xref: /qemu/include/exec/memory.h (revision ab930e80)
1 /*
2  * Physical memory management API
3  *
4  * Copyright 2011 Red Hat, Inc. and/or its affiliates
5  *
6  * Authors:
7  *  Avi Kivity <avi@redhat.com>
8  *
9  * This work is licensed under the terms of the GNU GPL, version 2.  See
10  * the COPYING file in the top-level directory.
11  *
12  */
13 
14 #ifndef MEMORY_H
15 #define MEMORY_H
16 
17 #ifndef CONFIG_USER_ONLY
18 
19 #include "exec/cpu-common.h"
20 #include "exec/hwaddr.h"
21 #include "exec/memattrs.h"
22 #include "exec/memop.h"
23 #include "exec/ramlist.h"
24 #include "qemu/bswap.h"
25 #include "qemu/queue.h"
26 #include "qemu/int128.h"
27 #include "qemu/notify.h"
28 #include "qom/object.h"
29 #include "qemu/rcu.h"
30 
31 #define RAM_ADDR_INVALID (~(ram_addr_t)0)
32 
33 #define MAX_PHYS_ADDR_SPACE_BITS 62
34 #define MAX_PHYS_ADDR            (((hwaddr)1 << MAX_PHYS_ADDR_SPACE_BITS) - 1)
35 
36 #define TYPE_MEMORY_REGION "memory-region"
37 DECLARE_INSTANCE_CHECKER(MemoryRegion, MEMORY_REGION,
38                          TYPE_MEMORY_REGION)
39 
40 #define TYPE_IOMMU_MEMORY_REGION "iommu-memory-region"
41 typedef struct IOMMUMemoryRegionClass IOMMUMemoryRegionClass;
42 DECLARE_OBJ_CHECKERS(IOMMUMemoryRegion, IOMMUMemoryRegionClass,
43                      IOMMU_MEMORY_REGION, TYPE_IOMMU_MEMORY_REGION)
44 
45 #define TYPE_RAM_DISCARD_MANAGER "qemu:ram-discard-manager"
46 typedef struct RamDiscardManagerClass RamDiscardManagerClass;
47 typedef struct RamDiscardManager RamDiscardManager;
48 DECLARE_OBJ_CHECKERS(RamDiscardManager, RamDiscardManagerClass,
49                      RAM_DISCARD_MANAGER, TYPE_RAM_DISCARD_MANAGER);
50 
51 #ifdef CONFIG_FUZZ
52 void fuzz_dma_read_cb(size_t addr,
53                       size_t len,
54                       MemoryRegion *mr);
55 #else
56 static inline void fuzz_dma_read_cb(size_t addr,
57                                     size_t len,
58                                     MemoryRegion *mr)
59 {
60     /* Do Nothing */
61 }
62 #endif
63 
64 /* Possible bits for global_dirty_log_{start|stop} */
65 
66 /* Dirty tracking enabled because migration is running */
67 #define GLOBAL_DIRTY_MIGRATION  (1U << 0)
68 
69 /* Dirty tracking enabled because measuring dirty rate */
70 #define GLOBAL_DIRTY_DIRTY_RATE (1U << 1)
71 
72 /* Dirty tracking enabled because dirty limit */
73 #define GLOBAL_DIRTY_LIMIT      (1U << 2)
74 
75 #define GLOBAL_DIRTY_MASK  (0x7)
76 
77 extern unsigned int global_dirty_tracking;
78 
79 typedef struct MemoryRegionOps MemoryRegionOps;
80 
81 struct ReservedRegion {
82     hwaddr low;
83     hwaddr high;
84     unsigned type;
85 };
86 
87 /**
88  * struct MemoryRegionSection: describes a fragment of a #MemoryRegion
89  *
90  * @mr: the region, or %NULL if empty
91  * @fv: the flat view of the address space the region is mapped in
92  * @offset_within_region: the beginning of the section, relative to @mr's start
93  * @size: the size of the section; will not exceed @mr's boundaries
94  * @offset_within_address_space: the address of the first byte of the section
95  *     relative to the region's address space
96  * @readonly: writes to this section are ignored
97  * @nonvolatile: this section is non-volatile
98  */
99 struct MemoryRegionSection {
100     Int128 size;
101     MemoryRegion *mr;
102     FlatView *fv;
103     hwaddr offset_within_region;
104     hwaddr offset_within_address_space;
105     bool readonly;
106     bool nonvolatile;
107 };
108 
109 typedef struct IOMMUTLBEntry IOMMUTLBEntry;
110 
111 /* See address_space_translate: bit 0 is read, bit 1 is write.  */
112 typedef enum {
113     IOMMU_NONE = 0,
114     IOMMU_RO   = 1,
115     IOMMU_WO   = 2,
116     IOMMU_RW   = 3,
117 } IOMMUAccessFlags;
118 
119 #define IOMMU_ACCESS_FLAG(r, w) (((r) ? IOMMU_RO : 0) | ((w) ? IOMMU_WO : 0))
120 
121 struct IOMMUTLBEntry {
122     AddressSpace    *target_as;
123     hwaddr           iova;
124     hwaddr           translated_addr;
125     hwaddr           addr_mask;  /* 0xfff = 4k translation */
126     IOMMUAccessFlags perm;
127 };
128 
129 /*
130  * Bitmap for different IOMMUNotifier capabilities. Each notifier can
131  * register with one or multiple IOMMU Notifier capability bit(s).
132  *
133  * Normally there're two use cases for the notifiers:
134  *
135  *   (1) When the device needs accurate synchronizations of the vIOMMU page
136  *       tables, it needs to register with both MAP|UNMAP notifies (which
137  *       is defined as IOMMU_NOTIFIER_IOTLB_EVENTS below).
138  *
139  *       Regarding to accurate synchronization, it's when the notified
140  *       device maintains a shadow page table and must be notified on each
141  *       guest MAP (page table entry creation) and UNMAP (invalidation)
142  *       events (e.g. VFIO). Both notifications must be accurate so that
143  *       the shadow page table is fully in sync with the guest view.
144  *
145  *   (2) When the device doesn't need accurate synchronizations of the
146  *       vIOMMU page tables, it needs to register only with UNMAP or
147  *       DEVIOTLB_UNMAP notifies.
148  *
149  *       It's when the device maintains a cache of IOMMU translations
150  *       (IOTLB) and is able to fill that cache by requesting translations
151  *       from the vIOMMU through a protocol similar to ATS (Address
152  *       Translation Service).
153  *
154  *       Note that in this mode the vIOMMU will not maintain a shadowed
155  *       page table for the address space, and the UNMAP messages can cover
156  *       more than the pages that used to get mapped.  The IOMMU notifiee
157  *       should be able to take care of over-sized invalidations.
158  */
159 typedef enum {
160     IOMMU_NOTIFIER_NONE = 0,
161     /* Notify cache invalidations */
162     IOMMU_NOTIFIER_UNMAP = 0x1,
163     /* Notify entry changes (newly created entries) */
164     IOMMU_NOTIFIER_MAP = 0x2,
165     /* Notify changes on device IOTLB entries */
166     IOMMU_NOTIFIER_DEVIOTLB_UNMAP = 0x04,
167 } IOMMUNotifierFlag;
168 
169 #define IOMMU_NOTIFIER_IOTLB_EVENTS (IOMMU_NOTIFIER_MAP | IOMMU_NOTIFIER_UNMAP)
170 #define IOMMU_NOTIFIER_DEVIOTLB_EVENTS IOMMU_NOTIFIER_DEVIOTLB_UNMAP
171 #define IOMMU_NOTIFIER_ALL (IOMMU_NOTIFIER_IOTLB_EVENTS | \
172                             IOMMU_NOTIFIER_DEVIOTLB_EVENTS)
173 
174 struct IOMMUNotifier;
175 typedef void (*IOMMUNotify)(struct IOMMUNotifier *notifier,
176                             IOMMUTLBEntry *data);
177 
178 struct IOMMUNotifier {
179     IOMMUNotify notify;
180     IOMMUNotifierFlag notifier_flags;
181     /* Notify for address space range start <= addr <= end */
182     hwaddr start;
183     hwaddr end;
184     int iommu_idx;
185     QLIST_ENTRY(IOMMUNotifier) node;
186 };
187 typedef struct IOMMUNotifier IOMMUNotifier;
188 
189 typedef struct IOMMUTLBEvent {
190     IOMMUNotifierFlag type;
191     IOMMUTLBEntry entry;
192 } IOMMUTLBEvent;
193 
194 /* RAM is pre-allocated and passed into qemu_ram_alloc_from_ptr */
195 #define RAM_PREALLOC   (1 << 0)
196 
197 /* RAM is mmap-ed with MAP_SHARED */
198 #define RAM_SHARED     (1 << 1)
199 
200 /* Only a portion of RAM (used_length) is actually used, and migrated.
201  * Resizing RAM while migrating can result in the migration being canceled.
202  */
203 #define RAM_RESIZEABLE (1 << 2)
204 
205 /* UFFDIO_ZEROPAGE is available on this RAMBlock to atomically
206  * zero the page and wake waiting processes.
207  * (Set during postcopy)
208  */
209 #define RAM_UF_ZEROPAGE (1 << 3)
210 
211 /* RAM can be migrated */
212 #define RAM_MIGRATABLE (1 << 4)
213 
214 /* RAM is a persistent kind memory */
215 #define RAM_PMEM (1 << 5)
216 
217 
218 /*
219  * UFFDIO_WRITEPROTECT is used on this RAMBlock to
220  * support 'write-tracking' migration type.
221  * Implies ram_state->ram_wt_enabled.
222  */
223 #define RAM_UF_WRITEPROTECT (1 << 6)
224 
225 /*
226  * RAM is mmap-ed with MAP_NORESERVE. When set, reserving swap space (or huge
227  * pages if applicable) is skipped: will bail out if not supported. When not
228  * set, the OS will do the reservation, if supported for the memory type.
229  */
230 #define RAM_NORESERVE (1 << 7)
231 
232 /* RAM that isn't accessible through normal means. */
233 #define RAM_PROTECTED (1 << 8)
234 
235 static inline void iommu_notifier_init(IOMMUNotifier *n, IOMMUNotify fn,
236                                        IOMMUNotifierFlag flags,
237                                        hwaddr start, hwaddr end,
238                                        int iommu_idx)
239 {
240     n->notify = fn;
241     n->notifier_flags = flags;
242     n->start = start;
243     n->end = end;
244     n->iommu_idx = iommu_idx;
245 }
246 
247 /*
248  * Memory region callbacks
249  */
250 struct MemoryRegionOps {
251     /* Read from the memory region. @addr is relative to @mr; @size is
252      * in bytes. */
253     uint64_t (*read)(void *opaque,
254                      hwaddr addr,
255                      unsigned size);
256     /* Write to the memory region. @addr is relative to @mr; @size is
257      * in bytes. */
258     void (*write)(void *opaque,
259                   hwaddr addr,
260                   uint64_t data,
261                   unsigned size);
262 
263     MemTxResult (*read_with_attrs)(void *opaque,
264                                    hwaddr addr,
265                                    uint64_t *data,
266                                    unsigned size,
267                                    MemTxAttrs attrs);
268     MemTxResult (*write_with_attrs)(void *opaque,
269                                     hwaddr addr,
270                                     uint64_t data,
271                                     unsigned size,
272                                     MemTxAttrs attrs);
273 
274     enum device_endian endianness;
275     /* Guest-visible constraints: */
276     struct {
277         /* If nonzero, specify bounds on access sizes beyond which a machine
278          * check is thrown.
279          */
280         unsigned min_access_size;
281         unsigned max_access_size;
282         /* If true, unaligned accesses are supported.  Otherwise unaligned
283          * accesses throw machine checks.
284          */
285          bool unaligned;
286         /*
287          * If present, and returns #false, the transaction is not accepted
288          * by the device (and results in machine dependent behaviour such
289          * as a machine check exception).
290          */
291         bool (*accepts)(void *opaque, hwaddr addr,
292                         unsigned size, bool is_write,
293                         MemTxAttrs attrs);
294     } valid;
295     /* Internal implementation constraints: */
296     struct {
297         /* If nonzero, specifies the minimum size implemented.  Smaller sizes
298          * will be rounded upwards and a partial result will be returned.
299          */
300         unsigned min_access_size;
301         /* If nonzero, specifies the maximum size implemented.  Larger sizes
302          * will be done as a series of accesses with smaller sizes.
303          */
304         unsigned max_access_size;
305         /* If true, unaligned accesses are supported.  Otherwise all accesses
306          * are converted to (possibly multiple) naturally aligned accesses.
307          */
308         bool unaligned;
309     } impl;
310 };
311 
312 typedef struct MemoryRegionClass {
313     /* private */
314     ObjectClass parent_class;
315 } MemoryRegionClass;
316 
317 
318 enum IOMMUMemoryRegionAttr {
319     IOMMU_ATTR_SPAPR_TCE_FD
320 };
321 
322 /*
323  * IOMMUMemoryRegionClass:
324  *
325  * All IOMMU implementations need to subclass TYPE_IOMMU_MEMORY_REGION
326  * and provide an implementation of at least the @translate method here
327  * to handle requests to the memory region. Other methods are optional.
328  *
329  * The IOMMU implementation must use the IOMMU notifier infrastructure
330  * to report whenever mappings are changed, by calling
331  * memory_region_notify_iommu() (or, if necessary, by calling
332  * memory_region_notify_iommu_one() for each registered notifier).
333  *
334  * Conceptually an IOMMU provides a mapping from input address
335  * to an output TLB entry. If the IOMMU is aware of memory transaction
336  * attributes and the output TLB entry depends on the transaction
337  * attributes, we represent this using IOMMU indexes. Each index
338  * selects a particular translation table that the IOMMU has:
339  *
340  *   @attrs_to_index returns the IOMMU index for a set of transaction attributes
341  *
342  *   @translate takes an input address and an IOMMU index
343  *
344  * and the mapping returned can only depend on the input address and the
345  * IOMMU index.
346  *
347  * Most IOMMUs don't care about the transaction attributes and support
348  * only a single IOMMU index. A more complex IOMMU might have one index
349  * for secure transactions and one for non-secure transactions.
350  */
351 struct IOMMUMemoryRegionClass {
352     /* private: */
353     MemoryRegionClass parent_class;
354 
355     /* public: */
356     /**
357      * @translate:
358      *
359      * Return a TLB entry that contains a given address.
360      *
361      * The IOMMUAccessFlags indicated via @flag are optional and may
362      * be specified as IOMMU_NONE to indicate that the caller needs
363      * the full translation information for both reads and writes. If
364      * the access flags are specified then the IOMMU implementation
365      * may use this as an optimization, to stop doing a page table
366      * walk as soon as it knows that the requested permissions are not
367      * allowed. If IOMMU_NONE is passed then the IOMMU must do the
368      * full page table walk and report the permissions in the returned
369      * IOMMUTLBEntry. (Note that this implies that an IOMMU may not
370      * return different mappings for reads and writes.)
371      *
372      * The returned information remains valid while the caller is
373      * holding the big QEMU lock or is inside an RCU critical section;
374      * if the caller wishes to cache the mapping beyond that it must
375      * register an IOMMU notifier so it can invalidate its cached
376      * information when the IOMMU mapping changes.
377      *
378      * @iommu: the IOMMUMemoryRegion
379      *
380      * @hwaddr: address to be translated within the memory region
381      *
382      * @flag: requested access permission
383      *
384      * @iommu_idx: IOMMU index for the translation
385      */
386     IOMMUTLBEntry (*translate)(IOMMUMemoryRegion *iommu, hwaddr addr,
387                                IOMMUAccessFlags flag, int iommu_idx);
388     /**
389      * @get_min_page_size:
390      *
391      * Returns minimum supported page size in bytes.
392      *
393      * If this method is not provided then the minimum is assumed to
394      * be TARGET_PAGE_SIZE.
395      *
396      * @iommu: the IOMMUMemoryRegion
397      */
398     uint64_t (*get_min_page_size)(IOMMUMemoryRegion *iommu);
399     /**
400      * @notify_flag_changed:
401      *
402      * Called when IOMMU Notifier flag changes (ie when the set of
403      * events which IOMMU users are requesting notification for changes).
404      * Optional method -- need not be provided if the IOMMU does not
405      * need to know exactly which events must be notified.
406      *
407      * @iommu: the IOMMUMemoryRegion
408      *
409      * @old_flags: events which previously needed to be notified
410      *
411      * @new_flags: events which now need to be notified
412      *
413      * Returns 0 on success, or a negative errno; in particular
414      * returns -EINVAL if the new flag bitmap is not supported by the
415      * IOMMU memory region. In case of failure, the error object
416      * must be created
417      */
418     int (*notify_flag_changed)(IOMMUMemoryRegion *iommu,
419                                IOMMUNotifierFlag old_flags,
420                                IOMMUNotifierFlag new_flags,
421                                Error **errp);
422     /**
423      * @replay:
424      *
425      * Called to handle memory_region_iommu_replay().
426      *
427      * The default implementation of memory_region_iommu_replay() is to
428      * call the IOMMU translate method for every page in the address space
429      * with flag == IOMMU_NONE and then call the notifier if translate
430      * returns a valid mapping. If this method is implemented then it
431      * overrides the default behaviour, and must provide the full semantics
432      * of memory_region_iommu_replay(), by calling @notifier for every
433      * translation present in the IOMMU.
434      *
435      * Optional method -- an IOMMU only needs to provide this method
436      * if the default is inefficient or produces undesirable side effects.
437      *
438      * Note: this is not related to record-and-replay functionality.
439      */
440     void (*replay)(IOMMUMemoryRegion *iommu, IOMMUNotifier *notifier);
441 
442     /**
443      * @get_attr:
444      *
445      * Get IOMMU misc attributes. This is an optional method that
446      * can be used to allow users of the IOMMU to get implementation-specific
447      * information. The IOMMU implements this method to handle calls
448      * by IOMMU users to memory_region_iommu_get_attr() by filling in
449      * the arbitrary data pointer for any IOMMUMemoryRegionAttr values that
450      * the IOMMU supports. If the method is unimplemented then
451      * memory_region_iommu_get_attr() will always return -EINVAL.
452      *
453      * @iommu: the IOMMUMemoryRegion
454      *
455      * @attr: attribute being queried
456      *
457      * @data: memory to fill in with the attribute data
458      *
459      * Returns 0 on success, or a negative errno; in particular
460      * returns -EINVAL for unrecognized or unimplemented attribute types.
461      */
462     int (*get_attr)(IOMMUMemoryRegion *iommu, enum IOMMUMemoryRegionAttr attr,
463                     void *data);
464 
465     /**
466      * @attrs_to_index:
467      *
468      * Return the IOMMU index to use for a given set of transaction attributes.
469      *
470      * Optional method: if an IOMMU only supports a single IOMMU index then
471      * the default implementation of memory_region_iommu_attrs_to_index()
472      * will return 0.
473      *
474      * The indexes supported by an IOMMU must be contiguous, starting at 0.
475      *
476      * @iommu: the IOMMUMemoryRegion
477      * @attrs: memory transaction attributes
478      */
479     int (*attrs_to_index)(IOMMUMemoryRegion *iommu, MemTxAttrs attrs);
480 
481     /**
482      * @num_indexes:
483      *
484      * Return the number of IOMMU indexes this IOMMU supports.
485      *
486      * Optional method: if this method is not provided, then
487      * memory_region_iommu_num_indexes() will return 1, indicating that
488      * only a single IOMMU index is supported.
489      *
490      * @iommu: the IOMMUMemoryRegion
491      */
492     int (*num_indexes)(IOMMUMemoryRegion *iommu);
493 
494     /**
495      * @iommu_set_page_size_mask:
496      *
497      * Restrict the page size mask that can be supported with a given IOMMU
498      * memory region. Used for example to propagate host physical IOMMU page
499      * size mask limitations to the virtual IOMMU.
500      *
501      * Optional method: if this method is not provided, then the default global
502      * page mask is used.
503      *
504      * @iommu: the IOMMUMemoryRegion
505      *
506      * @page_size_mask: a bitmask of supported page sizes. At least one bit,
507      * representing the smallest page size, must be set. Additional set bits
508      * represent supported block sizes. For example a host physical IOMMU that
509      * uses page tables with a page size of 4kB, and supports 2MB and 4GB
510      * blocks, will set mask 0x40201000. A granule of 4kB with indiscriminate
511      * block sizes is specified with mask 0xfffffffffffff000.
512      *
513      * Returns 0 on success, or a negative error. In case of failure, the error
514      * object must be created.
515      */
516      int (*iommu_set_page_size_mask)(IOMMUMemoryRegion *iommu,
517                                      uint64_t page_size_mask,
518                                      Error **errp);
519 };
520 
521 typedef struct RamDiscardListener RamDiscardListener;
522 typedef int (*NotifyRamPopulate)(RamDiscardListener *rdl,
523                                  MemoryRegionSection *section);
524 typedef void (*NotifyRamDiscard)(RamDiscardListener *rdl,
525                                  MemoryRegionSection *section);
526 
527 struct RamDiscardListener {
528     /*
529      * @notify_populate:
530      *
531      * Notification that previously discarded memory is about to get populated.
532      * Listeners are able to object. If any listener objects, already
533      * successfully notified listeners are notified about a discard again.
534      *
535      * @rdl: the #RamDiscardListener getting notified
536      * @section: the #MemoryRegionSection to get populated. The section
537      *           is aligned within the memory region to the minimum granularity
538      *           unless it would exceed the registered section.
539      *
540      * Returns 0 on success. If the notification is rejected by the listener,
541      * an error is returned.
542      */
543     NotifyRamPopulate notify_populate;
544 
545     /*
546      * @notify_discard:
547      *
548      * Notification that previously populated memory was discarded successfully
549      * and listeners should drop all references to such memory and prevent
550      * new population (e.g., unmap).
551      *
552      * @rdl: the #RamDiscardListener getting notified
553      * @section: the #MemoryRegionSection to get populated. The section
554      *           is aligned within the memory region to the minimum granularity
555      *           unless it would exceed the registered section.
556      */
557     NotifyRamDiscard notify_discard;
558 
559     /*
560      * @double_discard_supported:
561      *
562      * The listener suppors getting @notify_discard notifications that span
563      * already discarded parts.
564      */
565     bool double_discard_supported;
566 
567     MemoryRegionSection *section;
568     QLIST_ENTRY(RamDiscardListener) next;
569 };
570 
571 static inline void ram_discard_listener_init(RamDiscardListener *rdl,
572                                              NotifyRamPopulate populate_fn,
573                                              NotifyRamDiscard discard_fn,
574                                              bool double_discard_supported)
575 {
576     rdl->notify_populate = populate_fn;
577     rdl->notify_discard = discard_fn;
578     rdl->double_discard_supported = double_discard_supported;
579 }
580 
581 typedef int (*ReplayRamPopulate)(MemoryRegionSection *section, void *opaque);
582 typedef void (*ReplayRamDiscard)(MemoryRegionSection *section, void *opaque);
583 
584 /*
585  * RamDiscardManagerClass:
586  *
587  * A #RamDiscardManager coordinates which parts of specific RAM #MemoryRegion
588  * regions are currently populated to be used/accessed by the VM, notifying
589  * after parts were discarded (freeing up memory) and before parts will be
590  * populated (consuming memory), to be used/accessed by the VM.
591  *
592  * A #RamDiscardManager can only be set for a RAM #MemoryRegion while the
593  * #MemoryRegion isn't mapped yet; it cannot change while the #MemoryRegion is
594  * mapped.
595  *
596  * The #RamDiscardManager is intended to be used by technologies that are
597  * incompatible with discarding of RAM (e.g., VFIO, which may pin all
598  * memory inside a #MemoryRegion), and require proper coordination to only
599  * map the currently populated parts, to hinder parts that are expected to
600  * remain discarded from silently getting populated and consuming memory.
601  * Technologies that support discarding of RAM don't have to bother and can
602  * simply map the whole #MemoryRegion.
603  *
604  * An example #RamDiscardManager is virtio-mem, which logically (un)plugs
605  * memory within an assigned RAM #MemoryRegion, coordinated with the VM.
606  * Logically unplugging memory consists of discarding RAM. The VM agreed to not
607  * access unplugged (discarded) memory - especially via DMA. virtio-mem will
608  * properly coordinate with listeners before memory is plugged (populated),
609  * and after memory is unplugged (discarded).
610  *
611  * Listeners are called in multiples of the minimum granularity (unless it
612  * would exceed the registered range) and changes are aligned to the minimum
613  * granularity within the #MemoryRegion. Listeners have to prepare for memory
614  * becoming discarded in a different granularity than it was populated and the
615  * other way around.
616  */
617 struct RamDiscardManagerClass {
618     /* private */
619     InterfaceClass parent_class;
620 
621     /* public */
622 
623     /**
624      * @get_min_granularity:
625      *
626      * Get the minimum granularity in which listeners will get notified
627      * about changes within the #MemoryRegion via the #RamDiscardManager.
628      *
629      * @rdm: the #RamDiscardManager
630      * @mr: the #MemoryRegion
631      *
632      * Returns the minimum granularity.
633      */
634     uint64_t (*get_min_granularity)(const RamDiscardManager *rdm,
635                                     const MemoryRegion *mr);
636 
637     /**
638      * @is_populated:
639      *
640      * Check whether the given #MemoryRegionSection is completely populated
641      * (i.e., no parts are currently discarded) via the #RamDiscardManager.
642      * There are no alignment requirements.
643      *
644      * @rdm: the #RamDiscardManager
645      * @section: the #MemoryRegionSection
646      *
647      * Returns whether the given range is completely populated.
648      */
649     bool (*is_populated)(const RamDiscardManager *rdm,
650                          const MemoryRegionSection *section);
651 
652     /**
653      * @replay_populated:
654      *
655      * Call the #ReplayRamPopulate callback for all populated parts within the
656      * #MemoryRegionSection via the #RamDiscardManager.
657      *
658      * In case any call fails, no further calls are made.
659      *
660      * @rdm: the #RamDiscardManager
661      * @section: the #MemoryRegionSection
662      * @replay_fn: the #ReplayRamPopulate callback
663      * @opaque: pointer to forward to the callback
664      *
665      * Returns 0 on success, or a negative error if any notification failed.
666      */
667     int (*replay_populated)(const RamDiscardManager *rdm,
668                             MemoryRegionSection *section,
669                             ReplayRamPopulate replay_fn, void *opaque);
670 
671     /**
672      * @replay_discarded:
673      *
674      * Call the #ReplayRamDiscard callback for all discarded parts within the
675      * #MemoryRegionSection via the #RamDiscardManager.
676      *
677      * @rdm: the #RamDiscardManager
678      * @section: the #MemoryRegionSection
679      * @replay_fn: the #ReplayRamDiscard callback
680      * @opaque: pointer to forward to the callback
681      */
682     void (*replay_discarded)(const RamDiscardManager *rdm,
683                              MemoryRegionSection *section,
684                              ReplayRamDiscard replay_fn, void *opaque);
685 
686     /**
687      * @register_listener:
688      *
689      * Register a #RamDiscardListener for the given #MemoryRegionSection and
690      * immediately notify the #RamDiscardListener about all populated parts
691      * within the #MemoryRegionSection via the #RamDiscardManager.
692      *
693      * In case any notification fails, no further notifications are triggered
694      * and an error is logged.
695      *
696      * @rdm: the #RamDiscardManager
697      * @rdl: the #RamDiscardListener
698      * @section: the #MemoryRegionSection
699      */
700     void (*register_listener)(RamDiscardManager *rdm,
701                               RamDiscardListener *rdl,
702                               MemoryRegionSection *section);
703 
704     /**
705      * @unregister_listener:
706      *
707      * Unregister a previously registered #RamDiscardListener via the
708      * #RamDiscardManager after notifying the #RamDiscardListener about all
709      * populated parts becoming unpopulated within the registered
710      * #MemoryRegionSection.
711      *
712      * @rdm: the #RamDiscardManager
713      * @rdl: the #RamDiscardListener
714      */
715     void (*unregister_listener)(RamDiscardManager *rdm,
716                                 RamDiscardListener *rdl);
717 };
718 
719 uint64_t ram_discard_manager_get_min_granularity(const RamDiscardManager *rdm,
720                                                  const MemoryRegion *mr);
721 
722 bool ram_discard_manager_is_populated(const RamDiscardManager *rdm,
723                                       const MemoryRegionSection *section);
724 
725 int ram_discard_manager_replay_populated(const RamDiscardManager *rdm,
726                                          MemoryRegionSection *section,
727                                          ReplayRamPopulate replay_fn,
728                                          void *opaque);
729 
730 void ram_discard_manager_replay_discarded(const RamDiscardManager *rdm,
731                                           MemoryRegionSection *section,
732                                           ReplayRamDiscard replay_fn,
733                                           void *opaque);
734 
735 void ram_discard_manager_register_listener(RamDiscardManager *rdm,
736                                            RamDiscardListener *rdl,
737                                            MemoryRegionSection *section);
738 
739 void ram_discard_manager_unregister_listener(RamDiscardManager *rdm,
740                                              RamDiscardListener *rdl);
741 
742 bool memory_get_xlat_addr(IOMMUTLBEntry *iotlb, void **vaddr,
743                           ram_addr_t *ram_addr, bool *read_only,
744                           bool *mr_has_discard_manager);
745 
746 typedef struct CoalescedMemoryRange CoalescedMemoryRange;
747 typedef struct MemoryRegionIoeventfd MemoryRegionIoeventfd;
748 
749 /** MemoryRegion:
750  *
751  * A struct representing a memory region.
752  */
753 struct MemoryRegion {
754     Object parent_obj;
755 
756     /* private: */
757 
758     /* The following fields should fit in a cache line */
759     bool romd_mode;
760     bool ram;
761     bool subpage;
762     bool readonly; /* For RAM regions */
763     bool nonvolatile;
764     bool rom_device;
765     bool flush_coalesced_mmio;
766     uint8_t dirty_log_mask;
767     bool is_iommu;
768     RAMBlock *ram_block;
769     Object *owner;
770     /* owner as TYPE_DEVICE. Used for re-entrancy checks in MR access hotpath */
771     DeviceState *dev;
772 
773     const MemoryRegionOps *ops;
774     void *opaque;
775     MemoryRegion *container;
776     int mapped_via_alias; /* Mapped via an alias, container might be NULL */
777     Int128 size;
778     hwaddr addr;
779     void (*destructor)(MemoryRegion *mr);
780     uint64_t align;
781     bool terminates;
782     bool ram_device;
783     bool enabled;
784     bool warning_printed; /* For reservations */
785     uint8_t vga_logging_count;
786     MemoryRegion *alias;
787     hwaddr alias_offset;
788     int32_t priority;
789     QTAILQ_HEAD(, MemoryRegion) subregions;
790     QTAILQ_ENTRY(MemoryRegion) subregions_link;
791     QTAILQ_HEAD(, CoalescedMemoryRange) coalesced;
792     const char *name;
793     unsigned ioeventfd_nb;
794     MemoryRegionIoeventfd *ioeventfds;
795     RamDiscardManager *rdm; /* Only for RAM */
796 
797     /* For devices designed to perform re-entrant IO into their own IO MRs */
798     bool disable_reentrancy_guard;
799 };
800 
801 struct IOMMUMemoryRegion {
802     MemoryRegion parent_obj;
803 
804     QLIST_HEAD(, IOMMUNotifier) iommu_notify;
805     IOMMUNotifierFlag iommu_notify_flags;
806 };
807 
808 #define IOMMU_NOTIFIER_FOREACH(n, mr) \
809     QLIST_FOREACH((n), &(mr)->iommu_notify, node)
810 
811 /**
812  * struct MemoryListener: callbacks structure for updates to the physical memory map
813  *
814  * Allows a component to adjust to changes in the guest-visible memory map.
815  * Use with memory_listener_register() and memory_listener_unregister().
816  */
817 struct MemoryListener {
818     /**
819      * @begin:
820      *
821      * Called at the beginning of an address space update transaction.
822      * Followed by calls to #MemoryListener.region_add(),
823      * #MemoryListener.region_del(), #MemoryListener.region_nop(),
824      * #MemoryListener.log_start() and #MemoryListener.log_stop() in
825      * increasing address order.
826      *
827      * @listener: The #MemoryListener.
828      */
829     void (*begin)(MemoryListener *listener);
830 
831     /**
832      * @commit:
833      *
834      * Called at the end of an address space update transaction,
835      * after the last call to #MemoryListener.region_add(),
836      * #MemoryListener.region_del() or #MemoryListener.region_nop(),
837      * #MemoryListener.log_start() and #MemoryListener.log_stop().
838      *
839      * @listener: The #MemoryListener.
840      */
841     void (*commit)(MemoryListener *listener);
842 
843     /**
844      * @region_add:
845      *
846      * Called during an address space update transaction,
847      * for a section of the address space that is new in this address space
848      * space since the last transaction.
849      *
850      * @listener: The #MemoryListener.
851      * @section: The new #MemoryRegionSection.
852      */
853     void (*region_add)(MemoryListener *listener, MemoryRegionSection *section);
854 
855     /**
856      * @region_del:
857      *
858      * Called during an address space update transaction,
859      * for a section of the address space that has disappeared in the address
860      * space since the last transaction.
861      *
862      * @listener: The #MemoryListener.
863      * @section: The old #MemoryRegionSection.
864      */
865     void (*region_del)(MemoryListener *listener, MemoryRegionSection *section);
866 
867     /**
868      * @region_nop:
869      *
870      * Called during an address space update transaction,
871      * for a section of the address space that is in the same place in the address
872      * space as in the last transaction.
873      *
874      * @listener: The #MemoryListener.
875      * @section: The #MemoryRegionSection.
876      */
877     void (*region_nop)(MemoryListener *listener, MemoryRegionSection *section);
878 
879     /**
880      * @log_start:
881      *
882      * Called during an address space update transaction, after
883      * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
884      * #MemoryListener.region_nop(), if dirty memory logging clients have
885      * become active since the last transaction.
886      *
887      * @listener: The #MemoryListener.
888      * @section: The #MemoryRegionSection.
889      * @old: A bitmap of dirty memory logging clients that were active in
890      * the previous transaction.
891      * @new: A bitmap of dirty memory logging clients that are active in
892      * the current transaction.
893      */
894     void (*log_start)(MemoryListener *listener, MemoryRegionSection *section,
895                       int old, int new);
896 
897     /**
898      * @log_stop:
899      *
900      * Called during an address space update transaction, after
901      * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
902      * #MemoryListener.region_nop() and possibly after
903      * #MemoryListener.log_start(), if dirty memory logging clients have
904      * become inactive since the last transaction.
905      *
906      * @listener: The #MemoryListener.
907      * @section: The #MemoryRegionSection.
908      * @old: A bitmap of dirty memory logging clients that were active in
909      * the previous transaction.
910      * @new: A bitmap of dirty memory logging clients that are active in
911      * the current transaction.
912      */
913     void (*log_stop)(MemoryListener *listener, MemoryRegionSection *section,
914                      int old, int new);
915 
916     /**
917      * @log_sync:
918      *
919      * Called by memory_region_snapshot_and_clear_dirty() and
920      * memory_global_dirty_log_sync(), before accessing QEMU's "official"
921      * copy of the dirty memory bitmap for a #MemoryRegionSection.
922      *
923      * @listener: The #MemoryListener.
924      * @section: The #MemoryRegionSection.
925      */
926     void (*log_sync)(MemoryListener *listener, MemoryRegionSection *section);
927 
928     /**
929      * @log_sync_global:
930      *
931      * This is the global version of @log_sync when the listener does
932      * not have a way to synchronize the log with finer granularity.
933      * When the listener registers with @log_sync_global defined, then
934      * its @log_sync must be NULL.  Vice versa.
935      *
936      * @listener: The #MemoryListener.
937      */
938     void (*log_sync_global)(MemoryListener *listener);
939 
940     /**
941      * @log_clear:
942      *
943      * Called before reading the dirty memory bitmap for a
944      * #MemoryRegionSection.
945      *
946      * @listener: The #MemoryListener.
947      * @section: The #MemoryRegionSection.
948      */
949     void (*log_clear)(MemoryListener *listener, MemoryRegionSection *section);
950 
951     /**
952      * @log_global_start:
953      *
954      * Called by memory_global_dirty_log_start(), which
955      * enables the %DIRTY_LOG_MIGRATION client on all memory regions in
956      * the address space.  #MemoryListener.log_global_start() is also
957      * called when a #MemoryListener is added, if global dirty logging is
958      * active at that time.
959      *
960      * @listener: The #MemoryListener.
961      */
962     void (*log_global_start)(MemoryListener *listener);
963 
964     /**
965      * @log_global_stop:
966      *
967      * Called by memory_global_dirty_log_stop(), which
968      * disables the %DIRTY_LOG_MIGRATION client on all memory regions in
969      * the address space.
970      *
971      * @listener: The #MemoryListener.
972      */
973     void (*log_global_stop)(MemoryListener *listener);
974 
975     /**
976      * @log_global_after_sync:
977      *
978      * Called after reading the dirty memory bitmap
979      * for any #MemoryRegionSection.
980      *
981      * @listener: The #MemoryListener.
982      */
983     void (*log_global_after_sync)(MemoryListener *listener);
984 
985     /**
986      * @eventfd_add:
987      *
988      * Called during an address space update transaction,
989      * for a section of the address space that has had a new ioeventfd
990      * registration since the last transaction.
991      *
992      * @listener: The #MemoryListener.
993      * @section: The new #MemoryRegionSection.
994      * @match_data: The @match_data parameter for the new ioeventfd.
995      * @data: The @data parameter for the new ioeventfd.
996      * @e: The #EventNotifier parameter for the new ioeventfd.
997      */
998     void (*eventfd_add)(MemoryListener *listener, MemoryRegionSection *section,
999                         bool match_data, uint64_t data, EventNotifier *e);
1000 
1001     /**
1002      * @eventfd_del:
1003      *
1004      * Called during an address space update transaction,
1005      * for a section of the address space that has dropped an ioeventfd
1006      * registration since the last transaction.
1007      *
1008      * @listener: The #MemoryListener.
1009      * @section: The new #MemoryRegionSection.
1010      * @match_data: The @match_data parameter for the dropped ioeventfd.
1011      * @data: The @data parameter for the dropped ioeventfd.
1012      * @e: The #EventNotifier parameter for the dropped ioeventfd.
1013      */
1014     void (*eventfd_del)(MemoryListener *listener, MemoryRegionSection *section,
1015                         bool match_data, uint64_t data, EventNotifier *e);
1016 
1017     /**
1018      * @coalesced_io_add:
1019      *
1020      * Called during an address space update transaction,
1021      * for a section of the address space that has had a new coalesced
1022      * MMIO range registration since the last transaction.
1023      *
1024      * @listener: The #MemoryListener.
1025      * @section: The new #MemoryRegionSection.
1026      * @addr: The starting address for the coalesced MMIO range.
1027      * @len: The length of the coalesced MMIO range.
1028      */
1029     void (*coalesced_io_add)(MemoryListener *listener, MemoryRegionSection *section,
1030                                hwaddr addr, hwaddr len);
1031 
1032     /**
1033      * @coalesced_io_del:
1034      *
1035      * Called during an address space update transaction,
1036      * for a section of the address space that has dropped a coalesced
1037      * MMIO range since the last transaction.
1038      *
1039      * @listener: The #MemoryListener.
1040      * @section: The new #MemoryRegionSection.
1041      * @addr: The starting address for the coalesced MMIO range.
1042      * @len: The length of the coalesced MMIO range.
1043      */
1044     void (*coalesced_io_del)(MemoryListener *listener, MemoryRegionSection *section,
1045                                hwaddr addr, hwaddr len);
1046     /**
1047      * @priority:
1048      *
1049      * Govern the order in which memory listeners are invoked. Lower priorities
1050      * are invoked earlier for "add" or "start" callbacks, and later for "delete"
1051      * or "stop" callbacks.
1052      */
1053     unsigned priority;
1054 
1055     /**
1056      * @name:
1057      *
1058      * Name of the listener.  It can be used in contexts where we'd like to
1059      * identify one memory listener with the rest.
1060      */
1061     const char *name;
1062 
1063     /* private: */
1064     AddressSpace *address_space;
1065     QTAILQ_ENTRY(MemoryListener) link;
1066     QTAILQ_ENTRY(MemoryListener) link_as;
1067 };
1068 
1069 /**
1070  * struct AddressSpace: describes a mapping of addresses to #MemoryRegion objects
1071  */
1072 struct AddressSpace {
1073     /* private: */
1074     struct rcu_head rcu;
1075     char *name;
1076     MemoryRegion *root;
1077 
1078     /* Accessed via RCU.  */
1079     struct FlatView *current_map;
1080 
1081     int ioeventfd_nb;
1082     struct MemoryRegionIoeventfd *ioeventfds;
1083     QTAILQ_HEAD(, MemoryListener) listeners;
1084     QTAILQ_ENTRY(AddressSpace) address_spaces_link;
1085 };
1086 
1087 typedef struct AddressSpaceDispatch AddressSpaceDispatch;
1088 typedef struct FlatRange FlatRange;
1089 
1090 /* Flattened global view of current active memory hierarchy.  Kept in sorted
1091  * order.
1092  */
1093 struct FlatView {
1094     struct rcu_head rcu;
1095     unsigned ref;
1096     FlatRange *ranges;
1097     unsigned nr;
1098     unsigned nr_allocated;
1099     struct AddressSpaceDispatch *dispatch;
1100     MemoryRegion *root;
1101 };
1102 
1103 static inline FlatView *address_space_to_flatview(AddressSpace *as)
1104 {
1105     return qatomic_rcu_read(&as->current_map);
1106 }
1107 
1108 /**
1109  * typedef flatview_cb: callback for flatview_for_each_range()
1110  *
1111  * @start: start address of the range within the FlatView
1112  * @len: length of the range in bytes
1113  * @mr: MemoryRegion covering this range
1114  * @offset_in_region: offset of the first byte of the range within @mr
1115  * @opaque: data pointer passed to flatview_for_each_range()
1116  *
1117  * Returns: true to stop the iteration, false to keep going.
1118  */
1119 typedef bool (*flatview_cb)(Int128 start,
1120                             Int128 len,
1121                             const MemoryRegion *mr,
1122                             hwaddr offset_in_region,
1123                             void *opaque);
1124 
1125 /**
1126  * flatview_for_each_range: Iterate through a FlatView
1127  * @fv: the FlatView to iterate through
1128  * @cb: function to call for each range
1129  * @opaque: opaque data pointer to pass to @cb
1130  *
1131  * A FlatView is made up of a list of non-overlapping ranges, each of
1132  * which is a slice of a MemoryRegion. This function iterates through
1133  * each range in @fv, calling @cb. The callback function can terminate
1134  * iteration early by returning 'true'.
1135  */
1136 void flatview_for_each_range(FlatView *fv, flatview_cb cb, void *opaque);
1137 
1138 static inline bool MemoryRegionSection_eq(MemoryRegionSection *a,
1139                                           MemoryRegionSection *b)
1140 {
1141     return a->mr == b->mr &&
1142            a->fv == b->fv &&
1143            a->offset_within_region == b->offset_within_region &&
1144            a->offset_within_address_space == b->offset_within_address_space &&
1145            int128_eq(a->size, b->size) &&
1146            a->readonly == b->readonly &&
1147            a->nonvolatile == b->nonvolatile;
1148 }
1149 
1150 /**
1151  * memory_region_section_new_copy: Copy a memory region section
1152  *
1153  * Allocate memory for a new copy, copy the memory region section, and
1154  * properly take a reference on all relevant members.
1155  *
1156  * @s: the #MemoryRegionSection to copy
1157  */
1158 MemoryRegionSection *memory_region_section_new_copy(MemoryRegionSection *s);
1159 
1160 /**
1161  * memory_region_section_new_copy: Free a copied memory region section
1162  *
1163  * Free a copy of a memory section created via memory_region_section_new_copy().
1164  * properly dropping references on all relevant members.
1165  *
1166  * @s: the #MemoryRegionSection to copy
1167  */
1168 void memory_region_section_free_copy(MemoryRegionSection *s);
1169 
1170 /**
1171  * memory_region_init: Initialize a memory region
1172  *
1173  * The region typically acts as a container for other memory regions.  Use
1174  * memory_region_add_subregion() to add subregions.
1175  *
1176  * @mr: the #MemoryRegion to be initialized
1177  * @owner: the object that tracks the region's reference count
1178  * @name: used for debugging; not visible to the user or ABI
1179  * @size: size of the region; any subregions beyond this size will be clipped
1180  */
1181 void memory_region_init(MemoryRegion *mr,
1182                         Object *owner,
1183                         const char *name,
1184                         uint64_t size);
1185 
1186 /**
1187  * memory_region_ref: Add 1 to a memory region's reference count
1188  *
1189  * Whenever memory regions are accessed outside the BQL, they need to be
1190  * preserved against hot-unplug.  MemoryRegions actually do not have their
1191  * own reference count; they piggyback on a QOM object, their "owner".
1192  * This function adds a reference to the owner.
1193  *
1194  * All MemoryRegions must have an owner if they can disappear, even if the
1195  * device they belong to operates exclusively under the BQL.  This is because
1196  * the region could be returned at any time by memory_region_find, and this
1197  * is usually under guest control.
1198  *
1199  * @mr: the #MemoryRegion
1200  */
1201 void memory_region_ref(MemoryRegion *mr);
1202 
1203 /**
1204  * memory_region_unref: Remove 1 to a memory region's reference count
1205  *
1206  * Whenever memory regions are accessed outside the BQL, they need to be
1207  * preserved against hot-unplug.  MemoryRegions actually do not have their
1208  * own reference count; they piggyback on a QOM object, their "owner".
1209  * This function removes a reference to the owner and possibly destroys it.
1210  *
1211  * @mr: the #MemoryRegion
1212  */
1213 void memory_region_unref(MemoryRegion *mr);
1214 
1215 /**
1216  * memory_region_init_io: Initialize an I/O memory region.
1217  *
1218  * Accesses into the region will cause the callbacks in @ops to be called.
1219  * if @size is nonzero, subregions will be clipped to @size.
1220  *
1221  * @mr: the #MemoryRegion to be initialized.
1222  * @owner: the object that tracks the region's reference count
1223  * @ops: a structure containing read and write callbacks to be used when
1224  *       I/O is performed on the region.
1225  * @opaque: passed to the read and write callbacks of the @ops structure.
1226  * @name: used for debugging; not visible to the user or ABI
1227  * @size: size of the region.
1228  */
1229 void memory_region_init_io(MemoryRegion *mr,
1230                            Object *owner,
1231                            const MemoryRegionOps *ops,
1232                            void *opaque,
1233                            const char *name,
1234                            uint64_t size);
1235 
1236 /**
1237  * memory_region_init_ram_nomigrate:  Initialize RAM memory region.  Accesses
1238  *                                    into the region will modify memory
1239  *                                    directly.
1240  *
1241  * @mr: the #MemoryRegion to be initialized.
1242  * @owner: the object that tracks the region's reference count
1243  * @name: Region name, becomes part of RAMBlock name used in migration stream
1244  *        must be unique within any device
1245  * @size: size of the region.
1246  * @errp: pointer to Error*, to store an error if it happens.
1247  *
1248  * Note that this function does not do anything to cause the data in the
1249  * RAM memory region to be migrated; that is the responsibility of the caller.
1250  */
1251 void memory_region_init_ram_nomigrate(MemoryRegion *mr,
1252                                       Object *owner,
1253                                       const char *name,
1254                                       uint64_t size,
1255                                       Error **errp);
1256 
1257 /**
1258  * memory_region_init_ram_flags_nomigrate:  Initialize RAM memory region.
1259  *                                          Accesses into the region will
1260  *                                          modify memory directly.
1261  *
1262  * @mr: the #MemoryRegion to be initialized.
1263  * @owner: the object that tracks the region's reference count
1264  * @name: Region name, becomes part of RAMBlock name used in migration stream
1265  *        must be unique within any device
1266  * @size: size of the region.
1267  * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_NORESERVE.
1268  * @errp: pointer to Error*, to store an error if it happens.
1269  *
1270  * Note that this function does not do anything to cause the data in the
1271  * RAM memory region to be migrated; that is the responsibility of the caller.
1272  */
1273 void memory_region_init_ram_flags_nomigrate(MemoryRegion *mr,
1274                                             Object *owner,
1275                                             const char *name,
1276                                             uint64_t size,
1277                                             uint32_t ram_flags,
1278                                             Error **errp);
1279 
1280 /**
1281  * memory_region_init_resizeable_ram:  Initialize memory region with resizable
1282  *                                     RAM.  Accesses into the region will
1283  *                                     modify memory directly.  Only an initial
1284  *                                     portion of this RAM is actually used.
1285  *                                     Changing the size while migrating
1286  *                                     can result in the migration being
1287  *                                     canceled.
1288  *
1289  * @mr: the #MemoryRegion to be initialized.
1290  * @owner: the object that tracks the region's reference count
1291  * @name: Region name, becomes part of RAMBlock name used in migration stream
1292  *        must be unique within any device
1293  * @size: used size of the region.
1294  * @max_size: max size of the region.
1295  * @resized: callback to notify owner about used size change.
1296  * @errp: pointer to Error*, to store an error if it happens.
1297  *
1298  * Note that this function does not do anything to cause the data in the
1299  * RAM memory region to be migrated; that is the responsibility of the caller.
1300  */
1301 void memory_region_init_resizeable_ram(MemoryRegion *mr,
1302                                        Object *owner,
1303                                        const char *name,
1304                                        uint64_t size,
1305                                        uint64_t max_size,
1306                                        void (*resized)(const char*,
1307                                                        uint64_t length,
1308                                                        void *host),
1309                                        Error **errp);
1310 #ifdef CONFIG_POSIX
1311 
1312 /**
1313  * memory_region_init_ram_from_file:  Initialize RAM memory region with a
1314  *                                    mmap-ed backend.
1315  *
1316  * @mr: the #MemoryRegion to be initialized.
1317  * @owner: the object that tracks the region's reference count
1318  * @name: Region name, becomes part of RAMBlock name used in migration stream
1319  *        must be unique within any device
1320  * @size: size of the region.
1321  * @align: alignment of the region base address; if 0, the default alignment
1322  *         (getpagesize()) will be used.
1323  * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
1324  *             RAM_NORESERVE,
1325  * @path: the path in which to allocate the RAM.
1326  * @readonly: true to open @path for reading, false for read/write.
1327  * @errp: pointer to Error*, to store an error if it happens.
1328  *
1329  * Note that this function does not do anything to cause the data in the
1330  * RAM memory region to be migrated; that is the responsibility of the caller.
1331  */
1332 void memory_region_init_ram_from_file(MemoryRegion *mr,
1333                                       Object *owner,
1334                                       const char *name,
1335                                       uint64_t size,
1336                                       uint64_t align,
1337                                       uint32_t ram_flags,
1338                                       const char *path,
1339                                       bool readonly,
1340                                       Error **errp);
1341 
1342 /**
1343  * memory_region_init_ram_from_fd:  Initialize RAM memory region with a
1344  *                                  mmap-ed backend.
1345  *
1346  * @mr: the #MemoryRegion to be initialized.
1347  * @owner: the object that tracks the region's reference count
1348  * @name: the name of the region.
1349  * @size: size of the region.
1350  * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
1351  *             RAM_NORESERVE, RAM_PROTECTED.
1352  * @fd: the fd to mmap.
1353  * @offset: offset within the file referenced by fd
1354  * @errp: pointer to Error*, to store an error if it happens.
1355  *
1356  * Note that this function does not do anything to cause the data in the
1357  * RAM memory region to be migrated; that is the responsibility of the caller.
1358  */
1359 void memory_region_init_ram_from_fd(MemoryRegion *mr,
1360                                     Object *owner,
1361                                     const char *name,
1362                                     uint64_t size,
1363                                     uint32_t ram_flags,
1364                                     int fd,
1365                                     ram_addr_t offset,
1366                                     Error **errp);
1367 #endif
1368 
1369 /**
1370  * memory_region_init_ram_ptr:  Initialize RAM memory region from a
1371  *                              user-provided pointer.  Accesses into the
1372  *                              region will modify memory directly.
1373  *
1374  * @mr: the #MemoryRegion to be initialized.
1375  * @owner: the object that tracks the region's reference count
1376  * @name: Region name, becomes part of RAMBlock name used in migration stream
1377  *        must be unique within any device
1378  * @size: size of the region.
1379  * @ptr: memory to be mapped; must contain at least @size bytes.
1380  *
1381  * Note that this function does not do anything to cause the data in the
1382  * RAM memory region to be migrated; that is the responsibility of the caller.
1383  */
1384 void memory_region_init_ram_ptr(MemoryRegion *mr,
1385                                 Object *owner,
1386                                 const char *name,
1387                                 uint64_t size,
1388                                 void *ptr);
1389 
1390 /**
1391  * memory_region_init_ram_device_ptr:  Initialize RAM device memory region from
1392  *                                     a user-provided pointer.
1393  *
1394  * A RAM device represents a mapping to a physical device, such as to a PCI
1395  * MMIO BAR of an vfio-pci assigned device.  The memory region may be mapped
1396  * into the VM address space and access to the region will modify memory
1397  * directly.  However, the memory region should not be included in a memory
1398  * dump (device may not be enabled/mapped at the time of the dump), and
1399  * operations incompatible with manipulating MMIO should be avoided.  Replaces
1400  * skip_dump flag.
1401  *
1402  * @mr: the #MemoryRegion to be initialized.
1403  * @owner: the object that tracks the region's reference count
1404  * @name: the name of the region.
1405  * @size: size of the region.
1406  * @ptr: memory to be mapped; must contain at least @size bytes.
1407  *
1408  * Note that this function does not do anything to cause the data in the
1409  * RAM memory region to be migrated; that is the responsibility of the caller.
1410  * (For RAM device memory regions, migrating the contents rarely makes sense.)
1411  */
1412 void memory_region_init_ram_device_ptr(MemoryRegion *mr,
1413                                        Object *owner,
1414                                        const char *name,
1415                                        uint64_t size,
1416                                        void *ptr);
1417 
1418 /**
1419  * memory_region_init_alias: Initialize a memory region that aliases all or a
1420  *                           part of another memory region.
1421  *
1422  * @mr: the #MemoryRegion to be initialized.
1423  * @owner: the object that tracks the region's reference count
1424  * @name: used for debugging; not visible to the user or ABI
1425  * @orig: the region to be referenced; @mr will be equivalent to
1426  *        @orig between @offset and @offset + @size - 1.
1427  * @offset: start of the section in @orig to be referenced.
1428  * @size: size of the region.
1429  */
1430 void memory_region_init_alias(MemoryRegion *mr,
1431                               Object *owner,
1432                               const char *name,
1433                               MemoryRegion *orig,
1434                               hwaddr offset,
1435                               uint64_t size);
1436 
1437 /**
1438  * memory_region_init_rom_nomigrate: Initialize a ROM memory region.
1439  *
1440  * This has the same effect as calling memory_region_init_ram_nomigrate()
1441  * and then marking the resulting region read-only with
1442  * memory_region_set_readonly().
1443  *
1444  * Note that this function does not do anything to cause the data in the
1445  * RAM side of the memory region to be migrated; that is the responsibility
1446  * of the caller.
1447  *
1448  * @mr: the #MemoryRegion to be initialized.
1449  * @owner: the object that tracks the region's reference count
1450  * @name: Region name, becomes part of RAMBlock name used in migration stream
1451  *        must be unique within any device
1452  * @size: size of the region.
1453  * @errp: pointer to Error*, to store an error if it happens.
1454  */
1455 void memory_region_init_rom_nomigrate(MemoryRegion *mr,
1456                                       Object *owner,
1457                                       const char *name,
1458                                       uint64_t size,
1459                                       Error **errp);
1460 
1461 /**
1462  * memory_region_init_rom_device_nomigrate:  Initialize a ROM memory region.
1463  *                                 Writes are handled via callbacks.
1464  *
1465  * Note that this function does not do anything to cause the data in the
1466  * RAM side of the memory region to be migrated; that is the responsibility
1467  * of the caller.
1468  *
1469  * @mr: the #MemoryRegion to be initialized.
1470  * @owner: the object that tracks the region's reference count
1471  * @ops: callbacks for write access handling (must not be NULL).
1472  * @opaque: passed to the read and write callbacks of the @ops structure.
1473  * @name: Region name, becomes part of RAMBlock name used in migration stream
1474  *        must be unique within any device
1475  * @size: size of the region.
1476  * @errp: pointer to Error*, to store an error if it happens.
1477  */
1478 void memory_region_init_rom_device_nomigrate(MemoryRegion *mr,
1479                                              Object *owner,
1480                                              const MemoryRegionOps *ops,
1481                                              void *opaque,
1482                                              const char *name,
1483                                              uint64_t size,
1484                                              Error **errp);
1485 
1486 /**
1487  * memory_region_init_iommu: Initialize a memory region of a custom type
1488  * that translates addresses
1489  *
1490  * An IOMMU region translates addresses and forwards accesses to a target
1491  * memory region.
1492  *
1493  * The IOMMU implementation must define a subclass of TYPE_IOMMU_MEMORY_REGION.
1494  * @_iommu_mr should be a pointer to enough memory for an instance of
1495  * that subclass, @instance_size is the size of that subclass, and
1496  * @mrtypename is its name. This function will initialize @_iommu_mr as an
1497  * instance of the subclass, and its methods will then be called to handle
1498  * accesses to the memory region. See the documentation of
1499  * #IOMMUMemoryRegionClass for further details.
1500  *
1501  * @_iommu_mr: the #IOMMUMemoryRegion to be initialized
1502  * @instance_size: the IOMMUMemoryRegion subclass instance size
1503  * @mrtypename: the type name of the #IOMMUMemoryRegion
1504  * @owner: the object that tracks the region's reference count
1505  * @name: used for debugging; not visible to the user or ABI
1506  * @size: size of the region.
1507  */
1508 void memory_region_init_iommu(void *_iommu_mr,
1509                               size_t instance_size,
1510                               const char *mrtypename,
1511                               Object *owner,
1512                               const char *name,
1513                               uint64_t size);
1514 
1515 /**
1516  * memory_region_init_ram - Initialize RAM memory region.  Accesses into the
1517  *                          region will modify memory directly.
1518  *
1519  * @mr: the #MemoryRegion to be initialized
1520  * @owner: the object that tracks the region's reference count (must be
1521  *         TYPE_DEVICE or a subclass of TYPE_DEVICE, or NULL)
1522  * @name: name of the memory region
1523  * @size: size of the region in bytes
1524  * @errp: pointer to Error*, to store an error if it happens.
1525  *
1526  * This function allocates RAM for a board model or device, and
1527  * arranges for it to be migrated (by calling vmstate_register_ram()
1528  * if @owner is a DeviceState, or vmstate_register_ram_global() if
1529  * @owner is NULL).
1530  *
1531  * TODO: Currently we restrict @owner to being either NULL (for
1532  * global RAM regions with no owner) or devices, so that we can
1533  * give the RAM block a unique name for migration purposes.
1534  * We should lift this restriction and allow arbitrary Objects.
1535  * If you pass a non-NULL non-device @owner then we will assert.
1536  */
1537 void memory_region_init_ram(MemoryRegion *mr,
1538                             Object *owner,
1539                             const char *name,
1540                             uint64_t size,
1541                             Error **errp);
1542 
1543 /**
1544  * memory_region_init_rom: Initialize a ROM memory region.
1545  *
1546  * This has the same effect as calling memory_region_init_ram()
1547  * and then marking the resulting region read-only with
1548  * memory_region_set_readonly(). This includes arranging for the
1549  * contents to be migrated.
1550  *
1551  * TODO: Currently we restrict @owner to being either NULL (for
1552  * global RAM regions with no owner) or devices, so that we can
1553  * give the RAM block a unique name for migration purposes.
1554  * We should lift this restriction and allow arbitrary Objects.
1555  * If you pass a non-NULL non-device @owner then we will assert.
1556  *
1557  * @mr: the #MemoryRegion to be initialized.
1558  * @owner: the object that tracks the region's reference count
1559  * @name: Region name, becomes part of RAMBlock name used in migration stream
1560  *        must be unique within any device
1561  * @size: size of the region.
1562  * @errp: pointer to Error*, to store an error if it happens.
1563  */
1564 void memory_region_init_rom(MemoryRegion *mr,
1565                             Object *owner,
1566                             const char *name,
1567                             uint64_t size,
1568                             Error **errp);
1569 
1570 /**
1571  * memory_region_init_rom_device:  Initialize a ROM memory region.
1572  *                                 Writes are handled via callbacks.
1573  *
1574  * This function initializes a memory region backed by RAM for reads
1575  * and callbacks for writes, and arranges for the RAM backing to
1576  * be migrated (by calling vmstate_register_ram()
1577  * if @owner is a DeviceState, or vmstate_register_ram_global() if
1578  * @owner is NULL).
1579  *
1580  * TODO: Currently we restrict @owner to being either NULL (for
1581  * global RAM regions with no owner) or devices, so that we can
1582  * give the RAM block a unique name for migration purposes.
1583  * We should lift this restriction and allow arbitrary Objects.
1584  * If you pass a non-NULL non-device @owner then we will assert.
1585  *
1586  * @mr: the #MemoryRegion to be initialized.
1587  * @owner: the object that tracks the region's reference count
1588  * @ops: callbacks for write access handling (must not be NULL).
1589  * @opaque: passed to the read and write callbacks of the @ops structure.
1590  * @name: Region name, becomes part of RAMBlock name used in migration stream
1591  *        must be unique within any device
1592  * @size: size of the region.
1593  * @errp: pointer to Error*, to store an error if it happens.
1594  */
1595 void memory_region_init_rom_device(MemoryRegion *mr,
1596                                    Object *owner,
1597                                    const MemoryRegionOps *ops,
1598                                    void *opaque,
1599                                    const char *name,
1600                                    uint64_t size,
1601                                    Error **errp);
1602 
1603 
1604 /**
1605  * memory_region_owner: get a memory region's owner.
1606  *
1607  * @mr: the memory region being queried.
1608  */
1609 Object *memory_region_owner(MemoryRegion *mr);
1610 
1611 /**
1612  * memory_region_size: get a memory region's size.
1613  *
1614  * @mr: the memory region being queried.
1615  */
1616 uint64_t memory_region_size(MemoryRegion *mr);
1617 
1618 /**
1619  * memory_region_is_ram: check whether a memory region is random access
1620  *
1621  * Returns %true if a memory region is random access.
1622  *
1623  * @mr: the memory region being queried
1624  */
1625 static inline bool memory_region_is_ram(MemoryRegion *mr)
1626 {
1627     return mr->ram;
1628 }
1629 
1630 /**
1631  * memory_region_is_ram_device: check whether a memory region is a ram device
1632  *
1633  * Returns %true if a memory region is a device backed ram region
1634  *
1635  * @mr: the memory region being queried
1636  */
1637 bool memory_region_is_ram_device(MemoryRegion *mr);
1638 
1639 /**
1640  * memory_region_is_romd: check whether a memory region is in ROMD mode
1641  *
1642  * Returns %true if a memory region is a ROM device and currently set to allow
1643  * direct reads.
1644  *
1645  * @mr: the memory region being queried
1646  */
1647 static inline bool memory_region_is_romd(MemoryRegion *mr)
1648 {
1649     return mr->rom_device && mr->romd_mode;
1650 }
1651 
1652 /**
1653  * memory_region_is_protected: check whether a memory region is protected
1654  *
1655  * Returns %true if a memory region is protected RAM and cannot be accessed
1656  * via standard mechanisms, e.g. DMA.
1657  *
1658  * @mr: the memory region being queried
1659  */
1660 bool memory_region_is_protected(MemoryRegion *mr);
1661 
1662 /**
1663  * memory_region_get_iommu: check whether a memory region is an iommu
1664  *
1665  * Returns pointer to IOMMUMemoryRegion if a memory region is an iommu,
1666  * otherwise NULL.
1667  *
1668  * @mr: the memory region being queried
1669  */
1670 static inline IOMMUMemoryRegion *memory_region_get_iommu(MemoryRegion *mr)
1671 {
1672     if (mr->alias) {
1673         return memory_region_get_iommu(mr->alias);
1674     }
1675     if (mr->is_iommu) {
1676         return (IOMMUMemoryRegion *) mr;
1677     }
1678     return NULL;
1679 }
1680 
1681 /**
1682  * memory_region_get_iommu_class_nocheck: returns iommu memory region class
1683  *   if an iommu or NULL if not
1684  *
1685  * Returns pointer to IOMMUMemoryRegionClass if a memory region is an iommu,
1686  * otherwise NULL. This is fast path avoiding QOM checking, use with caution.
1687  *
1688  * @iommu_mr: the memory region being queried
1689  */
1690 static inline IOMMUMemoryRegionClass *memory_region_get_iommu_class_nocheck(
1691         IOMMUMemoryRegion *iommu_mr)
1692 {
1693     return (IOMMUMemoryRegionClass *) (((Object *)iommu_mr)->class);
1694 }
1695 
1696 #define memory_region_is_iommu(mr) (memory_region_get_iommu(mr) != NULL)
1697 
1698 /**
1699  * memory_region_iommu_get_min_page_size: get minimum supported page size
1700  * for an iommu
1701  *
1702  * Returns minimum supported page size for an iommu.
1703  *
1704  * @iommu_mr: the memory region being queried
1705  */
1706 uint64_t memory_region_iommu_get_min_page_size(IOMMUMemoryRegion *iommu_mr);
1707 
1708 /**
1709  * memory_region_notify_iommu: notify a change in an IOMMU translation entry.
1710  *
1711  * Note: for any IOMMU implementation, an in-place mapping change
1712  * should be notified with an UNMAP followed by a MAP.
1713  *
1714  * @iommu_mr: the memory region that was changed
1715  * @iommu_idx: the IOMMU index for the translation table which has changed
1716  * @event: TLB event with the new entry in the IOMMU translation table.
1717  *         The entry replaces all old entries for the same virtual I/O address
1718  *         range.
1719  */
1720 void memory_region_notify_iommu(IOMMUMemoryRegion *iommu_mr,
1721                                 int iommu_idx,
1722                                 IOMMUTLBEvent event);
1723 
1724 /**
1725  * memory_region_notify_iommu_one: notify a change in an IOMMU translation
1726  *                           entry to a single notifier
1727  *
1728  * This works just like memory_region_notify_iommu(), but it only
1729  * notifies a specific notifier, not all of them.
1730  *
1731  * @notifier: the notifier to be notified
1732  * @event: TLB event with the new entry in the IOMMU translation table.
1733  *         The entry replaces all old entries for the same virtual I/O address
1734  *         range.
1735  */
1736 void memory_region_notify_iommu_one(IOMMUNotifier *notifier,
1737                                     IOMMUTLBEvent *event);
1738 
1739 /**
1740  * memory_region_unmap_iommu_notifier_range: notify a unmap for an IOMMU
1741  *                                           translation that covers the
1742  *                                           range of a notifier
1743  *
1744  * @notifier: the notifier to be notified
1745  */
1746 void memory_region_unmap_iommu_notifier_range(IOMMUNotifier *notifier);
1747 
1748 
1749 /**
1750  * memory_region_register_iommu_notifier: register a notifier for changes to
1751  * IOMMU translation entries.
1752  *
1753  * Returns 0 on success, or a negative errno otherwise. In particular,
1754  * -EINVAL indicates that at least one of the attributes of the notifier
1755  * is not supported (flag/range) by the IOMMU memory region. In case of error
1756  * the error object must be created.
1757  *
1758  * @mr: the memory region to observe
1759  * @n: the IOMMUNotifier to be added; the notify callback receives a
1760  *     pointer to an #IOMMUTLBEntry as the opaque value; the pointer
1761  *     ceases to be valid on exit from the notifier.
1762  * @errp: pointer to Error*, to store an error if it happens.
1763  */
1764 int memory_region_register_iommu_notifier(MemoryRegion *mr,
1765                                           IOMMUNotifier *n, Error **errp);
1766 
1767 /**
1768  * memory_region_iommu_replay: replay existing IOMMU translations to
1769  * a notifier with the minimum page granularity returned by
1770  * mr->iommu_ops->get_page_size().
1771  *
1772  * Note: this is not related to record-and-replay functionality.
1773  *
1774  * @iommu_mr: the memory region to observe
1775  * @n: the notifier to which to replay iommu mappings
1776  */
1777 void memory_region_iommu_replay(IOMMUMemoryRegion *iommu_mr, IOMMUNotifier *n);
1778 
1779 /**
1780  * memory_region_unregister_iommu_notifier: unregister a notifier for
1781  * changes to IOMMU translation entries.
1782  *
1783  * @mr: the memory region which was observed and for which notity_stopped()
1784  *      needs to be called
1785  * @n: the notifier to be removed.
1786  */
1787 void memory_region_unregister_iommu_notifier(MemoryRegion *mr,
1788                                              IOMMUNotifier *n);
1789 
1790 /**
1791  * memory_region_iommu_get_attr: return an IOMMU attr if get_attr() is
1792  * defined on the IOMMU.
1793  *
1794  * Returns 0 on success, or a negative errno otherwise. In particular,
1795  * -EINVAL indicates that the IOMMU does not support the requested
1796  * attribute.
1797  *
1798  * @iommu_mr: the memory region
1799  * @attr: the requested attribute
1800  * @data: a pointer to the requested attribute data
1801  */
1802 int memory_region_iommu_get_attr(IOMMUMemoryRegion *iommu_mr,
1803                                  enum IOMMUMemoryRegionAttr attr,
1804                                  void *data);
1805 
1806 /**
1807  * memory_region_iommu_attrs_to_index: return the IOMMU index to
1808  * use for translations with the given memory transaction attributes.
1809  *
1810  * @iommu_mr: the memory region
1811  * @attrs: the memory transaction attributes
1812  */
1813 int memory_region_iommu_attrs_to_index(IOMMUMemoryRegion *iommu_mr,
1814                                        MemTxAttrs attrs);
1815 
1816 /**
1817  * memory_region_iommu_num_indexes: return the total number of IOMMU
1818  * indexes that this IOMMU supports.
1819  *
1820  * @iommu_mr: the memory region
1821  */
1822 int memory_region_iommu_num_indexes(IOMMUMemoryRegion *iommu_mr);
1823 
1824 /**
1825  * memory_region_iommu_set_page_size_mask: set the supported page
1826  * sizes for a given IOMMU memory region
1827  *
1828  * @iommu_mr: IOMMU memory region
1829  * @page_size_mask: supported page size mask
1830  * @errp: pointer to Error*, to store an error if it happens.
1831  */
1832 int memory_region_iommu_set_page_size_mask(IOMMUMemoryRegion *iommu_mr,
1833                                            uint64_t page_size_mask,
1834                                            Error **errp);
1835 
1836 /**
1837  * memory_region_name: get a memory region's name
1838  *
1839  * Returns the string that was used to initialize the memory region.
1840  *
1841  * @mr: the memory region being queried
1842  */
1843 const char *memory_region_name(const MemoryRegion *mr);
1844 
1845 /**
1846  * memory_region_is_logging: return whether a memory region is logging writes
1847  *
1848  * Returns %true if the memory region is logging writes for the given client
1849  *
1850  * @mr: the memory region being queried
1851  * @client: the client being queried
1852  */
1853 bool memory_region_is_logging(MemoryRegion *mr, uint8_t client);
1854 
1855 /**
1856  * memory_region_get_dirty_log_mask: return the clients for which a
1857  * memory region is logging writes.
1858  *
1859  * Returns a bitmap of clients, in which the DIRTY_MEMORY_* constants
1860  * are the bit indices.
1861  *
1862  * @mr: the memory region being queried
1863  */
1864 uint8_t memory_region_get_dirty_log_mask(MemoryRegion *mr);
1865 
1866 /**
1867  * memory_region_is_rom: check whether a memory region is ROM
1868  *
1869  * Returns %true if a memory region is read-only memory.
1870  *
1871  * @mr: the memory region being queried
1872  */
1873 static inline bool memory_region_is_rom(MemoryRegion *mr)
1874 {
1875     return mr->ram && mr->readonly;
1876 }
1877 
1878 /**
1879  * memory_region_is_nonvolatile: check whether a memory region is non-volatile
1880  *
1881  * Returns %true is a memory region is non-volatile memory.
1882  *
1883  * @mr: the memory region being queried
1884  */
1885 static inline bool memory_region_is_nonvolatile(MemoryRegion *mr)
1886 {
1887     return mr->nonvolatile;
1888 }
1889 
1890 /**
1891  * memory_region_get_fd: Get a file descriptor backing a RAM memory region.
1892  *
1893  * Returns a file descriptor backing a file-based RAM memory region,
1894  * or -1 if the region is not a file-based RAM memory region.
1895  *
1896  * @mr: the RAM or alias memory region being queried.
1897  */
1898 int memory_region_get_fd(MemoryRegion *mr);
1899 
1900 /**
1901  * memory_region_from_host: Convert a pointer into a RAM memory region
1902  * and an offset within it.
1903  *
1904  * Given a host pointer inside a RAM memory region (created with
1905  * memory_region_init_ram() or memory_region_init_ram_ptr()), return
1906  * the MemoryRegion and the offset within it.
1907  *
1908  * Use with care; by the time this function returns, the returned pointer is
1909  * not protected by RCU anymore.  If the caller is not within an RCU critical
1910  * section and does not hold the iothread lock, it must have other means of
1911  * protecting the pointer, such as a reference to the region that includes
1912  * the incoming ram_addr_t.
1913  *
1914  * @ptr: the host pointer to be converted
1915  * @offset: the offset within memory region
1916  */
1917 MemoryRegion *memory_region_from_host(void *ptr, ram_addr_t *offset);
1918 
1919 /**
1920  * memory_region_get_ram_ptr: Get a pointer into a RAM memory region.
1921  *
1922  * Returns a host pointer to a RAM memory region (created with
1923  * memory_region_init_ram() or memory_region_init_ram_ptr()).
1924  *
1925  * Use with care; by the time this function returns, the returned pointer is
1926  * not protected by RCU anymore.  If the caller is not within an RCU critical
1927  * section and does not hold the iothread lock, it must have other means of
1928  * protecting the pointer, such as a reference to the region that includes
1929  * the incoming ram_addr_t.
1930  *
1931  * @mr: the memory region being queried.
1932  */
1933 void *memory_region_get_ram_ptr(MemoryRegion *mr);
1934 
1935 /* memory_region_ram_resize: Resize a RAM region.
1936  *
1937  * Resizing RAM while migrating can result in the migration being canceled.
1938  * Care has to be taken if the guest might have already detected the memory.
1939  *
1940  * @mr: a memory region created with @memory_region_init_resizeable_ram.
1941  * @newsize: the new size the region
1942  * @errp: pointer to Error*, to store an error if it happens.
1943  */
1944 void memory_region_ram_resize(MemoryRegion *mr, ram_addr_t newsize,
1945                               Error **errp);
1946 
1947 /**
1948  * memory_region_msync: Synchronize selected address range of
1949  * a memory mapped region
1950  *
1951  * @mr: the memory region to be msync
1952  * @addr: the initial address of the range to be sync
1953  * @size: the size of the range to be sync
1954  */
1955 void memory_region_msync(MemoryRegion *mr, hwaddr addr, hwaddr size);
1956 
1957 /**
1958  * memory_region_writeback: Trigger cache writeback for
1959  * selected address range
1960  *
1961  * @mr: the memory region to be updated
1962  * @addr: the initial address of the range to be written back
1963  * @size: the size of the range to be written back
1964  */
1965 void memory_region_writeback(MemoryRegion *mr, hwaddr addr, hwaddr size);
1966 
1967 /**
1968  * memory_region_set_log: Turn dirty logging on or off for a region.
1969  *
1970  * Turns dirty logging on or off for a specified client (display, migration).
1971  * Only meaningful for RAM regions.
1972  *
1973  * @mr: the memory region being updated.
1974  * @log: whether dirty logging is to be enabled or disabled.
1975  * @client: the user of the logging information; %DIRTY_MEMORY_VGA only.
1976  */
1977 void memory_region_set_log(MemoryRegion *mr, bool log, unsigned client);
1978 
1979 /**
1980  * memory_region_set_dirty: Mark a range of bytes as dirty in a memory region.
1981  *
1982  * Marks a range of bytes as dirty, after it has been dirtied outside
1983  * guest code.
1984  *
1985  * @mr: the memory region being dirtied.
1986  * @addr: the address (relative to the start of the region) being dirtied.
1987  * @size: size of the range being dirtied.
1988  */
1989 void memory_region_set_dirty(MemoryRegion *mr, hwaddr addr,
1990                              hwaddr size);
1991 
1992 /**
1993  * memory_region_clear_dirty_bitmap - clear dirty bitmap for memory range
1994  *
1995  * This function is called when the caller wants to clear the remote
1996  * dirty bitmap of a memory range within the memory region.  This can
1997  * be used by e.g. KVM to manually clear dirty log when
1998  * KVM_CAP_MANUAL_DIRTY_LOG_PROTECT is declared support by the host
1999  * kernel.
2000  *
2001  * @mr:     the memory region to clear the dirty log upon
2002  * @start:  start address offset within the memory region
2003  * @len:    length of the memory region to clear dirty bitmap
2004  */
2005 void memory_region_clear_dirty_bitmap(MemoryRegion *mr, hwaddr start,
2006                                       hwaddr len);
2007 
2008 /**
2009  * memory_region_snapshot_and_clear_dirty: Get a snapshot of the dirty
2010  *                                         bitmap and clear it.
2011  *
2012  * Creates a snapshot of the dirty bitmap, clears the dirty bitmap and
2013  * returns the snapshot.  The snapshot can then be used to query dirty
2014  * status, using memory_region_snapshot_get_dirty.  Snapshotting allows
2015  * querying the same page multiple times, which is especially useful for
2016  * display updates where the scanlines often are not page aligned.
2017  *
2018  * The dirty bitmap region which gets copied into the snapshot (and
2019  * cleared afterwards) can be larger than requested.  The boundaries
2020  * are rounded up/down so complete bitmap longs (covering 64 pages on
2021  * 64bit hosts) can be copied over into the bitmap snapshot.  Which
2022  * isn't a problem for display updates as the extra pages are outside
2023  * the visible area, and in case the visible area changes a full
2024  * display redraw is due anyway.  Should other use cases for this
2025  * function emerge we might have to revisit this implementation
2026  * detail.
2027  *
2028  * Use g_free to release DirtyBitmapSnapshot.
2029  *
2030  * @mr: the memory region being queried.
2031  * @addr: the address (relative to the start of the region) being queried.
2032  * @size: the size of the range being queried.
2033  * @client: the user of the logging information; typically %DIRTY_MEMORY_VGA.
2034  */
2035 DirtyBitmapSnapshot *memory_region_snapshot_and_clear_dirty(MemoryRegion *mr,
2036                                                             hwaddr addr,
2037                                                             hwaddr size,
2038                                                             unsigned client);
2039 
2040 /**
2041  * memory_region_snapshot_get_dirty: Check whether a range of bytes is dirty
2042  *                                   in the specified dirty bitmap snapshot.
2043  *
2044  * @mr: the memory region being queried.
2045  * @snap: the dirty bitmap snapshot
2046  * @addr: the address (relative to the start of the region) being queried.
2047  * @size: the size of the range being queried.
2048  */
2049 bool memory_region_snapshot_get_dirty(MemoryRegion *mr,
2050                                       DirtyBitmapSnapshot *snap,
2051                                       hwaddr addr, hwaddr size);
2052 
2053 /**
2054  * memory_region_reset_dirty: Mark a range of pages as clean, for a specified
2055  *                            client.
2056  *
2057  * Marks a range of pages as no longer dirty.
2058  *
2059  * @mr: the region being updated.
2060  * @addr: the start of the subrange being cleaned.
2061  * @size: the size of the subrange being cleaned.
2062  * @client: the user of the logging information; %DIRTY_MEMORY_MIGRATION or
2063  *          %DIRTY_MEMORY_VGA.
2064  */
2065 void memory_region_reset_dirty(MemoryRegion *mr, hwaddr addr,
2066                                hwaddr size, unsigned client);
2067 
2068 /**
2069  * memory_region_flush_rom_device: Mark a range of pages dirty and invalidate
2070  *                                 TBs (for self-modifying code).
2071  *
2072  * The MemoryRegionOps->write() callback of a ROM device must use this function
2073  * to mark byte ranges that have been modified internally, such as by directly
2074  * accessing the memory returned by memory_region_get_ram_ptr().
2075  *
2076  * This function marks the range dirty and invalidates TBs so that TCG can
2077  * detect self-modifying code.
2078  *
2079  * @mr: the region being flushed.
2080  * @addr: the start, relative to the start of the region, of the range being
2081  *        flushed.
2082  * @size: the size, in bytes, of the range being flushed.
2083  */
2084 void memory_region_flush_rom_device(MemoryRegion *mr, hwaddr addr, hwaddr size);
2085 
2086 /**
2087  * memory_region_set_readonly: Turn a memory region read-only (or read-write)
2088  *
2089  * Allows a memory region to be marked as read-only (turning it into a ROM).
2090  * only useful on RAM regions.
2091  *
2092  * @mr: the region being updated.
2093  * @readonly: whether rhe region is to be ROM or RAM.
2094  */
2095 void memory_region_set_readonly(MemoryRegion *mr, bool readonly);
2096 
2097 /**
2098  * memory_region_set_nonvolatile: Turn a memory region non-volatile
2099  *
2100  * Allows a memory region to be marked as non-volatile.
2101  * only useful on RAM regions.
2102  *
2103  * @mr: the region being updated.
2104  * @nonvolatile: whether rhe region is to be non-volatile.
2105  */
2106 void memory_region_set_nonvolatile(MemoryRegion *mr, bool nonvolatile);
2107 
2108 /**
2109  * memory_region_rom_device_set_romd: enable/disable ROMD mode
2110  *
2111  * Allows a ROM device (initialized with memory_region_init_rom_device() to
2112  * set to ROMD mode (default) or MMIO mode.  When it is in ROMD mode, the
2113  * device is mapped to guest memory and satisfies read access directly.
2114  * When in MMIO mode, reads are forwarded to the #MemoryRegion.read function.
2115  * Writes are always handled by the #MemoryRegion.write function.
2116  *
2117  * @mr: the memory region to be updated
2118  * @romd_mode: %true to put the region into ROMD mode
2119  */
2120 void memory_region_rom_device_set_romd(MemoryRegion *mr, bool romd_mode);
2121 
2122 /**
2123  * memory_region_set_coalescing: Enable memory coalescing for the region.
2124  *
2125  * Enabled writes to a region to be queued for later processing. MMIO ->write
2126  * callbacks may be delayed until a non-coalesced MMIO is issued.
2127  * Only useful for IO regions.  Roughly similar to write-combining hardware.
2128  *
2129  * @mr: the memory region to be write coalesced
2130  */
2131 void memory_region_set_coalescing(MemoryRegion *mr);
2132 
2133 /**
2134  * memory_region_add_coalescing: Enable memory coalescing for a sub-range of
2135  *                               a region.
2136  *
2137  * Like memory_region_set_coalescing(), but works on a sub-range of a region.
2138  * Multiple calls can be issued coalesced disjoint ranges.
2139  *
2140  * @mr: the memory region to be updated.
2141  * @offset: the start of the range within the region to be coalesced.
2142  * @size: the size of the subrange to be coalesced.
2143  */
2144 void memory_region_add_coalescing(MemoryRegion *mr,
2145                                   hwaddr offset,
2146                                   uint64_t size);
2147 
2148 /**
2149  * memory_region_clear_coalescing: Disable MMIO coalescing for the region.
2150  *
2151  * Disables any coalescing caused by memory_region_set_coalescing() or
2152  * memory_region_add_coalescing().  Roughly equivalent to uncacheble memory
2153  * hardware.
2154  *
2155  * @mr: the memory region to be updated.
2156  */
2157 void memory_region_clear_coalescing(MemoryRegion *mr);
2158 
2159 /**
2160  * memory_region_set_flush_coalesced: Enforce memory coalescing flush before
2161  *                                    accesses.
2162  *
2163  * Ensure that pending coalesced MMIO request are flushed before the memory
2164  * region is accessed. This property is automatically enabled for all regions
2165  * passed to memory_region_set_coalescing() and memory_region_add_coalescing().
2166  *
2167  * @mr: the memory region to be updated.
2168  */
2169 void memory_region_set_flush_coalesced(MemoryRegion *mr);
2170 
2171 /**
2172  * memory_region_clear_flush_coalesced: Disable memory coalescing flush before
2173  *                                      accesses.
2174  *
2175  * Clear the automatic coalesced MMIO flushing enabled via
2176  * memory_region_set_flush_coalesced. Note that this service has no effect on
2177  * memory regions that have MMIO coalescing enabled for themselves. For them,
2178  * automatic flushing will stop once coalescing is disabled.
2179  *
2180  * @mr: the memory region to be updated.
2181  */
2182 void memory_region_clear_flush_coalesced(MemoryRegion *mr);
2183 
2184 /**
2185  * memory_region_add_eventfd: Request an eventfd to be triggered when a word
2186  *                            is written to a location.
2187  *
2188  * Marks a word in an IO region (initialized with memory_region_init_io())
2189  * as a trigger for an eventfd event.  The I/O callback will not be called.
2190  * The caller must be prepared to handle failure (that is, take the required
2191  * action if the callback _is_ called).
2192  *
2193  * @mr: the memory region being updated.
2194  * @addr: the address within @mr that is to be monitored
2195  * @size: the size of the access to trigger the eventfd
2196  * @match_data: whether to match against @data, instead of just @addr
2197  * @data: the data to match against the guest write
2198  * @e: event notifier to be triggered when @addr, @size, and @data all match.
2199  **/
2200 void memory_region_add_eventfd(MemoryRegion *mr,
2201                                hwaddr addr,
2202                                unsigned size,
2203                                bool match_data,
2204                                uint64_t data,
2205                                EventNotifier *e);
2206 
2207 /**
2208  * memory_region_del_eventfd: Cancel an eventfd.
2209  *
2210  * Cancels an eventfd trigger requested by a previous
2211  * memory_region_add_eventfd() call.
2212  *
2213  * @mr: the memory region being updated.
2214  * @addr: the address within @mr that is to be monitored
2215  * @size: the size of the access to trigger the eventfd
2216  * @match_data: whether to match against @data, instead of just @addr
2217  * @data: the data to match against the guest write
2218  * @e: event notifier to be triggered when @addr, @size, and @data all match.
2219  */
2220 void memory_region_del_eventfd(MemoryRegion *mr,
2221                                hwaddr addr,
2222                                unsigned size,
2223                                bool match_data,
2224                                uint64_t data,
2225                                EventNotifier *e);
2226 
2227 /**
2228  * memory_region_add_subregion: Add a subregion to a container.
2229  *
2230  * Adds a subregion at @offset.  The subregion may not overlap with other
2231  * subregions (except for those explicitly marked as overlapping).  A region
2232  * may only be added once as a subregion (unless removed with
2233  * memory_region_del_subregion()); use memory_region_init_alias() if you
2234  * want a region to be a subregion in multiple locations.
2235  *
2236  * @mr: the region to contain the new subregion; must be a container
2237  *      initialized with memory_region_init().
2238  * @offset: the offset relative to @mr where @subregion is added.
2239  * @subregion: the subregion to be added.
2240  */
2241 void memory_region_add_subregion(MemoryRegion *mr,
2242                                  hwaddr offset,
2243                                  MemoryRegion *subregion);
2244 /**
2245  * memory_region_add_subregion_overlap: Add a subregion to a container
2246  *                                      with overlap.
2247  *
2248  * Adds a subregion at @offset.  The subregion may overlap with other
2249  * subregions.  Conflicts are resolved by having a higher @priority hide a
2250  * lower @priority. Subregions without priority are taken as @priority 0.
2251  * A region may only be added once as a subregion (unless removed with
2252  * memory_region_del_subregion()); use memory_region_init_alias() if you
2253  * want a region to be a subregion in multiple locations.
2254  *
2255  * @mr: the region to contain the new subregion; must be a container
2256  *      initialized with memory_region_init().
2257  * @offset: the offset relative to @mr where @subregion is added.
2258  * @subregion: the subregion to be added.
2259  * @priority: used for resolving overlaps; highest priority wins.
2260  */
2261 void memory_region_add_subregion_overlap(MemoryRegion *mr,
2262                                          hwaddr offset,
2263                                          MemoryRegion *subregion,
2264                                          int priority);
2265 
2266 /**
2267  * memory_region_get_ram_addr: Get the ram address associated with a memory
2268  *                             region
2269  *
2270  * @mr: the region to be queried
2271  */
2272 ram_addr_t memory_region_get_ram_addr(MemoryRegion *mr);
2273 
2274 uint64_t memory_region_get_alignment(const MemoryRegion *mr);
2275 /**
2276  * memory_region_del_subregion: Remove a subregion.
2277  *
2278  * Removes a subregion from its container.
2279  *
2280  * @mr: the container to be updated.
2281  * @subregion: the region being removed; must be a current subregion of @mr.
2282  */
2283 void memory_region_del_subregion(MemoryRegion *mr,
2284                                  MemoryRegion *subregion);
2285 
2286 /*
2287  * memory_region_set_enabled: dynamically enable or disable a region
2288  *
2289  * Enables or disables a memory region.  A disabled memory region
2290  * ignores all accesses to itself and its subregions.  It does not
2291  * obscure sibling subregions with lower priority - it simply behaves as
2292  * if it was removed from the hierarchy.
2293  *
2294  * Regions default to being enabled.
2295  *
2296  * @mr: the region to be updated
2297  * @enabled: whether to enable or disable the region
2298  */
2299 void memory_region_set_enabled(MemoryRegion *mr, bool enabled);
2300 
2301 /*
2302  * memory_region_set_address: dynamically update the address of a region
2303  *
2304  * Dynamically updates the address of a region, relative to its container.
2305  * May be used on regions are currently part of a memory hierarchy.
2306  *
2307  * @mr: the region to be updated
2308  * @addr: new address, relative to container region
2309  */
2310 void memory_region_set_address(MemoryRegion *mr, hwaddr addr);
2311 
2312 /*
2313  * memory_region_set_size: dynamically update the size of a region.
2314  *
2315  * Dynamically updates the size of a region.
2316  *
2317  * @mr: the region to be updated
2318  * @size: used size of the region.
2319  */
2320 void memory_region_set_size(MemoryRegion *mr, uint64_t size);
2321 
2322 /*
2323  * memory_region_set_alias_offset: dynamically update a memory alias's offset
2324  *
2325  * Dynamically updates the offset into the target region that an alias points
2326  * to, as if the fourth argument to memory_region_init_alias() has changed.
2327  *
2328  * @mr: the #MemoryRegion to be updated; should be an alias.
2329  * @offset: the new offset into the target memory region
2330  */
2331 void memory_region_set_alias_offset(MemoryRegion *mr,
2332                                     hwaddr offset);
2333 
2334 /**
2335  * memory_region_present: checks if an address relative to a @container
2336  * translates into #MemoryRegion within @container
2337  *
2338  * Answer whether a #MemoryRegion within @container covers the address
2339  * @addr.
2340  *
2341  * @container: a #MemoryRegion within which @addr is a relative address
2342  * @addr: the area within @container to be searched
2343  */
2344 bool memory_region_present(MemoryRegion *container, hwaddr addr);
2345 
2346 /**
2347  * memory_region_is_mapped: returns true if #MemoryRegion is mapped
2348  * into another memory region, which does not necessarily imply that it is
2349  * mapped into an address space.
2350  *
2351  * @mr: a #MemoryRegion which should be checked if it's mapped
2352  */
2353 bool memory_region_is_mapped(MemoryRegion *mr);
2354 
2355 /**
2356  * memory_region_get_ram_discard_manager: get the #RamDiscardManager for a
2357  * #MemoryRegion
2358  *
2359  * The #RamDiscardManager cannot change while a memory region is mapped.
2360  *
2361  * @mr: the #MemoryRegion
2362  */
2363 RamDiscardManager *memory_region_get_ram_discard_manager(MemoryRegion *mr);
2364 
2365 /**
2366  * memory_region_has_ram_discard_manager: check whether a #MemoryRegion has a
2367  * #RamDiscardManager assigned
2368  *
2369  * @mr: the #MemoryRegion
2370  */
2371 static inline bool memory_region_has_ram_discard_manager(MemoryRegion *mr)
2372 {
2373     return !!memory_region_get_ram_discard_manager(mr);
2374 }
2375 
2376 /**
2377  * memory_region_set_ram_discard_manager: set the #RamDiscardManager for a
2378  * #MemoryRegion
2379  *
2380  * This function must not be called for a mapped #MemoryRegion, a #MemoryRegion
2381  * that does not cover RAM, or a #MemoryRegion that already has a
2382  * #RamDiscardManager assigned.
2383  *
2384  * @mr: the #MemoryRegion
2385  * @rdm: #RamDiscardManager to set
2386  */
2387 void memory_region_set_ram_discard_manager(MemoryRegion *mr,
2388                                            RamDiscardManager *rdm);
2389 
2390 /**
2391  * memory_region_find: translate an address/size relative to a
2392  * MemoryRegion into a #MemoryRegionSection.
2393  *
2394  * Locates the first #MemoryRegion within @mr that overlaps the range
2395  * given by @addr and @size.
2396  *
2397  * Returns a #MemoryRegionSection that describes a contiguous overlap.
2398  * It will have the following characteristics:
2399  * - @size = 0 iff no overlap was found
2400  * - @mr is non-%NULL iff an overlap was found
2401  *
2402  * Remember that in the return value the @offset_within_region is
2403  * relative to the returned region (in the .@mr field), not to the
2404  * @mr argument.
2405  *
2406  * Similarly, the .@offset_within_address_space is relative to the
2407  * address space that contains both regions, the passed and the
2408  * returned one.  However, in the special case where the @mr argument
2409  * has no container (and thus is the root of the address space), the
2410  * following will hold:
2411  * - @offset_within_address_space >= @addr
2412  * - @offset_within_address_space + .@size <= @addr + @size
2413  *
2414  * @mr: a MemoryRegion within which @addr is a relative address
2415  * @addr: start of the area within @as to be searched
2416  * @size: size of the area to be searched
2417  */
2418 MemoryRegionSection memory_region_find(MemoryRegion *mr,
2419                                        hwaddr addr, uint64_t size);
2420 
2421 /**
2422  * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2423  *
2424  * Synchronizes the dirty page log for all address spaces.
2425  */
2426 void memory_global_dirty_log_sync(void);
2427 
2428 /**
2429  * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2430  *
2431  * Synchronizes the vCPUs with a thread that is reading the dirty bitmap.
2432  * This function must be called after the dirty log bitmap is cleared, and
2433  * before dirty guest memory pages are read.  If you are using
2434  * #DirtyBitmapSnapshot, memory_region_snapshot_and_clear_dirty() takes
2435  * care of doing this.
2436  */
2437 void memory_global_after_dirty_log_sync(void);
2438 
2439 /**
2440  * memory_region_transaction_begin: Start a transaction.
2441  *
2442  * During a transaction, changes will be accumulated and made visible
2443  * only when the transaction ends (is committed).
2444  */
2445 void memory_region_transaction_begin(void);
2446 
2447 /**
2448  * memory_region_transaction_commit: Commit a transaction and make changes
2449  *                                   visible to the guest.
2450  */
2451 void memory_region_transaction_commit(void);
2452 
2453 /**
2454  * memory_listener_register: register callbacks to be called when memory
2455  *                           sections are mapped or unmapped into an address
2456  *                           space
2457  *
2458  * @listener: an object containing the callbacks to be called
2459  * @filter: if non-%NULL, only regions in this address space will be observed
2460  */
2461 void memory_listener_register(MemoryListener *listener, AddressSpace *filter);
2462 
2463 /**
2464  * memory_listener_unregister: undo the effect of memory_listener_register()
2465  *
2466  * @listener: an object containing the callbacks to be removed
2467  */
2468 void memory_listener_unregister(MemoryListener *listener);
2469 
2470 /**
2471  * memory_global_dirty_log_start: begin dirty logging for all regions
2472  *
2473  * @flags: purpose of starting dirty log, migration or dirty rate
2474  */
2475 void memory_global_dirty_log_start(unsigned int flags);
2476 
2477 /**
2478  * memory_global_dirty_log_stop: end dirty logging for all regions
2479  *
2480  * @flags: purpose of stopping dirty log, migration or dirty rate
2481  */
2482 void memory_global_dirty_log_stop(unsigned int flags);
2483 
2484 void mtree_info(bool flatview, bool dispatch_tree, bool owner, bool disabled);
2485 
2486 bool memory_region_access_valid(MemoryRegion *mr, hwaddr addr,
2487                                 unsigned size, bool is_write,
2488                                 MemTxAttrs attrs);
2489 
2490 /**
2491  * memory_region_dispatch_read: perform a read directly to the specified
2492  * MemoryRegion.
2493  *
2494  * @mr: #MemoryRegion to access
2495  * @addr: address within that region
2496  * @pval: pointer to uint64_t which the data is written to
2497  * @op: size, sign, and endianness of the memory operation
2498  * @attrs: memory transaction attributes to use for the access
2499  */
2500 MemTxResult memory_region_dispatch_read(MemoryRegion *mr,
2501                                         hwaddr addr,
2502                                         uint64_t *pval,
2503                                         MemOp op,
2504                                         MemTxAttrs attrs);
2505 /**
2506  * memory_region_dispatch_write: perform a write directly to the specified
2507  * MemoryRegion.
2508  *
2509  * @mr: #MemoryRegion to access
2510  * @addr: address within that region
2511  * @data: data to write
2512  * @op: size, sign, and endianness of the memory operation
2513  * @attrs: memory transaction attributes to use for the access
2514  */
2515 MemTxResult memory_region_dispatch_write(MemoryRegion *mr,
2516                                          hwaddr addr,
2517                                          uint64_t data,
2518                                          MemOp op,
2519                                          MemTxAttrs attrs);
2520 
2521 /**
2522  * address_space_init: initializes an address space
2523  *
2524  * @as: an uninitialized #AddressSpace
2525  * @root: a #MemoryRegion that routes addresses for the address space
2526  * @name: an address space name.  The name is only used for debugging
2527  *        output.
2528  */
2529 void address_space_init(AddressSpace *as, MemoryRegion *root, const char *name);
2530 
2531 /**
2532  * address_space_destroy: destroy an address space
2533  *
2534  * Releases all resources associated with an address space.  After an address space
2535  * is destroyed, its root memory region (given by address_space_init()) may be destroyed
2536  * as well.
2537  *
2538  * @as: address space to be destroyed
2539  */
2540 void address_space_destroy(AddressSpace *as);
2541 
2542 /**
2543  * address_space_remove_listeners: unregister all listeners of an address space
2544  *
2545  * Removes all callbacks previously registered with memory_listener_register()
2546  * for @as.
2547  *
2548  * @as: an initialized #AddressSpace
2549  */
2550 void address_space_remove_listeners(AddressSpace *as);
2551 
2552 /**
2553  * address_space_rw: read from or write to an address space.
2554  *
2555  * Return a MemTxResult indicating whether the operation succeeded
2556  * or failed (eg unassigned memory, device rejected the transaction,
2557  * IOMMU fault).
2558  *
2559  * @as: #AddressSpace to be accessed
2560  * @addr: address within that address space
2561  * @attrs: memory transaction attributes
2562  * @buf: buffer with the data transferred
2563  * @len: the number of bytes to read or write
2564  * @is_write: indicates the transfer direction
2565  */
2566 MemTxResult address_space_rw(AddressSpace *as, hwaddr addr,
2567                              MemTxAttrs attrs, void *buf,
2568                              hwaddr len, bool is_write);
2569 
2570 /**
2571  * address_space_write: write to address space.
2572  *
2573  * Return a MemTxResult indicating whether the operation succeeded
2574  * or failed (eg unassigned memory, device rejected the transaction,
2575  * IOMMU fault).
2576  *
2577  * @as: #AddressSpace to be accessed
2578  * @addr: address within that address space
2579  * @attrs: memory transaction attributes
2580  * @buf: buffer with the data transferred
2581  * @len: the number of bytes to write
2582  */
2583 MemTxResult address_space_write(AddressSpace *as, hwaddr addr,
2584                                 MemTxAttrs attrs,
2585                                 const void *buf, hwaddr len);
2586 
2587 /**
2588  * address_space_write_rom: write to address space, including ROM.
2589  *
2590  * This function writes to the specified address space, but will
2591  * write data to both ROM and RAM. This is used for non-guest
2592  * writes like writes from the gdb debug stub or initial loading
2593  * of ROM contents.
2594  *
2595  * Note that portions of the write which attempt to write data to
2596  * a device will be silently ignored -- only real RAM and ROM will
2597  * be written to.
2598  *
2599  * Return a MemTxResult indicating whether the operation succeeded
2600  * or failed (eg unassigned memory, device rejected the transaction,
2601  * IOMMU fault).
2602  *
2603  * @as: #AddressSpace to be accessed
2604  * @addr: address within that address space
2605  * @attrs: memory transaction attributes
2606  * @buf: buffer with the data transferred
2607  * @len: the number of bytes to write
2608  */
2609 MemTxResult address_space_write_rom(AddressSpace *as, hwaddr addr,
2610                                     MemTxAttrs attrs,
2611                                     const void *buf, hwaddr len);
2612 
2613 /* address_space_ld*: load from an address space
2614  * address_space_st*: store to an address space
2615  *
2616  * These functions perform a load or store of the byte, word,
2617  * longword or quad to the specified address within the AddressSpace.
2618  * The _le suffixed functions treat the data as little endian;
2619  * _be indicates big endian; no suffix indicates "same endianness
2620  * as guest CPU".
2621  *
2622  * The "guest CPU endianness" accessors are deprecated for use outside
2623  * target-* code; devices should be CPU-agnostic and use either the LE
2624  * or the BE accessors.
2625  *
2626  * @as #AddressSpace to be accessed
2627  * @addr: address within that address space
2628  * @val: data value, for stores
2629  * @attrs: memory transaction attributes
2630  * @result: location to write the success/failure of the transaction;
2631  *   if NULL, this information is discarded
2632  */
2633 
2634 #define SUFFIX
2635 #define ARG1         as
2636 #define ARG1_DECL    AddressSpace *as
2637 #include "exec/memory_ldst.h.inc"
2638 
2639 #define SUFFIX
2640 #define ARG1         as
2641 #define ARG1_DECL    AddressSpace *as
2642 #include "exec/memory_ldst_phys.h.inc"
2643 
2644 struct MemoryRegionCache {
2645     void *ptr;
2646     hwaddr xlat;
2647     hwaddr len;
2648     FlatView *fv;
2649     MemoryRegionSection mrs;
2650     bool is_write;
2651 };
2652 
2653 #define MEMORY_REGION_CACHE_INVALID ((MemoryRegionCache) { .mrs.mr = NULL })
2654 
2655 
2656 /* address_space_ld*_cached: load from a cached #MemoryRegion
2657  * address_space_st*_cached: store into a cached #MemoryRegion
2658  *
2659  * These functions perform a load or store of the byte, word,
2660  * longword or quad to the specified address.  The address is
2661  * a physical address in the AddressSpace, but it must lie within
2662  * a #MemoryRegion that was mapped with address_space_cache_init.
2663  *
2664  * The _le suffixed functions treat the data as little endian;
2665  * _be indicates big endian; no suffix indicates "same endianness
2666  * as guest CPU".
2667  *
2668  * The "guest CPU endianness" accessors are deprecated for use outside
2669  * target-* code; devices should be CPU-agnostic and use either the LE
2670  * or the BE accessors.
2671  *
2672  * @cache: previously initialized #MemoryRegionCache to be accessed
2673  * @addr: address within the address space
2674  * @val: data value, for stores
2675  * @attrs: memory transaction attributes
2676  * @result: location to write the success/failure of the transaction;
2677  *   if NULL, this information is discarded
2678  */
2679 
2680 #define SUFFIX       _cached_slow
2681 #define ARG1         cache
2682 #define ARG1_DECL    MemoryRegionCache *cache
2683 #include "exec/memory_ldst.h.inc"
2684 
2685 /* Inline fast path for direct RAM access.  */
2686 static inline uint8_t address_space_ldub_cached(MemoryRegionCache *cache,
2687     hwaddr addr, MemTxAttrs attrs, MemTxResult *result)
2688 {
2689     assert(addr < cache->len);
2690     if (likely(cache->ptr)) {
2691         return ldub_p(cache->ptr + addr);
2692     } else {
2693         return address_space_ldub_cached_slow(cache, addr, attrs, result);
2694     }
2695 }
2696 
2697 static inline void address_space_stb_cached(MemoryRegionCache *cache,
2698     hwaddr addr, uint8_t val, MemTxAttrs attrs, MemTxResult *result)
2699 {
2700     assert(addr < cache->len);
2701     if (likely(cache->ptr)) {
2702         stb_p(cache->ptr + addr, val);
2703     } else {
2704         address_space_stb_cached_slow(cache, addr, val, attrs, result);
2705     }
2706 }
2707 
2708 #define ENDIANNESS   _le
2709 #include "exec/memory_ldst_cached.h.inc"
2710 
2711 #define ENDIANNESS   _be
2712 #include "exec/memory_ldst_cached.h.inc"
2713 
2714 #define SUFFIX       _cached
2715 #define ARG1         cache
2716 #define ARG1_DECL    MemoryRegionCache *cache
2717 #include "exec/memory_ldst_phys.h.inc"
2718 
2719 /* address_space_cache_init: prepare for repeated access to a physical
2720  * memory region
2721  *
2722  * @cache: #MemoryRegionCache to be filled
2723  * @as: #AddressSpace to be accessed
2724  * @addr: address within that address space
2725  * @len: length of buffer
2726  * @is_write: indicates the transfer direction
2727  *
2728  * Will only work with RAM, and may map a subset of the requested range by
2729  * returning a value that is less than @len.  On failure, return a negative
2730  * errno value.
2731  *
2732  * Because it only works with RAM, this function can be used for
2733  * read-modify-write operations.  In this case, is_write should be %true.
2734  *
2735  * Note that addresses passed to the address_space_*_cached functions
2736  * are relative to @addr.
2737  */
2738 int64_t address_space_cache_init(MemoryRegionCache *cache,
2739                                  AddressSpace *as,
2740                                  hwaddr addr,
2741                                  hwaddr len,
2742                                  bool is_write);
2743 
2744 /**
2745  * address_space_cache_invalidate: complete a write to a #MemoryRegionCache
2746  *
2747  * @cache: The #MemoryRegionCache to operate on.
2748  * @addr: The first physical address that was written, relative to the
2749  * address that was passed to @address_space_cache_init.
2750  * @access_len: The number of bytes that were written starting at @addr.
2751  */
2752 void address_space_cache_invalidate(MemoryRegionCache *cache,
2753                                     hwaddr addr,
2754                                     hwaddr access_len);
2755 
2756 /**
2757  * address_space_cache_destroy: free a #MemoryRegionCache
2758  *
2759  * @cache: The #MemoryRegionCache whose memory should be released.
2760  */
2761 void address_space_cache_destroy(MemoryRegionCache *cache);
2762 
2763 /* address_space_get_iotlb_entry: translate an address into an IOTLB
2764  * entry. Should be called from an RCU critical section.
2765  */
2766 IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr,
2767                                             bool is_write, MemTxAttrs attrs);
2768 
2769 /* address_space_translate: translate an address range into an address space
2770  * into a MemoryRegion and an address range into that section.  Should be
2771  * called from an RCU critical section, to avoid that the last reference
2772  * to the returned region disappears after address_space_translate returns.
2773  *
2774  * @fv: #FlatView to be accessed
2775  * @addr: address within that address space
2776  * @xlat: pointer to address within the returned memory region section's
2777  * #MemoryRegion.
2778  * @len: pointer to length
2779  * @is_write: indicates the transfer direction
2780  * @attrs: memory attributes
2781  */
2782 MemoryRegion *flatview_translate(FlatView *fv,
2783                                  hwaddr addr, hwaddr *xlat,
2784                                  hwaddr *len, bool is_write,
2785                                  MemTxAttrs attrs);
2786 
2787 static inline MemoryRegion *address_space_translate(AddressSpace *as,
2788                                                     hwaddr addr, hwaddr *xlat,
2789                                                     hwaddr *len, bool is_write,
2790                                                     MemTxAttrs attrs)
2791 {
2792     return flatview_translate(address_space_to_flatview(as),
2793                               addr, xlat, len, is_write, attrs);
2794 }
2795 
2796 /* address_space_access_valid: check for validity of accessing an address
2797  * space range
2798  *
2799  * Check whether memory is assigned to the given address space range, and
2800  * access is permitted by any IOMMU regions that are active for the address
2801  * space.
2802  *
2803  * For now, addr and len should be aligned to a page size.  This limitation
2804  * will be lifted in the future.
2805  *
2806  * @as: #AddressSpace to be accessed
2807  * @addr: address within that address space
2808  * @len: length of the area to be checked
2809  * @is_write: indicates the transfer direction
2810  * @attrs: memory attributes
2811  */
2812 bool address_space_access_valid(AddressSpace *as, hwaddr addr, hwaddr len,
2813                                 bool is_write, MemTxAttrs attrs);
2814 
2815 /* address_space_map: map a physical memory region into a host virtual address
2816  *
2817  * May map a subset of the requested range, given by and returned in @plen.
2818  * May return %NULL and set *@plen to zero(0), if resources needed to perform
2819  * the mapping are exhausted.
2820  * Use only for reads OR writes - not for read-modify-write operations.
2821  * Use cpu_register_map_client() to know when retrying the map operation is
2822  * likely to succeed.
2823  *
2824  * @as: #AddressSpace to be accessed
2825  * @addr: address within that address space
2826  * @plen: pointer to length of buffer; updated on return
2827  * @is_write: indicates the transfer direction
2828  * @attrs: memory attributes
2829  */
2830 void *address_space_map(AddressSpace *as, hwaddr addr,
2831                         hwaddr *plen, bool is_write, MemTxAttrs attrs);
2832 
2833 /* address_space_unmap: Unmaps a memory region previously mapped by address_space_map()
2834  *
2835  * Will also mark the memory as dirty if @is_write == %true.  @access_len gives
2836  * the amount of memory that was actually read or written by the caller.
2837  *
2838  * @as: #AddressSpace used
2839  * @buffer: host pointer as returned by address_space_map()
2840  * @len: buffer length as returned by address_space_map()
2841  * @access_len: amount of data actually transferred
2842  * @is_write: indicates the transfer direction
2843  */
2844 void address_space_unmap(AddressSpace *as, void *buffer, hwaddr len,
2845                          bool is_write, hwaddr access_len);
2846 
2847 
2848 /* Internal functions, part of the implementation of address_space_read.  */
2849 MemTxResult address_space_read_full(AddressSpace *as, hwaddr addr,
2850                                     MemTxAttrs attrs, void *buf, hwaddr len);
2851 MemTxResult flatview_read_continue(FlatView *fv, hwaddr addr,
2852                                    MemTxAttrs attrs, void *buf,
2853                                    hwaddr len, hwaddr addr1, hwaddr l,
2854                                    MemoryRegion *mr);
2855 void *qemu_map_ram_ptr(RAMBlock *ram_block, ram_addr_t addr);
2856 
2857 /* Internal functions, part of the implementation of address_space_read_cached
2858  * and address_space_write_cached.  */
2859 MemTxResult address_space_read_cached_slow(MemoryRegionCache *cache,
2860                                            hwaddr addr, void *buf, hwaddr len);
2861 MemTxResult address_space_write_cached_slow(MemoryRegionCache *cache,
2862                                             hwaddr addr, const void *buf,
2863                                             hwaddr len);
2864 
2865 int memory_access_size(MemoryRegion *mr, unsigned l, hwaddr addr);
2866 bool prepare_mmio_access(MemoryRegion *mr);
2867 
2868 static inline bool memory_access_is_direct(MemoryRegion *mr, bool is_write)
2869 {
2870     if (is_write) {
2871         return memory_region_is_ram(mr) && !mr->readonly &&
2872                !mr->rom_device && !memory_region_is_ram_device(mr);
2873     } else {
2874         return (memory_region_is_ram(mr) && !memory_region_is_ram_device(mr)) ||
2875                memory_region_is_romd(mr);
2876     }
2877 }
2878 
2879 /**
2880  * address_space_read: read from an address space.
2881  *
2882  * Return a MemTxResult indicating whether the operation succeeded
2883  * or failed (eg unassigned memory, device rejected the transaction,
2884  * IOMMU fault).  Called within RCU critical section.
2885  *
2886  * @as: #AddressSpace to be accessed
2887  * @addr: address within that address space
2888  * @attrs: memory transaction attributes
2889  * @buf: buffer with the data transferred
2890  * @len: length of the data transferred
2891  */
2892 static inline __attribute__((__always_inline__))
2893 MemTxResult address_space_read(AddressSpace *as, hwaddr addr,
2894                                MemTxAttrs attrs, void *buf,
2895                                hwaddr len)
2896 {
2897     MemTxResult result = MEMTX_OK;
2898     hwaddr l, addr1;
2899     void *ptr;
2900     MemoryRegion *mr;
2901     FlatView *fv;
2902 
2903     if (__builtin_constant_p(len)) {
2904         if (len) {
2905             RCU_READ_LOCK_GUARD();
2906             fv = address_space_to_flatview(as);
2907             l = len;
2908             mr = flatview_translate(fv, addr, &addr1, &l, false, attrs);
2909             if (len == l && memory_access_is_direct(mr, false)) {
2910                 ptr = qemu_map_ram_ptr(mr->ram_block, addr1);
2911                 memcpy(buf, ptr, len);
2912             } else {
2913                 result = flatview_read_continue(fv, addr, attrs, buf, len,
2914                                                 addr1, l, mr);
2915             }
2916         }
2917     } else {
2918         result = address_space_read_full(as, addr, attrs, buf, len);
2919     }
2920     return result;
2921 }
2922 
2923 /**
2924  * address_space_read_cached: read from a cached RAM region
2925  *
2926  * @cache: Cached region to be addressed
2927  * @addr: address relative to the base of the RAM region
2928  * @buf: buffer with the data transferred
2929  * @len: length of the data transferred
2930  */
2931 static inline MemTxResult
2932 address_space_read_cached(MemoryRegionCache *cache, hwaddr addr,
2933                           void *buf, hwaddr len)
2934 {
2935     assert(addr < cache->len && len <= cache->len - addr);
2936     fuzz_dma_read_cb(cache->xlat + addr, len, cache->mrs.mr);
2937     if (likely(cache->ptr)) {
2938         memcpy(buf, cache->ptr + addr, len);
2939         return MEMTX_OK;
2940     } else {
2941         return address_space_read_cached_slow(cache, addr, buf, len);
2942     }
2943 }
2944 
2945 /**
2946  * address_space_write_cached: write to a cached RAM region
2947  *
2948  * @cache: Cached region to be addressed
2949  * @addr: address relative to the base of the RAM region
2950  * @buf: buffer with the data transferred
2951  * @len: length of the data transferred
2952  */
2953 static inline MemTxResult
2954 address_space_write_cached(MemoryRegionCache *cache, hwaddr addr,
2955                            const void *buf, hwaddr len)
2956 {
2957     assert(addr < cache->len && len <= cache->len - addr);
2958     if (likely(cache->ptr)) {
2959         memcpy(cache->ptr + addr, buf, len);
2960         return MEMTX_OK;
2961     } else {
2962         return address_space_write_cached_slow(cache, addr, buf, len);
2963     }
2964 }
2965 
2966 /**
2967  * address_space_set: Fill address space with a constant byte.
2968  *
2969  * Return a MemTxResult indicating whether the operation succeeded
2970  * or failed (eg unassigned memory, device rejected the transaction,
2971  * IOMMU fault).
2972  *
2973  * @as: #AddressSpace to be accessed
2974  * @addr: address within that address space
2975  * @c: constant byte to fill the memory
2976  * @len: the number of bytes to fill with the constant byte
2977  * @attrs: memory transaction attributes
2978  */
2979 MemTxResult address_space_set(AddressSpace *as, hwaddr addr,
2980                               uint8_t c, hwaddr len, MemTxAttrs attrs);
2981 
2982 #ifdef NEED_CPU_H
2983 /* enum device_endian to MemOp.  */
2984 static inline MemOp devend_memop(enum device_endian end)
2985 {
2986     QEMU_BUILD_BUG_ON(DEVICE_HOST_ENDIAN != DEVICE_LITTLE_ENDIAN &&
2987                       DEVICE_HOST_ENDIAN != DEVICE_BIG_ENDIAN);
2988 
2989 #if HOST_BIG_ENDIAN != TARGET_BIG_ENDIAN
2990     /* Swap if non-host endianness or native (target) endianness */
2991     return (end == DEVICE_HOST_ENDIAN) ? 0 : MO_BSWAP;
2992 #else
2993     const int non_host_endianness =
2994         DEVICE_LITTLE_ENDIAN ^ DEVICE_BIG_ENDIAN ^ DEVICE_HOST_ENDIAN;
2995 
2996     /* In this case, native (target) endianness needs no swap.  */
2997     return (end == non_host_endianness) ? MO_BSWAP : 0;
2998 #endif
2999 }
3000 #endif
3001 
3002 /*
3003  * Inhibit technologies that require discarding of pages in RAM blocks, e.g.,
3004  * to manage the actual amount of memory consumed by the VM (then, the memory
3005  * provided by RAM blocks might be bigger than the desired memory consumption).
3006  * This *must* be set if:
3007  * - Discarding parts of a RAM blocks does not result in the change being
3008  *   reflected in the VM and the pages getting freed.
3009  * - All memory in RAM blocks is pinned or duplicated, invaldiating any previous
3010  *   discards blindly.
3011  * - Discarding parts of a RAM blocks will result in integrity issues (e.g.,
3012  *   encrypted VMs).
3013  * Technologies that only temporarily pin the current working set of a
3014  * driver are fine, because we don't expect such pages to be discarded
3015  * (esp. based on guest action like balloon inflation).
3016  *
3017  * This is *not* to be used to protect from concurrent discards (esp.,
3018  * postcopy).
3019  *
3020  * Returns 0 if successful. Returns -EBUSY if a technology that relies on
3021  * discards to work reliably is active.
3022  */
3023 int ram_block_discard_disable(bool state);
3024 
3025 /*
3026  * See ram_block_discard_disable(): only disable uncoordinated discards,
3027  * keeping coordinated discards (via the RamDiscardManager) enabled.
3028  */
3029 int ram_block_uncoordinated_discard_disable(bool state);
3030 
3031 /*
3032  * Inhibit technologies that disable discarding of pages in RAM blocks.
3033  *
3034  * Returns 0 if successful. Returns -EBUSY if discards are already set to
3035  * broken.
3036  */
3037 int ram_block_discard_require(bool state);
3038 
3039 /*
3040  * See ram_block_discard_require(): only inhibit technologies that disable
3041  * uncoordinated discarding of pages in RAM blocks, allowing co-existance with
3042  * technologies that only inhibit uncoordinated discards (via the
3043  * RamDiscardManager).
3044  */
3045 int ram_block_coordinated_discard_require(bool state);
3046 
3047 /*
3048  * Test if any discarding of memory in ram blocks is disabled.
3049  */
3050 bool ram_block_discard_is_disabled(void);
3051 
3052 /*
3053  * Test if any discarding of memory in ram blocks is required to work reliably.
3054  */
3055 bool ram_block_discard_is_required(void);
3056 
3057 #endif
3058 
3059 #endif
3060