1/*-
2 * Copyright (c) 2000-2003 Tor Egge
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
13 *
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
24 * SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD: stable/10/sys/ufs/ffs/ffs_rawread.c 318267 2017-05-14 12:00:00Z kib $");
29
30#include <sys/param.h>
31#include <sys/systm.h>
32#include <sys/fcntl.h>
33#include <sys/file.h>
34#include <sys/stat.h>
35#include <sys/proc.h>
36#include <sys/limits.h>
37#include <sys/mount.h>
38#include <sys/namei.h>
39#include <sys/vnode.h>
40#include <sys/conf.h>
41#include <sys/filio.h>
42#include <sys/ttycom.h>
43#include <sys/bio.h>
44#include <sys/buf.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>
53
54#include <vm/vm.h>
55#include <vm/vm_extern.h>
56#include <vm/vm_object.h>
57#include <sys/kernel.h>
58#include <sys/sysctl.h>
59
60static int ffs_rawread_readahead(struct vnode *vp,
61				 caddr_t udata,
62				 off_t offset,
63				 size_t len,
64				 struct thread *td,
65				 struct buf *bp,
66				 caddr_t sa);
67static int ffs_rawread_main(struct vnode *vp,
68			    struct uio *uio);
69
70static int ffs_rawread_sync(struct vnode *vp);
71
72int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
73
74SYSCTL_DECL(_vfs_ffs);
75
76static int ffsrawbufcnt = 4;
77SYSCTL_INT(_vfs_ffs, OID_AUTO, ffsrawbufcnt, CTLFLAG_RD, &ffsrawbufcnt, 0,
78	   "Buffers available for raw reads");
79
80static int allowrawread = 1;
81SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0,
82	   "Flag to enable raw reads");
83
84static int rawreadahead = 1;
85SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0,
86	   "Flag to enable readahead for long raw reads");
87
88static void
89ffs_rawread_setup(void *arg __unused)
90{
91
92	ffsrawbufcnt = (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8;
93}
94SYSINIT(ffs_raw, SI_SUB_VM_CONF, SI_ORDER_ANY, ffs_rawread_setup, NULL);
95
96static int
97ffs_rawread_sync(struct vnode *vp)
98{
99	int error;
100	int upgraded;
101	struct bufobj *bo;
102	struct mount *mp;
103	vm_object_t obj;
104
105	/* Check for dirty mmap, pending writes and dirty buffers */
106	bo = &vp->v_bufobj;
107	BO_LOCK(bo);
108	VI_LOCK(vp);
109	if (bo->bo_numoutput > 0 ||
110	    bo->bo_dirty.bv_cnt > 0 ||
111	    ((obj = vp->v_object) != NULL &&
112	     (obj->flags & OBJ_MIGHTBEDIRTY) != 0)) {
113		VI_UNLOCK(vp);
114		BO_UNLOCK(bo);
115
116		if (vn_start_write(vp, &mp, V_NOWAIT) != 0) {
117			if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE)
118				upgraded = 1;
119			else
120				upgraded = 0;
121			VOP_UNLOCK(vp, 0);
122			(void) vn_start_write(vp, &mp, V_WAIT);
123			VOP_LOCK(vp, LK_EXCLUSIVE);
124		} else if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
125			upgraded = 1;
126			/* Upgrade to exclusive lock, this might block */
127			VOP_LOCK(vp, LK_UPGRADE);
128		} else
129			upgraded = 0;
130
131
132		VI_LOCK(vp);
133		/* Check if vnode was reclaimed while unlocked. */
134		if ((vp->v_iflag & VI_DOOMED) != 0) {
135			VI_UNLOCK(vp);
136			if (upgraded != 0)
137				VOP_LOCK(vp, LK_DOWNGRADE);
138			vn_finished_write(mp);
139			return (EIO);
140		}
141		/* Attempt to msync mmap() regions to clean dirty mmap */
142		if ((obj = vp->v_object) != NULL &&
143		    (obj->flags & OBJ_MIGHTBEDIRTY) != 0) {
144			VI_UNLOCK(vp);
145			VM_OBJECT_WLOCK(obj);
146			vm_object_page_clean(obj, 0, 0, OBJPC_SYNC);
147			VM_OBJECT_WUNLOCK(obj);
148		} else
149			VI_UNLOCK(vp);
150
151		/* Wait for pending writes to complete */
152		BO_LOCK(bo);
153		error = bufobj_wwait(&vp->v_bufobj, 0, 0);
154		if (error != 0) {
155			/* XXX: can't happen with a zero timeout ??? */
156			BO_UNLOCK(bo);
157			if (upgraded != 0)
158				VOP_LOCK(vp, LK_DOWNGRADE);
159			vn_finished_write(mp);
160			return (error);
161		}
162		/* Flush dirty buffers */
163		if (bo->bo_dirty.bv_cnt > 0) {
164			BO_UNLOCK(bo);
165			if ((error = ffs_syncvnode(vp, MNT_WAIT, 0)) != 0) {
166				if (upgraded != 0)
167					VOP_LOCK(vp, LK_DOWNGRADE);
168				vn_finished_write(mp);
169				return (error);
170			}
171			BO_LOCK(bo);
172			if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0)
173				panic("ffs_rawread_sync: dirty bufs");
174		}
175		BO_UNLOCK(bo);
176		if (upgraded != 0)
177			VOP_LOCK(vp, LK_DOWNGRADE);
178		vn_finished_write(mp);
179	} else {
180		VI_UNLOCK(vp);
181		BO_UNLOCK(bo);
182	}
183	return 0;
184}
185
186
187static int
188ffs_rawread_readahead(struct vnode *vp,
189		      caddr_t udata,
190		      off_t offset,
191		      size_t len,
192		      struct thread *td,
193		      struct buf *bp,
194		      caddr_t sa)
195{
196	int error;
197	u_int iolen;
198	off_t blockno;
199	int blockoff;
200	int bsize;
201	struct vnode *dp;
202	int bforwards;
203	struct inode *ip;
204	ufs2_daddr_t blkno;
205
206	bsize = vp->v_mount->mnt_stat.f_iosize;
207
208	ip = VTOI(vp);
209	dp = ip->i_devvp;
210
211	iolen = ((vm_offset_t) udata) & PAGE_MASK;
212	bp->b_bcount = len;
213	if (bp->b_bcount + iolen > bp->b_kvasize) {
214		bp->b_bcount = bp->b_kvasize;
215		if (iolen != 0)
216			bp->b_bcount -= PAGE_SIZE;
217	}
218	bp->b_flags = 0;	/* XXX necessary ? */
219	bp->b_iocmd = BIO_READ;
220	bp->b_iodone = bdone;
221	bp->b_data = udata;
222	bp->b_saveaddr = sa;
223	blockno = offset / bsize;
224	blockoff = (offset % bsize) / DEV_BSIZE;
225	if ((daddr_t) blockno != blockno) {
226		return EINVAL; /* blockno overflow */
227	}
228
229	bp->b_lblkno = bp->b_blkno = blockno;
230
231	error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, NULL, &bforwards, NULL);
232	if (error != 0)
233		return error;
234	if (blkno == -1) {
235
236		/* Fill holes with NULs to preserve semantics */
237
238		if (bp->b_bcount + blockoff * DEV_BSIZE > bsize)
239			bp->b_bcount = bsize - blockoff * DEV_BSIZE;
240		bp->b_bufsize = bp->b_bcount;
241
242		if (vmapbuf(bp, 1) < 0)
243			return EFAULT;
244
245		maybe_yield();
246		bzero(bp->b_data, bp->b_bufsize);
247
248		/* Mark operation completed (similar to bufdone()) */
249
250		bp->b_resid = 0;
251		bp->b_flags |= B_DONE;
252		return 0;
253	}
254	bp->b_blkno = blkno + blockoff;
255	bp->b_offset = bp->b_iooffset = (blkno + blockoff) * DEV_BSIZE;
256
257	if (bp->b_bcount + blockoff * DEV_BSIZE > bsize * (1 + bforwards))
258		bp->b_bcount = bsize * (1 + bforwards) - blockoff * DEV_BSIZE;
259	bp->b_bufsize = bp->b_bcount;
260
261	if (vmapbuf(bp, 1) < 0)
262		return EFAULT;
263
264	BO_STRATEGY(&dp->v_bufobj, bp);
265	return 0;
266}
267
268
269static int
270ffs_rawread_main(struct vnode *vp,
271		 struct uio *uio)
272{
273	int error, nerror;
274	struct buf *bp, *nbp, *tbp;
275	caddr_t sa, nsa, tsa;
276	u_int iolen;
277	caddr_t udata;
278	long resid;
279	off_t offset;
280	struct thread *td;
281
282	td = uio->uio_td ? uio->uio_td : curthread;
283	udata = uio->uio_iov->iov_base;
284	resid = uio->uio_resid;
285	offset = uio->uio_offset;
286
287	/*
288	 * keep the process from being swapped
289	 */
290	PHOLD(td->td_proc);
291
292	error = 0;
293	nerror = 0;
294
295	bp = NULL;
296	nbp = NULL;
297	sa = NULL;
298	nsa = NULL;
299
300	while (resid > 0) {
301
302		if (bp == NULL) { /* Setup first read */
303			/* XXX: Leave some bufs for swap */
304			bp = getpbuf(&ffsrawbufcnt);
305			sa = bp->b_data;
306			pbgetvp(vp, bp);
307			error = ffs_rawread_readahead(vp, udata, offset,
308						     resid, td, bp, sa);
309			if (error != 0)
310				break;
311
312			if (resid > bp->b_bufsize) { /* Setup fist readahead */
313				/* XXX: Leave bufs for swap */
314				if (rawreadahead != 0)
315					nbp = trypbuf(&ffsrawbufcnt);
316				else
317					nbp = NULL;
318				if (nbp != NULL) {
319					nsa = nbp->b_data;
320					pbgetvp(vp, nbp);
321
322					nerror = ffs_rawread_readahead(vp,
323								       udata +
324								       bp->b_bufsize,
325								       offset +
326								       bp->b_bufsize,
327								       resid -
328								       bp->b_bufsize,
329								       td,
330								       nbp,
331								       nsa);
332					if (nerror) {
333						pbrelvp(nbp);
334						relpbuf(nbp, &ffsrawbufcnt);
335						nbp = NULL;
336					}
337				}
338			}
339		}
340
341		bwait(bp, PRIBIO, "rawrd");
342		vunmapbuf(bp);
343
344		iolen = bp->b_bcount - bp->b_resid;
345		if (iolen == 0 && (bp->b_ioflags & BIO_ERROR) == 0) {
346			nerror = 0;	/* Ignore possible beyond EOF error */
347			break; /* EOF */
348		}
349
350		if ((bp->b_ioflags & BIO_ERROR) != 0) {
351			error = bp->b_error;
352			break;
353		}
354		resid -= iolen;
355		udata += iolen;
356		offset += iolen;
357		if (iolen < bp->b_bufsize) {
358			/* Incomplete read.  Try to read remaining part */
359			error = ffs_rawread_readahead(vp,
360						      udata,
361						      offset,
362						      bp->b_bufsize - iolen,
363						      td,
364						      bp,
365						      sa);
366			if (error != 0)
367				break;
368		} else if (nbp != NULL) { /* Complete read with readahead */
369
370			tbp = bp;
371			bp = nbp;
372			nbp = tbp;
373
374			tsa = sa;
375			sa = nsa;
376			nsa = tsa;
377
378			if (resid <= bp->b_bufsize) { /* No more readaheads */
379				pbrelvp(nbp);
380				relpbuf(nbp, &ffsrawbufcnt);
381				nbp = NULL;
382			} else { /* Setup next readahead */
383				nerror = ffs_rawread_readahead(vp,
384							       udata +
385							       bp->b_bufsize,
386							       offset +
387							       bp->b_bufsize,
388							       resid -
389							       bp->b_bufsize,
390							       td,
391							       nbp,
392							       nsa);
393				if (nerror != 0) {
394					pbrelvp(nbp);
395					relpbuf(nbp, &ffsrawbufcnt);
396					nbp = NULL;
397				}
398			}
399		} else if (nerror != 0) {/* Deferred Readahead error */
400			break;
401		}  else if (resid > 0) { /* More to read, no readahead */
402			error = ffs_rawread_readahead(vp, udata, offset,
403						      resid, td, bp, sa);
404			if (error != 0)
405				break;
406		}
407	}
408
409	if (bp != NULL) {
410		pbrelvp(bp);
411		relpbuf(bp, &ffsrawbufcnt);
412	}
413	if (nbp != NULL) {			/* Run down readahead buffer */
414		bwait(nbp, PRIBIO, "rawrd");
415		vunmapbuf(nbp);
416		pbrelvp(nbp);
417		relpbuf(nbp, &ffsrawbufcnt);
418	}
419
420	if (error == 0)
421		error = nerror;
422	PRELE(td->td_proc);
423	uio->uio_iov->iov_base = udata;
424	uio->uio_resid = resid;
425	uio->uio_offset = offset;
426	return error;
427}
428
429
430int
431ffs_rawread(struct vnode *vp,
432	    struct uio *uio,
433	    int *workdone)
434{
435	if (allowrawread != 0 &&
436	    uio->uio_iovcnt == 1 &&
437	    uio->uio_segflg == UIO_USERSPACE &&
438	    uio->uio_resid == uio->uio_iov->iov_len &&
439	    (((uio->uio_td != NULL) ? uio->uio_td : curthread)->td_pflags &
440	     TDP_DEADLKTREAT) == 0) {
441		int secsize;		/* Media sector size */
442		off_t filebytes;	/* Bytes left of file */
443		int blockbytes;		/* Bytes left of file in full blocks */
444		int partialbytes;	/* Bytes in last partial block */
445		int skipbytes;		/* Bytes not to read in ffs_rawread */
446		struct inode *ip;
447		int error;
448
449
450		/* Only handle sector aligned reads */
451		ip = VTOI(vp);
452		secsize = ip->i_devvp->v_bufobj.bo_bsize;
453		if ((uio->uio_offset & (secsize - 1)) == 0 &&
454		    (uio->uio_resid & (secsize - 1)) == 0) {
455
456			/* Sync dirty pages and buffers if needed */
457			error = ffs_rawread_sync(vp);
458			if (error != 0)
459				return error;
460
461			/* Check for end of file */
462			if (ip->i_size > uio->uio_offset) {
463				filebytes = ip->i_size - uio->uio_offset;
464
465				/* No special eof handling needed ? */
466				if (uio->uio_resid <= filebytes) {
467					*workdone = 1;
468					return ffs_rawread_main(vp, uio);
469				}
470
471				partialbytes = ((unsigned int) ip->i_size) %
472					ip->i_fs->fs_bsize;
473				blockbytes = (int) filebytes - partialbytes;
474				if (blockbytes > 0) {
475					skipbytes = uio->uio_resid -
476						blockbytes;
477					uio->uio_resid = blockbytes;
478					error = ffs_rawread_main(vp, uio);
479					uio->uio_resid += skipbytes;
480					if (error != 0)
481						return error;
482					/* Read remaining part using buffer */
483				}
484			}
485		}
486	}
487	*workdone = 0;
488	return 0;
489}
490