file.c 25.7 KB
Newer Older
1 2 3 4 5 6 7 8 9
/*
 * fs/kernfs/file.c - kernfs file implementation
 *
 * Copyright (c) 2001-3 Patrick Mochel
 * Copyright (c) 2007 SUSE Linux Products GmbH
 * Copyright (c) 2007, 2013 Tejun Heo <tj@kernel.org>
 *
 * This file is released under the GPLv2.
 */
10 11 12 13 14 15

#include <linux/fs.h>
#include <linux/seq_file.h>
#include <linux/slab.h>
#include <linux/poll.h>
#include <linux/pagemap.h>
16
#include <linux/sched/mm.h>
17
#include <linux/fsnotify.h>
18 19 20 21

#include "kernfs-internal.h"

/*
22
 * There's one kernfs_open_file for each open file and one kernfs_open_node
23
 * for each kernfs_node with one or more open files.
24
 *
25 26
 * kernfs_node->attr.open points to kernfs_open_node.  attr.open is
 * protected by kernfs_open_node_lock.
27 28
 *
 * filp->private_data points to seq_file whose ->private points to
29 30
 * kernfs_open_file.  kernfs_open_files are chained at
 * kernfs_open_node->files, which is protected by kernfs_open_file_mutex.
31
 */
32 33
static DEFINE_SPINLOCK(kernfs_open_node_lock);
static DEFINE_MUTEX(kernfs_open_file_mutex);
34

35
struct kernfs_open_node {
36 37 38
	atomic_t		refcnt;
	atomic_t		event;
	wait_queue_head_t	poll;
39
	struct list_head	files; /* goes through kernfs_open_file.list */
40 41
};

42 43 44 45 46 47 48 49 50 51 52 53 54
/*
 * kernfs_notify() may be called from any context and bounces notifications
 * through a work item.  To minimize space overhead in kernfs_node, the
 * pending queue is implemented as a singly linked list of kernfs_nodes.
 * The list is terminated with the self pointer so that whether a
 * kernfs_node is on the list or not can be determined by testing the next
 * pointer for NULL.
 */
#define KERNFS_NOTIFY_EOL			((void *)&kernfs_notify_list)

static DEFINE_SPINLOCK(kernfs_notify_lock);
static struct kernfs_node *kernfs_notify_list = KERNFS_NOTIFY_EOL;

55
static struct kernfs_open_file *kernfs_of(struct file *file)
56 57 58 59 60
{
	return ((struct seq_file *)file->private_data)->private;
}

/*
61
 * Determine the kernfs_ops for the given kernfs_node.  This function must
62 63
 * be called while holding an active reference.
 */
64
static const struct kernfs_ops *kernfs_ops(struct kernfs_node *kn)
65
{
66
	if (kn->flags & KERNFS_LOCKDEP)
67
		lockdep_assert_held(kn);
68
	return kn->attr.ops;
69 70
}

71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102
/*
 * As kernfs_seq_stop() is also called after kernfs_seq_start() or
 * kernfs_seq_next() failure, it needs to distinguish whether it's stopping
 * a seq_file iteration which is fully initialized with an active reference
 * or an aborted kernfs_seq_start() due to get_active failure.  The
 * position pointer is the only context for each seq_file iteration and
 * thus the stop condition should be encoded in it.  As the return value is
 * directly visible to userland, ERR_PTR(-ENODEV) is the only acceptable
 * choice to indicate get_active failure.
 *
 * Unfortunately, this is complicated due to the optional custom seq_file
 * operations which may return ERR_PTR(-ENODEV) too.  kernfs_seq_stop()
 * can't distinguish whether ERR_PTR(-ENODEV) is from get_active failure or
 * custom seq_file operations and thus can't decide whether put_active
 * should be performed or not only on ERR_PTR(-ENODEV).
 *
 * This is worked around by factoring out the custom seq_stop() and
 * put_active part into kernfs_seq_stop_active(), skipping it from
 * kernfs_seq_stop() if ERR_PTR(-ENODEV) while invoking it directly after
 * custom seq_file operations fail with ERR_PTR(-ENODEV) - this ensures
 * that kernfs_seq_stop_active() is skipped only after get_active failure.
 */
static void kernfs_seq_stop_active(struct seq_file *sf, void *v)
{
	struct kernfs_open_file *of = sf->private;
	const struct kernfs_ops *ops = kernfs_ops(of->kn);

	if (ops->seq_stop)
		ops->seq_stop(sf, v);
	kernfs_put_active(of->kn);
}

103 104
static void *kernfs_seq_start(struct seq_file *sf, loff_t *ppos)
{
105
	struct kernfs_open_file *of = sf->private;
106 107 108
	const struct kernfs_ops *ops;

	/*
109
	 * @of->mutex nests outside active ref and is primarily to ensure that
110 111 112
	 * the ops aren't called concurrently for the same open file.
	 */
	mutex_lock(&of->mutex);
113
	if (!kernfs_get_active(of->kn))
114 115
		return ERR_PTR(-ENODEV);

116
	ops = kernfs_ops(of->kn);
117
	if (ops->seq_start) {
118 119 120 121 122
		void *next = ops->seq_start(sf, ppos);
		/* see the comment above kernfs_seq_stop_active() */
		if (next == ERR_PTR(-ENODEV))
			kernfs_seq_stop_active(sf, next);
		return next;
123 124 125 126 127 128 129 130 131 132 133
	} else {
		/*
		 * The same behavior and code as single_open().  Returns
		 * !NULL if pos is at the beginning; otherwise, NULL.
		 */
		return NULL + !*ppos;
	}
}

static void *kernfs_seq_next(struct seq_file *sf, void *v, loff_t *ppos)
{
134
	struct kernfs_open_file *of = sf->private;
135
	const struct kernfs_ops *ops = kernfs_ops(of->kn);
136 137

	if (ops->seq_next) {
138 139 140 141 142
		void *next = ops->seq_next(sf, v, ppos);
		/* see the comment above kernfs_seq_stop_active() */
		if (next == ERR_PTR(-ENODEV))
			kernfs_seq_stop_active(sf, next);
		return next;
143 144 145 146 147 148 149 150 151 152 153 154
	} else {
		/*
		 * The same behavior and code as single_open(), always
		 * terminate after the initial read.
		 */
		++*ppos;
		return NULL;
	}
}

static void kernfs_seq_stop(struct seq_file *sf, void *v)
{
155
	struct kernfs_open_file *of = sf->private;
156

157 158
	if (v != ERR_PTR(-ENODEV))
		kernfs_seq_stop_active(sf, v);
159 160 161 162 163
	mutex_unlock(&of->mutex);
}

static int kernfs_seq_show(struct seq_file *sf, void *v)
{
164
	struct kernfs_open_file *of = sf->private;
165

166
	of->event = atomic_read(&of->kn->attr.open->event);
167

168
	return of->kn->attr.ops->seq_show(sf, v);
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
}

static const struct seq_operations kernfs_seq_ops = {
	.start = kernfs_seq_start,
	.next = kernfs_seq_next,
	.stop = kernfs_seq_stop,
	.show = kernfs_seq_show,
};

/*
 * As reading a bin file can have side-effects, the exact offset and bytes
 * specified in read(2) call should be passed to the read callback making
 * it difficult to use seq_file.  Implement simplistic custom buffering for
 * bin files.
 */
184
static ssize_t kernfs_file_direct_read(struct kernfs_open_file *of,
185 186 187 188 189 190 191
				       char __user *user_buf, size_t count,
				       loff_t *ppos)
{
	ssize_t len = min_t(size_t, count, PAGE_SIZE);
	const struct kernfs_ops *ops;
	char *buf;

192
	buf = of->prealloc_buf;
193 194 195
	if (buf)
		mutex_lock(&of->prealloc_mutex);
	else
196
		buf = kmalloc(len, GFP_KERNEL);
197 198 199 200
	if (!buf)
		return -ENOMEM;

	/*
201
	 * @of->mutex nests outside active ref and is used both to ensure that
202
	 * the ops aren't called concurrently for the same open file.
203 204
	 */
	mutex_lock(&of->mutex);
205
	if (!kernfs_get_active(of->kn)) {
206 207 208 209 210
		len = -ENODEV;
		mutex_unlock(&of->mutex);
		goto out_free;
	}

211
	of->event = atomic_read(&of->kn->attr.open->event);
212
	ops = kernfs_ops(of->kn);
213 214 215 216 217
	if (ops->read)
		len = ops->read(of, buf, len, *ppos);
	else
		len = -EINVAL;

218 219 220
	kernfs_put_active(of->kn);
	mutex_unlock(&of->mutex);

221
	if (len < 0)
222
		goto out_free;
223 224 225

	if (copy_to_user(user_buf, buf, len)) {
		len = -EFAULT;
226
		goto out_free;
227 228 229 230 231
	}

	*ppos += len;

 out_free:
232 233 234
	if (buf == of->prealloc_buf)
		mutex_unlock(&of->prealloc_mutex);
	else
235
		kfree(buf);
236 237 238 239
	return len;
}

/**
240
 * kernfs_fop_read - kernfs vfs read callback
241 242 243 244 245
 * @file: file pointer
 * @user_buf: data to write
 * @count: number of bytes
 * @ppos: starting offset
 */
246 247
static ssize_t kernfs_fop_read(struct file *file, char __user *user_buf,
			       size_t count, loff_t *ppos)
248
{
249
	struct kernfs_open_file *of = kernfs_of(file);
250

251
	if (of->kn->flags & KERNFS_HAS_SEQ_SHOW)
252 253 254 255 256 257
		return seq_read(file, user_buf, count, ppos);
	else
		return kernfs_file_direct_read(of, user_buf, count, ppos);
}

/**
258
 * kernfs_fop_write - kernfs vfs write callback
259 260 261 262 263 264 265 266 267 268 269 270 271 272
 * @file: file pointer
 * @user_buf: data to write
 * @count: number of bytes
 * @ppos: starting offset
 *
 * Copy data in from userland and pass it to the matching kernfs write
 * operation.
 *
 * There is no easy way for us to know if userspace is only doing a partial
 * write, so we don't support them. We expect the entire buffer to come on
 * the first write.  Hint: if you're writing a value, first read the file,
 * modify only the the value you're changing, then write entire buffer
 * back.
 */
273 274
static ssize_t kernfs_fop_write(struct file *file, const char __user *user_buf,
				size_t count, loff_t *ppos)
275
{
276
	struct kernfs_open_file *of = kernfs_of(file);
277
	const struct kernfs_ops *ops;
278
	ssize_t len;
279
	char *buf;
280

281
	if (of->atomic_write_len) {
282
		len = count;
283 284
		if (len > of->atomic_write_len)
			return -E2BIG;
285 286 287 288
	} else {
		len = min_t(size_t, count, PAGE_SIZE);
	}

289
	buf = of->prealloc_buf;
290 291 292
	if (buf)
		mutex_lock(&of->prealloc_mutex);
	else
293
		buf = kmalloc(len + 1, GFP_KERNEL);
294 295
	if (!buf)
		return -ENOMEM;
296

297 298 299 300 301 302
	if (copy_from_user(buf, user_buf, len)) {
		len = -EFAULT;
		goto out_free;
	}
	buf[len] = '\0';	/* guarantee string termination */

303
	/*
304
	 * @of->mutex nests outside active ref and is used both to ensure that
305
	 * the ops aren't called concurrently for the same open file.
306 307 308 309 310 311 312 313 314 315 316 317 318 319
	 */
	mutex_lock(&of->mutex);
	if (!kernfs_get_active(of->kn)) {
		mutex_unlock(&of->mutex);
		len = -ENODEV;
		goto out_free;
	}

	ops = kernfs_ops(of->kn);
	if (ops->write)
		len = ops->write(of, buf, len, *ppos);
	else
		len = -EINVAL;

320 321 322
	kernfs_put_active(of->kn);
	mutex_unlock(&of->mutex);

323 324
	if (len > 0)
		*ppos += len;
325

326
out_free:
327 328 329
	if (buf == of->prealloc_buf)
		mutex_unlock(&of->prealloc_mutex);
	else
330
		kfree(buf);
331 332 333 334 335 336
	return len;
}

static void kernfs_vma_open(struct vm_area_struct *vma)
{
	struct file *file = vma->vm_file;
337
	struct kernfs_open_file *of = kernfs_of(file);
338 339 340 341

	if (!of->vm_ops)
		return;

342
	if (!kernfs_get_active(of->kn))
343 344 345 346 347
		return;

	if (of->vm_ops->open)
		of->vm_ops->open(vma);

348
	kernfs_put_active(of->kn);
349 350
}

351
static vm_fault_t kernfs_vma_fault(struct vm_fault *vmf)
352
{
353
	struct file *file = vmf->vma->vm_file;
354
	struct kernfs_open_file *of = kernfs_of(file);
355
	vm_fault_t ret;
356 357 358 359

	if (!of->vm_ops)
		return VM_FAULT_SIGBUS;

360
	if (!kernfs_get_active(of->kn))
361 362 363 364
		return VM_FAULT_SIGBUS;

	ret = VM_FAULT_SIGBUS;
	if (of->vm_ops->fault)
365
		ret = of->vm_ops->fault(vmf);
366

367
	kernfs_put_active(of->kn);
368 369 370
	return ret;
}

371
static vm_fault_t kernfs_vma_page_mkwrite(struct vm_fault *vmf)
372
{
373
	struct file *file = vmf->vma->vm_file;
374
	struct kernfs_open_file *of = kernfs_of(file);
375
	vm_fault_t ret;
376 377 378 379

	if (!of->vm_ops)
		return VM_FAULT_SIGBUS;

380
	if (!kernfs_get_active(of->kn))
381 382 383 384
		return VM_FAULT_SIGBUS;

	ret = 0;
	if (of->vm_ops->page_mkwrite)
385
		ret = of->vm_ops->page_mkwrite(vmf);
386 387 388
	else
		file_update_time(file);

389
	kernfs_put_active(of->kn);
390 391 392 393 394 395 396
	return ret;
}

static int kernfs_vma_access(struct vm_area_struct *vma, unsigned long addr,
			     void *buf, int len, int write)
{
	struct file *file = vma->vm_file;
397
	struct kernfs_open_file *of = kernfs_of(file);
398 399 400 401 402
	int ret;

	if (!of->vm_ops)
		return -EINVAL;

403
	if (!kernfs_get_active(of->kn))
404 405 406 407 408 409
		return -EINVAL;

	ret = -EINVAL;
	if (of->vm_ops->access)
		ret = of->vm_ops->access(vma, addr, buf, len, write);

410
	kernfs_put_active(of->kn);
411 412 413 414 415 416 417 418
	return ret;
}

#ifdef CONFIG_NUMA
static int kernfs_vma_set_policy(struct vm_area_struct *vma,
				 struct mempolicy *new)
{
	struct file *file = vma->vm_file;
419
	struct kernfs_open_file *of = kernfs_of(file);
420 421 422 423 424
	int ret;

	if (!of->vm_ops)
		return 0;

425
	if (!kernfs_get_active(of->kn))
426 427 428 429 430 431
		return -EINVAL;

	ret = 0;
	if (of->vm_ops->set_policy)
		ret = of->vm_ops->set_policy(vma, new);

432
	kernfs_put_active(of->kn);
433 434 435 436 437 438 439
	return ret;
}

static struct mempolicy *kernfs_vma_get_policy(struct vm_area_struct *vma,
					       unsigned long addr)
{
	struct file *file = vma->vm_file;
440
	struct kernfs_open_file *of = kernfs_of(file);
441 442 443 444 445
	struct mempolicy *pol;

	if (!of->vm_ops)
		return vma->vm_policy;

446
	if (!kernfs_get_active(of->kn))
447 448 449 450 451 452
		return vma->vm_policy;

	pol = vma->vm_policy;
	if (of->vm_ops->get_policy)
		pol = of->vm_ops->get_policy(vma, addr);

453
	kernfs_put_active(of->kn);
454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469
	return pol;
}

#endif

static const struct vm_operations_struct kernfs_vm_ops = {
	.open		= kernfs_vma_open,
	.fault		= kernfs_vma_fault,
	.page_mkwrite	= kernfs_vma_page_mkwrite,
	.access		= kernfs_vma_access,
#ifdef CONFIG_NUMA
	.set_policy	= kernfs_vma_set_policy,
	.get_policy	= kernfs_vma_get_policy,
#endif
};

470
static int kernfs_fop_mmap(struct file *file, struct vm_area_struct *vma)
471
{
472
	struct kernfs_open_file *of = kernfs_of(file);
473 474 475
	const struct kernfs_ops *ops;
	int rc;

476 477 478 479 480 481 482
	/*
	 * mmap path and of->mutex are prone to triggering spurious lockdep
	 * warnings and we don't want to add spurious locking dependency
	 * between the two.  Check whether mmap is actually implemented
	 * without grabbing @of->mutex by testing HAS_MMAP flag.  See the
	 * comment in kernfs_file_open() for more details.
	 */
483
	if (!(of->kn->flags & KERNFS_HAS_MMAP))
484 485
		return -ENODEV;

486 487 488
	mutex_lock(&of->mutex);

	rc = -ENODEV;
489
	if (!kernfs_get_active(of->kn))
490 491
		goto out_unlock;

492
	ops = kernfs_ops(of->kn);
493
	rc = ops->mmap(of, vma);
494 495
	if (rc)
		goto out_put;
496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517

	/*
	 * PowerPC's pci_mmap of legacy_mem uses shmem_zero_setup()
	 * to satisfy versions of X which crash if the mmap fails: that
	 * substitutes a new vm_file, and we don't then want bin_vm_ops.
	 */
	if (vma->vm_file != file)
		goto out_put;

	rc = -EINVAL;
	if (of->mmapped && of->vm_ops != vma->vm_ops)
		goto out_put;

	/*
	 * It is not possible to successfully wrap close.
	 * So error if someone is trying to use close.
	 */
	rc = -EINVAL;
	if (vma->vm_ops && vma->vm_ops->close)
		goto out_put;

	rc = 0;
518
	of->mmapped = true;
519 520 521
	of->vm_ops = vma->vm_ops;
	vma->vm_ops = &kernfs_vm_ops;
out_put:
522
	kernfs_put_active(of->kn);
523 524 525 526 527 528 529
out_unlock:
	mutex_unlock(&of->mutex);

	return rc;
}

/**
530
 *	kernfs_get_open_node - get or create kernfs_open_node
531
 *	@kn: target kernfs_node
532
 *	@of: kernfs_open_file for this instance of open
533
 *
534 535
 *	If @kn->attr.open exists, increment its reference count; otherwise,
 *	create one.  @of is chained to the files list.
536 537 538 539 540 541 542
 *
 *	LOCKING:
 *	Kernel thread context (may sleep).
 *
 *	RETURNS:
 *	0 on success, -errno on failure.
 */
543 544
static int kernfs_get_open_node(struct kernfs_node *kn,
				struct kernfs_open_file *of)
545
{
546
	struct kernfs_open_node *on, *new_on = NULL;
547 548

 retry:
549 550
	mutex_lock(&kernfs_open_file_mutex);
	spin_lock_irq(&kernfs_open_node_lock);
551

552 553 554
	if (!kn->attr.open && new_on) {
		kn->attr.open = new_on;
		new_on = NULL;
555 556
	}

557 558 559 560
	on = kn->attr.open;
	if (on) {
		atomic_inc(&on->refcnt);
		list_add_tail(&of->list, &on->files);
561 562
	}

563 564
	spin_unlock_irq(&kernfs_open_node_lock);
	mutex_unlock(&kernfs_open_file_mutex);
565

566 567
	if (on) {
		kfree(new_on);
568 569 570 571
		return 0;
	}

	/* not there, initialize a new one and retry */
572 573
	new_on = kmalloc(sizeof(*new_on), GFP_KERNEL);
	if (!new_on)
574 575
		return -ENOMEM;

576 577 578 579
	atomic_set(&new_on->refcnt, 0);
	atomic_set(&new_on->event, 1);
	init_waitqueue_head(&new_on->poll);
	INIT_LIST_HEAD(&new_on->files);
580 581 582 583
	goto retry;
}

/**
584
 *	kernfs_put_open_node - put kernfs_open_node
585
 *	@kn: target kernfs_nodet
586
 *	@of: associated kernfs_open_file
587
 *
588
 *	Put @kn->attr.open and unlink @of from the files list.  If
589 590 591 592 593
 *	reference count reaches zero, disassociate and free it.
 *
 *	LOCKING:
 *	None.
 */
594 595
static void kernfs_put_open_node(struct kernfs_node *kn,
				 struct kernfs_open_file *of)
596
{
597
	struct kernfs_open_node *on = kn->attr.open;
598 599
	unsigned long flags;

600 601
	mutex_lock(&kernfs_open_file_mutex);
	spin_lock_irqsave(&kernfs_open_node_lock, flags);
602 603 604 605

	if (of)
		list_del(&of->list);

606
	if (atomic_dec_and_test(&on->refcnt))
607
		kn->attr.open = NULL;
608
	else
609
		on = NULL;
610

611 612
	spin_unlock_irqrestore(&kernfs_open_node_lock, flags);
	mutex_unlock(&kernfs_open_file_mutex);
613

614
	kfree(on);
615 616
}

617
static int kernfs_fop_open(struct inode *inode, struct file *file)
618
{
619
	struct kernfs_node *kn = inode->i_private;
620
	struct kernfs_root *root = kernfs_root(kn);
621
	const struct kernfs_ops *ops;
622
	struct kernfs_open_file *of;
623 624 625
	bool has_read, has_write, has_mmap;
	int error = -EACCES;

626
	if (!kernfs_get_active(kn))
627 628
		return -ENODEV;

629
	ops = kernfs_ops(kn);
630 631 632 633 634

	has_read = ops->seq_show || ops->read || ops->mmap;
	has_write = ops->write || ops->mmap;
	has_mmap = ops->mmap;

635 636 637 638 639
	/* see the flag definition for details */
	if (root->flags & KERNFS_ROOT_EXTRA_OPEN_PERM_CHECK) {
		if ((file->f_mode & FMODE_WRITE) &&
		    (!(inode->i_mode & S_IWUGO) || !has_write))
			goto err_out;
640

641 642 643 644
		if ((file->f_mode & FMODE_READ) &&
		    (!(inode->i_mode & S_IRUGO) || !has_read))
			goto err_out;
	}
645

646
	/* allocate a kernfs_open_file for the file */
647
	error = -ENOMEM;
648
	of = kzalloc(sizeof(struct kernfs_open_file), GFP_KERNEL);
649 650 651 652 653 654 655 656 657 658 659 660 661 662
	if (!of)
		goto err_out;

	/*
	 * The following is done to give a different lockdep key to
	 * @of->mutex for files which implement mmap.  This is a rather
	 * crude way to avoid false positive lockdep warning around
	 * mm->mmap_sem - mmap nests @of->mutex under mm->mmap_sem and
	 * reading /sys/block/sda/trace/act_mask grabs sr_mutex, under
	 * which mm->mmap_sem nests, while holding @of->mutex.  As each
	 * open file has a separate mutex, it's okay as long as those don't
	 * happen on the same file.  At this point, we can't easily give
	 * each file a separate locking class.  Let's differentiate on
	 * whether the file has mmap or not for now.
663 664 665
	 *
	 * Both paths of the branch look the same.  They're supposed to
	 * look that way and give @of->mutex different static lockdep keys.
666 667 668 669 670 671
	 */
	if (has_mmap)
		mutex_init(&of->mutex);
	else
		mutex_init(&of->mutex);

672
	of->kn = kn;
673 674
	of->file = file;

675 676 677 678 679 680
	/*
	 * Write path needs to atomic_write_len outside active reference.
	 * Cache it in open_file.  See kernfs_fop_write() for details.
	 */
	of->atomic_write_len = ops->atomic_write_len;

681 682 683 684 685 686 687 688
	error = -EINVAL;
	/*
	 * ->seq_show is incompatible with ->prealloc,
	 * as seq_read does its own allocation.
	 * ->read must be used instead.
	 */
	if (ops->prealloc && ops->seq_show)
		goto err_free;
689 690 691 692 693 694
	if (ops->prealloc) {
		int len = of->atomic_write_len ?: PAGE_SIZE;
		of->prealloc_buf = kmalloc(len + 1, GFP_KERNEL);
		error = -ENOMEM;
		if (!of->prealloc_buf)
			goto err_free;
695
		mutex_init(&of->prealloc_mutex);
696 697
	}

698 699 700 701 702 703 704 705 706 707 708 709
	/*
	 * Always instantiate seq_file even if read access doesn't use
	 * seq_file or is not requested.  This unifies private data access
	 * and readable regular files are the vast majority anyway.
	 */
	if (ops->seq_show)
		error = seq_open(file, &kernfs_seq_ops);
	else
		error = seq_open(file, NULL);
	if (error)
		goto err_free;

710 711
	of->seq_file = file->private_data;
	of->seq_file->private = of;
712 713 714 715 716

	/* seq_file clears PWRITE unconditionally, restore it if WRITE */
	if (file->f_mode & FMODE_WRITE)
		file->f_mode |= FMODE_PWRITE;

717 718
	/* make sure we have open node struct */
	error = kernfs_get_open_node(kn, of);
719
	if (error)
720 721 722 723 724 725 726 727
		goto err_seq_release;

	if (ops->open) {
		/* nobody has access to @of yet, skip @of->mutex */
		error = ops->open(of);
		if (error)
			goto err_put_node;
	}
728 729

	/* open succeeded, put active references */
730
	kernfs_put_active(kn);
731 732
	return 0;

733 734 735
err_put_node:
	kernfs_put_open_node(kn, of);
err_seq_release:
736 737
	seq_release(inode, file);
err_free:
738
	kfree(of->prealloc_buf);
739 740
	kfree(of);
err_out:
741
	kernfs_put_active(kn);
742 743 744
	return error;
}

745 746 747 748
/* used from release/drain to ensure that ->release() is called exactly once */
static void kernfs_release_file(struct kernfs_node *kn,
				struct kernfs_open_file *of)
{
749 750 751 752 753 754 755 756
	/*
	 * @of is guaranteed to have no other file operations in flight and
	 * we just want to synchronize release and drain paths.
	 * @kernfs_open_file_mutex is enough.  @of->mutex can't be used
	 * here because drain path may be called from places which can
	 * cause circular dependency.
	 */
	lockdep_assert_held(&kernfs_open_file_mutex);
757 758 759 760 761 762 763 764 765 766 767 768

	if (!of->released) {
		/*
		 * A file is never detached without being released and we
		 * need to be able to release files which are deactivated
		 * and being drained.  Don't use kernfs_ops().
		 */
		kn->attr.ops->release(of);
		of->released = true;
	}
}

769
static int kernfs_fop_release(struct inode *inode, struct file *filp)
770
{
771
	struct kernfs_node *kn = inode->i_private;
772
	struct kernfs_open_file *of = kernfs_of(filp);
773

774 775 776 777 778 779
	if (kn->flags & KERNFS_HAS_RELEASE) {
		mutex_lock(&kernfs_open_file_mutex);
		kernfs_release_file(kn, of);
		mutex_unlock(&kernfs_open_file_mutex);
	}

780
	kernfs_put_open_node(kn, of);
781
	seq_release(inode, filp);
782
	kfree(of->prealloc_buf);
783 784 785 786 787
	kfree(of);

	return 0;
}

788
void kernfs_drain_open_files(struct kernfs_node *kn)
789
{
790 791
	struct kernfs_open_node *on;
	struct kernfs_open_file *of;
792

793
	if (!(kn->flags & (KERNFS_HAS_MMAP | KERNFS_HAS_RELEASE)))
794 795
		return;

796 797 798 799 800 801
	spin_lock_irq(&kernfs_open_node_lock);
	on = kn->attr.open;
	if (on)
		atomic_inc(&on->refcnt);
	spin_unlock_irq(&kernfs_open_node_lock);
	if (!on)
802 803
		return;

804
	mutex_lock(&kernfs_open_file_mutex);
805

806
	list_for_each_entry(of, &on->files, list) {
807
		struct inode *inode = file_inode(of->file);
808 809 810 811

		if (kn->flags & KERNFS_HAS_MMAP)
			unmap_mapping_range(inode->i_mapping, 0, 0, 1);

812 813
		if (kn->flags & KERNFS_HAS_RELEASE)
			kernfs_release_file(kn, of);
814
	}
815

816
	mutex_unlock(&kernfs_open_file_mutex);
817

818
	kernfs_put_open_node(kn, NULL);
819 820
}

821 822
/*
 * Kernfs attribute files are pollable.  The idea is that you read
823 824 825
 * the content and then you use 'poll' or 'select' to wait for
 * the content to change.  When the content changes (assuming the
 * manager for the kobject supports notification), poll will
826
 * return EPOLLERR|EPOLLPRI, and select will return the fd whether
827 828 829 830 831 832 833 834
 * it is waiting for read, write, or exceptions.
 * Once poll/select indicates that the value has changed, you
 * need to close and re-open the file, or seek to 0 and read again.
 * Reminder: this only works for attributes which actively support
 * it, and it is not possible to test an attribute from userspace
 * to see if it supports poll (Neither 'poll' nor 'select' return
 * an appropriate error code).  When in doubt, set a suitable timeout value.
 */
835 836 837 838 839 840 841 842 843 844 845 846 847
__poll_t kernfs_generic_poll(struct kernfs_open_file *of, poll_table *wait)
{
	struct kernfs_node *kn = kernfs_dentry_node(of->file->f_path.dentry);
	struct kernfs_open_node *on = kn->attr.open;

	poll_wait(of->file, &on->poll, wait);

	if (of->event != atomic_read(&on->event))
		return DEFAULT_POLLMASK|EPOLLERR|EPOLLPRI;

	return DEFAULT_POLLMASK;
}

848
static __poll_t kernfs_fop_poll(struct file *filp, poll_table *wait)
849
{
850
	struct kernfs_open_file *of = kernfs_of(filp);
851
	struct kernfs_node *kn = kernfs_dentry_node(filp->f_path.dentry);
852
	__poll_t ret;
853

854
	if (!kernfs_get_active(kn))
855
		return DEFAULT_POLLMASK|EPOLLERR|EPOLLPRI;
856

857 858 859 860
	if (kn->attr.ops->poll)
		ret = kn->attr.ops->poll(of, wait);
	else
		ret = kernfs_generic_poll(of, wait);
861

862
	kernfs_put_active(kn);
863
	return ret;
864 865
}

866
static void kernfs_notify_workfn(struct work_struct *work)
867
{
868
	struct kernfs_node *kn;
869
	struct kernfs_open_node *on;
870
	struct kernfs_super_info *info;
871 872 873 874 875 876
repeat:
	/* pop one off the notify_list */
	spin_lock_irq(&kernfs_notify_lock);
	kn = kernfs_notify_list;
	if (kn == KERNFS_NOTIFY_EOL) {
		spin_unlock_irq(&kernfs_notify_lock);
877
		return;
878 879 880 881
	}
	kernfs_notify_list = kn->attr.notify_next;
	kn->attr.notify_next = NULL;
	spin_unlock_irq(&kernfs_notify_lock);
882 883

	/* kick poll */
884
	spin_lock_irq(&kernfs_open_node_lock);
885

886 887 888 889
	on = kn->attr.open;
	if (on) {
		atomic_inc(&on->event);
		wake_up_interruptible(&on->poll);
890 891
	}

892
	spin_unlock_irq(&kernfs_open_node_lock);
893 894 895 896

	/* kick fsnotify */
	mutex_lock(&kernfs_mutex);

897
	list_for_each_entry(info, &kernfs_root(kn)->supers, node) {
898
		struct kernfs_node *parent;
899 900
		struct inode *inode;

901 902 903 904 905 906
		/*
		 * We want fsnotify_modify() on @kn but as the
		 * modifications aren't originating from userland don't
		 * have the matching @file available.  Look up the inodes
		 * and generate the events manually.
		 */
907
		inode = ilookup(info->sb, kn->id.ino);
908 909 910
		if (!inode)
			continue;

911 912 913 914
		parent = kernfs_get_parent(kn);
		if (parent) {
			struct inode *p_inode;

915
			p_inode = ilookup(info->sb, parent->id.ino);
916 917 918 919 920 921 922
			if (p_inode) {
				fsnotify(p_inode, FS_MODIFY | FS_EVENT_ON_CHILD,
					 inode, FSNOTIFY_EVENT_INODE, kn->name, 0);
				iput(p_inode);
			}

			kernfs_put(parent);
923 924
		}

925 926
		fsnotify(inode, FS_MODIFY, inode, FSNOTIFY_EVENT_INODE,
			 kn->name, 0);
927 928 929 930
		iput(inode);
	}

	mutex_unlock(&kernfs_mutex);
931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957
	kernfs_put(kn);
	goto repeat;
}

/**
 * kernfs_notify - notify a kernfs file
 * @kn: file to notify
 *
 * Notify @kn such that poll(2) on @kn wakes up.  Maybe be called from any
 * context.
 */
void kernfs_notify(struct kernfs_node *kn)
{
	static DECLARE_WORK(kernfs_notify_work, kernfs_notify_workfn);
	unsigned long flags;

	if (WARN_ON(kernfs_type(kn) != KERNFS_FILE))
		return;

	spin_lock_irqsave(&kernfs_notify_lock, flags);
	if (!kn->attr.notify_next) {
		kernfs_get(kn);
		kn->attr.notify_next = kernfs_notify_list;
		kernfs_notify_list = kn;
		schedule_work(&kernfs_notify_work);
	}
	spin_unlock_irqrestore(&kernfs_notify_lock, flags);
958 959 960
}
EXPORT_SYMBOL_GPL(kernfs_notify);

961
const struct file_operations kernfs_file_fops = {
962 963
	.read		= kernfs_fop_read,
	.write		= kernfs_fop_write,
964
	.llseek		= generic_file_llseek,
965 966 967 968
	.mmap		= kernfs_fop_mmap,
	.open		= kernfs_fop_open,
	.release	= kernfs_fop_release,
	.poll		= kernfs_fop_poll,
969
	.fsync		= noop_fsync,
970 971 972
};

/**
973
 * __kernfs_create_file - kernfs internal function to create a file
974 975 976
 * @parent: directory to create the file in
 * @name: name of the file
 * @mode: mode of the file
977 978
 * @uid: uid of the file
 * @gid: gid of the file
979 980 981 982 983 984 985 986
 * @size: size of the file
 * @ops: kernfs operations for the file
 * @priv: private data for the file
 * @ns: optional namespace tag of the file
 * @key: lockdep key for the file's active_ref, %NULL to disable lockdep
 *
 * Returns the created node on success, ERR_PTR() value on error.
 */
987 988
struct kernfs_node *__kernfs_create_file(struct kernfs_node *parent,
					 const char *name,
989 990
					 umode_t mode, kuid_t uid, kgid_t gid,
					 loff_t size,
991 992 993
					 const struct kernfs_ops *ops,
					 void *priv, const void *ns,
					 struct lock_class_key *key)
994
{
995
	struct kernfs_node *kn;
996
	unsigned flags;
997 998
	int rc;

999 1000
	flags = KERNFS_FILE;

1001 1002
	kn = kernfs_new_node(parent, name, (mode & S_IALLUGO) | S_IFREG,
			     uid, gid, flags);
1003
	if (!kn)
1004 1005
		return ERR_PTR(-ENOMEM);

1006 1007 1008
	kn->attr.ops = ops;
	kn->attr.size = size;
	kn->ns = ns;
1009
	kn->priv = priv;
1010 1011 1012

#ifdef CONFIG_DEBUG_LOCK_ALLOC
	if (key) {
1013
		lockdep_init_map(&kn->dep_map, "kn->count", key, 0);
1014
		kn->flags |= KERNFS_LOCKDEP;
1015 1016 1017 1018
	}
#endif

	/*
1019
	 * kn->attr.ops is accesible only while holding active ref.  We
1020 1021 1022 1023
	 * need to know whether some ops are implemented outside active
	 * ref.  Cache their existence in flags.
	 */
	if (ops->seq_show)
1024
		kn->flags |= KERNFS_HAS_SEQ_SHOW;
1025
	if (ops->mmap)
1026
		kn->flags |= KERNFS_HAS_MMAP;
1027 1028
	if (ops->release)
		kn->flags |= KERNFS_HAS_RELEASE;
1029

1030
	rc = kernfs_add_one(kn);
1031
	if (rc) {
1032
		kernfs_put(kn);
1033 1034
		return ERR_PTR(rc);
	}
1035
	return kn;
1036
}
新手
引导
客服 返回
顶部