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. <info@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 * User-space Lustre headers.
28 #error Kernel files should not #include <liblustre.h>
31 * The userspace implementations of linux/spinlock.h vary; we just
32 * include our own for all of them
34 #define __LINUX_SPINLOCK_H
41 #ifdef HAVE_ASM_PAGE_H
42 # include <asm/page.h>
44 #ifdef HAVE_SYS_USER_H
45 # include <sys/user.h>
47 #ifdef HAVE_SYS_IOCTL_H
48 # include <sys/ioctl.h>
55 #include <sys/ioctl.h>
67 #include <libcfs/list.h>
68 #include <lnet/lnet.h>
69 #include <libcfs/kp30.h>
71 /* definitions for liblustre */
75 #define CFS_PAGE_SHIFT 12
76 #define CFS_PAGE_SIZE (1UL << CFS_PAGE_SHIFT)
77 #define CFS_PAGE_MASK (~((__u64)CFS_PAGE_SIZE-1))
78 #define loff_t long long
80 typedef unsigned short umode_t;
84 #ifndef CURRENT_SECONDS
85 # define CURRENT_SECONDS time(0)
89 #define ARRAY_SIZE(a) ((sizeof (a))/(sizeof ((a)[0])))
92 /* This is because lprocfs_status.h gets included here indirectly. It would
93 * be much better to just avoid lprocfs being included into liblustre entirely
94 * but that requires more header surgery than I can handle right now.
96 #ifndef smp_processor_id
97 #define smp_processor_id() 0
99 #ifndef num_online_cpus
100 #define num_online_cpus() 1
102 #ifndef num_possible_cpus
103 #define num_possible_cpus() 1
106 /* always adopt 2.5 definitions */
107 #define KERNEL_VERSION(a,b,c) ((a)*100+(b)*10+c)
108 #define LINUX_VERSION_CODE KERNEL_VERSION(2,6,5)
111 #define page_private(page) ((page)->private)
112 #define set_page_private(page, v) ((page)->private = (v))
116 static inline void inter_module_put(void *a)
121 void *inter_module_get(char *arg);
126 void (*ws_task)(void *arg);
130 static inline void prepare_work(struct work_struct *q, void (*t)(void *),
138 static inline void schedule_work(struct work_struct *q)
140 q->ws_task(q->ws_arg);
144 #define strnlen(a,b) strlen(a)
145 static inline void *kmalloc(int size, int prot)
149 #define vmalloc malloc
151 #define kfree(a) free(a)
153 #define GFP_HIGHUSER 1
156 #define IS_ERR(a) ((unsigned long)(a) > (unsigned long)-1000L)
157 #define PTR_ERR(a) ((long)(a))
158 #define ERR_PTR(a) ((void*)((long)(a)))
160 typedef int (read_proc_t)(char *page, char **start, off_t off,
161 int count, int *eof, void *data);
163 struct file; /* forward ref */
164 typedef int (write_proc_t)(struct file *file, const char *buffer,
165 unsigned long count, void *data);
167 #define NIPQUAD(addr) \
168 ((unsigned char *)&addr)[0], \
169 ((unsigned char *)&addr)[1], \
170 ((unsigned char *)&addr)[2], \
171 ((unsigned char *)&addr)[3]
173 #if defined(__LITTLE_ENDIAN)
174 #define HIPQUAD(addr) \
175 ((unsigned char *)&addr)[3], \
176 ((unsigned char *)&addr)[2], \
177 ((unsigned char *)&addr)[1], \
178 ((unsigned char *)&addr)[0]
179 #elif defined(__BIG_ENDIAN)
180 #define HIPQUAD NIPQUAD
182 #error "Undefined byteorder??"
183 #endif /* __LITTLE_ENDIAN */
187 /* a long can be more than 32 bits, so use BITS_PER_LONG
188 * to allow the compiler to adjust the bit shifting accordingly
191 /* test if bit nr is set in bitmap addr; returns previous value of bit nr */
192 static __inline__ int set_bit(int nr, long * addr)
196 addr += nr / BITS_PER_LONG;
197 mask = 1UL << (nr & (BITS_PER_LONG - 1));
198 nr = (mask & *addr) != 0;
203 /* clear bit nr in bitmap addr; returns previous value of bit nr*/
204 static __inline__ int clear_bit(int nr, long * addr)
208 addr += nr / BITS_PER_LONG;
209 mask = 1UL << (nr & (BITS_PER_LONG - 1));
210 nr = (mask & *addr) != 0;
215 static __inline__ int test_bit(int nr, const long * addr)
217 return ((1UL << (nr & (BITS_PER_LONG - 1))) & ((addr)[nr / BITS_PER_LONG])) != 0;
220 static __inline__ int ext2_set_bit(int nr, void *addr)
222 return set_bit(nr, (long*)addr);
225 static __inline__ int ext2_clear_bit(int nr, void *addr)
227 return clear_bit(nr, (long*)addr);
230 static __inline__ int ext2_test_bit(int nr, void *addr)
232 return test_bit(nr, (long*)addr);
241 static inline void MODULE_AUTHOR(char *name)
243 printf("%s\n", name);
245 #define MODULE_DESCRIPTION(name) MODULE_AUTHOR(name)
246 #define MODULE_LICENSE(name) MODULE_AUTHOR(name)
248 #define THIS_MODULE NULL
254 static inline int misc_register(void *foo)
259 static inline int misc_deregister(void *foo)
264 static inline int request_module(char *name)
269 #define __MOD_INC_USE_COUNT(m) do {} while (0)
270 #define __MOD_DEC_USE_COUNT(m) do {} while (0)
271 #define MOD_INC_USE_COUNT do {} while (0)
272 #define MOD_DEC_USE_COUNT do {} while (0)
273 static inline void __module_get(struct module *module)
277 static inline int try_module_get(struct module *module)
282 static inline void module_put(struct module *module)
286 /* module initialization */
287 extern int init_obdclass(void);
288 extern int ptlrpc_init(void);
289 extern int ldlm_init(void);
290 extern int osc_init(void);
291 extern int lov_init(void);
292 extern int mdc_init(void);
293 extern int lmv_init(void);
294 extern int mgc_init(void);
295 extern int echo_client_init(void);
301 #define EXPORT_SYMBOL(S)
305 typedef struct { } spinlock_t;
306 typedef __u64 kdev_t;
308 #define SPIN_LOCK_UNLOCKED (spinlock_t) { }
309 #define LASSERT_SPIN_LOCKED(lock) do {} while(0)
310 #define LASSERT_SEM_LOCKED(sem) do {} while(0)
312 static inline void spin_lock(spinlock_t *l) {return;}
313 static inline void spin_unlock(spinlock_t *l) {return;}
314 static inline void spin_lock_init(spinlock_t *l) {return;}
315 static inline void local_irq_save(unsigned long flag) {return;}
316 static inline void local_irq_restore(unsigned long flag) {return;}
317 static inline int spin_is_locked(spinlock_t *l) {return 1;}
319 static inline void spin_lock_bh(spinlock_t *l) {}
320 static inline void spin_unlock_bh(spinlock_t *l) {}
321 static inline void spin_lock_irqsave(spinlock_t *a, unsigned long b) {}
322 static inline void spin_unlock_irqrestore(spinlock_t *a, unsigned long b) {}
324 typedef spinlock_t rwlock_t;
325 #define RW_LOCK_UNLOCKED SPIN_LOCK_UNLOCKED
326 #define read_lock(l) spin_lock(l)
327 #define read_unlock(l) spin_unlock(l)
328 #define write_lock(l) spin_lock(l)
329 #define write_unlock(l) spin_unlock(l)
332 #define min(x,y) ((x)<(y) ? (x) : (y))
333 #define max(x,y) ((x)>(y) ? (x) : (y))
336 #define min_t(type,x,y) \
337 ({ type __x = (x); type __y = (y); __x < __y ? __x: __y; })
340 #define max_t(type,x,y) \
341 ({ type __x = (x); type __y = (y); __x > __y ? __x: __y; })
344 #define simple_strtol strtol
346 /* registering symbols */
348 #define ERESTARTSYS ERESTART
354 void get_random_bytes(void *ptr, int size);
358 /* memory size: used for some client tunables */
359 #define num_physpages (256 * 1024) /* 1GB */
361 static inline int copy_from_user(void *a,void *b, int c)
367 static inline int copy_to_user(void *a,void *b, int c)
378 #define SLAB_HWCACHE_ALIGN 0
379 static inline kmem_cache_t *
380 kmem_cache_create(const char *name, size_t objsize, size_t cdum,
382 void (*e)(void *, kmem_cache_t *, unsigned long),
383 void (*f)(void *, kmem_cache_t *, unsigned long))
386 c = malloc(sizeof(*c));
390 CDEBUG(D_MALLOC, "alloc slab cache %s at %p, objsize %d\n",
391 name, c, (int)objsize);
395 static inline int kmem_cache_destroy(kmem_cache_t *a)
397 CDEBUG(D_MALLOC, "destroy slab cache %p, objsize %u\n", a, a->size);
402 /* struct page decl moved out from here into portals/include/libcfs/user-prim.h */
405 #define PAGE_LIST_ENTRY list
406 #define PAGE_LIST(page) ((page)->list)
408 #define kmap(page) (page)->addr
409 #define kunmap(a) do {} while (0)
411 static inline cfs_page_t *alloc_pages(int mask, unsigned long order)
413 cfs_page_t *pg = malloc(sizeof(*pg));
417 #if 0 //#ifdef MAP_ANONYMOUS
418 pg->addr = mmap(0, PAGE_SIZE << order, PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, 0, 0);
420 pg->addr = malloc(CFS_PAGE_SIZE << order);
429 #define cfs_alloc_pages(mask, order) alloc_pages((mask), (order))
431 #define alloc_page(mask) alloc_pages((mask), 0)
432 #define cfs_alloc_page(mask) alloc_page(mask)
434 static inline void __free_pages(cfs_page_t *pg, int what)
436 #if 0 //#ifdef MAP_ANONYMOUS
437 munmap(pg->addr, PAGE_SIZE);
443 #define __cfs_free_pages(pg, order) __free_pages((pg), (order))
445 #define __free_page(page) __free_pages((page), 0)
446 #define free_page(page) __free_page(page)
447 #define __cfs_free_page(page) __cfs_free_pages((page), 0)
449 static inline cfs_page_t* __grab_cache_page(unsigned long index)
451 cfs_page_t *pg = alloc_pages(0, 0);
458 #define grab_cache_page(index) __grab_cache_page(index)
459 #define page_cache_release(page) __free_pages(page, 0)
462 #define do_div(a,b) \
464 unsigned long remainder;\
465 remainder = (a) % (b); \
471 #define ATTR_MODE 0x0001
472 #define ATTR_UID 0x0002
473 #define ATTR_GID 0x0004
474 #define ATTR_SIZE 0x0008
475 #define ATTR_ATIME 0x0010
476 #define ATTR_MTIME 0x0020
477 #define ATTR_CTIME 0x0040
478 #define ATTR_ATIME_SET 0x0080
479 #define ATTR_MTIME_SET 0x0100
480 #define ATTR_FORCE 0x0200 /* Not a change, but a change it */
481 #define ATTR_ATTR_FLAG 0x0400
482 #define ATTR_RAW 0x0800 /* file system, not vfs will massage attrs */
483 #define ATTR_FROM_OPEN 0x1000 /* called from open path, ie O_TRUNC */
484 #define ATTR_CTIME_SET 0x2000
485 #define ATTR_BLOCKS 0x4000
486 #define ATTR_KILL_SUID 0
487 #define ATTR_KILL_SGID 0
490 unsigned int ia_valid;
498 unsigned int ia_attr_flags;
501 #define ll_iattr iattr
503 #define IT_OPEN 0x0001
504 #define IT_CREAT 0x0002
505 #define IT_READDIR 0x0004
506 #define IT_GETATTR 0x0008
507 #define IT_LOOKUP 0x0010
508 #define IT_UNLINK 0x0020
509 #define IT_GETXATTR 0x0040
510 #define IT_EXEC 0x0080
511 #define IT_PIN 0x0100
513 #define IT_FL_LOCKED 0x0001
514 #define IT_FL_FOLLOWED 0x0002 /* set by vfs_follow_link */
516 #define INTENT_MAGIC 0x19620323
518 struct lustre_intent_data {
521 __u64 it_lock_handle;
526 struct lookup_intent {
528 void (*it_op_release)(struct lookup_intent *);
533 struct lustre_intent_data lustre;
537 static inline void intent_init(struct lookup_intent *it, int op, int flags)
539 memset(it, 0, sizeof(*it));
540 it->it_magic = INTENT_MAGIC;
542 it->it_flags = flags;
554 struct rw_semaphore {
563 /* use the macro's argument to avoid unused warnings */
564 #define down(a) do { (void)a; } while (0)
565 #define mutex_down(a) down(a)
566 #define up(a) do { (void)a; } while (0)
567 #define mutex_up(a) up(a)
568 #define down_read(a) do { (void)a; } while (0)
569 #define up_read(a) do { (void)a; } while (0)
570 #define down_write(a) do { (void)a; } while (0)
571 #define up_write(a) do { (void)a; } while (0)
572 #define sema_init(a,b) do { (void)a; } while (0)
573 #define init_rwsem(a) do { (void)a; } while (0)
574 #define DECLARE_MUTEX(name) \
575 struct semaphore name = { 1 }
576 static inline void init_MUTEX (struct semaphore *sem)
580 static inline void init_MUTEX_LOCKED (struct semaphore *sem)
585 #define init_mutex(s) init_MUTEX(s)
588 struct list_head sleepers;
592 struct list_head sleeping;
602 struct signal pending;
615 typedef struct task_struct cfs_task_t;
616 #define cfs_current() current
617 #define cfs_curproc_pid() (current->pid)
618 #define cfs_curproc_comm() (current->comm)
620 extern struct task_struct *current;
621 int in_group_p(gid_t gid);
622 static inline int capable(int cap)
624 if (current->cap_effective & (1 << cap))
630 #define set_current_state(foo) do { current->state = foo; } while (0)
632 #define init_waitqueue_entry(q,p) do { (q)->process = p; } while (0)
633 #define add_wait_queue(q,p) do { list_add(&(q)->sleepers, &(p)->sleeping); } while (0)
634 #define del_wait_queue(p) do { list_del(&(p)->sleeping); } while (0)
635 #define remove_wait_queue(q,p) do { list_del(&(p)->sleeping); } while (0)
637 #define DECLARE_WAIT_QUEUE_HEAD(HEAD) \
638 wait_queue_head_t HEAD = { \
639 .sleepers = LIST_HEAD_INIT(HEAD.sleepers) \
641 #define init_waitqueue_head(l) INIT_LIST_HEAD(&(l)->sleepers)
642 #define wake_up(l) do { int a; a++; } while (0)
643 #define TASK_INTERRUPTIBLE 0
644 #define TASK_UNINTERRUPTIBLE 1
645 #define TASK_RUNNING 2
647 #define wait_event_interruptible(wq, condition) \
649 struct l_wait_info lwi; \
650 int timeout = 100000000;/* for ever */ \
653 lwi = LWI_TIMEOUT(timeout, NULL, NULL); \
654 ret = l_wait_event(NULL, condition, &lwi); \
659 #define in_interrupt() (0)
661 #define schedule() do {} while (0)
662 static inline int schedule_timeout(signed long t)
667 #define lock_kernel() do {} while (0)
668 #define unlock_kernel() do {} while (0)
669 #define daemonize(l) do {} while (0)
670 #define sigfillset(l) do {} while (0)
671 #define recalc_sigpending(l) do {} while (0)
672 #define kernel_thread(l,m,n) LBUG()
674 #define USERMODEHELPER(path, argv, envp) (0)
675 #define SIGNAL_MASK_ASSERT()
678 #include <sys/time.h>
680 #error "liblustre's jiffies currently expects HZ to be 1"
684 unsigned long _ret = 0; \
686 if (gettimeofday(&tv, NULL) == 0) \
690 #define get_jiffies_64() (__u64)jiffies
691 #define time_after(a, b) ((long)(b) - (long)(a) < 0)
692 #define time_before(a, b) time_after(b,a)
693 #define time_after_eq(a,b) ((long)(a) - (long)(b) >= 0)
696 struct list_head tl_list;
697 void (*function)(unsigned long unused);
702 static inline int timer_pending(struct timer_list *l)
704 if (time_after(l->expires, jiffies))
710 static inline int init_timer(struct timer_list *l)
712 INIT_LIST_HEAD(&l->tl_list);
716 static inline void mod_timer(struct timer_list *l, int thetime)
718 l->expires = thetime;
721 static inline void del_timer(struct timer_list *l)
726 typedef struct { volatile int counter; } atomic_t;
728 #define ATOMIC_INIT(i) { (i) }
729 #define atomic_read(a) ((a)->counter)
730 #define atomic_set(a,b) do {(a)->counter = b; } while (0)
731 #define atomic_dec_and_test(a) ((--((a)->counter)) == 0)
732 #define atomic_dec_and_lock(a,b) ((--((a)->counter)) == 0)
733 #define atomic_inc(a) (((a)->counter)++)
734 #define atomic_dec(a) do { (a)->counter--; } while (0)
735 #define atomic_add(b,a) do {(a)->counter += b;} while (0)
736 #define atomic_sub(b,a) do {(a)->counter -= b;} while (0)
739 #define likely(exp) (exp)
742 #define unlikely(exp) (exp)
745 #define might_sleep()
746 #define might_sleep_if(c)
750 int test_and_set_bit(int nr, unsigned long *addr)
754 while (nr >= sizeof(long)) {
759 oldbit = (*addr) & (1 << nr);
765 int test_and_clear_bit(int nr, unsigned long *addr)
769 while (nr >= sizeof(long)) {
774 oldbit = (*addr) & (1 << nr);
779 /* FIXME sys/capability will finally included linux/fs.h thus
780 * cause numerous trouble on x86-64. as temporary solution for
781 * build broken at Cray, we copy definition we need from capability.h
785 typedef struct _cap_struct *cap_t;
786 typedef int cap_value_t;
797 #define CAP_DAC_OVERRIDE 1
798 #define CAP_DAC_READ_SEARCH 2
801 #define CAP_SYS_ADMIN 21
803 cap_t cap_get_proc(void);
804 int cap_get_flag(cap_t, cap_value_t, cap_flag_t, cap_flag_value_t *);
807 static inline int llog_init_commit_master(void) { return 0; }
808 static inline int llog_cleanup_commit_master(int force) { return 0; }
809 static inline void libcfs_run_lbug_upcall(char *file, const char *fn,
818 #define COMPLETION_INITIALIZER(work) \
819 { 0, __WAIT_QUEUE_HEAD_INITIALIZER((work).wait) }
821 #define DECLARE_COMPLETION(work) \
822 struct completion work = COMPLETION_INITIALIZER(work)
824 #define INIT_COMPLETION(x) ((x).done = 0)
826 static inline void init_completion(struct completion *x)
829 init_waitqueue_head(&x->wait);
832 struct liblustre_wait_callback {
833 struct list_head llwc_list;
834 const char *llwc_name;
835 int (*llwc_fn)(void *arg);
839 void *liblustre_register_wait_callback(const char *name,
840 int (*fn)(void *arg), void *arg);
841 void liblustre_deregister_wait_callback(void *notifier);
842 int liblustre_wait_event(int timeout);
844 void *liblustre_register_idle_callback(const char *name,
845 int (*fn)(void *arg), void *arg);
846 void liblustre_deregister_idle_callback(void *notifier);
847 void liblustre_wait_idle(void);
850 struct nfs_lock_info {
856 typedef struct file_lock {
857 struct file_lock *fl_next; /* singly linked list for this inode */
858 struct list_head fl_link; /* doubly linked list of all locks */
859 struct list_head fl_block; /* circular list of blocked processes */
863 struct file *fl_file;
864 unsigned char fl_flags;
865 unsigned char fl_type;
869 void (*fl_notify)(struct file_lock *); /* unblock callback */
870 void (*fl_insert)(struct file_lock *); /* lock insertion callback */
871 void (*fl_remove)(struct file_lock *); /* lock removal callback */
873 void *fl_fasync; /* for lease break notifications */
874 unsigned long fl_break_time; /* for nonblocking lease breaks */
877 struct nfs_lock_info nfs_fl;
881 #define cfs_flock_type(fl) ((fl)->fl_type)
882 #define cfs_flock_set_type(fl, type) do { (fl)->fl_type = (type); } while(0)
883 #define cfs_flock_pid(fl) ((fl)->fl_pid)
884 #define cfs_flock_set_pid(fl, pid) do { (fl)->fl_pid = (pid); } while(0)
885 #define cfs_flock_start(fl) ((fl)->fl_start)
886 #define cfs_flock_set_start(fl, start) do { (fl)->fl_start = (start); } while(0)
887 #define cfs_flock_end(fl) ((fl)->fl_end)
888 #define cfs_flock_set_end(fl, end) do { (fl)->fl_end = (end); } while(0)
891 #define INT_LIMIT(x) (~((x)1 << (sizeof(x)*8 - 1)))
892 #define OFFSET_MAX INT_LIMIT(loff_t)
895 /* XXX: defined in kernel */
904 struct posix_acl_entry {
906 unsigned short e_perm;
912 unsigned int a_count;
913 struct posix_acl_entry a_entries[0];
924 xattr_acl_entry a_entries[0];
927 static inline size_t xattr_acl_size(int count)
929 return sizeof(xattr_acl_header) + count * sizeof(xattr_acl_entry);
933 struct posix_acl * posix_acl_from_xattr(const void *value, size_t size)
939 int posix_acl_valid(const struct posix_acl *acl)
945 void posix_acl_release(struct posix_acl *acl)
950 #define ENOTSUPP ENOTSUP
953 typedef int mm_segment_t;
958 static inline mm_segment_t get_fs(void)
963 static inline void set_fs(mm_segment_t seg)
967 #include <obd_support.h>
968 #include <lustre/lustre_idl.h>
969 #include <lustre_lib.h>
970 #include <lustre_import.h>
971 #include <lustre_export.h>
972 #include <lustre_net.h>
974 /* Fast hashing routine for a long.
975 (C) 2002 William Lee Irwin III, IBM */
978 * Knuth recommends primes in approximately golden ratio to the maximum
979 * integer representable by a machine word for multiplicative hashing.
980 * Chuck Lever verified the effectiveness of this technique:
981 * http://www.citi.umich.edu/techreports/reports/citi-tr-00-1.pdf
983 * These primes are chosen to be bit-sparse, that is operations on
984 * them can use shifts and additions instead of multiplications for
985 * machines where multiplications are slow.
987 #if BITS_PER_LONG == 32
988 /* 2^31 + 2^29 - 2^25 + 2^22 - 2^19 - 2^16 + 1 */
989 #define GOLDEN_RATIO_PRIME 0x9e370001UL
990 #elif BITS_PER_LONG == 64
991 /* 2^63 + 2^61 - 2^57 + 2^54 - 2^51 - 2^18 + 1 */
992 #define GOLDEN_RATIO_PRIME 0x9e37fffffffc0001UL
994 #error Define GOLDEN_RATIO_PRIME for your wordsize.
997 static inline unsigned long hash_long(unsigned long val, unsigned int bits)
999 unsigned long hash = val;
1001 #if BITS_PER_LONG == 64
1002 /* Sigh, gcc can't optimise this alone like it does for 32 bits. */
1003 unsigned long n = hash;
1017 /* On some cpus multiply is faster, on others gcc will do shifts */
1018 hash *= GOLDEN_RATIO_PRIME;
1021 /* High bits are more random, so use them. */
1022 return hash >> (BITS_PER_LONG - bits);
1025 static inline unsigned long hash_ptr(void *ptr, unsigned int bits)
1027 return hash_long((unsigned long)ptr, bits);