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 $
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/fcntl.h>
35 #include <sys/mount.h>
36 #include <sys/namei.h>
37 #include <sys/vnode.h>
39 #include <sys/filio.h>
40 #include <sys/ttycom.h>
42 #include <ufs/ufs/quota.h>
43 #include <ufs/ufs/inode.h>
44 #include <ufs/ffs/fs.h>
46 #include <machine/limits.h>
48 #include <vm/vm_extern.h>
49 #include <vm/vm_object.h>
50 #include <sys/kernel.h>
51 #include <sys/sysctl.h>
53 static int ffs_rawread_readahead(struct vnode *vp,
60 static int ffs_rawread_main(struct vnode *vp,
63 static int ffs_rawread_sync(struct vnode *vp, struct proc *p);
65 int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
67 void ffs_rawread_setup(void);
69 static void ffs_rawreadwakeup(struct buf *bp);
72 SYSCTL_DECL(_vfs_ffs);
74 static int ffsrawbufcnt = 4;
75 SYSCTL_INT(_vfs_ffs, OID_AUTO, ffsrawbufcnt, CTLFLAG_RD, &ffsrawbufcnt, 0,
76 "Buffers available for raw reads");
78 static int allowrawread = 1;
79 SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0,
80 "Flag to enable raw reads");
82 static int rawreadahead = 1;
83 SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0,
84 "Flag to enable readahead for long raw reads");
88 ffs_rawread_setup(void)
90 ffsrawbufcnt = (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8;
95 ffs_rawread_sync(struct vnode *vp, struct proc *p)
101 /* Check for dirty mmap, pending writes and dirty buffers */
103 if (vp->v_numoutput > 0 ||
104 !TAILQ_EMPTY(&vp->v_dirtyblkhd) ||
105 (vp->v_flag & VOBJDIRTY) != 0) {
108 if (VOP_ISLOCKED(vp, p) != LK_EXCLUSIVE) {
110 /* Upgrade to exclusive lock, this might block */
111 VOP_LOCK(vp, LK_UPGRADE | LK_NOPAUSE, p);
115 /* Attempt to msync mmap() regions to clean dirty mmap */
116 if ((vp->v_flag & VOBJDIRTY) != 0) {
117 struct vm_object *obj;
118 if (VOP_GETVOBJECT(vp, &obj) == 0)
119 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC);
122 /* Wait for pending writes to complete */
124 while (vp->v_numoutput) {
125 vp->v_flag |= VBWAIT;
126 error = tsleep((caddr_t)&vp->v_numoutput,
132 VOP_LOCK(vp, LK_DOWNGRADE, p);
136 /* Flush dirty buffers */
137 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
139 if ((error = VOP_FSYNC(vp, NOCRED, MNT_WAIT, p)) != 0) {
141 VOP_LOCK(vp, LK_DOWNGRADE, p);
145 if (vp->v_numoutput > 0 ||
146 !TAILQ_EMPTY(&vp->v_dirtyblkhd))
147 panic("ffs_rawread_sync: dirty bufs");
151 VOP_LOCK(vp, LK_DOWNGRADE, p);
160 ffs_rawread_readahead(struct vnode *vp,
176 bsize = vp->v_mount->mnt_stat.f_iosize;
178 iolen = ((vm_offset_t) udata) & PAGE_MASK;
180 if (bp->b_bcount + iolen > bp->b_kvasize) {
181 bp->b_bcount = bp->b_kvasize;
183 bp->b_bcount -= PAGE_SIZE;
185 bp->b_flags = B_PHYS | B_CALL | B_READ;
186 bp->b_iodone = ffs_rawreadwakeup;
189 bp->b_offset = offset;
190 blockno = bp->b_offset / bsize;
191 blockoff = (bp->b_offset % bsize) / DEV_BSIZE;
192 if ((daddr_t) blockno != blockno) {
193 return EINVAL; /* blockno overflow */
196 bp->b_lblkno = bp->b_blkno = blockno;
198 error = VOP_BMAP(vp, bp->b_lblkno, &dp, &bp->b_blkno, &bforwards,
203 if (bp->b_blkno == -1) {
205 /* Fill holes with NULs to preserve semantics */
207 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize)
208 bp->b_bcount = bsize - blockoff * DEV_BSIZE;
209 bp->b_bufsize = bp->b_bcount;
214 if (ticks - switchticks >= hogticks)
216 bzero(bp->b_data, bp->b_bufsize);
218 /* Mark operation completed (similar to bufdone()) */
221 bp->b_flags |= B_DONE;
225 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize * (1 + bforwards))
226 bp->b_bcount = bsize * (1 + bforwards) - blockoff * DEV_BSIZE;
227 bp->b_bufsize = bp->b_bcount;
228 bp->b_blkno += blockoff;
229 bp->b_dev = dp->v_rdev;
234 (void) VOP_STRATEGY(dp, bp);
240 ffs_rawread_main(struct vnode *vp,
244 struct buf *bp, *nbp, *tbp;
245 caddr_t sa, nsa, tsa;
253 p = uio->uio_procp ? uio->uio_procp : curproc;
254 udata = uio->uio_iov->iov_base;
255 resid = uio->uio_resid;
256 offset = uio->uio_offset;
259 * keep the process from being swapped
273 if (bp == NULL) { /* Setup first read */
274 /* XXX: Leave some bufs for swap */
275 bp = getpbuf(&ffsrawbufcnt);
278 error = ffs_rawread_readahead(vp, udata, offset,
283 if (resid > bp->b_bufsize) { /* Setup fist readahead */
284 /* XXX: Leave bufs for swap */
285 if (rawreadahead != 0)
286 nbp = trypbuf(&ffsrawbufcnt);
293 nerror = ffs_rawread_readahead(vp,
304 relpbuf(nbp, &ffsrawbufcnt);
312 while ((bp->b_flags & B_DONE) == 0) {
313 tsleep((caddr_t)bp, PRIBIO, "rawrd", 0);
319 iolen = bp->b_bcount - bp->b_resid;
320 if (iolen == 0 && (bp->b_flags & B_ERROR) == 0) {
321 nerror = 0; /* Ignore possible beyond EOF error */
325 if ((bp->b_flags & B_ERROR) != 0) {
332 if (iolen < bp->b_bufsize) {
333 /* Incomplete read. Try to read remaining part */
334 error = ffs_rawread_readahead(vp,
337 bp->b_bufsize - iolen,
343 } else if (nbp != NULL) { /* Complete read with readahead */
353 if (resid <= bp->b_bufsize) { /* No more readaheads */
354 relpbuf(nbp, &ffsrawbufcnt);
356 } else { /* Setup next readahead */
357 nerror = ffs_rawread_readahead(vp,
368 relpbuf(nbp, &ffsrawbufcnt);
372 } else if (nerror != 0) {/* Deferred Readahead error */
374 } else if (resid > 0) { /* More to read, no readahead */
375 error = ffs_rawread_readahead(vp, udata, offset,
383 relpbuf(bp, &ffsrawbufcnt);
384 if (nbp != NULL) { /* Run down readahead buffer */
386 while ((nbp->b_flags & B_DONE) == 0) {
387 tsleep((caddr_t)nbp, PRIBIO, "rawrd", 0);
391 relpbuf(nbp, &ffsrawbufcnt);
397 uio->uio_iov->iov_base = udata;
398 uio->uio_resid = resid;
399 uio->uio_offset = offset;
405 ffs_rawread(struct vnode *vp,
409 if (allowrawread != 0 &&
410 uio->uio_iovcnt == 1 &&
411 uio->uio_segflg == UIO_USERSPACE &&
412 uio->uio_resid == uio->uio_iov->iov_len &&
413 (((uio->uio_procp != NULL) ? uio->uio_procp : curproc)->p_flag &
414 P_DEADLKTREAT) == 0) {
415 int secsize; /* Media sector size */
416 off_t filebytes; /* Bytes left of file */
417 int blockbytes; /* Bytes left of file in full blocks */
418 int partialbytes; /* Bytes in last partial block */
419 int skipbytes; /* Bytes not to read in ffs_rawread */
424 /* Only handle sector aligned reads */
426 secsize = ip->i_devvp->v_rdev->si_bsize_phys;
427 if ((uio->uio_offset & (secsize - 1)) == 0 &&
428 (uio->uio_resid & (secsize - 1)) == 0) {
430 /* Sync dirty pages and buffers if needed */
431 error = ffs_rawread_sync(vp,
432 (uio->uio_procp != NULL) ?
433 uio->uio_procp : curproc);
437 /* Check for end of file */
438 if (ip->i_size > uio->uio_offset) {
439 filebytes = ip->i_size - uio->uio_offset;
441 /* No special eof handling needed ? */
442 if (uio->uio_resid <= filebytes) {
444 return ffs_rawread_main(vp, uio);
447 partialbytes = ((unsigned int) ip->i_size) %
449 blockbytes = (int) filebytes - partialbytes;
450 if (blockbytes > 0) {
451 skipbytes = uio->uio_resid -
453 uio->uio_resid = blockbytes;
454 error = ffs_rawread_main(vp, uio);
455 uio->uio_resid += skipbytes;
458 /* Read remaining part using buffer */
469 ffs_rawreadwakeup(struct buf *bp)
471 wakeup((caddr_t) bp);