1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2001 Cluster File Systems, Inc. <braam@clusterfs.com>
6 * This file is part of Lustre, http://www.lustre.org.
8 * Lustre is free software; you can redistribute it and/or
9 * modify it under the terms of version 2 of the GNU General Public
10 * License as published by the Free Software Foundation.
12 * Lustre 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.
17 * You should have received a copy of the GNU General Public License
18 * along with Lustre; if not, write to the Free Software
19 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 * Basic Lustre library routines.
30 # include <sys/types.h>
32 # include <asm/semaphore.h>
33 # include <linux/rwsem.h>
34 # include <linux/sched.h>
35 # include <linux/signal.h>
36 # include <linux/types.h>
38 #include <linux/kp30.h>
39 #include <linux/lustre_idl.h>
40 #include <linux/lustre_cfg.h>
43 #if BITS_PER_LONG > 32
44 # define LI_POISON ((int)0x5a5a5a5a5a5a5a5a)
45 # define LL_POISON ((long)0x5a5a5a5a5a5a5a5a)
46 # define LP_POISON ((void *)(long)0x5a5a5a5a5a5a5a5a)
48 # define LI_POISON ((int)0x5a5a5a5a)
49 # define LL_POISON ((long)0x5a5a5a5a)
50 # define LP_POISON ((void *)(long)0x5a5a5a5a)
55 /* x86_64 has 64bit longs and defines u64 as long long */
56 #if BITS_PER_LONG > 32 && !defined(__x86_64__)
67 /* lustre_id output helper macros */
68 #define DLID4 "%lu/%lu/%lu/%lu"
71 (unsigned long)(id)->li_fid.lf_id, \
72 (unsigned long)(id)->li_fid.lf_group, \
73 (unsigned long)(id)->li_stc.u.e3s.l3s_ino, \
74 (unsigned long)(id)->li_stc.u.e3s.l3s_gen
77 * this flag is set in ->it->it_int_flags to show, that inode exists on client
78 * and no fid fetching flags needs to be set in server request.
80 #define LL_IT_EXIST (1 << 0)
83 struct ptlrpc_request;
87 #include <linux/lustre_ha.h>
88 #include <linux/lustre_net.h>
89 #include <linux/lustre_compat25.h>
90 #include <linux/lvfs.h>
92 int target_handle_connect(struct ptlrpc_request *req);
93 int target_handle_disconnect(struct ptlrpc_request *req);
94 void target_destroy_export(struct obd_export *exp);
95 int target_handle_reconnect(struct lustre_handle *conn, struct obd_export *exp,
96 struct obd_uuid *cluuid, int);
97 int target_handle_ping(struct ptlrpc_request *req);
98 void target_cancel_recovery_timer(struct obd_device *obd);
100 #define OBD_RECOVERY_TIMEOUT (obd_timeout * 5 * HZ / 2) /* *waves hands* */
101 void target_start_recovery_timer(struct obd_device *obd);
102 int target_start_recovery_thread(struct obd_device *obd,
103 svc_handler_t handler);
104 void target_stop_recovery_thread(struct obd_device *obd);
105 void target_cleanup_recovery(struct obd_device *obd);
106 int target_queue_recovery_request(struct ptlrpc_request *req,
107 struct obd_device *obd);
108 int target_queue_final_reply(struct ptlrpc_request *req, int rc);
109 void target_send_reply(struct ptlrpc_request *req, int rc, int fail_id);
113 int client_sanobd_setup(struct obd_device *obddev, obd_count len, void *buf);
114 struct client_obd *client_conn2cli(struct lustre_handle *conn);
116 struct mdc_open_data;
117 struct obd_client_handle {
118 struct lustre_handle och_fh;
119 struct llog_cookie och_cookie;
120 struct mdc_open_data *och_mod;
123 #define OBD_CLIENT_HANDLE_MAGIC 0xd15ea5ed
126 void statfs_pack(struct obd_statfs *osfs, struct kstatfs *sfs);
127 void statfs_unpack(struct kstatfs *sfs, struct obd_statfs *osfs);
132 struct task_struct *l_owner;
133 struct semaphore l_sem;
137 void l_lock_init(struct lustre_lock *);
138 void l_lock(struct lustre_lock *);
139 void l_unlock(struct lustre_lock *);
140 int l_has_lock(struct lustre_lock *);
146 #define OBD_IOCTL_VERSION 0x00010004
148 struct obd_ioctl_data {
150 uint32_t ioc_version;
156 struct obdo ioc_obdo1;
157 struct obdo ioc_obdo2;
162 uint32_t ioc_command;
168 /* buffers the kernel will treat as user pointers */
174 /* inline buffers for various arguments */
175 uint32_t ioc_inllen1;
177 uint32_t ioc_inllen2;
179 uint32_t ioc_inllen3;
181 uint32_t ioc_inllen4;
187 struct obd_ioctl_hdr {
189 uint32_t ioc_version;
192 static inline int obd_ioctl_packlen(struct obd_ioctl_data *data)
194 int len = size_round(sizeof(struct obd_ioctl_data));
195 len += size_round(data->ioc_inllen1);
196 len += size_round(data->ioc_inllen2);
197 len += size_round(data->ioc_inllen3);
198 len += size_round(data->ioc_inllen4);
203 static inline int obd_ioctl_is_invalid(struct obd_ioctl_data *data)
205 if (data->ioc_len > (1<<30)) {
206 CERROR("OBD ioctl: ioc_len larger than 1<<30\n");
209 if (data->ioc_inllen1 > (1<<30)) {
210 CERROR("OBD ioctl: ioc_inllen1 larger than 1<<30\n");
213 if (data->ioc_inllen2 > (1<<30)) {
214 CERROR("OBD ioctl: ioc_inllen2 larger than 1<<30\n");
217 if (data->ioc_inllen3 > (1<<30)) {
218 CERROR("OBD ioctl: ioc_inllen3 larger than 1<<30\n");
221 if (data->ioc_inllen4 > (1<<30)) {
222 CERROR("OBD ioctl: ioc_inllen4 larger than 1<<30\n");
225 if (data->ioc_inlbuf1 && !data->ioc_inllen1) {
226 CERROR("OBD ioctl: inlbuf1 pointer but 0 length\n");
229 if (data->ioc_inlbuf2 && !data->ioc_inllen2) {
230 CERROR("OBD ioctl: inlbuf2 pointer but 0 length\n");
233 if (data->ioc_inlbuf3 && !data->ioc_inllen3) {
234 CERROR("OBD ioctl: inlbuf3 pointer but 0 length\n");
237 if (data->ioc_inlbuf4 && !data->ioc_inllen4) {
238 CERROR("OBD ioctl: inlbuf4 pointer but 0 length\n");
241 if (data->ioc_pbuf1 && !data->ioc_plen1) {
242 CERROR("OBD ioctl: pbuf1 pointer but 0 length\n");
245 if (data->ioc_pbuf2 && !data->ioc_plen2) {
246 CERROR("OBD ioctl: pbuf2 pointer but 0 length\n");
249 if (data->ioc_plen1 && !data->ioc_pbuf1) {
250 CERROR("OBD ioctl: plen1 set but NULL pointer\n");
253 if (data->ioc_plen2 && !data->ioc_pbuf2) {
254 CERROR("OBD ioctl: plen2 set but NULL pointer\n");
257 if (obd_ioctl_packlen(data) != data->ioc_len) {
258 CERROR("OBD ioctl: packlen exceeds ioc_len (%d != %d)\n",
259 obd_ioctl_packlen(data), data->ioc_len);
266 static inline int obd_ioctl_pack(struct obd_ioctl_data *data, char **pbuf,
270 struct obd_ioctl_data *overlay;
271 data->ioc_len = obd_ioctl_packlen(data);
272 data->ioc_version = OBD_IOCTL_VERSION;
274 if (*pbuf && data->ioc_len > max)
277 *pbuf = malloc(data->ioc_len);
281 overlay = (struct obd_ioctl_data *)*pbuf;
282 memcpy(*pbuf, data, sizeof(*data));
284 ptr = overlay->ioc_bulk;
285 if (data->ioc_inlbuf1)
286 LOGL(data->ioc_inlbuf1, data->ioc_inllen1, ptr);
287 if (data->ioc_inlbuf2)
288 LOGL(data->ioc_inlbuf2, data->ioc_inllen2, ptr);
289 if (data->ioc_inlbuf3)
290 LOGL(data->ioc_inlbuf3, data->ioc_inllen3, ptr);
291 if (data->ioc_inlbuf4)
292 LOGL(data->ioc_inlbuf4, data->ioc_inllen4, ptr);
293 if (obd_ioctl_is_invalid(overlay))
299 static inline int obd_ioctl_unpack(struct obd_ioctl_data *data, char *pbuf,
303 struct obd_ioctl_data *overlay;
307 overlay = (struct obd_ioctl_data *)pbuf;
309 /* Preserve the caller's buffer pointers */
310 overlay->ioc_inlbuf1 = data->ioc_inlbuf1;
311 overlay->ioc_inlbuf2 = data->ioc_inlbuf2;
312 overlay->ioc_inlbuf3 = data->ioc_inlbuf3;
313 overlay->ioc_inlbuf4 = data->ioc_inlbuf4;
315 memcpy(data, pbuf, sizeof(*data));
317 ptr = overlay->ioc_bulk;
318 if (data->ioc_inlbuf1)
319 LOGU(data->ioc_inlbuf1, data->ioc_inllen1, ptr);
320 if (data->ioc_inlbuf2)
321 LOGU(data->ioc_inlbuf2, data->ioc_inllen2, ptr);
322 if (data->ioc_inlbuf3)
323 LOGU(data->ioc_inlbuf3, data->ioc_inllen3, ptr);
324 if (data->ioc_inlbuf4)
325 LOGU(data->ioc_inlbuf4, data->ioc_inllen4, ptr);
331 #include <linux/obd_support.h>
333 /* buffer MUST be at least the size of obd_ioctl_hdr */
334 static inline int obd_ioctl_getdata(char **buf, int *len, void *arg)
336 struct obd_ioctl_hdr hdr;
337 struct obd_ioctl_data *data;
342 err = copy_from_user(&hdr, (void *)arg, sizeof(hdr));
348 if (hdr.ioc_version != OBD_IOCTL_VERSION) {
349 CERROR("Version mismatch kernel vs application\n");
353 if (hdr.ioc_len > OBD_MAX_IOCTL_BUFFER) {
354 CERROR("User buffer len %d exceeds %d max buffer\n",
355 hdr.ioc_len, OBD_MAX_IOCTL_BUFFER);
359 if (hdr.ioc_len < sizeof(struct obd_ioctl_data)) {
360 CERROR("OBD: user buffer too small for ioctl\n");
364 /* XXX allocate this more intelligently, using kmalloc when
366 OBD_VMALLOC(*buf, hdr.ioc_len);
368 CERROR("Cannot allocate control buffer of len %d\n",
373 data = (struct obd_ioctl_data *)*buf;
375 err = copy_from_user(*buf, (void *)arg, hdr.ioc_len);
381 if (obd_ioctl_is_invalid(data)) {
382 CERROR("ioctl not correctly formatted\n");
386 if (data->ioc_inllen1) {
387 data->ioc_inlbuf1 = &data->ioc_bulk[0];
388 offset += size_round(data->ioc_inllen1);
391 if (data->ioc_inllen2) {
392 data->ioc_inlbuf2 = &data->ioc_bulk[0] + offset;
393 offset += size_round(data->ioc_inllen2);
396 if (data->ioc_inllen3) {
397 data->ioc_inlbuf3 = &data->ioc_bulk[0] + offset;
398 offset += size_round(data->ioc_inllen3);
401 if (data->ioc_inllen4) {
402 data->ioc_inlbuf4 = &data->ioc_bulk[0] + offset;
409 static inline void obd_ioctl_freedata(char *buf, int len)
418 #define OBD_IOC_CREATE _IOR ('f', 101, long)
419 #define OBD_IOC_DESTROY _IOW ('f', 104, long)
420 #define OBD_IOC_PREALLOCATE _IOWR('f', 105, long)
422 #define OBD_IOC_SETATTR _IOW ('f', 107, long)
423 #define OBD_IOC_GETATTR _IOR ('f', 108, long)
424 #define OBD_IOC_READ _IOWR('f', 109, long)
425 #define OBD_IOC_WRITE _IOWR('f', 110, long)
428 #define OBD_IOC_STATFS _IOWR('f', 113, long)
429 #define OBD_IOC_SYNC _IOR ('f', 114, long)
430 #define OBD_IOC_READ2 _IOWR('f', 115, long)
431 #define OBD_IOC_FORMAT _IOWR('f', 116, long)
432 #define OBD_IOC_PARTITION _IOWR('f', 117, long)
433 #define OBD_IOC_COPY _IOWR('f', 120, long)
434 #define OBD_IOC_MIGR _IOWR('f', 121, long)
435 #define OBD_IOC_PUNCH _IOWR('f', 122, long)
437 #define OBD_IOC_MODULE_DEBUG _IOWR('f', 124, long)
438 #define OBD_IOC_BRW_READ _IOWR('f', 125, long)
439 #define OBD_IOC_BRW_WRITE _IOWR('f', 126, long)
440 #define OBD_IOC_NAME2DEV _IOWR('f', 127, long)
441 #define OBD_IOC_UUID2DEV _IOWR('f', 130, long)
443 #define OBD_IOC_LOV_GET_CONFIG _IOWR('f', 132, long)
444 #define OBD_IOC_CLIENT_RECOVER _IOW ('f', 133, long)
446 #define OBD_IOC_PING _IOWR('f', 135, long)
448 #define OBD_IOC_DEC_FS_USE_COUNT _IO ('f', 139 )
449 #define OBD_IOC_NO_TRANSNO _IOW ('f', 140, long)
450 #define OBD_IOC_SET_READONLY _IOW ('f', 141, long)
451 #define OBD_IOC_ABORT_RECOVERY _IOR ('f', 142, long)
452 #define OBD_IOC_ROOT_SQUASH _IOWR('f', 143, long)
454 #define OBD_GET_VERSION _IOWR ('f', 144, long)
456 #define OBD_IOC_CLOSE_UUID _IOWR ('f', 147, long)
458 #define OBD_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
459 #define OBD_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
460 #define OBD_IOC_LOV_SETEA _IOW ('f', 156, long)
462 #define OBD_IOC_MOUNTOPT _IOWR('f', 170, long)
464 #define OBD_IOC_RECORD _IOWR('f', 180, long)
465 #define OBD_IOC_ENDRECORD _IOWR('f', 181, long)
466 #define OBD_IOC_PARSE _IOWR('f', 182, long)
467 #define OBD_IOC_DORECORD _IOWR('f', 183, long)
468 #define OBD_IOC_PROCESS_CFG _IOWR('f', 184, long)
469 #define OBD_IOC_DUMP_LOG _IOWR('f', 185, long)
470 #define OBD_IOC_CLEAR_LOG _IOWR('f', 186, long)
472 #define OBD_IOC_CATLOGLIST _IOWR('f', 190, long)
473 #define OBD_IOC_LLOG_INFO _IOWR('f', 191, long)
474 #define OBD_IOC_LLOG_PRINT _IOWR('f', 192, long)
475 #define OBD_IOC_LLOG_CANCEL _IOWR('f', 193, long)
476 #define OBD_IOC_LLOG_REMOVE _IOWR('f', 194, long)
477 #define OBD_IOC_LLOG_CHECK _IOWR('f', 195, long)
478 #define OBD_IOC_LLOG_CATINFO _IOWR('f', 196, long)
480 #define ECHO_IOC_GET_STRIPE _IOWR('f', 200, long)
481 #define ECHO_IOC_SET_STRIPE _IOWR('f', 201, long)
482 #define ECHO_IOC_ENQUEUE _IOWR('f', 202, long)
483 #define ECHO_IOC_CANCEL _IOWR('f', 203, long)
485 #define OBD_IOC_CMOBD_SYNC _IOWR('f', 210, long)
487 #define OBD_IOC_COBD_CON _IOWR('f', 220, long)
488 #define OBD_IOC_COBD_COFF _IOWR('f', 221, long)
489 #define OBD_IOC_COBD_CFLUSH _IOWR('f', 222, long)
491 #define OBD_IOC_SMFS_SNAP_ADD _IOWR('f', 230, long)
493 /* XXX _IOWR('f', 250, long) has been defined in
494 * portals/include/linux/kp30.h for debug, don't use it
497 /* Until such time as we get_info the per-stripe maximum from the OST,
498 * we define this to be 2T - 4k, which is the ext3 maxbytes. */
499 #define LUSTRE_STRIPE_MAXBYTES 0x1fffffff000ULL
501 #define CHECKSUM_BULK 0
502 #define POISON_BULK 0
505 static inline void ost_checksum(obd_count *cksum, void *addr, int len)
507 unsigned char *ptr = (unsigned char *)addr;
510 /* very stupid, but means I don't have to think about byte order */
514 *cksum = (*cksum << 2) + sum;
518 static inline int ll_insecure_random_int(void)
522 return (int)(t.tv_usec);
526 * l_wait_event is a flexible sleeping function, permitting simple caller
527 * configuration of interrupt and timeout sensitivity along with actions to
528 * be performed in the event of either exception.
530 * Common usage looks like this:
532 * struct l_wait_info lwi = LWI_TIMEOUT_INTR(timeout, timeout_handler,
533 * intr_handler, callback_data);
534 * rc = l_wait_event(waitq, condition, &lwi);
536 * (LWI_TIMEOUT and LWI_INTR macros are available for timeout- and
537 * interrupt-only variants, respectively.)
539 * If a timeout is specified, the timeout_handler will be invoked in the event
540 * that the timeout expires before the process is awakened. (Note that any
541 * waking of the process will restart the timeout, even if the condition is
542 * not satisfied and the process immediately returns to sleep. This might be
543 * considered a bug.) If the timeout_handler returns non-zero, l_wait_event
544 * will return -ETIMEDOUT and the caller will continue. If the handler returns
545 * zero instead, the process will go back to sleep until it is awakened by the
546 * waitq or some similar mechanism, or an interrupt occurs (if the caller has
547 * asked for interrupts to be detected). The timeout will only fire once, so
548 * callers should take care that a timeout_handler which returns zero will take
549 * future steps to awaken the process. N.B. that these steps must include
550 * making the provided condition become true.
552 * If the interrupt flag (lwi_signals) is non-zero, then the process will be
553 * interruptible, and will be awakened by any "killable" signal (SIGTERM,
554 * SIGKILL or SIGINT). If a timeout is also specified, then the process will
555 * only become interruptible _after_ the timeout has expired, though it can be
556 * awakened by a signal that was delivered before the timeout and is still
557 * pending when the timeout expires. If a timeout is not specified, the process
558 * will be interruptible at all times during l_wait_event.
563 int (*lwi_on_timeout)(void *);
565 void (*lwi_on_signal)(void *);
569 #define LWI_TIMEOUT(time, cb, data) \
570 ((struct l_wait_info) { \
572 lwi_on_timeout: cb, \
576 #define LWI_INTR(cb, data) \
577 ((struct l_wait_info) { \
583 #define LWI_TIMEOUT_INTR(time, time_cb, sig_cb, data) \
584 ((struct l_wait_info) { \
586 lwi_on_timeout: time_cb, \
588 lwi_on_signal: sig_cb, \
592 #define LUSTRE_FATAL_SIGS (sigmask(SIGKILL) | sigmask(SIGINT) | \
593 sigmask(SIGTERM) | sigmask(SIGQUIT) | \
597 static inline sigset_t l_w_e_set_sigs(int sigs)
600 unsigned long irqflags;
602 SIGNAL_MASK_LOCK(current, irqflags);
603 old = current->blocked;
604 siginitsetinv(¤t->blocked, sigs);
606 SIGNAL_MASK_UNLOCK(current, irqflags);
611 #define __l_wait_event(wq, condition, info, ret, excl) \
613 wait_queue_t __wait; \
614 int __timed_out = 0; \
615 unsigned long irqflags; \
618 init_waitqueue_entry(&__wait, current); \
620 add_wait_queue_exclusive(&wq, &__wait); \
622 add_wait_queue(&wq, &__wait); \
624 /* Block all signals (just the non-fatal ones if no timeout). */ \
625 if (info->lwi_signals && !info->lwi_timeout) \
626 blocked = l_w_e_set_sigs(LUSTRE_FATAL_SIGS); \
628 blocked = l_w_e_set_sigs(0); \
631 set_current_state(TASK_INTERRUPTIBLE); \
634 if (signal_pending(current)) { \
635 if (info->lwi_on_signal) \
636 info->lwi_on_signal(info->lwi_cb_data); \
640 if (info->lwi_timeout && !__timed_out) { \
641 if (schedule_timeout(info->lwi_timeout) == 0) { \
643 if (!info->lwi_on_timeout || \
644 info->lwi_on_timeout(info->lwi_cb_data)) { \
648 /* We'll take signals after a timeout. */ \
649 if (info->lwi_signals) \
650 (void)l_w_e_set_sigs(LUSTRE_FATAL_SIGS); \
657 SIGNAL_MASK_LOCK(current, irqflags); \
658 current->blocked = blocked; \
660 SIGNAL_MASK_UNLOCK(current, irqflags); \
662 current->state = TASK_RUNNING; \
663 remove_wait_queue(&wq, &__wait); \
666 #else /* !__KERNEL__ */
667 #define __l_wait_event(wq, condition, info, ret, excl) \
669 int timeout = info->lwi_timeout, elapse; \
670 int __timed_out = 0; \
677 if (liblustre_wait_event(timeout)) { \
680 elapse = (int) (time(NULL) - last); \
689 if (info->lwi_timeout && !__timed_out) { \
691 if (info->lwi_on_timeout == NULL || \
692 info->lwi_on_timeout(info->lwi_cb_data)) { \
700 #endif /* __KERNEL__ */
702 #define l_wait_event(wq, condition, info) \
705 struct l_wait_info *__info = (info); \
707 __l_wait_event(wq, condition, __info, __ret, 0); \
711 #define l_wait_event_exclusive(wq, condition, info) \
714 struct l_wait_info *__info = (info); \
716 __l_wait_event(wq, condition, __info, __ret, 1); \
720 #endif /* _LUSTRE_LIB_H */
722 #define LMD_MAGIC 0xbdacbdac
724 #define lmd_bad_magic(LMDP) \
726 struct lustre_mount_data *_lmd__ = (LMDP); \
729 CERROR("Missing mount data: " \
730 "check that /sbin/mount.lustre is installed.\n");\
732 } else if (_lmd__->lmd_magic != LMD_MAGIC) { \
733 CERROR("Invalid mount data (%#x != %#x): " \
734 "check that /sbin/mount.lustre is installed\n", \
735 _lmd__->lmd_magic, LMD_MAGIC); \