Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
7b718769 NS |
2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
3 | * All Rights Reserved. | |
1da177e4 | 4 | * |
7b718769 NS |
5 | * This program is free software; you can redistribute it and/or |
6 | * modify it under the terms of the GNU General Public License as | |
1da177e4 LT |
7 | * published by the Free Software Foundation. |
8 | * | |
7b718769 NS |
9 | * 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. | |
1da177e4 | 13 | * |
7b718769 NS |
14 | * 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 | |
1da177e4 LT |
17 | */ |
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 | ||
1da177e4 LT |
53 | #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> | |
1da177e4 LT |
75 | #include <linux/sort.h> |
76 | ||
77 | #include <asm/page.h> | |
78 | #include <asm/div64.h> | |
79 | #include <asm/param.h> | |
80 | #include <asm/uaccess.h> | |
81 | #include <asm/byteorder.h> | |
82 | #include <asm/unaligned.h> | |
83 | ||
84 | #include <xfs_behavior.h> | |
85 | #include <xfs_vfs.h> | |
86 | #include <xfs_cred.h> | |
87 | #include <xfs_vnode.h> | |
88 | #include <xfs_stats.h> | |
89 | #include <xfs_sysctl.h> | |
90 | #include <xfs_iops.h> | |
0829c360 | 91 | #include <xfs_aops.h> |
1da177e4 LT |
92 | #include <xfs_super.h> |
93 | #include <xfs_globals.h> | |
94 | #include <xfs_fs_subr.h> | |
95 | #include <xfs_lrw.h> | |
96 | #include <xfs_buf.h> | |
97 | ||
98 | /* | |
99 | * Feature macros (disable/enable) | |
100 | */ | |
101 | #undef HAVE_REFCACHE /* reference cache not needed for NFS in 2.6 */ | |
102 | #define HAVE_SENDFILE /* sendfile(2) exists in 2.6, but not in 2.4 */ | |
103 | ||
104 | /* | |
105 | * State flag for unwritten extent buffers. | |
106 | * | |
107 | * We need to be able to distinguish between these and delayed | |
108 | * allocate buffers within XFS. The generic IO path code does | |
109 | * not need to distinguish - we use the BH_Delay flag for both | |
110 | * delalloc and these ondisk-uninitialised buffers. | |
111 | */ | |
112 | BUFFER_FNS(PrivateStart, unwritten); | |
113 | static inline void set_buffer_unwritten_io(struct buffer_head *bh) | |
114 | { | |
115 | bh->b_end_io = linvfs_unwritten_done; | |
116 | } | |
117 | ||
118 | #define restricted_chown xfs_params.restrict_chown.val | |
119 | #define irix_sgid_inherit xfs_params.sgid_inherit.val | |
120 | #define irix_symlink_mode xfs_params.symlink_mode.val | |
121 | #define xfs_panic_mask xfs_params.panic_mask.val | |
122 | #define xfs_error_level xfs_params.error_level.val | |
123 | #define xfs_syncd_centisecs xfs_params.syncd_timer.val | |
124 | #define xfs_stats_clear xfs_params.stats_clear.val | |
125 | #define xfs_inherit_sync xfs_params.inherit_sync.val | |
126 | #define xfs_inherit_nodump xfs_params.inherit_nodump.val | |
127 | #define xfs_inherit_noatime xfs_params.inherit_noatim.val | |
128 | #define xfs_buf_timer_centisecs xfs_params.xfs_buf_timer.val | |
129 | #define xfs_buf_age_centisecs xfs_params.xfs_buf_age.val | |
130 | #define xfs_inherit_nosymlinks xfs_params.inherit_nosym.val | |
131 | #define xfs_rotorstep xfs_params.rotorstep.val | |
132 | ||
39c715b7 IM |
133 | #ifndef raw_smp_processor_id |
134 | #define raw_smp_processor_id() smp_processor_id() | |
1da177e4 | 135 | #endif |
39c715b7 | 136 | #define current_cpu() raw_smp_processor_id() |
1da177e4 LT |
137 | #define current_pid() (current->pid) |
138 | #define current_fsuid(cred) (current->fsuid) | |
139 | #define current_fsgid(cred) (current->fsgid) | |
140 | ||
141 | #define NBPP PAGE_SIZE | |
142 | #define DPPSHFT (PAGE_SHIFT - 9) | |
143 | #define NDPP (1 << (PAGE_SHIFT - 9)) | |
144 | #define dtop(DD) (((DD) + NDPP - 1) >> DPPSHFT) | |
145 | #define dtopt(DD) ((DD) >> DPPSHFT) | |
146 | #define dpoff(DD) ((DD) & (NDPP-1)) | |
147 | ||
148 | #define NBBY 8 /* number of bits per byte */ | |
149 | #define NBPC PAGE_SIZE /* Number of bytes per click */ | |
150 | #define BPCSHIFT PAGE_SHIFT /* LOG2(NBPC) if exact */ | |
151 | ||
152 | /* | |
153 | * Size of block device i/o is parameterized here. | |
154 | * Currently the system supports page-sized i/o. | |
155 | */ | |
156 | #define BLKDEV_IOSHIFT BPCSHIFT | |
157 | #define BLKDEV_IOSIZE (1<<BLKDEV_IOSHIFT) | |
158 | /* number of BB's per block device block */ | |
159 | #define BLKDEV_BB BTOBB(BLKDEV_IOSIZE) | |
160 | ||
161 | /* bytes to clicks */ | |
162 | #define btoc(x) (((__psunsigned_t)(x)+(NBPC-1))>>BPCSHIFT) | |
163 | #define btoct(x) ((__psunsigned_t)(x)>>BPCSHIFT) | |
164 | #define btoc64(x) (((__uint64_t)(x)+(NBPC-1))>>BPCSHIFT) | |
165 | #define btoct64(x) ((__uint64_t)(x)>>BPCSHIFT) | |
166 | #define io_btoc(x) (((__psunsigned_t)(x)+(IO_NBPC-1))>>IO_BPCSHIFT) | |
167 | #define io_btoct(x) ((__psunsigned_t)(x)>>IO_BPCSHIFT) | |
168 | ||
169 | /* off_t bytes to clicks */ | |
170 | #define offtoc(x) (((__uint64_t)(x)+(NBPC-1))>>BPCSHIFT) | |
171 | #define offtoct(x) ((xfs_off_t)(x)>>BPCSHIFT) | |
172 | ||
173 | /* clicks to off_t bytes */ | |
174 | #define ctooff(x) ((xfs_off_t)(x)<<BPCSHIFT) | |
175 | ||
176 | /* clicks to bytes */ | |
177 | #define ctob(x) ((__psunsigned_t)(x)<<BPCSHIFT) | |
178 | #define btoct(x) ((__psunsigned_t)(x)>>BPCSHIFT) | |
179 | #define ctob64(x) ((__uint64_t)(x)<<BPCSHIFT) | |
180 | #define io_ctob(x) ((__psunsigned_t)(x)<<IO_BPCSHIFT) | |
181 | ||
182 | /* bytes to clicks */ | |
183 | #define btoc(x) (((__psunsigned_t)(x)+(NBPC-1))>>BPCSHIFT) | |
184 | ||
1da177e4 LT |
185 | #ifndef ENOATTR |
186 | #define ENOATTR ENODATA /* Attribute not found */ | |
187 | #endif | |
188 | ||
189 | /* Note: EWRONGFS never visible outside the kernel */ | |
190 | #define EWRONGFS EINVAL /* Mount with wrong filesystem type */ | |
191 | ||
192 | /* | |
193 | * XXX EFSCORRUPTED needs a real value in errno.h. asm-i386/errno.h won't | |
194 | * return codes out of its known range in errno. | |
195 | * XXX Also note: needs to be < 1000 and fairly unique on Linux (mustn't | |
196 | * conflict with any code we use already or any code a driver may use) | |
197 | * XXX Some options (currently we do #2): | |
198 | * 1/ New error code ["Filesystem is corrupted", _after_ glibc updated] | |
199 | * 2/ 990 ["Unknown error 990"] | |
200 | * 3/ EUCLEAN ["Structure needs cleaning"] | |
201 | * 4/ Convert EFSCORRUPTED to EIO [just prior to return into userspace] | |
202 | */ | |
203 | #define EFSCORRUPTED 990 /* Filesystem is corrupted */ | |
204 | ||
205 | #define SYNCHRONIZE() barrier() | |
206 | #define __return_address __builtin_return_address(0) | |
207 | ||
208 | /* | |
209 | * IRIX (BSD) quotactl makes use of separate commands for user/group, | |
210 | * whereas on Linux the syscall encodes this information into the cmd | |
211 | * field (see the QCMD macro in quota.h). These macros help keep the | |
212 | * code portable - they are not visible from the syscall interface. | |
213 | */ | |
c8ad20ff NS |
214 | #define Q_XSETGQLIM XQM_CMD(8) /* set groups disk limits */ |
215 | #define Q_XGETGQUOTA XQM_CMD(9) /* get groups disk limits */ | |
216 | #define Q_XSETPQLIM XQM_CMD(10) /* set projects disk limits */ | |
217 | #define Q_XGETPQUOTA XQM_CMD(11) /* get projects disk limits */ | |
1da177e4 | 218 | |
6b3f6b5b | 219 | #define dfltprid 0 |
1da177e4 LT |
220 | #define MAXPATHLEN 1024 |
221 | ||
222 | #define MIN(a,b) (min(a,b)) | |
223 | #define MAX(a,b) (max(a,b)) | |
224 | #define howmany(x, y) (((x)+((y)-1))/(y)) | |
225 | #define roundup(x, y) ((((x)+((y)-1))/(y))*(y)) | |
226 | ||
cde410a9 NS |
227 | /* |
228 | * Various platform dependent calls that don't fit anywhere else | |
229 | */ | |
380b5dc0 | 230 | #define xfs_sort(a,n,s,fn) sort(a,n,s,fn,NULL) |
1da177e4 | 231 | #define xfs_stack_trace() dump_stack() |
1da177e4 LT |
232 | #define xfs_itruncate_data(ip, off) \ |
233 | (-vmtruncate(LINVFS_GET_IP(XFS_ITOV(ip)), (off))) | |
cde410a9 NS |
234 | #define xfs_statvfs_fsid(statp, mp) \ |
235 | ({ u64 id = huge_encode_dev((mp)->m_dev); \ | |
236 | __kernel_fsid_t *fsid = &(statp)->f_fsid; \ | |
237 | (fsid->val[0] = (u32)id, fsid->val[1] = (u32)(id >> 32)); }) | |
1da177e4 LT |
238 | |
239 | ||
240 | /* Move the kernel do_div definition off to one side */ | |
241 | ||
242 | #if defined __i386__ | |
243 | /* For ia32 we need to pull some tricks to get past various versions | |
244 | * of the compiler which do not like us using do_div in the middle | |
245 | * of large functions. | |
246 | */ | |
247 | static inline __u32 xfs_do_div(void *a, __u32 b, int n) | |
248 | { | |
249 | __u32 mod; | |
250 | ||
251 | switch (n) { | |
252 | case 4: | |
253 | mod = *(__u32 *)a % b; | |
254 | *(__u32 *)a = *(__u32 *)a / b; | |
255 | return mod; | |
256 | case 8: | |
257 | { | |
258 | unsigned long __upper, __low, __high, __mod; | |
259 | __u64 c = *(__u64 *)a; | |
260 | __upper = __high = c >> 32; | |
261 | __low = c; | |
262 | if (__high) { | |
263 | __upper = __high % (b); | |
264 | __high = __high / (b); | |
265 | } | |
266 | asm("divl %2":"=a" (__low), "=d" (__mod):"rm" (b), "0" (__low), "1" (__upper)); | |
267 | asm("":"=A" (c):"a" (__low),"d" (__high)); | |
268 | *(__u64 *)a = c; | |
269 | return __mod; | |
270 | } | |
271 | } | |
272 | ||
273 | /* NOTREACHED */ | |
274 | return 0; | |
275 | } | |
276 | ||
277 | /* Side effect free 64 bit mod operation */ | |
278 | static inline __u32 xfs_do_mod(void *a, __u32 b, int n) | |
279 | { | |
280 | switch (n) { | |
281 | case 4: | |
282 | return *(__u32 *)a % b; | |
283 | case 8: | |
284 | { | |
285 | unsigned long __upper, __low, __high, __mod; | |
286 | __u64 c = *(__u64 *)a; | |
287 | __upper = __high = c >> 32; | |
288 | __low = c; | |
289 | if (__high) { | |
290 | __upper = __high % (b); | |
291 | __high = __high / (b); | |
292 | } | |
293 | asm("divl %2":"=a" (__low), "=d" (__mod):"rm" (b), "0" (__low), "1" (__upper)); | |
294 | asm("":"=A" (c):"a" (__low),"d" (__high)); | |
295 | return __mod; | |
296 | } | |
297 | } | |
298 | ||
299 | /* NOTREACHED */ | |
300 | return 0; | |
301 | } | |
302 | #else | |
303 | static inline __u32 xfs_do_div(void *a, __u32 b, int n) | |
304 | { | |
305 | __u32 mod; | |
306 | ||
307 | switch (n) { | |
308 | case 4: | |
309 | mod = *(__u32 *)a % b; | |
310 | *(__u32 *)a = *(__u32 *)a / b; | |
311 | return mod; | |
312 | case 8: | |
313 | mod = do_div(*(__u64 *)a, b); | |
314 | return mod; | |
315 | } | |
316 | ||
317 | /* NOTREACHED */ | |
318 | return 0; | |
319 | } | |
320 | ||
321 | /* Side effect free 64 bit mod operation */ | |
322 | static inline __u32 xfs_do_mod(void *a, __u32 b, int n) | |
323 | { | |
324 | switch (n) { | |
325 | case 4: | |
326 | return *(__u32 *)a % b; | |
327 | case 8: | |
328 | { | |
329 | __u64 c = *(__u64 *)a; | |
330 | return do_div(c, b); | |
331 | } | |
332 | } | |
333 | ||
334 | /* NOTREACHED */ | |
335 | return 0; | |
336 | } | |
337 | #endif | |
338 | ||
339 | #undef do_div | |
340 | #define do_div(a, b) xfs_do_div(&(a), (b), sizeof(a)) | |
341 | #define do_mod(a, b) xfs_do_mod(&(a), (b), sizeof(a)) | |
342 | ||
343 | static inline __uint64_t roundup_64(__uint64_t x, __uint32_t y) | |
344 | { | |
345 | x += y - 1; | |
346 | do_div(x, y); | |
347 | return(x * y); | |
348 | } | |
349 | ||
1da177e4 | 350 | #endif /* __XFS_LINUX__ */ |