2 * Physical memory management API
4 * Copyright 2011 Red Hat, Inc. and/or its affiliates
7 * Avi Kivity <avi@redhat.com>
9 * This work is licensed under the terms of the GNU GPL, version 2. See
10 * the COPYING file in the top-level directory.
17 #ifndef CONFIG_USER_ONLY
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"
31 #define RAM_ADDR_INVALID (~(ram_addr_t)0)
33 #define MAX_PHYS_ADDR_SPACE_BITS 62
34 #define MAX_PHYS_ADDR (((hwaddr)1 << MAX_PHYS_ADDR_SPACE_BITS) - 1)
36 #define TYPE_MEMORY_REGION "memory-region"
37 DECLARE_INSTANCE_CHECKER(MemoryRegion
, MEMORY_REGION
,
40 #define TYPE_IOMMU_MEMORY_REGION "iommu-memory-region"
41 typedef struct IOMMUMemoryRegionClass IOMMUMemoryRegionClass
;
42 DECLARE_OBJ_CHECKERS(IOMMUMemoryRegion
, IOMMUMemoryRegionClass
,
43 IOMMU_MEMORY_REGION
, TYPE_IOMMU_MEMORY_REGION
)
45 #define TYPE_RAM_DISCARD_MANAGER "qemu:ram-discard-manager"
46 typedef struct RamDiscardManagerClass RamDiscardManagerClass
;
47 typedef struct RamDiscardManager RamDiscardManager
;
48 DECLARE_OBJ_CHECKERS(RamDiscardManager
, RamDiscardManagerClass
,
49 RAM_DISCARD_MANAGER
, TYPE_RAM_DISCARD_MANAGER
);
52 void fuzz_dma_read_cb(size_t addr
,
56 static inline void fuzz_dma_read_cb(size_t addr
,
64 /* Possible bits for global_dirty_log_{start|stop} */
66 /* Dirty tracking enabled because migration is running */
67 #define GLOBAL_DIRTY_MIGRATION (1U << 0)
69 /* Dirty tracking enabled because measuring dirty rate */
70 #define GLOBAL_DIRTY_DIRTY_RATE (1U << 1)
72 /* Dirty tracking enabled because dirty limit */
73 #define GLOBAL_DIRTY_LIMIT (1U << 2)
75 #define GLOBAL_DIRTY_MASK (0x7)
77 extern unsigned int global_dirty_tracking
;
79 typedef struct MemoryRegionOps MemoryRegionOps
;
81 struct ReservedRegion
{
88 * struct MemoryRegionSection: describes a fragment of a #MemoryRegion
90 * @mr: the region, or %NULL if empty
91 * @fv: the flat view of the address space the region is mapped in
92 * @offset_within_region: the beginning of the section, relative to @mr's start
93 * @size: the size of the section; will not exceed @mr's boundaries
94 * @offset_within_address_space: the address of the first byte of the section
95 * relative to the region's address space
96 * @readonly: writes to this section are ignored
97 * @nonvolatile: this section is non-volatile
99 struct MemoryRegionSection
{
103 hwaddr offset_within_region
;
104 hwaddr offset_within_address_space
;
109 typedef struct IOMMUTLBEntry IOMMUTLBEntry
;
111 /* See address_space_translate: bit 0 is read, bit 1 is write. */
119 #define IOMMU_ACCESS_FLAG(r, w) (((r) ? IOMMU_RO : 0) | ((w) ? IOMMU_WO : 0))
121 struct IOMMUTLBEntry
{
122 AddressSpace
*target_as
;
124 hwaddr translated_addr
;
125 hwaddr addr_mask
; /* 0xfff = 4k translation */
126 IOMMUAccessFlags perm
;
130 * Bitmap for different IOMMUNotifier capabilities. Each notifier can
131 * register with one or multiple IOMMU Notifier capability bit(s).
133 * Normally there're two use cases for the notifiers:
135 * (1) When the device needs accurate synchronizations of the vIOMMU page
136 * tables, it needs to register with both MAP|UNMAP notifies (which
137 * is defined as IOMMU_NOTIFIER_IOTLB_EVENTS below).
139 * Regarding to accurate synchronization, it's when the notified
140 * device maintains a shadow page table and must be notified on each
141 * guest MAP (page table entry creation) and UNMAP (invalidation)
142 * events (e.g. VFIO). Both notifications must be accurate so that
143 * the shadow page table is fully in sync with the guest view.
145 * (2) When the device doesn't need accurate synchronizations of the
146 * vIOMMU page tables, it needs to register only with UNMAP or
147 * DEVIOTLB_UNMAP notifies.
149 * It's when the device maintains a cache of IOMMU translations
150 * (IOTLB) and is able to fill that cache by requesting translations
151 * from the vIOMMU through a protocol similar to ATS (Address
152 * Translation Service).
154 * Note that in this mode the vIOMMU will not maintain a shadowed
155 * page table for the address space, and the UNMAP messages can cover
156 * more than the pages that used to get mapped. The IOMMU notifiee
157 * should be able to take care of over-sized invalidations.
160 IOMMU_NOTIFIER_NONE
= 0,
161 /* Notify cache invalidations */
162 IOMMU_NOTIFIER_UNMAP
= 0x1,
163 /* Notify entry changes (newly created entries) */
164 IOMMU_NOTIFIER_MAP
= 0x2,
165 /* Notify changes on device IOTLB entries */
166 IOMMU_NOTIFIER_DEVIOTLB_UNMAP
= 0x04,
169 #define IOMMU_NOTIFIER_IOTLB_EVENTS (IOMMU_NOTIFIER_MAP | IOMMU_NOTIFIER_UNMAP)
170 #define IOMMU_NOTIFIER_DEVIOTLB_EVENTS IOMMU_NOTIFIER_DEVIOTLB_UNMAP
171 #define IOMMU_NOTIFIER_ALL (IOMMU_NOTIFIER_IOTLB_EVENTS | \
172 IOMMU_NOTIFIER_DEVIOTLB_EVENTS)
174 struct IOMMUNotifier
;
175 typedef void (*IOMMUNotify
)(struct IOMMUNotifier
*notifier
,
176 IOMMUTLBEntry
*data
);
178 struct IOMMUNotifier
{
180 IOMMUNotifierFlag notifier_flags
;
181 /* Notify for address space range start <= addr <= end */
185 QLIST_ENTRY(IOMMUNotifier
) node
;
187 typedef struct IOMMUNotifier IOMMUNotifier
;
189 typedef struct IOMMUTLBEvent
{
190 IOMMUNotifierFlag type
;
194 /* RAM is pre-allocated and passed into qemu_ram_alloc_from_ptr */
195 #define RAM_PREALLOC (1 << 0)
197 /* RAM is mmap-ed with MAP_SHARED */
198 #define RAM_SHARED (1 << 1)
200 /* Only a portion of RAM (used_length) is actually used, and migrated.
201 * Resizing RAM while migrating can result in the migration being canceled.
203 #define RAM_RESIZEABLE (1 << 2)
205 /* UFFDIO_ZEROPAGE is available on this RAMBlock to atomically
206 * zero the page and wake waiting processes.
207 * (Set during postcopy)
209 #define RAM_UF_ZEROPAGE (1 << 3)
211 /* RAM can be migrated */
212 #define RAM_MIGRATABLE (1 << 4)
214 /* RAM is a persistent kind memory */
215 #define RAM_PMEM (1 << 5)
219 * UFFDIO_WRITEPROTECT is used on this RAMBlock to
220 * support 'write-tracking' migration type.
221 * Implies ram_state->ram_wt_enabled.
223 #define RAM_UF_WRITEPROTECT (1 << 6)
226 * RAM is mmap-ed with MAP_NORESERVE. When set, reserving swap space (or huge
227 * pages if applicable) is skipped: will bail out if not supported. When not
228 * set, the OS will do the reservation, if supported for the memory type.
230 #define RAM_NORESERVE (1 << 7)
232 /* RAM that isn't accessible through normal means. */
233 #define RAM_PROTECTED (1 << 8)
235 static inline void iommu_notifier_init(IOMMUNotifier
*n
, IOMMUNotify fn
,
236 IOMMUNotifierFlag flags
,
237 hwaddr start
, hwaddr end
,
241 n
->notifier_flags
= flags
;
244 n
->iommu_idx
= iommu_idx
;
248 * Memory region callbacks
250 struct MemoryRegionOps
{
251 /* Read from the memory region. @addr is relative to @mr; @size is
253 uint64_t (*read
)(void *opaque
,
256 /* Write to the memory region. @addr is relative to @mr; @size is
258 void (*write
)(void *opaque
,
263 MemTxResult (*read_with_attrs
)(void *opaque
,
268 MemTxResult (*write_with_attrs
)(void *opaque
,
274 enum device_endian endianness
;
275 /* Guest-visible constraints: */
277 /* If nonzero, specify bounds on access sizes beyond which a machine
280 unsigned min_access_size
;
281 unsigned max_access_size
;
282 /* If true, unaligned accesses are supported. Otherwise unaligned
283 * accesses throw machine checks.
287 * If present, and returns #false, the transaction is not accepted
288 * by the device (and results in machine dependent behaviour such
289 * as a machine check exception).
291 bool (*accepts
)(void *opaque
, hwaddr addr
,
292 unsigned size
, bool is_write
,
295 /* Internal implementation constraints: */
297 /* If nonzero, specifies the minimum size implemented. Smaller sizes
298 * will be rounded upwards and a partial result will be returned.
300 unsigned min_access_size
;
301 /* If nonzero, specifies the maximum size implemented. Larger sizes
302 * will be done as a series of accesses with smaller sizes.
304 unsigned max_access_size
;
305 /* If true, unaligned accesses are supported. Otherwise all accesses
306 * are converted to (possibly multiple) naturally aligned accesses.
312 typedef struct MemoryRegionClass
{
314 ObjectClass parent_class
;
318 enum IOMMUMemoryRegionAttr
{
319 IOMMU_ATTR_SPAPR_TCE_FD
323 * IOMMUMemoryRegionClass:
325 * All IOMMU implementations need to subclass TYPE_IOMMU_MEMORY_REGION
326 * and provide an implementation of at least the @translate method here
327 * to handle requests to the memory region. Other methods are optional.
329 * The IOMMU implementation must use the IOMMU notifier infrastructure
330 * to report whenever mappings are changed, by calling
331 * memory_region_notify_iommu() (or, if necessary, by calling
332 * memory_region_notify_iommu_one() for each registered notifier).
334 * Conceptually an IOMMU provides a mapping from input address
335 * to an output TLB entry. If the IOMMU is aware of memory transaction
336 * attributes and the output TLB entry depends on the transaction
337 * attributes, we represent this using IOMMU indexes. Each index
338 * selects a particular translation table that the IOMMU has:
340 * @attrs_to_index returns the IOMMU index for a set of transaction attributes
342 * @translate takes an input address and an IOMMU index
344 * and the mapping returned can only depend on the input address and the
347 * Most IOMMUs don't care about the transaction attributes and support
348 * only a single IOMMU index. A more complex IOMMU might have one index
349 * for secure transactions and one for non-secure transactions.
351 struct IOMMUMemoryRegionClass
{
353 MemoryRegionClass parent_class
;
359 * Return a TLB entry that contains a given address.
361 * The IOMMUAccessFlags indicated via @flag are optional and may
362 * be specified as IOMMU_NONE to indicate that the caller needs
363 * the full translation information for both reads and writes. If
364 * the access flags are specified then the IOMMU implementation
365 * may use this as an optimization, to stop doing a page table
366 * walk as soon as it knows that the requested permissions are not
367 * allowed. If IOMMU_NONE is passed then the IOMMU must do the
368 * full page table walk and report the permissions in the returned
369 * IOMMUTLBEntry. (Note that this implies that an IOMMU may not
370 * return different mappings for reads and writes.)
372 * The returned information remains valid while the caller is
373 * holding the big QEMU lock or is inside an RCU critical section;
374 * if the caller wishes to cache the mapping beyond that it must
375 * register an IOMMU notifier so it can invalidate its cached
376 * information when the IOMMU mapping changes.
378 * @iommu: the IOMMUMemoryRegion
380 * @hwaddr: address to be translated within the memory region
382 * @flag: requested access permission
384 * @iommu_idx: IOMMU index for the translation
386 IOMMUTLBEntry (*translate
)(IOMMUMemoryRegion
*iommu
, hwaddr addr
,
387 IOMMUAccessFlags flag
, int iommu_idx
);
389 * @get_min_page_size:
391 * Returns minimum supported page size in bytes.
393 * If this method is not provided then the minimum is assumed to
394 * be TARGET_PAGE_SIZE.
396 * @iommu: the IOMMUMemoryRegion
398 uint64_t (*get_min_page_size
)(IOMMUMemoryRegion
*iommu
);
400 * @notify_flag_changed:
402 * Called when IOMMU Notifier flag changes (ie when the set of
403 * events which IOMMU users are requesting notification for changes).
404 * Optional method -- need not be provided if the IOMMU does not
405 * need to know exactly which events must be notified.
407 * @iommu: the IOMMUMemoryRegion
409 * @old_flags: events which previously needed to be notified
411 * @new_flags: events which now need to be notified
413 * Returns 0 on success, or a negative errno; in particular
414 * returns -EINVAL if the new flag bitmap is not supported by the
415 * IOMMU memory region. In case of failure, the error object
418 int (*notify_flag_changed
)(IOMMUMemoryRegion
*iommu
,
419 IOMMUNotifierFlag old_flags
,
420 IOMMUNotifierFlag new_flags
,
425 * Called to handle memory_region_iommu_replay().
427 * The default implementation of memory_region_iommu_replay() is to
428 * call the IOMMU translate method for every page in the address space
429 * with flag == IOMMU_NONE and then call the notifier if translate
430 * returns a valid mapping. If this method is implemented then it
431 * overrides the default behaviour, and must provide the full semantics
432 * of memory_region_iommu_replay(), by calling @notifier for every
433 * translation present in the IOMMU.
435 * Optional method -- an IOMMU only needs to provide this method
436 * if the default is inefficient or produces undesirable side effects.
438 * Note: this is not related to record-and-replay functionality.
440 void (*replay
)(IOMMUMemoryRegion
*iommu
, IOMMUNotifier
*notifier
);
445 * Get IOMMU misc attributes. This is an optional method that
446 * can be used to allow users of the IOMMU to get implementation-specific
447 * information. The IOMMU implements this method to handle calls
448 * by IOMMU users to memory_region_iommu_get_attr() by filling in
449 * the arbitrary data pointer for any IOMMUMemoryRegionAttr values that
450 * the IOMMU supports. If the method is unimplemented then
451 * memory_region_iommu_get_attr() will always return -EINVAL.
453 * @iommu: the IOMMUMemoryRegion
455 * @attr: attribute being queried
457 * @data: memory to fill in with the attribute data
459 * Returns 0 on success, or a negative errno; in particular
460 * returns -EINVAL for unrecognized or unimplemented attribute types.
462 int (*get_attr
)(IOMMUMemoryRegion
*iommu
, enum IOMMUMemoryRegionAttr attr
,
468 * Return the IOMMU index to use for a given set of transaction attributes.
470 * Optional method: if an IOMMU only supports a single IOMMU index then
471 * the default implementation of memory_region_iommu_attrs_to_index()
474 * The indexes supported by an IOMMU must be contiguous, starting at 0.
476 * @iommu: the IOMMUMemoryRegion
477 * @attrs: memory transaction attributes
479 int (*attrs_to_index
)(IOMMUMemoryRegion
*iommu
, MemTxAttrs attrs
);
484 * Return the number of IOMMU indexes this IOMMU supports.
486 * Optional method: if this method is not provided, then
487 * memory_region_iommu_num_indexes() will return 1, indicating that
488 * only a single IOMMU index is supported.
490 * @iommu: the IOMMUMemoryRegion
492 int (*num_indexes
)(IOMMUMemoryRegion
*iommu
);
495 * @iommu_set_page_size_mask:
497 * Restrict the page size mask that can be supported with a given IOMMU
498 * memory region. Used for example to propagate host physical IOMMU page
499 * size mask limitations to the virtual IOMMU.
501 * Optional method: if this method is not provided, then the default global
504 * @iommu: the IOMMUMemoryRegion
506 * @page_size_mask: a bitmask of supported page sizes. At least one bit,
507 * representing the smallest page size, must be set. Additional set bits
508 * represent supported block sizes. For example a host physical IOMMU that
509 * uses page tables with a page size of 4kB, and supports 2MB and 4GB
510 * blocks, will set mask 0x40201000. A granule of 4kB with indiscriminate
511 * block sizes is specified with mask 0xfffffffffffff000.
513 * Returns 0 on success, or a negative error. In case of failure, the error
514 * object must be created.
516 int (*iommu_set_page_size_mask
)(IOMMUMemoryRegion
*iommu
,
517 uint64_t page_size_mask
,
521 typedef struct RamDiscardListener RamDiscardListener
;
522 typedef int (*NotifyRamPopulate
)(RamDiscardListener
*rdl
,
523 MemoryRegionSection
*section
);
524 typedef void (*NotifyRamDiscard
)(RamDiscardListener
*rdl
,
525 MemoryRegionSection
*section
);
527 struct RamDiscardListener
{
531 * Notification that previously discarded memory is about to get populated.
532 * Listeners are able to object. If any listener objects, already
533 * successfully notified listeners are notified about a discard again.
535 * @rdl: the #RamDiscardListener getting notified
536 * @section: the #MemoryRegionSection to get populated. The section
537 * is aligned within the memory region to the minimum granularity
538 * unless it would exceed the registered section.
540 * Returns 0 on success. If the notification is rejected by the listener,
541 * an error is returned.
543 NotifyRamPopulate notify_populate
;
548 * Notification that previously populated memory was discarded successfully
549 * and listeners should drop all references to such memory and prevent
550 * new population (e.g., unmap).
552 * @rdl: the #RamDiscardListener getting notified
553 * @section: the #MemoryRegionSection to get populated. The section
554 * is aligned within the memory region to the minimum granularity
555 * unless it would exceed the registered section.
557 NotifyRamDiscard notify_discard
;
560 * @double_discard_supported:
562 * The listener suppors getting @notify_discard notifications that span
563 * already discarded parts.
565 bool double_discard_supported
;
567 MemoryRegionSection
*section
;
568 QLIST_ENTRY(RamDiscardListener
) next
;
571 static inline void ram_discard_listener_init(RamDiscardListener
*rdl
,
572 NotifyRamPopulate populate_fn
,
573 NotifyRamDiscard discard_fn
,
574 bool double_discard_supported
)
576 rdl
->notify_populate
= populate_fn
;
577 rdl
->notify_discard
= discard_fn
;
578 rdl
->double_discard_supported
= double_discard_supported
;
581 typedef int (*ReplayRamPopulate
)(MemoryRegionSection
*section
, void *opaque
);
582 typedef void (*ReplayRamDiscard
)(MemoryRegionSection
*section
, void *opaque
);
585 * RamDiscardManagerClass:
587 * A #RamDiscardManager coordinates which parts of specific RAM #MemoryRegion
588 * regions are currently populated to be used/accessed by the VM, notifying
589 * after parts were discarded (freeing up memory) and before parts will be
590 * populated (consuming memory), to be used/accessed by the VM.
592 * A #RamDiscardManager can only be set for a RAM #MemoryRegion while the
593 * #MemoryRegion isn't mapped yet; it cannot change while the #MemoryRegion is
596 * The #RamDiscardManager is intended to be used by technologies that are
597 * incompatible with discarding of RAM (e.g., VFIO, which may pin all
598 * memory inside a #MemoryRegion), and require proper coordination to only
599 * map the currently populated parts, to hinder parts that are expected to
600 * remain discarded from silently getting populated and consuming memory.
601 * Technologies that support discarding of RAM don't have to bother and can
602 * simply map the whole #MemoryRegion.
604 * An example #RamDiscardManager is virtio-mem, which logically (un)plugs
605 * memory within an assigned RAM #MemoryRegion, coordinated with the VM.
606 * Logically unplugging memory consists of discarding RAM. The VM agreed to not
607 * access unplugged (discarded) memory - especially via DMA. virtio-mem will
608 * properly coordinate with listeners before memory is plugged (populated),
609 * and after memory is unplugged (discarded).
611 * Listeners are called in multiples of the minimum granularity (unless it
612 * would exceed the registered range) and changes are aligned to the minimum
613 * granularity within the #MemoryRegion. Listeners have to prepare for memory
614 * becoming discarded in a different granularity than it was populated and the
617 struct RamDiscardManagerClass
{
619 InterfaceClass parent_class
;
624 * @get_min_granularity:
626 * Get the minimum granularity in which listeners will get notified
627 * about changes within the #MemoryRegion via the #RamDiscardManager.
629 * @rdm: the #RamDiscardManager
630 * @mr: the #MemoryRegion
632 * Returns the minimum granularity.
634 uint64_t (*get_min_granularity
)(const RamDiscardManager
*rdm
,
635 const MemoryRegion
*mr
);
640 * Check whether the given #MemoryRegionSection is completely populated
641 * (i.e., no parts are currently discarded) via the #RamDiscardManager.
642 * There are no alignment requirements.
644 * @rdm: the #RamDiscardManager
645 * @section: the #MemoryRegionSection
647 * Returns whether the given range is completely populated.
649 bool (*is_populated
)(const RamDiscardManager
*rdm
,
650 const MemoryRegionSection
*section
);
655 * Call the #ReplayRamPopulate callback for all populated parts within the
656 * #MemoryRegionSection via the #RamDiscardManager.
658 * In case any call fails, no further calls are made.
660 * @rdm: the #RamDiscardManager
661 * @section: the #MemoryRegionSection
662 * @replay_fn: the #ReplayRamPopulate callback
663 * @opaque: pointer to forward to the callback
665 * Returns 0 on success, or a negative error if any notification failed.
667 int (*replay_populated
)(const RamDiscardManager
*rdm
,
668 MemoryRegionSection
*section
,
669 ReplayRamPopulate replay_fn
, void *opaque
);
674 * Call the #ReplayRamDiscard callback for all discarded parts within the
675 * #MemoryRegionSection via the #RamDiscardManager.
677 * @rdm: the #RamDiscardManager
678 * @section: the #MemoryRegionSection
679 * @replay_fn: the #ReplayRamDiscard callback
680 * @opaque: pointer to forward to the callback
682 void (*replay_discarded
)(const RamDiscardManager
*rdm
,
683 MemoryRegionSection
*section
,
684 ReplayRamDiscard replay_fn
, void *opaque
);
687 * @register_listener:
689 * Register a #RamDiscardListener for the given #MemoryRegionSection and
690 * immediately notify the #RamDiscardListener about all populated parts
691 * within the #MemoryRegionSection via the #RamDiscardManager.
693 * In case any notification fails, no further notifications are triggered
694 * and an error is logged.
696 * @rdm: the #RamDiscardManager
697 * @rdl: the #RamDiscardListener
698 * @section: the #MemoryRegionSection
700 void (*register_listener
)(RamDiscardManager
*rdm
,
701 RamDiscardListener
*rdl
,
702 MemoryRegionSection
*section
);
705 * @unregister_listener:
707 * Unregister a previously registered #RamDiscardListener via the
708 * #RamDiscardManager after notifying the #RamDiscardListener about all
709 * populated parts becoming unpopulated within the registered
710 * #MemoryRegionSection.
712 * @rdm: the #RamDiscardManager
713 * @rdl: the #RamDiscardListener
715 void (*unregister_listener
)(RamDiscardManager
*rdm
,
716 RamDiscardListener
*rdl
);
719 uint64_t ram_discard_manager_get_min_granularity(const RamDiscardManager
*rdm
,
720 const MemoryRegion
*mr
);
722 bool ram_discard_manager_is_populated(const RamDiscardManager
*rdm
,
723 const MemoryRegionSection
*section
);
725 int ram_discard_manager_replay_populated(const RamDiscardManager
*rdm
,
726 MemoryRegionSection
*section
,
727 ReplayRamPopulate replay_fn
,
730 void ram_discard_manager_replay_discarded(const RamDiscardManager
*rdm
,
731 MemoryRegionSection
*section
,
732 ReplayRamDiscard replay_fn
,
735 void ram_discard_manager_register_listener(RamDiscardManager
*rdm
,
736 RamDiscardListener
*rdl
,
737 MemoryRegionSection
*section
);
739 void ram_discard_manager_unregister_listener(RamDiscardManager
*rdm
,
740 RamDiscardListener
*rdl
);
742 bool memory_get_xlat_addr(IOMMUTLBEntry
*iotlb
, void **vaddr
,
743 ram_addr_t
*ram_addr
, bool *read_only
,
744 bool *mr_has_discard_manager
);
746 typedef struct CoalescedMemoryRange CoalescedMemoryRange
;
747 typedef struct MemoryRegionIoeventfd MemoryRegionIoeventfd
;
751 * A struct representing a memory region.
753 struct MemoryRegion
{
758 /* The following fields should fit in a cache line */
762 bool readonly
; /* For RAM regions */
765 bool flush_coalesced_mmio
;
766 uint8_t dirty_log_mask
;
770 /* owner as TYPE_DEVICE. Used for re-entrancy checks in MR access hotpath */
773 const MemoryRegionOps
*ops
;
775 MemoryRegion
*container
;
776 int mapped_via_alias
; /* Mapped via an alias, container might be NULL */
779 void (*destructor
)(MemoryRegion
*mr
);
784 bool warning_printed
; /* For reservations */
785 uint8_t vga_logging_count
;
789 QTAILQ_HEAD(, MemoryRegion
) subregions
;
790 QTAILQ_ENTRY(MemoryRegion
) subregions_link
;
791 QTAILQ_HEAD(, CoalescedMemoryRange
) coalesced
;
793 unsigned ioeventfd_nb
;
794 MemoryRegionIoeventfd
*ioeventfds
;
795 RamDiscardManager
*rdm
; /* Only for RAM */
797 /* For devices designed to perform re-entrant IO into their own IO MRs */
798 bool disable_reentrancy_guard
;
801 struct IOMMUMemoryRegion
{
802 MemoryRegion parent_obj
;
804 QLIST_HEAD(, IOMMUNotifier
) iommu_notify
;
805 IOMMUNotifierFlag iommu_notify_flags
;
808 #define IOMMU_NOTIFIER_FOREACH(n, mr) \
809 QLIST_FOREACH((n), &(mr)->iommu_notify, node)
812 * struct MemoryListener: callbacks structure for updates to the physical memory map
814 * Allows a component to adjust to changes in the guest-visible memory map.
815 * Use with memory_listener_register() and memory_listener_unregister().
817 struct MemoryListener
{
821 * Called at the beginning of an address space update transaction.
822 * Followed by calls to #MemoryListener.region_add(),
823 * #MemoryListener.region_del(), #MemoryListener.region_nop(),
824 * #MemoryListener.log_start() and #MemoryListener.log_stop() in
825 * increasing address order.
827 * @listener: The #MemoryListener.
829 void (*begin
)(MemoryListener
*listener
);
834 * Called at the end of an address space update transaction,
835 * after the last call to #MemoryListener.region_add(),
836 * #MemoryListener.region_del() or #MemoryListener.region_nop(),
837 * #MemoryListener.log_start() and #MemoryListener.log_stop().
839 * @listener: The #MemoryListener.
841 void (*commit
)(MemoryListener
*listener
);
846 * Called during an address space update transaction,
847 * for a section of the address space that is new in this address space
848 * space since the last transaction.
850 * @listener: The #MemoryListener.
851 * @section: The new #MemoryRegionSection.
853 void (*region_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
858 * Called during an address space update transaction,
859 * for a section of the address space that has disappeared in the address
860 * space since the last transaction.
862 * @listener: The #MemoryListener.
863 * @section: The old #MemoryRegionSection.
865 void (*region_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
870 * Called during an address space update transaction,
871 * for a section of the address space that is in the same place in the address
872 * space as in the last transaction.
874 * @listener: The #MemoryListener.
875 * @section: The #MemoryRegionSection.
877 void (*region_nop
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
882 * Called during an address space update transaction, after
883 * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
884 * #MemoryListener.region_nop(), if dirty memory logging clients have
885 * become active since the last transaction.
887 * @listener: The #MemoryListener.
888 * @section: The #MemoryRegionSection.
889 * @old: A bitmap of dirty memory logging clients that were active in
890 * the previous transaction.
891 * @new: A bitmap of dirty memory logging clients that are active in
892 * the current transaction.
894 void (*log_start
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
900 * Called during an address space update transaction, after
901 * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
902 * #MemoryListener.region_nop() and possibly after
903 * #MemoryListener.log_start(), if dirty memory logging clients have
904 * become inactive since the last transaction.
906 * @listener: The #MemoryListener.
907 * @section: The #MemoryRegionSection.
908 * @old: A bitmap of dirty memory logging clients that were active in
909 * the previous transaction.
910 * @new: A bitmap of dirty memory logging clients that are active in
911 * the current transaction.
913 void (*log_stop
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
919 * Called by memory_region_snapshot_and_clear_dirty() and
920 * memory_global_dirty_log_sync(), before accessing QEMU's "official"
921 * copy of the dirty memory bitmap for a #MemoryRegionSection.
923 * @listener: The #MemoryListener.
924 * @section: The #MemoryRegionSection.
926 void (*log_sync
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
931 * This is the global version of @log_sync when the listener does
932 * not have a way to synchronize the log with finer granularity.
933 * When the listener registers with @log_sync_global defined, then
934 * its @log_sync must be NULL. Vice versa.
936 * @listener: The #MemoryListener.
937 * @last_stage: The last stage to synchronize the log during migration.
938 * The caller should gurantee that the synchronization with true for
939 * @last_stage is triggered for once after all VCPUs have been stopped.
941 void (*log_sync_global
)(MemoryListener
*listener
, bool last_stage
);
946 * Called before reading the dirty memory bitmap for a
947 * #MemoryRegionSection.
949 * @listener: The #MemoryListener.
950 * @section: The #MemoryRegionSection.
952 void (*log_clear
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
957 * Called by memory_global_dirty_log_start(), which
958 * enables the %DIRTY_LOG_MIGRATION client on all memory regions in
959 * the address space. #MemoryListener.log_global_start() is also
960 * called when a #MemoryListener is added, if global dirty logging is
961 * active at that time.
963 * @listener: The #MemoryListener.
965 void (*log_global_start
)(MemoryListener
*listener
);
970 * Called by memory_global_dirty_log_stop(), which
971 * disables the %DIRTY_LOG_MIGRATION client on all memory regions in
974 * @listener: The #MemoryListener.
976 void (*log_global_stop
)(MemoryListener
*listener
);
979 * @log_global_after_sync:
981 * Called after reading the dirty memory bitmap
982 * for any #MemoryRegionSection.
984 * @listener: The #MemoryListener.
986 void (*log_global_after_sync
)(MemoryListener
*listener
);
991 * Called during an address space update transaction,
992 * for a section of the address space that has had a new ioeventfd
993 * registration since the last transaction.
995 * @listener: The #MemoryListener.
996 * @section: The new #MemoryRegionSection.
997 * @match_data: The @match_data parameter for the new ioeventfd.
998 * @data: The @data parameter for the new ioeventfd.
999 * @e: The #EventNotifier parameter for the new ioeventfd.
1001 void (*eventfd_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1002 bool match_data
, uint64_t data
, EventNotifier
*e
);
1007 * Called during an address space update transaction,
1008 * for a section of the address space that has dropped an ioeventfd
1009 * registration since the last transaction.
1011 * @listener: The #MemoryListener.
1012 * @section: The new #MemoryRegionSection.
1013 * @match_data: The @match_data parameter for the dropped ioeventfd.
1014 * @data: The @data parameter for the dropped ioeventfd.
1015 * @e: The #EventNotifier parameter for the dropped ioeventfd.
1017 void (*eventfd_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1018 bool match_data
, uint64_t data
, EventNotifier
*e
);
1021 * @coalesced_io_add:
1023 * Called during an address space update transaction,
1024 * for a section of the address space that has had a new coalesced
1025 * MMIO range registration since the last transaction.
1027 * @listener: The #MemoryListener.
1028 * @section: The new #MemoryRegionSection.
1029 * @addr: The starting address for the coalesced MMIO range.
1030 * @len: The length of the coalesced MMIO range.
1032 void (*coalesced_io_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1033 hwaddr addr
, hwaddr len
);
1036 * @coalesced_io_del:
1038 * Called during an address space update transaction,
1039 * for a section of the address space that has dropped a coalesced
1040 * MMIO range since the last transaction.
1042 * @listener: The #MemoryListener.
1043 * @section: The new #MemoryRegionSection.
1044 * @addr: The starting address for the coalesced MMIO range.
1045 * @len: The length of the coalesced MMIO range.
1047 void (*coalesced_io_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1048 hwaddr addr
, hwaddr len
);
1052 * Govern the order in which memory listeners are invoked. Lower priorities
1053 * are invoked earlier for "add" or "start" callbacks, and later for "delete"
1054 * or "stop" callbacks.
1061 * Name of the listener. It can be used in contexts where we'd like to
1062 * identify one memory listener with the rest.
1067 AddressSpace
*address_space
;
1068 QTAILQ_ENTRY(MemoryListener
) link
;
1069 QTAILQ_ENTRY(MemoryListener
) link_as
;
1073 * struct AddressSpace: describes a mapping of addresses to #MemoryRegion objects
1075 struct AddressSpace
{
1077 struct rcu_head rcu
;
1081 /* Accessed via RCU. */
1082 struct FlatView
*current_map
;
1085 struct MemoryRegionIoeventfd
*ioeventfds
;
1086 QTAILQ_HEAD(, MemoryListener
) listeners
;
1087 QTAILQ_ENTRY(AddressSpace
) address_spaces_link
;
1090 typedef struct AddressSpaceDispatch AddressSpaceDispatch
;
1091 typedef struct FlatRange FlatRange
;
1093 /* Flattened global view of current active memory hierarchy. Kept in sorted
1097 struct rcu_head rcu
;
1101 unsigned nr_allocated
;
1102 struct AddressSpaceDispatch
*dispatch
;
1106 static inline FlatView
*address_space_to_flatview(AddressSpace
*as
)
1108 return qatomic_rcu_read(&as
->current_map
);
1112 * typedef flatview_cb: callback for flatview_for_each_range()
1114 * @start: start address of the range within the FlatView
1115 * @len: length of the range in bytes
1116 * @mr: MemoryRegion covering this range
1117 * @offset_in_region: offset of the first byte of the range within @mr
1118 * @opaque: data pointer passed to flatview_for_each_range()
1120 * Returns: true to stop the iteration, false to keep going.
1122 typedef bool (*flatview_cb
)(Int128 start
,
1124 const MemoryRegion
*mr
,
1125 hwaddr offset_in_region
,
1129 * flatview_for_each_range: Iterate through a FlatView
1130 * @fv: the FlatView to iterate through
1131 * @cb: function to call for each range
1132 * @opaque: opaque data pointer to pass to @cb
1134 * A FlatView is made up of a list of non-overlapping ranges, each of
1135 * which is a slice of a MemoryRegion. This function iterates through
1136 * each range in @fv, calling @cb. The callback function can terminate
1137 * iteration early by returning 'true'.
1139 void flatview_for_each_range(FlatView
*fv
, flatview_cb cb
, void *opaque
);
1141 static inline bool MemoryRegionSection_eq(MemoryRegionSection
*a
,
1142 MemoryRegionSection
*b
)
1144 return a
->mr
== b
->mr
&&
1146 a
->offset_within_region
== b
->offset_within_region
&&
1147 a
->offset_within_address_space
== b
->offset_within_address_space
&&
1148 int128_eq(a
->size
, b
->size
) &&
1149 a
->readonly
== b
->readonly
&&
1150 a
->nonvolatile
== b
->nonvolatile
;
1154 * memory_region_section_new_copy: Copy a memory region section
1156 * Allocate memory for a new copy, copy the memory region section, and
1157 * properly take a reference on all relevant members.
1159 * @s: the #MemoryRegionSection to copy
1161 MemoryRegionSection
*memory_region_section_new_copy(MemoryRegionSection
*s
);
1164 * memory_region_section_new_copy: Free a copied memory region section
1166 * Free a copy of a memory section created via memory_region_section_new_copy().
1167 * properly dropping references on all relevant members.
1169 * @s: the #MemoryRegionSection to copy
1171 void memory_region_section_free_copy(MemoryRegionSection
*s
);
1174 * memory_region_init: Initialize a memory region
1176 * The region typically acts as a container for other memory regions. Use
1177 * memory_region_add_subregion() to add subregions.
1179 * @mr: the #MemoryRegion to be initialized
1180 * @owner: the object that tracks the region's reference count
1181 * @name: used for debugging; not visible to the user or ABI
1182 * @size: size of the region; any subregions beyond this size will be clipped
1184 void memory_region_init(MemoryRegion
*mr
,
1190 * memory_region_ref: Add 1 to a memory region's reference count
1192 * Whenever memory regions are accessed outside the BQL, they need to be
1193 * preserved against hot-unplug. MemoryRegions actually do not have their
1194 * own reference count; they piggyback on a QOM object, their "owner".
1195 * This function adds a reference to the owner.
1197 * All MemoryRegions must have an owner if they can disappear, even if the
1198 * device they belong to operates exclusively under the BQL. This is because
1199 * the region could be returned at any time by memory_region_find, and this
1200 * is usually under guest control.
1202 * @mr: the #MemoryRegion
1204 void memory_region_ref(MemoryRegion
*mr
);
1207 * memory_region_unref: Remove 1 to a memory region's reference count
1209 * Whenever memory regions are accessed outside the BQL, they need to be
1210 * preserved against hot-unplug. MemoryRegions actually do not have their
1211 * own reference count; they piggyback on a QOM object, their "owner".
1212 * This function removes a reference to the owner and possibly destroys it.
1214 * @mr: the #MemoryRegion
1216 void memory_region_unref(MemoryRegion
*mr
);
1219 * memory_region_init_io: Initialize an I/O memory region.
1221 * Accesses into the region will cause the callbacks in @ops to be called.
1222 * if @size is nonzero, subregions will be clipped to @size.
1224 * @mr: the #MemoryRegion to be initialized.
1225 * @owner: the object that tracks the region's reference count
1226 * @ops: a structure containing read and write callbacks to be used when
1227 * I/O is performed on the region.
1228 * @opaque: passed to the read and write callbacks of the @ops structure.
1229 * @name: used for debugging; not visible to the user or ABI
1230 * @size: size of the region.
1232 void memory_region_init_io(MemoryRegion
*mr
,
1234 const MemoryRegionOps
*ops
,
1240 * memory_region_init_ram_nomigrate: Initialize RAM memory region. Accesses
1241 * into the region will modify memory
1244 * @mr: the #MemoryRegion to be initialized.
1245 * @owner: the object that tracks the region's reference count
1246 * @name: Region name, becomes part of RAMBlock name used in migration stream
1247 * must be unique within any device
1248 * @size: size of the region.
1249 * @errp: pointer to Error*, to store an error if it happens.
1251 * Note that this function does not do anything to cause the data in the
1252 * RAM memory region to be migrated; that is the responsibility of the caller.
1254 void memory_region_init_ram_nomigrate(MemoryRegion
*mr
,
1261 * memory_region_init_ram_flags_nomigrate: Initialize RAM memory region.
1262 * Accesses into the region will
1263 * modify memory directly.
1265 * @mr: the #MemoryRegion to be initialized.
1266 * @owner: the object that tracks the region's reference count
1267 * @name: Region name, becomes part of RAMBlock name used in migration stream
1268 * must be unique within any device
1269 * @size: size of the region.
1270 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_NORESERVE.
1271 * @errp: pointer to Error*, to store an error if it happens.
1273 * Note that this function does not do anything to cause the data in the
1274 * RAM memory region to be migrated; that is the responsibility of the caller.
1276 void memory_region_init_ram_flags_nomigrate(MemoryRegion
*mr
,
1284 * memory_region_init_resizeable_ram: Initialize memory region with resizable
1285 * RAM. Accesses into the region will
1286 * modify memory directly. Only an initial
1287 * portion of this RAM is actually used.
1288 * Changing the size while migrating
1289 * can result in the migration being
1292 * @mr: the #MemoryRegion to be initialized.
1293 * @owner: the object that tracks the region's reference count
1294 * @name: Region name, becomes part of RAMBlock name used in migration stream
1295 * must be unique within any device
1296 * @size: used size of the region.
1297 * @max_size: max size of the region.
1298 * @resized: callback to notify owner about used size change.
1299 * @errp: pointer to Error*, to store an error if it happens.
1301 * Note that this function does not do anything to cause the data in the
1302 * RAM memory region to be migrated; that is the responsibility of the caller.
1304 void memory_region_init_resizeable_ram(MemoryRegion
*mr
,
1309 void (*resized
)(const char*,
1316 * memory_region_init_ram_from_file: Initialize RAM memory region with a
1319 * @mr: the #MemoryRegion to be initialized.
1320 * @owner: the object that tracks the region's reference count
1321 * @name: Region name, becomes part of RAMBlock name used in migration stream
1322 * must be unique within any device
1323 * @size: size of the region.
1324 * @align: alignment of the region base address; if 0, the default alignment
1325 * (getpagesize()) will be used.
1326 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
1328 * @path: the path in which to allocate the RAM.
1329 * @readonly: true to open @path for reading, false for read/write.
1330 * @errp: pointer to Error*, to store an error if it happens.
1332 * Note that this function does not do anything to cause the data in the
1333 * RAM memory region to be migrated; that is the responsibility of the caller.
1335 void memory_region_init_ram_from_file(MemoryRegion
*mr
,
1346 * memory_region_init_ram_from_fd: Initialize RAM memory region with a
1349 * @mr: the #MemoryRegion to be initialized.
1350 * @owner: the object that tracks the region's reference count
1351 * @name: the name of the region.
1352 * @size: size of the region.
1353 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
1354 * RAM_NORESERVE, RAM_PROTECTED.
1355 * @fd: the fd to mmap.
1356 * @offset: offset within the file referenced by fd
1357 * @errp: pointer to Error*, to store an error if it happens.
1359 * Note that this function does not do anything to cause the data in the
1360 * RAM memory region to be migrated; that is the responsibility of the caller.
1362 void memory_region_init_ram_from_fd(MemoryRegion
*mr
,
1373 * memory_region_init_ram_ptr: Initialize RAM memory region from a
1374 * user-provided pointer. Accesses into the
1375 * region will modify memory directly.
1377 * @mr: the #MemoryRegion to be initialized.
1378 * @owner: the object that tracks the region's reference count
1379 * @name: Region name, becomes part of RAMBlock name used in migration stream
1380 * must be unique within any device
1381 * @size: size of the region.
1382 * @ptr: memory to be mapped; must contain at least @size bytes.
1384 * Note that this function does not do anything to cause the data in the
1385 * RAM memory region to be migrated; that is the responsibility of the caller.
1387 void memory_region_init_ram_ptr(MemoryRegion
*mr
,
1394 * memory_region_init_ram_device_ptr: Initialize RAM device memory region from
1395 * a user-provided pointer.
1397 * A RAM device represents a mapping to a physical device, such as to a PCI
1398 * MMIO BAR of an vfio-pci assigned device. The memory region may be mapped
1399 * into the VM address space and access to the region will modify memory
1400 * directly. However, the memory region should not be included in a memory
1401 * dump (device may not be enabled/mapped at the time of the dump), and
1402 * operations incompatible with manipulating MMIO should be avoided. Replaces
1405 * @mr: the #MemoryRegion to be initialized.
1406 * @owner: the object that tracks the region's reference count
1407 * @name: the name of the region.
1408 * @size: size of the region.
1409 * @ptr: memory to be mapped; must contain at least @size bytes.
1411 * Note that this function does not do anything to cause the data in the
1412 * RAM memory region to be migrated; that is the responsibility of the caller.
1413 * (For RAM device memory regions, migrating the contents rarely makes sense.)
1415 void memory_region_init_ram_device_ptr(MemoryRegion
*mr
,
1422 * memory_region_init_alias: Initialize a memory region that aliases all or a
1423 * part of another memory region.
1425 * @mr: the #MemoryRegion to be initialized.
1426 * @owner: the object that tracks the region's reference count
1427 * @name: used for debugging; not visible to the user or ABI
1428 * @orig: the region to be referenced; @mr will be equivalent to
1429 * @orig between @offset and @offset + @size - 1.
1430 * @offset: start of the section in @orig to be referenced.
1431 * @size: size of the region.
1433 void memory_region_init_alias(MemoryRegion
*mr
,
1441 * memory_region_init_rom_nomigrate: Initialize a ROM memory region.
1443 * This has the same effect as calling memory_region_init_ram_nomigrate()
1444 * and then marking the resulting region read-only with
1445 * memory_region_set_readonly().
1447 * Note that this function does not do anything to cause the data in the
1448 * RAM side of the memory region to be migrated; that is the responsibility
1451 * @mr: the #MemoryRegion to be initialized.
1452 * @owner: the object that tracks the region's reference count
1453 * @name: Region name, becomes part of RAMBlock name used in migration stream
1454 * must be unique within any device
1455 * @size: size of the region.
1456 * @errp: pointer to Error*, to store an error if it happens.
1458 void memory_region_init_rom_nomigrate(MemoryRegion
*mr
,
1465 * memory_region_init_rom_device_nomigrate: Initialize a ROM memory region.
1466 * Writes are handled via callbacks.
1468 * Note that this function does not do anything to cause the data in the
1469 * RAM side of the memory region to be migrated; that is the responsibility
1472 * @mr: the #MemoryRegion to be initialized.
1473 * @owner: the object that tracks the region's reference count
1474 * @ops: callbacks for write access handling (must not be NULL).
1475 * @opaque: passed to the read and write callbacks of the @ops structure.
1476 * @name: Region name, becomes part of RAMBlock name used in migration stream
1477 * must be unique within any device
1478 * @size: size of the region.
1479 * @errp: pointer to Error*, to store an error if it happens.
1481 void memory_region_init_rom_device_nomigrate(MemoryRegion
*mr
,
1483 const MemoryRegionOps
*ops
,
1490 * memory_region_init_iommu: Initialize a memory region of a custom type
1491 * that translates addresses
1493 * An IOMMU region translates addresses and forwards accesses to a target
1496 * The IOMMU implementation must define a subclass of TYPE_IOMMU_MEMORY_REGION.
1497 * @_iommu_mr should be a pointer to enough memory for an instance of
1498 * that subclass, @instance_size is the size of that subclass, and
1499 * @mrtypename is its name. This function will initialize @_iommu_mr as an
1500 * instance of the subclass, and its methods will then be called to handle
1501 * accesses to the memory region. See the documentation of
1502 * #IOMMUMemoryRegionClass for further details.
1504 * @_iommu_mr: the #IOMMUMemoryRegion to be initialized
1505 * @instance_size: the IOMMUMemoryRegion subclass instance size
1506 * @mrtypename: the type name of the #IOMMUMemoryRegion
1507 * @owner: the object that tracks the region's reference count
1508 * @name: used for debugging; not visible to the user or ABI
1509 * @size: size of the region.
1511 void memory_region_init_iommu(void *_iommu_mr
,
1512 size_t instance_size
,
1513 const char *mrtypename
,
1519 * memory_region_init_ram - Initialize RAM memory region. Accesses into the
1520 * region will modify memory directly.
1522 * @mr: the #MemoryRegion to be initialized
1523 * @owner: the object that tracks the region's reference count (must be
1524 * TYPE_DEVICE or a subclass of TYPE_DEVICE, or NULL)
1525 * @name: name of the memory region
1526 * @size: size of the region in bytes
1527 * @errp: pointer to Error*, to store an error if it happens.
1529 * This function allocates RAM for a board model or device, and
1530 * arranges for it to be migrated (by calling vmstate_register_ram()
1531 * if @owner is a DeviceState, or vmstate_register_ram_global() if
1534 * TODO: Currently we restrict @owner to being either NULL (for
1535 * global RAM regions with no owner) or devices, so that we can
1536 * give the RAM block a unique name for migration purposes.
1537 * We should lift this restriction and allow arbitrary Objects.
1538 * If you pass a non-NULL non-device @owner then we will assert.
1540 void memory_region_init_ram(MemoryRegion
*mr
,
1547 * memory_region_init_rom: Initialize a ROM memory region.
1549 * This has the same effect as calling memory_region_init_ram()
1550 * and then marking the resulting region read-only with
1551 * memory_region_set_readonly(). This includes arranging for the
1552 * contents to be migrated.
1554 * TODO: Currently we restrict @owner to being either NULL (for
1555 * global RAM regions with no owner) or devices, so that we can
1556 * give the RAM block a unique name for migration purposes.
1557 * We should lift this restriction and allow arbitrary Objects.
1558 * If you pass a non-NULL non-device @owner then we will assert.
1560 * @mr: the #MemoryRegion to be initialized.
1561 * @owner: the object that tracks the region's reference count
1562 * @name: Region name, becomes part of RAMBlock name used in migration stream
1563 * must be unique within any device
1564 * @size: size of the region.
1565 * @errp: pointer to Error*, to store an error if it happens.
1567 void memory_region_init_rom(MemoryRegion
*mr
,
1574 * memory_region_init_rom_device: Initialize a ROM memory region.
1575 * Writes are handled via callbacks.
1577 * This function initializes a memory region backed by RAM for reads
1578 * and callbacks for writes, and arranges for the RAM backing to
1579 * be migrated (by calling vmstate_register_ram()
1580 * if @owner is a DeviceState, or vmstate_register_ram_global() if
1583 * TODO: Currently we restrict @owner to being either NULL (for
1584 * global RAM regions with no owner) or devices, so that we can
1585 * give the RAM block a unique name for migration purposes.
1586 * We should lift this restriction and allow arbitrary Objects.
1587 * If you pass a non-NULL non-device @owner then we will assert.
1589 * @mr: the #MemoryRegion to be initialized.
1590 * @owner: the object that tracks the region's reference count
1591 * @ops: callbacks for write access handling (must not be NULL).
1592 * @opaque: passed to the read and write callbacks of the @ops structure.
1593 * @name: Region name, becomes part of RAMBlock name used in migration stream
1594 * must be unique within any device
1595 * @size: size of the region.
1596 * @errp: pointer to Error*, to store an error if it happens.
1598 void memory_region_init_rom_device(MemoryRegion
*mr
,
1600 const MemoryRegionOps
*ops
,
1608 * memory_region_owner: get a memory region's owner.
1610 * @mr: the memory region being queried.
1612 Object
*memory_region_owner(MemoryRegion
*mr
);
1615 * memory_region_size: get a memory region's size.
1617 * @mr: the memory region being queried.
1619 uint64_t memory_region_size(MemoryRegion
*mr
);
1622 * memory_region_is_ram: check whether a memory region is random access
1624 * Returns %true if a memory region is random access.
1626 * @mr: the memory region being queried
1628 static inline bool memory_region_is_ram(MemoryRegion
*mr
)
1634 * memory_region_is_ram_device: check whether a memory region is a ram device
1636 * Returns %true if a memory region is a device backed ram region
1638 * @mr: the memory region being queried
1640 bool memory_region_is_ram_device(MemoryRegion
*mr
);
1643 * memory_region_is_romd: check whether a memory region is in ROMD mode
1645 * Returns %true if a memory region is a ROM device and currently set to allow
1648 * @mr: the memory region being queried
1650 static inline bool memory_region_is_romd(MemoryRegion
*mr
)
1652 return mr
->rom_device
&& mr
->romd_mode
;
1656 * memory_region_is_protected: check whether a memory region is protected
1658 * Returns %true if a memory region is protected RAM and cannot be accessed
1659 * via standard mechanisms, e.g. DMA.
1661 * @mr: the memory region being queried
1663 bool memory_region_is_protected(MemoryRegion
*mr
);
1666 * memory_region_get_iommu: check whether a memory region is an iommu
1668 * Returns pointer to IOMMUMemoryRegion if a memory region is an iommu,
1671 * @mr: the memory region being queried
1673 static inline IOMMUMemoryRegion
*memory_region_get_iommu(MemoryRegion
*mr
)
1676 return memory_region_get_iommu(mr
->alias
);
1679 return (IOMMUMemoryRegion
*) mr
;
1685 * memory_region_get_iommu_class_nocheck: returns iommu memory region class
1686 * if an iommu or NULL if not
1688 * Returns pointer to IOMMUMemoryRegionClass if a memory region is an iommu,
1689 * otherwise NULL. This is fast path avoiding QOM checking, use with caution.
1691 * @iommu_mr: the memory region being queried
1693 static inline IOMMUMemoryRegionClass
*memory_region_get_iommu_class_nocheck(
1694 IOMMUMemoryRegion
*iommu_mr
)
1696 return (IOMMUMemoryRegionClass
*) (((Object
*)iommu_mr
)->class);
1699 #define memory_region_is_iommu(mr) (memory_region_get_iommu(mr) != NULL)
1702 * memory_region_iommu_get_min_page_size: get minimum supported page size
1705 * Returns minimum supported page size for an iommu.
1707 * @iommu_mr: the memory region being queried
1709 uint64_t memory_region_iommu_get_min_page_size(IOMMUMemoryRegion
*iommu_mr
);
1712 * memory_region_notify_iommu: notify a change in an IOMMU translation entry.
1714 * Note: for any IOMMU implementation, an in-place mapping change
1715 * should be notified with an UNMAP followed by a MAP.
1717 * @iommu_mr: the memory region that was changed
1718 * @iommu_idx: the IOMMU index for the translation table which has changed
1719 * @event: TLB event with the new entry in the IOMMU translation table.
1720 * The entry replaces all old entries for the same virtual I/O address
1723 void memory_region_notify_iommu(IOMMUMemoryRegion
*iommu_mr
,
1725 IOMMUTLBEvent event
);
1728 * memory_region_notify_iommu_one: notify a change in an IOMMU translation
1729 * entry to a single notifier
1731 * This works just like memory_region_notify_iommu(), but it only
1732 * notifies a specific notifier, not all of them.
1734 * @notifier: the notifier to be notified
1735 * @event: TLB event with the new entry in the IOMMU translation table.
1736 * The entry replaces all old entries for the same virtual I/O address
1739 void memory_region_notify_iommu_one(IOMMUNotifier
*notifier
,
1740 IOMMUTLBEvent
*event
);
1743 * memory_region_unmap_iommu_notifier_range: notify a unmap for an IOMMU
1744 * translation that covers the
1745 * range of a notifier
1747 * @notifier: the notifier to be notified
1749 void memory_region_unmap_iommu_notifier_range(IOMMUNotifier
*notifier
);
1753 * memory_region_register_iommu_notifier: register a notifier for changes to
1754 * IOMMU translation entries.
1756 * Returns 0 on success, or a negative errno otherwise. In particular,
1757 * -EINVAL indicates that at least one of the attributes of the notifier
1758 * is not supported (flag/range) by the IOMMU memory region. In case of error
1759 * the error object must be created.
1761 * @mr: the memory region to observe
1762 * @n: the IOMMUNotifier to be added; the notify callback receives a
1763 * pointer to an #IOMMUTLBEntry as the opaque value; the pointer
1764 * ceases to be valid on exit from the notifier.
1765 * @errp: pointer to Error*, to store an error if it happens.
1767 int memory_region_register_iommu_notifier(MemoryRegion
*mr
,
1768 IOMMUNotifier
*n
, Error
**errp
);
1771 * memory_region_iommu_replay: replay existing IOMMU translations to
1772 * a notifier with the minimum page granularity returned by
1773 * mr->iommu_ops->get_page_size().
1775 * Note: this is not related to record-and-replay functionality.
1777 * @iommu_mr: the memory region to observe
1778 * @n: the notifier to which to replay iommu mappings
1780 void memory_region_iommu_replay(IOMMUMemoryRegion
*iommu_mr
, IOMMUNotifier
*n
);
1783 * memory_region_unregister_iommu_notifier: unregister a notifier for
1784 * changes to IOMMU translation entries.
1786 * @mr: the memory region which was observed and for which notity_stopped()
1787 * needs to be called
1788 * @n: the notifier to be removed.
1790 void memory_region_unregister_iommu_notifier(MemoryRegion
*mr
,
1794 * memory_region_iommu_get_attr: return an IOMMU attr if get_attr() is
1795 * defined on the IOMMU.
1797 * Returns 0 on success, or a negative errno otherwise. In particular,
1798 * -EINVAL indicates that the IOMMU does not support the requested
1801 * @iommu_mr: the memory region
1802 * @attr: the requested attribute
1803 * @data: a pointer to the requested attribute data
1805 int memory_region_iommu_get_attr(IOMMUMemoryRegion
*iommu_mr
,
1806 enum IOMMUMemoryRegionAttr attr
,
1810 * memory_region_iommu_attrs_to_index: return the IOMMU index to
1811 * use for translations with the given memory transaction attributes.
1813 * @iommu_mr: the memory region
1814 * @attrs: the memory transaction attributes
1816 int memory_region_iommu_attrs_to_index(IOMMUMemoryRegion
*iommu_mr
,
1820 * memory_region_iommu_num_indexes: return the total number of IOMMU
1821 * indexes that this IOMMU supports.
1823 * @iommu_mr: the memory region
1825 int memory_region_iommu_num_indexes(IOMMUMemoryRegion
*iommu_mr
);
1828 * memory_region_iommu_set_page_size_mask: set the supported page
1829 * sizes for a given IOMMU memory region
1831 * @iommu_mr: IOMMU memory region
1832 * @page_size_mask: supported page size mask
1833 * @errp: pointer to Error*, to store an error if it happens.
1835 int memory_region_iommu_set_page_size_mask(IOMMUMemoryRegion
*iommu_mr
,
1836 uint64_t page_size_mask
,
1840 * memory_region_name: get a memory region's name
1842 * Returns the string that was used to initialize the memory region.
1844 * @mr: the memory region being queried
1846 const char *memory_region_name(const MemoryRegion
*mr
);
1849 * memory_region_is_logging: return whether a memory region is logging writes
1851 * Returns %true if the memory region is logging writes for the given client
1853 * @mr: the memory region being queried
1854 * @client: the client being queried
1856 bool memory_region_is_logging(MemoryRegion
*mr
, uint8_t client
);
1859 * memory_region_get_dirty_log_mask: return the clients for which a
1860 * memory region is logging writes.
1862 * Returns a bitmap of clients, in which the DIRTY_MEMORY_* constants
1863 * are the bit indices.
1865 * @mr: the memory region being queried
1867 uint8_t memory_region_get_dirty_log_mask(MemoryRegion
*mr
);
1870 * memory_region_is_rom: check whether a memory region is ROM
1872 * Returns %true if a memory region is read-only memory.
1874 * @mr: the memory region being queried
1876 static inline bool memory_region_is_rom(MemoryRegion
*mr
)
1878 return mr
->ram
&& mr
->readonly
;
1882 * memory_region_is_nonvolatile: check whether a memory region is non-volatile
1884 * Returns %true is a memory region is non-volatile memory.
1886 * @mr: the memory region being queried
1888 static inline bool memory_region_is_nonvolatile(MemoryRegion
*mr
)
1890 return mr
->nonvolatile
;
1894 * memory_region_get_fd: Get a file descriptor backing a RAM memory region.
1896 * Returns a file descriptor backing a file-based RAM memory region,
1897 * or -1 if the region is not a file-based RAM memory region.
1899 * @mr: the RAM or alias memory region being queried.
1901 int memory_region_get_fd(MemoryRegion
*mr
);
1904 * memory_region_from_host: Convert a pointer into a RAM memory region
1905 * and an offset within it.
1907 * Given a host pointer inside a RAM memory region (created with
1908 * memory_region_init_ram() or memory_region_init_ram_ptr()), return
1909 * the MemoryRegion and the offset within it.
1911 * Use with care; by the time this function returns, the returned pointer is
1912 * not protected by RCU anymore. If the caller is not within an RCU critical
1913 * section and does not hold the iothread lock, it must have other means of
1914 * protecting the pointer, such as a reference to the region that includes
1915 * the incoming ram_addr_t.
1917 * @ptr: the host pointer to be converted
1918 * @offset: the offset within memory region
1920 MemoryRegion
*memory_region_from_host(void *ptr
, ram_addr_t
*offset
);
1923 * memory_region_get_ram_ptr: Get a pointer into a RAM memory region.
1925 * Returns a host pointer to a RAM memory region (created with
1926 * memory_region_init_ram() or memory_region_init_ram_ptr()).
1928 * Use with care; by the time this function returns, the returned pointer is
1929 * not protected by RCU anymore. If the caller is not within an RCU critical
1930 * section and does not hold the iothread lock, it must have other means of
1931 * protecting the pointer, such as a reference to the region that includes
1932 * the incoming ram_addr_t.
1934 * @mr: the memory region being queried.
1936 void *memory_region_get_ram_ptr(MemoryRegion
*mr
);
1938 /* memory_region_ram_resize: Resize a RAM region.
1940 * Resizing RAM while migrating can result in the migration being canceled.
1941 * Care has to be taken if the guest might have already detected the memory.
1943 * @mr: a memory region created with @memory_region_init_resizeable_ram.
1944 * @newsize: the new size the region
1945 * @errp: pointer to Error*, to store an error if it happens.
1947 void memory_region_ram_resize(MemoryRegion
*mr
, ram_addr_t newsize
,
1951 * memory_region_msync: Synchronize selected address range of
1952 * a memory mapped region
1954 * @mr: the memory region to be msync
1955 * @addr: the initial address of the range to be sync
1956 * @size: the size of the range to be sync
1958 void memory_region_msync(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
1961 * memory_region_writeback: Trigger cache writeback for
1962 * selected address range
1964 * @mr: the memory region to be updated
1965 * @addr: the initial address of the range to be written back
1966 * @size: the size of the range to be written back
1968 void memory_region_writeback(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
1971 * memory_region_set_log: Turn dirty logging on or off for a region.
1973 * Turns dirty logging on or off for a specified client (display, migration).
1974 * Only meaningful for RAM regions.
1976 * @mr: the memory region being updated.
1977 * @log: whether dirty logging is to be enabled or disabled.
1978 * @client: the user of the logging information; %DIRTY_MEMORY_VGA only.
1980 void memory_region_set_log(MemoryRegion
*mr
, bool log
, unsigned client
);
1983 * memory_region_set_dirty: Mark a range of bytes as dirty in a memory region.
1985 * Marks a range of bytes as dirty, after it has been dirtied outside
1988 * @mr: the memory region being dirtied.
1989 * @addr: the address (relative to the start of the region) being dirtied.
1990 * @size: size of the range being dirtied.
1992 void memory_region_set_dirty(MemoryRegion
*mr
, hwaddr addr
,
1996 * memory_region_clear_dirty_bitmap - clear dirty bitmap for memory range
1998 * This function is called when the caller wants to clear the remote
1999 * dirty bitmap of a memory range within the memory region. This can
2000 * be used by e.g. KVM to manually clear dirty log when
2001 * KVM_CAP_MANUAL_DIRTY_LOG_PROTECT is declared support by the host
2004 * @mr: the memory region to clear the dirty log upon
2005 * @start: start address offset within the memory region
2006 * @len: length of the memory region to clear dirty bitmap
2008 void memory_region_clear_dirty_bitmap(MemoryRegion
*mr
, hwaddr start
,
2012 * memory_region_snapshot_and_clear_dirty: Get a snapshot of the dirty
2013 * bitmap and clear it.
2015 * Creates a snapshot of the dirty bitmap, clears the dirty bitmap and
2016 * returns the snapshot. The snapshot can then be used to query dirty
2017 * status, using memory_region_snapshot_get_dirty. Snapshotting allows
2018 * querying the same page multiple times, which is especially useful for
2019 * display updates where the scanlines often are not page aligned.
2021 * The dirty bitmap region which gets copied into the snapshot (and
2022 * cleared afterwards) can be larger than requested. The boundaries
2023 * are rounded up/down so complete bitmap longs (covering 64 pages on
2024 * 64bit hosts) can be copied over into the bitmap snapshot. Which
2025 * isn't a problem for display updates as the extra pages are outside
2026 * the visible area, and in case the visible area changes a full
2027 * display redraw is due anyway. Should other use cases for this
2028 * function emerge we might have to revisit this implementation
2031 * Use g_free to release DirtyBitmapSnapshot.
2033 * @mr: the memory region being queried.
2034 * @addr: the address (relative to the start of the region) being queried.
2035 * @size: the size of the range being queried.
2036 * @client: the user of the logging information; typically %DIRTY_MEMORY_VGA.
2038 DirtyBitmapSnapshot
*memory_region_snapshot_and_clear_dirty(MemoryRegion
*mr
,
2044 * memory_region_snapshot_get_dirty: Check whether a range of bytes is dirty
2045 * in the specified dirty bitmap snapshot.
2047 * @mr: the memory region being queried.
2048 * @snap: the dirty bitmap snapshot
2049 * @addr: the address (relative to the start of the region) being queried.
2050 * @size: the size of the range being queried.
2052 bool memory_region_snapshot_get_dirty(MemoryRegion
*mr
,
2053 DirtyBitmapSnapshot
*snap
,
2054 hwaddr addr
, hwaddr size
);
2057 * memory_region_reset_dirty: Mark a range of pages as clean, for a specified
2060 * Marks a range of pages as no longer dirty.
2062 * @mr: the region being updated.
2063 * @addr: the start of the subrange being cleaned.
2064 * @size: the size of the subrange being cleaned.
2065 * @client: the user of the logging information; %DIRTY_MEMORY_MIGRATION or
2066 * %DIRTY_MEMORY_VGA.
2068 void memory_region_reset_dirty(MemoryRegion
*mr
, hwaddr addr
,
2069 hwaddr size
, unsigned client
);
2072 * memory_region_flush_rom_device: Mark a range of pages dirty and invalidate
2073 * TBs (for self-modifying code).
2075 * The MemoryRegionOps->write() callback of a ROM device must use this function
2076 * to mark byte ranges that have been modified internally, such as by directly
2077 * accessing the memory returned by memory_region_get_ram_ptr().
2079 * This function marks the range dirty and invalidates TBs so that TCG can
2080 * detect self-modifying code.
2082 * @mr: the region being flushed.
2083 * @addr: the start, relative to the start of the region, of the range being
2085 * @size: the size, in bytes, of the range being flushed.
2087 void memory_region_flush_rom_device(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
2090 * memory_region_set_readonly: Turn a memory region read-only (or read-write)
2092 * Allows a memory region to be marked as read-only (turning it into a ROM).
2093 * only useful on RAM regions.
2095 * @mr: the region being updated.
2096 * @readonly: whether rhe region is to be ROM or RAM.
2098 void memory_region_set_readonly(MemoryRegion
*mr
, bool readonly
);
2101 * memory_region_set_nonvolatile: Turn a memory region non-volatile
2103 * Allows a memory region to be marked as non-volatile.
2104 * only useful on RAM regions.
2106 * @mr: the region being updated.
2107 * @nonvolatile: whether rhe region is to be non-volatile.
2109 void memory_region_set_nonvolatile(MemoryRegion
*mr
, bool nonvolatile
);
2112 * memory_region_rom_device_set_romd: enable/disable ROMD mode
2114 * Allows a ROM device (initialized with memory_region_init_rom_device() to
2115 * set to ROMD mode (default) or MMIO mode. When it is in ROMD mode, the
2116 * device is mapped to guest memory and satisfies read access directly.
2117 * When in MMIO mode, reads are forwarded to the #MemoryRegion.read function.
2118 * Writes are always handled by the #MemoryRegion.write function.
2120 * @mr: the memory region to be updated
2121 * @romd_mode: %true to put the region into ROMD mode
2123 void memory_region_rom_device_set_romd(MemoryRegion
*mr
, bool romd_mode
);
2126 * memory_region_set_coalescing: Enable memory coalescing for the region.
2128 * Enabled writes to a region to be queued for later processing. MMIO ->write
2129 * callbacks may be delayed until a non-coalesced MMIO is issued.
2130 * Only useful for IO regions. Roughly similar to write-combining hardware.
2132 * @mr: the memory region to be write coalesced
2134 void memory_region_set_coalescing(MemoryRegion
*mr
);
2137 * memory_region_add_coalescing: Enable memory coalescing for a sub-range of
2140 * Like memory_region_set_coalescing(), but works on a sub-range of a region.
2141 * Multiple calls can be issued coalesced disjoint ranges.
2143 * @mr: the memory region to be updated.
2144 * @offset: the start of the range within the region to be coalesced.
2145 * @size: the size of the subrange to be coalesced.
2147 void memory_region_add_coalescing(MemoryRegion
*mr
,
2152 * memory_region_clear_coalescing: Disable MMIO coalescing for the region.
2154 * Disables any coalescing caused by memory_region_set_coalescing() or
2155 * memory_region_add_coalescing(). Roughly equivalent to uncacheble memory
2158 * @mr: the memory region to be updated.
2160 void memory_region_clear_coalescing(MemoryRegion
*mr
);
2163 * memory_region_set_flush_coalesced: Enforce memory coalescing flush before
2166 * Ensure that pending coalesced MMIO request are flushed before the memory
2167 * region is accessed. This property is automatically enabled for all regions
2168 * passed to memory_region_set_coalescing() and memory_region_add_coalescing().
2170 * @mr: the memory region to be updated.
2172 void memory_region_set_flush_coalesced(MemoryRegion
*mr
);
2175 * memory_region_clear_flush_coalesced: Disable memory coalescing flush before
2178 * Clear the automatic coalesced MMIO flushing enabled via
2179 * memory_region_set_flush_coalesced. Note that this service has no effect on
2180 * memory regions that have MMIO coalescing enabled for themselves. For them,
2181 * automatic flushing will stop once coalescing is disabled.
2183 * @mr: the memory region to be updated.
2185 void memory_region_clear_flush_coalesced(MemoryRegion
*mr
);
2188 * memory_region_add_eventfd: Request an eventfd to be triggered when a word
2189 * is written to a location.
2191 * Marks a word in an IO region (initialized with memory_region_init_io())
2192 * as a trigger for an eventfd event. The I/O callback will not be called.
2193 * The caller must be prepared to handle failure (that is, take the required
2194 * action if the callback _is_ called).
2196 * @mr: the memory region being updated.
2197 * @addr: the address within @mr that is to be monitored
2198 * @size: the size of the access to trigger the eventfd
2199 * @match_data: whether to match against @data, instead of just @addr
2200 * @data: the data to match against the guest write
2201 * @e: event notifier to be triggered when @addr, @size, and @data all match.
2203 void memory_region_add_eventfd(MemoryRegion
*mr
,
2211 * memory_region_del_eventfd: Cancel an eventfd.
2213 * Cancels an eventfd trigger requested by a previous
2214 * memory_region_add_eventfd() call.
2216 * @mr: the memory region being updated.
2217 * @addr: the address within @mr that is to be monitored
2218 * @size: the size of the access to trigger the eventfd
2219 * @match_data: whether to match against @data, instead of just @addr
2220 * @data: the data to match against the guest write
2221 * @e: event notifier to be triggered when @addr, @size, and @data all match.
2223 void memory_region_del_eventfd(MemoryRegion
*mr
,
2231 * memory_region_add_subregion: Add a subregion to a container.
2233 * Adds a subregion at @offset. The subregion may not overlap with other
2234 * subregions (except for those explicitly marked as overlapping). A region
2235 * may only be added once as a subregion (unless removed with
2236 * memory_region_del_subregion()); use memory_region_init_alias() if you
2237 * want a region to be a subregion in multiple locations.
2239 * @mr: the region to contain the new subregion; must be a container
2240 * initialized with memory_region_init().
2241 * @offset: the offset relative to @mr where @subregion is added.
2242 * @subregion: the subregion to be added.
2244 void memory_region_add_subregion(MemoryRegion
*mr
,
2246 MemoryRegion
*subregion
);
2248 * memory_region_add_subregion_overlap: Add a subregion to a container
2251 * Adds a subregion at @offset. The subregion may overlap with other
2252 * subregions. Conflicts are resolved by having a higher @priority hide a
2253 * lower @priority. Subregions without priority are taken as @priority 0.
2254 * A region may only be added once as a subregion (unless removed with
2255 * memory_region_del_subregion()); use memory_region_init_alias() if you
2256 * want a region to be a subregion in multiple locations.
2258 * @mr: the region to contain the new subregion; must be a container
2259 * initialized with memory_region_init().
2260 * @offset: the offset relative to @mr where @subregion is added.
2261 * @subregion: the subregion to be added.
2262 * @priority: used for resolving overlaps; highest priority wins.
2264 void memory_region_add_subregion_overlap(MemoryRegion
*mr
,
2266 MemoryRegion
*subregion
,
2270 * memory_region_get_ram_addr: Get the ram address associated with a memory
2273 * @mr: the region to be queried
2275 ram_addr_t
memory_region_get_ram_addr(MemoryRegion
*mr
);
2277 uint64_t memory_region_get_alignment(const MemoryRegion
*mr
);
2279 * memory_region_del_subregion: Remove a subregion.
2281 * Removes a subregion from its container.
2283 * @mr: the container to be updated.
2284 * @subregion: the region being removed; must be a current subregion of @mr.
2286 void memory_region_del_subregion(MemoryRegion
*mr
,
2287 MemoryRegion
*subregion
);
2290 * memory_region_set_enabled: dynamically enable or disable a region
2292 * Enables or disables a memory region. A disabled memory region
2293 * ignores all accesses to itself and its subregions. It does not
2294 * obscure sibling subregions with lower priority - it simply behaves as
2295 * if it was removed from the hierarchy.
2297 * Regions default to being enabled.
2299 * @mr: the region to be updated
2300 * @enabled: whether to enable or disable the region
2302 void memory_region_set_enabled(MemoryRegion
*mr
, bool enabled
);
2305 * memory_region_set_address: dynamically update the address of a region
2307 * Dynamically updates the address of a region, relative to its container.
2308 * May be used on regions are currently part of a memory hierarchy.
2310 * @mr: the region to be updated
2311 * @addr: new address, relative to container region
2313 void memory_region_set_address(MemoryRegion
*mr
, hwaddr addr
);
2316 * memory_region_set_size: dynamically update the size of a region.
2318 * Dynamically updates the size of a region.
2320 * @mr: the region to be updated
2321 * @size: used size of the region.
2323 void memory_region_set_size(MemoryRegion
*mr
, uint64_t size
);
2326 * memory_region_set_alias_offset: dynamically update a memory alias's offset
2328 * Dynamically updates the offset into the target region that an alias points
2329 * to, as if the fourth argument to memory_region_init_alias() has changed.
2331 * @mr: the #MemoryRegion to be updated; should be an alias.
2332 * @offset: the new offset into the target memory region
2334 void memory_region_set_alias_offset(MemoryRegion
*mr
,
2338 * memory_region_present: checks if an address relative to a @container
2339 * translates into #MemoryRegion within @container
2341 * Answer whether a #MemoryRegion within @container covers the address
2344 * @container: a #MemoryRegion within which @addr is a relative address
2345 * @addr: the area within @container to be searched
2347 bool memory_region_present(MemoryRegion
*container
, hwaddr addr
);
2350 * memory_region_is_mapped: returns true if #MemoryRegion is mapped
2351 * into another memory region, which does not necessarily imply that it is
2352 * mapped into an address space.
2354 * @mr: a #MemoryRegion which should be checked if it's mapped
2356 bool memory_region_is_mapped(MemoryRegion
*mr
);
2359 * memory_region_get_ram_discard_manager: get the #RamDiscardManager for a
2362 * The #RamDiscardManager cannot change while a memory region is mapped.
2364 * @mr: the #MemoryRegion
2366 RamDiscardManager
*memory_region_get_ram_discard_manager(MemoryRegion
*mr
);
2369 * memory_region_has_ram_discard_manager: check whether a #MemoryRegion has a
2370 * #RamDiscardManager assigned
2372 * @mr: the #MemoryRegion
2374 static inline bool memory_region_has_ram_discard_manager(MemoryRegion
*mr
)
2376 return !!memory_region_get_ram_discard_manager(mr
);
2380 * memory_region_set_ram_discard_manager: set the #RamDiscardManager for a
2383 * This function must not be called for a mapped #MemoryRegion, a #MemoryRegion
2384 * that does not cover RAM, or a #MemoryRegion that already has a
2385 * #RamDiscardManager assigned.
2387 * @mr: the #MemoryRegion
2388 * @rdm: #RamDiscardManager to set
2390 void memory_region_set_ram_discard_manager(MemoryRegion
*mr
,
2391 RamDiscardManager
*rdm
);
2394 * memory_region_find: translate an address/size relative to a
2395 * MemoryRegion into a #MemoryRegionSection.
2397 * Locates the first #MemoryRegion within @mr that overlaps the range
2398 * given by @addr and @size.
2400 * Returns a #MemoryRegionSection that describes a contiguous overlap.
2401 * It will have the following characteristics:
2402 * - @size = 0 iff no overlap was found
2403 * - @mr is non-%NULL iff an overlap was found
2405 * Remember that in the return value the @offset_within_region is
2406 * relative to the returned region (in the .@mr field), not to the
2409 * Similarly, the .@offset_within_address_space is relative to the
2410 * address space that contains both regions, the passed and the
2411 * returned one. However, in the special case where the @mr argument
2412 * has no container (and thus is the root of the address space), the
2413 * following will hold:
2414 * - @offset_within_address_space >= @addr
2415 * - @offset_within_address_space + .@size <= @addr + @size
2417 * @mr: a MemoryRegion within which @addr is a relative address
2418 * @addr: start of the area within @as to be searched
2419 * @size: size of the area to be searched
2421 MemoryRegionSection
memory_region_find(MemoryRegion
*mr
,
2422 hwaddr addr
, uint64_t size
);
2425 * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2427 * Synchronizes the dirty page log for all address spaces.
2429 * @last_stage: whether this is the last stage of live migration
2431 void memory_global_dirty_log_sync(bool last_stage
);
2434 * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2436 * Synchronizes the vCPUs with a thread that is reading the dirty bitmap.
2437 * This function must be called after the dirty log bitmap is cleared, and
2438 * before dirty guest memory pages are read. If you are using
2439 * #DirtyBitmapSnapshot, memory_region_snapshot_and_clear_dirty() takes
2440 * care of doing this.
2442 void memory_global_after_dirty_log_sync(void);
2445 * memory_region_transaction_begin: Start a transaction.
2447 * During a transaction, changes will be accumulated and made visible
2448 * only when the transaction ends (is committed).
2450 void memory_region_transaction_begin(void);
2453 * memory_region_transaction_commit: Commit a transaction and make changes
2454 * visible to the guest.
2456 void memory_region_transaction_commit(void);
2459 * memory_listener_register: register callbacks to be called when memory
2460 * sections are mapped or unmapped into an address
2463 * @listener: an object containing the callbacks to be called
2464 * @filter: if non-%NULL, only regions in this address space will be observed
2466 void memory_listener_register(MemoryListener
*listener
, AddressSpace
*filter
);
2469 * memory_listener_unregister: undo the effect of memory_listener_register()
2471 * @listener: an object containing the callbacks to be removed
2473 void memory_listener_unregister(MemoryListener
*listener
);
2476 * memory_global_dirty_log_start: begin dirty logging for all regions
2478 * @flags: purpose of starting dirty log, migration or dirty rate
2480 void memory_global_dirty_log_start(unsigned int flags
);
2483 * memory_global_dirty_log_stop: end dirty logging for all regions
2485 * @flags: purpose of stopping dirty log, migration or dirty rate
2487 void memory_global_dirty_log_stop(unsigned int flags
);
2489 void mtree_info(bool flatview
, bool dispatch_tree
, bool owner
, bool disabled
);
2491 bool memory_region_access_valid(MemoryRegion
*mr
, hwaddr addr
,
2492 unsigned size
, bool is_write
,
2496 * memory_region_dispatch_read: perform a read directly to the specified
2499 * @mr: #MemoryRegion to access
2500 * @addr: address within that region
2501 * @pval: pointer to uint64_t which the data is written to
2502 * @op: size, sign, and endianness of the memory operation
2503 * @attrs: memory transaction attributes to use for the access
2505 MemTxResult
memory_region_dispatch_read(MemoryRegion
*mr
,
2511 * memory_region_dispatch_write: perform a write directly to the specified
2514 * @mr: #MemoryRegion to access
2515 * @addr: address within that region
2516 * @data: data to write
2517 * @op: size, sign, and endianness of the memory operation
2518 * @attrs: memory transaction attributes to use for the access
2520 MemTxResult
memory_region_dispatch_write(MemoryRegion
*mr
,
2527 * address_space_init: initializes an address space
2529 * @as: an uninitialized #AddressSpace
2530 * @root: a #MemoryRegion that routes addresses for the address space
2531 * @name: an address space name. The name is only used for debugging
2534 void address_space_init(AddressSpace
*as
, MemoryRegion
*root
, const char *name
);
2537 * address_space_destroy: destroy an address space
2539 * Releases all resources associated with an address space. After an address space
2540 * is destroyed, its root memory region (given by address_space_init()) may be destroyed
2543 * @as: address space to be destroyed
2545 void address_space_destroy(AddressSpace
*as
);
2548 * address_space_remove_listeners: unregister all listeners of an address space
2550 * Removes all callbacks previously registered with memory_listener_register()
2553 * @as: an initialized #AddressSpace
2555 void address_space_remove_listeners(AddressSpace
*as
);
2558 * address_space_rw: read from or write to an address space.
2560 * Return a MemTxResult indicating whether the operation succeeded
2561 * or failed (eg unassigned memory, device rejected the transaction,
2564 * @as: #AddressSpace to be accessed
2565 * @addr: address within that address space
2566 * @attrs: memory transaction attributes
2567 * @buf: buffer with the data transferred
2568 * @len: the number of bytes to read or write
2569 * @is_write: indicates the transfer direction
2571 MemTxResult
address_space_rw(AddressSpace
*as
, hwaddr addr
,
2572 MemTxAttrs attrs
, void *buf
,
2573 hwaddr len
, bool is_write
);
2576 * address_space_write: write to address space.
2578 * Return a MemTxResult indicating whether the operation succeeded
2579 * or failed (eg unassigned memory, device rejected the transaction,
2582 * @as: #AddressSpace to be accessed
2583 * @addr: address within that address space
2584 * @attrs: memory transaction attributes
2585 * @buf: buffer with the data transferred
2586 * @len: the number of bytes to write
2588 MemTxResult
address_space_write(AddressSpace
*as
, hwaddr addr
,
2590 const void *buf
, hwaddr len
);
2593 * address_space_write_rom: write to address space, including ROM.
2595 * This function writes to the specified address space, but will
2596 * write data to both ROM and RAM. This is used for non-guest
2597 * writes like writes from the gdb debug stub or initial loading
2600 * Note that portions of the write which attempt to write data to
2601 * a device will be silently ignored -- only real RAM and ROM will
2604 * Return a MemTxResult indicating whether the operation succeeded
2605 * or failed (eg unassigned memory, device rejected the transaction,
2608 * @as: #AddressSpace to be accessed
2609 * @addr: address within that address space
2610 * @attrs: memory transaction attributes
2611 * @buf: buffer with the data transferred
2612 * @len: the number of bytes to write
2614 MemTxResult
address_space_write_rom(AddressSpace
*as
, hwaddr addr
,
2616 const void *buf
, hwaddr len
);
2618 /* address_space_ld*: load from an address space
2619 * address_space_st*: store to an address space
2621 * These functions perform a load or store of the byte, word,
2622 * longword or quad to the specified address within the AddressSpace.
2623 * The _le suffixed functions treat the data as little endian;
2624 * _be indicates big endian; no suffix indicates "same endianness
2627 * The "guest CPU endianness" accessors are deprecated for use outside
2628 * target-* code; devices should be CPU-agnostic and use either the LE
2629 * or the BE accessors.
2631 * @as #AddressSpace to be accessed
2632 * @addr: address within that address space
2633 * @val: data value, for stores
2634 * @attrs: memory transaction attributes
2635 * @result: location to write the success/failure of the transaction;
2636 * if NULL, this information is discarded
2641 #define ARG1_DECL AddressSpace *as
2642 #include "exec/memory_ldst.h.inc"
2646 #define ARG1_DECL AddressSpace *as
2647 #include "exec/memory_ldst_phys.h.inc"
2649 struct MemoryRegionCache
{
2654 MemoryRegionSection mrs
;
2658 #define MEMORY_REGION_CACHE_INVALID ((MemoryRegionCache) { .mrs.mr = NULL })
2661 /* address_space_ld*_cached: load from a cached #MemoryRegion
2662 * address_space_st*_cached: store into a cached #MemoryRegion
2664 * These functions perform a load or store of the byte, word,
2665 * longword or quad to the specified address. The address is
2666 * a physical address in the AddressSpace, but it must lie within
2667 * a #MemoryRegion that was mapped with address_space_cache_init.
2669 * The _le suffixed functions treat the data as little endian;
2670 * _be indicates big endian; no suffix indicates "same endianness
2673 * The "guest CPU endianness" accessors are deprecated for use outside
2674 * target-* code; devices should be CPU-agnostic and use either the LE
2675 * or the BE accessors.
2677 * @cache: previously initialized #MemoryRegionCache to be accessed
2678 * @addr: address within the address space
2679 * @val: data value, for stores
2680 * @attrs: memory transaction attributes
2681 * @result: location to write the success/failure of the transaction;
2682 * if NULL, this information is discarded
2685 #define SUFFIX _cached_slow
2687 #define ARG1_DECL MemoryRegionCache *cache
2688 #include "exec/memory_ldst.h.inc"
2690 /* Inline fast path for direct RAM access. */
2691 static inline uint8_t address_space_ldub_cached(MemoryRegionCache
*cache
,
2692 hwaddr addr
, MemTxAttrs attrs
, MemTxResult
*result
)
2694 assert(addr
< cache
->len
);
2695 if (likely(cache
->ptr
)) {
2696 return ldub_p(cache
->ptr
+ addr
);
2698 return address_space_ldub_cached_slow(cache
, addr
, attrs
, result
);
2702 static inline void address_space_stb_cached(MemoryRegionCache
*cache
,
2703 hwaddr addr
, uint8_t val
, MemTxAttrs attrs
, MemTxResult
*result
)
2705 assert(addr
< cache
->len
);
2706 if (likely(cache
->ptr
)) {
2707 stb_p(cache
->ptr
+ addr
, val
);
2709 address_space_stb_cached_slow(cache
, addr
, val
, attrs
, result
);
2713 #define ENDIANNESS _le
2714 #include "exec/memory_ldst_cached.h.inc"
2716 #define ENDIANNESS _be
2717 #include "exec/memory_ldst_cached.h.inc"
2719 #define SUFFIX _cached
2721 #define ARG1_DECL MemoryRegionCache *cache
2722 #include "exec/memory_ldst_phys.h.inc"
2724 /* address_space_cache_init: prepare for repeated access to a physical
2727 * @cache: #MemoryRegionCache to be filled
2728 * @as: #AddressSpace to be accessed
2729 * @addr: address within that address space
2730 * @len: length of buffer
2731 * @is_write: indicates the transfer direction
2733 * Will only work with RAM, and may map a subset of the requested range by
2734 * returning a value that is less than @len. On failure, return a negative
2737 * Because it only works with RAM, this function can be used for
2738 * read-modify-write operations. In this case, is_write should be %true.
2740 * Note that addresses passed to the address_space_*_cached functions
2741 * are relative to @addr.
2743 int64_t address_space_cache_init(MemoryRegionCache
*cache
,
2750 * address_space_cache_invalidate: complete a write to a #MemoryRegionCache
2752 * @cache: The #MemoryRegionCache to operate on.
2753 * @addr: The first physical address that was written, relative to the
2754 * address that was passed to @address_space_cache_init.
2755 * @access_len: The number of bytes that were written starting at @addr.
2757 void address_space_cache_invalidate(MemoryRegionCache
*cache
,
2762 * address_space_cache_destroy: free a #MemoryRegionCache
2764 * @cache: The #MemoryRegionCache whose memory should be released.
2766 void address_space_cache_destroy(MemoryRegionCache
*cache
);
2768 /* address_space_get_iotlb_entry: translate an address into an IOTLB
2769 * entry. Should be called from an RCU critical section.
2771 IOMMUTLBEntry
address_space_get_iotlb_entry(AddressSpace
*as
, hwaddr addr
,
2772 bool is_write
, MemTxAttrs attrs
);
2774 /* address_space_translate: translate an address range into an address space
2775 * into a MemoryRegion and an address range into that section. Should be
2776 * called from an RCU critical section, to avoid that the last reference
2777 * to the returned region disappears after address_space_translate returns.
2779 * @fv: #FlatView to be accessed
2780 * @addr: address within that address space
2781 * @xlat: pointer to address within the returned memory region section's
2783 * @len: pointer to length
2784 * @is_write: indicates the transfer direction
2785 * @attrs: memory attributes
2787 MemoryRegion
*flatview_translate(FlatView
*fv
,
2788 hwaddr addr
, hwaddr
*xlat
,
2789 hwaddr
*len
, bool is_write
,
2792 static inline MemoryRegion
*address_space_translate(AddressSpace
*as
,
2793 hwaddr addr
, hwaddr
*xlat
,
2794 hwaddr
*len
, bool is_write
,
2797 return flatview_translate(address_space_to_flatview(as
),
2798 addr
, xlat
, len
, is_write
, attrs
);
2801 /* address_space_access_valid: check for validity of accessing an address
2804 * Check whether memory is assigned to the given address space range, and
2805 * access is permitted by any IOMMU regions that are active for the address
2808 * For now, addr and len should be aligned to a page size. This limitation
2809 * will be lifted in the future.
2811 * @as: #AddressSpace to be accessed
2812 * @addr: address within that address space
2813 * @len: length of the area to be checked
2814 * @is_write: indicates the transfer direction
2815 * @attrs: memory attributes
2817 bool address_space_access_valid(AddressSpace
*as
, hwaddr addr
, hwaddr len
,
2818 bool is_write
, MemTxAttrs attrs
);
2820 /* address_space_map: map a physical memory region into a host virtual address
2822 * May map a subset of the requested range, given by and returned in @plen.
2823 * May return %NULL and set *@plen to zero(0), if resources needed to perform
2824 * the mapping are exhausted.
2825 * Use only for reads OR writes - not for read-modify-write operations.
2826 * Use cpu_register_map_client() to know when retrying the map operation is
2827 * likely to succeed.
2829 * @as: #AddressSpace to be accessed
2830 * @addr: address within that address space
2831 * @plen: pointer to length of buffer; updated on return
2832 * @is_write: indicates the transfer direction
2833 * @attrs: memory attributes
2835 void *address_space_map(AddressSpace
*as
, hwaddr addr
,
2836 hwaddr
*plen
, bool is_write
, MemTxAttrs attrs
);
2838 /* address_space_unmap: Unmaps a memory region previously mapped by address_space_map()
2840 * Will also mark the memory as dirty if @is_write == %true. @access_len gives
2841 * the amount of memory that was actually read or written by the caller.
2843 * @as: #AddressSpace used
2844 * @buffer: host pointer as returned by address_space_map()
2845 * @len: buffer length as returned by address_space_map()
2846 * @access_len: amount of data actually transferred
2847 * @is_write: indicates the transfer direction
2849 void address_space_unmap(AddressSpace
*as
, void *buffer
, hwaddr len
,
2850 bool is_write
, hwaddr access_len
);
2853 /* Internal functions, part of the implementation of address_space_read. */
2854 MemTxResult
address_space_read_full(AddressSpace
*as
, hwaddr addr
,
2855 MemTxAttrs attrs
, void *buf
, hwaddr len
);
2856 MemTxResult
flatview_read_continue(FlatView
*fv
, hwaddr addr
,
2857 MemTxAttrs attrs
, void *buf
,
2858 hwaddr len
, hwaddr addr1
, hwaddr l
,
2860 void *qemu_map_ram_ptr(RAMBlock
*ram_block
, ram_addr_t addr
);
2862 /* Internal functions, part of the implementation of address_space_read_cached
2863 * and address_space_write_cached. */
2864 MemTxResult
address_space_read_cached_slow(MemoryRegionCache
*cache
,
2865 hwaddr addr
, void *buf
, hwaddr len
);
2866 MemTxResult
address_space_write_cached_slow(MemoryRegionCache
*cache
,
2867 hwaddr addr
, const void *buf
,
2870 int memory_access_size(MemoryRegion
*mr
, unsigned l
, hwaddr addr
);
2871 bool prepare_mmio_access(MemoryRegion
*mr
);
2873 static inline bool memory_access_is_direct(MemoryRegion
*mr
, bool is_write
)
2876 return memory_region_is_ram(mr
) && !mr
->readonly
&&
2877 !mr
->rom_device
&& !memory_region_is_ram_device(mr
);
2879 return (memory_region_is_ram(mr
) && !memory_region_is_ram_device(mr
)) ||
2880 memory_region_is_romd(mr
);
2885 * address_space_read: read from an address space.
2887 * Return a MemTxResult indicating whether the operation succeeded
2888 * or failed (eg unassigned memory, device rejected the transaction,
2889 * IOMMU fault). Called within RCU critical section.
2891 * @as: #AddressSpace to be accessed
2892 * @addr: address within that address space
2893 * @attrs: memory transaction attributes
2894 * @buf: buffer with the data transferred
2895 * @len: length of the data transferred
2897 static inline __attribute__((__always_inline__
))
2898 MemTxResult
address_space_read(AddressSpace
*as
, hwaddr addr
,
2899 MemTxAttrs attrs
, void *buf
,
2902 MemTxResult result
= MEMTX_OK
;
2908 if (__builtin_constant_p(len
)) {
2910 RCU_READ_LOCK_GUARD();
2911 fv
= address_space_to_flatview(as
);
2913 mr
= flatview_translate(fv
, addr
, &addr1
, &l
, false, attrs
);
2914 if (len
== l
&& memory_access_is_direct(mr
, false)) {
2915 ptr
= qemu_map_ram_ptr(mr
->ram_block
, addr1
);
2916 memcpy(buf
, ptr
, len
);
2918 result
= flatview_read_continue(fv
, addr
, attrs
, buf
, len
,
2923 result
= address_space_read_full(as
, addr
, attrs
, buf
, len
);
2929 * address_space_read_cached: read from a cached RAM region
2931 * @cache: Cached region to be addressed
2932 * @addr: address relative to the base of the RAM region
2933 * @buf: buffer with the data transferred
2934 * @len: length of the data transferred
2936 static inline MemTxResult
2937 address_space_read_cached(MemoryRegionCache
*cache
, hwaddr addr
,
2938 void *buf
, hwaddr len
)
2940 assert(addr
< cache
->len
&& len
<= cache
->len
- addr
);
2941 fuzz_dma_read_cb(cache
->xlat
+ addr
, len
, cache
->mrs
.mr
);
2942 if (likely(cache
->ptr
)) {
2943 memcpy(buf
, cache
->ptr
+ addr
, len
);
2946 return address_space_read_cached_slow(cache
, addr
, buf
, len
);
2951 * address_space_write_cached: write to a cached RAM region
2953 * @cache: Cached region to be addressed
2954 * @addr: address relative to the base of the RAM region
2955 * @buf: buffer with the data transferred
2956 * @len: length of the data transferred
2958 static inline MemTxResult
2959 address_space_write_cached(MemoryRegionCache
*cache
, hwaddr addr
,
2960 const void *buf
, hwaddr len
)
2962 assert(addr
< cache
->len
&& len
<= cache
->len
- addr
);
2963 if (likely(cache
->ptr
)) {
2964 memcpy(cache
->ptr
+ addr
, buf
, len
);
2967 return address_space_write_cached_slow(cache
, addr
, buf
, len
);
2972 * address_space_set: Fill address space with a constant byte.
2974 * Return a MemTxResult indicating whether the operation succeeded
2975 * or failed (eg unassigned memory, device rejected the transaction,
2978 * @as: #AddressSpace to be accessed
2979 * @addr: address within that address space
2980 * @c: constant byte to fill the memory
2981 * @len: the number of bytes to fill with the constant byte
2982 * @attrs: memory transaction attributes
2984 MemTxResult
address_space_set(AddressSpace
*as
, hwaddr addr
,
2985 uint8_t c
, hwaddr len
, MemTxAttrs attrs
);
2988 /* enum device_endian to MemOp. */
2989 static inline MemOp
devend_memop(enum device_endian end
)
2991 QEMU_BUILD_BUG_ON(DEVICE_HOST_ENDIAN
!= DEVICE_LITTLE_ENDIAN
&&
2992 DEVICE_HOST_ENDIAN
!= DEVICE_BIG_ENDIAN
);
2994 #if HOST_BIG_ENDIAN != TARGET_BIG_ENDIAN
2995 /* Swap if non-host endianness or native (target) endianness */
2996 return (end
== DEVICE_HOST_ENDIAN
) ? 0 : MO_BSWAP
;
2998 const int non_host_endianness
=
2999 DEVICE_LITTLE_ENDIAN
^ DEVICE_BIG_ENDIAN
^ DEVICE_HOST_ENDIAN
;
3001 /* In this case, native (target) endianness needs no swap. */
3002 return (end
== non_host_endianness
) ? MO_BSWAP
: 0;
3008 * Inhibit technologies that require discarding of pages in RAM blocks, e.g.,
3009 * to manage the actual amount of memory consumed by the VM (then, the memory
3010 * provided by RAM blocks might be bigger than the desired memory consumption).
3011 * This *must* be set if:
3012 * - Discarding parts of a RAM blocks does not result in the change being
3013 * reflected in the VM and the pages getting freed.
3014 * - All memory in RAM blocks is pinned or duplicated, invaldiating any previous
3016 * - Discarding parts of a RAM blocks will result in integrity issues (e.g.,
3018 * Technologies that only temporarily pin the current working set of a
3019 * driver are fine, because we don't expect such pages to be discarded
3020 * (esp. based on guest action like balloon inflation).
3022 * This is *not* to be used to protect from concurrent discards (esp.,
3025 * Returns 0 if successful. Returns -EBUSY if a technology that relies on
3026 * discards to work reliably is active.
3028 int ram_block_discard_disable(bool state
);
3031 * See ram_block_discard_disable(): only disable uncoordinated discards,
3032 * keeping coordinated discards (via the RamDiscardManager) enabled.
3034 int ram_block_uncoordinated_discard_disable(bool state
);
3037 * Inhibit technologies that disable discarding of pages in RAM blocks.
3039 * Returns 0 if successful. Returns -EBUSY if discards are already set to
3042 int ram_block_discard_require(bool state
);
3045 * See ram_block_discard_require(): only inhibit technologies that disable
3046 * uncoordinated discarding of pages in RAM blocks, allowing co-existance with
3047 * technologies that only inhibit uncoordinated discards (via the
3048 * RamDiscardManager).
3050 int ram_block_coordinated_discard_require(bool state
);
3053 * Test if any discarding of memory in ram blocks is disabled.
3055 bool ram_block_discard_is_disabled(void);
3058 * Test if any discarding of memory in ram blocks is required to work reliably.
3060 bool ram_block_discard_is_required(void);