Use SYSREF to reference count struct vnode. v_usecount is now
[dragonfly.git] / sys / vfs / mfs / mfs_vnops.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. 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 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)mfs_vnops.c 8.11 (Berkeley) 5/22/95
34 * $FreeBSD: src/sys/ufs/mfs/mfs_vnops.c,v 1.47.2.1 2001/05/22 02:06:43 bp Exp $
3c37c940 35 * $DragonFly: src/sys/vfs/mfs/mfs_vnops.c,v 1.34 2007/05/06 19:23:34 dillon Exp $
984263bc
MD
36 */
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/kernel.h>
41#include <sys/proc.h>
42#include <sys/buf.h>
43#include <sys/vnode.h>
44#include <sys/malloc.h>
45#include <sys/sysproto.h>
46#include <sys/mman.h>
47#include <sys/conf.h>
48
d39c828b
MD
49#include <vm/vm.h>
50#include <vm/vm_object.h>
51#include <vm/vm_page.h>
52#include <vm/vm_pager.h>
53#include <vm/vnode_pager.h>
3020e3be 54
d39c828b 55#include <sys/buf2.h>
165dba55
DR
56#include <sys/thread2.h>
57
1f2de5d4
MD
58#include "mfsnode.h"
59#include "mfs_extern.h"
984263bc 60
a6ee311a
RG
61static int mfs_badop (struct vop_generic_args *);
62static int mfs_bmap (struct vop_bmap_args *);
63static int mfs_close (struct vop_close_args *);
64static int mfs_fsync (struct vop_fsync_args *);
65static int mfs_freeblks (struct vop_freeblks_args *);
66static int mfs_inactive (struct vop_inactive_args *); /* XXX */
67static int mfs_open (struct vop_open_args *);
68static int mfs_reclaim (struct vop_reclaim_args *); /* XXX */
69static int mfs_print (struct vop_print_args *); /* XXX */
70static int mfs_strategy (struct vop_strategy_args *); /* XXX */
71static int mfs_getpages (struct vop_getpages_args *); /* XXX */
984263bc 72/*
0961aa92
MD
73 * mfs vnode operations. Note: the vops here are used for the MFS block
74 * device, not for operations on files (MFS calls the ffs mount code for that)
984263bc 75 */
66a1ddf5
MD
76static struct vop_ops mfs_vnode_vops = {
77 .vop_default = mfs_badop,
78 .vop_bmap = mfs_bmap,
79 .vop_close = mfs_close,
80 .vop_freeblks = mfs_freeblks,
81 .vop_fsync = mfs_fsync,
82 .vop_getpages = mfs_getpages,
83 .vop_inactive = mfs_inactive,
84 .vop_ioctl = (void *)vop_enotty,
66a1ddf5
MD
85 .vop_open = mfs_open,
86 .vop_print = mfs_print,
87 .vop_reclaim = mfs_reclaim,
88 .vop_strategy = mfs_strategy,
984263bc 89};
984263bc 90
66a1ddf5
MD
91struct vop_ops *mfs_vnode_vops_p = &mfs_vnode_vops;
92
93VNODEOP_SET(mfs_vnode_vops);
984263bc
MD
94
95/*
96 * Vnode Operations.
97 *
98 * Open called to allow memory filesystem to initialize and
99 * validate before actual IO. Record our process identifier
100 * so we can tell when we are doing I/O to ourself.
77cf116b 101 *
e4c9c0c8
MD
102 * NOTE: new device sequencing. mounts check the device reference count
103 * before calling open, so we must associate the device in open and
104 * disassociate it in close rather then faking it when we created the vnode.
105 *
77cf116b
CP
106 * mfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
107 * struct thread *a_td)
984263bc
MD
108 */
109/* ARGSUSED */
110static int
77cf116b 111mfs_open(struct vop_open_args *ap)
984263bc 112{
e4c9c0c8
MD
113 struct vnode *vp = ap->a_vp;
114
115 if (vp->v_type != VCHR)
0cbed593 116 panic("mfs_open not VCHR");
e4c9c0c8 117 v_associate_rdev(vp, udev2dev(vp->v_udev, 0));
8ddc6004 118 return (vop_stdopen(ap));
984263bc
MD
119}
120
121static int
77cf116b 122mfs_fsync(struct vop_fsync_args *ap)
984263bc 123{
66a1ddf5 124 return (VOCALL(&spec_vnode_vops, &ap->a_head));
984263bc
MD
125}
126
127/*
128 * mfs_freeblks() - hook to allow us to free physical memory.
129 *
10f3fee5 130 * We implement the BUF_CMD_FREEBLKS strategy. We can't just madvise()
984263bc
MD
131 * here because we have to do it in the correct order vs other bio
132 * requests, so we queue it.
133 *
134 * Note: geteblk() sets B_INVAL. We leave it set to guarentee buffer
135 * throw-away on brelse()? XXX
77cf116b
CP
136 *
137 * mfs_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
984263bc 138 */
984263bc 139static int
77cf116b 140mfs_freeblks(struct vop_freeblks_args *ap)
984263bc
MD
141{
142 struct buf *bp;
e4c9c0c8 143 struct vnode *vp = ap->a_vp;
984263bc
MD
144
145 bp = geteblk(ap->a_length);
10f3fee5
MD
146 bp->b_flags |= B_ASYNC;
147 bp->b_cmd = BUF_CMD_FREEBLKS;
54078292 148 bp->b_bio1.bio_offset = ap->a_offset;
984263bc
MD
149 bp->b_bcount = ap->a_length;
150 BUF_KERNPROC(bp);
81b5c339 151 vn_strategy(vp, &bp->b_bio1);
984263bc
MD
152 return(0);
153}
154
155/*
156 * Pass I/O requests to the memory filesystem process.
77cf116b 157 *
81b5c339 158 * mfs_strategy(struct vnode *a_vp, struct bio *a_bio)
984263bc
MD
159 */
160static int
77cf116b 161mfs_strategy(struct vop_strategy_args *ap)
984263bc 162{
81b5c339
MD
163 struct bio *bio = ap->a_bio;
164 struct buf *bp = bio->bio_buf;
3ff2135f 165 struct mfsnode *mfsp;
dadab5e9 166 struct thread *td = curthread; /* XXX */
984263bc 167
81b5c339 168 mfsp = ap->a_vp->v_rdev->si_drv1;
e4c9c0c8
MD
169 if (mfsp == NULL) {
170 bp->b_error = ENXIO;
171 bp->b_flags |= B_ERROR;
81b5c339 172 biodone(bio);
e4c9c0c8
MD
173 return(0);
174 }
984263bc
MD
175
176 /*
177 * splbio required for queueing/dequeueing, in case of forwarded
178 * BPs from bio interrupts (?). It may not be necessary.
179 */
180
165dba55 181 crit_enter();
984263bc 182
dadab5e9 183 if (mfsp->mfs_td == NULL) {
984263bc 184 /*
10f3fee5
MD
185 * mini-root. Note: BUF_CMD_FREEBLKS not supported at the
186 * moment, since we do not know what kind of dataspace
187 * b_data is in.
984263bc
MD
188 */
189 caddr_t base;
190
54078292 191 base = mfsp->mfs_baseoff + bio->bio_offset;
10f3fee5
MD
192 switch(bp->b_cmd) {
193 case BUF_CMD_FREEBLKS:
194 break;
195 case BUF_CMD_READ:
984263bc 196 bcopy(base, bp->b_data, bp->b_bcount);
10f3fee5
MD
197 break;
198 case BUF_CMD_WRITE:
984263bc 199 bcopy(bp->b_data, base, bp->b_bcount);
10f3fee5
MD
200 break;
201 default:
202 panic("mfs: bad b_cmd %d\n", bp->b_cmd);
203 }
81b5c339 204 biodone(bio);
dadab5e9 205 } else if (mfsp->mfs_td == td) {
984263bc
MD
206 /*
207 * VOP to self
208 */
165dba55 209 crit_exit();
81b5c339 210 mfs_doio(bio, mfsp);
165dba55 211 crit_enter();
984263bc
MD
212 } else {
213 /*
214 * VOP from some other process, queue to MFS process and
215 * wake it up.
216 */
81b5c339 217 bioq_insert_tail(&mfsp->bio_queue, bio);
e4c9c0c8 218 wakeup((caddr_t)mfsp);
984263bc 219 }
165dba55 220 crit_exit();
984263bc
MD
221 return (0);
222}
223
224/*
225 * Memory file system I/O.
226 *
227 * Trivial on the HP since buffer has already been mapping into KVA space.
228 *
229 * Read and Write are handled with a simple copyin and copyout.
230 *
10f3fee5 231 * We also partially support VOP_FREEBLKS(). We can't implement
984263bc
MD
232 * completely -- for example, on fragments or inode metadata, but we can
233 * implement it for page-aligned requests.
234 */
235void
81b5c339 236mfs_doio(struct bio *bio, struct mfsnode *mfsp)
984263bc 237{
81b5c339 238 struct buf *bp = bio->bio_buf;
54078292 239 caddr_t base = mfsp->mfs_baseoff + bio->bio_offset;
10f3fee5 240 int bytes;
984263bc 241
10f3fee5
MD
242 switch(bp->b_cmd) {
243 case BUF_CMD_FREEBLKS:
984263bc 244 /*
10f3fee5 245 * Implement FREEBLKS, which allows the filesystem to tell
984263bc
MD
246 * a block device when blocks are no longer needed (like when
247 * a file is deleted). We use the hook to MADV_FREE the VM.
248 * This makes an MFS filesystem work as well or better then
249 * a sun-style swap-mounted filesystem.
250 */
10f3fee5 251 bytes = bp->b_bcount;
984263bc
MD
252
253 if ((vm_offset_t)base & PAGE_MASK) {
254 int n = PAGE_SIZE - ((vm_offset_t)base & PAGE_MASK);
255 bytes -= n;
256 base += n;
257 }
258 if (bytes > 0) {
259 struct madvise_args uap;
260
261 bytes &= ~PAGE_MASK;
262 if (bytes != 0) {
263 bzero(&uap, sizeof(uap));
264 uap.addr = base;
265 uap.len = bytes;
266 uap.behav = MADV_FREE;
753fd850 267 sys_madvise(&uap);
984263bc
MD
268 }
269 }
270 bp->b_error = 0;
10f3fee5
MD
271 break;
272 case BUF_CMD_READ:
984263bc
MD
273 /*
274 * Read data from our 'memory' disk
275 */
276 bp->b_error = copyin(base, bp->b_data, bp->b_bcount);
10f3fee5
MD
277 break;
278 case BUF_CMD_WRITE:
984263bc
MD
279 /*
280 * Write data to our 'memory' disk
281 */
282 bp->b_error = copyout(bp->b_data, base, bp->b_bcount);
10f3fee5
MD
283 break;
284 default:
285 panic("mfs: bad b_cmd %d\n", bp->b_cmd);
984263bc
MD
286 }
287 if (bp->b_error)
288 bp->b_flags |= B_ERROR;
81b5c339 289 biodone(bio);
984263bc
MD
290}
291
292/*
293 * This is a noop, simply returning what one has been given.
77cf116b 294 *
54078292
MD
295 * mfs_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp,
296 * off_t *a_doffsetp, int *a_runp, int *a_runb)
984263bc
MD
297 */
298static int
77cf116b 299mfs_bmap(struct vop_bmap_args *ap)
984263bc 300{
984263bc
MD
301 if (ap->a_vpp != NULL)
302 *ap->a_vpp = ap->a_vp;
54078292
MD
303 if (ap->a_doffsetp != NULL)
304 *ap->a_doffsetp = ap->a_loffset;
984263bc
MD
305 if (ap->a_runp != NULL)
306 *ap->a_runp = 0;
54078292
MD
307 if (ap->a_runb != NULL)
308 *ap->a_runb = 0;
984263bc
MD
309 return (0);
310}
311
312/*
313 * Memory filesystem close routine
77cf116b
CP
314 *
315 * mfs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
316 * struct thread *a_td)
984263bc
MD
317 */
318/* ARGSUSED */
319static int
77cf116b 320mfs_close(struct vop_close_args *ap)
984263bc 321{
3ff2135f
RG
322 struct vnode *vp = ap->a_vp;
323 struct mfsnode *mfsp = VTOMFS(vp);
81b5c339 324 struct bio *bio;
d39c828b 325 int error = 0;
984263bc
MD
326
327 /*
328 * Finish any pending I/O requests.
329 */
81b5c339
MD
330 while ((bio = bioq_first(&mfsp->bio_queue)) != NULL) {
331 bioq_remove(&mfsp->bio_queue, bio);
332 mfs_doio(bio, mfsp);
333 wakeup((caddr_t)bio->bio_buf);
984263bc 334 }
d39c828b 335
984263bc 336 /*
d39c828b 337 * We really only care about the last close
984263bc 338 */
d39c828b 339 if (vp->v_opencount > 1)
8ddc6004 340 goto done;
d39c828b
MD
341
342 /*
343 * Synchronize any remaining buffers and then destroy them.
344 */
87de5057 345 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0)
d39c828b
MD
346 goto done;
347
348 /*
349 * Get rid of the pseudo-backing object. Since the object is
350 * not directly memory mapped, we don't have to worry about
351 * synchronizing it.
352 */
353 if (vp->v_object)
354 vm_pager_deallocate(vp->v_object);
355
984263bc
MD
356 /*
357 * There should be no way to have any more uses of this
358 * vnode, so if we find any other uses, it is a panic.
359 */
3c37c940
MD
360 if (vp->v_sysref.refcnt > 1)
361 kprintf("mfs_close: ref count %d > 1\n", vp->v_sysref.refcnt);
362 if (vp->v_sysref.refcnt > 1 || (bioq_first(&mfsp->bio_queue) != NULL))
984263bc
MD
363 panic("mfs_close");
364 /*
365 * Send a request to the filesystem server to exit.
366 */
367 mfsp->mfs_active = 0;
e4c9c0c8
MD
368 v_release_rdev(vp);
369 if (mfsp->mfs_dev) {
370 destroy_dev(mfsp->mfs_dev);
371 mfsp->mfs_dev = NULL;
372 }
373 wakeup((caddr_t)mfsp);
8ddc6004
MD
374done:
375 vop_stdclose(ap);
376 return (error);
984263bc
MD
377}
378
379/*
380 * Memory filesystem inactive routine
77cf116b
CP
381 *
382 * mfs_inactive(struct vnode *a_vp, struct thread *a_td)
984263bc
MD
383 */
384/* ARGSUSED */
385static int
77cf116b 386mfs_inactive(struct vop_inactive_args *ap)
984263bc
MD
387{
388 struct vnode *vp = ap->a_vp;
389 struct mfsnode *mfsp = VTOMFS(vp);
390
81b5c339 391 if (bioq_first(&mfsp->bio_queue) != NULL)
984263bc 392 panic("mfs_inactive: not inactive (next buffer %p)",
81b5c339 393 bioq_first(&mfsp->bio_queue));
984263bc
MD
394 return (0);
395}
396
397/*
398 * Reclaim a memory filesystem devvp so that it can be reused.
77cf116b
CP
399 *
400 * mfs_reclaim(struct vnode *a_vp)
984263bc
MD
401 */
402static int
77cf116b 403mfs_reclaim(struct vop_reclaim_args *ap)
984263bc 404{
3ff2135f 405 struct vnode *vp = ap->a_vp;
984263bc
MD
406
407 FREE(vp->v_data, M_MFSNODE);
408 vp->v_data = NULL;
409 return (0);
410}
411
412/*
413 * Print out the contents of an mfsnode.
77cf116b
CP
414 *
415 * mfs_print(struct vnode *a_vp)
984263bc
MD
416 */
417static int
77cf116b 418mfs_print(struct vop_print_args *ap)
984263bc 419{
3ff2135f 420 struct mfsnode *mfsp = VTOMFS(ap->a_vp);
984263bc 421
086c1d7e 422 kprintf("tag VT_MFS, td %p, base %p, size %ld\n",
dadab5e9 423 mfsp->mfs_td, (void *)mfsp->mfs_baseoff, mfsp->mfs_size);
984263bc
MD
424 return (0);
425}
426
427/*
428 * Block device bad operation
429 */
430static int
431mfs_badop(struct vop_generic_args *ap)
432{
433 int i;
434
086c1d7e 435 kprintf("mfs_badop[%s]\n", ap->a_desc->sd_name);
984263bc 436 i = vop_defaultop(ap);
086c1d7e 437 kprintf("mfs_badop[%s] = %d\n", ap->a_desc->sd_name, i);
984263bc
MD
438 return (i);
439}
440
984263bc 441static int
77cf116b 442mfs_getpages(struct vop_getpages_args *ap)
984263bc 443{
66a1ddf5 444 return (VOCALL(&spec_vnode_vops, &ap->a_head));
984263bc 445}