virtio_balloon.c 19.7 KB
Newer Older
S
Sasha Levin 已提交
1 2
/*
 * Virtio balloon implementation, inspired by Dor Laor and Marcelo
R
Rusty Russell 已提交
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
 * Tosatti's implementations.
 *
 *  Copyright 2008 Rusty Russell IBM Corporation
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
S
Sasha Levin 已提交
21

R
Rusty Russell 已提交
22 23 24
#include <linux/virtio.h>
#include <linux/virtio_balloon.h>
#include <linux/swap.h>
25
#include <linux/workqueue.h>
26
#include <linux/delay.h>
27
#include <linux/slab.h>
28
#include <linux/module.h>
29
#include <linux/balloon_compaction.h>
30
#include <linux/oom.h>
31
#include <linux/wait.h>
32
#include <linux/mm.h>
33
#include <linux/mount.h>
34
#include <linux/magic.h>
R
Rusty Russell 已提交
35

36 37 38 39 40
/*
 * Balloon device works in 4K page units.  So each page is pointed to by
 * multiple balloon pages.  All memory counters in this driver are in balloon
 * page units.
 */
41 42
#define VIRTIO_BALLOON_PAGES_PER_PAGE (unsigned)(PAGE_SIZE >> VIRTIO_BALLOON_PFN_SHIFT)
#define VIRTIO_BALLOON_ARRAY_PFNS_MAX 256
43 44 45 46 47 48
#define OOM_VBALLOON_DEFAULT_PAGES 256
#define VIRTBALLOON_OOM_NOTIFY_PRIORITY 80

static int oom_pages = OOM_VBALLOON_DEFAULT_PAGES;
module_param(oom_pages, int, S_IRUSR | S_IWUSR);
MODULE_PARM_DESC(oom_pages, "pages to free on OOM");
49

50 51 52 53
#ifdef CONFIG_BALLOON_COMPACTION
static struct vfsmount *balloon_mnt;
#endif

54
struct virtio_balloon {
R
Rusty Russell 已提交
55
	struct virtio_device *vdev;
56
	struct virtqueue *inflate_vq, *deflate_vq, *stats_vq;
R
Rusty Russell 已提交
57

58
	/* The balloon servicing is delegated to a freezable workqueue. */
59 60
	struct work_struct update_balloon_stats_work;
	struct work_struct update_balloon_size_work;
R
Rusty Russell 已提交
61

62 63 64
	/* Prevent updating balloon when it is being canceled. */
	spinlock_t stop_update_lock;
	bool stop_update;
R
Rusty Russell 已提交
65 66

	/* Waiting for host to ack the pages we released. */
67
	wait_queue_head_t acked;
R
Rusty Russell 已提交
68

69
	/* Number of balloon pages we've told the Host we're not using. */
R
Rusty Russell 已提交
70
	unsigned int num_pages;
71
	/*
72 73
	 * The pages we've told the Host we're not using are enqueued
	 * at vb_dev_info->pages list.
74 75 76
	 * Each page on this list adds VIRTIO_BALLOON_PAGES_PER_PAGE
	 * to num_pages above.
	 */
77
	struct balloon_dev_info vb_dev_info;
78 79 80

	/* Synchronize access/update to this struct virtio_balloon elements */
	struct mutex balloon_lock;
R
Rusty Russell 已提交
81 82 83

	/* The array of pfns we tell the Host about. */
	unsigned int num_pfns;
84
	__virtio32 pfns[VIRTIO_BALLOON_ARRAY_PFNS_MAX];
85 86 87

	/* Memory statistics */
	struct virtio_balloon_stat stats[VIRTIO_BALLOON_S_NR];
88 89 90

	/* To register callback in oom notifier call chain */
	struct notifier_block nb;
R
Rusty Russell 已提交
91 92 93 94 95 96 97
};

static struct virtio_device_id id_table[] = {
	{ VIRTIO_ID_BALLOON, VIRTIO_DEV_ANY_ID },
	{ 0 },
};

98 99 100 101 102 103
static u32 page_to_balloon_pfn(struct page *page)
{
	unsigned long pfn = page_to_pfn(page);

	BUILD_BUG_ON(PAGE_SHIFT < VIRTIO_BALLOON_PFN_SHIFT);
	/* Convert pfn from Linux page size to balloon page size. */
104 105 106
	return pfn * VIRTIO_BALLOON_PAGES_PER_PAGE;
}

R
Rusty Russell 已提交
107 108
static void balloon_ack(struct virtqueue *vq)
{
109
	struct virtio_balloon *vb = vq->vdev->priv;
R
Rusty Russell 已提交
110

111
	wake_up(&vb->acked);
R
Rusty Russell 已提交
112 113 114 115 116
}

static void tell_host(struct virtio_balloon *vb, struct virtqueue *vq)
{
	struct scatterlist sg;
117
	unsigned int len;
R
Rusty Russell 已提交
118 119 120 121

	sg_init_one(&sg, vb->pfns, sizeof(vb->pfns[0]) * vb->num_pfns);

	/* We should always be able to add one buffer to an empty queue. */
122
	virtqueue_add_outbuf(vq, &sg, 1, vb, GFP_KERNEL);
123
	virtqueue_kick(vq);
R
Rusty Russell 已提交
124 125

	/* When host has read buffer, this completes via balloon_ack */
126
	wait_event(vb->acked, virtqueue_get_buf(vq, &len));
127

R
Rusty Russell 已提交
128 129
}

130 131
static void set_page_pfns(struct virtio_balloon *vb,
			  __virtio32 pfns[], struct page *page)
132 133 134
{
	unsigned int i;

W
Wei Wang 已提交
135 136 137 138
	/*
	 * Set balloon pfns pointing at this page.
	 * Note that the first pfn points at start of the page.
	 */
139
	for (i = 0; i < VIRTIO_BALLOON_PAGES_PER_PAGE; i++)
140 141
		pfns[i] = cpu_to_virtio32(vb->vdev,
					  page_to_balloon_pfn(page) + i);
142 143
}

144
static unsigned fill_balloon(struct virtio_balloon *vb, size_t num)
R
Rusty Russell 已提交
145
{
146
	unsigned num_allocated_pages;
147 148 149
	unsigned num_pfns;
	struct page *page;
	LIST_HEAD(pages);
150

R
Rusty Russell 已提交
151 152 153
	/* We can only do one array worth at a time. */
	num = min(num, ARRAY_SIZE(vb->pfns));

154 155 156
	for (num_pfns = 0; num_pfns < num;
	     num_pfns += VIRTIO_BALLOON_PAGES_PER_PAGE) {
		struct page *page = balloon_page_alloc();
157

R
Rusty Russell 已提交
158
		if (!page) {
159
			dev_info_ratelimited(&vb->vdev->dev,
160 161
					     "Out of puff! Can't get %u pages\n",
					     VIRTIO_BALLOON_PAGES_PER_PAGE);
R
Rusty Russell 已提交
162 163 164 165
			/* Sleep for at least 1/5 of a second before retry. */
			msleep(200);
			break;
		}
166 167 168 169 170 171 172 173 174 175 176

		balloon_page_push(&pages, page);
	}

	mutex_lock(&vb->balloon_lock);

	vb->num_pfns = 0;

	while ((page = balloon_page_pop(&pages))) {
		balloon_page_enqueue(&vb->vb_dev_info, page);

177
		set_page_pfns(vb, vb->pfns + vb->num_pfns, page);
178
		vb->num_pages += VIRTIO_BALLOON_PAGES_PER_PAGE;
179 180 181
		if (!virtio_has_feature(vb->vdev,
					VIRTIO_BALLOON_F_DEFLATE_ON_OOM))
			adjust_managed_page_count(page, -1);
182
		vb->num_pfns += VIRTIO_BALLOON_PAGES_PER_PAGE;
R
Rusty Russell 已提交
183 184
	}

185
	num_allocated_pages = vb->num_pfns;
186 187 188 189
	/* Did we get any? */
	if (vb->num_pfns != 0)
		tell_host(vb, vb->inflate_vq);
	mutex_unlock(&vb->balloon_lock);
190 191

	return num_allocated_pages;
R
Rusty Russell 已提交
192 193
}

194 195
static void release_pages_balloon(struct virtio_balloon *vb,
				 struct list_head *pages)
R
Rusty Russell 已提交
196
{
197
	struct page *page, *next;
R
Rusty Russell 已提交
198

199
	list_for_each_entry_safe(page, next, pages, lru) {
200 201 202
		if (!virtio_has_feature(vb->vdev,
					VIRTIO_BALLOON_F_DEFLATE_ON_OOM))
			adjust_managed_page_count(page, 1);
203
		list_del(&page->lru);
204
		put_page(page); /* balloon reference */
R
Rusty Russell 已提交
205 206 207
	}
}

208
static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)
R
Rusty Russell 已提交
209
{
210
	unsigned num_freed_pages;
R
Rusty Russell 已提交
211
	struct page *page;
212
	struct balloon_dev_info *vb_dev_info = &vb->vb_dev_info;
213
	LIST_HEAD(pages);
R
Rusty Russell 已提交
214 215 216 217

	/* We can only do one array worth at a time. */
	num = min(num, ARRAY_SIZE(vb->pfns));

218
	mutex_lock(&vb->balloon_lock);
219 220
	/* We can't release more pages than taken */
	num = min(num, (size_t)vb->num_pages);
221 222
	for (vb->num_pfns = 0; vb->num_pfns < num;
	     vb->num_pfns += VIRTIO_BALLOON_PAGES_PER_PAGE) {
223 224 225
		page = balloon_page_dequeue(vb_dev_info);
		if (!page)
			break;
226
		set_page_pfns(vb, vb->pfns + vb->num_pfns, page);
227
		list_add(&page->lru, &pages);
228
		vb->num_pages -= VIRTIO_BALLOON_PAGES_PER_PAGE;
R
Rusty Russell 已提交
229 230
	}

231
	num_freed_pages = vb->num_pfns;
232 233 234 235 236
	/*
	 * Note that if
	 * virtio_has_feature(vdev, VIRTIO_BALLOON_F_MUST_TELL_HOST);
	 * is true, we *have* to do it in this order
	 */
237 238
	if (vb->num_pfns != 0)
		tell_host(vb, vb->deflate_vq);
239
	release_pages_balloon(vb, &pages);
240
	mutex_unlock(&vb->balloon_lock);
241
	return num_freed_pages;
R
Rusty Russell 已提交
242 243
}

244 245 246 247
static inline void update_stat(struct virtio_balloon *vb, int idx,
			       u16 tag, u64 val)
{
	BUG_ON(idx >= VIRTIO_BALLOON_S_NR);
248 249
	vb->stats[idx].tag = cpu_to_virtio16(vb->vdev, tag);
	vb->stats[idx].val = cpu_to_virtio64(vb->vdev, val);
250 251 252 253
}

#define pages_to_bytes(x) ((u64)(x) << PAGE_SHIFT)

254
static unsigned int update_balloon_stats(struct virtio_balloon *vb)
255 256 257
{
	unsigned long events[NR_VM_EVENT_ITEMS];
	struct sysinfo i;
258
	unsigned int idx = 0;
259
	long available;
260
	unsigned long caches;
261 262 263 264

	all_vm_events(events);
	si_meminfo(&i);

265
	available = si_mem_available();
266
	caches = global_node_page_state(NR_FILE_PAGES);
267

268
#ifdef CONFIG_VM_EVENT_COUNTERS
269 270 271 272 273 274
	update_stat(vb, idx++, VIRTIO_BALLOON_S_SWAP_IN,
				pages_to_bytes(events[PSWPIN]));
	update_stat(vb, idx++, VIRTIO_BALLOON_S_SWAP_OUT,
				pages_to_bytes(events[PSWPOUT]));
	update_stat(vb, idx++, VIRTIO_BALLOON_S_MAJFLT, events[PGMAJFAULT]);
	update_stat(vb, idx++, VIRTIO_BALLOON_S_MINFLT, events[PGFAULT]);
275 276 277 278 279 280
#ifdef CONFIG_HUGETLB_PAGE
	update_stat(vb, idx++, VIRTIO_BALLOON_S_HTLB_PGALLOC,
		    events[HTLB_BUDDY_PGALLOC]);
	update_stat(vb, idx++, VIRTIO_BALLOON_S_HTLB_PGFAIL,
		    events[HTLB_BUDDY_PGALLOC_FAIL]);
#endif
281
#endif
282 283 284 285
	update_stat(vb, idx++, VIRTIO_BALLOON_S_MEMFREE,
				pages_to_bytes(i.freeram));
	update_stat(vb, idx++, VIRTIO_BALLOON_S_MEMTOT,
				pages_to_bytes(i.totalram));
286 287
	update_stat(vb, idx++, VIRTIO_BALLOON_S_AVAIL,
				pages_to_bytes(available));
288 289
	update_stat(vb, idx++, VIRTIO_BALLOON_S_CACHES,
				pages_to_bytes(caches));
290 291

	return idx;
292 293 294 295 296 297 298
}

/*
 * While most virtqueues communicate guest-initiated requests to the hypervisor,
 * the stats queue operates in reverse.  The driver initializes the virtqueue
 * with a single buffer.  From that point forward, all conversations consist of
 * a hypervisor request (a call to this function) which directs us to refill
299
 * the virtqueue with a fresh stats buffer.  Since stats collection can sleep,
300 301
 * we delegate the job to a freezable workqueue that will do the actual work via
 * stats_handle_request().
302
 */
303
static void stats_request(struct virtqueue *vq)
304
{
305
	struct virtio_balloon *vb = vq->vdev->priv;
306

307 308
	spin_lock(&vb->stop_update_lock);
	if (!vb->stop_update)
309
		queue_work(system_freezable_wq, &vb->update_balloon_stats_work);
310
	spin_unlock(&vb->stop_update_lock);
311 312 313 314 315 316
}

static void stats_handle_request(struct virtio_balloon *vb)
{
	struct virtqueue *vq;
	struct scatterlist sg;
317
	unsigned int len, num_stats;
318

319
	num_stats = update_balloon_stats(vb);
320

321
	vq = vb->stats_vq;
322 323
	if (!virtqueue_get_buf(vq, &len))
		return;
324
	sg_init_one(&sg, vb->stats, sizeof(vb->stats[0]) * num_stats);
325
	virtqueue_add_outbuf(vq, &sg, 1, vb, GFP_KERNEL);
326
	virtqueue_kick(vq);
327 328
}

R
Rusty Russell 已提交
329 330 331
static void virtballoon_changed(struct virtio_device *vdev)
{
	struct virtio_balloon *vb = vdev->priv;
332
	unsigned long flags;
R
Rusty Russell 已提交
333

334 335
	spin_lock_irqsave(&vb->stop_update_lock, flags);
	if (!vb->stop_update)
336
		queue_work(system_freezable_wq, &vb->update_balloon_size_work);
337
	spin_unlock_irqrestore(&vb->stop_update_lock, flags);
R
Rusty Russell 已提交
338 339
}

340
static inline s64 towards_target(struct virtio_balloon *vb)
R
Rusty Russell 已提交
341
{
D
David Gibson 已提交
342
	s64 target;
343
	u32 num_pages;
D
David Gibson 已提交
344

345 346
	virtio_cread(vb->vdev, struct virtio_balloon_config, num_pages,
		     &num_pages);
347

348 349 350 351 352
	/* Legacy balloon config space is LE, unlike all other devices. */
	if (!virtio_has_feature(vb->vdev, VIRTIO_F_VERSION_1))
		num_pages = le32_to_cpu((__force __le32)num_pages);

	target = num_pages;
D
David Gibson 已提交
353
	return target - vb->num_pages;
R
Rusty Russell 已提交
354 355 356 357
}

static void update_balloon_size(struct virtio_balloon *vb)
{
358 359 360 361 362
	u32 actual = vb->num_pages;

	/* Legacy balloon config space is LE, unlike all other devices. */
	if (!virtio_has_feature(vb->vdev, VIRTIO_F_VERSION_1))
		actual = (__force u32)cpu_to_le32(actual);
R
Rusty Russell 已提交
363

364
	virtio_cwrite(vb->vdev, struct virtio_balloon_config, actual,
365
		      &actual);
R
Rusty Russell 已提交
366 367
}

368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399
/*
 * virtballoon_oom_notify - release pages when system is under severe
 *			    memory pressure (called from out_of_memory())
 * @self : notifier block struct
 * @dummy: not used
 * @parm : returned - number of freed pages
 *
 * The balancing of memory by use of the virtio balloon should not cause
 * the termination of processes while there are pages in the balloon.
 * If virtio balloon manages to release some memory, it will make the
 * system return and retry the allocation that forced the OOM killer
 * to run.
 */
static int virtballoon_oom_notify(struct notifier_block *self,
				  unsigned long dummy, void *parm)
{
	struct virtio_balloon *vb;
	unsigned long *freed;
	unsigned num_freed_pages;

	vb = container_of(self, struct virtio_balloon, nb);
	if (!virtio_has_feature(vb->vdev, VIRTIO_BALLOON_F_DEFLATE_ON_OOM))
		return NOTIFY_OK;

	freed = parm;
	num_freed_pages = leak_balloon(vb, oom_pages);
	update_balloon_size(vb);
	*freed += num_freed_pages;

	return NOTIFY_OK;
}

400
static void update_balloon_stats_func(struct work_struct *work)
R
Rusty Russell 已提交
401
{
402
	struct virtio_balloon *vb;
403

404 405 406 407
	vb = container_of(work, struct virtio_balloon,
			  update_balloon_stats_work);
	stats_handle_request(vb);
}
408

409
static void update_balloon_size_func(struct work_struct *work)
R
Rusty Russell 已提交
410
{
411 412
	struct virtio_balloon *vb;
	s64 diff;
413

414 415
	vb = container_of(work, struct virtio_balloon,
			  update_balloon_size_work);
416
	diff = towards_target(vb);
417

418 419 420 421 422 423 424 425
	if (diff > 0)
		diff -= fill_balloon(vb, diff);
	else if (diff < 0)
		diff += leak_balloon(vb, -diff);
	update_balloon_size(vb);

	if (diff)
		queue_work(system_freezable_wq, work);
R
Rusty Russell 已提交
426 427
}

428
static int init_vqs(struct virtio_balloon *vb)
R
Rusty Russell 已提交
429
{
430
	struct virtqueue *vqs[3];
431
	vq_callback_t *callbacks[] = { balloon_ack, balloon_ack, stats_request };
432
	static const char * const names[] = { "inflate", "deflate", "stats" };
433
	int err, nvqs;
R
Rusty Russell 已提交
434

435 436 437 438
	/*
	 * We expect two virtqueues: inflate and deflate, and
	 * optionally stat.
	 */
439
	nvqs = virtio_has_feature(vb->vdev, VIRTIO_BALLOON_F_STATS_VQ) ? 3 : 2;
M
Michael S. Tsirkin 已提交
440
	err = virtio_find_vqs(vb->vdev, nvqs, vqs, callbacks, names, NULL);
441
	if (err)
442
		return err;
R
Rusty Russell 已提交
443

444 445
	vb->inflate_vq = vqs[0];
	vb->deflate_vq = vqs[1];
446 447
	if (virtio_has_feature(vb->vdev, VIRTIO_BALLOON_F_STATS_VQ)) {
		struct scatterlist sg;
448
		unsigned int num_stats;
449 450 451 452
		vb->stats_vq = vqs[2];

		/*
		 * Prime this virtqueue with one buffer so the hypervisor can
453
		 * use it to signal us later (it can't be broken yet!).
454
		 */
455
		num_stats = update_balloon_stats(vb);
456

457
		sg_init_one(&sg, vb->stats, sizeof(vb->stats[0]) * num_stats);
458
		if (virtqueue_add_outbuf(vb->stats_vq, &sg, 1, vb, GFP_KERNEL)
459
		    < 0)
460
			BUG();
461
		virtqueue_kick(vb->stats_vq);
462
	}
463 464 465
	return 0;
}

466 467 468 469
#ifdef CONFIG_BALLOON_COMPACTION
/*
 * virtballoon_migratepage - perform the balloon page migration on behalf of
 *			     a compation thread.     (called under page lock)
470
 * @vb_dev_info: the balloon device
471 472 473 474 475 476 477 478 479 480 481 482 483 484
 * @newpage: page that will replace the isolated page after migration finishes.
 * @page   : the isolated (old) page that is about to be migrated to newpage.
 * @mode   : compaction mode -- not used for balloon page migration.
 *
 * After a ballooned page gets isolated by compaction procedures, this is the
 * function that performs the page migration on behalf of a compaction thread
 * The page migration for virtio balloon is done in a simple swap fashion which
 * follows these two macro steps:
 *  1) insert newpage into vb->pages list and update the host about it;
 *  2) update the host about the old page removed from vb->pages list;
 *
 * This function preforms the balloon page migration task.
 * Called through balloon_mapping->a_ops->migratepage
 */
485
static int virtballoon_migratepage(struct balloon_dev_info *vb_dev_info,
486 487
		struct page *newpage, struct page *page, enum migrate_mode mode)
{
488 489
	struct virtio_balloon *vb = container_of(vb_dev_info,
			struct virtio_balloon, vb_dev_info);
490 491 492 493 494 495 496 497 498 499 500 501 502
	unsigned long flags;

	/*
	 * In order to avoid lock contention while migrating pages concurrently
	 * to leak_balloon() or fill_balloon() we just give up the balloon_lock
	 * this turn, as it is easier to retry the page migration later.
	 * This also prevents fill_balloon() getting stuck into a mutex
	 * recursion in the case it ends up triggering memory compaction
	 * while it is attempting to inflate the ballon.
	 */
	if (!mutex_trylock(&vb->balloon_lock))
		return -EAGAIN;

503 504
	get_page(newpage); /* balloon reference */

505 506
	/* balloon's page migration 1st step  -- inflate "newpage" */
	spin_lock_irqsave(&vb_dev_info->pages_lock, flags);
507
	balloon_page_insert(vb_dev_info, newpage);
508
	vb_dev_info->isolated_pages--;
509
	__count_vm_event(BALLOON_MIGRATE);
510 511
	spin_unlock_irqrestore(&vb_dev_info->pages_lock, flags);
	vb->num_pfns = VIRTIO_BALLOON_PAGES_PER_PAGE;
512
	set_page_pfns(vb, vb->pfns, newpage);
513 514
	tell_host(vb, vb->inflate_vq);

515
	/* balloon's page migration 2nd step -- deflate "page" */
516 517
	balloon_page_delete(page);
	vb->num_pfns = VIRTIO_BALLOON_PAGES_PER_PAGE;
518
	set_page_pfns(vb, vb->pfns, page);
519 520 521 522
	tell_host(vb, vb->deflate_vq);

	mutex_unlock(&vb->balloon_lock);

523 524
	put_page(page); /* balloon reference */

525
	return MIGRATEPAGE_SUCCESS;
526
}
527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544

static struct dentry *balloon_mount(struct file_system_type *fs_type,
		int flags, const char *dev_name, void *data)
{
	static const struct dentry_operations ops = {
		.d_dname = simple_dname,
	};

	return mount_pseudo(fs_type, "balloon-kvm:", NULL, &ops,
				BALLOON_KVM_MAGIC);
}

static struct file_system_type balloon_fs = {
	.name           = "balloon-kvm",
	.mount          = balloon_mount,
	.kill_sb        = kill_anon_super,
};

545 546
#endif /* CONFIG_BALLOON_COMPACTION */

547 548 549 550 551
static int virtballoon_probe(struct virtio_device *vdev)
{
	struct virtio_balloon *vb;
	int err;

552 553 554 555 556 557
	if (!vdev->config->get) {
		dev_err(&vdev->dev, "%s failure: config access disabled\n",
			__func__);
		return -EINVAL;
	}

558 559 560 561 562 563
	vdev->priv = vb = kmalloc(sizeof(*vb), GFP_KERNEL);
	if (!vb) {
		err = -ENOMEM;
		goto out;
	}

564 565
	INIT_WORK(&vb->update_balloon_stats_work, update_balloon_stats_func);
	INIT_WORK(&vb->update_balloon_size_work, update_balloon_size_func);
566 567
	spin_lock_init(&vb->stop_update_lock);
	vb->stop_update = false;
568
	vb->num_pages = 0;
569
	mutex_init(&vb->balloon_lock);
570
	init_waitqueue_head(&vb->acked);
571 572
	vb->vdev = vdev;

573
	balloon_devinfo_init(&vb->vb_dev_info);
574

575 576
	err = init_vqs(vb);
	if (err)
577
		goto out_free_vb;
R
Rusty Russell 已提交
578

579 580 581 582
	vb->nb.notifier_call = virtballoon_oom_notify;
	vb->nb.priority = VIRTBALLOON_OOM_NOTIFY_PRIORITY;
	err = register_oom_notifier(&vb->nb);
	if (err < 0)
583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603
		goto out_del_vqs;

#ifdef CONFIG_BALLOON_COMPACTION
	balloon_mnt = kern_mount(&balloon_fs);
	if (IS_ERR(balloon_mnt)) {
		err = PTR_ERR(balloon_mnt);
		unregister_oom_notifier(&vb->nb);
		goto out_del_vqs;
	}

	vb->vb_dev_info.migratepage = virtballoon_migratepage;
	vb->vb_dev_info.inode = alloc_anon_inode(balloon_mnt->mnt_sb);
	if (IS_ERR(vb->vb_dev_info.inode)) {
		err = PTR_ERR(vb->vb_dev_info.inode);
		kern_unmount(balloon_mnt);
		unregister_oom_notifier(&vb->nb);
		vb->vb_dev_info.inode = NULL;
		goto out_del_vqs;
	}
	vb->vb_dev_info.inode->i_mapping->a_ops = &balloon_aops;
#endif
604

605 606
	virtio_device_ready(vdev);

607 608
	if (towards_target(vb))
		virtballoon_changed(vdev);
R
Rusty Russell 已提交
609 610
	return 0;

611
out_del_vqs:
612
	vdev->config->del_vqs(vdev);
R
Rusty Russell 已提交
613 614 615 616 617 618
out_free_vb:
	kfree(vb);
out:
	return err;
}

619
static void remove_common(struct virtio_balloon *vb)
R
Rusty Russell 已提交
620 621 622 623
{
	/* There might be pages left in the balloon: free them. */
	while (vb->num_pages)
		leak_balloon(vb, vb->num_pages);
624
	update_balloon_size(vb);
R
Rusty Russell 已提交
625 626

	/* Now we reset the device so we can clean up the queues. */
627
	vb->vdev->config->reset(vb->vdev);
R
Rusty Russell 已提交
628

629 630 631
	vb->vdev->config->del_vqs(vb->vdev);
}

632
static void virtballoon_remove(struct virtio_device *vdev)
633 634 635
{
	struct virtio_balloon *vb = vdev->priv;

636
	unregister_oom_notifier(&vb->nb);
637 638 639 640

	spin_lock_irq(&vb->stop_update_lock);
	vb->stop_update = true;
	spin_unlock_irq(&vb->stop_update_lock);
641 642
	cancel_work_sync(&vb->update_balloon_size_work);
	cancel_work_sync(&vb->update_balloon_stats_work);
643

644
	remove_common(vb);
645
#ifdef CONFIG_BALLOON_COMPACTION
646 647
	if (vb->vb_dev_info.inode)
		iput(vb->vb_dev_info.inode);
648 649 650

	kern_unmount(balloon_mnt);
#endif
R
Rusty Russell 已提交
651 652 653
	kfree(vb);
}

654
#ifdef CONFIG_PM_SLEEP
655 656
static int virtballoon_freeze(struct virtio_device *vdev)
{
657 658
	struct virtio_balloon *vb = vdev->priv;

659
	/*
660
	 * The workqueue is already frozen by the PM core before this
661 662
	 * function is called.
	 */
663
	remove_common(vb);
664 665 666
	return 0;
}

667
static int virtballoon_restore(struct virtio_device *vdev)
668 669 670 671 672 673 674 675
{
	struct virtio_balloon *vb = vdev->priv;
	int ret;

	ret = init_vqs(vdev->priv);
	if (ret)
		return ret;

676 677
	virtio_device_ready(vdev);

678 679
	if (towards_target(vb))
		virtballoon_changed(vdev);
680 681 682
	update_balloon_size(vb);
	return 0;
}
683 684
#endif

685 686 687 688 689 690
static int virtballoon_validate(struct virtio_device *vdev)
{
	__virtio_clear_bit(vdev, VIRTIO_F_IOMMU_PLATFORM);
	return 0;
}

691 692 693
static unsigned int features[] = {
	VIRTIO_BALLOON_F_MUST_TELL_HOST,
	VIRTIO_BALLOON_F_STATS_VQ,
694
	VIRTIO_BALLOON_F_DEFLATE_ON_OOM,
695
};
696

697
static struct virtio_driver virtio_balloon_driver = {
698 699
	.feature_table = features,
	.feature_table_size = ARRAY_SIZE(features),
R
Rusty Russell 已提交
700 701 702
	.driver.name =	KBUILD_MODNAME,
	.driver.owner =	THIS_MODULE,
	.id_table =	id_table,
703
	.validate =	virtballoon_validate,
R
Rusty Russell 已提交
704
	.probe =	virtballoon_probe,
705
	.remove =	virtballoon_remove,
R
Rusty Russell 已提交
706
	.config_changed = virtballoon_changed,
707
#ifdef CONFIG_PM_SLEEP
708 709 710
	.freeze	=	virtballoon_freeze,
	.restore =	virtballoon_restore,
#endif
R
Rusty Russell 已提交
711 712
};

713
module_virtio_driver(virtio_balloon_driver);
R
Rusty Russell 已提交
714 715 716
MODULE_DEVICE_TABLE(virtio, id_table);
MODULE_DESCRIPTION("Virtio balloon driver");
MODULE_LICENSE("GPL");