4 * Copyright IBM, Corp. 2012
7 * Christian Borntraeger <borntraeger@de.ibm.com>
8 * Heinz Graalfs <graalfs@linux.vnet.ibm.com>
10 * This work is licensed under the terms of the GNU GPL, version 2 or (at your
11 * option) any later version. See the COPYING file in the top-level directory.
15 #include "qemu/osdep.h"
16 #include "qapi/error.h"
18 #include "sysemu/kvm.h"
19 #include "exec/memory.h"
20 #include "sysemu/sysemu.h"
21 #include "exec/address-spaces.h"
22 #include "hw/boards.h"
23 #include "hw/s390x/sclp.h"
24 #include "hw/s390x/event-facility.h"
25 #include "hw/s390x/s390-pci-bus.h"
27 static inline SCLPDevice
*get_sclp_device(void)
29 return SCLP(object_resolve_path_type("", TYPE_SCLP
, NULL
));
32 /* Provide information about the configuration, CPUs and storage */
33 static void read_SCP_info(SCLPDevice
*sclp
, SCCB
*sccb
)
35 ReadInfo
*read_info
= (ReadInfo
*) sccb
;
36 MachineState
*machine
= MACHINE(qdev_get_machine());
37 sclpMemoryHotplugDev
*mhd
= get_sclp_memory_hotplug_dev();
42 int slots
= MIN(machine
->ram_slots
, s390_get_memslot_count(kvm_state
));
49 read_info
->entries_cpu
= cpu_to_be16(cpu_count
);
50 read_info
->offset_cpu
= cpu_to_be16(offsetof(ReadInfo
, entries
));
51 read_info
->highest_cpu
= cpu_to_be16(max_cpus
);
53 for (i
= 0; i
< cpu_count
; i
++) {
54 read_info
->entries
[i
].address
= i
;
55 read_info
->entries
[i
].type
= 0;
58 read_info
->facilities
= cpu_to_be64(SCLP_HAS_CPU_INFO
|
59 SCLP_HAS_PCI_RECONFIG
);
61 /* Memory Hotplug is only supported for the ccw machine type */
63 mhd
->standby_subregion_size
= MEM_SECTION_SIZE
;
64 /* Deduct the memory slot already used for core */
66 while ((mhd
->standby_subregion_size
* (slots
- 1)
67 < mhd
->standby_mem_size
)) {
68 mhd
->standby_subregion_size
= mhd
->standby_subregion_size
<< 1;
72 * Initialize mapping of guest standby memory sections indicating which
73 * are and are not online. Assume all standby memory begins offline.
75 if (mhd
->standby_state_map
== 0) {
76 if (mhd
->standby_mem_size
% mhd
->standby_subregion_size
) {
77 mhd
->standby_state_map
= g_malloc0((mhd
->standby_mem_size
/
78 mhd
->standby_subregion_size
+ 1) *
79 (mhd
->standby_subregion_size
/
82 mhd
->standby_state_map
= g_malloc0(mhd
->standby_mem_size
/
86 mhd
->padded_ram_size
= ram_size
+ mhd
->pad_size
;
87 mhd
->rzm
= 1 << mhd
->increment_size
;
89 read_info
->facilities
|= cpu_to_be64(SCLP_FC_ASSIGN_ATTACH_READ_STOR
);
92 rnsize
= 1 << (sclp
->increment_size
- 20);
94 read_info
->rnsize
= rnsize
;
96 read_info
->rnsize
= 0;
97 read_info
->rnsize2
= cpu_to_be32(rnsize
);
100 rnmax
= machine
->maxram_size
>> sclp
->increment_size
;
101 if (rnmax
< 0x10000) {
102 read_info
->rnmax
= cpu_to_be16(rnmax
);
104 read_info
->rnmax
= cpu_to_be16(0);
105 read_info
->rnmax2
= cpu_to_be64(rnmax
);
108 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_NORMAL_READ_COMPLETION
);
111 static void read_storage_element0_info(SCLPDevice
*sclp
, SCCB
*sccb
)
114 int subincrement_id
= SCLP_STARTING_SUBINCREMENT_ID
;
115 ReadStorageElementInfo
*storage_info
= (ReadStorageElementInfo
*) sccb
;
116 sclpMemoryHotplugDev
*mhd
= get_sclp_memory_hotplug_dev();
119 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_INVALID_SCLP_COMMAND
);
123 if ((ram_size
>> mhd
->increment_size
) >= 0x10000) {
124 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_SCCB_BOUNDARY_VIOLATION
);
128 /* Return information regarding core memory */
129 storage_info
->max_id
= cpu_to_be16(mhd
->standby_mem_size
? 1 : 0);
130 assigned
= ram_size
>> mhd
->increment_size
;
131 storage_info
->assigned
= cpu_to_be16(assigned
);
133 for (i
= 0; i
< assigned
; i
++) {
134 storage_info
->entries
[i
] = cpu_to_be32(subincrement_id
);
135 subincrement_id
+= SCLP_INCREMENT_UNIT
;
137 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_NORMAL_READ_COMPLETION
);
140 static void read_storage_element1_info(SCLPDevice
*sclp
, SCCB
*sccb
)
142 ReadStorageElementInfo
*storage_info
= (ReadStorageElementInfo
*) sccb
;
143 sclpMemoryHotplugDev
*mhd
= get_sclp_memory_hotplug_dev();
146 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_INVALID_SCLP_COMMAND
);
150 if ((mhd
->standby_mem_size
>> mhd
->increment_size
) >= 0x10000) {
151 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_SCCB_BOUNDARY_VIOLATION
);
155 /* Return information regarding standby memory */
156 storage_info
->max_id
= cpu_to_be16(mhd
->standby_mem_size
? 1 : 0);
157 storage_info
->assigned
= cpu_to_be16(mhd
->standby_mem_size
>>
158 mhd
->increment_size
);
159 storage_info
->standby
= cpu_to_be16(mhd
->standby_mem_size
>>
160 mhd
->increment_size
);
161 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_STANDBY_READ_COMPLETION
);
164 static void attach_storage_element(SCLPDevice
*sclp
, SCCB
*sccb
,
167 int i
, assigned
, subincrement_id
;
168 AttachStorageElement
*attach_info
= (AttachStorageElement
*) sccb
;
169 sclpMemoryHotplugDev
*mhd
= get_sclp_memory_hotplug_dev();
172 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_INVALID_SCLP_COMMAND
);
177 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_INVALID_SCLP_COMMAND
);
181 assigned
= mhd
->standby_mem_size
>> mhd
->increment_size
;
182 attach_info
->assigned
= cpu_to_be16(assigned
);
183 subincrement_id
= ((ram_size
>> mhd
->increment_size
) << 16)
184 + SCLP_STARTING_SUBINCREMENT_ID
;
185 for (i
= 0; i
< assigned
; i
++) {
186 attach_info
->entries
[i
] = cpu_to_be32(subincrement_id
);
187 subincrement_id
+= SCLP_INCREMENT_UNIT
;
189 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_NORMAL_COMPLETION
);
192 static void assign_storage(SCLPDevice
*sclp
, SCCB
*sccb
)
194 MemoryRegion
*mr
= NULL
;
195 uint64_t this_subregion_size
;
196 AssignStorage
*assign_info
= (AssignStorage
*) sccb
;
197 sclpMemoryHotplugDev
*mhd
= get_sclp_memory_hotplug_dev();
198 ram_addr_t assign_addr
;
199 MemoryRegion
*sysmem
= get_system_memory();
202 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_INVALID_SCLP_COMMAND
);
205 assign_addr
= (assign_info
->rn
- 1) * mhd
->rzm
;
207 if ((assign_addr
% MEM_SECTION_SIZE
== 0) &&
208 (assign_addr
>= mhd
->padded_ram_size
)) {
209 /* Re-use existing memory region if found */
210 mr
= memory_region_find(sysmem
, assign_addr
, 1).mr
;
211 memory_region_unref(mr
);
214 MemoryRegion
*standby_ram
= g_new(MemoryRegion
, 1);
216 /* offset to align to standby_subregion_size for allocation */
217 ram_addr_t offset
= assign_addr
-
218 (assign_addr
- mhd
->padded_ram_size
)
219 % mhd
->standby_subregion_size
;
221 /* strlen("standby.ram") + 4 (Max of KVM_MEMORY_SLOTS) + NULL */
223 snprintf(id
, 16, "standby.ram%d",
224 (int)((offset
- mhd
->padded_ram_size
) /
225 mhd
->standby_subregion_size
) + 1);
227 /* Allocate a subregion of the calculated standby_subregion_size */
228 if (offset
+ mhd
->standby_subregion_size
>
229 mhd
->padded_ram_size
+ mhd
->standby_mem_size
) {
230 this_subregion_size
= mhd
->padded_ram_size
+
231 mhd
->standby_mem_size
- offset
;
233 this_subregion_size
= mhd
->standby_subregion_size
;
236 memory_region_init_ram(standby_ram
, NULL
, id
, this_subregion_size
,
238 /* This is a hack to make memory hotunplug work again. Once we have
239 * subdevices, we have to unparent them when unassigning memory,
240 * instead of doing it via the ref count of the MemoryRegion. */
241 object_ref(OBJECT(standby_ram
));
242 object_unparent(OBJECT(standby_ram
));
243 vmstate_register_ram_global(standby_ram
);
244 memory_region_add_subregion(sysmem
, offset
, standby_ram
);
246 /* The specified subregion is no longer in standby */
247 mhd
->standby_state_map
[(assign_addr
- mhd
->padded_ram_size
)
248 / MEM_SECTION_SIZE
] = 1;
250 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_NORMAL_COMPLETION
);
253 static void unassign_storage(SCLPDevice
*sclp
, SCCB
*sccb
)
255 MemoryRegion
*mr
= NULL
;
256 AssignStorage
*assign_info
= (AssignStorage
*) sccb
;
257 sclpMemoryHotplugDev
*mhd
= get_sclp_memory_hotplug_dev();
258 ram_addr_t unassign_addr
;
259 MemoryRegion
*sysmem
= get_system_memory();
262 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_INVALID_SCLP_COMMAND
);
265 unassign_addr
= (assign_info
->rn
- 1) * mhd
->rzm
;
267 /* if the addr is a multiple of 256 MB */
268 if ((unassign_addr
% MEM_SECTION_SIZE
== 0) &&
269 (unassign_addr
>= mhd
->padded_ram_size
)) {
270 mhd
->standby_state_map
[(unassign_addr
-
271 mhd
->padded_ram_size
) / MEM_SECTION_SIZE
] = 0;
273 /* find the specified memory region and destroy it */
274 mr
= memory_region_find(sysmem
, unassign_addr
, 1).mr
;
275 memory_region_unref(mr
);
278 int is_removable
= 1;
279 ram_addr_t map_offset
= (unassign_addr
- mhd
->padded_ram_size
-
280 (unassign_addr
- mhd
->padded_ram_size
)
281 % mhd
->standby_subregion_size
);
282 /* Mark all affected subregions as 'standby' once again */
284 i
< (mhd
->standby_subregion_size
/ MEM_SECTION_SIZE
);
287 if (mhd
->standby_state_map
[i
+ map_offset
/ MEM_SECTION_SIZE
]) {
293 memory_region_del_subregion(sysmem
, mr
);
294 object_unref(OBJECT(mr
));
298 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_NORMAL_COMPLETION
);
301 /* Provide information about the CPU */
302 static void sclp_read_cpu_info(SCLPDevice
*sclp
, SCCB
*sccb
)
304 ReadCpuInfo
*cpu_info
= (ReadCpuInfo
*) sccb
;
313 cpu_info
->nr_configured
= cpu_to_be16(cpu_count
);
314 cpu_info
->offset_configured
= cpu_to_be16(offsetof(ReadCpuInfo
, entries
));
315 cpu_info
->nr_standby
= cpu_to_be16(0);
317 /* The standby offset is 16-byte for each CPU */
318 cpu_info
->offset_standby
= cpu_to_be16(cpu_info
->offset_configured
319 + cpu_info
->nr_configured
*sizeof(CPUEntry
));
321 for (i
= 0; i
< cpu_count
; i
++) {
322 cpu_info
->entries
[i
].address
= i
;
323 cpu_info
->entries
[i
].type
= 0;
326 sccb
->h
.response_code
= cpu_to_be16(SCLP_RC_NORMAL_READ_COMPLETION
);
329 static void sclp_execute(SCLPDevice
*sclp
, SCCB
*sccb
, uint32_t code
)
331 SCLPDeviceClass
*sclp_c
= SCLP_GET_CLASS(sclp
);
332 SCLPEventFacility
*ef
= sclp
->event_facility
;
333 SCLPEventFacilityClass
*efc
= EVENT_FACILITY_GET_CLASS(ef
);
335 switch (code
& SCLP_CMD_CODE_MASK
) {
336 case SCLP_CMDW_READ_SCP_INFO
:
337 case SCLP_CMDW_READ_SCP_INFO_FORCED
:
338 sclp_c
->read_SCP_info(sclp
, sccb
);
340 case SCLP_CMDW_READ_CPU_INFO
:
341 sclp_c
->read_cpu_info(sclp
, sccb
);
343 case SCLP_READ_STORAGE_ELEMENT_INFO
:
345 sclp_c
->read_storage_element1_info(sclp
, sccb
);
347 sclp_c
->read_storage_element0_info(sclp
, sccb
);
350 case SCLP_ATTACH_STORAGE_ELEMENT
:
351 sclp_c
->attach_storage_element(sclp
, sccb
, (code
& 0xff00) >> 8);
353 case SCLP_ASSIGN_STORAGE
:
354 sclp_c
->assign_storage(sclp
, sccb
);
356 case SCLP_UNASSIGN_STORAGE
:
357 sclp_c
->unassign_storage(sclp
, sccb
);
359 case SCLP_CMDW_CONFIGURE_PCI
:
360 s390_pci_sclp_configure(sccb
);
362 case SCLP_CMDW_DECONFIGURE_PCI
:
363 s390_pci_sclp_deconfigure(sccb
);
366 efc
->command_handler(ef
, sccb
, code
);
371 int sclp_service_call(CPUS390XState
*env
, uint64_t sccb
, uint32_t code
)
373 SCLPDevice
*sclp
= get_sclp_device();
374 SCLPDeviceClass
*sclp_c
= SCLP_GET_CLASS(sclp
);
378 hwaddr sccb_len
= sizeof(SCCB
);
380 /* first some basic checks on program checks */
381 if (env
->psw
.mask
& PSW_MASK_PSTATE
) {
385 if (cpu_physical_memory_is_io(sccb
)) {
389 if ((sccb
& ~0x1fffUL
) == 0 || (sccb
& ~0x1fffUL
) == env
->psa
390 || (sccb
& ~0x7ffffff8UL
) != 0) {
391 r
= -PGM_SPECIFICATION
;
396 * we want to work on a private copy of the sccb, to prevent guests
397 * from playing dirty tricks by modifying the memory content after
398 * the host has checked the values
400 cpu_physical_memory_read(sccb
, &work_sccb
, sccb_len
);
402 /* Valid sccb sizes */
403 if (be16_to_cpu(work_sccb
.h
.length
) < sizeof(SCCBHeader
) ||
404 be16_to_cpu(work_sccb
.h
.length
) > SCCB_SIZE
) {
405 r
= -PGM_SPECIFICATION
;
409 sclp_c
->execute(sclp
, (SCCB
*)&work_sccb
, code
);
411 cpu_physical_memory_write(sccb
, &work_sccb
,
412 be16_to_cpu(work_sccb
.h
.length
));
414 sclp_c
->service_interrupt(sclp
, sccb
);
420 static void service_interrupt(SCLPDevice
*sclp
, uint32_t sccb
)
422 SCLPEventFacility
*ef
= sclp
->event_facility
;
423 SCLPEventFacilityClass
*efc
= EVENT_FACILITY_GET_CLASS(ef
);
425 uint32_t param
= sccb
& ~3;
427 /* Indicate whether an event is still pending */
428 param
|= efc
->event_pending(ef
) ? 1 : 0;
431 /* No need to send an interrupt, there's nothing to be notified about */
434 s390_sclp_extint(param
);
437 void sclp_service_interrupt(uint32_t sccb
)
439 SCLPDevice
*sclp
= get_sclp_device();
440 SCLPDeviceClass
*sclp_c
= SCLP_GET_CLASS(sclp
);
442 sclp_c
->service_interrupt(sclp
, sccb
);
445 /* qemu object creation and initialization functions */
447 void s390_sclp_init(void)
449 Object
*new = object_new(TYPE_SCLP
);
451 object_property_add_child(qdev_get_machine(), TYPE_SCLP
, new,
453 object_unref(OBJECT(new));
454 qdev_init_nofail(DEVICE(new));
457 static void sclp_realize(DeviceState
*dev
, Error
**errp
)
459 MachineState
*machine
= MACHINE(qdev_get_machine());
460 SCLPDevice
*sclp
= SCLP(dev
);
465 object_property_set_bool(OBJECT(sclp
->event_facility
), true, "realized",
471 * qdev_device_add searches the sysbus for TYPE_SCLP_EVENTS_BUS. As long
472 * as we can't find a fitting bus via the qom tree, we have to add the
473 * event facility to the sysbus, so e.g. a sclp console can be created.
475 qdev_set_parent_bus(DEVICE(sclp
->event_facility
), sysbus_get_default());
477 ret
= s390_set_memory_limit(machine
->maxram_size
, &hw_limit
);
479 error_setg(&err
, "qemu: host supports a maximum of %" PRIu64
" GB",
482 error_setg(&err
, "qemu: setting the guest size failed");
486 error_propagate(errp
, err
);
489 static void sclp_memory_init(SCLPDevice
*sclp
)
491 MachineState
*machine
= MACHINE(qdev_get_machine());
492 ram_addr_t initial_mem
= machine
->ram_size
;
493 ram_addr_t max_mem
= machine
->maxram_size
;
494 ram_addr_t standby_mem
= max_mem
- initial_mem
;
495 ram_addr_t pad_mem
= 0;
496 int increment_size
= 20;
498 /* The storage increment size is a multiple of 1M and is a power of 2.
499 * The number of storage increments must be MAX_STORAGE_INCREMENTS or fewer.
500 * The variable 'increment_size' is an exponent of 2 that can be
501 * used to calculate the size (in bytes) of an increment. */
502 while ((initial_mem
>> increment_size
) > MAX_STORAGE_INCREMENTS
) {
505 if (machine
->ram_slots
) {
506 while ((standby_mem
>> increment_size
) > MAX_STORAGE_INCREMENTS
) {
510 sclp
->increment_size
= increment_size
;
512 /* The core and standby memory areas need to be aligned with
513 * the increment size. In effect, this can cause the
514 * user-specified memory size to be rounded down to align
515 * with the nearest increment boundary. */
516 initial_mem
= initial_mem
>> increment_size
<< increment_size
;
517 standby_mem
= standby_mem
>> increment_size
<< increment_size
;
519 /* If the size of ram is not on a MEM_SECTION_SIZE boundary,
520 calculate the pad size necessary to force this boundary. */
521 if (machine
->ram_slots
&& standby_mem
) {
522 sclpMemoryHotplugDev
*mhd
= init_sclp_memory_hotplug_dev();
524 if (initial_mem
% MEM_SECTION_SIZE
) {
525 pad_mem
= MEM_SECTION_SIZE
- initial_mem
% MEM_SECTION_SIZE
;
527 mhd
->increment_size
= increment_size
;
528 mhd
->pad_size
= pad_mem
;
529 mhd
->standby_mem_size
= standby_mem
;
531 machine
->ram_size
= initial_mem
;
532 machine
->maxram_size
= initial_mem
+ pad_mem
+ standby_mem
;
533 /* let's propagate the changed ram size into the global variable. */
534 ram_size
= initial_mem
;
537 static void sclp_init(Object
*obj
)
539 SCLPDevice
*sclp
= SCLP(obj
);
542 new = object_new(TYPE_SCLP_EVENT_FACILITY
);
543 object_property_add_child(obj
, TYPE_SCLP_EVENT_FACILITY
, new, NULL
);
545 sclp
->event_facility
= EVENT_FACILITY(new);
547 sclp_memory_init(sclp
);
550 static void sclp_class_init(ObjectClass
*oc
, void *data
)
552 SCLPDeviceClass
*sc
= SCLP_CLASS(oc
);
553 DeviceClass
*dc
= DEVICE_CLASS(oc
);
555 dc
->desc
= "SCLP (Service-Call Logical Processor)";
556 dc
->realize
= sclp_realize
;
557 dc
->hotpluggable
= false;
558 set_bit(DEVICE_CATEGORY_MISC
, dc
->categories
);
560 sc
->read_SCP_info
= read_SCP_info
;
561 sc
->read_storage_element0_info
= read_storage_element0_info
;
562 sc
->read_storage_element1_info
= read_storage_element1_info
;
563 sc
->attach_storage_element
= attach_storage_element
;
564 sc
->assign_storage
= assign_storage
;
565 sc
->unassign_storage
= unassign_storage
;
566 sc
->read_cpu_info
= sclp_read_cpu_info
;
567 sc
->execute
= sclp_execute
;
568 sc
->service_interrupt
= service_interrupt
;
571 static TypeInfo sclp_info
= {
573 .parent
= TYPE_DEVICE
,
574 .instance_init
= sclp_init
,
575 .instance_size
= sizeof(SCLPDevice
),
576 .class_init
= sclp_class_init
,
577 .class_size
= sizeof(SCLPDeviceClass
),
580 sclpMemoryHotplugDev
*init_sclp_memory_hotplug_dev(void)
583 dev
= qdev_create(NULL
, TYPE_SCLP_MEMORY_HOTPLUG_DEV
);
584 object_property_add_child(qdev_get_machine(),
585 TYPE_SCLP_MEMORY_HOTPLUG_DEV
,
587 qdev_init_nofail(dev
);
588 return SCLP_MEMORY_HOTPLUG_DEV(object_resolve_path(
589 TYPE_SCLP_MEMORY_HOTPLUG_DEV
, NULL
));
592 sclpMemoryHotplugDev
*get_sclp_memory_hotplug_dev(void)
594 return SCLP_MEMORY_HOTPLUG_DEV(object_resolve_path(
595 TYPE_SCLP_MEMORY_HOTPLUG_DEV
, NULL
));
598 static void sclp_memory_hotplug_dev_class_init(ObjectClass
*klass
,
601 DeviceClass
*dc
= DEVICE_CLASS(klass
);
603 set_bit(DEVICE_CATEGORY_MISC
, dc
->categories
);
606 static TypeInfo sclp_memory_hotplug_dev_info
= {
607 .name
= TYPE_SCLP_MEMORY_HOTPLUG_DEV
,
608 .parent
= TYPE_SYS_BUS_DEVICE
,
609 .instance_size
= sizeof(sclpMemoryHotplugDev
),
610 .class_init
= sclp_memory_hotplug_dev_class_init
,
613 static void register_types(void)
615 type_register_static(&sclp_memory_hotplug_dev_info
);
616 type_register_static(&sclp_info
);
618 type_init(register_types
);