2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2000-2003 Tor Egge
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/fcntl.h>
38 #include <sys/limits.h>
39 #include <sys/mount.h>
40 #include <sys/namei.h>
41 #include <sys/vnode.h>
43 #include <sys/filio.h>
44 #include <sys/ttycom.h>
47 #include <sys/rwlock.h>
48 #include <ufs/ufs/extattr.h>
49 #include <ufs/ufs/quota.h>
50 #include <ufs/ufs/inode.h>
51 #include <ufs/ufs/ufsmount.h>
52 #include <ufs/ufs/ufs_extern.h>
53 #include <ufs/ffs/fs.h>
54 #include <ufs/ffs/ffs_extern.h>
57 #include <vm/vm_extern.h>
58 #include <vm/vm_object.h>
59 #include <sys/kernel.h>
60 #include <sys/sysctl.h>
62 static int ffs_rawread_readahead(struct vnode *vp,
68 static int ffs_rawread_main(struct vnode *vp,
71 static int ffs_rawread_sync(struct vnode *vp);
73 int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
75 SYSCTL_DECL(_vfs_ffs);
77 static uma_zone_t ffsraw_pbuf_zone;
79 static int allowrawread = 1;
80 SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0,
81 "Flag to enable raw reads");
83 static int rawreadahead = 1;
84 SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0,
85 "Flag to enable readahead for long raw reads");
88 ffs_rawread_setup(void *arg __unused)
91 ffsraw_pbuf_zone = pbuf_zsecond_create("ffsrawpbuf",
92 (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8);
94 SYSINIT(ffs_raw, SI_SUB_VM_CONF, SI_ORDER_ANY, ffs_rawread_setup, NULL);
97 ffs_rawread_sync(struct vnode *vp)
105 /* Check for dirty mmap, pending writes and dirty buffers */
109 if (bo->bo_numoutput > 0 ||
110 bo->bo_dirty.bv_cnt > 0 ||
111 ((obj = vp->v_object) != NULL &&
112 vm_object_mightbedirty(obj))) {
116 if (vn_start_write(vp, &mp, V_NOWAIT) != 0) {
117 if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE)
122 (void) vn_start_write(vp, &mp, V_WAIT);
123 VOP_LOCK(vp, LK_EXCLUSIVE);
124 } else if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
126 /* Upgrade to exclusive lock, this might block */
127 VOP_LOCK(vp, LK_UPGRADE);
133 /* Check if vnode was reclaimed while unlocked. */
134 if (VN_IS_DOOMED(vp)) {
137 VOP_LOCK(vp, LK_DOWNGRADE);
138 vn_finished_write(mp);
141 /* Attempt to msync mmap() regions to clean dirty mmap */
142 if ((obj = vp->v_object) != NULL &&
143 vm_object_mightbedirty(obj)) {
145 VM_OBJECT_WLOCK(obj);
146 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC);
147 VM_OBJECT_WUNLOCK(obj);
151 /* Wait for pending writes to complete */
153 error = bufobj_wwait(&vp->v_bufobj, 0, 0);
155 /* XXX: can't happen with a zero timeout ??? */
158 VOP_LOCK(vp, LK_DOWNGRADE);
159 vn_finished_write(mp);
162 /* Flush dirty buffers */
163 if (bo->bo_dirty.bv_cnt > 0) {
165 if ((error = ffs_syncvnode(vp, MNT_WAIT, 0)) != 0) {
167 VOP_LOCK(vp, LK_DOWNGRADE);
168 vn_finished_write(mp);
172 if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0)
173 panic("ffs_rawread_sync: dirty bufs");
177 VOP_LOCK(vp, LK_DOWNGRADE);
178 vn_finished_write(mp);
188 ffs_rawread_readahead(struct vnode *vp,
205 bsize = vp->v_mount->mnt_stat.f_iosize;
210 iolen = ((vm_offset_t) udata) & PAGE_MASK;
212 if (bp->b_bcount + iolen > bp->b_kvasize) {
213 bp->b_bcount = bp->b_kvasize;
215 bp->b_bcount -= PAGE_SIZE;
217 bp->b_flags = 0; /* XXX necessary ? */
218 bp->b_iocmd = BIO_READ;
219 bp->b_iodone = bdone;
221 blockno = offset / bsize;
222 blockoff = (offset % bsize) / DEV_BSIZE;
223 if ((daddr_t) blockno != blockno) {
224 return EINVAL; /* blockno overflow */
227 bp->b_lblkno = bp->b_blkno = blockno;
229 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, NULL, &bforwards, NULL);
234 /* Fill holes with NULs to preserve semantics */
236 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize)
237 bp->b_bcount = bsize - blockoff * DEV_BSIZE;
238 bp->b_bufsize = bp->b_bcount;
240 if (vmapbuf(bp, 1) < 0)
244 bzero(bp->b_data, bp->b_bufsize);
246 /* Mark operation completed (similar to bufdone()) */
249 bp->b_flags |= B_DONE;
252 bp->b_blkno = blkno + blockoff;
253 bp->b_offset = bp->b_iooffset = (blkno + blockoff) * DEV_BSIZE;
255 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize * (1 + bforwards))
256 bp->b_bcount = bsize * (1 + bforwards) - blockoff * DEV_BSIZE;
257 bp->b_bufsize = bp->b_bcount;
259 if (vmapbuf(bp, 1) < 0)
262 BO_STRATEGY(&dp->v_bufobj, bp);
268 ffs_rawread_main(struct vnode *vp,
272 struct buf *bp, *nbp, *tbp;
279 td = uio->uio_td ? uio->uio_td : curthread;
280 udata = uio->uio_iov->iov_base;
281 resid = uio->uio_resid;
282 offset = uio->uio_offset;
285 * keep the process from being swapped
297 if (bp == NULL) { /* Setup first read */
298 bp = uma_zalloc(ffsraw_pbuf_zone, M_WAITOK);
300 error = ffs_rawread_readahead(vp, udata, offset,
305 if (resid > bp->b_bufsize) { /* Setup fist readahead */
306 if (rawreadahead != 0)
307 nbp = uma_zalloc(ffsraw_pbuf_zone,
314 nerror = ffs_rawread_readahead(vp,
325 uma_zfree(ffsraw_pbuf_zone,
333 bwait(bp, PRIBIO, "rawrd");
336 iolen = bp->b_bcount - bp->b_resid;
337 if (iolen == 0 && (bp->b_ioflags & BIO_ERROR) == 0) {
338 nerror = 0; /* Ignore possible beyond EOF error */
342 if ((bp->b_ioflags & BIO_ERROR) != 0) {
349 if (iolen < bp->b_bufsize) {
350 /* Incomplete read. Try to read remaining part */
351 error = ffs_rawread_readahead(vp,
354 bp->b_bufsize - iolen,
359 } else if (nbp != NULL) { /* Complete read with readahead */
365 if (resid <= bp->b_bufsize) { /* No more readaheads */
367 uma_zfree(ffsraw_pbuf_zone, nbp);
369 } else { /* Setup next readahead */
370 nerror = ffs_rawread_readahead(vp,
381 uma_zfree(ffsraw_pbuf_zone, nbp);
385 } else if (nerror != 0) {/* Deferred Readahead error */
387 } else if (resid > 0) { /* More to read, no readahead */
388 error = ffs_rawread_readahead(vp, udata, offset,
397 uma_zfree(ffsraw_pbuf_zone, bp);
399 if (nbp != NULL) { /* Run down readahead buffer */
400 bwait(nbp, PRIBIO, "rawrd");
403 uma_zfree(ffsraw_pbuf_zone, nbp);
409 uio->uio_iov->iov_base = udata;
410 uio->uio_resid = resid;
411 uio->uio_offset = offset;
417 ffs_rawread(struct vnode *vp,
421 if (allowrawread != 0 &&
422 uio->uio_iovcnt == 1 &&
423 uio->uio_segflg == UIO_USERSPACE &&
424 uio->uio_resid == uio->uio_iov->iov_len &&
425 (((uio->uio_td != NULL) ? uio->uio_td : curthread)->td_pflags &
426 TDP_DEADLKTREAT) == 0) {
427 int secsize; /* Media sector size */
428 off_t filebytes; /* Bytes left of file */
429 int blockbytes; /* Bytes left of file in full blocks */
430 int partialbytes; /* Bytes in last partial block */
431 int skipbytes; /* Bytes not to read in ffs_rawread */
436 /* Only handle sector aligned reads */
438 secsize = ITODEVVP(ip)->v_bufobj.bo_bsize;
439 if ((uio->uio_offset & (secsize - 1)) == 0 &&
440 (uio->uio_resid & (secsize - 1)) == 0) {
442 /* Sync dirty pages and buffers if needed */
443 error = ffs_rawread_sync(vp);
447 /* Check for end of file */
448 if (ip->i_size > uio->uio_offset) {
449 filebytes = ip->i_size - uio->uio_offset;
451 /* No special eof handling needed ? */
452 if (uio->uio_resid <= filebytes) {
454 return ffs_rawread_main(vp, uio);
457 partialbytes = ((unsigned int) ip->i_size) %
459 blockbytes = (int) filebytes - partialbytes;
460 if (blockbytes > 0) {
461 skipbytes = uio->uio_resid -
463 uio->uio_resid = blockbytes;
464 error = ffs_rawread_main(vp, uio);
465 uio->uio_resid += skipbytes;
468 /* Read remaining part using buffer */