4 * Copyright 2011 IBM Corp.
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
14 #include <linux/cpu.h>
15 #include <linux/errno.h>
16 #include <linux/sched.h>
17 #include <linux/kernel.h>
18 #include <linux/tty.h>
19 #include <linux/reboot.h>
20 #include <linux/init.h>
21 #include <linux/console.h>
22 #include <linux/delay.h>
23 #include <linux/irq.h>
24 #include <linux/seq_file.h>
26 #include <linux/of_fdt.h>
27 #include <linux/interrupt.h>
28 #include <linux/bug.h>
29 #include <linux/pci.h>
30 #include <linux/cpufreq.h>
32 #include <asm/machdep.h>
33 #include <asm/firmware.h>
37 #include <asm/kexec.h>
40 #include <asm/setup.h>
41 #include <asm/security_features.h>
46 static bool fw_feature_is(const char *state
, const char *name
,
47 struct device_node
*fw_features
)
49 struct device_node
*np
;
52 np
= of_get_child_by_name(fw_features
, name
);
54 rc
= of_property_read_bool(np
, state
);
61 static void init_fw_feat_flags(struct device_node
*np
)
63 if (fw_feature_is("enabled", "inst-spec-barrier-ori31,31,0", np
))
64 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31
);
66 if (fw_feature_is("enabled", "fw-bcctrl-serialized", np
))
67 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED
);
69 if (fw_feature_is("enabled", "inst-l1d-flush-ori30,30,0", np
))
70 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30
);
72 if (fw_feature_is("enabled", "inst-l1d-flush-trig2", np
))
73 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2
);
75 if (fw_feature_is("enabled", "fw-l1d-thread-split", np
))
76 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV
);
78 if (fw_feature_is("enabled", "fw-count-cache-disabled", np
))
79 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED
);
82 * The features below are enabled by default, so we instead look to see
83 * if firmware has *disabled* them, and clear them if so.
85 if (fw_feature_is("disabled", "speculation-policy-favor-security", np
))
86 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY
);
88 if (fw_feature_is("disabled", "needs-l1d-flush-msr-pr-0-to-1", np
))
89 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR
);
91 if (fw_feature_is("disabled", "needs-l1d-flush-msr-hv-1-to-0", np
))
92 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV
);
94 if (fw_feature_is("disabled", "needs-spec-barrier-for-bound-checks", np
))
95 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR
);
98 static void pnv_setup_rfi_flush(void)
100 struct device_node
*np
, *fw_features
;
101 enum l1d_flush_type type
;
104 /* Default to fallback in case fw-features are not available */
105 type
= L1D_FLUSH_FALLBACK
;
107 np
= of_find_node_by_name(NULL
, "ibm,opal");
108 fw_features
= of_get_child_by_name(np
, "fw-features");
112 init_fw_feat_flags(fw_features
);
113 of_node_put(fw_features
);
115 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2
))
116 type
= L1D_FLUSH_MTTRIG
;
118 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30
))
119 type
= L1D_FLUSH_ORI
;
122 enable
= security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY
) && \
123 (security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR
) || \
124 security_ftr_enabled(SEC_FTR_L1D_FLUSH_HV
));
126 setup_rfi_flush(type
, enable
);
129 static void __init
pnv_setup_arch(void)
131 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT
);
133 pnv_setup_rfi_flush();
142 /* Setup RTC and NVRAM callbacks */
143 if (firmware_has_feature(FW_FEATURE_OPAL
))
146 /* Enable NAP mode */
152 static void __init
pnv_init(void)
155 * Initialize the LPC bus now so that legacy serial
156 * ports can be found on it
160 #ifdef CONFIG_HVC_OPAL
161 if (firmware_has_feature(FW_FEATURE_OPAL
))
162 hvc_opal_init_early();
165 add_preferred_console("hvc", 0, NULL
);
168 static void __init
pnv_init_IRQ(void)
170 /* Try using a XIVE if available, otherwise use a XICS */
171 if (!xive_native_init())
174 WARN_ON(!ppc_md
.get_irq
);
177 static void pnv_show_cpuinfo(struct seq_file
*m
)
179 struct device_node
*root
;
180 const char *model
= "";
182 root
= of_find_node_by_path("/");
184 model
= of_get_property(root
, "model", NULL
);
185 seq_printf(m
, "machine\t\t: PowerNV %s\n", model
);
186 if (firmware_has_feature(FW_FEATURE_OPAL
))
187 seq_printf(m
, "firmware\t: OPAL\n");
189 seq_printf(m
, "firmware\t: BML\n");
192 seq_printf(m
, "MMU\t\t: Radix\n");
194 seq_printf(m
, "MMU\t\t: Hash\n");
197 static void pnv_prepare_going_down(void)
200 * Disable all notifiers from OPAL, we can't
201 * service interrupts anymore anyway
203 opal_event_shutdown();
205 /* Soft disable interrupts */
209 * Return secondary CPUs to firwmare if a flash update
210 * is pending otherwise we will get all sort of error
211 * messages about CPU being stuck etc.. This will also
212 * have the side effect of hard disabling interrupts so
213 * past this point, the kernel is effectively dead.
215 opal_flash_term_callback();
218 static void __noreturn
pnv_restart(char *cmd
)
222 pnv_prepare_going_down();
226 rc
= opal_cec_reboot();
227 else if (strcmp(cmd
, "full") == 0)
228 rc
= opal_cec_reboot2(OPAL_REBOOT_FULL_IPL
, NULL
);
230 rc
= OPAL_UNSUPPORTED
;
232 if (rc
== OPAL_BUSY
|| rc
== OPAL_BUSY_EVENT
) {
233 /* Opal is busy wait for some time and retry */
234 opal_poll_events(NULL
);
237 } else if (cmd
&& rc
) {
238 /* Unknown error while issuing reboot */
239 if (rc
== OPAL_UNSUPPORTED
)
240 pr_err("Unsupported '%s' reboot.\n", cmd
);
242 pr_err("Unable to issue '%s' reboot. Err=%ld\n",
244 pr_info("Forcing a cec-reboot\n");
248 } else if (rc
!= OPAL_SUCCESS
) {
249 /* Unknown error while issuing cec-reboot */
250 pr_err("Unable to reboot. Err=%ld\n", rc
);
253 } while (rc
== OPAL_BUSY
|| rc
== OPAL_BUSY_EVENT
);
256 opal_poll_events(NULL
);
259 static void __noreturn
pnv_power_off(void)
263 pnv_prepare_going_down();
265 while (rc
== OPAL_BUSY
|| rc
== OPAL_BUSY_EVENT
) {
266 rc
= opal_cec_power_down(0);
267 if (rc
== OPAL_BUSY_EVENT
)
268 opal_poll_events(NULL
);
273 opal_poll_events(NULL
);
276 static void __noreturn
pnv_halt(void)
281 static void pnv_progress(char *s
, unsigned short hex
)
285 static void pnv_shutdown(void)
287 /* Let the PCI code clear up IODA tables */
291 * Stop OPAL activity: Unregister all OPAL interrupts so they
292 * don't fire up while we kexec and make sure all potentially
293 * DMA'ing ops are complete (such as dump retrieval).
298 #ifdef CONFIG_KEXEC_CORE
299 static void pnv_kexec_wait_secondaries_down(void)
301 int my_cpu
, i
, notified
= -1;
305 for_each_online_cpu(i
) {
307 int64_t rc
, timeout
= 1000;
313 rc
= opal_query_cpu_status(get_hard_smp_processor_id(i
),
315 if (rc
!= OPAL_SUCCESS
|| status
!= OPAL_THREAD_STARTED
)
319 printk(KERN_INFO
"kexec: waiting for cpu %d "
320 "(physical %d) to enter OPAL\n",
321 i
, paca
[i
].hw_cpu_id
);
326 * On crash secondaries might be unreachable or hung,
327 * so timeout if we've waited too long
330 if (timeout
-- == 0) {
331 printk(KERN_ERR
"kexec: timed out waiting for "
332 "cpu %d (physical %d) to enter OPAL\n",
333 i
, paca
[i
].hw_cpu_id
);
340 static void pnv_kexec_cpu_down(int crash_shutdown
, int secondary
)
345 xive_kexec_teardown_cpu(secondary
);
347 xics_kexec_teardown_cpu(secondary
);
349 /* On OPAL, we return all CPUs to firmware */
350 if (!firmware_has_feature(FW_FEATURE_OPAL
))
354 /* Return secondary CPUs to firmware on OPAL v3 */
356 get_paca()->kexec_state
= KEXEC_STATE_REAL_MODE
;
359 /* Return the CPU to OPAL */
362 /* Primary waits for the secondaries to have reached OPAL */
363 pnv_kexec_wait_secondaries_down();
365 /* Switch XIVE back to emulation mode */
370 * We might be running as little-endian - now that interrupts
371 * are disabled, reset the HILE bit to big-endian so we don't
372 * take interrupts in the wrong endian later
374 * We reinit to enable both radix and hash on P9 to ensure
375 * the mode used by the next kernel is always supported.
377 reinit_flags
= OPAL_REINIT_CPUS_HILE_BE
;
378 if (cpu_has_feature(CPU_FTR_ARCH_300
))
379 reinit_flags
|= OPAL_REINIT_CPUS_MMU_RADIX
|
380 OPAL_REINIT_CPUS_MMU_HASH
;
381 opal_reinit_cpus(reinit_flags
);
384 #endif /* CONFIG_KEXEC_CORE */
386 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
387 static unsigned long pnv_memory_block_size(void)
389 return 256UL * 1024 * 1024;
393 static void __init
pnv_setup_machdep_opal(void)
395 ppc_md
.get_boot_time
= opal_get_boot_time
;
396 ppc_md
.restart
= pnv_restart
;
397 pm_power_off
= pnv_power_off
;
398 ppc_md
.halt
= pnv_halt
;
399 /* ppc_md.system_reset_exception gets filled in by pnv_smp_init() */
400 ppc_md
.machine_check_exception
= opal_machine_check
;
401 ppc_md
.mce_check_early_recovery
= opal_mce_check_early_recovery
;
402 ppc_md
.hmi_exception_early
= opal_hmi_exception_early
;
403 ppc_md
.handle_hmi_exception
= opal_handle_hmi_exception
;
406 static int __init
pnv_probe(void)
408 if (!of_machine_is_compatible("ibm,powernv"))
411 if (firmware_has_feature(FW_FEATURE_OPAL
))
412 pnv_setup_machdep_opal();
414 pr_debug("PowerNV detected !\n");
421 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
422 void __init
pnv_tm_init(void)
424 if (!firmware_has_feature(FW_FEATURE_OPAL
) ||
425 !pvr_version_is(PVR_POWER9
) ||
426 early_cpu_has_feature(CPU_FTR_TM
))
429 if (opal_reinit_cpus(OPAL_REINIT_CPUS_TM_SUSPEND_DISABLED
) != OPAL_SUCCESS
)
432 pr_info("Enabling TM (Transactional Memory) with Suspend Disabled\n");
433 cur_cpu_spec
->cpu_features
|= CPU_FTR_TM
;
434 /* Make sure "normal" HTM is off (it should be) */
435 cur_cpu_spec
->cpu_user_features2
&= ~PPC_FEATURE2_HTM
;
436 /* Turn on no suspend mode, and HTM no SC */
437 cur_cpu_spec
->cpu_user_features2
|= PPC_FEATURE2_HTM_NO_SUSPEND
| \
438 PPC_FEATURE2_HTM_NOSC
;
439 tm_suspend_disabled
= true;
441 #endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
444 * Returns the cpu frequency for 'cpu' in Hz. This is used by
447 static unsigned long pnv_get_proc_freq(unsigned int cpu
)
449 unsigned long ret_freq
;
451 ret_freq
= cpufreq_get(cpu
) * 1000ul;
454 * If the backend cpufreq driver does not exist,
455 * then fallback to old way of reporting the clockrate.
458 ret_freq
= ppc_proc_freq
;
462 define_machine(powernv
) {
465 .setup_arch
= pnv_setup_arch
,
466 .init_IRQ
= pnv_init_IRQ
,
467 .show_cpuinfo
= pnv_show_cpuinfo
,
468 .get_proc_freq
= pnv_get_proc_freq
,
469 .progress
= pnv_progress
,
470 .machine_shutdown
= pnv_shutdown
,
472 .calibrate_decr
= generic_calibrate_decr
,
473 #ifdef CONFIG_KEXEC_CORE
474 .kexec_cpu_down
= pnv_kexec_cpu_down
,
476 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
477 .memory_block_size
= pnv_memory_block_size
,