1/*-
2 * Copyright (c) 2000-2003 Tor Egge
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD: head/sys/ufs/ffs/ffs_rawread.c 177493 2008-03-22 09:15:16Z jeff $");
29
30#include <sys/param.h>
31#include <sys/systm.h>
32#include <sys/fcntl.h>
33#include <sys/file.h>
34#include <sys/stat.h>
35#include <sys/proc.h>
36#include <sys/limits.h>
37#include <sys/mount.h>
38#include <sys/namei.h>
39#include <sys/vnode.h>
40#include <sys/conf.h>
41#include <sys/filio.h>
42#include <sys/ttycom.h>
43#include <sys/bio.h>
44#include <sys/buf.h>
45#include <ufs/ufs/extattr.h>
46#include <ufs/ufs/quota.h>
47#include <ufs/ufs/inode.h>
48#include <ufs/ufs/ufsmount.h>
49#include <ufs/ufs/ufs_extern.h>
50#include <ufs/ffs/fs.h>
51#include <ufs/ffs/ffs_extern.h>
52
53#include <vm/vm.h>
54#include <vm/vm_extern.h>
55#include <vm/vm_object.h>
56#include <sys/kernel.h>
57#include <sys/sysctl.h>
58
59static int ffs_rawread_readahead(struct vnode *vp,
60 caddr_t udata,
61 off_t offset,
62 size_t len,
63 struct thread *td,
64 struct buf *bp,
65 caddr_t sa);
66static int ffs_rawread_main(struct vnode *vp,
67 struct uio *uio);
68
69static int ffs_rawread_sync(struct vnode *vp);
70
71int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
72
73void ffs_rawread_setup(void);
74
75SYSCTL_DECL(_vfs_ffs);
76
77static int ffsrawbufcnt = 4;
78SYSCTL_INT(_vfs_ffs, OID_AUTO, ffsrawbufcnt, CTLFLAG_RD, &ffsrawbufcnt, 0,
79 "Buffers available for raw reads");
80
81static int allowrawread = 1;
82SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0,
83 "Flag to enable raw reads");
84
85static int rawreadahead = 1;
86SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0,
87 "Flag to enable readahead for long raw reads");
88
89
90void
91ffs_rawread_setup(void)
92{
93 ffsrawbufcnt = (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8;
94}
95
96
97static int
98ffs_rawread_sync(struct vnode *vp)
99{
100 int error;
101 int upgraded;
102 struct bufobj *bo;
103 struct mount *mp;
104
105 /* Check for dirty mmap, pending writes and dirty buffers */
106 bo = &vp->v_bufobj;
107 BO_LOCK(bo);
108 VI_LOCK(vp);
109 if (bo->bo_numoutput > 0 ||
110 bo->bo_dirty.bv_cnt > 0 ||
111 (vp->v_iflag & VI_OBJDIRTY) != 0) {
112 VI_UNLOCK(vp);
113 BO_UNLOCK(bo);
114
115 if (vn_start_write(vp, &mp, V_NOWAIT) != 0) {
116 if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE)
117 upgraded = 1;
118 else
119 upgraded = 0;
120 VOP_UNLOCK(vp, 0);
121 (void) vn_start_write(vp, &mp, V_WAIT);
122 VOP_LOCK(vp, LK_EXCLUSIVE);
123 } else if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
124 upgraded = 1;
125 /* Upgrade to exclusive lock, this might block */
126 VOP_LOCK(vp, LK_UPGRADE);
127 } else
128 upgraded = 0;
129
130
131 VI_LOCK(vp);
132 /* Check if vnode was reclaimed while unlocked. */
133 if ((vp->v_iflag & VI_DOOMED) != 0) {
134 VI_UNLOCK(vp);
135 if (upgraded != 0)
136 VOP_LOCK(vp, LK_DOWNGRADE);
137 vn_finished_write(mp);
138 return (EIO);
139 }
140 /* Attempt to msync mmap() regions to clean dirty mmap */
141 if ((vp->v_iflag & VI_OBJDIRTY) != 0) {
142 VI_UNLOCK(vp);
143 if (vp->v_object != NULL) {
144 VM_OBJECT_LOCK(vp->v_object);
145 vm_object_page_clean(vp->v_object, 0, 0, OBJPC_SYNC);
146 VM_OBJECT_UNLOCK(vp->v_object);
147 }
148 } else
149 VI_UNLOCK(vp);
150
151 /* Wait for pending writes to complete */
152 BO_LOCK(bo);
153 error = bufobj_wwait(&vp->v_bufobj, 0, 0);
154 if (error != 0) {
155 /* XXX: can't happen with a zero timeout ??? */
156 BO_UNLOCK(bo);
157 if (upgraded != 0)
158 VOP_LOCK(vp, LK_DOWNGRADE);
159 vn_finished_write(mp);
160 return (error);
161 }
162 /* Flush dirty buffers */
163 if (bo->bo_dirty.bv_cnt > 0) {
164 BO_UNLOCK(bo);
165 if ((error = ffs_syncvnode(vp, MNT_WAIT)) != 0) {
166 if (upgraded != 0)
167 VOP_LOCK(vp, LK_DOWNGRADE);
168 vn_finished_write(mp);
169 return (error);
170 }
171 BO_LOCK(bo);
172 if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0)
173 panic("ffs_rawread_sync: dirty bufs");
174 }
175 BO_UNLOCK(bo);
176 if (upgraded != 0)
177 VOP_LOCK(vp, LK_DOWNGRADE);
178 vn_finished_write(mp);
179 } else {
180 VI_UNLOCK(vp);
181 BO_UNLOCK(bo);
182 }
183 return 0;
184}
185
186
187static int
188ffs_rawread_readahead(struct vnode *vp,
189 caddr_t udata,
190 off_t offset,
191 size_t len,
192 struct thread *td,
193 struct buf *bp,
194 caddr_t sa)
195{
196 int error;
197 u_int iolen;
198 off_t blockno;
199 int blockoff;
200 int bsize;
201 struct vnode *dp;
202 int bforwards;
203 struct inode *ip;
204 ufs2_daddr_t blkno;
205
206 bsize = vp->v_mount->mnt_stat.f_iosize;
207
208 ip = VTOI(vp);
209 dp = ip->i_devvp;
210
211 iolen = ((vm_offset_t) udata) & PAGE_MASK;
212 bp->b_bcount = len;
213 if (bp->b_bcount + iolen > bp->b_kvasize) {
214 bp->b_bcount = bp->b_kvasize;
215 if (iolen != 0)
216 bp->b_bcount -= PAGE_SIZE;
217 }
218 bp->b_flags = 0; /* XXX necessary ? */
219 bp->b_iocmd = BIO_READ;
220 bp->b_iodone = bdone;
221 bp->b_data = udata;
222 bp->b_saveaddr = sa;
223 blockno = offset / bsize;
224 blockoff = (offset % bsize) / DEV_BSIZE;
225 if ((daddr_t) blockno != blockno) {
226 return EINVAL; /* blockno overflow */
227 }
228
229 bp->b_lblkno = bp->b_blkno = blockno;
230
231 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, NULL, &bforwards, NULL);
232 if (error != 0)
233 return error;
234 if (blkno == -1) {
235
236 /* Fill holes with NULs to preserve semantics */
237
238 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize)
239 bp->b_bcount = bsize - blockoff * DEV_BSIZE;
240 bp->b_bufsize = bp->b_bcount;
241
242 if (vmapbuf(bp) < 0)
243 return EFAULT;
244
245 if (ticks - PCPU_GET(switchticks) >= hogticks)
246 uio_yield();
247 bzero(bp->b_data, bp->b_bufsize);
248
249 /* Mark operation completed (similar to bufdone()) */
250
251 bp->b_resid = 0;
252 bp->b_flags |= B_DONE;
253 return 0;
254 }
255 bp->b_blkno = blkno + blockoff;
256 bp->b_offset = bp->b_iooffset = (blkno + blockoff) * DEV_BSIZE;
257
258 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize * (1 + bforwards))
259 bp->b_bcount = bsize * (1 + bforwards) - blockoff * DEV_BSIZE;
260 bp->b_bufsize = bp->b_bcount;
261
262 if (vmapbuf(bp) < 0)
263 return EFAULT;
264
265 BO_STRATEGY(&dp->v_bufobj, bp);
266 return 0;
267}
268
269
270static int
271ffs_rawread_main(struct vnode *vp,
272 struct uio *uio)
273{
274 int error, nerror;
275 struct buf *bp, *nbp, *tbp;
276 caddr_t sa, nsa, tsa;
277 u_int iolen;
278 int spl;
279 caddr_t udata;
280 long resid;
281 off_t offset;
282 struct thread *td;
283
284 td = uio->uio_td ? uio->uio_td : curthread;
285 udata = uio->uio_iov->iov_base;
286 resid = uio->uio_resid;
287 offset = uio->uio_offset;
288
289 /*
290 * keep the process from being swapped
291 */
292 PHOLD(td->td_proc);
293
294 error = 0;
295 nerror = 0;
296
297 bp = NULL;
298 nbp = NULL;
299 sa = NULL;
300 nsa = NULL;
301
302 while (resid > 0) {
303
304 if (bp == NULL) { /* Setup first read */
305 /* XXX: Leave some bufs for swap */
306 bp = getpbuf(&ffsrawbufcnt);
307 sa = bp->b_data;
308 pbgetvp(vp, bp);
309 error = ffs_rawread_readahead(vp, udata, offset,
310 resid, td, bp, sa);
311 if (error != 0)
312 break;
313
314 if (resid > bp->b_bufsize) { /* Setup fist readahead */
315 /* XXX: Leave bufs for swap */
316 if (rawreadahead != 0)
317 nbp = trypbuf(&ffsrawbufcnt);
318 else
319 nbp = NULL;
320 if (nbp != NULL) {
321 nsa = nbp->b_data;
322 pbgetvp(vp, nbp);
323
324 nerror = ffs_rawread_readahead(vp,
325 udata +
326 bp->b_bufsize,
327 offset +
328 bp->b_bufsize,
329 resid -
330 bp->b_bufsize,
331 td,
332 nbp,
333 nsa);
334 if (nerror) {
335 pbrelvp(nbp);
336 relpbuf(nbp, &ffsrawbufcnt);
337 nbp = NULL;
338 }
339 }
340 }
341 }
342
343 spl = splbio();
344 bwait(bp, PRIBIO, "rawrd");
345 splx(spl);
346
347 vunmapbuf(bp);
348
349 iolen = bp->b_bcount - bp->b_resid;
350 if (iolen == 0 && (bp->b_ioflags & BIO_ERROR) == 0) {
351 nerror = 0; /* Ignore possible beyond EOF error */
352 break; /* EOF */
353 }
354
355 if ((bp->b_ioflags & BIO_ERROR) != 0) {
356 error = bp->b_error;
357 break;
358 }
359 resid -= iolen;
360 udata += iolen;
361 offset += iolen;
362 if (iolen < bp->b_bufsize) {
363 /* Incomplete read. Try to read remaining part */
364 error = ffs_rawread_readahead(vp,
365 udata,
366 offset,
367 bp->b_bufsize - iolen,
368 td,
369 bp,
370 sa);
371 if (error != 0)
372 break;
373 } else if (nbp != NULL) { /* Complete read with readahead */
374
375 tbp = bp;
376 bp = nbp;
377 nbp = tbp;
378
379 tsa = sa;
380 sa = nsa;
381 nsa = tsa;
382
383 if (resid <= bp->b_bufsize) { /* No more readaheads */
384 pbrelvp(nbp);
385 relpbuf(nbp, &ffsrawbufcnt);
386 nbp = NULL;
387 } else { /* Setup next readahead */
388 nerror = ffs_rawread_readahead(vp,
389 udata +
390 bp->b_bufsize,
391 offset +
392 bp->b_bufsize,
393 resid -
394 bp->b_bufsize,
395 td,
396 nbp,
397 nsa);
398 if (nerror != 0) {
399 pbrelvp(nbp);
400 relpbuf(nbp, &ffsrawbufcnt);
401 nbp = NULL;
402 }
403 }
404 } else if (nerror != 0) {/* Deferred Readahead error */
405 break;
406 } else if (resid > 0) { /* More to read, no readahead */
407 error = ffs_rawread_readahead(vp, udata, offset,
408 resid, td, bp, sa);
409 if (error != 0)
410 break;
411 }
412 }
413
414 if (bp != NULL) {
415 pbrelvp(bp);
416 relpbuf(bp, &ffsrawbufcnt);
417 }
418 if (nbp != NULL) { /* Run down readahead buffer */
419 spl = splbio();
420 bwait(nbp, PRIBIO, "rawrd");
421 splx(spl);
422 vunmapbuf(nbp);
423 pbrelvp(nbp);
424 relpbuf(nbp, &ffsrawbufcnt);
425 }
426
427 if (error == 0)
428 error = nerror;
429 PRELE(td->td_proc);
430 uio->uio_iov->iov_base = udata;
431 uio->uio_resid = resid;
432 uio->uio_offset = offset;
433 return error;
434}
435
436
437int
438ffs_rawread(struct vnode *vp,
439 struct uio *uio,
440 int *workdone)
441{
442 if (allowrawread != 0 &&
443 uio->uio_iovcnt == 1 &&
444 uio->uio_segflg == UIO_USERSPACE &&
445 uio->uio_resid == uio->uio_iov->iov_len &&
446 (((uio->uio_td != NULL) ? uio->uio_td : curthread)->td_pflags &
447 TDP_DEADLKTREAT) == 0) {
448 int secsize; /* Media sector size */
449 off_t filebytes; /* Bytes left of file */
450 int blockbytes; /* Bytes left of file in full blocks */
451 int partialbytes; /* Bytes in last partial block */
452 int skipbytes; /* Bytes not to read in ffs_rawread */
453 struct inode *ip;
454 int error;
455
456
457 /* Only handle sector aligned reads */
458 ip = VTOI(vp);
459 secsize = ip->i_devvp->v_bufobj.bo_bsize;
460 if ((uio->uio_offset & (secsize - 1)) == 0 &&
461 (uio->uio_resid & (secsize - 1)) == 0) {
462
463 /* Sync dirty pages and buffers if needed */
464 error = ffs_rawread_sync(vp);
465 if (error != 0)
466 return error;
467
468 /* Check for end of file */
469 if (ip->i_size > uio->uio_offset) {
470 filebytes = ip->i_size - uio->uio_offset;
471
472 /* No special eof handling needed ? */
473 if (uio->uio_resid <= filebytes) {
474 *workdone = 1;
475 return ffs_rawread_main(vp, uio);
476 }
477
478 partialbytes = ((unsigned int) ip->i_size) %
479 ip->i_fs->fs_bsize;
480 blockbytes = (int) filebytes - partialbytes;
481 if (blockbytes > 0) {
482 skipbytes = uio->uio_resid -
483 blockbytes;
484 uio->uio_resid = blockbytes;
485 error = ffs_rawread_main(vp, uio);
486 uio->uio_resid += skipbytes;
487 if (error != 0)
488 return error;
489 /* Read remaining part using buffer */
490 }
491 }
492 }
493 }
494 *workdone = 0;
495 return 0;
496}