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).
134 IOMMU_NOTIFIER_NONE
= 0,
135 /* Notify cache invalidations */
136 IOMMU_NOTIFIER_UNMAP
= 0x1,
137 /* Notify entry changes (newly created entries) */
138 IOMMU_NOTIFIER_MAP
= 0x2,
139 /* Notify changes on device IOTLB entries */
140 IOMMU_NOTIFIER_DEVIOTLB_UNMAP
= 0x04,
143 #define IOMMU_NOTIFIER_IOTLB_EVENTS (IOMMU_NOTIFIER_MAP | IOMMU_NOTIFIER_UNMAP)
144 #define IOMMU_NOTIFIER_DEVIOTLB_EVENTS IOMMU_NOTIFIER_DEVIOTLB_UNMAP
145 #define IOMMU_NOTIFIER_ALL (IOMMU_NOTIFIER_IOTLB_EVENTS | \
146 IOMMU_NOTIFIER_DEVIOTLB_EVENTS)
148 struct IOMMUNotifier
;
149 typedef void (*IOMMUNotify
)(struct IOMMUNotifier
*notifier
,
150 IOMMUTLBEntry
*data
);
152 struct IOMMUNotifier
{
154 IOMMUNotifierFlag notifier_flags
;
155 /* Notify for address space range start <= addr <= end */
159 QLIST_ENTRY(IOMMUNotifier
) node
;
161 typedef struct IOMMUNotifier IOMMUNotifier
;
163 typedef struct IOMMUTLBEvent
{
164 IOMMUNotifierFlag type
;
168 /* RAM is pre-allocated and passed into qemu_ram_alloc_from_ptr */
169 #define RAM_PREALLOC (1 << 0)
171 /* RAM is mmap-ed with MAP_SHARED */
172 #define RAM_SHARED (1 << 1)
174 /* Only a portion of RAM (used_length) is actually used, and migrated.
175 * Resizing RAM while migrating can result in the migration being canceled.
177 #define RAM_RESIZEABLE (1 << 2)
179 /* UFFDIO_ZEROPAGE is available on this RAMBlock to atomically
180 * zero the page and wake waiting processes.
181 * (Set during postcopy)
183 #define RAM_UF_ZEROPAGE (1 << 3)
185 /* RAM can be migrated */
186 #define RAM_MIGRATABLE (1 << 4)
188 /* RAM is a persistent kind memory */
189 #define RAM_PMEM (1 << 5)
193 * UFFDIO_WRITEPROTECT is used on this RAMBlock to
194 * support 'write-tracking' migration type.
195 * Implies ram_state->ram_wt_enabled.
197 #define RAM_UF_WRITEPROTECT (1 << 6)
200 * RAM is mmap-ed with MAP_NORESERVE. When set, reserving swap space (or huge
201 * pages if applicable) is skipped: will bail out if not supported. When not
202 * set, the OS will do the reservation, if supported for the memory type.
204 #define RAM_NORESERVE (1 << 7)
206 /* RAM that isn't accessible through normal means. */
207 #define RAM_PROTECTED (1 << 8)
209 static inline void iommu_notifier_init(IOMMUNotifier
*n
, IOMMUNotify fn
,
210 IOMMUNotifierFlag flags
,
211 hwaddr start
, hwaddr end
,
215 n
->notifier_flags
= flags
;
218 n
->iommu_idx
= iommu_idx
;
222 * Memory region callbacks
224 struct MemoryRegionOps
{
225 /* Read from the memory region. @addr is relative to @mr; @size is
227 uint64_t (*read
)(void *opaque
,
230 /* Write to the memory region. @addr is relative to @mr; @size is
232 void (*write
)(void *opaque
,
237 MemTxResult (*read_with_attrs
)(void *opaque
,
242 MemTxResult (*write_with_attrs
)(void *opaque
,
248 enum device_endian endianness
;
249 /* Guest-visible constraints: */
251 /* If nonzero, specify bounds on access sizes beyond which a machine
254 unsigned min_access_size
;
255 unsigned max_access_size
;
256 /* If true, unaligned accesses are supported. Otherwise unaligned
257 * accesses throw machine checks.
261 * If present, and returns #false, the transaction is not accepted
262 * by the device (and results in machine dependent behaviour such
263 * as a machine check exception).
265 bool (*accepts
)(void *opaque
, hwaddr addr
,
266 unsigned size
, bool is_write
,
269 /* Internal implementation constraints: */
271 /* If nonzero, specifies the minimum size implemented. Smaller sizes
272 * will be rounded upwards and a partial result will be returned.
274 unsigned min_access_size
;
275 /* If nonzero, specifies the maximum size implemented. Larger sizes
276 * will be done as a series of accesses with smaller sizes.
278 unsigned max_access_size
;
279 /* If true, unaligned accesses are supported. Otherwise all accesses
280 * are converted to (possibly multiple) naturally aligned accesses.
286 typedef struct MemoryRegionClass
{
288 ObjectClass parent_class
;
292 enum IOMMUMemoryRegionAttr
{
293 IOMMU_ATTR_SPAPR_TCE_FD
297 * IOMMUMemoryRegionClass:
299 * All IOMMU implementations need to subclass TYPE_IOMMU_MEMORY_REGION
300 * and provide an implementation of at least the @translate method here
301 * to handle requests to the memory region. Other methods are optional.
303 * The IOMMU implementation must use the IOMMU notifier infrastructure
304 * to report whenever mappings are changed, by calling
305 * memory_region_notify_iommu() (or, if necessary, by calling
306 * memory_region_notify_iommu_one() for each registered notifier).
308 * Conceptually an IOMMU provides a mapping from input address
309 * to an output TLB entry. If the IOMMU is aware of memory transaction
310 * attributes and the output TLB entry depends on the transaction
311 * attributes, we represent this using IOMMU indexes. Each index
312 * selects a particular translation table that the IOMMU has:
314 * @attrs_to_index returns the IOMMU index for a set of transaction attributes
316 * @translate takes an input address and an IOMMU index
318 * and the mapping returned can only depend on the input address and the
321 * Most IOMMUs don't care about the transaction attributes and support
322 * only a single IOMMU index. A more complex IOMMU might have one index
323 * for secure transactions and one for non-secure transactions.
325 struct IOMMUMemoryRegionClass
{
327 MemoryRegionClass parent_class
;
333 * Return a TLB entry that contains a given address.
335 * The IOMMUAccessFlags indicated via @flag are optional and may
336 * be specified as IOMMU_NONE to indicate that the caller needs
337 * the full translation information for both reads and writes. If
338 * the access flags are specified then the IOMMU implementation
339 * may use this as an optimization, to stop doing a page table
340 * walk as soon as it knows that the requested permissions are not
341 * allowed. If IOMMU_NONE is passed then the IOMMU must do the
342 * full page table walk and report the permissions in the returned
343 * IOMMUTLBEntry. (Note that this implies that an IOMMU may not
344 * return different mappings for reads and writes.)
346 * The returned information remains valid while the caller is
347 * holding the big QEMU lock or is inside an RCU critical section;
348 * if the caller wishes to cache the mapping beyond that it must
349 * register an IOMMU notifier so it can invalidate its cached
350 * information when the IOMMU mapping changes.
352 * @iommu: the IOMMUMemoryRegion
354 * @hwaddr: address to be translated within the memory region
356 * @flag: requested access permission
358 * @iommu_idx: IOMMU index for the translation
360 IOMMUTLBEntry (*translate
)(IOMMUMemoryRegion
*iommu
, hwaddr addr
,
361 IOMMUAccessFlags flag
, int iommu_idx
);
363 * @get_min_page_size:
365 * Returns minimum supported page size in bytes.
367 * If this method is not provided then the minimum is assumed to
368 * be TARGET_PAGE_SIZE.
370 * @iommu: the IOMMUMemoryRegion
372 uint64_t (*get_min_page_size
)(IOMMUMemoryRegion
*iommu
);
374 * @notify_flag_changed:
376 * Called when IOMMU Notifier flag changes (ie when the set of
377 * events which IOMMU users are requesting notification for changes).
378 * Optional method -- need not be provided if the IOMMU does not
379 * need to know exactly which events must be notified.
381 * @iommu: the IOMMUMemoryRegion
383 * @old_flags: events which previously needed to be notified
385 * @new_flags: events which now need to be notified
387 * Returns 0 on success, or a negative errno; in particular
388 * returns -EINVAL if the new flag bitmap is not supported by the
389 * IOMMU memory region. In case of failure, the error object
392 int (*notify_flag_changed
)(IOMMUMemoryRegion
*iommu
,
393 IOMMUNotifierFlag old_flags
,
394 IOMMUNotifierFlag new_flags
,
399 * Called to handle memory_region_iommu_replay().
401 * The default implementation of memory_region_iommu_replay() is to
402 * call the IOMMU translate method for every page in the address space
403 * with flag == IOMMU_NONE and then call the notifier if translate
404 * returns a valid mapping. If this method is implemented then it
405 * overrides the default behaviour, and must provide the full semantics
406 * of memory_region_iommu_replay(), by calling @notifier for every
407 * translation present in the IOMMU.
409 * Optional method -- an IOMMU only needs to provide this method
410 * if the default is inefficient or produces undesirable side effects.
412 * Note: this is not related to record-and-replay functionality.
414 void (*replay
)(IOMMUMemoryRegion
*iommu
, IOMMUNotifier
*notifier
);
419 * Get IOMMU misc attributes. This is an optional method that
420 * can be used to allow users of the IOMMU to get implementation-specific
421 * information. The IOMMU implements this method to handle calls
422 * by IOMMU users to memory_region_iommu_get_attr() by filling in
423 * the arbitrary data pointer for any IOMMUMemoryRegionAttr values that
424 * the IOMMU supports. If the method is unimplemented then
425 * memory_region_iommu_get_attr() will always return -EINVAL.
427 * @iommu: the IOMMUMemoryRegion
429 * @attr: attribute being queried
431 * @data: memory to fill in with the attribute data
433 * Returns 0 on success, or a negative errno; in particular
434 * returns -EINVAL for unrecognized or unimplemented attribute types.
436 int (*get_attr
)(IOMMUMemoryRegion
*iommu
, enum IOMMUMemoryRegionAttr attr
,
442 * Return the IOMMU index to use for a given set of transaction attributes.
444 * Optional method: if an IOMMU only supports a single IOMMU index then
445 * the default implementation of memory_region_iommu_attrs_to_index()
448 * The indexes supported by an IOMMU must be contiguous, starting at 0.
450 * @iommu: the IOMMUMemoryRegion
451 * @attrs: memory transaction attributes
453 int (*attrs_to_index
)(IOMMUMemoryRegion
*iommu
, MemTxAttrs attrs
);
458 * Return the number of IOMMU indexes this IOMMU supports.
460 * Optional method: if this method is not provided, then
461 * memory_region_iommu_num_indexes() will return 1, indicating that
462 * only a single IOMMU index is supported.
464 * @iommu: the IOMMUMemoryRegion
466 int (*num_indexes
)(IOMMUMemoryRegion
*iommu
);
469 * @iommu_set_page_size_mask:
471 * Restrict the page size mask that can be supported with a given IOMMU
472 * memory region. Used for example to propagate host physical IOMMU page
473 * size mask limitations to the virtual IOMMU.
475 * Optional method: if this method is not provided, then the default global
478 * @iommu: the IOMMUMemoryRegion
480 * @page_size_mask: a bitmask of supported page sizes. At least one bit,
481 * representing the smallest page size, must be set. Additional set bits
482 * represent supported block sizes. For example a host physical IOMMU that
483 * uses page tables with a page size of 4kB, and supports 2MB and 4GB
484 * blocks, will set mask 0x40201000. A granule of 4kB with indiscriminate
485 * block sizes is specified with mask 0xfffffffffffff000.
487 * Returns 0 on success, or a negative error. In case of failure, the error
488 * object must be created.
490 int (*iommu_set_page_size_mask
)(IOMMUMemoryRegion
*iommu
,
491 uint64_t page_size_mask
,
495 typedef struct RamDiscardListener RamDiscardListener
;
496 typedef int (*NotifyRamPopulate
)(RamDiscardListener
*rdl
,
497 MemoryRegionSection
*section
);
498 typedef void (*NotifyRamDiscard
)(RamDiscardListener
*rdl
,
499 MemoryRegionSection
*section
);
501 struct RamDiscardListener
{
505 * Notification that previously discarded memory is about to get populated.
506 * Listeners are able to object. If any listener objects, already
507 * successfully notified listeners are notified about a discard again.
509 * @rdl: the #RamDiscardListener getting notified
510 * @section: the #MemoryRegionSection to get populated. The section
511 * is aligned within the memory region to the minimum granularity
512 * unless it would exceed the registered section.
514 * Returns 0 on success. If the notification is rejected by the listener,
515 * an error is returned.
517 NotifyRamPopulate notify_populate
;
522 * Notification that previously populated memory was discarded successfully
523 * and listeners should drop all references to such memory and prevent
524 * new population (e.g., unmap).
526 * @rdl: the #RamDiscardListener getting notified
527 * @section: the #MemoryRegionSection to get populated. The section
528 * is aligned within the memory region to the minimum granularity
529 * unless it would exceed the registered section.
531 NotifyRamDiscard notify_discard
;
534 * @double_discard_supported:
536 * The listener suppors getting @notify_discard notifications that span
537 * already discarded parts.
539 bool double_discard_supported
;
541 MemoryRegionSection
*section
;
542 QLIST_ENTRY(RamDiscardListener
) next
;
545 static inline void ram_discard_listener_init(RamDiscardListener
*rdl
,
546 NotifyRamPopulate populate_fn
,
547 NotifyRamDiscard discard_fn
,
548 bool double_discard_supported
)
550 rdl
->notify_populate
= populate_fn
;
551 rdl
->notify_discard
= discard_fn
;
552 rdl
->double_discard_supported
= double_discard_supported
;
555 typedef int (*ReplayRamPopulate
)(MemoryRegionSection
*section
, void *opaque
);
556 typedef void (*ReplayRamDiscard
)(MemoryRegionSection
*section
, void *opaque
);
559 * RamDiscardManagerClass:
561 * A #RamDiscardManager coordinates which parts of specific RAM #MemoryRegion
562 * regions are currently populated to be used/accessed by the VM, notifying
563 * after parts were discarded (freeing up memory) and before parts will be
564 * populated (consuming memory), to be used/acessed by the VM.
566 * A #RamDiscardManager can only be set for a RAM #MemoryRegion while the
567 * #MemoryRegion isn't mapped yet; it cannot change while the #MemoryRegion is
570 * The #RamDiscardManager is intended to be used by technologies that are
571 * incompatible with discarding of RAM (e.g., VFIO, which may pin all
572 * memory inside a #MemoryRegion), and require proper coordination to only
573 * map the currently populated parts, to hinder parts that are expected to
574 * remain discarded from silently getting populated and consuming memory.
575 * Technologies that support discarding of RAM don't have to bother and can
576 * simply map the whole #MemoryRegion.
578 * An example #RamDiscardManager is virtio-mem, which logically (un)plugs
579 * memory within an assigned RAM #MemoryRegion, coordinated with the VM.
580 * Logically unplugging memory consists of discarding RAM. The VM agreed to not
581 * access unplugged (discarded) memory - especially via DMA. virtio-mem will
582 * properly coordinate with listeners before memory is plugged (populated),
583 * and after memory is unplugged (discarded).
585 * Listeners are called in multiples of the minimum granularity (unless it
586 * would exceed the registered range) and changes are aligned to the minimum
587 * granularity within the #MemoryRegion. Listeners have to prepare for memory
588 * becomming discarded in a different granularity than it was populated and the
591 struct RamDiscardManagerClass
{
593 InterfaceClass parent_class
;
598 * @get_min_granularity:
600 * Get the minimum granularity in which listeners will get notified
601 * about changes within the #MemoryRegion via the #RamDiscardManager.
603 * @rdm: the #RamDiscardManager
604 * @mr: the #MemoryRegion
606 * Returns the minimum granularity.
608 uint64_t (*get_min_granularity
)(const RamDiscardManager
*rdm
,
609 const MemoryRegion
*mr
);
614 * Check whether the given #MemoryRegionSection is completely populated
615 * (i.e., no parts are currently discarded) via the #RamDiscardManager.
616 * There are no alignment requirements.
618 * @rdm: the #RamDiscardManager
619 * @section: the #MemoryRegionSection
621 * Returns whether the given range is completely populated.
623 bool (*is_populated
)(const RamDiscardManager
*rdm
,
624 const MemoryRegionSection
*section
);
629 * Call the #ReplayRamPopulate callback for all populated parts within the
630 * #MemoryRegionSection via the #RamDiscardManager.
632 * In case any call fails, no further calls are made.
634 * @rdm: the #RamDiscardManager
635 * @section: the #MemoryRegionSection
636 * @replay_fn: the #ReplayRamPopulate callback
637 * @opaque: pointer to forward to the callback
639 * Returns 0 on success, or a negative error if any notification failed.
641 int (*replay_populated
)(const RamDiscardManager
*rdm
,
642 MemoryRegionSection
*section
,
643 ReplayRamPopulate replay_fn
, void *opaque
);
648 * Call the #ReplayRamDiscard callback for all discarded parts within the
649 * #MemoryRegionSection via the #RamDiscardManager.
651 * @rdm: the #RamDiscardManager
652 * @section: the #MemoryRegionSection
653 * @replay_fn: the #ReplayRamDiscard callback
654 * @opaque: pointer to forward to the callback
656 void (*replay_discarded
)(const RamDiscardManager
*rdm
,
657 MemoryRegionSection
*section
,
658 ReplayRamDiscard replay_fn
, void *opaque
);
661 * @register_listener:
663 * Register a #RamDiscardListener for the given #MemoryRegionSection and
664 * immediately notify the #RamDiscardListener about all populated parts
665 * within the #MemoryRegionSection via the #RamDiscardManager.
667 * In case any notification fails, no further notifications are triggered
668 * and an error is logged.
670 * @rdm: the #RamDiscardManager
671 * @rdl: the #RamDiscardListener
672 * @section: the #MemoryRegionSection
674 void (*register_listener
)(RamDiscardManager
*rdm
,
675 RamDiscardListener
*rdl
,
676 MemoryRegionSection
*section
);
679 * @unregister_listener:
681 * Unregister a previously registered #RamDiscardListener via the
682 * #RamDiscardManager after notifying the #RamDiscardListener about all
683 * populated parts becoming unpopulated within the registered
684 * #MemoryRegionSection.
686 * @rdm: the #RamDiscardManager
687 * @rdl: the #RamDiscardListener
689 void (*unregister_listener
)(RamDiscardManager
*rdm
,
690 RamDiscardListener
*rdl
);
693 uint64_t ram_discard_manager_get_min_granularity(const RamDiscardManager
*rdm
,
694 const MemoryRegion
*mr
);
696 bool ram_discard_manager_is_populated(const RamDiscardManager
*rdm
,
697 const MemoryRegionSection
*section
);
699 int ram_discard_manager_replay_populated(const RamDiscardManager
*rdm
,
700 MemoryRegionSection
*section
,
701 ReplayRamPopulate replay_fn
,
704 void ram_discard_manager_replay_discarded(const RamDiscardManager
*rdm
,
705 MemoryRegionSection
*section
,
706 ReplayRamDiscard replay_fn
,
709 void ram_discard_manager_register_listener(RamDiscardManager
*rdm
,
710 RamDiscardListener
*rdl
,
711 MemoryRegionSection
*section
);
713 void ram_discard_manager_unregister_listener(RamDiscardManager
*rdm
,
714 RamDiscardListener
*rdl
);
716 typedef struct CoalescedMemoryRange CoalescedMemoryRange
;
717 typedef struct MemoryRegionIoeventfd MemoryRegionIoeventfd
;
721 * A struct representing a memory region.
723 struct MemoryRegion
{
728 /* The following fields should fit in a cache line */
732 bool readonly
; /* For RAM regions */
735 bool flush_coalesced_mmio
;
736 uint8_t dirty_log_mask
;
741 const MemoryRegionOps
*ops
;
743 MemoryRegion
*container
;
744 int mapped_via_alias
; /* Mapped via an alias, container might be NULL */
747 void (*destructor
)(MemoryRegion
*mr
);
752 bool warning_printed
; /* For reservations */
753 uint8_t vga_logging_count
;
757 QTAILQ_HEAD(, MemoryRegion
) subregions
;
758 QTAILQ_ENTRY(MemoryRegion
) subregions_link
;
759 QTAILQ_HEAD(, CoalescedMemoryRange
) coalesced
;
761 unsigned ioeventfd_nb
;
762 MemoryRegionIoeventfd
*ioeventfds
;
763 RamDiscardManager
*rdm
; /* Only for RAM */
766 struct IOMMUMemoryRegion
{
767 MemoryRegion parent_obj
;
769 QLIST_HEAD(, IOMMUNotifier
) iommu_notify
;
770 IOMMUNotifierFlag iommu_notify_flags
;
773 #define IOMMU_NOTIFIER_FOREACH(n, mr) \
774 QLIST_FOREACH((n), &(mr)->iommu_notify, node)
777 * struct MemoryListener: callbacks structure for updates to the physical memory map
779 * Allows a component to adjust to changes in the guest-visible memory map.
780 * Use with memory_listener_register() and memory_listener_unregister().
782 struct MemoryListener
{
786 * Called at the beginning of an address space update transaction.
787 * Followed by calls to #MemoryListener.region_add(),
788 * #MemoryListener.region_del(), #MemoryListener.region_nop(),
789 * #MemoryListener.log_start() and #MemoryListener.log_stop() in
790 * increasing address order.
792 * @listener: The #MemoryListener.
794 void (*begin
)(MemoryListener
*listener
);
799 * Called at the end of an address space update transaction,
800 * after the last call to #MemoryListener.region_add(),
801 * #MemoryListener.region_del() or #MemoryListener.region_nop(),
802 * #MemoryListener.log_start() and #MemoryListener.log_stop().
804 * @listener: The #MemoryListener.
806 void (*commit
)(MemoryListener
*listener
);
811 * Called during an address space update transaction,
812 * for a section of the address space that is new in this address space
813 * space since the last transaction.
815 * @listener: The #MemoryListener.
816 * @section: The new #MemoryRegionSection.
818 void (*region_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
823 * Called during an address space update transaction,
824 * for a section of the address space that has disappeared in the address
825 * space since the last transaction.
827 * @listener: The #MemoryListener.
828 * @section: The old #MemoryRegionSection.
830 void (*region_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
835 * Called during an address space update transaction,
836 * for a section of the address space that is in the same place in the address
837 * space as in the last transaction.
839 * @listener: The #MemoryListener.
840 * @section: The #MemoryRegionSection.
842 void (*region_nop
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
847 * Called during an address space update transaction, after
848 * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
849 * #MemoryListener.region_nop(), if dirty memory logging clients have
850 * become active since the last transaction.
852 * @listener: The #MemoryListener.
853 * @section: The #MemoryRegionSection.
854 * @old: A bitmap of dirty memory logging clients that were active in
855 * the previous transaction.
856 * @new: A bitmap of dirty memory logging clients that are active in
857 * the current transaction.
859 void (*log_start
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
865 * Called during an address space update transaction, after
866 * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
867 * #MemoryListener.region_nop() and possibly after
868 * #MemoryListener.log_start(), if dirty memory logging clients have
869 * become inactive since the last transaction.
871 * @listener: The #MemoryListener.
872 * @section: The #MemoryRegionSection.
873 * @old: A bitmap of dirty memory logging clients that were active in
874 * the previous transaction.
875 * @new: A bitmap of dirty memory logging clients that are active in
876 * the current transaction.
878 void (*log_stop
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
884 * Called by memory_region_snapshot_and_clear_dirty() and
885 * memory_global_dirty_log_sync(), before accessing QEMU's "official"
886 * copy of the dirty memory bitmap for a #MemoryRegionSection.
888 * @listener: The #MemoryListener.
889 * @section: The #MemoryRegionSection.
891 void (*log_sync
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
896 * This is the global version of @log_sync when the listener does
897 * not have a way to synchronize the log with finer granularity.
898 * When the listener registers with @log_sync_global defined, then
899 * its @log_sync must be NULL. Vice versa.
901 * @listener: The #MemoryListener.
903 void (*log_sync_global
)(MemoryListener
*listener
);
908 * Called before reading the dirty memory bitmap for a
909 * #MemoryRegionSection.
911 * @listener: The #MemoryListener.
912 * @section: The #MemoryRegionSection.
914 void (*log_clear
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
919 * Called by memory_global_dirty_log_start(), which
920 * enables the %DIRTY_LOG_MIGRATION client on all memory regions in
921 * the address space. #MemoryListener.log_global_start() is also
922 * called when a #MemoryListener is added, if global dirty logging is
923 * active at that time.
925 * @listener: The #MemoryListener.
927 void (*log_global_start
)(MemoryListener
*listener
);
932 * Called by memory_global_dirty_log_stop(), which
933 * disables the %DIRTY_LOG_MIGRATION client on all memory regions in
936 * @listener: The #MemoryListener.
938 void (*log_global_stop
)(MemoryListener
*listener
);
941 * @log_global_after_sync:
943 * Called after reading the dirty memory bitmap
944 * for any #MemoryRegionSection.
946 * @listener: The #MemoryListener.
948 void (*log_global_after_sync
)(MemoryListener
*listener
);
953 * Called during an address space update transaction,
954 * for a section of the address space that has had a new ioeventfd
955 * registration since the last transaction.
957 * @listener: The #MemoryListener.
958 * @section: The new #MemoryRegionSection.
959 * @match_data: The @match_data parameter for the new ioeventfd.
960 * @data: The @data parameter for the new ioeventfd.
961 * @e: The #EventNotifier parameter for the new ioeventfd.
963 void (*eventfd_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
964 bool match_data
, uint64_t data
, EventNotifier
*e
);
969 * Called during an address space update transaction,
970 * for a section of the address space that has dropped an ioeventfd
971 * registration since the last transaction.
973 * @listener: The #MemoryListener.
974 * @section: The new #MemoryRegionSection.
975 * @match_data: The @match_data parameter for the dropped ioeventfd.
976 * @data: The @data parameter for the dropped ioeventfd.
977 * @e: The #EventNotifier parameter for the dropped ioeventfd.
979 void (*eventfd_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
980 bool match_data
, uint64_t data
, EventNotifier
*e
);
985 * Called during an address space update transaction,
986 * for a section of the address space that has had a new coalesced
987 * MMIO range registration since the last transaction.
989 * @listener: The #MemoryListener.
990 * @section: The new #MemoryRegionSection.
991 * @addr: The starting address for the coalesced MMIO range.
992 * @len: The length of the coalesced MMIO range.
994 void (*coalesced_io_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
995 hwaddr addr
, hwaddr len
);
1000 * Called during an address space update transaction,
1001 * for a section of the address space that has dropped a coalesced
1002 * MMIO range since the last transaction.
1004 * @listener: The #MemoryListener.
1005 * @section: The new #MemoryRegionSection.
1006 * @addr: The starting address for the coalesced MMIO range.
1007 * @len: The length of the coalesced MMIO range.
1009 void (*coalesced_io_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1010 hwaddr addr
, hwaddr len
);
1014 * Govern the order in which memory listeners are invoked. Lower priorities
1015 * are invoked earlier for "add" or "start" callbacks, and later for "delete"
1016 * or "stop" callbacks.
1023 * Name of the listener. It can be used in contexts where we'd like to
1024 * identify one memory listener with the rest.
1029 AddressSpace
*address_space
;
1030 QTAILQ_ENTRY(MemoryListener
) link
;
1031 QTAILQ_ENTRY(MemoryListener
) link_as
;
1035 * struct AddressSpace: describes a mapping of addresses to #MemoryRegion objects
1037 struct AddressSpace
{
1039 struct rcu_head rcu
;
1043 /* Accessed via RCU. */
1044 struct FlatView
*current_map
;
1047 struct MemoryRegionIoeventfd
*ioeventfds
;
1048 QTAILQ_HEAD(, MemoryListener
) listeners
;
1049 QTAILQ_ENTRY(AddressSpace
) address_spaces_link
;
1052 typedef struct AddressSpaceDispatch AddressSpaceDispatch
;
1053 typedef struct FlatRange FlatRange
;
1055 /* Flattened global view of current active memory hierarchy. Kept in sorted
1059 struct rcu_head rcu
;
1063 unsigned nr_allocated
;
1064 struct AddressSpaceDispatch
*dispatch
;
1068 static inline FlatView
*address_space_to_flatview(AddressSpace
*as
)
1070 return qatomic_rcu_read(&as
->current_map
);
1074 * typedef flatview_cb: callback for flatview_for_each_range()
1076 * @start: start address of the range within the FlatView
1077 * @len: length of the range in bytes
1078 * @mr: MemoryRegion covering this range
1079 * @offset_in_region: offset of the first byte of the range within @mr
1080 * @opaque: data pointer passed to flatview_for_each_range()
1082 * Returns: true to stop the iteration, false to keep going.
1084 typedef bool (*flatview_cb
)(Int128 start
,
1086 const MemoryRegion
*mr
,
1087 hwaddr offset_in_region
,
1091 * flatview_for_each_range: Iterate through a FlatView
1092 * @fv: the FlatView to iterate through
1093 * @cb: function to call for each range
1094 * @opaque: opaque data pointer to pass to @cb
1096 * A FlatView is made up of a list of non-overlapping ranges, each of
1097 * which is a slice of a MemoryRegion. This function iterates through
1098 * each range in @fv, calling @cb. The callback function can terminate
1099 * iteration early by returning 'true'.
1101 void flatview_for_each_range(FlatView
*fv
, flatview_cb cb
, void *opaque
);
1103 static inline bool MemoryRegionSection_eq(MemoryRegionSection
*a
,
1104 MemoryRegionSection
*b
)
1106 return a
->mr
== b
->mr
&&
1108 a
->offset_within_region
== b
->offset_within_region
&&
1109 a
->offset_within_address_space
== b
->offset_within_address_space
&&
1110 int128_eq(a
->size
, b
->size
) &&
1111 a
->readonly
== b
->readonly
&&
1112 a
->nonvolatile
== b
->nonvolatile
;
1116 * memory_region_section_new_copy: Copy a memory region section
1118 * Allocate memory for a new copy, copy the memory region section, and
1119 * properly take a reference on all relevant members.
1121 * @s: the #MemoryRegionSection to copy
1123 MemoryRegionSection
*memory_region_section_new_copy(MemoryRegionSection
*s
);
1126 * memory_region_section_new_copy: Free a copied memory region section
1128 * Free a copy of a memory section created via memory_region_section_new_copy().
1129 * properly dropping references on all relevant members.
1131 * @s: the #MemoryRegionSection to copy
1133 void memory_region_section_free_copy(MemoryRegionSection
*s
);
1136 * memory_region_init: Initialize a memory region
1138 * The region typically acts as a container for other memory regions. Use
1139 * memory_region_add_subregion() to add subregions.
1141 * @mr: the #MemoryRegion to be initialized
1142 * @owner: the object that tracks the region's reference count
1143 * @name: used for debugging; not visible to the user or ABI
1144 * @size: size of the region; any subregions beyond this size will be clipped
1146 void memory_region_init(MemoryRegion
*mr
,
1152 * memory_region_ref: Add 1 to a memory region's reference count
1154 * Whenever memory regions are accessed outside the BQL, they need to be
1155 * preserved against hot-unplug. MemoryRegions actually do not have their
1156 * own reference count; they piggyback on a QOM object, their "owner".
1157 * This function adds a reference to the owner.
1159 * All MemoryRegions must have an owner if they can disappear, even if the
1160 * device they belong to operates exclusively under the BQL. This is because
1161 * the region could be returned at any time by memory_region_find, and this
1162 * is usually under guest control.
1164 * @mr: the #MemoryRegion
1166 void memory_region_ref(MemoryRegion
*mr
);
1169 * memory_region_unref: Remove 1 to a memory region's reference count
1171 * Whenever memory regions are accessed outside the BQL, they need to be
1172 * preserved against hot-unplug. MemoryRegions actually do not have their
1173 * own reference count; they piggyback on a QOM object, their "owner".
1174 * This function removes a reference to the owner and possibly destroys it.
1176 * @mr: the #MemoryRegion
1178 void memory_region_unref(MemoryRegion
*mr
);
1181 * memory_region_init_io: Initialize an I/O memory region.
1183 * Accesses into the region will cause the callbacks in @ops to be called.
1184 * if @size is nonzero, subregions will be clipped to @size.
1186 * @mr: the #MemoryRegion to be initialized.
1187 * @owner: the object that tracks the region's reference count
1188 * @ops: a structure containing read and write callbacks to be used when
1189 * I/O is performed on the region.
1190 * @opaque: passed to the read and write callbacks of the @ops structure.
1191 * @name: used for debugging; not visible to the user or ABI
1192 * @size: size of the region.
1194 void memory_region_init_io(MemoryRegion
*mr
,
1196 const MemoryRegionOps
*ops
,
1202 * memory_region_init_ram_nomigrate: Initialize RAM memory region. Accesses
1203 * into the region will modify memory
1206 * @mr: the #MemoryRegion to be initialized.
1207 * @owner: the object that tracks the region's reference count
1208 * @name: Region name, becomes part of RAMBlock name used in migration stream
1209 * must be unique within any device
1210 * @size: size of the region.
1211 * @errp: pointer to Error*, to store an error if it happens.
1213 * Note that this function does not do anything to cause the data in the
1214 * RAM memory region to be migrated; that is the responsibility of the caller.
1216 void memory_region_init_ram_nomigrate(MemoryRegion
*mr
,
1223 * memory_region_init_ram_flags_nomigrate: Initialize RAM memory region.
1224 * Accesses into the region will
1225 * modify memory directly.
1227 * @mr: the #MemoryRegion to be initialized.
1228 * @owner: the object that tracks the region's reference count
1229 * @name: Region name, becomes part of RAMBlock name used in migration stream
1230 * must be unique within any device
1231 * @size: size of the region.
1232 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_NORESERVE.
1233 * @errp: pointer to Error*, to store an error if it happens.
1235 * Note that this function does not do anything to cause the data in the
1236 * RAM memory region to be migrated; that is the responsibility of the caller.
1238 void memory_region_init_ram_flags_nomigrate(MemoryRegion
*mr
,
1246 * memory_region_init_resizeable_ram: Initialize memory region with resizeable
1247 * RAM. Accesses into the region will
1248 * modify memory directly. Only an initial
1249 * portion of this RAM is actually used.
1250 * Changing the size while migrating
1251 * can result in the migration being
1254 * @mr: the #MemoryRegion to be initialized.
1255 * @owner: the object that tracks the region's reference count
1256 * @name: Region name, becomes part of RAMBlock name used in migration stream
1257 * must be unique within any device
1258 * @size: used size of the region.
1259 * @max_size: max size of the region.
1260 * @resized: callback to notify owner about used size change.
1261 * @errp: pointer to Error*, to store an error if it happens.
1263 * Note that this function does not do anything to cause the data in the
1264 * RAM memory region to be migrated; that is the responsibility of the caller.
1266 void memory_region_init_resizeable_ram(MemoryRegion
*mr
,
1271 void (*resized
)(const char*,
1278 * memory_region_init_ram_from_file: Initialize RAM memory region with a
1281 * @mr: the #MemoryRegion to be initialized.
1282 * @owner: the object that tracks the region's reference count
1283 * @name: Region name, becomes part of RAMBlock name used in migration stream
1284 * must be unique within any device
1285 * @size: size of the region.
1286 * @align: alignment of the region base address; if 0, the default alignment
1287 * (getpagesize()) will be used.
1288 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
1290 * @path: the path in which to allocate the RAM.
1291 * @readonly: true to open @path for reading, false for read/write.
1292 * @errp: pointer to Error*, to store an error if it happens.
1294 * Note that this function does not do anything to cause the data in the
1295 * RAM memory region to be migrated; that is the responsibility of the caller.
1297 void memory_region_init_ram_from_file(MemoryRegion
*mr
,
1308 * memory_region_init_ram_from_fd: Initialize RAM memory region with a
1311 * @mr: the #MemoryRegion to be initialized.
1312 * @owner: the object that tracks the region's reference count
1313 * @name: the name of the region.
1314 * @size: size of the region.
1315 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
1316 * RAM_NORESERVE, RAM_PROTECTED.
1317 * @fd: the fd to mmap.
1318 * @offset: offset within the file referenced by fd
1319 * @errp: pointer to Error*, to store an error if it happens.
1321 * Note that this function does not do anything to cause the data in the
1322 * RAM memory region to be migrated; that is the responsibility of the caller.
1324 void memory_region_init_ram_from_fd(MemoryRegion
*mr
,
1335 * memory_region_init_ram_ptr: Initialize RAM memory region from a
1336 * user-provided pointer. Accesses into the
1337 * region will modify memory directly.
1339 * @mr: the #MemoryRegion to be initialized.
1340 * @owner: the object that tracks the region's reference count
1341 * @name: Region name, becomes part of RAMBlock name used in migration stream
1342 * must be unique within any device
1343 * @size: size of the region.
1344 * @ptr: memory to be mapped; must contain at least @size bytes.
1346 * Note that this function does not do anything to cause the data in the
1347 * RAM memory region to be migrated; that is the responsibility of the caller.
1349 void memory_region_init_ram_ptr(MemoryRegion
*mr
,
1356 * memory_region_init_ram_device_ptr: Initialize RAM device memory region from
1357 * a user-provided pointer.
1359 * A RAM device represents a mapping to a physical device, such as to a PCI
1360 * MMIO BAR of an vfio-pci assigned device. The memory region may be mapped
1361 * into the VM address space and access to the region will modify memory
1362 * directly. However, the memory region should not be included in a memory
1363 * dump (device may not be enabled/mapped at the time of the dump), and
1364 * operations incompatible with manipulating MMIO should be avoided. Replaces
1367 * @mr: the #MemoryRegion to be initialized.
1368 * @owner: the object that tracks the region's reference count
1369 * @name: the name of the region.
1370 * @size: size of the region.
1371 * @ptr: memory to be mapped; must contain at least @size bytes.
1373 * Note that this function does not do anything to cause the data in the
1374 * RAM memory region to be migrated; that is the responsibility of the caller.
1375 * (For RAM device memory regions, migrating the contents rarely makes sense.)
1377 void memory_region_init_ram_device_ptr(MemoryRegion
*mr
,
1384 * memory_region_init_alias: Initialize a memory region that aliases all or a
1385 * part of another memory region.
1387 * @mr: the #MemoryRegion to be initialized.
1388 * @owner: the object that tracks the region's reference count
1389 * @name: used for debugging; not visible to the user or ABI
1390 * @orig: the region to be referenced; @mr will be equivalent to
1391 * @orig between @offset and @offset + @size - 1.
1392 * @offset: start of the section in @orig to be referenced.
1393 * @size: size of the region.
1395 void memory_region_init_alias(MemoryRegion
*mr
,
1403 * memory_region_init_rom_nomigrate: Initialize a ROM memory region.
1405 * This has the same effect as calling memory_region_init_ram_nomigrate()
1406 * and then marking the resulting region read-only with
1407 * memory_region_set_readonly().
1409 * Note that this function does not do anything to cause the data in the
1410 * RAM side of the memory region to be migrated; that is the responsibility
1413 * @mr: the #MemoryRegion to be initialized.
1414 * @owner: the object that tracks the region's reference count
1415 * @name: Region name, becomes part of RAMBlock name used in migration stream
1416 * must be unique within any device
1417 * @size: size of the region.
1418 * @errp: pointer to Error*, to store an error if it happens.
1420 void memory_region_init_rom_nomigrate(MemoryRegion
*mr
,
1427 * memory_region_init_rom_device_nomigrate: Initialize a ROM memory region.
1428 * Writes are handled via callbacks.
1430 * Note that this function does not do anything to cause the data in the
1431 * RAM side of the memory region to be migrated; that is the responsibility
1434 * @mr: the #MemoryRegion to be initialized.
1435 * @owner: the object that tracks the region's reference count
1436 * @ops: callbacks for write access handling (must not be NULL).
1437 * @opaque: passed to the read and write callbacks of the @ops structure.
1438 * @name: Region name, becomes part of RAMBlock name used in migration stream
1439 * must be unique within any device
1440 * @size: size of the region.
1441 * @errp: pointer to Error*, to store an error if it happens.
1443 void memory_region_init_rom_device_nomigrate(MemoryRegion
*mr
,
1445 const MemoryRegionOps
*ops
,
1452 * memory_region_init_iommu: Initialize a memory region of a custom type
1453 * that translates addresses
1455 * An IOMMU region translates addresses and forwards accesses to a target
1458 * The IOMMU implementation must define a subclass of TYPE_IOMMU_MEMORY_REGION.
1459 * @_iommu_mr should be a pointer to enough memory for an instance of
1460 * that subclass, @instance_size is the size of that subclass, and
1461 * @mrtypename is its name. This function will initialize @_iommu_mr as an
1462 * instance of the subclass, and its methods will then be called to handle
1463 * accesses to the memory region. See the documentation of
1464 * #IOMMUMemoryRegionClass for further details.
1466 * @_iommu_mr: the #IOMMUMemoryRegion to be initialized
1467 * @instance_size: the IOMMUMemoryRegion subclass instance size
1468 * @mrtypename: the type name of the #IOMMUMemoryRegion
1469 * @owner: the object that tracks the region's reference count
1470 * @name: used for debugging; not visible to the user or ABI
1471 * @size: size of the region.
1473 void memory_region_init_iommu(void *_iommu_mr
,
1474 size_t instance_size
,
1475 const char *mrtypename
,
1481 * memory_region_init_ram - Initialize RAM memory region. Accesses into the
1482 * region will modify memory directly.
1484 * @mr: the #MemoryRegion to be initialized
1485 * @owner: the object that tracks the region's reference count (must be
1486 * TYPE_DEVICE or a subclass of TYPE_DEVICE, or NULL)
1487 * @name: name of the memory region
1488 * @size: size of the region in bytes
1489 * @errp: pointer to Error*, to store an error if it happens.
1491 * This function allocates RAM for a board model or device, and
1492 * arranges for it to be migrated (by calling vmstate_register_ram()
1493 * if @owner is a DeviceState, or vmstate_register_ram_global() if
1496 * TODO: Currently we restrict @owner to being either NULL (for
1497 * global RAM regions with no owner) or devices, so that we can
1498 * give the RAM block a unique name for migration purposes.
1499 * We should lift this restriction and allow arbitrary Objects.
1500 * If you pass a non-NULL non-device @owner then we will assert.
1502 void memory_region_init_ram(MemoryRegion
*mr
,
1509 * memory_region_init_rom: Initialize a ROM memory region.
1511 * This has the same effect as calling memory_region_init_ram()
1512 * and then marking the resulting region read-only with
1513 * memory_region_set_readonly(). This includes arranging for the
1514 * contents to be migrated.
1516 * TODO: Currently we restrict @owner to being either NULL (for
1517 * global RAM regions with no owner) or devices, so that we can
1518 * give the RAM block a unique name for migration purposes.
1519 * We should lift this restriction and allow arbitrary Objects.
1520 * If you pass a non-NULL non-device @owner then we will assert.
1522 * @mr: the #MemoryRegion to be initialized.
1523 * @owner: the object that tracks the region's reference count
1524 * @name: Region name, becomes part of RAMBlock name used in migration stream
1525 * must be unique within any device
1526 * @size: size of the region.
1527 * @errp: pointer to Error*, to store an error if it happens.
1529 void memory_region_init_rom(MemoryRegion
*mr
,
1536 * memory_region_init_rom_device: Initialize a ROM memory region.
1537 * Writes are handled via callbacks.
1539 * This function initializes a memory region backed by RAM for reads
1540 * and callbacks for writes, and arranges for the RAM backing to
1541 * be migrated (by calling vmstate_register_ram()
1542 * if @owner is a DeviceState, or vmstate_register_ram_global() if
1545 * TODO: Currently we restrict @owner to being either NULL (for
1546 * global RAM regions with no owner) or devices, so that we can
1547 * give the RAM block a unique name for migration purposes.
1548 * We should lift this restriction and allow arbitrary Objects.
1549 * If you pass a non-NULL non-device @owner then we will assert.
1551 * @mr: the #MemoryRegion to be initialized.
1552 * @owner: the object that tracks the region's reference count
1553 * @ops: callbacks for write access handling (must not be NULL).
1554 * @opaque: passed to the read and write callbacks of the @ops structure.
1555 * @name: Region name, becomes part of RAMBlock name used in migration stream
1556 * must be unique within any device
1557 * @size: size of the region.
1558 * @errp: pointer to Error*, to store an error if it happens.
1560 void memory_region_init_rom_device(MemoryRegion
*mr
,
1562 const MemoryRegionOps
*ops
,
1570 * memory_region_owner: get a memory region's owner.
1572 * @mr: the memory region being queried.
1574 Object
*memory_region_owner(MemoryRegion
*mr
);
1577 * memory_region_size: get a memory region's size.
1579 * @mr: the memory region being queried.
1581 uint64_t memory_region_size(MemoryRegion
*mr
);
1584 * memory_region_is_ram: check whether a memory region is random access
1586 * Returns %true if a memory region is random access.
1588 * @mr: the memory region being queried
1590 static inline bool memory_region_is_ram(MemoryRegion
*mr
)
1596 * memory_region_is_ram_device: check whether a memory region is a ram device
1598 * Returns %true if a memory region is a device backed ram region
1600 * @mr: the memory region being queried
1602 bool memory_region_is_ram_device(MemoryRegion
*mr
);
1605 * memory_region_is_romd: check whether a memory region is in ROMD mode
1607 * Returns %true if a memory region is a ROM device and currently set to allow
1610 * @mr: the memory region being queried
1612 static inline bool memory_region_is_romd(MemoryRegion
*mr
)
1614 return mr
->rom_device
&& mr
->romd_mode
;
1618 * memory_region_is_protected: check whether a memory region is protected
1620 * Returns %true if a memory region is protected RAM and cannot be accessed
1621 * via standard mechanisms, e.g. DMA.
1623 * @mr: the memory region being queried
1625 bool memory_region_is_protected(MemoryRegion
*mr
);
1628 * memory_region_get_iommu: check whether a memory region is an iommu
1630 * Returns pointer to IOMMUMemoryRegion if a memory region is an iommu,
1633 * @mr: the memory region being queried
1635 static inline IOMMUMemoryRegion
*memory_region_get_iommu(MemoryRegion
*mr
)
1638 return memory_region_get_iommu(mr
->alias
);
1641 return (IOMMUMemoryRegion
*) mr
;
1647 * memory_region_get_iommu_class_nocheck: returns iommu memory region class
1648 * if an iommu or NULL if not
1650 * Returns pointer to IOMMUMemoryRegionClass if a memory region is an iommu,
1651 * otherwise NULL. This is fast path avoiding QOM checking, use with caution.
1653 * @iommu_mr: the memory region being queried
1655 static inline IOMMUMemoryRegionClass
*memory_region_get_iommu_class_nocheck(
1656 IOMMUMemoryRegion
*iommu_mr
)
1658 return (IOMMUMemoryRegionClass
*) (((Object
*)iommu_mr
)->class);
1661 #define memory_region_is_iommu(mr) (memory_region_get_iommu(mr) != NULL)
1664 * memory_region_iommu_get_min_page_size: get minimum supported page size
1667 * Returns minimum supported page size for an iommu.
1669 * @iommu_mr: the memory region being queried
1671 uint64_t memory_region_iommu_get_min_page_size(IOMMUMemoryRegion
*iommu_mr
);
1674 * memory_region_notify_iommu: notify a change in an IOMMU translation entry.
1676 * Note: for any IOMMU implementation, an in-place mapping change
1677 * should be notified with an UNMAP followed by a MAP.
1679 * @iommu_mr: the memory region that was changed
1680 * @iommu_idx: the IOMMU index for the translation table which has changed
1681 * @event: TLB event with the new entry in the IOMMU translation table.
1682 * The entry replaces all old entries for the same virtual I/O address
1685 void memory_region_notify_iommu(IOMMUMemoryRegion
*iommu_mr
,
1687 IOMMUTLBEvent event
);
1690 * memory_region_notify_iommu_one: notify a change in an IOMMU translation
1691 * entry to a single notifier
1693 * This works just like memory_region_notify_iommu(), but it only
1694 * notifies a specific notifier, not all of them.
1696 * @notifier: the notifier to be notified
1697 * @event: TLB event with the new entry in the IOMMU translation table.
1698 * The entry replaces all old entries for the same virtual I/O address
1701 void memory_region_notify_iommu_one(IOMMUNotifier
*notifier
,
1702 IOMMUTLBEvent
*event
);
1705 * memory_region_register_iommu_notifier: register a notifier for changes to
1706 * IOMMU translation entries.
1708 * Returns 0 on success, or a negative errno otherwise. In particular,
1709 * -EINVAL indicates that at least one of the attributes of the notifier
1710 * is not supported (flag/range) by the IOMMU memory region. In case of error
1711 * the error object must be created.
1713 * @mr: the memory region to observe
1714 * @n: the IOMMUNotifier to be added; the notify callback receives a
1715 * pointer to an #IOMMUTLBEntry as the opaque value; the pointer
1716 * ceases to be valid on exit from the notifier.
1717 * @errp: pointer to Error*, to store an error if it happens.
1719 int memory_region_register_iommu_notifier(MemoryRegion
*mr
,
1720 IOMMUNotifier
*n
, Error
**errp
);
1723 * memory_region_iommu_replay: replay existing IOMMU translations to
1724 * a notifier with the minimum page granularity returned by
1725 * mr->iommu_ops->get_page_size().
1727 * Note: this is not related to record-and-replay functionality.
1729 * @iommu_mr: the memory region to observe
1730 * @n: the notifier to which to replay iommu mappings
1732 void memory_region_iommu_replay(IOMMUMemoryRegion
*iommu_mr
, IOMMUNotifier
*n
);
1735 * memory_region_unregister_iommu_notifier: unregister a notifier for
1736 * changes to IOMMU translation entries.
1738 * @mr: the memory region which was observed and for which notity_stopped()
1739 * needs to be called
1740 * @n: the notifier to be removed.
1742 void memory_region_unregister_iommu_notifier(MemoryRegion
*mr
,
1746 * memory_region_iommu_get_attr: return an IOMMU attr if get_attr() is
1747 * defined on the IOMMU.
1749 * Returns 0 on success, or a negative errno otherwise. In particular,
1750 * -EINVAL indicates that the IOMMU does not support the requested
1753 * @iommu_mr: the memory region
1754 * @attr: the requested attribute
1755 * @data: a pointer to the requested attribute data
1757 int memory_region_iommu_get_attr(IOMMUMemoryRegion
*iommu_mr
,
1758 enum IOMMUMemoryRegionAttr attr
,
1762 * memory_region_iommu_attrs_to_index: return the IOMMU index to
1763 * use for translations with the given memory transaction attributes.
1765 * @iommu_mr: the memory region
1766 * @attrs: the memory transaction attributes
1768 int memory_region_iommu_attrs_to_index(IOMMUMemoryRegion
*iommu_mr
,
1772 * memory_region_iommu_num_indexes: return the total number of IOMMU
1773 * indexes that this IOMMU supports.
1775 * @iommu_mr: the memory region
1777 int memory_region_iommu_num_indexes(IOMMUMemoryRegion
*iommu_mr
);
1780 * memory_region_iommu_set_page_size_mask: set the supported page
1781 * sizes for a given IOMMU memory region
1783 * @iommu_mr: IOMMU memory region
1784 * @page_size_mask: supported page size mask
1785 * @errp: pointer to Error*, to store an error if it happens.
1787 int memory_region_iommu_set_page_size_mask(IOMMUMemoryRegion
*iommu_mr
,
1788 uint64_t page_size_mask
,
1792 * memory_region_name: get a memory region's name
1794 * Returns the string that was used to initialize the memory region.
1796 * @mr: the memory region being queried
1798 const char *memory_region_name(const MemoryRegion
*mr
);
1801 * memory_region_is_logging: return whether a memory region is logging writes
1803 * Returns %true if the memory region is logging writes for the given client
1805 * @mr: the memory region being queried
1806 * @client: the client being queried
1808 bool memory_region_is_logging(MemoryRegion
*mr
, uint8_t client
);
1811 * memory_region_get_dirty_log_mask: return the clients for which a
1812 * memory region is logging writes.
1814 * Returns a bitmap of clients, in which the DIRTY_MEMORY_* constants
1815 * are the bit indices.
1817 * @mr: the memory region being queried
1819 uint8_t memory_region_get_dirty_log_mask(MemoryRegion
*mr
);
1822 * memory_region_is_rom: check whether a memory region is ROM
1824 * Returns %true if a memory region is read-only memory.
1826 * @mr: the memory region being queried
1828 static inline bool memory_region_is_rom(MemoryRegion
*mr
)
1830 return mr
->ram
&& mr
->readonly
;
1834 * memory_region_is_nonvolatile: check whether a memory region is non-volatile
1836 * Returns %true is a memory region is non-volatile memory.
1838 * @mr: the memory region being queried
1840 static inline bool memory_region_is_nonvolatile(MemoryRegion
*mr
)
1842 return mr
->nonvolatile
;
1846 * memory_region_get_fd: Get a file descriptor backing a RAM memory region.
1848 * Returns a file descriptor backing a file-based RAM memory region,
1849 * or -1 if the region is not a file-based RAM memory region.
1851 * @mr: the RAM or alias memory region being queried.
1853 int memory_region_get_fd(MemoryRegion
*mr
);
1856 * memory_region_from_host: Convert a pointer into a RAM memory region
1857 * and an offset within it.
1859 * Given a host pointer inside a RAM memory region (created with
1860 * memory_region_init_ram() or memory_region_init_ram_ptr()), return
1861 * the MemoryRegion and the offset within it.
1863 * Use with care; by the time this function returns, the returned pointer is
1864 * not protected by RCU anymore. If the caller is not within an RCU critical
1865 * section and does not hold the iothread lock, it must have other means of
1866 * protecting the pointer, such as a reference to the region that includes
1867 * the incoming ram_addr_t.
1869 * @ptr: the host pointer to be converted
1870 * @offset: the offset within memory region
1872 MemoryRegion
*memory_region_from_host(void *ptr
, ram_addr_t
*offset
);
1875 * memory_region_get_ram_ptr: Get a pointer into a RAM memory region.
1877 * Returns a host pointer to a RAM memory region (created with
1878 * memory_region_init_ram() or memory_region_init_ram_ptr()).
1880 * Use with care; by the time this function returns, the returned pointer is
1881 * not protected by RCU anymore. If the caller is not within an RCU critical
1882 * section and does not hold the iothread lock, it must have other means of
1883 * protecting the pointer, such as a reference to the region that includes
1884 * the incoming ram_addr_t.
1886 * @mr: the memory region being queried.
1888 void *memory_region_get_ram_ptr(MemoryRegion
*mr
);
1890 /* memory_region_ram_resize: Resize a RAM region.
1892 * Resizing RAM while migrating can result in the migration being canceled.
1893 * Care has to be taken if the guest might have already detected the memory.
1895 * @mr: a memory region created with @memory_region_init_resizeable_ram.
1896 * @newsize: the new size the region
1897 * @errp: pointer to Error*, to store an error if it happens.
1899 void memory_region_ram_resize(MemoryRegion
*mr
, ram_addr_t newsize
,
1903 * memory_region_msync: Synchronize selected address range of
1904 * a memory mapped region
1906 * @mr: the memory region to be msync
1907 * @addr: the initial address of the range to be sync
1908 * @size: the size of the range to be sync
1910 void memory_region_msync(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
1913 * memory_region_writeback: Trigger cache writeback for
1914 * selected address range
1916 * @mr: the memory region to be updated
1917 * @addr: the initial address of the range to be written back
1918 * @size: the size of the range to be written back
1920 void memory_region_writeback(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
1923 * memory_region_set_log: Turn dirty logging on or off for a region.
1925 * Turns dirty logging on or off for a specified client (display, migration).
1926 * Only meaningful for RAM regions.
1928 * @mr: the memory region being updated.
1929 * @log: whether dirty logging is to be enabled or disabled.
1930 * @client: the user of the logging information; %DIRTY_MEMORY_VGA only.
1932 void memory_region_set_log(MemoryRegion
*mr
, bool log
, unsigned client
);
1935 * memory_region_set_dirty: Mark a range of bytes as dirty in a memory region.
1937 * Marks a range of bytes as dirty, after it has been dirtied outside
1940 * @mr: the memory region being dirtied.
1941 * @addr: the address (relative to the start of the region) being dirtied.
1942 * @size: size of the range being dirtied.
1944 void memory_region_set_dirty(MemoryRegion
*mr
, hwaddr addr
,
1948 * memory_region_clear_dirty_bitmap - clear dirty bitmap for memory range
1950 * This function is called when the caller wants to clear the remote
1951 * dirty bitmap of a memory range within the memory region. This can
1952 * be used by e.g. KVM to manually clear dirty log when
1953 * KVM_CAP_MANUAL_DIRTY_LOG_PROTECT is declared support by the host
1956 * @mr: the memory region to clear the dirty log upon
1957 * @start: start address offset within the memory region
1958 * @len: length of the memory region to clear dirty bitmap
1960 void memory_region_clear_dirty_bitmap(MemoryRegion
*mr
, hwaddr start
,
1964 * memory_region_snapshot_and_clear_dirty: Get a snapshot of the dirty
1965 * bitmap and clear it.
1967 * Creates a snapshot of the dirty bitmap, clears the dirty bitmap and
1968 * returns the snapshot. The snapshot can then be used to query dirty
1969 * status, using memory_region_snapshot_get_dirty. Snapshotting allows
1970 * querying the same page multiple times, which is especially useful for
1971 * display updates where the scanlines often are not page aligned.
1973 * The dirty bitmap region which gets copyed into the snapshot (and
1974 * cleared afterwards) can be larger than requested. The boundaries
1975 * are rounded up/down so complete bitmap longs (covering 64 pages on
1976 * 64bit hosts) can be copied over into the bitmap snapshot. Which
1977 * isn't a problem for display updates as the extra pages are outside
1978 * the visible area, and in case the visible area changes a full
1979 * display redraw is due anyway. Should other use cases for this
1980 * function emerge we might have to revisit this implementation
1983 * Use g_free to release DirtyBitmapSnapshot.
1985 * @mr: the memory region being queried.
1986 * @addr: the address (relative to the start of the region) being queried.
1987 * @size: the size of the range being queried.
1988 * @client: the user of the logging information; typically %DIRTY_MEMORY_VGA.
1990 DirtyBitmapSnapshot
*memory_region_snapshot_and_clear_dirty(MemoryRegion
*mr
,
1996 * memory_region_snapshot_get_dirty: Check whether a range of bytes is dirty
1997 * in the specified dirty bitmap snapshot.
1999 * @mr: the memory region being queried.
2000 * @snap: the dirty bitmap snapshot
2001 * @addr: the address (relative to the start of the region) being queried.
2002 * @size: the size of the range being queried.
2004 bool memory_region_snapshot_get_dirty(MemoryRegion
*mr
,
2005 DirtyBitmapSnapshot
*snap
,
2006 hwaddr addr
, hwaddr size
);
2009 * memory_region_reset_dirty: Mark a range of pages as clean, for a specified
2012 * Marks a range of pages as no longer dirty.
2014 * @mr: the region being updated.
2015 * @addr: the start of the subrange being cleaned.
2016 * @size: the size of the subrange being cleaned.
2017 * @client: the user of the logging information; %DIRTY_MEMORY_MIGRATION or
2018 * %DIRTY_MEMORY_VGA.
2020 void memory_region_reset_dirty(MemoryRegion
*mr
, hwaddr addr
,
2021 hwaddr size
, unsigned client
);
2024 * memory_region_flush_rom_device: Mark a range of pages dirty and invalidate
2025 * TBs (for self-modifying code).
2027 * The MemoryRegionOps->write() callback of a ROM device must use this function
2028 * to mark byte ranges that have been modified internally, such as by directly
2029 * accessing the memory returned by memory_region_get_ram_ptr().
2031 * This function marks the range dirty and invalidates TBs so that TCG can
2032 * detect self-modifying code.
2034 * @mr: the region being flushed.
2035 * @addr: the start, relative to the start of the region, of the range being
2037 * @size: the size, in bytes, of the range being flushed.
2039 void memory_region_flush_rom_device(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
2042 * memory_region_set_readonly: Turn a memory region read-only (or read-write)
2044 * Allows a memory region to be marked as read-only (turning it into a ROM).
2045 * only useful on RAM regions.
2047 * @mr: the region being updated.
2048 * @readonly: whether rhe region is to be ROM or RAM.
2050 void memory_region_set_readonly(MemoryRegion
*mr
, bool readonly
);
2053 * memory_region_set_nonvolatile: Turn a memory region non-volatile
2055 * Allows a memory region to be marked as non-volatile.
2056 * only useful on RAM regions.
2058 * @mr: the region being updated.
2059 * @nonvolatile: whether rhe region is to be non-volatile.
2061 void memory_region_set_nonvolatile(MemoryRegion
*mr
, bool nonvolatile
);
2064 * memory_region_rom_device_set_romd: enable/disable ROMD mode
2066 * Allows a ROM device (initialized with memory_region_init_rom_device() to
2067 * set to ROMD mode (default) or MMIO mode. When it is in ROMD mode, the
2068 * device is mapped to guest memory and satisfies read access directly.
2069 * When in MMIO mode, reads are forwarded to the #MemoryRegion.read function.
2070 * Writes are always handled by the #MemoryRegion.write function.
2072 * @mr: the memory region to be updated
2073 * @romd_mode: %true to put the region into ROMD mode
2075 void memory_region_rom_device_set_romd(MemoryRegion
*mr
, bool romd_mode
);
2078 * memory_region_set_coalescing: Enable memory coalescing for the region.
2080 * Enabled writes to a region to be queued for later processing. MMIO ->write
2081 * callbacks may be delayed until a non-coalesced MMIO is issued.
2082 * Only useful for IO regions. Roughly similar to write-combining hardware.
2084 * @mr: the memory region to be write coalesced
2086 void memory_region_set_coalescing(MemoryRegion
*mr
);
2089 * memory_region_add_coalescing: Enable memory coalescing for a sub-range of
2092 * Like memory_region_set_coalescing(), but works on a sub-range of a region.
2093 * Multiple calls can be issued coalesced disjoint ranges.
2095 * @mr: the memory region to be updated.
2096 * @offset: the start of the range within the region to be coalesced.
2097 * @size: the size of the subrange to be coalesced.
2099 void memory_region_add_coalescing(MemoryRegion
*mr
,
2104 * memory_region_clear_coalescing: Disable MMIO coalescing for the region.
2106 * Disables any coalescing caused by memory_region_set_coalescing() or
2107 * memory_region_add_coalescing(). Roughly equivalent to uncacheble memory
2110 * @mr: the memory region to be updated.
2112 void memory_region_clear_coalescing(MemoryRegion
*mr
);
2115 * memory_region_set_flush_coalesced: Enforce memory coalescing flush before
2118 * Ensure that pending coalesced MMIO request are flushed before the memory
2119 * region is accessed. This property is automatically enabled for all regions
2120 * passed to memory_region_set_coalescing() and memory_region_add_coalescing().
2122 * @mr: the memory region to be updated.
2124 void memory_region_set_flush_coalesced(MemoryRegion
*mr
);
2127 * memory_region_clear_flush_coalesced: Disable memory coalescing flush before
2130 * Clear the automatic coalesced MMIO flushing enabled via
2131 * memory_region_set_flush_coalesced. Note that this service has no effect on
2132 * memory regions that have MMIO coalescing enabled for themselves. For them,
2133 * automatic flushing will stop once coalescing is disabled.
2135 * @mr: the memory region to be updated.
2137 void memory_region_clear_flush_coalesced(MemoryRegion
*mr
);
2140 * memory_region_add_eventfd: Request an eventfd to be triggered when a word
2141 * is written to a location.
2143 * Marks a word in an IO region (initialized with memory_region_init_io())
2144 * as a trigger for an eventfd event. The I/O callback will not be called.
2145 * The caller must be prepared to handle failure (that is, take the required
2146 * action if the callback _is_ called).
2148 * @mr: the memory region being updated.
2149 * @addr: the address within @mr that is to be monitored
2150 * @size: the size of the access to trigger the eventfd
2151 * @match_data: whether to match against @data, instead of just @addr
2152 * @data: the data to match against the guest write
2153 * @e: event notifier to be triggered when @addr, @size, and @data all match.
2155 void memory_region_add_eventfd(MemoryRegion
*mr
,
2163 * memory_region_del_eventfd: Cancel an eventfd.
2165 * Cancels an eventfd trigger requested by a previous
2166 * memory_region_add_eventfd() call.
2168 * @mr: the memory region being updated.
2169 * @addr: the address within @mr that is to be monitored
2170 * @size: the size of the access to trigger the eventfd
2171 * @match_data: whether to match against @data, instead of just @addr
2172 * @data: the data to match against the guest write
2173 * @e: event notifier to be triggered when @addr, @size, and @data all match.
2175 void memory_region_del_eventfd(MemoryRegion
*mr
,
2183 * memory_region_add_subregion: Add a subregion to a container.
2185 * Adds a subregion at @offset. The subregion may not overlap with other
2186 * subregions (except for those explicitly marked as overlapping). A region
2187 * may only be added once as a subregion (unless removed with
2188 * memory_region_del_subregion()); use memory_region_init_alias() if you
2189 * want a region to be a subregion in multiple locations.
2191 * @mr: the region to contain the new subregion; must be a container
2192 * initialized with memory_region_init().
2193 * @offset: the offset relative to @mr where @subregion is added.
2194 * @subregion: the subregion to be added.
2196 void memory_region_add_subregion(MemoryRegion
*mr
,
2198 MemoryRegion
*subregion
);
2200 * memory_region_add_subregion_overlap: Add a subregion to a container
2203 * Adds a subregion at @offset. The subregion may overlap with other
2204 * subregions. Conflicts are resolved by having a higher @priority hide a
2205 * lower @priority. Subregions without priority are taken as @priority 0.
2206 * A region may only be added once as a subregion (unless removed with
2207 * memory_region_del_subregion()); use memory_region_init_alias() if you
2208 * want a region to be a subregion in multiple locations.
2210 * @mr: the region to contain the new subregion; must be a container
2211 * initialized with memory_region_init().
2212 * @offset: the offset relative to @mr where @subregion is added.
2213 * @subregion: the subregion to be added.
2214 * @priority: used for resolving overlaps; highest priority wins.
2216 void memory_region_add_subregion_overlap(MemoryRegion
*mr
,
2218 MemoryRegion
*subregion
,
2222 * memory_region_get_ram_addr: Get the ram address associated with a memory
2225 * @mr: the region to be queried
2227 ram_addr_t
memory_region_get_ram_addr(MemoryRegion
*mr
);
2229 uint64_t memory_region_get_alignment(const MemoryRegion
*mr
);
2231 * memory_region_del_subregion: Remove a subregion.
2233 * Removes a subregion from its container.
2235 * @mr: the container to be updated.
2236 * @subregion: the region being removed; must be a current subregion of @mr.
2238 void memory_region_del_subregion(MemoryRegion
*mr
,
2239 MemoryRegion
*subregion
);
2242 * memory_region_set_enabled: dynamically enable or disable a region
2244 * Enables or disables a memory region. A disabled memory region
2245 * ignores all accesses to itself and its subregions. It does not
2246 * obscure sibling subregions with lower priority - it simply behaves as
2247 * if it was removed from the hierarchy.
2249 * Regions default to being enabled.
2251 * @mr: the region to be updated
2252 * @enabled: whether to enable or disable the region
2254 void memory_region_set_enabled(MemoryRegion
*mr
, bool enabled
);
2257 * memory_region_set_address: dynamically update the address of a region
2259 * Dynamically updates the address of a region, relative to its container.
2260 * May be used on regions are currently part of a memory hierarchy.
2262 * @mr: the region to be updated
2263 * @addr: new address, relative to container region
2265 void memory_region_set_address(MemoryRegion
*mr
, hwaddr addr
);
2268 * memory_region_set_size: dynamically update the size of a region.
2270 * Dynamically updates the size of a region.
2272 * @mr: the region to be updated
2273 * @size: used size of the region.
2275 void memory_region_set_size(MemoryRegion
*mr
, uint64_t size
);
2278 * memory_region_set_alias_offset: dynamically update a memory alias's offset
2280 * Dynamically updates the offset into the target region that an alias points
2281 * to, as if the fourth argument to memory_region_init_alias() has changed.
2283 * @mr: the #MemoryRegion to be updated; should be an alias.
2284 * @offset: the new offset into the target memory region
2286 void memory_region_set_alias_offset(MemoryRegion
*mr
,
2290 * memory_region_present: checks if an address relative to a @container
2291 * translates into #MemoryRegion within @container
2293 * Answer whether a #MemoryRegion within @container covers the address
2296 * @container: a #MemoryRegion within which @addr is a relative address
2297 * @addr: the area within @container to be searched
2299 bool memory_region_present(MemoryRegion
*container
, hwaddr addr
);
2302 * memory_region_is_mapped: returns true if #MemoryRegion is mapped
2303 * into another memory region, which does not necessarily imply that it is
2304 * mapped into an address space.
2306 * @mr: a #MemoryRegion which should be checked if it's mapped
2308 bool memory_region_is_mapped(MemoryRegion
*mr
);
2311 * memory_region_get_ram_discard_manager: get the #RamDiscardManager for a
2314 * The #RamDiscardManager cannot change while a memory region is mapped.
2316 * @mr: the #MemoryRegion
2318 RamDiscardManager
*memory_region_get_ram_discard_manager(MemoryRegion
*mr
);
2321 * memory_region_has_ram_discard_manager: check whether a #MemoryRegion has a
2322 * #RamDiscardManager assigned
2324 * @mr: the #MemoryRegion
2326 static inline bool memory_region_has_ram_discard_manager(MemoryRegion
*mr
)
2328 return !!memory_region_get_ram_discard_manager(mr
);
2332 * memory_region_set_ram_discard_manager: set the #RamDiscardManager for a
2335 * This function must not be called for a mapped #MemoryRegion, a #MemoryRegion
2336 * that does not cover RAM, or a #MemoryRegion that already has a
2337 * #RamDiscardManager assigned.
2339 * @mr: the #MemoryRegion
2340 * @rdm: #RamDiscardManager to set
2342 void memory_region_set_ram_discard_manager(MemoryRegion
*mr
,
2343 RamDiscardManager
*rdm
);
2346 * memory_region_find: translate an address/size relative to a
2347 * MemoryRegion into a #MemoryRegionSection.
2349 * Locates the first #MemoryRegion within @mr that overlaps the range
2350 * given by @addr and @size.
2352 * Returns a #MemoryRegionSection that describes a contiguous overlap.
2353 * It will have the following characteristics:
2354 * - @size = 0 iff no overlap was found
2355 * - @mr is non-%NULL iff an overlap was found
2357 * Remember that in the return value the @offset_within_region is
2358 * relative to the returned region (in the .@mr field), not to the
2361 * Similarly, the .@offset_within_address_space is relative to the
2362 * address space that contains both regions, the passed and the
2363 * returned one. However, in the special case where the @mr argument
2364 * has no container (and thus is the root of the address space), the
2365 * following will hold:
2366 * - @offset_within_address_space >= @addr
2367 * - @offset_within_address_space + .@size <= @addr + @size
2369 * @mr: a MemoryRegion within which @addr is a relative address
2370 * @addr: start of the area within @as to be searched
2371 * @size: size of the area to be searched
2373 MemoryRegionSection
memory_region_find(MemoryRegion
*mr
,
2374 hwaddr addr
, uint64_t size
);
2377 * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2379 * Synchronizes the dirty page log for all address spaces.
2381 void memory_global_dirty_log_sync(void);
2384 * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2386 * Synchronizes the vCPUs with a thread that is reading the dirty bitmap.
2387 * This function must be called after the dirty log bitmap is cleared, and
2388 * before dirty guest memory pages are read. If you are using
2389 * #DirtyBitmapSnapshot, memory_region_snapshot_and_clear_dirty() takes
2390 * care of doing this.
2392 void memory_global_after_dirty_log_sync(void);
2395 * memory_region_transaction_begin: Start a transaction.
2397 * During a transaction, changes will be accumulated and made visible
2398 * only when the transaction ends (is committed).
2400 void memory_region_transaction_begin(void);
2403 * memory_region_transaction_commit: Commit a transaction and make changes
2404 * visible to the guest.
2406 void memory_region_transaction_commit(void);
2409 * memory_listener_register: register callbacks to be called when memory
2410 * sections are mapped or unmapped into an address
2413 * @listener: an object containing the callbacks to be called
2414 * @filter: if non-%NULL, only regions in this address space will be observed
2416 void memory_listener_register(MemoryListener
*listener
, AddressSpace
*filter
);
2419 * memory_listener_unregister: undo the effect of memory_listener_register()
2421 * @listener: an object containing the callbacks to be removed
2423 void memory_listener_unregister(MemoryListener
*listener
);
2426 * memory_global_dirty_log_start: begin dirty logging for all regions
2428 * @flags: purpose of starting dirty log, migration or dirty rate
2430 void memory_global_dirty_log_start(unsigned int flags
);
2433 * memory_global_dirty_log_stop: end dirty logging for all regions
2435 * @flags: purpose of stopping dirty log, migration or dirty rate
2437 void memory_global_dirty_log_stop(unsigned int flags
);
2439 void mtree_info(bool flatview
, bool dispatch_tree
, bool owner
, bool disabled
);
2442 * memory_region_dispatch_read: perform a read directly to the specified
2445 * @mr: #MemoryRegion to access
2446 * @addr: address within that region
2447 * @pval: pointer to uint64_t which the data is written to
2448 * @op: size, sign, and endianness of the memory operation
2449 * @attrs: memory transaction attributes to use for the access
2451 MemTxResult
memory_region_dispatch_read(MemoryRegion
*mr
,
2457 * memory_region_dispatch_write: perform a write directly to the specified
2460 * @mr: #MemoryRegion to access
2461 * @addr: address within that region
2462 * @data: data to write
2463 * @op: size, sign, and endianness of the memory operation
2464 * @attrs: memory transaction attributes to use for the access
2466 MemTxResult
memory_region_dispatch_write(MemoryRegion
*mr
,
2473 * address_space_init: initializes an address space
2475 * @as: an uninitialized #AddressSpace
2476 * @root: a #MemoryRegion that routes addresses for the address space
2477 * @name: an address space name. The name is only used for debugging
2480 void address_space_init(AddressSpace
*as
, MemoryRegion
*root
, const char *name
);
2483 * address_space_destroy: destroy an address space
2485 * Releases all resources associated with an address space. After an address space
2486 * is destroyed, its root memory region (given by address_space_init()) may be destroyed
2489 * @as: address space to be destroyed
2491 void address_space_destroy(AddressSpace
*as
);
2494 * address_space_remove_listeners: unregister all listeners of an address space
2496 * Removes all callbacks previously registered with memory_listener_register()
2499 * @as: an initialized #AddressSpace
2501 void address_space_remove_listeners(AddressSpace
*as
);
2504 * address_space_rw: read from or write to an address space.
2506 * Return a MemTxResult indicating whether the operation succeeded
2507 * or failed (eg unassigned memory, device rejected the transaction,
2510 * @as: #AddressSpace to be accessed
2511 * @addr: address within that address space
2512 * @attrs: memory transaction attributes
2513 * @buf: buffer with the data transferred
2514 * @len: the number of bytes to read or write
2515 * @is_write: indicates the transfer direction
2517 MemTxResult
address_space_rw(AddressSpace
*as
, hwaddr addr
,
2518 MemTxAttrs attrs
, void *buf
,
2519 hwaddr len
, bool is_write
);
2522 * address_space_write: write to address space.
2524 * Return a MemTxResult indicating whether the operation succeeded
2525 * or failed (eg unassigned memory, device rejected the transaction,
2528 * @as: #AddressSpace to be accessed
2529 * @addr: address within that address space
2530 * @attrs: memory transaction attributes
2531 * @buf: buffer with the data transferred
2532 * @len: the number of bytes to write
2534 MemTxResult
address_space_write(AddressSpace
*as
, hwaddr addr
,
2536 const void *buf
, hwaddr len
);
2539 * address_space_write_rom: write to address space, including ROM.
2541 * This function writes to the specified address space, but will
2542 * write data to both ROM and RAM. This is used for non-guest
2543 * writes like writes from the gdb debug stub or initial loading
2546 * Note that portions of the write which attempt to write data to
2547 * a device will be silently ignored -- only real RAM and ROM will
2550 * Return a MemTxResult indicating whether the operation succeeded
2551 * or failed (eg unassigned memory, device rejected the transaction,
2554 * @as: #AddressSpace to be accessed
2555 * @addr: address within that address space
2556 * @attrs: memory transaction attributes
2557 * @buf: buffer with the data transferred
2558 * @len: the number of bytes to write
2560 MemTxResult
address_space_write_rom(AddressSpace
*as
, hwaddr addr
,
2562 const void *buf
, hwaddr len
);
2564 /* address_space_ld*: load from an address space
2565 * address_space_st*: store to an address space
2567 * These functions perform a load or store of the byte, word,
2568 * longword or quad to the specified address within the AddressSpace.
2569 * The _le suffixed functions treat the data as little endian;
2570 * _be indicates big endian; no suffix indicates "same endianness
2573 * The "guest CPU endianness" accessors are deprecated for use outside
2574 * target-* code; devices should be CPU-agnostic and use either the LE
2575 * or the BE accessors.
2577 * @as #AddressSpace to be accessed
2578 * @addr: address within that address space
2579 * @val: data value, for stores
2580 * @attrs: memory transaction attributes
2581 * @result: location to write the success/failure of the transaction;
2582 * if NULL, this information is discarded
2587 #define ARG1_DECL AddressSpace *as
2588 #include "exec/memory_ldst.h.inc"
2592 #define ARG1_DECL AddressSpace *as
2593 #include "exec/memory_ldst_phys.h.inc"
2595 struct MemoryRegionCache
{
2600 MemoryRegionSection mrs
;
2604 #define MEMORY_REGION_CACHE_INVALID ((MemoryRegionCache) { .mrs.mr = NULL })
2607 /* address_space_ld*_cached: load from a cached #MemoryRegion
2608 * address_space_st*_cached: store into a cached #MemoryRegion
2610 * These functions perform a load or store of the byte, word,
2611 * longword or quad to the specified address. The address is
2612 * a physical address in the AddressSpace, but it must lie within
2613 * a #MemoryRegion that was mapped with address_space_cache_init.
2615 * The _le suffixed functions treat the data as little endian;
2616 * _be indicates big endian; no suffix indicates "same endianness
2619 * The "guest CPU endianness" accessors are deprecated for use outside
2620 * target-* code; devices should be CPU-agnostic and use either the LE
2621 * or the BE accessors.
2623 * @cache: previously initialized #MemoryRegionCache to be accessed
2624 * @addr: address within the address space
2625 * @val: data value, for stores
2626 * @attrs: memory transaction attributes
2627 * @result: location to write the success/failure of the transaction;
2628 * if NULL, this information is discarded
2631 #define SUFFIX _cached_slow
2633 #define ARG1_DECL MemoryRegionCache *cache
2634 #include "exec/memory_ldst.h.inc"
2636 /* Inline fast path for direct RAM access. */
2637 static inline uint8_t address_space_ldub_cached(MemoryRegionCache
*cache
,
2638 hwaddr addr
, MemTxAttrs attrs
, MemTxResult
*result
)
2640 assert(addr
< cache
->len
);
2641 if (likely(cache
->ptr
)) {
2642 return ldub_p(cache
->ptr
+ addr
);
2644 return address_space_ldub_cached_slow(cache
, addr
, attrs
, result
);
2648 static inline void address_space_stb_cached(MemoryRegionCache
*cache
,
2649 hwaddr addr
, uint8_t val
, MemTxAttrs attrs
, MemTxResult
*result
)
2651 assert(addr
< cache
->len
);
2652 if (likely(cache
->ptr
)) {
2653 stb_p(cache
->ptr
+ addr
, val
);
2655 address_space_stb_cached_slow(cache
, addr
, val
, attrs
, result
);
2659 #define ENDIANNESS _le
2660 #include "exec/memory_ldst_cached.h.inc"
2662 #define ENDIANNESS _be
2663 #include "exec/memory_ldst_cached.h.inc"
2665 #define SUFFIX _cached
2667 #define ARG1_DECL MemoryRegionCache *cache
2668 #include "exec/memory_ldst_phys.h.inc"
2670 /* address_space_cache_init: prepare for repeated access to a physical
2673 * @cache: #MemoryRegionCache to be filled
2674 * @as: #AddressSpace to be accessed
2675 * @addr: address within that address space
2676 * @len: length of buffer
2677 * @is_write: indicates the transfer direction
2679 * Will only work with RAM, and may map a subset of the requested range by
2680 * returning a value that is less than @len. On failure, return a negative
2683 * Because it only works with RAM, this function can be used for
2684 * read-modify-write operations. In this case, is_write should be %true.
2686 * Note that addresses passed to the address_space_*_cached functions
2687 * are relative to @addr.
2689 int64_t address_space_cache_init(MemoryRegionCache
*cache
,
2696 * address_space_cache_invalidate: complete a write to a #MemoryRegionCache
2698 * @cache: The #MemoryRegionCache to operate on.
2699 * @addr: The first physical address that was written, relative to the
2700 * address that was passed to @address_space_cache_init.
2701 * @access_len: The number of bytes that were written starting at @addr.
2703 void address_space_cache_invalidate(MemoryRegionCache
*cache
,
2708 * address_space_cache_destroy: free a #MemoryRegionCache
2710 * @cache: The #MemoryRegionCache whose memory should be released.
2712 void address_space_cache_destroy(MemoryRegionCache
*cache
);
2714 /* address_space_get_iotlb_entry: translate an address into an IOTLB
2715 * entry. Should be called from an RCU critical section.
2717 IOMMUTLBEntry
address_space_get_iotlb_entry(AddressSpace
*as
, hwaddr addr
,
2718 bool is_write
, MemTxAttrs attrs
);
2720 /* address_space_translate: translate an address range into an address space
2721 * into a MemoryRegion and an address range into that section. Should be
2722 * called from an RCU critical section, to avoid that the last reference
2723 * to the returned region disappears after address_space_translate returns.
2725 * @fv: #FlatView to be accessed
2726 * @addr: address within that address space
2727 * @xlat: pointer to address within the returned memory region section's
2729 * @len: pointer to length
2730 * @is_write: indicates the transfer direction
2731 * @attrs: memory attributes
2733 MemoryRegion
*flatview_translate(FlatView
*fv
,
2734 hwaddr addr
, hwaddr
*xlat
,
2735 hwaddr
*len
, bool is_write
,
2738 static inline MemoryRegion
*address_space_translate(AddressSpace
*as
,
2739 hwaddr addr
, hwaddr
*xlat
,
2740 hwaddr
*len
, bool is_write
,
2743 return flatview_translate(address_space_to_flatview(as
),
2744 addr
, xlat
, len
, is_write
, attrs
);
2747 /* address_space_access_valid: check for validity of accessing an address
2750 * Check whether memory is assigned to the given address space range, and
2751 * access is permitted by any IOMMU regions that are active for the address
2754 * For now, addr and len should be aligned to a page size. This limitation
2755 * will be lifted in the future.
2757 * @as: #AddressSpace to be accessed
2758 * @addr: address within that address space
2759 * @len: length of the area to be checked
2760 * @is_write: indicates the transfer direction
2761 * @attrs: memory attributes
2763 bool address_space_access_valid(AddressSpace
*as
, hwaddr addr
, hwaddr len
,
2764 bool is_write
, MemTxAttrs attrs
);
2766 /* address_space_map: map a physical memory region into a host virtual address
2768 * May map a subset of the requested range, given by and returned in @plen.
2769 * May return %NULL and set *@plen to zero(0), if resources needed to perform
2770 * the mapping are exhausted.
2771 * Use only for reads OR writes - not for read-modify-write operations.
2772 * Use cpu_register_map_client() to know when retrying the map operation is
2773 * likely to succeed.
2775 * @as: #AddressSpace to be accessed
2776 * @addr: address within that address space
2777 * @plen: pointer to length of buffer; updated on return
2778 * @is_write: indicates the transfer direction
2779 * @attrs: memory attributes
2781 void *address_space_map(AddressSpace
*as
, hwaddr addr
,
2782 hwaddr
*plen
, bool is_write
, MemTxAttrs attrs
);
2784 /* address_space_unmap: Unmaps a memory region previously mapped by address_space_map()
2786 * Will also mark the memory as dirty if @is_write == %true. @access_len gives
2787 * the amount of memory that was actually read or written by the caller.
2789 * @as: #AddressSpace used
2790 * @buffer: host pointer as returned by address_space_map()
2791 * @len: buffer length as returned by address_space_map()
2792 * @access_len: amount of data actually transferred
2793 * @is_write: indicates the transfer direction
2795 void address_space_unmap(AddressSpace
*as
, void *buffer
, hwaddr len
,
2796 bool is_write
, hwaddr access_len
);
2799 /* Internal functions, part of the implementation of address_space_read. */
2800 MemTxResult
address_space_read_full(AddressSpace
*as
, hwaddr addr
,
2801 MemTxAttrs attrs
, void *buf
, hwaddr len
);
2802 MemTxResult
flatview_read_continue(FlatView
*fv
, hwaddr addr
,
2803 MemTxAttrs attrs
, void *buf
,
2804 hwaddr len
, hwaddr addr1
, hwaddr l
,
2806 void *qemu_map_ram_ptr(RAMBlock
*ram_block
, ram_addr_t addr
);
2808 /* Internal functions, part of the implementation of address_space_read_cached
2809 * and address_space_write_cached. */
2810 MemTxResult
address_space_read_cached_slow(MemoryRegionCache
*cache
,
2811 hwaddr addr
, void *buf
, hwaddr len
);
2812 MemTxResult
address_space_write_cached_slow(MemoryRegionCache
*cache
,
2813 hwaddr addr
, const void *buf
,
2816 int memory_access_size(MemoryRegion
*mr
, unsigned l
, hwaddr addr
);
2817 bool prepare_mmio_access(MemoryRegion
*mr
);
2819 static inline bool memory_access_is_direct(MemoryRegion
*mr
, bool is_write
)
2822 return memory_region_is_ram(mr
) && !mr
->readonly
&&
2823 !mr
->rom_device
&& !memory_region_is_ram_device(mr
);
2825 return (memory_region_is_ram(mr
) && !memory_region_is_ram_device(mr
)) ||
2826 memory_region_is_romd(mr
);
2831 * address_space_read: read from an address space.
2833 * Return a MemTxResult indicating whether the operation succeeded
2834 * or failed (eg unassigned memory, device rejected the transaction,
2835 * IOMMU fault). Called within RCU critical section.
2837 * @as: #AddressSpace to be accessed
2838 * @addr: address within that address space
2839 * @attrs: memory transaction attributes
2840 * @buf: buffer with the data transferred
2841 * @len: length of the data transferred
2843 static inline __attribute__((__always_inline__
))
2844 MemTxResult
address_space_read(AddressSpace
*as
, hwaddr addr
,
2845 MemTxAttrs attrs
, void *buf
,
2848 MemTxResult result
= MEMTX_OK
;
2854 if (__builtin_constant_p(len
)) {
2856 RCU_READ_LOCK_GUARD();
2857 fv
= address_space_to_flatview(as
);
2859 mr
= flatview_translate(fv
, addr
, &addr1
, &l
, false, attrs
);
2860 if (len
== l
&& memory_access_is_direct(mr
, false)) {
2861 ptr
= qemu_map_ram_ptr(mr
->ram_block
, addr1
);
2862 memcpy(buf
, ptr
, len
);
2864 result
= flatview_read_continue(fv
, addr
, attrs
, buf
, len
,
2869 result
= address_space_read_full(as
, addr
, attrs
, buf
, len
);
2875 * address_space_read_cached: read from a cached RAM region
2877 * @cache: Cached region to be addressed
2878 * @addr: address relative to the base of the RAM region
2879 * @buf: buffer with the data transferred
2880 * @len: length of the data transferred
2882 static inline MemTxResult
2883 address_space_read_cached(MemoryRegionCache
*cache
, hwaddr addr
,
2884 void *buf
, hwaddr len
)
2886 assert(addr
< cache
->len
&& len
<= cache
->len
- addr
);
2887 fuzz_dma_read_cb(cache
->xlat
+ addr
, len
, cache
->mrs
.mr
);
2888 if (likely(cache
->ptr
)) {
2889 memcpy(buf
, cache
->ptr
+ addr
, len
);
2892 return address_space_read_cached_slow(cache
, addr
, buf
, len
);
2897 * address_space_write_cached: write to a cached RAM region
2899 * @cache: Cached region to be addressed
2900 * @addr: address relative to the base of the RAM region
2901 * @buf: buffer with the data transferred
2902 * @len: length of the data transferred
2904 static inline MemTxResult
2905 address_space_write_cached(MemoryRegionCache
*cache
, hwaddr addr
,
2906 const void *buf
, hwaddr len
)
2908 assert(addr
< cache
->len
&& len
<= cache
->len
- addr
);
2909 if (likely(cache
->ptr
)) {
2910 memcpy(cache
->ptr
+ addr
, buf
, len
);
2913 return address_space_write_cached_slow(cache
, addr
, buf
, len
);
2918 * address_space_set: Fill address space with a constant byte.
2920 * Return a MemTxResult indicating whether the operation succeeded
2921 * or failed (eg unassigned memory, device rejected the transaction,
2924 * @as: #AddressSpace to be accessed
2925 * @addr: address within that address space
2926 * @c: constant byte to fill the memory
2927 * @len: the number of bytes to fill with the constant byte
2928 * @attrs: memory transaction attributes
2930 MemTxResult
address_space_set(AddressSpace
*as
, hwaddr addr
,
2931 uint8_t c
, hwaddr len
, MemTxAttrs attrs
);
2934 /* enum device_endian to MemOp. */
2935 static inline MemOp
devend_memop(enum device_endian end
)
2937 QEMU_BUILD_BUG_ON(DEVICE_HOST_ENDIAN
!= DEVICE_LITTLE_ENDIAN
&&
2938 DEVICE_HOST_ENDIAN
!= DEVICE_BIG_ENDIAN
);
2940 #if HOST_BIG_ENDIAN != TARGET_BIG_ENDIAN
2941 /* Swap if non-host endianness or native (target) endianness */
2942 return (end
== DEVICE_HOST_ENDIAN
) ? 0 : MO_BSWAP
;
2944 const int non_host_endianness
=
2945 DEVICE_LITTLE_ENDIAN
^ DEVICE_BIG_ENDIAN
^ DEVICE_HOST_ENDIAN
;
2947 /* In this case, native (target) endianness needs no swap. */
2948 return (end
== non_host_endianness
) ? MO_BSWAP
: 0;
2954 * Inhibit technologies that require discarding of pages in RAM blocks, e.g.,
2955 * to manage the actual amount of memory consumed by the VM (then, the memory
2956 * provided by RAM blocks might be bigger than the desired memory consumption).
2957 * This *must* be set if:
2958 * - Discarding parts of a RAM blocks does not result in the change being
2959 * reflected in the VM and the pages getting freed.
2960 * - All memory in RAM blocks is pinned or duplicated, invaldiating any previous
2962 * - Discarding parts of a RAM blocks will result in integrity issues (e.g.,
2964 * Technologies that only temporarily pin the current working set of a
2965 * driver are fine, because we don't expect such pages to be discarded
2966 * (esp. based on guest action like balloon inflation).
2968 * This is *not* to be used to protect from concurrent discards (esp.,
2971 * Returns 0 if successful. Returns -EBUSY if a technology that relies on
2972 * discards to work reliably is active.
2974 int ram_block_discard_disable(bool state
);
2977 * See ram_block_discard_disable(): only disable uncoordinated discards,
2978 * keeping coordinated discards (via the RamDiscardManager) enabled.
2980 int ram_block_uncoordinated_discard_disable(bool state
);
2983 * Inhibit technologies that disable discarding of pages in RAM blocks.
2985 * Returns 0 if successful. Returns -EBUSY if discards are already set to
2988 int ram_block_discard_require(bool state
);
2991 * See ram_block_discard_require(): only inhibit technologies that disable
2992 * uncoordinated discarding of pages in RAM blocks, allowing co-existance with
2993 * technologies that only inhibit uncoordinated discards (via the
2994 * RamDiscardManager).
2996 int ram_block_coordinated_discard_require(bool state
);
2999 * Test if any discarding of memory in ram blocks is disabled.
3001 bool ram_block_discard_is_disabled(void);
3004 * Test if any discarding of memory in ram blocks is required to work reliably.
3006 bool ram_block_discard_is_required(void);