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
37 #include <libcfs/list.h>
38 #include <lnet/lnet.h>
39 #include <libcfs/kp30.h>
40 #include <libcfs/user-bitops.h>
46 #ifdef HAVE_ASM_PAGE_H
47 # include <asm/page.h>
49 #ifdef HAVE_SYS_USER_H
50 # include <sys/user.h>
52 #ifdef HAVE_SYS_IOCTL_H
53 # include <sys/ioctl.h>
57 #include <sys/ioctl.h>
69 #include <libcfs/list.h>
70 #include <lnet/lnet.h>
71 #include <libcfs/kp30.h>
72 #include <libcfs/user-bitops.h>
78 /* definitions for liblustre */
82 #define loff_t long long
84 typedef unsigned short umode_t;
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);
170 /* a long can be more than 32 bits, so use BITS_PER_LONG
171 * to allow the compiler to adjust the bit shifting accordingly
174 static __inline__ int ext2_set_bit(int nr, void *addr)
176 return set_bit(nr, addr);
179 static __inline__ int ext2_clear_bit(int nr, void *addr)
181 return clear_bit(nr, addr);
184 static __inline__ int ext2_test_bit(int nr, void *addr)
186 return test_bit(nr, addr);
195 static inline void MODULE_AUTHOR(char *name)
197 printf("%s\n", name);
199 #define MODULE_DESCRIPTION(name) MODULE_AUTHOR(name)
200 #define MODULE_LICENSE(name) MODULE_AUTHOR(name)
202 #define THIS_MODULE NULL
208 static inline int misc_register(void *foo)
213 static inline int misc_deregister(void *foo)
218 static inline int request_module(char *name)
223 #define __MOD_INC_USE_COUNT(m) do {} while (0)
224 #define __MOD_DEC_USE_COUNT(m) do {} while (0)
225 #define MOD_INC_USE_COUNT do {} while (0)
226 #define MOD_DEC_USE_COUNT do {} while (0)
227 static inline void __module_get(struct module *module)
231 static inline int try_module_get(struct module *module)
236 static inline void module_put(struct module *module)
240 /* module initialization */
241 extern int init_obdclass(void);
242 extern int ptlrpc_init(void);
243 extern int ldlm_init(void);
244 extern int osc_init(void);
245 extern int lov_init(void);
246 extern int mdc_init(void);
247 extern int lmv_init(void);
248 extern int mgc_init(void);
249 extern int echo_client_init(void);
255 #define EXPORT_SYMBOL(S)
259 typedef struct { } spinlock_t;
260 typedef __u64 kdev_t;
262 #define SPIN_LOCK_UNLOCKED (spinlock_t) { }
263 #define LASSERT_SPIN_LOCKED(lock) do {} while(0)
264 #define LASSERT_SEM_LOCKED(sem) do {} while(0)
266 static inline void spin_lock(spinlock_t *l) {return;}
267 static inline void spin_unlock(spinlock_t *l) {return;}
268 static inline void spin_lock_init(spinlock_t *l) {return;}
269 static inline void local_irq_save(unsigned long flag) {return;}
270 static inline void local_irq_restore(unsigned long flag) {return;}
271 static inline int spin_is_locked(spinlock_t *l) {return 1;}
273 static inline void spin_lock_bh(spinlock_t *l) {}
274 static inline void spin_unlock_bh(spinlock_t *l) {}
275 static inline void spin_lock_irqsave(spinlock_t *a, unsigned long b) {}
276 static inline void spin_unlock_irqrestore(spinlock_t *a, unsigned long b) {}
278 typedef spinlock_t rwlock_t;
279 #define RW_LOCK_UNLOCKED SPIN_LOCK_UNLOCKED
280 #define read_lock(l) spin_lock(l)
281 #define read_unlock(l) spin_unlock(l)
282 #define write_lock(l) spin_lock(l)
283 #define write_unlock(l) spin_unlock(l)
284 #define rwlock_init(l) spin_lock_init(l)
286 #define min(x,y) ((x)<(y) ? (x) : (y))
287 #define max(x,y) ((x)>(y) ? (x) : (y))
290 #define min_t(type,x,y) \
291 ({ type __x = (x); type __y = (y); __x < __y ? __x: __y; })
294 #define max_t(type,x,y) \
295 ({ type __x = (x); type __y = (y); __x > __y ? __x: __y; })
298 #define simple_strtol strtol
300 /* registering symbols */
302 #define ERESTARTSYS ERESTART
308 void get_random_bytes(void *ptr, int size);
312 /* memory size: used for some client tunables */
313 #define num_physpages (256 * 1024) /* 1GB */
315 static inline int copy_from_user(void *a,void *b, int c)
321 static inline int copy_to_user(void *a,void *b, int c)
332 #define SLAB_HWCACHE_ALIGN 0
333 static inline kmem_cache_t *
334 kmem_cache_create(const char *name, size_t objsize, size_t cdum,
336 void (*e)(void *, kmem_cache_t *, unsigned long),
337 void (*f)(void *, kmem_cache_t *, unsigned long))
340 c = malloc(sizeof(*c));
344 CDEBUG(D_MALLOC, "alloc slab cache %s at %p, objsize %d\n",
345 name, c, (int)objsize);
349 static inline int kmem_cache_destroy(kmem_cache_t *a)
351 CDEBUG(D_MALLOC, "destroy slab cache %p, objsize %u\n", a, a->size);
356 /* struct page decl moved out from here into portals/include/libcfs/user-prim.h */
359 #define PAGE_LIST_ENTRY list
360 #define PAGE_LIST(page) ((page)->list)
362 #define kmap(page) (page)->addr
363 #define kunmap(a) do {} while (0)
365 static inline cfs_page_t *alloc_pages(int mask, unsigned long order)
367 cfs_page_t *pg = malloc(sizeof(*pg));
371 #if 0 //#ifdef MAP_ANONYMOUS
372 pg->addr = mmap(0, PAGE_SIZE << order, PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, 0, 0);
374 pg->addr = malloc(CFS_PAGE_SIZE << order);
383 #define cfs_alloc_pages(mask, order) alloc_pages((mask), (order))
385 #define alloc_page(mask) alloc_pages((mask), 0)
386 #define cfs_alloc_page(mask) alloc_page(mask)
388 static inline void __free_pages(cfs_page_t *pg, int what)
390 #if 0 //#ifdef MAP_ANONYMOUS
391 munmap(pg->addr, PAGE_SIZE);
397 #define __cfs_free_pages(pg, order) __free_pages((pg), (order))
399 #define __free_page(page) __free_pages((page), 0)
400 #define free_page(page) __free_page(page)
401 #define __cfs_free_page(page) __cfs_free_pages((page), 0)
403 static inline cfs_page_t* __grab_cache_page(unsigned long index)
405 cfs_page_t *pg = alloc_pages(0, 0);
412 #define grab_cache_page(index) __grab_cache_page(index)
413 #define page_cache_release(page) __free_pages(page, 0)
416 #define do_div(a,b) \
418 unsigned long remainder;\
419 remainder = (a) % (b); \
425 #define ATTR_MODE 0x0001
426 #define ATTR_UID 0x0002
427 #define ATTR_GID 0x0004
428 #define ATTR_SIZE 0x0008
429 #define ATTR_ATIME 0x0010
430 #define ATTR_MTIME 0x0020
431 #define ATTR_CTIME 0x0040
432 #define ATTR_ATIME_SET 0x0080
433 #define ATTR_MTIME_SET 0x0100
434 #define ATTR_FORCE 0x0200 /* Not a change, but a change it */
435 #define ATTR_ATTR_FLAG 0x0400
436 #define ATTR_RAW 0x0800 /* file system, not vfs will massage attrs */
437 #define ATTR_FROM_OPEN 0x1000 /* called from open path, ie O_TRUNC */
438 #define ATTR_CTIME_SET 0x2000
439 #define ATTR_BLOCKS 0x4000
440 #define ATTR_KILL_SUID 0
441 #define ATTR_KILL_SGID 0
444 unsigned int ia_valid;
452 unsigned int ia_attr_flags;
455 #define ll_iattr iattr
457 #define IT_OPEN 0x0001
458 #define IT_CREAT 0x0002
459 #define IT_READDIR 0x0004
460 #define IT_GETATTR 0x0008
461 #define IT_LOOKUP 0x0010
462 #define IT_UNLINK 0x0020
463 #define IT_GETXATTR 0x0040
464 #define IT_EXEC 0x0080
465 #define IT_PIN 0x0100
467 #define IT_FL_LOCKED 0x0001
468 #define IT_FL_FOLLOWED 0x0002 /* set by vfs_follow_link */
470 #define INTENT_MAGIC 0x19620323
472 struct lustre_intent_data {
475 __u64 it_lock_handle;
480 struct lookup_intent {
482 void (*it_op_release)(struct lookup_intent *);
487 struct lustre_intent_data lustre;
491 static inline void intent_init(struct lookup_intent *it, int op, int flags)
493 memset(it, 0, sizeof(*it));
494 it->it_magic = INTENT_MAGIC;
496 it->it_flags = flags;
508 struct rw_semaphore {
517 /* use the macro's argument to avoid unused warnings */
518 #define down(a) do { (void)a; } while (0)
519 #define mutex_down(a) down(a)
520 #define up(a) do { (void)a; } while (0)
521 #define mutex_up(a) up(a)
522 #define down_read(a) do { (void)a; } while (0)
523 #define up_read(a) do { (void)a; } while (0)
524 #define down_write(a) do { (void)a; } while (0)
525 #define up_write(a) do { (void)a; } while (0)
526 #define sema_init(a,b) do { (void)a; } while (0)
527 #define init_rwsem(a) do { (void)a; } while (0)
528 #define DECLARE_MUTEX(name) \
529 struct semaphore name = { 1 }
530 static inline void init_MUTEX (struct semaphore *sem)
534 static inline void init_MUTEX_LOCKED (struct semaphore *sem)
539 #define init_mutex(s) init_MUTEX(s)
542 struct list_head sleepers;
546 struct list_head sleeping;
556 struct signal pending;
569 typedef struct task_struct cfs_task_t;
570 #define cfs_current() current
571 #define cfs_curproc_pid() (current->pid)
572 #define cfs_curproc_comm() (current->comm)
574 extern struct task_struct *current;
575 int in_group_p(gid_t gid);
576 static inline int capable(int cap)
578 if (current->cap_effective & (1 << cap))
584 #define set_current_state(foo) do { current->state = foo; } while (0)
586 #define init_waitqueue_entry(q,p) do { (q)->process = p; } while (0)
587 #define add_wait_queue(q,p) do { list_add(&(q)->sleepers, &(p)->sleeping); } while (0)
588 #define del_wait_queue(p) do { list_del(&(p)->sleeping); } while (0)
589 #define remove_wait_queue(q,p) do { list_del(&(p)->sleeping); } while (0)
591 #define DECLARE_WAIT_QUEUE_HEAD(HEAD) \
592 wait_queue_head_t HEAD = { \
593 .sleepers = CFS_LIST_HEAD_INIT(HEAD.sleepers) \
595 #define init_waitqueue_head(l) CFS_INIT_LIST_HEAD(&(l)->sleepers)
596 #define wake_up(l) do { int a; a++; } while (0)
597 #define TASK_INTERRUPTIBLE 0
598 #define TASK_UNINTERRUPTIBLE 1
599 #define TASK_RUNNING 2
601 #define wait_event_interruptible(wq, condition) \
603 struct l_wait_info lwi; \
604 int timeout = 100000000;/* for ever */ \
607 lwi = LWI_TIMEOUT(timeout, NULL, NULL); \
608 ret = l_wait_event(NULL, condition, &lwi); \
613 #define in_interrupt() (0)
615 #define schedule() do {} while (0)
616 static inline int schedule_timeout(signed long t)
621 #define lock_kernel() do {} while (0)
622 #define unlock_kernel() do {} while (0)
623 #define daemonize(l) do {} while (0)
624 #define sigfillset(l) do {} while (0)
625 #define recalc_sigpending(l) do {} while (0)
626 #define kernel_thread(l,m,n) LBUG()
628 #define USERMODEHELPER(path, argv, envp) (0)
629 #define SIGNAL_MASK_ASSERT()
632 #include <sys/time.h>
634 #error "liblustre's jiffies currently expects HZ to be 1"
638 unsigned long _ret = 0; \
640 if (gettimeofday(&tv, NULL) == 0) \
644 #define get_jiffies_64() (__u64)jiffies
645 #define time_after(a, b) ((long)(b) - (long)(a) < 0)
646 #define time_before(a, b) time_after(b,a)
647 #define time_after_eq(a,b) ((long)(a) - (long)(b) >= 0)
650 struct list_head tl_list;
651 void (*function)(unsigned long unused);
656 static inline int timer_pending(struct timer_list *l)
658 if (time_after(l->expires, jiffies))
664 static inline int init_timer(struct timer_list *l)
666 CFS_INIT_LIST_HEAD(&l->tl_list);
670 static inline void mod_timer(struct timer_list *l, int thetime)
672 l->expires = thetime;
675 static inline void del_timer(struct timer_list *l)
680 typedef struct { volatile int counter; } atomic_t;
682 #define ATOMIC_INIT(i) { (i) }
684 #define atomic_read(a) ((a)->counter)
685 #define atomic_set(a,b) do {(a)->counter = b; } while (0)
686 #define atomic_dec_and_test(a) ((--((a)->counter)) == 0)
687 #define atomic_dec_and_lock(a,b) ((--((a)->counter)) == 0)
688 #define atomic_inc(a) (((a)->counter)++)
689 #define atomic_dec(a) do { (a)->counter--; } while (0)
690 #define atomic_add(b,a) do {(a)->counter += b;} while (0)
691 #define atomic_add_return(n,a) ((a)->counter += n)
692 #define atomic_inc_return(a) atomic_add_return(1,a)
693 #define atomic_sub(b,a) do {(a)->counter -= b;} while (0)
694 #define atomic_sub_return(n,a) ((a)->counter -= n)
695 #define atomic_dec_return(a) atomic_sub_return(1,a)
698 #define likely(exp) (exp)
701 #define unlikely(exp) (exp)
704 #define might_sleep()
705 #define might_sleep_if(c)
709 int test_and_set_bit(int nr, unsigned long *addr)
713 while (nr >= sizeof(long)) {
718 oldbit = (*addr) & (1 << nr);
724 int test_and_clear_bit(int nr, unsigned long *addr)
728 while (nr >= sizeof(long)) {
733 oldbit = (*addr) & (1 << nr);
738 /* FIXME sys/capability will finally included linux/fs.h thus
739 * cause numerous trouble on x86-64. as temporary solution for
740 * build broken at Cray, we copy definition we need from capability.h
744 typedef struct _cap_struct *cap_t;
745 typedef int cap_value_t;
756 #define CAP_DAC_OVERRIDE 1
757 #define CAP_DAC_READ_SEARCH 2
760 #define CAP_SYS_ADMIN 21
762 cap_t cap_get_proc(void);
763 int cap_get_flag(cap_t, cap_value_t, cap_flag_t, cap_flag_value_t *);
765 static inline void libcfs_run_lbug_upcall(char *file, const char *fn,
774 #define COMPLETION_INITIALIZER(work) \
775 { 0, __WAIT_QUEUE_HEAD_INITIALIZER((work).wait) }
777 #define DECLARE_COMPLETION(work) \
778 struct completion work = COMPLETION_INITIALIZER(work)
780 #define INIT_COMPLETION(x) ((x).done = 0)
782 static inline void init_completion(struct completion *x)
785 init_waitqueue_head(&x->wait);
788 struct liblustre_wait_callback {
789 struct list_head llwc_list;
790 const char *llwc_name;
791 int (*llwc_fn)(void *arg);
795 void *liblustre_register_wait_callback(const char *name,
796 int (*fn)(void *arg), void *arg);
797 void liblustre_deregister_wait_callback(void *notifier);
798 int liblustre_wait_event(int timeout);
800 void *liblustre_register_idle_callback(const char *name,
801 int (*fn)(void *arg), void *arg);
802 void liblustre_deregister_idle_callback(void *notifier);
803 void liblustre_wait_idle(void);
806 struct nfs_lock_info {
812 typedef struct file_lock {
813 struct file_lock *fl_next; /* singly linked list for this inode */
814 struct list_head fl_link; /* doubly linked list of all locks */
815 struct list_head fl_block; /* circular list of blocked processes */
819 struct file *fl_file;
820 unsigned char fl_flags;
821 unsigned char fl_type;
825 void (*fl_notify)(struct file_lock *); /* unblock callback */
826 void (*fl_insert)(struct file_lock *); /* lock insertion callback */
827 void (*fl_remove)(struct file_lock *); /* lock removal callback */
829 void *fl_fasync; /* for lease break notifications */
830 unsigned long fl_break_time; /* for nonblocking lease breaks */
833 struct nfs_lock_info nfs_fl;
837 #define cfs_flock_type(fl) ((fl)->fl_type)
838 #define cfs_flock_set_type(fl, type) do { (fl)->fl_type = (type); } while(0)
839 #define cfs_flock_pid(fl) ((fl)->fl_pid)
840 #define cfs_flock_set_pid(fl, pid) do { (fl)->fl_pid = (pid); } while(0)
841 #define cfs_flock_start(fl) ((fl)->fl_start)
842 #define cfs_flock_set_start(fl, start) do { (fl)->fl_start = (start); } while(0)
843 #define cfs_flock_end(fl) ((fl)->fl_end)
844 #define cfs_flock_set_end(fl, end) do { (fl)->fl_end = (end); } while(0)
847 #define INT_LIMIT(x) (~((x)1 << (sizeof(x)*8 - 1)))
848 #define OFFSET_MAX INT_LIMIT(loff_t)
851 /* XXX: defined in kernel */
860 struct posix_acl_entry {
862 unsigned short e_perm;
868 unsigned int a_count;
869 struct posix_acl_entry a_entries[0];
880 xattr_acl_entry a_entries[0];
883 static inline size_t xattr_acl_size(int count)
885 return sizeof(xattr_acl_header) + count * sizeof(xattr_acl_entry);
889 struct posix_acl * posix_acl_from_xattr(const void *value, size_t size)
895 int posix_acl_valid(const struct posix_acl *acl)
901 void posix_acl_release(struct posix_acl *acl)
905 #ifdef LIBLUSTRE_POSIX_ACL
906 # ifndef posix_acl_xattr_entry
907 # define posix_acl_xattr_entry xattr_acl_entry
909 # ifndef posix_acl_xattr_header
910 # define posix_acl_xattr_header xattr_acl_header
912 # ifndef posix_acl_xattr_size
913 # define posix_acl_xattr_size(entry) xattr_acl_size(entry)
915 # ifndef CONFIG_FS_POSIX_ACL
916 # define CONFIG_FS_POSIX_ACL 1
921 #define ENOTSUPP ENOTSUP
924 typedef int mm_segment_t;
929 static inline mm_segment_t get_fs(void)
934 static inline void set_fs(mm_segment_t seg)
938 #include <obd_support.h>
939 #include <lustre/lustre_idl.h>
940 #include <lustre_lib.h>
941 #include <lustre_import.h>
942 #include <lustre_export.h>
943 #include <lustre_net.h>
945 /* Fast hashing routine for a long.
946 (C) 2002 William Lee Irwin III, IBM */
949 * Knuth recommends primes in approximately golden ratio to the maximum
950 * integer representable by a machine word for multiplicative hashing.
951 * Chuck Lever verified the effectiveness of this technique:
952 * http://www.citi.umich.edu/techreports/reports/citi-tr-00-1.pdf
954 * These primes are chosen to be bit-sparse, that is operations on
955 * them can use shifts and additions instead of multiplications for
956 * machines where multiplications are slow.
958 #if BITS_PER_LONG == 32
959 /* 2^31 + 2^29 - 2^25 + 2^22 - 2^19 - 2^16 + 1 */
960 #define GOLDEN_RATIO_PRIME 0x9e370001UL
961 #elif BITS_PER_LONG == 64
962 /* 2^63 + 2^61 - 2^57 + 2^54 - 2^51 - 2^18 + 1 */
963 #define GOLDEN_RATIO_PRIME 0x9e37fffffffc0001UL
965 #error Define GOLDEN_RATIO_PRIME for your wordsize.
968 static inline unsigned long hash_long(unsigned long val, unsigned int bits)
970 unsigned long hash = val;
972 #if BITS_PER_LONG == 64
973 /* Sigh, gcc can't optimise this alone like it does for 32 bits. */
974 unsigned long n = hash;
988 /* On some cpus multiply is faster, on others gcc will do shifts */
989 hash *= GOLDEN_RATIO_PRIME;
992 /* High bits are more random, so use them. */
993 return hash >> (BITS_PER_LONG - bits);
996 static inline unsigned long hash_ptr(void *ptr, unsigned int bits)
998 return hash_long((unsigned long)ptr, bits);