2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
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
7 * published by the Free Software Foundation.
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.
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
21 #include <linux/list.h>
22 #include <linux/types.h>
23 #include <linux/spinlock.h>
24 #include <asm/system.h>
27 #include <linux/buffer_head.h>
28 #include <linux/uio.h>
34 #define XFS_BUF_DADDR_NULL ((xfs_daddr_t) (-1LL))
36 #define xfs_buf_ctob(pp) ((pp) * PAGE_CACHE_SIZE)
37 #define xfs_buf_btoc(dd) (((dd) + PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT)
38 #define xfs_buf_btoct(dd) ((dd) >> PAGE_CACHE_SHIFT)
39 #define xfs_buf_poff(aa) ((aa) & ~PAGE_CACHE_MASK)
42 XBRW_READ
= 1, /* transfer into target memory */
43 XBRW_WRITE
= 2, /* transfer from target memory */
44 XBRW_ZERO
= 3, /* Zero target memory */
47 #define XBF_READ (1 << 0) /* buffer intended for reading from device */
48 #define XBF_WRITE (1 << 1) /* buffer intended for writing to device */
49 #define XBF_MAPPED (1 << 2) /* buffer mapped (b_addr valid) */
50 #define XBF_ASYNC (1 << 4) /* initiator will not wait for completion */
51 #define XBF_DONE (1 << 5) /* all pages in the buffer uptodate */
52 #define XBF_DELWRI (1 << 6) /* buffer has dirty pages */
53 #define XBF_STALE (1 << 7) /* buffer has been staled, do not find it */
54 #define XBF_ORDERED (1 << 11)/* use ordered writes */
55 #define XBF_READ_AHEAD (1 << 12)/* asynchronous read-ahead */
56 #define XBF_LOG_BUFFER (1 << 13)/* this is a buffer used for the log */
58 /* flags used only as arguments to access routines */
59 #define XBF_LOCK (1 << 14)/* lock requested */
60 #define XBF_TRYLOCK (1 << 15)/* lock requested, but do not wait */
61 #define XBF_DONT_BLOCK (1 << 16)/* do not block in current thread */
63 /* flags used only internally */
64 #define _XBF_PAGE_CACHE (1 << 17)/* backed by pagecache */
65 #define _XBF_PAGES (1 << 18)/* backed by refcounted pages */
66 #define _XBF_RUN_QUEUES (1 << 19)/* run block device task queue */
67 #define _XBF_DELWRI_Q (1 << 21)/* buffer on delwri queue */
70 * Special flag for supporting metadata blocks smaller than a FSB.
72 * In this case we can have multiple xfs_buf_t on a single page and
73 * need to lock out concurrent xfs_buf_t readers as they only
74 * serialise access to the buffer.
76 * If the FSB size >= PAGE_CACHE_SIZE case, we have no serialisation
77 * between reads of the page. Hence we can have one thread read the
78 * page and modify it, but then race with another thread that thinks
79 * the page is not up-to-date and hence reads it again.
81 * The result is that the first modifcation to the page is lost.
82 * This sort of AGF/AGI reading race can happen when unlinking inodes
83 * that require truncation and results in the AGI unlinked list
84 * modifications being lost.
86 #define _XBF_PAGE_LOCKED (1 << 22)
88 typedef unsigned int xfs_buf_flags_t
;
90 #define XFS_BUF_FLAGS \
91 { XBF_READ, "READ" }, \
92 { XBF_WRITE, "WRITE" }, \
93 { XBF_MAPPED, "MAPPED" }, \
94 { XBF_ASYNC, "ASYNC" }, \
95 { XBF_DONE, "DONE" }, \
96 { XBF_DELWRI, "DELWRI" }, \
97 { XBF_STALE, "STALE" }, \
98 { XBF_ORDERED, "ORDERED" }, \
99 { XBF_READ_AHEAD, "READ_AHEAD" }, \
100 { XBF_LOCK, "LOCK" }, /* should never be set */\
101 { XBF_TRYLOCK, "TRYLOCK" }, /* ditto */\
102 { XBF_DONT_BLOCK, "DONT_BLOCK" }, /* ditto */\
103 { _XBF_PAGE_CACHE, "PAGE_CACHE" }, \
104 { _XBF_PAGES, "PAGES" }, \
105 { _XBF_RUN_QUEUES, "RUN_QUEUES" }, \
106 { _XBF_DELWRI_Q, "DELWRI_Q" }, \
107 { _XBF_PAGE_LOCKED, "PAGE_LOCKED" }
113 } xfs_buftarg_flags_t
;
115 typedef struct xfs_bufhash
{
116 struct list_head bh_list
;
120 typedef struct xfs_buftarg
{
122 struct block_device
*bt_bdev
;
123 struct address_space
*bt_mapping
;
124 struct xfs_mount
*bt_mount
;
125 unsigned int bt_bsize
;
126 unsigned int bt_sshift
;
129 /* per device delwri queue */
130 struct task_struct
*bt_task
;
131 struct list_head bt_list
;
132 struct list_head bt_delwrite_queue
;
133 spinlock_t bt_delwrite_lock
;
134 unsigned long bt_flags
;
138 * xfs_buf_t: Buffer structure for pagecache-based buffers
140 * This buffer structure is used by the pagecache buffer management routines
141 * to refer to an assembly of pages forming a logical buffer.
143 * The buffer structure is used on a temporary basis only, and discarded when
144 * released. The real data storage is recorded in the pagecache. Buffers are
145 * hashed to the block device on which the file system resides.
149 typedef void (*xfs_buf_iodone_t
)(struct xfs_buf
*);
150 typedef void (*xfs_buf_relse_t
)(struct xfs_buf
*);
151 typedef int (*xfs_buf_bdstrat_t
)(struct xfs_buf
*);
155 typedef struct xfs_buf
{
157 * first cacheline holds all the fields needed for an uncontended cache
158 * hit to be fully processed. The semaphore straddles the cacheline
159 * boundary, but the counter and lock sits on the first cacheline,
160 * which is the only bit that is touched if we hit the semaphore
161 * fast-path on locking.
163 struct rb_node b_rbnode
; /* rbtree node */
164 xfs_off_t b_file_offset
; /* offset in file */
165 size_t b_buffer_length
;/* size of buffer in bytes */
166 atomic_t b_hold
; /* reference count */
167 xfs_buf_flags_t b_flags
; /* status flags */
168 struct semaphore b_sema
; /* semaphore for lockables */
170 wait_queue_head_t b_waiters
; /* unpin waiters */
171 struct list_head b_list
;
172 struct xfs_perag
*b_pag
; /* contains rbtree root */
173 xfs_buftarg_t
*b_target
; /* buffer target (device) */
174 xfs_daddr_t b_bn
; /* block number for I/O */
175 size_t b_count_desired
;/* desired transfer size */
176 void *b_addr
; /* virtual address of buffer */
177 struct work_struct b_iodone_work
;
178 xfs_buf_iodone_t b_iodone
; /* I/O completion function */
179 xfs_buf_relse_t b_relse
; /* releasing function */
180 struct completion b_iowait
; /* queue for I/O waiters */
183 struct page
**b_pages
; /* array of page pointers */
184 struct page
*b_page_array
[XB_PAGES
]; /* inline pages */
185 unsigned long b_queuetime
; /* time buffer was queued */
186 atomic_t b_pin_count
; /* pin count */
187 atomic_t b_io_remaining
; /* #outstanding I/O requests */
188 unsigned int b_page_count
; /* size of page array */
189 unsigned int b_offset
; /* page offset in first page */
190 unsigned short b_error
; /* error code on I/O */
191 #ifdef XFS_BUF_LOCK_TRACKING
197 /* Finding and Reading Buffers */
198 extern xfs_buf_t
*_xfs_buf_find(xfs_buftarg_t
*, xfs_off_t
, size_t,
199 xfs_buf_flags_t
, xfs_buf_t
*);
200 #define xfs_incore(buftarg,blkno,len,lockit) \
201 _xfs_buf_find(buftarg, blkno ,len, lockit, NULL)
203 extern xfs_buf_t
*xfs_buf_get(xfs_buftarg_t
*, xfs_off_t
, size_t,
205 extern xfs_buf_t
*xfs_buf_read(xfs_buftarg_t
*, xfs_off_t
, size_t,
208 extern xfs_buf_t
*xfs_buf_get_empty(size_t, xfs_buftarg_t
*);
209 extern xfs_buf_t
*xfs_buf_get_uncached(struct xfs_buftarg
*, size_t, int);
210 extern int xfs_buf_associate_memory(xfs_buf_t
*, void *, size_t);
211 extern void xfs_buf_hold(xfs_buf_t
*);
212 extern void xfs_buf_readahead(xfs_buftarg_t
*, xfs_off_t
, size_t);
213 struct xfs_buf
*xfs_buf_read_uncached(struct xfs_mount
*mp
,
214 struct xfs_buftarg
*target
,
215 xfs_daddr_t daddr
, size_t length
, int flags
);
217 /* Releasing Buffers */
218 extern void xfs_buf_free(xfs_buf_t
*);
219 extern void xfs_buf_rele(xfs_buf_t
*);
221 /* Locking and Unlocking Buffers */
222 extern int xfs_buf_cond_lock(xfs_buf_t
*);
223 extern int xfs_buf_lock_value(xfs_buf_t
*);
224 extern void xfs_buf_lock(xfs_buf_t
*);
225 extern void xfs_buf_unlock(xfs_buf_t
*);
227 /* Buffer Read and Write Routines */
228 extern int xfs_bwrite(struct xfs_mount
*mp
, struct xfs_buf
*bp
);
229 extern void xfs_bdwrite(void *mp
, xfs_buf_t
*bp
);
231 extern void xfsbdstrat(struct xfs_mount
*, struct xfs_buf
*);
232 extern int xfs_bdstrat_cb(struct xfs_buf
*);
234 extern void xfs_buf_ioend(xfs_buf_t
*, int);
235 extern void xfs_buf_ioerror(xfs_buf_t
*, int);
236 extern int xfs_buf_iorequest(xfs_buf_t
*);
237 extern int xfs_buf_iowait(xfs_buf_t
*);
238 extern void xfs_buf_iomove(xfs_buf_t
*, size_t, size_t, void *,
240 #define xfs_buf_zero(bp, off, len) \
241 xfs_buf_iomove((bp), (off), (len), NULL, XBRW_ZERO)
243 static inline int xfs_buf_geterror(xfs_buf_t
*bp
)
245 return bp
? bp
->b_error
: ENOMEM
;
248 /* Buffer Utility Routines */
249 extern xfs_caddr_t
xfs_buf_offset(xfs_buf_t
*, size_t);
251 /* Delayed Write Buffer Routines */
252 extern void xfs_buf_delwri_dequeue(xfs_buf_t
*);
253 extern void xfs_buf_delwri_promote(xfs_buf_t
*);
255 /* Buffer Daemon Setup Routines */
256 extern int xfs_buf_init(void);
257 extern void xfs_buf_terminate(void);
259 #define xfs_buf_target_name(target) \
260 ({ char __b[BDEVNAME_SIZE]; bdevname((target)->bt_bdev, __b); __b; })
263 #define XFS_BUF_BFLAGS(bp) ((bp)->b_flags)
264 #define XFS_BUF_ZEROFLAGS(bp) ((bp)->b_flags &= \
265 ~(XBF_READ|XBF_WRITE|XBF_ASYNC|XBF_DELWRI|XBF_ORDERED))
267 #define XFS_BUF_STALE(bp) ((bp)->b_flags |= XBF_STALE)
268 #define XFS_BUF_UNSTALE(bp) ((bp)->b_flags &= ~XBF_STALE)
269 #define XFS_BUF_ISSTALE(bp) ((bp)->b_flags & XBF_STALE)
270 #define XFS_BUF_SUPER_STALE(bp) do { \
272 xfs_buf_delwri_dequeue(bp); \
276 #define XFS_BUF_DELAYWRITE(bp) ((bp)->b_flags |= XBF_DELWRI)
277 #define XFS_BUF_UNDELAYWRITE(bp) xfs_buf_delwri_dequeue(bp)
278 #define XFS_BUF_ISDELAYWRITE(bp) ((bp)->b_flags & XBF_DELWRI)
280 #define XFS_BUF_ERROR(bp,no) xfs_buf_ioerror(bp,no)
281 #define XFS_BUF_GETERROR(bp) xfs_buf_geterror(bp)
282 #define XFS_BUF_ISERROR(bp) (xfs_buf_geterror(bp) ? 1 : 0)
284 #define XFS_BUF_DONE(bp) ((bp)->b_flags |= XBF_DONE)
285 #define XFS_BUF_UNDONE(bp) ((bp)->b_flags &= ~XBF_DONE)
286 #define XFS_BUF_ISDONE(bp) ((bp)->b_flags & XBF_DONE)
288 #define XFS_BUF_BUSY(bp) do { } while (0)
289 #define XFS_BUF_UNBUSY(bp) do { } while (0)
290 #define XFS_BUF_ISBUSY(bp) (1)
292 #define XFS_BUF_ASYNC(bp) ((bp)->b_flags |= XBF_ASYNC)
293 #define XFS_BUF_UNASYNC(bp) ((bp)->b_flags &= ~XBF_ASYNC)
294 #define XFS_BUF_ISASYNC(bp) ((bp)->b_flags & XBF_ASYNC)
296 #define XFS_BUF_ORDERED(bp) ((bp)->b_flags |= XBF_ORDERED)
297 #define XFS_BUF_UNORDERED(bp) ((bp)->b_flags &= ~XBF_ORDERED)
298 #define XFS_BUF_ISORDERED(bp) ((bp)->b_flags & XBF_ORDERED)
300 #define XFS_BUF_HOLD(bp) xfs_buf_hold(bp)
301 #define XFS_BUF_READ(bp) ((bp)->b_flags |= XBF_READ)
302 #define XFS_BUF_UNREAD(bp) ((bp)->b_flags &= ~XBF_READ)
303 #define XFS_BUF_ISREAD(bp) ((bp)->b_flags & XBF_READ)
305 #define XFS_BUF_WRITE(bp) ((bp)->b_flags |= XBF_WRITE)
306 #define XFS_BUF_UNWRITE(bp) ((bp)->b_flags &= ~XBF_WRITE)
307 #define XFS_BUF_ISWRITE(bp) ((bp)->b_flags & XBF_WRITE)
309 #define XFS_BUF_IODONE_FUNC(bp) ((bp)->b_iodone)
310 #define XFS_BUF_SET_IODONE_FUNC(bp, func) ((bp)->b_iodone = (func))
311 #define XFS_BUF_CLR_IODONE_FUNC(bp) ((bp)->b_iodone = NULL)
313 #define XFS_BUF_FSPRIVATE(bp, type) ((type)(bp)->b_fspriv)
314 #define XFS_BUF_SET_FSPRIVATE(bp, val) ((bp)->b_fspriv = (void*)(val))
315 #define XFS_BUF_FSPRIVATE2(bp, type) ((type)(bp)->b_fspriv2)
316 #define XFS_BUF_SET_FSPRIVATE2(bp, val) ((bp)->b_fspriv2 = (void*)(val))
317 #define XFS_BUF_SET_START(bp) do { } while (0)
318 #define XFS_BUF_SET_BRELSE_FUNC(bp, func) ((bp)->b_relse = (func))
320 #define XFS_BUF_PTR(bp) (xfs_caddr_t)((bp)->b_addr)
321 #define XFS_BUF_SET_PTR(bp, val, cnt) xfs_buf_associate_memory(bp, val, cnt)
322 #define XFS_BUF_ADDR(bp) ((bp)->b_bn)
323 #define XFS_BUF_SET_ADDR(bp, bno) ((bp)->b_bn = (xfs_daddr_t)(bno))
324 #define XFS_BUF_OFFSET(bp) ((bp)->b_file_offset)
325 #define XFS_BUF_SET_OFFSET(bp, off) ((bp)->b_file_offset = (off))
326 #define XFS_BUF_COUNT(bp) ((bp)->b_count_desired)
327 #define XFS_BUF_SET_COUNT(bp, cnt) ((bp)->b_count_desired = (cnt))
328 #define XFS_BUF_SIZE(bp) ((bp)->b_buffer_length)
329 #define XFS_BUF_SET_SIZE(bp, cnt) ((bp)->b_buffer_length = (cnt))
331 #define XFS_BUF_SET_VTYPE_REF(bp, type, ref) do { } while (0)
332 #define XFS_BUF_SET_VTYPE(bp, type) do { } while (0)
333 #define XFS_BUF_SET_REF(bp, ref) do { } while (0)
335 #define XFS_BUF_ISPINNED(bp) atomic_read(&((bp)->b_pin_count))
337 #define XFS_BUF_VALUSEMA(bp) xfs_buf_lock_value(bp)
338 #define XFS_BUF_CPSEMA(bp) (xfs_buf_cond_lock(bp) == 0)
339 #define XFS_BUF_VSEMA(bp) xfs_buf_unlock(bp)
340 #define XFS_BUF_PSEMA(bp,x) xfs_buf_lock(bp)
341 #define XFS_BUF_FINISH_IOWAIT(bp) complete(&bp->b_iowait);
343 #define XFS_BUF_SET_TARGET(bp, target) ((bp)->b_target = (target))
344 #define XFS_BUF_TARGET(bp) ((bp)->b_target)
345 #define XFS_BUFTARG_NAME(target) xfs_buf_target_name(target)
347 static inline void xfs_buf_relse(xfs_buf_t
*bp
)
355 * Handling of buftargs.
357 extern xfs_buftarg_t
*xfs_alloc_buftarg(struct xfs_mount
*,
358 struct block_device
*, int, const char *);
359 extern void xfs_free_buftarg(struct xfs_mount
*, struct xfs_buftarg
*);
360 extern void xfs_wait_buftarg(xfs_buftarg_t
*);
361 extern int xfs_setsize_buftarg(xfs_buftarg_t
*, unsigned int, unsigned int);
362 extern int xfs_flush_buftarg(xfs_buftarg_t
*, int);
364 #ifdef CONFIG_KDB_MODULES
365 extern struct list_head
*xfs_get_buftarg_list(void);
368 #define xfs_getsize_buftarg(buftarg) block_size((buftarg)->bt_bdev)
369 #define xfs_readonly_buftarg(buftarg) bdev_read_only((buftarg)->bt_bdev)
371 #define xfs_binval(buftarg) xfs_flush_buftarg(buftarg, 1)
372 #define XFS_bflush(buftarg) xfs_flush_buftarg(buftarg, 1)
374 #endif /* __XFS_BUF_H__ */