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 1.12 2005/08/03 16:36:33 hmp 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 thread *td, struct buf *bp,
56 caddr_t sa, int *baseticks);
57 static int ffs_rawread_main(struct vnode *vp,
60 static int ffs_rawread_sync(struct vnode *vp, struct thread *td);
62 int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
64 void ffs_rawread_setup(void);
66 static void ffs_rawreadwakeup(struct buf *bp);
69 SYSCTL_DECL(_vfs_ffs);
71 static int ffsrawbufcnt = 4;
72 SYSCTL_INT(_vfs_ffs, OID_AUTO, ffsrawbufcnt, CTLFLAG_RD, &ffsrawbufcnt, 0,
73 "Buffers available for raw reads");
75 static int allowrawread = 1;
76 SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0,
77 "Flag to enable raw reads");
79 static int rawreadahead = 1;
80 SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0,
81 "Flag to enable readahead for long raw reads");
85 ffs_rawread_setup(void)
87 ffsrawbufcnt = (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8;
92 ffs_rawread_sync(struct vnode *vp, struct thread *td)
97 /* Check for dirty mmap, pending writes and dirty buffers */
99 if (vp->v_numoutput > 0 ||
100 !RB_EMPTY(&vp->v_rbdirty_tree) ||
101 (vp->v_flag & VOBJDIRTY) != 0) {
104 if (VOP_ISLOCKED(vp, td) != LK_EXCLUSIVE) {
106 /* Upgrade to exclusive lock, this might block */
107 VOP_LOCK(vp, LK_UPGRADE | LK_NOPAUSE, td);
111 /* Attempt to msync mmap() regions to clean dirty mmap */
112 if ((vp->v_flag & VOBJDIRTY) != 0) {
113 struct vm_object *obj;
114 if (VOP_GETVOBJECT(vp, &obj) == 0)
115 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC);
118 /* Wait for pending writes to complete */
120 while (vp->v_numoutput) {
121 vp->v_flag |= VBWAIT;
122 error = tsleep((caddr_t)&vp->v_numoutput, 0,
127 VOP_LOCK(vp, LK_DOWNGRADE, td);
131 /* Flush dirty buffers */
132 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
134 if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) != 0) {
136 VOP_LOCK(vp, LK_DOWNGRADE, td);
140 if (vp->v_numoutput > 0 ||
141 !RB_EMPTY(&vp->v_rbdirty_tree))
142 panic("ffs_rawread_sync: dirty bufs");
146 VOP_LOCK(vp, LK_DOWNGRADE, td);
155 ffs_rawread_readahead(struct vnode *vp, caddr_t udata, off_t offset,
156 size_t len, struct thread *td, struct buf *bp,
157 caddr_t sa, int *baseticks)
167 bsize = vp->v_mount->mnt_stat.f_iosize;
169 iolen = ((vm_offset_t) udata) & PAGE_MASK;
171 if (bp->b_bcount + iolen > bp->b_kvasize) {
172 bp->b_bcount = bp->b_kvasize;
174 bp->b_bcount -= PAGE_SIZE;
176 bp->b_flags = B_PHYS | B_READ;
177 bp->b_iodone = ffs_rawreadwakeup;
180 bp->b_offset = offset;
181 blockno = bp->b_offset / bsize;
182 blockoff = (bp->b_offset % bsize) / DEV_BSIZE;
183 if ((daddr_t) blockno != blockno) {
184 return EINVAL; /* blockno overflow */
187 bp->b_lblkno = bp->b_blkno = blockno;
189 error = VOP_BMAP(vp, bp->b_lblkno, &dp, &bp->b_blkno, &bforwards,
194 if (bp->b_blkno == -1) {
196 /* Fill holes with NULs to preserve semantics */
198 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize)
199 bp->b_bcount = bsize - blockoff * DEV_BSIZE;
200 bp->b_bufsize = bp->b_bcount;
205 if (ticks - *baseticks >= hogticks) {
209 bzero(bp->b_data, bp->b_bufsize);
211 /* Mark operation completed (similar to bufdone()) */
214 bp->b_flags |= B_DONE;
218 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize * (1 + bforwards))
219 bp->b_bcount = bsize * (1 + bforwards) - blockoff * DEV_BSIZE;
220 bp->b_bufsize = bp->b_bcount;
221 bp->b_blkno += blockoff;
222 bp->b_dev = dp->v_rdev;
227 (void) VOP_STRATEGY(dp, bp);
233 ffs_rawread_main(struct vnode *vp, struct uio *uio)
236 struct buf *bp, *nbp, *tbp;
237 caddr_t sa, nsa, tsa;
239 int baseticks = ticks;
245 td = uio->uio_td ? uio->uio_td : curthread;
246 udata = uio->uio_iov->iov_base;
247 resid = uio->uio_resid;
248 offset = uio->uio_offset;
251 * keep the process from being swapped
253 if (uio->uio_td && uio->uio_td->td_proc)
254 PHOLD(uio->uio_td->td_proc);
266 if (bp == NULL) { /* Setup first read */
267 /* XXX: Leave some bufs for swap */
268 bp = getpbuf(&ffsrawbufcnt);
271 error = ffs_rawread_readahead(vp, udata, offset, resid,
272 td, bp, sa, &baseticks);
276 if (resid > bp->b_bufsize) { /* Setup fist readahead */
277 /* XXX: Leave bufs for swap */
278 if (rawreadahead != 0)
279 nbp = trypbuf(&ffsrawbufcnt);
286 nerror = ffs_rawread_readahead(
288 udata + bp->b_bufsize,
289 offset + bp->b_bufsize,
290 resid - bp->b_bufsize,
294 relpbuf(nbp, &ffsrawbufcnt);
302 while ((bp->b_flags & B_DONE) == 0) {
303 tsleep((caddr_t)bp, 0, "rawrd", 0);
309 iolen = bp->b_bcount - bp->b_resid;
310 if (iolen == 0 && (bp->b_flags & B_ERROR) == 0) {
311 nerror = 0; /* Ignore possible beyond EOF error */
315 if ((bp->b_flags & B_ERROR) != 0) {
322 if (iolen < bp->b_bufsize) {
323 /* Incomplete read. Try to read remaining part */
324 error = ffs_rawread_readahead(
326 bp->b_bufsize - iolen, td, bp,
330 } else if (nbp != NULL) { /* Complete read with readahead */
340 if (resid <= bp->b_bufsize) { /* No more readaheads */
341 relpbuf(nbp, &ffsrawbufcnt);
343 } else { /* Setup next readahead */
344 nerror = ffs_rawread_readahead(
345 vp, udata + bp->b_bufsize,
346 offset + bp->b_bufsize,
347 resid - bp->b_bufsize,
348 td, nbp, nsa, &baseticks);
350 relpbuf(nbp, &ffsrawbufcnt);
354 } else if (nerror != 0) {/* Deferred Readahead error */
356 } else if (resid > 0) { /* More to read, no readahead */
357 error = ffs_rawread_readahead(vp, udata, offset,
366 relpbuf(bp, &ffsrawbufcnt);
367 if (nbp != NULL) { /* Run down readahead buffer */
369 while ((nbp->b_flags & B_DONE) == 0) {
370 tsleep((caddr_t)nbp, 0, "rawrd", 0);
374 relpbuf(nbp, &ffsrawbufcnt);
379 if (uio->uio_td && uio->uio_td->td_proc)
380 PRELE(uio->uio_td->td_proc);
381 uio->uio_iov->iov_base = udata;
382 uio->uio_resid = resid;
383 uio->uio_offset = offset;
389 ffs_rawread(struct vnode *vp,
393 if (allowrawread != 0 &&
394 uio->uio_iovcnt == 1 &&
395 uio->uio_segflg == UIO_USERSPACE &&
396 uio->uio_resid == uio->uio_iov->iov_len &&
397 (((uio->uio_td != NULL) ? uio->uio_td : curthread)->td_flags &
398 TDF_DEADLKTREAT) == 0) {
399 int secsize; /* Media sector size */
400 off_t filebytes; /* Bytes left of file */
401 int blockbytes; /* Bytes left of file in full blocks */
402 int partialbytes; /* Bytes in last partial block */
403 int skipbytes; /* Bytes not to read in ffs_rawread */
408 /* Only handle sector aligned reads */
410 secsize = ip->i_devvp->v_rdev->si_bsize_phys;
411 if ((uio->uio_offset & (secsize - 1)) == 0 &&
412 (uio->uio_resid & (secsize - 1)) == 0) {
414 /* Sync dirty pages and buffers if needed */
415 error = ffs_rawread_sync(vp,
416 (uio->uio_td != NULL) ?
417 uio->uio_td : curthread);
421 /* Check for end of file */
422 if (ip->i_size > uio->uio_offset) {
423 filebytes = ip->i_size - uio->uio_offset;
425 /* No special eof handling needed ? */
426 if (uio->uio_resid <= filebytes) {
428 return ffs_rawread_main(vp, uio);
431 partialbytes = ((unsigned int) ip->i_size) %
433 blockbytes = (int) filebytes - partialbytes;
434 if (blockbytes > 0) {
435 skipbytes = uio->uio_resid -
437 uio->uio_resid = blockbytes;
438 error = ffs_rawread_main(vp, uio);
439 uio->uio_resid += skipbytes;
442 /* Read remaining part using buffer */
453 ffs_rawreadwakeup(struct buf *bp)
455 wakeup((caddr_t) bp);