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>
70 #include <libcfs/user-bitops.h>
72 /* definitions for liblustre */
76 #define CFS_PAGE_SHIFT 12
77 #define CFS_PAGE_SIZE (1UL << CFS_PAGE_SHIFT)
78 #define CFS_PAGE_MASK (~((__u64)CFS_PAGE_SIZE-1))
79 #define loff_t long long
81 typedef unsigned short umode_t;
85 #ifndef CURRENT_SECONDS
86 # define CURRENT_SECONDS time(0)
90 #define ARRAY_SIZE(a) ((sizeof (a))/(sizeof ((a)[0])))
93 /* This is because lprocfs_status.h gets included here indirectly. It would
94 * be much better to just avoid lprocfs being included into liblustre entirely
95 * but that requires more header surgery than I can handle right now.
97 #ifndef smp_processor_id
98 #define smp_processor_id() 0
100 #ifndef num_online_cpus
101 #define num_online_cpus() 1
103 #ifndef num_possible_cpus
104 #define num_possible_cpus() 1
107 /* always adopt 2.5 definitions */
108 #define KERNEL_VERSION(a,b,c) ((a)*100+(b)*10+c)
109 #define LINUX_VERSION_CODE KERNEL_VERSION(2,6,5)
112 #define page_private(page) ((page)->private)
113 #define set_page_private(page, v) ((page)->private = (v))
117 static inline void inter_module_put(void *a)
122 void *inter_module_get(char *arg);
127 void (*ws_task)(void *arg);
131 static inline void prepare_work(struct work_struct *q, void (*t)(void *),
139 static inline void schedule_work(struct work_struct *q)
141 q->ws_task(q->ws_arg);
145 #define strnlen(a,b) strlen(a)
146 static inline void *kmalloc(int size, int prot)
150 #define vmalloc malloc
152 #define kfree(a) free(a)
154 #define GFP_HIGHUSER 1
157 #define IS_ERR(a) ((unsigned long)(a) > (unsigned long)-1000L)
158 #define PTR_ERR(a) ((long)(a))
159 #define ERR_PTR(a) ((void*)((long)(a)))
161 typedef int (read_proc_t)(char *page, char **start, off_t off,
162 int count, int *eof, void *data);
164 struct file; /* forward ref */
165 typedef int (write_proc_t)(struct file *file, const char *buffer,
166 unsigned long count, void *data);
168 #define NIPQUAD(addr) \
169 ((unsigned char *)&addr)[0], \
170 ((unsigned char *)&addr)[1], \
171 ((unsigned char *)&addr)[2], \
172 ((unsigned char *)&addr)[3]
174 #if defined(__LITTLE_ENDIAN)
175 #define HIPQUAD(addr) \
176 ((unsigned char *)&addr)[3], \
177 ((unsigned char *)&addr)[2], \
178 ((unsigned char *)&addr)[1], \
179 ((unsigned char *)&addr)[0]
180 #elif defined(__BIG_ENDIAN)
181 #define HIPQUAD NIPQUAD
183 #error "Undefined byteorder??"
184 #endif /* __LITTLE_ENDIAN */
188 /* a long can be more than 32 bits, so use BITS_PER_LONG
189 * to allow the compiler to adjust the bit shifting accordingly
192 static __inline__ int ext2_set_bit(int nr, void *addr)
194 return set_bit(nr, (long*)addr);
197 static __inline__ int ext2_clear_bit(int nr, void *addr)
199 return clear_bit(nr, (long*)addr);
202 static __inline__ int ext2_test_bit(int nr, void *addr)
204 return test_bit(nr, (long*)addr);
213 static inline void MODULE_AUTHOR(char *name)
215 printf("%s\n", name);
217 #define MODULE_DESCRIPTION(name) MODULE_AUTHOR(name)
218 #define MODULE_LICENSE(name) MODULE_AUTHOR(name)
220 #define THIS_MODULE NULL
226 static inline int misc_register(void *foo)
231 static inline int misc_deregister(void *foo)
236 static inline int request_module(char *name)
241 #define __MOD_INC_USE_COUNT(m) do {} while (0)
242 #define __MOD_DEC_USE_COUNT(m) do {} while (0)
243 #define MOD_INC_USE_COUNT do {} while (0)
244 #define MOD_DEC_USE_COUNT do {} while (0)
245 static inline void __module_get(struct module *module)
249 static inline int try_module_get(struct module *module)
254 static inline void module_put(struct module *module)
258 /* module initialization */
259 extern int init_obdclass(void);
260 extern int ptlrpc_init(void);
261 extern int ldlm_init(void);
262 extern int osc_init(void);
263 extern int lov_init(void);
264 extern int mdc_init(void);
265 extern int lmv_init(void);
266 extern int mgc_init(void);
267 extern int echo_client_init(void);
273 #define EXPORT_SYMBOL(S)
277 typedef struct { } spinlock_t;
278 typedef __u64 kdev_t;
280 #define SPIN_LOCK_UNLOCKED (spinlock_t) { }
281 #define LASSERT_SPIN_LOCKED(lock) do {} while(0)
282 #define LASSERT_SEM_LOCKED(sem) do {} while(0)
284 static inline void spin_lock(spinlock_t *l) {return;}
285 static inline void spin_unlock(spinlock_t *l) {return;}
286 static inline void spin_lock_init(spinlock_t *l) {return;}
287 static inline void local_irq_save(unsigned long flag) {return;}
288 static inline void local_irq_restore(unsigned long flag) {return;}
289 static inline int spin_is_locked(spinlock_t *l) {return 1;}
291 static inline void spin_lock_bh(spinlock_t *l) {}
292 static inline void spin_unlock_bh(spinlock_t *l) {}
293 static inline void spin_lock_irqsave(spinlock_t *a, unsigned long b) {}
294 static inline void spin_unlock_irqrestore(spinlock_t *a, unsigned long b) {}
296 typedef spinlock_t rwlock_t;
297 #define RW_LOCK_UNLOCKED SPIN_LOCK_UNLOCKED
298 #define read_lock(l) spin_lock(l)
299 #define read_unlock(l) spin_unlock(l)
300 #define write_lock(l) spin_lock(l)
301 #define write_unlock(l) spin_unlock(l)
304 #define min(x,y) ((x)<(y) ? (x) : (y))
305 #define max(x,y) ((x)>(y) ? (x) : (y))
308 #define min_t(type,x,y) \
309 ({ type __x = (x); type __y = (y); __x < __y ? __x: __y; })
312 #define max_t(type,x,y) \
313 ({ type __x = (x); type __y = (y); __x > __y ? __x: __y; })
316 #define simple_strtol strtol
318 /* registering symbols */
320 #define ERESTARTSYS ERESTART
326 void get_random_bytes(void *ptr, int size);
330 /* memory size: used for some client tunables */
331 #define num_physpages (256 * 1024) /* 1GB */
333 static inline int copy_from_user(void *a,void *b, int c)
339 static inline int copy_to_user(void *a,void *b, int c)
350 #define SLAB_HWCACHE_ALIGN 0
351 static inline kmem_cache_t *
352 kmem_cache_create(const char *name, size_t objsize, size_t cdum,
354 void (*e)(void *, kmem_cache_t *, unsigned long),
355 void (*f)(void *, kmem_cache_t *, unsigned long))
358 c = malloc(sizeof(*c));
362 CDEBUG(D_MALLOC, "alloc slab cache %s at %p, objsize %d\n",
363 name, c, (int)objsize);
367 static inline int kmem_cache_destroy(kmem_cache_t *a)
369 CDEBUG(D_MALLOC, "destroy slab cache %p, objsize %u\n", a, a->size);
374 /* struct page decl moved out from here into portals/include/libcfs/user-prim.h */
377 #define PAGE_LIST_ENTRY list
378 #define PAGE_LIST(page) ((page)->list)
380 #define kmap(page) (page)->addr
381 #define kunmap(a) do {} while (0)
383 static inline cfs_page_t *alloc_pages(int mask, unsigned long order)
385 cfs_page_t *pg = malloc(sizeof(*pg));
389 #if 0 //#ifdef MAP_ANONYMOUS
390 pg->addr = mmap(0, PAGE_SIZE << order, PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, 0, 0);
392 pg->addr = malloc(CFS_PAGE_SIZE << order);
401 #define cfs_alloc_pages(mask, order) alloc_pages((mask), (order))
403 #define alloc_page(mask) alloc_pages((mask), 0)
404 #define cfs_alloc_page(mask) alloc_page(mask)
406 static inline void __free_pages(cfs_page_t *pg, int what)
408 #if 0 //#ifdef MAP_ANONYMOUS
409 munmap(pg->addr, PAGE_SIZE);
415 #define __cfs_free_pages(pg, order) __free_pages((pg), (order))
417 #define __free_page(page) __free_pages((page), 0)
418 #define free_page(page) __free_page(page)
419 #define __cfs_free_page(page) __cfs_free_pages((page), 0)
421 static inline cfs_page_t* __grab_cache_page(unsigned long index)
423 cfs_page_t *pg = alloc_pages(0, 0);
430 #define grab_cache_page(index) __grab_cache_page(index)
431 #define page_cache_release(page) __free_pages(page, 0)
434 #define do_div(a,b) \
436 unsigned long remainder;\
437 remainder = (a) % (b); \
443 #define ATTR_MODE 0x0001
444 #define ATTR_UID 0x0002
445 #define ATTR_GID 0x0004
446 #define ATTR_SIZE 0x0008
447 #define ATTR_ATIME 0x0010
448 #define ATTR_MTIME 0x0020
449 #define ATTR_CTIME 0x0040
450 #define ATTR_ATIME_SET 0x0080
451 #define ATTR_MTIME_SET 0x0100
452 #define ATTR_FORCE 0x0200 /* Not a change, but a change it */
453 #define ATTR_ATTR_FLAG 0x0400
454 #define ATTR_RAW 0x0800 /* file system, not vfs will massage attrs */
455 #define ATTR_FROM_OPEN 0x1000 /* called from open path, ie O_TRUNC */
456 #define ATTR_CTIME_SET 0x2000
457 #define ATTR_BLOCKS 0x4000
458 #define ATTR_KILL_SUID 0
459 #define ATTR_KILL_SGID 0
462 unsigned int ia_valid;
470 unsigned int ia_attr_flags;
473 #define ll_iattr iattr
475 #define IT_OPEN 0x0001
476 #define IT_CREAT 0x0002
477 #define IT_READDIR 0x0004
478 #define IT_GETATTR 0x0008
479 #define IT_LOOKUP 0x0010
480 #define IT_UNLINK 0x0020
481 #define IT_GETXATTR 0x0040
482 #define IT_EXEC 0x0080
483 #define IT_PIN 0x0100
485 #define IT_FL_LOCKED 0x0001
486 #define IT_FL_FOLLOWED 0x0002 /* set by vfs_follow_link */
488 #define INTENT_MAGIC 0x19620323
490 struct lustre_intent_data {
493 __u64 it_lock_handle;
498 struct lookup_intent {
500 void (*it_op_release)(struct lookup_intent *);
505 struct lustre_intent_data lustre;
509 static inline void intent_init(struct lookup_intent *it, int op, int flags)
511 memset(it, 0, sizeof(*it));
512 it->it_magic = INTENT_MAGIC;
514 it->it_flags = flags;
526 struct rw_semaphore {
535 /* use the macro's argument to avoid unused warnings */
536 #define down(a) do { (void)a; } while (0)
537 #define mutex_down(a) down(a)
538 #define up(a) do { (void)a; } while (0)
539 #define mutex_up(a) up(a)
540 #define down_read(a) do { (void)a; } while (0)
541 #define up_read(a) do { (void)a; } while (0)
542 #define down_write(a) do { (void)a; } while (0)
543 #define up_write(a) do { (void)a; } while (0)
544 #define sema_init(a,b) do { (void)a; } while (0)
545 #define init_rwsem(a) do { (void)a; } while (0)
546 #define DECLARE_MUTEX(name) \
547 struct semaphore name = { 1 }
548 static inline void init_MUTEX (struct semaphore *sem)
552 static inline void init_MUTEX_LOCKED (struct semaphore *sem)
557 #define init_mutex(s) init_MUTEX(s)
560 struct list_head sleepers;
564 struct list_head sleeping;
574 struct signal pending;
587 typedef struct task_struct cfs_task_t;
588 #define cfs_current() current
589 #define cfs_curproc_pid() (current->pid)
590 #define cfs_curproc_comm() (current->comm)
592 extern struct task_struct *current;
593 int in_group_p(gid_t gid);
594 static inline int capable(int cap)
596 if (current->cap_effective & (1 << cap))
602 #define set_current_state(foo) do { current->state = foo; } while (0)
604 #define init_waitqueue_entry(q,p) do { (q)->process = p; } while (0)
605 #define add_wait_queue(q,p) do { list_add(&(q)->sleepers, &(p)->sleeping); } while (0)
606 #define del_wait_queue(p) do { list_del(&(p)->sleeping); } while (0)
607 #define remove_wait_queue(q,p) do { list_del(&(p)->sleeping); } while (0)
609 #define DECLARE_WAIT_QUEUE_HEAD(HEAD) \
610 wait_queue_head_t HEAD = { \
611 .sleepers = LIST_HEAD_INIT(HEAD.sleepers) \
613 #define init_waitqueue_head(l) INIT_LIST_HEAD(&(l)->sleepers)
614 #define wake_up(l) do { int a; a++; } while (0)
615 #define TASK_INTERRUPTIBLE 0
616 #define TASK_UNINTERRUPTIBLE 1
617 #define TASK_RUNNING 2
619 #define wait_event_interruptible(wq, condition) \
621 struct l_wait_info lwi; \
622 int timeout = 100000000;/* for ever */ \
625 lwi = LWI_TIMEOUT(timeout, NULL, NULL); \
626 ret = l_wait_event(NULL, condition, &lwi); \
631 #define in_interrupt() (0)
633 #define schedule() do {} while (0)
634 static inline int schedule_timeout(signed long t)
639 #define lock_kernel() do {} while (0)
640 #define unlock_kernel() do {} while (0)
641 #define daemonize(l) do {} while (0)
642 #define sigfillset(l) do {} while (0)
643 #define recalc_sigpending(l) do {} while (0)
644 #define kernel_thread(l,m,n) LBUG()
646 #define USERMODEHELPER(path, argv, envp) (0)
647 #define SIGNAL_MASK_ASSERT()
650 #include <sys/time.h>
652 #error "liblustre's jiffies currently expects HZ to be 1"
656 unsigned long _ret = 0; \
658 if (gettimeofday(&tv, NULL) == 0) \
662 #define get_jiffies_64() (__u64)jiffies
663 #define time_after(a, b) ((long)(b) - (long)(a) < 0)
664 #define time_before(a, b) time_after(b,a)
665 #define time_after_eq(a,b) ((long)(a) - (long)(b) >= 0)
668 struct list_head tl_list;
669 void (*function)(unsigned long unused);
674 static inline int timer_pending(struct timer_list *l)
676 if (time_after(l->expires, jiffies))
682 static inline int init_timer(struct timer_list *l)
684 INIT_LIST_HEAD(&l->tl_list);
688 static inline void mod_timer(struct timer_list *l, int thetime)
690 l->expires = thetime;
693 static inline void del_timer(struct timer_list *l)
698 typedef struct { volatile int counter; } atomic_t;
700 #define ATOMIC_INIT(i) { (i) }
702 #define atomic_read(a) ((a)->counter)
703 #define atomic_set(a,b) do {(a)->counter = b; } while (0)
704 #define atomic_dec_and_test(a) ((--((a)->counter)) == 0)
705 #define atomic_dec_and_lock(a,b) ((--((a)->counter)) == 0)
706 #define atomic_inc(a) (((a)->counter)++)
707 #define atomic_dec(a) do { (a)->counter--; } while (0)
708 #define atomic_add(b,a) do {(a)->counter += b;} while (0)
709 #define atomic_add_return(n,a) ((a)->counter = n)
710 #define atomic_inc_return(a) atomic_add_return(1,a)
711 #define atomic_sub(b,a) do {(a)->counter -= b;} while (0)
712 #define atomic_sub_return(n,a) ((a)->counter -= n)
713 #define atomic_dec_return(a) atomic_sub_return(1,a)
716 #define likely(exp) (exp)
719 #define unlikely(exp) (exp)
722 #define might_sleep()
723 #define might_sleep_if(c)
727 int test_and_set_bit(int nr, unsigned long *addr)
731 while (nr >= sizeof(long)) {
736 oldbit = (*addr) & (1 << nr);
742 int test_and_clear_bit(int nr, unsigned long *addr)
746 while (nr >= sizeof(long)) {
751 oldbit = (*addr) & (1 << nr);
756 /* FIXME sys/capability will finally included linux/fs.h thus
757 * cause numerous trouble on x86-64. as temporary solution for
758 * build broken at Cray, we copy definition we need from capability.h
762 typedef struct _cap_struct *cap_t;
763 typedef int cap_value_t;
774 #define CAP_DAC_OVERRIDE 1
775 #define CAP_DAC_READ_SEARCH 2
778 #define CAP_SYS_ADMIN 21
780 cap_t cap_get_proc(void);
781 int cap_get_flag(cap_t, cap_value_t, cap_flag_t, cap_flag_value_t *);
784 static inline int llog_init_commit_master(void) { return 0; }
785 static inline int llog_cleanup_commit_master(int force) { return 0; }
786 static inline void libcfs_run_lbug_upcall(char *file, const char *fn,
795 #define COMPLETION_INITIALIZER(work) \
796 { 0, __WAIT_QUEUE_HEAD_INITIALIZER((work).wait) }
798 #define DECLARE_COMPLETION(work) \
799 struct completion work = COMPLETION_INITIALIZER(work)
801 #define INIT_COMPLETION(x) ((x).done = 0)
803 static inline void init_completion(struct completion *x)
806 init_waitqueue_head(&x->wait);
809 struct liblustre_wait_callback {
810 struct list_head llwc_list;
811 const char *llwc_name;
812 int (*llwc_fn)(void *arg);
816 void *liblustre_register_wait_callback(const char *name,
817 int (*fn)(void *arg), void *arg);
818 void liblustre_deregister_wait_callback(void *notifier);
819 int liblustre_wait_event(int timeout);
821 void *liblustre_register_idle_callback(const char *name,
822 int (*fn)(void *arg), void *arg);
823 void liblustre_deregister_idle_callback(void *notifier);
824 void liblustre_wait_idle(void);
827 struct nfs_lock_info {
833 typedef struct file_lock {
834 struct file_lock *fl_next; /* singly linked list for this inode */
835 struct list_head fl_link; /* doubly linked list of all locks */
836 struct list_head fl_block; /* circular list of blocked processes */
840 struct file *fl_file;
841 unsigned char fl_flags;
842 unsigned char fl_type;
846 void (*fl_notify)(struct file_lock *); /* unblock callback */
847 void (*fl_insert)(struct file_lock *); /* lock insertion callback */
848 void (*fl_remove)(struct file_lock *); /* lock removal callback */
850 void *fl_fasync; /* for lease break notifications */
851 unsigned long fl_break_time; /* for nonblocking lease breaks */
854 struct nfs_lock_info nfs_fl;
858 #define cfs_flock_type(fl) ((fl)->fl_type)
859 #define cfs_flock_set_type(fl, type) do { (fl)->fl_type = (type); } while(0)
860 #define cfs_flock_pid(fl) ((fl)->fl_pid)
861 #define cfs_flock_set_pid(fl, pid) do { (fl)->fl_pid = (pid); } while(0)
862 #define cfs_flock_start(fl) ((fl)->fl_start)
863 #define cfs_flock_set_start(fl, start) do { (fl)->fl_start = (start); } while(0)
864 #define cfs_flock_end(fl) ((fl)->fl_end)
865 #define cfs_flock_set_end(fl, end) do { (fl)->fl_end = (end); } while(0)
868 #define INT_LIMIT(x) (~((x)1 << (sizeof(x)*8 - 1)))
869 #define OFFSET_MAX INT_LIMIT(loff_t)
872 /* XXX: defined in kernel */
881 struct posix_acl_entry {
883 unsigned short e_perm;
889 unsigned int a_count;
890 struct posix_acl_entry a_entries[0];
901 xattr_acl_entry a_entries[0];
904 static inline size_t xattr_acl_size(int count)
906 return sizeof(xattr_acl_header) + count * sizeof(xattr_acl_entry);
910 struct posix_acl * posix_acl_from_xattr(const void *value, size_t size)
916 int posix_acl_valid(const struct posix_acl *acl)
922 void posix_acl_release(struct posix_acl *acl)
926 #ifdef LIBLUSTRE_POSIX_ACL
927 # ifndef posix_acl_xattr_entry
928 # define posix_acl_xattr_entry xattr_acl_entry
930 # ifndef posix_acl_xattr_header
931 # define posix_acl_xattr_header xattr_acl_header
933 # ifndef posix_acl_xattr_size
934 # define posix_acl_xattr_size(entry) xattr_acl_size(entry)
936 # ifndef CONFIG_FS_POSIX_ACL
937 # define CONFIG_FS_POSIX_ACL 1
942 #define ENOTSUPP ENOTSUP
945 typedef int mm_segment_t;
950 static inline mm_segment_t get_fs(void)
955 static inline void set_fs(mm_segment_t seg)
959 #include <obd_support.h>
960 #include <lustre/lustre_idl.h>
961 #include <lustre_lib.h>
962 #include <lustre_import.h>
963 #include <lustre_export.h>
964 #include <lustre_net.h>
966 /* Fast hashing routine for a long.
967 (C) 2002 William Lee Irwin III, IBM */
970 * Knuth recommends primes in approximately golden ratio to the maximum
971 * integer representable by a machine word for multiplicative hashing.
972 * Chuck Lever verified the effectiveness of this technique:
973 * http://www.citi.umich.edu/techreports/reports/citi-tr-00-1.pdf
975 * These primes are chosen to be bit-sparse, that is operations on
976 * them can use shifts and additions instead of multiplications for
977 * machines where multiplications are slow.
979 #if BITS_PER_LONG == 32
980 /* 2^31 + 2^29 - 2^25 + 2^22 - 2^19 - 2^16 + 1 */
981 #define GOLDEN_RATIO_PRIME 0x9e370001UL
982 #elif BITS_PER_LONG == 64
983 /* 2^63 + 2^61 - 2^57 + 2^54 - 2^51 - 2^18 + 1 */
984 #define GOLDEN_RATIO_PRIME 0x9e37fffffffc0001UL
986 #error Define GOLDEN_RATIO_PRIME for your wordsize.
989 static inline unsigned long hash_long(unsigned long val, unsigned int bits)
991 unsigned long hash = val;
993 #if BITS_PER_LONG == 64
994 /* Sigh, gcc can't optimise this alone like it does for 32 bits. */
995 unsigned long n = hash;
1009 /* On some cpus multiply is faster, on others gcc will do shifts */
1010 hash *= GOLDEN_RATIO_PRIME;
1013 /* High bits are more random, so use them. */
1014 return hash >> (BITS_PER_LONG - bits);
1017 static inline unsigned long hash_ptr(void *ptr, unsigned int bits)
1019 return hash_long((unsigned long)ptr, bits);