]> git.saurik.com Git - apple/xnu.git/blame - bsd/kern/kern_symfile.c
xnu-4903.270.47.tar.gz
[apple/xnu.git] / bsd / kern / kern_symfile.c
CommitLineData
1c79356b 1/*
2d21ac55 2 * Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
5d5c5d0d 3 *
2d21ac55 4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
0a7de745 5 *
2d21ac55
A
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
0a7de745 14 *
2d21ac55
A
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
0a7de745 17 *
2d21ac55
A
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
8f6c56a5
A
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
2d21ac55
A
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
0a7de745 25 *
2d21ac55 26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
1c79356b
A
27 */
28/* Copyright (c) 1998 Apple Computer, Inc. All rights reserved.
29 *
30 * File: bsd/kern/kern_symfile.c
31 *
1c79356b 32 * HISTORY
1c79356b
A
33 */
34
35#include <mach/vm_param.h>
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/signalvar.h>
40#include <sys/resourcevar.h>
41#include <sys/namei.h>
91447636
A
42#include <sys/vnode_internal.h>
43#include <sys/proc_internal.h>
44#include <sys/kauth.h>
1c79356b
A
45#include <sys/timeb.h>
46#include <sys/times.h>
1c79356b 47#include <sys/acct.h>
91447636 48#include <sys/file_internal.h>
1c79356b
A
49#include <sys/uio.h>
50#include <sys/kernel.h>
51#include <sys/stat.h>
91447636
A
52#include <sys/disk.h>
53#include <sys/conf.h>
3e170ce0 54#include <sys/content_protection.h>
d9a64523 55#include <sys/fsctl.h>
1c79356b
A
56
57#include <mach-o/loader.h>
58#include <mach-o/nlist.h>
59
91447636 60#include <kern/kalloc.h>
1c79356b 61#include <vm/vm_kern.h>
91447636 62#include <pexpert/pexpert.h>
3e170ce0 63#include <IOKit/IOPolledInterface.h>
1c79356b 64
d9a64523
A
65#define HIBERNATE_MIN_PHYSICAL_LBA (34)
66#define HIBERNATE_MIN_FILE_SIZE (1024*1024)
67
2d21ac55
A
68/* This function is called from kern_sysctl in the current process context;
69 * it is exported with the System6.0.exports, but this appears to be a legacy
70 * export, as there are no internal consumers.
1c79356b 71 */
2d21ac55 72int
7ddcb079
A
73get_kernel_symfile(__unused proc_t p, __unused char const **symfile);
74int
2d21ac55 75get_kernel_symfile(__unused proc_t p, __unused char const **symfile)
1c79356b 76{
0a7de745 77 return KERN_FAILURE;
0b4e3aa0 78}
91447636 79
0a7de745
A
80struct kern_direct_file_io_ref_t {
81 vfs_context_t ctx;
82 struct vnode * vp;
83 dev_t device;
84 uint32_t blksize;
85 off_t filelength;
86 char cf;
87 char pinned;
88 char frozen;
89 char wbcranged;
3a60a9f5
A
90};
91
92
0a7de745
A
93static int
94file_ioctl(void * p1, void * p2, u_long theIoctl, caddr_t result)
3a60a9f5 95{
0a7de745 96 dev_t device = *(dev_t*) p1;
3a60a9f5 97
0a7de745
A
98 return (*bdevsw[major(device)].d_ioctl)
99 (device, theIoctl, result, S_IFBLK, p2);
3a60a9f5
A
100}
101
0a7de745
A
102static int
103device_ioctl(void * p1, __unused void * p2, u_long theIoctl, caddr_t result)
3a60a9f5 104{
0a7de745 105 return VNOP_IOCTL(p1, theIoctl, result, 0, p2);
3a60a9f5
A
106}
107
316670eb
A
108static int
109kern_ioctl_file_extents(struct kern_direct_file_io_ref_t * ref, u_long theIoctl, off_t offset, off_t end)
110{
0a7de745
A
111 int error = 0;
112 int (*do_ioctl)(void * p1, void * p2, u_long theIoctl, caddr_t result);
113 void * p1;
114 void * p2;
115 uint64_t fileblk;
116 size_t filechunk;
117 dk_extent_t extent;
118 dk_unmap_t unmap;
119 _dk_cs_pin_t pin;
120
121 bzero(&extent, sizeof(dk_extent_t));
122 bzero(&unmap, sizeof(dk_unmap_t));
123 bzero(&pin, sizeof(pin));
124 if (ref->vp->v_type == VREG) {
125 p1 = &ref->device;
126 p2 = kernproc;
127 do_ioctl = &file_ioctl;
128 } else {
129 /* Partition. */
130 p1 = ref->vp;
131 p2 = ref->ctx;
132 do_ioctl = &device_ioctl;
133 }
134
135 if (_DKIOCCSPINEXTENT == theIoctl) {
136 /* Tell CS the image size, so it knows whether to place the subsequent pins SSD/HDD */
137 pin.cp_extent.length = end;
138 pin.cp_flags = _DKIOCCSHIBERNATEIMGSIZE;
139 (void) do_ioctl(p1, p2, _DKIOCCSPINEXTENT, (caddr_t)&pin);
140 } else if (_DKIOCCSUNPINEXTENT == theIoctl) {
141 /* Tell CS hibernation is done, so it can stop blocking overlapping writes */
142 pin.cp_flags = _DKIOCCSPINDISCARDBLACKLIST;
143 (void) do_ioctl(p1, p2, _DKIOCCSUNPINEXTENT, (caddr_t)&pin);
144 }
145
146 for (; offset < end; offset += filechunk) {
147 if (ref->vp->v_type == VREG) {
148 daddr64_t blkno;
149 filechunk = 1 * 1024 * 1024 * 1024;
150 if (filechunk > (size_t)(end - offset)) {
151 filechunk = (size_t)(end - offset);
152 }
153 error = VNOP_BLOCKMAP(ref->vp, offset, filechunk, &blkno,
154 &filechunk, NULL, VNODE_WRITE | VNODE_BLOCKMAP_NO_TRACK, NULL);
155 if (error) {
156 break;
157 }
158 if (-1LL == blkno) {
159 continue;
160 }
161 fileblk = blkno * ref->blksize;
162 } else if ((ref->vp->v_type == VBLK) || (ref->vp->v_type == VCHR)) {
163 fileblk = offset;
164 filechunk = ref->filelength;
165 }
166
167 if (DKIOCUNMAP == theIoctl) {
168 extent.offset = fileblk;
169 extent.length = filechunk;
170 unmap.extents = &extent;
171 unmap.extentsCount = 1;
172 error = do_ioctl(p1, p2, theIoctl, (caddr_t)&unmap);
173// printf("DKIOCUNMAP(%d) 0x%qx, 0x%qx\n", error, extent.offset, extent.length);
174 } else if (_DKIOCCSPINEXTENT == theIoctl) {
175 pin.cp_extent.offset = fileblk;
176 pin.cp_extent.length = filechunk;
177 pin.cp_flags = _DKIOCCSPINFORHIBERNATION;
178 error = do_ioctl(p1, p2, theIoctl, (caddr_t)&pin);
179 if (error && (ENOTTY != error)) {
180 printf("_DKIOCCSPINEXTENT(%d) 0x%qx, 0x%qx\n", error, pin.cp_extent.offset, pin.cp_extent.length);
181 }
182 } else if (_DKIOCCSUNPINEXTENT == theIoctl) {
183 pin.cp_extent.offset = fileblk;
184 pin.cp_extent.length = filechunk;
185 pin.cp_flags = _DKIOCCSPINFORHIBERNATION;
186 error = do_ioctl(p1, p2, theIoctl, (caddr_t)&pin);
187 if (error && (ENOTTY != error)) {
188 printf("_DKIOCCSUNPINEXTENT(%d) 0x%qx, 0x%qx\n", error, pin.cp_extent.offset, pin.cp_extent.length);
189 }
190 } else {
191 error = EINVAL;
192 }
193
194 if (error) {
195 break;
196 }
197 }
198 return error;
316670eb
A
199}
200
3e170ce0 201extern uint32_t freespace_mb(vnode_t vp);
7ddcb079 202
3a60a9f5 203struct kern_direct_file_io_ref_t *
0a7de745
A
204kern_open_file_for_direct_io(const char * name,
205 uint32_t iflags,
206 kern_get_file_extents_callback_t callback,
207 void * callback_ref,
208 off_t set_file_size,
209 off_t fs_free_size,
210 off_t write_file_offset,
211 void * write_file_addr,
212 size_t write_file_len,
213 dev_t * partition_device_result,
214 dev_t * image_device_result,
215 uint64_t * partitionbase_result,
216 uint64_t * maxiocount_result,
217 uint32_t * oflags)
3a60a9f5 218{
0a7de745
A
219 struct kern_direct_file_io_ref_t * ref;
220
221 proc_t p;
222 struct vnode_attr va;
223 dk_apfs_wbc_range_t wbc_range;
224 int error;
225 off_t f_offset;
226 uint64_t fileblk;
227 size_t filechunk;
228 uint64_t physoffset, minoffset;
229 dev_t device;
230 dev_t target = 0;
231 int isssd = 0;
232 uint32_t flags = 0;
233 uint32_t blksize;
234 off_t maxiocount, count, segcount, wbctotal;
235 boolean_t locked = FALSE;
236 int fmode, cmode;
237 struct nameidata nd;
238 u_int32_t ndflags;
239 off_t mpFree;
240
241 int (*do_ioctl)(void * p1, void * p2, u_long theIoctl, caddr_t result);
242 void * p1 = NULL;
243 void * p2 = NULL;
244
3a60a9f5 245 error = EFAULT;
0a7de745
A
246
247 ref = (struct kern_direct_file_io_ref_t *) kalloc(sizeof(struct kern_direct_file_io_ref_t));
248 if (!ref) {
249 error = EFAULT;
250 goto out;
251 }
252
253 bzero(ref, sizeof(*ref));
254 p = kernproc;
255 ref->ctx = vfs_context_kernel();
256
257 fmode = (kIOPolledFileCreate & iflags) ? (O_CREAT | FWRITE) : FWRITE;
258 cmode = S_IRUSR | S_IWUSR;
259 ndflags = NOFOLLOW;
260 NDINIT(&nd, LOOKUP, OP_OPEN, ndflags, UIO_SYSSPACE, CAST_USER_ADDR_T(name), ref->ctx);
261 VATTR_INIT(&va);
262 VATTR_SET(&va, va_mode, cmode);
263 VATTR_SET(&va, va_dataprotect_flags, VA_DP_RAWENCRYPTED);
264 VATTR_SET(&va, va_dataprotect_class, PROTECTION_CLASS_D);
265 if ((error = vn_open_auth(&nd, &fmode, &va))) {
266 kprintf("vn_open_auth(fmode: %d, cmode: %d) failed with error: %d\n", fmode, cmode, error);
267 goto out;
268 }
269
270 ref->vp = nd.ni_vp;
271 if (ref->vp->v_type == VREG) {
272 vnode_lock_spin(ref->vp);
273 SET(ref->vp->v_flag, VSWAP);
274 vnode_unlock(ref->vp);
275 }
276
277 if (write_file_addr && write_file_len) {
278 if ((error = kern_write_file(ref, write_file_offset, write_file_addr, write_file_len, IO_SKIP_ENCRYPTION))) {
279 kprintf("kern_write_file() failed with error: %d\n", error);
280 goto out;
281 }
282 }
283
284 VATTR_INIT(&va);
285 VATTR_WANTED(&va, va_rdev);
286 VATTR_WANTED(&va, va_fsid);
287 VATTR_WANTED(&va, va_devid);
288 VATTR_WANTED(&va, va_data_size);
289 VATTR_WANTED(&va, va_data_alloc);
290 VATTR_WANTED(&va, va_nlink);
291 error = EFAULT;
292 if (vnode_getattr(ref->vp, &va, ref->ctx)) {
293 goto out;
294 }
295
296 wbctotal = 0;
297 mpFree = freespace_mb(ref->vp);
298 mpFree <<= 20;
299 kprintf("kern_direct_file(%s): vp size %qd, alloc %qd, mp free %qd, keep free %qd\n",
300 name, va.va_data_size, va.va_data_alloc, mpFree, fs_free_size);
301
302 if (ref->vp->v_type == VREG) {
303 /* Don't dump files with links. */
304 if (va.va_nlink != 1) {
305 goto out;
306 }
307
308 device = (VATTR_IS_SUPPORTED(&va, va_devid)) ? va.va_devid : va.va_fsid;
309 ref->filelength = va.va_data_size;
310
311 p1 = &device;
312 p2 = p;
313 do_ioctl = &file_ioctl;
314
315 if (kIOPolledFileHibernate & iflags) {
316 error = do_ioctl(p1, p2, DKIOCAPFSGETWBCRANGE, (caddr_t) &wbc_range);
317 ref->wbcranged = (error == 0);
318 }
319 if (ref->wbcranged) {
320 uint32_t idx;
321 assert(wbc_range.count <= (sizeof(wbc_range.extents) / sizeof(wbc_range.extents[0])));
322 for (idx = 0; idx < wbc_range.count; idx++) {
323 wbctotal += wbc_range.extents[idx].length;
324 }
325 kprintf("kern_direct_file(%s): wbc %qd\n", name, wbctotal);
326 if (wbctotal) {
327 target = wbc_range.dev;
328 }
329 }
330
331 if (set_file_size) {
332 if (wbctotal) {
333 if (wbctotal >= set_file_size) {
334 set_file_size = HIBERNATE_MIN_FILE_SIZE;
335 } else {
336 set_file_size -= wbctotal;
337 if (set_file_size < HIBERNATE_MIN_FILE_SIZE) {
338 set_file_size = HIBERNATE_MIN_FILE_SIZE;
339 }
340 }
341 }
342 if (fs_free_size) {
343 mpFree += va.va_data_alloc;
344 if ((mpFree < set_file_size) || ((mpFree - set_file_size) < fs_free_size)) {
345 error = ENOSPC;
346 goto out;
347 }
348 }
349 error = vnode_setsize(ref->vp, set_file_size, IO_NOZEROFILL | IO_NOAUTH, ref->ctx);
350 if (error) {
351 goto out;
352 }
353 ref->filelength = set_file_size;
354 }
355 } else if ((ref->vp->v_type == VBLK) || (ref->vp->v_type == VCHR)) {
356 /* Partition. */
357 device = va.va_rdev;
358
359 p1 = ref->vp;
360 p2 = ref->ctx;
361 do_ioctl = &device_ioctl;
362 } else {
363 /* Don't dump to non-regular files. */
364 error = EFAULT;
365 goto out;
366 }
367 ref->device = device;
368
369 // probe for CF
370 dk_corestorage_info_t cs_info;
371 memset(&cs_info, 0, sizeof(dk_corestorage_info_t));
372 error = do_ioctl(p1, p2, DKIOCCORESTORAGE, (caddr_t)&cs_info);
373 ref->cf = (error == 0) && (cs_info.flags & DK_CORESTORAGE_ENABLE_HOTFILES);
374
375 // get block size
376
377 error = do_ioctl(p1, p2, DKIOCGETBLOCKSIZE, (caddr_t) &ref->blksize);
378 if (error) {
379 goto out;
380 }
381
382 minoffset = HIBERNATE_MIN_PHYSICAL_LBA * ref->blksize;
383
384 if (ref->vp->v_type != VREG) {
385 error = do_ioctl(p1, p2, DKIOCGETBLOCKCOUNT, (caddr_t) &fileblk);
386 if (error) {
387 goto out;
fe8ab488 388 }
0a7de745
A
389 ref->filelength = fileblk * ref->blksize;
390 }
391
392 // pin logical extents, CS version
393
394 error = kern_ioctl_file_extents(ref, _DKIOCCSPINEXTENT, 0, ref->filelength);
395 if (error && (ENOTTY != error)) {
396 goto out;
397 }
398 ref->pinned = (error == 0);
399
400 // pin logical extents, apfs version
401
402 error = VNOP_IOCTL(ref->vp, FSCTL_FREEZE_EXTENTS, NULL, 0, ref->ctx);
403 if (error && (ENOTTY != error)) {
404 goto out;
405 }
406 ref->frozen = (error == 0);
407
408 // generate the block list
409
410 error = do_ioctl(p1, p2, DKIOCLOCKPHYSICALEXTENTS, NULL);
411 if (error) {
412 goto out;
413 }
414 locked = TRUE;
415
416 f_offset = 0;
417 for (; f_offset < ref->filelength; f_offset += filechunk) {
418 if (ref->vp->v_type == VREG) {
419 filechunk = 1 * 1024 * 1024 * 1024;
420 daddr64_t blkno;
421
422 error = VNOP_BLOCKMAP(ref->vp, f_offset, filechunk, &blkno,
423 &filechunk, NULL, VNODE_WRITE | VNODE_BLOCKMAP_NO_TRACK, NULL);
424 if (error) {
425 goto out;
426 }
427 if (-1LL == blkno) {
428 continue;
429 }
430 fileblk = blkno * ref->blksize;
431 } else if ((ref->vp->v_type == VBLK) || (ref->vp->v_type == VCHR)) {
432 fileblk = f_offset;
433 filechunk = f_offset ? 0 : ref->filelength;
434 }
435
436 physoffset = 0;
437 while (physoffset < filechunk) {
438 dk_physical_extent_t getphysreq;
439 bzero(&getphysreq, sizeof(getphysreq));
440
441 getphysreq.offset = fileblk + physoffset;
442 getphysreq.length = (filechunk - physoffset);
443 error = do_ioctl(p1, p2, DKIOCGETPHYSICALEXTENT, (caddr_t) &getphysreq);
444 if (error) {
445 goto out;
446 }
447 if (!target) {
448 target = getphysreq.dev;
449 } else if (target != getphysreq.dev) {
450 error = ENOTSUP;
451 goto out;
452 }
453
454 assert(getphysreq.offset >= minoffset);
d9a64523 455
39236c6e 456#if HIBFRAGMENT
0a7de745
A
457 uint64_t rev;
458 for (rev = 4096; rev <= getphysreq.length; rev += 4096) {
459 callback(callback_ref, getphysreq.offset + getphysreq.length - rev, 4096);
460 }
39236c6e 461#else
0a7de745 462 callback(callback_ref, getphysreq.offset, getphysreq.length);
39236c6e 463#endif
0a7de745
A
464 physoffset += getphysreq.length;
465 }
466 }
467 if (ref->wbcranged) {
468 uint32_t idx;
469 for (idx = 0; idx < wbc_range.count; idx++) {
470 assert(wbc_range.extents[idx].offset >= minoffset);
471 callback(callback_ref, wbc_range.extents[idx].offset, wbc_range.extents[idx].length);
472 }
473 }
474 callback(callback_ref, 0ULL, 0ULL);
475
476 if (ref->vp->v_type == VREG) {
477 p1 = &target;
478 } else {
479 p1 = &target;
480 p2 = p;
481 do_ioctl = &file_ioctl;
482 }
483
484 // get partition base
485
486 if (partitionbase_result) {
487 error = do_ioctl(p1, p2, DKIOCGETBASE, (caddr_t) partitionbase_result);
488 if (error) {
489 goto out;
490 }
491 }
492
493 // get block size & constraints
494
495 error = do_ioctl(p1, p2, DKIOCGETBLOCKSIZE, (caddr_t) &blksize);
496 if (error) {
497 goto out;
498 }
499
500 maxiocount = 1 * 1024 * 1024 * 1024;
501
502 error = do_ioctl(p1, p2, DKIOCGETMAXBLOCKCOUNTREAD, (caddr_t) &count);
503 if (error) {
504 count = 0;
505 }
506 count *= blksize;
507 if (count && (count < maxiocount)) {
508 maxiocount = count;
509 }
510
511 error = do_ioctl(p1, p2, DKIOCGETMAXBLOCKCOUNTWRITE, (caddr_t) &count);
512 if (error) {
513 count = 0;
514 }
515 count *= blksize;
516 if (count && (count < maxiocount)) {
517 maxiocount = count;
518 }
519
520 error = do_ioctl(p1, p2, DKIOCGETMAXBYTECOUNTREAD, (caddr_t) &count);
521 if (error) {
522 count = 0;
523 }
524 if (count && (count < maxiocount)) {
525 maxiocount = count;
526 }
527
528 error = do_ioctl(p1, p2, DKIOCGETMAXBYTECOUNTWRITE, (caddr_t) &count);
529 if (error) {
530 count = 0;
531 }
532 if (count && (count < maxiocount)) {
533 maxiocount = count;
534 }
535
536 error = do_ioctl(p1, p2, DKIOCGETMAXSEGMENTBYTECOUNTREAD, (caddr_t) &count);
537 if (!error) {
538 error = do_ioctl(p1, p2, DKIOCGETMAXSEGMENTCOUNTREAD, (caddr_t) &segcount);
539 }
540 if (error) {
541 count = segcount = 0;
542 }
543 count *= segcount;
544 if (count && (count < maxiocount)) {
545 maxiocount = count;
546 }
547
548 error = do_ioctl(p1, p2, DKIOCGETMAXSEGMENTBYTECOUNTWRITE, (caddr_t) &count);
549 if (!error) {
550 error = do_ioctl(p1, p2, DKIOCGETMAXSEGMENTCOUNTWRITE, (caddr_t) &segcount);
551 }
552 if (error) {
553 count = segcount = 0;
554 }
555 count *= segcount;
556 if (count && (count < maxiocount)) {
557 maxiocount = count;
558 }
559
560 kprintf("max io 0x%qx bytes\n", maxiocount);
561 if (maxiocount_result) {
562 *maxiocount_result = maxiocount;
563 }
564
565 error = do_ioctl(p1, p2, DKIOCISSOLIDSTATE, (caddr_t)&isssd);
566 if (!error && isssd) {
567 flags |= kIOPolledFileSSD;
568 }
569
570 if (partition_device_result) {
571 *partition_device_result = device;
572 }
573 if (image_device_result) {
574 *image_device_result = target;
575 }
576 if (oflags) {
577 *oflags = flags;
578 }
579
580 if ((ref->vp->v_type == VBLK) || (ref->vp->v_type == VCHR)) {
581 vnode_close(ref->vp, FWRITE, ref->ctx);
582 ref->vp = NULLVP;
583 ref->ctx = NULL;
584 }
3e170ce0 585
3a60a9f5 586out:
0a7de745
A
587 printf("kern_open_file_for_direct_io(%p, %d)\n", ref, error);
588
589
590 if (error && locked) {
591 p1 = &device;
592 (void) do_ioctl(p1, p2, DKIOCUNLOCKPHYSICALEXTENTS, NULL);
593 }
594
595 if (error && ref) {
596 if (ref->vp) {
597 (void) kern_ioctl_file_extents(ref, _DKIOCCSUNPINEXTENT, 0, (ref->pinned && ref->cf) ? ref->filelength : 0);
598
599 if (ref->frozen) {
600 (void) VNOP_IOCTL(ref->vp, FSCTL_THAW_EXTENTS, NULL, 0, ref->ctx);
601 }
602 if (ref->wbcranged) {
603 (void) do_ioctl(p1, p2, DKIOCAPFSRELEASEWBCRANGE, (caddr_t) NULL);
604 }
605 vnode_close(ref->vp, FWRITE, ref->ctx);
606 ref->vp = NULLVP;
607 }
608 ref->ctx = NULL;
609 kfree(ref, sizeof(struct kern_direct_file_io_ref_t));
610 ref = NULL;
611 }
612
613 return ref;
3a60a9f5
A
614}
615
616int
3e170ce0 617kern_write_file(struct kern_direct_file_io_ref_t * ref, off_t offset, void * addr, size_t len, int ioflag)
3a60a9f5 618{
0a7de745
A
619 return vn_rdwr(UIO_WRITE, ref->vp,
620 addr, len, offset,
621 UIO_SYSSPACE, ioflag | IO_SYNC | IO_NODELOCKED | IO_UNIT,
622 vfs_context_ucred(ref->ctx), (int *) 0,
623 vfs_context_proc(ref->ctx));
3a60a9f5
A
624}
625
3e170ce0
A
626int
627kern_read_file(struct kern_direct_file_io_ref_t * ref, off_t offset, void * addr, size_t len, int ioflag)
628{
0a7de745
A
629 return vn_rdwr(UIO_READ, ref->vp,
630 addr, len, offset,
631 UIO_SYSSPACE, ioflag | IO_SYNC | IO_NODELOCKED | IO_UNIT,
632 vfs_context_ucred(ref->ctx), (int *) 0,
633 vfs_context_proc(ref->ctx));
3e170ce0
A
634}
635
636
637struct mount *
638kern_file_mount(struct kern_direct_file_io_ref_t * ref)
639{
0a7de745 640 return ref->vp->v_mount;
3e170ce0 641}
7ddcb079 642
3a60a9f5 643void
6d2010ae 644kern_close_file_for_direct_io(struct kern_direct_file_io_ref_t * ref,
0a7de745
A
645 off_t write_offset, void * addr, size_t write_length,
646 off_t discard_offset, off_t discard_end)
3a60a9f5 647{
0a7de745
A
648 int error;
649 printf("kern_close_file_for_direct_io(%p)\n", ref);
650
651 if (!ref) {
652 return;
653 }
654
655 if (ref->vp) {
656 int (*do_ioctl)(void * p1, void * p2, u_long theIoctl, caddr_t result);
657 void * p1;
658 void * p2;
659
660 discard_offset = ((discard_offset + ref->blksize - 1) & ~(((off_t) ref->blksize) - 1));
661 discard_end = ((discard_end) & ~(((off_t) ref->blksize) - 1));
662
663 if (ref->vp->v_type == VREG) {
664 p1 = &ref->device;
665 p2 = kernproc;
666 do_ioctl = &file_ioctl;
667 } else {
668 /* Partition. */
669 p1 = ref->vp;
670 p2 = ref->ctx;
671 do_ioctl = &device_ioctl;
672 }
673 (void) do_ioctl(p1, p2, DKIOCUNLOCKPHYSICALEXTENTS, NULL);
fe8ab488 674
3e170ce0
A
675 //XXX If unmapping extents then don't also need to unpin; except ...
676 //XXX if file unaligned (HFS 4k / Fusion 128k) then pin is superset and
677 //XXX unmap is subset, so save extra walk over file extents (and the risk
678 //XXX that CF drain starts) vs leaving partial units pinned to SSD
679 //XXX (until whatever was sharing also unmaps). Err on cleaning up fully.
680 boolean_t will_unmap = (!ref->pinned || ref->cf) && (discard_end > discard_offset);
681 boolean_t will_unpin = (ref->pinned && ref->cf /* && !will_unmap */);
fe8ab488 682
3e170ce0
A
683 (void) kern_ioctl_file_extents(ref, _DKIOCCSUNPINEXTENT, 0, (will_unpin) ? ref->filelength : 0);
684
0a7de745
A
685 if (will_unmap) {
686 (void) kern_ioctl_file_extents(ref, DKIOCUNMAP, discard_offset, (ref->cf) ? ref->filelength : discard_end);
687 }
3e170ce0 688
0a7de745
A
689 if (ref->frozen) {
690 (void) VNOP_IOCTL(ref->vp, FSCTL_THAW_EXTENTS, NULL, 0, ref->ctx);
691 }
692 if (ref->wbcranged) {
693 (void) do_ioctl(p1, p2, DKIOCAPFSRELEASEWBCRANGE, (caddr_t) NULL);
694 }
d9a64523 695
0a7de745
A
696 if (addr && write_length) {
697 (void) kern_write_file(ref, write_offset, addr, write_length, IO_SKIP_ENCRYPTION);
698 }
6d2010ae 699
0a7de745 700 error = vnode_close(ref->vp, FWRITE, ref->ctx);
6d2010ae 701
0a7de745
A
702 ref->vp = NULLVP;
703 kprintf("vnode_close(%d)\n", error);
00867663 704
0a7de745 705 }
00867663 706
0a7de745 707 ref->ctx = NULL;
00867663 708
0a7de745 709 kfree(ref, sizeof(struct kern_direct_file_io_ref_t));
3a60a9f5 710}