relay.c 32.1 KB
Newer Older
1 2 3
/*
 * Public API and common code for kernel->userspace relay file support.
 *
4
 * See Documentation/filesystems/relay.txt for an overview.
5 6 7 8 9
 *
 * Copyright (C) 2002-2005 - Tom Zanussi (zanussi@us.ibm.com), IBM Corp
 * Copyright (C) 1999-2005 - Karim Yaghmour (karim@opersys.com)
 *
 * Moved to kernel/relay.c by Paul Mundt, 2006.
10 11
 * November 2006 - CPU hotplug support by Mathieu Desnoyers
 * 	(mathieu.desnoyers@polymtl.ca)
12 13 14 15 16 17
 *
 * This file is released under the GPL.
 */
#include <linux/errno.h>
#include <linux/stddef.h>
#include <linux/slab.h>
18
#include <linux/export.h>
19 20 21 22
#include <linux/string.h>
#include <linux/relay.h>
#include <linux/vmalloc.h>
#include <linux/mm.h>
23
#include <linux/cpu.h>
24
#include <linux/splice.h>
25 26 27 28

/* list of open channels, for cpu hotplug */
static DEFINE_MUTEX(relay_channels_mutex);
static LIST_HEAD(relay_channels);
29 30 31 32 33 34 35 36 37 38 39

/*
 * close() vm_op implementation for relay file mapping.
 */
static void relay_file_mmap_close(struct vm_area_struct *vma)
{
	struct rchan_buf *buf = vma->vm_private_data;
	buf->chan->cb->buf_unmapped(buf, vma->vm_file);
}

/*
N
Nick Piggin 已提交
40
 * fault() vm_op implementation for relay file mapping.
41
 */
N
Nick Piggin 已提交
42
static int relay_buf_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
43 44 45
{
	struct page *page;
	struct rchan_buf *buf = vma->vm_private_data;
N
Nick Piggin 已提交
46
	pgoff_t pgoff = vmf->pgoff;
47 48

	if (!buf)
N
Nick Piggin 已提交
49
		return VM_FAULT_OOM;
50

N
Nick Piggin 已提交
51
	page = vmalloc_to_page(buf->start + (pgoff << PAGE_SHIFT));
52
	if (!page)
N
Nick Piggin 已提交
53
		return VM_FAULT_SIGBUS;
54
	get_page(page);
N
Nick Piggin 已提交
55
	vmf->page = page;
56

N
Nick Piggin 已提交
57
	return 0;
58 59 60 61 62
}

/*
 * vm_ops for relay file mappings.
 */
63
static const struct vm_operations_struct relay_file_mmap_ops = {
N
Nick Piggin 已提交
64
	.fault = relay_buf_fault,
65 66 67
	.close = relay_file_mmap_close,
};

68 69 70 71 72
/*
 * allocate an array of pointers of struct page
 */
static struct page **relay_alloc_page_array(unsigned int n_pages)
{
73 74 75 76
	const size_t pa_size = n_pages * sizeof(struct page *);
	if (pa_size > PAGE_SIZE)
		return vzalloc(pa_size);
	return kzalloc(pa_size, GFP_KERNEL);
77 78 79 80 81 82 83
}

/*
 * free an array of pointers of struct page
 */
static void relay_free_page_array(struct page **array)
{
84
	kvfree(array);
85 86
}

87 88 89 90 91 92 93 94 95
/**
 *	relay_mmap_buf: - mmap channel buffer to process address space
 *	@buf: relay channel buffer
 *	@vma: vm_area_struct describing memory to be mapped
 *
 *	Returns 0 if ok, negative on error
 *
 *	Caller should already have grabbed mmap_sem.
 */
96
static int relay_mmap_buf(struct rchan_buf *buf, struct vm_area_struct *vma)
97 98 99 100 101 102 103 104 105 106 107
{
	unsigned long length = vma->vm_end - vma->vm_start;
	struct file *filp = vma->vm_file;

	if (!buf)
		return -EBADF;

	if (length != (unsigned long)buf->chan->alloc_size)
		return -EINVAL;

	vma->vm_ops = &relay_file_mmap_ops;
108
	vma->vm_flags |= VM_DONTEXPAND;
109 110 111 112 113 114 115 116 117 118 119
	vma->vm_private_data = buf;
	buf->chan->cb->buf_mapped(buf, filp);

	return 0;
}

/**
 *	relay_alloc_buf - allocate a channel buffer
 *	@buf: the buffer struct
 *	@size: total size of the buffer
 *
120
 *	Returns a pointer to the resulting buffer, %NULL if unsuccessful. The
121
 *	passed in size will get page aligned, if it isn't already.
122
 */
123
static void *relay_alloc_buf(struct rchan_buf *buf, size_t *size)
124 125 126 127
{
	void *mem;
	unsigned int i, j, n_pages;

128 129
	*size = PAGE_ALIGN(*size);
	n_pages = *size >> PAGE_SHIFT;
130

131
	buf->page_array = relay_alloc_page_array(n_pages);
132 133 134 135 136 137 138
	if (!buf->page_array)
		return NULL;

	for (i = 0; i < n_pages; i++) {
		buf->page_array[i] = alloc_page(GFP_KERNEL);
		if (unlikely(!buf->page_array[i]))
			goto depopulate;
T
Tom Zanussi 已提交
139
		set_page_private(buf->page_array[i], (unsigned long)buf);
140 141 142 143 144
	}
	mem = vmap(buf->page_array, n_pages, VM_MAP, PAGE_KERNEL);
	if (!mem)
		goto depopulate;

145
	memset(mem, 0, *size);
146 147 148 149 150 151
	buf->page_count = n_pages;
	return mem;

depopulate:
	for (j = 0; j < i; j++)
		__free_page(buf->page_array[j]);
152
	relay_free_page_array(buf->page_array);
153 154 155 156 157
	return NULL;
}

/**
 *	relay_create_buf - allocate and initialize a channel buffer
158
 *	@chan: the relay channel
159
 *
160
 *	Returns channel buffer if successful, %NULL otherwise.
161
 */
162
static struct rchan_buf *relay_create_buf(struct rchan *chan)
163
{
164 165 166
	struct rchan_buf *buf;

	if (chan->n_subbufs > UINT_MAX / sizeof(size_t *))
167 168
		return NULL;

169 170 171
	buf = kzalloc(sizeof(struct rchan_buf), GFP_KERNEL);
	if (!buf)
		return NULL;
172 173 174 175
	buf->padding = kmalloc(chan->n_subbufs * sizeof(size_t *), GFP_KERNEL);
	if (!buf->padding)
		goto free_buf;

176
	buf->start = relay_alloc_buf(buf, &chan->alloc_size);
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
	if (!buf->start)
		goto free_buf;

	buf->chan = chan;
	kref_get(&buf->chan->kref);
	return buf;

free_buf:
	kfree(buf->padding);
	kfree(buf);
	return NULL;
}

/**
 *	relay_destroy_channel - free the channel struct
192
 *	@kref: target kernel reference that contains the relay channel
193 194 195
 *
 *	Should only be called from kref_put().
 */
196
static void relay_destroy_channel(struct kref *kref)
197 198 199 200 201 202 203 204 205
{
	struct rchan *chan = container_of(kref, struct rchan, kref);
	kfree(chan);
}

/**
 *	relay_destroy_buf - destroy an rchan_buf struct and associated buffer
 *	@buf: the buffer struct
 */
206
static void relay_destroy_buf(struct rchan_buf *buf)
207 208 209 210 211 212 213 214
{
	struct rchan *chan = buf->chan;
	unsigned int i;

	if (likely(buf->start)) {
		vunmap(buf->start);
		for (i = 0; i < buf->page_count; i++)
			__free_page(buf->page_array[i]);
215
		relay_free_page_array(buf->page_array);
216
	}
217
	*per_cpu_ptr(chan->buf, buf->cpu) = NULL;
218 219 220 221 222 223 224
	kfree(buf->padding);
	kfree(buf);
	kref_put(&chan->kref, relay_destroy_channel);
}

/**
 *	relay_remove_buf - remove a channel buffer
225
 *	@kref: target kernel reference that contains the relay buffer
226
 *
227
 *	Removes the file from the filesystem, which also frees the
228 229 230
 *	rchan_buf_struct and the channel buffer.  Should only be called from
 *	kref_put().
 */
231
static void relay_remove_buf(struct kref *kref)
232 233 234 235 236 237 238 239 240 241 242
{
	struct rchan_buf *buf = container_of(kref, struct rchan_buf, kref);
	relay_destroy_buf(buf);
}

/**
 *	relay_buf_empty - boolean, is the channel buffer empty?
 *	@buf: channel buffer
 *
 *	Returns 1 if the buffer is empty, 0 otherwise.
 */
243
static int relay_buf_empty(struct rchan_buf *buf)
244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304
{
	return (buf->subbufs_produced - buf->subbufs_consumed) ? 0 : 1;
}

/**
 *	relay_buf_full - boolean, is the channel buffer full?
 *	@buf: channel buffer
 *
 *	Returns 1 if the buffer is full, 0 otherwise.
 */
int relay_buf_full(struct rchan_buf *buf)
{
	size_t ready = buf->subbufs_produced - buf->subbufs_consumed;
	return (ready >= buf->chan->n_subbufs) ? 1 : 0;
}
EXPORT_SYMBOL_GPL(relay_buf_full);

/*
 * High-level relay kernel API and associated functions.
 */

/*
 * rchan_callback implementations defining default channel behavior.  Used
 * in place of corresponding NULL values in client callback struct.
 */

/*
 * subbuf_start() default callback.  Does nothing.
 */
static int subbuf_start_default_callback (struct rchan_buf *buf,
					  void *subbuf,
					  void *prev_subbuf,
					  size_t prev_padding)
{
	if (relay_buf_full(buf))
		return 0;

	return 1;
}

/*
 * buf_mapped() default callback.  Does nothing.
 */
static void buf_mapped_default_callback(struct rchan_buf *buf,
					struct file *filp)
{
}

/*
 * buf_unmapped() default callback.  Does nothing.
 */
static void buf_unmapped_default_callback(struct rchan_buf *buf,
					  struct file *filp)
{
}

/*
 * create_buf_file_create() default callback.  Does nothing.
 */
static struct dentry *create_buf_file_default_callback(const char *filename,
						       struct dentry *parent,
A
Al Viro 已提交
305
						       umode_t mode,
306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
						       struct rchan_buf *buf,
						       int *is_global)
{
	return NULL;
}

/*
 * remove_buf_file() default callback.  Does nothing.
 */
static int remove_buf_file_default_callback(struct dentry *dentry)
{
	return -EINVAL;
}

/* relay channel default callbacks */
static struct rchan_callbacks default_channel_callbacks = {
	.subbuf_start = subbuf_start_default_callback,
	.buf_mapped = buf_mapped_default_callback,
	.buf_unmapped = buf_unmapped_default_callback,
	.create_buf_file = create_buf_file_default_callback,
	.remove_buf_file = remove_buf_file_default_callback,
};

/**
 *	wakeup_readers - wake up readers waiting on a channel
331
 *	@data: contains the channel buffer
332
 *
333
 *	This is the timer function used to defer reader waking.
334
 */
335
static void wakeup_readers(unsigned long data)
336
{
337
	struct rchan_buf *buf = (struct rchan_buf *)data;
338 339 340 341 342 343 344 345
	wake_up_interruptible(&buf->read_wait);
}

/**
 *	__relay_reset - reset a channel buffer
 *	@buf: the channel buffer
 *	@init: 1 if this is a first-time initialization
 *
346
 *	See relay_reset() for description of effect.
347
 */
A
Andrew Morton 已提交
348
static void __relay_reset(struct rchan_buf *buf, unsigned int init)
349 350 351 352 353 354
{
	size_t i;

	if (init) {
		init_waitqueue_head(&buf->read_wait);
		kref_init(&buf->kref);
355 356 357
		setup_timer(&buf->timer, wakeup_readers, (unsigned long)buf);
	} else
		del_timer_sync(&buf->timer);
358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379

	buf->subbufs_produced = 0;
	buf->subbufs_consumed = 0;
	buf->bytes_consumed = 0;
	buf->finalized = 0;
	buf->data = buf->start;
	buf->offset = 0;

	for (i = 0; i < buf->chan->n_subbufs; i++)
		buf->padding[i] = 0;

	buf->chan->cb->subbuf_start(buf, buf->data, NULL, 0);
}

/**
 *	relay_reset - reset the channel
 *	@chan: the channel
 *
 *	This has the effect of erasing all data from all channel buffers
 *	and restarting the channel in its initial state.  The buffers
 *	are not freed, so any mappings are still in effect.
 *
380
 *	NOTE. Care should be taken that the channel isn't actually
381 382 383 384
 *	being used by anything when this call is made.
 */
void relay_reset(struct rchan *chan)
{
385
	struct rchan_buf *buf;
386 387 388 389 390
	unsigned int i;

	if (!chan)
		return;

391 392
	if (chan->is_global && (buf = *per_cpu_ptr(chan->buf, 0))) {
		__relay_reset(buf, 0);
393
		return;
394
	}
395 396

	mutex_lock(&relay_channels_mutex);
L
Lai Jiangshan 已提交
397
	for_each_possible_cpu(i)
398 399
		if ((buf = *per_cpu_ptr(chan->buf, i)))
			__relay_reset(buf, 0);
400
	mutex_unlock(&relay_channels_mutex);
401 402 403
}
EXPORT_SYMBOL_GPL(relay_reset);

404 405 406 407
static inline void relay_set_buf_dentry(struct rchan_buf *buf,
					struct dentry *dentry)
{
	buf->dentry = dentry;
408
	d_inode(buf->dentry)->i_size = buf->early_bytes;
409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
}

static struct dentry *relay_create_buf_file(struct rchan *chan,
					    struct rchan_buf *buf,
					    unsigned int cpu)
{
	struct dentry *dentry;
	char *tmpname;

	tmpname = kzalloc(NAME_MAX + 1, GFP_KERNEL);
	if (!tmpname)
		return NULL;
	snprintf(tmpname, NAME_MAX, "%s%d", chan->base_filename, cpu);

	/* Create file in fs */
	dentry = chan->cb->create_buf_file(tmpname, chan->parent,
					   S_IRUSR, buf,
					   &chan->is_global);

	kfree(tmpname);

	return dentry;
}

433
/*
434 435
 *	relay_open_buf - create a new relay channel buffer
 *
436
 *	used by relay_open() and CPU hotplug.
437
 */
438
static struct rchan_buf *relay_open_buf(struct rchan *chan, unsigned int cpu)
439
{
440
 	struct rchan_buf *buf = NULL;
441 442
	struct dentry *dentry;

443
 	if (chan->is_global)
444
		return *per_cpu_ptr(chan->buf, 0);
445 446 447

	buf = relay_create_buf(chan);
	if (!buf)
448 449 450 451 452 453 454
		return NULL;

	if (chan->has_base_filename) {
		dentry = relay_create_buf_file(chan, buf, cpu);
		if (!dentry)
			goto free_buf;
		relay_set_buf_dentry(buf, dentry);
455 456 457 458 459 460 461
	} else {
		/* Only retrieve global info, nothing more, nothing less */
		dentry = chan->cb->create_buf_file(NULL, NULL,
						   S_IRUSR, buf,
						   &chan->is_global);
		if (WARN_ON(dentry))
			goto free_buf;
462
	}
463 464 465

 	buf->cpu = cpu;
 	__relay_reset(buf, 1);
466

467
 	if(chan->is_global) {
468
		*per_cpu_ptr(chan->buf, 0) = buf;
469 470 471
 		buf->cpu = 0;
  	}

472
	return buf;
473 474 475

free_buf:
 	relay_destroy_buf(buf);
476
	return NULL;
477 478 479 480 481 482 483 484 485 486
}

/**
 *	relay_close_buf - close a channel buffer
 *	@buf: channel buffer
 *
 *	Marks the buffer finalized and restores the default callbacks.
 *	The channel buffer and channel buffer data structure are then freed
 *	automatically when the last reference is given up.
 */
A
Andrew Morton 已提交
487
static void relay_close_buf(struct rchan_buf *buf)
488 489
{
	buf->finalized = 1;
490
	del_timer_sync(&buf->timer);
491
	buf->chan->cb->remove_buf_file(buf->dentry);
492 493 494
	kref_put(&buf->kref, relay_remove_buf);
}

A
Andrew Morton 已提交
495
static void setup_callbacks(struct rchan *chan,
496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
				   struct rchan_callbacks *cb)
{
	if (!cb) {
		chan->cb = &default_channel_callbacks;
		return;
	}

	if (!cb->subbuf_start)
		cb->subbuf_start = subbuf_start_default_callback;
	if (!cb->buf_mapped)
		cb->buf_mapped = buf_mapped_default_callback;
	if (!cb->buf_unmapped)
		cb->buf_unmapped = buf_unmapped_default_callback;
	if (!cb->create_buf_file)
		cb->create_buf_file = create_buf_file_default_callback;
	if (!cb->remove_buf_file)
		cb->remove_buf_file = remove_buf_file_default_callback;
	chan->cb = cb;
}

516
int relay_prepare_cpu(unsigned int cpu)
517 518
{
	struct rchan *chan;
519
	struct rchan_buf *buf;
520

521 522 523 524 525 526 527 528 529
	mutex_lock(&relay_channels_mutex);
	list_for_each_entry(chan, &relay_channels, list) {
		if ((buf = *per_cpu_ptr(chan->buf, cpu)))
			continue;
		buf = relay_open_buf(chan, cpu);
		if (!buf) {
			pr_err("relay: cpu %d buffer creation failed\n", cpu);
			mutex_unlock(&relay_channels_mutex);
			return -ENOMEM;
530
		}
531
		*per_cpu_ptr(chan->buf, cpu) = buf;
532
	}
533 534
	mutex_unlock(&relay_channels_mutex);
	return 0;
535 536
}

537 538
/**
 *	relay_open - create a new relay channel
539 540
 *	@base_filename: base name of files to create, %NULL for buffering only
 *	@parent: dentry of parent directory, %NULL for root directory or buffer
541 542 543
 *	@subbuf_size: size of sub-buffers
 *	@n_subbufs: number of sub-buffers
 *	@cb: client callback functions
544
 *	@private_data: user-defined data
545
 *
546
 *	Returns channel pointer if successful, %NULL otherwise.
547 548 549 550
 *
 *	Creates a channel buffer for each cpu using the sizes and
 *	attributes specified.  The created channel buffer files
 *	will be named base_filename0...base_filenameN-1.  File
551
 *	permissions will be %S_IRUSR.
552 553 554 555
 *
 *	If opening a buffer (@parent = NULL) that you later wish to register
 *	in a filesystem, call relay_late_setup_files() once the @parent dentry
 *	is available.
556 557 558 559 560
 */
struct rchan *relay_open(const char *base_filename,
			 struct dentry *parent,
			 size_t subbuf_size,
			 size_t n_subbufs,
561 562
			 struct rchan_callbacks *cb,
			 void *private_data)
563 564 565
{
	unsigned int i;
	struct rchan *chan;
566
	struct rchan_buf *buf;
567 568 569

	if (!(subbuf_size && n_subbufs))
		return NULL;
570 571
	if (subbuf_size > UINT_MAX / n_subbufs)
		return NULL;
572

573
	chan = kzalloc(sizeof(struct rchan), GFP_KERNEL);
574 575 576
	if (!chan)
		return NULL;

577
	chan->buf = alloc_percpu(struct rchan_buf *);
578 579 580
	chan->version = RELAYFS_CHANNEL_VERSION;
	chan->n_subbufs = n_subbufs;
	chan->subbuf_size = subbuf_size;
581
	chan->alloc_size = PAGE_ALIGN(subbuf_size * n_subbufs);
582 583
	chan->parent = parent;
	chan->private_data = private_data;
584 585 586 587
	if (base_filename) {
		chan->has_base_filename = 1;
		strlcpy(chan->base_filename, base_filename, NAME_MAX);
	}
588 589 590
	setup_callbacks(chan, cb);
	kref_init(&chan->kref);

591
	mutex_lock(&relay_channels_mutex);
592
	for_each_online_cpu(i) {
593 594
		buf = relay_open_buf(chan, i);
		if (!buf)
595
			goto free_bufs;
596
		*per_cpu_ptr(chan->buf, i) = buf;
597
	}
598 599
	list_add(&chan->list, &relay_channels);
	mutex_unlock(&relay_channels_mutex);
600 601 602 603

	return chan;

free_bufs:
L
Lai Jiangshan 已提交
604
	for_each_possible_cpu(i) {
605 606
		if ((buf = *per_cpu_ptr(chan->buf, i)))
			relay_close_buf(buf);
607 608 609
	}

	kref_put(&chan->kref, relay_destroy_channel);
610
	mutex_unlock(&relay_channels_mutex);
611
	kfree(chan);
612 613 614 615
	return NULL;
}
EXPORT_SYMBOL_GPL(relay_open);

616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636
struct rchan_percpu_buf_dispatcher {
	struct rchan_buf *buf;
	struct dentry *dentry;
};

/* Called in atomic context. */
static void __relay_set_buf_dentry(void *info)
{
	struct rchan_percpu_buf_dispatcher *p = info;

	relay_set_buf_dentry(p->buf, p->dentry);
}

/**
 *	relay_late_setup_files - triggers file creation
 *	@chan: channel to operate on
 *	@base_filename: base name of files to create
 *	@parent: dentry of parent directory, %NULL for root directory
 *
 *	Returns 0 if successful, non-zero otherwise.
 *
637 638 639 640 641 642
 *	Use to setup files for a previously buffer-only channel created
 *	by relay_open() with a NULL parent dentry.
 *
 *	For example, this is useful for perfomring early tracing in kernel,
 *	before VFS is up and then exposing the early results once the dentry
 *	is available.
643 644 645 646 647 648 649 650 651
 */
int relay_late_setup_files(struct rchan *chan,
			   const char *base_filename,
			   struct dentry *parent)
{
	int err = 0;
	unsigned int i, curr_cpu;
	unsigned long flags;
	struct dentry *dentry;
652
	struct rchan_buf *buf;
653 654 655 656 657 658 659 660 661
	struct rchan_percpu_buf_dispatcher disp;

	if (!chan || !base_filename)
		return -EINVAL;

	strlcpy(chan->base_filename, base_filename, NAME_MAX);

	mutex_lock(&relay_channels_mutex);
	/* Is chan already set up? */
662 663
	if (unlikely(chan->has_base_filename)) {
		mutex_unlock(&relay_channels_mutex);
664
		return -EEXIST;
665
	}
666 667
	chan->has_base_filename = 1;
	chan->parent = parent;
668 669 670

	if (chan->is_global) {
		err = -EINVAL;
671 672 673
		buf = *per_cpu_ptr(chan->buf, 0);
		if (!WARN_ON_ONCE(!buf)) {
			dentry = relay_create_buf_file(chan, buf, 0);
674
			if (dentry && !WARN_ON_ONCE(!chan->is_global)) {
675
				relay_set_buf_dentry(buf, dentry);
676 677 678 679 680 681 682
				err = 0;
			}
		}
		mutex_unlock(&relay_channels_mutex);
		return err;
	}

683 684 685 686 687 688 689
	curr_cpu = get_cpu();
	/*
	 * The CPU hotplug notifier ran before us and created buffers with
	 * no files associated. So it's safe to call relay_setup_buf_file()
	 * on all currently online CPUs.
	 */
	for_each_online_cpu(i) {
690 691
		buf = *per_cpu_ptr(chan->buf, i);
		if (unlikely(!buf)) {
692
			WARN_ONCE(1, KERN_ERR "CPU has no buffer!\n");
693 694 695 696
			err = -EINVAL;
			break;
		}

697
		dentry = relay_create_buf_file(chan, buf, i);
698 699 700 701 702 703 704
		if (unlikely(!dentry)) {
			err = -EINVAL;
			break;
		}

		if (curr_cpu == i) {
			local_irq_save(flags);
705
			relay_set_buf_dentry(buf, dentry);
706 707
			local_irq_restore(flags);
		} else {
708
			disp.buf = buf;
709 710 711 712 713 714 715 716 717 718 719 720 721 722 723
			disp.dentry = dentry;
			smp_mb();
			/* relay_channels_mutex must be held, so wait. */
			err = smp_call_function_single(i,
						       __relay_set_buf_dentry,
						       &disp, 1);
		}
		if (unlikely(err))
			break;
	}
	put_cpu();
	mutex_unlock(&relay_channels_mutex);

	return err;
}
724
EXPORT_SYMBOL_GPL(relay_late_setup_files);
725

726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748
/**
 *	relay_switch_subbuf - switch to a new sub-buffer
 *	@buf: channel buffer
 *	@length: size of current event
 *
 *	Returns either the length passed in or 0 if full.
 *
 *	Performs sub-buffer-switch tasks such as invoking callbacks,
 *	updating padding counts, waking up readers, etc.
 */
size_t relay_switch_subbuf(struct rchan_buf *buf, size_t length)
{
	void *old, *new;
	size_t old_subbuf, new_subbuf;

	if (unlikely(length > buf->chan->subbuf_size))
		goto toobig;

	if (buf->offset != buf->chan->subbuf_size + 1) {
		buf->prev_padding = buf->chan->subbuf_size - buf->offset;
		old_subbuf = buf->subbufs_produced % buf->chan->n_subbufs;
		buf->padding[old_subbuf] = buf->prev_padding;
		buf->subbufs_produced++;
749
		if (buf->dentry)
750
			d_inode(buf->dentry)->i_size +=
751 752 753 754 755
				buf->chan->subbuf_size -
				buf->padding[old_subbuf];
		else
			buf->early_bytes += buf->chan->subbuf_size -
					    buf->padding[old_subbuf];
756
		smp_mb();
757 758 759 760 761 762 763
		if (waitqueue_active(&buf->read_wait))
			/*
			 * Calling wake_up_interruptible() from here
			 * will deadlock if we happen to be logging
			 * from the scheduler (trying to re-grab
			 * rq->lock), so defer it.
			 */
I
Ingo Molnar 已提交
764
			mod_timer(&buf->timer, jiffies + 1);
765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798
	}

	old = buf->data;
	new_subbuf = buf->subbufs_produced % buf->chan->n_subbufs;
	new = buf->start + new_subbuf * buf->chan->subbuf_size;
	buf->offset = 0;
	if (!buf->chan->cb->subbuf_start(buf, new, old, buf->prev_padding)) {
		buf->offset = buf->chan->subbuf_size + 1;
		return 0;
	}
	buf->data = new;
	buf->padding[new_subbuf] = 0;

	if (unlikely(length + buf->offset > buf->chan->subbuf_size))
		goto toobig;

	return length;

toobig:
	buf->chan->last_toobig = length;
	return 0;
}
EXPORT_SYMBOL_GPL(relay_switch_subbuf);

/**
 *	relay_subbufs_consumed - update the buffer's sub-buffers-consumed count
 *	@chan: the channel
 *	@cpu: the cpu associated with the channel buffer to update
 *	@subbufs_consumed: number of sub-buffers to add to current buf's count
 *
 *	Adds to the channel buffer's consumed sub-buffer count.
 *	subbufs_consumed should be the number of sub-buffers newly consumed,
 *	not the total consumed.
 *
799
 *	NOTE. Kernel clients don't need to call this function if the channel
800 801 802 803 804 805 806 807 808 809 810
 *	mode is 'overwrite'.
 */
void relay_subbufs_consumed(struct rchan *chan,
			    unsigned int cpu,
			    size_t subbufs_consumed)
{
	struct rchan_buf *buf;

	if (!chan)
		return;

811 812
	buf = *per_cpu_ptr(chan->buf, cpu);
	if (cpu >= NR_CPUS || !buf || subbufs_consumed > chan->n_subbufs)
813 814
		return;

815
	if (subbufs_consumed > buf->subbufs_produced - buf->subbufs_consumed)
816
		buf->subbufs_consumed = buf->subbufs_produced;
817 818
	else
		buf->subbufs_consumed += subbufs_consumed;
819 820 821 822 823 824 825 826 827 828 829
}
EXPORT_SYMBOL_GPL(relay_subbufs_consumed);

/**
 *	relay_close - close the channel
 *	@chan: the channel
 *
 *	Closes all channel buffers and frees the channel.
 */
void relay_close(struct rchan *chan)
{
830
	struct rchan_buf *buf;
831 832 833 834 835
	unsigned int i;

	if (!chan)
		return;

836
	mutex_lock(&relay_channels_mutex);
837 838
	if (chan->is_global && (buf = *per_cpu_ptr(chan->buf, 0)))
		relay_close_buf(buf);
839 840
	else
		for_each_possible_cpu(i)
841 842
			if ((buf = *per_cpu_ptr(chan->buf, i)))
				relay_close_buf(buf);
843 844 845 846 847 848

	if (chan->last_toobig)
		printk(KERN_WARNING "relay: one or more items not logged "
		       "[item size (%Zd) > sub-buffer size (%Zd)]\n",
		       chan->last_toobig, chan->subbuf_size);

849
	list_del(&chan->list);
850
	kref_put(&chan->kref, relay_destroy_channel);
851
	mutex_unlock(&relay_channels_mutex);
852 853 854 855 856 857 858
}
EXPORT_SYMBOL_GPL(relay_close);

/**
 *	relay_flush - close the channel
 *	@chan: the channel
 *
859
 *	Flushes all channel buffers, i.e. forces buffer switch.
860 861 862
 */
void relay_flush(struct rchan *chan)
{
863
	struct rchan_buf *buf;
864 865 866 867 868
	unsigned int i;

	if (!chan)
		return;

869 870
	if (chan->is_global && (buf = *per_cpu_ptr(chan->buf, 0))) {
		relay_switch_subbuf(buf, 0);
871
		return;
872
	}
873 874 875

	mutex_lock(&relay_channels_mutex);
	for_each_possible_cpu(i)
876 877
		if ((buf = *per_cpu_ptr(chan->buf, i)))
			relay_switch_subbuf(buf, 0);
878
	mutex_unlock(&relay_channels_mutex);
879 880 881 882 883 884 885 886 887 888 889 890
}
EXPORT_SYMBOL_GPL(relay_flush);

/**
 *	relay_file_open - open file op for relay files
 *	@inode: the inode
 *	@filp: the file
 *
 *	Increments the channel buffer refcount.
 */
static int relay_file_open(struct inode *inode, struct file *filp)
{
891
	struct rchan_buf *buf = inode->i_private;
892 893 894
	kref_get(&buf->kref);
	filp->private_data = buf;

895
	return nonseekable_open(inode, filp);
896 897 898 899 900 901 902
}

/**
 *	relay_file_mmap - mmap file op for relay files
 *	@filp: the file
 *	@vma: the vma describing what to map
 *
903
 *	Calls upon relay_mmap_buf() to map the file into user space.
904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950
 */
static int relay_file_mmap(struct file *filp, struct vm_area_struct *vma)
{
	struct rchan_buf *buf = filp->private_data;
	return relay_mmap_buf(buf, vma);
}

/**
 *	relay_file_poll - poll file op for relay files
 *	@filp: the file
 *	@wait: poll table
 *
 *	Poll implemention.
 */
static unsigned int relay_file_poll(struct file *filp, poll_table *wait)
{
	unsigned int mask = 0;
	struct rchan_buf *buf = filp->private_data;

	if (buf->finalized)
		return POLLERR;

	if (filp->f_mode & FMODE_READ) {
		poll_wait(filp, &buf->read_wait, wait);
		if (!relay_buf_empty(buf))
			mask |= POLLIN | POLLRDNORM;
	}

	return mask;
}

/**
 *	relay_file_release - release file op for relay files
 *	@inode: the inode
 *	@filp: the file
 *
 *	Decrements the channel refcount, as the filesystem is
 *	no longer using it.
 */
static int relay_file_release(struct inode *inode, struct file *filp)
{
	struct rchan_buf *buf = filp->private_data;
	kref_put(&buf->kref, relay_remove_buf);

	return 0;
}

951
/*
952 953 954 955 956 957 958 959 960 961
 *	relay_file_read_consume - update the consumed count for the buffer
 */
static void relay_file_read_consume(struct rchan_buf *buf,
				    size_t read_pos,
				    size_t bytes_consumed)
{
	size_t subbuf_size = buf->chan->subbuf_size;
	size_t n_subbufs = buf->chan->n_subbufs;
	size_t read_subbuf;

962 963 964 965
	if (buf->subbufs_produced == buf->subbufs_consumed &&
	    buf->offset == buf->bytes_consumed)
		return;

966 967 968 969 970 971
	if (buf->bytes_consumed + bytes_consumed > subbuf_size) {
		relay_subbufs_consumed(buf->chan, buf->cpu, 1);
		buf->bytes_consumed = 0;
	}

	buf->bytes_consumed += bytes_consumed;
M
Masami Hiramatsu 已提交
972 973 974 975
	if (!read_pos)
		read_subbuf = buf->subbufs_consumed % n_subbufs;
	else
		read_subbuf = read_pos / buf->chan->subbuf_size;
976 977 978 979 980 981 982 983 984
	if (buf->bytes_consumed + buf->padding[read_subbuf] == subbuf_size) {
		if ((read_subbuf == buf->subbufs_produced % n_subbufs) &&
		    (buf->offset == subbuf_size))
			return;
		relay_subbufs_consumed(buf->chan, buf->cpu, 1);
		buf->bytes_consumed = 0;
	}
}

985
/*
986 987 988 989 990 991
 *	relay_file_read_avail - boolean, are there unconsumed bytes available?
 */
static int relay_file_read_avail(struct rchan_buf *buf, size_t read_pos)
{
	size_t subbuf_size = buf->chan->subbuf_size;
	size_t n_subbufs = buf->chan->n_subbufs;
992 993
	size_t produced = buf->subbufs_produced;
	size_t consumed = buf->subbufs_consumed;
994

995
	relay_file_read_consume(buf, read_pos, 0);
996

997 998
	consumed = buf->subbufs_consumed;

999 1000 1001 1002
	if (unlikely(buf->offset > subbuf_size)) {
		if (produced == consumed)
			return 0;
		return 1;
1003 1004
	}

1005
	if (unlikely(produced - consumed >= n_subbufs)) {
M
Masami Hiramatsu 已提交
1006
		consumed = produced - n_subbufs + 1;
1007
		buf->subbufs_consumed = consumed;
M
Masami Hiramatsu 已提交
1008
		buf->bytes_consumed = 0;
1009
	}
D
Daniel Walker 已提交
1010

1011 1012 1013 1014 1015
	produced = (produced % n_subbufs) * subbuf_size + buf->offset;
	consumed = (consumed % n_subbufs) * subbuf_size + buf->bytes_consumed;

	if (consumed > produced)
		produced += n_subbufs * subbuf_size;
D
Daniel Walker 已提交
1016

1017 1018 1019 1020
	if (consumed == produced) {
		if (buf->offset == subbuf_size &&
		    buf->subbufs_produced > buf->subbufs_consumed)
			return 1;
1021
		return 0;
1022
	}
1023 1024 1025 1026 1027 1028

	return 1;
}

/**
 *	relay_file_read_subbuf_avail - return bytes available in sub-buffer
1029 1030
 *	@read_pos: file read position
 *	@buf: relay channel buffer
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
 */
static size_t relay_file_read_subbuf_avail(size_t read_pos,
					   struct rchan_buf *buf)
{
	size_t padding, avail = 0;
	size_t read_subbuf, read_offset, write_subbuf, write_offset;
	size_t subbuf_size = buf->chan->subbuf_size;

	write_subbuf = (buf->data - buf->start) / subbuf_size;
	write_offset = buf->offset > subbuf_size ? subbuf_size : buf->offset;
	read_subbuf = read_pos / subbuf_size;
	read_offset = read_pos % subbuf_size;
	padding = buf->padding[read_subbuf];

	if (read_subbuf == write_subbuf) {
		if (read_offset + padding < write_offset)
			avail = write_offset - (read_offset + padding);
	} else
		avail = (subbuf_size - padding) - read_offset;

	return avail;
}

/**
 *	relay_file_read_start_pos - find the first available byte to read
1056 1057
 *	@read_pos: file read position
 *	@buf: relay channel buffer
1058
 *
1059
 *	If the @read_pos is in the middle of padding, return the
1060 1061 1062 1063 1064 1065 1066 1067 1068
 *	position of the first actually available byte, otherwise
 *	return the original value.
 */
static size_t relay_file_read_start_pos(size_t read_pos,
					struct rchan_buf *buf)
{
	size_t read_subbuf, padding, padding_start, padding_end;
	size_t subbuf_size = buf->chan->subbuf_size;
	size_t n_subbufs = buf->chan->n_subbufs;
D
David Wilder 已提交
1069
	size_t consumed = buf->subbufs_consumed % n_subbufs;
1070

D
David Wilder 已提交
1071 1072
	if (!read_pos)
		read_pos = consumed * subbuf_size + buf->bytes_consumed;
1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086
	read_subbuf = read_pos / subbuf_size;
	padding = buf->padding[read_subbuf];
	padding_start = (read_subbuf + 1) * subbuf_size - padding;
	padding_end = (read_subbuf + 1) * subbuf_size;
	if (read_pos >= padding_start && read_pos < padding_end) {
		read_subbuf = (read_subbuf + 1) % n_subbufs;
		read_pos = read_subbuf * subbuf_size;
	}

	return read_pos;
}

/**
 *	relay_file_read_end_pos - return the new read position
1087 1088 1089
 *	@read_pos: file read position
 *	@buf: relay channel buffer
 *	@count: number of bytes to be read
1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
 */
static size_t relay_file_read_end_pos(struct rchan_buf *buf,
				      size_t read_pos,
				      size_t count)
{
	size_t read_subbuf, padding, end_pos;
	size_t subbuf_size = buf->chan->subbuf_size;
	size_t n_subbufs = buf->chan->n_subbufs;

	read_subbuf = read_pos / subbuf_size;
	padding = buf->padding[read_subbuf];
	if (read_pos % subbuf_size + count + padding == subbuf_size)
		end_pos = (read_subbuf + 1) * subbuf_size;
	else
		end_pos = read_pos + count;
	if (end_pos >= subbuf_size * n_subbufs)
		end_pos = 0;

	return end_pos;
}

A
Al Viro 已提交
1111 1112 1113 1114
static ssize_t relay_file_read(struct file *filp,
			       char __user *buffer,
			       size_t count,
			       loff_t *ppos)
1115
{
1116 1117
	struct rchan_buf *buf = filp->private_data;
	size_t read_start, avail;
A
Al Viro 已提交
1118
	size_t written = 0;
1119
	int ret;
1120

A
Al Viro 已提交
1121
	if (!count)
1122 1123
		return 0;

A
Al Viro 已提交
1124
	inode_lock(file_inode(filp));
1125
	do {
A
Al Viro 已提交
1126 1127
		void *from;

1128 1129 1130 1131 1132 1133
		if (!relay_file_read_avail(buf, *ppos))
			break;

		read_start = relay_file_read_start_pos(*ppos, buf);
		avail = relay_file_read_subbuf_avail(read_start, buf);
		if (!avail)
1134 1135
			break;

A
Al Viro 已提交
1136 1137 1138 1139
		avail = min(count, avail);
		from = buf->start + read_start;
		ret = avail;
		if (copy_to_user(buffer, from, avail))
1140 1141
			break;

A
Al Viro 已提交
1142 1143 1144
		buffer += ret;
		written += ret;
		count -= ret;
1145

A
Al Viro 已提交
1146 1147 1148 1149
		relay_file_read_consume(buf, read_start, ret);
		*ppos = relay_file_read_end_pos(buf, read_start, ret);
	} while (count);
	inode_unlock(file_inode(filp));
1150

A
Al Viro 已提交
1151
	return written;
1152 1153
}

1154 1155 1156 1157 1158 1159 1160 1161 1162 1163
static void relay_consume_bytes(struct rchan_buf *rbuf, int bytes_consumed)
{
	rbuf->bytes_consumed += bytes_consumed;

	if (rbuf->bytes_consumed >= rbuf->chan->subbuf_size) {
		relay_subbufs_consumed(rbuf->chan, rbuf->cpu, 1);
		rbuf->bytes_consumed %= rbuf->chan->subbuf_size;
	}
}

T
Tom Zanussi 已提交
1164 1165
static void relay_pipe_buf_release(struct pipe_inode_info *pipe,
				   struct pipe_buffer *buf)
1166
{
T
Tom Zanussi 已提交
1167 1168 1169
	struct rchan_buf *rbuf;

	rbuf = (struct rchan_buf *)page_private(buf->page);
1170
	relay_consume_bytes(rbuf, buf->private);
T
Tom Zanussi 已提交
1171 1172
}

1173
static const struct pipe_buf_operations relay_pipe_buf_ops = {
T
Tom Zanussi 已提交
1174
	.can_merge = 0,
1175
	.confirm = generic_pipe_buf_confirm,
T
Tom Zanussi 已提交
1176 1177 1178 1179 1180
	.release = relay_pipe_buf_release,
	.steal = generic_pipe_buf_steal,
	.get = generic_pipe_buf_get,
};

1181 1182 1183 1184
static void relay_page_release(struct splice_pipe_desc *spd, unsigned int i)
{
}

T
Tom Zanussi 已提交
1185
/*
T
Tom Zanussi 已提交
1186 1187
 *	subbuf_splice_actor - splice up to one subbuf's worth of data
 */
1188
static ssize_t subbuf_splice_actor(struct file *in,
T
Tom Zanussi 已提交
1189 1190 1191 1192 1193 1194
			       loff_t *ppos,
			       struct pipe_inode_info *pipe,
			       size_t len,
			       unsigned int flags,
			       int *nonpad_ret)
{
1195
	unsigned int pidx, poff, total_len, subbuf_pages, nr_pages;
T
Tom Zanussi 已提交
1196 1197
	struct rchan_buf *rbuf = in->private_data;
	unsigned int subbuf_size = rbuf->chan->subbuf_size;
1198 1199 1200
	uint64_t pos = (uint64_t) *ppos;
	uint32_t alloc_size = (uint32_t) rbuf->chan->alloc_size;
	size_t read_start = (size_t) do_div(pos, alloc_size);
T
Tom Zanussi 已提交
1201 1202 1203
	size_t read_subbuf = read_start / subbuf_size;
	size_t padding = rbuf->padding[read_subbuf];
	size_t nonpad_end = read_subbuf * subbuf_size + subbuf_size - padding;
1204 1205
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
1206 1207 1208
	struct splice_pipe_desc spd = {
		.pages = pages,
		.nr_pages = 0,
1209
		.nr_pages_max = PIPE_DEF_BUFFERS,
1210 1211 1212
		.partial = partial,
		.flags = flags,
		.ops = &relay_pipe_buf_ops,
1213
		.spd_release = relay_page_release,
1214
	};
1215
	ssize_t ret;
T
Tom Zanussi 已提交
1216 1217 1218

	if (rbuf->subbufs_produced == rbuf->subbufs_consumed)
		return 0;
1219 1220
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;
T
Tom Zanussi 已提交
1221

1222 1223 1224 1225 1226
	/*
	 * Adjust read len, if longer than what is available
	 */
	if (len > (subbuf_size - read_start % subbuf_size))
		len = subbuf_size - read_start % subbuf_size;
T
Tom Zanussi 已提交
1227 1228 1229 1230

	subbuf_pages = rbuf->chan->alloc_size >> PAGE_SHIFT;
	pidx = (read_start / PAGE_SIZE) % subbuf_pages;
	poff = read_start & ~PAGE_MASK;
1231
	nr_pages = min_t(unsigned int, subbuf_pages, spd.nr_pages_max);
T
Tom Zanussi 已提交
1232

1233
	for (total_len = 0; spd.nr_pages < nr_pages; spd.nr_pages++) {
1234 1235
		unsigned int this_len, this_end, private;
		unsigned int cur_pos = read_start + total_len;
T
Tom Zanussi 已提交
1236

1237
		if (!len)
T
Tom Zanussi 已提交
1238 1239
			break;

1240 1241
		this_len = min_t(unsigned long, len, PAGE_SIZE - poff);
		private = this_len;
T
Tom Zanussi 已提交
1242

1243 1244
		spd.pages[spd.nr_pages] = rbuf->page_array[pidx];
		spd.partial[spd.nr_pages].offset = poff;
T
Tom Zanussi 已提交
1245

1246 1247 1248 1249
		this_end = cur_pos + this_len;
		if (this_end >= nonpad_end) {
			this_len = nonpad_end - cur_pos;
			private = this_len + padding;
T
Tom Zanussi 已提交
1250
		}
1251 1252
		spd.partial[spd.nr_pages].len = this_len;
		spd.partial[spd.nr_pages].private = private;
T
Tom Zanussi 已提交
1253

1254 1255 1256 1257
		len -= this_len;
		total_len += this_len;
		poff = 0;
		pidx = (pidx + 1) % subbuf_pages;
T
Tom Zanussi 已提交
1258

1259 1260
		if (this_end >= nonpad_end) {
			spd.nr_pages++;
T
Tom Zanussi 已提交
1261 1262 1263 1264
			break;
		}
	}

1265
	ret = 0;
1266
	if (!spd.nr_pages)
1267
		goto out;
T
Tom Zanussi 已提交
1268

1269 1270
	ret = *nonpad_ret = splice_to_pipe(pipe, &spd);
	if (ret < 0 || ret < total_len)
1271
		goto out;
T
Tom Zanussi 已提交
1272

1273 1274 1275
        if (read_start + ret == nonpad_end)
                ret += padding;

1276
out:
1277 1278
	splice_shrink_spd(&spd);
	return ret;
T
Tom Zanussi 已提交
1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293
}

static ssize_t relay_file_splice_read(struct file *in,
				      loff_t *ppos,
				      struct pipe_inode_info *pipe,
				      size_t len,
				      unsigned int flags)
{
	ssize_t spliced;
	int ret;
	int nonpad_ret = 0;

	ret = 0;
	spliced = 0;

1294
	while (len && !spliced) {
T
Tom Zanussi 已提交
1295 1296 1297 1298
		ret = subbuf_splice_actor(in, ppos, pipe, len, flags, &nonpad_ret);
		if (ret < 0)
			break;
		else if (!ret) {
1299
			if (flags & SPLICE_F_NONBLOCK)
T
Tom Zanussi 已提交
1300
				ret = -EAGAIN;
1301
			break;
T
Tom Zanussi 已提交
1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
		}

		*ppos += ret;
		if (ret > len)
			len = 0;
		else
			len -= ret;
		spliced += nonpad_ret;
		nonpad_ret = 0;
	}

	if (spliced)
		return spliced;

	return ret;
1317 1318
}

1319
const struct file_operations relay_file_operations = {
1320 1321 1322 1323 1324 1325
	.open		= relay_file_open,
	.poll		= relay_file_poll,
	.mmap		= relay_file_mmap,
	.read		= relay_file_read,
	.llseek		= no_llseek,
	.release	= relay_file_release,
T
Tom Zanussi 已提交
1326
	.splice_read	= relay_file_splice_read,
1327 1328
};
EXPORT_SYMBOL_GPL(relay_file_operations);