2 * QEMU PAPR Storage Class Memory Interfaces
4 * Copyright (c) 2019-2020, IBM Corporation.
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 #include "qemu/osdep.h"
25 #include "qapi/error.h"
26 #include "hw/ppc/spapr_drc.h"
27 #include "hw/ppc/spapr_nvdimm.h"
28 #include "hw/mem/nvdimm.h"
29 #include "qemu/nvdimm-utils.h"
30 #include "qemu/option.h"
31 #include "hw/ppc/fdt.h"
32 #include "qemu/range.h"
33 #include "sysemu/sysemu.h"
34 #include "hw/ppc/spapr_numa.h"
36 bool spapr_nvdimm_validate(HotplugHandler
*hotplug_dev
, NVDIMMDevice
*nvdimm
,
37 uint64_t size
, Error
**errp
)
39 const MachineClass
*mc
= MACHINE_GET_CLASS(hotplug_dev
);
40 const MachineState
*ms
= MACHINE(hotplug_dev
);
41 const char *nvdimm_opt
= qemu_opt_get(qemu_get_machine_opts(), "nvdimm");
42 g_autofree
char *uuidstr
= NULL
;
46 if (!mc
->nvdimm_supported
) {
47 error_setg(errp
, "NVDIMM hotplug not supported for this machine");
52 * NVDIMM support went live in 5.1 without considering that, in
53 * other archs, the user needs to enable NVDIMM support with the
54 * 'nvdimm' machine option and the default behavior is NVDIMM
55 * support disabled. It is too late to roll back to the standard
56 * behavior without breaking 5.1 guests. What we can do is to
57 * ensure that, if the user sets nvdimm=off, we error out
58 * regardless of being 5.1 or newer.
60 if (!ms
->nvdimms_state
->is_enabled
&& nvdimm_opt
) {
61 error_setg(errp
, "nvdimm device found but 'nvdimm=off' was set");
65 if (object_property_get_int(OBJECT(nvdimm
), NVDIMM_LABEL_SIZE_PROP
,
67 error_setg(errp
, "PAPR requires NVDIMM devices to have label-size set");
71 if (size
% SPAPR_MINIMUM_SCM_BLOCK_SIZE
) {
72 error_setg(errp
, "PAPR requires NVDIMM memory size (excluding label)"
73 " to be a multiple of %" PRIu64
"MB",
74 SPAPR_MINIMUM_SCM_BLOCK_SIZE
/ MiB
);
78 uuidstr
= object_property_get_str(OBJECT(nvdimm
), NVDIMM_UUID_PROP
,
80 ret
= qemu_uuid_parse(uuidstr
, &uuid
);
83 if (qemu_uuid_is_null(&uuid
)) {
84 error_setg(errp
, "NVDIMM device requires the uuid to be set");
92 bool spapr_add_nvdimm(DeviceState
*dev
, uint64_t slot
, Error
**errp
)
95 bool hotplugged
= spapr_drc_hotplugged(dev
);
97 drc
= spapr_drc_by_id(TYPE_SPAPR_DRC_PMEM
, slot
);
100 if (!spapr_drc_attach(drc
, dev
, errp
)) {
105 spapr_hotplug_req_add_by_index(drc
);
110 static int spapr_dt_nvdimm(SpaprMachineState
*spapr
, void *fdt
,
111 int parent_offset
, NVDIMMDevice
*nvdimm
)
117 uint32_t node
= object_property_get_uint(OBJECT(nvdimm
), PC_DIMM_NODE_PROP
,
119 uint64_t slot
= object_property_get_uint(OBJECT(nvdimm
), PC_DIMM_SLOT_PROP
,
121 uint64_t lsize
= nvdimm
->label_size
;
122 uint64_t size
= object_property_get_int(OBJECT(nvdimm
), PC_DIMM_SIZE_PROP
,
125 drc
= spapr_drc_by_id(TYPE_SPAPR_DRC_PMEM
, slot
);
128 drc_idx
= spapr_drc_index(drc
);
130 buf
= g_strdup_printf("ibm,pmemory@%x", drc_idx
);
131 child_offset
= fdt_add_subnode(fdt
, parent_offset
, buf
);
136 _FDT((fdt_setprop_cell(fdt
, child_offset
, "reg", drc_idx
)));
137 _FDT((fdt_setprop_string(fdt
, child_offset
, "compatible", "ibm,pmemory")));
138 _FDT((fdt_setprop_string(fdt
, child_offset
, "device_type", "ibm,pmemory")));
140 spapr_numa_write_associativity_dt(spapr
, fdt
, child_offset
, node
);
142 buf
= qemu_uuid_unparse_strdup(&nvdimm
->uuid
);
143 _FDT((fdt_setprop_string(fdt
, child_offset
, "ibm,unit-guid", buf
)));
146 _FDT((fdt_setprop_cell(fdt
, child_offset
, "ibm,my-drc-index", drc_idx
)));
148 _FDT((fdt_setprop_u64(fdt
, child_offset
, "ibm,block-size",
149 SPAPR_MINIMUM_SCM_BLOCK_SIZE
)));
150 _FDT((fdt_setprop_u64(fdt
, child_offset
, "ibm,number-of-blocks",
151 size
/ SPAPR_MINIMUM_SCM_BLOCK_SIZE
)));
152 _FDT((fdt_setprop_cell(fdt
, child_offset
, "ibm,metadata-size", lsize
)));
154 _FDT((fdt_setprop_string(fdt
, child_offset
, "ibm,pmem-application",
155 "operating-system")));
156 _FDT(fdt_setprop(fdt
, child_offset
, "ibm,cache-flush-required", NULL
, 0));
161 int spapr_pmem_dt_populate(SpaprDrc
*drc
, SpaprMachineState
*spapr
,
162 void *fdt
, int *fdt_start_offset
, Error
**errp
)
164 NVDIMMDevice
*nvdimm
= NVDIMM(drc
->dev
);
166 *fdt_start_offset
= spapr_dt_nvdimm(spapr
, fdt
, 0, nvdimm
);
171 void spapr_dt_persistent_memory(SpaprMachineState
*spapr
, void *fdt
)
173 int offset
= fdt_subnode_offset(fdt
, 0, "persistent-memory");
174 GSList
*iter
, *nvdimms
= nvdimm_get_device_list();
177 offset
= fdt_add_subnode(fdt
, 0, "persistent-memory");
179 _FDT((fdt_setprop_cell(fdt
, offset
, "#address-cells", 0x1)));
180 _FDT((fdt_setprop_cell(fdt
, offset
, "#size-cells", 0x0)));
181 _FDT((fdt_setprop_string(fdt
, offset
, "device_type",
182 "ibm,persistent-memory")));
185 /* Create DT entries for cold plugged NVDIMM devices */
186 for (iter
= nvdimms
; iter
; iter
= iter
->next
) {
187 NVDIMMDevice
*nvdimm
= iter
->data
;
189 spapr_dt_nvdimm(spapr
, fdt
, offset
, nvdimm
);
191 g_slist_free(nvdimms
);
196 static target_ulong
h_scm_read_metadata(PowerPCCPU
*cpu
,
197 SpaprMachineState
*spapr
,
201 uint32_t drc_index
= args
[0];
202 uint64_t offset
= args
[1];
203 uint64_t len
= args
[2];
204 SpaprDrc
*drc
= spapr_drc_by_index(drc_index
);
205 NVDIMMDevice
*nvdimm
;
208 uint8_t buf
[8] = { 0 };
210 if (!drc
|| !drc
->dev
||
211 spapr_drc_type(drc
) != SPAPR_DR_CONNECTOR_TYPE_PMEM
) {
215 if (len
!= 1 && len
!= 2 &&
216 len
!= 4 && len
!= 8) {
220 nvdimm
= NVDIMM(drc
->dev
);
221 if ((offset
+ len
< offset
) ||
222 (nvdimm
->label_size
< len
+ offset
)) {
226 ddc
= NVDIMM_GET_CLASS(nvdimm
);
227 ddc
->read_label_data(nvdimm
, buf
, len
, offset
);
234 data
= lduw_be_p(buf
);
237 data
= ldl_be_p(buf
);
240 data
= ldq_be_p(buf
);
243 g_assert_not_reached();
251 static target_ulong
h_scm_write_metadata(PowerPCCPU
*cpu
,
252 SpaprMachineState
*spapr
,
256 uint32_t drc_index
= args
[0];
257 uint64_t offset
= args
[1];
258 uint64_t data
= args
[2];
259 uint64_t len
= args
[3];
260 SpaprDrc
*drc
= spapr_drc_by_index(drc_index
);
261 NVDIMMDevice
*nvdimm
;
263 uint8_t buf
[8] = { 0 };
265 if (!drc
|| !drc
->dev
||
266 spapr_drc_type(drc
) != SPAPR_DR_CONNECTOR_TYPE_PMEM
) {
270 if (len
!= 1 && len
!= 2 &&
271 len
!= 4 && len
!= 8) {
275 nvdimm
= NVDIMM(drc
->dev
);
276 if ((offset
+ len
< offset
) ||
277 (nvdimm
->label_size
< len
+ offset
)) {
283 if (data
& 0xffffffffffffff00) {
289 if (data
& 0xffffffffffff0000) {
295 if (data
& 0xffffffff00000000) {
304 g_assert_not_reached();
307 ddc
= NVDIMM_GET_CLASS(nvdimm
);
308 ddc
->write_label_data(nvdimm
, buf
, len
, offset
);
313 static target_ulong
h_scm_bind_mem(PowerPCCPU
*cpu
, SpaprMachineState
*spapr
,
314 target_ulong opcode
, target_ulong
*args
)
316 uint32_t drc_index
= args
[0];
317 uint64_t starting_idx
= args
[1];
318 uint64_t no_of_scm_blocks_to_bind
= args
[2];
319 uint64_t target_logical_mem_addr
= args
[3];
320 uint64_t continue_token
= args
[4];
322 uint64_t total_no_of_scm_blocks
;
323 SpaprDrc
*drc
= spapr_drc_by_index(drc_index
);
325 NVDIMMDevice
*nvdimm
;
327 if (!drc
|| !drc
->dev
||
328 spapr_drc_type(drc
) != SPAPR_DR_CONNECTOR_TYPE_PMEM
) {
333 * Currently continue token should be zero qemu has already bound
334 * everything and this hcall doesnt return H_BUSY.
336 if (continue_token
> 0) {
340 /* Currently qemu assigns the address. */
341 if (target_logical_mem_addr
!= 0xffffffffffffffff) {
345 nvdimm
= NVDIMM(drc
->dev
);
347 size
= object_property_get_uint(OBJECT(nvdimm
),
348 PC_DIMM_SIZE_PROP
, &error_abort
);
350 total_no_of_scm_blocks
= size
/ SPAPR_MINIMUM_SCM_BLOCK_SIZE
;
352 if (starting_idx
> total_no_of_scm_blocks
) {
356 if (((starting_idx
+ no_of_scm_blocks_to_bind
) < starting_idx
) ||
357 ((starting_idx
+ no_of_scm_blocks_to_bind
) > total_no_of_scm_blocks
)) {
361 addr
= object_property_get_uint(OBJECT(nvdimm
),
362 PC_DIMM_ADDR_PROP
, &error_abort
);
364 addr
+= starting_idx
* SPAPR_MINIMUM_SCM_BLOCK_SIZE
;
366 /* Already bound, Return target logical address in R5 */
368 args
[2] = no_of_scm_blocks_to_bind
;
373 static target_ulong
h_scm_unbind_mem(PowerPCCPU
*cpu
, SpaprMachineState
*spapr
,
374 target_ulong opcode
, target_ulong
*args
)
376 uint32_t drc_index
= args
[0];
377 uint64_t starting_scm_logical_addr
= args
[1];
378 uint64_t no_of_scm_blocks_to_unbind
= args
[2];
379 uint64_t continue_token
= args
[3];
380 uint64_t size_to_unbind
;
381 Range blockrange
= range_empty
;
382 Range nvdimmrange
= range_empty
;
383 SpaprDrc
*drc
= spapr_drc_by_index(drc_index
);
384 NVDIMMDevice
*nvdimm
;
387 if (!drc
|| !drc
->dev
||
388 spapr_drc_type(drc
) != SPAPR_DR_CONNECTOR_TYPE_PMEM
) {
392 /* continue_token should be zero as this hcall doesn't return H_BUSY. */
393 if (continue_token
> 0) {
397 /* Check if starting_scm_logical_addr is block aligned */
398 if (!QEMU_IS_ALIGNED(starting_scm_logical_addr
,
399 SPAPR_MINIMUM_SCM_BLOCK_SIZE
)) {
403 size_to_unbind
= no_of_scm_blocks_to_unbind
* SPAPR_MINIMUM_SCM_BLOCK_SIZE
;
404 if (no_of_scm_blocks_to_unbind
== 0 || no_of_scm_blocks_to_unbind
!=
405 size_to_unbind
/ SPAPR_MINIMUM_SCM_BLOCK_SIZE
) {
409 nvdimm
= NVDIMM(drc
->dev
);
410 size
= object_property_get_int(OBJECT(nvdimm
), PC_DIMM_SIZE_PROP
,
412 addr
= object_property_get_int(OBJECT(nvdimm
), PC_DIMM_ADDR_PROP
,
415 range_init_nofail(&nvdimmrange
, addr
, size
);
416 range_init_nofail(&blockrange
, starting_scm_logical_addr
, size_to_unbind
);
418 if (!range_contains_range(&nvdimmrange
, &blockrange
)) {
422 args
[1] = no_of_scm_blocks_to_unbind
;
424 /* let unplug take care of actual unbind */
428 #define H_UNBIND_SCOPE_ALL 0x1
429 #define H_UNBIND_SCOPE_DRC 0x2
431 static target_ulong
h_scm_unbind_all(PowerPCCPU
*cpu
, SpaprMachineState
*spapr
,
432 target_ulong opcode
, target_ulong
*args
)
434 uint64_t target_scope
= args
[0];
435 uint32_t drc_index
= args
[1];
436 uint64_t continue_token
= args
[2];
437 NVDIMMDevice
*nvdimm
;
439 uint64_t no_of_scm_blocks_unbound
= 0;
441 /* continue_token should be zero as this hcall doesn't return H_BUSY. */
442 if (continue_token
> 0) {
446 if (target_scope
== H_UNBIND_SCOPE_DRC
) {
447 SpaprDrc
*drc
= spapr_drc_by_index(drc_index
);
449 if (!drc
|| !drc
->dev
||
450 spapr_drc_type(drc
) != SPAPR_DR_CONNECTOR_TYPE_PMEM
) {
454 nvdimm
= NVDIMM(drc
->dev
);
455 size
= object_property_get_int(OBJECT(nvdimm
), PC_DIMM_SIZE_PROP
,
458 no_of_scm_blocks_unbound
= size
/ SPAPR_MINIMUM_SCM_BLOCK_SIZE
;
459 } else if (target_scope
== H_UNBIND_SCOPE_ALL
) {
460 GSList
*list
, *nvdimms
;
462 nvdimms
= nvdimm_get_device_list();
463 for (list
= nvdimms
; list
; list
= list
->next
) {
465 size
= object_property_get_int(OBJECT(nvdimm
), PC_DIMM_SIZE_PROP
,
468 no_of_scm_blocks_unbound
+= size
/ SPAPR_MINIMUM_SCM_BLOCK_SIZE
;
470 g_slist_free(nvdimms
);
475 args
[1] = no_of_scm_blocks_unbound
;
477 /* let unplug take care of actual unbind */
481 static void spapr_scm_register_types(void)
483 /* qemu/scm specific hcalls */
484 spapr_register_hypercall(H_SCM_READ_METADATA
, h_scm_read_metadata
);
485 spapr_register_hypercall(H_SCM_WRITE_METADATA
, h_scm_write_metadata
);
486 spapr_register_hypercall(H_SCM_BIND_MEM
, h_scm_bind_mem
);
487 spapr_register_hypercall(H_SCM_UNBIND_MEM
, h_scm_unbind_mem
);
488 spapr_register_hypercall(H_SCM_UNBIND_ALL
, h_scm_unbind_all
);
491 type_init(spapr_scm_register_types
)