]> git.saurik.com Git - apple/xnu.git/blob - bsd/kern/kern_symfile.c
xnu-7195.101.1.tar.gz
[apple/xnu.git] / bsd / kern / kern_symfile.c
1 /*
2 * Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
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.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
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
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
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.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28 /* Copyright (c) 1998 Apple Computer, Inc. All rights reserved.
29 *
30 * File: bsd/kern/kern_symfile.c
31 *
32 * HISTORY
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>
42 #include <sys/vnode_internal.h>
43 #include <sys/proc_internal.h>
44 #include <sys/kauth.h>
45 #include <sys/timeb.h>
46 #include <sys/times.h>
47 #include <sys/acct.h>
48 #include <sys/file_internal.h>
49 #include <sys/uio.h>
50 #include <sys/kernel.h>
51 #include <sys/stat.h>
52 #include <sys/disk.h>
53 #include <sys/conf.h>
54 #include <sys/content_protection.h>
55 #include <sys/fsctl.h>
56
57 #include <mach-o/loader.h>
58 #include <mach-o/nlist.h>
59
60 #include <kern/kalloc.h>
61 #include <vm/vm_kern.h>
62 #include <pexpert/pexpert.h>
63 #include <IOKit/IOPolledInterface.h>
64
65 #define HIBERNATE_MIN_PHYSICAL_LBA_512 (34)
66 #define HIBERNATE_MIN_PHYSICAL_LBA_4096 (6)
67 #define HIBERNATE_MIN_FILE_SIZE (1024*1024)
68
69 /* This function is called from kern_sysctl in the current process context;
70 * it is exported with the System6.0.exports, but this appears to be a legacy
71 * export, as there are no internal consumers.
72 */
73 int
74 get_kernel_symfile(__unused proc_t p, __unused char const **symfile);
75 int
76 get_kernel_symfile(__unused proc_t p, __unused char const **symfile)
77 {
78 return KERN_FAILURE;
79 }
80
81 struct kern_direct_file_io_ref_t {
82 vfs_context_t ctx;
83 struct vnode * vp;
84 dev_t device;
85 uint32_t blksize;
86 off_t filelength;
87 char cf;
88 char pinned;
89 char frozen;
90 char wbcranged;
91 };
92
93
94 static int
95 file_ioctl(void * p1, void * p2, u_long theIoctl, caddr_t result)
96 {
97 dev_t device = *(dev_t*) p1;
98
99 return (*bdevsw[major(device)].d_ioctl)
100 (device, theIoctl, result, S_IFBLK, p2);
101 }
102
103 static int
104 device_ioctl(void * p1, __unused void * p2, u_long theIoctl, caddr_t result)
105 {
106 return VNOP_IOCTL(p1, theIoctl, result, 0, p2);
107 }
108
109 static int
110 kern_ioctl_file_extents(struct kern_direct_file_io_ref_t * ref, u_long theIoctl, off_t offset, off_t end)
111 {
112 int error = 0;
113 int (*do_ioctl)(void * p1, void * p2, u_long theIoctl, caddr_t result);
114 void * p1;
115 void * p2;
116 uint64_t fileblk;
117 size_t filechunk;
118 dk_extent_t extent;
119 dk_unmap_t unmap;
120 _dk_cs_pin_t pin;
121
122 bzero(&extent, sizeof(dk_extent_t));
123 bzero(&unmap, sizeof(dk_unmap_t));
124 bzero(&pin, sizeof(pin));
125 if (ref->vp->v_type == VREG) {
126 p1 = &ref->device;
127 p2 = kernproc;
128 do_ioctl = &file_ioctl;
129 } else {
130 /* Partition. */
131 p1 = ref->vp;
132 p2 = ref->ctx;
133 do_ioctl = &device_ioctl;
134 }
135
136 if (_DKIOCCSPINEXTENT == theIoctl) {
137 /* Tell CS the image size, so it knows whether to place the subsequent pins SSD/HDD */
138 pin.cp_extent.length = end;
139 pin.cp_flags = _DKIOCCSHIBERNATEIMGSIZE;
140 (void) do_ioctl(p1, p2, _DKIOCCSPINEXTENT, (caddr_t)&pin);
141 } else if (_DKIOCCSUNPINEXTENT == theIoctl) {
142 /* Tell CS hibernation is done, so it can stop blocking overlapping writes */
143 pin.cp_flags = _DKIOCCSPINDISCARDBLACKLIST;
144 (void) do_ioctl(p1, p2, _DKIOCCSUNPINEXTENT, (caddr_t)&pin);
145 }
146
147 for (; offset < end; offset += filechunk) {
148 if (ref->vp->v_type == VREG) {
149 daddr64_t blkno;
150 filechunk = 1 * 1024 * 1024 * 1024;
151 if (filechunk > (size_t)(end - offset)) {
152 filechunk = (size_t)(end - offset);
153 }
154 error = VNOP_BLOCKMAP(ref->vp, offset, filechunk, &blkno,
155 &filechunk, NULL, VNODE_WRITE | VNODE_BLOCKMAP_NO_TRACK, NULL);
156 if (error) {
157 break;
158 }
159 if (-1LL == blkno) {
160 continue;
161 }
162 fileblk = blkno * ref->blksize;
163 } else if ((ref->vp->v_type == VBLK) || (ref->vp->v_type == VCHR)) {
164 fileblk = offset;
165 filechunk = (unsigned long)((ref->filelength > ULONG_MAX) ? ULONG_MAX: ref->filelength);
166 }
167
168 if (DKIOCUNMAP == theIoctl) {
169 extent.offset = fileblk;
170 extent.length = filechunk;
171 unmap.extents = &extent;
172 unmap.extentsCount = 1;
173 error = do_ioctl(p1, p2, theIoctl, (caddr_t)&unmap);
174 // printf("DKIOCUNMAP(%d) 0x%qx, 0x%qx\n", error, extent.offset, extent.length);
175 } else if (_DKIOCCSPINEXTENT == theIoctl) {
176 pin.cp_extent.offset = fileblk;
177 pin.cp_extent.length = filechunk;
178 pin.cp_flags = _DKIOCCSPINFORHIBERNATION;
179 error = do_ioctl(p1, p2, theIoctl, (caddr_t)&pin);
180 if (error && (ENOTTY != error)) {
181 printf("_DKIOCCSPINEXTENT(%d) 0x%qx, 0x%qx\n", error, pin.cp_extent.offset, pin.cp_extent.length);
182 }
183 } else if (_DKIOCCSUNPINEXTENT == theIoctl) {
184 pin.cp_extent.offset = fileblk;
185 pin.cp_extent.length = filechunk;
186 pin.cp_flags = _DKIOCCSPINFORHIBERNATION;
187 error = do_ioctl(p1, p2, theIoctl, (caddr_t)&pin);
188 if (error && (ENOTTY != error)) {
189 printf("_DKIOCCSUNPINEXTENT(%d) 0x%qx, 0x%qx\n", error, pin.cp_extent.offset, pin.cp_extent.length);
190 }
191 } else {
192 error = EINVAL;
193 }
194
195 if (error) {
196 break;
197 }
198 }
199 return error;
200 }
201
202 extern uint32_t freespace_mb(vnode_t vp);
203
204 struct kern_direct_file_io_ref_t *
205 kern_open_file_for_direct_io(const char * name,
206 uint32_t iflags,
207 kern_get_file_extents_callback_t callback,
208 void * callback_ref,
209 off_t set_file_size,
210 off_t fs_free_size,
211 off_t write_file_offset,
212 void * write_file_addr,
213 size_t write_file_len,
214 dev_t * partition_device_result,
215 dev_t * image_device_result,
216 uint64_t * partitionbase_result,
217 uint64_t * maxiocount_result,
218 uint32_t * oflags)
219 {
220 struct kern_direct_file_io_ref_t * ref;
221
222 proc_t p;
223 struct vnode_attr va;
224 dk_apfs_wbc_range_t wbc_range;
225 int error;
226 off_t f_offset;
227 uint64_t fileblk;
228 size_t filechunk;
229 uint64_t physoffset, minoffset;
230 dev_t device;
231 dev_t target = 0;
232 int isssd = 0;
233 uint32_t flags = 0;
234 uint32_t blksize;
235 off_t maxiocount, count, segcount, wbctotal;
236 boolean_t locked = FALSE;
237 int fmode;
238 mode_t cmode;
239 struct nameidata nd;
240 u_int32_t ndflags;
241 off_t mpFree;
242
243 int (*do_ioctl)(void * p1, void * p2, u_long theIoctl, caddr_t result);
244 void * p1 = NULL;
245 void * p2 = NULL;
246
247 error = EFAULT;
248
249 ref = (struct kern_direct_file_io_ref_t *) kalloc(sizeof(struct kern_direct_file_io_ref_t));
250 if (!ref) {
251 error = EFAULT;
252 goto out;
253 }
254
255 bzero(ref, sizeof(*ref));
256 p = kernproc;
257 ref->ctx = vfs_context_kernel();
258
259 fmode = (kIOPolledFileCreate & iflags) ? (O_CREAT | FWRITE) : FWRITE;
260 cmode = S_IRUSR | S_IWUSR;
261 ndflags = NOFOLLOW;
262 NDINIT(&nd, LOOKUP, OP_OPEN, ndflags, UIO_SYSSPACE, CAST_USER_ADDR_T(name), ref->ctx);
263 VATTR_INIT(&va);
264 VATTR_SET(&va, va_mode, cmode);
265 VATTR_SET(&va, va_dataprotect_flags, VA_DP_RAWENCRYPTED);
266 VATTR_SET(&va, va_dataprotect_class, PROTECTION_CLASS_D);
267 if ((error = vn_open_auth(&nd, &fmode, &va))) {
268 kprintf("vn_open_auth(fmode: %d, cmode: %d) failed with error: %d\n", fmode, cmode, error);
269 goto out;
270 }
271
272 ref->vp = nd.ni_vp;
273 if (ref->vp->v_type == VREG) {
274 vnode_lock_spin(ref->vp);
275 SET(ref->vp->v_flag, VSWAP);
276 vnode_unlock(ref->vp);
277 }
278
279 if (write_file_addr && write_file_len) {
280 if ((error = kern_write_file(ref, write_file_offset, write_file_addr, write_file_len, IO_SKIP_ENCRYPTION))) {
281 kprintf("kern_write_file() failed with error: %d\n", error);
282 goto out;
283 }
284 }
285
286 VATTR_INIT(&va);
287 VATTR_WANTED(&va, va_rdev);
288 VATTR_WANTED(&va, va_fsid);
289 VATTR_WANTED(&va, va_devid);
290 VATTR_WANTED(&va, va_data_size);
291 VATTR_WANTED(&va, va_data_alloc);
292 VATTR_WANTED(&va, va_nlink);
293 error = EFAULT;
294 if (vnode_getattr(ref->vp, &va, ref->ctx)) {
295 goto out;
296 }
297
298 wbctotal = 0;
299 mpFree = freespace_mb(ref->vp);
300 mpFree <<= 20;
301 kprintf("kern_direct_file(%s): vp size %qd, alloc %qd, mp free %qd, keep free %qd\n",
302 name, va.va_data_size, va.va_data_alloc, mpFree, fs_free_size);
303
304 if (ref->vp->v_type == VREG) {
305 /* Don't dump files with links. */
306 if (va.va_nlink != 1) {
307 goto out;
308 }
309
310 device = (VATTR_IS_SUPPORTED(&va, va_devid)) ? va.va_devid : va.va_fsid;
311 ref->filelength = va.va_data_size;
312
313 p1 = &device;
314 p2 = p;
315 do_ioctl = &file_ioctl;
316
317 if (kIOPolledFileHibernate & iflags) {
318 error = do_ioctl(p1, p2, DKIOCAPFSGETWBCRANGE, (caddr_t) &wbc_range);
319 ref->wbcranged = (error == 0);
320 }
321 if (ref->wbcranged) {
322 uint32_t idx;
323 assert(wbc_range.count <= (sizeof(wbc_range.extents) / sizeof(wbc_range.extents[0])));
324 for (idx = 0; idx < wbc_range.count; idx++) {
325 wbctotal += wbc_range.extents[idx].length;
326 }
327 kprintf("kern_direct_file(%s): wbc %qd\n", name, wbctotal);
328 if (wbctotal) {
329 target = wbc_range.dev;
330 }
331 }
332
333 if (set_file_size) {
334 if (wbctotal) {
335 if (wbctotal >= set_file_size) {
336 set_file_size = HIBERNATE_MIN_FILE_SIZE;
337 } else {
338 set_file_size -= wbctotal;
339 if (set_file_size < HIBERNATE_MIN_FILE_SIZE) {
340 set_file_size = HIBERNATE_MIN_FILE_SIZE;
341 }
342 }
343 }
344 if (fs_free_size) {
345 mpFree += va.va_data_alloc;
346 if ((mpFree < set_file_size) || ((mpFree - set_file_size) < fs_free_size)) {
347 error = ENOSPC;
348 goto out;
349 }
350 }
351 error = vnode_setsize(ref->vp, set_file_size, IO_NOZEROFILL | IO_NOAUTH, ref->ctx);
352 if (error) {
353 goto out;
354 }
355 ref->filelength = set_file_size;
356 }
357 } else if ((ref->vp->v_type == VBLK) || (ref->vp->v_type == VCHR)) {
358 /* Partition. */
359 device = va.va_rdev;
360
361 p1 = ref->vp;
362 p2 = ref->ctx;
363 do_ioctl = &device_ioctl;
364 } else {
365 /* Don't dump to non-regular files. */
366 error = EFAULT;
367 goto out;
368 }
369 ref->device = device;
370
371 // probe for CF
372 dk_corestorage_info_t cs_info;
373 memset(&cs_info, 0, sizeof(dk_corestorage_info_t));
374 error = do_ioctl(p1, p2, DKIOCCORESTORAGE, (caddr_t)&cs_info);
375 ref->cf = (error == 0) && (cs_info.flags & DK_CORESTORAGE_ENABLE_HOTFILES);
376
377 // get block size
378
379 error = do_ioctl(p1, p2, DKIOCGETBLOCKSIZE, (caddr_t) &ref->blksize);
380 if (error) {
381 goto out;
382 }
383
384 if (ref->blksize == 4096) {
385 minoffset = HIBERNATE_MIN_PHYSICAL_LBA_4096 * ref->blksize;
386 } else {
387 minoffset = HIBERNATE_MIN_PHYSICAL_LBA_512 * ref->blksize;
388 }
389
390 if (ref->vp->v_type != VREG) {
391 error = do_ioctl(p1, p2, DKIOCGETBLOCKCOUNT, (caddr_t) &fileblk);
392 if (error) {
393 goto out;
394 }
395 ref->filelength = fileblk * ref->blksize;
396 }
397
398 // pin logical extents, CS version
399
400 error = kern_ioctl_file_extents(ref, _DKIOCCSPINEXTENT, 0, ref->filelength);
401 if (error && (ENOTTY != error)) {
402 goto out;
403 }
404 ref->pinned = (error == 0);
405
406 // pin logical extents, apfs version
407
408 error = VNOP_IOCTL(ref->vp, FSCTL_FREEZE_EXTENTS, NULL, 0, ref->ctx);
409 if (error && (ENOTTY != error)) {
410 goto out;
411 }
412 ref->frozen = (error == 0);
413
414 // generate the block list
415
416 error = do_ioctl(p1, p2, DKIOCLOCKPHYSICALEXTENTS, NULL);
417 if (error) {
418 goto out;
419 }
420 locked = TRUE;
421
422 f_offset = 0;
423 for (; f_offset < ref->filelength; f_offset += filechunk) {
424 if (ref->vp->v_type == VREG) {
425 filechunk = 1 * 1024 * 1024 * 1024;
426 daddr64_t blkno;
427
428 error = VNOP_BLOCKMAP(ref->vp, f_offset, filechunk, &blkno,
429 &filechunk, NULL, VNODE_WRITE | VNODE_BLOCKMAP_NO_TRACK, NULL);
430 if (error) {
431 goto out;
432 }
433 if (-1LL == blkno) {
434 continue;
435 }
436 fileblk = blkno * ref->blksize;
437 } else if ((ref->vp->v_type == VBLK) || (ref->vp->v_type == VCHR)) {
438 fileblk = f_offset;
439 filechunk = f_offset ? 0 : (unsigned long)ref->filelength;
440 }
441
442 physoffset = 0;
443 while (physoffset < filechunk) {
444 dk_physical_extent_t getphysreq;
445 bzero(&getphysreq, sizeof(getphysreq));
446
447 getphysreq.offset = fileblk + physoffset;
448 getphysreq.length = (filechunk - physoffset);
449 error = do_ioctl(p1, p2, DKIOCGETPHYSICALEXTENT, (caddr_t) &getphysreq);
450 if (error) {
451 goto out;
452 }
453 if (!target) {
454 target = getphysreq.dev;
455 } else if (target != getphysreq.dev) {
456 error = ENOTSUP;
457 goto out;
458 }
459
460 assert(getphysreq.offset >= minoffset);
461
462 #if HIBFRAGMENT
463 uint64_t rev;
464 for (rev = 4096; rev <= getphysreq.length; rev += 4096) {
465 callback(callback_ref, getphysreq.offset + getphysreq.length - rev, 4096);
466 }
467 #else
468 callback(callback_ref, getphysreq.offset, getphysreq.length);
469 #endif
470 physoffset += getphysreq.length;
471 }
472 }
473 if (ref->wbcranged) {
474 uint32_t idx;
475 for (idx = 0; idx < wbc_range.count; idx++) {
476 assert(wbc_range.extents[idx].offset >= minoffset);
477 callback(callback_ref, wbc_range.extents[idx].offset, wbc_range.extents[idx].length);
478 }
479 }
480 callback(callback_ref, 0ULL, 0ULL);
481
482 if (ref->vp->v_type == VREG) {
483 p1 = &target;
484 } else {
485 p1 = &target;
486 p2 = p;
487 do_ioctl = &file_ioctl;
488 }
489
490 // get partition base
491
492 if (partitionbase_result) {
493 error = do_ioctl(p1, p2, DKIOCGETBASE, (caddr_t) partitionbase_result);
494 if (error) {
495 goto out;
496 }
497 }
498
499 // get block size & constraints
500
501 error = do_ioctl(p1, p2, DKIOCGETBLOCKSIZE, (caddr_t) &blksize);
502 if (error) {
503 goto out;
504 }
505
506 maxiocount = 1 * 1024 * 1024 * 1024;
507
508 error = do_ioctl(p1, p2, DKIOCGETMAXBLOCKCOUNTREAD, (caddr_t) &count);
509 if (error) {
510 count = 0;
511 }
512 count *= blksize;
513 if (count && (count < maxiocount)) {
514 maxiocount = count;
515 }
516
517 error = do_ioctl(p1, p2, DKIOCGETMAXBLOCKCOUNTWRITE, (caddr_t) &count);
518 if (error) {
519 count = 0;
520 }
521 count *= blksize;
522 if (count && (count < maxiocount)) {
523 maxiocount = count;
524 }
525
526 error = do_ioctl(p1, p2, DKIOCGETMAXBYTECOUNTREAD, (caddr_t) &count);
527 if (error) {
528 count = 0;
529 }
530 if (count && (count < maxiocount)) {
531 maxiocount = count;
532 }
533
534 error = do_ioctl(p1, p2, DKIOCGETMAXBYTECOUNTWRITE, (caddr_t) &count);
535 if (error) {
536 count = 0;
537 }
538 if (count && (count < maxiocount)) {
539 maxiocount = count;
540 }
541
542 error = do_ioctl(p1, p2, DKIOCGETMAXSEGMENTBYTECOUNTREAD, (caddr_t) &count);
543 if (!error) {
544 error = do_ioctl(p1, p2, DKIOCGETMAXSEGMENTCOUNTREAD, (caddr_t) &segcount);
545 }
546 if (error) {
547 count = segcount = 0;
548 }
549 count *= segcount;
550 if (count && (count < maxiocount)) {
551 maxiocount = count;
552 }
553
554 error = do_ioctl(p1, p2, DKIOCGETMAXSEGMENTBYTECOUNTWRITE, (caddr_t) &count);
555 if (!error) {
556 error = do_ioctl(p1, p2, DKIOCGETMAXSEGMENTCOUNTWRITE, (caddr_t) &segcount);
557 }
558 if (error) {
559 count = segcount = 0;
560 }
561 count *= segcount;
562 if (count && (count < maxiocount)) {
563 maxiocount = count;
564 }
565
566 kprintf("max io 0x%qx bytes\n", maxiocount);
567 if (maxiocount_result) {
568 *maxiocount_result = maxiocount;
569 }
570
571 error = do_ioctl(p1, p2, DKIOCISSOLIDSTATE, (caddr_t)&isssd);
572 if (!error && isssd) {
573 flags |= kIOPolledFileSSD;
574 }
575
576 if (partition_device_result) {
577 *partition_device_result = device;
578 }
579 if (image_device_result) {
580 *image_device_result = target;
581 }
582 if (oflags) {
583 *oflags = flags;
584 }
585
586 if ((ref->vp->v_type == VBLK) || (ref->vp->v_type == VCHR)) {
587 vnode_close(ref->vp, FWRITE, ref->ctx);
588 ref->vp = NULLVP;
589 ref->ctx = NULL;
590 }
591
592 out:
593 printf("kern_open_file_for_direct_io(%p, %d)\n", ref, error);
594
595
596 if (error && locked) {
597 p1 = &device;
598 (void) do_ioctl(p1, p2, DKIOCUNLOCKPHYSICALEXTENTS, NULL);
599 }
600
601 if (error && ref) {
602 if (ref->vp) {
603 (void) kern_ioctl_file_extents(ref, _DKIOCCSUNPINEXTENT, 0, (ref->pinned && ref->cf) ? ref->filelength : 0);
604
605 if (ref->frozen) {
606 (void) VNOP_IOCTL(ref->vp, FSCTL_THAW_EXTENTS, NULL, 0, ref->ctx);
607 }
608 if (ref->wbcranged) {
609 (void) do_ioctl(p1, p2, DKIOCAPFSRELEASEWBCRANGE, (caddr_t) NULL);
610 }
611 vnode_close(ref->vp, FWRITE, ref->ctx);
612 ref->vp = NULLVP;
613 }
614 ref->ctx = NULL;
615 kfree(ref, sizeof(struct kern_direct_file_io_ref_t));
616 ref = NULL;
617 }
618
619 return ref;
620 }
621
622 int
623 kern_write_file(struct kern_direct_file_io_ref_t * ref, off_t offset, void * addr, size_t len, int ioflag)
624 {
625 assert(len <= INT32_MAX);
626 return vn_rdwr(UIO_WRITE, ref->vp,
627 addr, (int)len, offset,
628 UIO_SYSSPACE, ioflag | IO_SYNC | IO_NODELOCKED | IO_UNIT,
629 vfs_context_ucred(ref->ctx), (int *) 0,
630 vfs_context_proc(ref->ctx));
631 }
632
633 int
634 kern_read_file(struct kern_direct_file_io_ref_t * ref, off_t offset, void * addr, size_t len, int ioflag)
635 {
636 assert(len <= INT32_MAX);
637 return vn_rdwr(UIO_READ, ref->vp,
638 addr, (int)len, offset,
639 UIO_SYSSPACE, ioflag | IO_SYNC | IO_NODELOCKED | IO_UNIT,
640 vfs_context_ucred(ref->ctx), (int *) 0,
641 vfs_context_proc(ref->ctx));
642 }
643
644
645 struct mount *
646 kern_file_mount(struct kern_direct_file_io_ref_t * ref)
647 {
648 return ref->vp->v_mount;
649 }
650
651 void
652 kern_close_file_for_direct_io(struct kern_direct_file_io_ref_t * ref,
653 off_t write_offset, void * addr, size_t write_length,
654 off_t discard_offset, off_t discard_end)
655 {
656 int error;
657 printf("kern_close_file_for_direct_io(%p)\n", ref);
658
659 if (!ref) {
660 return;
661 }
662
663 if (ref->vp) {
664 int (*do_ioctl)(void * p1, void * p2, u_long theIoctl, caddr_t result);
665 void * p1;
666 void * p2;
667
668 discard_offset = ((discard_offset + ref->blksize - 1) & ~(((off_t) ref->blksize) - 1));
669 discard_end = ((discard_end) & ~(((off_t) ref->blksize) - 1));
670
671 if (ref->vp->v_type == VREG) {
672 p1 = &ref->device;
673 p2 = kernproc;
674 do_ioctl = &file_ioctl;
675 } else {
676 /* Partition. */
677 p1 = ref->vp;
678 p2 = ref->ctx;
679 do_ioctl = &device_ioctl;
680 }
681 (void) do_ioctl(p1, p2, DKIOCUNLOCKPHYSICALEXTENTS, NULL);
682
683 //XXX If unmapping extents then don't also need to unpin; except ...
684 //XXX if file unaligned (HFS 4k / Fusion 128k) then pin is superset and
685 //XXX unmap is subset, so save extra walk over file extents (and the risk
686 //XXX that CF drain starts) vs leaving partial units pinned to SSD
687 //XXX (until whatever was sharing also unmaps). Err on cleaning up fully.
688 boolean_t will_unmap = (!ref->pinned || ref->cf) && (discard_end > discard_offset);
689 boolean_t will_unpin = (ref->pinned && ref->cf /* && !will_unmap */);
690
691 (void) kern_ioctl_file_extents(ref, _DKIOCCSUNPINEXTENT, 0, (will_unpin) ? ref->filelength : 0);
692
693 if (will_unmap) {
694 (void) kern_ioctl_file_extents(ref, DKIOCUNMAP, discard_offset, (ref->cf) ? ref->filelength : discard_end);
695 }
696
697 if (ref->frozen) {
698 (void) VNOP_IOCTL(ref->vp, FSCTL_THAW_EXTENTS, NULL, 0, ref->ctx);
699 }
700 if (ref->wbcranged) {
701 (void) do_ioctl(p1, p2, DKIOCAPFSRELEASEWBCRANGE, (caddr_t) NULL);
702 }
703
704 if (addr && write_length) {
705 (void) kern_write_file(ref, write_offset, addr, write_length, IO_SKIP_ENCRYPTION);
706 }
707
708 error = vnode_close(ref->vp, FWRITE, ref->ctx);
709
710 ref->vp = NULLVP;
711 kprintf("vnode_close(%d)\n", error);
712
713 }
714
715 ref->ctx = NULL;
716
717 kfree(ref, sizeof(struct kern_direct_file_io_ref_t));
718 }