1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22/* ONC_PLUS EXTRACT START */
23/*
24 * Copyright (c) 1994, 2010, Oracle and/or its affiliates. All rights reserved.
25 */
26
27/*	Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T	*/
28/*	  All Rights Reserved  	*/
29
30/*
31 * Portions of this source code were derived from Berkeley 4.3 BSD
32 * under license from the Regents of the University of California.
33 */
34
35/* ONC_PLUS EXTRACT END */
36
37#include <sys/param.h>
38#include <sys/isa_defs.h>
39#include <sys/types.h>
40#include <sys/sysmacros.h>
41#include <sys/systm.h>
42#include <sys/errno.h>
43#include <sys/fcntl.h>
44/* ONC_PLUS EXTRACT START */
45#include <sys/flock.h>
46/* ONC_PLUS EXTRACT END */
47#include <sys/vnode.h>
48#include <sys/file.h>
49#include <sys/mode.h>
50#include <sys/proc.h>
51#include <sys/filio.h>
52#include <sys/share.h>
53#include <sys/debug.h>
54#include <sys/rctl.h>
55#include <sys/nbmlock.h>
56
57#include <sys/cmn_err.h>
58
59/* ONC_PLUS EXTRACT START */
60static int flock_check(vnode_t *, flock64_t *, offset_t, offset_t);
61static int flock_get_start(vnode_t *, flock64_t *, offset_t, u_offset_t *);
62static void fd_too_big(proc_t *);
63
64/*
65 * File control.
66 */
67int
68fcntl(int fdes, int cmd, intptr_t arg)
69{
70	int iarg;
71	int error = 0;
72	int retval;
73	proc_t *p;
74	file_t *fp;
75	vnode_t *vp;
76	u_offset_t offset;
77	u_offset_t start;
78	struct vattr vattr;
79	int in_crit;
80	int flag;
81	struct flock sbf;
82	struct flock64 bf;
83	struct o_flock obf;
84	struct flock64_32 bf64_32;
85	struct fshare fsh;
86	struct shrlock shr;
87	struct shr_locowner shr_own;
88	offset_t maxoffset;
89	model_t datamodel;
90	int fdres;
91
92#if defined(_ILP32) && !defined(lint) && defined(_SYSCALL32)
93	ASSERT(sizeof (struct flock) == sizeof (struct flock32));
94	ASSERT(sizeof (struct flock64) == sizeof (struct flock64_32));
95#endif
96#if defined(_LP64) && !defined(lint) && defined(_SYSCALL32)
97	ASSERT(sizeof (struct flock) == sizeof (struct flock64_64));
98	ASSERT(sizeof (struct flock64) == sizeof (struct flock64_64));
99#endif
100
101	/*
102	 * First, for speed, deal with the subset of cases
103	 * that do not require getf() / releasef().
104	 */
105	switch (cmd) {
106	case F_GETFD:
107		if ((error = f_getfd_error(fdes, &flag)) == 0)
108			retval = flag;
109		goto out;
110
111	case F_SETFD:
112		error = f_setfd_error(fdes, (int)arg);
113		retval = 0;
114		goto out;
115
116	case F_GETFL:
117		if ((error = f_getfl(fdes, &flag)) == 0) {
118			retval = (flag & (FMASK | FASYNC));
119			if ((flag & (FSEARCH | FEXEC)) == 0)
120				retval += FOPEN;
121			else
122				retval |= (flag & (FSEARCH | FEXEC));
123		}
124		goto out;
125
126	case F_GETXFL:
127		if ((error = f_getfl(fdes, &flag)) == 0) {
128			retval = flag;
129			if ((flag & (FSEARCH | FEXEC)) == 0)
130				retval += FOPEN;
131		}
132		goto out;
133
134	case F_BADFD:
135		if ((error = f_badfd(fdes, &fdres, (int)arg)) == 0)
136			retval = fdres;
137		goto out;
138	}
139
140	/*
141	 * Second, for speed, deal with the subset of cases that
142	 * require getf() / releasef() but do not require copyin.
143	 */
144	if ((fp = getf(fdes)) == NULL) {
145		error = EBADF;
146		goto out;
147	}
148	iarg = (int)arg;
149
150	switch (cmd) {
151/* ONC_PLUS EXTRACT END */
152
153	case F_DUPFD:
154		p = curproc;
155		if ((uint_t)iarg >= p->p_fno_ctl) {
156			if (iarg >= 0)
157				fd_too_big(p);
158			error = EINVAL;
159			goto done;
160		}
161		/*
162		 * We need to increment the f_count reference counter
163		 * before allocating a new file descriptor.
164		 * Doing it other way round opens a window for race condition
165		 * with closeandsetf() on the target file descriptor which can
166		 * close the file still referenced by the original
167		 * file descriptor.
168		 */
169		mutex_enter(&fp->f_tlock);
170		fp->f_count++;
171		mutex_exit(&fp->f_tlock);
172		if ((retval = ufalloc_file(iarg, fp)) == -1) {
173			/*
174			 * New file descriptor can't be allocated.
175			 * Revert the reference count.
176			 */
177			mutex_enter(&fp->f_tlock);
178			fp->f_count--;
179			mutex_exit(&fp->f_tlock);
180			error = EMFILE;
181		}
182		goto done;
183
184	case F_DUP2FD:
185		p = curproc;
186		if (fdes == iarg) {
187			retval = iarg;
188		} else if ((uint_t)iarg >= p->p_fno_ctl) {
189			if (iarg >= 0)
190				fd_too_big(p);
191			error = EBADF;
192		} else {
193			/*
194			 * We can't hold our getf(fdes) across the call to
195			 * closeandsetf() because it creates a window for
196			 * deadlock: if one thread is doing dup2(a, b) while
197			 * another is doing dup2(b, a), each one will block
198			 * waiting for the other to call releasef().  The
199			 * solution is to increment the file reference count
200			 * (which we have to do anyway), then releasef(fdes),
201			 * then closeandsetf().  Incrementing f_count ensures
202			 * that fp won't disappear after we call releasef().
203			 * When closeandsetf() fails, we try avoid calling
204			 * closef() because of all the side effects.
205			 */
206			mutex_enter(&fp->f_tlock);
207			fp->f_count++;
208			mutex_exit(&fp->f_tlock);
209			releasef(fdes);
210			if ((error = closeandsetf(iarg, fp)) == 0) {
211				retval = iarg;
212			} else {
213				mutex_enter(&fp->f_tlock);
214				if (fp->f_count > 1) {
215					fp->f_count--;
216					mutex_exit(&fp->f_tlock);
217				} else {
218					mutex_exit(&fp->f_tlock);
219					(void) closef(fp);
220				}
221			}
222			goto out;
223		}
224		goto done;
225
226	case F_SETFL:
227		vp = fp->f_vnode;
228		flag = fp->f_flag;
229		if ((iarg & (FNONBLOCK|FNDELAY)) == (FNONBLOCK|FNDELAY))
230			iarg &= ~FNDELAY;
231		if ((error = VOP_SETFL(vp, flag, iarg, fp->f_cred, NULL)) ==
232		    0) {
233			iarg &= FMASK;
234			mutex_enter(&fp->f_tlock);
235			fp->f_flag &= ~FMASK | (FREAD|FWRITE);
236			fp->f_flag |= (iarg - FOPEN) & ~(FREAD|FWRITE);
237			mutex_exit(&fp->f_tlock);
238		}
239		retval = 0;
240		goto done;
241	}
242
243	/*
244	 * Finally, deal with the expensive cases.
245	 */
246	retval = 0;
247	in_crit = 0;
248	maxoffset = MAXOFF_T;
249	datamodel = DATAMODEL_NATIVE;
250#if defined(_SYSCALL32_IMPL)
251	if ((datamodel = get_udatamodel()) == DATAMODEL_ILP32)
252		maxoffset = MAXOFF32_T;
253#endif
254
255	vp = fp->f_vnode;
256	flag = fp->f_flag;
257	offset = fp->f_offset;
258
259	switch (cmd) {
260/* ONC_PLUS EXTRACT START */
261	/*
262	 * The file system and vnode layers understand and implement
263	 * locking with flock64 structures. So here once we pass through
264	 * the test for compatibility as defined by LFS API, (for F_SETLK,
265	 * F_SETLKW, F_GETLK, F_GETLKW, F_FREESP) we transform
266	 * the flock structure to a flock64 structure and send it to the
267	 * lower layers. Similarly in case of GETLK the returned flock64
268	 * structure is transformed to a flock structure if everything fits
269	 * in nicely, otherwise we return EOVERFLOW.
270	 */
271
272	case F_GETLK:
273	case F_O_GETLK:
274	case F_SETLK:
275	case F_SETLKW:
276	case F_SETLK_NBMAND:
277
278		/*
279		 * Copy in input fields only.
280		 */
281
282		if (cmd == F_O_GETLK) {
283			if (datamodel != DATAMODEL_ILP32) {
284				error = EINVAL;
285				break;
286			}
287
288			if (copyin((void *)arg, &obf, sizeof (obf))) {
289				error = EFAULT;
290				break;
291			}
292			bf.l_type = obf.l_type;
293			bf.l_whence = obf.l_whence;
294			bf.l_start = (off64_t)obf.l_start;
295			bf.l_len = (off64_t)obf.l_len;
296			bf.l_sysid = (int)obf.l_sysid;
297			bf.l_pid = obf.l_pid;
298		} else if (datamodel == DATAMODEL_NATIVE) {
299			if (copyin((void *)arg, &sbf, sizeof (sbf))) {
300				error = EFAULT;
301				break;
302			}
303			/*
304			 * XXX	In an LP64 kernel with an LP64 application
305			 *	there's no need to do a structure copy here
306			 *	struct flock == struct flock64. However,
307			 *	we did it this way to avoid more conditional
308			 *	compilation.
309			 */
310			bf.l_type = sbf.l_type;
311			bf.l_whence = sbf.l_whence;
312			bf.l_start = (off64_t)sbf.l_start;
313			bf.l_len = (off64_t)sbf.l_len;
314			bf.l_sysid = sbf.l_sysid;
315			bf.l_pid = sbf.l_pid;
316		}
317#if defined(_SYSCALL32_IMPL)
318		else {
319			struct flock32 sbf32;
320			if (copyin((void *)arg, &sbf32, sizeof (sbf32))) {
321				error = EFAULT;
322				break;
323			}
324			bf.l_type = sbf32.l_type;
325			bf.l_whence = sbf32.l_whence;
326			bf.l_start = (off64_t)sbf32.l_start;
327			bf.l_len = (off64_t)sbf32.l_len;
328			bf.l_sysid = sbf32.l_sysid;
329			bf.l_pid = sbf32.l_pid;
330		}
331#endif /* _SYSCALL32_IMPL */
332
333		/*
334		 * 64-bit support: check for overflow for 32-bit lock ops
335		 */
336		if ((error = flock_check(vp, &bf, offset, maxoffset)) != 0)
337			break;
338
339		/*
340		 * Not all of the filesystems understand F_O_GETLK, and
341		 * there's no need for them to know.  Map it to F_GETLK.
342		 */
343		if ((error = VOP_FRLOCK(vp, (cmd == F_O_GETLK) ? F_GETLK : cmd,
344		    &bf, flag, offset, NULL, fp->f_cred, NULL)) != 0)
345			break;
346
347		/*
348		 * If command is GETLK and no lock is found, only
349		 * the type field is changed.
350		 */
351		if ((cmd == F_O_GETLK || cmd == F_GETLK) &&
352		    bf.l_type == F_UNLCK) {
353			/* l_type always first entry, always a short */
354			if (copyout(&bf.l_type, &((struct flock *)arg)->l_type,
355			    sizeof (bf.l_type)))
356				error = EFAULT;
357			break;
358		}
359
360		if (cmd == F_O_GETLK) {
361			/*
362			 * Return an SVR3 flock structure to the user.
363			 */
364			obf.l_type = (int16_t)bf.l_type;
365			obf.l_whence = (int16_t)bf.l_whence;
366			obf.l_start = (int32_t)bf.l_start;
367			obf.l_len = (int32_t)bf.l_len;
368			if (bf.l_sysid > SHRT_MAX || bf.l_pid > SHRT_MAX) {
369				/*
370				 * One or both values for the above fields
371				 * is too large to store in an SVR3 flock
372				 * structure.
373				 */
374				error = EOVERFLOW;
375				break;
376			}
377			obf.l_sysid = (int16_t)bf.l_sysid;
378			obf.l_pid = (int16_t)bf.l_pid;
379			if (copyout(&obf, (void *)arg, sizeof (obf)))
380				error = EFAULT;
381		} else if (cmd == F_GETLK) {
382			/*
383			 * Copy out SVR4 flock.
384			 */
385			int i;
386
387			if (bf.l_start > maxoffset || bf.l_len > maxoffset) {
388				error = EOVERFLOW;
389				break;
390			}
391
392			if (datamodel == DATAMODEL_NATIVE) {
393				for (i = 0; i < 4; i++)
394					sbf.l_pad[i] = 0;
395				/*
396				 * XXX	In an LP64 kernel with an LP64
397				 *	application there's no need to do a
398				 *	structure copy here as currently
399				 *	struct flock == struct flock64.
400				 *	We did it this way to avoid more
401				 *	conditional compilation.
402				 */
403				sbf.l_type = bf.l_type;
404				sbf.l_whence = bf.l_whence;
405				sbf.l_start = (off_t)bf.l_start;
406				sbf.l_len = (off_t)bf.l_len;
407				sbf.l_sysid = bf.l_sysid;
408				sbf.l_pid = bf.l_pid;
409				if (copyout(&sbf, (void *)arg, sizeof (sbf)))
410					error = EFAULT;
411			}
412#if defined(_SYSCALL32_IMPL)
413			else {
414				struct flock32 sbf32;
415				if (bf.l_start > MAXOFF32_T ||
416				    bf.l_len > MAXOFF32_T) {
417					error = EOVERFLOW;
418					break;
419				}
420				for (i = 0; i < 4; i++)
421					sbf32.l_pad[i] = 0;
422				sbf32.l_type = (int16_t)bf.l_type;
423				sbf32.l_whence = (int16_t)bf.l_whence;
424				sbf32.l_start = (off32_t)bf.l_start;
425				sbf32.l_len = (off32_t)bf.l_len;
426				sbf32.l_sysid = (int32_t)bf.l_sysid;
427				sbf32.l_pid = (pid32_t)bf.l_pid;
428				if (copyout(&sbf32,
429				    (void *)arg, sizeof (sbf32)))
430					error = EFAULT;
431			}
432#endif
433		}
434		break;
435/* ONC_PLUS EXTRACT END */
436
437	case F_CHKFL:
438		/*
439		 * This is for internal use only, to allow the vnode layer
440		 * to validate a flags setting before applying it.  User
441		 * programs can't issue it.
442		 */
443		error = EINVAL;
444		break;
445
446	case F_ALLOCSP:
447	case F_FREESP:
448	case F_ALLOCSP64:
449	case F_FREESP64:
450		/*
451		 * Test for not-a-regular-file (and returning EINVAL)
452		 * before testing for open-for-writing (and returning EBADF).
453		 * This is relied upon by posix_fallocate() in libc.
454		 */
455		if (vp->v_type != VREG) {
456			error = EINVAL;
457			break;
458		}
459
460		if ((flag & FWRITE) == 0) {
461			error = EBADF;
462			break;
463		}
464
465		if (datamodel != DATAMODEL_ILP32 &&
466		    (cmd == F_ALLOCSP64 || cmd == F_FREESP64)) {
467			error = EINVAL;
468			break;
469		}
470
471#if defined(_ILP32) || defined(_SYSCALL32_IMPL)
472		if (datamodel == DATAMODEL_ILP32 &&
473		    (cmd == F_ALLOCSP || cmd == F_FREESP)) {
474			struct flock32 sbf32;
475			/*
476			 * For compatibility we overlay an SVR3 flock on an SVR4
477			 * flock.  This works because the input field offsets
478			 * in "struct flock" were preserved.
479			 */
480			if (copyin((void *)arg, &sbf32, sizeof (sbf32))) {
481				error = EFAULT;
482				break;
483			} else {
484				bf.l_type = sbf32.l_type;
485				bf.l_whence = sbf32.l_whence;
486				bf.l_start = (off64_t)sbf32.l_start;
487				bf.l_len = (off64_t)sbf32.l_len;
488				bf.l_sysid = sbf32.l_sysid;
489				bf.l_pid = sbf32.l_pid;
490			}
491		}
492#endif /* _ILP32 || _SYSCALL32_IMPL */
493
494#if defined(_LP64)
495		if (datamodel == DATAMODEL_LP64 &&
496		    (cmd == F_ALLOCSP || cmd == F_FREESP)) {
497			if (copyin((void *)arg, &bf, sizeof (bf))) {
498				error = EFAULT;
499				break;
500			}
501		}
502#endif /* defined(_LP64) */
503
504#if !defined(_LP64) || defined(_SYSCALL32_IMPL)
505		if (datamodel == DATAMODEL_ILP32 &&
506		    (cmd == F_ALLOCSP64 || cmd == F_FREESP64)) {
507			if (copyin((void *)arg, &bf64_32, sizeof (bf64_32))) {
508				error = EFAULT;
509				break;
510			} else {
511				/*
512				 * Note that the size of flock64 is different in
513				 * the ILP32 and LP64 models, due to the l_pad
514				 * field. We do not want to assume that the
515				 * flock64 structure is laid out the same in
516				 * ILP32 and LP64 environments, so we will
517				 * copy in the ILP32 version of flock64
518				 * explicitly and copy it to the native
519				 * flock64 structure.
520				 */
521				bf.l_type = (short)bf64_32.l_type;
522				bf.l_whence = (short)bf64_32.l_whence;
523				bf.l_start = bf64_32.l_start;
524				bf.l_len = bf64_32.l_len;
525				bf.l_sysid = (int)bf64_32.l_sysid;
526				bf.l_pid = (pid_t)bf64_32.l_pid;
527			}
528		}
529#endif /* !defined(_LP64) || defined(_SYSCALL32_IMPL) */
530
531		if (cmd == F_ALLOCSP || cmd == F_FREESP)
532			error = flock_check(vp, &bf, offset, maxoffset);
533		else if (cmd == F_ALLOCSP64 || cmd == F_FREESP64)
534			error = flock_check(vp, &bf, offset, MAXOFFSET_T);
535		if (error)
536			break;
537
538		if (vp->v_type == VREG && bf.l_len == 0 &&
539		    bf.l_start > OFFSET_MAX(fp)) {
540			error = EFBIG;
541			break;
542		}
543
544		/*
545		 * Make sure that there are no conflicting non-blocking
546		 * mandatory locks in the region being manipulated. If
547		 * there are such locks then return EACCES.
548		 */
549		if ((error = flock_get_start(vp, &bf, offset, &start)) != 0)
550			break;
551
552		if (nbl_need_check(vp)) {
553			u_offset_t	begin;
554			ssize_t		length;
555
556			nbl_start_crit(vp, RW_READER);
557			in_crit = 1;
558			vattr.va_mask = AT_SIZE;
559			if ((error = VOP_GETATTR(vp, &vattr, 0, CRED(), NULL))
560			    != 0)
561				break;
562			begin = start > vattr.va_size ? vattr.va_size : start;
563			length = vattr.va_size > start ? vattr.va_size - start :
564			    start - vattr.va_size;
565			if (nbl_conflict(vp, NBL_WRITE, begin, length, 0,
566			    NULL)) {
567				error = EACCES;
568				break;
569			}
570		}
571
572		if (cmd == F_ALLOCSP64)
573			cmd = F_ALLOCSP;
574		else if (cmd == F_FREESP64)
575			cmd = F_FREESP;
576
577		error = VOP_SPACE(vp, cmd, &bf, flag, offset, fp->f_cred, NULL);
578
579		break;
580
581#if !defined(_LP64) || defined(_SYSCALL32_IMPL)
582/* ONC_PLUS EXTRACT START */
583	case F_GETLK64:
584	case F_SETLK64:
585	case F_SETLKW64:
586	case F_SETLK64_NBMAND:
587		/*
588		 * Large Files: Here we set cmd as *LK and send it to
589		 * lower layers. *LK64 is only for the user land.
590		 * Most of the comments described above for F_SETLK
591		 * applies here too.
592		 * Large File support is only needed for ILP32 apps!
593		 */
594		if (datamodel != DATAMODEL_ILP32) {
595			error = EINVAL;
596			break;
597		}
598
599		if (cmd == F_GETLK64)
600			cmd = F_GETLK;
601		else if (cmd == F_SETLK64)
602			cmd = F_SETLK;
603		else if (cmd == F_SETLKW64)
604			cmd = F_SETLKW;
605		else if (cmd == F_SETLK64_NBMAND)
606			cmd = F_SETLK_NBMAND;
607
608		/*
609		 * Note that the size of flock64 is different in the ILP32
610		 * and LP64 models, due to the sucking l_pad field.
611		 * We do not want to assume that the flock64 structure is
612		 * laid out in the same in ILP32 and LP64 environments, so
613		 * we will copy in the ILP32 version of flock64 explicitly
614		 * and copy it to the native flock64 structure.
615		 */
616
617		if (copyin((void *)arg, &bf64_32, sizeof (bf64_32))) {
618			error = EFAULT;
619			break;
620		}
621
622		bf.l_type = (short)bf64_32.l_type;
623		bf.l_whence = (short)bf64_32.l_whence;
624		bf.l_start = bf64_32.l_start;
625		bf.l_len = bf64_32.l_len;
626		bf.l_sysid = (int)bf64_32.l_sysid;
627		bf.l_pid = (pid_t)bf64_32.l_pid;
628
629		if ((error = flock_check(vp, &bf, offset, MAXOFFSET_T)) != 0)
630			break;
631
632		if ((error = VOP_FRLOCK(vp, cmd, &bf, flag, offset,
633		    NULL, fp->f_cred, NULL)) != 0)
634			break;
635
636		if ((cmd == F_GETLK) && bf.l_type == F_UNLCK) {
637			if (copyout(&bf.l_type, &((struct flock *)arg)->l_type,
638			    sizeof (bf.l_type)))
639				error = EFAULT;
640			break;
641		}
642
643		if (cmd == F_GETLK) {
644			int i;
645
646			/*
647			 * We do not want to assume that the flock64 structure
648			 * is laid out in the same in ILP32 and LP64
649			 * environments, so we will copy out the ILP32 version
650			 * of flock64 explicitly after copying the native
651			 * flock64 structure to it.
652			 */
653			for (i = 0; i < 4; i++)
654				bf64_32.l_pad[i] = 0;
655			bf64_32.l_type = (int16_t)bf.l_type;
656			bf64_32.l_whence = (int16_t)bf.l_whence;
657			bf64_32.l_start = bf.l_start;
658			bf64_32.l_len = bf.l_len;
659			bf64_32.l_sysid = (int32_t)bf.l_sysid;
660			bf64_32.l_pid = (pid32_t)bf.l_pid;
661			if (copyout(&bf64_32, (void *)arg, sizeof (bf64_32)))
662				error = EFAULT;
663		}
664		break;
665/* ONC_PLUS EXTRACT END */
666#endif /* !defined(_LP64) || defined(_SYSCALL32_IMPL) */
667
668/* ONC_PLUS EXTRACT START */
669	case F_SHARE:
670	case F_SHARE_NBMAND:
671	case F_UNSHARE:
672
673		/*
674		 * Copy in input fields only.
675		 */
676		if (copyin((void *)arg, &fsh, sizeof (fsh))) {
677			error = EFAULT;
678			break;
679		}
680
681		/*
682		 * Local share reservations always have this simple form
683		 */
684		shr.s_access = fsh.f_access;
685		shr.s_deny = fsh.f_deny;
686		shr.s_sysid = 0;
687		shr.s_pid = ttoproc(curthread)->p_pid;
688		shr_own.sl_pid = shr.s_pid;
689		shr_own.sl_id = fsh.f_id;
690		shr.s_own_len = sizeof (shr_own);
691		shr.s_owner = (caddr_t)&shr_own;
692		error = VOP_SHRLOCK(vp, cmd, &shr, flag, fp->f_cred, NULL);
693/* ONC_PLUS EXTRACT END */
694		break;
695
696	default:
697		error = EINVAL;
698		break;
699	}
700
701	if (in_crit)
702		nbl_end_crit(vp);
703
704done:
705	releasef(fdes);
706out:
707	if (error)
708		return (set_errno(error));
709	return (retval);
710}
711
712/* ONC_PLUS EXTRACT START */
713int
714flock_check(vnode_t *vp, flock64_t *flp, offset_t offset, offset_t max)
715{
716	struct vattr	vattr;
717	int	error;
718	u_offset_t start, end;
719
720	/*
721	 * Determine the starting point of the request
722	 */
723	switch (flp->l_whence) {
724	case 0:		/* SEEK_SET */
725		start = (u_offset_t)flp->l_start;
726		if (start > max)
727			return (EINVAL);
728		break;
729	case 1:		/* SEEK_CUR */
730		if (flp->l_start > (max - offset))
731			return (EOVERFLOW);
732		start = (u_offset_t)(flp->l_start + offset);
733		if (start > max)
734			return (EINVAL);
735		break;
736	case 2:		/* SEEK_END */
737		vattr.va_mask = AT_SIZE;
738		if (error = VOP_GETATTR(vp, &vattr, 0, CRED(), NULL))
739			return (error);
740		if (flp->l_start > (max - (offset_t)vattr.va_size))
741			return (EOVERFLOW);
742		start = (u_offset_t)(flp->l_start + (offset_t)vattr.va_size);
743		if (start > max)
744			return (EINVAL);
745		break;
746	default:
747		return (EINVAL);
748	}
749
750	/*
751	 * Determine the range covered by the request.
752	 */
753	if (flp->l_len == 0)
754		end = MAXEND;
755	else if ((offset_t)flp->l_len > 0) {
756		if (flp->l_len > (max - start + 1))
757			return (EOVERFLOW);
758		end = (u_offset_t)(start + (flp->l_len - 1));
759		ASSERT(end <= max);
760	} else {
761		/*
762		 * Negative length; why do we even allow this ?
763		 * Because this allows easy specification of
764		 * the last n bytes of the file.
765		 */
766		end = start;
767		start += (u_offset_t)flp->l_len;
768		(start)++;
769		if (start > max)
770			return (EINVAL);
771		ASSERT(end <= max);
772	}
773	ASSERT(start <= max);
774	if (flp->l_type == F_UNLCK && flp->l_len > 0 &&
775	    end == (offset_t)max) {
776		flp->l_len = 0;
777	}
778	if (start  > end)
779		return (EINVAL);
780	return (0);
781}
782
783static int
784flock_get_start(vnode_t *vp, flock64_t *flp, offset_t offset, u_offset_t *start)
785{
786	struct vattr	vattr;
787	int	error;
788
789	/*
790	 * Determine the starting point of the request. Assume that it is
791	 * a valid starting point.
792	 */
793	switch (flp->l_whence) {
794	case 0:		/* SEEK_SET */
795		*start = (u_offset_t)flp->l_start;
796		break;
797	case 1:		/* SEEK_CUR */
798		*start = (u_offset_t)(flp->l_start + offset);
799		break;
800	case 2:		/* SEEK_END */
801		vattr.va_mask = AT_SIZE;
802		if (error = VOP_GETATTR(vp, &vattr, 0, CRED(), NULL))
803			return (error);
804		*start = (u_offset_t)(flp->l_start + (offset_t)vattr.va_size);
805		break;
806	default:
807		return (EINVAL);
808	}
809
810	return (0);
811}
812
813/*
814 * Take rctl action when the requested file descriptor is too big.
815 */
816static void
817fd_too_big(proc_t *p)
818{
819	mutex_enter(&p->p_lock);
820	(void) rctl_action(rctlproc_legacy[RLIMIT_NOFILE],
821	    p->p_rctls, p, RCA_SAFE);
822	mutex_exit(&p->p_lock);
823}
824/* ONC_PLUS EXTRACT END */
825