blob: 52b2405395313da95fc241c6e83fd47e0f14f268 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Nathan Scott7b718762005-11-02 14:58:39 +11002 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 *
Nathan Scott7b718762005-11-02 14:58:39 +11005 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 * published by the Free Software Foundation.
8 *
Nathan Scott7b718762005-11-02 14:58:39 +11009 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 *
Nathan Scott7b718762005-11-02 14:58:39 +110014 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
Linus Torvalds1da177e2005-04-16 15:20:36 -070017 */
18#ifndef __XFS_LINUX__
19#define __XFS_LINUX__
20
21#include <linux/types.h>
22#include <linux/config.h>
23
24/*
25 * Some types are conditional depending on the target system.
26 * XFS_BIG_BLKNOS needs block layer disk addresses to be 64 bits.
27 * XFS_BIG_INUMS needs the VFS inode number to be 64 bits, as well
28 * as requiring XFS_BIG_BLKNOS to be set.
29 */
30#if defined(CONFIG_LBD) || (BITS_PER_LONG == 64)
31# define XFS_BIG_BLKNOS 1
32# if BITS_PER_LONG == 64
33# define XFS_BIG_INUMS 1
34# else
35# define XFS_BIG_INUMS 0
36# endif
37#else
38# define XFS_BIG_BLKNOS 0
39# define XFS_BIG_INUMS 0
40#endif
41
42#include <xfs_types.h>
43#include <xfs_arch.h>
44
45#include <kmem.h>
46#include <mrlock.h>
47#include <spin.h>
48#include <sv.h>
49#include <mutex.h>
50#include <sema.h>
51#include <time.h>
52
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#include <support/ktrace.h>
54#include <support/debug.h>
55#include <support/move.h>
56#include <support/uuid.h>
57
58#include <linux/mm.h>
59#include <linux/kernel.h>
60#include <linux/blkdev.h>
61#include <linux/slab.h>
62#include <linux/module.h>
63#include <linux/file.h>
64#include <linux/swap.h>
65#include <linux/errno.h>
66#include <linux/sched.h>
67#include <linux/bitops.h>
68#include <linux/major.h>
69#include <linux/pagemap.h>
70#include <linux/vfs.h>
71#include <linux/seq_file.h>
72#include <linux/init.h>
73#include <linux/list.h>
74#include <linux/proc_fs.h>
75#include <linux/version.h>
76#include <linux/sort.h>
77
78#include <asm/page.h>
79#include <asm/div64.h>
80#include <asm/param.h>
81#include <asm/uaccess.h>
82#include <asm/byteorder.h>
83#include <asm/unaligned.h>
84
85#include <xfs_behavior.h>
86#include <xfs_vfs.h>
87#include <xfs_cred.h>
88#include <xfs_vnode.h>
89#include <xfs_stats.h>
90#include <xfs_sysctl.h>
91#include <xfs_iops.h>
Christoph Hellwig0829c362005-09-02 16:58:49 +100092#include <xfs_aops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#include <xfs_super.h>
94#include <xfs_globals.h>
95#include <xfs_fs_subr.h>
96#include <xfs_lrw.h>
97#include <xfs_buf.h>
98
99/*
100 * Feature macros (disable/enable)
101 */
102#undef HAVE_REFCACHE /* reference cache not needed for NFS in 2.6 */
103#define HAVE_SENDFILE /* sendfile(2) exists in 2.6, but not in 2.4 */
104
105/*
106 * State flag for unwritten extent buffers.
107 *
108 * We need to be able to distinguish between these and delayed
109 * allocate buffers within XFS. The generic IO path code does
110 * not need to distinguish - we use the BH_Delay flag for both
111 * delalloc and these ondisk-uninitialised buffers.
112 */
113BUFFER_FNS(PrivateStart, unwritten);
114static inline void set_buffer_unwritten_io(struct buffer_head *bh)
115{
116 bh->b_end_io = linvfs_unwritten_done;
117}
118
119#define restricted_chown xfs_params.restrict_chown.val
120#define irix_sgid_inherit xfs_params.sgid_inherit.val
121#define irix_symlink_mode xfs_params.symlink_mode.val
122#define xfs_panic_mask xfs_params.panic_mask.val
123#define xfs_error_level xfs_params.error_level.val
124#define xfs_syncd_centisecs xfs_params.syncd_timer.val
125#define xfs_stats_clear xfs_params.stats_clear.val
126#define xfs_inherit_sync xfs_params.inherit_sync.val
127#define xfs_inherit_nodump xfs_params.inherit_nodump.val
128#define xfs_inherit_noatime xfs_params.inherit_noatim.val
129#define xfs_buf_timer_centisecs xfs_params.xfs_buf_timer.val
130#define xfs_buf_age_centisecs xfs_params.xfs_buf_age.val
131#define xfs_inherit_nosymlinks xfs_params.inherit_nosym.val
132#define xfs_rotorstep xfs_params.rotorstep.val
133
Ingo Molnar39c715b2005-06-21 17:14:34 -0700134#ifndef raw_smp_processor_id
135#define raw_smp_processor_id() smp_processor_id()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136#endif
Ingo Molnar39c715b2005-06-21 17:14:34 -0700137#define current_cpu() raw_smp_processor_id()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138#define current_pid() (current->pid)
139#define current_fsuid(cred) (current->fsuid)
140#define current_fsgid(cred) (current->fsgid)
141
142#define NBPP PAGE_SIZE
143#define DPPSHFT (PAGE_SHIFT - 9)
144#define NDPP (1 << (PAGE_SHIFT - 9))
145#define dtop(DD) (((DD) + NDPP - 1) >> DPPSHFT)
146#define dtopt(DD) ((DD) >> DPPSHFT)
147#define dpoff(DD) ((DD) & (NDPP-1))
148
149#define NBBY 8 /* number of bits per byte */
150#define NBPC PAGE_SIZE /* Number of bytes per click */
151#define BPCSHIFT PAGE_SHIFT /* LOG2(NBPC) if exact */
152
153/*
154 * Size of block device i/o is parameterized here.
155 * Currently the system supports page-sized i/o.
156 */
157#define BLKDEV_IOSHIFT BPCSHIFT
158#define BLKDEV_IOSIZE (1<<BLKDEV_IOSHIFT)
159/* number of BB's per block device block */
160#define BLKDEV_BB BTOBB(BLKDEV_IOSIZE)
161
162/* bytes to clicks */
163#define btoc(x) (((__psunsigned_t)(x)+(NBPC-1))>>BPCSHIFT)
164#define btoct(x) ((__psunsigned_t)(x)>>BPCSHIFT)
165#define btoc64(x) (((__uint64_t)(x)+(NBPC-1))>>BPCSHIFT)
166#define btoct64(x) ((__uint64_t)(x)>>BPCSHIFT)
167#define io_btoc(x) (((__psunsigned_t)(x)+(IO_NBPC-1))>>IO_BPCSHIFT)
168#define io_btoct(x) ((__psunsigned_t)(x)>>IO_BPCSHIFT)
169
170/* off_t bytes to clicks */
171#define offtoc(x) (((__uint64_t)(x)+(NBPC-1))>>BPCSHIFT)
172#define offtoct(x) ((xfs_off_t)(x)>>BPCSHIFT)
173
174/* clicks to off_t bytes */
175#define ctooff(x) ((xfs_off_t)(x)<<BPCSHIFT)
176
177/* clicks to bytes */
178#define ctob(x) ((__psunsigned_t)(x)<<BPCSHIFT)
179#define btoct(x) ((__psunsigned_t)(x)>>BPCSHIFT)
180#define ctob64(x) ((__uint64_t)(x)<<BPCSHIFT)
181#define io_ctob(x) ((__psunsigned_t)(x)<<IO_BPCSHIFT)
182
183/* bytes to clicks */
184#define btoc(x) (((__psunsigned_t)(x)+(NBPC-1))>>BPCSHIFT)
185
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186#ifndef ENOATTR
187#define ENOATTR ENODATA /* Attribute not found */
188#endif
189
190/* Note: EWRONGFS never visible outside the kernel */
191#define EWRONGFS EINVAL /* Mount with wrong filesystem type */
192
193/*
194 * XXX EFSCORRUPTED needs a real value in errno.h. asm-i386/errno.h won't
195 * return codes out of its known range in errno.
196 * XXX Also note: needs to be < 1000 and fairly unique on Linux (mustn't
197 * conflict with any code we use already or any code a driver may use)
198 * XXX Some options (currently we do #2):
199 * 1/ New error code ["Filesystem is corrupted", _after_ glibc updated]
200 * 2/ 990 ["Unknown error 990"]
201 * 3/ EUCLEAN ["Structure needs cleaning"]
202 * 4/ Convert EFSCORRUPTED to EIO [just prior to return into userspace]
203 */
204#define EFSCORRUPTED 990 /* Filesystem is corrupted */
205
206#define SYNCHRONIZE() barrier()
207#define __return_address __builtin_return_address(0)
208
209/*
210 * IRIX (BSD) quotactl makes use of separate commands for user/group,
211 * whereas on Linux the syscall encodes this information into the cmd
212 * field (see the QCMD macro in quota.h). These macros help keep the
213 * code portable - they are not visible from the syscall interface.
214 */
Nathan Scottc8ad20f2005-06-21 15:38:48 +1000215#define Q_XSETGQLIM XQM_CMD(8) /* set groups disk limits */
216#define Q_XGETGQUOTA XQM_CMD(9) /* get groups disk limits */
217#define Q_XSETPQLIM XQM_CMD(10) /* set projects disk limits */
218#define Q_XGETPQUOTA XQM_CMD(11) /* get projects disk limits */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
220/* IRIX uses a dynamic sizing algorithm (ndquot = 200 + numprocs*2) */
221/* we may well need to fine-tune this if it ever becomes an issue. */
222#define DQUOT_MAX_HEURISTIC 1024 /* NR_DQUOTS */
223#define ndquot DQUOT_MAX_HEURISTIC
224
225/* IRIX uses the current size of the name cache to guess a good value */
226/* - this isn't the same but is a good enough starting point for now. */
227#define DQUOT_HASH_HEURISTIC files_stat.nr_files
228
229/* IRIX inodes maintain the project ID also, zero this field on Linux */
230#define DEFAULT_PROJID 0
231#define dfltprid DEFAULT_PROJID
232
233#define MAXPATHLEN 1024
234
235#define MIN(a,b) (min(a,b))
236#define MAX(a,b) (max(a,b))
237#define howmany(x, y) (((x)+((y)-1))/(y))
238#define roundup(x, y) ((((x)+((y)-1))/(y))*(y))
239
Nathan Scottcde410a2005-09-05 11:47:01 +1000240/*
241 * Various platform dependent calls that don't fit anywhere else
242 */
Nathan Scott380b5dc2005-11-02 11:43:18 +1100243#define xfs_sort(a,n,s,fn) sort(a,n,s,fn,NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244#define xfs_stack_trace() dump_stack()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245#define xfs_itruncate_data(ip, off) \
246 (-vmtruncate(LINVFS_GET_IP(XFS_ITOV(ip)), (off)))
Nathan Scottcde410a2005-09-05 11:47:01 +1000247#define xfs_statvfs_fsid(statp, mp) \
248 ({ u64 id = huge_encode_dev((mp)->m_dev); \
249 __kernel_fsid_t *fsid = &(statp)->f_fsid; \
250 (fsid->val[0] = (u32)id, fsid->val[1] = (u32)(id >> 32)); })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251
252
253/* Move the kernel do_div definition off to one side */
254
255#if defined __i386__
256/* For ia32 we need to pull some tricks to get past various versions
257 * of the compiler which do not like us using do_div in the middle
258 * of large functions.
259 */
260static inline __u32 xfs_do_div(void *a, __u32 b, int n)
261{
262 __u32 mod;
263
264 switch (n) {
265 case 4:
266 mod = *(__u32 *)a % b;
267 *(__u32 *)a = *(__u32 *)a / b;
268 return mod;
269 case 8:
270 {
271 unsigned long __upper, __low, __high, __mod;
272 __u64 c = *(__u64 *)a;
273 __upper = __high = c >> 32;
274 __low = c;
275 if (__high) {
276 __upper = __high % (b);
277 __high = __high / (b);
278 }
279 asm("divl %2":"=a" (__low), "=d" (__mod):"rm" (b), "0" (__low), "1" (__upper));
280 asm("":"=A" (c):"a" (__low),"d" (__high));
281 *(__u64 *)a = c;
282 return __mod;
283 }
284 }
285
286 /* NOTREACHED */
287 return 0;
288}
289
290/* Side effect free 64 bit mod operation */
291static inline __u32 xfs_do_mod(void *a, __u32 b, int n)
292{
293 switch (n) {
294 case 4:
295 return *(__u32 *)a % b;
296 case 8:
297 {
298 unsigned long __upper, __low, __high, __mod;
299 __u64 c = *(__u64 *)a;
300 __upper = __high = c >> 32;
301 __low = c;
302 if (__high) {
303 __upper = __high % (b);
304 __high = __high / (b);
305 }
306 asm("divl %2":"=a" (__low), "=d" (__mod):"rm" (b), "0" (__low), "1" (__upper));
307 asm("":"=A" (c):"a" (__low),"d" (__high));
308 return __mod;
309 }
310 }
311
312 /* NOTREACHED */
313 return 0;
314}
315#else
316static inline __u32 xfs_do_div(void *a, __u32 b, int n)
317{
318 __u32 mod;
319
320 switch (n) {
321 case 4:
322 mod = *(__u32 *)a % b;
323 *(__u32 *)a = *(__u32 *)a / b;
324 return mod;
325 case 8:
326 mod = do_div(*(__u64 *)a, b);
327 return mod;
328 }
329
330 /* NOTREACHED */
331 return 0;
332}
333
334/* Side effect free 64 bit mod operation */
335static inline __u32 xfs_do_mod(void *a, __u32 b, int n)
336{
337 switch (n) {
338 case 4:
339 return *(__u32 *)a % b;
340 case 8:
341 {
342 __u64 c = *(__u64 *)a;
343 return do_div(c, b);
344 }
345 }
346
347 /* NOTREACHED */
348 return 0;
349}
350#endif
351
352#undef do_div
353#define do_div(a, b) xfs_do_div(&(a), (b), sizeof(a))
354#define do_mod(a, b) xfs_do_mod(&(a), (b), sizeof(a))
355
356static inline __uint64_t roundup_64(__uint64_t x, __uint32_t y)
357{
358 x += y - 1;
359 do_div(x, y);
360 return(x * y);
361}
362
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363#endif /* __XFS_LINUX__ */