1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Virtio-mem device driver.
4 *
5 * Copyright Red Hat, Inc. 2020
6 *
7 * Author(s): David Hildenbrand <david@redhat.com>
8 */
9
10#include <linux/virtio.h>
11#include <linux/virtio_mem.h>
12#include <linux/workqueue.h>
13#include <linux/slab.h>
14#include <linux/module.h>
15#include <linux/mm.h>
16#include <linux/memory_hotplug.h>
17#include <linux/memory.h>
18#include <linux/hrtimer.h>
19#include <linux/crash_dump.h>
20#include <linux/mutex.h>
21#include <linux/bitmap.h>
22#include <linux/lockdep.h>
23
24#include <acpi/acpi_numa.h>
25
26static bool unplug_online = true;
27module_param(unplug_online, bool, 0644);
28MODULE_PARM_DESC(unplug_online, "Try to unplug online memory");
29
30enum virtio_mem_mb_state {
31	/* Unplugged, not added to Linux. Can be reused later. */
32	VIRTIO_MEM_MB_STATE_UNUSED = 0,
33	/* (Partially) plugged, not added to Linux. Error on add_memory(). */
34	VIRTIO_MEM_MB_STATE_PLUGGED,
35	/* Fully plugged, fully added to Linux, offline. */
36	VIRTIO_MEM_MB_STATE_OFFLINE,
37	/* Partially plugged, fully added to Linux, offline. */
38	VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL,
39	/* Fully plugged, fully added to Linux, online. */
40	VIRTIO_MEM_MB_STATE_ONLINE,
41	/* Partially plugged, fully added to Linux, online. */
42	VIRTIO_MEM_MB_STATE_ONLINE_PARTIAL,
43	VIRTIO_MEM_MB_STATE_COUNT
44};
45
46struct virtio_mem {
47	struct virtio_device *vdev;
48
49	/* We might first have to unplug all memory when starting up. */
50	bool unplug_all_required;
51
52	/* Workqueue that processes the plug/unplug requests. */
53	struct work_struct wq;
54	atomic_t config_changed;
55
56	/* Virtqueue for guest->host requests. */
57	struct virtqueue *vq;
58
59	/* Wait for a host response to a guest request. */
60	wait_queue_head_t host_resp;
61
62	/* Space for one guest request and the host response. */
63	struct virtio_mem_req req;
64	struct virtio_mem_resp resp;
65
66	/* The current size of the device. */
67	uint64_t plugged_size;
68	/* The requested size of the device. */
69	uint64_t requested_size;
70
71	/* The device block size (for communicating with the device). */
72	uint64_t device_block_size;
73	/* The translated node id. NUMA_NO_NODE in case not specified. */
74	int nid;
75	/* Physical start address of the memory region. */
76	uint64_t addr;
77	/* Maximum region size in bytes. */
78	uint64_t region_size;
79
80	/* The subblock size. */
81	uint64_t subblock_size;
82	/* The number of subblocks per memory block. */
83	uint32_t nb_sb_per_mb;
84
85	/* Id of the first memory block of this device. */
86	unsigned long first_mb_id;
87	/* Id of the last memory block of this device. */
88	unsigned long last_mb_id;
89	/* Id of the last usable memory block of this device. */
90	unsigned long last_usable_mb_id;
91	/* Id of the next memory bock to prepare when needed. */
92	unsigned long next_mb_id;
93
94	/* The parent resource for all memory added via this device. */
95	struct resource *parent_resource;
96	/*
97	 * Copy of "System RAM (virtio_mem)" to be used for
98	 * add_memory_driver_managed().
99	 */
100	const char *resource_name;
101
102	/* Summary of all memory block states. */
103	unsigned long nb_mb_state[VIRTIO_MEM_MB_STATE_COUNT];
104#define VIRTIO_MEM_NB_OFFLINE_THRESHOLD		10
105
106	/*
107	 * One byte state per memory block.
108	 *
109	 * Allocated via vmalloc(). When preparing new blocks, resized
110	 * (alloc+copy+free) when needed (crossing pages with the next mb).
111	 * (when crossing pages).
112	 *
113	 * With 128MB memory blocks, we have states for 512GB of memory in one
114	 * page.
115	 */
116	uint8_t *mb_state;
117
118	/*
119	 * $nb_sb_per_mb bit per memory block. Handled similar to mb_state.
120	 *
121	 * With 4MB subblocks, we manage 128GB of memory in one page.
122	 */
123	unsigned long *sb_bitmap;
124
125	/*
126	 * Mutex that protects the nb_mb_state, mb_state, and sb_bitmap.
127	 *
128	 * When this lock is held the pointers can't change, ONLINE and
129	 * OFFLINE blocks can't change the state and no subblocks will get
130	 * plugged/unplugged.
131	 */
132	struct mutex hotplug_mutex;
133	bool hotplug_active;
134
135	/* An error occurred we cannot handle - stop processing requests. */
136	bool broken;
137
138	/* The driver is being removed. */
139	spinlock_t removal_lock;
140	bool removing;
141
142	/* Timer for retrying to plug/unplug memory. */
143	struct hrtimer retry_timer;
144	unsigned int retry_timer_ms;
145#define VIRTIO_MEM_RETRY_TIMER_MIN_MS		50000
146#define VIRTIO_MEM_RETRY_TIMER_MAX_MS		300000
147
148	/* Memory notifier (online/offline events). */
149	struct notifier_block memory_notifier;
150
151	/* Next device in the list of virtio-mem devices. */
152	struct list_head next;
153};
154
155/*
156 * We have to share a single online_page callback among all virtio-mem
157 * devices. We use RCU to iterate the list in the callback.
158 */
159static DEFINE_MUTEX(virtio_mem_mutex);
160static LIST_HEAD(virtio_mem_devices);
161
162static void virtio_mem_online_page_cb(struct page *page, unsigned int order);
163
164/*
165 * Register a virtio-mem device so it will be considered for the online_page
166 * callback.
167 */
168static int register_virtio_mem_device(struct virtio_mem *vm)
169{
170	int rc = 0;
171
172	/* First device registers the callback. */
173	mutex_lock(&virtio_mem_mutex);
174	if (list_empty(&virtio_mem_devices))
175		rc = set_online_page_callback(&virtio_mem_online_page_cb);
176	if (!rc)
177		list_add_rcu(&vm->next, &virtio_mem_devices);
178	mutex_unlock(&virtio_mem_mutex);
179
180	return rc;
181}
182
183/*
184 * Unregister a virtio-mem device so it will no longer be considered for the
185 * online_page callback.
186 */
187static void unregister_virtio_mem_device(struct virtio_mem *vm)
188{
189	/* Last device unregisters the callback. */
190	mutex_lock(&virtio_mem_mutex);
191	list_del_rcu(&vm->next);
192	if (list_empty(&virtio_mem_devices))
193		restore_online_page_callback(&virtio_mem_online_page_cb);
194	mutex_unlock(&virtio_mem_mutex);
195
196	synchronize_rcu();
197}
198
199/*
200 * Calculate the memory block id of a given address.
201 */
202static unsigned long virtio_mem_phys_to_mb_id(unsigned long addr)
203{
204	return addr / memory_block_size_bytes();
205}
206
207/*
208 * Calculate the physical start address of a given memory block id.
209 */
210static unsigned long virtio_mem_mb_id_to_phys(unsigned long mb_id)
211{
212	return mb_id * memory_block_size_bytes();
213}
214
215/*
216 * Calculate the subblock id of a given address.
217 */
218static unsigned long virtio_mem_phys_to_sb_id(struct virtio_mem *vm,
219					      unsigned long addr)
220{
221	const unsigned long mb_id = virtio_mem_phys_to_mb_id(addr);
222	const unsigned long mb_addr = virtio_mem_mb_id_to_phys(mb_id);
223
224	return (addr - mb_addr) / vm->subblock_size;
225}
226
227/*
228 * Set the state of a memory block, taking care of the state counter.
229 */
230static void virtio_mem_mb_set_state(struct virtio_mem *vm, unsigned long mb_id,
231				    enum virtio_mem_mb_state state)
232{
233	const unsigned long idx = mb_id - vm->first_mb_id;
234	enum virtio_mem_mb_state old_state;
235
236	old_state = vm->mb_state[idx];
237	vm->mb_state[idx] = state;
238
239	BUG_ON(vm->nb_mb_state[old_state] == 0);
240	vm->nb_mb_state[old_state]--;
241	vm->nb_mb_state[state]++;
242}
243
244/*
245 * Get the state of a memory block.
246 */
247static enum virtio_mem_mb_state virtio_mem_mb_get_state(struct virtio_mem *vm,
248							unsigned long mb_id)
249{
250	const unsigned long idx = mb_id - vm->first_mb_id;
251
252	return vm->mb_state[idx];
253}
254
255/*
256 * Prepare the state array for the next memory block.
257 */
258static int virtio_mem_mb_state_prepare_next_mb(struct virtio_mem *vm)
259{
260	unsigned long old_bytes = vm->next_mb_id - vm->first_mb_id + 1;
261	unsigned long new_bytes = vm->next_mb_id - vm->first_mb_id + 2;
262	int old_pages = PFN_UP(old_bytes);
263	int new_pages = PFN_UP(new_bytes);
264	uint8_t *new_mb_state;
265
266	if (vm->mb_state && old_pages == new_pages)
267		return 0;
268
269	new_mb_state = vzalloc(new_pages * PAGE_SIZE);
270	if (!new_mb_state)
271		return -ENOMEM;
272
273	mutex_lock(&vm->hotplug_mutex);
274	if (vm->mb_state)
275		memcpy(new_mb_state, vm->mb_state, old_pages * PAGE_SIZE);
276	vfree(vm->mb_state);
277	vm->mb_state = new_mb_state;
278	mutex_unlock(&vm->hotplug_mutex);
279
280	return 0;
281}
282
283#define virtio_mem_for_each_mb_state(_vm, _mb_id, _state) \
284	for (_mb_id = _vm->first_mb_id; \
285	     _mb_id < _vm->next_mb_id && _vm->nb_mb_state[_state]; \
286	     _mb_id++) \
287		if (virtio_mem_mb_get_state(_vm, _mb_id) == _state)
288
289#define virtio_mem_for_each_mb_state_rev(_vm, _mb_id, _state) \
290	for (_mb_id = _vm->next_mb_id - 1; \
291	     _mb_id >= _vm->first_mb_id && _vm->nb_mb_state[_state]; \
292	     _mb_id--) \
293		if (virtio_mem_mb_get_state(_vm, _mb_id) == _state)
294
295/*
296 * Mark all selected subblocks plugged.
297 *
298 * Will not modify the state of the memory block.
299 */
300static void virtio_mem_mb_set_sb_plugged(struct virtio_mem *vm,
301					 unsigned long mb_id, int sb_id,
302					 int count)
303{
304	const int bit = (mb_id - vm->first_mb_id) * vm->nb_sb_per_mb + sb_id;
305
306	__bitmap_set(vm->sb_bitmap, bit, count);
307}
308
309/*
310 * Mark all selected subblocks unplugged.
311 *
312 * Will not modify the state of the memory block.
313 */
314static void virtio_mem_mb_set_sb_unplugged(struct virtio_mem *vm,
315					   unsigned long mb_id, int sb_id,
316					   int count)
317{
318	const int bit = (mb_id - vm->first_mb_id) * vm->nb_sb_per_mb + sb_id;
319
320	__bitmap_clear(vm->sb_bitmap, bit, count);
321}
322
323/*
324 * Test if all selected subblocks are plugged.
325 */
326static bool virtio_mem_mb_test_sb_plugged(struct virtio_mem *vm,
327					  unsigned long mb_id, int sb_id,
328					  int count)
329{
330	const int bit = (mb_id - vm->first_mb_id) * vm->nb_sb_per_mb + sb_id;
331
332	if (count == 1)
333		return test_bit(bit, vm->sb_bitmap);
334
335	/* TODO: Helper similar to bitmap_set() */
336	return find_next_zero_bit(vm->sb_bitmap, bit + count, bit) >=
337	       bit + count;
338}
339
340/*
341 * Test if all selected subblocks are unplugged.
342 */
343static bool virtio_mem_mb_test_sb_unplugged(struct virtio_mem *vm,
344					    unsigned long mb_id, int sb_id,
345					    int count)
346{
347	const int bit = (mb_id - vm->first_mb_id) * vm->nb_sb_per_mb + sb_id;
348
349	/* TODO: Helper similar to bitmap_set() */
350	return find_next_bit(vm->sb_bitmap, bit + count, bit) >= bit + count;
351}
352
353/*
354 * Find the first unplugged subblock. Returns vm->nb_sb_per_mb in case there is
355 * none.
356 */
357static int virtio_mem_mb_first_unplugged_sb(struct virtio_mem *vm,
358					    unsigned long mb_id)
359{
360	const int bit = (mb_id - vm->first_mb_id) * vm->nb_sb_per_mb;
361
362	return find_next_zero_bit(vm->sb_bitmap, bit + vm->nb_sb_per_mb, bit) -
363	       bit;
364}
365
366/*
367 * Prepare the subblock bitmap for the next memory block.
368 */
369static int virtio_mem_sb_bitmap_prepare_next_mb(struct virtio_mem *vm)
370{
371	const unsigned long old_nb_mb = vm->next_mb_id - vm->first_mb_id;
372	const unsigned long old_nb_bits = old_nb_mb * vm->nb_sb_per_mb;
373	const unsigned long new_nb_bits = (old_nb_mb + 1) * vm->nb_sb_per_mb;
374	int old_pages = PFN_UP(BITS_TO_LONGS(old_nb_bits) * sizeof(long));
375	int new_pages = PFN_UP(BITS_TO_LONGS(new_nb_bits) * sizeof(long));
376	unsigned long *new_sb_bitmap, *old_sb_bitmap;
377
378	if (vm->sb_bitmap && old_pages == new_pages)
379		return 0;
380
381	new_sb_bitmap = vzalloc(new_pages * PAGE_SIZE);
382	if (!new_sb_bitmap)
383		return -ENOMEM;
384
385	mutex_lock(&vm->hotplug_mutex);
386	if (new_sb_bitmap)
387		memcpy(new_sb_bitmap, vm->sb_bitmap, old_pages * PAGE_SIZE);
388
389	old_sb_bitmap = vm->sb_bitmap;
390	vm->sb_bitmap = new_sb_bitmap;
391	mutex_unlock(&vm->hotplug_mutex);
392
393	vfree(old_sb_bitmap);
394	return 0;
395}
396
397/*
398 * Try to add a memory block to Linux. This will usually only fail
399 * if out of memory.
400 *
401 * Must not be called with the vm->hotplug_mutex held (possible deadlock with
402 * onlining code).
403 *
404 * Will not modify the state of the memory block.
405 */
406static int virtio_mem_mb_add(struct virtio_mem *vm, unsigned long mb_id)
407{
408	const uint64_t addr = virtio_mem_mb_id_to_phys(mb_id);
409	int nid = vm->nid;
410
411	if (nid == NUMA_NO_NODE)
412		nid = memory_add_physaddr_to_nid(addr);
413
414	/*
415	 * When force-unloading the driver and we still have memory added to
416	 * Linux, the resource name has to stay.
417	 */
418	if (!vm->resource_name) {
419		vm->resource_name = kstrdup_const("System RAM (virtio_mem)",
420						  GFP_KERNEL);
421		if (!vm->resource_name)
422			return -ENOMEM;
423	}
424
425	dev_dbg(&vm->vdev->dev, "adding memory block: %lu\n", mb_id);
426	return add_memory_driver_managed(nid, addr, memory_block_size_bytes(),
427					 vm->resource_name,
428					 MEMHP_MERGE_RESOURCE);
429}
430
431/*
432 * Try to remove a memory block from Linux. Will only fail if the memory block
433 * is not offline.
434 *
435 * Must not be called with the vm->hotplug_mutex held (possible deadlock with
436 * onlining code).
437 *
438 * Will not modify the state of the memory block.
439 */
440static int virtio_mem_mb_remove(struct virtio_mem *vm, unsigned long mb_id)
441{
442	const uint64_t addr = virtio_mem_mb_id_to_phys(mb_id);
443	int nid = vm->nid;
444
445	if (nid == NUMA_NO_NODE)
446		nid = memory_add_physaddr_to_nid(addr);
447
448	dev_dbg(&vm->vdev->dev, "removing memory block: %lu\n", mb_id);
449	return remove_memory(nid, addr, memory_block_size_bytes());
450}
451
452/*
453 * Try to offline and remove a memory block from Linux.
454 *
455 * Must not be called with the vm->hotplug_mutex held (possible deadlock with
456 * onlining code).
457 *
458 * Will not modify the state of the memory block.
459 */
460static int virtio_mem_mb_offline_and_remove(struct virtio_mem *vm,
461					    unsigned long mb_id)
462{
463	const uint64_t addr = virtio_mem_mb_id_to_phys(mb_id);
464	int nid = vm->nid;
465
466	if (nid == NUMA_NO_NODE)
467		nid = memory_add_physaddr_to_nid(addr);
468
469	dev_dbg(&vm->vdev->dev, "offlining and removing memory block: %lu\n",
470		mb_id);
471	return offline_and_remove_memory(nid, addr, memory_block_size_bytes());
472}
473
474/*
475 * Trigger the workqueue so the device can perform its magic.
476 */
477static void virtio_mem_retry(struct virtio_mem *vm)
478{
479	unsigned long flags;
480
481	spin_lock_irqsave(&vm->removal_lock, flags);
482	if (!vm->removing)
483		queue_work(system_freezable_wq, &vm->wq);
484	spin_unlock_irqrestore(&vm->removal_lock, flags);
485}
486
487static int virtio_mem_translate_node_id(struct virtio_mem *vm, uint16_t node_id)
488{
489	int node = NUMA_NO_NODE;
490
491#if defined(CONFIG_ACPI_NUMA)
492	if (virtio_has_feature(vm->vdev, VIRTIO_MEM_F_ACPI_PXM))
493		node = pxm_to_node(node_id);
494#endif
495	return node;
496}
497
498/*
499 * Test if a virtio-mem device overlaps with the given range. Can be called
500 * from (notifier) callbacks lockless.
501 */
502static bool virtio_mem_overlaps_range(struct virtio_mem *vm,
503				      unsigned long start, unsigned long size)
504{
505	unsigned long dev_start = virtio_mem_mb_id_to_phys(vm->first_mb_id);
506	unsigned long dev_end = virtio_mem_mb_id_to_phys(vm->last_mb_id) +
507				memory_block_size_bytes();
508
509	return start < dev_end && dev_start < start + size;
510}
511
512/*
513 * Test if a virtio-mem device owns a memory block. Can be called from
514 * (notifier) callbacks lockless.
515 */
516static bool virtio_mem_owned_mb(struct virtio_mem *vm, unsigned long mb_id)
517{
518	return mb_id >= vm->first_mb_id && mb_id <= vm->last_mb_id;
519}
520
521static int virtio_mem_notify_going_online(struct virtio_mem *vm,
522					  unsigned long mb_id)
523{
524	switch (virtio_mem_mb_get_state(vm, mb_id)) {
525	case VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL:
526	case VIRTIO_MEM_MB_STATE_OFFLINE:
527		return NOTIFY_OK;
528	default:
529		break;
530	}
531	dev_warn_ratelimited(&vm->vdev->dev,
532			     "memory block onlining denied\n");
533	return NOTIFY_BAD;
534}
535
536static void virtio_mem_notify_offline(struct virtio_mem *vm,
537				      unsigned long mb_id)
538{
539	switch (virtio_mem_mb_get_state(vm, mb_id)) {
540	case VIRTIO_MEM_MB_STATE_ONLINE_PARTIAL:
541		virtio_mem_mb_set_state(vm, mb_id,
542					VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL);
543		break;
544	case VIRTIO_MEM_MB_STATE_ONLINE:
545		virtio_mem_mb_set_state(vm, mb_id,
546					VIRTIO_MEM_MB_STATE_OFFLINE);
547		break;
548	default:
549		BUG();
550		break;
551	}
552
553	/*
554	 * Trigger the workqueue, maybe we can now unplug memory. Also,
555	 * when we offline and remove a memory block, this will re-trigger
556	 * us immediately - which is often nice because the removal of
557	 * the memory block (e.g., memmap) might have freed up memory
558	 * on other memory blocks we manage.
559	 */
560	virtio_mem_retry(vm);
561}
562
563static void virtio_mem_notify_online(struct virtio_mem *vm, unsigned long mb_id)
564{
565	unsigned long nb_offline;
566
567	switch (virtio_mem_mb_get_state(vm, mb_id)) {
568	case VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL:
569		virtio_mem_mb_set_state(vm, mb_id,
570					VIRTIO_MEM_MB_STATE_ONLINE_PARTIAL);
571		break;
572	case VIRTIO_MEM_MB_STATE_OFFLINE:
573		virtio_mem_mb_set_state(vm, mb_id, VIRTIO_MEM_MB_STATE_ONLINE);
574		break;
575	default:
576		BUG();
577		break;
578	}
579	nb_offline = vm->nb_mb_state[VIRTIO_MEM_MB_STATE_OFFLINE] +
580		     vm->nb_mb_state[VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL];
581
582	/* see if we can add new blocks now that we onlined one block */
583	if (nb_offline == VIRTIO_MEM_NB_OFFLINE_THRESHOLD - 1)
584		virtio_mem_retry(vm);
585}
586
587static void virtio_mem_notify_going_offline(struct virtio_mem *vm,
588					    unsigned long mb_id)
589{
590	const unsigned long nr_pages = PFN_DOWN(vm->subblock_size);
591	struct page *page;
592	unsigned long pfn;
593	int sb_id, i;
594
595	for (sb_id = 0; sb_id < vm->nb_sb_per_mb; sb_id++) {
596		if (virtio_mem_mb_test_sb_plugged(vm, mb_id, sb_id, 1))
597			continue;
598		/*
599		 * Drop our reference to the pages so the memory can get
600		 * offlined and add the unplugged pages to the managed
601		 * page counters (so offlining code can correctly subtract
602		 * them again).
603		 */
604		pfn = PFN_DOWN(virtio_mem_mb_id_to_phys(mb_id) +
605			       sb_id * vm->subblock_size);
606		adjust_managed_page_count(pfn_to_page(pfn), nr_pages);
607		for (i = 0; i < nr_pages; i++) {
608			page = pfn_to_page(pfn + i);
609			if (WARN_ON(!page_ref_dec_and_test(page)))
610				dump_page(page, "unplugged page referenced");
611		}
612	}
613}
614
615static void virtio_mem_notify_cancel_offline(struct virtio_mem *vm,
616					     unsigned long mb_id)
617{
618	const unsigned long nr_pages = PFN_DOWN(vm->subblock_size);
619	unsigned long pfn;
620	int sb_id, i;
621
622	for (sb_id = 0; sb_id < vm->nb_sb_per_mb; sb_id++) {
623		if (virtio_mem_mb_test_sb_plugged(vm, mb_id, sb_id, 1))
624			continue;
625		/*
626		 * Get the reference we dropped when going offline and
627		 * subtract the unplugged pages from the managed page
628		 * counters.
629		 */
630		pfn = PFN_DOWN(virtio_mem_mb_id_to_phys(mb_id) +
631			       sb_id * vm->subblock_size);
632		adjust_managed_page_count(pfn_to_page(pfn), -nr_pages);
633		for (i = 0; i < nr_pages; i++)
634			page_ref_inc(pfn_to_page(pfn + i));
635	}
636}
637
638/*
639 * This callback will either be called synchronously from add_memory() or
640 * asynchronously (e.g., triggered via user space). We have to be careful
641 * with locking when calling add_memory().
642 */
643static int virtio_mem_memory_notifier_cb(struct notifier_block *nb,
644					 unsigned long action, void *arg)
645{
646	struct virtio_mem *vm = container_of(nb, struct virtio_mem,
647					     memory_notifier);
648	struct memory_notify *mhp = arg;
649	const unsigned long start = PFN_PHYS(mhp->start_pfn);
650	const unsigned long size = PFN_PHYS(mhp->nr_pages);
651	const unsigned long mb_id = virtio_mem_phys_to_mb_id(start);
652	int rc = NOTIFY_OK;
653
654	if (!virtio_mem_overlaps_range(vm, start, size))
655		return NOTIFY_DONE;
656
657	/*
658	 * Memory is onlined/offlined in memory block granularity. We cannot
659	 * cross virtio-mem device boundaries and memory block boundaries. Bail
660	 * out if this ever changes.
661	 */
662	if (WARN_ON_ONCE(size != memory_block_size_bytes() ||
663			 !IS_ALIGNED(start, memory_block_size_bytes())))
664		return NOTIFY_BAD;
665
666	/*
667	 * Avoid circular locking lockdep warnings. We lock the mutex
668	 * e.g., in MEM_GOING_ONLINE and unlock it in MEM_ONLINE. The
669	 * blocking_notifier_call_chain() has it's own lock, which gets unlocked
670	 * between both notifier calls and will bail out. False positive.
671	 */
672	lockdep_off();
673
674	switch (action) {
675	case MEM_GOING_OFFLINE:
676		mutex_lock(&vm->hotplug_mutex);
677		if (vm->removing) {
678			rc = notifier_from_errno(-EBUSY);
679			mutex_unlock(&vm->hotplug_mutex);
680			break;
681		}
682		vm->hotplug_active = true;
683		virtio_mem_notify_going_offline(vm, mb_id);
684		break;
685	case MEM_GOING_ONLINE:
686		mutex_lock(&vm->hotplug_mutex);
687		if (vm->removing) {
688			rc = notifier_from_errno(-EBUSY);
689			mutex_unlock(&vm->hotplug_mutex);
690			break;
691		}
692		vm->hotplug_active = true;
693		rc = virtio_mem_notify_going_online(vm, mb_id);
694		break;
695	case MEM_OFFLINE:
696		virtio_mem_notify_offline(vm, mb_id);
697		vm->hotplug_active = false;
698		mutex_unlock(&vm->hotplug_mutex);
699		break;
700	case MEM_ONLINE:
701		virtio_mem_notify_online(vm, mb_id);
702		vm->hotplug_active = false;
703		mutex_unlock(&vm->hotplug_mutex);
704		break;
705	case MEM_CANCEL_OFFLINE:
706		if (!vm->hotplug_active)
707			break;
708		virtio_mem_notify_cancel_offline(vm, mb_id);
709		vm->hotplug_active = false;
710		mutex_unlock(&vm->hotplug_mutex);
711		break;
712	case MEM_CANCEL_ONLINE:
713		if (!vm->hotplug_active)
714			break;
715		vm->hotplug_active = false;
716		mutex_unlock(&vm->hotplug_mutex);
717		break;
718	default:
719		break;
720	}
721
722	lockdep_on();
723
724	return rc;
725}
726
727/*
728 * Set a range of pages PG_offline. Remember pages that were never onlined
729 * (via generic_online_page()) using PageDirty().
730 */
731static void virtio_mem_set_fake_offline(unsigned long pfn,
732					unsigned int nr_pages, bool onlined)
733{
734	for (; nr_pages--; pfn++) {
735		struct page *page = pfn_to_page(pfn);
736
737		__SetPageOffline(page);
738		if (!onlined) {
739			SetPageDirty(page);
740			/* FIXME: remove after cleanups */
741			ClearPageReserved(page);
742		}
743	}
744}
745
746/*
747 * Clear PG_offline from a range of pages. If the pages were never onlined,
748 * (via generic_online_page()), clear PageDirty().
749 */
750static void virtio_mem_clear_fake_offline(unsigned long pfn,
751					  unsigned int nr_pages, bool onlined)
752{
753	for (; nr_pages--; pfn++) {
754		struct page *page = pfn_to_page(pfn);
755
756		__ClearPageOffline(page);
757		if (!onlined)
758			ClearPageDirty(page);
759	}
760}
761
762/*
763 * Release a range of fake-offline pages to the buddy, effectively
764 * fake-onlining them.
765 */
766static void virtio_mem_fake_online(unsigned long pfn, unsigned int nr_pages)
767{
768	const int order = MAX_ORDER - 1;
769	int i;
770
771	/*
772	 * We are always called with subblock granularity, which is at least
773	 * aligned to MAX_ORDER - 1.
774	 */
775	for (i = 0; i < nr_pages; i += 1 << order) {
776		struct page *page = pfn_to_page(pfn + i);
777
778		/*
779		 * If the page is PageDirty(), it was kept fake-offline when
780		 * onlining the memory block. Otherwise, it was allocated
781		 * using alloc_contig_range(). All pages in a subblock are
782		 * alike.
783		 */
784		if (PageDirty(page)) {
785			virtio_mem_clear_fake_offline(pfn + i, 1 << order,
786						      false);
787			generic_online_page(page, order);
788		} else {
789			virtio_mem_clear_fake_offline(pfn + i, 1 << order,
790						      true);
791			free_contig_range(pfn + i, 1 << order);
792			adjust_managed_page_count(page, 1 << order);
793		}
794	}
795}
796
797static void virtio_mem_online_page_cb(struct page *page, unsigned int order)
798{
799	const unsigned long addr = page_to_phys(page);
800	const unsigned long mb_id = virtio_mem_phys_to_mb_id(addr);
801	struct virtio_mem *vm;
802	int sb_id;
803
804	/*
805	 * We exploit here that subblocks have at least MAX_ORDER - 1
806	 * size/alignment and that this callback is is called with such a
807	 * size/alignment. So we cannot cross subblocks and therefore
808	 * also not memory blocks.
809	 */
810	rcu_read_lock();
811	list_for_each_entry_rcu(vm, &virtio_mem_devices, next) {
812		if (!virtio_mem_owned_mb(vm, mb_id))
813			continue;
814
815		sb_id = virtio_mem_phys_to_sb_id(vm, addr);
816		/*
817		 * If plugged, online the pages, otherwise, set them fake
818		 * offline (PageOffline).
819		 */
820		if (virtio_mem_mb_test_sb_plugged(vm, mb_id, sb_id, 1))
821			generic_online_page(page, order);
822		else
823			virtio_mem_set_fake_offline(PFN_DOWN(addr), 1 << order,
824						    false);
825		rcu_read_unlock();
826		return;
827	}
828	rcu_read_unlock();
829
830	/* not virtio-mem memory, but e.g., a DIMM. online it */
831	generic_online_page(page, order);
832}
833
834static uint64_t virtio_mem_send_request(struct virtio_mem *vm,
835					const struct virtio_mem_req *req)
836{
837	struct scatterlist *sgs[2], sg_req, sg_resp;
838	unsigned int len;
839	int rc;
840
841	/* don't use the request residing on the stack (vaddr) */
842	vm->req = *req;
843
844	/* out: buffer for request */
845	sg_init_one(&sg_req, &vm->req, sizeof(vm->req));
846	sgs[0] = &sg_req;
847
848	/* in: buffer for response */
849	sg_init_one(&sg_resp, &vm->resp, sizeof(vm->resp));
850	sgs[1] = &sg_resp;
851
852	rc = virtqueue_add_sgs(vm->vq, sgs, 1, 1, vm, GFP_KERNEL);
853	if (rc < 0)
854		return rc;
855
856	virtqueue_kick(vm->vq);
857
858	/* wait for a response */
859	wait_event(vm->host_resp, virtqueue_get_buf(vm->vq, &len));
860
861	return virtio16_to_cpu(vm->vdev, vm->resp.type);
862}
863
864static int virtio_mem_send_plug_request(struct virtio_mem *vm, uint64_t addr,
865					uint64_t size)
866{
867	const uint64_t nb_vm_blocks = size / vm->device_block_size;
868	const struct virtio_mem_req req = {
869		.type = cpu_to_virtio16(vm->vdev, VIRTIO_MEM_REQ_PLUG),
870		.u.plug.addr = cpu_to_virtio64(vm->vdev, addr),
871		.u.plug.nb_blocks = cpu_to_virtio16(vm->vdev, nb_vm_blocks),
872	};
873
874	if (atomic_read(&vm->config_changed))
875		return -EAGAIN;
876
877	switch (virtio_mem_send_request(vm, &req)) {
878	case VIRTIO_MEM_RESP_ACK:
879		vm->plugged_size += size;
880		return 0;
881	case VIRTIO_MEM_RESP_NACK:
882		return -EAGAIN;
883	case VIRTIO_MEM_RESP_BUSY:
884		return -ETXTBSY;
885	case VIRTIO_MEM_RESP_ERROR:
886		return -EINVAL;
887	default:
888		return -ENOMEM;
889	}
890}
891
892static int virtio_mem_send_unplug_request(struct virtio_mem *vm, uint64_t addr,
893					  uint64_t size)
894{
895	const uint64_t nb_vm_blocks = size / vm->device_block_size;
896	const struct virtio_mem_req req = {
897		.type = cpu_to_virtio16(vm->vdev, VIRTIO_MEM_REQ_UNPLUG),
898		.u.unplug.addr = cpu_to_virtio64(vm->vdev, addr),
899		.u.unplug.nb_blocks = cpu_to_virtio16(vm->vdev, nb_vm_blocks),
900	};
901
902	if (atomic_read(&vm->config_changed))
903		return -EAGAIN;
904
905	switch (virtio_mem_send_request(vm, &req)) {
906	case VIRTIO_MEM_RESP_ACK:
907		vm->plugged_size -= size;
908		return 0;
909	case VIRTIO_MEM_RESP_BUSY:
910		return -ETXTBSY;
911	case VIRTIO_MEM_RESP_ERROR:
912		return -EINVAL;
913	default:
914		return -ENOMEM;
915	}
916}
917
918static int virtio_mem_send_unplug_all_request(struct virtio_mem *vm)
919{
920	const struct virtio_mem_req req = {
921		.type = cpu_to_virtio16(vm->vdev, VIRTIO_MEM_REQ_UNPLUG_ALL),
922	};
923
924	switch (virtio_mem_send_request(vm, &req)) {
925	case VIRTIO_MEM_RESP_ACK:
926		vm->unplug_all_required = false;
927		vm->plugged_size = 0;
928		/* usable region might have shrunk */
929		atomic_set(&vm->config_changed, 1);
930		return 0;
931	case VIRTIO_MEM_RESP_BUSY:
932		return -ETXTBSY;
933	default:
934		return -ENOMEM;
935	}
936}
937
938/*
939 * Plug selected subblocks. Updates the plugged state, but not the state
940 * of the memory block.
941 */
942static int virtio_mem_mb_plug_sb(struct virtio_mem *vm, unsigned long mb_id,
943				 int sb_id, int count)
944{
945	const uint64_t addr = virtio_mem_mb_id_to_phys(mb_id) +
946			      sb_id * vm->subblock_size;
947	const uint64_t size = count * vm->subblock_size;
948	int rc;
949
950	dev_dbg(&vm->vdev->dev, "plugging memory block: %lu : %i - %i\n", mb_id,
951		sb_id, sb_id + count - 1);
952
953	rc = virtio_mem_send_plug_request(vm, addr, size);
954	if (!rc)
955		virtio_mem_mb_set_sb_plugged(vm, mb_id, sb_id, count);
956	return rc;
957}
958
959/*
960 * Unplug selected subblocks. Updates the plugged state, but not the state
961 * of the memory block.
962 */
963static int virtio_mem_mb_unplug_sb(struct virtio_mem *vm, unsigned long mb_id,
964				   int sb_id, int count)
965{
966	const uint64_t addr = virtio_mem_mb_id_to_phys(mb_id) +
967			      sb_id * vm->subblock_size;
968	const uint64_t size = count * vm->subblock_size;
969	int rc;
970
971	dev_dbg(&vm->vdev->dev, "unplugging memory block: %lu : %i - %i\n",
972		mb_id, sb_id, sb_id + count - 1);
973
974	rc = virtio_mem_send_unplug_request(vm, addr, size);
975	if (!rc)
976		virtio_mem_mb_set_sb_unplugged(vm, mb_id, sb_id, count);
977	return rc;
978}
979
980/*
981 * Unplug the desired number of plugged subblocks of a offline or not-added
982 * memory block. Will fail if any subblock cannot get unplugged (instead of
983 * skipping it).
984 *
985 * Will not modify the state of the memory block.
986 *
987 * Note: can fail after some subblocks were unplugged.
988 */
989static int virtio_mem_mb_unplug_any_sb(struct virtio_mem *vm,
990				       unsigned long mb_id, uint64_t *nb_sb)
991{
992	int sb_id, count;
993	int rc;
994
995	sb_id = vm->nb_sb_per_mb - 1;
996	while (*nb_sb) {
997		/* Find the next candidate subblock */
998		while (sb_id >= 0 &&
999		       virtio_mem_mb_test_sb_unplugged(vm, mb_id, sb_id, 1))
1000			sb_id--;
1001		if (sb_id < 0)
1002			break;
1003		/* Try to unplug multiple subblocks at a time */
1004		count = 1;
1005		while (count < *nb_sb && sb_id > 0 &&
1006		       virtio_mem_mb_test_sb_plugged(vm, mb_id, sb_id - 1, 1)) {
1007			count++;
1008			sb_id--;
1009		}
1010
1011		rc = virtio_mem_mb_unplug_sb(vm, mb_id, sb_id, count);
1012		if (rc)
1013			return rc;
1014		*nb_sb -= count;
1015		sb_id--;
1016	}
1017
1018	return 0;
1019}
1020
1021/*
1022 * Unplug all plugged subblocks of an offline or not-added memory block.
1023 *
1024 * Will not modify the state of the memory block.
1025 *
1026 * Note: can fail after some subblocks were unplugged.
1027 */
1028static int virtio_mem_mb_unplug(struct virtio_mem *vm, unsigned long mb_id)
1029{
1030	uint64_t nb_sb = vm->nb_sb_per_mb;
1031
1032	return virtio_mem_mb_unplug_any_sb(vm, mb_id, &nb_sb);
1033}
1034
1035/*
1036 * Prepare tracking data for the next memory block.
1037 */
1038static int virtio_mem_prepare_next_mb(struct virtio_mem *vm,
1039				      unsigned long *mb_id)
1040{
1041	int rc;
1042
1043	if (vm->next_mb_id > vm->last_usable_mb_id)
1044		return -ENOSPC;
1045
1046	/* Resize the state array if required. */
1047	rc = virtio_mem_mb_state_prepare_next_mb(vm);
1048	if (rc)
1049		return rc;
1050
1051	/* Resize the subblock bitmap if required. */
1052	rc = virtio_mem_sb_bitmap_prepare_next_mb(vm);
1053	if (rc)
1054		return rc;
1055
1056	vm->nb_mb_state[VIRTIO_MEM_MB_STATE_UNUSED]++;
1057	*mb_id = vm->next_mb_id++;
1058	return 0;
1059}
1060
1061/*
1062 * Don't add too many blocks that are not onlined yet to avoid running OOM.
1063 */
1064static bool virtio_mem_too_many_mb_offline(struct virtio_mem *vm)
1065{
1066	unsigned long nb_offline;
1067
1068	nb_offline = vm->nb_mb_state[VIRTIO_MEM_MB_STATE_OFFLINE] +
1069		     vm->nb_mb_state[VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL];
1070	return nb_offline >= VIRTIO_MEM_NB_OFFLINE_THRESHOLD;
1071}
1072
1073/*
1074 * Try to plug the desired number of subblocks and add the memory block
1075 * to Linux.
1076 *
1077 * Will modify the state of the memory block.
1078 */
1079static int virtio_mem_mb_plug_and_add(struct virtio_mem *vm,
1080				      unsigned long mb_id,
1081				      uint64_t *nb_sb)
1082{
1083	const int count = min_t(int, *nb_sb, vm->nb_sb_per_mb);
1084	int rc, rc2;
1085
1086	if (WARN_ON_ONCE(!count))
1087		return -EINVAL;
1088
1089	/*
1090	 * Plug the requested number of subblocks before adding it to linux,
1091	 * so that onlining will directly online all plugged subblocks.
1092	 */
1093	rc = virtio_mem_mb_plug_sb(vm, mb_id, 0, count);
1094	if (rc)
1095		return rc;
1096
1097	/*
1098	 * Mark the block properly offline before adding it to Linux,
1099	 * so the memory notifiers will find the block in the right state.
1100	 */
1101	if (count == vm->nb_sb_per_mb)
1102		virtio_mem_mb_set_state(vm, mb_id,
1103					VIRTIO_MEM_MB_STATE_OFFLINE);
1104	else
1105		virtio_mem_mb_set_state(vm, mb_id,
1106					VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL);
1107
1108	/* Add the memory block to linux - if that fails, try to unplug. */
1109	rc = virtio_mem_mb_add(vm, mb_id);
1110	if (rc) {
1111		enum virtio_mem_mb_state new_state = VIRTIO_MEM_MB_STATE_UNUSED;
1112
1113		dev_err(&vm->vdev->dev,
1114			"adding memory block %lu failed with %d\n", mb_id, rc);
1115		rc2 = virtio_mem_mb_unplug_sb(vm, mb_id, 0, count);
1116
1117		/*
1118		 * TODO: Linux MM does not properly clean up yet in all cases
1119		 * where adding of memory failed - especially on -ENOMEM.
1120		 */
1121		if (rc2)
1122			new_state = VIRTIO_MEM_MB_STATE_PLUGGED;
1123		virtio_mem_mb_set_state(vm, mb_id, new_state);
1124		return rc;
1125	}
1126
1127	*nb_sb -= count;
1128	return 0;
1129}
1130
1131/*
1132 * Try to plug the desired number of subblocks of a memory block that
1133 * is already added to Linux.
1134 *
1135 * Will modify the state of the memory block.
1136 *
1137 * Note: Can fail after some subblocks were successfully plugged.
1138 */
1139static int virtio_mem_mb_plug_any_sb(struct virtio_mem *vm, unsigned long mb_id,
1140				     uint64_t *nb_sb, bool online)
1141{
1142	unsigned long pfn, nr_pages;
1143	int sb_id, count;
1144	int rc;
1145
1146	if (WARN_ON_ONCE(!*nb_sb))
1147		return -EINVAL;
1148
1149	while (*nb_sb) {
1150		sb_id = virtio_mem_mb_first_unplugged_sb(vm, mb_id);
1151		if (sb_id >= vm->nb_sb_per_mb)
1152			break;
1153		count = 1;
1154		while (count < *nb_sb &&
1155		       sb_id + count < vm->nb_sb_per_mb &&
1156		       !virtio_mem_mb_test_sb_plugged(vm, mb_id, sb_id + count,
1157						      1))
1158			count++;
1159
1160		rc = virtio_mem_mb_plug_sb(vm, mb_id, sb_id, count);
1161		if (rc)
1162			return rc;
1163		*nb_sb -= count;
1164		if (!online)
1165			continue;
1166
1167		/* fake-online the pages if the memory block is online */
1168		pfn = PFN_DOWN(virtio_mem_mb_id_to_phys(mb_id) +
1169			       sb_id * vm->subblock_size);
1170		nr_pages = PFN_DOWN(count * vm->subblock_size);
1171		virtio_mem_fake_online(pfn, nr_pages);
1172	}
1173
1174	if (virtio_mem_mb_test_sb_plugged(vm, mb_id, 0, vm->nb_sb_per_mb)) {
1175		if (online)
1176			virtio_mem_mb_set_state(vm, mb_id,
1177						VIRTIO_MEM_MB_STATE_ONLINE);
1178		else
1179			virtio_mem_mb_set_state(vm, mb_id,
1180						VIRTIO_MEM_MB_STATE_OFFLINE);
1181	}
1182
1183	return 0;
1184}
1185
1186/*
1187 * Try to plug the requested amount of memory.
1188 */
1189static int virtio_mem_plug_request(struct virtio_mem *vm, uint64_t diff)
1190{
1191	uint64_t nb_sb = diff / vm->subblock_size;
1192	unsigned long mb_id;
1193	int rc;
1194
1195	if (!nb_sb)
1196		return 0;
1197
1198	/* Don't race with onlining/offlining */
1199	mutex_lock(&vm->hotplug_mutex);
1200
1201	/* Try to plug subblocks of partially plugged online blocks. */
1202	virtio_mem_for_each_mb_state(vm, mb_id,
1203				     VIRTIO_MEM_MB_STATE_ONLINE_PARTIAL) {
1204		rc = virtio_mem_mb_plug_any_sb(vm, mb_id, &nb_sb, true);
1205		if (rc || !nb_sb)
1206			goto out_unlock;
1207		cond_resched();
1208	}
1209
1210	/* Try to plug subblocks of partially plugged offline blocks. */
1211	virtio_mem_for_each_mb_state(vm, mb_id,
1212				     VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL) {
1213		rc = virtio_mem_mb_plug_any_sb(vm, mb_id, &nb_sb, false);
1214		if (rc || !nb_sb)
1215			goto out_unlock;
1216		cond_resched();
1217	}
1218
1219	/*
1220	 * We won't be working on online/offline memory blocks from this point,
1221	 * so we can't race with memory onlining/offlining. Drop the mutex.
1222	 */
1223	mutex_unlock(&vm->hotplug_mutex);
1224
1225	/* Try to plug and add unused blocks */
1226	virtio_mem_for_each_mb_state(vm, mb_id, VIRTIO_MEM_MB_STATE_UNUSED) {
1227		if (virtio_mem_too_many_mb_offline(vm))
1228			return -ENOSPC;
1229
1230		rc = virtio_mem_mb_plug_and_add(vm, mb_id, &nb_sb);
1231		if (rc || !nb_sb)
1232			return rc;
1233		cond_resched();
1234	}
1235
1236	/* Try to prepare, plug and add new blocks */
1237	while (nb_sb) {
1238		if (virtio_mem_too_many_mb_offline(vm))
1239			return -ENOSPC;
1240
1241		rc = virtio_mem_prepare_next_mb(vm, &mb_id);
1242		if (rc)
1243			return rc;
1244		rc = virtio_mem_mb_plug_and_add(vm, mb_id, &nb_sb);
1245		if (rc)
1246			return rc;
1247		cond_resched();
1248	}
1249
1250	return 0;
1251out_unlock:
1252	mutex_unlock(&vm->hotplug_mutex);
1253	return rc;
1254}
1255
1256/*
1257 * Unplug the desired number of plugged subblocks of an offline memory block.
1258 * Will fail if any subblock cannot get unplugged (instead of skipping it).
1259 *
1260 * Will modify the state of the memory block. Might temporarily drop the
1261 * hotplug_mutex.
1262 *
1263 * Note: Can fail after some subblocks were successfully unplugged.
1264 */
1265static int virtio_mem_mb_unplug_any_sb_offline(struct virtio_mem *vm,
1266					       unsigned long mb_id,
1267					       uint64_t *nb_sb)
1268{
1269	int rc;
1270
1271	rc = virtio_mem_mb_unplug_any_sb(vm, mb_id, nb_sb);
1272
1273	/* some subblocks might have been unplugged even on failure */
1274	if (!virtio_mem_mb_test_sb_plugged(vm, mb_id, 0, vm->nb_sb_per_mb))
1275		virtio_mem_mb_set_state(vm, mb_id,
1276					VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL);
1277	if (rc)
1278		return rc;
1279
1280	if (virtio_mem_mb_test_sb_unplugged(vm, mb_id, 0, vm->nb_sb_per_mb)) {
1281		/*
1282		 * Remove the block from Linux - this should never fail.
1283		 * Hinder the block from getting onlined by marking it
1284		 * unplugged. Temporarily drop the mutex, so
1285		 * any pending GOING_ONLINE requests can be serviced/rejected.
1286		 */
1287		virtio_mem_mb_set_state(vm, mb_id,
1288					VIRTIO_MEM_MB_STATE_UNUSED);
1289
1290		mutex_unlock(&vm->hotplug_mutex);
1291		rc = virtio_mem_mb_remove(vm, mb_id);
1292		BUG_ON(rc);
1293		mutex_lock(&vm->hotplug_mutex);
1294	}
1295	return 0;
1296}
1297
1298/*
1299 * Unplug the given plugged subblocks of an online memory block.
1300 *
1301 * Will modify the state of the memory block.
1302 */
1303static int virtio_mem_mb_unplug_sb_online(struct virtio_mem *vm,
1304					  unsigned long mb_id, int sb_id,
1305					  int count)
1306{
1307	const unsigned long nr_pages = PFN_DOWN(vm->subblock_size) * count;
1308	unsigned long start_pfn;
1309	int rc;
1310
1311	start_pfn = PFN_DOWN(virtio_mem_mb_id_to_phys(mb_id) +
1312			     sb_id * vm->subblock_size);
1313	rc = alloc_contig_range(start_pfn, start_pfn + nr_pages,
1314				MIGRATE_MOVABLE, GFP_KERNEL);
1315	if (rc == -ENOMEM)
1316		/* whoops, out of memory */
1317		return rc;
1318	if (rc)
1319		return -EBUSY;
1320
1321	/* Mark it as fake-offline before unplugging it */
1322	virtio_mem_set_fake_offline(start_pfn, nr_pages, true);
1323	adjust_managed_page_count(pfn_to_page(start_pfn), -nr_pages);
1324
1325	/* Try to unplug the allocated memory */
1326	rc = virtio_mem_mb_unplug_sb(vm, mb_id, sb_id, count);
1327	if (rc) {
1328		/* Return the memory to the buddy. */
1329		virtio_mem_fake_online(start_pfn, nr_pages);
1330		return rc;
1331	}
1332
1333	virtio_mem_mb_set_state(vm, mb_id,
1334				VIRTIO_MEM_MB_STATE_ONLINE_PARTIAL);
1335	return 0;
1336}
1337
1338/*
1339 * Unplug the desired number of plugged subblocks of an online memory block.
1340 * Will skip subblock that are busy.
1341 *
1342 * Will modify the state of the memory block. Might temporarily drop the
1343 * hotplug_mutex.
1344 *
1345 * Note: Can fail after some subblocks were successfully unplugged. Can
1346 *       return 0 even if subblocks were busy and could not get unplugged.
1347 */
1348static int virtio_mem_mb_unplug_any_sb_online(struct virtio_mem *vm,
1349					      unsigned long mb_id,
1350					      uint64_t *nb_sb)
1351{
1352	int rc, sb_id;
1353
1354	/* If possible, try to unplug the complete block in one shot. */
1355	if (*nb_sb >= vm->nb_sb_per_mb &&
1356	    virtio_mem_mb_test_sb_plugged(vm, mb_id, 0, vm->nb_sb_per_mb)) {
1357		rc = virtio_mem_mb_unplug_sb_online(vm, mb_id, 0,
1358						    vm->nb_sb_per_mb);
1359		if (!rc) {
1360			*nb_sb -= vm->nb_sb_per_mb;
1361			goto unplugged;
1362		} else if (rc != -EBUSY)
1363			return rc;
1364	}
1365
1366	/* Fallback to single subblocks. */
1367	for (sb_id = vm->nb_sb_per_mb - 1; sb_id >= 0 && *nb_sb; sb_id--) {
1368		/* Find the next candidate subblock */
1369		while (sb_id >= 0 &&
1370		       !virtio_mem_mb_test_sb_plugged(vm, mb_id, sb_id, 1))
1371			sb_id--;
1372		if (sb_id < 0)
1373			break;
1374
1375		rc = virtio_mem_mb_unplug_sb_online(vm, mb_id, sb_id, 1);
1376		if (rc == -EBUSY)
1377			continue;
1378		else if (rc)
1379			return rc;
1380		*nb_sb -= 1;
1381	}
1382
1383unplugged:
1384	/*
1385	 * Once all subblocks of a memory block were unplugged, offline and
1386	 * remove it. This will usually not fail, as no memory is in use
1387	 * anymore - however some other notifiers might NACK the request.
1388	 */
1389	if (virtio_mem_mb_test_sb_unplugged(vm, mb_id, 0, vm->nb_sb_per_mb)) {
1390		mutex_unlock(&vm->hotplug_mutex);
1391		rc = virtio_mem_mb_offline_and_remove(vm, mb_id);
1392		mutex_lock(&vm->hotplug_mutex);
1393		if (!rc)
1394			virtio_mem_mb_set_state(vm, mb_id,
1395						VIRTIO_MEM_MB_STATE_UNUSED);
1396	}
1397
1398	return 0;
1399}
1400
1401/*
1402 * Try to unplug the requested amount of memory.
1403 */
1404static int virtio_mem_unplug_request(struct virtio_mem *vm, uint64_t diff)
1405{
1406	uint64_t nb_sb = diff / vm->subblock_size;
1407	unsigned long mb_id;
1408	int rc;
1409
1410	if (!nb_sb)
1411		return 0;
1412
1413	/*
1414	 * We'll drop the mutex a couple of times when it is safe to do so.
1415	 * This might result in some blocks switching the state (online/offline)
1416	 * and we could miss them in this run - we will retry again later.
1417	 */
1418	mutex_lock(&vm->hotplug_mutex);
1419
1420	/* Try to unplug subblocks of partially plugged offline blocks. */
1421	virtio_mem_for_each_mb_state_rev(vm, mb_id,
1422					 VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL) {
1423		rc = virtio_mem_mb_unplug_any_sb_offline(vm, mb_id,
1424							 &nb_sb);
1425		if (rc || !nb_sb)
1426			goto out_unlock;
1427		cond_resched();
1428	}
1429
1430	/* Try to unplug subblocks of plugged offline blocks. */
1431	virtio_mem_for_each_mb_state_rev(vm, mb_id,
1432					 VIRTIO_MEM_MB_STATE_OFFLINE) {
1433		rc = virtio_mem_mb_unplug_any_sb_offline(vm, mb_id,
1434							 &nb_sb);
1435		if (rc || !nb_sb)
1436			goto out_unlock;
1437		cond_resched();
1438	}
1439
1440	if (!unplug_online) {
1441		mutex_unlock(&vm->hotplug_mutex);
1442		return 0;
1443	}
1444
1445	/* Try to unplug subblocks of partially plugged online blocks. */
1446	virtio_mem_for_each_mb_state_rev(vm, mb_id,
1447					 VIRTIO_MEM_MB_STATE_ONLINE_PARTIAL) {
1448		rc = virtio_mem_mb_unplug_any_sb_online(vm, mb_id,
1449							&nb_sb);
1450		if (rc || !nb_sb)
1451			goto out_unlock;
1452		mutex_unlock(&vm->hotplug_mutex);
1453		cond_resched();
1454		mutex_lock(&vm->hotplug_mutex);
1455	}
1456
1457	/* Try to unplug subblocks of plugged online blocks. */
1458	virtio_mem_for_each_mb_state_rev(vm, mb_id,
1459					 VIRTIO_MEM_MB_STATE_ONLINE) {
1460		rc = virtio_mem_mb_unplug_any_sb_online(vm, mb_id,
1461							&nb_sb);
1462		if (rc || !nb_sb)
1463			goto out_unlock;
1464		mutex_unlock(&vm->hotplug_mutex);
1465		cond_resched();
1466		mutex_lock(&vm->hotplug_mutex);
1467	}
1468
1469	mutex_unlock(&vm->hotplug_mutex);
1470	return nb_sb ? -EBUSY : 0;
1471out_unlock:
1472	mutex_unlock(&vm->hotplug_mutex);
1473	return rc;
1474}
1475
1476/*
1477 * Try to unplug all blocks that couldn't be unplugged before, for example,
1478 * because the hypervisor was busy.
1479 */
1480static int virtio_mem_unplug_pending_mb(struct virtio_mem *vm)
1481{
1482	unsigned long mb_id;
1483	int rc;
1484
1485	virtio_mem_for_each_mb_state(vm, mb_id, VIRTIO_MEM_MB_STATE_PLUGGED) {
1486		rc = virtio_mem_mb_unplug(vm, mb_id);
1487		if (rc)
1488			return rc;
1489		virtio_mem_mb_set_state(vm, mb_id, VIRTIO_MEM_MB_STATE_UNUSED);
1490	}
1491
1492	return 0;
1493}
1494
1495/*
1496 * Update all parts of the config that could have changed.
1497 */
1498static void virtio_mem_refresh_config(struct virtio_mem *vm)
1499{
1500	const uint64_t phys_limit = 1UL << MAX_PHYSMEM_BITS;
1501	uint64_t new_plugged_size, usable_region_size, end_addr;
1502
1503	/* the plugged_size is just a reflection of what _we_ did previously */
1504	virtio_cread_le(vm->vdev, struct virtio_mem_config, plugged_size,
1505			&new_plugged_size);
1506	if (WARN_ON_ONCE(new_plugged_size != vm->plugged_size))
1507		vm->plugged_size = new_plugged_size;
1508
1509	/* calculate the last usable memory block id */
1510	virtio_cread_le(vm->vdev, struct virtio_mem_config,
1511			usable_region_size, &usable_region_size);
1512	end_addr = vm->addr + usable_region_size;
1513	end_addr = min(end_addr, phys_limit);
1514	vm->last_usable_mb_id = virtio_mem_phys_to_mb_id(end_addr) - 1;
1515
1516	/* see if there is a request to change the size */
1517	virtio_cread_le(vm->vdev, struct virtio_mem_config, requested_size,
1518			&vm->requested_size);
1519
1520	dev_info(&vm->vdev->dev, "plugged size: 0x%llx", vm->plugged_size);
1521	dev_info(&vm->vdev->dev, "requested size: 0x%llx", vm->requested_size);
1522}
1523
1524/*
1525 * Workqueue function for handling plug/unplug requests and config updates.
1526 */
1527static void virtio_mem_run_wq(struct work_struct *work)
1528{
1529	struct virtio_mem *vm = container_of(work, struct virtio_mem, wq);
1530	uint64_t diff;
1531	int rc;
1532
1533	hrtimer_cancel(&vm->retry_timer);
1534
1535	if (vm->broken)
1536		return;
1537
1538retry:
1539	rc = 0;
1540
1541	/* Make sure we start with a clean state if there are leftovers. */
1542	if (unlikely(vm->unplug_all_required))
1543		rc = virtio_mem_send_unplug_all_request(vm);
1544
1545	if (atomic_read(&vm->config_changed)) {
1546		atomic_set(&vm->config_changed, 0);
1547		virtio_mem_refresh_config(vm);
1548	}
1549
1550	/* Unplug any leftovers from previous runs */
1551	if (!rc)
1552		rc = virtio_mem_unplug_pending_mb(vm);
1553
1554	if (!rc && vm->requested_size != vm->plugged_size) {
1555		if (vm->requested_size > vm->plugged_size) {
1556			diff = vm->requested_size - vm->plugged_size;
1557			rc = virtio_mem_plug_request(vm, diff);
1558		} else {
1559			diff = vm->plugged_size - vm->requested_size;
1560			rc = virtio_mem_unplug_request(vm, diff);
1561		}
1562	}
1563
1564	switch (rc) {
1565	case 0:
1566		vm->retry_timer_ms = VIRTIO_MEM_RETRY_TIMER_MIN_MS;
1567		break;
1568	case -ENOSPC:
1569		/*
1570		 * We cannot add any more memory (alignment, physical limit)
1571		 * or we have too many offline memory blocks.
1572		 */
1573		break;
1574	case -ETXTBSY:
1575		/*
1576		 * The hypervisor cannot process our request right now
1577		 * (e.g., out of memory, migrating);
1578		 */
1579	case -EBUSY:
1580		/*
1581		 * We cannot free up any memory to unplug it (all plugged memory
1582		 * is busy).
1583		 */
1584	case -ENOMEM:
1585		/* Out of memory, try again later. */
1586		hrtimer_start(&vm->retry_timer, ms_to_ktime(vm->retry_timer_ms),
1587			      HRTIMER_MODE_REL);
1588		break;
1589	case -EAGAIN:
1590		/* Retry immediately (e.g., the config changed). */
1591		goto retry;
1592	default:
1593		/* Unknown error, mark as broken */
1594		dev_err(&vm->vdev->dev,
1595			"unknown error, marking device broken: %d\n", rc);
1596		vm->broken = true;
1597	}
1598}
1599
1600static enum hrtimer_restart virtio_mem_timer_expired(struct hrtimer *timer)
1601{
1602	struct virtio_mem *vm = container_of(timer, struct virtio_mem,
1603					     retry_timer);
1604
1605	virtio_mem_retry(vm);
1606	vm->retry_timer_ms = min_t(unsigned int, vm->retry_timer_ms * 2,
1607				   VIRTIO_MEM_RETRY_TIMER_MAX_MS);
1608	return HRTIMER_NORESTART;
1609}
1610
1611static void virtio_mem_handle_response(struct virtqueue *vq)
1612{
1613	struct virtio_mem *vm = vq->vdev->priv;
1614
1615	wake_up(&vm->host_resp);
1616}
1617
1618static int virtio_mem_init_vq(struct virtio_mem *vm)
1619{
1620	struct virtqueue *vq;
1621
1622	vq = virtio_find_single_vq(vm->vdev, virtio_mem_handle_response,
1623				   "guest-request");
1624	if (IS_ERR(vq))
1625		return PTR_ERR(vq);
1626	vm->vq = vq;
1627
1628	return 0;
1629}
1630
1631static int virtio_mem_init(struct virtio_mem *vm)
1632{
1633	const uint64_t phys_limit = 1UL << MAX_PHYSMEM_BITS;
1634	uint16_t node_id;
1635
1636	if (!vm->vdev->config->get) {
1637		dev_err(&vm->vdev->dev, "config access disabled\n");
1638		return -EINVAL;
1639	}
1640
1641	/*
1642	 * We don't want to (un)plug or reuse any memory when in kdump. The
1643	 * memory is still accessible (but not mapped).
1644	 */
1645	if (is_kdump_kernel()) {
1646		dev_warn(&vm->vdev->dev, "disabled in kdump kernel\n");
1647		return -EBUSY;
1648	}
1649
1650	/* Fetch all properties that can't change. */
1651	virtio_cread_le(vm->vdev, struct virtio_mem_config, plugged_size,
1652			&vm->plugged_size);
1653	virtio_cread_le(vm->vdev, struct virtio_mem_config, block_size,
1654			&vm->device_block_size);
1655	virtio_cread_le(vm->vdev, struct virtio_mem_config, node_id,
1656			&node_id);
1657	vm->nid = virtio_mem_translate_node_id(vm, node_id);
1658	virtio_cread_le(vm->vdev, struct virtio_mem_config, addr, &vm->addr);
1659	virtio_cread_le(vm->vdev, struct virtio_mem_config, region_size,
1660			&vm->region_size);
1661
1662	/*
1663	 * We always hotplug memory in memory block granularity. This way,
1664	 * we have to wait for exactly one memory block to online.
1665	 */
1666	if (vm->device_block_size > memory_block_size_bytes()) {
1667		dev_err(&vm->vdev->dev,
1668			"The block size is not supported (too big).\n");
1669		return -EINVAL;
1670	}
1671
1672	/* bad device setup - warn only */
1673	if (!IS_ALIGNED(vm->addr, memory_block_size_bytes()))
1674		dev_warn(&vm->vdev->dev,
1675			 "The alignment of the physical start address can make some memory unusable.\n");
1676	if (!IS_ALIGNED(vm->addr + vm->region_size, memory_block_size_bytes()))
1677		dev_warn(&vm->vdev->dev,
1678			 "The alignment of the physical end address can make some memory unusable.\n");
1679	if (vm->addr + vm->region_size > phys_limit)
1680		dev_warn(&vm->vdev->dev,
1681			 "Some memory is not addressable. This can make some memory unusable.\n");
1682
1683	/*
1684	 * Calculate the subblock size:
1685	 * - At least MAX_ORDER - 1 / pageblock_order.
1686	 * - At least the device block size.
1687	 * In the worst case, a single subblock per memory block.
1688	 */
1689	vm->subblock_size = PAGE_SIZE * 1ul << max_t(uint32_t, MAX_ORDER - 1,
1690						     pageblock_order);
1691	vm->subblock_size = max_t(uint64_t, vm->device_block_size,
1692				  vm->subblock_size);
1693	vm->nb_sb_per_mb = memory_block_size_bytes() / vm->subblock_size;
1694
1695	/* Round up to the next full memory block */
1696	vm->first_mb_id = virtio_mem_phys_to_mb_id(vm->addr - 1 +
1697						   memory_block_size_bytes());
1698	vm->next_mb_id = vm->first_mb_id;
1699	vm->last_mb_id = virtio_mem_phys_to_mb_id(vm->addr +
1700			 vm->region_size) - 1;
1701
1702	dev_info(&vm->vdev->dev, "start address: 0x%llx", vm->addr);
1703	dev_info(&vm->vdev->dev, "region size: 0x%llx", vm->region_size);
1704	dev_info(&vm->vdev->dev, "device block size: 0x%llx",
1705		 (unsigned long long)vm->device_block_size);
1706	dev_info(&vm->vdev->dev, "memory block size: 0x%lx",
1707		 memory_block_size_bytes());
1708	dev_info(&vm->vdev->dev, "subblock size: 0x%llx",
1709		 (unsigned long long)vm->subblock_size);
1710	if (vm->nid != NUMA_NO_NODE)
1711		dev_info(&vm->vdev->dev, "nid: %d", vm->nid);
1712
1713	return 0;
1714}
1715
1716static int virtio_mem_create_resource(struct virtio_mem *vm)
1717{
1718	/*
1719	 * When force-unloading the driver and removing the device, we
1720	 * could have a garbage pointer. Duplicate the string.
1721	 */
1722	const char *name = kstrdup(dev_name(&vm->vdev->dev), GFP_KERNEL);
1723
1724	if (!name)
1725		return -ENOMEM;
1726
1727	vm->parent_resource = __request_mem_region(vm->addr, vm->region_size,
1728						   name, IORESOURCE_SYSTEM_RAM);
1729	if (!vm->parent_resource) {
1730		kfree(name);
1731		dev_warn(&vm->vdev->dev, "could not reserve device region\n");
1732		dev_info(&vm->vdev->dev,
1733			 "reloading the driver is not supported\n");
1734		return -EBUSY;
1735	}
1736
1737	/* The memory is not actually busy - make add_memory() work. */
1738	vm->parent_resource->flags &= ~IORESOURCE_BUSY;
1739	return 0;
1740}
1741
1742static void virtio_mem_delete_resource(struct virtio_mem *vm)
1743{
1744	const char *name;
1745
1746	if (!vm->parent_resource)
1747		return;
1748
1749	name = vm->parent_resource->name;
1750	release_resource(vm->parent_resource);
1751	kfree(vm->parent_resource);
1752	kfree(name);
1753	vm->parent_resource = NULL;
1754}
1755
1756static int virtio_mem_probe(struct virtio_device *vdev)
1757{
1758	struct virtio_mem *vm;
1759	int rc;
1760
1761	BUILD_BUG_ON(sizeof(struct virtio_mem_req) != 24);
1762	BUILD_BUG_ON(sizeof(struct virtio_mem_resp) != 10);
1763
1764	vdev->priv = vm = kzalloc(sizeof(*vm), GFP_KERNEL);
1765	if (!vm)
1766		return -ENOMEM;
1767
1768	init_waitqueue_head(&vm->host_resp);
1769	vm->vdev = vdev;
1770	INIT_WORK(&vm->wq, virtio_mem_run_wq);
1771	mutex_init(&vm->hotplug_mutex);
1772	INIT_LIST_HEAD(&vm->next);
1773	spin_lock_init(&vm->removal_lock);
1774	hrtimer_init(&vm->retry_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1775	vm->retry_timer.function = virtio_mem_timer_expired;
1776	vm->retry_timer_ms = VIRTIO_MEM_RETRY_TIMER_MIN_MS;
1777
1778	/* register the virtqueue */
1779	rc = virtio_mem_init_vq(vm);
1780	if (rc)
1781		goto out_free_vm;
1782
1783	/* initialize the device by querying the config */
1784	rc = virtio_mem_init(vm);
1785	if (rc)
1786		goto out_del_vq;
1787
1788	/* create the parent resource for all memory */
1789	rc = virtio_mem_create_resource(vm);
1790	if (rc)
1791		goto out_del_vq;
1792
1793	/*
1794	 * If we still have memory plugged, we have to unplug all memory first.
1795	 * Registering our parent resource makes sure that this memory isn't
1796	 * actually in use (e.g., trying to reload the driver).
1797	 */
1798	if (vm->plugged_size) {
1799		vm->unplug_all_required = 1;
1800		dev_info(&vm->vdev->dev, "unplugging all memory is required\n");
1801	}
1802
1803	/* register callbacks */
1804	vm->memory_notifier.notifier_call = virtio_mem_memory_notifier_cb;
1805	rc = register_memory_notifier(&vm->memory_notifier);
1806	if (rc)
1807		goto out_del_resource;
1808	rc = register_virtio_mem_device(vm);
1809	if (rc)
1810		goto out_unreg_mem;
1811
1812	virtio_device_ready(vdev);
1813
1814	/* trigger a config update to start processing the requested_size */
1815	atomic_set(&vm->config_changed, 1);
1816	queue_work(system_freezable_wq, &vm->wq);
1817
1818	return 0;
1819out_unreg_mem:
1820	unregister_memory_notifier(&vm->memory_notifier);
1821out_del_resource:
1822	virtio_mem_delete_resource(vm);
1823out_del_vq:
1824	vdev->config->del_vqs(vdev);
1825out_free_vm:
1826	kfree(vm);
1827	vdev->priv = NULL;
1828
1829	return rc;
1830}
1831
1832static void virtio_mem_remove(struct virtio_device *vdev)
1833{
1834	struct virtio_mem *vm = vdev->priv;
1835	unsigned long mb_id;
1836	int rc;
1837
1838	/*
1839	 * Make sure the workqueue won't be triggered anymore and no memory
1840	 * blocks can be onlined/offlined until we're finished here.
1841	 */
1842	mutex_lock(&vm->hotplug_mutex);
1843	spin_lock_irq(&vm->removal_lock);
1844	vm->removing = true;
1845	spin_unlock_irq(&vm->removal_lock);
1846	mutex_unlock(&vm->hotplug_mutex);
1847
1848	/* wait until the workqueue stopped */
1849	cancel_work_sync(&vm->wq);
1850	hrtimer_cancel(&vm->retry_timer);
1851
1852	/*
1853	 * After we unregistered our callbacks, user space can online partially
1854	 * plugged offline blocks. Make sure to remove them.
1855	 */
1856	virtio_mem_for_each_mb_state(vm, mb_id,
1857				     VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL) {
1858		rc = virtio_mem_mb_remove(vm, mb_id);
1859		BUG_ON(rc);
1860		virtio_mem_mb_set_state(vm, mb_id, VIRTIO_MEM_MB_STATE_UNUSED);
1861	}
1862	/*
1863	 * After we unregistered our callbacks, user space can no longer
1864	 * offline partially plugged online memory blocks. No need to worry
1865	 * about them.
1866	 */
1867
1868	/* unregister callbacks */
1869	unregister_virtio_mem_device(vm);
1870	unregister_memory_notifier(&vm->memory_notifier);
1871
1872	/*
1873	 * There is no way we could reliably remove all memory we have added to
1874	 * the system. And there is no way to stop the driver/device from going
1875	 * away. Warn at least.
1876	 */
1877	if (vm->nb_mb_state[VIRTIO_MEM_MB_STATE_OFFLINE] ||
1878	    vm->nb_mb_state[VIRTIO_MEM_MB_STATE_OFFLINE_PARTIAL] ||
1879	    vm->nb_mb_state[VIRTIO_MEM_MB_STATE_ONLINE] ||
1880	    vm->nb_mb_state[VIRTIO_MEM_MB_STATE_ONLINE_PARTIAL]) {
1881		dev_warn(&vdev->dev, "device still has system memory added\n");
1882	} else {
1883		virtio_mem_delete_resource(vm);
1884		kfree_const(vm->resource_name);
1885	}
1886
1887	/* remove all tracking data - no locking needed */
1888	vfree(vm->mb_state);
1889	vfree(vm->sb_bitmap);
1890
1891	/* reset the device and cleanup the queues */
1892	vdev->config->reset(vdev);
1893	vdev->config->del_vqs(vdev);
1894
1895	kfree(vm);
1896	vdev->priv = NULL;
1897}
1898
1899static void virtio_mem_config_changed(struct virtio_device *vdev)
1900{
1901	struct virtio_mem *vm = vdev->priv;
1902
1903	atomic_set(&vm->config_changed, 1);
1904	virtio_mem_retry(vm);
1905}
1906
1907#ifdef CONFIG_PM_SLEEP
1908static int virtio_mem_freeze(struct virtio_device *vdev)
1909{
1910	/*
1911	 * When restarting the VM, all memory is usually unplugged. Don't
1912	 * allow to suspend/hibernate.
1913	 */
1914	dev_err(&vdev->dev, "save/restore not supported.\n");
1915	return -EPERM;
1916}
1917
1918static int virtio_mem_restore(struct virtio_device *vdev)
1919{
1920	return -EPERM;
1921}
1922#endif
1923
1924static unsigned int virtio_mem_features[] = {
1925#if defined(CONFIG_NUMA) && defined(CONFIG_ACPI_NUMA)
1926	VIRTIO_MEM_F_ACPI_PXM,
1927#endif
1928};
1929
1930static const struct virtio_device_id virtio_mem_id_table[] = {
1931	{ VIRTIO_ID_MEM, VIRTIO_DEV_ANY_ID },
1932	{ 0 },
1933};
1934
1935static struct virtio_driver virtio_mem_driver = {
1936	.feature_table = virtio_mem_features,
1937	.feature_table_size = ARRAY_SIZE(virtio_mem_features),
1938	.driver.name = KBUILD_MODNAME,
1939	.driver.owner = THIS_MODULE,
1940	.id_table = virtio_mem_id_table,
1941	.probe = virtio_mem_probe,
1942	.remove = virtio_mem_remove,
1943	.config_changed = virtio_mem_config_changed,
1944#ifdef CONFIG_PM_SLEEP
1945	.freeze	=	virtio_mem_freeze,
1946	.restore =	virtio_mem_restore,
1947#endif
1948};
1949
1950module_virtio_driver(virtio_mem_driver);
1951MODULE_DEVICE_TABLE(virtio, virtio_mem_id_table);
1952MODULE_AUTHOR("David Hildenbrand <david@redhat.com>");
1953MODULE_DESCRIPTION("Virtio-mem driver");
1954MODULE_LICENSE("GPL");
1955