1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * PowerPC64 LPAR Configuration Information Driver
5 * Dave Engebretsen engebret@us.ibm.com
6 * Copyright (c) 2003 Dave Engebretsen
7 * Will Schmidt willschm@us.ibm.com
8 * SPLPAR updates, Copyright (c) 2003 Will Schmidt IBM Corporation.
9 * seq_file updates, Copyright (c) 2004 Will Schmidt IBM Corporation.
10 * Nathan Lynch nathanl@austin.ibm.com
11 * Added lparcfg_write, Copyright (C) 2004 Nathan Lynch IBM Corporation.
13 * This driver creates a proc file at /proc/ppc64/lparcfg which contains
14 * keyword - value pairs that specify the configuration of the partition.
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/errno.h>
20 #include <linux/proc_fs.h>
21 #include <linux/init.h>
22 #include <linux/seq_file.h>
23 #include <linux/slab.h>
24 #include <linux/uaccess.h>
25 #include <linux/hugetlb.h>
26 #include <asm/lppaca.h>
27 #include <asm/hvcall.h>
28 #include <asm/firmware.h>
32 #include <asm/vdso_datapage.h>
35 #include <asm/machdep.h>
36 #include <asm/drmem.h>
41 * This isn't a module but we expose that to userspace
42 * via /proc so leave the definitions here
44 #define MODULE_VERS "1.9"
45 #define MODULE_NAME "lparcfg"
47 /* #define LPARCFG_DEBUG */
50 * Track sum of all purrs across all processors. This is used to further
51 * calculate usage values by different applications
53 static void cpu_get_purr(void *arg)
55 atomic64_t *sum = arg;
57 atomic64_add(mfspr(SPRN_PURR), sum);
60 static unsigned long get_purr(void)
62 atomic64_t purr = ATOMIC64_INIT(0);
64 on_each_cpu(cpu_get_purr, &purr, 1);
66 return atomic64_read(&purr);
70 * Methods used to fetch LPAR data when running on a pSeries platform.
73 struct hvcall_ppp_data {
75 u64 unallocated_entitlement;
80 u8 unallocated_weight;
81 u16 active_procs_in_pool;
82 u16 active_system_procs;
83 u16 phys_platform_procs;
84 u32 max_proc_cap_avail;
85 u32 entitled_proc_cap_avail;
89 * H_GET_PPP hcall returns info in 4 parms.
90 * entitled_capacity,unallocated_capacity,
91 * aggregation, resource_capability).
93 * R4 = Entitled Processor Capacity Percentage.
94 * R5 = Unallocated Processor Capacity Percentage.
95 * R6 (AABBCCDDEEFFGGHH).
100 * R7 (IIJJKKLLMMNNOOPP).
102 * XX - bit 0-6 reserved (0). bit 7 is Capped indicator.
103 * XX - variable processor Capacity Weight
104 * XX - Unallocated Variable Processor Capacity Weight.
105 * XXXX - Active processors in Physical Processor Pool.
106 * XXXX - Processors active on platform.
107 * R8 (QQQQRRRRRRSSSSSS). if ibm,partition-performance-parameters-level >= 1
108 * XXXX - Physical platform procs allocated to virtualization.
109 * XXXXXX - Max procs capacity % available to the partitions pool.
110 * XXXXXX - Entitled procs capacity % available to the
113 static unsigned int h_get_ppp(struct hvcall_ppp_data *ppp_data)
116 unsigned long retbuf[PLPAR_HCALL9_BUFSIZE];
118 rc = plpar_hcall9(H_GET_PPP, retbuf);
120 ppp_data->entitlement = retbuf[0];
121 ppp_data->unallocated_entitlement = retbuf[1];
123 ppp_data->group_num = (retbuf[2] >> 2 * 8) & 0xffff;
124 ppp_data->pool_num = retbuf[2] & 0xffff;
126 ppp_data->capped = (retbuf[3] >> 6 * 8) & 0x01;
127 ppp_data->weight = (retbuf[3] >> 5 * 8) & 0xff;
128 ppp_data->unallocated_weight = (retbuf[3] >> 4 * 8) & 0xff;
129 ppp_data->active_procs_in_pool = (retbuf[3] >> 2 * 8) & 0xffff;
130 ppp_data->active_system_procs = retbuf[3] & 0xffff;
132 ppp_data->phys_platform_procs = retbuf[4] >> 6 * 8;
133 ppp_data->max_proc_cap_avail = (retbuf[4] >> 3 * 8) & 0xffffff;
134 ppp_data->entitled_proc_cap_avail = retbuf[4] & 0xffffff;
139 static void show_gpci_data(struct seq_file *m)
141 struct hv_gpci_request_buffer *buf;
142 unsigned int affinity_score;
145 buf = kmalloc(sizeof(*buf), GFP_KERNEL);
150 * Show the local LPAR's affinity score.
152 * 0xB1 selects the Affinity_Domain_Info_By_Partition subcall.
153 * The score is at byte 0xB in the output buffer.
155 memset(&buf->params, 0, sizeof(buf->params));
156 buf->params.counter_request = cpu_to_be32(0xB1);
157 buf->params.starting_index = cpu_to_be32(-1); /* local LPAR */
158 buf->params.counter_info_version_in = 0x5; /* v5+ for score */
159 ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO, virt_to_phys(buf),
161 if (ret != H_SUCCESS) {
162 pr_debug("hcall failed: H_GET_PERF_COUNTER_INFO: %ld, %x\n",
163 ret, be32_to_cpu(buf->params.detail_rc));
166 affinity_score = buf->bytes[0xB];
167 seq_printf(m, "partition_affinity_score=%u\n", affinity_score);
172 static unsigned h_pic(unsigned long *pool_idle_time,
173 unsigned long *num_procs)
176 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
178 rc = plpar_hcall(H_PIC, retbuf);
180 *pool_idle_time = retbuf[0];
181 *num_procs = retbuf[1];
188 * Parse out the data returned from h_get_ppp and h_pic
190 static void parse_ppp_data(struct seq_file *m)
192 struct hvcall_ppp_data ppp_data;
193 struct device_node *root;
194 const __be32 *perf_level;
197 rc = h_get_ppp(&ppp_data);
201 seq_printf(m, "partition_entitled_capacity=%lld\n",
202 ppp_data.entitlement);
203 seq_printf(m, "group=%d\n", ppp_data.group_num);
204 seq_printf(m, "system_active_processors=%d\n",
205 ppp_data.active_system_procs);
207 /* pool related entries are appropriate for shared configs */
208 if (lppaca_shared_proc()) {
209 unsigned long pool_idle_time, pool_procs;
211 seq_printf(m, "pool=%d\n", ppp_data.pool_num);
213 /* report pool_capacity in percentage */
214 seq_printf(m, "pool_capacity=%d\n",
215 ppp_data.active_procs_in_pool * 100);
217 h_pic(&pool_idle_time, &pool_procs);
218 seq_printf(m, "pool_idle_time=%ld\n", pool_idle_time);
219 seq_printf(m, "pool_num_procs=%ld\n", pool_procs);
222 seq_printf(m, "unallocated_capacity_weight=%d\n",
223 ppp_data.unallocated_weight);
224 seq_printf(m, "capacity_weight=%d\n", ppp_data.weight);
225 seq_printf(m, "capped=%d\n", ppp_data.capped);
226 seq_printf(m, "unallocated_capacity=%lld\n",
227 ppp_data.unallocated_entitlement);
229 /* The last bits of information returned from h_get_ppp are only
230 * valid if the ibm,partition-performance-parameters-level
233 root = of_find_node_by_path("/");
235 perf_level = of_get_property(root,
236 "ibm,partition-performance-parameters-level",
238 if (perf_level && (be32_to_cpup(perf_level) >= 1)) {
240 "physical_procs_allocated_to_virtualization=%d\n",
241 ppp_data.phys_platform_procs);
242 seq_printf(m, "max_proc_capacity_available=%d\n",
243 ppp_data.max_proc_cap_avail);
244 seq_printf(m, "entitled_proc_capacity_available=%d\n",
245 ppp_data.entitled_proc_cap_avail);
254 * Parse out data returned from h_get_mpp
256 static void parse_mpp_data(struct seq_file *m)
258 struct hvcall_mpp_data mpp_data;
261 rc = h_get_mpp(&mpp_data);
265 seq_printf(m, "entitled_memory=%ld\n", mpp_data.entitled_mem);
267 if (mpp_data.mapped_mem != -1)
268 seq_printf(m, "mapped_entitled_memory=%ld\n",
269 mpp_data.mapped_mem);
271 seq_printf(m, "entitled_memory_group_number=%d\n", mpp_data.group_num);
272 seq_printf(m, "entitled_memory_pool_number=%d\n", mpp_data.pool_num);
274 seq_printf(m, "entitled_memory_weight=%d\n", mpp_data.mem_weight);
275 seq_printf(m, "unallocated_entitled_memory_weight=%d\n",
276 mpp_data.unallocated_mem_weight);
277 seq_printf(m, "unallocated_io_mapping_entitlement=%ld\n",
278 mpp_data.unallocated_entitlement);
280 if (mpp_data.pool_size != -1)
281 seq_printf(m, "entitled_memory_pool_size=%ld bytes\n",
284 seq_printf(m, "entitled_memory_loan_request=%ld\n",
285 mpp_data.loan_request);
287 seq_printf(m, "backing_memory=%ld bytes\n", mpp_data.backing_mem);
292 * Parse out data returned from h_get_mpp_x
294 static void parse_mpp_x_data(struct seq_file *m)
296 struct hvcall_mpp_x_data mpp_x_data;
298 if (!firmware_has_feature(FW_FEATURE_XCMO))
300 if (h_get_mpp_x(&mpp_x_data))
303 seq_printf(m, "coalesced_bytes=%ld\n", mpp_x_data.coalesced_bytes);
305 if (mpp_x_data.pool_coalesced_bytes)
306 seq_printf(m, "pool_coalesced_bytes=%ld\n",
307 mpp_x_data.pool_coalesced_bytes);
308 if (mpp_x_data.pool_purr_cycles)
309 seq_printf(m, "coalesce_pool_purr=%ld\n", mpp_x_data.pool_purr_cycles);
310 if (mpp_x_data.pool_spurr_cycles)
311 seq_printf(m, "coalesce_pool_spurr=%ld\n", mpp_x_data.pool_spurr_cycles);
314 #define SPLPAR_CHARACTERISTICS_TOKEN 20
315 #define SPLPAR_MAXLENGTH 1026*(sizeof(char))
318 * parse_system_parameter_string()
319 * Retrieve the potential_processors, max_entitled_capacity and friends
320 * through the get-system-parameter rtas call. Replace keyword strings as
323 static void parse_system_parameter_string(struct seq_file *m)
325 const s32 token = rtas_token("ibm,get-system-parameter");
328 unsigned char *local_buffer = kmalloc(SPLPAR_MAXLENGTH, GFP_KERNEL);
330 printk(KERN_ERR "%s %s kmalloc failure at line %d\n",
331 __FILE__, __func__, __LINE__);
336 spin_lock(&rtas_data_buf_lock);
337 memset(rtas_data_buf, 0, SPLPAR_MAXLENGTH);
338 call_status = rtas_call(token, 3, 1, NULL, SPLPAR_CHARACTERISTICS_TOKEN,
339 __pa(rtas_data_buf), RTAS_DATA_BUF_SIZE);
340 memcpy(local_buffer, rtas_data_buf, SPLPAR_MAXLENGTH);
341 local_buffer[SPLPAR_MAXLENGTH - 1] = '\0';
342 spin_unlock(&rtas_data_buf_lock);
343 } while (rtas_busy_delay(call_status));
345 if (call_status != 0) {
347 "%s %s Error calling get-system-parameter (0x%x)\n",
348 __FILE__, __func__, call_status);
352 char *workbuffer = kzalloc(SPLPAR_MAXLENGTH, GFP_KERNEL);
354 printk(KERN_ERR "%s %s kmalloc failure at line %d\n",
355 __FILE__, __func__, __LINE__);
360 printk(KERN_INFO "success calling get-system-parameter\n");
362 splpar_strlen = local_buffer[0] * 256 + local_buffer[1];
363 local_buffer += 2; /* step over strlen value */
367 while ((*local_buffer) && (idx < splpar_strlen)) {
368 workbuffer[w_idx++] = local_buffer[idx++];
369 if ((local_buffer[idx] == ',')
370 || (local_buffer[idx] == '\0')) {
371 workbuffer[w_idx] = '\0';
373 /* avoid the empty string */
374 seq_printf(m, "%s\n", workbuffer);
376 memset(workbuffer, 0, SPLPAR_MAXLENGTH);
377 idx++; /* skip the comma */
379 } else if (local_buffer[idx] == '=') {
380 /* code here to replace workbuffer contents
381 with different keyword strings */
382 if (0 == strcmp(workbuffer, "MaxEntCap")) {
384 "partition_max_entitled_capacity");
385 w_idx = strlen(workbuffer);
387 if (0 == strcmp(workbuffer, "MaxPlatProcs")) {
389 "system_potential_processors");
390 w_idx = strlen(workbuffer);
395 local_buffer -= 2; /* back up over strlen value */
400 /* Return the number of processors in the system.
401 * This function reads through the device tree and counts
402 * the virtual processors, this does not include threads.
404 static int lparcfg_count_active_processors(void)
406 struct device_node *cpus_dn;
409 for_each_node_by_type(cpus_dn, "cpu") {
411 printk(KERN_ERR "cpus_dn %p\n", cpus_dn);
418 static void pseries_cmo_data(struct seq_file *m)
421 unsigned long cmo_faults = 0;
422 unsigned long cmo_fault_time = 0;
424 seq_printf(m, "cmo_enabled=%d\n", firmware_has_feature(FW_FEATURE_CMO));
426 if (!firmware_has_feature(FW_FEATURE_CMO))
429 for_each_possible_cpu(cpu) {
430 cmo_faults += be64_to_cpu(lppaca_of(cpu).cmo_faults);
431 cmo_fault_time += be64_to_cpu(lppaca_of(cpu).cmo_fault_time);
434 seq_printf(m, "cmo_faults=%lu\n", cmo_faults);
435 seq_printf(m, "cmo_fault_time_usec=%lu\n",
436 cmo_fault_time / tb_ticks_per_usec);
437 seq_printf(m, "cmo_primary_psp=%d\n", cmo_get_primary_psp());
438 seq_printf(m, "cmo_secondary_psp=%d\n", cmo_get_secondary_psp());
439 seq_printf(m, "cmo_page_size=%lu\n", cmo_get_page_size());
442 static void splpar_dispatch_data(struct seq_file *m)
445 unsigned long dispatches = 0;
446 unsigned long dispatch_dispersions = 0;
448 for_each_possible_cpu(cpu) {
449 dispatches += be32_to_cpu(lppaca_of(cpu).yield_count);
450 dispatch_dispersions +=
451 be32_to_cpu(lppaca_of(cpu).dispersion_count);
454 seq_printf(m, "dispatches=%lu\n", dispatches);
455 seq_printf(m, "dispatch_dispersions=%lu\n", dispatch_dispersions);
458 static void parse_em_data(struct seq_file *m)
460 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
462 if (firmware_has_feature(FW_FEATURE_LPAR) &&
463 plpar_hcall(H_GET_EM_PARMS, retbuf) == H_SUCCESS)
464 seq_printf(m, "power_mode_data=%016lx\n", retbuf[0]);
467 static void maxmem_data(struct seq_file *m)
469 unsigned long maxmem = 0;
471 maxmem += (unsigned long)drmem_info->n_lmbs * drmem_info->lmb_size;
472 maxmem += hugetlb_total_pages() * PAGE_SIZE;
474 seq_printf(m, "MaxMem=%lu\n", maxmem);
477 static int pseries_lparcfg_data(struct seq_file *m, void *v)
479 int partition_potential_processors;
480 int partition_active_processors;
481 struct device_node *rtas_node;
482 const __be32 *lrdrp = NULL;
484 rtas_node = of_find_node_by_path("/rtas");
486 lrdrp = of_get_property(rtas_node, "ibm,lrdr-capacity", NULL);
489 partition_potential_processors = vdso_data->processorCount;
491 partition_potential_processors = be32_to_cpup(lrdrp + 4);
493 of_node_put(rtas_node);
495 partition_active_processors = lparcfg_count_active_processors();
497 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
498 /* this call handles the ibm,get-system-parameter contents */
499 parse_system_parameter_string(m);
504 splpar_dispatch_data(m);
506 seq_printf(m, "purr=%ld\n", get_purr());
507 seq_printf(m, "tbr=%ld\n", mftb());
508 } else { /* non SPLPAR case */
510 seq_printf(m, "system_active_processors=%d\n",
511 partition_potential_processors);
513 seq_printf(m, "system_potential_processors=%d\n",
514 partition_potential_processors);
516 seq_printf(m, "partition_max_entitled_capacity=%d\n",
517 partition_potential_processors * 100);
519 seq_printf(m, "partition_entitled_capacity=%d\n",
520 partition_active_processors * 100);
525 seq_printf(m, "partition_active_processors=%d\n",
526 partition_active_processors);
528 seq_printf(m, "partition_potential_processors=%d\n",
529 partition_potential_processors);
531 seq_printf(m, "shared_processor_mode=%d\n",
532 lppaca_shared_proc());
534 #ifdef CONFIG_PPC_BOOK3S_64
535 seq_printf(m, "slb_size=%d\n", mmu_slb_size);
543 static ssize_t update_ppp(u64 *entitlement, u8 *weight)
545 struct hvcall_ppp_data ppp_data;
550 /* Get our current parameters */
551 retval = h_get_ppp(&ppp_data);
556 new_weight = ppp_data.weight;
557 new_entitled = *entitlement;
559 new_weight = *weight;
560 new_entitled = ppp_data.entitlement;
564 pr_debug("%s: current_entitled = %llu, current_weight = %u\n",
565 __func__, ppp_data.entitlement, ppp_data.weight);
567 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
568 __func__, new_entitled, new_weight);
570 retval = plpar_hcall_norets(H_SET_PPP, new_entitled, new_weight);
577 * Update the memory entitlement and weight for the partition. Caller must
578 * specify either a new entitlement or weight, not both, to be updated
579 * since the h_set_mpp call takes both entitlement and weight as parameters.
581 static ssize_t update_mpp(u64 *entitlement, u8 *weight)
583 struct hvcall_mpp_data mpp_data;
589 /* Check with vio to ensure the new memory entitlement
592 rc = vio_cmo_entitlement_update(*entitlement);
597 rc = h_get_mpp(&mpp_data);
602 new_weight = mpp_data.mem_weight;
603 new_entitled = *entitlement;
605 new_weight = *weight;
606 new_entitled = mpp_data.entitled_mem;
610 pr_debug("%s: current_entitled = %lu, current_weight = %u\n",
611 __func__, mpp_data.entitled_mem, mpp_data.mem_weight);
613 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
614 __func__, new_entitled, new_weight);
616 rc = plpar_hcall_norets(H_SET_MPP, new_entitled, new_weight);
621 * Interface for changing system parameters (variable capacity weight
622 * and entitled capacity). Format of input is "param_name=value";
623 * anything after value is ignored. Valid parameters at this time are
624 * "partition_entitled_capacity" and "capacity_weight". We use
625 * H_SET_PPP to alter parameters.
627 * This function should be invoked only on systems with
630 static ssize_t lparcfg_write(struct file *file, const char __user * buf,
631 size_t count, loff_t * off)
635 u64 new_entitled, *new_entitled_ptr = &new_entitled;
636 u8 new_weight, *new_weight_ptr = &new_weight;
639 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
642 if (count > sizeof(kbuf))
645 if (copy_from_user(kbuf, buf, count))
648 kbuf[count - 1] = '\0';
649 tmp = strchr(kbuf, '=');
655 if (!strcmp(kbuf, "partition_entitled_capacity")) {
657 *new_entitled_ptr = (u64) simple_strtoul(tmp, &endp, 10);
661 retval = update_ppp(new_entitled_ptr, NULL);
662 } else if (!strcmp(kbuf, "capacity_weight")) {
664 *new_weight_ptr = (u8) simple_strtoul(tmp, &endp, 10);
668 retval = update_ppp(NULL, new_weight_ptr);
669 } else if (!strcmp(kbuf, "entitled_memory")) {
671 *new_entitled_ptr = (u64) simple_strtoul(tmp, &endp, 10);
675 retval = update_mpp(new_entitled_ptr, NULL);
676 } else if (!strcmp(kbuf, "entitled_memory_weight")) {
678 *new_weight_ptr = (u8) simple_strtoul(tmp, &endp, 10);
682 retval = update_mpp(NULL, new_weight_ptr);
686 if (retval == H_SUCCESS || retval == H_CONSTRAINED) {
688 } else if (retval == H_BUSY) {
690 } else if (retval == H_HARDWARE) {
692 } else if (retval == H_PARAMETER) {
699 static int lparcfg_data(struct seq_file *m, void *v)
701 struct device_node *rootdn;
702 const char *model = "";
703 const char *system_id = "";
705 const __be32 *lp_index_ptr;
706 unsigned int lp_index = 0;
708 seq_printf(m, "%s %s\n", MODULE_NAME, MODULE_VERS);
710 rootdn = of_find_node_by_path("/");
712 tmp = of_get_property(rootdn, "model", NULL);
715 tmp = of_get_property(rootdn, "system-id", NULL);
718 lp_index_ptr = of_get_property(rootdn, "ibm,partition-no",
721 lp_index = be32_to_cpup(lp_index_ptr);
724 seq_printf(m, "serial_number=%s\n", system_id);
725 seq_printf(m, "system_type=%s\n", model);
726 seq_printf(m, "partition_id=%d\n", (int)lp_index);
728 return pseries_lparcfg_data(m, v);
731 static int lparcfg_open(struct inode *inode, struct file *file)
733 return single_open(file, lparcfg_data, NULL);
736 static const struct proc_ops lparcfg_proc_ops = {
737 .proc_read = seq_read,
738 .proc_write = lparcfg_write,
739 .proc_open = lparcfg_open,
740 .proc_release = single_release,
741 .proc_lseek = seq_lseek,
744 static int __init lparcfg_init(void)
748 /* Allow writing if we have FW_FEATURE_SPLPAR */
749 if (firmware_has_feature(FW_FEATURE_SPLPAR))
752 if (!proc_create("powerpc/lparcfg", mode, NULL, &lparcfg_proc_ops)) {
753 printk(KERN_ERR "Failed to create powerpc/lparcfg\n");
758 machine_device_initcall(pseries, lparcfg_init);