1 // SPDX-License-Identifier: GPL-2.0-or-later
4 * Procedures for interfacing to the RTAS on CHRP machines.
6 * Peter Bergner, IBM March 2001.
7 * Copyright (C) 2001 IBM.
10 #define pr_fmt(fmt) "rtas: " fmt
12 #include <linux/bsearch.h>
13 #include <linux/capability.h>
14 #include <linux/delay.h>
15 #include <linux/export.h>
16 #include <linux/init.h>
17 #include <linux/kconfig.h>
18 #include <linux/kernel.h>
19 #include <linux/lockdep.h>
20 #include <linux/memblock.h>
22 #include <linux/of_fdt.h>
23 #include <linux/reboot.h>
24 #include <linux/sched.h>
25 #include <linux/security.h>
26 #include <linux/slab.h>
27 #include <linux/spinlock.h>
28 #include <linux/stdarg.h>
29 #include <linux/syscalls.h>
30 #include <linux/types.h>
31 #include <linux/uaccess.h>
32 #include <linux/xarray.h>
34 #include <asm/delay.h>
35 #include <asm/firmware.h>
36 #include <asm/interrupt.h>
37 #include <asm/machdep.h>
40 #include <asm/rtas-work-area.h>
43 #include <asm/trace.h>
47 /* Indexes into the args buffer, -1 if not used */
53 * Assumed buffer size per the spec if the function does not
54 * have a size parameter, e.g. ibm,errinjct. 0 if unused.
60 * struct rtas_function - Descriptor for RTAS functions.
62 * @token: Value of @name if it exists under the /rtas node.
63 * @name: Function name.
64 * @filter: If non-NULL, invoking this function via the rtas syscall is
65 * generally allowed, and @filter describes constraints on the
66 * arguments. See also @banned_for_syscall_on_le.
67 * @banned_for_syscall_on_le: Set when call via sys_rtas is generally allowed
68 * but specifically restricted on ppc64le. Such
69 * functions are believed to have no users on
70 * ppc64le, and we want to keep it that way. It does
71 * not make sense for this to be set when @filter
74 struct rtas_function {
76 const bool banned_for_syscall_on_le:1;
77 const char * const name;
78 const struct rtas_filter *filter;
81 static struct rtas_function rtas_function_table[] __ro_after_init = {
82 [RTAS_FNIDX__CHECK_EXCEPTION] = {
83 .name = "check-exception",
85 [RTAS_FNIDX__DISPLAY_CHARACTER] = {
86 .name = "display-character",
87 .filter = &(const struct rtas_filter) {
88 .buf_idx1 = -1, .size_idx1 = -1,
89 .buf_idx2 = -1, .size_idx2 = -1,
92 [RTAS_FNIDX__EVENT_SCAN] = {
95 [RTAS_FNIDX__FREEZE_TIME_BASE] = {
96 .name = "freeze-time-base",
98 [RTAS_FNIDX__GET_POWER_LEVEL] = {
99 .name = "get-power-level",
100 .filter = &(const struct rtas_filter) {
101 .buf_idx1 = -1, .size_idx1 = -1,
102 .buf_idx2 = -1, .size_idx2 = -1,
105 [RTAS_FNIDX__GET_SENSOR_STATE] = {
106 .name = "get-sensor-state",
107 .filter = &(const struct rtas_filter) {
108 .buf_idx1 = -1, .size_idx1 = -1,
109 .buf_idx2 = -1, .size_idx2 = -1,
112 [RTAS_FNIDX__GET_TERM_CHAR] = {
113 .name = "get-term-char",
115 [RTAS_FNIDX__GET_TIME_OF_DAY] = {
116 .name = "get-time-of-day",
117 .filter = &(const struct rtas_filter) {
118 .buf_idx1 = -1, .size_idx1 = -1,
119 .buf_idx2 = -1, .size_idx2 = -1,
122 [RTAS_FNIDX__IBM_ACTIVATE_FIRMWARE] = {
123 .name = "ibm,activate-firmware",
124 .filter = &(const struct rtas_filter) {
125 .buf_idx1 = -1, .size_idx1 = -1,
126 .buf_idx2 = -1, .size_idx2 = -1,
129 [RTAS_FNIDX__IBM_CBE_START_PTCAL] = {
130 .name = "ibm,cbe-start-ptcal",
132 [RTAS_FNIDX__IBM_CBE_STOP_PTCAL] = {
133 .name = "ibm,cbe-stop-ptcal",
135 [RTAS_FNIDX__IBM_CHANGE_MSI] = {
136 .name = "ibm,change-msi",
138 [RTAS_FNIDX__IBM_CLOSE_ERRINJCT] = {
139 .name = "ibm,close-errinjct",
140 .filter = &(const struct rtas_filter) {
141 .buf_idx1 = -1, .size_idx1 = -1,
142 .buf_idx2 = -1, .size_idx2 = -1,
145 [RTAS_FNIDX__IBM_CONFIGURE_BRIDGE] = {
146 .name = "ibm,configure-bridge",
148 [RTAS_FNIDX__IBM_CONFIGURE_CONNECTOR] = {
149 .name = "ibm,configure-connector",
150 .filter = &(const struct rtas_filter) {
151 .buf_idx1 = 0, .size_idx1 = -1,
152 .buf_idx2 = 1, .size_idx2 = -1,
156 [RTAS_FNIDX__IBM_CONFIGURE_KERNEL_DUMP] = {
157 .name = "ibm,configure-kernel-dump",
159 [RTAS_FNIDX__IBM_CONFIGURE_PE] = {
160 .name = "ibm,configure-pe",
162 [RTAS_FNIDX__IBM_CREATE_PE_DMA_WINDOW] = {
163 .name = "ibm,create-pe-dma-window",
165 [RTAS_FNIDX__IBM_DISPLAY_MESSAGE] = {
166 .name = "ibm,display-message",
167 .filter = &(const struct rtas_filter) {
168 .buf_idx1 = 0, .size_idx1 = -1,
169 .buf_idx2 = -1, .size_idx2 = -1,
172 [RTAS_FNIDX__IBM_ERRINJCT] = {
173 .name = "ibm,errinjct",
174 .filter = &(const struct rtas_filter) {
175 .buf_idx1 = 2, .size_idx1 = -1,
176 .buf_idx2 = -1, .size_idx2 = -1,
180 [RTAS_FNIDX__IBM_EXTI2C] = {
181 .name = "ibm,exti2c",
183 [RTAS_FNIDX__IBM_GET_CONFIG_ADDR_INFO] = {
184 .name = "ibm,get-config-addr-info",
186 [RTAS_FNIDX__IBM_GET_CONFIG_ADDR_INFO2] = {
187 .name = "ibm,get-config-addr-info2",
188 .filter = &(const struct rtas_filter) {
189 .buf_idx1 = -1, .size_idx1 = -1,
190 .buf_idx2 = -1, .size_idx2 = -1,
193 [RTAS_FNIDX__IBM_GET_DYNAMIC_SENSOR_STATE] = {
194 .name = "ibm,get-dynamic-sensor-state",
195 .filter = &(const struct rtas_filter) {
196 .buf_idx1 = 1, .size_idx1 = -1,
197 .buf_idx2 = -1, .size_idx2 = -1,
200 [RTAS_FNIDX__IBM_GET_INDICES] = {
201 .name = "ibm,get-indices",
202 .filter = &(const struct rtas_filter) {
203 .buf_idx1 = 2, .size_idx1 = 3,
204 .buf_idx2 = -1, .size_idx2 = -1,
207 [RTAS_FNIDX__IBM_GET_RIO_TOPOLOGY] = {
208 .name = "ibm,get-rio-topology",
210 [RTAS_FNIDX__IBM_GET_SYSTEM_PARAMETER] = {
211 .name = "ibm,get-system-parameter",
212 .filter = &(const struct rtas_filter) {
213 .buf_idx1 = 1, .size_idx1 = 2,
214 .buf_idx2 = -1, .size_idx2 = -1,
217 [RTAS_FNIDX__IBM_GET_VPD] = {
218 .name = "ibm,get-vpd",
219 .filter = &(const struct rtas_filter) {
220 .buf_idx1 = 0, .size_idx1 = -1,
221 .buf_idx2 = 1, .size_idx2 = 2,
224 [RTAS_FNIDX__IBM_GET_XIVE] = {
225 .name = "ibm,get-xive",
227 [RTAS_FNIDX__IBM_INT_OFF] = {
228 .name = "ibm,int-off",
230 [RTAS_FNIDX__IBM_INT_ON] = {
231 .name = "ibm,int-on",
233 [RTAS_FNIDX__IBM_IO_QUIESCE_ACK] = {
234 .name = "ibm,io-quiesce-ack",
236 [RTAS_FNIDX__IBM_LPAR_PERFTOOLS] = {
237 .name = "ibm,lpar-perftools",
238 .filter = &(const struct rtas_filter) {
239 .buf_idx1 = 2, .size_idx1 = 3,
240 .buf_idx2 = -1, .size_idx2 = -1,
243 [RTAS_FNIDX__IBM_MANAGE_FLASH_IMAGE] = {
244 .name = "ibm,manage-flash-image",
246 [RTAS_FNIDX__IBM_MANAGE_STORAGE_PRESERVATION] = {
247 .name = "ibm,manage-storage-preservation",
249 [RTAS_FNIDX__IBM_NMI_INTERLOCK] = {
250 .name = "ibm,nmi-interlock",
252 [RTAS_FNIDX__IBM_NMI_REGISTER] = {
253 .name = "ibm,nmi-register",
255 [RTAS_FNIDX__IBM_OPEN_ERRINJCT] = {
256 .name = "ibm,open-errinjct",
257 .filter = &(const struct rtas_filter) {
258 .buf_idx1 = -1, .size_idx1 = -1,
259 .buf_idx2 = -1, .size_idx2 = -1,
262 [RTAS_FNIDX__IBM_OPEN_SRIOV_ALLOW_UNFREEZE] = {
263 .name = "ibm,open-sriov-allow-unfreeze",
265 [RTAS_FNIDX__IBM_OPEN_SRIOV_MAP_PE_NUMBER] = {
266 .name = "ibm,open-sriov-map-pe-number",
268 [RTAS_FNIDX__IBM_OS_TERM] = {
269 .name = "ibm,os-term",
271 [RTAS_FNIDX__IBM_PARTNER_CONTROL] = {
272 .name = "ibm,partner-control",
274 [RTAS_FNIDX__IBM_PHYSICAL_ATTESTATION] = {
275 .name = "ibm,physical-attestation",
276 .filter = &(const struct rtas_filter) {
277 .buf_idx1 = 0, .size_idx1 = 1,
278 .buf_idx2 = -1, .size_idx2 = -1,
281 [RTAS_FNIDX__IBM_PLATFORM_DUMP] = {
282 .name = "ibm,platform-dump",
283 .filter = &(const struct rtas_filter) {
284 .buf_idx1 = 4, .size_idx1 = 5,
285 .buf_idx2 = -1, .size_idx2 = -1,
288 [RTAS_FNIDX__IBM_POWER_OFF_UPS] = {
289 .name = "ibm,power-off-ups",
291 [RTAS_FNIDX__IBM_QUERY_INTERRUPT_SOURCE_NUMBER] = {
292 .name = "ibm,query-interrupt-source-number",
294 [RTAS_FNIDX__IBM_QUERY_PE_DMA_WINDOW] = {
295 .name = "ibm,query-pe-dma-window",
297 [RTAS_FNIDX__IBM_READ_PCI_CONFIG] = {
298 .name = "ibm,read-pci-config",
300 [RTAS_FNIDX__IBM_READ_SLOT_RESET_STATE] = {
301 .name = "ibm,read-slot-reset-state",
302 .filter = &(const struct rtas_filter) {
303 .buf_idx1 = -1, .size_idx1 = -1,
304 .buf_idx2 = -1, .size_idx2 = -1,
307 [RTAS_FNIDX__IBM_READ_SLOT_RESET_STATE2] = {
308 .name = "ibm,read-slot-reset-state2",
310 [RTAS_FNIDX__IBM_REMOVE_PE_DMA_WINDOW] = {
311 .name = "ibm,remove-pe-dma-window",
313 [RTAS_FNIDX__IBM_RESET_PE_DMA_WINDOWS] = {
314 .name = "ibm,reset-pe-dma-windows",
316 [RTAS_FNIDX__IBM_SCAN_LOG_DUMP] = {
317 .name = "ibm,scan-log-dump",
318 .filter = &(const struct rtas_filter) {
319 .buf_idx1 = 0, .size_idx1 = 1,
320 .buf_idx2 = -1, .size_idx2 = -1,
323 [RTAS_FNIDX__IBM_SET_DYNAMIC_INDICATOR] = {
324 .name = "ibm,set-dynamic-indicator",
325 .filter = &(const struct rtas_filter) {
326 .buf_idx1 = 2, .size_idx1 = -1,
327 .buf_idx2 = -1, .size_idx2 = -1,
330 [RTAS_FNIDX__IBM_SET_EEH_OPTION] = {
331 .name = "ibm,set-eeh-option",
332 .filter = &(const struct rtas_filter) {
333 .buf_idx1 = -1, .size_idx1 = -1,
334 .buf_idx2 = -1, .size_idx2 = -1,
337 [RTAS_FNIDX__IBM_SET_SLOT_RESET] = {
338 .name = "ibm,set-slot-reset",
340 [RTAS_FNIDX__IBM_SET_SYSTEM_PARAMETER] = {
341 .name = "ibm,set-system-parameter",
342 .filter = &(const struct rtas_filter) {
343 .buf_idx1 = 1, .size_idx1 = -1,
344 .buf_idx2 = -1, .size_idx2 = -1,
347 [RTAS_FNIDX__IBM_SET_XIVE] = {
348 .name = "ibm,set-xive",
350 [RTAS_FNIDX__IBM_SLOT_ERROR_DETAIL] = {
351 .name = "ibm,slot-error-detail",
353 [RTAS_FNIDX__IBM_SUSPEND_ME] = {
354 .name = "ibm,suspend-me",
355 .banned_for_syscall_on_le = true,
356 .filter = &(const struct rtas_filter) {
357 .buf_idx1 = -1, .size_idx1 = -1,
358 .buf_idx2 = -1, .size_idx2 = -1,
361 [RTAS_FNIDX__IBM_TUNE_DMA_PARMS] = {
362 .name = "ibm,tune-dma-parms",
364 [RTAS_FNIDX__IBM_UPDATE_FLASH_64_AND_REBOOT] = {
365 .name = "ibm,update-flash-64-and-reboot",
367 [RTAS_FNIDX__IBM_UPDATE_NODES] = {
368 .name = "ibm,update-nodes",
369 .banned_for_syscall_on_le = true,
370 .filter = &(const struct rtas_filter) {
371 .buf_idx1 = 0, .size_idx1 = -1,
372 .buf_idx2 = -1, .size_idx2 = -1,
376 [RTAS_FNIDX__IBM_UPDATE_PROPERTIES] = {
377 .name = "ibm,update-properties",
378 .banned_for_syscall_on_le = true,
379 .filter = &(const struct rtas_filter) {
380 .buf_idx1 = 0, .size_idx1 = -1,
381 .buf_idx2 = -1, .size_idx2 = -1,
385 [RTAS_FNIDX__IBM_VALIDATE_FLASH_IMAGE] = {
386 .name = "ibm,validate-flash-image",
388 [RTAS_FNIDX__IBM_WRITE_PCI_CONFIG] = {
389 .name = "ibm,write-pci-config",
391 [RTAS_FNIDX__NVRAM_FETCH] = {
392 .name = "nvram-fetch",
394 [RTAS_FNIDX__NVRAM_STORE] = {
395 .name = "nvram-store",
397 [RTAS_FNIDX__POWER_OFF] = {
400 [RTAS_FNIDX__PUT_TERM_CHAR] = {
401 .name = "put-term-char",
403 [RTAS_FNIDX__QUERY_CPU_STOPPED_STATE] = {
404 .name = "query-cpu-stopped-state",
406 [RTAS_FNIDX__READ_PCI_CONFIG] = {
407 .name = "read-pci-config",
409 [RTAS_FNIDX__RTAS_LAST_ERROR] = {
410 .name = "rtas-last-error",
412 [RTAS_FNIDX__SET_INDICATOR] = {
413 .name = "set-indicator",
414 .filter = &(const struct rtas_filter) {
415 .buf_idx1 = -1, .size_idx1 = -1,
416 .buf_idx2 = -1, .size_idx2 = -1,
419 [RTAS_FNIDX__SET_POWER_LEVEL] = {
420 .name = "set-power-level",
421 .filter = &(const struct rtas_filter) {
422 .buf_idx1 = -1, .size_idx1 = -1,
423 .buf_idx2 = -1, .size_idx2 = -1,
426 [RTAS_FNIDX__SET_TIME_FOR_POWER_ON] = {
427 .name = "set-time-for-power-on",
428 .filter = &(const struct rtas_filter) {
429 .buf_idx1 = -1, .size_idx1 = -1,
430 .buf_idx2 = -1, .size_idx2 = -1,
433 [RTAS_FNIDX__SET_TIME_OF_DAY] = {
434 .name = "set-time-of-day",
435 .filter = &(const struct rtas_filter) {
436 .buf_idx1 = -1, .size_idx1 = -1,
437 .buf_idx2 = -1, .size_idx2 = -1,
440 [RTAS_FNIDX__START_CPU] = {
443 [RTAS_FNIDX__STOP_SELF] = {
446 [RTAS_FNIDX__SYSTEM_REBOOT] = {
447 .name = "system-reboot",
449 [RTAS_FNIDX__THAW_TIME_BASE] = {
450 .name = "thaw-time-base",
452 [RTAS_FNIDX__WRITE_PCI_CONFIG] = {
453 .name = "write-pci-config",
458 * Nearly all RTAS calls need to be serialized. All uses of the
459 * default rtas_args block must hold rtas_lock.
461 * Exceptions to the RTAS serialization requirement (e.g. stop-self)
462 * must use a separate rtas_args structure.
464 static DEFINE_RAW_SPINLOCK(rtas_lock);
465 static struct rtas_args rtas_args;
468 * rtas_function_token() - RTAS function token lookup.
469 * @handle: Function handle, e.g. RTAS_FN_EVENT_SCAN.
471 * Context: Any context.
472 * Return: the token value for the function if implemented by this platform,
473 * otherwise RTAS_UNKNOWN_SERVICE.
475 s32 rtas_function_token(const rtas_fn_handle_t handle)
477 const size_t index = handle.index;
478 const bool out_of_bounds = index >= ARRAY_SIZE(rtas_function_table);
480 if (WARN_ONCE(out_of_bounds, "invalid function index %zu", index))
481 return RTAS_UNKNOWN_SERVICE;
483 * Various drivers attempt token lookups on non-RTAS
487 return RTAS_UNKNOWN_SERVICE;
489 return rtas_function_table[index].token;
491 EXPORT_SYMBOL_GPL(rtas_function_token);
493 static int rtas_function_cmp(const void *a, const void *b)
495 const struct rtas_function *f1 = a;
496 const struct rtas_function *f2 = b;
498 return strcmp(f1->name, f2->name);
502 * Boot-time initialization of the function table needs the lookup to
503 * return a non-const-qualified object. Use rtas_name_to_function()
504 * in all other contexts.
506 static struct rtas_function *__rtas_name_to_function(const char *name)
508 const struct rtas_function key = {
511 struct rtas_function *found;
513 found = bsearch(&key, rtas_function_table, ARRAY_SIZE(rtas_function_table),
514 sizeof(rtas_function_table[0]), rtas_function_cmp);
519 static const struct rtas_function *rtas_name_to_function(const char *name)
521 return __rtas_name_to_function(name);
524 static DEFINE_XARRAY(rtas_token_to_function_xarray);
526 static int __init rtas_token_to_function_xarray_init(void)
530 for (size_t i = 0; i < ARRAY_SIZE(rtas_function_table); ++i) {
531 const struct rtas_function *func = &rtas_function_table[i];
532 const s32 token = func->token;
534 if (token == RTAS_UNKNOWN_SERVICE)
537 err = xa_err(xa_store(&rtas_token_to_function_xarray,
538 token, (void *)func, GFP_KERNEL));
545 arch_initcall(rtas_token_to_function_xarray_init);
547 static const struct rtas_function *rtas_token_to_function(s32 token)
549 const struct rtas_function *func;
551 if (WARN_ONCE(token < 0, "invalid token %d", token))
554 func = xa_load(&rtas_token_to_function_xarray, token);
556 if (WARN_ONCE(!func, "unexpected failed lookup for token %d", token))
562 /* This is here deliberately so it's only used in this file */
563 void enter_rtas(unsigned long);
565 static void __do_enter_rtas(struct rtas_args *args)
567 enter_rtas(__pa(args));
568 srr_regs_clobbered(); /* rtas uses SRRs, invalidate */
571 static void __do_enter_rtas_trace(struct rtas_args *args)
573 const char *name = NULL;
575 if (args == &rtas_args)
576 lockdep_assert_held(&rtas_lock);
578 * If the tracepoints that consume the function name aren't
579 * active, avoid the lookup.
581 if ((trace_rtas_input_enabled() || trace_rtas_output_enabled())) {
582 const s32 token = be32_to_cpu(args->token);
583 const struct rtas_function *func = rtas_token_to_function(token);
588 trace_rtas_input(args, name);
589 trace_rtas_ll_entry(args);
591 __do_enter_rtas(args);
593 trace_rtas_ll_exit(args);
594 trace_rtas_output(args, name);
597 static void do_enter_rtas(struct rtas_args *args)
599 const unsigned long msr = mfmsr();
601 * Situations where we want to skip any active tracepoints for
604 * 1. The last code executed on an offline CPU as it stops,
605 * i.e. we're about to call stop-self. The tracepoints'
606 * function name lookup uses xarray, which uses RCU, which
607 * isn't valid to call on an offline CPU. Any events
608 * emitted on an offline CPU will be discarded anyway.
610 * 2. In real mode, as when invoking ibm,nmi-interlock from
611 * the pseries MCE handler. We cannot count on trace
612 * buffers or the entries in rtas_token_to_function_xarray
613 * to be contained in the RMO.
615 const unsigned long mask = MSR_IR | MSR_DR;
616 const bool can_trace = likely(cpu_online(raw_smp_processor_id()) &&
617 (msr & mask) == mask);
619 * Make sure MSR[RI] is currently enabled as it will be forced later
622 BUG_ON(!(msr & MSR_RI));
624 BUG_ON(!irqs_disabled());
626 hard_irq_disable(); /* Ensure MSR[EE] is disabled on PPC64 */
629 __do_enter_rtas_trace(args);
631 __do_enter_rtas(args);
636 DEFINE_SPINLOCK(rtas_data_buf_lock);
637 EXPORT_SYMBOL_GPL(rtas_data_buf_lock);
639 char rtas_data_buf[RTAS_DATA_BUF_SIZE] __aligned(SZ_4K);
640 EXPORT_SYMBOL_GPL(rtas_data_buf);
642 unsigned long rtas_rmo_buf;
645 * If non-NULL, this gets called when the kernel terminates.
646 * This is done like this so rtas_flash can be a module.
648 void (*rtas_flash_term_hook)(int);
649 EXPORT_SYMBOL_GPL(rtas_flash_term_hook);
652 * call_rtas_display_status and call_rtas_display_status_delay
653 * are designed only for very early low-level debugging, which
654 * is why the token is hard-coded to 10.
656 static void call_rtas_display_status(unsigned char c)
663 raw_spin_lock_irqsave(&rtas_lock, flags);
664 rtas_call_unlocked(&rtas_args, 10, 1, 1, NULL, c);
665 raw_spin_unlock_irqrestore(&rtas_lock, flags);
668 static void call_rtas_display_status_delay(char c)
670 static int pending_newline = 0; /* did last write end with unprinted newline? */
671 static int width = 16;
675 call_rtas_display_status(' ');
680 if (pending_newline) {
681 call_rtas_display_status('\r');
682 call_rtas_display_status('\n');
686 call_rtas_display_status(c);
692 void __init udbg_init_rtas_panel(void)
694 udbg_putc = call_rtas_display_status_delay;
697 #ifdef CONFIG_UDBG_RTAS_CONSOLE
699 /* If you think you're dying before early_init_dt_scan_rtas() does its
700 * work, you can hard code the token values for your firmware here and
701 * hardcode rtas.base/entry etc.
703 static unsigned int rtas_putchar_token = RTAS_UNKNOWN_SERVICE;
704 static unsigned int rtas_getchar_token = RTAS_UNKNOWN_SERVICE;
706 static void udbg_rtascon_putc(char c)
713 /* Add CRs before LFs */
715 udbg_rtascon_putc('\r');
717 /* if there is more than one character to be displayed, wait a bit */
718 for (tries = 0; tries < 16; tries++) {
719 if (rtas_call(rtas_putchar_token, 1, 1, NULL, c) == 0)
725 static int udbg_rtascon_getc_poll(void)
732 if (rtas_call(rtas_getchar_token, 0, 2, &c))
738 static int udbg_rtascon_getc(void)
742 while ((c = udbg_rtascon_getc_poll()) == -1)
749 void __init udbg_init_rtas_console(void)
751 udbg_putc = udbg_rtascon_putc;
752 udbg_getc = udbg_rtascon_getc;
753 udbg_getc_poll = udbg_rtascon_getc_poll;
755 #endif /* CONFIG_UDBG_RTAS_CONSOLE */
757 void rtas_progress(char *s, unsigned short hex)
759 struct device_node *root;
763 static int display_character, set_indicator;
764 static int display_width, display_lines, form_feed;
765 static const int *row_width;
766 static DEFINE_SPINLOCK(progress_lock);
767 static int current_line;
768 static int pending_newline = 0; /* did last write end with unprinted newline? */
773 if (display_width == 0) {
774 display_width = 0x10;
775 if ((root = of_find_node_by_path("/rtas"))) {
776 if ((p = of_get_property(root,
777 "ibm,display-line-length", NULL)))
778 display_width = be32_to_cpu(*p);
779 if ((p = of_get_property(root,
780 "ibm,form-feed", NULL)))
781 form_feed = be32_to_cpu(*p);
782 if ((p = of_get_property(root,
783 "ibm,display-number-of-lines", NULL)))
784 display_lines = be32_to_cpu(*p);
785 row_width = of_get_property(root,
786 "ibm,display-truncation-length", NULL);
789 display_character = rtas_function_token(RTAS_FN_DISPLAY_CHARACTER);
790 set_indicator = rtas_function_token(RTAS_FN_SET_INDICATOR);
793 if (display_character == RTAS_UNKNOWN_SERVICE) {
794 /* use hex display if available */
795 if (set_indicator != RTAS_UNKNOWN_SERVICE)
796 rtas_call(set_indicator, 3, 1, NULL, 6, 0, hex);
800 spin_lock(&progress_lock);
803 * Last write ended with newline, but we didn't print it since
804 * it would just clear the bottom line of output. Print it now
807 * If no newline is pending and form feed is supported, clear the
808 * display with a form feed; otherwise, print a CR to start output
809 * at the beginning of the line.
811 if (pending_newline) {
812 rtas_call(display_character, 1, 1, NULL, '\r');
813 rtas_call(display_character, 1, 1, NULL, '\n');
818 rtas_call(display_character, 1, 1, NULL,
821 rtas_call(display_character, 1, 1, NULL, '\r');
825 width = row_width[current_line];
827 width = display_width;
830 if (*os == '\n' || *os == '\r') {
831 /* If newline is the last character, save it
832 * until next call to avoid bumping up the
835 if (*os == '\n' && !os[1]) {
838 if (current_line > display_lines-1)
839 current_line = display_lines-1;
840 spin_unlock(&progress_lock);
844 /* RTAS wants CR-LF, not just LF */
847 rtas_call(display_character, 1, 1, NULL, '\r');
848 rtas_call(display_character, 1, 1, NULL, '\n');
850 /* CR might be used to re-draw a line, so we'll
851 * leave it alone and not add LF.
853 rtas_call(display_character, 1, 1, NULL, *os);
857 width = row_width[current_line];
859 width = display_width;
862 rtas_call(display_character, 1, 1, NULL, *os);
867 /* if we overwrite the screen length */
869 while ((*os != 0) && (*os != '\n') && (*os != '\r'))
873 spin_unlock(&progress_lock);
875 EXPORT_SYMBOL_GPL(rtas_progress); /* needed by rtas_flash module */
877 int rtas_token(const char *service)
879 const struct rtas_function *func;
882 if (rtas.dev == NULL)
883 return RTAS_UNKNOWN_SERVICE;
885 func = rtas_name_to_function(service);
889 * The caller is looking up a name that is not known to be an
890 * RTAS function. Either it's a function that needs to be
891 * added to the table, or they're misusing rtas_token() to
892 * access non-function properties of the /rtas node. Warn and
893 * fall back to the legacy behavior.
895 WARN_ONCE(1, "unknown function `%s`, should it be added to rtas_function_table?\n",
898 tokp = of_get_property(rtas.dev, service, NULL);
899 return tokp ? be32_to_cpu(*tokp) : RTAS_UNKNOWN_SERVICE;
901 EXPORT_SYMBOL_GPL(rtas_token);
903 int rtas_service_present(const char *service)
905 return rtas_token(service) != RTAS_UNKNOWN_SERVICE;
908 #ifdef CONFIG_RTAS_ERROR_LOGGING
910 static u32 rtas_error_log_max __ro_after_init = RTAS_ERROR_LOG_MAX;
913 * Return the firmware-specified size of the error log buffer
914 * for all rtas calls that require an error buffer argument.
915 * This includes 'check-exception' and 'rtas-last-error'.
917 int rtas_get_error_log_max(void)
919 return rtas_error_log_max;
922 static void __init init_error_log_max(void)
924 static const char propname[] __initconst = "rtas-error-log-max";
927 if (of_property_read_u32(rtas.dev, propname, &max)) {
928 pr_warn("%s not found, using default of %u\n",
929 propname, RTAS_ERROR_LOG_MAX);
930 max = RTAS_ERROR_LOG_MAX;
933 if (max > RTAS_ERROR_LOG_MAX) {
934 pr_warn("%s = %u, clamping max error log size to %u\n",
935 propname, max, RTAS_ERROR_LOG_MAX);
936 max = RTAS_ERROR_LOG_MAX;
939 rtas_error_log_max = max;
943 static char rtas_err_buf[RTAS_ERROR_LOG_MAX];
945 /** Return a copy of the detailed error text associated with the
946 * most recent failed call to rtas. Because the error text
947 * might go stale if there are any other intervening rtas calls,
948 * this routine must be called atomically with whatever produced
949 * the error (i.e. with rtas_lock still held from the previous call).
951 static char *__fetch_rtas_last_error(char *altbuf)
953 const s32 token = rtas_function_token(RTAS_FN_RTAS_LAST_ERROR);
954 struct rtas_args err_args, save_args;
958 lockdep_assert_held(&rtas_lock);
963 bufsz = rtas_get_error_log_max();
965 err_args.token = cpu_to_be32(token);
966 err_args.nargs = cpu_to_be32(2);
967 err_args.nret = cpu_to_be32(1);
968 err_args.args[0] = cpu_to_be32(__pa(rtas_err_buf));
969 err_args.args[1] = cpu_to_be32(bufsz);
970 err_args.args[2] = 0;
972 save_args = rtas_args;
973 rtas_args = err_args;
975 do_enter_rtas(&rtas_args);
977 err_args = rtas_args;
978 rtas_args = save_args;
980 /* Log the error in the unlikely case that there was one. */
981 if (unlikely(err_args.args[2] == 0)) {
986 if (slab_is_available())
987 buf = kmalloc(RTAS_ERROR_LOG_MAX, GFP_ATOMIC);
990 memmove(buf, rtas_err_buf, RTAS_ERROR_LOG_MAX);
996 #define get_errorlog_buffer() kmalloc(RTAS_ERROR_LOG_MAX, GFP_KERNEL)
998 #else /* CONFIG_RTAS_ERROR_LOGGING */
999 #define __fetch_rtas_last_error(x) NULL
1000 #define get_errorlog_buffer() NULL
1001 static void __init init_error_log_max(void) {}
1006 va_rtas_call_unlocked(struct rtas_args *args, int token, int nargs, int nret,
1011 args->token = cpu_to_be32(token);
1012 args->nargs = cpu_to_be32(nargs);
1013 args->nret = cpu_to_be32(nret);
1014 args->rets = &(args->args[nargs]);
1016 for (i = 0; i < nargs; ++i)
1017 args->args[i] = cpu_to_be32(va_arg(list, __u32));
1019 for (i = 0; i < nret; ++i)
1022 do_enter_rtas(args);
1026 * rtas_call_unlocked() - Invoke an RTAS firmware function without synchronization.
1027 * @args: RTAS parameter block to be used for the call, must obey RTAS addressing
1029 * @token: Identifies the function being invoked.
1030 * @nargs: Number of input parameters. Does not include token.
1031 * @nret: Number of output parameters, including the call status.
1032 * @....: List of @nargs input parameters.
1034 * Invokes the RTAS function indicated by @token, which the caller
1035 * should obtain via rtas_function_token().
1037 * This function is similar to rtas_call(), but must be used with a
1038 * limited set of RTAS calls specifically exempted from the general
1039 * requirement that only one RTAS call may be in progress at any
1040 * time. Examples include stop-self and ibm,nmi-interlock.
1042 void rtas_call_unlocked(struct rtas_args *args, int token, int nargs, int nret, ...)
1046 va_start(list, nret);
1047 va_rtas_call_unlocked(args, token, nargs, nret, list);
1051 static bool token_is_restricted_errinjct(s32 token)
1053 return token == rtas_function_token(RTAS_FN_IBM_OPEN_ERRINJCT) ||
1054 token == rtas_function_token(RTAS_FN_IBM_ERRINJCT);
1058 * rtas_call() - Invoke an RTAS firmware function.
1059 * @token: Identifies the function being invoked.
1060 * @nargs: Number of input parameters. Does not include token.
1061 * @nret: Number of output parameters, including the call status.
1062 * @outputs: Array of @nret output words.
1063 * @....: List of @nargs input parameters.
1065 * Invokes the RTAS function indicated by @token, which the caller
1066 * should obtain via rtas_function_token().
1068 * The @nargs and @nret arguments must match the number of input and
1069 * output parameters specified for the RTAS function.
1071 * rtas_call() returns RTAS status codes, not conventional Linux errno
1072 * values. Callers must translate any failure to an appropriate errno
1073 * in syscall context. Most callers of RTAS functions that can return
1074 * -2 or 990x should use rtas_busy_delay() to correctly handle those
1075 * statuses before calling again.
1077 * The return value descriptions are adapted from 7.2.8 [RTAS] Return
1078 * Codes of the PAPR and CHRP specifications.
1080 * Context: Process context preferably, interrupt context if
1081 * necessary. Acquires an internal spinlock and may perform
1082 * GFP_ATOMIC slab allocation in error path. Unsafe for NMI
1085 * * 0 - RTAS function call succeeded.
1086 * * -1 - RTAS function encountered a hardware or
1087 * platform error, or the token is invalid,
1088 * or the function is restricted by kernel policy.
1089 * * -2 - Specs say "A necessary hardware device was busy,
1090 * and the requested function could not be
1091 * performed. The operation should be retried at
1092 * a later time." This is misleading, at least with
1093 * respect to current RTAS implementations. What it
1094 * usually means in practice is that the function
1095 * could not be completed while meeting RTAS's
1096 * deadline for returning control to the OS (250us
1097 * for PAPR/PowerVM, typically), but the call may be
1098 * immediately reattempted to resume work on it.
1099 * * -3 - Parameter error.
1100 * * -7 - Unexpected state change.
1101 * * 9000...9899 - Vendor-specific success codes.
1102 * * 9900...9905 - Advisory extended delay. Caller should try
1103 * again after ~10^x ms has elapsed, where x is
1104 * the last digit of the status [0-5]. Again going
1105 * beyond the PAPR text, 990x on PowerVM indicates
1106 * contention for RTAS-internal resources. Other
1107 * RTAS call sequences in progress should be
1108 * allowed to complete before reattempting the
1110 * * -9000 - Multi-level isolation error.
1111 * * -9999...-9004 - Vendor-specific error codes.
1112 * * Additional negative values - Function-specific error.
1113 * * Additional positive values - Function-specific success.
1115 int rtas_call(int token, int nargs, int nret, int *outputs, ...)
1117 struct pin_cookie cookie;
1120 unsigned long flags;
1121 struct rtas_args *args;
1122 char *buff_copy = NULL;
1125 if (!rtas.entry || token == RTAS_UNKNOWN_SERVICE)
1128 if (token_is_restricted_errinjct(token)) {
1130 * It would be nicer to not discard the error value
1131 * from security_locked_down(), but callers expect an
1132 * RTAS status, not an errno.
1134 if (security_locked_down(LOCKDOWN_RTAS_ERROR_INJECTION))
1138 if ((mfmsr() & (MSR_IR|MSR_DR)) != (MSR_IR|MSR_DR)) {
1143 raw_spin_lock_irqsave(&rtas_lock, flags);
1144 cookie = lockdep_pin_lock(&rtas_lock);
1146 /* We use the global rtas args buffer */
1149 va_start(list, outputs);
1150 va_rtas_call_unlocked(args, token, nargs, nret, list);
1153 /* A -1 return code indicates that the last command couldn't
1154 be completed due to a hardware error. */
1155 if (be32_to_cpu(args->rets[0]) == -1)
1156 buff_copy = __fetch_rtas_last_error(NULL);
1158 if (nret > 1 && outputs != NULL)
1159 for (i = 0; i < nret-1; ++i)
1160 outputs[i] = be32_to_cpu(args->rets[i + 1]);
1161 ret = (nret > 0) ? be32_to_cpu(args->rets[0]) : 0;
1163 lockdep_unpin_lock(&rtas_lock, cookie);
1164 raw_spin_unlock_irqrestore(&rtas_lock, flags);
1167 log_error(buff_copy, ERR_TYPE_RTAS_LOG, 0);
1168 if (slab_is_available())
1173 EXPORT_SYMBOL_GPL(rtas_call);
1176 * rtas_busy_delay_time() - From an RTAS status value, calculate the
1177 * suggested delay time in milliseconds.
1179 * @status: a value returned from rtas_call() or similar APIs which return
1180 * the status of a RTAS function call.
1182 * Context: Any context.
1185 * * 100000 - If @status is 9905.
1186 * * 10000 - If @status is 9904.
1187 * * 1000 - If @status is 9903.
1188 * * 100 - If @status is 9902.
1189 * * 10 - If @status is 9901.
1190 * * 1 - If @status is either 9900 or -2. This is "wrong" for -2, but
1191 * some callers depend on this behavior, and the worst outcome
1192 * is that they will delay for longer than necessary.
1193 * * 0 - If @status is not a busy or extended delay value.
1195 unsigned int rtas_busy_delay_time(int status)
1198 unsigned int ms = 0;
1200 if (status == RTAS_BUSY) {
1202 } else if (status >= RTAS_EXTENDED_DELAY_MIN &&
1203 status <= RTAS_EXTENDED_DELAY_MAX) {
1204 order = status - RTAS_EXTENDED_DELAY_MIN;
1205 for (ms = 1; order > 0; order--)
1213 * Early boot fallback for rtas_busy_delay().
1215 static bool __init rtas_busy_delay_early(int status)
1217 static size_t successive_ext_delays __initdata;
1221 case RTAS_EXTENDED_DELAY_MIN...RTAS_EXTENDED_DELAY_MAX:
1223 * In the unlikely case that we receive an extended
1224 * delay status in early boot, the OS is probably not
1225 * the cause, and there's nothing we can do to clear
1226 * the condition. Best we can do is delay for a bit
1227 * and hope it's transient. Lie to the caller if it
1228 * seems like we're stuck in a retry loop.
1232 successive_ext_delays += 1;
1233 if (successive_ext_delays > 1000) {
1234 pr_err("too many extended delays, giving up\n");
1237 successive_ext_delays = 0;
1242 successive_ext_delays = 0;
1246 successive_ext_delays = 0;
1254 * rtas_busy_delay() - helper for RTAS busy and extended delay statuses
1256 * @status: a value returned from rtas_call() or similar APIs which return
1257 * the status of a RTAS function call.
1259 * Context: Process context. May sleep or schedule.
1262 * * true - @status is RTAS_BUSY or an extended delay hint. The
1263 * caller may assume that the CPU has been yielded if necessary,
1264 * and that an appropriate delay for @status has elapsed.
1265 * Generally the caller should reattempt the RTAS call which
1268 * * false - @status is not @RTAS_BUSY nor an extended delay hint. The
1269 * caller is responsible for handling @status.
1271 bool __ref rtas_busy_delay(int status)
1277 * Can't do timed sleeps before timekeeping is up.
1279 if (system_state < SYSTEM_SCHEDULING)
1280 return rtas_busy_delay_early(status);
1283 case RTAS_EXTENDED_DELAY_MIN...RTAS_EXTENDED_DELAY_MAX:
1285 ms = rtas_busy_delay_time(status);
1287 * The extended delay hint can be as high as 100 seconds.
1288 * Surely any function returning such a status is either
1289 * buggy or isn't going to be significantly slowed by us
1290 * polling at 1HZ. Clamp the sleep time to one second.
1292 ms = clamp(ms, 1U, 1000U);
1294 * The delay hint is an order-of-magnitude suggestion, not
1295 * a minimum. It is fine, possibly even advantageous, for
1296 * us to pause for less time than hinted. For small values,
1297 * use usleep_range() to ensure we don't sleep much longer
1298 * than actually needed.
1300 * See Documentation/timers/timers-howto.rst for
1301 * explanation of the threshold used here. In effect we use
1302 * usleep_range() for 9900 and 9901, msleep() for
1306 usleep_range(ms * 100, ms * 1000);
1313 * We should call again immediately if there's no other
1321 * Not a busy or extended delay status; the caller should
1322 * handle @status itself. Ensure we warn on misuses in
1323 * atomic context regardless.
1331 EXPORT_SYMBOL_GPL(rtas_busy_delay);
1333 static int rtas_error_rc(int rtas_rc)
1338 case -1: /* Hardware Error */
1341 case -3: /* Bad indicator/domain/etc */
1344 case -9000: /* Isolation error */
1347 case -9001: /* Outstanding TCE/PTE */
1350 case -9002: /* No usable slot */
1354 pr_err("%s: unexpected error %d\n", __func__, rtas_rc);
1361 int rtas_get_power_level(int powerdomain, int *level)
1363 int token = rtas_function_token(RTAS_FN_GET_POWER_LEVEL);
1366 if (token == RTAS_UNKNOWN_SERVICE)
1369 while ((rc = rtas_call(token, 1, 2, level, powerdomain)) == RTAS_BUSY)
1373 return rtas_error_rc(rc);
1376 EXPORT_SYMBOL_GPL(rtas_get_power_level);
1378 int rtas_set_power_level(int powerdomain, int level, int *setlevel)
1380 int token = rtas_function_token(RTAS_FN_SET_POWER_LEVEL);
1383 if (token == RTAS_UNKNOWN_SERVICE)
1387 rc = rtas_call(token, 2, 2, setlevel, powerdomain, level);
1388 } while (rtas_busy_delay(rc));
1391 return rtas_error_rc(rc);
1394 EXPORT_SYMBOL_GPL(rtas_set_power_level);
1396 int rtas_get_sensor(int sensor, int index, int *state)
1398 int token = rtas_function_token(RTAS_FN_GET_SENSOR_STATE);
1401 if (token == RTAS_UNKNOWN_SERVICE)
1405 rc = rtas_call(token, 2, 2, state, sensor, index);
1406 } while (rtas_busy_delay(rc));
1409 return rtas_error_rc(rc);
1412 EXPORT_SYMBOL_GPL(rtas_get_sensor);
1414 int rtas_get_sensor_fast(int sensor, int index, int *state)
1416 int token = rtas_function_token(RTAS_FN_GET_SENSOR_STATE);
1419 if (token == RTAS_UNKNOWN_SERVICE)
1422 rc = rtas_call(token, 2, 2, state, sensor, index);
1423 WARN_ON(rc == RTAS_BUSY || (rc >= RTAS_EXTENDED_DELAY_MIN &&
1424 rc <= RTAS_EXTENDED_DELAY_MAX));
1427 return rtas_error_rc(rc);
1431 bool rtas_indicator_present(int token, int *maxindex)
1433 int proplen, count, i;
1434 const struct indicator_elem {
1439 indicators = of_get_property(rtas.dev, "rtas-indicators", &proplen);
1443 count = proplen / sizeof(struct indicator_elem);
1445 for (i = 0; i < count; i++) {
1446 if (__be32_to_cpu(indicators[i].token) != token)
1449 *maxindex = __be32_to_cpu(indicators[i].maxindex);
1456 int rtas_set_indicator(int indicator, int index, int new_value)
1458 int token = rtas_function_token(RTAS_FN_SET_INDICATOR);
1461 if (token == RTAS_UNKNOWN_SERVICE)
1465 rc = rtas_call(token, 3, 1, NULL, indicator, index, new_value);
1466 } while (rtas_busy_delay(rc));
1469 return rtas_error_rc(rc);
1472 EXPORT_SYMBOL_GPL(rtas_set_indicator);
1475 * Ignoring RTAS extended delay
1477 int rtas_set_indicator_fast(int indicator, int index, int new_value)
1479 int token = rtas_function_token(RTAS_FN_SET_INDICATOR);
1482 if (token == RTAS_UNKNOWN_SERVICE)
1485 rc = rtas_call(token, 3, 1, NULL, indicator, index, new_value);
1487 WARN_ON(rc == RTAS_BUSY || (rc >= RTAS_EXTENDED_DELAY_MIN &&
1488 rc <= RTAS_EXTENDED_DELAY_MAX));
1491 return rtas_error_rc(rc);
1497 * rtas_ibm_suspend_me() - Call ibm,suspend-me to suspend the LPAR.
1499 * @fw_status: RTAS call status will be placed here if not NULL.
1501 * rtas_ibm_suspend_me() should be called only on a CPU which has
1502 * received H_CONTINUE from the H_JOIN hcall. All other active CPUs
1503 * should be waiting to return from H_JOIN.
1505 * rtas_ibm_suspend_me() may suspend execution of the OS
1506 * indefinitely. Callers should take appropriate measures upon return, such as
1507 * resetting watchdog facilities.
1509 * Callers may choose to retry this call if @fw_status is
1510 * %RTAS_THREADS_ACTIVE.
1513 * 0 - The partition has resumed from suspend, possibly after
1514 * migration to a different host.
1515 * -ECANCELED - The operation was aborted.
1516 * -EAGAIN - There were other CPUs not in H_JOIN at the time of the call.
1517 * -EBUSY - Some other condition prevented the suspend from succeeding.
1518 * -EIO - Hardware/platform error.
1520 int rtas_ibm_suspend_me(int *fw_status)
1522 int token = rtas_function_token(RTAS_FN_IBM_SUSPEND_ME);
1526 fwrc = rtas_call(token, 0, 1, NULL);
1532 case RTAS_SUSPEND_ABORTED:
1535 case RTAS_THREADS_ACTIVE:
1538 case RTAS_NOT_SUSPENDABLE:
1539 case RTAS_OUTSTANDING_COPROC:
1554 void __noreturn rtas_restart(char *cmd)
1556 if (rtas_flash_term_hook)
1557 rtas_flash_term_hook(SYS_RESTART);
1558 pr_emerg("system-reboot returned %d\n",
1559 rtas_call(rtas_function_token(RTAS_FN_SYSTEM_REBOOT), 0, 1, NULL));
1563 void rtas_power_off(void)
1565 if (rtas_flash_term_hook)
1566 rtas_flash_term_hook(SYS_POWER_OFF);
1567 /* allow power on only with power button press */
1568 pr_emerg("power-off returned %d\n",
1569 rtas_call(rtas_function_token(RTAS_FN_POWER_OFF), 2, 1, NULL, -1, -1));
1573 void __noreturn rtas_halt(void)
1575 if (rtas_flash_term_hook)
1576 rtas_flash_term_hook(SYS_HALT);
1577 /* allow power on only with power button press */
1578 pr_emerg("power-off returned %d\n",
1579 rtas_call(rtas_function_token(RTAS_FN_POWER_OFF), 2, 1, NULL, -1, -1));
1583 /* Must be in the RMO region, so we place it here */
1584 static char rtas_os_term_buf[2048];
1585 static bool ibm_extended_os_term;
1587 void rtas_os_term(char *str)
1589 s32 token = rtas_function_token(RTAS_FN_IBM_OS_TERM);
1593 * Firmware with the ibm,extended-os-term property is guaranteed
1594 * to always return from an ibm,os-term call. Earlier versions without
1595 * this property may terminate the partition which we want to avoid
1596 * since it interferes with panic_timeout.
1599 if (token == RTAS_UNKNOWN_SERVICE || !ibm_extended_os_term)
1602 snprintf(rtas_os_term_buf, 2048, "OS panic: %s", str);
1605 * Keep calling as long as RTAS returns a "try again" status,
1606 * but don't use rtas_busy_delay(), which potentially
1610 status = rtas_call(token, 1, 1, NULL, __pa(rtas_os_term_buf));
1611 } while (rtas_busy_delay_time(status));
1614 pr_emerg("ibm,os-term call failed %d\n", status);
1618 * rtas_activate_firmware() - Activate a new version of firmware.
1620 * Context: This function may sleep.
1622 * Activate a new version of partition firmware. The OS must call this
1623 * after resuming from a partition hibernation or migration in order
1624 * to maintain the ability to perform live firmware updates. It's not
1625 * catastrophic for this method to be absent or to fail; just log the
1626 * condition in that case.
1628 void rtas_activate_firmware(void)
1630 int token = rtas_function_token(RTAS_FN_IBM_ACTIVATE_FIRMWARE);
1633 if (token == RTAS_UNKNOWN_SERVICE) {
1634 pr_notice("ibm,activate-firmware method unavailable\n");
1639 fwrc = rtas_call(token, 0, 1, NULL);
1640 } while (rtas_busy_delay(fwrc));
1643 pr_err("ibm,activate-firmware failed (%i)\n", fwrc);
1647 * get_pseries_errorlog() - Find a specific pseries error log in an RTAS
1648 * extended event log.
1649 * @log: RTAS error/event log
1650 * @section_id: two character section identifier
1652 * Return: A pointer to the specified errorlog or NULL if not found.
1654 noinstr struct pseries_errorlog *get_pseries_errorlog(struct rtas_error_log *log,
1655 uint16_t section_id)
1657 struct rtas_ext_event_log_v6 *ext_log =
1658 (struct rtas_ext_event_log_v6 *)log->buffer;
1659 struct pseries_errorlog *sect;
1660 unsigned char *p, *log_end;
1661 uint32_t ext_log_length = rtas_error_extended_log_length(log);
1662 uint8_t log_format = rtas_ext_event_log_format(ext_log);
1663 uint32_t company_id = rtas_ext_event_company_id(ext_log);
1665 /* Check that we understand the format */
1666 if (ext_log_length < sizeof(struct rtas_ext_event_log_v6) ||
1667 log_format != RTAS_V6EXT_LOG_FORMAT_EVENT_LOG ||
1668 company_id != RTAS_V6EXT_COMPANY_ID_IBM)
1671 log_end = log->buffer + ext_log_length;
1672 p = ext_log->vendor_log;
1674 while (p < log_end) {
1675 sect = (struct pseries_errorlog *)p;
1676 if (pseries_errorlog_id(sect) == section_id)
1678 p += pseries_errorlog_length(sect);
1685 * The sys_rtas syscall, as originally designed, allows root to pass
1686 * arbitrary physical addresses to RTAS calls. A number of RTAS calls
1687 * can be abused to write to arbitrary memory and do other things that
1688 * are potentially harmful to system integrity, and thus should only
1689 * be used inside the kernel and not exposed to userspace.
1691 * All known legitimate users of the sys_rtas syscall will only ever
1692 * pass addresses that fall within the RMO buffer, and use a known
1693 * subset of RTAS calls.
1695 * Accordingly, we filter RTAS requests to check that the call is
1696 * permitted, and that provided pointers fall within the RMO buffer.
1697 * If a function is allowed to be invoked via the syscall, then its
1698 * entry in the rtas_functions table points to a rtas_filter that
1699 * describes its constraints, with the indexes of the parameters which
1700 * are expected to contain addresses and sizes of buffers allocated
1701 * inside the RMO buffer.
1704 static bool in_rmo_buf(u32 base, u32 end)
1706 return base >= rtas_rmo_buf &&
1707 base < (rtas_rmo_buf + RTAS_USER_REGION_SIZE) &&
1709 end >= rtas_rmo_buf &&
1710 end < (rtas_rmo_buf + RTAS_USER_REGION_SIZE);
1713 static bool block_rtas_call(int token, int nargs,
1714 struct rtas_args *args)
1716 const struct rtas_function *func;
1717 const struct rtas_filter *f;
1718 const bool is_platform_dump = token == rtas_function_token(RTAS_FN_IBM_PLATFORM_DUMP);
1719 const bool is_config_conn = token == rtas_function_token(RTAS_FN_IBM_CONFIGURE_CONNECTOR);
1720 u32 base, size, end;
1723 * If this token doesn't correspond to a function the kernel
1724 * understands, you're not allowed to call it.
1726 func = rtas_token_to_function(token);
1730 * And only functions with filters attached are allowed.
1736 * And some functions aren't allowed on LE.
1738 if (IS_ENABLED(CONFIG_CPU_LITTLE_ENDIAN) && func->banned_for_syscall_on_le)
1741 if (f->buf_idx1 != -1) {
1742 base = be32_to_cpu(args->args[f->buf_idx1]);
1743 if (f->size_idx1 != -1)
1744 size = be32_to_cpu(args->args[f->size_idx1]);
1745 else if (f->fixed_size)
1746 size = f->fixed_size;
1750 end = base + size - 1;
1753 * Special case for ibm,platform-dump - NULL buffer
1754 * address is used to indicate end of dump processing
1756 if (is_platform_dump && base == 0)
1759 if (!in_rmo_buf(base, end))
1763 if (f->buf_idx2 != -1) {
1764 base = be32_to_cpu(args->args[f->buf_idx2]);
1765 if (f->size_idx2 != -1)
1766 size = be32_to_cpu(args->args[f->size_idx2]);
1767 else if (f->fixed_size)
1768 size = f->fixed_size;
1771 end = base + size - 1;
1774 * Special case for ibm,configure-connector where the
1777 if (is_config_conn && base == 0)
1780 if (!in_rmo_buf(base, end))
1786 pr_err_ratelimited("sys_rtas: RTAS call blocked - exploit attempt?\n");
1787 pr_err_ratelimited("sys_rtas: token=0x%x, nargs=%d (called by %s)\n",
1788 token, nargs, current->comm);
1792 /* We assume to be passed big endian arguments */
1793 SYSCALL_DEFINE1(rtas, struct rtas_args __user *, uargs)
1795 struct pin_cookie cookie;
1796 struct rtas_args args;
1797 unsigned long flags;
1798 char *buff_copy, *errbuf = NULL;
1799 int nargs, nret, token;
1801 if (!capable(CAP_SYS_ADMIN))
1807 if (copy_from_user(&args, uargs, 3 * sizeof(u32)) != 0)
1810 nargs = be32_to_cpu(args.nargs);
1811 nret = be32_to_cpu(args.nret);
1812 token = be32_to_cpu(args.token);
1814 if (nargs >= ARRAY_SIZE(args.args)
1815 || nret > ARRAY_SIZE(args.args)
1816 || nargs + nret > ARRAY_SIZE(args.args))
1820 if (copy_from_user(args.args, uargs->args,
1821 nargs * sizeof(rtas_arg_t)) != 0)
1824 if (token == RTAS_UNKNOWN_SERVICE)
1827 args.rets = &args.args[nargs];
1828 memset(args.rets, 0, nret * sizeof(rtas_arg_t));
1830 if (block_rtas_call(token, nargs, &args))
1833 if (token_is_restricted_errinjct(token)) {
1836 err = security_locked_down(LOCKDOWN_RTAS_ERROR_INJECTION);
1841 /* Need to handle ibm,suspend_me call specially */
1842 if (token == rtas_function_token(RTAS_FN_IBM_SUSPEND_ME)) {
1845 * rtas_ibm_suspend_me assumes the streamid handle is in cpu
1846 * endian, or at least the hcall within it requires it.
1849 u64 handle = ((u64)be32_to_cpu(args.args[0]) << 32)
1850 | be32_to_cpu(args.args[1]);
1851 rc = rtas_syscall_dispatch_ibm_suspend_me(handle);
1853 args.rets[0] = cpu_to_be32(RTAS_NOT_SUSPENDABLE);
1854 else if (rc == -EIO)
1855 args.rets[0] = cpu_to_be32(-1);
1861 buff_copy = get_errorlog_buffer();
1863 raw_spin_lock_irqsave(&rtas_lock, flags);
1864 cookie = lockdep_pin_lock(&rtas_lock);
1867 do_enter_rtas(&rtas_args);
1870 /* A -1 return code indicates that the last command couldn't
1871 be completed due to a hardware error. */
1872 if (be32_to_cpu(args.rets[0]) == -1)
1873 errbuf = __fetch_rtas_last_error(buff_copy);
1875 lockdep_unpin_lock(&rtas_lock, cookie);
1876 raw_spin_unlock_irqrestore(&rtas_lock, flags);
1880 log_error(errbuf, ERR_TYPE_RTAS_LOG, 0);
1885 /* Copy out args. */
1886 if (copy_to_user(uargs->args + nargs,
1888 nret * sizeof(rtas_arg_t)) != 0)
1894 static void __init rtas_function_table_init(void)
1896 struct property *prop;
1898 for (size_t i = 0; i < ARRAY_SIZE(rtas_function_table); ++i) {
1899 struct rtas_function *curr = &rtas_function_table[i];
1900 struct rtas_function *prior;
1903 curr->token = RTAS_UNKNOWN_SERVICE;
1908 * Ensure table is sorted correctly for binary search
1909 * on function names.
1911 prior = &rtas_function_table[i - 1];
1913 cmp = strcmp(prior->name, curr->name);
1918 pr_err("'%s' has duplicate function table entries\n",
1921 pr_err("function table unsorted: '%s' wrongly precedes '%s'\n",
1922 prior->name, curr->name);
1926 for_each_property_of_node(rtas.dev, prop) {
1927 struct rtas_function *func;
1929 if (prop->length != sizeof(u32))
1932 func = __rtas_name_to_function(prop->name);
1936 func->token = be32_to_cpup((__be32 *)prop->value);
1938 pr_debug("function %s has token %u\n", func->name, func->token);
1943 * Call early during boot, before mem init, to retrieve the RTAS
1944 * information from the device-tree and allocate the RMO buffer for userland
1947 void __init rtas_initialize(void)
1949 unsigned long rtas_region = RTAS_INSTANTIATE_MAX;
1950 u32 base, size, entry;
1951 int no_base, no_size, no_entry;
1953 /* Get RTAS dev node and fill up our "rtas" structure with infos
1956 rtas.dev = of_find_node_by_name(NULL, "rtas");
1960 no_base = of_property_read_u32(rtas.dev, "linux,rtas-base", &base);
1961 no_size = of_property_read_u32(rtas.dev, "rtas-size", &size);
1962 if (no_base || no_size) {
1963 of_node_put(rtas.dev);
1970 no_entry = of_property_read_u32(rtas.dev, "linux,rtas-entry", &entry);
1971 rtas.entry = no_entry ? rtas.base : entry;
1973 init_error_log_max();
1975 /* Must be called before any function token lookups */
1976 rtas_function_table_init();
1979 * Discover this now to avoid a device tree lookup in the
1982 ibm_extended_os_term = of_property_read_bool(rtas.dev, "ibm,extended-os-term");
1984 /* If RTAS was found, allocate the RMO buffer for it and look for
1985 * the stop-self token if any
1988 if (firmware_has_feature(FW_FEATURE_LPAR))
1989 rtas_region = min(ppc64_rma_size, RTAS_INSTANTIATE_MAX);
1991 rtas_rmo_buf = memblock_phys_alloc_range(RTAS_USER_REGION_SIZE, PAGE_SIZE,
1994 panic("ERROR: RTAS: Failed to allocate %lx bytes below %pa\n",
1995 PAGE_SIZE, &rtas_region);
1997 rtas_work_area_reserve_arena(rtas_region);
2000 int __init early_init_dt_scan_rtas(unsigned long node,
2001 const char *uname, int depth, void *data)
2003 const u32 *basep, *entryp, *sizep;
2005 if (depth != 1 || strcmp(uname, "rtas") != 0)
2008 basep = of_get_flat_dt_prop(node, "linux,rtas-base", NULL);
2009 entryp = of_get_flat_dt_prop(node, "linux,rtas-entry", NULL);
2010 sizep = of_get_flat_dt_prop(node, "rtas-size", NULL);
2013 /* need this feature to decide the crashkernel offset */
2014 if (of_get_flat_dt_prop(node, "ibm,hypertas-functions", NULL))
2015 powerpc_firmware_features |= FW_FEATURE_LPAR;
2018 if (basep && entryp && sizep) {
2020 rtas.entry = *entryp;
2024 #ifdef CONFIG_UDBG_RTAS_CONSOLE
2025 basep = of_get_flat_dt_prop(node, "put-term-char", NULL);
2027 rtas_putchar_token = *basep;
2029 basep = of_get_flat_dt_prop(node, "get-term-char", NULL);
2031 rtas_getchar_token = *basep;
2033 if (rtas_putchar_token != RTAS_UNKNOWN_SERVICE &&
2034 rtas_getchar_token != RTAS_UNKNOWN_SERVICE)
2035 udbg_init_rtas_console();
2043 static DEFINE_RAW_SPINLOCK(timebase_lock);
2044 static u64 timebase = 0;
2046 void rtas_give_timebase(void)
2048 unsigned long flags;
2050 raw_spin_lock_irqsave(&timebase_lock, flags);
2052 rtas_call(rtas_function_token(RTAS_FN_FREEZE_TIME_BASE), 0, 1, NULL);
2053 timebase = get_tb();
2054 raw_spin_unlock(&timebase_lock);
2058 rtas_call(rtas_function_token(RTAS_FN_THAW_TIME_BASE), 0, 1, NULL);
2059 local_irq_restore(flags);
2062 void rtas_take_timebase(void)
2066 raw_spin_lock(&timebase_lock);
2067 set_tb(timebase >> 32, timebase & 0xffffffff);
2069 raw_spin_unlock(&timebase_lock);