2 * Copyright (C) 2010 Citrix Ltd.
4 * This work is licensed under the terms of the GNU GPL, version 2. See
5 * the COPYING file in the top-level directory.
7 * Contributions after 2012-01-13 are licensed under the terms of the
8 * GNU GPL, version 2 or (at your option) any later version.
11 #include "qemu/osdep.h"
14 #include "hw/pci/pci.h"
15 #include "hw/pci/pci_host.h"
16 #include "hw/i386/pc.h"
19 #include "hw/i386/apic-msidef.h"
20 #include "hw/xen/xen_common.h"
21 #include "hw/xen/xen-legacy-backend.h"
22 #include "hw/xen/xen-bus.h"
23 #include "qapi/error.h"
24 #include "qapi/qapi-commands-misc.h"
25 #include "qemu/error-report.h"
26 #include "qemu/main-loop.h"
27 #include "qemu/range.h"
28 #include "sysemu/runstate.h"
29 #include "sysemu/sysemu.h"
30 #include "sysemu/xen-mapcache.h"
32 #include "exec/address-spaces.h"
34 #include <xen/hvm/ioreq.h>
35 #include <xen/hvm/e820.h>
37 //#define DEBUG_XEN_HVM
40 #define DPRINTF(fmt, ...) \
41 do { fprintf(stderr, "xen: " fmt, ## __VA_ARGS__); } while (0)
43 #define DPRINTF(fmt, ...) \
47 static MemoryRegion ram_memory
, ram_640k
, ram_lo
, ram_hi
;
48 static MemoryRegion
*framebuffer
;
49 static bool xen_in_migration
;
51 /* Compatibility with older version */
53 /* This allows QEMU to build on a system that has Xen 4.5 or earlier
54 * installed. This here (not in hw/xen/xen_common.h) because xen/hvm/ioreq.h
55 * needs to be included before this block and hw/xen/xen_common.h needs to
56 * be included before xen/hvm/ioreq.h
58 #ifndef IOREQ_TYPE_VMWARE_PORT
59 #define IOREQ_TYPE_VMWARE_PORT 3
67 typedef struct vmware_regs vmware_regs_t
;
69 struct shared_vmport_iopage
{
70 struct vmware_regs vcpu_vmport_regs
[1];
72 typedef struct shared_vmport_iopage shared_vmport_iopage_t
;
75 static inline uint32_t xen_vcpu_eport(shared_iopage_t
*shared_page
, int i
)
77 return shared_page
->vcpu_ioreq
[i
].vp_eport
;
79 static inline ioreq_t
*xen_vcpu_ioreq(shared_iopage_t
*shared_page
, int vcpu
)
81 return &shared_page
->vcpu_ioreq
[vcpu
];
84 #define BUFFER_IO_MAX_DELAY 100
86 typedef struct XenPhysmap
{
92 QLIST_ENTRY(XenPhysmap
) list
;
95 static QLIST_HEAD(, XenPhysmap
) xen_physmap
;
97 typedef struct XenPciDevice
{
100 QLIST_ENTRY(XenPciDevice
) entry
;
103 typedef struct XenIOState
{
105 shared_iopage_t
*shared_page
;
106 shared_vmport_iopage_t
*shared_vmport_page
;
107 buffered_iopage_t
*buffered_io_page
;
108 QEMUTimer
*buffered_io_timer
;
109 CPUState
**cpu_by_vcpu_id
;
110 /* the evtchn port for polling the notification, */
111 evtchn_port_t
*ioreq_local_port
;
112 /* evtchn remote and local ports for buffered io */
113 evtchn_port_t bufioreq_remote_port
;
114 evtchn_port_t bufioreq_local_port
;
115 /* the evtchn fd for polling */
116 xenevtchn_handle
*xce_handle
;
117 /* which vcpu we are serving */
120 struct xs_handle
*xenstore
;
121 MemoryListener memory_listener
;
122 MemoryListener io_listener
;
123 QLIST_HEAD(, XenPciDevice
) dev_list
;
124 DeviceListener device_listener
;
125 hwaddr free_phys_offset
;
126 const XenPhysmap
*log_for_dirtybit
;
127 /* Buffer used by xen_sync_dirty_bitmap */
128 unsigned long *dirty_bitmap
;
135 /* Xen specific function for piix pci */
137 int xen_pci_slot_get_pirq(PCIDevice
*pci_dev
, int irq_num
)
139 return irq_num
+ ((pci_dev
->devfn
>> 3) << 2);
142 void xen_piix3_set_irq(void *opaque
, int irq_num
, int level
)
144 xen_set_pci_intx_level(xen_domid
, 0, 0, irq_num
>> 2,
148 void xen_piix_pci_write_config_client(uint32_t address
, uint32_t val
, int len
)
152 /* Scan for updates to PCI link routes (0x60-0x63). */
153 for (i
= 0; i
< len
; i
++) {
154 uint8_t v
= (val
>> (8 * i
)) & 0xff;
159 if (((address
+ i
) >= 0x60) && ((address
+ i
) <= 0x63)) {
160 xen_set_pci_link_route(xen_domid
, address
+ i
- 0x60, v
);
165 int xen_is_pirq_msi(uint32_t msi_data
)
167 /* If vector is 0, the msi is remapped into a pirq, passed as
170 return ((msi_data
& MSI_DATA_VECTOR_MASK
) >> MSI_DATA_VECTOR_SHIFT
) == 0;
173 void xen_hvm_inject_msi(uint64_t addr
, uint32_t data
)
175 xen_inject_msi(xen_domid
, addr
, data
);
178 static void xen_suspend_notifier(Notifier
*notifier
, void *data
)
180 xc_set_hvm_param(xen_xc
, xen_domid
, HVM_PARAM_ACPI_S_STATE
, 3);
183 /* Xen Interrupt Controller */
185 static void xen_set_irq(void *opaque
, int irq
, int level
)
187 xen_set_isa_irq_level(xen_domid
, irq
, level
);
190 qemu_irq
*xen_interrupt_controller_init(void)
192 return qemu_allocate_irqs(xen_set_irq
, NULL
, 16);
197 static void xen_ram_init(PCMachineState
*pcms
,
198 ram_addr_t ram_size
, MemoryRegion
**ram_memory_p
)
200 MemoryRegion
*sysmem
= get_system_memory();
201 ram_addr_t block_len
;
202 uint64_t user_lowmem
= object_property_get_uint(qdev_get_machine(),
203 PC_MACHINE_MAX_RAM_BELOW_4G
,
206 /* Handle the machine opt max-ram-below-4g. It is basically doing
207 * min(xen limit, user limit).
210 user_lowmem
= HVM_BELOW_4G_RAM_END
; /* default */
212 if (HVM_BELOW_4G_RAM_END
<= user_lowmem
) {
213 user_lowmem
= HVM_BELOW_4G_RAM_END
;
216 if (ram_size
>= user_lowmem
) {
217 pcms
->above_4g_mem_size
= ram_size
- user_lowmem
;
218 pcms
->below_4g_mem_size
= user_lowmem
;
220 pcms
->above_4g_mem_size
= 0;
221 pcms
->below_4g_mem_size
= ram_size
;
223 if (!pcms
->above_4g_mem_size
) {
224 block_len
= ram_size
;
227 * Xen does not allocate the memory continuously, it keeps a
228 * hole of the size computed above or passed in.
230 block_len
= (1ULL << 32) + pcms
->above_4g_mem_size
;
232 memory_region_init_ram(&ram_memory
, NULL
, "xen.ram", block_len
,
234 *ram_memory_p
= &ram_memory
;
236 memory_region_init_alias(&ram_640k
, NULL
, "xen.ram.640k",
237 &ram_memory
, 0, 0xa0000);
238 memory_region_add_subregion(sysmem
, 0, &ram_640k
);
239 /* Skip of the VGA IO memory space, it will be registered later by the VGA
242 * The area between 0xc0000 and 0x100000 will be used by SeaBIOS to load
243 * the Options ROM, so it is registered here as RAM.
245 memory_region_init_alias(&ram_lo
, NULL
, "xen.ram.lo",
246 &ram_memory
, 0xc0000,
247 pcms
->below_4g_mem_size
- 0xc0000);
248 memory_region_add_subregion(sysmem
, 0xc0000, &ram_lo
);
249 if (pcms
->above_4g_mem_size
> 0) {
250 memory_region_init_alias(&ram_hi
, NULL
, "xen.ram.hi",
251 &ram_memory
, 0x100000000ULL
,
252 pcms
->above_4g_mem_size
);
253 memory_region_add_subregion(sysmem
, 0x100000000ULL
, &ram_hi
);
257 void xen_ram_alloc(ram_addr_t ram_addr
, ram_addr_t size
, MemoryRegion
*mr
,
260 unsigned long nr_pfn
;
264 if (runstate_check(RUN_STATE_INMIGRATE
)) {
265 /* RAM already populated in Xen */
266 fprintf(stderr
, "%s: do not alloc "RAM_ADDR_FMT
267 " bytes of ram at "RAM_ADDR_FMT
" when runstate is INMIGRATE\n",
268 __func__
, size
, ram_addr
);
272 if (mr
== &ram_memory
) {
276 trace_xen_ram_alloc(ram_addr
, size
);
278 nr_pfn
= size
>> TARGET_PAGE_BITS
;
279 pfn_list
= g_malloc(sizeof (*pfn_list
) * nr_pfn
);
281 for (i
= 0; i
< nr_pfn
; i
++) {
282 pfn_list
[i
] = (ram_addr
>> TARGET_PAGE_BITS
) + i
;
285 if (xc_domain_populate_physmap_exact(xen_xc
, xen_domid
, nr_pfn
, 0, 0, pfn_list
)) {
286 error_setg(errp
, "xen: failed to populate ram at " RAM_ADDR_FMT
,
293 static XenPhysmap
*get_physmapping(hwaddr start_addr
, ram_addr_t size
)
295 XenPhysmap
*physmap
= NULL
;
297 start_addr
&= TARGET_PAGE_MASK
;
299 QLIST_FOREACH(physmap
, &xen_physmap
, list
) {
300 if (range_covers_byte(physmap
->start_addr
, physmap
->size
, start_addr
)) {
307 static hwaddr
xen_phys_offset_to_gaddr(hwaddr phys_offset
, ram_addr_t size
)
309 hwaddr addr
= phys_offset
& TARGET_PAGE_MASK
;
310 XenPhysmap
*physmap
= NULL
;
312 QLIST_FOREACH(physmap
, &xen_physmap
, list
) {
313 if (range_covers_byte(physmap
->phys_offset
, physmap
->size
, addr
)) {
314 return physmap
->start_addr
+ (phys_offset
- physmap
->phys_offset
);
321 #ifdef XEN_COMPAT_PHYSMAP
322 static int xen_save_physmap(XenIOState
*state
, XenPhysmap
*physmap
)
324 char path
[80], value
[17];
326 snprintf(path
, sizeof(path
),
327 "/local/domain/0/device-model/%d/physmap/%"PRIx64
"/start_addr",
328 xen_domid
, (uint64_t)physmap
->phys_offset
);
329 snprintf(value
, sizeof(value
), "%"PRIx64
, (uint64_t)physmap
->start_addr
);
330 if (!xs_write(state
->xenstore
, 0, path
, value
, strlen(value
))) {
333 snprintf(path
, sizeof(path
),
334 "/local/domain/0/device-model/%d/physmap/%"PRIx64
"/size",
335 xen_domid
, (uint64_t)physmap
->phys_offset
);
336 snprintf(value
, sizeof(value
), "%"PRIx64
, (uint64_t)physmap
->size
);
337 if (!xs_write(state
->xenstore
, 0, path
, value
, strlen(value
))) {
341 snprintf(path
, sizeof(path
),
342 "/local/domain/0/device-model/%d/physmap/%"PRIx64
"/name",
343 xen_domid
, (uint64_t)physmap
->phys_offset
);
344 if (!xs_write(state
->xenstore
, 0, path
,
345 physmap
->name
, strlen(physmap
->name
))) {
352 static int xen_save_physmap(XenIOState
*state
, XenPhysmap
*physmap
)
358 static int xen_add_to_physmap(XenIOState
*state
,
362 hwaddr offset_within_region
)
364 unsigned long nr_pages
;
366 XenPhysmap
*physmap
= NULL
;
367 hwaddr pfn
, start_gpfn
;
368 hwaddr phys_offset
= memory_region_get_ram_addr(mr
);
371 if (get_physmapping(start_addr
, size
)) {
378 /* Xen can only handle a single dirty log region for now and we want
379 * the linear framebuffer to be that region.
380 * Avoid tracking any regions that is not videoram and avoid tracking
381 * the legacy vga region. */
382 if (mr
== framebuffer
&& start_addr
> 0xbffff) {
388 DPRINTF("mapping vram to %"HWADDR_PRIx
" - %"HWADDR_PRIx
"\n",
389 start_addr
, start_addr
+ size
);
391 mr_name
= memory_region_name(mr
);
393 physmap
= g_malloc(sizeof(XenPhysmap
));
395 physmap
->start_addr
= start_addr
;
396 physmap
->size
= size
;
397 physmap
->name
= mr_name
;
398 physmap
->phys_offset
= phys_offset
;
400 QLIST_INSERT_HEAD(&xen_physmap
, physmap
, list
);
402 if (runstate_check(RUN_STATE_INMIGRATE
)) {
403 /* Now when we have a physmap entry we can replace a dummy mapping with
404 * a real one of guest foreign memory. */
405 uint8_t *p
= xen_replace_cache_entry(phys_offset
, start_addr
, size
);
406 assert(p
&& p
== memory_region_get_ram_ptr(mr
));
411 pfn
= phys_offset
>> TARGET_PAGE_BITS
;
412 start_gpfn
= start_addr
>> TARGET_PAGE_BITS
;
413 nr_pages
= size
>> TARGET_PAGE_BITS
;
414 rc
= xendevicemodel_relocate_memory(xen_dmod
, xen_domid
, nr_pages
, pfn
,
417 int saved_errno
= errno
;
419 error_report("relocate_memory %lu pages from GFN %"HWADDR_PRIx
420 " to GFN %"HWADDR_PRIx
" failed: %s",
421 nr_pages
, pfn
, start_gpfn
, strerror(saved_errno
));
426 rc
= xendevicemodel_pin_memory_cacheattr(xen_dmod
, xen_domid
,
427 start_addr
>> TARGET_PAGE_BITS
,
428 (start_addr
+ size
- 1) >> TARGET_PAGE_BITS
,
429 XEN_DOMCTL_MEM_CACHEATTR_WB
);
431 error_report("pin_memory_cacheattr failed: %s", strerror(errno
));
433 return xen_save_physmap(state
, physmap
);
436 static int xen_remove_from_physmap(XenIOState
*state
,
441 XenPhysmap
*physmap
= NULL
;
442 hwaddr phys_offset
= 0;
444 physmap
= get_physmapping(start_addr
, size
);
445 if (physmap
== NULL
) {
449 phys_offset
= physmap
->phys_offset
;
450 size
= physmap
->size
;
452 DPRINTF("unmapping vram to %"HWADDR_PRIx
" - %"HWADDR_PRIx
", at "
453 "%"HWADDR_PRIx
"\n", start_addr
, start_addr
+ size
, phys_offset
);
455 size
>>= TARGET_PAGE_BITS
;
456 start_addr
>>= TARGET_PAGE_BITS
;
457 phys_offset
>>= TARGET_PAGE_BITS
;
458 rc
= xendevicemodel_relocate_memory(xen_dmod
, xen_domid
, size
, start_addr
,
461 int saved_errno
= errno
;
463 error_report("relocate_memory "RAM_ADDR_FMT
" pages"
464 " from GFN %"HWADDR_PRIx
465 " to GFN %"HWADDR_PRIx
" failed: %s",
466 size
, start_addr
, phys_offset
, strerror(saved_errno
));
471 QLIST_REMOVE(physmap
, list
);
472 if (state
->log_for_dirtybit
== physmap
) {
473 state
->log_for_dirtybit
= NULL
;
474 g_free(state
->dirty_bitmap
);
475 state
->dirty_bitmap
= NULL
;
482 static void xen_set_memory(struct MemoryListener
*listener
,
483 MemoryRegionSection
*section
,
486 XenIOState
*state
= container_of(listener
, XenIOState
, memory_listener
);
487 hwaddr start_addr
= section
->offset_within_address_space
;
488 ram_addr_t size
= int128_get64(section
->size
);
489 bool log_dirty
= memory_region_is_logging(section
->mr
, DIRTY_MEMORY_VGA
);
490 hvmmem_type_t mem_type
;
492 if (section
->mr
== &ram_memory
) {
496 xen_map_memory_section(xen_domid
, state
->ioservid
,
499 xen_unmap_memory_section(xen_domid
, state
->ioservid
,
504 if (!memory_region_is_ram(section
->mr
)) {
508 if (log_dirty
!= add
) {
512 trace_xen_client_set_memory(start_addr
, size
, log_dirty
);
514 start_addr
&= TARGET_PAGE_MASK
;
515 size
= TARGET_PAGE_ALIGN(size
);
518 if (!memory_region_is_rom(section
->mr
)) {
519 xen_add_to_physmap(state
, start_addr
, size
,
520 section
->mr
, section
->offset_within_region
);
522 mem_type
= HVMMEM_ram_ro
;
523 if (xen_set_mem_type(xen_domid
, mem_type
,
524 start_addr
>> TARGET_PAGE_BITS
,
525 size
>> TARGET_PAGE_BITS
)) {
526 DPRINTF("xen_set_mem_type error, addr: "TARGET_FMT_plx
"\n",
531 if (xen_remove_from_physmap(state
, start_addr
, size
) < 0) {
532 DPRINTF("physmapping does not exist at "TARGET_FMT_plx
"\n", start_addr
);
537 static void xen_region_add(MemoryListener
*listener
,
538 MemoryRegionSection
*section
)
540 memory_region_ref(section
->mr
);
541 xen_set_memory(listener
, section
, true);
544 static void xen_region_del(MemoryListener
*listener
,
545 MemoryRegionSection
*section
)
547 xen_set_memory(listener
, section
, false);
548 memory_region_unref(section
->mr
);
551 static void xen_io_add(MemoryListener
*listener
,
552 MemoryRegionSection
*section
)
554 XenIOState
*state
= container_of(listener
, XenIOState
, io_listener
);
555 MemoryRegion
*mr
= section
->mr
;
557 if (mr
->ops
== &unassigned_io_ops
) {
561 memory_region_ref(mr
);
563 xen_map_io_section(xen_domid
, state
->ioservid
, section
);
566 static void xen_io_del(MemoryListener
*listener
,
567 MemoryRegionSection
*section
)
569 XenIOState
*state
= container_of(listener
, XenIOState
, io_listener
);
570 MemoryRegion
*mr
= section
->mr
;
572 if (mr
->ops
== &unassigned_io_ops
) {
576 xen_unmap_io_section(xen_domid
, state
->ioservid
, section
);
578 memory_region_unref(mr
);
581 static void xen_device_realize(DeviceListener
*listener
,
584 XenIOState
*state
= container_of(listener
, XenIOState
, device_listener
);
586 if (object_dynamic_cast(OBJECT(dev
), TYPE_PCI_DEVICE
)) {
587 PCIDevice
*pci_dev
= PCI_DEVICE(dev
);
588 XenPciDevice
*xendev
= g_new(XenPciDevice
, 1);
590 xendev
->pci_dev
= pci_dev
;
591 xendev
->sbdf
= PCI_BUILD_BDF(pci_dev_bus_num(pci_dev
),
593 QLIST_INSERT_HEAD(&state
->dev_list
, xendev
, entry
);
595 xen_map_pcidev(xen_domid
, state
->ioservid
, pci_dev
);
599 static void xen_device_unrealize(DeviceListener
*listener
,
602 XenIOState
*state
= container_of(listener
, XenIOState
, device_listener
);
604 if (object_dynamic_cast(OBJECT(dev
), TYPE_PCI_DEVICE
)) {
605 PCIDevice
*pci_dev
= PCI_DEVICE(dev
);
606 XenPciDevice
*xendev
, *next
;
608 xen_unmap_pcidev(xen_domid
, state
->ioservid
, pci_dev
);
610 QLIST_FOREACH_SAFE(xendev
, &state
->dev_list
, entry
, next
) {
611 if (xendev
->pci_dev
== pci_dev
) {
612 QLIST_REMOVE(xendev
, entry
);
620 static void xen_sync_dirty_bitmap(XenIOState
*state
,
624 hwaddr npages
= size
>> TARGET_PAGE_BITS
;
625 const int width
= sizeof(unsigned long) * 8;
626 size_t bitmap_size
= DIV_ROUND_UP(npages
, width
);
628 const XenPhysmap
*physmap
= NULL
;
630 physmap
= get_physmapping(start_addr
, size
);
631 if (physmap
== NULL
) {
636 if (state
->log_for_dirtybit
== NULL
) {
637 state
->log_for_dirtybit
= physmap
;
638 state
->dirty_bitmap
= g_new(unsigned long, bitmap_size
);
639 } else if (state
->log_for_dirtybit
!= physmap
) {
640 /* Only one range for dirty bitmap can be tracked. */
644 rc
= xen_track_dirty_vram(xen_domid
, start_addr
>> TARGET_PAGE_BITS
,
645 npages
, state
->dirty_bitmap
);
648 #define ENODATA ENOENT
650 if (errno
== ENODATA
) {
651 memory_region_set_dirty(framebuffer
, 0, size
);
652 DPRINTF("xen: track_dirty_vram failed (0x" TARGET_FMT_plx
653 ", 0x" TARGET_FMT_plx
"): %s\n",
654 start_addr
, start_addr
+ size
, strerror(errno
));
659 for (i
= 0; i
< bitmap_size
; i
++) {
660 unsigned long map
= state
->dirty_bitmap
[i
];
664 memory_region_set_dirty(framebuffer
,
665 (i
* width
+ j
) * TARGET_PAGE_SIZE
,
671 static void xen_log_start(MemoryListener
*listener
,
672 MemoryRegionSection
*section
,
675 XenIOState
*state
= container_of(listener
, XenIOState
, memory_listener
);
677 if (new & ~old
& (1 << DIRTY_MEMORY_VGA
)) {
678 xen_sync_dirty_bitmap(state
, section
->offset_within_address_space
,
679 int128_get64(section
->size
));
683 static void xen_log_stop(MemoryListener
*listener
, MemoryRegionSection
*section
,
686 XenIOState
*state
= container_of(listener
, XenIOState
, memory_listener
);
688 if (old
& ~new & (1 << DIRTY_MEMORY_VGA
)) {
689 state
->log_for_dirtybit
= NULL
;
690 g_free(state
->dirty_bitmap
);
691 state
->dirty_bitmap
= NULL
;
692 /* Disable dirty bit tracking */
693 xen_track_dirty_vram(xen_domid
, 0, 0, NULL
);
697 static void xen_log_sync(MemoryListener
*listener
, MemoryRegionSection
*section
)
699 XenIOState
*state
= container_of(listener
, XenIOState
, memory_listener
);
701 xen_sync_dirty_bitmap(state
, section
->offset_within_address_space
,
702 int128_get64(section
->size
));
705 static void xen_log_global_start(MemoryListener
*listener
)
708 xen_in_migration
= true;
712 static void xen_log_global_stop(MemoryListener
*listener
)
714 xen_in_migration
= false;
717 static MemoryListener xen_memory_listener
= {
718 .region_add
= xen_region_add
,
719 .region_del
= xen_region_del
,
720 .log_start
= xen_log_start
,
721 .log_stop
= xen_log_stop
,
722 .log_sync
= xen_log_sync
,
723 .log_global_start
= xen_log_global_start
,
724 .log_global_stop
= xen_log_global_stop
,
728 static MemoryListener xen_io_listener
= {
729 .region_add
= xen_io_add
,
730 .region_del
= xen_io_del
,
734 static DeviceListener xen_device_listener
= {
735 .realize
= xen_device_realize
,
736 .unrealize
= xen_device_unrealize
,
739 /* get the ioreq packets from share mem */
740 static ioreq_t
*cpu_get_ioreq_from_shared_memory(XenIOState
*state
, int vcpu
)
742 ioreq_t
*req
= xen_vcpu_ioreq(state
->shared_page
, vcpu
);
744 if (req
->state
!= STATE_IOREQ_READY
) {
745 DPRINTF("I/O request not ready: "
746 "%x, ptr: %x, port: %"PRIx64
", "
747 "data: %"PRIx64
", count: %u, size: %u\n",
748 req
->state
, req
->data_is_ptr
, req
->addr
,
749 req
->data
, req
->count
, req
->size
);
753 xen_rmb(); /* see IOREQ_READY /then/ read contents of ioreq */
755 req
->state
= STATE_IOREQ_INPROCESS
;
759 /* use poll to get the port notification */
760 /* ioreq_vec--out,the */
761 /* retval--the number of ioreq packet */
762 static ioreq_t
*cpu_get_ioreq(XenIOState
*state
)
764 MachineState
*ms
= MACHINE(qdev_get_machine());
765 unsigned int max_cpus
= ms
->smp
.max_cpus
;
769 port
= xenevtchn_pending(state
->xce_handle
);
770 if (port
== state
->bufioreq_local_port
) {
771 timer_mod(state
->buffered_io_timer
,
772 BUFFER_IO_MAX_DELAY
+ qemu_clock_get_ms(QEMU_CLOCK_REALTIME
));
777 for (i
= 0; i
< max_cpus
; i
++) {
778 if (state
->ioreq_local_port
[i
] == port
) {
784 hw_error("Fatal error while trying to get io event!\n");
787 /* unmask the wanted port again */
788 xenevtchn_unmask(state
->xce_handle
, port
);
790 /* get the io packet from shared memory */
791 state
->send_vcpu
= i
;
792 return cpu_get_ioreq_from_shared_memory(state
, i
);
795 /* read error or read nothing */
799 static uint32_t do_inp(uint32_t addr
, unsigned long size
)
803 return cpu_inb(addr
);
805 return cpu_inw(addr
);
807 return cpu_inl(addr
);
809 hw_error("inp: bad size: %04x %lx", addr
, size
);
813 static void do_outp(uint32_t addr
,
814 unsigned long size
, uint32_t val
)
818 return cpu_outb(addr
, val
);
820 return cpu_outw(addr
, val
);
822 return cpu_outl(addr
, val
);
824 hw_error("outp: bad size: %04x %lx", addr
, size
);
829 * Helper functions which read/write an object from/to physical guest
830 * memory, as part of the implementation of an ioreq.
833 * cpu_physical_memory_rw(addr + (req->df ? -1 : +1) * req->size * i,
834 * val, req->size, 0/1)
835 * except without the integer overflow problems.
837 static void rw_phys_req_item(hwaddr addr
,
838 ioreq_t
*req
, uint32_t i
, void *val
, int rw
)
840 /* Do everything unsigned so overflow just results in a truncated result
841 * and accesses to undesired parts of guest memory, which is up
843 hwaddr offset
= (hwaddr
)req
->size
* i
;
849 cpu_physical_memory_rw(addr
, val
, req
->size
, rw
);
852 static inline void read_phys_req_item(hwaddr addr
,
853 ioreq_t
*req
, uint32_t i
, void *val
)
855 rw_phys_req_item(addr
, req
, i
, val
, 0);
857 static inline void write_phys_req_item(hwaddr addr
,
858 ioreq_t
*req
, uint32_t i
, void *val
)
860 rw_phys_req_item(addr
, req
, i
, val
, 1);
864 static void cpu_ioreq_pio(ioreq_t
*req
)
868 trace_cpu_ioreq_pio(req
, req
->dir
, req
->df
, req
->data_is_ptr
, req
->addr
,
869 req
->data
, req
->count
, req
->size
);
871 if (req
->size
> sizeof(uint32_t)) {
872 hw_error("PIO: bad size (%u)", req
->size
);
875 if (req
->dir
== IOREQ_READ
) {
876 if (!req
->data_is_ptr
) {
877 req
->data
= do_inp(req
->addr
, req
->size
);
878 trace_cpu_ioreq_pio_read_reg(req
, req
->data
, req
->addr
,
883 for (i
= 0; i
< req
->count
; i
++) {
884 tmp
= do_inp(req
->addr
, req
->size
);
885 write_phys_req_item(req
->data
, req
, i
, &tmp
);
888 } else if (req
->dir
== IOREQ_WRITE
) {
889 if (!req
->data_is_ptr
) {
890 trace_cpu_ioreq_pio_write_reg(req
, req
->data
, req
->addr
,
892 do_outp(req
->addr
, req
->size
, req
->data
);
894 for (i
= 0; i
< req
->count
; i
++) {
897 read_phys_req_item(req
->data
, req
, i
, &tmp
);
898 do_outp(req
->addr
, req
->size
, tmp
);
904 static void cpu_ioreq_move(ioreq_t
*req
)
908 trace_cpu_ioreq_move(req
, req
->dir
, req
->df
, req
->data_is_ptr
, req
->addr
,
909 req
->data
, req
->count
, req
->size
);
911 if (req
->size
> sizeof(req
->data
)) {
912 hw_error("MMIO: bad size (%u)", req
->size
);
915 if (!req
->data_is_ptr
) {
916 if (req
->dir
== IOREQ_READ
) {
917 for (i
= 0; i
< req
->count
; i
++) {
918 read_phys_req_item(req
->addr
, req
, i
, &req
->data
);
920 } else if (req
->dir
== IOREQ_WRITE
) {
921 for (i
= 0; i
< req
->count
; i
++) {
922 write_phys_req_item(req
->addr
, req
, i
, &req
->data
);
928 if (req
->dir
== IOREQ_READ
) {
929 for (i
= 0; i
< req
->count
; i
++) {
930 read_phys_req_item(req
->addr
, req
, i
, &tmp
);
931 write_phys_req_item(req
->data
, req
, i
, &tmp
);
933 } else if (req
->dir
== IOREQ_WRITE
) {
934 for (i
= 0; i
< req
->count
; i
++) {
935 read_phys_req_item(req
->data
, req
, i
, &tmp
);
936 write_phys_req_item(req
->addr
, req
, i
, &tmp
);
942 static void cpu_ioreq_config(XenIOState
*state
, ioreq_t
*req
)
944 uint32_t sbdf
= req
->addr
>> 32;
945 uint32_t reg
= req
->addr
;
946 XenPciDevice
*xendev
;
948 if (req
->size
!= sizeof(uint8_t) && req
->size
!= sizeof(uint16_t) &&
949 req
->size
!= sizeof(uint32_t)) {
950 hw_error("PCI config access: bad size (%u)", req
->size
);
953 if (req
->count
!= 1) {
954 hw_error("PCI config access: bad count (%u)", req
->count
);
957 QLIST_FOREACH(xendev
, &state
->dev_list
, entry
) {
958 if (xendev
->sbdf
!= sbdf
) {
962 if (!req
->data_is_ptr
) {
963 if (req
->dir
== IOREQ_READ
) {
964 req
->data
= pci_host_config_read_common(
965 xendev
->pci_dev
, reg
, PCI_CONFIG_SPACE_SIZE
,
967 trace_cpu_ioreq_config_read(req
, xendev
->sbdf
, reg
,
968 req
->size
, req
->data
);
969 } else if (req
->dir
== IOREQ_WRITE
) {
970 trace_cpu_ioreq_config_write(req
, xendev
->sbdf
, reg
,
971 req
->size
, req
->data
);
972 pci_host_config_write_common(
973 xendev
->pci_dev
, reg
, PCI_CONFIG_SPACE_SIZE
,
974 req
->data
, req
->size
);
979 if (req
->dir
== IOREQ_READ
) {
980 tmp
= pci_host_config_read_common(
981 xendev
->pci_dev
, reg
, PCI_CONFIG_SPACE_SIZE
,
983 trace_cpu_ioreq_config_read(req
, xendev
->sbdf
, reg
,
985 write_phys_req_item(req
->data
, req
, 0, &tmp
);
986 } else if (req
->dir
== IOREQ_WRITE
) {
987 read_phys_req_item(req
->data
, req
, 0, &tmp
);
988 trace_cpu_ioreq_config_write(req
, xendev
->sbdf
, reg
,
990 pci_host_config_write_common(
991 xendev
->pci_dev
, reg
, PCI_CONFIG_SPACE_SIZE
,
998 static void regs_to_cpu(vmware_regs_t
*vmport_regs
, ioreq_t
*req
)
1003 cpu
= X86_CPU(current_cpu
);
1005 env
->regs
[R_EAX
] = req
->data
;
1006 env
->regs
[R_EBX
] = vmport_regs
->ebx
;
1007 env
->regs
[R_ECX
] = vmport_regs
->ecx
;
1008 env
->regs
[R_EDX
] = vmport_regs
->edx
;
1009 env
->regs
[R_ESI
] = vmport_regs
->esi
;
1010 env
->regs
[R_EDI
] = vmport_regs
->edi
;
1013 static void regs_from_cpu(vmware_regs_t
*vmport_regs
)
1015 X86CPU
*cpu
= X86_CPU(current_cpu
);
1016 CPUX86State
*env
= &cpu
->env
;
1018 vmport_regs
->ebx
= env
->regs
[R_EBX
];
1019 vmport_regs
->ecx
= env
->regs
[R_ECX
];
1020 vmport_regs
->edx
= env
->regs
[R_EDX
];
1021 vmport_regs
->esi
= env
->regs
[R_ESI
];
1022 vmport_regs
->edi
= env
->regs
[R_EDI
];
1025 static void handle_vmport_ioreq(XenIOState
*state
, ioreq_t
*req
)
1027 vmware_regs_t
*vmport_regs
;
1029 assert(state
->shared_vmport_page
);
1031 &state
->shared_vmport_page
->vcpu_vmport_regs
[state
->send_vcpu
];
1032 QEMU_BUILD_BUG_ON(sizeof(*req
) < sizeof(*vmport_regs
));
1034 current_cpu
= state
->cpu_by_vcpu_id
[state
->send_vcpu
];
1035 regs_to_cpu(vmport_regs
, req
);
1037 regs_from_cpu(vmport_regs
);
1041 static void handle_ioreq(XenIOState
*state
, ioreq_t
*req
)
1043 trace_handle_ioreq(req
, req
->type
, req
->dir
, req
->df
, req
->data_is_ptr
,
1044 req
->addr
, req
->data
, req
->count
, req
->size
);
1046 if (!req
->data_is_ptr
&& (req
->dir
== IOREQ_WRITE
) &&
1047 (req
->size
< sizeof (target_ulong
))) {
1048 req
->data
&= ((target_ulong
) 1 << (8 * req
->size
)) - 1;
1051 if (req
->dir
== IOREQ_WRITE
)
1052 trace_handle_ioreq_write(req
, req
->type
, req
->df
, req
->data_is_ptr
,
1053 req
->addr
, req
->data
, req
->count
, req
->size
);
1055 switch (req
->type
) {
1056 case IOREQ_TYPE_PIO
:
1059 case IOREQ_TYPE_COPY
:
1060 cpu_ioreq_move(req
);
1062 case IOREQ_TYPE_VMWARE_PORT
:
1063 handle_vmport_ioreq(state
, req
);
1065 case IOREQ_TYPE_TIMEOFFSET
:
1067 case IOREQ_TYPE_INVALIDATE
:
1068 xen_invalidate_map_cache();
1070 case IOREQ_TYPE_PCI_CONFIG
:
1071 cpu_ioreq_config(state
, req
);
1074 hw_error("Invalid ioreq type 0x%x\n", req
->type
);
1076 if (req
->dir
== IOREQ_READ
) {
1077 trace_handle_ioreq_read(req
, req
->type
, req
->df
, req
->data_is_ptr
,
1078 req
->addr
, req
->data
, req
->count
, req
->size
);
1082 static int handle_buffered_iopage(XenIOState
*state
)
1084 buffered_iopage_t
*buf_page
= state
->buffered_io_page
;
1085 buf_ioreq_t
*buf_req
= NULL
;
1093 memset(&req
, 0x00, sizeof(req
));
1094 req
.state
= STATE_IOREQ_READY
;
1096 req
.dir
= IOREQ_WRITE
;
1099 uint32_t rdptr
= buf_page
->read_pointer
, wrptr
;
1102 wrptr
= buf_page
->write_pointer
;
1104 if (rdptr
!= buf_page
->read_pointer
) {
1107 if (rdptr
== wrptr
) {
1110 buf_req
= &buf_page
->buf_ioreq
[rdptr
% IOREQ_BUFFER_SLOT_NUM
];
1111 req
.size
= 1U << buf_req
->size
;
1112 req
.addr
= buf_req
->addr
;
1113 req
.data
= buf_req
->data
;
1114 req
.type
= buf_req
->type
;
1116 qw
= (req
.size
== 8);
1118 if (rdptr
+ 1 == wrptr
) {
1119 hw_error("Incomplete quad word buffered ioreq");
1121 buf_req
= &buf_page
->buf_ioreq
[(rdptr
+ 1) %
1122 IOREQ_BUFFER_SLOT_NUM
];
1123 req
.data
|= ((uint64_t)buf_req
->data
) << 32;
1127 handle_ioreq(state
, &req
);
1129 /* Only req.data may get updated by handle_ioreq(), albeit even that
1130 * should not happen as such data would never make it to the guest (we
1131 * can only usefully see writes here after all).
1133 assert(req
.state
== STATE_IOREQ_READY
);
1134 assert(req
.count
== 1);
1135 assert(req
.dir
== IOREQ_WRITE
);
1136 assert(!req
.data_is_ptr
);
1138 atomic_add(&buf_page
->read_pointer
, qw
+ 1);
1144 static void handle_buffered_io(void *opaque
)
1146 XenIOState
*state
= opaque
;
1148 if (handle_buffered_iopage(state
)) {
1149 timer_mod(state
->buffered_io_timer
,
1150 BUFFER_IO_MAX_DELAY
+ qemu_clock_get_ms(QEMU_CLOCK_REALTIME
));
1152 timer_del(state
->buffered_io_timer
);
1153 xenevtchn_unmask(state
->xce_handle
, state
->bufioreq_local_port
);
1157 static void cpu_handle_ioreq(void *opaque
)
1159 XenIOState
*state
= opaque
;
1160 ioreq_t
*req
= cpu_get_ioreq(state
);
1162 handle_buffered_iopage(state
);
1164 ioreq_t copy
= *req
;
1167 handle_ioreq(state
, ©
);
1168 req
->data
= copy
.data
;
1170 if (req
->state
!= STATE_IOREQ_INPROCESS
) {
1171 fprintf(stderr
, "Badness in I/O request ... not in service?!: "
1172 "%x, ptr: %x, port: %"PRIx64
", "
1173 "data: %"PRIx64
", count: %u, size: %u, type: %u\n",
1174 req
->state
, req
->data_is_ptr
, req
->addr
,
1175 req
->data
, req
->count
, req
->size
, req
->type
);
1176 destroy_hvm_domain(false);
1180 xen_wmb(); /* Update ioreq contents /then/ update state. */
1183 * We do this before we send the response so that the tools
1184 * have the opportunity to pick up on the reset before the
1185 * guest resumes and does a hlt with interrupts disabled which
1186 * causes Xen to powerdown the domain.
1188 if (runstate_is_running()) {
1189 ShutdownCause request
;
1191 if (qemu_shutdown_requested_get()) {
1192 destroy_hvm_domain(false);
1194 request
= qemu_reset_requested_get();
1196 qemu_system_reset(request
);
1197 destroy_hvm_domain(true);
1201 req
->state
= STATE_IORESP_READY
;
1202 xenevtchn_notify(state
->xce_handle
,
1203 state
->ioreq_local_port
[state
->send_vcpu
]);
1207 static void xen_main_loop_prepare(XenIOState
*state
)
1211 if (state
->xce_handle
!= NULL
) {
1212 evtchn_fd
= xenevtchn_fd(state
->xce_handle
);
1215 state
->buffered_io_timer
= timer_new_ms(QEMU_CLOCK_REALTIME
, handle_buffered_io
,
1218 if (evtchn_fd
!= -1) {
1219 CPUState
*cpu_state
;
1221 DPRINTF("%s: Init cpu_by_vcpu_id\n", __func__
);
1222 CPU_FOREACH(cpu_state
) {
1223 DPRINTF("%s: cpu_by_vcpu_id[%d]=%p\n",
1224 __func__
, cpu_state
->cpu_index
, cpu_state
);
1225 state
->cpu_by_vcpu_id
[cpu_state
->cpu_index
] = cpu_state
;
1227 qemu_set_fd_handler(evtchn_fd
, cpu_handle_ioreq
, NULL
, state
);
1232 static void xen_hvm_change_state_handler(void *opaque
, int running
,
1235 XenIOState
*state
= opaque
;
1238 xen_main_loop_prepare(state
);
1241 xen_set_ioreq_server_state(xen_domid
,
1243 (rstate
== RUN_STATE_RUNNING
));
1246 static void xen_exit_notifier(Notifier
*n
, void *data
)
1248 XenIOState
*state
= container_of(n
, XenIOState
, exit
);
1250 xen_destroy_ioreq_server(xen_domid
, state
->ioservid
);
1252 xenevtchn_close(state
->xce_handle
);
1253 xs_daemon_close(state
->xenstore
);
1256 #ifdef XEN_COMPAT_PHYSMAP
1257 static void xen_read_physmap(XenIOState
*state
)
1259 XenPhysmap
*physmap
= NULL
;
1260 unsigned int len
, num
, i
;
1261 char path
[80], *value
= NULL
;
1262 char **entries
= NULL
;
1264 snprintf(path
, sizeof(path
),
1265 "/local/domain/0/device-model/%d/physmap", xen_domid
);
1266 entries
= xs_directory(state
->xenstore
, 0, path
, &num
);
1267 if (entries
== NULL
)
1270 for (i
= 0; i
< num
; i
++) {
1271 physmap
= g_malloc(sizeof (XenPhysmap
));
1272 physmap
->phys_offset
= strtoull(entries
[i
], NULL
, 16);
1273 snprintf(path
, sizeof(path
),
1274 "/local/domain/0/device-model/%d/physmap/%s/start_addr",
1275 xen_domid
, entries
[i
]);
1276 value
= xs_read(state
->xenstore
, 0, path
, &len
);
1277 if (value
== NULL
) {
1281 physmap
->start_addr
= strtoull(value
, NULL
, 16);
1284 snprintf(path
, sizeof(path
),
1285 "/local/domain/0/device-model/%d/physmap/%s/size",
1286 xen_domid
, entries
[i
]);
1287 value
= xs_read(state
->xenstore
, 0, path
, &len
);
1288 if (value
== NULL
) {
1292 physmap
->size
= strtoull(value
, NULL
, 16);
1295 snprintf(path
, sizeof(path
),
1296 "/local/domain/0/device-model/%d/physmap/%s/name",
1297 xen_domid
, entries
[i
]);
1298 physmap
->name
= xs_read(state
->xenstore
, 0, path
, &len
);
1300 QLIST_INSERT_HEAD(&xen_physmap
, physmap
, list
);
1305 static void xen_read_physmap(XenIOState
*state
)
1310 static void xen_wakeup_notifier(Notifier
*notifier
, void *data
)
1312 xc_set_hvm_param(xen_xc
, xen_domid
, HVM_PARAM_ACPI_S_STATE
, 0);
1315 static int xen_map_ioreq_server(XenIOState
*state
)
1318 xenforeignmemory_resource_handle
*fres
;
1319 xen_pfn_t ioreq_pfn
;
1320 xen_pfn_t bufioreq_pfn
;
1321 evtchn_port_t bufioreq_evtchn
;
1325 * Attempt to map using the resource API and fall back to normal
1326 * foreign mapping if this is not supported.
1328 QEMU_BUILD_BUG_ON(XENMEM_resource_ioreq_server_frame_bufioreq
!= 0);
1329 QEMU_BUILD_BUG_ON(XENMEM_resource_ioreq_server_frame_ioreq(0) != 1);
1330 fres
= xenforeignmemory_map_resource(xen_fmem
, xen_domid
,
1331 XENMEM_resource_ioreq_server
,
1332 state
->ioservid
, 0, 2,
1334 PROT_READ
| PROT_WRITE
, 0);
1336 trace_xen_map_resource_ioreq(state
->ioservid
, addr
);
1337 state
->buffered_io_page
= addr
;
1338 state
->shared_page
= addr
+ TARGET_PAGE_SIZE
;
1339 } else if (errno
!= EOPNOTSUPP
) {
1340 error_report("failed to map ioreq server resources: error %d handle=%p",
1345 rc
= xen_get_ioreq_server_info(xen_domid
, state
->ioservid
,
1346 (state
->shared_page
== NULL
) ?
1348 (state
->buffered_io_page
== NULL
) ?
1349 &bufioreq_pfn
: NULL
,
1352 error_report("failed to get ioreq server info: error %d handle=%p",
1357 if (state
->shared_page
== NULL
) {
1358 DPRINTF("shared page at pfn %lx\n", ioreq_pfn
);
1360 state
->shared_page
= xenforeignmemory_map(xen_fmem
, xen_domid
,
1361 PROT_READ
| PROT_WRITE
,
1362 1, &ioreq_pfn
, NULL
);
1363 if (state
->shared_page
== NULL
) {
1364 error_report("map shared IO page returned error %d handle=%p",
1369 if (state
->buffered_io_page
== NULL
) {
1370 DPRINTF("buffered io page at pfn %lx\n", bufioreq_pfn
);
1372 state
->buffered_io_page
= xenforeignmemory_map(xen_fmem
, xen_domid
,
1373 PROT_READ
| PROT_WRITE
,
1376 if (state
->buffered_io_page
== NULL
) {
1377 error_report("map buffered IO page returned error %d", errno
);
1382 if (state
->shared_page
== NULL
|| state
->buffered_io_page
== NULL
) {
1386 DPRINTF("buffered io evtchn is %x\n", bufioreq_evtchn
);
1388 state
->bufioreq_remote_port
= bufioreq_evtchn
;
1393 void xen_hvm_init(PCMachineState
*pcms
, MemoryRegion
**ram_memory
)
1395 MachineState
*ms
= MACHINE(pcms
);
1396 unsigned int max_cpus
= ms
->smp
.max_cpus
;
1398 xen_pfn_t ioreq_pfn
;
1401 state
= g_malloc0(sizeof (XenIOState
));
1403 state
->xce_handle
= xenevtchn_open(NULL
, 0);
1404 if (state
->xce_handle
== NULL
) {
1405 perror("xen: event channel open");
1409 state
->xenstore
= xs_daemon_open();
1410 if (state
->xenstore
== NULL
) {
1411 perror("xen: xenstore open");
1415 xen_create_ioreq_server(xen_domid
, &state
->ioservid
);
1417 state
->exit
.notify
= xen_exit_notifier
;
1418 qemu_add_exit_notifier(&state
->exit
);
1420 state
->suspend
.notify
= xen_suspend_notifier
;
1421 qemu_register_suspend_notifier(&state
->suspend
);
1423 state
->wakeup
.notify
= xen_wakeup_notifier
;
1424 qemu_register_wakeup_notifier(&state
->wakeup
);
1427 * Register wake-up support in QMP query-current-machine API
1429 qemu_register_wakeup_support();
1431 rc
= xen_map_ioreq_server(state
);
1436 rc
= xen_get_vmport_regs_pfn(xen_xc
, xen_domid
, &ioreq_pfn
);
1438 DPRINTF("shared vmport page at pfn %lx\n", ioreq_pfn
);
1439 state
->shared_vmport_page
=
1440 xenforeignmemory_map(xen_fmem
, xen_domid
, PROT_READ
|PROT_WRITE
,
1441 1, &ioreq_pfn
, NULL
);
1442 if (state
->shared_vmport_page
== NULL
) {
1443 error_report("map shared vmport IO page returned error %d handle=%p",
1447 } else if (rc
!= -ENOSYS
) {
1448 error_report("get vmport regs pfn returned error %d, rc=%d",
1453 /* Note: cpus is empty at this point in init */
1454 state
->cpu_by_vcpu_id
= g_malloc0(max_cpus
* sizeof(CPUState
*));
1456 rc
= xen_set_ioreq_server_state(xen_domid
, state
->ioservid
, true);
1458 error_report("failed to enable ioreq server info: error %d handle=%p",
1463 state
->ioreq_local_port
= g_malloc0(max_cpus
* sizeof (evtchn_port_t
));
1465 /* FIXME: how about if we overflow the page here? */
1466 for (i
= 0; i
< max_cpus
; i
++) {
1467 rc
= xenevtchn_bind_interdomain(state
->xce_handle
, xen_domid
,
1468 xen_vcpu_eport(state
->shared_page
, i
));
1470 error_report("shared evtchn %d bind error %d", i
, errno
);
1473 state
->ioreq_local_port
[i
] = rc
;
1476 rc
= xenevtchn_bind_interdomain(state
->xce_handle
, xen_domid
,
1477 state
->bufioreq_remote_port
);
1479 error_report("buffered evtchn bind error %d", errno
);
1482 state
->bufioreq_local_port
= rc
;
1484 /* Init RAM management */
1485 #ifdef XEN_COMPAT_PHYSMAP
1486 xen_map_cache_init(xen_phys_offset_to_gaddr
, state
);
1488 xen_map_cache_init(NULL
, state
);
1490 xen_ram_init(pcms
, ram_size
, ram_memory
);
1492 qemu_add_vm_change_state_handler(xen_hvm_change_state_handler
, state
);
1494 state
->memory_listener
= xen_memory_listener
;
1495 memory_listener_register(&state
->memory_listener
, &address_space_memory
);
1496 state
->log_for_dirtybit
= NULL
;
1498 state
->io_listener
= xen_io_listener
;
1499 memory_listener_register(&state
->io_listener
, &address_space_io
);
1501 state
->device_listener
= xen_device_listener
;
1502 QLIST_INIT(&state
->dev_list
);
1503 device_listener_register(&state
->device_listener
);
1507 /* Initialize backend core & drivers */
1508 if (xen_be_init() != 0) {
1509 error_report("xen backend core setup failed");
1512 xen_be_register_common();
1514 QLIST_INIT(&xen_physmap
);
1515 xen_read_physmap(state
);
1517 /* Disable ACPI build because Xen handles it */
1518 pcms
->acpi_build_enabled
= false;
1523 error_report("xen hardware virtual machine initialisation failed");
1527 void destroy_hvm_domain(bool reboot
)
1529 xc_interface
*xc_handle
;
1533 unsigned int reason
= reboot
? SHUTDOWN_reboot
: SHUTDOWN_poweroff
;
1536 rc
= xendevicemodel_shutdown(xen_dmod
, xen_domid
, reason
);
1540 if (errno
!= ENOTTY
/* old Xen */) {
1541 perror("xendevicemodel_shutdown failed");
1543 /* well, try the old thing then */
1546 xc_handle
= xc_interface_open(0, 0, 0);
1547 if (xc_handle
== NULL
) {
1548 fprintf(stderr
, "Cannot acquire xenctrl handle\n");
1550 sts
= xc_domain_shutdown(xc_handle
, xen_domid
, reason
);
1552 fprintf(stderr
, "xc_domain_shutdown failed to issue %s, "
1553 "sts %d, %s\n", reboot
? "reboot" : "poweroff",
1554 sts
, strerror(errno
));
1556 fprintf(stderr
, "Issued domain %d %s\n", xen_domid
,
1557 reboot
? "reboot" : "poweroff");
1559 xc_interface_close(xc_handle
);
1563 void xen_register_framebuffer(MemoryRegion
*mr
)
1568 void xen_shutdown_fatal_error(const char *fmt
, ...)
1573 vfprintf(stderr
, fmt
, ap
);
1575 fprintf(stderr
, "Will destroy the domain.\n");
1576 /* destroy the domain */
1577 qemu_system_shutdown_request(SHUTDOWN_CAUSE_HOST_ERROR
);
1580 void xen_hvm_modified_memory(ram_addr_t start
, ram_addr_t length
)
1582 if (unlikely(xen_in_migration
)) {
1584 ram_addr_t start_pfn
, nb_pages
;
1586 start
= xen_phys_offset_to_gaddr(start
, length
);
1589 length
= TARGET_PAGE_SIZE
;
1591 start_pfn
= start
>> TARGET_PAGE_BITS
;
1592 nb_pages
= ((start
+ length
+ TARGET_PAGE_SIZE
- 1) >> TARGET_PAGE_BITS
)
1594 rc
= xen_modified_memory(xen_domid
, start_pfn
, nb_pages
);
1597 "%s failed for "RAM_ADDR_FMT
" ("RAM_ADDR_FMT
"): %i, %s\n",
1598 __func__
, start
, nb_pages
, errno
, strerror(errno
));
1603 void qmp_xen_set_global_dirty_log(bool enable
, Error
**errp
)
1606 memory_global_dirty_log_start();
1608 memory_global_dirty_log_stop();