2 * Copyright (c) 2000-2003 Tor Egge
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.
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
26 * $FreeBSD: src/sys/ufs/ffs/ffs_rawread.c,v 1.3.2.2 2003/05/29 06:15:35 alc Exp $
27 * $DragonFly: src/sys/vfs/ufs/ffs_rawread.c,v 1.28 2008/06/19 23:27:39 dillon Exp $
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/fcntl.h>
36 #include <sys/mount.h>
37 #include <sys/namei.h>
38 #include <sys/vnode.h>
40 #include <sys/filio.h>
41 #include <sys/ttycom.h>
47 #include <machine/limits.h>
49 #include <vm/vm_extern.h>
50 #include <vm/vm_object.h>
51 #include <sys/kernel.h>
52 #include <sys/sysctl.h>
54 static int ffs_rawread_readahead(struct vnode *vp, caddr_t udata, off_t offset,
55 size_t len, struct buf *bp);
56 static int ffs_rawread_main(struct vnode *vp,
59 static int ffs_rawread_sync(struct vnode *vp);
61 int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
63 void ffs_rawread_setup(void);
65 SYSCTL_DECL(_vfs_ffs);
67 static int ffsrawbufcnt = 4;
68 SYSCTL_INT(_vfs_ffs, OID_AUTO, ffsrawbufcnt, CTLFLAG_RD, &ffsrawbufcnt, 0,
69 "Buffers available for raw reads");
71 static int allowrawread = 1;
72 SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0,
73 "Flag to enable raw reads");
75 static int rawreadahead = 1;
76 SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0,
77 "Flag to enable readahead for long raw reads");
81 ffs_rawread_setup(void)
83 ffsrawbufcnt = (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8;
88 ffs_rawread_sync(struct vnode *vp)
94 * Check for dirty mmap, pending writes and dirty buffers
96 lwkt_gettoken(&vp->v_token);
97 if (bio_track_active(&vp->v_track_write) ||
98 !RB_EMPTY(&vp->v_rbdirty_tree) ||
99 (vp->v_flag & VOBJDIRTY) != 0) {
100 if (vn_islocked(vp) != LK_EXCLUSIVE) {
102 /* Upgrade to exclusive lock, this might block */
103 vn_lock(vp, LK_UPGRADE);
107 /* Attempt to msync mmap() regions to clean dirty mmap */
108 if ((vp->v_flag & VOBJDIRTY) != 0) {
109 struct vm_object *obj;
110 if ((obj = vp->v_object) != NULL)
111 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC);
114 /* Wait for pending writes to complete */
115 error = bio_track_wait(&vp->v_track_write, 0, 0);
118 vn_lock(vp, LK_DOWNGRADE);
121 /* Flush dirty buffers */
122 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
123 if ((error = VOP_FSYNC(vp, MNT_WAIT, 0)) != 0) {
125 vn_lock(vp, LK_DOWNGRADE);
128 if (bio_track_active(&vp->v_track_write) ||
129 !RB_EMPTY(&vp->v_rbdirty_tree))
130 panic("ffs_rawread_sync: dirty bufs");
133 vn_lock(vp, LK_DOWNGRADE);
138 lwkt_reltoken(&vp->v_token);
144 ffs_rawread_readahead(struct vnode *vp, caddr_t udata, off_t loffset,
145 size_t len, struct buf *bp)
154 bsize = vp->v_mount->mnt_stat.f_iosize;
157 * Make sure it fits into the pbuf
159 iolen = (int)(intptr_t)udata & PAGE_MASK;
160 if (len + iolen > bp->b_kvasize) {
167 * Raw disk address is in bio2, but we wait for it to
170 bp->b_flags &= ~B_ERROR;
171 bp->b_loffset = loffset;
172 bp->b_bio2.bio_offset = NOOFFSET;
173 bp->b_bio1.bio_done = biodone_sync;
174 bp->b_bio1.bio_flags |= BIO_SYNC;
176 blockoff = (loffset % bsize) / DEV_BSIZE;
178 error = VOP_BMAP(vp, bp->b_loffset, &bp->b_bio2.bio_offset,
179 &bforwards, NULL, BUF_CMD_READ);
182 dp = VTOI(vp)->i_devvp;
183 if (bp->b_bio2.bio_offset == NOOFFSET) {
185 * Fill holes with NULs to preserve semantics
187 if (len + blockoff * DEV_BSIZE > bsize)
188 len = bsize - blockoff * DEV_BSIZE;
190 if (vmapbuf(bp, udata, len) < 0)
194 bzero(bp->b_data, bp->b_bcount);
196 /* Mark operation completed (similar to bufdone()) */
202 if (len + blockoff * DEV_BSIZE > bforwards)
203 len = bforwards - blockoff * DEV_BSIZE;
204 bp->b_bio2.bio_offset += blockoff * DEV_BSIZE;
206 if (vmapbuf(bp, udata, len) < 0)
210 * Access the block device layer using the device vnode (dp) and
211 * the translated block number (bio2) instead of the logical block
214 * Even though we are bypassing the vnode layer, we still
215 * want the vnode state to indicate that an I/O on its behalf
218 bp->b_cmd = BUF_CMD_READ;
219 bio_start_transaction(&bp->b_bio1, &vp->v_track_read);
220 vn_strategy(dp, &bp->b_bio2);
225 ffs_rawread_main(struct vnode *vp, struct uio *uio)
228 struct buf *bp, *nbp, *tbp;
234 udata = uio->uio_iov->iov_base;
235 resid = uio->uio_resid;
236 offset = uio->uio_offset;
246 if (bp == NULL) { /* Setup first read */
247 /* XXX: Leave some bufs for swap */
248 bp = getpbuf_kva(&ffsrawbufcnt);
249 error = ffs_rawread_readahead(vp, udata, offset,
254 if (resid > bp->b_bufsize) { /* Setup fist readahead */
255 /* XXX: Leave bufs for swap */
256 if (rawreadahead != 0)
257 nbp = trypbuf_kva(&ffsrawbufcnt);
261 nerror = ffs_rawread_readahead(
263 udata + bp->b_bufsize,
264 offset + bp->b_bufsize,
265 resid - bp->b_bufsize,
268 relpbuf(nbp, &ffsrawbufcnt);
275 biowait(&bp->b_bio1, "rawrd");
279 iolen = bp->b_bcount - bp->b_resid;
280 if (iolen == 0 && (bp->b_flags & B_ERROR) == 0) {
281 nerror = 0; /* Ignore possible beyond EOF error */
285 if ((bp->b_flags & B_ERROR) != 0) {
289 clearbiocache(&bp->b_bio2);
293 if (iolen < bp->b_bufsize) {
294 /* Incomplete read. Try to read remaining part */
295 error = ffs_rawread_readahead(
297 bp->b_bufsize - iolen, bp);
300 } else if (nbp != NULL) { /* Complete read with readahead */
306 clearbiocache(&nbp->b_bio2);
308 if (resid <= bp->b_bufsize) { /* No more readaheads */
309 relpbuf(nbp, &ffsrawbufcnt);
311 } else { /* Setup next readahead */
312 nerror = ffs_rawread_readahead(
313 vp, udata + bp->b_bufsize,
314 offset + bp->b_bufsize,
315 resid - bp->b_bufsize,
318 relpbuf(nbp, &ffsrawbufcnt);
322 } else if (nerror != 0) {/* Deferred Readahead error */
324 } else if (resid > 0) { /* More to read, no readahead */
325 error = ffs_rawread_readahead(vp, udata, offset,
333 relpbuf(bp, &ffsrawbufcnt);
334 if (nbp != NULL) { /* Run down readahead buffer */
335 biowait(&nbp->b_bio1, "rawrd");
337 relpbuf(nbp, &ffsrawbufcnt);
342 uio->uio_iov->iov_base = udata;
343 uio->uio_resid = resid;
344 uio->uio_offset = offset;
350 ffs_rawread(struct vnode *vp,
354 if (allowrawread != 0 &&
355 uio->uio_iovcnt == 1 &&
356 uio->uio_segflg == UIO_USERSPACE &&
357 uio->uio_resid == uio->uio_iov->iov_len &&
358 (curthread->td_flags & TDF_DEADLKTREAT) == 0) {
359 int secsize; /* Media sector size */
360 off_t filebytes; /* Bytes left of file */
361 int blockbytes; /* Bytes left of file in full blocks */
362 int partialbytes; /* Bytes in last partial block */
363 int skipbytes; /* Bytes not to read in ffs_rawread */
368 /* Only handle sector aligned reads */
370 secsize = ip->i_devvp->v_rdev->si_bsize_phys;
371 if ((uio->uio_offset & (secsize - 1)) == 0 &&
372 (uio->uio_resid & (secsize - 1)) == 0) {
374 /* Sync dirty pages and buffers if needed */
375 error = ffs_rawread_sync(vp);
379 /* Check for end of file */
380 if (ip->i_size > uio->uio_offset) {
381 filebytes = ip->i_size - uio->uio_offset;
383 /* No special eof handling needed ? */
384 if (uio->uio_resid <= filebytes) {
386 return ffs_rawread_main(vp, uio);
389 partialbytes = ((unsigned int) ip->i_size) %
391 blockbytes = (int) filebytes - partialbytes;
392 if (blockbytes > 0) {
393 skipbytes = uio->uio_resid -
395 uio->uio_resid = blockbytes;
396 error = ffs_rawread_main(vp, uio);
397 uio->uio_resid += skipbytes;
400 /* Read remaining part using buffer */