3 * Procedures for interfacing to the RTAS on CHRP machines.
5 * Peter Bergner, IBM March 2001.
6 * Copyright (C) 2001 IBM.
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
15 #include <linux/kernel.h>
16 #include <linux/types.h>
17 #include <linux/spinlock.h>
18 #include <linux/export.h>
19 #include <linux/init.h>
20 #include <linux/capability.h>
21 #include <linux/delay.h>
22 #include <linux/cpu.h>
23 #include <linux/smp.h>
24 #include <linux/completion.h>
25 #include <linux/cpumask.h>
26 #include <linux/memblock.h>
27 #include <linux/slab.h>
28 #include <linux/reboot.h>
32 #include <asm/hvcall.h>
33 #include <asm/machdep.h>
34 #include <asm/firmware.h>
36 #include <asm/param.h>
37 #include <asm/delay.h>
38 #include <linux/uaccess.h>
40 #include <asm/syscalls.h>
42 #include <linux/atomic.h>
45 #include <asm/topology.h>
47 /* This is here deliberately so it's only used in this file */
48 void enter_rtas(unsigned long);
50 struct rtas_t rtas = {
51 .lock = __ARCH_SPIN_LOCK_UNLOCKED
55 DEFINE_SPINLOCK(rtas_data_buf_lock);
56 EXPORT_SYMBOL(rtas_data_buf_lock);
58 char rtas_data_buf[RTAS_DATA_BUF_SIZE] __cacheline_aligned;
59 EXPORT_SYMBOL(rtas_data_buf);
61 unsigned long rtas_rmo_buf;
64 * If non-NULL, this gets called when the kernel terminates.
65 * This is done like this so rtas_flash can be a module.
67 void (*rtas_flash_term_hook)(int);
68 EXPORT_SYMBOL(rtas_flash_term_hook);
70 /* RTAS use home made raw locking instead of spin_lock_irqsave
71 * because those can be called from within really nasty contexts
72 * such as having the timebase stopped which would lockup with
73 * normal locks and spinlock debugging enabled
75 static unsigned long lock_rtas(void)
79 local_irq_save(flags);
81 arch_spin_lock_flags(&rtas.lock, flags);
85 static void unlock_rtas(unsigned long flags)
87 arch_spin_unlock(&rtas.lock);
88 local_irq_restore(flags);
93 * call_rtas_display_status and call_rtas_display_status_delay
94 * are designed only for very early low-level debugging, which
95 * is why the token is hard-coded to 10.
97 static void call_rtas_display_status(unsigned char c)
105 rtas_call_unlocked(&rtas.args, 10, 1, 1, NULL, c);
109 static void call_rtas_display_status_delay(char c)
111 static int pending_newline = 0; /* did last write end with unprinted newline? */
112 static int width = 16;
116 call_rtas_display_status(' ');
121 if (pending_newline) {
122 call_rtas_display_status('\r');
123 call_rtas_display_status('\n');
127 call_rtas_display_status(c);
133 void __init udbg_init_rtas_panel(void)
135 udbg_putc = call_rtas_display_status_delay;
138 #ifdef CONFIG_UDBG_RTAS_CONSOLE
140 /* If you think you're dying before early_init_dt_scan_rtas() does its
141 * work, you can hard code the token values for your firmware here and
142 * hardcode rtas.base/entry etc.
144 static unsigned int rtas_putchar_token = RTAS_UNKNOWN_SERVICE;
145 static unsigned int rtas_getchar_token = RTAS_UNKNOWN_SERVICE;
147 static void udbg_rtascon_putc(char c)
154 /* Add CRs before LFs */
156 udbg_rtascon_putc('\r');
158 /* if there is more than one character to be displayed, wait a bit */
159 for (tries = 0; tries < 16; tries++) {
160 if (rtas_call(rtas_putchar_token, 1, 1, NULL, c) == 0)
166 static int udbg_rtascon_getc_poll(void)
173 if (rtas_call(rtas_getchar_token, 0, 2, &c))
179 static int udbg_rtascon_getc(void)
183 while ((c = udbg_rtascon_getc_poll()) == -1)
190 void __init udbg_init_rtas_console(void)
192 udbg_putc = udbg_rtascon_putc;
193 udbg_getc = udbg_rtascon_getc;
194 udbg_getc_poll = udbg_rtascon_getc_poll;
196 #endif /* CONFIG_UDBG_RTAS_CONSOLE */
198 void rtas_progress(char *s, unsigned short hex)
200 struct device_node *root;
204 static int display_character, set_indicator;
205 static int display_width, display_lines, form_feed;
206 static const int *row_width;
207 static DEFINE_SPINLOCK(progress_lock);
208 static int current_line;
209 static int pending_newline = 0; /* did last write end with unprinted newline? */
214 if (display_width == 0) {
215 display_width = 0x10;
216 if ((root = of_find_node_by_path("/rtas"))) {
217 if ((p = of_get_property(root,
218 "ibm,display-line-length", NULL)))
219 display_width = be32_to_cpu(*p);
220 if ((p = of_get_property(root,
221 "ibm,form-feed", NULL)))
222 form_feed = be32_to_cpu(*p);
223 if ((p = of_get_property(root,
224 "ibm,display-number-of-lines", NULL)))
225 display_lines = be32_to_cpu(*p);
226 row_width = of_get_property(root,
227 "ibm,display-truncation-length", NULL);
230 display_character = rtas_token("display-character");
231 set_indicator = rtas_token("set-indicator");
234 if (display_character == RTAS_UNKNOWN_SERVICE) {
235 /* use hex display if available */
236 if (set_indicator != RTAS_UNKNOWN_SERVICE)
237 rtas_call(set_indicator, 3, 1, NULL, 6, 0, hex);
241 spin_lock(&progress_lock);
244 * Last write ended with newline, but we didn't print it since
245 * it would just clear the bottom line of output. Print it now
248 * If no newline is pending and form feed is supported, clear the
249 * display with a form feed; otherwise, print a CR to start output
250 * at the beginning of the line.
252 if (pending_newline) {
253 rtas_call(display_character, 1, 1, NULL, '\r');
254 rtas_call(display_character, 1, 1, NULL, '\n');
259 rtas_call(display_character, 1, 1, NULL,
262 rtas_call(display_character, 1, 1, NULL, '\r');
266 width = row_width[current_line];
268 width = display_width;
271 if (*os == '\n' || *os == '\r') {
272 /* If newline is the last character, save it
273 * until next call to avoid bumping up the
276 if (*os == '\n' && !os[1]) {
279 if (current_line > display_lines-1)
280 current_line = display_lines-1;
281 spin_unlock(&progress_lock);
285 /* RTAS wants CR-LF, not just LF */
288 rtas_call(display_character, 1, 1, NULL, '\r');
289 rtas_call(display_character, 1, 1, NULL, '\n');
291 /* CR might be used to re-draw a line, so we'll
292 * leave it alone and not add LF.
294 rtas_call(display_character, 1, 1, NULL, *os);
298 width = row_width[current_line];
300 width = display_width;
303 rtas_call(display_character, 1, 1, NULL, *os);
308 /* if we overwrite the screen length */
310 while ((*os != 0) && (*os != '\n') && (*os != '\r'))
314 spin_unlock(&progress_lock);
316 EXPORT_SYMBOL(rtas_progress); /* needed by rtas_flash module */
318 int rtas_token(const char *service)
321 if (rtas.dev == NULL)
322 return RTAS_UNKNOWN_SERVICE;
323 tokp = of_get_property(rtas.dev, service, NULL);
324 return tokp ? be32_to_cpu(*tokp) : RTAS_UNKNOWN_SERVICE;
326 EXPORT_SYMBOL(rtas_token);
328 int rtas_service_present(const char *service)
330 return rtas_token(service) != RTAS_UNKNOWN_SERVICE;
332 EXPORT_SYMBOL(rtas_service_present);
334 #ifdef CONFIG_RTAS_ERROR_LOGGING
336 * Return the firmware-specified size of the error log buffer
337 * for all rtas calls that require an error buffer argument.
338 * This includes 'check-exception' and 'rtas-last-error'.
340 int rtas_get_error_log_max(void)
342 static int rtas_error_log_max;
343 if (rtas_error_log_max)
344 return rtas_error_log_max;
346 rtas_error_log_max = rtas_token ("rtas-error-log-max");
347 if ((rtas_error_log_max == RTAS_UNKNOWN_SERVICE) ||
348 (rtas_error_log_max > RTAS_ERROR_LOG_MAX)) {
349 printk (KERN_WARNING "RTAS: bad log buffer size %d\n",
351 rtas_error_log_max = RTAS_ERROR_LOG_MAX;
353 return rtas_error_log_max;
355 EXPORT_SYMBOL(rtas_get_error_log_max);
358 static char rtas_err_buf[RTAS_ERROR_LOG_MAX];
359 static int rtas_last_error_token;
361 /** Return a copy of the detailed error text associated with the
362 * most recent failed call to rtas. Because the error text
363 * might go stale if there are any other intervening rtas calls,
364 * this routine must be called atomically with whatever produced
365 * the error (i.e. with rtas.lock still held from the previous call).
367 static char *__fetch_rtas_last_error(char *altbuf)
369 struct rtas_args err_args, save_args;
373 if (rtas_last_error_token == -1)
376 bufsz = rtas_get_error_log_max();
378 err_args.token = cpu_to_be32(rtas_last_error_token);
379 err_args.nargs = cpu_to_be32(2);
380 err_args.nret = cpu_to_be32(1);
381 err_args.args[0] = cpu_to_be32(__pa(rtas_err_buf));
382 err_args.args[1] = cpu_to_be32(bufsz);
383 err_args.args[2] = 0;
385 save_args = rtas.args;
386 rtas.args = err_args;
388 enter_rtas(__pa(&rtas.args));
390 err_args = rtas.args;
391 rtas.args = save_args;
393 /* Log the error in the unlikely case that there was one. */
394 if (unlikely(err_args.args[2] == 0)) {
399 if (slab_is_available())
400 buf = kmalloc(RTAS_ERROR_LOG_MAX, GFP_ATOMIC);
403 memmove(buf, rtas_err_buf, RTAS_ERROR_LOG_MAX);
409 #define get_errorlog_buffer() kmalloc(RTAS_ERROR_LOG_MAX, GFP_KERNEL)
411 #else /* CONFIG_RTAS_ERROR_LOGGING */
412 #define __fetch_rtas_last_error(x) NULL
413 #define get_errorlog_buffer() NULL
418 va_rtas_call_unlocked(struct rtas_args *args, int token, int nargs, int nret,
423 args->token = cpu_to_be32(token);
424 args->nargs = cpu_to_be32(nargs);
425 args->nret = cpu_to_be32(nret);
426 args->rets = &(args->args[nargs]);
428 for (i = 0; i < nargs; ++i)
429 args->args[i] = cpu_to_be32(va_arg(list, __u32));
431 for (i = 0; i < nret; ++i)
434 enter_rtas(__pa(args));
437 void rtas_call_unlocked(struct rtas_args *args, int token, int nargs, int nret, ...)
441 va_start(list, nret);
442 va_rtas_call_unlocked(args, token, nargs, nret, list);
446 int rtas_call(int token, int nargs, int nret, int *outputs, ...)
451 struct rtas_args *rtas_args;
452 char *buff_copy = NULL;
455 if (!rtas.entry || token == RTAS_UNKNOWN_SERVICE)
460 /* We use the global rtas args buffer */
461 rtas_args = &rtas.args;
463 va_start(list, outputs);
464 va_rtas_call_unlocked(rtas_args, token, nargs, nret, list);
467 /* A -1 return code indicates that the last command couldn't
468 be completed due to a hardware error. */
469 if (be32_to_cpu(rtas_args->rets[0]) == -1)
470 buff_copy = __fetch_rtas_last_error(NULL);
472 if (nret > 1 && outputs != NULL)
473 for (i = 0; i < nret-1; ++i)
474 outputs[i] = be32_to_cpu(rtas_args->rets[i+1]);
475 ret = (nret > 0)? be32_to_cpu(rtas_args->rets[0]): 0;
480 log_error(buff_copy, ERR_TYPE_RTAS_LOG, 0);
481 if (slab_is_available())
486 EXPORT_SYMBOL(rtas_call);
488 /* For RTAS_BUSY (-2), delay for 1 millisecond. For an extended busy status
489 * code of 990n, perform the hinted delay of 10^n (last digit) milliseconds.
491 unsigned int rtas_busy_delay_time(int status)
496 if (status == RTAS_BUSY) {
498 } else if (status >= RTAS_EXTENDED_DELAY_MIN &&
499 status <= RTAS_EXTENDED_DELAY_MAX) {
500 order = status - RTAS_EXTENDED_DELAY_MIN;
501 for (ms = 1; order > 0; order--)
507 EXPORT_SYMBOL(rtas_busy_delay_time);
509 /* For an RTAS busy status code, perform the hinted delay. */
510 unsigned int rtas_busy_delay(int status)
515 ms = rtas_busy_delay_time(status);
516 if (ms && need_resched())
521 EXPORT_SYMBOL(rtas_busy_delay);
523 static int rtas_error_rc(int rtas_rc)
528 case -1: /* Hardware Error */
531 case -3: /* Bad indicator/domain/etc */
534 case -9000: /* Isolation error */
537 case -9001: /* Outstanding TCE/PTE */
540 case -9002: /* No usable slot */
544 printk(KERN_ERR "%s: unexpected RTAS error %d\n",
552 int rtas_get_power_level(int powerdomain, int *level)
554 int token = rtas_token("get-power-level");
557 if (token == RTAS_UNKNOWN_SERVICE)
560 while ((rc = rtas_call(token, 1, 2, level, powerdomain)) == RTAS_BUSY)
564 return rtas_error_rc(rc);
567 EXPORT_SYMBOL(rtas_get_power_level);
569 int rtas_set_power_level(int powerdomain, int level, int *setlevel)
571 int token = rtas_token("set-power-level");
574 if (token == RTAS_UNKNOWN_SERVICE)
578 rc = rtas_call(token, 2, 2, setlevel, powerdomain, level);
579 } while (rtas_busy_delay(rc));
582 return rtas_error_rc(rc);
585 EXPORT_SYMBOL(rtas_set_power_level);
587 int rtas_get_sensor(int sensor, int index, int *state)
589 int token = rtas_token("get-sensor-state");
592 if (token == RTAS_UNKNOWN_SERVICE)
596 rc = rtas_call(token, 2, 2, state, sensor, index);
597 } while (rtas_busy_delay(rc));
600 return rtas_error_rc(rc);
603 EXPORT_SYMBOL(rtas_get_sensor);
605 int rtas_get_sensor_fast(int sensor, int index, int *state)
607 int token = rtas_token("get-sensor-state");
610 if (token == RTAS_UNKNOWN_SERVICE)
613 rc = rtas_call(token, 2, 2, state, sensor, index);
614 WARN_ON(rc == RTAS_BUSY || (rc >= RTAS_EXTENDED_DELAY_MIN &&
615 rc <= RTAS_EXTENDED_DELAY_MAX));
618 return rtas_error_rc(rc);
622 bool rtas_indicator_present(int token, int *maxindex)
624 int proplen, count, i;
625 const struct indicator_elem {
630 indicators = of_get_property(rtas.dev, "rtas-indicators", &proplen);
634 count = proplen / sizeof(struct indicator_elem);
636 for (i = 0; i < count; i++) {
637 if (__be32_to_cpu(indicators[i].token) != token)
640 *maxindex = __be32_to_cpu(indicators[i].maxindex);
646 EXPORT_SYMBOL(rtas_indicator_present);
648 int rtas_set_indicator(int indicator, int index, int new_value)
650 int token = rtas_token("set-indicator");
653 if (token == RTAS_UNKNOWN_SERVICE)
657 rc = rtas_call(token, 3, 1, NULL, indicator, index, new_value);
658 } while (rtas_busy_delay(rc));
661 return rtas_error_rc(rc);
664 EXPORT_SYMBOL(rtas_set_indicator);
667 * Ignoring RTAS extended delay
669 int rtas_set_indicator_fast(int indicator, int index, int new_value)
672 int token = rtas_token("set-indicator");
674 if (token == RTAS_UNKNOWN_SERVICE)
677 rc = rtas_call(token, 3, 1, NULL, indicator, index, new_value);
679 WARN_ON(rc == RTAS_BUSY || (rc >= RTAS_EXTENDED_DELAY_MIN &&
680 rc <= RTAS_EXTENDED_DELAY_MAX));
683 return rtas_error_rc(rc);
688 void __noreturn rtas_restart(char *cmd)
690 if (rtas_flash_term_hook)
691 rtas_flash_term_hook(SYS_RESTART);
692 printk("RTAS system-reboot returned %d\n",
693 rtas_call(rtas_token("system-reboot"), 0, 1, NULL));
697 void rtas_power_off(void)
699 if (rtas_flash_term_hook)
700 rtas_flash_term_hook(SYS_POWER_OFF);
701 /* allow power on only with power button press */
702 printk("RTAS power-off returned %d\n",
703 rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1));
707 void __noreturn rtas_halt(void)
709 if (rtas_flash_term_hook)
710 rtas_flash_term_hook(SYS_HALT);
711 /* allow power on only with power button press */
712 printk("RTAS power-off returned %d\n",
713 rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1));
717 /* Must be in the RMO region, so we place it here */
718 static char rtas_os_term_buf[2048];
719 static s32 ibm_os_term_token = RTAS_UNKNOWN_SERVICE;
721 void rtas_os_term(char *str)
726 * Firmware with the ibm,extended-os-term property is guaranteed
727 * to always return from an ibm,os-term call. Earlier versions without
728 * this property may terminate the partition which we want to avoid
729 * since it interferes with panic_timeout.
731 if (ibm_os_term_token == RTAS_UNKNOWN_SERVICE)
734 snprintf(rtas_os_term_buf, 2048, "OS panic: %s", str);
737 * Keep calling as long as RTAS returns a "try again" status,
738 * but don't use rtas_busy_delay(), which potentially
742 status = rtas_call(ibm_os_term_token, 1, 1, NULL,
743 __pa(rtas_os_term_buf));
744 } while (rtas_busy_delay_time(status));
747 printk(KERN_EMERG "ibm,os-term call failed %d\n", status);
750 static int ibm_suspend_me_token = RTAS_UNKNOWN_SERVICE;
751 #ifdef CONFIG_PPC_PSERIES
752 static int __rtas_suspend_last_cpu(struct rtas_suspend_me_data *data, int wake_when_done)
754 u16 slb_size = mmu_slb_size;
755 int rc = H_MULTI_THREADS_ACTIVE;
758 slb_set_size(SLB_MIN_SIZE);
759 printk(KERN_DEBUG "calling ibm,suspend-me on cpu %i\n", smp_processor_id());
761 while (rc == H_MULTI_THREADS_ACTIVE && !atomic_read(&data->done) &&
762 !atomic_read(&data->error))
763 rc = rtas_call(data->token, 0, 1, NULL);
765 if (rc || atomic_read(&data->error)) {
766 printk(KERN_DEBUG "ibm,suspend-me returned %d\n", rc);
767 slb_set_size(slb_size);
770 if (atomic_read(&data->error))
771 rc = atomic_read(&data->error);
773 atomic_set(&data->error, rc);
774 pSeries_coalesce_init();
776 if (wake_when_done) {
777 atomic_set(&data->done, 1);
779 for_each_online_cpu(cpu)
780 plpar_hcall_norets(H_PROD, get_hard_smp_processor_id(cpu));
783 if (atomic_dec_return(&data->working) == 0)
784 complete(data->complete);
789 int rtas_suspend_last_cpu(struct rtas_suspend_me_data *data)
791 atomic_inc(&data->working);
792 return __rtas_suspend_last_cpu(data, 0);
795 static int __rtas_suspend_cpu(struct rtas_suspend_me_data *data, int wake_when_done)
798 unsigned long msr_save;
801 atomic_inc(&data->working);
803 /* really need to ensure MSR.EE is off for H_JOIN */
805 mtmsr(msr_save & ~(MSR_EE));
807 while (rc == H_SUCCESS && !atomic_read(&data->done) && !atomic_read(&data->error))
808 rc = plpar_hcall_norets(H_JOIN);
812 if (rc == H_SUCCESS) {
813 /* This cpu was prodded and the suspend is complete. */
815 } else if (rc == H_CONTINUE) {
816 /* All other cpus are in H_JOIN, this cpu does
819 return __rtas_suspend_last_cpu(data, wake_when_done);
821 printk(KERN_ERR "H_JOIN on cpu %i failed with rc = %ld\n",
822 smp_processor_id(), rc);
823 atomic_set(&data->error, rc);
826 if (wake_when_done) {
827 atomic_set(&data->done, 1);
829 /* This cpu did the suspend or got an error; in either case,
830 * we need to prod all other other cpus out of join state.
831 * Extra prods are harmless.
833 for_each_online_cpu(cpu)
834 plpar_hcall_norets(H_PROD, get_hard_smp_processor_id(cpu));
837 if (atomic_dec_return(&data->working) == 0)
838 complete(data->complete);
842 int rtas_suspend_cpu(struct rtas_suspend_me_data *data)
844 return __rtas_suspend_cpu(data, 0);
847 static void rtas_percpu_suspend_me(void *info)
849 __rtas_suspend_cpu((struct rtas_suspend_me_data *)info, 1);
852 enum rtas_cpu_state {
858 static int rtas_cpu_state_change_mask(enum rtas_cpu_state state,
861 if (!cpumask_empty(cpus)) {
868 /* On return cpumask will be altered to indicate CPUs changed.
869 * CPUs with states changed will be set in the mask,
870 * CPUs with status unchanged will be unset in the mask. */
871 static int rtas_cpu_state_change_mask(enum rtas_cpu_state state,
878 if (cpumask_empty(cpus))
881 for_each_cpu(cpu, cpus) {
882 struct device *dev = get_cpu_device(cpu);
886 cpuret = device_offline(dev);
889 cpuret = device_online(dev);
893 pr_debug("%s: cpu_%s for cpu#%d returned %d.\n",
895 ((state == UP) ? "up" : "down"),
900 /* clear bits for unchanged cpus, return */
901 cpumask_shift_right(cpus, cpus, cpu);
902 cpumask_shift_left(cpus, cpus, cpu);
905 /* clear bit for unchanged cpu, continue */
906 cpumask_clear_cpu(cpu, cpus);
915 int rtas_online_cpus_mask(cpumask_var_t cpus)
919 ret = rtas_cpu_state_change_mask(UP, cpus);
922 cpumask_var_t tmp_mask;
924 if (!alloc_cpumask_var(&tmp_mask, GFP_KERNEL))
927 /* Use tmp_mask to preserve cpus mask from first failure */
928 cpumask_copy(tmp_mask, cpus);
929 rtas_offline_cpus_mask(tmp_mask);
930 free_cpumask_var(tmp_mask);
935 EXPORT_SYMBOL(rtas_online_cpus_mask);
937 int rtas_offline_cpus_mask(cpumask_var_t cpus)
939 return rtas_cpu_state_change_mask(DOWN, cpus);
941 EXPORT_SYMBOL(rtas_offline_cpus_mask);
943 int rtas_ibm_suspend_me(u64 handle)
947 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
948 struct rtas_suspend_me_data data;
949 DECLARE_COMPLETION_ONSTACK(done);
950 cpumask_var_t offline_mask;
953 if (!rtas_service_present("ibm,suspend-me"))
956 /* Make sure the state is valid */
957 rc = plpar_hcall(H_VASI_STATE, retbuf, handle);
962 printk(KERN_ERR "rtas_ibm_suspend_me: vasi_state returned %ld\n",rc);
964 } else if (state == H_VASI_ENABLED) {
966 } else if (state != H_VASI_SUSPENDING) {
967 printk(KERN_ERR "rtas_ibm_suspend_me: vasi_state returned state %ld\n",
972 if (!alloc_cpumask_var(&offline_mask, GFP_KERNEL))
975 atomic_set(&data.working, 0);
976 atomic_set(&data.done, 0);
977 atomic_set(&data.error, 0);
978 data.token = rtas_token("ibm,suspend-me");
979 data.complete = &done;
981 lock_device_hotplug();
983 /* All present CPUs must be online */
984 cpumask_andnot(offline_mask, cpu_present_mask, cpu_online_mask);
985 cpuret = rtas_online_cpus_mask(offline_mask);
987 pr_err("%s: Could not bring present CPUs online.\n", __func__);
988 atomic_set(&data.error, cpuret);
992 cpu_hotplug_disable();
993 stop_topology_update();
995 /* Call function on all CPUs. One of us will make the
998 if (on_each_cpu(rtas_percpu_suspend_me, &data, 0))
999 atomic_set(&data.error, -EINVAL);
1001 wait_for_completion(&done);
1003 if (atomic_read(&data.error) != 0)
1004 printk(KERN_ERR "Error doing global join\n");
1006 start_topology_update();
1007 cpu_hotplug_enable();
1009 /* Take down CPUs not online prior to suspend */
1010 cpuret = rtas_offline_cpus_mask(offline_mask);
1012 pr_warn("%s: Could not restore CPUs to offline state.\n",
1016 unlock_device_hotplug();
1017 free_cpumask_var(offline_mask);
1018 return atomic_read(&data.error);
1020 #else /* CONFIG_PPC_PSERIES */
1021 int rtas_ibm_suspend_me(u64 handle)
1028 * Find a specific pseries error log in an RTAS extended event log.
1029 * @log: RTAS error/event log
1030 * @section_id: two character section identifier
1032 * Returns a pointer to the specified errorlog or NULL if not found.
1034 struct pseries_errorlog *get_pseries_errorlog(struct rtas_error_log *log,
1035 uint16_t section_id)
1037 struct rtas_ext_event_log_v6 *ext_log =
1038 (struct rtas_ext_event_log_v6 *)log->buffer;
1039 struct pseries_errorlog *sect;
1040 unsigned char *p, *log_end;
1041 uint32_t ext_log_length = rtas_error_extended_log_length(log);
1042 uint8_t log_format = rtas_ext_event_log_format(ext_log);
1043 uint32_t company_id = rtas_ext_event_company_id(ext_log);
1045 /* Check that we understand the format */
1046 if (ext_log_length < sizeof(struct rtas_ext_event_log_v6) ||
1047 log_format != RTAS_V6EXT_LOG_FORMAT_EVENT_LOG ||
1048 company_id != RTAS_V6EXT_COMPANY_ID_IBM)
1051 log_end = log->buffer + ext_log_length;
1052 p = ext_log->vendor_log;
1054 while (p < log_end) {
1055 sect = (struct pseries_errorlog *)p;
1056 if (pseries_errorlog_id(sect) == section_id)
1058 p += pseries_errorlog_length(sect);
1064 #ifdef CONFIG_PPC_RTAS_FILTER
1067 * The sys_rtas syscall, as originally designed, allows root to pass
1068 * arbitrary physical addresses to RTAS calls. A number of RTAS calls
1069 * can be abused to write to arbitrary memory and do other things that
1070 * are potentially harmful to system integrity, and thus should only
1071 * be used inside the kernel and not exposed to userspace.
1073 * All known legitimate users of the sys_rtas syscall will only ever
1074 * pass addresses that fall within the RMO buffer, and use a known
1075 * subset of RTAS calls.
1077 * Accordingly, we filter RTAS requests to check that the call is
1078 * permitted, and that provided pointers fall within the RMO buffer.
1079 * The rtas_filters list contains an entry for each permitted call,
1080 * with the indexes of the parameters which are expected to contain
1081 * addresses and sizes of buffers allocated inside the RMO buffer.
1083 struct rtas_filter {
1086 /* Indexes into the args buffer, -1 if not used */
1095 static struct rtas_filter rtas_filters[] __ro_after_init = {
1096 { "ibm,activate-firmware", -1, -1, -1, -1, -1 },
1097 { "ibm,configure-connector", -1, 0, -1, 1, -1, 4096 }, /* Special cased */
1098 { "display-character", -1, -1, -1, -1, -1 },
1099 { "ibm,display-message", -1, 0, -1, -1, -1 },
1100 { "ibm,errinjct", -1, 2, -1, -1, -1, 1024 },
1101 { "ibm,close-errinjct", -1, -1, -1, -1, -1 },
1102 { "ibm,open-errinjct", -1, -1, -1, -1, -1 },
1103 { "ibm,get-config-addr-info2", -1, -1, -1, -1, -1 },
1104 { "ibm,get-dynamic-sensor-state", -1, 1, -1, -1, -1 },
1105 { "ibm,get-indices", -1, 2, 3, -1, -1 },
1106 { "get-power-level", -1, -1, -1, -1, -1 },
1107 { "get-sensor-state", -1, -1, -1, -1, -1 },
1108 { "ibm,get-system-parameter", -1, 1, 2, -1, -1 },
1109 { "get-time-of-day", -1, -1, -1, -1, -1 },
1110 { "ibm,get-vpd", -1, 0, -1, 1, 2 },
1111 { "ibm,lpar-perftools", -1, 2, 3, -1, -1 },
1112 { "ibm,platform-dump", -1, 4, 5, -1, -1 }, /* Special cased */
1113 { "ibm,read-slot-reset-state", -1, -1, -1, -1, -1 },
1114 { "ibm,scan-log-dump", -1, 0, 1, -1, -1 },
1115 { "ibm,set-dynamic-indicator", -1, 2, -1, -1, -1 },
1116 { "ibm,set-eeh-option", -1, -1, -1, -1, -1 },
1117 { "set-indicator", -1, -1, -1, -1, -1 },
1118 { "set-power-level", -1, -1, -1, -1, -1 },
1119 { "set-time-for-power-on", -1, -1, -1, -1, -1 },
1120 { "ibm,set-system-parameter", -1, 1, -1, -1, -1 },
1121 { "set-time-of-day", -1, -1, -1, -1, -1 },
1122 { "ibm,suspend-me", -1, -1, -1, -1, -1 },
1123 { "ibm,update-nodes", -1, 0, -1, -1, -1, 4096 },
1124 { "ibm,update-properties", -1, 0, -1, -1, -1, 4096 },
1125 { "ibm,physical-attestation", -1, 0, 1, -1, -1 },
1128 static bool in_rmo_buf(u32 base, u32 end)
1130 return base >= rtas_rmo_buf &&
1131 base < (rtas_rmo_buf + RTAS_RMOBUF_MAX) &&
1133 end >= rtas_rmo_buf &&
1134 end < (rtas_rmo_buf + RTAS_RMOBUF_MAX);
1137 static bool block_rtas_call(int token, int nargs,
1138 struct rtas_args *args)
1142 for (i = 0; i < ARRAY_SIZE(rtas_filters); i++) {
1143 struct rtas_filter *f = &rtas_filters[i];
1144 u32 base, size, end;
1146 if (token != f->token)
1149 if (f->buf_idx1 != -1) {
1150 base = be32_to_cpu(args->args[f->buf_idx1]);
1151 if (f->size_idx1 != -1)
1152 size = be32_to_cpu(args->args[f->size_idx1]);
1153 else if (f->fixed_size)
1154 size = f->fixed_size;
1158 end = base + size - 1;
1161 * Special case for ibm,platform-dump - NULL buffer
1162 * address is used to indicate end of dump processing
1164 if (!strcmp(f->name, "ibm,platform-dump") &&
1168 if (!in_rmo_buf(base, end))
1172 if (f->buf_idx2 != -1) {
1173 base = be32_to_cpu(args->args[f->buf_idx2]);
1174 if (f->size_idx2 != -1)
1175 size = be32_to_cpu(args->args[f->size_idx2]);
1176 else if (f->fixed_size)
1177 size = f->fixed_size;
1180 end = base + size - 1;
1183 * Special case for ibm,configure-connector where the
1186 if (!strcmp(f->name, "ibm,configure-connector") &&
1190 if (!in_rmo_buf(base, end))
1198 pr_err_ratelimited("sys_rtas: RTAS call blocked - exploit attempt?\n");
1199 pr_err_ratelimited("sys_rtas: token=0x%x, nargs=%d (called by %s)\n",
1200 token, nargs, current->comm);
1206 static bool block_rtas_call(int token, int nargs,
1207 struct rtas_args *args)
1212 #endif /* CONFIG_PPC_RTAS_FILTER */
1214 /* We assume to be passed big endian arguments */
1215 asmlinkage int ppc_rtas(struct rtas_args __user *uargs)
1217 struct rtas_args args;
1218 unsigned long flags;
1219 char *buff_copy, *errbuf = NULL;
1220 int nargs, nret, token;
1222 if (!capable(CAP_SYS_ADMIN))
1228 if (copy_from_user(&args, uargs, 3 * sizeof(u32)) != 0)
1231 nargs = be32_to_cpu(args.nargs);
1232 nret = be32_to_cpu(args.nret);
1233 token = be32_to_cpu(args.token);
1235 if (nargs >= ARRAY_SIZE(args.args)
1236 || nret > ARRAY_SIZE(args.args)
1237 || nargs + nret > ARRAY_SIZE(args.args))
1241 if (copy_from_user(args.args, uargs->args,
1242 nargs * sizeof(rtas_arg_t)) != 0)
1245 if (token == RTAS_UNKNOWN_SERVICE)
1248 args.rets = &args.args[nargs];
1249 memset(args.rets, 0, nret * sizeof(rtas_arg_t));
1251 if (block_rtas_call(token, nargs, &args))
1254 /* Need to handle ibm,suspend_me call specially */
1255 if (token == ibm_suspend_me_token) {
1258 * rtas_ibm_suspend_me assumes the streamid handle is in cpu
1259 * endian, or at least the hcall within it requires it.
1262 u64 handle = ((u64)be32_to_cpu(args.args[0]) << 32)
1263 | be32_to_cpu(args.args[1]);
1264 rc = rtas_ibm_suspend_me(handle);
1266 args.rets[0] = cpu_to_be32(RTAS_NOT_SUSPENDABLE);
1267 else if (rc == -EIO)
1268 args.rets[0] = cpu_to_be32(-1);
1274 buff_copy = get_errorlog_buffer();
1276 flags = lock_rtas();
1279 enter_rtas(__pa(&rtas.args));
1282 /* A -1 return code indicates that the last command couldn't
1283 be completed due to a hardware error. */
1284 if (be32_to_cpu(args.rets[0]) == -1)
1285 errbuf = __fetch_rtas_last_error(buff_copy);
1291 log_error(errbuf, ERR_TYPE_RTAS_LOG, 0);
1296 /* Copy out args. */
1297 if (copy_to_user(uargs->args + nargs,
1299 nret * sizeof(rtas_arg_t)) != 0)
1306 * Call early during boot, before mem init, to retrieve the RTAS
1307 * information from the device-tree and allocate the RMO buffer for userland
1310 void __init rtas_initialize(void)
1312 unsigned long rtas_region = RTAS_INSTANTIATE_MAX;
1313 u32 base, size, entry;
1314 int no_base, no_size, no_entry;
1315 #ifdef CONFIG_PPC_RTAS_FILTER
1319 /* Get RTAS dev node and fill up our "rtas" structure with infos
1322 rtas.dev = of_find_node_by_name(NULL, "rtas");
1326 no_base = of_property_read_u32(rtas.dev, "linux,rtas-base", &base);
1327 no_size = of_property_read_u32(rtas.dev, "rtas-size", &size);
1328 if (no_base || no_size) {
1329 of_node_put(rtas.dev);
1336 no_entry = of_property_read_u32(rtas.dev, "linux,rtas-entry", &entry);
1337 rtas.entry = no_entry ? rtas.base : entry;
1340 * Discover these now to avoid device tree lookups in the
1343 if (of_property_read_bool(rtas.dev, "ibm,extended-os-term"))
1344 ibm_os_term_token = rtas_token("ibm,os-term");
1346 /* If RTAS was found, allocate the RMO buffer for it and look for
1347 * the stop-self token if any
1350 if (firmware_has_feature(FW_FEATURE_LPAR)) {
1351 rtas_region = min(ppc64_rma_size, RTAS_INSTANTIATE_MAX);
1352 ibm_suspend_me_token = rtas_token("ibm,suspend-me");
1355 rtas_rmo_buf = memblock_alloc_base(RTAS_RMOBUF_MAX, PAGE_SIZE, rtas_region);
1357 #ifdef CONFIG_RTAS_ERROR_LOGGING
1358 rtas_last_error_token = rtas_token("rtas-last-error");
1361 #ifdef CONFIG_PPC_RTAS_FILTER
1362 for (i = 0; i < ARRAY_SIZE(rtas_filters); i++) {
1363 rtas_filters[i].token = rtas_token(rtas_filters[i].name);
1368 int __init early_init_dt_scan_rtas(unsigned long node,
1369 const char *uname, int depth, void *data)
1371 const u32 *basep, *entryp, *sizep;
1373 if (depth != 1 || strcmp(uname, "rtas") != 0)
1376 basep = of_get_flat_dt_prop(node, "linux,rtas-base", NULL);
1377 entryp = of_get_flat_dt_prop(node, "linux,rtas-entry", NULL);
1378 sizep = of_get_flat_dt_prop(node, "rtas-size", NULL);
1381 /* need this feature to decide the crashkernel offset */
1382 if (of_get_flat_dt_prop(node, "ibm,hypertas-functions", NULL))
1383 powerpc_firmware_features |= FW_FEATURE_LPAR;
1386 if (basep && entryp && sizep) {
1388 rtas.entry = *entryp;
1392 #ifdef CONFIG_UDBG_RTAS_CONSOLE
1393 basep = of_get_flat_dt_prop(node, "put-term-char", NULL);
1395 rtas_putchar_token = *basep;
1397 basep = of_get_flat_dt_prop(node, "get-term-char", NULL);
1399 rtas_getchar_token = *basep;
1401 if (rtas_putchar_token != RTAS_UNKNOWN_SERVICE &&
1402 rtas_getchar_token != RTAS_UNKNOWN_SERVICE)
1403 udbg_init_rtas_console();
1411 static arch_spinlock_t timebase_lock;
1412 static u64 timebase = 0;
1414 void rtas_give_timebase(void)
1416 unsigned long flags;
1418 local_irq_save(flags);
1420 arch_spin_lock(&timebase_lock);
1421 rtas_call(rtas_token("freeze-time-base"), 0, 1, NULL);
1422 timebase = get_tb();
1423 arch_spin_unlock(&timebase_lock);
1427 rtas_call(rtas_token("thaw-time-base"), 0, 1, NULL);
1428 local_irq_restore(flags);
1431 void rtas_take_timebase(void)
1435 arch_spin_lock(&timebase_lock);
1436 set_tb(timebase >> 32, timebase & 0xffffffff);
1438 arch_spin_unlock(&timebase_lock);