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
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/fcntl.h>
36 #include <sys/limits.h>
37 #include <sys/mount.h>
38 #include <sys/namei.h>
39 #include <sys/vnode.h>
41 #include <sys/filio.h>
42 #include <sys/ttycom.h>
45 #include <sys/rwlock.h>
46 #include <ufs/ufs/extattr.h>
47 #include <ufs/ufs/quota.h>
48 #include <ufs/ufs/inode.h>
49 #include <ufs/ufs/ufsmount.h>
50 #include <ufs/ufs/ufs_extern.h>
51 #include <ufs/ffs/fs.h>
52 #include <ufs/ffs/ffs_extern.h>
55 #include <vm/vm_extern.h>
56 #include <vm/vm_object.h>
57 #include <sys/kernel.h>
58 #include <sys/sysctl.h>
60 static int ffs_rawread_readahead(struct vnode *vp,
66 static int ffs_rawread_main(struct vnode *vp,
69 static int ffs_rawread_sync(struct vnode *vp);
71 int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
73 SYSCTL_DECL(_vfs_ffs);
75 static int ffsrawbufcnt = 4;
76 SYSCTL_INT(_vfs_ffs, OID_AUTO, ffsrawbufcnt, CTLFLAG_RD, &ffsrawbufcnt, 0,
77 "Buffers available for raw reads");
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 ffsrawbufcnt = (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8;
93 SYSINIT(ffs_raw, SI_SUB_VM_CONF, SI_ORDER_ANY, ffs_rawread_setup, NULL);
96 ffs_rawread_sync(struct vnode *vp)
104 /* Check for dirty mmap, pending writes and dirty buffers */
108 if (bo->bo_numoutput > 0 ||
109 bo->bo_dirty.bv_cnt > 0 ||
110 ((obj = vp->v_object) != NULL &&
111 (obj->flags & OBJ_MIGHTBEDIRTY) != 0)) {
115 if (vn_start_write(vp, &mp, V_NOWAIT) != 0) {
116 if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE)
121 (void) vn_start_write(vp, &mp, V_WAIT);
122 VOP_LOCK(vp, LK_EXCLUSIVE);
123 } else if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
125 /* Upgrade to exclusive lock, this might block */
126 VOP_LOCK(vp, LK_UPGRADE);
132 /* Check if vnode was reclaimed while unlocked. */
133 if ((vp->v_iflag & VI_DOOMED) != 0) {
136 VOP_LOCK(vp, LK_DOWNGRADE);
137 vn_finished_write(mp);
140 /* Attempt to msync mmap() regions to clean dirty mmap */
141 if ((obj = vp->v_object) != NULL &&
142 (obj->flags & OBJ_MIGHTBEDIRTY) != 0) {
144 VM_OBJECT_WLOCK(obj);
145 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC);
146 VM_OBJECT_WUNLOCK(obj);
150 /* Wait for pending writes to complete */
152 error = bufobj_wwait(&vp->v_bufobj, 0, 0);
154 /* XXX: can't happen with a zero timeout ??? */
157 VOP_LOCK(vp, LK_DOWNGRADE);
158 vn_finished_write(mp);
161 /* Flush dirty buffers */
162 if (bo->bo_dirty.bv_cnt > 0) {
164 if ((error = ffs_syncvnode(vp, MNT_WAIT, 0)) != 0) {
166 VOP_LOCK(vp, LK_DOWNGRADE);
167 vn_finished_write(mp);
171 if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0)
172 panic("ffs_rawread_sync: dirty bufs");
176 VOP_LOCK(vp, LK_DOWNGRADE);
177 vn_finished_write(mp);
187 ffs_rawread_readahead(struct vnode *vp,
204 bsize = vp->v_mount->mnt_stat.f_iosize;
209 iolen = ((vm_offset_t) udata) & PAGE_MASK;
211 if (bp->b_bcount + iolen > bp->b_kvasize) {
212 bp->b_bcount = bp->b_kvasize;
214 bp->b_bcount -= PAGE_SIZE;
216 bp->b_flags = 0; /* XXX necessary ? */
217 bp->b_iocmd = BIO_READ;
218 bp->b_iodone = bdone;
220 blockno = offset / bsize;
221 blockoff = (offset % bsize) / DEV_BSIZE;
222 if ((daddr_t) blockno != blockno) {
223 return EINVAL; /* blockno overflow */
226 bp->b_lblkno = bp->b_blkno = blockno;
228 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, NULL, &bforwards, NULL);
233 /* Fill holes with NULs to preserve semantics */
235 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize)
236 bp->b_bcount = bsize - blockoff * DEV_BSIZE;
237 bp->b_bufsize = bp->b_bcount;
239 if (vmapbuf(bp, 1) < 0)
243 bzero(bp->b_data, bp->b_bufsize);
245 /* Mark operation completed (similar to bufdone()) */
248 bp->b_flags |= B_DONE;
251 bp->b_blkno = blkno + blockoff;
252 bp->b_offset = bp->b_iooffset = (blkno + blockoff) * DEV_BSIZE;
254 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize * (1 + bforwards))
255 bp->b_bcount = bsize * (1 + bforwards) - blockoff * DEV_BSIZE;
256 bp->b_bufsize = bp->b_bcount;
258 if (vmapbuf(bp, 1) < 0)
261 BO_STRATEGY(&dp->v_bufobj, bp);
267 ffs_rawread_main(struct vnode *vp,
271 struct buf *bp, *nbp, *tbp;
278 td = uio->uio_td ? uio->uio_td : curthread;
279 udata = uio->uio_iov->iov_base;
280 resid = uio->uio_resid;
281 offset = uio->uio_offset;
284 * keep the process from being swapped
296 if (bp == NULL) { /* Setup first read */
297 /* XXX: Leave some bufs for swap */
298 bp = getpbuf(&ffsrawbufcnt);
300 error = ffs_rawread_readahead(vp, udata, offset,
305 if (resid > bp->b_bufsize) { /* Setup fist readahead */
306 /* XXX: Leave bufs for swap */
307 if (rawreadahead != 0)
308 nbp = trypbuf(&ffsrawbufcnt);
314 nerror = ffs_rawread_readahead(vp,
325 relpbuf(nbp, &ffsrawbufcnt);
332 bwait(bp, PRIBIO, "rawrd");
335 iolen = bp->b_bcount - bp->b_resid;
336 if (iolen == 0 && (bp->b_ioflags & BIO_ERROR) == 0) {
337 nerror = 0; /* Ignore possible beyond EOF error */
341 if ((bp->b_ioflags & BIO_ERROR) != 0) {
348 if (iolen < bp->b_bufsize) {
349 /* Incomplete read. Try to read remaining part */
350 error = ffs_rawread_readahead(vp,
353 bp->b_bufsize - iolen,
358 } else if (nbp != NULL) { /* Complete read with readahead */
364 if (resid <= bp->b_bufsize) { /* No more readaheads */
366 relpbuf(nbp, &ffsrawbufcnt);
368 } else { /* Setup next readahead */
369 nerror = ffs_rawread_readahead(vp,
380 relpbuf(nbp, &ffsrawbufcnt);
384 } else if (nerror != 0) {/* Deferred Readahead error */
386 } else if (resid > 0) { /* More to read, no readahead */
387 error = ffs_rawread_readahead(vp, udata, offset,
396 relpbuf(bp, &ffsrawbufcnt);
398 if (nbp != NULL) { /* Run down readahead buffer */
399 bwait(nbp, PRIBIO, "rawrd");
402 relpbuf(nbp, &ffsrawbufcnt);
408 uio->uio_iov->iov_base = udata;
409 uio->uio_resid = resid;
410 uio->uio_offset = offset;
416 ffs_rawread(struct vnode *vp,
420 if (allowrawread != 0 &&
421 uio->uio_iovcnt == 1 &&
422 uio->uio_segflg == UIO_USERSPACE &&
423 uio->uio_resid == uio->uio_iov->iov_len &&
424 (((uio->uio_td != NULL) ? uio->uio_td : curthread)->td_pflags &
425 TDP_DEADLKTREAT) == 0) {
426 int secsize; /* Media sector size */
427 off_t filebytes; /* Bytes left of file */
428 int blockbytes; /* Bytes left of file in full blocks */
429 int partialbytes; /* Bytes in last partial block */
430 int skipbytes; /* Bytes not to read in ffs_rawread */
435 /* Only handle sector aligned reads */
437 secsize = ITODEVVP(ip)->v_bufobj.bo_bsize;
438 if ((uio->uio_offset & (secsize - 1)) == 0 &&
439 (uio->uio_resid & (secsize - 1)) == 0) {
441 /* Sync dirty pages and buffers if needed */
442 error = ffs_rawread_sync(vp);
446 /* Check for end of file */
447 if (ip->i_size > uio->uio_offset) {
448 filebytes = ip->i_size - uio->uio_offset;
450 /* No special eof handling needed ? */
451 if (uio->uio_resid <= filebytes) {
453 return ffs_rawread_main(vp, uio);
456 partialbytes = ((unsigned int) ip->i_size) %
458 blockbytes = (int) filebytes - partialbytes;
459 if (blockbytes > 0) {
460 skipbytes = uio->uio_resid -
462 uio->uio_resid = blockbytes;
463 error = ffs_rawread_main(vp, uio);
464 uio->uio_resid += skipbytes;
467 /* Read remaining part using buffer */