lparcfg.c

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