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 <linux/uaccess.h>
29 #include <asm/lppaca.h>
30 #include <asm/hvcall.h>
31 #include <asm/firmware.h>
35 #include <asm/vdso_datapage.h>
38 #include <asm/machdep.h>
43 * This isn't a module but we expose that to userspace
44 * via /proc so leave the definitions here
46 #define MODULE_VERS "1.9"
47 #define MODULE_NAME "lparcfg"
49 /* #define LPARCFG_DEBUG */
52 * Track sum of all purrs across all processors. This is used to further
53 * calculate usage values by different applications
55 static unsigned long get_purr(void)
57 unsigned long sum_purr
= 0;
60 for_each_possible_cpu(cpu
) {
63 cu
= &per_cpu(cpu_usage_array
, cpu
);
64 sum_purr
+= cu
->current_tb
;
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 unsigned h_pic(unsigned long *pool_idle_time
,
140 unsigned long *num_procs
)
143 unsigned long retbuf
[PLPAR_HCALL_BUFSIZE
];
145 rc
= plpar_hcall(H_PIC
, retbuf
);
147 *pool_idle_time
= retbuf
[0];
148 *num_procs
= retbuf
[1];
155 * Parse out the data returned from h_get_ppp and h_pic
157 static void parse_ppp_data(struct seq_file
*m
)
159 struct hvcall_ppp_data ppp_data
;
160 struct device_node
*root
;
161 const __be32
*perf_level
;
164 rc
= h_get_ppp(&ppp_data
);
168 seq_printf(m
, "partition_entitled_capacity=%lld\n",
169 ppp_data
.entitlement
);
170 seq_printf(m
, "group=%d\n", ppp_data
.group_num
);
171 seq_printf(m
, "system_active_processors=%d\n",
172 ppp_data
.active_system_procs
);
174 /* pool related entries are appropriate for shared configs */
175 if (lppaca_shared_proc(get_lppaca())) {
176 unsigned long pool_idle_time
, pool_procs
;
178 seq_printf(m
, "pool=%d\n", ppp_data
.pool_num
);
180 /* report pool_capacity in percentage */
181 seq_printf(m
, "pool_capacity=%d\n",
182 ppp_data
.active_procs_in_pool
* 100);
184 h_pic(&pool_idle_time
, &pool_procs
);
185 seq_printf(m
, "pool_idle_time=%ld\n", pool_idle_time
);
186 seq_printf(m
, "pool_num_procs=%ld\n", pool_procs
);
189 seq_printf(m
, "unallocated_capacity_weight=%d\n",
190 ppp_data
.unallocated_weight
);
191 seq_printf(m
, "capacity_weight=%d\n", ppp_data
.weight
);
192 seq_printf(m
, "capped=%d\n", ppp_data
.capped
);
193 seq_printf(m
, "unallocated_capacity=%lld\n",
194 ppp_data
.unallocated_entitlement
);
196 /* The last bits of information returned from h_get_ppp are only
197 * valid if the ibm,partition-performance-parameters-level
200 root
= of_find_node_by_path("/");
202 perf_level
= of_get_property(root
,
203 "ibm,partition-performance-parameters-level",
205 if (perf_level
&& (be32_to_cpup(perf_level
) >= 1)) {
207 "physical_procs_allocated_to_virtualization=%d\n",
208 ppp_data
.phys_platform_procs
);
209 seq_printf(m
, "max_proc_capacity_available=%d\n",
210 ppp_data
.max_proc_cap_avail
);
211 seq_printf(m
, "entitled_proc_capacity_available=%d\n",
212 ppp_data
.entitled_proc_cap_avail
);
221 * Parse out data returned from h_get_mpp
223 static void parse_mpp_data(struct seq_file
*m
)
225 struct hvcall_mpp_data mpp_data
;
228 rc
= h_get_mpp(&mpp_data
);
232 seq_printf(m
, "entitled_memory=%ld\n", mpp_data
.entitled_mem
);
234 if (mpp_data
.mapped_mem
!= -1)
235 seq_printf(m
, "mapped_entitled_memory=%ld\n",
236 mpp_data
.mapped_mem
);
238 seq_printf(m
, "entitled_memory_group_number=%d\n", mpp_data
.group_num
);
239 seq_printf(m
, "entitled_memory_pool_number=%d\n", mpp_data
.pool_num
);
241 seq_printf(m
, "entitled_memory_weight=%d\n", mpp_data
.mem_weight
);
242 seq_printf(m
, "unallocated_entitled_memory_weight=%d\n",
243 mpp_data
.unallocated_mem_weight
);
244 seq_printf(m
, "unallocated_io_mapping_entitlement=%ld\n",
245 mpp_data
.unallocated_entitlement
);
247 if (mpp_data
.pool_size
!= -1)
248 seq_printf(m
, "entitled_memory_pool_size=%ld bytes\n",
251 seq_printf(m
, "entitled_memory_loan_request=%ld\n",
252 mpp_data
.loan_request
);
254 seq_printf(m
, "backing_memory=%ld bytes\n", mpp_data
.backing_mem
);
259 * Parse out data returned from h_get_mpp_x
261 static void parse_mpp_x_data(struct seq_file
*m
)
263 struct hvcall_mpp_x_data mpp_x_data
;
265 if (!firmware_has_feature(FW_FEATURE_XCMO
))
267 if (h_get_mpp_x(&mpp_x_data
))
270 seq_printf(m
, "coalesced_bytes=%ld\n", mpp_x_data
.coalesced_bytes
);
272 if (mpp_x_data
.pool_coalesced_bytes
)
273 seq_printf(m
, "pool_coalesced_bytes=%ld\n",
274 mpp_x_data
.pool_coalesced_bytes
);
275 if (mpp_x_data
.pool_purr_cycles
)
276 seq_printf(m
, "coalesce_pool_purr=%ld\n", mpp_x_data
.pool_purr_cycles
);
277 if (mpp_x_data
.pool_spurr_cycles
)
278 seq_printf(m
, "coalesce_pool_spurr=%ld\n", mpp_x_data
.pool_spurr_cycles
);
281 #define SPLPAR_CHARACTERISTICS_TOKEN 20
282 #define SPLPAR_MAXLENGTH 1026*(sizeof(char))
285 * parse_system_parameter_string()
286 * Retrieve the potential_processors, max_entitled_capacity and friends
287 * through the get-system-parameter rtas call. Replace keyword strings as
290 static void parse_system_parameter_string(struct seq_file
*m
)
294 unsigned char *local_buffer
= kmalloc(SPLPAR_MAXLENGTH
, GFP_KERNEL
);
296 printk(KERN_ERR
"%s %s kmalloc failure at line %d\n",
297 __FILE__
, __func__
, __LINE__
);
301 spin_lock(&rtas_data_buf_lock
);
302 memset(rtas_data_buf
, 0, SPLPAR_MAXLENGTH
);
303 call_status
= rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
305 SPLPAR_CHARACTERISTICS_TOKEN
,
308 memcpy(local_buffer
, rtas_data_buf
, SPLPAR_MAXLENGTH
);
309 local_buffer
[SPLPAR_MAXLENGTH
- 1] = '\0';
310 spin_unlock(&rtas_data_buf_lock
);
312 if (call_status
!= 0) {
314 "%s %s Error calling get-system-parameter (0x%x)\n",
315 __FILE__
, __func__
, call_status
);
319 char *workbuffer
= kzalloc(SPLPAR_MAXLENGTH
, GFP_KERNEL
);
321 printk(KERN_ERR
"%s %s kmalloc failure at line %d\n",
322 __FILE__
, __func__
, __LINE__
);
327 printk(KERN_INFO
"success calling get-system-parameter\n");
329 splpar_strlen
= local_buffer
[0] * 256 + local_buffer
[1];
330 local_buffer
+= 2; /* step over strlen value */
334 while ((*local_buffer
) && (idx
< splpar_strlen
)) {
335 workbuffer
[w_idx
++] = local_buffer
[idx
++];
336 if ((local_buffer
[idx
] == ',')
337 || (local_buffer
[idx
] == '\0')) {
338 workbuffer
[w_idx
] = '\0';
340 /* avoid the empty string */
341 seq_printf(m
, "%s\n", workbuffer
);
343 memset(workbuffer
, 0, SPLPAR_MAXLENGTH
);
344 idx
++; /* skip the comma */
346 } else if (local_buffer
[idx
] == '=') {
347 /* code here to replace workbuffer contents
348 with different keyword strings */
349 if (0 == strcmp(workbuffer
, "MaxEntCap")) {
351 "partition_max_entitled_capacity");
352 w_idx
= strlen(workbuffer
);
354 if (0 == strcmp(workbuffer
, "MaxPlatProcs")) {
356 "system_potential_processors");
357 w_idx
= strlen(workbuffer
);
362 local_buffer
-= 2; /* back up over strlen value */
367 /* Return the number of processors in the system.
368 * This function reads through the device tree and counts
369 * the virtual processors, this does not include threads.
371 static int lparcfg_count_active_processors(void)
373 struct device_node
*cpus_dn
= NULL
;
376 while ((cpus_dn
= of_find_node_by_type(cpus_dn
, "cpu"))) {
378 printk(KERN_ERR
"cpus_dn %p\n", cpus_dn
);
385 static void pseries_cmo_data(struct seq_file
*m
)
388 unsigned long cmo_faults
= 0;
389 unsigned long cmo_fault_time
= 0;
391 seq_printf(m
, "cmo_enabled=%d\n", firmware_has_feature(FW_FEATURE_CMO
));
393 if (!firmware_has_feature(FW_FEATURE_CMO
))
396 for_each_possible_cpu(cpu
) {
397 cmo_faults
+= be64_to_cpu(lppaca_of(cpu
).cmo_faults
);
398 cmo_fault_time
+= be64_to_cpu(lppaca_of(cpu
).cmo_fault_time
);
401 seq_printf(m
, "cmo_faults=%lu\n", cmo_faults
);
402 seq_printf(m
, "cmo_fault_time_usec=%lu\n",
403 cmo_fault_time
/ tb_ticks_per_usec
);
404 seq_printf(m
, "cmo_primary_psp=%d\n", cmo_get_primary_psp());
405 seq_printf(m
, "cmo_secondary_psp=%d\n", cmo_get_secondary_psp());
406 seq_printf(m
, "cmo_page_size=%lu\n", cmo_get_page_size());
409 static void splpar_dispatch_data(struct seq_file
*m
)
412 unsigned long dispatches
= 0;
413 unsigned long dispatch_dispersions
= 0;
415 for_each_possible_cpu(cpu
) {
416 dispatches
+= be32_to_cpu(lppaca_of(cpu
).yield_count
);
417 dispatch_dispersions
+=
418 be32_to_cpu(lppaca_of(cpu
).dispersion_count
);
421 seq_printf(m
, "dispatches=%lu\n", dispatches
);
422 seq_printf(m
, "dispatch_dispersions=%lu\n", dispatch_dispersions
);
425 static void parse_em_data(struct seq_file
*m
)
427 unsigned long retbuf
[PLPAR_HCALL_BUFSIZE
];
429 if (firmware_has_feature(FW_FEATURE_LPAR
) &&
430 plpar_hcall(H_GET_EM_PARMS
, retbuf
) == H_SUCCESS
)
431 seq_printf(m
, "power_mode_data=%016lx\n", retbuf
[0]);
434 static int pseries_lparcfg_data(struct seq_file
*m
, void *v
)
436 int partition_potential_processors
;
437 int partition_active_processors
;
438 struct device_node
*rtas_node
;
439 const __be32
*lrdrp
= NULL
;
441 rtas_node
= of_find_node_by_path("/rtas");
443 lrdrp
= of_get_property(rtas_node
, "ibm,lrdr-capacity", NULL
);
446 partition_potential_processors
= vdso_data
->processorCount
;
448 partition_potential_processors
= be32_to_cpup(lrdrp
+ 4);
450 of_node_put(rtas_node
);
452 partition_active_processors
= lparcfg_count_active_processors();
454 if (firmware_has_feature(FW_FEATURE_SPLPAR
)) {
455 /* this call handles the ibm,get-system-parameter contents */
456 parse_system_parameter_string(m
);
461 splpar_dispatch_data(m
);
463 seq_printf(m
, "purr=%ld\n", get_purr());
464 } else { /* non SPLPAR case */
466 seq_printf(m
, "system_active_processors=%d\n",
467 partition_potential_processors
);
469 seq_printf(m
, "system_potential_processors=%d\n",
470 partition_potential_processors
);
472 seq_printf(m
, "partition_max_entitled_capacity=%d\n",
473 partition_potential_processors
* 100);
475 seq_printf(m
, "partition_entitled_capacity=%d\n",
476 partition_active_processors
* 100);
479 seq_printf(m
, "partition_active_processors=%d\n",
480 partition_active_processors
);
482 seq_printf(m
, "partition_potential_processors=%d\n",
483 partition_potential_processors
);
485 seq_printf(m
, "shared_processor_mode=%d\n",
486 lppaca_shared_proc(get_lppaca()));
488 #ifdef CONFIG_PPC_STD_MMU_64
489 seq_printf(m
, "slb_size=%d\n", mmu_slb_size
);
496 static ssize_t
update_ppp(u64
*entitlement
, u8
*weight
)
498 struct hvcall_ppp_data ppp_data
;
503 /* Get our current parameters */
504 retval
= h_get_ppp(&ppp_data
);
509 new_weight
= ppp_data
.weight
;
510 new_entitled
= *entitlement
;
512 new_weight
= *weight
;
513 new_entitled
= ppp_data
.entitlement
;
517 pr_debug("%s: current_entitled = %llu, current_weight = %u\n",
518 __func__
, ppp_data
.entitlement
, ppp_data
.weight
);
520 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
521 __func__
, new_entitled
, new_weight
);
523 retval
= plpar_hcall_norets(H_SET_PPP
, new_entitled
, new_weight
);
530 * Update the memory entitlement and weight for the partition. Caller must
531 * specify either a new entitlement or weight, not both, to be updated
532 * since the h_set_mpp call takes both entitlement and weight as parameters.
534 static ssize_t
update_mpp(u64
*entitlement
, u8
*weight
)
536 struct hvcall_mpp_data mpp_data
;
542 /* Check with vio to ensure the new memory entitlement
545 rc
= vio_cmo_entitlement_update(*entitlement
);
550 rc
= h_get_mpp(&mpp_data
);
555 new_weight
= mpp_data
.mem_weight
;
556 new_entitled
= *entitlement
;
558 new_weight
= *weight
;
559 new_entitled
= mpp_data
.entitled_mem
;
563 pr_debug("%s: current_entitled = %lu, current_weight = %u\n",
564 __func__
, mpp_data
.entitled_mem
, mpp_data
.mem_weight
);
566 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
567 __func__
, new_entitled
, new_weight
);
569 rc
= plpar_hcall_norets(H_SET_MPP
, new_entitled
, new_weight
);
574 * Interface for changing system parameters (variable capacity weight
575 * and entitled capacity). Format of input is "param_name=value";
576 * anything after value is ignored. Valid parameters at this time are
577 * "partition_entitled_capacity" and "capacity_weight". We use
578 * H_SET_PPP to alter parameters.
580 * This function should be invoked only on systems with
583 static ssize_t
lparcfg_write(struct file
*file
, const char __user
* buf
,
584 size_t count
, loff_t
* off
)
589 u64 new_entitled
, *new_entitled_ptr
= &new_entitled
;
590 u8 new_weight
, *new_weight_ptr
= &new_weight
;
593 if (!firmware_has_feature(FW_FEATURE_SPLPAR
))
599 if (copy_from_user(kbuf
, buf
, count
))
602 kbuf
[count
- 1] = '\0';
603 tmp
= strchr(kbuf
, '=');
609 if (!strcmp(kbuf
, "partition_entitled_capacity")) {
611 *new_entitled_ptr
= (u64
) simple_strtoul(tmp
, &endp
, 10);
615 retval
= update_ppp(new_entitled_ptr
, NULL
);
616 } else if (!strcmp(kbuf
, "capacity_weight")) {
618 *new_weight_ptr
= (u8
) simple_strtoul(tmp
, &endp
, 10);
622 retval
= update_ppp(NULL
, new_weight_ptr
);
623 } else if (!strcmp(kbuf
, "entitled_memory")) {
625 *new_entitled_ptr
= (u64
) simple_strtoul(tmp
, &endp
, 10);
629 retval
= update_mpp(new_entitled_ptr
, NULL
);
630 } else if (!strcmp(kbuf
, "entitled_memory_weight")) {
632 *new_weight_ptr
= (u8
) simple_strtoul(tmp
, &endp
, 10);
636 retval
= update_mpp(NULL
, new_weight_ptr
);
640 if (retval
== H_SUCCESS
|| retval
== H_CONSTRAINED
) {
642 } else if (retval
== H_BUSY
) {
644 } else if (retval
== H_HARDWARE
) {
646 } else if (retval
== H_PARAMETER
) {
653 static int lparcfg_data(struct seq_file
*m
, void *v
)
655 struct device_node
*rootdn
;
656 const char *model
= "";
657 const char *system_id
= "";
659 const __be32
*lp_index_ptr
;
660 unsigned int lp_index
= 0;
662 seq_printf(m
, "%s %s\n", MODULE_NAME
, MODULE_VERS
);
664 rootdn
= of_find_node_by_path("/");
666 tmp
= of_get_property(rootdn
, "model", NULL
);
669 tmp
= of_get_property(rootdn
, "system-id", NULL
);
672 lp_index_ptr
= of_get_property(rootdn
, "ibm,partition-no",
675 lp_index
= be32_to_cpup(lp_index_ptr
);
678 seq_printf(m
, "serial_number=%s\n", system_id
);
679 seq_printf(m
, "system_type=%s\n", model
);
680 seq_printf(m
, "partition_id=%d\n", (int)lp_index
);
682 return pseries_lparcfg_data(m
, v
);
685 static int lparcfg_open(struct inode
*inode
, struct file
*file
)
687 return single_open(file
, lparcfg_data
, NULL
);
690 static const struct file_operations lparcfg_fops
= {
692 .write
= lparcfg_write
,
693 .open
= lparcfg_open
,
694 .release
= single_release
,
698 static int __init
lparcfg_init(void)
700 umode_t mode
= S_IRUSR
| S_IRGRP
| S_IROTH
;
702 /* Allow writing if we have FW_FEATURE_SPLPAR */
703 if (firmware_has_feature(FW_FEATURE_SPLPAR
))
706 if (!proc_create("powerpc/lparcfg", mode
, NULL
, &lparcfg_fops
)) {
707 printk(KERN_ERR
"Failed to create powerpc/lparcfg\n");
712 machine_device_initcall(pseries
, lparcfg_init
);