2 * PowerPC64 LPAR Configuration Information Driver
4 * Dave Engebretsen engebret@us.ibm.com
5 * Copyright (c) 2003 Dave Engebretsen
6 * Will Schmidt willschm@us.ibm.com
7 * SPLPAR updates, Copyright (c) 2003 Will Schmidt IBM Corporation.
8 * seq_file updates, Copyright (c) 2004 Will Schmidt IBM Corporation.
9 * Nathan Lynch nathanl@austin.ibm.com
10 * Added lparcfg_write, Copyright (C) 2004 Nathan Lynch IBM Corporation.
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
17 * This driver creates a proc file at /proc/ppc64/lparcfg which contains
18 * keyword - value pairs that specify the configuration of the partition.
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/errno.h>
24 #include <linux/proc_fs.h>
25 #include <linux/init.h>
26 #include <linux/seq_file.h>
27 #include <linux/slab.h>
28 #include <asm/uaccess.h>
29 #include <asm/lppaca.h>
30 #include <asm/hvcall.h>
31 #include <asm/firmware.h>
33 #include <asm/system.h>
36 #include <asm/vdso_datapage.h>
40 #define MODULE_VERS "1.9"
41 #define MODULE_NAME "lparcfg"
43 /* #define LPARCFG_DEBUG */
45 static struct proc_dir_entry *proc_ppc64_lparcfg;
48 * Track sum of all purrs across all processors. This is used to further
49 * calculate usage values by different applications
51 static unsigned long get_purr(void)
53 unsigned long sum_purr = 0;
56 for_each_possible_cpu(cpu) {
59 cu = &per_cpu(cpu_usage_array, cpu);
60 sum_purr += cu->current_tb;
66 * Methods used to fetch LPAR data when running on a pSeries platform.
69 struct hvcall_ppp_data {
71 u64 unallocated_entitlement;
76 u8 unallocated_weight;
77 u16 active_procs_in_pool;
78 u16 active_system_procs;
79 u16 phys_platform_procs;
80 u32 max_proc_cap_avail;
81 u32 entitled_proc_cap_avail;
85 * H_GET_PPP hcall returns info in 4 parms.
86 * entitled_capacity,unallocated_capacity,
87 * aggregation, resource_capability).
89 * R4 = Entitled Processor Capacity Percentage.
90 * R5 = Unallocated Processor Capacity Percentage.
91 * R6 (AABBCCDDEEFFGGHH).
96 * R7 (IIJJKKLLMMNNOOPP).
98 * XX - bit 0-6 reserved (0). bit 7 is Capped indicator.
99 * XX - variable processor Capacity Weight
100 * XX - Unallocated Variable Processor Capacity Weight.
101 * XXXX - Active processors in Physical Processor Pool.
102 * XXXX - Processors active on platform.
103 * R8 (QQQQRRRRRRSSSSSS). if ibm,partition-performance-parameters-level >= 1
104 * XXXX - Physical platform procs allocated to virtualization.
105 * XXXXXX - Max procs capacity % available to the partitions pool.
106 * XXXXXX - Entitled procs capacity % available to the
109 static unsigned int h_get_ppp(struct hvcall_ppp_data *ppp_data)
112 unsigned long retbuf[PLPAR_HCALL9_BUFSIZE];
114 rc = plpar_hcall9(H_GET_PPP, retbuf);
116 ppp_data->entitlement = retbuf[0];
117 ppp_data->unallocated_entitlement = retbuf[1];
119 ppp_data->group_num = (retbuf[2] >> 2 * 8) & 0xffff;
120 ppp_data->pool_num = retbuf[2] & 0xffff;
122 ppp_data->capped = (retbuf[3] >> 6 * 8) & 0x01;
123 ppp_data->weight = (retbuf[3] >> 5 * 8) & 0xff;
124 ppp_data->unallocated_weight = (retbuf[3] >> 4 * 8) & 0xff;
125 ppp_data->active_procs_in_pool = (retbuf[3] >> 2 * 8) & 0xffff;
126 ppp_data->active_system_procs = retbuf[3] & 0xffff;
128 ppp_data->phys_platform_procs = retbuf[4] >> 6 * 8;
129 ppp_data->max_proc_cap_avail = (retbuf[4] >> 3 * 8) & 0xffffff;
130 ppp_data->entitled_proc_cap_avail = retbuf[4] & 0xffffff;
135 static unsigned h_pic(unsigned long *pool_idle_time,
136 unsigned long *num_procs)
139 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
141 rc = plpar_hcall(H_PIC, retbuf);
143 *pool_idle_time = retbuf[0];
144 *num_procs = retbuf[1];
151 * Parse out the data returned from h_get_ppp and h_pic
153 static void parse_ppp_data(struct seq_file *m)
155 struct hvcall_ppp_data ppp_data;
156 struct device_node *root;
157 const int *perf_level;
160 rc = h_get_ppp(&ppp_data);
164 seq_printf(m, "partition_entitled_capacity=%lld\n",
165 ppp_data.entitlement);
166 seq_printf(m, "group=%d\n", ppp_data.group_num);
167 seq_printf(m, "system_active_processors=%d\n",
168 ppp_data.active_system_procs);
170 /* pool related entries are appropriate for shared configs */
171 if (lppaca_of(0).shared_proc) {
172 unsigned long pool_idle_time, pool_procs;
174 seq_printf(m, "pool=%d\n", ppp_data.pool_num);
176 /* report pool_capacity in percentage */
177 seq_printf(m, "pool_capacity=%d\n",
178 ppp_data.active_procs_in_pool * 100);
180 h_pic(&pool_idle_time, &pool_procs);
181 seq_printf(m, "pool_idle_time=%ld\n", pool_idle_time);
182 seq_printf(m, "pool_num_procs=%ld\n", pool_procs);
185 seq_printf(m, "unallocated_capacity_weight=%d\n",
186 ppp_data.unallocated_weight);
187 seq_printf(m, "capacity_weight=%d\n", ppp_data.weight);
188 seq_printf(m, "capped=%d\n", ppp_data.capped);
189 seq_printf(m, "unallocated_capacity=%lld\n",
190 ppp_data.unallocated_entitlement);
192 /* The last bits of information returned from h_get_ppp are only
193 * valid if the ibm,partition-performance-parameters-level
196 root = of_find_node_by_path("/");
198 perf_level = of_get_property(root,
199 "ibm,partition-performance-parameters-level",
201 if (perf_level && (*perf_level >= 1)) {
203 "physical_procs_allocated_to_virtualization=%d\n",
204 ppp_data.phys_platform_procs);
205 seq_printf(m, "max_proc_capacity_available=%d\n",
206 ppp_data.max_proc_cap_avail);
207 seq_printf(m, "entitled_proc_capacity_available=%d\n",
208 ppp_data.entitled_proc_cap_avail);
217 * Parse out data returned from h_get_mpp
219 static void parse_mpp_data(struct seq_file *m)
221 struct hvcall_mpp_data mpp_data;
224 rc = h_get_mpp(&mpp_data);
228 seq_printf(m, "entitled_memory=%ld\n", mpp_data.entitled_mem);
230 if (mpp_data.mapped_mem != -1)
231 seq_printf(m, "mapped_entitled_memory=%ld\n",
232 mpp_data.mapped_mem);
234 seq_printf(m, "entitled_memory_group_number=%d\n", mpp_data.group_num);
235 seq_printf(m, "entitled_memory_pool_number=%d\n", mpp_data.pool_num);
237 seq_printf(m, "entitled_memory_weight=%d\n", mpp_data.mem_weight);
238 seq_printf(m, "unallocated_entitled_memory_weight=%d\n",
239 mpp_data.unallocated_mem_weight);
240 seq_printf(m, "unallocated_io_mapping_entitlement=%ld\n",
241 mpp_data.unallocated_entitlement);
243 if (mpp_data.pool_size != -1)
244 seq_printf(m, "entitled_memory_pool_size=%ld bytes\n",
247 seq_printf(m, "entitled_memory_loan_request=%ld\n",
248 mpp_data.loan_request);
250 seq_printf(m, "backing_memory=%ld bytes\n", mpp_data.backing_mem);
255 * Parse out data returned from h_get_mpp_x
257 static void parse_mpp_x_data(struct seq_file *m)
259 struct hvcall_mpp_x_data mpp_x_data;
261 if (!firmware_has_feature(FW_FEATURE_XCMO))
263 if (h_get_mpp_x(&mpp_x_data))
266 seq_printf(m, "coalesced_bytes=%ld\n", mpp_x_data.coalesced_bytes);
268 if (mpp_x_data.pool_coalesced_bytes)
269 seq_printf(m, "pool_coalesced_bytes=%ld\n",
270 mpp_x_data.pool_coalesced_bytes);
271 if (mpp_x_data.pool_purr_cycles)
272 seq_printf(m, "coalesce_pool_purr=%ld\n", mpp_x_data.pool_purr_cycles);
273 if (mpp_x_data.pool_spurr_cycles)
274 seq_printf(m, "coalesce_pool_spurr=%ld\n", mpp_x_data.pool_spurr_cycles);
277 #define SPLPAR_CHARACTERISTICS_TOKEN 20
278 #define SPLPAR_MAXLENGTH 1026*(sizeof(char))
281 * parse_system_parameter_string()
282 * Retrieve the potential_processors, max_entitled_capacity and friends
283 * through the get-system-parameter rtas call. Replace keyword strings as
286 static void parse_system_parameter_string(struct seq_file *m)
290 unsigned char *local_buffer = kmalloc(SPLPAR_MAXLENGTH, GFP_KERNEL);
292 printk(KERN_ERR "%s %s kmalloc failure at line %d\n",
293 __FILE__, __func__, __LINE__);
297 spin_lock(&rtas_data_buf_lock);
298 memset(rtas_data_buf, 0, SPLPAR_MAXLENGTH);
299 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
301 SPLPAR_CHARACTERISTICS_TOKEN,
304 memcpy(local_buffer, rtas_data_buf, SPLPAR_MAXLENGTH);
305 spin_unlock(&rtas_data_buf_lock);
307 if (call_status != 0) {
309 "%s %s Error calling get-system-parameter (0x%x)\n",
310 __FILE__, __func__, call_status);
314 char *workbuffer = kzalloc(SPLPAR_MAXLENGTH, GFP_KERNEL);
316 printk(KERN_ERR "%s %s kmalloc failure at line %d\n",
317 __FILE__, __func__, __LINE__);
322 printk(KERN_INFO "success calling get-system-parameter\n");
324 splpar_strlen = local_buffer[0] * 256 + local_buffer[1];
325 local_buffer += 2; /* step over strlen value */
329 while ((*local_buffer) && (idx < splpar_strlen)) {
330 workbuffer[w_idx++] = local_buffer[idx++];
331 if ((local_buffer[idx] == ',')
332 || (local_buffer[idx] == '\0')) {
333 workbuffer[w_idx] = '\0';
335 /* avoid the empty string */
336 seq_printf(m, "%s\n", workbuffer);
338 memset(workbuffer, 0, SPLPAR_MAXLENGTH);
339 idx++; /* skip the comma */
341 } else if (local_buffer[idx] == '=') {
342 /* code here to replace workbuffer contents
343 with different keyword strings */
344 if (0 == strcmp(workbuffer, "MaxEntCap")) {
346 "partition_max_entitled_capacity");
347 w_idx = strlen(workbuffer);
349 if (0 == strcmp(workbuffer, "MaxPlatProcs")) {
351 "system_potential_processors");
352 w_idx = strlen(workbuffer);
357 local_buffer -= 2; /* back up over strlen value */
362 /* Return the number of processors in the system.
363 * This function reads through the device tree and counts
364 * the virtual processors, this does not include threads.
366 static int lparcfg_count_active_processors(void)
368 struct device_node *cpus_dn = NULL;
371 while ((cpus_dn = of_find_node_by_type(cpus_dn, "cpu"))) {
373 printk(KERN_ERR "cpus_dn %p\n", cpus_dn);
380 static void pseries_cmo_data(struct seq_file *m)
383 unsigned long cmo_faults = 0;
384 unsigned long cmo_fault_time = 0;
386 seq_printf(m, "cmo_enabled=%d\n", firmware_has_feature(FW_FEATURE_CMO));
388 if (!firmware_has_feature(FW_FEATURE_CMO))
391 for_each_possible_cpu(cpu) {
392 cmo_faults += lppaca_of(cpu).cmo_faults;
393 cmo_fault_time += lppaca_of(cpu).cmo_fault_time;
396 seq_printf(m, "cmo_faults=%lu\n", cmo_faults);
397 seq_printf(m, "cmo_fault_time_usec=%lu\n",
398 cmo_fault_time / tb_ticks_per_usec);
399 seq_printf(m, "cmo_primary_psp=%d\n", cmo_get_primary_psp());
400 seq_printf(m, "cmo_secondary_psp=%d\n", cmo_get_secondary_psp());
401 seq_printf(m, "cmo_page_size=%lu\n", cmo_get_page_size());
404 static void splpar_dispatch_data(struct seq_file *m)
407 unsigned long dispatches = 0;
408 unsigned long dispatch_dispersions = 0;
410 for_each_possible_cpu(cpu) {
411 dispatches += lppaca_of(cpu).yield_count;
412 dispatch_dispersions += lppaca_of(cpu).dispersion_count;
415 seq_printf(m, "dispatches=%lu\n", dispatches);
416 seq_printf(m, "dispatch_dispersions=%lu\n", dispatch_dispersions);
419 static void parse_em_data(struct seq_file *m)
421 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
423 if (plpar_hcall(H_GET_EM_PARMS, retbuf) == H_SUCCESS)
424 seq_printf(m, "power_mode_data=%016lx\n", retbuf[0]);
427 static int pseries_lparcfg_data(struct seq_file *m, void *v)
429 int partition_potential_processors;
430 int partition_active_processors;
431 struct device_node *rtas_node;
432 const int *lrdrp = NULL;
434 rtas_node = of_find_node_by_path("/rtas");
436 lrdrp = of_get_property(rtas_node, "ibm,lrdr-capacity", NULL);
439 partition_potential_processors = vdso_data->processorCount;
441 partition_potential_processors = *(lrdrp + 4);
443 of_node_put(rtas_node);
445 partition_active_processors = lparcfg_count_active_processors();
447 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
448 /* this call handles the ibm,get-system-parameter contents */
449 parse_system_parameter_string(m);
454 splpar_dispatch_data(m);
456 seq_printf(m, "purr=%ld\n", get_purr());
457 } else { /* non SPLPAR case */
459 seq_printf(m, "system_active_processors=%d\n",
460 partition_potential_processors);
462 seq_printf(m, "system_potential_processors=%d\n",
463 partition_potential_processors);
465 seq_printf(m, "partition_max_entitled_capacity=%d\n",
466 partition_potential_processors * 100);
468 seq_printf(m, "partition_entitled_capacity=%d\n",
469 partition_active_processors * 100);
472 seq_printf(m, "partition_active_processors=%d\n",
473 partition_active_processors);
475 seq_printf(m, "partition_potential_processors=%d\n",
476 partition_potential_processors);
478 seq_printf(m, "shared_processor_mode=%d\n", lppaca_of(0).shared_proc);
480 seq_printf(m, "slb_size=%d\n", mmu_slb_size);
487 static ssize_t update_ppp(u64 *entitlement, u8 *weight)
489 struct hvcall_ppp_data ppp_data;
494 /* Get our current parameters */
495 retval = h_get_ppp(&ppp_data);
500 new_weight = ppp_data.weight;
501 new_entitled = *entitlement;
503 new_weight = *weight;
504 new_entitled = ppp_data.entitlement;
508 pr_debug("%s: current_entitled = %llu, current_weight = %u\n",
509 __func__, ppp_data.entitlement, ppp_data.weight);
511 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
512 __func__, new_entitled, new_weight);
514 retval = plpar_hcall_norets(H_SET_PPP, new_entitled, new_weight);
521 * Update the memory entitlement and weight for the partition. Caller must
522 * specify either a new entitlement or weight, not both, to be updated
523 * since the h_set_mpp call takes both entitlement and weight as parameters.
525 static ssize_t update_mpp(u64 *entitlement, u8 *weight)
527 struct hvcall_mpp_data mpp_data;
533 /* Check with vio to ensure the new memory entitlement
536 rc = vio_cmo_entitlement_update(*entitlement);
541 rc = h_get_mpp(&mpp_data);
546 new_weight = mpp_data.mem_weight;
547 new_entitled = *entitlement;
549 new_weight = *weight;
550 new_entitled = mpp_data.entitled_mem;
554 pr_debug("%s: current_entitled = %lu, current_weight = %u\n",
555 __func__, mpp_data.entitled_mem, mpp_data.mem_weight);
557 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
558 __func__, new_entitled, new_weight);
560 rc = plpar_hcall_norets(H_SET_MPP, new_entitled, new_weight);
565 * Interface for changing system parameters (variable capacity weight
566 * and entitled capacity). Format of input is "param_name=value";
567 * anything after value is ignored. Valid parameters at this time are
568 * "partition_entitled_capacity" and "capacity_weight". We use
569 * H_SET_PPP to alter parameters.
571 * This function should be invoked only on systems with
574 static ssize_t lparcfg_write(struct file *file, const char __user * buf,
575 size_t count, loff_t * off)
580 u64 new_entitled, *new_entitled_ptr = &new_entitled;
581 u8 new_weight, *new_weight_ptr = &new_weight;
584 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
590 if (copy_from_user(kbuf, buf, count))
593 kbuf[count - 1] = '\0';
594 tmp = strchr(kbuf, '=');
600 if (!strcmp(kbuf, "partition_entitled_capacity")) {
602 *new_entitled_ptr = (u64) simple_strtoul(tmp, &endp, 10);
606 retval = update_ppp(new_entitled_ptr, NULL);
607 } else if (!strcmp(kbuf, "capacity_weight")) {
609 *new_weight_ptr = (u8) simple_strtoul(tmp, &endp, 10);
613 retval = update_ppp(NULL, new_weight_ptr);
614 } else if (!strcmp(kbuf, "entitled_memory")) {
616 *new_entitled_ptr = (u64) simple_strtoul(tmp, &endp, 10);
620 retval = update_mpp(new_entitled_ptr, NULL);
621 } else if (!strcmp(kbuf, "entitled_memory_weight")) {
623 *new_weight_ptr = (u8) simple_strtoul(tmp, &endp, 10);
627 retval = update_mpp(NULL, new_weight_ptr);
631 if (retval == H_SUCCESS || retval == H_CONSTRAINED) {
633 } else if (retval == H_BUSY) {
635 } else if (retval == H_HARDWARE) {
637 } else if (retval == H_PARAMETER) {
644 static int lparcfg_data(struct seq_file *m, void *v)
646 struct device_node *rootdn;
647 const char *model = "";
648 const char *system_id = "";
650 const unsigned int *lp_index_ptr;
651 unsigned int lp_index = 0;
653 seq_printf(m, "%s %s\n", MODULE_NAME, MODULE_VERS);
655 rootdn = of_find_node_by_path("/");
657 tmp = of_get_property(rootdn, "model", NULL);
660 tmp = of_get_property(rootdn, "system-id", NULL);
663 lp_index_ptr = of_get_property(rootdn, "ibm,partition-no",
666 lp_index = *lp_index_ptr;
669 seq_printf(m, "serial_number=%s\n", system_id);
670 seq_printf(m, "system_type=%s\n", model);
671 seq_printf(m, "partition_id=%d\n", (int)lp_index);
673 return pseries_lparcfg_data(m, v);
676 static int lparcfg_open(struct inode *inode, struct file *file)
678 return single_open(file, lparcfg_data, NULL);
681 static const struct file_operations lparcfg_fops = {
682 .owner = THIS_MODULE,
684 .write = lparcfg_write,
685 .open = lparcfg_open,
686 .release = single_release,
690 static int __init lparcfg_init(void)
692 struct proc_dir_entry *ent;
693 umode_t mode = S_IRUSR | S_IRGRP | S_IROTH;
695 /* Allow writing if we have FW_FEATURE_SPLPAR */
696 if (firmware_has_feature(FW_FEATURE_SPLPAR))
699 ent = proc_create("powerpc/lparcfg", mode, NULL, &lparcfg_fops);
701 printk(KERN_ERR "Failed to create powerpc/lparcfg\n");
705 proc_ppc64_lparcfg = ent;
709 static void __exit lparcfg_cleanup(void)
711 if (proc_ppc64_lparcfg)
712 remove_proc_entry("lparcfg", proc_ppc64_lparcfg->parent);
715 module_init(lparcfg_init);
716 module_exit(lparcfg_cleanup);
717 MODULE_DESCRIPTION("Interface for LPAR configuration data");
718 MODULE_AUTHOR("Dave Engebretsen");
719 MODULE_LICENSE("GPL");