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/range.h"
28 #include "qemu/notify.h"
29 #include "qom/object.h"
32 #define RAM_ADDR_INVALID (~(ram_addr_t)0)
34 #define MAX_PHYS_ADDR_SPACE_BITS 62
35 #define MAX_PHYS_ADDR (((hwaddr)1 << MAX_PHYS_ADDR_SPACE_BITS) - 1)
37 #define TYPE_MEMORY_REGION "memory-region"
38 DECLARE_INSTANCE_CHECKER(MemoryRegion
, MEMORY_REGION
,
41 #define TYPE_IOMMU_MEMORY_REGION "iommu-memory-region"
42 typedef struct IOMMUMemoryRegionClass IOMMUMemoryRegionClass
;
43 DECLARE_OBJ_CHECKERS(IOMMUMemoryRegion
, IOMMUMemoryRegionClass
,
44 IOMMU_MEMORY_REGION
, TYPE_IOMMU_MEMORY_REGION
)
46 #define TYPE_RAM_DISCARD_MANAGER "ram-discard-manager"
47 typedef struct RamDiscardManagerClass RamDiscardManagerClass
;
48 typedef struct RamDiscardManager RamDiscardManager
;
49 DECLARE_OBJ_CHECKERS(RamDiscardManager
, RamDiscardManagerClass
,
50 RAM_DISCARD_MANAGER
, TYPE_RAM_DISCARD_MANAGER
);
53 void fuzz_dma_read_cb(size_t addr
,
57 static inline void fuzz_dma_read_cb(size_t addr
,
65 /* Possible bits for global_dirty_log_{start|stop} */
67 /* Dirty tracking enabled because migration is running */
68 #define GLOBAL_DIRTY_MIGRATION (1U << 0)
70 /* Dirty tracking enabled because measuring dirty rate */
71 #define GLOBAL_DIRTY_DIRTY_RATE (1U << 1)
73 /* Dirty tracking enabled because dirty limit */
74 #define GLOBAL_DIRTY_LIMIT (1U << 2)
76 #define GLOBAL_DIRTY_MASK (0x7)
78 extern unsigned int global_dirty_tracking
;
80 typedef struct MemoryRegionOps MemoryRegionOps
;
82 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
98 * @unmergeable: this section should not get merged with adjacent sections
100 struct MemoryRegionSection
{
104 hwaddr offset_within_region
;
105 hwaddr offset_within_address_space
;
111 typedef struct IOMMUTLBEntry IOMMUTLBEntry
;
113 /* See address_space_translate: bit 0 is read, bit 1 is write. */
121 #define IOMMU_ACCESS_FLAG(r, w) (((r) ? IOMMU_RO : 0) | ((w) ? IOMMU_WO : 0))
123 struct IOMMUTLBEntry
{
124 AddressSpace
*target_as
;
126 hwaddr translated_addr
;
127 hwaddr addr_mask
; /* 0xfff = 4k translation */
128 IOMMUAccessFlags perm
;
132 * Bitmap for different IOMMUNotifier capabilities. Each notifier can
133 * register with one or multiple IOMMU Notifier capability bit(s).
135 * Normally there're two use cases for the notifiers:
137 * (1) When the device needs accurate synchronizations of the vIOMMU page
138 * tables, it needs to register with both MAP|UNMAP notifies (which
139 * is defined as IOMMU_NOTIFIER_IOTLB_EVENTS below).
141 * Regarding to accurate synchronization, it's when the notified
142 * device maintains a shadow page table and must be notified on each
143 * guest MAP (page table entry creation) and UNMAP (invalidation)
144 * events (e.g. VFIO). Both notifications must be accurate so that
145 * the shadow page table is fully in sync with the guest view.
147 * (2) When the device doesn't need accurate synchronizations of the
148 * vIOMMU page tables, it needs to register only with UNMAP or
149 * DEVIOTLB_UNMAP notifies.
151 * It's when the device maintains a cache of IOMMU translations
152 * (IOTLB) and is able to fill that cache by requesting translations
153 * from the vIOMMU through a protocol similar to ATS (Address
154 * Translation Service).
156 * Note that in this mode the vIOMMU will not maintain a shadowed
157 * page table for the address space, and the UNMAP messages can cover
158 * more than the pages that used to get mapped. The IOMMU notifiee
159 * should be able to take care of over-sized invalidations.
162 IOMMU_NOTIFIER_NONE
= 0,
163 /* Notify cache invalidations */
164 IOMMU_NOTIFIER_UNMAP
= 0x1,
165 /* Notify entry changes (newly created entries) */
166 IOMMU_NOTIFIER_MAP
= 0x2,
167 /* Notify changes on device IOTLB entries */
168 IOMMU_NOTIFIER_DEVIOTLB_UNMAP
= 0x04,
171 #define IOMMU_NOTIFIER_IOTLB_EVENTS (IOMMU_NOTIFIER_MAP | IOMMU_NOTIFIER_UNMAP)
172 #define IOMMU_NOTIFIER_DEVIOTLB_EVENTS IOMMU_NOTIFIER_DEVIOTLB_UNMAP
173 #define IOMMU_NOTIFIER_ALL (IOMMU_NOTIFIER_IOTLB_EVENTS | \
174 IOMMU_NOTIFIER_DEVIOTLB_EVENTS)
176 struct IOMMUNotifier
;
177 typedef void (*IOMMUNotify
)(struct IOMMUNotifier
*notifier
,
178 IOMMUTLBEntry
*data
);
180 struct IOMMUNotifier
{
182 IOMMUNotifierFlag notifier_flags
;
183 /* Notify for address space range start <= addr <= end */
187 QLIST_ENTRY(IOMMUNotifier
) node
;
189 typedef struct IOMMUNotifier IOMMUNotifier
;
191 typedef struct IOMMUTLBEvent
{
192 IOMMUNotifierFlag type
;
196 /* RAM is pre-allocated and passed into qemu_ram_alloc_from_ptr */
197 #define RAM_PREALLOC (1 << 0)
199 /* RAM is mmap-ed with MAP_SHARED */
200 #define RAM_SHARED (1 << 1)
202 /* Only a portion of RAM (used_length) is actually used, and migrated.
203 * Resizing RAM while migrating can result in the migration being canceled.
205 #define RAM_RESIZEABLE (1 << 2)
207 /* UFFDIO_ZEROPAGE is available on this RAMBlock to atomically
208 * zero the page and wake waiting processes.
209 * (Set during postcopy)
211 #define RAM_UF_ZEROPAGE (1 << 3)
213 /* RAM can be migrated */
214 #define RAM_MIGRATABLE (1 << 4)
216 /* RAM is a persistent kind memory */
217 #define RAM_PMEM (1 << 5)
221 * UFFDIO_WRITEPROTECT is used on this RAMBlock to
222 * support 'write-tracking' migration type.
223 * Implies ram_state->ram_wt_enabled.
225 #define RAM_UF_WRITEPROTECT (1 << 6)
228 * RAM is mmap-ed with MAP_NORESERVE. When set, reserving swap space (or huge
229 * pages if applicable) is skipped: will bail out if not supported. When not
230 * set, the OS will do the reservation, if supported for the memory type.
232 #define RAM_NORESERVE (1 << 7)
234 /* RAM that isn't accessible through normal means. */
235 #define RAM_PROTECTED (1 << 8)
237 /* RAM is an mmap-ed named file */
238 #define RAM_NAMED_FILE (1 << 9)
240 /* RAM is mmap-ed read-only */
241 #define RAM_READONLY (1 << 10)
243 /* RAM FD is opened read-only */
244 #define RAM_READONLY_FD (1 << 11)
246 /* RAM can be private that has kvm guest memfd backend */
247 #define RAM_GUEST_MEMFD (1 << 12)
249 static inline void iommu_notifier_init(IOMMUNotifier
*n
, IOMMUNotify fn
,
250 IOMMUNotifierFlag flags
,
251 hwaddr start
, hwaddr end
,
255 n
->notifier_flags
= flags
;
258 n
->iommu_idx
= iommu_idx
;
262 * Memory region callbacks
264 struct MemoryRegionOps
{
265 /* Read from the memory region. @addr is relative to @mr; @size is
267 uint64_t (*read
)(void *opaque
,
270 /* Write to the memory region. @addr is relative to @mr; @size is
272 void (*write
)(void *opaque
,
277 MemTxResult (*read_with_attrs
)(void *opaque
,
282 MemTxResult (*write_with_attrs
)(void *opaque
,
288 enum device_endian endianness
;
289 /* Guest-visible constraints: */
291 /* If nonzero, specify bounds on access sizes beyond which a machine
294 unsigned min_access_size
;
295 unsigned max_access_size
;
296 /* If true, unaligned accesses are supported. Otherwise unaligned
297 * accesses throw machine checks.
301 * If present, and returns #false, the transaction is not accepted
302 * by the device (and results in machine dependent behaviour such
303 * as a machine check exception).
305 bool (*accepts
)(void *opaque
, hwaddr addr
,
306 unsigned size
, bool is_write
,
309 /* Internal implementation constraints: */
311 /* If nonzero, specifies the minimum size implemented. Smaller sizes
312 * will be rounded upwards and a partial result will be returned.
314 unsigned min_access_size
;
315 /* If nonzero, specifies the maximum size implemented. Larger sizes
316 * will be done as a series of accesses with smaller sizes.
318 unsigned max_access_size
;
319 /* If true, unaligned accesses are supported. Otherwise all accesses
320 * are converted to (possibly multiple) naturally aligned accesses.
326 typedef struct MemoryRegionClass
{
328 ObjectClass parent_class
;
332 enum IOMMUMemoryRegionAttr
{
333 IOMMU_ATTR_SPAPR_TCE_FD
337 * IOMMUMemoryRegionClass:
339 * All IOMMU implementations need to subclass TYPE_IOMMU_MEMORY_REGION
340 * and provide an implementation of at least the @translate method here
341 * to handle requests to the memory region. Other methods are optional.
343 * The IOMMU implementation must use the IOMMU notifier infrastructure
344 * to report whenever mappings are changed, by calling
345 * memory_region_notify_iommu() (or, if necessary, by calling
346 * memory_region_notify_iommu_one() for each registered notifier).
348 * Conceptually an IOMMU provides a mapping from input address
349 * to an output TLB entry. If the IOMMU is aware of memory transaction
350 * attributes and the output TLB entry depends on the transaction
351 * attributes, we represent this using IOMMU indexes. Each index
352 * selects a particular translation table that the IOMMU has:
354 * @attrs_to_index returns the IOMMU index for a set of transaction attributes
356 * @translate takes an input address and an IOMMU index
358 * and the mapping returned can only depend on the input address and the
361 * Most IOMMUs don't care about the transaction attributes and support
362 * only a single IOMMU index. A more complex IOMMU might have one index
363 * for secure transactions and one for non-secure transactions.
365 struct IOMMUMemoryRegionClass
{
367 MemoryRegionClass parent_class
;
373 * Return a TLB entry that contains a given address.
375 * The IOMMUAccessFlags indicated via @flag are optional and may
376 * be specified as IOMMU_NONE to indicate that the caller needs
377 * the full translation information for both reads and writes. If
378 * the access flags are specified then the IOMMU implementation
379 * may use this as an optimization, to stop doing a page table
380 * walk as soon as it knows that the requested permissions are not
381 * allowed. If IOMMU_NONE is passed then the IOMMU must do the
382 * full page table walk and report the permissions in the returned
383 * IOMMUTLBEntry. (Note that this implies that an IOMMU may not
384 * return different mappings for reads and writes.)
386 * The returned information remains valid while the caller is
387 * holding the big QEMU lock or is inside an RCU critical section;
388 * if the caller wishes to cache the mapping beyond that it must
389 * register an IOMMU notifier so it can invalidate its cached
390 * information when the IOMMU mapping changes.
392 * @iommu: the IOMMUMemoryRegion
394 * @hwaddr: address to be translated within the memory region
396 * @flag: requested access permission
398 * @iommu_idx: IOMMU index for the translation
400 IOMMUTLBEntry (*translate
)(IOMMUMemoryRegion
*iommu
, hwaddr addr
,
401 IOMMUAccessFlags flag
, int iommu_idx
);
403 * @get_min_page_size:
405 * Returns minimum supported page size in bytes.
407 * If this method is not provided then the minimum is assumed to
408 * be TARGET_PAGE_SIZE.
410 * @iommu: the IOMMUMemoryRegion
412 uint64_t (*get_min_page_size
)(IOMMUMemoryRegion
*iommu
);
414 * @notify_flag_changed:
416 * Called when IOMMU Notifier flag changes (ie when the set of
417 * events which IOMMU users are requesting notification for changes).
418 * Optional method -- need not be provided if the IOMMU does not
419 * need to know exactly which events must be notified.
421 * @iommu: the IOMMUMemoryRegion
423 * @old_flags: events which previously needed to be notified
425 * @new_flags: events which now need to be notified
427 * Returns 0 on success, or a negative errno; in particular
428 * returns -EINVAL if the new flag bitmap is not supported by the
429 * IOMMU memory region. In case of failure, the error object
432 int (*notify_flag_changed
)(IOMMUMemoryRegion
*iommu
,
433 IOMMUNotifierFlag old_flags
,
434 IOMMUNotifierFlag new_flags
,
439 * Called to handle memory_region_iommu_replay().
441 * The default implementation of memory_region_iommu_replay() is to
442 * call the IOMMU translate method for every page in the address space
443 * with flag == IOMMU_NONE and then call the notifier if translate
444 * returns a valid mapping. If this method is implemented then it
445 * overrides the default behaviour, and must provide the full semantics
446 * of memory_region_iommu_replay(), by calling @notifier for every
447 * translation present in the IOMMU.
449 * Optional method -- an IOMMU only needs to provide this method
450 * if the default is inefficient or produces undesirable side effects.
452 * Note: this is not related to record-and-replay functionality.
454 void (*replay
)(IOMMUMemoryRegion
*iommu
, IOMMUNotifier
*notifier
);
459 * Get IOMMU misc attributes. This is an optional method that
460 * can be used to allow users of the IOMMU to get implementation-specific
461 * information. The IOMMU implements this method to handle calls
462 * by IOMMU users to memory_region_iommu_get_attr() by filling in
463 * the arbitrary data pointer for any IOMMUMemoryRegionAttr values that
464 * the IOMMU supports. If the method is unimplemented then
465 * memory_region_iommu_get_attr() will always return -EINVAL.
467 * @iommu: the IOMMUMemoryRegion
469 * @attr: attribute being queried
471 * @data: memory to fill in with the attribute data
473 * Returns 0 on success, or a negative errno; in particular
474 * returns -EINVAL for unrecognized or unimplemented attribute types.
476 int (*get_attr
)(IOMMUMemoryRegion
*iommu
, enum IOMMUMemoryRegionAttr attr
,
482 * Return the IOMMU index to use for a given set of transaction attributes.
484 * Optional method: if an IOMMU only supports a single IOMMU index then
485 * the default implementation of memory_region_iommu_attrs_to_index()
488 * The indexes supported by an IOMMU must be contiguous, starting at 0.
490 * @iommu: the IOMMUMemoryRegion
491 * @attrs: memory transaction attributes
493 int (*attrs_to_index
)(IOMMUMemoryRegion
*iommu
, MemTxAttrs attrs
);
498 * Return the number of IOMMU indexes this IOMMU supports.
500 * Optional method: if this method is not provided, then
501 * memory_region_iommu_num_indexes() will return 1, indicating that
502 * only a single IOMMU index is supported.
504 * @iommu: the IOMMUMemoryRegion
506 int (*num_indexes
)(IOMMUMemoryRegion
*iommu
);
509 * @iommu_set_page_size_mask:
511 * Restrict the page size mask that can be supported with a given IOMMU
512 * memory region. Used for example to propagate host physical IOMMU page
513 * size mask limitations to the virtual IOMMU.
515 * Optional method: if this method is not provided, then the default global
518 * @iommu: the IOMMUMemoryRegion
520 * @page_size_mask: a bitmask of supported page sizes. At least one bit,
521 * representing the smallest page size, must be set. Additional set bits
522 * represent supported block sizes. For example a host physical IOMMU that
523 * uses page tables with a page size of 4kB, and supports 2MB and 4GB
524 * blocks, will set mask 0x40201000. A granule of 4kB with indiscriminate
525 * block sizes is specified with mask 0xfffffffffffff000.
527 * Returns 0 on success, or a negative error. In case of failure, the error
528 * object must be created.
530 int (*iommu_set_page_size_mask
)(IOMMUMemoryRegion
*iommu
,
531 uint64_t page_size_mask
,
535 typedef struct RamDiscardListener RamDiscardListener
;
536 typedef int (*NotifyRamPopulate
)(RamDiscardListener
*rdl
,
537 MemoryRegionSection
*section
);
538 typedef void (*NotifyRamDiscard
)(RamDiscardListener
*rdl
,
539 MemoryRegionSection
*section
);
541 struct RamDiscardListener
{
545 * Notification that previously discarded memory is about to get populated.
546 * Listeners are able to object. If any listener objects, already
547 * successfully notified listeners are notified about a discard again.
549 * @rdl: the #RamDiscardListener getting notified
550 * @section: the #MemoryRegionSection to get populated. The section
551 * is aligned within the memory region to the minimum granularity
552 * unless it would exceed the registered section.
554 * Returns 0 on success. If the notification is rejected by the listener,
555 * an error is returned.
557 NotifyRamPopulate notify_populate
;
562 * Notification that previously populated memory was discarded successfully
563 * and listeners should drop all references to such memory and prevent
564 * new population (e.g., unmap).
566 * @rdl: the #RamDiscardListener getting notified
567 * @section: the #MemoryRegionSection to get populated. The section
568 * is aligned within the memory region to the minimum granularity
569 * unless it would exceed the registered section.
571 NotifyRamDiscard notify_discard
;
574 * @double_discard_supported:
576 * The listener suppors getting @notify_discard notifications that span
577 * already discarded parts.
579 bool double_discard_supported
;
581 MemoryRegionSection
*section
;
582 QLIST_ENTRY(RamDiscardListener
) next
;
585 static inline void ram_discard_listener_init(RamDiscardListener
*rdl
,
586 NotifyRamPopulate populate_fn
,
587 NotifyRamDiscard discard_fn
,
588 bool double_discard_supported
)
590 rdl
->notify_populate
= populate_fn
;
591 rdl
->notify_discard
= discard_fn
;
592 rdl
->double_discard_supported
= double_discard_supported
;
595 typedef int (*ReplayRamPopulate
)(MemoryRegionSection
*section
, void *opaque
);
596 typedef void (*ReplayRamDiscard
)(MemoryRegionSection
*section
, void *opaque
);
599 * RamDiscardManagerClass:
601 * A #RamDiscardManager coordinates which parts of specific RAM #MemoryRegion
602 * regions are currently populated to be used/accessed by the VM, notifying
603 * after parts were discarded (freeing up memory) and before parts will be
604 * populated (consuming memory), to be used/accessed by the VM.
606 * A #RamDiscardManager can only be set for a RAM #MemoryRegion while the
607 * #MemoryRegion isn't mapped into an address space yet (either directly
608 * or via an alias); it cannot change while the #MemoryRegion is
609 * mapped into an address space.
611 * The #RamDiscardManager is intended to be used by technologies that are
612 * incompatible with discarding of RAM (e.g., VFIO, which may pin all
613 * memory inside a #MemoryRegion), and require proper coordination to only
614 * map the currently populated parts, to hinder parts that are expected to
615 * remain discarded from silently getting populated and consuming memory.
616 * Technologies that support discarding of RAM don't have to bother and can
617 * simply map the whole #MemoryRegion.
619 * An example #RamDiscardManager is virtio-mem, which logically (un)plugs
620 * memory within an assigned RAM #MemoryRegion, coordinated with the VM.
621 * Logically unplugging memory consists of discarding RAM. The VM agreed to not
622 * access unplugged (discarded) memory - especially via DMA. virtio-mem will
623 * properly coordinate with listeners before memory is plugged (populated),
624 * and after memory is unplugged (discarded).
626 * Listeners are called in multiples of the minimum granularity (unless it
627 * would exceed the registered range) and changes are aligned to the minimum
628 * granularity within the #MemoryRegion. Listeners have to prepare for memory
629 * becoming discarded in a different granularity than it was populated and the
632 struct RamDiscardManagerClass
{
634 InterfaceClass parent_class
;
639 * @get_min_granularity:
641 * Get the minimum granularity in which listeners will get notified
642 * about changes within the #MemoryRegion via the #RamDiscardManager.
644 * @rdm: the #RamDiscardManager
645 * @mr: the #MemoryRegion
647 * Returns the minimum granularity.
649 uint64_t (*get_min_granularity
)(const RamDiscardManager
*rdm
,
650 const MemoryRegion
*mr
);
655 * Check whether the given #MemoryRegionSection is completely populated
656 * (i.e., no parts are currently discarded) via the #RamDiscardManager.
657 * There are no alignment requirements.
659 * @rdm: the #RamDiscardManager
660 * @section: the #MemoryRegionSection
662 * Returns whether the given range is completely populated.
664 bool (*is_populated
)(const RamDiscardManager
*rdm
,
665 const MemoryRegionSection
*section
);
670 * Call the #ReplayRamPopulate callback for all populated parts within the
671 * #MemoryRegionSection via the #RamDiscardManager.
673 * In case any call fails, no further calls are made.
675 * @rdm: the #RamDiscardManager
676 * @section: the #MemoryRegionSection
677 * @replay_fn: the #ReplayRamPopulate callback
678 * @opaque: pointer to forward to the callback
680 * Returns 0 on success, or a negative error if any notification failed.
682 int (*replay_populated
)(const RamDiscardManager
*rdm
,
683 MemoryRegionSection
*section
,
684 ReplayRamPopulate replay_fn
, void *opaque
);
689 * Call the #ReplayRamDiscard callback for all discarded parts within the
690 * #MemoryRegionSection via the #RamDiscardManager.
692 * @rdm: the #RamDiscardManager
693 * @section: the #MemoryRegionSection
694 * @replay_fn: the #ReplayRamDiscard callback
695 * @opaque: pointer to forward to the callback
697 void (*replay_discarded
)(const RamDiscardManager
*rdm
,
698 MemoryRegionSection
*section
,
699 ReplayRamDiscard replay_fn
, void *opaque
);
702 * @register_listener:
704 * Register a #RamDiscardListener for the given #MemoryRegionSection and
705 * immediately notify the #RamDiscardListener about all populated parts
706 * within the #MemoryRegionSection via the #RamDiscardManager.
708 * In case any notification fails, no further notifications are triggered
709 * and an error is logged.
711 * @rdm: the #RamDiscardManager
712 * @rdl: the #RamDiscardListener
713 * @section: the #MemoryRegionSection
715 void (*register_listener
)(RamDiscardManager
*rdm
,
716 RamDiscardListener
*rdl
,
717 MemoryRegionSection
*section
);
720 * @unregister_listener:
722 * Unregister a previously registered #RamDiscardListener via the
723 * #RamDiscardManager after notifying the #RamDiscardListener about all
724 * populated parts becoming unpopulated within the registered
725 * #MemoryRegionSection.
727 * @rdm: the #RamDiscardManager
728 * @rdl: the #RamDiscardListener
730 void (*unregister_listener
)(RamDiscardManager
*rdm
,
731 RamDiscardListener
*rdl
);
734 uint64_t ram_discard_manager_get_min_granularity(const RamDiscardManager
*rdm
,
735 const MemoryRegion
*mr
);
737 bool ram_discard_manager_is_populated(const RamDiscardManager
*rdm
,
738 const MemoryRegionSection
*section
);
740 int ram_discard_manager_replay_populated(const RamDiscardManager
*rdm
,
741 MemoryRegionSection
*section
,
742 ReplayRamPopulate replay_fn
,
745 void ram_discard_manager_replay_discarded(const RamDiscardManager
*rdm
,
746 MemoryRegionSection
*section
,
747 ReplayRamDiscard replay_fn
,
750 void ram_discard_manager_register_listener(RamDiscardManager
*rdm
,
751 RamDiscardListener
*rdl
,
752 MemoryRegionSection
*section
);
754 void ram_discard_manager_unregister_listener(RamDiscardManager
*rdm
,
755 RamDiscardListener
*rdl
);
758 * memory_get_xlat_addr: Extract addresses from a TLB entry
760 * @iotlb: pointer to an #IOMMUTLBEntry
761 * @vaddr: virtual address
762 * @ram_addr: RAM address
763 * @read_only: indicates if writes are allowed
764 * @mr_has_discard_manager: indicates memory is controlled by a
766 * @errp: pointer to Error*, to store an error if it happens.
768 * Return: true on success, else false setting @errp with error.
770 bool memory_get_xlat_addr(IOMMUTLBEntry
*iotlb
, void **vaddr
,
771 ram_addr_t
*ram_addr
, bool *read_only
,
772 bool *mr_has_discard_manager
, Error
**errp
);
774 typedef struct CoalescedMemoryRange CoalescedMemoryRange
;
775 typedef struct MemoryRegionIoeventfd MemoryRegionIoeventfd
;
779 * A struct representing a memory region.
781 struct MemoryRegion
{
786 /* The following fields should fit in a cache line */
790 bool readonly
; /* For RAM regions */
793 bool flush_coalesced_mmio
;
795 uint8_t dirty_log_mask
;
799 /* owner as TYPE_DEVICE. Used for re-entrancy checks in MR access hotpath */
802 const MemoryRegionOps
*ops
;
804 MemoryRegion
*container
;
805 int mapped_via_alias
; /* Mapped via an alias, container might be NULL */
808 void (*destructor
)(MemoryRegion
*mr
);
813 bool warning_printed
; /* For reservations */
814 uint8_t vga_logging_count
;
818 QTAILQ_HEAD(, MemoryRegion
) subregions
;
819 QTAILQ_ENTRY(MemoryRegion
) subregions_link
;
820 QTAILQ_HEAD(, CoalescedMemoryRange
) coalesced
;
822 unsigned ioeventfd_nb
;
823 MemoryRegionIoeventfd
*ioeventfds
;
824 RamDiscardManager
*rdm
; /* Only for RAM */
826 /* For devices designed to perform re-entrant IO into their own IO MRs */
827 bool disable_reentrancy_guard
;
830 struct IOMMUMemoryRegion
{
831 MemoryRegion parent_obj
;
833 QLIST_HEAD(, IOMMUNotifier
) iommu_notify
;
834 IOMMUNotifierFlag iommu_notify_flags
;
837 #define IOMMU_NOTIFIER_FOREACH(n, mr) \
838 QLIST_FOREACH((n), &(mr)->iommu_notify, node)
840 #define MEMORY_LISTENER_PRIORITY_MIN 0
841 #define MEMORY_LISTENER_PRIORITY_ACCEL 10
842 #define MEMORY_LISTENER_PRIORITY_DEV_BACKEND 10
845 * struct MemoryListener: callbacks structure for updates to the physical memory map
847 * Allows a component to adjust to changes in the guest-visible memory map.
848 * Use with memory_listener_register() and memory_listener_unregister().
850 struct MemoryListener
{
854 * Called at the beginning of an address space update transaction.
855 * Followed by calls to #MemoryListener.region_add(),
856 * #MemoryListener.region_del(), #MemoryListener.region_nop(),
857 * #MemoryListener.log_start() and #MemoryListener.log_stop() in
858 * increasing address order.
860 * @listener: The #MemoryListener.
862 void (*begin
)(MemoryListener
*listener
);
867 * Called at the end of an address space update transaction,
868 * after the last call to #MemoryListener.region_add(),
869 * #MemoryListener.region_del() or #MemoryListener.region_nop(),
870 * #MemoryListener.log_start() and #MemoryListener.log_stop().
872 * @listener: The #MemoryListener.
874 void (*commit
)(MemoryListener
*listener
);
879 * Called during an address space update transaction,
880 * for a section of the address space that is new in this address space
881 * space since the last transaction.
883 * @listener: The #MemoryListener.
884 * @section: The new #MemoryRegionSection.
886 void (*region_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
891 * Called during an address space update transaction,
892 * for a section of the address space that has disappeared in the address
893 * space since the last transaction.
895 * @listener: The #MemoryListener.
896 * @section: The old #MemoryRegionSection.
898 void (*region_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
903 * Called during an address space update transaction,
904 * for a section of the address space that is in the same place in the address
905 * space as in the last transaction.
907 * @listener: The #MemoryListener.
908 * @section: The #MemoryRegionSection.
910 void (*region_nop
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
915 * Called during an address space update transaction, after
916 * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
917 * #MemoryListener.region_nop(), if dirty memory logging clients have
918 * become active since the last transaction.
920 * @listener: The #MemoryListener.
921 * @section: The #MemoryRegionSection.
922 * @old: A bitmap of dirty memory logging clients that were active in
923 * the previous transaction.
924 * @new: A bitmap of dirty memory logging clients that are active in
925 * the current transaction.
927 void (*log_start
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
928 int old_val
, int new_val
);
933 * Called during an address space update transaction, after
934 * one of #MemoryListener.region_add(), #MemoryListener.region_del() or
935 * #MemoryListener.region_nop() and possibly after
936 * #MemoryListener.log_start(), if dirty memory logging clients have
937 * become inactive since the last transaction.
939 * @listener: The #MemoryListener.
940 * @section: The #MemoryRegionSection.
941 * @old: A bitmap of dirty memory logging clients that were active in
942 * the previous transaction.
943 * @new: A bitmap of dirty memory logging clients that are active in
944 * the current transaction.
946 void (*log_stop
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
947 int old_val
, int new_val
);
952 * Called by memory_region_snapshot_and_clear_dirty() and
953 * memory_global_dirty_log_sync(), before accessing QEMU's "official"
954 * copy of the dirty memory bitmap for a #MemoryRegionSection.
956 * @listener: The #MemoryListener.
957 * @section: The #MemoryRegionSection.
959 void (*log_sync
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
964 * This is the global version of @log_sync when the listener does
965 * not have a way to synchronize the log with finer granularity.
966 * When the listener registers with @log_sync_global defined, then
967 * its @log_sync must be NULL. Vice versa.
969 * @listener: The #MemoryListener.
970 * @last_stage: The last stage to synchronize the log during migration.
971 * The caller should guarantee that the synchronization with true for
972 * @last_stage is triggered for once after all VCPUs have been stopped.
974 void (*log_sync_global
)(MemoryListener
*listener
, bool last_stage
);
979 * Called before reading the dirty memory bitmap for a
980 * #MemoryRegionSection.
982 * @listener: The #MemoryListener.
983 * @section: The #MemoryRegionSection.
985 void (*log_clear
)(MemoryListener
*listener
, MemoryRegionSection
*section
);
990 * Called by memory_global_dirty_log_start(), which
991 * enables the %DIRTY_LOG_MIGRATION client on all memory regions in
992 * the address space. #MemoryListener.log_global_start() is also
993 * called when a #MemoryListener is added, if global dirty logging is
994 * active at that time.
996 * @listener: The #MemoryListener.
997 * @errp: pointer to Error*, to store an error if it happens.
999 * Return: true on success, else false setting @errp with error.
1001 bool (*log_global_start
)(MemoryListener
*listener
, Error
**errp
);
1006 * Called by memory_global_dirty_log_stop(), which
1007 * disables the %DIRTY_LOG_MIGRATION client on all memory regions in
1008 * the address space.
1010 * @listener: The #MemoryListener.
1012 void (*log_global_stop
)(MemoryListener
*listener
);
1015 * @log_global_after_sync:
1017 * Called after reading the dirty memory bitmap
1018 * for any #MemoryRegionSection.
1020 * @listener: The #MemoryListener.
1022 void (*log_global_after_sync
)(MemoryListener
*listener
);
1027 * Called during an address space update transaction,
1028 * for a section of the address space that has had a new ioeventfd
1029 * registration since the last transaction.
1031 * @listener: The #MemoryListener.
1032 * @section: The new #MemoryRegionSection.
1033 * @match_data: The @match_data parameter for the new ioeventfd.
1034 * @data: The @data parameter for the new ioeventfd.
1035 * @e: The #EventNotifier parameter for the new ioeventfd.
1037 void (*eventfd_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1038 bool match_data
, uint64_t data
, EventNotifier
*e
);
1043 * Called during an address space update transaction,
1044 * for a section of the address space that has dropped an ioeventfd
1045 * registration since the last transaction.
1047 * @listener: The #MemoryListener.
1048 * @section: The new #MemoryRegionSection.
1049 * @match_data: The @match_data parameter for the dropped ioeventfd.
1050 * @data: The @data parameter for the dropped ioeventfd.
1051 * @e: The #EventNotifier parameter for the dropped ioeventfd.
1053 void (*eventfd_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1054 bool match_data
, uint64_t data
, EventNotifier
*e
);
1057 * @coalesced_io_add:
1059 * Called during an address space update transaction,
1060 * for a section of the address space that has had a new coalesced
1061 * MMIO range registration since the last transaction.
1063 * @listener: The #MemoryListener.
1064 * @section: The new #MemoryRegionSection.
1065 * @addr: The starting address for the coalesced MMIO range.
1066 * @len: The length of the coalesced MMIO range.
1068 void (*coalesced_io_add
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1069 hwaddr addr
, hwaddr len
);
1072 * @coalesced_io_del:
1074 * Called during an address space update transaction,
1075 * for a section of the address space that has dropped a coalesced
1076 * MMIO range since the last transaction.
1078 * @listener: The #MemoryListener.
1079 * @section: The new #MemoryRegionSection.
1080 * @addr: The starting address for the coalesced MMIO range.
1081 * @len: The length of the coalesced MMIO range.
1083 void (*coalesced_io_del
)(MemoryListener
*listener
, MemoryRegionSection
*section
,
1084 hwaddr addr
, hwaddr len
);
1088 * Govern the order in which memory listeners are invoked. Lower priorities
1089 * are invoked earlier for "add" or "start" callbacks, and later for "delete"
1090 * or "stop" callbacks.
1097 * Name of the listener. It can be used in contexts where we'd like to
1098 * identify one memory listener with the rest.
1103 AddressSpace
*address_space
;
1104 QTAILQ_ENTRY(MemoryListener
) link
;
1105 QTAILQ_ENTRY(MemoryListener
) link_as
;
1108 typedef struct AddressSpaceMapClient
{
1110 QLIST_ENTRY(AddressSpaceMapClient
) link
;
1111 } AddressSpaceMapClient
;
1122 * struct AddressSpace: describes a mapping of addresses to #MemoryRegion objects
1124 struct AddressSpace
{
1126 struct rcu_head rcu
;
1130 /* Accessed via RCU. */
1131 struct FlatView
*current_map
;
1134 int ioeventfd_notifiers
;
1135 struct MemoryRegionIoeventfd
*ioeventfds
;
1136 QTAILQ_HEAD(, MemoryListener
) listeners
;
1137 QTAILQ_ENTRY(AddressSpace
) address_spaces_link
;
1139 /* Bounce buffer to use for this address space. */
1140 BounceBuffer bounce
;
1141 /* List of callbacks to invoke when buffers free up */
1142 QemuMutex map_client_list_lock
;
1143 QLIST_HEAD(, AddressSpaceMapClient
) map_client_list
;
1146 typedef struct AddressSpaceDispatch AddressSpaceDispatch
;
1147 typedef struct FlatRange FlatRange
;
1149 /* Flattened global view of current active memory hierarchy. Kept in sorted
1153 struct rcu_head rcu
;
1157 unsigned nr_allocated
;
1158 struct AddressSpaceDispatch
*dispatch
;
1162 static inline FlatView
*address_space_to_flatview(AddressSpace
*as
)
1164 return qatomic_rcu_read(&as
->current_map
);
1168 * typedef flatview_cb: callback for flatview_for_each_range()
1170 * @start: start address of the range within the FlatView
1171 * @len: length of the range in bytes
1172 * @mr: MemoryRegion covering this range
1173 * @offset_in_region: offset of the first byte of the range within @mr
1174 * @opaque: data pointer passed to flatview_for_each_range()
1176 * Returns: true to stop the iteration, false to keep going.
1178 typedef bool (*flatview_cb
)(Int128 start
,
1180 const MemoryRegion
*mr
,
1181 hwaddr offset_in_region
,
1185 * flatview_for_each_range: Iterate through a FlatView
1186 * @fv: the FlatView to iterate through
1187 * @cb: function to call for each range
1188 * @opaque: opaque data pointer to pass to @cb
1190 * A FlatView is made up of a list of non-overlapping ranges, each of
1191 * which is a slice of a MemoryRegion. This function iterates through
1192 * each range in @fv, calling @cb. The callback function can terminate
1193 * iteration early by returning 'true'.
1195 void flatview_for_each_range(FlatView
*fv
, flatview_cb cb
, void *opaque
);
1197 static inline bool MemoryRegionSection_eq(MemoryRegionSection
*a
,
1198 MemoryRegionSection
*b
)
1200 return a
->mr
== b
->mr
&&
1202 a
->offset_within_region
== b
->offset_within_region
&&
1203 a
->offset_within_address_space
== b
->offset_within_address_space
&&
1204 int128_eq(a
->size
, b
->size
) &&
1205 a
->readonly
== b
->readonly
&&
1206 a
->nonvolatile
== b
->nonvolatile
;
1210 * memory_region_section_new_copy: Copy a memory region section
1212 * Allocate memory for a new copy, copy the memory region section, and
1213 * properly take a reference on all relevant members.
1215 * @s: the #MemoryRegionSection to copy
1217 MemoryRegionSection
*memory_region_section_new_copy(MemoryRegionSection
*s
);
1220 * memory_region_section_new_copy: Free a copied memory region section
1222 * Free a copy of a memory section created via memory_region_section_new_copy().
1223 * properly dropping references on all relevant members.
1225 * @s: the #MemoryRegionSection to copy
1227 void memory_region_section_free_copy(MemoryRegionSection
*s
);
1230 * memory_region_init: Initialize a memory region
1232 * The region typically acts as a container for other memory regions. Use
1233 * memory_region_add_subregion() to add subregions.
1235 * @mr: the #MemoryRegion to be initialized
1236 * @owner: the object that tracks the region's reference count
1237 * @name: used for debugging; not visible to the user or ABI
1238 * @size: size of the region; any subregions beyond this size will be clipped
1240 void memory_region_init(MemoryRegion
*mr
,
1246 * memory_region_ref: Add 1 to a memory region's reference count
1248 * Whenever memory regions are accessed outside the BQL, they need to be
1249 * preserved against hot-unplug. MemoryRegions actually do not have their
1250 * own reference count; they piggyback on a QOM object, their "owner".
1251 * This function adds a reference to the owner.
1253 * All MemoryRegions must have an owner if they can disappear, even if the
1254 * device they belong to operates exclusively under the BQL. This is because
1255 * the region could be returned at any time by memory_region_find, and this
1256 * is usually under guest control.
1258 * @mr: the #MemoryRegion
1260 void memory_region_ref(MemoryRegion
*mr
);
1263 * memory_region_unref: Remove 1 to a memory region's reference count
1265 * Whenever memory regions are accessed outside the BQL, they need to be
1266 * preserved against hot-unplug. MemoryRegions actually do not have their
1267 * own reference count; they piggyback on a QOM object, their "owner".
1268 * This function removes a reference to the owner and possibly destroys it.
1270 * @mr: the #MemoryRegion
1272 void memory_region_unref(MemoryRegion
*mr
);
1275 * memory_region_init_io: Initialize an I/O memory region.
1277 * Accesses into the region will cause the callbacks in @ops to be called.
1278 * if @size is nonzero, subregions will be clipped to @size.
1280 * @mr: the #MemoryRegion to be initialized.
1281 * @owner: the object that tracks the region's reference count
1282 * @ops: a structure containing read and write callbacks to be used when
1283 * I/O is performed on the region.
1284 * @opaque: passed to the read and write callbacks of the @ops structure.
1285 * @name: used for debugging; not visible to the user or ABI
1286 * @size: size of the region.
1288 void memory_region_init_io(MemoryRegion
*mr
,
1290 const MemoryRegionOps
*ops
,
1296 * memory_region_init_ram_nomigrate: Initialize RAM memory region. Accesses
1297 * into the region will modify memory
1300 * @mr: the #MemoryRegion to be initialized.
1301 * @owner: the object that tracks the region's reference count
1302 * @name: Region name, becomes part of RAMBlock name used in migration stream
1303 * must be unique within any device
1304 * @size: size of the region.
1305 * @errp: pointer to Error*, to store an error if it happens.
1307 * Note that this function does not do anything to cause the data in the
1308 * RAM memory region to be migrated; that is the responsibility of the caller.
1310 * Return: true on success, else false setting @errp with error.
1312 bool memory_region_init_ram_nomigrate(MemoryRegion
*mr
,
1319 * memory_region_init_ram_flags_nomigrate: Initialize RAM memory region.
1320 * Accesses into the region will
1321 * modify memory directly.
1323 * @mr: the #MemoryRegion to be initialized.
1324 * @owner: the object that tracks the region's reference count
1325 * @name: Region name, becomes part of RAMBlock name used in migration stream
1326 * must be unique within any device
1327 * @size: size of the region.
1328 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_NORESERVE,
1330 * @errp: pointer to Error*, to store an error if it happens.
1332 * Note that this function does not do anything to cause the data in the
1333 * RAM memory region to be migrated; that is the responsibility of the caller.
1335 * Return: true on success, else false setting @errp with error.
1337 bool memory_region_init_ram_flags_nomigrate(MemoryRegion
*mr
,
1345 * memory_region_init_resizeable_ram: Initialize memory region with resizable
1346 * RAM. Accesses into the region will
1347 * modify memory directly. Only an initial
1348 * portion of this RAM is actually used.
1349 * Changing the size while migrating
1350 * can result in the migration being
1353 * @mr: the #MemoryRegion to be initialized.
1354 * @owner: the object that tracks the region's reference count
1355 * @name: Region name, becomes part of RAMBlock name used in migration stream
1356 * must be unique within any device
1357 * @size: used size of the region.
1358 * @max_size: max size of the region.
1359 * @resized: callback to notify owner about used size change.
1360 * @errp: pointer to Error*, to store an error if it happens.
1362 * Note that this function does not do anything to cause the data in the
1363 * RAM memory region to be migrated; that is the responsibility of the caller.
1365 * Return: true on success, else false setting @errp with error.
1367 bool memory_region_init_resizeable_ram(MemoryRegion
*mr
,
1372 void (*resized
)(const char*,
1379 * memory_region_init_ram_from_file: Initialize RAM memory region with a
1382 * @mr: the #MemoryRegion to be initialized.
1383 * @owner: the object that tracks the region's reference count
1384 * @name: Region name, becomes part of RAMBlock name used in migration stream
1385 * must be unique within any device
1386 * @size: size of the region.
1387 * @align: alignment of the region base address; if 0, the default alignment
1388 * (getpagesize()) will be used.
1389 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
1390 * RAM_NORESERVE, RAM_PROTECTED, RAM_NAMED_FILE, RAM_READONLY,
1391 * RAM_READONLY_FD, RAM_GUEST_MEMFD
1392 * @path: the path in which to allocate the RAM.
1393 * @offset: offset within the file referenced by path
1394 * @errp: pointer to Error*, to store an error if it happens.
1396 * Note that this function does not do anything to cause the data in the
1397 * RAM memory region to be migrated; that is the responsibility of the caller.
1399 * Return: true on success, else false setting @errp with error.
1401 bool memory_region_init_ram_from_file(MemoryRegion
*mr
,
1412 * memory_region_init_ram_from_fd: Initialize RAM memory region with a
1415 * @mr: the #MemoryRegion to be initialized.
1416 * @owner: the object that tracks the region's reference count
1417 * @name: the name of the region.
1418 * @size: size of the region.
1419 * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
1420 * RAM_NORESERVE, RAM_PROTECTED, RAM_NAMED_FILE, RAM_READONLY,
1421 * RAM_READONLY_FD, RAM_GUEST_MEMFD
1422 * @fd: the fd to mmap.
1423 * @offset: offset within the file referenced by fd
1424 * @errp: pointer to Error*, to store an error if it happens.
1426 * Note that this function does not do anything to cause the data in the
1427 * RAM memory region to be migrated; that is the responsibility of the caller.
1429 * Return: true on success, else false setting @errp with error.
1431 bool memory_region_init_ram_from_fd(MemoryRegion
*mr
,
1442 * memory_region_init_ram_ptr: Initialize RAM memory region from a
1443 * user-provided pointer. Accesses into the
1444 * region will modify memory directly.
1446 * @mr: the #MemoryRegion to be initialized.
1447 * @owner: the object that tracks the region's reference count
1448 * @name: Region name, becomes part of RAMBlock name used in migration stream
1449 * must be unique within any device
1450 * @size: size of the region.
1451 * @ptr: memory to be mapped; must contain at least @size bytes.
1453 * Note that this function does not do anything to cause the data in the
1454 * RAM memory region to be migrated; that is the responsibility of the caller.
1456 void memory_region_init_ram_ptr(MemoryRegion
*mr
,
1463 * memory_region_init_ram_device_ptr: Initialize RAM device memory region from
1464 * a user-provided pointer.
1466 * A RAM device represents a mapping to a physical device, such as to a PCI
1467 * MMIO BAR of an vfio-pci assigned device. The memory region may be mapped
1468 * into the VM address space and access to the region will modify memory
1469 * directly. However, the memory region should not be included in a memory
1470 * dump (device may not be enabled/mapped at the time of the dump), and
1471 * operations incompatible with manipulating MMIO should be avoided. Replaces
1474 * @mr: the #MemoryRegion to be initialized.
1475 * @owner: the object that tracks the region's reference count
1476 * @name: the name of the region.
1477 * @size: size of the region.
1478 * @ptr: memory to be mapped; must contain at least @size bytes.
1480 * Note that this function does not do anything to cause the data in the
1481 * RAM memory region to be migrated; that is the responsibility of the caller.
1482 * (For RAM device memory regions, migrating the contents rarely makes sense.)
1484 void memory_region_init_ram_device_ptr(MemoryRegion
*mr
,
1491 * memory_region_init_alias: Initialize a memory region that aliases all or a
1492 * part of another memory region.
1494 * @mr: the #MemoryRegion to be initialized.
1495 * @owner: the object that tracks the region's reference count
1496 * @name: used for debugging; not visible to the user or ABI
1497 * @orig: the region to be referenced; @mr will be equivalent to
1498 * @orig between @offset and @offset + @size - 1.
1499 * @offset: start of the section in @orig to be referenced.
1500 * @size: size of the region.
1502 void memory_region_init_alias(MemoryRegion
*mr
,
1510 * memory_region_init_rom_nomigrate: Initialize a ROM memory region.
1512 * This has the same effect as calling memory_region_init_ram_nomigrate()
1513 * and then marking the resulting region read-only with
1514 * memory_region_set_readonly().
1516 * Note that this function does not do anything to cause the data in the
1517 * RAM side of the memory region to be migrated; that is the responsibility
1520 * @mr: the #MemoryRegion to be initialized.
1521 * @owner: the object that tracks the region's reference count
1522 * @name: Region name, becomes part of RAMBlock name used in migration stream
1523 * must be unique within any device
1524 * @size: size of the region.
1525 * @errp: pointer to Error*, to store an error if it happens.
1527 * Return: true on success, else false setting @errp with error.
1529 bool memory_region_init_rom_nomigrate(MemoryRegion
*mr
,
1536 * memory_region_init_rom_device_nomigrate: Initialize a ROM memory region.
1537 * Writes are handled via callbacks.
1539 * Note that this function does not do anything to cause the data in the
1540 * RAM side of the memory region to be migrated; that is the responsibility
1543 * @mr: the #MemoryRegion to be initialized.
1544 * @owner: the object that tracks the region's reference count
1545 * @ops: callbacks for write access handling (must not be NULL).
1546 * @opaque: passed to the read and write callbacks of the @ops structure.
1547 * @name: Region name, becomes part of RAMBlock name used in migration stream
1548 * must be unique within any device
1549 * @size: size of the region.
1550 * @errp: pointer to Error*, to store an error if it happens.
1552 * Return: true on success, else false setting @errp with error.
1554 bool memory_region_init_rom_device_nomigrate(MemoryRegion
*mr
,
1556 const MemoryRegionOps
*ops
,
1563 * memory_region_init_iommu: Initialize a memory region of a custom type
1564 * that translates addresses
1566 * An IOMMU region translates addresses and forwards accesses to a target
1569 * The IOMMU implementation must define a subclass of TYPE_IOMMU_MEMORY_REGION.
1570 * @_iommu_mr should be a pointer to enough memory for an instance of
1571 * that subclass, @instance_size is the size of that subclass, and
1572 * @mrtypename is its name. This function will initialize @_iommu_mr as an
1573 * instance of the subclass, and its methods will then be called to handle
1574 * accesses to the memory region. See the documentation of
1575 * #IOMMUMemoryRegionClass for further details.
1577 * @_iommu_mr: the #IOMMUMemoryRegion to be initialized
1578 * @instance_size: the IOMMUMemoryRegion subclass instance size
1579 * @mrtypename: the type name of the #IOMMUMemoryRegion
1580 * @owner: the object that tracks the region's reference count
1581 * @name: used for debugging; not visible to the user or ABI
1582 * @size: size of the region.
1584 void memory_region_init_iommu(void *_iommu_mr
,
1585 size_t instance_size
,
1586 const char *mrtypename
,
1592 * memory_region_init_ram - Initialize RAM memory region. Accesses into the
1593 * region will modify memory directly.
1595 * @mr: the #MemoryRegion to be initialized
1596 * @owner: the object that tracks the region's reference count (must be
1597 * TYPE_DEVICE or a subclass of TYPE_DEVICE, or NULL)
1598 * @name: name of the memory region
1599 * @size: size of the region in bytes
1600 * @errp: pointer to Error*, to store an error if it happens.
1602 * This function allocates RAM for a board model or device, and
1603 * arranges for it to be migrated (by calling vmstate_register_ram()
1604 * if @owner is a DeviceState, or vmstate_register_ram_global() if
1607 * TODO: Currently we restrict @owner to being either NULL (for
1608 * global RAM regions with no owner) or devices, so that we can
1609 * give the RAM block a unique name for migration purposes.
1610 * We should lift this restriction and allow arbitrary Objects.
1611 * If you pass a non-NULL non-device @owner then we will assert.
1613 * Return: true on success, else false setting @errp with error.
1615 bool memory_region_init_ram(MemoryRegion
*mr
,
1621 bool memory_region_init_ram_guest_memfd(MemoryRegion
*mr
,
1628 * memory_region_init_rom: Initialize a ROM memory region.
1630 * This has the same effect as calling memory_region_init_ram()
1631 * and then marking the resulting region read-only with
1632 * memory_region_set_readonly(). This includes arranging for the
1633 * contents to be migrated.
1635 * TODO: Currently we restrict @owner to being either NULL (for
1636 * global RAM regions with no owner) or devices, so that we can
1637 * give the RAM block a unique name for migration purposes.
1638 * We should lift this restriction and allow arbitrary Objects.
1639 * If you pass a non-NULL non-device @owner then we will assert.
1641 * @mr: the #MemoryRegion to be initialized.
1642 * @owner: the object that tracks the region's reference count
1643 * @name: Region name, becomes part of RAMBlock name used in migration stream
1644 * must be unique within any device
1645 * @size: size of the region.
1646 * @errp: pointer to Error*, to store an error if it happens.
1648 * Return: true on success, else false setting @errp with error.
1650 bool memory_region_init_rom(MemoryRegion
*mr
,
1657 * memory_region_init_rom_device: Initialize a ROM memory region.
1658 * Writes are handled via callbacks.
1660 * This function initializes a memory region backed by RAM for reads
1661 * and callbacks for writes, and arranges for the RAM backing to
1662 * be migrated (by calling vmstate_register_ram()
1663 * if @owner is a DeviceState, or vmstate_register_ram_global() if
1666 * TODO: Currently we restrict @owner to being either NULL (for
1667 * global RAM regions with no owner) or devices, so that we can
1668 * give the RAM block a unique name for migration purposes.
1669 * We should lift this restriction and allow arbitrary Objects.
1670 * If you pass a non-NULL non-device @owner then we will assert.
1672 * @mr: the #MemoryRegion to be initialized.
1673 * @owner: the object that tracks the region's reference count
1674 * @ops: callbacks for write access handling (must not be NULL).
1675 * @opaque: passed to the read and write callbacks of the @ops structure.
1676 * @name: Region name, becomes part of RAMBlock name used in migration stream
1677 * must be unique within any device
1678 * @size: size of the region.
1679 * @errp: pointer to Error*, to store an error if it happens.
1681 * Return: true on success, else false setting @errp with error.
1683 bool memory_region_init_rom_device(MemoryRegion
*mr
,
1685 const MemoryRegionOps
*ops
,
1693 * memory_region_owner: get a memory region's owner.
1695 * @mr: the memory region being queried.
1697 Object
*memory_region_owner(MemoryRegion
*mr
);
1700 * memory_region_size: get a memory region's size.
1702 * @mr: the memory region being queried.
1704 uint64_t memory_region_size(MemoryRegion
*mr
);
1707 * memory_region_is_ram: check whether a memory region is random access
1709 * Returns %true if a memory region is random access.
1711 * @mr: the memory region being queried
1713 static inline bool memory_region_is_ram(MemoryRegion
*mr
)
1719 * memory_region_is_ram_device: check whether a memory region is a ram device
1721 * Returns %true if a memory region is a device backed ram region
1723 * @mr: the memory region being queried
1725 bool memory_region_is_ram_device(MemoryRegion
*mr
);
1728 * memory_region_is_romd: check whether a memory region is in ROMD mode
1730 * Returns %true if a memory region is a ROM device and currently set to allow
1733 * @mr: the memory region being queried
1735 static inline bool memory_region_is_romd(MemoryRegion
*mr
)
1737 return mr
->rom_device
&& mr
->romd_mode
;
1741 * memory_region_is_protected: check whether a memory region is protected
1743 * Returns %true if a memory region is protected RAM and cannot be accessed
1744 * via standard mechanisms, e.g. DMA.
1746 * @mr: the memory region being queried
1748 bool memory_region_is_protected(MemoryRegion
*mr
);
1751 * memory_region_has_guest_memfd: check whether a memory region has guest_memfd
1754 * Returns %true if a memory region's ram_block has valid guest_memfd assigned.
1756 * @mr: the memory region being queried
1758 bool memory_region_has_guest_memfd(MemoryRegion
*mr
);
1761 * memory_region_get_iommu: check whether a memory region is an iommu
1763 * Returns pointer to IOMMUMemoryRegion if a memory region is an iommu,
1766 * @mr: the memory region being queried
1768 static inline IOMMUMemoryRegion
*memory_region_get_iommu(MemoryRegion
*mr
)
1771 return memory_region_get_iommu(mr
->alias
);
1774 return (IOMMUMemoryRegion
*) mr
;
1780 * memory_region_get_iommu_class_nocheck: returns iommu memory region class
1781 * if an iommu or NULL if not
1783 * Returns pointer to IOMMUMemoryRegionClass if a memory region is an iommu,
1784 * otherwise NULL. This is fast path avoiding QOM checking, use with caution.
1786 * @iommu_mr: the memory region being queried
1788 static inline IOMMUMemoryRegionClass
*memory_region_get_iommu_class_nocheck(
1789 IOMMUMemoryRegion
*iommu_mr
)
1791 return (IOMMUMemoryRegionClass
*) (((Object
*)iommu_mr
)->class);
1794 #define memory_region_is_iommu(mr) (memory_region_get_iommu(mr) != NULL)
1797 * memory_region_iommu_get_min_page_size: get minimum supported page size
1800 * Returns minimum supported page size for an iommu.
1802 * @iommu_mr: the memory region being queried
1804 uint64_t memory_region_iommu_get_min_page_size(IOMMUMemoryRegion
*iommu_mr
);
1807 * memory_region_notify_iommu: notify a change in an IOMMU translation entry.
1809 * Note: for any IOMMU implementation, an in-place mapping change
1810 * should be notified with an UNMAP followed by a MAP.
1812 * @iommu_mr: the memory region that was changed
1813 * @iommu_idx: the IOMMU index for the translation table which has changed
1814 * @event: TLB event with the new entry in the IOMMU translation table.
1815 * The entry replaces all old entries for the same virtual I/O address
1818 void memory_region_notify_iommu(IOMMUMemoryRegion
*iommu_mr
,
1820 const IOMMUTLBEvent event
);
1823 * memory_region_notify_iommu_one: notify a change in an IOMMU translation
1824 * entry to a single notifier
1826 * This works just like memory_region_notify_iommu(), but it only
1827 * notifies a specific notifier, not all of them.
1829 * @notifier: the notifier to be notified
1830 * @event: TLB event with the new entry in the IOMMU translation table.
1831 * The entry replaces all old entries for the same virtual I/O address
1834 void memory_region_notify_iommu_one(IOMMUNotifier
*notifier
,
1835 const IOMMUTLBEvent
*event
);
1838 * memory_region_unmap_iommu_notifier_range: notify a unmap for an IOMMU
1839 * translation that covers the
1840 * range of a notifier
1842 * @notifier: the notifier to be notified
1844 void memory_region_unmap_iommu_notifier_range(IOMMUNotifier
*notifier
);
1848 * memory_region_register_iommu_notifier: register a notifier for changes to
1849 * IOMMU translation entries.
1851 * Returns 0 on success, or a negative errno otherwise. In particular,
1852 * -EINVAL indicates that at least one of the attributes of the notifier
1853 * is not supported (flag/range) by the IOMMU memory region. In case of error
1854 * the error object must be created.
1856 * @mr: the memory region to observe
1857 * @n: the IOMMUNotifier to be added; the notify callback receives a
1858 * pointer to an #IOMMUTLBEntry as the opaque value; the pointer
1859 * ceases to be valid on exit from the notifier.
1860 * @errp: pointer to Error*, to store an error if it happens.
1862 int memory_region_register_iommu_notifier(MemoryRegion
*mr
,
1863 IOMMUNotifier
*n
, Error
**errp
);
1866 * memory_region_iommu_replay: replay existing IOMMU translations to
1867 * a notifier with the minimum page granularity returned by
1868 * mr->iommu_ops->get_page_size().
1870 * Note: this is not related to record-and-replay functionality.
1872 * @iommu_mr: the memory region to observe
1873 * @n: the notifier to which to replay iommu mappings
1875 void memory_region_iommu_replay(IOMMUMemoryRegion
*iommu_mr
, IOMMUNotifier
*n
);
1878 * memory_region_unregister_iommu_notifier: unregister a notifier for
1879 * changes to IOMMU translation entries.
1881 * @mr: the memory region which was observed and for which notity_stopped()
1882 * needs to be called
1883 * @n: the notifier to be removed.
1885 void memory_region_unregister_iommu_notifier(MemoryRegion
*mr
,
1889 * memory_region_iommu_get_attr: return an IOMMU attr if get_attr() is
1890 * defined on the IOMMU.
1892 * Returns 0 on success, or a negative errno otherwise. In particular,
1893 * -EINVAL indicates that the IOMMU does not support the requested
1896 * @iommu_mr: the memory region
1897 * @attr: the requested attribute
1898 * @data: a pointer to the requested attribute data
1900 int memory_region_iommu_get_attr(IOMMUMemoryRegion
*iommu_mr
,
1901 enum IOMMUMemoryRegionAttr attr
,
1905 * memory_region_iommu_attrs_to_index: return the IOMMU index to
1906 * use for translations with the given memory transaction attributes.
1908 * @iommu_mr: the memory region
1909 * @attrs: the memory transaction attributes
1911 int memory_region_iommu_attrs_to_index(IOMMUMemoryRegion
*iommu_mr
,
1915 * memory_region_iommu_num_indexes: return the total number of IOMMU
1916 * indexes that this IOMMU supports.
1918 * @iommu_mr: the memory region
1920 int memory_region_iommu_num_indexes(IOMMUMemoryRegion
*iommu_mr
);
1923 * memory_region_iommu_set_page_size_mask: set the supported page
1924 * sizes for a given IOMMU memory region
1926 * @iommu_mr: IOMMU memory region
1927 * @page_size_mask: supported page size mask
1928 * @errp: pointer to Error*, to store an error if it happens.
1930 int memory_region_iommu_set_page_size_mask(IOMMUMemoryRegion
*iommu_mr
,
1931 uint64_t page_size_mask
,
1935 * memory_region_name: get a memory region's name
1937 * Returns the string that was used to initialize the memory region.
1939 * @mr: the memory region being queried
1941 const char *memory_region_name(const MemoryRegion
*mr
);
1944 * memory_region_is_logging: return whether a memory region is logging writes
1946 * Returns %true if the memory region is logging writes for the given client
1948 * @mr: the memory region being queried
1949 * @client: the client being queried
1951 bool memory_region_is_logging(MemoryRegion
*mr
, uint8_t client
);
1954 * memory_region_get_dirty_log_mask: return the clients for which a
1955 * memory region is logging writes.
1957 * Returns a bitmap of clients, in which the DIRTY_MEMORY_* constants
1958 * are the bit indices.
1960 * @mr: the memory region being queried
1962 uint8_t memory_region_get_dirty_log_mask(MemoryRegion
*mr
);
1965 * memory_region_is_rom: check whether a memory region is ROM
1967 * Returns %true if a memory region is read-only memory.
1969 * @mr: the memory region being queried
1971 static inline bool memory_region_is_rom(MemoryRegion
*mr
)
1973 return mr
->ram
&& mr
->readonly
;
1977 * memory_region_is_nonvolatile: check whether a memory region is non-volatile
1979 * Returns %true is a memory region is non-volatile memory.
1981 * @mr: the memory region being queried
1983 static inline bool memory_region_is_nonvolatile(MemoryRegion
*mr
)
1985 return mr
->nonvolatile
;
1989 * memory_region_get_fd: Get a file descriptor backing a RAM memory region.
1991 * Returns a file descriptor backing a file-based RAM memory region,
1992 * or -1 if the region is not a file-based RAM memory region.
1994 * @mr: the RAM or alias memory region being queried.
1996 int memory_region_get_fd(MemoryRegion
*mr
);
1999 * memory_region_from_host: Convert a pointer into a RAM memory region
2000 * and an offset within it.
2002 * Given a host pointer inside a RAM memory region (created with
2003 * memory_region_init_ram() or memory_region_init_ram_ptr()), return
2004 * the MemoryRegion and the offset within it.
2006 * Use with care; by the time this function returns, the returned pointer is
2007 * not protected by RCU anymore. If the caller is not within an RCU critical
2008 * section and does not hold the BQL, it must have other means of
2009 * protecting the pointer, such as a reference to the region that includes
2010 * the incoming ram_addr_t.
2012 * @ptr: the host pointer to be converted
2013 * @offset: the offset within memory region
2015 MemoryRegion
*memory_region_from_host(void *ptr
, ram_addr_t
*offset
);
2018 * memory_region_get_ram_ptr: Get a pointer into a RAM memory region.
2020 * Returns a host pointer to a RAM memory region (created with
2021 * memory_region_init_ram() or memory_region_init_ram_ptr()).
2023 * Use with care; by the time this function returns, the returned pointer is
2024 * not protected by RCU anymore. If the caller is not within an RCU critical
2025 * section and does not hold the BQL, it must have other means of
2026 * protecting the pointer, such as a reference to the region that includes
2027 * the incoming ram_addr_t.
2029 * @mr: the memory region being queried.
2031 void *memory_region_get_ram_ptr(MemoryRegion
*mr
);
2033 /* memory_region_ram_resize: Resize a RAM region.
2035 * Resizing RAM while migrating can result in the migration being canceled.
2036 * Care has to be taken if the guest might have already detected the memory.
2038 * @mr: a memory region created with @memory_region_init_resizeable_ram.
2039 * @newsize: the new size the region
2040 * @errp: pointer to Error*, to store an error if it happens.
2042 void memory_region_ram_resize(MemoryRegion
*mr
, ram_addr_t newsize
,
2046 * memory_region_msync: Synchronize selected address range of
2047 * a memory mapped region
2049 * @mr: the memory region to be msync
2050 * @addr: the initial address of the range to be sync
2051 * @size: the size of the range to be sync
2053 void memory_region_msync(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
2056 * memory_region_writeback: Trigger cache writeback for
2057 * selected address range
2059 * @mr: the memory region to be updated
2060 * @addr: the initial address of the range to be written back
2061 * @size: the size of the range to be written back
2063 void memory_region_writeback(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
2066 * memory_region_set_log: Turn dirty logging on or off for a region.
2068 * Turns dirty logging on or off for a specified client (display, migration).
2069 * Only meaningful for RAM regions.
2071 * @mr: the memory region being updated.
2072 * @log: whether dirty logging is to be enabled or disabled.
2073 * @client: the user of the logging information; %DIRTY_MEMORY_VGA only.
2075 void memory_region_set_log(MemoryRegion
*mr
, bool log
, unsigned client
);
2078 * memory_region_set_dirty: Mark a range of bytes as dirty in a memory region.
2080 * Marks a range of bytes as dirty, after it has been dirtied outside
2083 * @mr: the memory region being dirtied.
2084 * @addr: the address (relative to the start of the region) being dirtied.
2085 * @size: size of the range being dirtied.
2087 void memory_region_set_dirty(MemoryRegion
*mr
, hwaddr addr
,
2091 * memory_region_clear_dirty_bitmap - clear dirty bitmap for memory range
2093 * This function is called when the caller wants to clear the remote
2094 * dirty bitmap of a memory range within the memory region. This can
2095 * be used by e.g. KVM to manually clear dirty log when
2096 * KVM_CAP_MANUAL_DIRTY_LOG_PROTECT is declared support by the host
2099 * @mr: the memory region to clear the dirty log upon
2100 * @start: start address offset within the memory region
2101 * @len: length of the memory region to clear dirty bitmap
2103 void memory_region_clear_dirty_bitmap(MemoryRegion
*mr
, hwaddr start
,
2107 * memory_region_snapshot_and_clear_dirty: Get a snapshot of the dirty
2108 * bitmap and clear it.
2110 * Creates a snapshot of the dirty bitmap, clears the dirty bitmap and
2111 * returns the snapshot. The snapshot can then be used to query dirty
2112 * status, using memory_region_snapshot_get_dirty. Snapshotting allows
2113 * querying the same page multiple times, which is especially useful for
2114 * display updates where the scanlines often are not page aligned.
2116 * The dirty bitmap region which gets copied into the snapshot (and
2117 * cleared afterwards) can be larger than requested. The boundaries
2118 * are rounded up/down so complete bitmap longs (covering 64 pages on
2119 * 64bit hosts) can be copied over into the bitmap snapshot. Which
2120 * isn't a problem for display updates as the extra pages are outside
2121 * the visible area, and in case the visible area changes a full
2122 * display redraw is due anyway. Should other use cases for this
2123 * function emerge we might have to revisit this implementation
2126 * Use g_free to release DirtyBitmapSnapshot.
2128 * @mr: the memory region being queried.
2129 * @addr: the address (relative to the start of the region) being queried.
2130 * @size: the size of the range being queried.
2131 * @client: the user of the logging information; typically %DIRTY_MEMORY_VGA.
2133 DirtyBitmapSnapshot
*memory_region_snapshot_and_clear_dirty(MemoryRegion
*mr
,
2139 * memory_region_snapshot_get_dirty: Check whether a range of bytes is dirty
2140 * in the specified dirty bitmap snapshot.
2142 * @mr: the memory region being queried.
2143 * @snap: the dirty bitmap snapshot
2144 * @addr: the address (relative to the start of the region) being queried.
2145 * @size: the size of the range being queried.
2147 bool memory_region_snapshot_get_dirty(MemoryRegion
*mr
,
2148 DirtyBitmapSnapshot
*snap
,
2149 hwaddr addr
, hwaddr size
);
2152 * memory_region_reset_dirty: Mark a range of pages as clean, for a specified
2155 * Marks a range of pages as no longer dirty.
2157 * @mr: the region being updated.
2158 * @addr: the start of the subrange being cleaned.
2159 * @size: the size of the subrange being cleaned.
2160 * @client: the user of the logging information; %DIRTY_MEMORY_MIGRATION or
2161 * %DIRTY_MEMORY_VGA.
2163 void memory_region_reset_dirty(MemoryRegion
*mr
, hwaddr addr
,
2164 hwaddr size
, unsigned client
);
2167 * memory_region_flush_rom_device: Mark a range of pages dirty and invalidate
2168 * TBs (for self-modifying code).
2170 * The MemoryRegionOps->write() callback of a ROM device must use this function
2171 * to mark byte ranges that have been modified internally, such as by directly
2172 * accessing the memory returned by memory_region_get_ram_ptr().
2174 * This function marks the range dirty and invalidates TBs so that TCG can
2175 * detect self-modifying code.
2177 * @mr: the region being flushed.
2178 * @addr: the start, relative to the start of the region, of the range being
2180 * @size: the size, in bytes, of the range being flushed.
2182 void memory_region_flush_rom_device(MemoryRegion
*mr
, hwaddr addr
, hwaddr size
);
2185 * memory_region_set_readonly: Turn a memory region read-only (or read-write)
2187 * Allows a memory region to be marked as read-only (turning it into a ROM).
2188 * only useful on RAM regions.
2190 * @mr: the region being updated.
2191 * @readonly: whether rhe region is to be ROM or RAM.
2193 void memory_region_set_readonly(MemoryRegion
*mr
, bool readonly
);
2196 * memory_region_set_nonvolatile: Turn a memory region non-volatile
2198 * Allows a memory region to be marked as non-volatile.
2199 * only useful on RAM regions.
2201 * @mr: the region being updated.
2202 * @nonvolatile: whether rhe region is to be non-volatile.
2204 void memory_region_set_nonvolatile(MemoryRegion
*mr
, bool nonvolatile
);
2207 * memory_region_rom_device_set_romd: enable/disable ROMD mode
2209 * Allows a ROM device (initialized with memory_region_init_rom_device() to
2210 * set to ROMD mode (default) or MMIO mode. When it is in ROMD mode, the
2211 * device is mapped to guest memory and satisfies read access directly.
2212 * When in MMIO mode, reads are forwarded to the #MemoryRegion.read function.
2213 * Writes are always handled by the #MemoryRegion.write function.
2215 * @mr: the memory region to be updated
2216 * @romd_mode: %true to put the region into ROMD mode
2218 void memory_region_rom_device_set_romd(MemoryRegion
*mr
, bool romd_mode
);
2221 * memory_region_set_coalescing: Enable memory coalescing for the region.
2223 * Enabled writes to a region to be queued for later processing. MMIO ->write
2224 * callbacks may be delayed until a non-coalesced MMIO is issued.
2225 * Only useful for IO regions. Roughly similar to write-combining hardware.
2227 * @mr: the memory region to be write coalesced
2229 void memory_region_set_coalescing(MemoryRegion
*mr
);
2232 * memory_region_add_coalescing: Enable memory coalescing for a sub-range of
2235 * Like memory_region_set_coalescing(), but works on a sub-range of a region.
2236 * Multiple calls can be issued coalesced disjoint ranges.
2238 * @mr: the memory region to be updated.
2239 * @offset: the start of the range within the region to be coalesced.
2240 * @size: the size of the subrange to be coalesced.
2242 void memory_region_add_coalescing(MemoryRegion
*mr
,
2247 * memory_region_clear_coalescing: Disable MMIO coalescing for the region.
2249 * Disables any coalescing caused by memory_region_set_coalescing() or
2250 * memory_region_add_coalescing(). Roughly equivalent to uncacheble memory
2253 * @mr: the memory region to be updated.
2255 void memory_region_clear_coalescing(MemoryRegion
*mr
);
2258 * memory_region_set_flush_coalesced: Enforce memory coalescing flush before
2261 * Ensure that pending coalesced MMIO request are flushed before the memory
2262 * region is accessed. This property is automatically enabled for all regions
2263 * passed to memory_region_set_coalescing() and memory_region_add_coalescing().
2265 * @mr: the memory region to be updated.
2267 void memory_region_set_flush_coalesced(MemoryRegion
*mr
);
2270 * memory_region_clear_flush_coalesced: Disable memory coalescing flush before
2273 * Clear the automatic coalesced MMIO flushing enabled via
2274 * memory_region_set_flush_coalesced. Note that this service has no effect on
2275 * memory regions that have MMIO coalescing enabled for themselves. For them,
2276 * automatic flushing will stop once coalescing is disabled.
2278 * @mr: the memory region to be updated.
2280 void memory_region_clear_flush_coalesced(MemoryRegion
*mr
);
2283 * memory_region_add_eventfd: Request an eventfd to be triggered when a word
2284 * is written to a location.
2286 * Marks a word in an IO region (initialized with memory_region_init_io())
2287 * as a trigger for an eventfd event. The I/O callback will not be called.
2288 * The caller must be prepared to handle failure (that is, take the required
2289 * action if the callback _is_ called).
2291 * @mr: the memory region being updated.
2292 * @addr: the address within @mr that is to be monitored
2293 * @size: the size of the access to trigger the eventfd
2294 * @match_data: whether to match against @data, instead of just @addr
2295 * @data: the data to match against the guest write
2296 * @e: event notifier to be triggered when @addr, @size, and @data all match.
2298 void memory_region_add_eventfd(MemoryRegion
*mr
,
2306 * memory_region_del_eventfd: Cancel an eventfd.
2308 * Cancels an eventfd trigger requested by a previous
2309 * memory_region_add_eventfd() call.
2311 * @mr: the memory region being updated.
2312 * @addr: the address within @mr that is to be monitored
2313 * @size: the size of the access to trigger the eventfd
2314 * @match_data: whether to match against @data, instead of just @addr
2315 * @data: the data to match against the guest write
2316 * @e: event notifier to be triggered when @addr, @size, and @data all match.
2318 void memory_region_del_eventfd(MemoryRegion
*mr
,
2326 * memory_region_add_subregion: Add a subregion to a container.
2328 * Adds a subregion at @offset. The subregion may not overlap with other
2329 * subregions (except for those explicitly marked as overlapping). A region
2330 * may only be added once as a subregion (unless removed with
2331 * memory_region_del_subregion()); use memory_region_init_alias() if you
2332 * want a region to be a subregion in multiple locations.
2334 * @mr: the region to contain the new subregion; must be a container
2335 * initialized with memory_region_init().
2336 * @offset: the offset relative to @mr where @subregion is added.
2337 * @subregion: the subregion to be added.
2339 void memory_region_add_subregion(MemoryRegion
*mr
,
2341 MemoryRegion
*subregion
);
2343 * memory_region_add_subregion_overlap: Add a subregion to a container
2346 * Adds a subregion at @offset. The subregion may overlap with other
2347 * subregions. Conflicts are resolved by having a higher @priority hide a
2348 * lower @priority. Subregions without priority are taken as @priority 0.
2349 * A region may only be added once as a subregion (unless removed with
2350 * memory_region_del_subregion()); use memory_region_init_alias() if you
2351 * want a region to be a subregion in multiple locations.
2353 * @mr: the region to contain the new subregion; must be a container
2354 * initialized with memory_region_init().
2355 * @offset: the offset relative to @mr where @subregion is added.
2356 * @subregion: the subregion to be added.
2357 * @priority: used for resolving overlaps; highest priority wins.
2359 void memory_region_add_subregion_overlap(MemoryRegion
*mr
,
2361 MemoryRegion
*subregion
,
2365 * memory_region_get_ram_addr: Get the ram address associated with a memory
2368 * @mr: the region to be queried
2370 ram_addr_t
memory_region_get_ram_addr(MemoryRegion
*mr
);
2372 uint64_t memory_region_get_alignment(const MemoryRegion
*mr
);
2374 * memory_region_del_subregion: Remove a subregion.
2376 * Removes a subregion from its container.
2378 * @mr: the container to be updated.
2379 * @subregion: the region being removed; must be a current subregion of @mr.
2381 void memory_region_del_subregion(MemoryRegion
*mr
,
2382 MemoryRegion
*subregion
);
2385 * memory_region_set_enabled: dynamically enable or disable a region
2387 * Enables or disables a memory region. A disabled memory region
2388 * ignores all accesses to itself and its subregions. It does not
2389 * obscure sibling subregions with lower priority - it simply behaves as
2390 * if it was removed from the hierarchy.
2392 * Regions default to being enabled.
2394 * @mr: the region to be updated
2395 * @enabled: whether to enable or disable the region
2397 void memory_region_set_enabled(MemoryRegion
*mr
, bool enabled
);
2400 * memory_region_set_address: dynamically update the address of a region
2402 * Dynamically updates the address of a region, relative to its container.
2403 * May be used on regions are currently part of a memory hierarchy.
2405 * @mr: the region to be updated
2406 * @addr: new address, relative to container region
2408 void memory_region_set_address(MemoryRegion
*mr
, hwaddr addr
);
2411 * memory_region_set_size: dynamically update the size of a region.
2413 * Dynamically updates the size of a region.
2415 * @mr: the region to be updated
2416 * @size: used size of the region.
2418 void memory_region_set_size(MemoryRegion
*mr
, uint64_t size
);
2421 * memory_region_set_alias_offset: dynamically update a memory alias's offset
2423 * Dynamically updates the offset into the target region that an alias points
2424 * to, as if the fourth argument to memory_region_init_alias() has changed.
2426 * @mr: the #MemoryRegion to be updated; should be an alias.
2427 * @offset: the new offset into the target memory region
2429 void memory_region_set_alias_offset(MemoryRegion
*mr
,
2433 * memory_region_set_unmergeable: Set a memory region unmergeable
2435 * Mark a memory region unmergeable, resulting in the memory region (or
2436 * everything contained in a memory region container) not getting merged when
2437 * simplifying the address space and notifying memory listeners. Consequently,
2438 * memory listeners will never get notified about ranges that are larger than
2439 * the original memory regions.
2441 * This is primarily useful when multiple aliases to a RAM memory region are
2442 * mapped into a memory region container, and updates (e.g., enable/disable or
2443 * map/unmap) of individual memory region aliases are not supposed to affect
2444 * other memory regions in the same container.
2446 * @mr: the #MemoryRegion to be updated
2447 * @unmergeable: whether to mark the #MemoryRegion unmergeable
2449 void memory_region_set_unmergeable(MemoryRegion
*mr
, bool unmergeable
);
2452 * memory_region_present: checks if an address relative to a @container
2453 * translates into #MemoryRegion within @container
2455 * Answer whether a #MemoryRegion within @container covers the address
2458 * @container: a #MemoryRegion within which @addr is a relative address
2459 * @addr: the area within @container to be searched
2461 bool memory_region_present(MemoryRegion
*container
, hwaddr addr
);
2464 * memory_region_is_mapped: returns true if #MemoryRegion is mapped
2465 * into another memory region, which does not necessarily imply that it is
2466 * mapped into an address space.
2468 * @mr: a #MemoryRegion which should be checked if it's mapped
2470 bool memory_region_is_mapped(MemoryRegion
*mr
);
2473 * memory_region_get_ram_discard_manager: get the #RamDiscardManager for a
2476 * The #RamDiscardManager cannot change while a memory region is mapped.
2478 * @mr: the #MemoryRegion
2480 RamDiscardManager
*memory_region_get_ram_discard_manager(MemoryRegion
*mr
);
2483 * memory_region_has_ram_discard_manager: check whether a #MemoryRegion has a
2484 * #RamDiscardManager assigned
2486 * @mr: the #MemoryRegion
2488 static inline bool memory_region_has_ram_discard_manager(MemoryRegion
*mr
)
2490 return !!memory_region_get_ram_discard_manager(mr
);
2494 * memory_region_set_ram_discard_manager: set the #RamDiscardManager for a
2497 * This function must not be called for a mapped #MemoryRegion, a #MemoryRegion
2498 * that does not cover RAM, or a #MemoryRegion that already has a
2499 * #RamDiscardManager assigned.
2501 * @mr: the #MemoryRegion
2502 * @rdm: #RamDiscardManager to set
2504 void memory_region_set_ram_discard_manager(MemoryRegion
*mr
,
2505 RamDiscardManager
*rdm
);
2508 * memory_region_find: translate an address/size relative to a
2509 * MemoryRegion into a #MemoryRegionSection.
2511 * Locates the first #MemoryRegion within @mr that overlaps the range
2512 * given by @addr and @size.
2514 * Returns a #MemoryRegionSection that describes a contiguous overlap.
2515 * It will have the following characteristics:
2516 * - @size = 0 iff no overlap was found
2517 * - @mr is non-%NULL iff an overlap was found
2519 * Remember that in the return value the @offset_within_region is
2520 * relative to the returned region (in the .@mr field), not to the
2523 * Similarly, the .@offset_within_address_space is relative to the
2524 * address space that contains both regions, the passed and the
2525 * returned one. However, in the special case where the @mr argument
2526 * has no container (and thus is the root of the address space), the
2527 * following will hold:
2528 * - @offset_within_address_space >= @addr
2529 * - @offset_within_address_space + .@size <= @addr + @size
2531 * @mr: a MemoryRegion within which @addr is a relative address
2532 * @addr: start of the area within @as to be searched
2533 * @size: size of the area to be searched
2535 MemoryRegionSection
memory_region_find(MemoryRegion
*mr
,
2536 hwaddr addr
, uint64_t size
);
2539 * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2541 * Synchronizes the dirty page log for all address spaces.
2543 * @last_stage: whether this is the last stage of live migration
2545 void memory_global_dirty_log_sync(bool last_stage
);
2548 * memory_global_dirty_log_sync: synchronize the dirty log for all memory
2550 * Synchronizes the vCPUs with a thread that is reading the dirty bitmap.
2551 * This function must be called after the dirty log bitmap is cleared, and
2552 * before dirty guest memory pages are read. If you are using
2553 * #DirtyBitmapSnapshot, memory_region_snapshot_and_clear_dirty() takes
2554 * care of doing this.
2556 void memory_global_after_dirty_log_sync(void);
2559 * memory_region_transaction_begin: Start a transaction.
2561 * During a transaction, changes will be accumulated and made visible
2562 * only when the transaction ends (is committed).
2564 void memory_region_transaction_begin(void);
2567 * memory_region_transaction_commit: Commit a transaction and make changes
2568 * visible to the guest.
2570 void memory_region_transaction_commit(void);
2573 * memory_listener_register: register callbacks to be called when memory
2574 * sections are mapped or unmapped into an address
2577 * @listener: an object containing the callbacks to be called
2578 * @filter: if non-%NULL, only regions in this address space will be observed
2580 void memory_listener_register(MemoryListener
*listener
, AddressSpace
*filter
);
2583 * memory_listener_unregister: undo the effect of memory_listener_register()
2585 * @listener: an object containing the callbacks to be removed
2587 void memory_listener_unregister(MemoryListener
*listener
);
2590 * memory_global_dirty_log_start: begin dirty logging for all regions
2592 * @flags: purpose of starting dirty log, migration or dirty rate
2593 * @errp: pointer to Error*, to store an error if it happens.
2595 * Return: true on success, else false setting @errp with error.
2597 bool memory_global_dirty_log_start(unsigned int flags
, Error
**errp
);
2600 * memory_global_dirty_log_stop: end dirty logging for all regions
2602 * @flags: purpose of stopping dirty log, migration or dirty rate
2604 void memory_global_dirty_log_stop(unsigned int flags
);
2606 void mtree_info(bool flatview
, bool dispatch_tree
, bool owner
, bool disabled
);
2608 bool memory_region_access_valid(MemoryRegion
*mr
, hwaddr addr
,
2609 unsigned size
, bool is_write
,
2613 * memory_region_dispatch_read: perform a read directly to the specified
2616 * @mr: #MemoryRegion to access
2617 * @addr: address within that region
2618 * @pval: pointer to uint64_t which the data is written to
2619 * @op: size, sign, and endianness of the memory operation
2620 * @attrs: memory transaction attributes to use for the access
2622 MemTxResult
memory_region_dispatch_read(MemoryRegion
*mr
,
2628 * memory_region_dispatch_write: perform a write directly to the specified
2631 * @mr: #MemoryRegion to access
2632 * @addr: address within that region
2633 * @data: data to write
2634 * @op: size, sign, and endianness of the memory operation
2635 * @attrs: memory transaction attributes to use for the access
2637 MemTxResult
memory_region_dispatch_write(MemoryRegion
*mr
,
2644 * address_space_init: initializes an address space
2646 * @as: an uninitialized #AddressSpace
2647 * @root: a #MemoryRegion that routes addresses for the address space
2648 * @name: an address space name. The name is only used for debugging
2651 void address_space_init(AddressSpace
*as
, MemoryRegion
*root
, const char *name
);
2654 * address_space_destroy: destroy an address space
2656 * Releases all resources associated with an address space. After an address space
2657 * is destroyed, its root memory region (given by address_space_init()) may be destroyed
2660 * @as: address space to be destroyed
2662 void address_space_destroy(AddressSpace
*as
);
2665 * address_space_remove_listeners: unregister all listeners of an address space
2667 * Removes all callbacks previously registered with memory_listener_register()
2670 * @as: an initialized #AddressSpace
2672 void address_space_remove_listeners(AddressSpace
*as
);
2675 * address_space_rw: read from or write to an address space.
2677 * Return a MemTxResult indicating whether the operation succeeded
2678 * or failed (eg unassigned memory, device rejected the transaction,
2681 * @as: #AddressSpace to be accessed
2682 * @addr: address within that address space
2683 * @attrs: memory transaction attributes
2684 * @buf: buffer with the data transferred
2685 * @len: the number of bytes to read or write
2686 * @is_write: indicates the transfer direction
2688 MemTxResult
address_space_rw(AddressSpace
*as
, hwaddr addr
,
2689 MemTxAttrs attrs
, void *buf
,
2690 hwaddr len
, bool is_write
);
2693 * address_space_write: write to address space.
2695 * Return a MemTxResult indicating whether the operation succeeded
2696 * or failed (eg unassigned memory, device rejected the transaction,
2699 * @as: #AddressSpace to be accessed
2700 * @addr: address within that address space
2701 * @attrs: memory transaction attributes
2702 * @buf: buffer with the data transferred
2703 * @len: the number of bytes to write
2705 MemTxResult
address_space_write(AddressSpace
*as
, hwaddr addr
,
2707 const void *buf
, hwaddr len
);
2710 * address_space_write_rom: write to address space, including ROM.
2712 * This function writes to the specified address space, but will
2713 * write data to both ROM and RAM. This is used for non-guest
2714 * writes like writes from the gdb debug stub or initial loading
2717 * Note that portions of the write which attempt to write data to
2718 * a device will be silently ignored -- only real RAM and ROM will
2721 * Return a MemTxResult indicating whether the operation succeeded
2722 * or failed (eg unassigned memory, device rejected the transaction,
2725 * @as: #AddressSpace to be accessed
2726 * @addr: address within that address space
2727 * @attrs: memory transaction attributes
2728 * @buf: buffer with the data transferred
2729 * @len: the number of bytes to write
2731 MemTxResult
address_space_write_rom(AddressSpace
*as
, hwaddr addr
,
2733 const void *buf
, hwaddr len
);
2735 /* address_space_ld*: load from an address space
2736 * address_space_st*: store to an address space
2738 * These functions perform a load or store of the byte, word,
2739 * longword or quad to the specified address within the AddressSpace.
2740 * The _le suffixed functions treat the data as little endian;
2741 * _be indicates big endian; no suffix indicates "same endianness
2744 * The "guest CPU endianness" accessors are deprecated for use outside
2745 * target-* code; devices should be CPU-agnostic and use either the LE
2746 * or the BE accessors.
2748 * @as #AddressSpace to be accessed
2749 * @addr: address within that address space
2750 * @val: data value, for stores
2751 * @attrs: memory transaction attributes
2752 * @result: location to write the success/failure of the transaction;
2753 * if NULL, this information is discarded
2758 #define ARG1_DECL AddressSpace *as
2759 #include "exec/memory_ldst.h.inc"
2763 #define ARG1_DECL AddressSpace *as
2764 #include "exec/memory_ldst_phys.h.inc"
2766 struct MemoryRegionCache
{
2771 MemoryRegionSection mrs
;
2775 /* address_space_ld*_cached: load from a cached #MemoryRegion
2776 * address_space_st*_cached: store into a cached #MemoryRegion
2778 * These functions perform a load or store of the byte, word,
2779 * longword or quad to the specified address. The address is
2780 * a physical address in the AddressSpace, but it must lie within
2781 * a #MemoryRegion that was mapped with address_space_cache_init.
2783 * The _le suffixed functions treat the data as little endian;
2784 * _be indicates big endian; no suffix indicates "same endianness
2787 * The "guest CPU endianness" accessors are deprecated for use outside
2788 * target-* code; devices should be CPU-agnostic and use either the LE
2789 * or the BE accessors.
2791 * @cache: previously initialized #MemoryRegionCache to be accessed
2792 * @addr: address within the address space
2793 * @val: data value, for stores
2794 * @attrs: memory transaction attributes
2795 * @result: location to write the success/failure of the transaction;
2796 * if NULL, this information is discarded
2799 #define SUFFIX _cached_slow
2801 #define ARG1_DECL MemoryRegionCache *cache
2802 #include "exec/memory_ldst.h.inc"
2804 /* Inline fast path for direct RAM access. */
2805 static inline uint8_t address_space_ldub_cached(MemoryRegionCache
*cache
,
2806 hwaddr addr
, MemTxAttrs attrs
, MemTxResult
*result
)
2808 assert(addr
< cache
->len
);
2809 if (likely(cache
->ptr
)) {
2810 return ldub_p(cache
->ptr
+ addr
);
2812 return address_space_ldub_cached_slow(cache
, addr
, attrs
, result
);
2816 static inline void address_space_stb_cached(MemoryRegionCache
*cache
,
2817 hwaddr addr
, uint8_t val
, MemTxAttrs attrs
, MemTxResult
*result
)
2819 assert(addr
< cache
->len
);
2820 if (likely(cache
->ptr
)) {
2821 stb_p(cache
->ptr
+ addr
, val
);
2823 address_space_stb_cached_slow(cache
, addr
, val
, attrs
, result
);
2827 #define ENDIANNESS _le
2828 #include "exec/memory_ldst_cached.h.inc"
2830 #define ENDIANNESS _be
2831 #include "exec/memory_ldst_cached.h.inc"
2833 #define SUFFIX _cached
2835 #define ARG1_DECL MemoryRegionCache *cache
2836 #include "exec/memory_ldst_phys.h.inc"
2838 /* address_space_cache_init: prepare for repeated access to a physical
2841 * @cache: #MemoryRegionCache to be filled
2842 * @as: #AddressSpace to be accessed
2843 * @addr: address within that address space
2844 * @len: length of buffer
2845 * @is_write: indicates the transfer direction
2847 * Will only work with RAM, and may map a subset of the requested range by
2848 * returning a value that is less than @len. On failure, return a negative
2851 * Because it only works with RAM, this function can be used for
2852 * read-modify-write operations. In this case, is_write should be %true.
2854 * Note that addresses passed to the address_space_*_cached functions
2855 * are relative to @addr.
2857 int64_t address_space_cache_init(MemoryRegionCache
*cache
,
2864 * address_space_cache_init_empty: Initialize empty #MemoryRegionCache
2866 * @cache: The #MemoryRegionCache to operate on.
2868 * Initializes #MemoryRegionCache structure without memory region attached.
2869 * Cache initialized this way can only be safely destroyed, but not used.
2871 static inline void address_space_cache_init_empty(MemoryRegionCache
*cache
)
2873 cache
->mrs
.mr
= NULL
;
2874 /* There is no real need to initialize fv, but it makes Coverity happy. */
2879 * address_space_cache_invalidate: complete a write to a #MemoryRegionCache
2881 * @cache: The #MemoryRegionCache to operate on.
2882 * @addr: The first physical address that was written, relative to the
2883 * address that was passed to @address_space_cache_init.
2884 * @access_len: The number of bytes that were written starting at @addr.
2886 void address_space_cache_invalidate(MemoryRegionCache
*cache
,
2891 * address_space_cache_destroy: free a #MemoryRegionCache
2893 * @cache: The #MemoryRegionCache whose memory should be released.
2895 void address_space_cache_destroy(MemoryRegionCache
*cache
);
2897 /* address_space_get_iotlb_entry: translate an address into an IOTLB
2898 * entry. Should be called from an RCU critical section.
2900 IOMMUTLBEntry
address_space_get_iotlb_entry(AddressSpace
*as
, hwaddr addr
,
2901 bool is_write
, MemTxAttrs attrs
);
2903 /* address_space_translate: translate an address range into an address space
2904 * into a MemoryRegion and an address range into that section. Should be
2905 * called from an RCU critical section, to avoid that the last reference
2906 * to the returned region disappears after address_space_translate returns.
2908 * @fv: #FlatView to be accessed
2909 * @addr: address within that address space
2910 * @xlat: pointer to address within the returned memory region section's
2912 * @len: pointer to length
2913 * @is_write: indicates the transfer direction
2914 * @attrs: memory attributes
2916 MemoryRegion
*flatview_translate(FlatView
*fv
,
2917 hwaddr addr
, hwaddr
*xlat
,
2918 hwaddr
*len
, bool is_write
,
2921 static inline MemoryRegion
*address_space_translate(AddressSpace
*as
,
2922 hwaddr addr
, hwaddr
*xlat
,
2923 hwaddr
*len
, bool is_write
,
2926 return flatview_translate(address_space_to_flatview(as
),
2927 addr
, xlat
, len
, is_write
, attrs
);
2930 /* address_space_access_valid: check for validity of accessing an address
2933 * Check whether memory is assigned to the given address space range, and
2934 * access is permitted by any IOMMU regions that are active for the address
2937 * For now, addr and len should be aligned to a page size. This limitation
2938 * will be lifted in the future.
2940 * @as: #AddressSpace to be accessed
2941 * @addr: address within that address space
2942 * @len: length of the area to be checked
2943 * @is_write: indicates the transfer direction
2944 * @attrs: memory attributes
2946 bool address_space_access_valid(AddressSpace
*as
, hwaddr addr
, hwaddr len
,
2947 bool is_write
, MemTxAttrs attrs
);
2949 /* address_space_map: map a physical memory region into a host virtual address
2951 * May map a subset of the requested range, given by and returned in @plen.
2952 * May return %NULL and set *@plen to zero(0), if resources needed to perform
2953 * the mapping are exhausted.
2954 * Use only for reads OR writes - not for read-modify-write operations.
2955 * Use address_space_register_map_client() to know when retrying the map
2956 * operation is likely to succeed.
2958 * @as: #AddressSpace to be accessed
2959 * @addr: address within that address space
2960 * @plen: pointer to length of buffer; updated on return
2961 * @is_write: indicates the transfer direction
2962 * @attrs: memory attributes
2964 void *address_space_map(AddressSpace
*as
, hwaddr addr
,
2965 hwaddr
*plen
, bool is_write
, MemTxAttrs attrs
);
2967 /* address_space_unmap: Unmaps a memory region previously mapped by address_space_map()
2969 * Will also mark the memory as dirty if @is_write == %true. @access_len gives
2970 * the amount of memory that was actually read or written by the caller.
2972 * @as: #AddressSpace used
2973 * @buffer: host pointer as returned by address_space_map()
2974 * @len: buffer length as returned by address_space_map()
2975 * @access_len: amount of data actually transferred
2976 * @is_write: indicates the transfer direction
2978 void address_space_unmap(AddressSpace
*as
, void *buffer
, hwaddr len
,
2979 bool is_write
, hwaddr access_len
);
2982 * address_space_register_map_client: Register a callback to invoke when
2983 * resources for address_space_map() are available again.
2985 * address_space_map may fail when there are not enough resources available,
2986 * such as when bounce buffer memory would exceed the limit. The callback can
2987 * be used to retry the address_space_map operation. Note that the callback
2988 * gets automatically removed after firing.
2990 * @as: #AddressSpace to be accessed
2991 * @bh: callback to invoke when address_space_map() retry is appropriate
2993 void address_space_register_map_client(AddressSpace
*as
, QEMUBH
*bh
);
2996 * address_space_unregister_map_client: Unregister a callback that has
2997 * previously been registered and not fired yet.
2999 * @as: #AddressSpace to be accessed
3000 * @bh: callback to unregister
3002 void address_space_unregister_map_client(AddressSpace
*as
, QEMUBH
*bh
);
3004 /* Internal functions, part of the implementation of address_space_read. */
3005 MemTxResult
address_space_read_full(AddressSpace
*as
, hwaddr addr
,
3006 MemTxAttrs attrs
, void *buf
, hwaddr len
);
3007 MemTxResult
flatview_read_continue(FlatView
*fv
, hwaddr addr
,
3008 MemTxAttrs attrs
, void *buf
,
3009 hwaddr len
, hwaddr addr1
, hwaddr l
,
3011 void *qemu_map_ram_ptr(RAMBlock
*ram_block
, ram_addr_t addr
);
3013 /* Internal functions, part of the implementation of address_space_read_cached
3014 * and address_space_write_cached. */
3015 MemTxResult
address_space_read_cached_slow(MemoryRegionCache
*cache
,
3016 hwaddr addr
, void *buf
, hwaddr len
);
3017 MemTxResult
address_space_write_cached_slow(MemoryRegionCache
*cache
,
3018 hwaddr addr
, const void *buf
,
3021 int memory_access_size(MemoryRegion
*mr
, unsigned l
, hwaddr addr
);
3022 bool prepare_mmio_access(MemoryRegion
*mr
);
3024 static inline bool memory_access_is_direct(MemoryRegion
*mr
, bool is_write
)
3027 return memory_region_is_ram(mr
) && !mr
->readonly
&&
3028 !mr
->rom_device
&& !memory_region_is_ram_device(mr
);
3030 return (memory_region_is_ram(mr
) && !memory_region_is_ram_device(mr
)) ||
3031 memory_region_is_romd(mr
);
3036 * address_space_read: read from an address space.
3038 * Return a MemTxResult indicating whether the operation succeeded
3039 * or failed (eg unassigned memory, device rejected the transaction,
3040 * IOMMU fault). Called within RCU critical section.
3042 * @as: #AddressSpace to be accessed
3043 * @addr: address within that address space
3044 * @attrs: memory transaction attributes
3045 * @buf: buffer with the data transferred
3046 * @len: length of the data transferred
3048 static inline __attribute__((__always_inline__
))
3049 MemTxResult
address_space_read(AddressSpace
*as
, hwaddr addr
,
3050 MemTxAttrs attrs
, void *buf
,
3053 MemTxResult result
= MEMTX_OK
;
3059 if (__builtin_constant_p(len
)) {
3061 RCU_READ_LOCK_GUARD();
3062 fv
= address_space_to_flatview(as
);
3064 mr
= flatview_translate(fv
, addr
, &addr1
, &l
, false, attrs
);
3065 if (len
== l
&& memory_access_is_direct(mr
, false)) {
3066 ptr
= qemu_map_ram_ptr(mr
->ram_block
, addr1
);
3067 memcpy(buf
, ptr
, len
);
3069 result
= flatview_read_continue(fv
, addr
, attrs
, buf
, len
,
3074 result
= address_space_read_full(as
, addr
, attrs
, buf
, len
);
3080 * address_space_read_cached: read from a cached RAM region
3082 * @cache: Cached region to be addressed
3083 * @addr: address relative to the base of the RAM region
3084 * @buf: buffer with the data transferred
3085 * @len: length of the data transferred
3087 static inline MemTxResult
3088 address_space_read_cached(MemoryRegionCache
*cache
, hwaddr addr
,
3089 void *buf
, hwaddr len
)
3091 assert(addr
< cache
->len
&& len
<= cache
->len
- addr
);
3092 fuzz_dma_read_cb(cache
->xlat
+ addr
, len
, cache
->mrs
.mr
);
3093 if (likely(cache
->ptr
)) {
3094 memcpy(buf
, cache
->ptr
+ addr
, len
);
3097 return address_space_read_cached_slow(cache
, addr
, buf
, len
);
3102 * address_space_write_cached: write to a cached RAM region
3104 * @cache: Cached region to be addressed
3105 * @addr: address relative to the base of the RAM region
3106 * @buf: buffer with the data transferred
3107 * @len: length of the data transferred
3109 static inline MemTxResult
3110 address_space_write_cached(MemoryRegionCache
*cache
, hwaddr addr
,
3111 const void *buf
, hwaddr len
)
3113 assert(addr
< cache
->len
&& len
<= cache
->len
- addr
);
3114 if (likely(cache
->ptr
)) {
3115 memcpy(cache
->ptr
+ addr
, buf
, len
);
3118 return address_space_write_cached_slow(cache
, addr
, buf
, len
);
3123 * address_space_set: Fill address space with a constant byte.
3125 * Return a MemTxResult indicating whether the operation succeeded
3126 * or failed (eg unassigned memory, device rejected the transaction,
3129 * @as: #AddressSpace to be accessed
3130 * @addr: address within that address space
3131 * @c: constant byte to fill the memory
3132 * @len: the number of bytes to fill with the constant byte
3133 * @attrs: memory transaction attributes
3135 MemTxResult
address_space_set(AddressSpace
*as
, hwaddr addr
,
3136 uint8_t c
, hwaddr len
, MemTxAttrs attrs
);
3138 #ifdef COMPILING_PER_TARGET
3139 /* enum device_endian to MemOp. */
3140 static inline MemOp
devend_memop(enum device_endian end
)
3142 QEMU_BUILD_BUG_ON(DEVICE_HOST_ENDIAN
!= DEVICE_LITTLE_ENDIAN
&&
3143 DEVICE_HOST_ENDIAN
!= DEVICE_BIG_ENDIAN
);
3145 #if HOST_BIG_ENDIAN != TARGET_BIG_ENDIAN
3146 /* Swap if non-host endianness or native (target) endianness */
3147 return (end
== DEVICE_HOST_ENDIAN
) ? 0 : MO_BSWAP
;
3149 const int non_host_endianness
=
3150 DEVICE_LITTLE_ENDIAN
^ DEVICE_BIG_ENDIAN
^ DEVICE_HOST_ENDIAN
;
3152 /* In this case, native (target) endianness needs no swap. */
3153 return (end
== non_host_endianness
) ? MO_BSWAP
: 0;
3156 #endif /* COMPILING_PER_TARGET */
3159 * Inhibit technologies that require discarding of pages in RAM blocks, e.g.,
3160 * to manage the actual amount of memory consumed by the VM (then, the memory
3161 * provided by RAM blocks might be bigger than the desired memory consumption).
3162 * This *must* be set if:
3163 * - Discarding parts of a RAM blocks does not result in the change being
3164 * reflected in the VM and the pages getting freed.
3165 * - All memory in RAM blocks is pinned or duplicated, invaldiating any previous
3167 * - Discarding parts of a RAM blocks will result in integrity issues (e.g.,
3169 * Technologies that only temporarily pin the current working set of a
3170 * driver are fine, because we don't expect such pages to be discarded
3171 * (esp. based on guest action like balloon inflation).
3173 * This is *not* to be used to protect from concurrent discards (esp.,
3176 * Returns 0 if successful. Returns -EBUSY if a technology that relies on
3177 * discards to work reliably is active.
3179 int ram_block_discard_disable(bool state
);
3182 * See ram_block_discard_disable(): only disable uncoordinated discards,
3183 * keeping coordinated discards (via the RamDiscardManager) enabled.
3185 int ram_block_uncoordinated_discard_disable(bool state
);
3188 * Inhibit technologies that disable discarding of pages in RAM blocks.
3190 * Returns 0 if successful. Returns -EBUSY if discards are already set to
3193 int ram_block_discard_require(bool state
);
3196 * See ram_block_discard_require(): only inhibit technologies that disable
3197 * uncoordinated discarding of pages in RAM blocks, allowing co-existence with
3198 * technologies that only inhibit uncoordinated discards (via the
3199 * RamDiscardManager).
3201 int ram_block_coordinated_discard_require(bool state
);
3204 * Test if any discarding of memory in ram blocks is disabled.
3206 bool ram_block_discard_is_disabled(void);
3209 * Test if any discarding of memory in ram blocks is required to work reliably.
3211 bool ram_block_discard_is_required(void);