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>
35 #include <asm/vdso_datapage.h>
38 #include <asm/machdep.h>
42 * This isn't a module but we expose that to userspace
43 * via /proc so leave the definitions here
45 #define MODULE_VERS "1.9"
46 #define MODULE_NAME "lparcfg"
48 /* #define LPARCFG_DEBUG */
51 * Track sum of all purrs across all processors. This is used to further
52 * calculate usage values by different applications
54 static unsigned long get_purr(void)
56 unsigned long sum_purr
= 0;
59 for_each_possible_cpu(cpu
) {
62 cu
= &per_cpu(cpu_usage_array
, cpu
);
63 sum_purr
+= cu
->current_tb
;
69 * Methods used to fetch LPAR data when running on a pSeries platform.
72 struct hvcall_ppp_data
{
74 u64 unallocated_entitlement
;
79 u8 unallocated_weight
;
80 u16 active_procs_in_pool
;
81 u16 active_system_procs
;
82 u16 phys_platform_procs
;
83 u32 max_proc_cap_avail
;
84 u32 entitled_proc_cap_avail
;
88 * H_GET_PPP hcall returns info in 4 parms.
89 * entitled_capacity,unallocated_capacity,
90 * aggregation, resource_capability).
92 * R4 = Entitled Processor Capacity Percentage.
93 * R5 = Unallocated Processor Capacity Percentage.
94 * R6 (AABBCCDDEEFFGGHH).
99 * R7 (IIJJKKLLMMNNOOPP).
101 * XX - bit 0-6 reserved (0). bit 7 is Capped indicator.
102 * XX - variable processor Capacity Weight
103 * XX - Unallocated Variable Processor Capacity Weight.
104 * XXXX - Active processors in Physical Processor Pool.
105 * XXXX - Processors active on platform.
106 * R8 (QQQQRRRRRRSSSSSS). if ibm,partition-performance-parameters-level >= 1
107 * XXXX - Physical platform procs allocated to virtualization.
108 * XXXXXX - Max procs capacity % available to the partitions pool.
109 * XXXXXX - Entitled procs capacity % available to the
112 static unsigned int h_get_ppp(struct hvcall_ppp_data
*ppp_data
)
115 unsigned long retbuf
[PLPAR_HCALL9_BUFSIZE
];
117 rc
= plpar_hcall9(H_GET_PPP
, retbuf
);
119 ppp_data
->entitlement
= retbuf
[0];
120 ppp_data
->unallocated_entitlement
= retbuf
[1];
122 ppp_data
->group_num
= (retbuf
[2] >> 2 * 8) & 0xffff;
123 ppp_data
->pool_num
= retbuf
[2] & 0xffff;
125 ppp_data
->capped
= (retbuf
[3] >> 6 * 8) & 0x01;
126 ppp_data
->weight
= (retbuf
[3] >> 5 * 8) & 0xff;
127 ppp_data
->unallocated_weight
= (retbuf
[3] >> 4 * 8) & 0xff;
128 ppp_data
->active_procs_in_pool
= (retbuf
[3] >> 2 * 8) & 0xffff;
129 ppp_data
->active_system_procs
= retbuf
[3] & 0xffff;
131 ppp_data
->phys_platform_procs
= retbuf
[4] >> 6 * 8;
132 ppp_data
->max_proc_cap_avail
= (retbuf
[4] >> 3 * 8) & 0xffffff;
133 ppp_data
->entitled_proc_cap_avail
= retbuf
[4] & 0xffffff;
138 static unsigned h_pic(unsigned long *pool_idle_time
,
139 unsigned long *num_procs
)
142 unsigned long retbuf
[PLPAR_HCALL_BUFSIZE
];
144 rc
= plpar_hcall(H_PIC
, retbuf
);
146 *pool_idle_time
= retbuf
[0];
147 *num_procs
= retbuf
[1];
154 * Parse out the data returned from h_get_ppp and h_pic
156 static void parse_ppp_data(struct seq_file
*m
)
158 struct hvcall_ppp_data ppp_data
;
159 struct device_node
*root
;
160 const __be32
*perf_level
;
163 rc
= h_get_ppp(&ppp_data
);
167 seq_printf(m
, "partition_entitled_capacity=%lld\n",
168 ppp_data
.entitlement
);
169 seq_printf(m
, "group=%d\n", ppp_data
.group_num
);
170 seq_printf(m
, "system_active_processors=%d\n",
171 ppp_data
.active_system_procs
);
173 /* pool related entries are appropriate for shared configs */
174 if (lppaca_shared_proc(get_lppaca())) {
175 unsigned long pool_idle_time
, pool_procs
;
177 seq_printf(m
, "pool=%d\n", ppp_data
.pool_num
);
179 /* report pool_capacity in percentage */
180 seq_printf(m
, "pool_capacity=%d\n",
181 ppp_data
.active_procs_in_pool
* 100);
183 h_pic(&pool_idle_time
, &pool_procs
);
184 seq_printf(m
, "pool_idle_time=%ld\n", pool_idle_time
);
185 seq_printf(m
, "pool_num_procs=%ld\n", pool_procs
);
188 seq_printf(m
, "unallocated_capacity_weight=%d\n",
189 ppp_data
.unallocated_weight
);
190 seq_printf(m
, "capacity_weight=%d\n", ppp_data
.weight
);
191 seq_printf(m
, "capped=%d\n", ppp_data
.capped
);
192 seq_printf(m
, "unallocated_capacity=%lld\n",
193 ppp_data
.unallocated_entitlement
);
195 /* The last bits of information returned from h_get_ppp are only
196 * valid if the ibm,partition-performance-parameters-level
199 root
= of_find_node_by_path("/");
201 perf_level
= of_get_property(root
,
202 "ibm,partition-performance-parameters-level",
204 if (perf_level
&& (be32_to_cpup(perf_level
) >= 1)) {
206 "physical_procs_allocated_to_virtualization=%d\n",
207 ppp_data
.phys_platform_procs
);
208 seq_printf(m
, "max_proc_capacity_available=%d\n",
209 ppp_data
.max_proc_cap_avail
);
210 seq_printf(m
, "entitled_proc_capacity_available=%d\n",
211 ppp_data
.entitled_proc_cap_avail
);
220 * Parse out data returned from h_get_mpp
222 static void parse_mpp_data(struct seq_file
*m
)
224 struct hvcall_mpp_data mpp_data
;
227 rc
= h_get_mpp(&mpp_data
);
231 seq_printf(m
, "entitled_memory=%ld\n", mpp_data
.entitled_mem
);
233 if (mpp_data
.mapped_mem
!= -1)
234 seq_printf(m
, "mapped_entitled_memory=%ld\n",
235 mpp_data
.mapped_mem
);
237 seq_printf(m
, "entitled_memory_group_number=%d\n", mpp_data
.group_num
);
238 seq_printf(m
, "entitled_memory_pool_number=%d\n", mpp_data
.pool_num
);
240 seq_printf(m
, "entitled_memory_weight=%d\n", mpp_data
.mem_weight
);
241 seq_printf(m
, "unallocated_entitled_memory_weight=%d\n",
242 mpp_data
.unallocated_mem_weight
);
243 seq_printf(m
, "unallocated_io_mapping_entitlement=%ld\n",
244 mpp_data
.unallocated_entitlement
);
246 if (mpp_data
.pool_size
!= -1)
247 seq_printf(m
, "entitled_memory_pool_size=%ld bytes\n",
250 seq_printf(m
, "entitled_memory_loan_request=%ld\n",
251 mpp_data
.loan_request
);
253 seq_printf(m
, "backing_memory=%ld bytes\n", mpp_data
.backing_mem
);
258 * Parse out data returned from h_get_mpp_x
260 static void parse_mpp_x_data(struct seq_file
*m
)
262 struct hvcall_mpp_x_data mpp_x_data
;
264 if (!firmware_has_feature(FW_FEATURE_XCMO
))
266 if (h_get_mpp_x(&mpp_x_data
))
269 seq_printf(m
, "coalesced_bytes=%ld\n", mpp_x_data
.coalesced_bytes
);
271 if (mpp_x_data
.pool_coalesced_bytes
)
272 seq_printf(m
, "pool_coalesced_bytes=%ld\n",
273 mpp_x_data
.pool_coalesced_bytes
);
274 if (mpp_x_data
.pool_purr_cycles
)
275 seq_printf(m
, "coalesce_pool_purr=%ld\n", mpp_x_data
.pool_purr_cycles
);
276 if (mpp_x_data
.pool_spurr_cycles
)
277 seq_printf(m
, "coalesce_pool_spurr=%ld\n", mpp_x_data
.pool_spurr_cycles
);
280 #define SPLPAR_CHARACTERISTICS_TOKEN 20
281 #define SPLPAR_MAXLENGTH 1026*(sizeof(char))
284 * parse_system_parameter_string()
285 * Retrieve the potential_processors, max_entitled_capacity and friends
286 * through the get-system-parameter rtas call. Replace keyword strings as
289 static void parse_system_parameter_string(struct seq_file
*m
)
293 unsigned char *local_buffer
= kmalloc(SPLPAR_MAXLENGTH
, GFP_KERNEL
);
295 printk(KERN_ERR
"%s %s kmalloc failure at line %d\n",
296 __FILE__
, __func__
, __LINE__
);
300 spin_lock(&rtas_data_buf_lock
);
301 memset(rtas_data_buf
, 0, SPLPAR_MAXLENGTH
);
302 call_status
= rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
304 SPLPAR_CHARACTERISTICS_TOKEN
,
307 memcpy(local_buffer
, rtas_data_buf
, SPLPAR_MAXLENGTH
);
308 local_buffer
[SPLPAR_MAXLENGTH
- 1] = '\0';
309 spin_unlock(&rtas_data_buf_lock
);
311 if (call_status
!= 0) {
313 "%s %s Error calling get-system-parameter (0x%x)\n",
314 __FILE__
, __func__
, call_status
);
318 char *workbuffer
= kzalloc(SPLPAR_MAXLENGTH
, GFP_KERNEL
);
320 printk(KERN_ERR
"%s %s kmalloc failure at line %d\n",
321 __FILE__
, __func__
, __LINE__
);
326 printk(KERN_INFO
"success calling get-system-parameter\n");
328 splpar_strlen
= local_buffer
[0] * 256 + local_buffer
[1];
329 local_buffer
+= 2; /* step over strlen value */
333 while ((*local_buffer
) && (idx
< splpar_strlen
)) {
334 workbuffer
[w_idx
++] = local_buffer
[idx
++];
335 if ((local_buffer
[idx
] == ',')
336 || (local_buffer
[idx
] == '\0')) {
337 workbuffer
[w_idx
] = '\0';
339 /* avoid the empty string */
340 seq_printf(m
, "%s\n", workbuffer
);
342 memset(workbuffer
, 0, SPLPAR_MAXLENGTH
);
343 idx
++; /* skip the comma */
345 } else if (local_buffer
[idx
] == '=') {
346 /* code here to replace workbuffer contents
347 with different keyword strings */
348 if (0 == strcmp(workbuffer
, "MaxEntCap")) {
350 "partition_max_entitled_capacity");
351 w_idx
= strlen(workbuffer
);
353 if (0 == strcmp(workbuffer
, "MaxPlatProcs")) {
355 "system_potential_processors");
356 w_idx
= strlen(workbuffer
);
361 local_buffer
-= 2; /* back up over strlen value */
366 /* Return the number of processors in the system.
367 * This function reads through the device tree and counts
368 * the virtual processors, this does not include threads.
370 static int lparcfg_count_active_processors(void)
372 struct device_node
*cpus_dn
= NULL
;
375 while ((cpus_dn
= of_find_node_by_type(cpus_dn
, "cpu"))) {
377 printk(KERN_ERR
"cpus_dn %p\n", cpus_dn
);
384 static void pseries_cmo_data(struct seq_file
*m
)
387 unsigned long cmo_faults
= 0;
388 unsigned long cmo_fault_time
= 0;
390 seq_printf(m
, "cmo_enabled=%d\n", firmware_has_feature(FW_FEATURE_CMO
));
392 if (!firmware_has_feature(FW_FEATURE_CMO
))
395 for_each_possible_cpu(cpu
) {
396 cmo_faults
+= be64_to_cpu(lppaca_of(cpu
).cmo_faults
);
397 cmo_fault_time
+= be64_to_cpu(lppaca_of(cpu
).cmo_fault_time
);
400 seq_printf(m
, "cmo_faults=%lu\n", cmo_faults
);
401 seq_printf(m
, "cmo_fault_time_usec=%lu\n",
402 cmo_fault_time
/ tb_ticks_per_usec
);
403 seq_printf(m
, "cmo_primary_psp=%d\n", cmo_get_primary_psp());
404 seq_printf(m
, "cmo_secondary_psp=%d\n", cmo_get_secondary_psp());
405 seq_printf(m
, "cmo_page_size=%lu\n", cmo_get_page_size());
408 static void splpar_dispatch_data(struct seq_file
*m
)
411 unsigned long dispatches
= 0;
412 unsigned long dispatch_dispersions
= 0;
414 for_each_possible_cpu(cpu
) {
415 dispatches
+= be32_to_cpu(lppaca_of(cpu
).yield_count
);
416 dispatch_dispersions
+=
417 be32_to_cpu(lppaca_of(cpu
).dispersion_count
);
420 seq_printf(m
, "dispatches=%lu\n", dispatches
);
421 seq_printf(m
, "dispatch_dispersions=%lu\n", dispatch_dispersions
);
424 static void parse_em_data(struct seq_file
*m
)
426 unsigned long retbuf
[PLPAR_HCALL_BUFSIZE
];
428 if (firmware_has_feature(FW_FEATURE_LPAR
) &&
429 plpar_hcall(H_GET_EM_PARMS
, retbuf
) == H_SUCCESS
)
430 seq_printf(m
, "power_mode_data=%016lx\n", retbuf
[0]);
433 static int pseries_lparcfg_data(struct seq_file
*m
, void *v
)
435 int partition_potential_processors
;
436 int partition_active_processors
;
437 struct device_node
*rtas_node
;
438 const __be32
*lrdrp
= NULL
;
440 rtas_node
= of_find_node_by_path("/rtas");
442 lrdrp
= of_get_property(rtas_node
, "ibm,lrdr-capacity", NULL
);
445 partition_potential_processors
= vdso_data
->processorCount
;
447 partition_potential_processors
= be32_to_cpup(lrdrp
+ 4);
449 of_node_put(rtas_node
);
451 partition_active_processors
= lparcfg_count_active_processors();
453 if (firmware_has_feature(FW_FEATURE_SPLPAR
)) {
454 /* this call handles the ibm,get-system-parameter contents */
455 parse_system_parameter_string(m
);
460 splpar_dispatch_data(m
);
462 seq_printf(m
, "purr=%ld\n", get_purr());
463 } else { /* non SPLPAR case */
465 seq_printf(m
, "system_active_processors=%d\n",
466 partition_potential_processors
);
468 seq_printf(m
, "system_potential_processors=%d\n",
469 partition_potential_processors
);
471 seq_printf(m
, "partition_max_entitled_capacity=%d\n",
472 partition_potential_processors
* 100);
474 seq_printf(m
, "partition_entitled_capacity=%d\n",
475 partition_active_processors
* 100);
478 seq_printf(m
, "partition_active_processors=%d\n",
479 partition_active_processors
);
481 seq_printf(m
, "partition_potential_processors=%d\n",
482 partition_potential_processors
);
484 seq_printf(m
, "shared_processor_mode=%d\n",
485 lppaca_shared_proc(get_lppaca()));
487 seq_printf(m
, "slb_size=%d\n", mmu_slb_size
);
494 static ssize_t
update_ppp(u64
*entitlement
, u8
*weight
)
496 struct hvcall_ppp_data ppp_data
;
501 /* Get our current parameters */
502 retval
= h_get_ppp(&ppp_data
);
507 new_weight
= ppp_data
.weight
;
508 new_entitled
= *entitlement
;
510 new_weight
= *weight
;
511 new_entitled
= ppp_data
.entitlement
;
515 pr_debug("%s: current_entitled = %llu, current_weight = %u\n",
516 __func__
, ppp_data
.entitlement
, ppp_data
.weight
);
518 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
519 __func__
, new_entitled
, new_weight
);
521 retval
= plpar_hcall_norets(H_SET_PPP
, new_entitled
, new_weight
);
528 * Update the memory entitlement and weight for the partition. Caller must
529 * specify either a new entitlement or weight, not both, to be updated
530 * since the h_set_mpp call takes both entitlement and weight as parameters.
532 static ssize_t
update_mpp(u64
*entitlement
, u8
*weight
)
534 struct hvcall_mpp_data mpp_data
;
540 /* Check with vio to ensure the new memory entitlement
543 rc
= vio_cmo_entitlement_update(*entitlement
);
548 rc
= h_get_mpp(&mpp_data
);
553 new_weight
= mpp_data
.mem_weight
;
554 new_entitled
= *entitlement
;
556 new_weight
= *weight
;
557 new_entitled
= mpp_data
.entitled_mem
;
561 pr_debug("%s: current_entitled = %lu, current_weight = %u\n",
562 __func__
, mpp_data
.entitled_mem
, mpp_data
.mem_weight
);
564 pr_debug("%s: new_entitled = %llu, new_weight = %u\n",
565 __func__
, new_entitled
, new_weight
);
567 rc
= plpar_hcall_norets(H_SET_MPP
, new_entitled
, new_weight
);
572 * Interface for changing system parameters (variable capacity weight
573 * and entitled capacity). Format of input is "param_name=value";
574 * anything after value is ignored. Valid parameters at this time are
575 * "partition_entitled_capacity" and "capacity_weight". We use
576 * H_SET_PPP to alter parameters.
578 * This function should be invoked only on systems with
581 static ssize_t
lparcfg_write(struct file
*file
, const char __user
* buf
,
582 size_t count
, loff_t
* off
)
587 u64 new_entitled
, *new_entitled_ptr
= &new_entitled
;
588 u8 new_weight
, *new_weight_ptr
= &new_weight
;
591 if (!firmware_has_feature(FW_FEATURE_SPLPAR
))
597 if (copy_from_user(kbuf
, buf
, count
))
600 kbuf
[count
- 1] = '\0';
601 tmp
= strchr(kbuf
, '=');
607 if (!strcmp(kbuf
, "partition_entitled_capacity")) {
609 *new_entitled_ptr
= (u64
) simple_strtoul(tmp
, &endp
, 10);
613 retval
= update_ppp(new_entitled_ptr
, NULL
);
614 } else if (!strcmp(kbuf
, "capacity_weight")) {
616 *new_weight_ptr
= (u8
) simple_strtoul(tmp
, &endp
, 10);
620 retval
= update_ppp(NULL
, new_weight_ptr
);
621 } else if (!strcmp(kbuf
, "entitled_memory")) {
623 *new_entitled_ptr
= (u64
) simple_strtoul(tmp
, &endp
, 10);
627 retval
= update_mpp(new_entitled_ptr
, NULL
);
628 } else if (!strcmp(kbuf
, "entitled_memory_weight")) {
630 *new_weight_ptr
= (u8
) simple_strtoul(tmp
, &endp
, 10);
634 retval
= update_mpp(NULL
, new_weight_ptr
);
638 if (retval
== H_SUCCESS
|| retval
== H_CONSTRAINED
) {
640 } else if (retval
== H_BUSY
) {
642 } else if (retval
== H_HARDWARE
) {
644 } else if (retval
== H_PARAMETER
) {
651 static int lparcfg_data(struct seq_file
*m
, void *v
)
653 struct device_node
*rootdn
;
654 const char *model
= "";
655 const char *system_id
= "";
657 const __be32
*lp_index_ptr
;
658 unsigned int lp_index
= 0;
660 seq_printf(m
, "%s %s\n", MODULE_NAME
, MODULE_VERS
);
662 rootdn
= of_find_node_by_path("/");
664 tmp
= of_get_property(rootdn
, "model", NULL
);
667 tmp
= of_get_property(rootdn
, "system-id", NULL
);
670 lp_index_ptr
= of_get_property(rootdn
, "ibm,partition-no",
673 lp_index
= be32_to_cpup(lp_index_ptr
);
676 seq_printf(m
, "serial_number=%s\n", system_id
);
677 seq_printf(m
, "system_type=%s\n", model
);
678 seq_printf(m
, "partition_id=%d\n", (int)lp_index
);
680 return pseries_lparcfg_data(m
, v
);
683 static int lparcfg_open(struct inode
*inode
, struct file
*file
)
685 return single_open(file
, lparcfg_data
, NULL
);
688 static const struct file_operations lparcfg_fops
= {
690 .write
= lparcfg_write
,
691 .open
= lparcfg_open
,
692 .release
= single_release
,
696 static int __init
lparcfg_init(void)
698 umode_t mode
= S_IRUSR
| S_IRGRP
| S_IROTH
;
700 /* Allow writing if we have FW_FEATURE_SPLPAR */
701 if (firmware_has_feature(FW_FEATURE_SPLPAR
))
704 if (!proc_create("powerpc/lparcfg", mode
, NULL
, &lparcfg_fops
)) {
705 printk(KERN_ERR
"Failed to create powerpc/lparcfg\n");
710 machine_device_initcall(pseries
, lparcfg_init
);