1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * PowerPC64 LPAR Configuration Information Driver
4 *
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.
12 *
13 * This driver creates a proc file at /proc/ppc64/lparcfg which contains
14 * keyword - value pairs that specify the configuration of the partition.
15 */
16
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 <asm/papr-sysparm.h>
23 #include <linux/seq_file.h>
24 #include <linux/slab.h>
25 #include <linux/string.h>
26 #include <linux/uaccess.h>
27 #include <linux/hugetlb.h>
28 #include <asm/lppaca.h>
29 #include <asm/hvcall.h>
30 #include <asm/firmware.h>
31 #include <asm/rtas.h>
32 #include <asm/time.h>
33 #include <asm/vio.h>
34 #include <asm/mmu.h>
35 #include <asm/machdep.h>
36 #include <asm/drmem.h>
37
38 #include "pseries.h"
39 #include "vas.h" /* pseries_vas_dlpar_cpu() */
40
41 /*
42 * This isn't a module but we expose that to userspace
43 * via /proc so leave the definitions here
44 */
45 #define MODULE_VERS "1.9"
46 #define MODULE_NAME "lparcfg"
47
48 /* #define LPARCFG_DEBUG */
49
50 /*
51 * Track sum of all purrs across all processors. This is used to further
52 * calculate usage values by different applications
53 */
cpu_get_purr(void * arg)54 static void cpu_get_purr(void *arg)
55 {
56 atomic64_t *sum = arg;
57
58 atomic64_add(mfspr(SPRN_PURR), sum);
59 }
60
get_purr(void)61 static unsigned long get_purr(void)
62 {
63 atomic64_t purr = ATOMIC64_INIT(0);
64
65 on_each_cpu(cpu_get_purr, &purr, 1);
66
67 return atomic64_read(&purr);
68 }
69
70 /*
71 * Methods used to fetch LPAR data when running on a pSeries platform.
72 */
73
74 struct hvcall_ppp_data {
75 u64 entitlement;
76 u64 unallocated_entitlement;
77 u16 group_num;
78 u16 pool_num;
79 u8 capped;
80 u8 weight;
81 u8 unallocated_weight;
82 u8 resource_group_index;
83 u16 active_procs_in_resource_group;
84 u16 active_procs_in_pool;
85 u16 active_system_procs;
86 u16 phys_platform_procs;
87 u32 max_proc_cap_avail;
88 u32 entitled_proc_cap_avail;
89 };
90
91 /*
92 * H_GET_PPP hcall returns info in 5 parms.
93 * entitled_capacity,unallocated_capacity,
94 * aggregation, resource_capability).
95 *
96 * R4 = Entitled Processor Capacity Percentage.
97 * R5 = Unallocated Processor Capacity Percentage.
98 * R6 (AABBCCDDEEFFGGHH).
99 * XXXX - reserved (0)
100 * XXXX - Active Cores in Resource Group
101 * XXXX - Group Number
102 * XXXX - Pool Number.
103 * R7 (IIJJKKLLMMNNOOPP).
104 * XX - Resource group Number
105 * XX - bit 0-6 reserved (0). bit 7 is Capped indicator.
106 * XX - variable processor Capacity Weight
107 * XX - Unallocated Variable Processor Capacity Weight.
108 * XXXX - Active processors in Physical Processor Pool.
109 * XXXX - Processors active on platform.
110 * R8 (QQQQRRRRRRSSSSSS). if ibm,partition-performance-parameters-level >= 1
111 * XXXX - Physical platform procs allocated to virtualization.
112 * XXXXXX - Max procs capacity % available to the partitions pool.
113 * XXXXXX - Entitled procs capacity % available to the
114 * partitions pool.
115 */
h_get_ppp(struct hvcall_ppp_data * ppp_data)116 static unsigned int h_get_ppp(struct hvcall_ppp_data *ppp_data)
117 {
118 unsigned long retbuf[PLPAR_HCALL9_BUFSIZE] = {0};
119 long rc;
120
121 rc = plpar_hcall9(H_GET_PPP, retbuf);
122
123 ppp_data->entitlement = retbuf[0];
124 ppp_data->unallocated_entitlement = retbuf[1];
125
126 ppp_data->active_procs_in_resource_group = (retbuf[2] >> 4 * 8) & 0xffff;
127 ppp_data->group_num = (retbuf[2] >> 2 * 8) & 0xffff;
128 ppp_data->pool_num = retbuf[2] & 0xffff;
129
130 ppp_data->resource_group_index = (retbuf[3] >> 7 * 8) & 0xff;
131 ppp_data->capped = (retbuf[3] >> 6 * 8) & 0x01;
132 ppp_data->weight = (retbuf[3] >> 5 * 8) & 0xff;
133 ppp_data->unallocated_weight = (retbuf[3] >> 4 * 8) & 0xff;
134 ppp_data->active_procs_in_pool = (retbuf[3] >> 2 * 8) & 0xffff;
135 ppp_data->active_system_procs = retbuf[3] & 0xffff;
136
137 ppp_data->phys_platform_procs = retbuf[4] >> 6 * 8;
138 ppp_data->max_proc_cap_avail = (retbuf[4] >> 3 * 8) & 0xffffff;
139 ppp_data->entitled_proc_cap_avail = retbuf[4] & 0xffffff;
140
141 return rc;
142 }
143
show_gpci_data(struct seq_file * m)144 static void show_gpci_data(struct seq_file *m)
145 {
146 struct hv_gpci_request_buffer *buf;
147 unsigned int affinity_score;
148 long ret;
149
150 buf = kmalloc_obj(*buf);
151 if (buf == NULL)
152 return;
153
154 /*
155 * Show the local LPAR's affinity score.
156 *
157 * 0xB1 selects the Affinity_Domain_Info_By_Partition subcall.
158 * The score is at byte 0xB in the output buffer.
159 */
160 memset(&buf->params, 0, sizeof(buf->params));
161 buf->params.counter_request = cpu_to_be32(0xB1);
162 buf->params.starting_index = cpu_to_be32(-1); /* local LPAR */
163 buf->params.counter_info_version_in = 0x5; /* v5+ for score */
164 ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO, virt_to_phys(buf),
165 sizeof(*buf));
166 if (ret != H_SUCCESS) {
167 pr_debug("hcall failed: H_GET_PERF_COUNTER_INFO: %ld, %x\n",
168 ret, be32_to_cpu(buf->params.detail_rc));
169 goto out;
170 }
171 affinity_score = buf->bytes[0xB];
172 seq_printf(m, "partition_affinity_score=%u\n", affinity_score);
173 out:
174 kfree(buf);
175 }
176
h_pic(unsigned long * pool_idle_time,unsigned long * num_procs)177 static long h_pic(unsigned long *pool_idle_time,
178 unsigned long *num_procs)
179 {
180 long rc;
181 unsigned long retbuf[PLPAR_HCALL_BUFSIZE] = {0};
182
183 rc = plpar_hcall(H_PIC, retbuf);
184
185 if (pool_idle_time)
186 *pool_idle_time = retbuf[0];
187 if (num_procs)
188 *num_procs = retbuf[1];
189
190 return rc;
191 }
192
193 unsigned long boot_pool_idle_time;
194
195 /*
196 * parse_ppp_data
197 * Parse out the data returned from h_get_ppp and h_pic
198 */
parse_ppp_data(struct seq_file * m)199 static void parse_ppp_data(struct seq_file *m)
200 {
201 struct hvcall_ppp_data ppp_data;
202 struct device_node *root;
203 const __be32 *perf_level;
204 long rc;
205
206 rc = h_get_ppp(&ppp_data);
207 if (rc)
208 return;
209
210 seq_printf(m, "partition_entitled_capacity=%lld\n",
211 ppp_data.entitlement);
212 seq_printf(m, "group=%d\n", ppp_data.group_num);
213 seq_printf(m, "system_active_processors=%d\n",
214 ppp_data.active_system_procs);
215
216 /* pool related entries are appropriate for shared configs */
217 if (lppaca_shared_proc()) {
218 unsigned long pool_idle_time, pool_procs;
219
220 seq_printf(m, "pool=%d\n", ppp_data.pool_num);
221
222 /* report pool_capacity in percentage */
223 seq_printf(m, "pool_capacity=%d\n",
224 ppp_data.active_procs_in_pool * 100);
225
226 /* In case h_pic call is not successful, this would result in
227 * APP values being wrong in tools like lparstat.
228 */
229
230 if (h_pic(&pool_idle_time, &pool_procs) == H_SUCCESS) {
231 seq_printf(m, "pool_idle_time=%ld\n", pool_idle_time);
232 seq_printf(m, "pool_num_procs=%ld\n", pool_procs);
233 seq_printf(m, "boot_pool_idle_time=%ld\n", boot_pool_idle_time);
234 }
235 }
236
237 seq_printf(m, "unallocated_capacity_weight=%d\n",
238 ppp_data.unallocated_weight);
239 seq_printf(m, "capacity_weight=%d\n", ppp_data.weight);
240 seq_printf(m, "capped=%d\n", ppp_data.capped);
241 seq_printf(m, "unallocated_capacity=%lld\n",
242 ppp_data.unallocated_entitlement);
243
244 if (ppp_data.active_procs_in_resource_group) {
245 seq_printf(m, "resource_group_number=%d\n",
246 ppp_data.resource_group_index);
247 seq_printf(m, "resource_group_active_processors=%d\n",
248 ppp_data.active_procs_in_resource_group);
249 }
250
251 /* The last bits of information returned from h_get_ppp are only
252 * valid if the ibm,partition-performance-parameters-level
253 * property is >= 1.
254 */
255 root = of_find_node_by_path("/");
256 if (root) {
257 perf_level = of_get_property(root,
258 "ibm,partition-performance-parameters-level",
259 NULL);
260 if (perf_level && (be32_to_cpup(perf_level) >= 1)) {
261 seq_printf(m,
262 "physical_procs_allocated_to_virtualization=%d\n",
263 ppp_data.phys_platform_procs);
264 seq_printf(m, "max_proc_capacity_available=%d\n",
265 ppp_data.max_proc_cap_avail);
266 seq_printf(m, "entitled_proc_capacity_available=%d\n",
267 ppp_data.entitled_proc_cap_avail);
268 }
269
270 of_node_put(root);
271 }
272 }
273
274 /**
275 * parse_mpp_data
276 * Parse out data returned from h_get_mpp
277 */
parse_mpp_data(struct seq_file * m)278 static void parse_mpp_data(struct seq_file *m)
279 {
280 struct hvcall_mpp_data mpp_data;
281 int rc;
282
283 rc = h_get_mpp(&mpp_data);
284 if (rc)
285 return;
286
287 seq_printf(m, "entitled_memory=%ld\n", mpp_data.entitled_mem);
288
289 if (mpp_data.mapped_mem != -1)
290 seq_printf(m, "mapped_entitled_memory=%ld\n",
291 mpp_data.mapped_mem);
292
293 seq_printf(m, "entitled_memory_group_number=%d\n", mpp_data.group_num);
294 seq_printf(m, "entitled_memory_pool_number=%d\n", mpp_data.pool_num);
295
296 seq_printf(m, "entitled_memory_weight=%d\n", mpp_data.mem_weight);
297 seq_printf(m, "unallocated_entitled_memory_weight=%d\n",
298 mpp_data.unallocated_mem_weight);
299 seq_printf(m, "unallocated_io_mapping_entitlement=%ld\n",
300 mpp_data.unallocated_entitlement);
301
302 if (mpp_data.pool_size != -1)
303 seq_printf(m, "entitled_memory_pool_size=%ld bytes\n",
304 mpp_data.pool_size);
305
306 seq_printf(m, "entitled_memory_loan_request=%ld\n",
307 mpp_data.loan_request);
308
309 seq_printf(m, "backing_memory=%ld bytes\n", mpp_data.backing_mem);
310 }
311
312 /**
313 * parse_mpp_x_data
314 * Parse out data returned from h_get_mpp_x
315 */
parse_mpp_x_data(struct seq_file * m)316 static void parse_mpp_x_data(struct seq_file *m)
317 {
318 struct hvcall_mpp_x_data mpp_x_data;
319
320 if (!firmware_has_feature(FW_FEATURE_XCMO))
321 return;
322 if (h_get_mpp_x(&mpp_x_data))
323 return;
324
325 seq_printf(m, "coalesced_bytes=%ld\n", mpp_x_data.coalesced_bytes);
326
327 if (mpp_x_data.pool_coalesced_bytes)
328 seq_printf(m, "pool_coalesced_bytes=%ld\n",
329 mpp_x_data.pool_coalesced_bytes);
330 if (mpp_x_data.pool_purr_cycles)
331 seq_printf(m, "coalesce_pool_purr=%ld\n", mpp_x_data.pool_purr_cycles);
332 if (mpp_x_data.pool_spurr_cycles)
333 seq_printf(m, "coalesce_pool_spurr=%ld\n", mpp_x_data.pool_spurr_cycles);
334 }
335
336 /*
337 * Read the lpar name using the RTAS ibm,get-system-parameter call.
338 *
339 * The name read through this call is updated if changes are made by the end
340 * user on the hypervisor side.
341 *
342 * Some hypervisor (like Qemu) may not provide this value. In that case, a non
343 * null value is returned.
344 */
read_rtas_lpar_name(struct seq_file * m)345 static int read_rtas_lpar_name(struct seq_file *m)
346 {
347 struct papr_sysparm_buf *buf;
348 int err;
349
350 buf = papr_sysparm_buf_alloc();
351 if (!buf)
352 return -ENOMEM;
353
354 err = papr_sysparm_get(PAPR_SYSPARM_LPAR_NAME, buf);
355 if (!err)
356 seq_printf(m, "partition_name=%s\n", buf->val);
357
358 papr_sysparm_buf_free(buf);
359 return err;
360 }
361
362 /*
363 * Read the LPAR name from the Device Tree.
364 *
365 * The value read in the DT is not updated if the end-user is touching the LPAR
366 * name on the hypervisor side.
367 */
read_dt_lpar_name(struct seq_file * m)368 static int read_dt_lpar_name(struct seq_file *m)
369 {
370 struct device_node *root = of_find_node_by_path("/");
371 const char *name;
372 int ret;
373
374 ret = of_property_read_string(root, "ibm,partition-name", &name);
375 of_node_put(root);
376 if (ret)
377 return -ENOENT;
378
379 seq_printf(m, "partition_name=%s\n", name);
380 return 0;
381 }
382
read_lpar_name(struct seq_file * m)383 static void read_lpar_name(struct seq_file *m)
384 {
385 if (read_rtas_lpar_name(m))
386 read_dt_lpar_name(m);
387 }
388
389 #define SPLPAR_MAXLENGTH 1026*(sizeof(char))
390
391 /*
392 * parse_system_parameter_string()
393 * Retrieve the potential_processors, max_entitled_capacity and friends
394 * through the get-system-parameter rtas call. Replace keyword strings as
395 * necessary.
396 */
parse_system_parameter_string(struct seq_file * m)397 static void parse_system_parameter_string(struct seq_file *m)
398 {
399 struct papr_sysparm_buf *buf;
400
401 buf = papr_sysparm_buf_alloc();
402 if (!buf)
403 return;
404
405 if (papr_sysparm_get(PAPR_SYSPARM_SHARED_PROC_LPAR_ATTRS, buf)) {
406 goto out_free;
407 } else {
408 const char *local_buffer;
409 int splpar_strlen;
410 int idx, w_idx;
411 char *workbuffer = kzalloc(SPLPAR_MAXLENGTH, GFP_KERNEL);
412
413 if (!workbuffer)
414 goto out_free;
415
416 splpar_strlen = be16_to_cpu(buf->len);
417 local_buffer = buf->val;
418
419 w_idx = 0;
420 idx = 0;
421 while ((*local_buffer) && (idx < splpar_strlen)) {
422 workbuffer[w_idx++] = local_buffer[idx++];
423 if ((local_buffer[idx] == ',')
424 || (local_buffer[idx] == '\0')) {
425 workbuffer[w_idx] = '\0';
426 if (w_idx) {
427 /* avoid the empty string */
428 seq_printf(m, "%s\n", workbuffer);
429 }
430 memset(workbuffer, 0, SPLPAR_MAXLENGTH);
431 idx++; /* skip the comma */
432 w_idx = 0;
433 } else if (local_buffer[idx] == '=') {
434 /* code here to replace workbuffer contents
435 * with different keyword strings. Truncation
436 * by strscpy is deliberately ignored because
437 * SPLPAR_MAXLENGTH >= maximum string size.
438 */
439 if (!strcmp(workbuffer, "MaxEntCap"))
440 w_idx = strscpy(workbuffer,
441 "partition_max_entitled_capacity",
442 SPLPAR_MAXLENGTH);
443 if (!strcmp(workbuffer, "MaxPlatProcs"))
444 w_idx = strscpy(workbuffer,
445 "system_potential_processors",
446 SPLPAR_MAXLENGTH);
447 }
448 }
449 kfree(workbuffer);
450 local_buffer -= 2; /* back up over strlen value */
451 }
452 out_free:
453 papr_sysparm_buf_free(buf);
454 }
455
456 /* Return the number of processors in the system.
457 * This function reads through the device tree and counts
458 * the virtual processors, this does not include threads.
459 */
lparcfg_count_active_processors(void)460 static int lparcfg_count_active_processors(void)
461 {
462 struct device_node *cpus_dn;
463 int count = 0;
464
465 for_each_node_by_type(cpus_dn, "cpu") {
466 #ifdef LPARCFG_DEBUG
467 printk(KERN_ERR "cpus_dn %p\n", cpus_dn);
468 #endif
469 count++;
470 }
471 return count;
472 }
473
pseries_cmo_data(struct seq_file * m)474 static void pseries_cmo_data(struct seq_file *m)
475 {
476 int cpu;
477 unsigned long cmo_faults = 0;
478 unsigned long cmo_fault_time = 0;
479
480 seq_printf(m, "cmo_enabled=%d\n", firmware_has_feature(FW_FEATURE_CMO));
481
482 if (!firmware_has_feature(FW_FEATURE_CMO))
483 return;
484
485 for_each_possible_cpu(cpu) {
486 cmo_faults += be64_to_cpu(lppaca_of(cpu).cmo_faults);
487 cmo_fault_time += be64_to_cpu(lppaca_of(cpu).cmo_fault_time);
488 }
489
490 seq_printf(m, "cmo_faults=%lu\n", cmo_faults);
491 seq_printf(m, "cmo_fault_time_usec=%lu\n",
492 cmo_fault_time / tb_ticks_per_usec);
493 seq_printf(m, "cmo_primary_psp=%d\n", cmo_get_primary_psp());
494 seq_printf(m, "cmo_secondary_psp=%d\n", cmo_get_secondary_psp());
495 seq_printf(m, "cmo_page_size=%lu\n", cmo_get_page_size());
496 }
497
splpar_dispatch_data(struct seq_file * m)498 static void splpar_dispatch_data(struct seq_file *m)
499 {
500 int cpu;
501 unsigned long dispatches = 0;
502 unsigned long dispatch_dispersions = 0;
503
504 for_each_possible_cpu(cpu) {
505 dispatches += be32_to_cpu(lppaca_of(cpu).yield_count);
506 dispatch_dispersions +=
507 be32_to_cpu(lppaca_of(cpu).dispersion_count);
508 }
509
510 seq_printf(m, "dispatches=%lu\n", dispatches);
511 seq_printf(m, "dispatch_dispersions=%lu\n", dispatch_dispersions);
512 }
513
parse_em_data(struct seq_file * m)514 static void parse_em_data(struct seq_file *m)
515 {
516 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
517
518 if (firmware_has_feature(FW_FEATURE_LPAR) &&
519 plpar_hcall(H_GET_EM_PARMS, retbuf) == H_SUCCESS)
520 seq_printf(m, "power_mode_data=%016lx\n", retbuf[0]);
521 }
522
maxmem_data(struct seq_file * m)523 static void maxmem_data(struct seq_file *m)
524 {
525 unsigned long maxmem = 0;
526
527 maxmem += (unsigned long)drmem_info->n_lmbs * drmem_info->lmb_size;
528 maxmem += hugetlb_total_pages() * PAGE_SIZE;
529
530 seq_printf(m, "MaxMem=%lu\n", maxmem);
531 }
532
pseries_lparcfg_data(struct seq_file * m,void * v)533 static int pseries_lparcfg_data(struct seq_file *m, void *v)
534 {
535 int partition_potential_processors;
536 int partition_active_processors;
537 struct device_node *rtas_node;
538 const __be32 *lrdrp = NULL;
539
540 rtas_node = of_find_node_by_path("/rtas");
541 if (rtas_node)
542 lrdrp = of_get_property(rtas_node, "ibm,lrdr-capacity", NULL);
543
544 if (lrdrp == NULL) {
545 partition_potential_processors = num_possible_cpus();
546 } else {
547 partition_potential_processors = be32_to_cpup(lrdrp + 4);
548 }
549 of_node_put(rtas_node);
550
551 partition_active_processors = lparcfg_count_active_processors();
552
553 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
554 /* this call handles the ibm,get-system-parameter contents */
555 read_lpar_name(m);
556 parse_system_parameter_string(m);
557 parse_ppp_data(m);
558 parse_mpp_data(m);
559 parse_mpp_x_data(m);
560 pseries_cmo_data(m);
561 splpar_dispatch_data(m);
562
563 seq_printf(m, "purr=%ld\n", get_purr());
564 seq_printf(m, "tbr=%ld\n", mftb());
565 } else { /* non SPLPAR case */
566
567 seq_printf(m, "system_active_processors=%d\n",
568 partition_active_processors);
569
570 seq_printf(m, "system_potential_processors=%d\n",
571 partition_potential_processors);
572
573 seq_printf(m, "partition_max_entitled_capacity=%d\n",
574 partition_potential_processors * 100);
575
576 seq_printf(m, "partition_entitled_capacity=%d\n",
577 partition_active_processors * 100);
578 }
579
580 show_gpci_data(m);
581
582 seq_printf(m, "partition_active_processors=%d\n",
583 partition_active_processors);
584
585 seq_printf(m, "partition_potential_processors=%d\n",
586 partition_potential_processors);
587
588 seq_printf(m, "shared_processor_mode=%d\n",
589 lppaca_shared_proc());
590
591 #ifdef CONFIG_PPC_64S_HASH_MMU
592 if (!radix_enabled())
593 seq_printf(m, "slb_size=%d\n", mmu_slb_size);
594 #endif
595 parse_em_data(m);
596 maxmem_data(m);
597
598 seq_printf(m, "security_flavor=%u\n", pseries_security_flavor);
599
600 return 0;
601 }
602
update_ppp(u64 * entitlement,u8 * weight)603 static ssize_t update_ppp(u64 *entitlement, u8 *weight)
604 {
605 struct hvcall_ppp_data ppp_data;
606 u8 new_weight;
607 u64 new_entitled;
608 ssize_t retval;
609
610 /* Get our current parameters */
611 retval = h_get_ppp(&ppp_data);
612 if (retval)
613 return retval;
614
615 if (entitlement) {
616 new_weight = ppp_data.weight;
617 new_entitled = *entitlement;
618 } else if (weight) {
619 new_weight = *weight;
620 new_entitled = ppp_data.entitlement;
621 } else
622 return -EINVAL;
623
624 pr_debug("%s: current_entitled = %llu, current_weight = %u\n",
625 __func__, ppp_data.entitlement, ppp_data.weight);
626
627 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
628 __func__, new_entitled, new_weight);
629
630 retval = plpar_hcall_norets(H_SET_PPP, new_entitled, new_weight);
631 return retval;
632 }
633
634 /**
635 * update_mpp
636 *
637 * Update the memory entitlement and weight for the partition. Caller must
638 * specify either a new entitlement or weight, not both, to be updated
639 * since the h_set_mpp call takes both entitlement and weight as parameters.
640 */
update_mpp(u64 * entitlement,u8 * weight)641 static ssize_t update_mpp(u64 *entitlement, u8 *weight)
642 {
643 struct hvcall_mpp_data mpp_data;
644 u64 new_entitled;
645 u8 new_weight;
646 ssize_t rc;
647
648 if (entitlement) {
649 /* Check with vio to ensure the new memory entitlement
650 * can be handled.
651 */
652 rc = vio_cmo_entitlement_update(*entitlement);
653 if (rc)
654 return rc;
655 }
656
657 rc = h_get_mpp(&mpp_data);
658 if (rc)
659 return rc;
660
661 if (entitlement) {
662 new_weight = mpp_data.mem_weight;
663 new_entitled = *entitlement;
664 } else if (weight) {
665 new_weight = *weight;
666 new_entitled = mpp_data.entitled_mem;
667 } else
668 return -EINVAL;
669
670 pr_debug("%s: current_entitled = %lu, current_weight = %u\n",
671 __func__, mpp_data.entitled_mem, mpp_data.mem_weight);
672
673 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
674 __func__, new_entitled, new_weight);
675
676 rc = plpar_hcall_norets(H_SET_MPP, new_entitled, new_weight);
677 return rc;
678 }
679
680 /*
681 * Interface for changing system parameters (variable capacity weight
682 * and entitled capacity). Format of input is "param_name=value";
683 * anything after value is ignored. Valid parameters at this time are
684 * "partition_entitled_capacity" and "capacity_weight". We use
685 * H_SET_PPP to alter parameters.
686 *
687 * This function should be invoked only on systems with
688 * FW_FEATURE_SPLPAR.
689 */
lparcfg_write(struct file * file,const char __user * buf,size_t count,loff_t * off)690 static ssize_t lparcfg_write(struct file *file, const char __user * buf,
691 size_t count, loff_t * off)
692 {
693 char kbuf[64];
694 char *tmp;
695 u64 new_entitled, *new_entitled_ptr = &new_entitled;
696 u8 new_weight, *new_weight_ptr = &new_weight;
697 ssize_t retval;
698
699 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
700 return -EINVAL;
701
702 if (count == 0 || count > sizeof(kbuf))
703 return -EINVAL;
704
705 if (copy_from_user(kbuf, buf, count))
706 return -EFAULT;
707
708 kbuf[count - 1] = '\0';
709 tmp = strchr(kbuf, '=');
710 if (!tmp)
711 return -EINVAL;
712
713 *tmp++ = '\0';
714
715 if (!strcmp(kbuf, "partition_entitled_capacity")) {
716 char *endp;
717 *new_entitled_ptr = (u64) simple_strtoul(tmp, &endp, 10);
718 if (endp == tmp)
719 return -EINVAL;
720
721 retval = update_ppp(new_entitled_ptr, NULL);
722
723 if (retval == H_SUCCESS || retval == H_CONSTRAINED) {
724 /*
725 * The hypervisor assigns VAS resources based
726 * on entitled capacity for shared mode.
727 * Reconfig VAS windows based on DLPAR CPU events.
728 */
729 if (pseries_vas_dlpar_cpu() != 0)
730 retval = H_HARDWARE;
731 }
732 } else if (!strcmp(kbuf, "capacity_weight")) {
733 char *endp;
734 *new_weight_ptr = (u8) simple_strtoul(tmp, &endp, 10);
735 if (endp == tmp)
736 return -EINVAL;
737
738 retval = update_ppp(NULL, new_weight_ptr);
739 } else if (!strcmp(kbuf, "entitled_memory")) {
740 char *endp;
741 *new_entitled_ptr = (u64) simple_strtoul(tmp, &endp, 10);
742 if (endp == tmp)
743 return -EINVAL;
744
745 retval = update_mpp(new_entitled_ptr, NULL);
746 } else if (!strcmp(kbuf, "entitled_memory_weight")) {
747 char *endp;
748 *new_weight_ptr = (u8) simple_strtoul(tmp, &endp, 10);
749 if (endp == tmp)
750 return -EINVAL;
751
752 retval = update_mpp(NULL, new_weight_ptr);
753 } else
754 return -EINVAL;
755
756 if (retval == H_SUCCESS || retval == H_CONSTRAINED) {
757 retval = count;
758 } else if (retval == H_BUSY) {
759 retval = -EBUSY;
760 } else if (retval == H_HARDWARE) {
761 retval = -EIO;
762 } else if (retval == H_PARAMETER) {
763 retval = -EINVAL;
764 }
765
766 return retval;
767 }
768
lparcfg_data(struct seq_file * m,void * v)769 static int lparcfg_data(struct seq_file *m, void *v)
770 {
771 struct device_node *rootdn;
772 const char *model = "";
773 const char *system_id = "";
774 const char *tmp;
775 const __be32 *lp_index_ptr;
776 unsigned int lp_index = 0;
777
778 seq_printf(m, "%s %s\n", MODULE_NAME, MODULE_VERS);
779
780 rootdn = of_find_node_by_path("/");
781 if (rootdn) {
782 tmp = of_get_property(rootdn, "model", NULL);
783 if (tmp)
784 model = tmp;
785 tmp = of_get_property(rootdn, "system-id", NULL);
786 if (tmp)
787 system_id = tmp;
788 lp_index_ptr = of_get_property(rootdn, "ibm,partition-no",
789 NULL);
790 if (lp_index_ptr)
791 lp_index = be32_to_cpup(lp_index_ptr);
792 of_node_put(rootdn);
793 }
794 seq_printf(m, "serial_number=%s\n", system_id);
795 seq_printf(m, "system_type=%s\n", model);
796 seq_printf(m, "partition_id=%d\n", (int)lp_index);
797
798 return pseries_lparcfg_data(m, v);
799 }
800
lparcfg_open(struct inode * inode,struct file * file)801 static int lparcfg_open(struct inode *inode, struct file *file)
802 {
803 return single_open(file, lparcfg_data, NULL);
804 }
805
806 static const struct proc_ops lparcfg_proc_ops = {
807 .proc_read = seq_read,
808 .proc_write = lparcfg_write,
809 .proc_open = lparcfg_open,
810 .proc_release = single_release,
811 .proc_lseek = seq_lseek,
812 };
813
lparcfg_init(void)814 static int __init lparcfg_init(void)
815 {
816 umode_t mode = 0444;
817 long retval;
818
819 /* Allow writing if we have FW_FEATURE_SPLPAR */
820 if (firmware_has_feature(FW_FEATURE_SPLPAR))
821 mode |= 0200;
822
823 if (!proc_create("powerpc/lparcfg", mode, NULL, &lparcfg_proc_ops)) {
824 printk(KERN_ERR "Failed to create powerpc/lparcfg\n");
825 return -EIO;
826 }
827
828 /* If this call fails, it would result in APP values
829 * being wrong for since boot reports of lparstat
830 */
831 retval = h_pic(&boot_pool_idle_time, NULL);
832
833 if (retval != H_SUCCESS)
834 pr_debug("H_PIC failed during lparcfg init retval: %ld\n",
835 retval);
836
837 return 0;
838 }
839 machine_device_initcall(pseries, lparcfg_init);
840