xref: /qemu/include/exec/memory.h (revision 922d42bb)
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 "qemu:memory-region"
37 DECLARE_INSTANCE_CHECKER(MemoryRegion, MEMORY_REGION,
38                          TYPE_MEMORY_REGION)
39 
40 #define TYPE_IOMMU_MEMORY_REGION "qemu:iommu-memory-region"
41 typedef struct IOMMUMemoryRegionClass IOMMUMemoryRegionClass;
42 DECLARE_OBJ_CHECKERS(IOMMUMemoryRegion, IOMMUMemoryRegionClass,
43                      IOMMU_MEMORY_REGION, TYPE_IOMMU_MEMORY_REGION)
44 
45 #ifdef CONFIG_FUZZ
46 void fuzz_dma_read_cb(size_t addr,
47                       size_t len,
48                       MemoryRegion *mr,
49                       bool is_write);
50 #else
51 static inline void fuzz_dma_read_cb(size_t addr,
52                                     size_t len,
53                                     MemoryRegion *mr,
54                                     bool is_write)
55 {
56     /* Do Nothing */
57 }
58 #endif
59 
60 extern bool global_dirty_log;
61 
62 typedef struct MemoryRegionOps MemoryRegionOps;
63 
64 struct ReservedRegion {
65     hwaddr low;
66     hwaddr high;
67     unsigned type;
68 };
69 
70 typedef struct IOMMUTLBEntry IOMMUTLBEntry;
71 
72 /* See address_space_translate: bit 0 is read, bit 1 is write.  */
73 typedef enum {
74     IOMMU_NONE = 0,
75     IOMMU_RO   = 1,
76     IOMMU_WO   = 2,
77     IOMMU_RW   = 3,
78 } IOMMUAccessFlags;
79 
80 #define IOMMU_ACCESS_FLAG(r, w) (((r) ? IOMMU_RO : 0) | ((w) ? IOMMU_WO : 0))
81 
82 struct IOMMUTLBEntry {
83     AddressSpace    *target_as;
84     hwaddr           iova;
85     hwaddr           translated_addr;
86     hwaddr           addr_mask;  /* 0xfff = 4k translation */
87     IOMMUAccessFlags perm;
88 };
89 
90 /*
91  * Bitmap for different IOMMUNotifier capabilities. Each notifier can
92  * register with one or multiple IOMMU Notifier capability bit(s).
93  */
94 typedef enum {
95     IOMMU_NOTIFIER_NONE = 0,
96     /* Notify cache invalidations */
97     IOMMU_NOTIFIER_UNMAP = 0x1,
98     /* Notify entry changes (newly created entries) */
99     IOMMU_NOTIFIER_MAP = 0x2,
100 } IOMMUNotifierFlag;
101 
102 #define IOMMU_NOTIFIER_ALL (IOMMU_NOTIFIER_MAP | IOMMU_NOTIFIER_UNMAP)
103 
104 struct IOMMUNotifier;
105 typedef void (*IOMMUNotify)(struct IOMMUNotifier *notifier,
106                             IOMMUTLBEntry *data);
107 
108 struct IOMMUNotifier {
109     IOMMUNotify notify;
110     IOMMUNotifierFlag notifier_flags;
111     /* Notify for address space range start <= addr <= end */
112     hwaddr start;
113     hwaddr end;
114     int iommu_idx;
115     QLIST_ENTRY(IOMMUNotifier) node;
116 };
117 typedef struct IOMMUNotifier IOMMUNotifier;
118 
119 /* RAM is pre-allocated and passed into qemu_ram_alloc_from_ptr */
120 #define RAM_PREALLOC   (1 << 0)
121 
122 /* RAM is mmap-ed with MAP_SHARED */
123 #define RAM_SHARED     (1 << 1)
124 
125 /* Only a portion of RAM (used_length) is actually used, and migrated.
126  * This used_length size can change across reboots.
127  */
128 #define RAM_RESIZEABLE (1 << 2)
129 
130 /* UFFDIO_ZEROPAGE is available on this RAMBlock to atomically
131  * zero the page and wake waiting processes.
132  * (Set during postcopy)
133  */
134 #define RAM_UF_ZEROPAGE (1 << 3)
135 
136 /* RAM can be migrated */
137 #define RAM_MIGRATABLE (1 << 4)
138 
139 /* RAM is a persistent kind memory */
140 #define RAM_PMEM (1 << 5)
141 
142 static inline void iommu_notifier_init(IOMMUNotifier *n, IOMMUNotify fn,
143                                        IOMMUNotifierFlag flags,
144                                        hwaddr start, hwaddr end,
145                                        int iommu_idx)
146 {
147     n->notify = fn;
148     n->notifier_flags = flags;
149     n->start = start;
150     n->end = end;
151     n->iommu_idx = iommu_idx;
152 }
153 
154 /*
155  * Memory region callbacks
156  */
157 struct MemoryRegionOps {
158     /* Read from the memory region. @addr is relative to @mr; @size is
159      * in bytes. */
160     uint64_t (*read)(void *opaque,
161                      hwaddr addr,
162                      unsigned size);
163     /* Write to the memory region. @addr is relative to @mr; @size is
164      * in bytes. */
165     void (*write)(void *opaque,
166                   hwaddr addr,
167                   uint64_t data,
168                   unsigned size);
169 
170     MemTxResult (*read_with_attrs)(void *opaque,
171                                    hwaddr addr,
172                                    uint64_t *data,
173                                    unsigned size,
174                                    MemTxAttrs attrs);
175     MemTxResult (*write_with_attrs)(void *opaque,
176                                     hwaddr addr,
177                                     uint64_t data,
178                                     unsigned size,
179                                     MemTxAttrs attrs);
180 
181     enum device_endian endianness;
182     /* Guest-visible constraints: */
183     struct {
184         /* If nonzero, specify bounds on access sizes beyond which a machine
185          * check is thrown.
186          */
187         unsigned min_access_size;
188         unsigned max_access_size;
189         /* If true, unaligned accesses are supported.  Otherwise unaligned
190          * accesses throw machine checks.
191          */
192          bool unaligned;
193         /*
194          * If present, and returns #false, the transaction is not accepted
195          * by the device (and results in machine dependent behaviour such
196          * as a machine check exception).
197          */
198         bool (*accepts)(void *opaque, hwaddr addr,
199                         unsigned size, bool is_write,
200                         MemTxAttrs attrs);
201     } valid;
202     /* Internal implementation constraints: */
203     struct {
204         /* If nonzero, specifies the minimum size implemented.  Smaller sizes
205          * will be rounded upwards and a partial result will be returned.
206          */
207         unsigned min_access_size;
208         /* If nonzero, specifies the maximum size implemented.  Larger sizes
209          * will be done as a series of accesses with smaller sizes.
210          */
211         unsigned max_access_size;
212         /* If true, unaligned accesses are supported.  Otherwise all accesses
213          * are converted to (possibly multiple) naturally aligned accesses.
214          */
215         bool unaligned;
216     } impl;
217 };
218 
219 typedef struct MemoryRegionClass {
220     /* private */
221     ObjectClass parent_class;
222 } MemoryRegionClass;
223 
224 
225 enum IOMMUMemoryRegionAttr {
226     IOMMU_ATTR_SPAPR_TCE_FD
227 };
228 
229 /*
230  * IOMMUMemoryRegionClass:
231  *
232  * All IOMMU implementations need to subclass TYPE_IOMMU_MEMORY_REGION
233  * and provide an implementation of at least the @translate method here
234  * to handle requests to the memory region. Other methods are optional.
235  *
236  * The IOMMU implementation must use the IOMMU notifier infrastructure
237  * to report whenever mappings are changed, by calling
238  * memory_region_notify_iommu() (or, if necessary, by calling
239  * memory_region_notify_one() for each registered notifier).
240  *
241  * Conceptually an IOMMU provides a mapping from input address
242  * to an output TLB entry. If the IOMMU is aware of memory transaction
243  * attributes and the output TLB entry depends on the transaction
244  * attributes, we represent this using IOMMU indexes. Each index
245  * selects a particular translation table that the IOMMU has:
246  *
247  *   @attrs_to_index returns the IOMMU index for a set of transaction attributes
248  *
249  *   @translate takes an input address and an IOMMU index
250  *
251  * and the mapping returned can only depend on the input address and the
252  * IOMMU index.
253  *
254  * Most IOMMUs don't care about the transaction attributes and support
255  * only a single IOMMU index. A more complex IOMMU might have one index
256  * for secure transactions and one for non-secure transactions.
257  */
258 struct IOMMUMemoryRegionClass {
259     /* private: */
260     MemoryRegionClass parent_class;
261 
262     /* public: */
263     /**
264      * @translate:
265      *
266      * Return a TLB entry that contains a given address.
267      *
268      * The IOMMUAccessFlags indicated via @flag are optional and may
269      * be specified as IOMMU_NONE to indicate that the caller needs
270      * the full translation information for both reads and writes. If
271      * the access flags are specified then the IOMMU implementation
272      * may use this as an optimization, to stop doing a page table
273      * walk as soon as it knows that the requested permissions are not
274      * allowed. If IOMMU_NONE is passed then the IOMMU must do the
275      * full page table walk and report the permissions in the returned
276      * IOMMUTLBEntry. (Note that this implies that an IOMMU may not
277      * return different mappings for reads and writes.)
278      *
279      * The returned information remains valid while the caller is
280      * holding the big QEMU lock or is inside an RCU critical section;
281      * if the caller wishes to cache the mapping beyond that it must
282      * register an IOMMU notifier so it can invalidate its cached
283      * information when the IOMMU mapping changes.
284      *
285      * @iommu: the IOMMUMemoryRegion
286      *
287      * @hwaddr: address to be translated within the memory region
288      *
289      * @flag: requested access permission
290      *
291      * @iommu_idx: IOMMU index for the translation
292      */
293     IOMMUTLBEntry (*translate)(IOMMUMemoryRegion *iommu, hwaddr addr,
294                                IOMMUAccessFlags flag, int iommu_idx);
295     /**
296      * @get_min_page_size:
297      *
298      * Returns minimum supported page size in bytes.
299      *
300      * If this method is not provided then the minimum is assumed to
301      * be TARGET_PAGE_SIZE.
302      *
303      * @iommu: the IOMMUMemoryRegion
304      */
305     uint64_t (*get_min_page_size)(IOMMUMemoryRegion *iommu);
306     /**
307      * @notify_flag_changed:
308      *
309      * Called when IOMMU Notifier flag changes (ie when the set of
310      * events which IOMMU users are requesting notification for changes).
311      * Optional method -- need not be provided if the IOMMU does not
312      * need to know exactly which events must be notified.
313      *
314      * @iommu: the IOMMUMemoryRegion
315      *
316      * @old_flags: events which previously needed to be notified
317      *
318      * @new_flags: events which now need to be notified
319      *
320      * Returns 0 on success, or a negative errno; in particular
321      * returns -EINVAL if the new flag bitmap is not supported by the
322      * IOMMU memory region. In case of failure, the error object
323      * must be created
324      */
325     int (*notify_flag_changed)(IOMMUMemoryRegion *iommu,
326                                IOMMUNotifierFlag old_flags,
327                                IOMMUNotifierFlag new_flags,
328                                Error **errp);
329     /**
330      * @replay:
331      *
332      * Called to handle memory_region_iommu_replay().
333      *
334      * The default implementation of memory_region_iommu_replay() is to
335      * call the IOMMU translate method for every page in the address space
336      * with flag == IOMMU_NONE and then call the notifier if translate
337      * returns a valid mapping. If this method is implemented then it
338      * overrides the default behaviour, and must provide the full semantics
339      * of memory_region_iommu_replay(), by calling @notifier for every
340      * translation present in the IOMMU.
341      *
342      * Optional method -- an IOMMU only needs to provide this method
343      * if the default is inefficient or produces undesirable side effects.
344      *
345      * Note: this is not related to record-and-replay functionality.
346      */
347     void (*replay)(IOMMUMemoryRegion *iommu, IOMMUNotifier *notifier);
348 
349     /**
350      * @get_attr:
351      *
352      * Get IOMMU misc attributes. This is an optional method that
353      * can be used to allow users of the IOMMU to get implementation-specific
354      * information. The IOMMU implements this method to handle calls
355      * by IOMMU users to memory_region_iommu_get_attr() by filling in
356      * the arbitrary data pointer for any IOMMUMemoryRegionAttr values that
357      * the IOMMU supports. If the method is unimplemented then
358      * memory_region_iommu_get_attr() will always return -EINVAL.
359      *
360      * @iommu: the IOMMUMemoryRegion
361      *
362      * @attr: attribute being queried
363      *
364      * @data: memory to fill in with the attribute data
365      *
366      * Returns 0 on success, or a negative errno; in particular
367      * returns -EINVAL for unrecognized or unimplemented attribute types.
368      */
369     int (*get_attr)(IOMMUMemoryRegion *iommu, enum IOMMUMemoryRegionAttr attr,
370                     void *data);
371 
372     /**
373      * @attrs_to_index:
374      *
375      * Return the IOMMU index to use for a given set of transaction attributes.
376      *
377      * Optional method: if an IOMMU only supports a single IOMMU index then
378      * the default implementation of memory_region_iommu_attrs_to_index()
379      * will return 0.
380      *
381      * The indexes supported by an IOMMU must be contiguous, starting at 0.
382      *
383      * @iommu: the IOMMUMemoryRegion
384      * @attrs: memory transaction attributes
385      */
386     int (*attrs_to_index)(IOMMUMemoryRegion *iommu, MemTxAttrs attrs);
387 
388     /**
389      * @num_indexes:
390      *
391      * Return the number of IOMMU indexes this IOMMU supports.
392      *
393      * Optional method: if this method is not provided, then
394      * memory_region_iommu_num_indexes() will return 1, indicating that
395      * only a single IOMMU index is supported.
396      *
397      * @iommu: the IOMMUMemoryRegion
398      */
399     int (*num_indexes)(IOMMUMemoryRegion *iommu);
400 
401     /**
402      * @iommu_set_page_size_mask:
403      *
404      * Restrict the page size mask that can be supported with a given IOMMU
405      * memory region. Used for example to propagate host physical IOMMU page
406      * size mask limitations to the virtual IOMMU.
407      *
408      * Optional method: if this method is not provided, then the default global
409      * page mask is used.
410      *
411      * @iommu: the IOMMUMemoryRegion
412      *
413      * @page_size_mask: a bitmask of supported page sizes. At least one bit,
414      * representing the smallest page size, must be set. Additional set bits
415      * represent supported block sizes. For example a host physical IOMMU that
416      * uses page tables with a page size of 4kB, and supports 2MB and 4GB
417      * blocks, will set mask 0x40201000. A granule of 4kB with indiscriminate
418      * block sizes is specified with mask 0xfffffffffffff000.
419      *
420      * Returns 0 on success, or a negative error. In case of failure, the error
421      * object must be created.
422      */
423      int (*iommu_set_page_size_mask)(IOMMUMemoryRegion *iommu,
424                                      uint64_t page_size_mask,
425                                      Error **errp);
426 };
427 
428 typedef struct CoalescedMemoryRange CoalescedMemoryRange;
429 typedef struct MemoryRegionIoeventfd MemoryRegionIoeventfd;
430 
431 /** MemoryRegion:
432  *
433  * A struct representing a memory region.
434  */
435 struct MemoryRegion {
436     Object parent_obj;
437 
438     /* private: */
439 
440     /* The following fields should fit in a cache line */
441     bool romd_mode;
442     bool ram;
443     bool subpage;
444     bool readonly; /* For RAM regions */
445     bool nonvolatile;
446     bool rom_device;
447     bool flush_coalesced_mmio;
448     uint8_t dirty_log_mask;
449     bool is_iommu;
450     RAMBlock *ram_block;
451     Object *owner;
452 
453     const MemoryRegionOps *ops;
454     void *opaque;
455     MemoryRegion *container;
456     Int128 size;
457     hwaddr addr;
458     void (*destructor)(MemoryRegion *mr);
459     uint64_t align;
460     bool terminates;
461     bool ram_device;
462     bool enabled;
463     bool warning_printed; /* For reservations */
464     uint8_t vga_logging_count;
465     MemoryRegion *alias;
466     hwaddr alias_offset;
467     int32_t priority;
468     QTAILQ_HEAD(, MemoryRegion) subregions;
469     QTAILQ_ENTRY(MemoryRegion) subregions_link;
470     QTAILQ_HEAD(, CoalescedMemoryRange) coalesced;
471     const char *name;
472     unsigned ioeventfd_nb;
473     MemoryRegionIoeventfd *ioeventfds;
474 };
475 
476 struct IOMMUMemoryRegion {
477     MemoryRegion parent_obj;
478 
479     QLIST_HEAD(, IOMMUNotifier) iommu_notify;
480     IOMMUNotifierFlag iommu_notify_flags;
481 };
482 
483 #define IOMMU_NOTIFIER_FOREACH(n, mr) \
484     QLIST_FOREACH((n), &(mr)->iommu_notify, node)
485 
486 /**
487  * struct MemoryListener: callbacks structure for updates to the physical memory map
488  *
489  * Allows a component to adjust to changes in the guest-visible memory map.
490  * Use with memory_listener_register() and memory_listener_unregister().
491  */
492 struct MemoryListener {
493     /**
494      * @begin:
495      *
496      * Called at the beginning of an address space update transaction.
497      * Followed by calls to #MemoryListener.region_add(),
498      * #MemoryListener.region_del(), #MemoryListener.region_nop(),
499      * #MemoryListener.log_start() and #MemoryListener.log_stop() in
500      * increasing address order.
501      *
502      * @listener: The #MemoryListener.
503      */
504     void (*begin)(MemoryListener *listener);
505 
506     /**
507      * @commit:
508      *
509      * Called at the end of an address space update transaction,
510      * after the last call to #MemoryListener.region_add(),
511      * #MemoryListener.region_del() or #MemoryListener.region_nop(),
512      * #MemoryListener.log_start() and #MemoryListener.log_stop().
513      *
514      * @listener: The #MemoryListener.
515      */
516     void (*commit)(MemoryListener *listener);
517 
518     /**
519      * @region_add:
520      *
521      * Called during an address space update transaction,
522      * for a section of the address space that is new in this address space
523      * space since the last transaction.
524      *
525      * @listener: The #MemoryListener.
526      * @section: The new #MemoryRegionSection.
527      */
528     void (*region_add)(MemoryListener *listener, MemoryRegionSection *section);
529 
530     /**
531      * @region_del:
532      *
533      * Called during an address space update transaction,
534      * for a section of the address space that has disappeared in the address
535      * space since the last transaction.
536      *
537      * @listener: The #MemoryListener.
538      * @section: The old #MemoryRegionSection.
539      */
540     void (*region_del)(MemoryListener *listener, MemoryRegionSection *section);
541 
542     /**
543      * @region_nop:
544      *
545      * Called during an address space update transaction,
546      * for a section of the address space that is in the same place in the address
547      * space as in the last transaction.
548      *
549      * @listener: The #MemoryListener.
550      * @section: The #MemoryRegionSection.
551      */
552     void (*region_nop)(MemoryListener *listener, MemoryRegionSection *section);
553 
554     /**
555      * @log_start:
556      *
557      * Called during an address space update transaction, after
558      * one of #MemoryListener.region_add(),#MemoryListener.region_del() or
559      * #MemoryListener.region_nop(), if dirty memory logging clients have
560      * become active since the last transaction.
561      *
562      * @listener: The #MemoryListener.
563      * @section: The #MemoryRegionSection.
564      * @old: A bitmap of dirty memory logging clients that were active in
565      * the previous transaction.
566      * @new: A bitmap of dirty memory logging clients that are active in
567      * the current transaction.
568      */
569     void (*log_start)(MemoryListener *listener, MemoryRegionSection *section,
570                       int old, int new);
571 
572     /**
573      * @log_stop:
574      *
575      * Called during an address space update transaction, after
576      * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
577      * #MemoryListener.region_nop() and possibly after
578      * #MemoryListener.log_start(), if dirty memory logging clients have
579      * become inactive since the last transaction.
580      *
581      * @listener: The #MemoryListener.
582      * @section: The #MemoryRegionSection.
583      * @old: A bitmap of dirty memory logging clients that were active in
584      * the previous transaction.
585      * @new: A bitmap of dirty memory logging clients that are active in
586      * the current transaction.
587      */
588     void (*log_stop)(MemoryListener *listener, MemoryRegionSection *section,
589                      int old, int new);
590 
591     /**
592      * @log_sync:
593      *
594      * Called by memory_region_snapshot_and_clear_dirty() and
595      * memory_global_dirty_log_sync(), before accessing QEMU's "official"
596      * copy of the dirty memory bitmap for a #MemoryRegionSection.
597      *
598      * @listener: The #MemoryListener.
599      * @section: The #MemoryRegionSection.
600      */
601     void (*log_sync)(MemoryListener *listener, MemoryRegionSection *section);
602 
603     /**
604      * @log_clear:
605      *
606      * Called before reading the dirty memory bitmap for a
607      * #MemoryRegionSection.
608      *
609      * @listener: The #MemoryListener.
610      * @section: The #MemoryRegionSection.
611      */
612     void (*log_clear)(MemoryListener *listener, MemoryRegionSection *section);
613 
614     /**
615      * @log_global_start:
616      *
617      * Called by memory_global_dirty_log_start(), which
618      * enables the %DIRTY_LOG_MIGRATION client on all memory regions in
619      * the address space.  #MemoryListener.log_global_start() is also
620      * called when a #MemoryListener is added, if global dirty logging is
621      * active at that time.
622      *
623      * @listener: The #MemoryListener.
624      */
625     void (*log_global_start)(MemoryListener *listener);
626 
627     /**
628      * @log_global_stop:
629      *
630      * Called by memory_global_dirty_log_stop(), which
631      * disables the %DIRTY_LOG_MIGRATION client on all memory regions in
632      * the address space.
633      *
634      * @listener: The #MemoryListener.
635      */
636     void (*log_global_stop)(MemoryListener *listener);
637 
638     /**
639      * @log_global_after_sync:
640      *
641      * Called after reading the dirty memory bitmap
642      * for any #MemoryRegionSection.
643      *
644      * @listener: The #MemoryListener.
645      */
646     void (*log_global_after_sync)(MemoryListener *listener);
647 
648     /**
649      * @eventfd_add:
650      *
651      * Called during an address space update transaction,
652      * for a section of the address space that has had a new ioeventfd
653      * registration since the last transaction.
654      *
655      * @listener: The #MemoryListener.
656      * @section: The new #MemoryRegionSection.
657      * @match_data: The @match_data parameter for the new ioeventfd.
658      * @data: The @data parameter for the new ioeventfd.
659      * @e: The #EventNotifier parameter for the new ioeventfd.
660      */
661     void (*eventfd_add)(MemoryListener *listener, MemoryRegionSection *section,
662                         bool match_data, uint64_t data, EventNotifier *e);
663 
664     /**
665      * @eventfd_del:
666      *
667      * Called during an address space update transaction,
668      * for a section of the address space that has dropped an ioeventfd
669      * registration since the last transaction.
670      *
671      * @listener: The #MemoryListener.
672      * @section: The new #MemoryRegionSection.
673      * @match_data: The @match_data parameter for the dropped ioeventfd.
674      * @data: The @data parameter for the dropped ioeventfd.
675      * @e: The #EventNotifier parameter for the dropped ioeventfd.
676      */
677     void (*eventfd_del)(MemoryListener *listener, MemoryRegionSection *section,
678                         bool match_data, uint64_t data, EventNotifier *e);
679 
680     /**
681      * @coalesced_io_add:
682      *
683      * Called during an address space update transaction,
684      * for a section of the address space that has had a new coalesced
685      * MMIO range registration since the last transaction.
686      *
687      * @listener: The #MemoryListener.
688      * @section: The new #MemoryRegionSection.
689      * @addr: The starting address for the coalesced MMIO range.
690      * @len: The length of the coalesced MMIO range.
691      */
692     void (*coalesced_io_add)(MemoryListener *listener, MemoryRegionSection *section,
693                                hwaddr addr, hwaddr len);
694 
695     /**
696      * @coalesced_io_del:
697      *
698      * Called during an address space update transaction,
699      * for a section of the address space that has dropped a coalesced
700      * MMIO range since the last transaction.
701      *
702      * @listener: The #MemoryListener.
703      * @section: The new #MemoryRegionSection.
704      * @addr: The starting address for the coalesced MMIO range.
705      * @len: The length of the coalesced MMIO range.
706      */
707     void (*coalesced_io_del)(MemoryListener *listener, MemoryRegionSection *section,
708                                hwaddr addr, hwaddr len);
709     /**
710      * @priority:
711      *
712      * Govern the order in which memory listeners are invoked. Lower priorities
713      * are invoked earlier for "add" or "start" callbacks, and later for "delete"
714      * or "stop" callbacks.
715      */
716     unsigned priority;
717 
718     /* private: */
719     AddressSpace *address_space;
720     QTAILQ_ENTRY(MemoryListener) link;
721     QTAILQ_ENTRY(MemoryListener) link_as;
722 };
723 
724 /**
725  * struct AddressSpace: describes a mapping of addresses to #MemoryRegion objects
726  */
727 struct AddressSpace {
728     /* private: */
729     struct rcu_head rcu;
730     char *name;
731     MemoryRegion *root;
732 
733     /* Accessed via RCU.  */
734     struct FlatView *current_map;
735 
736     int ioeventfd_nb;
737     struct MemoryRegionIoeventfd *ioeventfds;
738     QTAILQ_HEAD(, MemoryListener) listeners;
739     QTAILQ_ENTRY(AddressSpace) address_spaces_link;
740 };
741 
742 typedef struct AddressSpaceDispatch AddressSpaceDispatch;
743 typedef struct FlatRange FlatRange;
744 
745 /* Flattened global view of current active memory hierarchy.  Kept in sorted
746  * order.
747  */
748 struct FlatView {
749     struct rcu_head rcu;
750     unsigned ref;
751     FlatRange *ranges;
752     unsigned nr;
753     unsigned nr_allocated;
754     struct AddressSpaceDispatch *dispatch;
755     MemoryRegion *root;
756 };
757 
758 static inline FlatView *address_space_to_flatview(AddressSpace *as)
759 {
760     return qatomic_rcu_read(&as->current_map);
761 }
762 
763 typedef int (*flatview_cb)(Int128 start,
764                            Int128 len,
765                            const MemoryRegion*, void*);
766 
767 void flatview_for_each_range(FlatView *fv, flatview_cb cb , void *opaque);
768 
769 /**
770  * struct MemoryRegionSection: describes a fragment of a #MemoryRegion
771  *
772  * @mr: the region, or %NULL if empty
773  * @fv: the flat view of the address space the region is mapped in
774  * @offset_within_region: the beginning of the section, relative to @mr's start
775  * @size: the size of the section; will not exceed @mr's boundaries
776  * @offset_within_address_space: the address of the first byte of the section
777  *     relative to the region's address space
778  * @readonly: writes to this section are ignored
779  * @nonvolatile: this section is non-volatile
780  */
781 struct MemoryRegionSection {
782     Int128 size;
783     MemoryRegion *mr;
784     FlatView *fv;
785     hwaddr offset_within_region;
786     hwaddr offset_within_address_space;
787     bool readonly;
788     bool nonvolatile;
789 };
790 
791 static inline bool MemoryRegionSection_eq(MemoryRegionSection *a,
792                                           MemoryRegionSection *b)
793 {
794     return a->mr == b->mr &&
795            a->fv == b->fv &&
796            a->offset_within_region == b->offset_within_region &&
797            a->offset_within_address_space == b->offset_within_address_space &&
798            int128_eq(a->size, b->size) &&
799            a->readonly == b->readonly &&
800            a->nonvolatile == b->nonvolatile;
801 }
802 
803 /**
804  * memory_region_init: Initialize a memory region
805  *
806  * The region typically acts as a container for other memory regions.  Use
807  * memory_region_add_subregion() to add subregions.
808  *
809  * @mr: the #MemoryRegion to be initialized
810  * @owner: the object that tracks the region's reference count
811  * @name: used for debugging; not visible to the user or ABI
812  * @size: size of the region; any subregions beyond this size will be clipped
813  */
814 void memory_region_init(MemoryRegion *mr,
815                         struct Object *owner,
816                         const char *name,
817                         uint64_t size);
818 
819 /**
820  * memory_region_ref: Add 1 to a memory region's reference count
821  *
822  * Whenever memory regions are accessed outside the BQL, they need to be
823  * preserved against hot-unplug.  MemoryRegions actually do not have their
824  * own reference count; they piggyback on a QOM object, their "owner".
825  * This function adds a reference to the owner.
826  *
827  * All MemoryRegions must have an owner if they can disappear, even if the
828  * device they belong to operates exclusively under the BQL.  This is because
829  * the region could be returned at any time by memory_region_find, and this
830  * is usually under guest control.
831  *
832  * @mr: the #MemoryRegion
833  */
834 void memory_region_ref(MemoryRegion *mr);
835 
836 /**
837  * memory_region_unref: Remove 1 to a memory region's reference count
838  *
839  * Whenever memory regions are accessed outside the BQL, they need to be
840  * preserved against hot-unplug.  MemoryRegions actually do not have their
841  * own reference count; they piggyback on a QOM object, their "owner".
842  * This function removes a reference to the owner and possibly destroys it.
843  *
844  * @mr: the #MemoryRegion
845  */
846 void memory_region_unref(MemoryRegion *mr);
847 
848 /**
849  * memory_region_init_io: Initialize an I/O memory region.
850  *
851  * Accesses into the region will cause the callbacks in @ops to be called.
852  * if @size is nonzero, subregions will be clipped to @size.
853  *
854  * @mr: the #MemoryRegion to be initialized.
855  * @owner: the object that tracks the region's reference count
856  * @ops: a structure containing read and write callbacks to be used when
857  *       I/O is performed on the region.
858  * @opaque: passed to the read and write callbacks of the @ops structure.
859  * @name: used for debugging; not visible to the user or ABI
860  * @size: size of the region.
861  */
862 void memory_region_init_io(MemoryRegion *mr,
863                            struct Object *owner,
864                            const MemoryRegionOps *ops,
865                            void *opaque,
866                            const char *name,
867                            uint64_t size);
868 
869 /**
870  * memory_region_init_ram_nomigrate:  Initialize RAM memory region.  Accesses
871  *                                    into the region will modify memory
872  *                                    directly.
873  *
874  * @mr: the #MemoryRegion to be initialized.
875  * @owner: the object that tracks the region's reference count
876  * @name: Region name, becomes part of RAMBlock name used in migration stream
877  *        must be unique within any device
878  * @size: size of the region.
879  * @errp: pointer to Error*, to store an error if it happens.
880  *
881  * Note that this function does not do anything to cause the data in the
882  * RAM memory region to be migrated; that is the responsibility of the caller.
883  */
884 void memory_region_init_ram_nomigrate(MemoryRegion *mr,
885                                       struct Object *owner,
886                                       const char *name,
887                                       uint64_t size,
888                                       Error **errp);
889 
890 /**
891  * memory_region_init_ram_shared_nomigrate:  Initialize RAM memory region.
892  *                                           Accesses into the region will
893  *                                           modify memory directly.
894  *
895  * @mr: the #MemoryRegion to be initialized.
896  * @owner: the object that tracks the region's reference count
897  * @name: Region name, becomes part of RAMBlock name used in migration stream
898  *        must be unique within any device
899  * @size: size of the region.
900  * @share: allow remapping RAM to different addresses
901  * @errp: pointer to Error*, to store an error if it happens.
902  *
903  * Note that this function is similar to memory_region_init_ram_nomigrate.
904  * The only difference is part of the RAM region can be remapped.
905  */
906 void memory_region_init_ram_shared_nomigrate(MemoryRegion *mr,
907                                              struct Object *owner,
908                                              const char *name,
909                                              uint64_t size,
910                                              bool share,
911                                              Error **errp);
912 
913 /**
914  * memory_region_init_resizeable_ram:  Initialize memory region with resizeable
915  *                                     RAM.  Accesses into the region will
916  *                                     modify memory directly.  Only an initial
917  *                                     portion of this RAM is actually used.
918  *                                     The used size can change across reboots.
919  *
920  * @mr: the #MemoryRegion to be initialized.
921  * @owner: the object that tracks the region's reference count
922  * @name: Region name, becomes part of RAMBlock name used in migration stream
923  *        must be unique within any device
924  * @size: used size of the region.
925  * @max_size: max size of the region.
926  * @resized: callback to notify owner about used size change.
927  * @errp: pointer to Error*, to store an error if it happens.
928  *
929  * Note that this function does not do anything to cause the data in the
930  * RAM memory region to be migrated; that is the responsibility of the caller.
931  */
932 void memory_region_init_resizeable_ram(MemoryRegion *mr,
933                                        struct Object *owner,
934                                        const char *name,
935                                        uint64_t size,
936                                        uint64_t max_size,
937                                        void (*resized)(const char*,
938                                                        uint64_t length,
939                                                        void *host),
940                                        Error **errp);
941 #ifdef CONFIG_POSIX
942 
943 /**
944  * memory_region_init_ram_from_file:  Initialize RAM memory region with a
945  *                                    mmap-ed backend.
946  *
947  * @mr: the #MemoryRegion to be initialized.
948  * @owner: the object that tracks the region's reference count
949  * @name: Region name, becomes part of RAMBlock name used in migration stream
950  *        must be unique within any device
951  * @size: size of the region.
952  * @align: alignment of the region base address; if 0, the default alignment
953  *         (getpagesize()) will be used.
954  * @ram_flags: Memory region features:
955  *             - RAM_SHARED: memory must be mmaped with the MAP_SHARED flag
956  *             - RAM_PMEM: the memory is persistent memory
957  *             Other bits are ignored now.
958  * @path: the path in which to allocate the RAM.
959  * @errp: pointer to Error*, to store an error if it happens.
960  *
961  * Note that this function does not do anything to cause the data in the
962  * RAM memory region to be migrated; that is the responsibility of the caller.
963  */
964 void memory_region_init_ram_from_file(MemoryRegion *mr,
965                                       struct Object *owner,
966                                       const char *name,
967                                       uint64_t size,
968                                       uint64_t align,
969                                       uint32_t ram_flags,
970                                       const char *path,
971                                       Error **errp);
972 
973 /**
974  * memory_region_init_ram_from_fd:  Initialize RAM memory region with a
975  *                                  mmap-ed backend.
976  *
977  * @mr: the #MemoryRegion to be initialized.
978  * @owner: the object that tracks the region's reference count
979  * @name: the name of the region.
980  * @size: size of the region.
981  * @share: %true if memory must be mmaped with the MAP_SHARED flag
982  * @fd: the fd to mmap.
983  * @errp: pointer to Error*, to store an error if it happens.
984  *
985  * Note that this function does not do anything to cause the data in the
986  * RAM memory region to be migrated; that is the responsibility of the caller.
987  */
988 void memory_region_init_ram_from_fd(MemoryRegion *mr,
989                                     struct Object *owner,
990                                     const char *name,
991                                     uint64_t size,
992                                     bool share,
993                                     int fd,
994                                     Error **errp);
995 #endif
996 
997 /**
998  * memory_region_init_ram_ptr:  Initialize RAM memory region from a
999  *                              user-provided pointer.  Accesses into the
1000  *                              region will modify memory directly.
1001  *
1002  * @mr: the #MemoryRegion to be initialized.
1003  * @owner: the object that tracks the region's reference count
1004  * @name: Region name, becomes part of RAMBlock name used in migration stream
1005  *        must be unique within any device
1006  * @size: size of the region.
1007  * @ptr: memory to be mapped; must contain at least @size bytes.
1008  *
1009  * Note that this function does not do anything to cause the data in the
1010  * RAM memory region to be migrated; that is the responsibility of the caller.
1011  */
1012 void memory_region_init_ram_ptr(MemoryRegion *mr,
1013                                 struct Object *owner,
1014                                 const char *name,
1015                                 uint64_t size,
1016                                 void *ptr);
1017 
1018 /**
1019  * memory_region_init_ram_device_ptr:  Initialize RAM device memory region from
1020  *                                     a user-provided pointer.
1021  *
1022  * A RAM device represents a mapping to a physical device, such as to a PCI
1023  * MMIO BAR of an vfio-pci assigned device.  The memory region may be mapped
1024  * into the VM address space and access to the region will modify memory
1025  * directly.  However, the memory region should not be included in a memory
1026  * dump (device may not be enabled/mapped at the time of the dump), and
1027  * operations incompatible with manipulating MMIO should be avoided.  Replaces
1028  * skip_dump flag.
1029  *
1030  * @mr: the #MemoryRegion to be initialized.
1031  * @owner: the object that tracks the region's reference count
1032  * @name: the name of the region.
1033  * @size: size of the region.
1034  * @ptr: memory to be mapped; must contain at least @size bytes.
1035  *
1036  * Note that this function does not do anything to cause the data in the
1037  * RAM memory region to be migrated; that is the responsibility of the caller.
1038  * (For RAM device memory regions, migrating the contents rarely makes sense.)
1039  */
1040 void memory_region_init_ram_device_ptr(MemoryRegion *mr,
1041                                        struct Object *owner,
1042                                        const char *name,
1043                                        uint64_t size,
1044                                        void *ptr);
1045 
1046 /**
1047  * memory_region_init_alias: Initialize a memory region that aliases all or a
1048  *                           part of another memory region.
1049  *
1050  * @mr: the #MemoryRegion to be initialized.
1051  * @owner: the object that tracks the region's reference count
1052  * @name: used for debugging; not visible to the user or ABI
1053  * @orig: the region to be referenced; @mr will be equivalent to
1054  *        @orig between @offset and @offset + @size - 1.
1055  * @offset: start of the section in @orig to be referenced.
1056  * @size: size of the region.
1057  */
1058 void memory_region_init_alias(MemoryRegion *mr,
1059                               struct Object *owner,
1060                               const char *name,
1061                               MemoryRegion *orig,
1062                               hwaddr offset,
1063                               uint64_t size);
1064 
1065 /**
1066  * memory_region_init_rom_nomigrate: Initialize a ROM memory region.
1067  *
1068  * This has the same effect as calling memory_region_init_ram_nomigrate()
1069  * and then marking the resulting region read-only with
1070  * memory_region_set_readonly().
1071  *
1072  * Note that this function does not do anything to cause the data in the
1073  * RAM side of the memory region to be migrated; that is the responsibility
1074  * of the caller.
1075  *
1076  * @mr: the #MemoryRegion to be initialized.
1077  * @owner: the object that tracks the region's reference count
1078  * @name: Region name, becomes part of RAMBlock name used in migration stream
1079  *        must be unique within any device
1080  * @size: size of the region.
1081  * @errp: pointer to Error*, to store an error if it happens.
1082  */
1083 void memory_region_init_rom_nomigrate(MemoryRegion *mr,
1084                                       struct Object *owner,
1085                                       const char *name,
1086                                       uint64_t size,
1087                                       Error **errp);
1088 
1089 /**
1090  * memory_region_init_rom_device_nomigrate:  Initialize a ROM memory region.
1091  *                                 Writes are handled via callbacks.
1092  *
1093  * Note that this function does not do anything to cause the data in the
1094  * RAM side of the memory region to be migrated; that is the responsibility
1095  * of the caller.
1096  *
1097  * @mr: the #MemoryRegion to be initialized.
1098  * @owner: the object that tracks the region's reference count
1099  * @ops: callbacks for write access handling (must not be NULL).
1100  * @opaque: passed to the read and write callbacks of the @ops structure.
1101  * @name: Region name, becomes part of RAMBlock name used in migration stream
1102  *        must be unique within any device
1103  * @size: size of the region.
1104  * @errp: pointer to Error*, to store an error if it happens.
1105  */
1106 void memory_region_init_rom_device_nomigrate(MemoryRegion *mr,
1107                                              struct Object *owner,
1108                                              const MemoryRegionOps *ops,
1109                                              void *opaque,
1110                                              const char *name,
1111                                              uint64_t size,
1112                                              Error **errp);
1113 
1114 /**
1115  * memory_region_init_iommu: Initialize a memory region of a custom type
1116  * that translates addresses
1117  *
1118  * An IOMMU region translates addresses and forwards accesses to a target
1119  * memory region.
1120  *
1121  * The IOMMU implementation must define a subclass of TYPE_IOMMU_MEMORY_REGION.
1122  * @_iommu_mr should be a pointer to enough memory for an instance of
1123  * that subclass, @instance_size is the size of that subclass, and
1124  * @mrtypename is its name. This function will initialize @_iommu_mr as an
1125  * instance of the subclass, and its methods will then be called to handle
1126  * accesses to the memory region. See the documentation of
1127  * #IOMMUMemoryRegionClass for further details.
1128  *
1129  * @_iommu_mr: the #IOMMUMemoryRegion to be initialized
1130  * @instance_size: the IOMMUMemoryRegion subclass instance size
1131  * @mrtypename: the type name of the #IOMMUMemoryRegion
1132  * @owner: the object that tracks the region's reference count
1133  * @name: used for debugging; not visible to the user or ABI
1134  * @size: size of the region.
1135  */
1136 void memory_region_init_iommu(void *_iommu_mr,
1137                               size_t instance_size,
1138                               const char *mrtypename,
1139                               Object *owner,
1140                               const char *name,
1141                               uint64_t size);
1142 
1143 /**
1144  * memory_region_init_ram - Initialize RAM memory region.  Accesses into the
1145  *                          region will modify memory directly.
1146  *
1147  * @mr: the #MemoryRegion to be initialized
1148  * @owner: the object that tracks the region's reference count (must be
1149  *         TYPE_DEVICE or a subclass of TYPE_DEVICE, or NULL)
1150  * @name: name of the memory region
1151  * @size: size of the region in bytes
1152  * @errp: pointer to Error*, to store an error if it happens.
1153  *
1154  * This function allocates RAM for a board model or device, and
1155  * arranges for it to be migrated (by calling vmstate_register_ram()
1156  * if @owner is a DeviceState, or vmstate_register_ram_global() if
1157  * @owner is NULL).
1158  *
1159  * TODO: Currently we restrict @owner to being either NULL (for
1160  * global RAM regions with no owner) or devices, so that we can
1161  * give the RAM block a unique name for migration purposes.
1162  * We should lift this restriction and allow arbitrary Objects.
1163  * If you pass a non-NULL non-device @owner then we will assert.
1164  */
1165 void memory_region_init_ram(MemoryRegion *mr,
1166                             struct Object *owner,
1167                             const char *name,
1168                             uint64_t size,
1169                             Error **errp);
1170 
1171 /**
1172  * memory_region_init_rom: Initialize a ROM memory region.
1173  *
1174  * This has the same effect as calling memory_region_init_ram()
1175  * and then marking the resulting region read-only with
1176  * memory_region_set_readonly(). This includes arranging for the
1177  * contents to be migrated.
1178  *
1179  * TODO: Currently we restrict @owner to being either NULL (for
1180  * global RAM regions with no owner) or devices, so that we can
1181  * give the RAM block a unique name for migration purposes.
1182  * We should lift this restriction and allow arbitrary Objects.
1183  * If you pass a non-NULL non-device @owner then we will assert.
1184  *
1185  * @mr: the #MemoryRegion to be initialized.
1186  * @owner: the object that tracks the region's reference count
1187  * @name: Region name, becomes part of RAMBlock name used in migration stream
1188  *        must be unique within any device
1189  * @size: size of the region.
1190  * @errp: pointer to Error*, to store an error if it happens.
1191  */
1192 void memory_region_init_rom(MemoryRegion *mr,
1193                             struct Object *owner,
1194                             const char *name,
1195                             uint64_t size,
1196                             Error **errp);
1197 
1198 /**
1199  * memory_region_init_rom_device:  Initialize a ROM memory region.
1200  *                                 Writes are handled via callbacks.
1201  *
1202  * This function initializes a memory region backed by RAM for reads
1203  * and callbacks for writes, and arranges for the RAM backing to
1204  * be migrated (by calling vmstate_register_ram()
1205  * if @owner is a DeviceState, or vmstate_register_ram_global() if
1206  * @owner is NULL).
1207  *
1208  * TODO: Currently we restrict @owner to being either NULL (for
1209  * global RAM regions with no owner) or devices, so that we can
1210  * give the RAM block a unique name for migration purposes.
1211  * We should lift this restriction and allow arbitrary Objects.
1212  * If you pass a non-NULL non-device @owner then we will assert.
1213  *
1214  * @mr: the #MemoryRegion to be initialized.
1215  * @owner: the object that tracks the region's reference count
1216  * @ops: callbacks for write access handling (must not be NULL).
1217  * @opaque: passed to the read and write callbacks of the @ops structure.
1218  * @name: Region name, becomes part of RAMBlock name used in migration stream
1219  *        must be unique within any device
1220  * @size: size of the region.
1221  * @errp: pointer to Error*, to store an error if it happens.
1222  */
1223 void memory_region_init_rom_device(MemoryRegion *mr,
1224                                    struct Object *owner,
1225                                    const MemoryRegionOps *ops,
1226                                    void *opaque,
1227                                    const char *name,
1228                                    uint64_t size,
1229                                    Error **errp);
1230 
1231 
1232 /**
1233  * memory_region_owner: get a memory region's owner.
1234  *
1235  * @mr: the memory region being queried.
1236  */
1237 struct Object *memory_region_owner(MemoryRegion *mr);
1238 
1239 /**
1240  * memory_region_size: get a memory region's size.
1241  *
1242  * @mr: the memory region being queried.
1243  */
1244 uint64_t memory_region_size(MemoryRegion *mr);
1245 
1246 /**
1247  * memory_region_is_ram: check whether a memory region is random access
1248  *
1249  * Returns %true if a memory region is random access.
1250  *
1251  * @mr: the memory region being queried
1252  */
1253 static inline bool memory_region_is_ram(MemoryRegion *mr)
1254 {
1255     return mr->ram;
1256 }
1257 
1258 /**
1259  * memory_region_is_ram_device: check whether a memory region is a ram device
1260  *
1261  * Returns %true if a memory region is a device backed ram region
1262  *
1263  * @mr: the memory region being queried
1264  */
1265 bool memory_region_is_ram_device(MemoryRegion *mr);
1266 
1267 /**
1268  * memory_region_is_romd: check whether a memory region is in ROMD mode
1269  *
1270  * Returns %true if a memory region is a ROM device and currently set to allow
1271  * direct reads.
1272  *
1273  * @mr: the memory region being queried
1274  */
1275 static inline bool memory_region_is_romd(MemoryRegion *mr)
1276 {
1277     return mr->rom_device && mr->romd_mode;
1278 }
1279 
1280 /**
1281  * memory_region_get_iommu: check whether a memory region is an iommu
1282  *
1283  * Returns pointer to IOMMUMemoryRegion if a memory region is an iommu,
1284  * otherwise NULL.
1285  *
1286  * @mr: the memory region being queried
1287  */
1288 static inline IOMMUMemoryRegion *memory_region_get_iommu(MemoryRegion *mr)
1289 {
1290     if (mr->alias) {
1291         return memory_region_get_iommu(mr->alias);
1292     }
1293     if (mr->is_iommu) {
1294         return (IOMMUMemoryRegion *) mr;
1295     }
1296     return NULL;
1297 }
1298 
1299 /**
1300  * memory_region_get_iommu_class_nocheck: returns iommu memory region class
1301  *   if an iommu or NULL if not
1302  *
1303  * Returns pointer to IOMMUMemoryRegionClass if a memory region is an iommu,
1304  * otherwise NULL. This is fast path avoiding QOM checking, use with caution.
1305  *
1306  * @iommu_mr: the memory region being queried
1307  */
1308 static inline IOMMUMemoryRegionClass *memory_region_get_iommu_class_nocheck(
1309         IOMMUMemoryRegion *iommu_mr)
1310 {
1311     return (IOMMUMemoryRegionClass *) (((Object *)iommu_mr)->class);
1312 }
1313 
1314 #define memory_region_is_iommu(mr) (memory_region_get_iommu(mr) != NULL)
1315 
1316 /**
1317  * memory_region_iommu_get_min_page_size: get minimum supported page size
1318  * for an iommu
1319  *
1320  * Returns minimum supported page size for an iommu.
1321  *
1322  * @iommu_mr: the memory region being queried
1323  */
1324 uint64_t memory_region_iommu_get_min_page_size(IOMMUMemoryRegion *iommu_mr);
1325 
1326 /**
1327  * memory_region_notify_iommu: notify a change in an IOMMU translation entry.
1328  *
1329  * The notification type will be decided by entry.perm bits:
1330  *
1331  * - For UNMAP (cache invalidation) notifies: set entry.perm to IOMMU_NONE.
1332  * - For MAP (newly added entry) notifies: set entry.perm to the
1333  *   permission of the page (which is definitely !IOMMU_NONE).
1334  *
1335  * Note: for any IOMMU implementation, an in-place mapping change
1336  * should be notified with an UNMAP followed by a MAP.
1337  *
1338  * @iommu_mr: the memory region that was changed
1339  * @iommu_idx: the IOMMU index for the translation table which has changed
1340  * @entry: the new entry in the IOMMU translation table.  The entry
1341  *         replaces all old entries for the same virtual I/O address range.
1342  *         Deleted entries have .@perm == 0.
1343  */
1344 void memory_region_notify_iommu(IOMMUMemoryRegion *iommu_mr,
1345                                 int iommu_idx,
1346                                 IOMMUTLBEntry entry);
1347 
1348 /**
1349  * memory_region_notify_one: notify a change in an IOMMU translation
1350  *                           entry to a single notifier
1351  *
1352  * This works just like memory_region_notify_iommu(), but it only
1353  * notifies a specific notifier, not all of them.
1354  *
1355  * @notifier: the notifier to be notified
1356  * @entry: the new entry in the IOMMU translation table.  The entry
1357  *         replaces all old entries for the same virtual I/O address range.
1358  *         Deleted entries have .@perm == 0.
1359  */
1360 void memory_region_notify_one(IOMMUNotifier *notifier,
1361                               IOMMUTLBEntry *entry);
1362 
1363 /**
1364  * memory_region_register_iommu_notifier: register a notifier for changes to
1365  * IOMMU translation entries.
1366  *
1367  * Returns 0 on success, or a negative errno otherwise. In particular,
1368  * -EINVAL indicates that at least one of the attributes of the notifier
1369  * is not supported (flag/range) by the IOMMU memory region. In case of error
1370  * the error object must be created.
1371  *
1372  * @mr: the memory region to observe
1373  * @n: the IOMMUNotifier to be added; the notify callback receives a
1374  *     pointer to an #IOMMUTLBEntry as the opaque value; the pointer
1375  *     ceases to be valid on exit from the notifier.
1376  * @errp: pointer to Error*, to store an error if it happens.
1377  */
1378 int memory_region_register_iommu_notifier(MemoryRegion *mr,
1379                                           IOMMUNotifier *n, Error **errp);
1380 
1381 /**
1382  * memory_region_iommu_replay: replay existing IOMMU translations to
1383  * a notifier with the minimum page granularity returned by
1384  * mr->iommu_ops->get_page_size().
1385  *
1386  * Note: this is not related to record-and-replay functionality.
1387  *
1388  * @iommu_mr: the memory region to observe
1389  * @n: the notifier to which to replay iommu mappings
1390  */
1391 void memory_region_iommu_replay(IOMMUMemoryRegion *iommu_mr, IOMMUNotifier *n);
1392 
1393 /**
1394  * memory_region_unregister_iommu_notifier: unregister a notifier for
1395  * changes to IOMMU translation entries.
1396  *
1397  * @mr: the memory region which was observed and for which notity_stopped()
1398  *      needs to be called
1399  * @n: the notifier to be removed.
1400  */
1401 void memory_region_unregister_iommu_notifier(MemoryRegion *mr,
1402                                              IOMMUNotifier *n);
1403 
1404 /**
1405  * memory_region_iommu_get_attr: return an IOMMU attr if get_attr() is
1406  * defined on the IOMMU.
1407  *
1408  * Returns 0 on success, or a negative errno otherwise. In particular,
1409  * -EINVAL indicates that the IOMMU does not support the requested
1410  * attribute.
1411  *
1412  * @iommu_mr: the memory region
1413  * @attr: the requested attribute
1414  * @data: a pointer to the requested attribute data
1415  */
1416 int memory_region_iommu_get_attr(IOMMUMemoryRegion *iommu_mr,
1417                                  enum IOMMUMemoryRegionAttr attr,
1418                                  void *data);
1419 
1420 /**
1421  * memory_region_iommu_attrs_to_index: return the IOMMU index to
1422  * use for translations with the given memory transaction attributes.
1423  *
1424  * @iommu_mr: the memory region
1425  * @attrs: the memory transaction attributes
1426  */
1427 int memory_region_iommu_attrs_to_index(IOMMUMemoryRegion *iommu_mr,
1428                                        MemTxAttrs attrs);
1429 
1430 /**
1431  * memory_region_iommu_num_indexes: return the total number of IOMMU
1432  * indexes that this IOMMU supports.
1433  *
1434  * @iommu_mr: the memory region
1435  */
1436 int memory_region_iommu_num_indexes(IOMMUMemoryRegion *iommu_mr);
1437 
1438 /**
1439  * memory_region_iommu_set_page_size_mask: set the supported page
1440  * sizes for a given IOMMU memory region
1441  *
1442  * @iommu_mr: IOMMU memory region
1443  * @page_size_mask: supported page size mask
1444  * @errp: pointer to Error*, to store an error if it happens.
1445  */
1446 int memory_region_iommu_set_page_size_mask(IOMMUMemoryRegion *iommu_mr,
1447                                            uint64_t page_size_mask,
1448                                            Error **errp);
1449 
1450 /**
1451  * memory_region_name: get a memory region's name
1452  *
1453  * Returns the string that was used to initialize the memory region.
1454  *
1455  * @mr: the memory region being queried
1456  */
1457 const char *memory_region_name(const MemoryRegion *mr);
1458 
1459 /**
1460  * memory_region_is_logging: return whether a memory region is logging writes
1461  *
1462  * Returns %true if the memory region is logging writes for the given client
1463  *
1464  * @mr: the memory region being queried
1465  * @client: the client being queried
1466  */
1467 bool memory_region_is_logging(MemoryRegion *mr, uint8_t client);
1468 
1469 /**
1470  * memory_region_get_dirty_log_mask: return the clients for which a
1471  * memory region is logging writes.
1472  *
1473  * Returns a bitmap of clients, in which the DIRTY_MEMORY_* constants
1474  * are the bit indices.
1475  *
1476  * @mr: the memory region being queried
1477  */
1478 uint8_t memory_region_get_dirty_log_mask(MemoryRegion *mr);
1479 
1480 /**
1481  * memory_region_is_rom: check whether a memory region is ROM
1482  *
1483  * Returns %true if a memory region is read-only memory.
1484  *
1485  * @mr: the memory region being queried
1486  */
1487 static inline bool memory_region_is_rom(MemoryRegion *mr)
1488 {
1489     return mr->ram && mr->readonly;
1490 }
1491 
1492 /**
1493  * memory_region_is_nonvolatile: check whether a memory region is non-volatile
1494  *
1495  * Returns %true is a memory region is non-volatile memory.
1496  *
1497  * @mr: the memory region being queried
1498  */
1499 static inline bool memory_region_is_nonvolatile(MemoryRegion *mr)
1500 {
1501     return mr->nonvolatile;
1502 }
1503 
1504 /**
1505  * memory_region_get_fd: Get a file descriptor backing a RAM memory region.
1506  *
1507  * Returns a file descriptor backing a file-based RAM memory region,
1508  * or -1 if the region is not a file-based RAM memory region.
1509  *
1510  * @mr: the RAM or alias memory region being queried.
1511  */
1512 int memory_region_get_fd(MemoryRegion *mr);
1513 
1514 /**
1515  * memory_region_from_host: Convert a pointer into a RAM memory region
1516  * and an offset within it.
1517  *
1518  * Given a host pointer inside a RAM memory region (created with
1519  * memory_region_init_ram() or memory_region_init_ram_ptr()), return
1520  * the MemoryRegion and the offset within it.
1521  *
1522  * Use with care; by the time this function returns, the returned pointer is
1523  * not protected by RCU anymore.  If the caller is not within an RCU critical
1524  * section and does not hold the iothread lock, it must have other means of
1525  * protecting the pointer, such as a reference to the region that includes
1526  * the incoming ram_addr_t.
1527  *
1528  * @ptr: the host pointer to be converted
1529  * @offset: the offset within memory region
1530  */
1531 MemoryRegion *memory_region_from_host(void *ptr, ram_addr_t *offset);
1532 
1533 /**
1534  * memory_region_get_ram_ptr: Get a pointer into a RAM memory region.
1535  *
1536  * Returns a host pointer to a RAM memory region (created with
1537  * memory_region_init_ram() or memory_region_init_ram_ptr()).
1538  *
1539  * Use with care; by the time this function returns, the returned pointer is
1540  * not protected by RCU anymore.  If the caller is not within an RCU critical
1541  * section and does not hold the iothread lock, it must have other means of
1542  * protecting the pointer, such as a reference to the region that includes
1543  * the incoming ram_addr_t.
1544  *
1545  * @mr: the memory region being queried.
1546  */
1547 void *memory_region_get_ram_ptr(MemoryRegion *mr);
1548 
1549 /* memory_region_ram_resize: Resize a RAM region.
1550  *
1551  * Only legal before guest might have detected the memory size: e.g. on
1552  * incoming migration, or right after reset.
1553  *
1554  * @mr: a memory region created with @memory_region_init_resizeable_ram.
1555  * @newsize: the new size the region
1556  * @errp: pointer to Error*, to store an error if it happens.
1557  */
1558 void memory_region_ram_resize(MemoryRegion *mr, ram_addr_t newsize,
1559                               Error **errp);
1560 
1561 /**
1562  * memory_region_msync: Synchronize selected address range of
1563  * a memory mapped region
1564  *
1565  * @mr: the memory region to be msync
1566  * @addr: the initial address of the range to be sync
1567  * @size: the size of the range to be sync
1568  */
1569 void memory_region_msync(MemoryRegion *mr, hwaddr addr, hwaddr size);
1570 
1571 /**
1572  * memory_region_writeback: Trigger cache writeback for
1573  * selected address range
1574  *
1575  * @mr: the memory region to be updated
1576  * @addr: the initial address of the range to be written back
1577  * @size: the size of the range to be written back
1578  */
1579 void memory_region_writeback(MemoryRegion *mr, hwaddr addr, hwaddr size);
1580 
1581 /**
1582  * memory_region_set_log: Turn dirty logging on or off for a region.
1583  *
1584  * Turns dirty logging on or off for a specified client (display, migration).
1585  * Only meaningful for RAM regions.
1586  *
1587  * @mr: the memory region being updated.
1588  * @log: whether dirty logging is to be enabled or disabled.
1589  * @client: the user of the logging information; %DIRTY_MEMORY_VGA only.
1590  */
1591 void memory_region_set_log(MemoryRegion *mr, bool log, unsigned client);
1592 
1593 /**
1594  * memory_region_set_dirty: Mark a range of bytes as dirty in a memory region.
1595  *
1596  * Marks a range of bytes as dirty, after it has been dirtied outside
1597  * guest code.
1598  *
1599  * @mr: the memory region being dirtied.
1600  * @addr: the address (relative to the start of the region) being dirtied.
1601  * @size: size of the range being dirtied.
1602  */
1603 void memory_region_set_dirty(MemoryRegion *mr, hwaddr addr,
1604                              hwaddr size);
1605 
1606 /**
1607  * memory_region_clear_dirty_bitmap - clear dirty bitmap for memory range
1608  *
1609  * This function is called when the caller wants to clear the remote
1610  * dirty bitmap of a memory range within the memory region.  This can
1611  * be used by e.g. KVM to manually clear dirty log when
1612  * KVM_CAP_MANUAL_DIRTY_LOG_PROTECT is declared support by the host
1613  * kernel.
1614  *
1615  * @mr:     the memory region to clear the dirty log upon
1616  * @start:  start address offset within the memory region
1617  * @len:    length of the memory region to clear dirty bitmap
1618  */
1619 void memory_region_clear_dirty_bitmap(MemoryRegion *mr, hwaddr start,
1620                                       hwaddr len);
1621 
1622 /**
1623  * memory_region_snapshot_and_clear_dirty: Get a snapshot of the dirty
1624  *                                         bitmap and clear it.
1625  *
1626  * Creates a snapshot of the dirty bitmap, clears the dirty bitmap and
1627  * returns the snapshot.  The snapshot can then be used to query dirty
1628  * status, using memory_region_snapshot_get_dirty.  Snapshotting allows
1629  * querying the same page multiple times, which is especially useful for
1630  * display updates where the scanlines often are not page aligned.
1631  *
1632  * The dirty bitmap region which gets copyed into the snapshot (and
1633  * cleared afterwards) can be larger than requested.  The boundaries
1634  * are rounded up/down so complete bitmap longs (covering 64 pages on
1635  * 64bit hosts) can be copied over into the bitmap snapshot.  Which
1636  * isn't a problem for display updates as the extra pages are outside
1637  * the visible area, and in case the visible area changes a full
1638  * display redraw is due anyway.  Should other use cases for this
1639  * function emerge we might have to revisit this implementation
1640  * detail.
1641  *
1642  * Use g_free to release DirtyBitmapSnapshot.
1643  *
1644  * @mr: the memory region being queried.
1645  * @addr: the address (relative to the start of the region) being queried.
1646  * @size: the size of the range being queried.
1647  * @client: the user of the logging information; typically %DIRTY_MEMORY_VGA.
1648  */
1649 DirtyBitmapSnapshot *memory_region_snapshot_and_clear_dirty(MemoryRegion *mr,
1650                                                             hwaddr addr,
1651                                                             hwaddr size,
1652                                                             unsigned client);
1653 
1654 /**
1655  * memory_region_snapshot_get_dirty: Check whether a range of bytes is dirty
1656  *                                   in the specified dirty bitmap snapshot.
1657  *
1658  * @mr: the memory region being queried.
1659  * @snap: the dirty bitmap snapshot
1660  * @addr: the address (relative to the start of the region) being queried.
1661  * @size: the size of the range being queried.
1662  */
1663 bool memory_region_snapshot_get_dirty(MemoryRegion *mr,
1664                                       DirtyBitmapSnapshot *snap,
1665                                       hwaddr addr, hwaddr size);
1666 
1667 /**
1668  * memory_region_reset_dirty: Mark a range of pages as clean, for a specified
1669  *                            client.
1670  *
1671  * Marks a range of pages as no longer dirty.
1672  *
1673  * @mr: the region being updated.
1674  * @addr: the start of the subrange being cleaned.
1675  * @size: the size of the subrange being cleaned.
1676  * @client: the user of the logging information; %DIRTY_MEMORY_MIGRATION or
1677  *          %DIRTY_MEMORY_VGA.
1678  */
1679 void memory_region_reset_dirty(MemoryRegion *mr, hwaddr addr,
1680                                hwaddr size, unsigned client);
1681 
1682 /**
1683  * memory_region_flush_rom_device: Mark a range of pages dirty and invalidate
1684  *                                 TBs (for self-modifying code).
1685  *
1686  * The MemoryRegionOps->write() callback of a ROM device must use this function
1687  * to mark byte ranges that have been modified internally, such as by directly
1688  * accessing the memory returned by memory_region_get_ram_ptr().
1689  *
1690  * This function marks the range dirty and invalidates TBs so that TCG can
1691  * detect self-modifying code.
1692  *
1693  * @mr: the region being flushed.
1694  * @addr: the start, relative to the start of the region, of the range being
1695  *        flushed.
1696  * @size: the size, in bytes, of the range being flushed.
1697  */
1698 void memory_region_flush_rom_device(MemoryRegion *mr, hwaddr addr, hwaddr size);
1699 
1700 /**
1701  * memory_region_set_readonly: Turn a memory region read-only (or read-write)
1702  *
1703  * Allows a memory region to be marked as read-only (turning it into a ROM).
1704  * only useful on RAM regions.
1705  *
1706  * @mr: the region being updated.
1707  * @readonly: whether rhe region is to be ROM or RAM.
1708  */
1709 void memory_region_set_readonly(MemoryRegion *mr, bool readonly);
1710 
1711 /**
1712  * memory_region_set_nonvolatile: Turn a memory region non-volatile
1713  *
1714  * Allows a memory region to be marked as non-volatile.
1715  * only useful on RAM regions.
1716  *
1717  * @mr: the region being updated.
1718  * @nonvolatile: whether rhe region is to be non-volatile.
1719  */
1720 void memory_region_set_nonvolatile(MemoryRegion *mr, bool nonvolatile);
1721 
1722 /**
1723  * memory_region_rom_device_set_romd: enable/disable ROMD mode
1724  *
1725  * Allows a ROM device (initialized with memory_region_init_rom_device() to
1726  * set to ROMD mode (default) or MMIO mode.  When it is in ROMD mode, the
1727  * device is mapped to guest memory and satisfies read access directly.
1728  * When in MMIO mode, reads are forwarded to the #MemoryRegion.read function.
1729  * Writes are always handled by the #MemoryRegion.write function.
1730  *
1731  * @mr: the memory region to be updated
1732  * @romd_mode: %true to put the region into ROMD mode
1733  */
1734 void memory_region_rom_device_set_romd(MemoryRegion *mr, bool romd_mode);
1735 
1736 /**
1737  * memory_region_set_coalescing: Enable memory coalescing for the region.
1738  *
1739  * Enabled writes to a region to be queued for later processing. MMIO ->write
1740  * callbacks may be delayed until a non-coalesced MMIO is issued.
1741  * Only useful for IO regions.  Roughly similar to write-combining hardware.
1742  *
1743  * @mr: the memory region to be write coalesced
1744  */
1745 void memory_region_set_coalescing(MemoryRegion *mr);
1746 
1747 /**
1748  * memory_region_add_coalescing: Enable memory coalescing for a sub-range of
1749  *                               a region.
1750  *
1751  * Like memory_region_set_coalescing(), but works on a sub-range of a region.
1752  * Multiple calls can be issued coalesced disjoint ranges.
1753  *
1754  * @mr: the memory region to be updated.
1755  * @offset: the start of the range within the region to be coalesced.
1756  * @size: the size of the subrange to be coalesced.
1757  */
1758 void memory_region_add_coalescing(MemoryRegion *mr,
1759                                   hwaddr offset,
1760                                   uint64_t size);
1761 
1762 /**
1763  * memory_region_clear_coalescing: Disable MMIO coalescing for the region.
1764  *
1765  * Disables any coalescing caused by memory_region_set_coalescing() or
1766  * memory_region_add_coalescing().  Roughly equivalent to uncacheble memory
1767  * hardware.
1768  *
1769  * @mr: the memory region to be updated.
1770  */
1771 void memory_region_clear_coalescing(MemoryRegion *mr);
1772 
1773 /**
1774  * memory_region_set_flush_coalesced: Enforce memory coalescing flush before
1775  *                                    accesses.
1776  *
1777  * Ensure that pending coalesced MMIO request are flushed before the memory
1778  * region is accessed. This property is automatically enabled for all regions
1779  * passed to memory_region_set_coalescing() and memory_region_add_coalescing().
1780  *
1781  * @mr: the memory region to be updated.
1782  */
1783 void memory_region_set_flush_coalesced(MemoryRegion *mr);
1784 
1785 /**
1786  * memory_region_clear_flush_coalesced: Disable memory coalescing flush before
1787  *                                      accesses.
1788  *
1789  * Clear the automatic coalesced MMIO flushing enabled via
1790  * memory_region_set_flush_coalesced. Note that this service has no effect on
1791  * memory regions that have MMIO coalescing enabled for themselves. For them,
1792  * automatic flushing will stop once coalescing is disabled.
1793  *
1794  * @mr: the memory region to be updated.
1795  */
1796 void memory_region_clear_flush_coalesced(MemoryRegion *mr);
1797 
1798 /**
1799  * memory_region_add_eventfd: Request an eventfd to be triggered when a word
1800  *                            is written to a location.
1801  *
1802  * Marks a word in an IO region (initialized with memory_region_init_io())
1803  * as a trigger for an eventfd event.  The I/O callback will not be called.
1804  * The caller must be prepared to handle failure (that is, take the required
1805  * action if the callback _is_ called).
1806  *
1807  * @mr: the memory region being updated.
1808  * @addr: the address within @mr that is to be monitored
1809  * @size: the size of the access to trigger the eventfd
1810  * @match_data: whether to match against @data, instead of just @addr
1811  * @data: the data to match against the guest write
1812  * @e: event notifier to be triggered when @addr, @size, and @data all match.
1813  **/
1814 void memory_region_add_eventfd(MemoryRegion *mr,
1815                                hwaddr addr,
1816                                unsigned size,
1817                                bool match_data,
1818                                uint64_t data,
1819                                EventNotifier *e);
1820 
1821 /**
1822  * memory_region_del_eventfd: Cancel an eventfd.
1823  *
1824  * Cancels an eventfd trigger requested by a previous
1825  * memory_region_add_eventfd() call.
1826  *
1827  * @mr: the memory region being updated.
1828  * @addr: the address within @mr that is to be monitored
1829  * @size: the size of the access to trigger the eventfd
1830  * @match_data: whether to match against @data, instead of just @addr
1831  * @data: the data to match against the guest write
1832  * @e: event notifier to be triggered when @addr, @size, and @data all match.
1833  */
1834 void memory_region_del_eventfd(MemoryRegion *mr,
1835                                hwaddr addr,
1836                                unsigned size,
1837                                bool match_data,
1838                                uint64_t data,
1839                                EventNotifier *e);
1840 
1841 /**
1842  * memory_region_add_subregion: Add a subregion to a container.
1843  *
1844  * Adds a subregion at @offset.  The subregion may not overlap with other
1845  * subregions (except for those explicitly marked as overlapping).  A region
1846  * may only be added once as a subregion (unless removed with
1847  * memory_region_del_subregion()); use memory_region_init_alias() if you
1848  * want a region to be a subregion in multiple locations.
1849  *
1850  * @mr: the region to contain the new subregion; must be a container
1851  *      initialized with memory_region_init().
1852  * @offset: the offset relative to @mr where @subregion is added.
1853  * @subregion: the subregion to be added.
1854  */
1855 void memory_region_add_subregion(MemoryRegion *mr,
1856                                  hwaddr offset,
1857                                  MemoryRegion *subregion);
1858 /**
1859  * memory_region_add_subregion_overlap: Add a subregion to a container
1860  *                                      with overlap.
1861  *
1862  * Adds a subregion at @offset.  The subregion may overlap with other
1863  * subregions.  Conflicts are resolved by having a higher @priority hide a
1864  * lower @priority. Subregions without priority are taken as @priority 0.
1865  * A region may only be added once as a subregion (unless removed with
1866  * memory_region_del_subregion()); use memory_region_init_alias() if you
1867  * want a region to be a subregion in multiple locations.
1868  *
1869  * @mr: the region to contain the new subregion; must be a container
1870  *      initialized with memory_region_init().
1871  * @offset: the offset relative to @mr where @subregion is added.
1872  * @subregion: the subregion to be added.
1873  * @priority: used for resolving overlaps; highest priority wins.
1874  */
1875 void memory_region_add_subregion_overlap(MemoryRegion *mr,
1876                                          hwaddr offset,
1877                                          MemoryRegion *subregion,
1878                                          int priority);
1879 
1880 /**
1881  * memory_region_get_ram_addr: Get the ram address associated with a memory
1882  *                             region
1883  *
1884  * @mr: the region to be queried
1885  */
1886 ram_addr_t memory_region_get_ram_addr(MemoryRegion *mr);
1887 
1888 uint64_t memory_region_get_alignment(const MemoryRegion *mr);
1889 /**
1890  * memory_region_del_subregion: Remove a subregion.
1891  *
1892  * Removes a subregion from its container.
1893  *
1894  * @mr: the container to be updated.
1895  * @subregion: the region being removed; must be a current subregion of @mr.
1896  */
1897 void memory_region_del_subregion(MemoryRegion *mr,
1898                                  MemoryRegion *subregion);
1899 
1900 /*
1901  * memory_region_set_enabled: dynamically enable or disable a region
1902  *
1903  * Enables or disables a memory region.  A disabled memory region
1904  * ignores all accesses to itself and its subregions.  It does not
1905  * obscure sibling subregions with lower priority - it simply behaves as
1906  * if it was removed from the hierarchy.
1907  *
1908  * Regions default to being enabled.
1909  *
1910  * @mr: the region to be updated
1911  * @enabled: whether to enable or disable the region
1912  */
1913 void memory_region_set_enabled(MemoryRegion *mr, bool enabled);
1914 
1915 /*
1916  * memory_region_set_address: dynamically update the address of a region
1917  *
1918  * Dynamically updates the address of a region, relative to its container.
1919  * May be used on regions are currently part of a memory hierarchy.
1920  *
1921  * @mr: the region to be updated
1922  * @addr: new address, relative to container region
1923  */
1924 void memory_region_set_address(MemoryRegion *mr, hwaddr addr);
1925 
1926 /*
1927  * memory_region_set_size: dynamically update the size of a region.
1928  *
1929  * Dynamically updates the size of a region.
1930  *
1931  * @mr: the region to be updated
1932  * @size: used size of the region.
1933  */
1934 void memory_region_set_size(MemoryRegion *mr, uint64_t size);
1935 
1936 /*
1937  * memory_region_set_alias_offset: dynamically update a memory alias's offset
1938  *
1939  * Dynamically updates the offset into the target region that an alias points
1940  * to, as if the fourth argument to memory_region_init_alias() has changed.
1941  *
1942  * @mr: the #MemoryRegion to be updated; should be an alias.
1943  * @offset: the new offset into the target memory region
1944  */
1945 void memory_region_set_alias_offset(MemoryRegion *mr,
1946                                     hwaddr offset);
1947 
1948 /**
1949  * memory_region_present: checks if an address relative to a @container
1950  * translates into #MemoryRegion within @container
1951  *
1952  * Answer whether a #MemoryRegion within @container covers the address
1953  * @addr.
1954  *
1955  * @container: a #MemoryRegion within which @addr is a relative address
1956  * @addr: the area within @container to be searched
1957  */
1958 bool memory_region_present(MemoryRegion *container, hwaddr addr);
1959 
1960 /**
1961  * memory_region_is_mapped: returns true if #MemoryRegion is mapped
1962  * into any address space.
1963  *
1964  * @mr: a #MemoryRegion which should be checked if it's mapped
1965  */
1966 bool memory_region_is_mapped(MemoryRegion *mr);
1967 
1968 /**
1969  * memory_region_find: translate an address/size relative to a
1970  * MemoryRegion into a #MemoryRegionSection.
1971  *
1972  * Locates the first #MemoryRegion within @mr that overlaps the range
1973  * given by @addr and @size.
1974  *
1975  * Returns a #MemoryRegionSection that describes a contiguous overlap.
1976  * It will have the following characteristics:
1977  * - @size = 0 iff no overlap was found
1978  * - @mr is non-%NULL iff an overlap was found
1979  *
1980  * Remember that in the return value the @offset_within_region is
1981  * relative to the returned region (in the .@mr field), not to the
1982  * @mr argument.
1983  *
1984  * Similarly, the .@offset_within_address_space is relative to the
1985  * address space that contains both regions, the passed and the
1986  * returned one.  However, in the special case where the @mr argument
1987  * has no container (and thus is the root of the address space), the
1988  * following will hold:
1989  * - @offset_within_address_space >= @addr
1990  * - @offset_within_address_space + .@size <= @addr + @size
1991  *
1992  * @mr: a MemoryRegion within which @addr is a relative address
1993  * @addr: start of the area within @as to be searched
1994  * @size: size of the area to be searched
1995  */
1996 MemoryRegionSection memory_region_find(MemoryRegion *mr,
1997                                        hwaddr addr, uint64_t size);
1998 
1999 /**
2000  * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2001  *
2002  * Synchronizes the dirty page log for all address spaces.
2003  */
2004 void memory_global_dirty_log_sync(void);
2005 
2006 /**
2007  * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2008  *
2009  * Synchronizes the vCPUs with a thread that is reading the dirty bitmap.
2010  * This function must be called after the dirty log bitmap is cleared, and
2011  * before dirty guest memory pages are read.  If you are using
2012  * #DirtyBitmapSnapshot, memory_region_snapshot_and_clear_dirty() takes
2013  * care of doing this.
2014  */
2015 void memory_global_after_dirty_log_sync(void);
2016 
2017 /**
2018  * memory_region_transaction_begin: Start a transaction.
2019  *
2020  * During a transaction, changes will be accumulated and made visible
2021  * only when the transaction ends (is committed).
2022  */
2023 void memory_region_transaction_begin(void);
2024 
2025 /**
2026  * memory_region_transaction_commit: Commit a transaction and make changes
2027  *                                   visible to the guest.
2028  */
2029 void memory_region_transaction_commit(void);
2030 
2031 /**
2032  * memory_listener_register: register callbacks to be called when memory
2033  *                           sections are mapped or unmapped into an address
2034  *                           space
2035  *
2036  * @listener: an object containing the callbacks to be called
2037  * @filter: if non-%NULL, only regions in this address space will be observed
2038  */
2039 void memory_listener_register(MemoryListener *listener, AddressSpace *filter);
2040 
2041 /**
2042  * memory_listener_unregister: undo the effect of memory_listener_register()
2043  *
2044  * @listener: an object containing the callbacks to be removed
2045  */
2046 void memory_listener_unregister(MemoryListener *listener);
2047 
2048 /**
2049  * memory_global_dirty_log_start: begin dirty logging for all regions
2050  */
2051 void memory_global_dirty_log_start(void);
2052 
2053 /**
2054  * memory_global_dirty_log_stop: end dirty logging for all regions
2055  */
2056 void memory_global_dirty_log_stop(void);
2057 
2058 void mtree_info(bool flatview, bool dispatch_tree, bool owner, bool disabled);
2059 
2060 /**
2061  * memory_region_dispatch_read: perform a read directly to the specified
2062  * MemoryRegion.
2063  *
2064  * @mr: #MemoryRegion to access
2065  * @addr: address within that region
2066  * @pval: pointer to uint64_t which the data is written to
2067  * @op: size, sign, and endianness of the memory operation
2068  * @attrs: memory transaction attributes to use for the access
2069  */
2070 MemTxResult memory_region_dispatch_read(MemoryRegion *mr,
2071                                         hwaddr addr,
2072                                         uint64_t *pval,
2073                                         MemOp op,
2074                                         MemTxAttrs attrs);
2075 /**
2076  * memory_region_dispatch_write: perform a write directly to the specified
2077  * MemoryRegion.
2078  *
2079  * @mr: #MemoryRegion to access
2080  * @addr: address within that region
2081  * @data: data to write
2082  * @op: size, sign, and endianness of the memory operation
2083  * @attrs: memory transaction attributes to use for the access
2084  */
2085 MemTxResult memory_region_dispatch_write(MemoryRegion *mr,
2086                                          hwaddr addr,
2087                                          uint64_t data,
2088                                          MemOp op,
2089                                          MemTxAttrs attrs);
2090 
2091 /**
2092  * address_space_init: initializes an address space
2093  *
2094  * @as: an uninitialized #AddressSpace
2095  * @root: a #MemoryRegion that routes addresses for the address space
2096  * @name: an address space name.  The name is only used for debugging
2097  *        output.
2098  */
2099 void address_space_init(AddressSpace *as, MemoryRegion *root, const char *name);
2100 
2101 /**
2102  * address_space_destroy: destroy an address space
2103  *
2104  * Releases all resources associated with an address space.  After an address space
2105  * is destroyed, its root memory region (given by address_space_init()) may be destroyed
2106  * as well.
2107  *
2108  * @as: address space to be destroyed
2109  */
2110 void address_space_destroy(AddressSpace *as);
2111 
2112 /**
2113  * address_space_remove_listeners: unregister all listeners of an address space
2114  *
2115  * Removes all callbacks previously registered with memory_listener_register()
2116  * for @as.
2117  *
2118  * @as: an initialized #AddressSpace
2119  */
2120 void address_space_remove_listeners(AddressSpace *as);
2121 
2122 /**
2123  * address_space_rw: read from or write to an address space.
2124  *
2125  * Return a MemTxResult indicating whether the operation succeeded
2126  * or failed (eg unassigned memory, device rejected the transaction,
2127  * IOMMU fault).
2128  *
2129  * @as: #AddressSpace to be accessed
2130  * @addr: address within that address space
2131  * @attrs: memory transaction attributes
2132  * @buf: buffer with the data transferred
2133  * @len: the number of bytes to read or write
2134  * @is_write: indicates the transfer direction
2135  */
2136 MemTxResult address_space_rw(AddressSpace *as, hwaddr addr,
2137                              MemTxAttrs attrs, void *buf,
2138                              hwaddr len, bool is_write);
2139 
2140 /**
2141  * address_space_write: write to address space.
2142  *
2143  * Return a MemTxResult indicating whether the operation succeeded
2144  * or failed (eg unassigned memory, device rejected the transaction,
2145  * IOMMU fault).
2146  *
2147  * @as: #AddressSpace to be accessed
2148  * @addr: address within that address space
2149  * @attrs: memory transaction attributes
2150  * @buf: buffer with the data transferred
2151  * @len: the number of bytes to write
2152  */
2153 MemTxResult address_space_write(AddressSpace *as, hwaddr addr,
2154                                 MemTxAttrs attrs,
2155                                 const void *buf, hwaddr len);
2156 
2157 /**
2158  * address_space_write_rom: write to address space, including ROM.
2159  *
2160  * This function writes to the specified address space, but will
2161  * write data to both ROM and RAM. This is used for non-guest
2162  * writes like writes from the gdb debug stub or initial loading
2163  * of ROM contents.
2164  *
2165  * Note that portions of the write which attempt to write data to
2166  * a device will be silently ignored -- only real RAM and ROM will
2167  * be written to.
2168  *
2169  * Return a MemTxResult indicating whether the operation succeeded
2170  * or failed (eg unassigned memory, device rejected the transaction,
2171  * IOMMU fault).
2172  *
2173  * @as: #AddressSpace to be accessed
2174  * @addr: address within that address space
2175  * @attrs: memory transaction attributes
2176  * @buf: buffer with the data transferred
2177  * @len: the number of bytes to write
2178  */
2179 MemTxResult address_space_write_rom(AddressSpace *as, hwaddr addr,
2180                                     MemTxAttrs attrs,
2181                                     const void *buf, hwaddr len);
2182 
2183 /* address_space_ld*: load from an address space
2184  * address_space_st*: store to an address space
2185  *
2186  * These functions perform a load or store of the byte, word,
2187  * longword or quad to the specified address within the AddressSpace.
2188  * The _le suffixed functions treat the data as little endian;
2189  * _be indicates big endian; no suffix indicates "same endianness
2190  * as guest CPU".
2191  *
2192  * The "guest CPU endianness" accessors are deprecated for use outside
2193  * target-* code; devices should be CPU-agnostic and use either the LE
2194  * or the BE accessors.
2195  *
2196  * @as #AddressSpace to be accessed
2197  * @addr: address within that address space
2198  * @val: data value, for stores
2199  * @attrs: memory transaction attributes
2200  * @result: location to write the success/failure of the transaction;
2201  *   if NULL, this information is discarded
2202  */
2203 
2204 #define SUFFIX
2205 #define ARG1         as
2206 #define ARG1_DECL    AddressSpace *as
2207 #include "exec/memory_ldst.h.inc"
2208 
2209 #define SUFFIX
2210 #define ARG1         as
2211 #define ARG1_DECL    AddressSpace *as
2212 #include "exec/memory_ldst_phys.h.inc"
2213 
2214 struct MemoryRegionCache {
2215     void *ptr;
2216     hwaddr xlat;
2217     hwaddr len;
2218     FlatView *fv;
2219     MemoryRegionSection mrs;
2220     bool is_write;
2221 };
2222 
2223 #define MEMORY_REGION_CACHE_INVALID ((MemoryRegionCache) { .mrs.mr = NULL })
2224 
2225 
2226 /* address_space_ld*_cached: load from a cached #MemoryRegion
2227  * address_space_st*_cached: store into a cached #MemoryRegion
2228  *
2229  * These functions perform a load or store of the byte, word,
2230  * longword or quad to the specified address.  The address is
2231  * a physical address in the AddressSpace, but it must lie within
2232  * a #MemoryRegion that was mapped with address_space_cache_init.
2233  *
2234  * The _le suffixed functions treat the data as little endian;
2235  * _be indicates big endian; no suffix indicates "same endianness
2236  * as guest CPU".
2237  *
2238  * The "guest CPU endianness" accessors are deprecated for use outside
2239  * target-* code; devices should be CPU-agnostic and use either the LE
2240  * or the BE accessors.
2241  *
2242  * @cache: previously initialized #MemoryRegionCache to be accessed
2243  * @addr: address within the address space
2244  * @val: data value, for stores
2245  * @attrs: memory transaction attributes
2246  * @result: location to write the success/failure of the transaction;
2247  *   if NULL, this information is discarded
2248  */
2249 
2250 #define SUFFIX       _cached_slow
2251 #define ARG1         cache
2252 #define ARG1_DECL    MemoryRegionCache *cache
2253 #include "exec/memory_ldst.h.inc"
2254 
2255 /* Inline fast path for direct RAM access.  */
2256 static inline uint8_t address_space_ldub_cached(MemoryRegionCache *cache,
2257     hwaddr addr, MemTxAttrs attrs, MemTxResult *result)
2258 {
2259     assert(addr < cache->len);
2260     if (likely(cache->ptr)) {
2261         return ldub_p(cache->ptr + addr);
2262     } else {
2263         return address_space_ldub_cached_slow(cache, addr, attrs, result);
2264     }
2265 }
2266 
2267 static inline void address_space_stb_cached(MemoryRegionCache *cache,
2268     hwaddr addr, uint32_t val, MemTxAttrs attrs, MemTxResult *result)
2269 {
2270     assert(addr < cache->len);
2271     if (likely(cache->ptr)) {
2272         stb_p(cache->ptr + addr, val);
2273     } else {
2274         address_space_stb_cached_slow(cache, addr, val, attrs, result);
2275     }
2276 }
2277 
2278 #define ENDIANNESS   _le
2279 #include "exec/memory_ldst_cached.h.inc"
2280 
2281 #define ENDIANNESS   _be
2282 #include "exec/memory_ldst_cached.h.inc"
2283 
2284 #define SUFFIX       _cached
2285 #define ARG1         cache
2286 #define ARG1_DECL    MemoryRegionCache *cache
2287 #include "exec/memory_ldst_phys.h.inc"
2288 
2289 /* address_space_cache_init: prepare for repeated access to a physical
2290  * memory region
2291  *
2292  * @cache: #MemoryRegionCache to be filled
2293  * @as: #AddressSpace to be accessed
2294  * @addr: address within that address space
2295  * @len: length of buffer
2296  * @is_write: indicates the transfer direction
2297  *
2298  * Will only work with RAM, and may map a subset of the requested range by
2299  * returning a value that is less than @len.  On failure, return a negative
2300  * errno value.
2301  *
2302  * Because it only works with RAM, this function can be used for
2303  * read-modify-write operations.  In this case, is_write should be %true.
2304  *
2305  * Note that addresses passed to the address_space_*_cached functions
2306  * are relative to @addr.
2307  */
2308 int64_t address_space_cache_init(MemoryRegionCache *cache,
2309                                  AddressSpace *as,
2310                                  hwaddr addr,
2311                                  hwaddr len,
2312                                  bool is_write);
2313 
2314 /**
2315  * address_space_cache_invalidate: complete a write to a #MemoryRegionCache
2316  *
2317  * @cache: The #MemoryRegionCache to operate on.
2318  * @addr: The first physical address that was written, relative to the
2319  * address that was passed to @address_space_cache_init.
2320  * @access_len: The number of bytes that were written starting at @addr.
2321  */
2322 void address_space_cache_invalidate(MemoryRegionCache *cache,
2323                                     hwaddr addr,
2324                                     hwaddr access_len);
2325 
2326 /**
2327  * address_space_cache_destroy: free a #MemoryRegionCache
2328  *
2329  * @cache: The #MemoryRegionCache whose memory should be released.
2330  */
2331 void address_space_cache_destroy(MemoryRegionCache *cache);
2332 
2333 /* address_space_get_iotlb_entry: translate an address into an IOTLB
2334  * entry. Should be called from an RCU critical section.
2335  */
2336 IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr,
2337                                             bool is_write, MemTxAttrs attrs);
2338 
2339 /* address_space_translate: translate an address range into an address space
2340  * into a MemoryRegion and an address range into that section.  Should be
2341  * called from an RCU critical section, to avoid that the last reference
2342  * to the returned region disappears after address_space_translate returns.
2343  *
2344  * @fv: #FlatView to be accessed
2345  * @addr: address within that address space
2346  * @xlat: pointer to address within the returned memory region section's
2347  * #MemoryRegion.
2348  * @len: pointer to length
2349  * @is_write: indicates the transfer direction
2350  * @attrs: memory attributes
2351  */
2352 MemoryRegion *flatview_translate(FlatView *fv,
2353                                  hwaddr addr, hwaddr *xlat,
2354                                  hwaddr *len, bool is_write,
2355                                  MemTxAttrs attrs);
2356 
2357 static inline MemoryRegion *address_space_translate(AddressSpace *as,
2358                                                     hwaddr addr, hwaddr *xlat,
2359                                                     hwaddr *len, bool is_write,
2360                                                     MemTxAttrs attrs)
2361 {
2362     return flatview_translate(address_space_to_flatview(as),
2363                               addr, xlat, len, is_write, attrs);
2364 }
2365 
2366 /* address_space_access_valid: check for validity of accessing an address
2367  * space range
2368  *
2369  * Check whether memory is assigned to the given address space range, and
2370  * access is permitted by any IOMMU regions that are active for the address
2371  * space.
2372  *
2373  * For now, addr and len should be aligned to a page size.  This limitation
2374  * will be lifted in the future.
2375  *
2376  * @as: #AddressSpace to be accessed
2377  * @addr: address within that address space
2378  * @len: length of the area to be checked
2379  * @is_write: indicates the transfer direction
2380  * @attrs: memory attributes
2381  */
2382 bool address_space_access_valid(AddressSpace *as, hwaddr addr, hwaddr len,
2383                                 bool is_write, MemTxAttrs attrs);
2384 
2385 /* address_space_map: map a physical memory region into a host virtual address
2386  *
2387  * May map a subset of the requested range, given by and returned in @plen.
2388  * May return %NULL and set *@plen to zero(0), if resources needed to perform
2389  * the mapping are exhausted.
2390  * Use only for reads OR writes - not for read-modify-write operations.
2391  * Use cpu_register_map_client() to know when retrying the map operation is
2392  * likely to succeed.
2393  *
2394  * @as: #AddressSpace to be accessed
2395  * @addr: address within that address space
2396  * @plen: pointer to length of buffer; updated on return
2397  * @is_write: indicates the transfer direction
2398  * @attrs: memory attributes
2399  */
2400 void *address_space_map(AddressSpace *as, hwaddr addr,
2401                         hwaddr *plen, bool is_write, MemTxAttrs attrs);
2402 
2403 /* address_space_unmap: Unmaps a memory region previously mapped by address_space_map()
2404  *
2405  * Will also mark the memory as dirty if @is_write == %true.  @access_len gives
2406  * the amount of memory that was actually read or written by the caller.
2407  *
2408  * @as: #AddressSpace used
2409  * @buffer: host pointer as returned by address_space_map()
2410  * @len: buffer length as returned by address_space_map()
2411  * @access_len: amount of data actually transferred
2412  * @is_write: indicates the transfer direction
2413  */
2414 void address_space_unmap(AddressSpace *as, void *buffer, hwaddr len,
2415                          bool is_write, hwaddr access_len);
2416 
2417 
2418 /* Internal functions, part of the implementation of address_space_read.  */
2419 MemTxResult address_space_read_full(AddressSpace *as, hwaddr addr,
2420                                     MemTxAttrs attrs, void *buf, hwaddr len);
2421 MemTxResult flatview_read_continue(FlatView *fv, hwaddr addr,
2422                                    MemTxAttrs attrs, void *buf,
2423                                    hwaddr len, hwaddr addr1, hwaddr l,
2424                                    MemoryRegion *mr);
2425 void *qemu_map_ram_ptr(RAMBlock *ram_block, ram_addr_t addr);
2426 
2427 /* Internal functions, part of the implementation of address_space_read_cached
2428  * and address_space_write_cached.  */
2429 MemTxResult address_space_read_cached_slow(MemoryRegionCache *cache,
2430                                            hwaddr addr, void *buf, hwaddr len);
2431 MemTxResult address_space_write_cached_slow(MemoryRegionCache *cache,
2432                                             hwaddr addr, const void *buf,
2433                                             hwaddr len);
2434 
2435 static inline bool memory_access_is_direct(MemoryRegion *mr, bool is_write)
2436 {
2437     if (is_write) {
2438         return memory_region_is_ram(mr) && !mr->readonly &&
2439                !mr->rom_device && !memory_region_is_ram_device(mr);
2440     } else {
2441         return (memory_region_is_ram(mr) && !memory_region_is_ram_device(mr)) ||
2442                memory_region_is_romd(mr);
2443     }
2444 }
2445 
2446 /**
2447  * address_space_read: read from an address space.
2448  *
2449  * Return a MemTxResult indicating whether the operation succeeded
2450  * or failed (eg unassigned memory, device rejected the transaction,
2451  * IOMMU fault).  Called within RCU critical section.
2452  *
2453  * @as: #AddressSpace to be accessed
2454  * @addr: address within that address space
2455  * @attrs: memory transaction attributes
2456  * @buf: buffer with the data transferred
2457  * @len: length of the data transferred
2458  */
2459 static inline __attribute__((__always_inline__))
2460 MemTxResult address_space_read(AddressSpace *as, hwaddr addr,
2461                                MemTxAttrs attrs, void *buf,
2462                                hwaddr len)
2463 {
2464     MemTxResult result = MEMTX_OK;
2465     hwaddr l, addr1;
2466     void *ptr;
2467     MemoryRegion *mr;
2468     FlatView *fv;
2469 
2470     if (__builtin_constant_p(len)) {
2471         if (len) {
2472             RCU_READ_LOCK_GUARD();
2473             fv = address_space_to_flatview(as);
2474             l = len;
2475             mr = flatview_translate(fv, addr, &addr1, &l, false, attrs);
2476             if (len == l && memory_access_is_direct(mr, false)) {
2477                 ptr = qemu_map_ram_ptr(mr->ram_block, addr1);
2478                 memcpy(buf, ptr, len);
2479             } else {
2480                 result = flatview_read_continue(fv, addr, attrs, buf, len,
2481                                                 addr1, l, mr);
2482             }
2483         }
2484     } else {
2485         result = address_space_read_full(as, addr, attrs, buf, len);
2486     }
2487     return result;
2488 }
2489 
2490 /**
2491  * address_space_read_cached: read from a cached RAM region
2492  *
2493  * @cache: Cached region to be addressed
2494  * @addr: address relative to the base of the RAM region
2495  * @buf: buffer with the data transferred
2496  * @len: length of the data transferred
2497  */
2498 static inline MemTxResult
2499 address_space_read_cached(MemoryRegionCache *cache, hwaddr addr,
2500                           void *buf, hwaddr len)
2501 {
2502     assert(addr < cache->len && len <= cache->len - addr);
2503     fuzz_dma_read_cb(cache->xlat + addr, len, cache->mrs.mr, false);
2504     if (likely(cache->ptr)) {
2505         memcpy(buf, cache->ptr + addr, len);
2506         return MEMTX_OK;
2507     } else {
2508         return address_space_read_cached_slow(cache, addr, buf, len);
2509     }
2510 }
2511 
2512 /**
2513  * address_space_write_cached: write to a cached RAM region
2514  *
2515  * @cache: Cached region to be addressed
2516  * @addr: address relative to the base of the RAM region
2517  * @buf: buffer with the data transferred
2518  * @len: length of the data transferred
2519  */
2520 static inline MemTxResult
2521 address_space_write_cached(MemoryRegionCache *cache, hwaddr addr,
2522                            const void *buf, hwaddr len)
2523 {
2524     assert(addr < cache->len && len <= cache->len - addr);
2525     if (likely(cache->ptr)) {
2526         memcpy(cache->ptr + addr, buf, len);
2527         return MEMTX_OK;
2528     } else {
2529         return address_space_write_cached_slow(cache, addr, buf, len);
2530     }
2531 }
2532 
2533 #ifdef NEED_CPU_H
2534 /* enum device_endian to MemOp.  */
2535 static inline MemOp devend_memop(enum device_endian end)
2536 {
2537     QEMU_BUILD_BUG_ON(DEVICE_HOST_ENDIAN != DEVICE_LITTLE_ENDIAN &&
2538                       DEVICE_HOST_ENDIAN != DEVICE_BIG_ENDIAN);
2539 
2540 #if defined(HOST_WORDS_BIGENDIAN) != defined(TARGET_WORDS_BIGENDIAN)
2541     /* Swap if non-host endianness or native (target) endianness */
2542     return (end == DEVICE_HOST_ENDIAN) ? 0 : MO_BSWAP;
2543 #else
2544     const int non_host_endianness =
2545         DEVICE_LITTLE_ENDIAN ^ DEVICE_BIG_ENDIAN ^ DEVICE_HOST_ENDIAN;
2546 
2547     /* In this case, native (target) endianness needs no swap.  */
2548     return (end == non_host_endianness) ? MO_BSWAP : 0;
2549 #endif
2550 }
2551 #endif
2552 
2553 /*
2554  * Inhibit technologies that require discarding of pages in RAM blocks, e.g.,
2555  * to manage the actual amount of memory consumed by the VM (then, the memory
2556  * provided by RAM blocks might be bigger than the desired memory consumption).
2557  * This *must* be set if:
2558  * - Discarding parts of a RAM blocks does not result in the change being
2559  *   reflected in the VM and the pages getting freed.
2560  * - All memory in RAM blocks is pinned or duplicated, invaldiating any previous
2561  *   discards blindly.
2562  * - Discarding parts of a RAM blocks will result in integrity issues (e.g.,
2563  *   encrypted VMs).
2564  * Technologies that only temporarily pin the current working set of a
2565  * driver are fine, because we don't expect such pages to be discarded
2566  * (esp. based on guest action like balloon inflation).
2567  *
2568  * This is *not* to be used to protect from concurrent discards (esp.,
2569  * postcopy).
2570  *
2571  * Returns 0 if successful. Returns -EBUSY if a technology that relies on
2572  * discards to work reliably is active.
2573  */
2574 int ram_block_discard_disable(bool state);
2575 
2576 /*
2577  * Inhibit technologies that disable discarding of pages in RAM blocks.
2578  *
2579  * Returns 0 if successful. Returns -EBUSY if discards are already set to
2580  * broken.
2581  */
2582 int ram_block_discard_require(bool state);
2583 
2584 /*
2585  * Test if discarding of memory in ram blocks is disabled.
2586  */
2587 bool ram_block_discard_is_disabled(void);
2588 
2589 /*
2590  * Test if discarding of memory in ram blocks is required to work reliably.
2591  */
2592 bool ram_block_discard_is_required(void);
2593 
2594 #endif
2595 
2596 #endif
2597