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 1.27 2007/08/13 17:31:56 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, int *baseticks);
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 static void ffs_rawreadwakeup(struct bio *bio);
68 SYSCTL_DECL(_vfs_ffs);
70 static int ffsrawbufcnt = 4;
71 SYSCTL_INT(_vfs_ffs, OID_AUTO, ffsrawbufcnt, CTLFLAG_RD, &ffsrawbufcnt, 0,
72 "Buffers available for raw reads");
74 static int allowrawread = 1;
75 SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0,
76 "Flag to enable raw reads");
78 static int rawreadahead = 1;
79 SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0,
80 "Flag to enable readahead for long raw reads");
84 ffs_rawread_setup(void)
86 ffsrawbufcnt = (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8;
91 ffs_rawread_sync(struct vnode *vp)
96 /* Check for dirty mmap, pending writes and dirty buffers */
98 if (vp->v_track_write.bk_active > 0 ||
99 !RB_EMPTY(&vp->v_rbdirty_tree) ||
100 (vp->v_flag & VOBJDIRTY) != 0) {
103 if (vn_islocked(vp) != LK_EXCLUSIVE) {
105 /* Upgrade to exclusive lock, this might block */
106 vn_lock(vp, LK_UPGRADE);
110 /* Attempt to msync mmap() regions to clean dirty mmap */
111 if ((vp->v_flag & VOBJDIRTY) != 0) {
112 struct vm_object *obj;
113 if ((obj = vp->v_object) != NULL)
114 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC);
117 /* Wait for pending writes to complete */
119 while (vp->v_track_write.bk_active) {
120 vp->v_track_write.bk_waitflag = 1;
121 error = tsleep(&vp->v_track_write, 0, "rawrdfls", 0);
125 vn_lock(vp, LK_DOWNGRADE);
129 /* Flush dirty buffers */
130 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
132 if ((error = VOP_FSYNC(vp, MNT_WAIT)) != 0) {
134 vn_lock(vp, LK_DOWNGRADE);
138 if (vp->v_track_write.bk_active > 0 ||
139 !RB_EMPTY(&vp->v_rbdirty_tree))
140 panic("ffs_rawread_sync: dirty bufs");
144 vn_lock(vp, LK_DOWNGRADE);
153 ffs_rawread_readahead(struct vnode *vp, caddr_t udata, off_t loffset,
154 size_t len, struct buf *bp, int *baseticks)
163 bsize = vp->v_mount->mnt_stat.f_iosize;
166 * Make sure it fits into the pbuf
168 iolen = (int)(intptr_t)udata & PAGE_MASK;
169 if (len + iolen > bp->b_kvasize) {
174 bp->b_flags &= ~B_ERROR;
175 bp->b_loffset = loffset;
176 bp->b_bio2.bio_offset = NOOFFSET;
177 bp->b_bio2.bio_done = ffs_rawreadwakeup;
179 blockoff = (loffset % bsize) / DEV_BSIZE;
181 error = VOP_BMAP(vp, bp->b_loffset, &bp->b_bio2.bio_offset,
185 dp = VTOI(vp)->i_devvp;
186 if (bp->b_bio2.bio_offset == NOOFFSET) {
188 * Fill holes with NULs to preserve semantics
190 if (len + blockoff * DEV_BSIZE > bsize)
191 len = bsize - blockoff * DEV_BSIZE;
193 if (vmapbuf(bp, udata, len) < 0)
196 if (ticks - *baseticks >= hogticks) {
200 bzero(bp->b_data, bp->b_bcount);
202 /* Mark operation completed (similar to bufdone()) */
208 if (len + blockoff * DEV_BSIZE > bforwards)
209 len = bforwards - blockoff * DEV_BSIZE;
210 bp->b_bio2.bio_offset += blockoff * DEV_BSIZE;
212 if (vmapbuf(bp, udata, len) < 0)
216 * Access the block device layer using the device vnode (dp) and
217 * the translated block number (bio2) instead of the logical block
220 * Even though we are bypassing the vnode layer, we still
221 * want the vnode state to indicate that an I/O on its behalf
224 bp->b_cmd = BUF_CMD_READ;
225 bio_start_transaction(&bp->b_bio1, &vp->v_track_read);
226 vn_strategy(dp, &bp->b_bio2);
231 ffs_rawread_main(struct vnode *vp, struct uio *uio)
234 struct buf *bp, *nbp, *tbp;
236 int baseticks = ticks;
241 udata = uio->uio_iov->iov_base;
242 resid = uio->uio_resid;
243 offset = uio->uio_offset;
253 if (bp == NULL) { /* Setup first read */
254 /* XXX: Leave some bufs for swap */
255 bp = getpbuf(&ffsrawbufcnt);
256 error = ffs_rawread_readahead(vp, udata, offset, resid,
261 if (resid > bp->b_bufsize) { /* Setup fist readahead */
262 /* XXX: Leave bufs for swap */
263 if (rawreadahead != 0)
264 nbp = trypbuf(&ffsrawbufcnt);
268 nerror = ffs_rawread_readahead(
270 udata + bp->b_bufsize,
271 offset + bp->b_bufsize,
272 resid - bp->b_bufsize,
275 relpbuf(nbp, &ffsrawbufcnt);
283 while (bp->b_cmd != BUF_CMD_DONE)
284 tsleep((caddr_t)&bp->b_bio2, 0, "rawrd", 0);
289 iolen = bp->b_bcount - bp->b_resid;
290 if (iolen == 0 && (bp->b_flags & B_ERROR) == 0) {
291 nerror = 0; /* Ignore possible beyond EOF error */
295 if ((bp->b_flags & B_ERROR) != 0) {
299 clearbiocache(&bp->b_bio2);
303 if (iolen < bp->b_bufsize) {
304 /* Incomplete read. Try to read remaining part */
305 error = ffs_rawread_readahead(
307 bp->b_bufsize - iolen, bp, &baseticks);
310 } else if (nbp != NULL) { /* Complete read with readahead */
316 clearbiocache(&nbp->b_bio2);
318 if (resid <= bp->b_bufsize) { /* No more readaheads */
319 relpbuf(nbp, &ffsrawbufcnt);
321 } else { /* Setup next readahead */
322 nerror = ffs_rawread_readahead(
323 vp, udata + bp->b_bufsize,
324 offset + bp->b_bufsize,
325 resid - bp->b_bufsize,
328 relpbuf(nbp, &ffsrawbufcnt);
332 } else if (nerror != 0) {/* Deferred Readahead error */
334 } else if (resid > 0) { /* More to read, no readahead */
335 error = ffs_rawread_readahead(vp, udata, offset,
344 relpbuf(bp, &ffsrawbufcnt);
345 if (nbp != NULL) { /* Run down readahead buffer */
347 while (nbp->b_cmd != BUF_CMD_DONE)
348 tsleep(&nbp->b_bio2, 0, "rawrd", 0);
351 relpbuf(nbp, &ffsrawbufcnt);
356 uio->uio_iov->iov_base = udata;
357 uio->uio_resid = resid;
358 uio->uio_offset = offset;
364 ffs_rawread(struct vnode *vp,
368 if (allowrawread != 0 &&
369 uio->uio_iovcnt == 1 &&
370 uio->uio_segflg == UIO_USERSPACE &&
371 uio->uio_resid == uio->uio_iov->iov_len &&
372 (curthread->td_flags & TDF_DEADLKTREAT) == 0) {
373 int secsize; /* Media sector size */
374 off_t filebytes; /* Bytes left of file */
375 int blockbytes; /* Bytes left of file in full blocks */
376 int partialbytes; /* Bytes in last partial block */
377 int skipbytes; /* Bytes not to read in ffs_rawread */
382 /* Only handle sector aligned reads */
384 secsize = ip->i_devvp->v_rdev->si_bsize_phys;
385 if ((uio->uio_offset & (secsize - 1)) == 0 &&
386 (uio->uio_resid & (secsize - 1)) == 0) {
388 /* Sync dirty pages and buffers if needed */
389 error = ffs_rawread_sync(vp);
393 /* Check for end of file */
394 if (ip->i_size > uio->uio_offset) {
395 filebytes = ip->i_size - uio->uio_offset;
397 /* No special eof handling needed ? */
398 if (uio->uio_resid <= filebytes) {
400 return ffs_rawread_main(vp, uio);
403 partialbytes = ((unsigned int) ip->i_size) %
405 blockbytes = (int) filebytes - partialbytes;
406 if (blockbytes > 0) {
407 skipbytes = uio->uio_resid -
409 uio->uio_resid = blockbytes;
410 error = ffs_rawread_main(vp, uio);
411 uio->uio_resid += skipbytes;
414 /* Read remaining part using buffer */
425 ffs_rawreadwakeup(struct bio *bio)
427 bio->bio_buf->b_cmd = BUF_CMD_DONE;