOC_FID2PATH _IOWR ('f', 150, OBD_IOC_DATA_TYPE)
517 /* see also <lustre/lustre_user.h> for ioctls 151-153 */
518 /* OBD_IOC_LOV_SETSTRIPE: See also LL_IOC_LOV_SETSTRIPE */
519 #define OBD_IOC_LOV_SETSTRIPE _IOW ('f', 154, OBD_IOC_DATA_TYPE)
520 /* OBD_IOC_LOV_GETSTRIPE: See also LL_IOC_LOV_GETSTRIPE */
521 #define OBD_IOC_LOV_GETSTRIPE _IOW ('f', 155, OBD_IOC_DATA_TYPE)
522 /* OBD_IOC_LOV_SETEA: See also LL_IOC_LOV_SETEA */
523 #define OBD_IOC_LOV_SETEA _IOW ('f', 156, OBD_IOC_DATA_TYPE)
524 /* see <lustre/lustre_user.h> for ioctls 157-159 */
525 /* OBD_IOC_QUOTACHECK: See also LL_IOC_QUOTACHECK */
526 #define OBD_IOC_QUOTACHECK _IOW ('f', 160, int)
527 /* OBD_IOC_POLL_QUOTACHECK: See also LL_IOC_POLL_QUOTACHECK */
528 #define OBD_IOC_POLL_QUOTACHECK _IOR ('f', 161, struct if_quotacheck *)
529 /* OBD_IOC_QUOTACTL: See also LL_IOC_QUOTACTL */
530 #define OBD_IOC_QUOTACTL _IOWR('f', 162, struct if_quotactl)
531 /* see also <lustre/lustre_user.h> for ioctls 163-176 */
532 #define OBD_IOC_CHANGELOG_REG _IOW ('f', 177, struct obd_ioctl_data)
533 #define OBD_IOC_CHANGELOG_DEREG _IOW ('f', 178, struct obd_ioctl_data)
534 #define OBD_IOC_CHANGELOG_CLEAR _IOW ('f', 179, struct obd_ioctl_data)
535 #define OBD_IOC_RECORD _IOWR('f', 180, OBD_IOC_DATA_TYPE)
536 #define OBD_IOC_ENDRECORD _IOWR('f', 181, OBD_IOC_DATA_TYPE)
537 #define OBD_IOC_PARSE _IOWR('f', 182, OBD_IOC_DATA_TYPE)
538 #define OBD_IOC_DORECORD _IOWR('f', 183, OBD_IOC_DATA_TYPE)
539 #define OBD_IOC_PROCESS_CFG _IOWR('f', 184, OBD_IOC_DATA_TYPE)
540 #define OBD_IOC_DUMP_LOG _IOWR('f', 185, OBD_IOC_DATA_TYPE)
541 #define OBD_IOC_CLEAR_LOG _IOWR('f', 186, OBD_IOC_DATA_TYPE)
542 #define OBD_IOC_PARAM _IOW ('f', 187, OBD_IOC_DATA_TYPE)
543 #define OBD_IOC_POOL _IOWR('f', 188, OBD_IOC_DATA_TYPE)
545 #define OBD_IOC_CATLOGLIST _IOWR('f', 190, OBD_IOC_DATA_TYPE)
546 #define OBD_IOC_LLOG_INFO _IOWR('f', 191, OBD_IOC_DATA_TYPE)
547 #define OBD_IOC_LLOG_PRINT _IOWR('f', 192, OBD_IOC_DATA_TYPE)
548 #define OBD_IOC_LLOG_CANCEL _IOWR('f', 193, OBD_IOC_DATA_TYPE)
549 #define OBD_IOC_LLOG_REMOVE _IOWR('f', 194, OBD_IOC_DATA_TYPE)
550 #define OBD_IOC_LLOG_CHECK _IOWR('f', 195, OBD_IOC_DATA_TYPE)
551 #define OBD_IOC_LLOG_CATINFO _IOWR('f', 196, OBD_IOC_DATA_TYPE)
553 #define ECHO_IOC_GET_STRIPE _IOWR('f', 200, OBD_IOC_DATA_TYPE)
554 #define ECHO_IOC_SET_STRIPE _IOWR('f', 201, OBD_IOC_DATA_TYPE)
555 #define ECHO_IOC_ENQUEUE _IOWR('f', 202, OBD_IOC_DATA_TYPE)
556 #define ECHO_IOC_CANCEL _IOWR('f', 203, OBD_IOC_DATA_TYPE)
558 #define OBD_IOC_GET_OBJ_VERSION _IOR('f', 210, OBD_IOC_DATA_TYPE)
560 #define OBD_IOC_GET_MNTOPT _IOW('f', 220, mntopt_t)
562 #define OBD_IOC_ECHO_MD _IOR('f', 221, struct obd_ioctl_data)
563 #define OBD_IOC_ECHO_ALLOC_SEQ _IOWR('f', 222, struct obd_ioctl_data)
564 /* XXX _IOWR('f', 250, long) has been defined in
565 * libcfs/include/libcfs/libcfs_private.h for debug, don't use it
568 /* Until such time as we get_info the per-stripe maximum from the OST,
569 * we define this to be 2T - 4k, which is the ext3 maxbytes. */
570 #define LUSTRE_STRIPE_MAXBYTES 0x1fffffff000ULL
572 /* Special values for remove LOV EA from disk */
573 #define LOVEA_DELETE_VALUES(size, count, offset) (size == 0 && count == 0 && \
574 offset == (typeof(offset))(-1))
576 /* #define POISON_BULK 0 */
579 * l_wait_event is a flexible sleeping function, permitting simple caller
580 * configuration of interrupt and timeout sensitivity along with actions to
581 * be performed in the event of either exception.
583 * The first form of usage looks like this:
585 * struct l_wait_info lwi = LWI_TIMEOUT_INTR(timeout, timeout_handler,
586 * intr_handler, callback_data);
587 * rc = l_wait_event(waitq, condition, &lwi);
589 * l_wait_event() makes the current process wait on 'waitq' until 'condition'
590 * is TRUE or a "killable" signal (SIGTERM, SIKGILL, SIGINT) is pending. It
591 * returns 0 to signify 'condition' is TRUE, but if a signal wakes it before
592 * 'condition' becomes true, it optionally calls the specified 'intr_handler'
593 * if not NULL, and returns -EINTR.
595 * If a non-zero timeout is specified, signals are ignored until the timeout
596 * has expired. At this time, if 'timeout_handler' is not NULL it is called.
597 * If it returns FALSE l_wait_event() continues to wait as described above with
598 * signals enabled. Otherwise it returns -ETIMEDOUT.
600 * LWI_INTR(intr_handler, callback_data) is shorthand for
601 * LWI_TIMEOUT_INTR(0, NULL, intr_handler, callback_data)
603 * The second form of usage looks like this:
605 * struct l_wait_info lwi = LWI_TIMEOUT(timeout, timeout_handler);
606 * rc = l_wait_event(waitq, condition, &lwi);
608 * This form is the same as the first except that it COMPLETELY IGNORES
609 * SIGNALS. The caller must therefore beware that if 'timeout' is zero, or if
610 * 'timeout_handler' is not NULL and returns FALSE, then the ONLY thing that
611 * can unblock the current process is 'condition' becoming TRUE.
613 * Another form of usage is:
614 * struct l_wait_info lwi = LWI_TIMEOUT_INTERVAL(timeout, interval,
616 * rc = l_wait_event(waitq, condition, &lwi);
617 * This is the same as previous case, but condition is checked once every
618 * 'interval' jiffies (if non-zero).
620 * Subtle synchronization point: this macro does *not* necessary takes
621 * wait-queue spin-lock before returning, and, hence, following idiom is safe
622 * ONLY when caller provides some external locking:
626 * l_wait_event(&obj->wq, ....); (1)
628 * wake_up(&obj->wq): (2)
629 * spin_lock(&q->lock); (2.1)
630 * __wake_up_common(q, ...); (2.2)
631 * spin_unlock(&q->lock, flags); (2.3)
633 * OBD_FREE_PTR(obj); (3)
635 * As l_wait_event() may "short-cut" execution and return without taking
636 * wait-queue spin-lock, some additional synchronization is necessary to
637 * guarantee that step (3) can begin only after (2.3) finishes.
639 * XXX nikita: some ptlrpc daemon threads have races of that sort.
642 static inline int back_to_sleep(void *arg)
647 #define LWI_ON_SIGNAL_NOOP ((void (*)(void *))(-1))
650 cfs_duration_t lwi_timeout;
651 cfs_duration_t lwi_interval;
653 int (*lwi_on_timeout)(void *);
654 void (*lwi_on_signal)(void *);
658 /* NB: LWI_TIMEOUT ignores signals completely */
659 #define LWI_TIMEOUT(time, cb, data) \
660 ((struct l_wait_info) { \
661 .lwi_timeout = time, \
662 .lwi_on_timeout = cb, \
663 .lwi_cb_data = data, \
665 .lwi_allow_intr = 0 \
668 #define LWI_TIMEOUT_INTERVAL(time, interval, cb, data) \
669 ((struct l_wait_info) { \
670 .lwi_timeout = time, \
671 .lwi_on_timeout = cb, \
672 .lwi_cb_data = data, \
673 .lwi_interval = interval, \
674 .lwi_allow_intr = 0 \
677 #define LWI_TIMEOUT_INTR(time, time_cb, sig_cb, data) \
678 ((struct l_wait_info) { \
679 .lwi_timeout = time, \
680 .lwi_on_timeout = time_cb, \
681 .lwi_on_signal = sig_cb, \
682 .lwi_cb_data = data, \
684 .lwi_allow_intr = 0 \
687 #define LWI_TIMEOUT_INTR_ALL(time, time_cb, sig_cb, data) \
688 ((struct l_wait_info) { \
689 .lwi_timeout = time, \
690 .lwi_on_timeout = time_cb, \
691 .lwi_on_signal = sig_cb, \
692 .lwi_cb_data = data, \
694 .lwi_allow_intr = 1 \
697 #define LWI_INTR(cb, data) LWI_TIMEOUT_INTR(0, NULL, cb, data)
702 * wait for @condition to become true, but no longer than timeout, specified
705 #define __l_wait_event(wq, condition, info, ret, l_add_wait) \
707 cfs_waitlink_t __wait; \
708 cfs_duration_t __timeout = info->lwi_timeout; \
709 cfs_sigset_t __blocked; \
710 int __allow_intr = info->lwi_allow_intr; \
716 cfs_waitlink_init(&__wait); \
717 l_add_wait(&wq, &__wait); \
719 /* Block all signals (just the non-fatal ones if no timeout). */ \
720 if (info->lwi_on_signal != NULL && (__timeout == 0 || __allow_intr)) \
721 __blocked = cfs_block_sigsinv(LUSTRE_FATAL_SIGS); \
723 __blocked = cfs_block_sigsinv(0); \
728 __wstate = info->lwi_on_signal != NULL && \
729 (__timeout == 0 || __allow_intr) ? \
730 CFS_TASK_INTERRUPTIBLE : CFS_TASK_UNINT; \
732 cfs_set_current_state(CFS_TASK_INTERRUPTIBLE); \
737 if (__timeout == 0) { \
738 cfs_waitq_wait(&__wait, __wstate); \
740 cfs_duration_t interval = info->lwi_interval? \
741 min_t(cfs_duration_t, \
742 info->lwi_interval,__timeout):\
744 cfs_duration_t remaining = cfs_waitq_timedwait(&__wait,\
747 __timeout = cfs_time_sub(__timeout, \
748 cfs_time_sub(interval, remaining));\
749 if (__timeout == 0) { \
750 if (info->lwi_on_timeout == NULL || \
751 info->lwi_on_timeout(info->lwi_cb_data)) { \
755 /* Take signals after the timeout expires. */ \
756 if (info->lwi_on_signal != NULL) \
757 (void)cfs_block_sigsinv(LUSTRE_FATAL_SIGS);\
763 if (cfs_signal_pending()) { \
764 if (info->lwi_on_signal != NULL && \
765 (__timeout == 0 || __allow_intr)) { \
766 if (info->lwi_on_signal != LWI_ON_SIGNAL_NOOP) \
767 info->lwi_on_signal(info->lwi_cb_data);\
771 /* We have to do this here because some signals */ \
772 /* are not blockable - ie from strace(1). */ \
773 /* In these cases we want to schedule_timeout() */ \
774 /* again, because we don't want that to return */ \
775 /* -EINTR when the RPC actually succeeded. */ \
776 /* the RECALC_SIGPENDING below will deliver the */ \
777 /* signal properly. */ \
778 cfs_clear_sigpending(); \
782 cfs_block_sigs(__blocked); \
784 cfs_set_current_state(CFS_TASK_RUNNING); \
785 cfs_waitq_del(&wq, &__wait); \
788 #else /* !__KERNEL__ */
789 #define __l_wait_event(wq, condition, info, ret, l_add_wait) \
791 long __timeout = info->lwi_timeout; \
794 int __timed_out = 0; \
795 int __interval = obd_timeout; \
801 if (__timeout != 0) \
802 __then = time(NULL); \
804 if (__timeout && __timeout < __interval) \
805 __interval = __timeout; \
806 if (info->lwi_interval && info->lwi_interval < __interval) \
807 __interval = info->lwi_interval; \
809 while (!(condition)) { \
810 liblustre_wait_event(__interval); \
814 if (!__timed_out && info->lwi_timeout != 0) { \
815 __now = time(NULL); \
816 __timeout -= __now - __then; \
824 if (info->lwi_on_timeout == NULL || \
825 info->lwi_on_timeout(info->lwi_cb_data)) { \
831 SET_BUT_UNUSED(wq); \
834 #endif /* __KERNEL__ */
837 #define l_wait_event(wq, condition, info) \
840 struct l_wait_info *__info = (info); \
842 __l_wait_event(wq, condition, __info, \
843 __ret, cfs_waitq_add); \
847 #define l_wait_event_exclusive(wq, condition, info) \
850 struct l_wait_info *__info = (info); \
852 __l_wait_event(wq, condition, __info, \
853 __ret, cfs_waitq_add_exclusive); \
857 #define l_wait_event_exclusive_head(wq, condition, info) \
860 struct l_wait_info *__info = (info); \
862 __l_wait_event(wq, condition, __info, \
863 __ret, cfs_waitq_add_exclusive_head); \
867 #define l_wait_condition(wq, condition) \
869 struct l_wait_info lwi = { 0 }; \
870 l_wait_event(wq, condition, &lwi); \
873 #define l_wait_condition_exclusive(wq, condition) \
875 struct l_wait_info lwi = { 0 }; \
876 l_wait_event_exclusive(wq, condition, &lwi); \
879 #define l_wait_condition_exclusive_head(wq, condition) \
881 struct l_wait_info lwi = { 0 }; \
882 l_wait_event_exclusive_head(wq, condition, &lwi); \
886 #define LIBLUSTRE_CLIENT (0)
888 #define LIBLUSTRE_CLIENT (1)
893 #endif /* _LUSTRE_LIB_H */