3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)ufs_readwrite.c 8.11 (Berkeley) 5/8/95
34 * $FreeBSD: src/sys/ufs/ufs/ufs_readwrite.c,v 1.65.2.14 2003/04/04 22:21:29 tegge Exp $
37 #define BLKSIZE(a, b, c) blksize(a, b, c)
42 #include <vm/vm_object.h>
43 #include <vm/vm_pager.h>
44 #include <vm/vm_map.h>
45 #include <vm/vnode_pager.h>
46 #include <sys/event.h>
47 #include <sys/vmmeter.h>
48 #include <sys/sysctl.h>
49 #include <vm/vm_page2.h>
51 #include "opt_directio.h"
53 #define VN_KNOTE(vp, b) \
54 KNOTE((struct klist *)&vp->v_pollinfo.vpi_kqinfo.ki_note, (b))
57 extern int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
60 SYSCTL_DECL(_vfs_ffs);
63 * Vnode op for reading.
65 * ffs_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
66 * struct ucred *a_cred)
70 ffs_read(struct vop_read_args *ap)
78 int xfersize, blkoffset;
79 int error, orig_resid;
84 seqcount = ap->a_ioflag >> 16;
87 ioflag = ap->a_ioflag;
89 if ((ioflag & IO_DIRECT) != 0) {
92 error = ffs_rawread(vp, uio, &workdone);
93 if (error || workdone)
99 if (uio->uio_rw != UIO_READ)
100 panic("ffs_read: mode");
102 if (vp->v_type == VLNK) {
103 if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
104 panic("ffs_read: short symlink");
105 } else if (vp->v_type != VREG && vp->v_type != VDIR)
106 panic("ffs_read: type %d", vp->v_type);
109 if ((uint64_t)uio->uio_offset > fs->fs_maxfilesize)
112 orig_resid = uio->uio_resid;
116 bytesinfile = ip->i_size - uio->uio_offset;
117 if (bytesinfile <= 0) {
118 if ((vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
119 ip->i_flag |= IN_ACCESS;
124 * Ok so we couldn't do it all in one vm trick...
125 * so cycle around trying smaller bites..
127 for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
128 if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
131 error = ffs_blkatoff_ra(vp, uio->uio_offset, NULL,
137 * If IO_DIRECT then set B_DIRECT for the buffer. This
138 * will cause us to attempt to release the buffer later on
139 * and will cause the buffer cache to attempt to free the
142 if (ioflag & IO_DIRECT)
143 bp->b_flags |= B_DIRECT;
146 * We should only get non-zero b_resid when an I/O error
147 * has occurred, which should cause us to break above.
148 * However, if the short read did not cause an error,
149 * then we want to ensure that we do not uiomove bad
150 * or uninitialized data.
152 * XXX b_resid is only valid when an actual I/O has occured
153 * and may be incorrect if the buffer is B_CACHE or if the
154 * last op on the buffer was a failed write. This KASSERT
155 * is a precursor to removing it from the UFS code.
157 KASSERT(bp->b_resid == 0, ("bp->b_resid != 0"));
160 * Calculate how much data we can copy
162 blkoffset = blkoff(fs, uio->uio_offset);
163 xfersize = bp->b_bufsize - blkoffset;
164 if (xfersize > uio->uio_resid)
165 xfersize = uio->uio_resid;
166 if (xfersize > bytesinfile)
167 xfersize = bytesinfile;
169 panic("ufs_readwrite: impossible xfersize: %d",
174 * otherwise use the general form
176 error = uiomovebp(bp, bp->b_data + blkoffset, xfersize, uio);
181 if ((ioflag & (IO_VMIO|IO_DIRECT)) &&
182 (LIST_FIRST(&bp->b_dep) == NULL)) {
184 * If there are no dependencies, and it's VMIO,
185 * then we don't need the buf, mark it available
186 * for freeing. The VM has the data.
188 bp->b_flags |= B_RELBUF;
192 * Otherwise let whoever
193 * made the request take care of
194 * freeing it. We just queue
195 * it onto another list.
202 * This can only happen in the case of an error
203 * because the loop above resets bp to NULL on each iteration
204 * and on normal completion has not set a new value into it.
205 * so it must have come from a 'break' statement
208 if ((ioflag & (IO_VMIO|IO_DIRECT)) &&
209 (LIST_FIRST(&bp->b_dep) == NULL)) {
210 bp->b_flags |= B_RELBUF;
217 if ((error == 0 || uio->uio_resid != orig_resid) &&
218 (vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
219 ip->i_flag |= IN_ACCESS;
224 * Vnode op for writing.
226 * ffs_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
227 * struct ucred *a_cred)
230 ffs_write(struct vop_write_args *ap)
241 int blkoffset, error, extended, flags, ioflag, resid, size, xfersize;
245 seqcount = ap->a_ioflag >> 16;
246 ioflag = ap->a_ioflag;
252 if (uio->uio_rw != UIO_WRITE)
253 panic("ffs_write: mode");
256 switch (vp->v_type) {
258 if (ioflag & IO_APPEND)
259 uio->uio_offset = ip->i_size;
260 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
266 panic("ffs_write: dir write");
269 panic("ffs_write: type %p %d (%d,%d)", vp, (int)vp->v_type,
270 (int)uio->uio_offset,
276 if (uio->uio_offset < 0 ||
277 (uint64_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize) {
281 * Maybe this should be above the vnode op call, but so long as
282 * file servers have no limits, I don't think it matters.
285 if (vp->v_type == VREG && td && td->td_proc &&
286 uio->uio_offset + uio->uio_resid >
287 td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
288 lwpsignal(td->td_proc, td->td_lwp, SIGXFSZ);
292 resid = uio->uio_resid;
296 * NOTE! These B_ flags are actually balloc-only flags, not buffer
297 * flags. They are similar to the BA_ flags in fbsd.
299 if (seqcount > B_SEQMAX)
300 flags = B_SEQMAX << B_SEQSHIFT;
302 flags = seqcount << B_SEQSHIFT;
303 if ((ioflag & IO_SYNC) && !DOINGASYNC(vp))
306 for (error = 0; uio->uio_resid > 0;) {
307 lbn = lblkno(fs, uio->uio_offset);
308 blkoffset = blkoff(fs, uio->uio_offset);
309 xfersize = fs->fs_bsize - blkoffset;
310 if (uio->uio_resid < xfersize)
311 xfersize = uio->uio_resid;
313 if (uio->uio_offset + xfersize > ip->i_size) {
314 nsize = uio->uio_offset + xfersize;
315 nvnode_pager_setsize(vp, nsize,
316 blkoffresize(fs, nsize), blkoff(fs, nsize));
321 * If doing a dummy write to flush the buffer for a
322 * putpages we must perform a read-before-write to
323 * fill in any missing spots and clear any invalid
324 * areas. Otherwise a multi-page buffer may not properly
327 * We must clear any invalid areas
329 if (uio->uio_segflg == UIO_NOCOPY) {
330 error = ffs_blkatoff(vp, uio->uio_offset, NULL, &bp);
338 * We must clear invalid areas.
340 if (xfersize < fs->fs_bsize || uio->uio_segflg == UIO_NOCOPY)
344 /* XXX is uio->uio_offset the right thing here? */
345 error = VOP_BALLOC(vp, uio->uio_offset, xfersize,
346 ap->a_cred, flags, &bp);
350 * If the buffer is not valid and we did not clear garbage
351 * out above, we have to do so here even though the write
352 * covers the entire buffer in order to avoid a mmap()/write
353 * race where another process may see the garbage prior to
354 * the uiomove() for a write replacing it.
356 if ((bp->b_flags & B_CACHE) == 0 && (flags & B_CLRBUF) == 0)
358 if (ioflag & IO_DIRECT)
359 bp->b_flags |= B_DIRECT;
360 if ((ioflag & (IO_SYNC|IO_INVAL)) == (IO_SYNC|IO_INVAL))
361 bp->b_flags |= B_NOCACHE;
363 if (uio->uio_offset + xfersize > ip->i_size) {
364 ip->i_size = uio->uio_offset + xfersize;
368 size = BLKSIZE(fs, ip, lbn) - bp->b_resid;
372 error = uiomovebp(bp, bp->b_data + blkoffset, xfersize, uio);
373 if ((ioflag & (IO_VMIO|IO_DIRECT)) &&
374 (LIST_FIRST(&bp->b_dep) == NULL)) {
375 bp->b_flags |= B_RELBUF;
379 * If IO_SYNC each buffer is written synchronously. Otherwise
380 * if we have a severe page deficiency write the buffer
381 * asynchronously. Otherwise try to cluster, and if that
382 * doesn't do it then either do an async write (if O_DIRECT),
383 * or a delayed write (if not).
386 if (ioflag & IO_SYNC) {
388 } else if (vm_page_count_severe() ||
389 buf_dirty_count_severe() ||
390 (ioflag & IO_ASYNC)) {
391 bp->b_flags |= B_CLUSTEROK;
393 } else if (xfersize + blkoffset == fs->fs_bsize) {
394 if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
395 bp->b_flags |= B_CLUSTEROK;
396 cluster_write(bp, (off_t)ip->i_size, fs->fs_bsize, seqcount);
400 } else if (ioflag & IO_DIRECT) {
401 bp->b_flags |= B_CLUSTEROK;
404 bp->b_flags |= B_CLUSTEROK;
407 if (error || xfersize == 0)
409 ip->i_flag |= IN_CHANGE | IN_UPDATE;
412 * If we successfully wrote any data, and we are not the superuser
413 * we clear the setuid and setgid bits as a precaution against
416 if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
417 ip->i_mode &= ~(ISUID | ISGID);
418 if (resid > uio->uio_resid)
419 VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
421 if (ioflag & IO_UNIT) {
422 (void)ffs_truncate(vp, osize, ioflag & IO_SYNC,
424 uio->uio_offset -= resid - uio->uio_resid;
425 uio->uio_resid = resid;
427 } else if (resid > uio->uio_resid && (ioflag & IO_SYNC)) {
428 error = ffs_update(vp, 1);