xref: /linux/drivers/platform/x86/intel/speed_select_if/isst_tpmi_core.c (revision fbddf68d7b4e1e6da7a78dd7fbd8ec376536584a)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * isst_tpmi.c: SST TPMI interface core
4  *
5  * Copyright (c) 2023, Intel Corporation.
6  * All Rights Reserved.
7  *
8  * This information will be useful to understand flows:
9  * In the current generation of platforms, TPMI is supported via OOB
10  * PCI device. This PCI device has one instance per CPU package.
11  * There is a unique TPMI ID for SST. Each TPMI ID also has multiple
12  * entries, representing per power domain information.
13  *
14  * There is one dev file for complete SST information and control same as the
15  * prior generation of hardware. User spaces don't need to know how the
16  * information is presented by the hardware. The TPMI core module implements
17  * the hardware mapping.
18  */
19 
20 #define dev_fmt(fmt) "tpmi_sst: " fmt
21 
22 #include <linux/auxiliary_bus.h>
23 #include <linux/delay.h>
24 #include <linux/intel_tpmi.h>
25 #include <linux/intel_vsec.h>
26 #include <linux/fs.h>
27 #include <linux/io.h>
28 #include <linux/kernel.h>
29 #include <linux/minmax.h>
30 #include <linux/module.h>
31 #include <asm/msr.h>
32 #include <uapi/linux/isst_if.h>
33 
34 #include "isst_tpmi_core.h"
35 #include "isst_if_common.h"
36 
37 /* Supported SST hardware version by this driver */
38 #define ISST_MAJOR_VERSION	0
39 #define ISST_MINOR_VERSION	2
40 
41 /*
42  * Used to indicate if value read from MMIO needs to get multiplied
43  * to get to a standard unit or not.
44  */
45 #define SST_MUL_FACTOR_NONE    1
46 
47 /* Define 100 as a scaling factor frequency ratio to frequency conversion */
48 #define SST_MUL_FACTOR_FREQ    100
49 
50 /* All SST regs are 64 bit size */
51 #define SST_REG_SIZE   8
52 
53 /**
54  * struct sst_header -	SST main header
55  * @interface_version:	Version number for this interface
56  * @cap_mask:		Bitmask of the supported sub features. 1=the sub feature is enabled.
57  *			0=disabled.
58  *			Bit[8]= SST_CP enable (1), disable (0)
59  *			bit[9]= SST_PP enable (1), disable (0)
60  *			other bits are reserved for future use
61  * @cp_offset:		Qword (8 bytes) offset to the SST_CP register bank
62  * @pp_offset:		Qword (8 bytes) offset to the SST_PP register bank
63  * @reserved:		Reserved for future use
64  *
65  * This register allows SW to discover SST capability and the offsets to SST-CP
66  * and SST-PP register banks.
67  */
68 struct sst_header {
69 	u8 interface_version;
70 	u8 cap_mask;
71 	u8 cp_offset;
72 	u8 pp_offset;
73 	u32 reserved;
74 } __packed;
75 
76 /**
77  * struct cp_header -	SST-CP (core-power) header
78  * @feature_id:		0=SST-CP, 1=SST-PP, 2=SST-BF, 3=SST-TF
79  * @feature_rev:	Interface Version number for this SST feature
80  * @ratio_unit:		Frequency ratio unit. 00: 100MHz. All others are reserved
81  * @reserved:		Reserved for future use
82  *
83  * This structure is used store SST-CP header. This is packed to the same
84  * format as defined in the specifications.
85  */
86 struct cp_header {
87 	u64 feature_id :4;
88 	u64 feature_rev :8;
89 	u64 ratio_unit :2;
90 	u64 reserved :50;
91 } __packed;
92 
93 /**
94  * struct pp_header -	SST-PP (Perf profile) header
95  * @feature_id:		0=SST-CP, 1=SST-PP, 2=SST-BF, 3=SST-TF
96  * @feature_rev:	Interface Version number for this SST feature
97  * @level_en_mask:	SST-PP level enable/disable fuse mask
98  * @allowed_level_mask:	Allowed level mask used for dynamic config level switching
99  * @reserved0:		Reserved for future use
100  * @ratio_unit:		Frequency ratio unit. 00: 100MHz. All others are reserved
101  * @block_size:		Size of PP block in Qword unit (8 bytes)
102  * @dynamic_switch:	If set (1), dynamic switching of SST PP is supported
103  * @memory_ratio_unit:	Memory Controller frequency ratio unit. 00: 100MHz, others reserved
104  * @reserved1:		Reserved for future use
105  *
106  * This structure is used store SST-PP header. This is packed to the same
107  * format as defined in the specifications.
108  */
109 struct pp_header {
110 	u64 feature_id :4;
111 	u64 feature_rev :8;
112 	u64 level_en_mask :8;
113 	u64 allowed_level_mask :8;
114 	u64 reserved0 :4;
115 	u64 ratio_unit :2;
116 	u64 block_size :8;
117 	u64 dynamic_switch :1;
118 	u64 memory_ratio_unit :2;
119 	u64 reserved1 :19;
120 } __packed;
121 
122 /**
123  * struct feature_offset -	Offsets to SST-PP features
124  * @pp_offset:		Qword offset within PP level for the SST_PP register bank
125  * @bf_offset:		Qword offset within PP level for the SST_BF register bank
126  * @tf_offset:		Qword offset within PP level for the SST_TF register bank
127  * @reserved:		Reserved for future use
128  *
129  * This structure is used store offsets for SST features in the register bank.
130  * This is packed to the same format as defined in the specifications.
131  */
132 struct feature_offset {
133 	u64 pp_offset :8;
134 	u64 bf_offset :8;
135 	u64 tf_offset :8;
136 	u64 reserved :40;
137 } __packed;
138 
139 /**
140  * struct levels_offset -	Offsets to each SST PP level
141  * @sst_pp_level0_offset:	Qword offset to the register block of PP level 0
142  * @sst_pp_level1_offset:	Qword offset to the register block of PP level 1
143  * @sst_pp_level2_offset:	Qword offset to the register block of PP level 2
144  * @sst_pp_level3_offset:	Qword offset to the register block of PP level 3
145  * @sst_pp_level4_offset:	Qword offset to the register block of PP level 4
146  * @reserved:			Reserved for future use
147  *
148  * This structure is used store offsets of SST PP levels in the register bank.
149  * This is packed to the same format as defined in the specifications.
150  */
151 struct levels_offset {
152 	u64 sst_pp_level0_offset :8;
153 	u64 sst_pp_level1_offset :8;
154 	u64 sst_pp_level2_offset :8;
155 	u64 sst_pp_level3_offset :8;
156 	u64 sst_pp_level4_offset :8;
157 	u64 reserved :24;
158 } __packed;
159 
160 /**
161  * struct pp_control_offset -	Offsets for SST PP controls
162  * @perf_level:		A SST-PP level that SW intends to switch to
163  * @perf_level_lock:	SST-PP level select lock. 0 - unlocked. 1 - locked till next reset
164  * @resvd0:		Reserved for future use
165  * @current_state:	Bit mask to control the enable(1)/disable(0) state of each feature
166  *			of the current PP level, bit 0 = BF, bit 1 = TF, bit 2-7 = reserved
167  * @reserved:		Reserved for future use
168  *
169  * This structure is used store offsets of SST PP controls in the register bank.
170  * This is packed to the same format as defined in the specifications.
171  */
172 struct pp_control_offset {
173 	u64 perf_level :3;
174 	u64 perf_level_lock :1;
175 	u64 resvd0 :4;
176 	u64 current_state :8;
177 	u64 reserved :48;
178 } __packed;
179 
180 /**
181  * struct pp_status_offset -	Offsets for SST PP status fields
182  * @sst_pp_level:	Returns the current SST-PP level
183  * @sst_pp_lock:	Returns the lock bit setting of perf_level_lock in pp_control_offset
184  * @error_type:		Returns last error of SST-PP level change request. 0: no error,
185  *			1: level change not allowed, others: reserved
186  * @feature_state:	Bit mask to indicate the enable(1)/disable(0) state of each feature of the
187  *			current PP level. bit 0 = BF, bit 1 = TF, bit 2-7 reserved
188  * @reserved0:		Reserved for future use
189  * @feature_error_type: Returns last error of the specific feature. Three error_type bits per
190  *			feature. i.e. ERROR_TYPE[2:0] for BF, ERROR_TYPE[5:3] for TF, etc.
191  *			0x0: no error, 0x1: The specific feature is not supported by the hardware.
192  *			0x2-0x6: Reserved. 0x7: feature state change is not allowed.
193  * @reserved1:		Reserved for future use
194  *
195  * This structure is used store offsets of SST PP status in the register bank.
196  * This is packed to the same format as defined in the specifications.
197  */
198 struct pp_status_offset {
199 	u64 sst_pp_level :3;
200 	u64 sst_pp_lock :1;
201 	u64 error_type :4;
202 	u64 feature_state :8;
203 	u64 reserved0 :16;
204 	u64 feature_error_type : 24;
205 	u64 reserved1 :8;
206 } __packed;
207 
208 /**
209  * struct perf_level -	Used to store perf level and mmio offset
210  * @mmio_offset:	mmio offset for a perf level
211  * @level:		perf level for this offset
212  *
213  * This structure is used store final mmio offset of each perf level from the
214  * SST base mmio offset.
215  */
216 struct perf_level {
217 	int mmio_offset;
218 	int level;
219 };
220 
221 /**
222  * struct tpmi_per_power_domain_info -	Store per power_domain SST info
223  * @package_id:		Package id for this power_domain
224  * @power_domain_id:	Power domain id, Each entry from the SST-TPMI instance is a power_domain.
225  * @max_level:		Max possible PP level possible for this power_domain
226  * @ratio_unit:		Ratio unit for converting to MHz
227  * @avx_levels:		Number of AVX levels
228  * @pp_block_size:	Block size from PP header
229  * @sst_header:		Store SST header for this power_domain
230  * @cp_header:		Store SST-CP header for this power_domain
231  * @pp_header:		Store SST-PP header for this power_domain
232  * @perf_levels:	Pointer to each perf level to map level to mmio offset
233  * @feature_offsets:	Store feature offsets for each PP-level
234  * @control_offset:	Store the control offset for each PP-level
235  * @status_offset:	Store the status offset for each PP-level
236  * @sst_base:		Mapped SST base IO memory
237  * @auxdev:		Auxiliary device instance enumerated this instance
238  * @saved_sst_cp_control: Save SST-CP control configuration to store restore for suspend/resume
239  * @saved_clos_configs:	Save SST-CP CLOS configuration to store restore for suspend/resume
240  * @saved_clos_assocs:	Save SST-CP CLOS association to store restore for suspend/resume
241  * @saved_pp_control:	Save SST-PP control information to store restore for suspend/resume
242  * @write_blocked:	Write operation is blocked, so can't change SST state
243  *
244  * This structure is used store complete SST information for a power_domain. This information
245  * is used to read/write request for any SST IOCTL. Each physical CPU package can have multiple
246  * power_domains. Each power domain describes its own SST information and has its own controls.
247  */
248 struct tpmi_per_power_domain_info {
249 	int package_id;
250 	int power_domain_id;
251 	int max_level;
252 	int ratio_unit;
253 	int avx_levels;
254 	int pp_block_size;
255 	struct sst_header sst_header;
256 	struct cp_header cp_header;
257 	struct pp_header pp_header;
258 	struct perf_level *perf_levels;
259 	struct feature_offset feature_offsets;
260 	struct pp_control_offset control_offset;
261 	struct pp_status_offset status_offset;
262 	void __iomem *sst_base;
263 	struct auxiliary_device *auxdev;
264 	u64 saved_sst_cp_control;
265 	u64 saved_clos_configs[4];
266 	u64 saved_clos_assocs[4];
267 	u64 saved_pp_control;
268 	bool write_blocked;
269 };
270 
271 /* Supported maximum partitions */
272 #define SST_MAX_PARTITIONS	2
273 
274 /**
275  * struct tpmi_sst_struct -	Store sst info for a package
276  * @package_id:			Package id for this aux device instance
277  * @number_of_power_domains:	Number of power_domains pointed by power_domain_info pointer
278  * @power_domain_info:		Pointer to power domains information
279  * @cdie_mask:			Mask of compute dies present in a partition from hardware.
280  *				This mask is not present in the version 1 information header.
281  * @io_dies:			Number of IO dies in a partition. This will be 0 for TPMI
282  *				version 1 information header.
283  * @partition_mask:		Mask of all partitions.
284  * @partition_mask_current:	Current partition mask as some may have been unbound.
285  *
286  * This structure is used store full SST information for a package.
287  * Each package has one or multiple OOB PCI devices. Each package can contain multiple
288  * power domains.
289  */
290 struct tpmi_sst_struct {
291 	int package_id;
292 	struct tpmi_per_power_domain_info *power_domain_info[SST_MAX_PARTITIONS];
293 	u16 cdie_mask[SST_MAX_PARTITIONS];
294 	u8 number_of_power_domains[SST_MAX_PARTITIONS];
295 	u8 io_dies[SST_MAX_PARTITIONS];
296 	u8 partition_mask;
297 	u8 partition_mask_current;
298 };
299 
300 /**
301  * struct tpmi_sst_common_struct -	Store all SST instances
302  * @max_index:		Maximum instances currently present
303  * @sst_inst:		Pointer to per package instance
304  *
305  * Stores every SST Package instance.
306  */
307 struct tpmi_sst_common_struct {
308 	int max_index;
309 	struct tpmi_sst_struct **sst_inst;
310 };
311 
312 /*
313  * Each IOCTL request is processed under this lock. Also used to protect
314  * registration functions and common data structures.
315  */
316 static DEFINE_MUTEX(isst_tpmi_dev_lock);
317 
318 /* Usage count to track, number of TPMI SST instances registered to this core. */
319 static int isst_core_usage_count;
320 
321 /* Stores complete SST information for every package and power_domain */
322 static struct tpmi_sst_common_struct isst_common;
323 
324 #define SST_MAX_AVX_LEVELS	3
325 
326 #define SST_PP_OFFSET_0		8
327 #define SST_PP_OFFSET_1		16
328 #define SST_PP_OFFSET_SIZE	8
329 
330 static int sst_add_perf_profiles(struct auxiliary_device *auxdev,
331 				 struct tpmi_per_power_domain_info *pd_info,
332 				 int levels)
333 {
334 	struct device *dev = &auxdev->dev;
335 	u64 perf_level_offsets;
336 	int i;
337 
338 	pd_info->perf_levels = devm_kcalloc(dev, levels, sizeof(struct perf_level), GFP_KERNEL);
339 	if (!pd_info->perf_levels)
340 		return 0;
341 
342 	pd_info->ratio_unit = pd_info->pp_header.ratio_unit;
343 	pd_info->avx_levels = SST_MAX_AVX_LEVELS;
344 	pd_info->pp_block_size = pd_info->pp_header.block_size;
345 
346 	/* Read PP Offset 0: Get feature offset with PP level */
347 	*((u64 *)&pd_info->feature_offsets) = readq(pd_info->sst_base +
348 						    pd_info->sst_header.pp_offset +
349 						    SST_PP_OFFSET_0);
350 
351 	perf_level_offsets = readq(pd_info->sst_base + pd_info->sst_header.pp_offset +
352 				   SST_PP_OFFSET_1);
353 
354 	for (i = 0; i < levels; ++i) {
355 		u64 offset;
356 
357 		offset = perf_level_offsets & (0xffULL << (i * SST_PP_OFFSET_SIZE));
358 		offset >>= (i * 8);
359 		offset &= 0xff;
360 		offset *= 8; /* Convert to byte from QWORD offset */
361 		pd_info->perf_levels[i].mmio_offset = pd_info->sst_header.pp_offset + offset;
362 	}
363 
364 	return 0;
365 }
366 
367 static int sst_main(struct auxiliary_device *auxdev, struct tpmi_per_power_domain_info *pd_info)
368 {
369 	struct device *dev = &auxdev->dev;
370 	int i, mask, levels;
371 
372 	*((u64 *)&pd_info->sst_header) = readq(pd_info->sst_base);
373 	pd_info->sst_header.cp_offset *= 8;
374 	pd_info->sst_header.pp_offset *= 8;
375 
376 	if (pd_info->sst_header.interface_version == TPMI_VERSION_INVALID)
377 		return -ENODEV;
378 
379 	if (TPMI_MAJOR_VERSION(pd_info->sst_header.interface_version) != ISST_MAJOR_VERSION) {
380 		dev_err(dev, "SST: Unsupported major version:%lx\n",
381 			TPMI_MAJOR_VERSION(pd_info->sst_header.interface_version));
382 		return -ENODEV;
383 	}
384 
385 	if (TPMI_MINOR_VERSION(pd_info->sst_header.interface_version) > ISST_MINOR_VERSION)
386 		dev_info(dev, "SST: Ignore: Unsupported minor version:%lx\n",
387 			 TPMI_MINOR_VERSION(pd_info->sst_header.interface_version));
388 
389 	/* Read SST CP Header */
390 	*((u64 *)&pd_info->cp_header) = readq(pd_info->sst_base + pd_info->sst_header.cp_offset);
391 
392 	/* Read PP header */
393 	*((u64 *)&pd_info->pp_header) = readq(pd_info->sst_base + pd_info->sst_header.pp_offset);
394 
395 	mask = 0x01;
396 	levels = 0;
397 	for (i = 0; i < 8; ++i) {
398 		if (pd_info->pp_header.level_en_mask & mask)
399 			levels = i;
400 		mask <<= 1;
401 	}
402 	pd_info->max_level = levels;
403 	sst_add_perf_profiles(auxdev, pd_info, levels + 1);
404 
405 	return 0;
406 }
407 
408 static u8 isst_instance_count(struct tpmi_sst_struct *sst_inst)
409 {
410 	u8 i, max_part, count = 0;
411 
412 	/* Partition mask starts from bit 0 and contains 1s only */
413 	max_part = hweight8(sst_inst->partition_mask);
414 	for (i = 0; i < max_part; i++)
415 		count += sst_inst->number_of_power_domains[i];
416 
417 	return count;
418 }
419 
420 /**
421  * map_cdies() - Map user domain ID to compute domain ID
422  * @sst_inst: TPMI Instance
423  * @id: User domain ID
424  * @partition: Resolved partition
425  *
426  * Helper function to map_partition_power_domain_id() to resolve compute
427  * domain ID and partition. Use hardware provided cdie_mask for a partition
428  * as is to resolve a compute domain ID.
429  *
430  * Return: %-EINVAL on error, otherwise mapped domain ID >= 0.
431  */
432 static int map_cdies(struct tpmi_sst_struct *sst_inst, u8 id, u8 *partition)
433 {
434 	u8 i, max_part;
435 
436 	max_part = hweight8(sst_inst->partition_mask);
437 	for (i = 0; i < max_part; i++) {
438 		if (!(sst_inst->cdie_mask[i] & BIT(id)))
439 			continue;
440 
441 		*partition = i;
442 		return id - ffs(sst_inst->cdie_mask[i]) + 1;
443 	}
444 
445 	return -EINVAL;
446 }
447 
448 /**
449  * map_partition_power_domain_id() - Map user domain ID to partition domain ID
450  * @sst_inst: TPMI Instance
451  * @id: User domain ID
452  * @partition: Resolved partition
453  *
454  * In a partitioned system a CPU package has two separate MMIO ranges (Under
455  * two PCI devices). But the CPU package compute die/power domain IDs are
456  * unique in a package. User space can get compute die/power domain ID from
457  * CPUID and MSR 0x54 for a CPU. So, those IDs need to be preserved even if
458  * they are present in two different partitions with its own order.
459  *
460  * For example for command ISST_IF_COUNT_TPMI_INSTANCES, the valid_mask
461  * is 111111b for a 4 compute and 2 IO dies system. This is presented as
462  * provided by the hardware in a non-partitioned system with the following
463  * order:
464  *	I1-I0-C3-C2-C1-C0
465  * Here: "C": for compute and "I" for IO die.
466  * Compute dies are always present first in TPMI instances, as they have
467  * to map to the real power domain/die ID of a system. In a non-partitioned
468  * system there is no way to identify compute and IO die boundaries from
469  * this driver without reading each CPU's mapping.
470  *
471  * The same order needs to be preserved, even if those compute dies are
472  * distributed among multiple partitions. For example:
473  * Partition 1 can contain: I1-C1-C0
474  * Partition 2 can contain: I2-C3-C2
475  *
476  * This will require a conversion of user space IDs to the actual index into
477  * array of stored power domains for each partition. For the above example
478  * this function will return partition and index as follows:
479  *
480  * =============	=========	=====	========
481  * User space ID	Partition	Index	Die type
482  * =============	=========	=====	========
483  * 0			0		0	Compute
484  * 1			0		1	Compute
485  * 2			1		0	Compute
486  * 3			1		1	Compute
487  * 4			0		2	IO
488  * 5			1		2	IO
489  * =============	=========	=====	========
490  *
491  * Return: %-EINVAL on error, otherwise mapped domain ID >= 0.
492  */
493 static int map_partition_power_domain_id(struct tpmi_sst_struct *sst_inst, u8 id, u8 *partition)
494 {
495 	u8 i, io_start_id, max_part;
496 
497 	*partition = 0;
498 
499 	/* If any PCI device for partition is unbound, treat this as failure */
500 	if (sst_inst->partition_mask != sst_inst->partition_mask_current)
501 		return -EINVAL;
502 
503 	max_part = hweight8(sst_inst->partition_mask);
504 
505 	/* IO Index begin here */
506 	io_start_id = fls(sst_inst->cdie_mask[max_part - 1]);
507 
508 	if (id < io_start_id)
509 		return map_cdies(sst_inst, id, partition);
510 
511 	for (i = 0; i < max_part; i++) {
512 		u8 io_id;
513 
514 		io_id = id - io_start_id;
515 		if (io_id < sst_inst->io_dies[i]) {
516 			u8 cdie_range;
517 
518 			cdie_range = fls(sst_inst->cdie_mask[i]) - ffs(sst_inst->cdie_mask[i]) + 1;
519 			*partition = i;
520 			return cdie_range + io_id;
521 		}
522 		io_start_id += sst_inst->io_dies[i];
523 	}
524 
525 	return -EINVAL;
526 }
527 
528 /*
529  * Map a package and power_domain id to SST information structure unique for a power_domain.
530  * The caller should call under isst_tpmi_dev_lock.
531  */
532 static struct tpmi_per_power_domain_info *get_instance(int pkg_id, int power_domain_id)
533 {
534 	struct tpmi_per_power_domain_info *power_domain_info;
535 	struct tpmi_sst_struct *sst_inst;
536 	u8 part;
537 
538 	if (!in_range(pkg_id, 0, topology_max_packages()) || pkg_id > isst_common.max_index)
539 		return NULL;
540 
541 	sst_inst = isst_common.sst_inst[pkg_id];
542 	if (!sst_inst)
543 		return NULL;
544 
545 	power_domain_id = map_partition_power_domain_id(sst_inst, power_domain_id, &part);
546 	if (power_domain_id < 0)
547 		return NULL;
548 
549 	power_domain_info = &sst_inst->power_domain_info[part][power_domain_id];
550 
551 	if (power_domain_info && !power_domain_info->sst_base)
552 		return NULL;
553 
554 	return power_domain_info;
555 }
556 
557 static bool disable_dynamic_sst_features(void)
558 {
559 	u64 value;
560 
561 	if (!static_cpu_has(X86_FEATURE_HWP))
562 		return true;
563 
564 	rdmsrq(MSR_PM_ENABLE, value);
565 	return !(value & 0x1);
566 }
567 
568 #define _read_cp_info(name_str, name, offset, start, width, mult_factor)\
569 {\
570 	u64 val, mask;\
571 	\
572 	val = readq(power_domain_info->sst_base + power_domain_info->sst_header.cp_offset +\
573 			(offset));\
574 	mask = GENMASK_ULL((start + width - 1), start);\
575 	val &= mask; \
576 	val >>= start;\
577 	name = (val * mult_factor);\
578 }
579 
580 #define _write_cp_info(name_str, name, offset, start, width, div_factor)\
581 {\
582 	u64 val, mask;\
583 	\
584 	val = readq(power_domain_info->sst_base +\
585 		    power_domain_info->sst_header.cp_offset + (offset));\
586 	mask = GENMASK_ULL((start + width - 1), start);\
587 	val &= ~mask;\
588 	val |= (name / div_factor) << start;\
589 	writeq(val, power_domain_info->sst_base + power_domain_info->sst_header.cp_offset +\
590 		(offset));\
591 }
592 
593 #define	SST_CP_CONTROL_OFFSET	8
594 #define	SST_CP_STATUS_OFFSET	16
595 
596 #define SST_CP_ENABLE_START		0
597 #define SST_CP_ENABLE_WIDTH		1
598 
599 #define SST_CP_PRIORITY_TYPE_START	1
600 #define SST_CP_PRIORITY_TYPE_WIDTH	1
601 
602 static long isst_if_core_power_state(void __user *argp)
603 {
604 	struct tpmi_per_power_domain_info *power_domain_info;
605 	struct isst_core_power core_power;
606 
607 	if (copy_from_user(&core_power, argp, sizeof(core_power)))
608 		return -EFAULT;
609 
610 	if (core_power.get_set && disable_dynamic_sst_features())
611 		return -EFAULT;
612 
613 	power_domain_info = get_instance(core_power.socket_id, core_power.power_domain_id);
614 	if (!power_domain_info)
615 		return -EINVAL;
616 
617 	if (core_power.get_set) {
618 		if (power_domain_info->write_blocked || !capable(CAP_SYS_ADMIN))
619 			return -EPERM;
620 
621 		_write_cp_info("cp_enable", core_power.enable, SST_CP_CONTROL_OFFSET,
622 			       SST_CP_ENABLE_START, SST_CP_ENABLE_WIDTH, SST_MUL_FACTOR_NONE)
623 		_write_cp_info("cp_prio_type", core_power.priority_type, SST_CP_CONTROL_OFFSET,
624 			       SST_CP_PRIORITY_TYPE_START, SST_CP_PRIORITY_TYPE_WIDTH,
625 			       SST_MUL_FACTOR_NONE)
626 	} else {
627 		/* get */
628 		_read_cp_info("cp_enable", core_power.enable, SST_CP_STATUS_OFFSET,
629 			      SST_CP_ENABLE_START, SST_CP_ENABLE_WIDTH, SST_MUL_FACTOR_NONE)
630 		_read_cp_info("cp_prio_type", core_power.priority_type, SST_CP_STATUS_OFFSET,
631 			      SST_CP_PRIORITY_TYPE_START, SST_CP_PRIORITY_TYPE_WIDTH,
632 			      SST_MUL_FACTOR_NONE)
633 		core_power.supported = !!(power_domain_info->sst_header.cap_mask & BIT(0));
634 		if (copy_to_user(argp, &core_power, sizeof(core_power)))
635 			return -EFAULT;
636 	}
637 
638 	return 0;
639 }
640 
641 #define SST_CLOS_CONFIG_0_OFFSET	24
642 
643 #define SST_CLOS_CONFIG_PRIO_START	4
644 #define SST_CLOS_CONFIG_PRIO_WIDTH	4
645 
646 #define SST_CLOS_CONFIG_MIN_START	8
647 #define SST_CLOS_CONFIG_MIN_WIDTH	8
648 
649 #define SST_CLOS_CONFIG_MAX_START	16
650 #define SST_CLOS_CONFIG_MAX_WIDTH	8
651 
652 static long isst_if_clos_param(void __user *argp)
653 {
654 	struct tpmi_per_power_domain_info *power_domain_info;
655 	struct isst_clos_param clos_param;
656 
657 	if (copy_from_user(&clos_param, argp, sizeof(clos_param)))
658 		return -EFAULT;
659 
660 	power_domain_info = get_instance(clos_param.socket_id, clos_param.power_domain_id);
661 	if (!power_domain_info)
662 		return -EINVAL;
663 
664 	if (clos_param.get_set) {
665 		if (power_domain_info->write_blocked || !capable(CAP_SYS_ADMIN))
666 			return -EPERM;
667 
668 		_write_cp_info("clos.min_freq", clos_param.min_freq_mhz,
669 			       (SST_CLOS_CONFIG_0_OFFSET + clos_param.clos * SST_REG_SIZE),
670 			       SST_CLOS_CONFIG_MIN_START, SST_CLOS_CONFIG_MIN_WIDTH,
671 			       SST_MUL_FACTOR_FREQ);
672 		_write_cp_info("clos.max_freq", clos_param.max_freq_mhz,
673 			       (SST_CLOS_CONFIG_0_OFFSET + clos_param.clos * SST_REG_SIZE),
674 			       SST_CLOS_CONFIG_MAX_START, SST_CLOS_CONFIG_MAX_WIDTH,
675 			       SST_MUL_FACTOR_FREQ);
676 		_write_cp_info("clos.prio", clos_param.prop_prio,
677 			       (SST_CLOS_CONFIG_0_OFFSET + clos_param.clos * SST_REG_SIZE),
678 			       SST_CLOS_CONFIG_PRIO_START, SST_CLOS_CONFIG_PRIO_WIDTH,
679 			       SST_MUL_FACTOR_NONE);
680 	} else {
681 		/* get */
682 		_read_cp_info("clos.min_freq", clos_param.min_freq_mhz,
683 				(SST_CLOS_CONFIG_0_OFFSET + clos_param.clos * SST_REG_SIZE),
684 				SST_CLOS_CONFIG_MIN_START, SST_CLOS_CONFIG_MIN_WIDTH,
685 				SST_MUL_FACTOR_FREQ)
686 		_read_cp_info("clos.max_freq", clos_param.max_freq_mhz,
687 				(SST_CLOS_CONFIG_0_OFFSET + clos_param.clos * SST_REG_SIZE),
688 				SST_CLOS_CONFIG_MAX_START, SST_CLOS_CONFIG_MAX_WIDTH,
689 				SST_MUL_FACTOR_FREQ)
690 		_read_cp_info("clos.prio", clos_param.prop_prio,
691 				(SST_CLOS_CONFIG_0_OFFSET + clos_param.clos * SST_REG_SIZE),
692 				SST_CLOS_CONFIG_PRIO_START, SST_CLOS_CONFIG_PRIO_WIDTH,
693 				SST_MUL_FACTOR_NONE)
694 
695 		if (copy_to_user(argp, &clos_param, sizeof(clos_param)))
696 			return -EFAULT;
697 	}
698 
699 	return 0;
700 }
701 
702 #define SST_CLOS_ASSOC_0_OFFSET		56
703 #define SST_CLOS_ASSOC_CPUS_PER_REG	16
704 #define SST_CLOS_ASSOC_BITS_PER_CPU	4
705 
706 static long isst_if_clos_assoc(void __user *argp)
707 {
708 	struct isst_if_clos_assoc_cmds assoc_cmds;
709 	unsigned char __user *ptr;
710 	int i;
711 
712 	/* Each multi command has u16 command count as the first field */
713 	if (copy_from_user(&assoc_cmds, argp, sizeof(assoc_cmds)))
714 		return -EFAULT;
715 
716 	if (!assoc_cmds.cmd_count || assoc_cmds.cmd_count > ISST_IF_CMD_LIMIT)
717 		return -EINVAL;
718 
719 	ptr = argp + offsetof(struct isst_if_clos_assoc_cmds, assoc_info);
720 	for (i = 0; i < assoc_cmds.cmd_count; ++i) {
721 		struct tpmi_per_power_domain_info *power_domain_info;
722 		struct isst_if_clos_assoc clos_assoc;
723 		int punit_id, punit_cpu_no, pkg_id;
724 		struct tpmi_sst_struct *sst_inst;
725 		int offset, shift, cpu;
726 		u64 val, mask, clos;
727 		u8 part;
728 
729 		if (copy_from_user(&clos_assoc, ptr, sizeof(clos_assoc)))
730 			return -EFAULT;
731 
732 		if (clos_assoc.socket_id > topology_max_packages())
733 			return -EINVAL;
734 
735 		cpu = clos_assoc.logical_cpu;
736 		clos = clos_assoc.clos;
737 
738 		if (assoc_cmds.punit_cpu_map)
739 			punit_cpu_no = cpu;
740 		else
741 			return -EOPNOTSUPP;
742 
743 		if (punit_cpu_no < 0)
744 			return -EINVAL;
745 
746 		punit_id = clos_assoc.power_domain_id;
747 		pkg_id = clos_assoc.socket_id;
748 
749 		sst_inst = isst_common.sst_inst[pkg_id];
750 
751 		punit_id = map_partition_power_domain_id(sst_inst, punit_id, &part);
752 		if (punit_id < 0)
753 			return -EINVAL;
754 
755 		power_domain_info = &sst_inst->power_domain_info[part][punit_id];
756 
757 		if (assoc_cmds.get_set && (power_domain_info->write_blocked ||
758 					   !capable(CAP_SYS_ADMIN)))
759 			return -EPERM;
760 
761 		offset = SST_CLOS_ASSOC_0_OFFSET +
762 				(punit_cpu_no / SST_CLOS_ASSOC_CPUS_PER_REG) * SST_REG_SIZE;
763 		shift = punit_cpu_no % SST_CLOS_ASSOC_CPUS_PER_REG;
764 		shift *= SST_CLOS_ASSOC_BITS_PER_CPU;
765 
766 		val = readq(power_domain_info->sst_base +
767 				power_domain_info->sst_header.cp_offset + offset);
768 		if (assoc_cmds.get_set) {
769 			mask = GENMASK_ULL((shift + SST_CLOS_ASSOC_BITS_PER_CPU - 1), shift);
770 			val &= ~mask;
771 			val |= (clos << shift);
772 			writeq(val, power_domain_info->sst_base +
773 					power_domain_info->sst_header.cp_offset + offset);
774 		} else {
775 			val >>= shift;
776 			clos_assoc.clos = val & GENMASK(SST_CLOS_ASSOC_BITS_PER_CPU - 1, 0);
777 			if (copy_to_user(ptr, &clos_assoc, sizeof(clos_assoc)))
778 				return -EFAULT;
779 		}
780 
781 		ptr += sizeof(clos_assoc);
782 	}
783 
784 	return 0;
785 }
786 
787 #define _read_pp_info(name_str, name, offset, start, width, mult_factor)\
788 {\
789 	u64 val, _mask;\
790 	\
791 	val = readq(power_domain_info->sst_base + power_domain_info->sst_header.pp_offset +\
792 		    (offset));\
793 	_mask = GENMASK_ULL((start + width - 1), start);\
794 	val &= _mask;\
795 	val >>= start;\
796 	name = (val * mult_factor);\
797 }
798 
799 #define _write_pp_info(name_str, name, offset, start, width, div_factor)\
800 {\
801 	u64 val, _mask;\
802 	\
803 	val = readq(power_domain_info->sst_base + power_domain_info->sst_header.pp_offset +\
804 		    (offset));\
805 	_mask = GENMASK((start + width - 1), start);\
806 	val &= ~_mask;\
807 	val |= (name / div_factor) << start;\
808 	writeq(val, power_domain_info->sst_base + power_domain_info->sst_header.pp_offset +\
809 	      (offset));\
810 }
811 
812 #define _read_bf_level_info(name_str, name, level, offset, start, width, mult_factor)\
813 {\
814 	u64 val, _mask;\
815 	\
816 	val = readq(power_domain_info->sst_base +\
817 		    power_domain_info->perf_levels[level].mmio_offset +\
818 		(power_domain_info->feature_offsets.bf_offset * 8) + (offset));\
819 	_mask = GENMASK_ULL((start + width - 1), start);\
820 	val &= _mask; \
821 	val >>= start;\
822 	name = (val * mult_factor);\
823 }
824 
825 #define _read_tf_level_info(name_str, name, level, offset, start, width, mult_factor)\
826 {\
827 	u64 val, _mask;\
828 	\
829 	val = readq(power_domain_info->sst_base +\
830 		    power_domain_info->perf_levels[level].mmio_offset +\
831 		(power_domain_info->feature_offsets.tf_offset * 8) + (offset));\
832 	_mask = GENMASK_ULL((start + width - 1), start);\
833 	val &= _mask; \
834 	val >>= start;\
835 	name = (val * mult_factor);\
836 }
837 
838 #define SST_PP_STATUS_OFFSET	32
839 
840 #define SST_PP_LEVEL_START	0
841 #define SST_PP_LEVEL_WIDTH	3
842 
843 #define SST_PP_LOCK_START	3
844 #define SST_PP_LOCK_WIDTH	1
845 
846 #define SST_PP_FEATURE_STATE_START	8
847 #define SST_PP_FEATURE_STATE_WIDTH	8
848 
849 #define SST_BF_FEATURE_SUPPORTED_START	12
850 #define SST_BF_FEATURE_SUPPORTED_WIDTH	1
851 
852 #define SST_TF_FEATURE_SUPPORTED_START	12
853 #define SST_TF_FEATURE_SUPPORTED_WIDTH	1
854 
855 static int isst_if_get_perf_level(void __user *argp)
856 {
857 	struct isst_perf_level_info perf_level;
858 	struct tpmi_per_power_domain_info *power_domain_info;
859 	unsigned long level_mask;
860 	u8 level, support;
861 
862 	if (copy_from_user(&perf_level, argp, sizeof(perf_level)))
863 		return -EFAULT;
864 
865 	power_domain_info = get_instance(perf_level.socket_id, perf_level.power_domain_id);
866 	if (!power_domain_info)
867 		return -EINVAL;
868 
869 	perf_level.max_level = power_domain_info->max_level;
870 	perf_level.level_mask = power_domain_info->pp_header.level_en_mask;
871 	perf_level.feature_rev = power_domain_info->pp_header.feature_rev;
872 	_read_pp_info("current_level", perf_level.current_level, SST_PP_STATUS_OFFSET,
873 		      SST_PP_LEVEL_START, SST_PP_LEVEL_WIDTH, SST_MUL_FACTOR_NONE)
874 	_read_pp_info("locked", perf_level.locked, SST_PP_STATUS_OFFSET,
875 		      SST_PP_LOCK_START, SST_PP_LOCK_WIDTH, SST_MUL_FACTOR_NONE)
876 	_read_pp_info("feature_state", perf_level.feature_state, SST_PP_STATUS_OFFSET,
877 		      SST_PP_FEATURE_STATE_START, SST_PP_FEATURE_STATE_WIDTH, SST_MUL_FACTOR_NONE)
878 	perf_level.enabled = !!(power_domain_info->sst_header.cap_mask & BIT(1));
879 
880 	level_mask = perf_level.level_mask;
881 	perf_level.sst_bf_support = 0;
882 	for_each_set_bit(level, &level_mask, BITS_PER_BYTE) {
883 		/*
884 		 * Read BF support for a level. Read output is updated
885 		 * to "support" variable by the below macro.
886 		 */
887 		_read_bf_level_info("bf_support", support, level, 0, SST_BF_FEATURE_SUPPORTED_START,
888 				    SST_BF_FEATURE_SUPPORTED_WIDTH, SST_MUL_FACTOR_NONE);
889 
890 		/* If supported set the bit for the level */
891 		if (support)
892 			perf_level.sst_bf_support |= BIT(level);
893 	}
894 
895 	perf_level.sst_tf_support = 0;
896 	for_each_set_bit(level, &level_mask, BITS_PER_BYTE) {
897 		/*
898 		 * Read TF support for a level. Read output is updated
899 		 * to "support" variable by the below macro.
900 		 */
901 		_read_tf_level_info("tf_support", support, level, 0, SST_TF_FEATURE_SUPPORTED_START,
902 				    SST_TF_FEATURE_SUPPORTED_WIDTH, SST_MUL_FACTOR_NONE);
903 
904 		/* If supported set the bit for the level */
905 		if (support)
906 			perf_level.sst_tf_support |= BIT(level);
907 	}
908 
909 	if (copy_to_user(argp, &perf_level, sizeof(perf_level)))
910 		return -EFAULT;
911 
912 	return 0;
913 }
914 
915 #define SST_PP_CONTROL_OFFSET		24
916 #define SST_PP_LEVEL_CHANGE_TIME_MS	5
917 #define SST_PP_LEVEL_CHANGE_RETRY_COUNT	3
918 
919 static int isst_if_set_perf_level(void __user *argp)
920 {
921 	struct isst_perf_level_control perf_level;
922 	struct tpmi_per_power_domain_info *power_domain_info;
923 	int level, retry = 0;
924 
925 	if (disable_dynamic_sst_features())
926 		return -EFAULT;
927 
928 	if (copy_from_user(&perf_level, argp, sizeof(perf_level)))
929 		return -EFAULT;
930 
931 	power_domain_info = get_instance(perf_level.socket_id, perf_level.power_domain_id);
932 	if (!power_domain_info)
933 		return -EINVAL;
934 
935 	if (power_domain_info->write_blocked || !capable(CAP_SYS_ADMIN))
936 		return -EPERM;
937 
938 	if (!(power_domain_info->pp_header.allowed_level_mask & BIT(perf_level.level)))
939 		return -EINVAL;
940 
941 	_read_pp_info("current_level", level, SST_PP_STATUS_OFFSET,
942 		      SST_PP_LEVEL_START, SST_PP_LEVEL_WIDTH, SST_MUL_FACTOR_NONE)
943 
944 	/* If the requested new level is same as the current level, reject */
945 	if (perf_level.level == level)
946 		return -EINVAL;
947 
948 	_write_pp_info("perf_level", perf_level.level, SST_PP_CONTROL_OFFSET,
949 		       SST_PP_LEVEL_START, SST_PP_LEVEL_WIDTH, SST_MUL_FACTOR_NONE)
950 
951 	/* It is possible that firmware is busy (although unlikely), so retry */
952 	do {
953 		/* Give time to FW to process */
954 		msleep(SST_PP_LEVEL_CHANGE_TIME_MS);
955 
956 		_read_pp_info("current_level", level, SST_PP_STATUS_OFFSET,
957 			      SST_PP_LEVEL_START, SST_PP_LEVEL_WIDTH, SST_MUL_FACTOR_NONE)
958 
959 		/* Check if the new level is active */
960 		if (perf_level.level == level)
961 			break;
962 
963 	} while (retry++ < SST_PP_LEVEL_CHANGE_RETRY_COUNT);
964 
965 	/* If the level change didn't happen, return fault */
966 	if (perf_level.level != level)
967 		return -EFAULT;
968 
969 	/* Reset the feature state on level change */
970 	_write_pp_info("perf_feature", 0, SST_PP_CONTROL_OFFSET,
971 		       SST_PP_FEATURE_STATE_START, SST_PP_FEATURE_STATE_WIDTH,
972 		       SST_MUL_FACTOR_NONE)
973 
974 	/* Give time to FW to process */
975 	msleep(SST_PP_LEVEL_CHANGE_TIME_MS);
976 
977 	return 0;
978 }
979 
980 static int isst_if_set_perf_feature(void __user *argp)
981 {
982 	struct isst_perf_feature_control perf_feature;
983 	struct tpmi_per_power_domain_info *power_domain_info;
984 
985 	if (disable_dynamic_sst_features())
986 		return -EFAULT;
987 
988 	if (copy_from_user(&perf_feature, argp, sizeof(perf_feature)))
989 		return -EFAULT;
990 
991 	power_domain_info = get_instance(perf_feature.socket_id, perf_feature.power_domain_id);
992 	if (!power_domain_info)
993 		return -EINVAL;
994 
995 	if (power_domain_info->write_blocked || !capable(CAP_SYS_ADMIN))
996 		return -EPERM;
997 
998 	_write_pp_info("perf_feature", perf_feature.feature, SST_PP_CONTROL_OFFSET,
999 		       SST_PP_FEATURE_STATE_START, SST_PP_FEATURE_STATE_WIDTH,
1000 		       SST_MUL_FACTOR_NONE)
1001 
1002 	return 0;
1003 }
1004 
1005 #define _read_pp_level_info(name_str, name, level, offset, start, width, mult_factor)\
1006 {\
1007 	u64 val, _mask;\
1008 	\
1009 	val = readq(power_domain_info->sst_base +\
1010 		    power_domain_info->perf_levels[level].mmio_offset +\
1011 		(power_domain_info->feature_offsets.pp_offset * 8) + (offset));\
1012 	_mask = GENMASK_ULL((start + width - 1), start);\
1013 	val &= _mask; \
1014 	val >>= start;\
1015 	name = (val * mult_factor);\
1016 }
1017 
1018 #define SST_PP_INFO_0_OFFSET	0
1019 #define SST_PP_INFO_1_OFFSET	8
1020 #define SST_PP_INFO_2_OFFSET	16
1021 #define SST_PP_INFO_3_OFFSET	24
1022 
1023 /* SST_PP_INFO_4_OFFSET to SST_PP_INFO_9_OFFSET are trl levels */
1024 #define SST_PP_INFO_4_OFFSET	32
1025 
1026 #define SST_PP_INFO_10_OFFSET	80
1027 #define SST_PP_INFO_11_OFFSET	88
1028 #define SST_PP_INFO_12_OFFSET	96
1029 
1030 #define SST_PP_P1_SSE_START	0
1031 #define SST_PP_P1_SSE_WIDTH	8
1032 
1033 #define SST_PP_P1_AVX2_START	8
1034 #define SST_PP_P1_AVX2_WIDTH	8
1035 
1036 #define SST_PP_P1_AVX512_START	16
1037 #define SST_PP_P1_AVX512_WIDTH	8
1038 
1039 #define SST_PP_P1_AMX_START	24
1040 #define SST_PP_P1_AMX_WIDTH	8
1041 
1042 #define SST_PP_TDP_START	32
1043 #define SST_PP_TDP_WIDTH	15
1044 
1045 #define SST_PP_T_PROCHOT_START	47
1046 #define SST_PP_T_PROCHOT_WIDTH	8
1047 
1048 #define SST_PP_MAX_MEMORY_FREQ_START	55
1049 #define SST_PP_MAX_MEMORY_FREQ_WIDTH	7
1050 
1051 #define SST_PP_COOLING_TYPE_START	62
1052 #define SST_PP_COOLING_TYPE_WIDTH	2
1053 
1054 #define SST_PP_TRL_0_RATIO_0_START	0
1055 #define SST_PP_TRL_0_RATIO_0_WIDTH	8
1056 
1057 #define SST_PP_TRL_CORES_BUCKET_0_START	0
1058 #define SST_PP_TRL_CORES_BUCKET_0_WIDTH	8
1059 
1060 #define SST_PP_CORE_RATIO_P0_START	0
1061 #define SST_PP_CORE_RATIO_P0_WIDTH	8
1062 
1063 #define SST_PP_CORE_RATIO_P1_START	8
1064 #define SST_PP_CORE_RATIO_P1_WIDTH	8
1065 
1066 #define SST_PP_CORE_RATIO_PN_START	16
1067 #define SST_PP_CORE_RATIO_PN_WIDTH	8
1068 
1069 #define SST_PP_CORE_RATIO_PM_START	24
1070 #define SST_PP_CORE_RATIO_PM_WIDTH	8
1071 
1072 #define SST_PP_CORE_RATIO_P0_FABRIC_START	32
1073 #define SST_PP_CORE_RATIO_P0_FABRIC_WIDTH	8
1074 
1075 #define SST_PP_CORE_RATIO_P1_FABRIC_START	40
1076 #define SST_PP_CORE_RATIO_P1_FABRIC_WIDTH	8
1077 
1078 #define SST_PP_CORE_RATIO_PM_FABRIC_START	48
1079 #define SST_PP_CORE_RATIO_PM_FABRIC_WIDTH	8
1080 
1081 #define SST_PP_CORE_RATIO_P0_FABRIC_1_START	0
1082 #define SST_PP_CORE_RATIO_P0_FABRIC_1_WIDTH	8
1083 
1084 #define SST_PP_CORE_RATIO_P1_FABRIC_1_START	8
1085 #define SST_PP_CORE_RATIO_P1_FABRIC_1_WIDTH	8
1086 
1087 #define SST_PP_CORE_RATIO_PM_FABRIC_1_START	16
1088 #define SST_PP_CORE_RATIO_PM_FABRIC_1_WIDTH	8
1089 
1090 static int isst_if_get_perf_level_info(void __user *argp)
1091 {
1092 	struct isst_perf_level_data_info perf_level;
1093 	struct tpmi_per_power_domain_info *power_domain_info;
1094 	int i, j;
1095 
1096 	if (copy_from_user(&perf_level, argp, sizeof(perf_level)))
1097 		return -EFAULT;
1098 
1099 	power_domain_info = get_instance(perf_level.socket_id, perf_level.power_domain_id);
1100 	if (!power_domain_info)
1101 		return -EINVAL;
1102 
1103 	if (perf_level.level > power_domain_info->max_level)
1104 		return -EINVAL;
1105 
1106 	if (!(power_domain_info->pp_header.level_en_mask & BIT(perf_level.level)))
1107 		return -EINVAL;
1108 
1109 	_read_pp_level_info("tdp_ratio", perf_level.tdp_ratio, perf_level.level,
1110 			    SST_PP_INFO_0_OFFSET, SST_PP_P1_SSE_START, SST_PP_P1_SSE_WIDTH,
1111 			    SST_MUL_FACTOR_NONE)
1112 	_read_pp_level_info("base_freq_mhz", perf_level.base_freq_mhz, perf_level.level,
1113 			    SST_PP_INFO_0_OFFSET, SST_PP_P1_SSE_START, SST_PP_P1_SSE_WIDTH,
1114 			    SST_MUL_FACTOR_FREQ)
1115 	_read_pp_level_info("base_freq_avx2_mhz", perf_level.base_freq_avx2_mhz, perf_level.level,
1116 			    SST_PP_INFO_0_OFFSET, SST_PP_P1_AVX2_START, SST_PP_P1_AVX2_WIDTH,
1117 			    SST_MUL_FACTOR_FREQ)
1118 	_read_pp_level_info("base_freq_avx512_mhz", perf_level.base_freq_avx512_mhz,
1119 			    perf_level.level, SST_PP_INFO_0_OFFSET, SST_PP_P1_AVX512_START,
1120 			    SST_PP_P1_AVX512_WIDTH, SST_MUL_FACTOR_FREQ)
1121 	_read_pp_level_info("base_freq_amx_mhz", perf_level.base_freq_amx_mhz, perf_level.level,
1122 			    SST_PP_INFO_0_OFFSET, SST_PP_P1_AMX_START, SST_PP_P1_AMX_WIDTH,
1123 			    SST_MUL_FACTOR_FREQ)
1124 
1125 	_read_pp_level_info("thermal_design_power_w", perf_level.thermal_design_power_w,
1126 			    perf_level.level, SST_PP_INFO_1_OFFSET, SST_PP_TDP_START,
1127 			    SST_PP_TDP_WIDTH, SST_MUL_FACTOR_NONE)
1128 	perf_level.thermal_design_power_w /= 8; /* units are in 1/8th watt */
1129 	_read_pp_level_info("tjunction_max_c", perf_level.tjunction_max_c, perf_level.level,
1130 			    SST_PP_INFO_1_OFFSET, SST_PP_T_PROCHOT_START, SST_PP_T_PROCHOT_WIDTH,
1131 			    SST_MUL_FACTOR_NONE)
1132 	_read_pp_level_info("max_memory_freq_mhz", perf_level.max_memory_freq_mhz,
1133 			    perf_level.level, SST_PP_INFO_1_OFFSET, SST_PP_MAX_MEMORY_FREQ_START,
1134 			    SST_PP_MAX_MEMORY_FREQ_WIDTH, SST_MUL_FACTOR_FREQ)
1135 	_read_pp_level_info("cooling_type", perf_level.cooling_type, perf_level.level,
1136 			    SST_PP_INFO_1_OFFSET, SST_PP_COOLING_TYPE_START,
1137 			    SST_PP_COOLING_TYPE_WIDTH, SST_MUL_FACTOR_NONE)
1138 
1139 	for (i = 0; i < TRL_MAX_LEVELS; ++i) {
1140 		for (j = 0; j < TRL_MAX_BUCKETS; ++j)
1141 			_read_pp_level_info("trl*_bucket*_freq_mhz",
1142 					    perf_level.trl_freq_mhz[i][j], perf_level.level,
1143 					    SST_PP_INFO_4_OFFSET + (i * SST_PP_TRL_0_RATIO_0_WIDTH),
1144 					    j * SST_PP_TRL_0_RATIO_0_WIDTH,
1145 					    SST_PP_TRL_0_RATIO_0_WIDTH,
1146 					    SST_MUL_FACTOR_FREQ);
1147 	}
1148 
1149 	for (i = 0; i < TRL_MAX_BUCKETS; ++i)
1150 		_read_pp_level_info("bucket*_core_count", perf_level.bucket_core_counts[i],
1151 				    perf_level.level, SST_PP_INFO_10_OFFSET,
1152 				    SST_PP_TRL_CORES_BUCKET_0_WIDTH * i,
1153 				    SST_PP_TRL_CORES_BUCKET_0_WIDTH, SST_MUL_FACTOR_NONE)
1154 
1155 	perf_level.max_buckets = TRL_MAX_BUCKETS;
1156 	perf_level.max_trl_levels = TRL_MAX_LEVELS;
1157 
1158 	_read_pp_level_info("p0_freq_mhz", perf_level.p0_freq_mhz, perf_level.level,
1159 			    SST_PP_INFO_11_OFFSET, SST_PP_CORE_RATIO_P0_START,
1160 			    SST_PP_CORE_RATIO_P0_WIDTH, SST_MUL_FACTOR_FREQ)
1161 	_read_pp_level_info("p1_freq_mhz", perf_level.p1_freq_mhz, perf_level.level,
1162 			    SST_PP_INFO_11_OFFSET, SST_PP_CORE_RATIO_P1_START,
1163 			    SST_PP_CORE_RATIO_P1_WIDTH, SST_MUL_FACTOR_FREQ)
1164 	_read_pp_level_info("pn_freq_mhz", perf_level.pn_freq_mhz, perf_level.level,
1165 			    SST_PP_INFO_11_OFFSET, SST_PP_CORE_RATIO_PN_START,
1166 			    SST_PP_CORE_RATIO_PN_WIDTH, SST_MUL_FACTOR_FREQ)
1167 	_read_pp_level_info("pm_freq_mhz", perf_level.pm_freq_mhz, perf_level.level,
1168 			    SST_PP_INFO_11_OFFSET, SST_PP_CORE_RATIO_PM_START,
1169 			    SST_PP_CORE_RATIO_PM_WIDTH, SST_MUL_FACTOR_FREQ)
1170 	_read_pp_level_info("p0_fabric_freq_mhz", perf_level.p0_fabric_freq_mhz,
1171 			    perf_level.level, SST_PP_INFO_11_OFFSET,
1172 			    SST_PP_CORE_RATIO_P0_FABRIC_START,
1173 			    SST_PP_CORE_RATIO_P0_FABRIC_WIDTH, SST_MUL_FACTOR_FREQ)
1174 	_read_pp_level_info("p1_fabric_freq_mhz", perf_level.p1_fabric_freq_mhz,
1175 			    perf_level.level, SST_PP_INFO_11_OFFSET,
1176 			    SST_PP_CORE_RATIO_P1_FABRIC_START,
1177 			    SST_PP_CORE_RATIO_P1_FABRIC_WIDTH, SST_MUL_FACTOR_FREQ)
1178 	_read_pp_level_info("pm_fabric_freq_mhz", perf_level.pm_fabric_freq_mhz,
1179 			    perf_level.level, SST_PP_INFO_11_OFFSET,
1180 			    SST_PP_CORE_RATIO_PM_FABRIC_START,
1181 			    SST_PP_CORE_RATIO_PM_FABRIC_WIDTH, SST_MUL_FACTOR_FREQ)
1182 
1183 	if (copy_to_user(argp, &perf_level, sizeof(perf_level)))
1184 		return -EFAULT;
1185 
1186 	return 0;
1187 }
1188 
1189 static int isst_if_get_perf_level_fabric_info(void __user *argp)
1190 {
1191 	struct isst_perf_level_fabric_info perf_level_fabric;
1192 	struct tpmi_per_power_domain_info *power_domain_info;
1193 	int start = SST_PP_CORE_RATIO_P0_FABRIC_START;
1194 	int width = SST_PP_CORE_RATIO_P0_FABRIC_WIDTH;
1195 	int offset = SST_PP_INFO_11_OFFSET;
1196 	int i;
1197 
1198 	if (copy_from_user(&perf_level_fabric, argp, sizeof(perf_level_fabric)))
1199 		return -EFAULT;
1200 
1201 	power_domain_info = get_instance(perf_level_fabric.socket_id,
1202 					 perf_level_fabric.power_domain_id);
1203 	if (!power_domain_info)
1204 		return -EINVAL;
1205 
1206 	if (perf_level_fabric.level > power_domain_info->max_level)
1207 		return -EINVAL;
1208 
1209 	if (power_domain_info->pp_header.feature_rev < 2)
1210 		return -EINVAL;
1211 
1212 	if (!(power_domain_info->pp_header.level_en_mask & BIT(perf_level_fabric.level)))
1213 		return -EINVAL;
1214 
1215 	/* For revision 2, maximum number of fabrics is 2 */
1216 	perf_level_fabric.max_fabrics = 2;
1217 
1218 	for (i = 0; i < perf_level_fabric.max_fabrics; i++) {
1219 		_read_pp_level_info("p0_fabric_freq_mhz", perf_level_fabric.p0_fabric_freq_mhz[i],
1220 				    perf_level_fabric.level, offset, start, width,
1221 				    SST_MUL_FACTOR_FREQ)
1222 		start += width;
1223 
1224 		_read_pp_level_info("p1_fabric_freq_mhz", perf_level_fabric.p1_fabric_freq_mhz[i],
1225 				    perf_level_fabric.level, offset, start, width,
1226 				    SST_MUL_FACTOR_FREQ)
1227 		start += width;
1228 
1229 		_read_pp_level_info("pm_fabric_freq_mhz", perf_level_fabric.pm_fabric_freq_mhz[i],
1230 				    perf_level_fabric.level, offset, start, width,
1231 				    SST_MUL_FACTOR_FREQ)
1232 		offset = SST_PP_INFO_12_OFFSET;
1233 		start = SST_PP_CORE_RATIO_P0_FABRIC_1_START;
1234 	}
1235 
1236 	if (copy_to_user(argp, &perf_level_fabric, sizeof(perf_level_fabric)))
1237 		return -EFAULT;
1238 
1239 	return 0;
1240 }
1241 
1242 #define SST_PP_FUSED_CORE_COUNT_START	0
1243 #define SST_PP_FUSED_CORE_COUNT_WIDTH	8
1244 
1245 #define SST_PP_RSLVD_CORE_COUNT_START	8
1246 #define SST_PP_RSLVD_CORE_COUNT_WIDTH	8
1247 
1248 #define SST_PP_RSLVD_CORE_MASK_START	0
1249 #define SST_PP_RSLVD_CORE_MASK_WIDTH	64
1250 
1251 static int isst_if_get_perf_level_mask(void __user *argp)
1252 {
1253 	static struct isst_perf_level_cpu_mask cpumask;
1254 	struct tpmi_per_power_domain_info *power_domain_info;
1255 	u64 mask;
1256 
1257 	if (copy_from_user(&cpumask, argp, sizeof(cpumask)))
1258 		return -EFAULT;
1259 
1260 	power_domain_info = get_instance(cpumask.socket_id, cpumask.power_domain_id);
1261 	if (!power_domain_info)
1262 		return -EINVAL;
1263 
1264 	_read_pp_level_info("mask", mask, cpumask.level, SST_PP_INFO_2_OFFSET,
1265 			    SST_PP_RSLVD_CORE_MASK_START, SST_PP_RSLVD_CORE_MASK_WIDTH,
1266 			    SST_MUL_FACTOR_NONE)
1267 
1268 	cpumask.mask = mask;
1269 
1270 	if (!cpumask.punit_cpu_map)
1271 		return -EOPNOTSUPP;
1272 
1273 	if (copy_to_user(argp, &cpumask, sizeof(cpumask)))
1274 		return -EFAULT;
1275 
1276 	return 0;
1277 }
1278 
1279 #define SST_BF_INFO_0_OFFSET	0
1280 #define SST_BF_INFO_1_OFFSET	8
1281 
1282 #define SST_BF_P1_HIGH_START	13
1283 #define SST_BF_P1_HIGH_WIDTH	8
1284 
1285 #define SST_BF_P1_LOW_START	21
1286 #define SST_BF_P1_LOW_WIDTH	8
1287 
1288 #define SST_BF_T_PROHOT_START	38
1289 #define SST_BF_T_PROHOT_WIDTH	8
1290 
1291 #define SST_BF_TDP_START	46
1292 #define SST_BF_TDP_WIDTH	15
1293 
1294 static int isst_if_get_base_freq_info(void __user *argp)
1295 {
1296 	static struct isst_base_freq_info base_freq;
1297 	struct tpmi_per_power_domain_info *power_domain_info;
1298 
1299 	if (copy_from_user(&base_freq, argp, sizeof(base_freq)))
1300 		return -EFAULT;
1301 
1302 	power_domain_info = get_instance(base_freq.socket_id, base_freq.power_domain_id);
1303 	if (!power_domain_info)
1304 		return -EINVAL;
1305 
1306 	if (base_freq.level > power_domain_info->max_level)
1307 		return -EINVAL;
1308 
1309 	_read_bf_level_info("p1_high", base_freq.high_base_freq_mhz, base_freq.level,
1310 			    SST_BF_INFO_0_OFFSET, SST_BF_P1_HIGH_START, SST_BF_P1_HIGH_WIDTH,
1311 			    SST_MUL_FACTOR_FREQ)
1312 	_read_bf_level_info("p1_low", base_freq.low_base_freq_mhz, base_freq.level,
1313 			    SST_BF_INFO_0_OFFSET, SST_BF_P1_LOW_START, SST_BF_P1_LOW_WIDTH,
1314 			    SST_MUL_FACTOR_FREQ)
1315 	_read_bf_level_info("BF-TJ", base_freq.tjunction_max_c, base_freq.level,
1316 			    SST_BF_INFO_0_OFFSET, SST_BF_T_PROHOT_START, SST_BF_T_PROHOT_WIDTH,
1317 			    SST_MUL_FACTOR_NONE)
1318 	_read_bf_level_info("BF-tdp", base_freq.thermal_design_power_w, base_freq.level,
1319 			    SST_BF_INFO_0_OFFSET, SST_BF_TDP_START, SST_BF_TDP_WIDTH,
1320 			    SST_MUL_FACTOR_NONE)
1321 	base_freq.thermal_design_power_w /= 8; /*unit = 1/8th watt*/
1322 
1323 	if (copy_to_user(argp, &base_freq, sizeof(base_freq)))
1324 		return -EFAULT;
1325 
1326 	return 0;
1327 }
1328 
1329 #define P1_HI_CORE_MASK_START	0
1330 #define P1_HI_CORE_MASK_WIDTH	64
1331 
1332 static int isst_if_get_base_freq_mask(void __user *argp)
1333 {
1334 	static struct isst_perf_level_cpu_mask cpumask;
1335 	struct tpmi_per_power_domain_info *power_domain_info;
1336 	u64 mask;
1337 
1338 	if (copy_from_user(&cpumask, argp, sizeof(cpumask)))
1339 		return -EFAULT;
1340 
1341 	power_domain_info = get_instance(cpumask.socket_id, cpumask.power_domain_id);
1342 	if (!power_domain_info)
1343 		return -EINVAL;
1344 
1345 	_read_bf_level_info("BF-cpumask", mask, cpumask.level, SST_BF_INFO_1_OFFSET,
1346 			    P1_HI_CORE_MASK_START, P1_HI_CORE_MASK_WIDTH,
1347 			    SST_MUL_FACTOR_NONE)
1348 
1349 	cpumask.mask = mask;
1350 
1351 	if (!cpumask.punit_cpu_map)
1352 		return -EOPNOTSUPP;
1353 
1354 	if (copy_to_user(argp, &cpumask, sizeof(cpumask)))
1355 		return -EFAULT;
1356 
1357 	return 0;
1358 }
1359 
1360 static int isst_if_get_tpmi_instance_count(void __user *argp)
1361 {
1362 	struct isst_tpmi_instance_count tpmi_inst;
1363 	struct tpmi_sst_struct *sst_inst;
1364 	int i;
1365 
1366 	if (copy_from_user(&tpmi_inst, argp, sizeof(tpmi_inst)))
1367 		return -EFAULT;
1368 
1369 	if (tpmi_inst.socket_id >= topology_max_packages())
1370 		return -EINVAL;
1371 
1372 	sst_inst = isst_common.sst_inst[tpmi_inst.socket_id];
1373 
1374 	tpmi_inst.count = isst_instance_count(sst_inst);
1375 
1376 	tpmi_inst.valid_mask = 0;
1377 	for (i = 0; i < tpmi_inst.count; i++) {
1378 		struct tpmi_per_power_domain_info *pd_info;
1379 		u8 part;
1380 		int pd;
1381 
1382 		pd = map_partition_power_domain_id(sst_inst, i, &part);
1383 		if (pd < 0)
1384 			continue;
1385 
1386 		pd_info = &sst_inst->power_domain_info[part][pd];
1387 		if (pd_info->sst_base)
1388 			tpmi_inst.valid_mask |= BIT(i);
1389 	}
1390 
1391 	if (!tpmi_inst.valid_mask)
1392 		tpmi_inst.count = 0;
1393 
1394 	if (copy_to_user(argp, &tpmi_inst, sizeof(tpmi_inst)))
1395 		return -EFAULT;
1396 
1397 	return 0;
1398 }
1399 
1400 #define SST_TF_INFO_0_OFFSET	0
1401 #define SST_TF_INFO_1_OFFSET	8
1402 #define SST_TF_INFO_2_OFFSET	16
1403 #define SST_TF_INFO_8_OFFSET	64
1404 #define SST_TF_INFO_8_BUCKETS	3
1405 
1406 #define SST_TF_MAX_LP_CLIP_RATIOS	TRL_MAX_LEVELS
1407 
1408 #define SST_TF_FEATURE_REV_START	4
1409 #define SST_TF_FEATURE_REV_WIDTH	8
1410 
1411 #define SST_TF_LP_CLIP_RATIO_0_START	16
1412 #define SST_TF_LP_CLIP_RATIO_0_WIDTH	8
1413 
1414 #define SST_TF_RATIO_0_START	0
1415 #define SST_TF_RATIO_0_WIDTH	8
1416 
1417 #define SST_TF_NUM_CORE_0_START 0
1418 #define SST_TF_NUM_CORE_0_WIDTH 8
1419 
1420 #define SST_TF_NUM_MOD_0_START	0
1421 #define SST_TF_NUM_MOD_0_WIDTH	16
1422 
1423 static int isst_if_get_turbo_freq_info(void __user *argp)
1424 {
1425 	static struct isst_turbo_freq_info turbo_freq;
1426 	struct tpmi_per_power_domain_info *power_domain_info;
1427 	u8 feature_rev;
1428 	int i, j;
1429 
1430 	if (copy_from_user(&turbo_freq, argp, sizeof(turbo_freq)))
1431 		return -EFAULT;
1432 
1433 	power_domain_info = get_instance(turbo_freq.socket_id, turbo_freq.power_domain_id);
1434 	if (!power_domain_info)
1435 		return -EINVAL;
1436 
1437 	if (turbo_freq.level > power_domain_info->max_level)
1438 		return -EINVAL;
1439 
1440 	turbo_freq.max_buckets = TRL_MAX_BUCKETS;
1441 	turbo_freq.max_trl_levels = TRL_MAX_LEVELS;
1442 	turbo_freq.max_clip_freqs = SST_TF_MAX_LP_CLIP_RATIOS;
1443 
1444 	_read_tf_level_info("feature_rev", feature_rev, turbo_freq.level,
1445 			    SST_TF_INFO_0_OFFSET, SST_TF_FEATURE_REV_START,
1446 			    SST_TF_FEATURE_REV_WIDTH, SST_MUL_FACTOR_NONE);
1447 
1448 	for (i = 0; i < turbo_freq.max_clip_freqs; ++i)
1449 		_read_tf_level_info("lp_clip*", turbo_freq.lp_clip_freq_mhz[i],
1450 				    turbo_freq.level, SST_TF_INFO_0_OFFSET,
1451 				    SST_TF_LP_CLIP_RATIO_0_START +
1452 				    (i * SST_TF_LP_CLIP_RATIO_0_WIDTH),
1453 				    SST_TF_LP_CLIP_RATIO_0_WIDTH, SST_MUL_FACTOR_FREQ)
1454 
1455 	for (i = 0; i < TRL_MAX_LEVELS; ++i) {
1456 		for (j = 0; j < TRL_MAX_BUCKETS; ++j)
1457 			_read_tf_level_info("cydn*_bucket_*_trl",
1458 					    turbo_freq.trl_freq_mhz[i][j], turbo_freq.level,
1459 					    SST_TF_INFO_2_OFFSET + (i * SST_TF_RATIO_0_WIDTH),
1460 					    j * SST_TF_RATIO_0_WIDTH, SST_TF_RATIO_0_WIDTH,
1461 					    SST_MUL_FACTOR_FREQ)
1462 	}
1463 
1464 	if (feature_rev >= 2) {
1465 		bool has_tf_info_8 = false;
1466 
1467 		for (i = 0; i < SST_TF_INFO_8_BUCKETS; ++i) {
1468 			_read_tf_level_info("bucket_*_mod_count", turbo_freq.bucket_core_counts[i],
1469 					    turbo_freq.level, SST_TF_INFO_8_OFFSET,
1470 					    SST_TF_NUM_MOD_0_WIDTH * i, SST_TF_NUM_MOD_0_WIDTH,
1471 					    SST_MUL_FACTOR_NONE)
1472 
1473 			if (turbo_freq.bucket_core_counts[i])
1474 				has_tf_info_8 = true;
1475 		}
1476 
1477 		if (has_tf_info_8)
1478 			goto done_core_count;
1479 	}
1480 
1481 	for (i = 0; i < TRL_MAX_BUCKETS; ++i)
1482 		_read_tf_level_info("bucket_*_core_count", turbo_freq.bucket_core_counts[i],
1483 				    turbo_freq.level, SST_TF_INFO_1_OFFSET,
1484 				    SST_TF_NUM_CORE_0_WIDTH * i, SST_TF_NUM_CORE_0_WIDTH,
1485 				    SST_MUL_FACTOR_NONE)
1486 
1487 
1488 done_core_count:
1489 
1490 	if (copy_to_user(argp, &turbo_freq, sizeof(turbo_freq)))
1491 		return -EFAULT;
1492 
1493 	return 0;
1494 }
1495 
1496 static long isst_if_def_ioctl(struct file *file, unsigned int cmd,
1497 			      unsigned long arg)
1498 {
1499 	void __user *argp = (void __user *)arg;
1500 	long ret = -ENOTTY;
1501 
1502 	mutex_lock(&isst_tpmi_dev_lock);
1503 	switch (cmd) {
1504 	case ISST_IF_COUNT_TPMI_INSTANCES:
1505 		ret = isst_if_get_tpmi_instance_count(argp);
1506 		break;
1507 	case ISST_IF_CORE_POWER_STATE:
1508 		ret = isst_if_core_power_state(argp);
1509 		break;
1510 	case ISST_IF_CLOS_PARAM:
1511 		ret = isst_if_clos_param(argp);
1512 		break;
1513 	case ISST_IF_CLOS_ASSOC:
1514 		ret = isst_if_clos_assoc(argp);
1515 		break;
1516 	case ISST_IF_PERF_LEVELS:
1517 		ret = isst_if_get_perf_level(argp);
1518 		break;
1519 	case ISST_IF_PERF_SET_LEVEL:
1520 		ret = isst_if_set_perf_level(argp);
1521 		break;
1522 	case ISST_IF_PERF_SET_FEATURE:
1523 		ret = isst_if_set_perf_feature(argp);
1524 		break;
1525 	case ISST_IF_GET_PERF_LEVEL_INFO:
1526 		ret = isst_if_get_perf_level_info(argp);
1527 		break;
1528 	case ISST_IF_GET_PERF_LEVEL_FABRIC_INFO:
1529 		ret = isst_if_get_perf_level_fabric_info(argp);
1530 		break;
1531 	case ISST_IF_GET_PERF_LEVEL_CPU_MASK:
1532 		ret = isst_if_get_perf_level_mask(argp);
1533 		break;
1534 	case ISST_IF_GET_BASE_FREQ_INFO:
1535 		ret = isst_if_get_base_freq_info(argp);
1536 		break;
1537 	case ISST_IF_GET_BASE_FREQ_CPU_MASK:
1538 		ret = isst_if_get_base_freq_mask(argp);
1539 		break;
1540 	case ISST_IF_GET_TURBO_FREQ_INFO:
1541 		ret = isst_if_get_turbo_freq_info(argp);
1542 		break;
1543 	default:
1544 		break;
1545 	}
1546 	mutex_unlock(&isst_tpmi_dev_lock);
1547 
1548 	return ret;
1549 }
1550 
1551 #define TPMI_SST_AUTO_SUSPEND_DELAY_MS	2000
1552 
1553 int tpmi_sst_dev_add(struct auxiliary_device *auxdev)
1554 {
1555 	struct tpmi_per_power_domain_info *pd_info;
1556 	bool read_blocked = 0, write_blocked = 0;
1557 	struct oobmsm_plat_info *plat_info;
1558 	struct device *dev = &auxdev->dev;
1559 	struct tpmi_sst_struct *tpmi_sst;
1560 	u8 i, num_resources, io_die_cnt;
1561 	int ret, pkg = 0, inst = 0;
1562 	bool first_enum = false;
1563 	u16 cdie_mask;
1564 	u8 partition;
1565 
1566 	ret = tpmi_get_feature_status(auxdev, TPMI_ID_SST, &read_blocked, &write_blocked);
1567 	if (ret)
1568 		dev_info(dev, "Can't read feature status: ignoring read/write blocked status\n");
1569 
1570 	if (read_blocked) {
1571 		dev_info(dev, "Firmware has blocked reads, exiting\n");
1572 		return -ENODEV;
1573 	}
1574 
1575 	plat_info = tpmi_get_platform_data(auxdev);
1576 	if (!plat_info) {
1577 		dev_err(dev, "No platform info\n");
1578 		return -EINVAL;
1579 	}
1580 
1581 	pkg = plat_info->package_id;
1582 	if (pkg >= topology_max_packages()) {
1583 		dev_err(dev, "Invalid package id :%x\n", pkg);
1584 		return -EINVAL;
1585 	}
1586 
1587 	partition = plat_info->partition;
1588 	if (partition >= SST_MAX_PARTITIONS) {
1589 		dev_err(&auxdev->dev, "Invalid partition :%x\n", partition);
1590 		return -EINVAL;
1591 	}
1592 
1593 	num_resources = tpmi_get_resource_count(auxdev);
1594 
1595 	if (!num_resources)
1596 		return -EINVAL;
1597 
1598 	mutex_lock(&isst_tpmi_dev_lock);
1599 
1600 	if (isst_common.sst_inst[pkg]) {
1601 		tpmi_sst = isst_common.sst_inst[pkg];
1602 	} else {
1603 		/*
1604 		 * tpmi_sst instance is for a package. So needs to be
1605 		 * allocated only once for both partitions. We can't use
1606 		 * devm_* allocation here as each partition is a
1607 		 * different device, which can be unbound.
1608 		 */
1609 		tpmi_sst = kzalloc_obj(*tpmi_sst);
1610 		if (!tpmi_sst) {
1611 			ret = -ENOMEM;
1612 			goto unlock_exit;
1613 		}
1614 		first_enum = true;
1615 	}
1616 
1617 	ret = 0;
1618 
1619 	pd_info = devm_kcalloc(dev, num_resources, sizeof(*pd_info), GFP_KERNEL);
1620 	if (!pd_info) {
1621 		ret = -ENOMEM;
1622 		goto unlock_free;
1623 	}
1624 
1625 	/* Get the IO die count, if cdie_mask is present */
1626 	if (plat_info->cdie_mask) {
1627 		u8 cdie_range;
1628 
1629 		cdie_mask = plat_info->cdie_mask;
1630 		cdie_range = fls(cdie_mask) - ffs(cdie_mask) + 1;
1631 		io_die_cnt = num_resources - cdie_range;
1632 	} else {
1633 		/*
1634 		 * This is a synthetic mask, careful when assuming that
1635 		 * they are compute dies only.
1636 		 */
1637 		cdie_mask = (1 << num_resources) - 1;
1638 		io_die_cnt = 0;
1639 	}
1640 
1641 	for (i = 0; i < num_resources; ++i) {
1642 		struct resource *res;
1643 
1644 		res = tpmi_get_resource_at_index(auxdev, i);
1645 		if (!res) {
1646 			pd_info[i].sst_base = NULL;
1647 			continue;
1648 		}
1649 
1650 		pd_info[i].package_id = pkg;
1651 		pd_info[i].power_domain_id = i;
1652 		pd_info[i].auxdev = auxdev;
1653 		pd_info[i].write_blocked = write_blocked;
1654 		pd_info[i].sst_base = devm_ioremap_resource(dev, res);
1655 		if (IS_ERR(pd_info[i].sst_base)) {
1656 			ret = PTR_ERR(pd_info[i].sst_base);
1657 			goto unlock_free;
1658 		}
1659 
1660 		if (sst_main(auxdev, &pd_info[i])) {
1661 			/*
1662 			 * This entry is not valid, hardware can partially
1663 			 * populate dies. In this case MMIO will have 0xFFs.
1664 			 * Also possible some pre-production hardware has
1665 			 * invalid data. But don't fail and continue to use
1666 			 * other dies with valid data.
1667 			 */
1668 			devm_iounmap(dev, pd_info[i].sst_base);
1669 			pd_info[i].sst_base = NULL;
1670 			continue;
1671 		}
1672 
1673 		++inst;
1674 	}
1675 
1676 	if (!inst) {
1677 		ret = -ENODEV;
1678 		goto unlock_free;
1679 	}
1680 
1681 	tpmi_sst->package_id = pkg;
1682 
1683 	tpmi_sst->power_domain_info[partition] = pd_info;
1684 	tpmi_sst->number_of_power_domains[partition] = num_resources;
1685 	tpmi_sst->cdie_mask[partition] = cdie_mask;
1686 	tpmi_sst->io_dies[partition] = io_die_cnt;
1687 	tpmi_sst->partition_mask |= BIT(partition);
1688 	tpmi_sst->partition_mask_current |= BIT(partition);
1689 
1690 	auxiliary_set_drvdata(auxdev, tpmi_sst);
1691 
1692 	if (isst_common.max_index < pkg)
1693 		isst_common.max_index = pkg;
1694 	isst_common.sst_inst[pkg] = tpmi_sst;
1695 
1696 unlock_free:
1697 	if (ret && first_enum)
1698 		kfree(tpmi_sst);
1699 unlock_exit:
1700 	mutex_unlock(&isst_tpmi_dev_lock);
1701 
1702 	return ret;
1703 }
1704 EXPORT_SYMBOL_NS_GPL(tpmi_sst_dev_add, "INTEL_TPMI_SST");
1705 
1706 void tpmi_sst_dev_remove(struct auxiliary_device *auxdev)
1707 {
1708 	struct tpmi_sst_struct *tpmi_sst = auxiliary_get_drvdata(auxdev);
1709 	struct oobmsm_plat_info *plat_info;
1710 
1711 	plat_info = tpmi_get_platform_data(auxdev);
1712 	if (!plat_info)
1713 		return;
1714 
1715 	mutex_lock(&isst_tpmi_dev_lock);
1716 	tpmi_sst->power_domain_info[plat_info->partition] = NULL;
1717 	tpmi_sst->partition_mask_current &= ~BIT(plat_info->partition);
1718 	/* Free the package instance when the all partitions are removed */
1719 	if (!tpmi_sst->partition_mask_current) {
1720 		isst_common.sst_inst[tpmi_sst->package_id] = NULL;
1721 		kfree(tpmi_sst);
1722 	}
1723 	mutex_unlock(&isst_tpmi_dev_lock);
1724 }
1725 EXPORT_SYMBOL_NS_GPL(tpmi_sst_dev_remove, "INTEL_TPMI_SST");
1726 
1727 #define SST_PP_CAP_CP_ENABLE	BIT(0)
1728 #define SST_PP_CAP_PP_ENABLE	BIT(1)
1729 
1730 void tpmi_sst_dev_suspend(struct auxiliary_device *auxdev)
1731 {
1732 	struct tpmi_sst_struct *tpmi_sst = auxiliary_get_drvdata(auxdev);
1733 	struct tpmi_per_power_domain_info *power_domain_info, *pd_info;
1734 	struct oobmsm_plat_info *plat_info;
1735 	void __iomem *cp_base;
1736 	int num_resources, i;
1737 
1738 	plat_info = tpmi_get_platform_data(auxdev);
1739 	if (!plat_info)
1740 		return;
1741 
1742 	power_domain_info = tpmi_sst->power_domain_info[plat_info->partition];
1743 	num_resources = tpmi_sst->number_of_power_domains[plat_info->partition];
1744 
1745 	for (i = 0; i < num_resources; i++) {
1746 		pd_info = &power_domain_info[i];
1747 		if (!pd_info || !pd_info->sst_base)
1748 			continue;
1749 
1750 		if (!(pd_info->sst_header.cap_mask & SST_PP_CAP_CP_ENABLE))
1751 			goto process_pp_suspend;
1752 
1753 		cp_base = pd_info->sst_base + pd_info->sst_header.cp_offset;
1754 		pd_info->saved_sst_cp_control = readq(cp_base + SST_CP_CONTROL_OFFSET);
1755 		memcpy_fromio(pd_info->saved_clos_configs, cp_base + SST_CLOS_CONFIG_0_OFFSET,
1756 			      sizeof(pd_info->saved_clos_configs));
1757 		memcpy_fromio(pd_info->saved_clos_assocs, cp_base + SST_CLOS_ASSOC_0_OFFSET,
1758 			      sizeof(pd_info->saved_clos_assocs));
1759 
1760 process_pp_suspend:
1761 		if (!(pd_info->sst_header.cap_mask & SST_PP_CAP_PP_ENABLE))
1762 			continue;
1763 
1764 		pd_info->saved_pp_control = readq(pd_info->sst_base +
1765 						  pd_info->sst_header.pp_offset +
1766 						  SST_PP_CONTROL_OFFSET);
1767 	}
1768 }
1769 EXPORT_SYMBOL_NS_GPL(tpmi_sst_dev_suspend, "INTEL_TPMI_SST");
1770 
1771 void tpmi_sst_dev_resume(struct auxiliary_device *auxdev)
1772 {
1773 	struct tpmi_sst_struct *tpmi_sst = auxiliary_get_drvdata(auxdev);
1774 	struct tpmi_per_power_domain_info *power_domain_info, *pd_info;
1775 	struct oobmsm_plat_info *plat_info;
1776 	void __iomem *cp_base;
1777 	int num_resources, i;
1778 
1779 	plat_info = tpmi_get_platform_data(auxdev);
1780 	if (!plat_info)
1781 		return;
1782 
1783 	power_domain_info = tpmi_sst->power_domain_info[plat_info->partition];
1784 	num_resources = tpmi_sst->number_of_power_domains[plat_info->partition];
1785 
1786 	for (i = 0; i < num_resources; i++) {
1787 		pd_info = &power_domain_info[i];
1788 		if (!pd_info || !pd_info->sst_base)
1789 			continue;
1790 
1791 		if (!(pd_info->sst_header.cap_mask & SST_PP_CAP_CP_ENABLE))
1792 			goto process_pp_resume;
1793 
1794 		cp_base = pd_info->sst_base + pd_info->sst_header.cp_offset;
1795 		writeq(pd_info->saved_sst_cp_control, cp_base + SST_CP_CONTROL_OFFSET);
1796 		memcpy_toio(cp_base + SST_CLOS_CONFIG_0_OFFSET, pd_info->saved_clos_configs,
1797 			    sizeof(pd_info->saved_clos_configs));
1798 		memcpy_toio(cp_base + SST_CLOS_ASSOC_0_OFFSET, pd_info->saved_clos_assocs,
1799 			    sizeof(pd_info->saved_clos_assocs));
1800 
1801 process_pp_resume:
1802 		if (!(pd_info->sst_header.cap_mask & SST_PP_CAP_PP_ENABLE))
1803 			continue;
1804 
1805 		writeq(pd_info->saved_pp_control, power_domain_info->sst_base +
1806 		       pd_info->sst_header.pp_offset + SST_PP_CONTROL_OFFSET);
1807 	}
1808 }
1809 EXPORT_SYMBOL_NS_GPL(tpmi_sst_dev_resume, "INTEL_TPMI_SST");
1810 
1811 #define ISST_TPMI_API_VERSION	0x03
1812 
1813 int tpmi_sst_init(void)
1814 {
1815 	struct isst_if_cmd_cb cb;
1816 	int ret = 0;
1817 
1818 	mutex_lock(&isst_tpmi_dev_lock);
1819 
1820 	if (isst_core_usage_count) {
1821 		++isst_core_usage_count;
1822 		goto init_done;
1823 	}
1824 
1825 	isst_common.sst_inst = kzalloc_objs(*isst_common.sst_inst,
1826 					    topology_max_packages());
1827 	if (!isst_common.sst_inst) {
1828 		ret = -ENOMEM;
1829 		goto init_done;
1830 	}
1831 
1832 	memset(&cb, 0, sizeof(cb));
1833 	cb.cmd_size = sizeof(struct isst_if_io_reg);
1834 	cb.offset = offsetof(struct isst_if_io_regs, io_reg);
1835 	cb.cmd_callback = NULL;
1836 	cb.api_version = ISST_TPMI_API_VERSION;
1837 	cb.def_ioctl = isst_if_def_ioctl;
1838 	cb.owner = THIS_MODULE;
1839 	ret = isst_if_cdev_register(ISST_IF_DEV_TPMI, &cb);
1840 	if (ret)
1841 		kfree(isst_common.sst_inst);
1842 	else
1843 		++isst_core_usage_count;
1844 init_done:
1845 	mutex_unlock(&isst_tpmi_dev_lock);
1846 	return ret;
1847 }
1848 EXPORT_SYMBOL_NS_GPL(tpmi_sst_init, "INTEL_TPMI_SST");
1849 
1850 void tpmi_sst_exit(void)
1851 {
1852 	mutex_lock(&isst_tpmi_dev_lock);
1853 	if (isst_core_usage_count)
1854 		--isst_core_usage_count;
1855 
1856 	if (!isst_core_usage_count) {
1857 		isst_if_cdev_unregister(ISST_IF_DEV_TPMI);
1858 		kfree(isst_common.sst_inst);
1859 	}
1860 	mutex_unlock(&isst_tpmi_dev_lock);
1861 }
1862 EXPORT_SYMBOL_NS_GPL(tpmi_sst_exit, "INTEL_TPMI_SST");
1863 
1864 MODULE_IMPORT_NS("INTEL_TPMI");
1865 MODULE_IMPORT_NS("INTEL_TPMI_POWER_DOMAIN");
1866 
1867 MODULE_DESCRIPTION("ISST TPMI interface module");
1868 MODULE_LICENSE("GPL");
1869