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-spec-barrier-ori31,31,0", 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
;
108 np
= of_find_node_by_name(NULL
, "ibm,opal");
109 fw_features
= of_get_child_by_name(np
, "fw-features");
113 init_fw_feat_flags(fw_features
);
115 np
= of_get_child_by_name(fw_features
, "inst-l1d-flush-trig2");
116 if (np
&& of_property_read_bool(np
, "enabled"))
117 type
= L1D_FLUSH_MTTRIG
;
121 np
= of_get_child_by_name(fw_features
, "inst-l1d-flush-ori30,30,0");
122 if (np
&& of_property_read_bool(np
, "enabled"))
123 type
= L1D_FLUSH_ORI
;
127 /* Enable unless firmware says NOT to */
129 np
= of_get_child_by_name(fw_features
, "needs-l1d-flush-msr-hv-1-to-0");
130 if (np
&& of_property_read_bool(np
, "disabled"))
135 np
= of_get_child_by_name(fw_features
, "needs-l1d-flush-msr-pr-0-to-1");
136 if (np
&& of_property_read_bool(np
, "disabled"))
139 np
= of_get_child_by_name(fw_features
, "speculation-policy-favor-security");
140 if (np
&& of_property_read_bool(np
, "disabled"))
144 of_node_put(fw_features
);
147 setup_rfi_flush(type
, enable
> 0);
150 static void __init
pnv_setup_arch(void)
152 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT
);
154 pnv_setup_rfi_flush();
162 /* Setup RTC and NVRAM callbacks */
163 if (firmware_has_feature(FW_FEATURE_OPAL
))
166 /* Enable NAP mode */
172 static void __init
pnv_init(void)
175 * Initialize the LPC bus now so that legacy serial
176 * ports can be found on it
180 #ifdef CONFIG_HVC_OPAL
181 if (firmware_has_feature(FW_FEATURE_OPAL
))
182 hvc_opal_init_early();
185 add_preferred_console("hvc", 0, NULL
);
188 static void __init
pnv_init_IRQ(void)
190 /* Try using a XIVE if available, otherwise use a XICS */
191 if (!xive_native_init())
194 WARN_ON(!ppc_md
.get_irq
);
197 static void pnv_show_cpuinfo(struct seq_file
*m
)
199 struct device_node
*root
;
200 const char *model
= "";
202 root
= of_find_node_by_path("/");
204 model
= of_get_property(root
, "model", NULL
);
205 seq_printf(m
, "machine\t\t: PowerNV %s\n", model
);
206 if (firmware_has_feature(FW_FEATURE_OPAL
))
207 seq_printf(m
, "firmware\t: OPAL\n");
209 seq_printf(m
, "firmware\t: BML\n");
212 seq_printf(m
, "MMU\t\t: Radix\n");
214 seq_printf(m
, "MMU\t\t: Hash\n");
217 static void pnv_prepare_going_down(void)
220 * Disable all notifiers from OPAL, we can't
221 * service interrupts anymore anyway
223 opal_event_shutdown();
225 /* Soft disable interrupts */
229 * Return secondary CPUs to firwmare if a flash update
230 * is pending otherwise we will get all sort of error
231 * messages about CPU being stuck etc.. This will also
232 * have the side effect of hard disabling interrupts so
233 * past this point, the kernel is effectively dead.
235 opal_flash_term_callback();
238 static void __noreturn
pnv_restart(char *cmd
)
242 pnv_prepare_going_down();
244 while (rc
== OPAL_BUSY
|| rc
== OPAL_BUSY_EVENT
) {
245 rc
= opal_cec_reboot();
246 if (rc
== OPAL_BUSY_EVENT
)
247 opal_poll_events(NULL
);
252 opal_poll_events(NULL
);
255 static void __noreturn
pnv_power_off(void)
259 pnv_prepare_going_down();
261 while (rc
== OPAL_BUSY
|| rc
== OPAL_BUSY_EVENT
) {
262 rc
= opal_cec_power_down(0);
263 if (rc
== OPAL_BUSY_EVENT
)
264 opal_poll_events(NULL
);
269 opal_poll_events(NULL
);
272 static void __noreturn
pnv_halt(void)
277 static void pnv_progress(char *s
, unsigned short hex
)
281 static void pnv_shutdown(void)
283 /* Let the PCI code clear up IODA tables */
287 * Stop OPAL activity: Unregister all OPAL interrupts so they
288 * don't fire up while we kexec and make sure all potentially
289 * DMA'ing ops are complete (such as dump retrieval).
294 #ifdef CONFIG_KEXEC_CORE
295 static void pnv_kexec_wait_secondaries_down(void)
297 int my_cpu
, i
, notified
= -1;
301 for_each_online_cpu(i
) {
303 int64_t rc
, timeout
= 1000;
309 rc
= opal_query_cpu_status(get_hard_smp_processor_id(i
),
311 if (rc
!= OPAL_SUCCESS
|| status
!= OPAL_THREAD_STARTED
)
315 printk(KERN_INFO
"kexec: waiting for cpu %d "
316 "(physical %d) to enter OPAL\n",
317 i
, paca
[i
].hw_cpu_id
);
322 * On crash secondaries might be unreachable or hung,
323 * so timeout if we've waited too long
326 if (timeout
-- == 0) {
327 printk(KERN_ERR
"kexec: timed out waiting for "
328 "cpu %d (physical %d) to enter OPAL\n",
329 i
, paca
[i
].hw_cpu_id
);
336 static void pnv_kexec_cpu_down(int crash_shutdown
, int secondary
)
341 xive_kexec_teardown_cpu(secondary
);
343 xics_kexec_teardown_cpu(secondary
);
345 /* On OPAL, we return all CPUs to firmware */
346 if (!firmware_has_feature(FW_FEATURE_OPAL
))
350 /* Return secondary CPUs to firmware on OPAL v3 */
352 get_paca()->kexec_state
= KEXEC_STATE_REAL_MODE
;
355 /* Return the CPU to OPAL */
358 /* Primary waits for the secondaries to have reached OPAL */
359 pnv_kexec_wait_secondaries_down();
361 /* Switch XIVE back to emulation mode */
366 * We might be running as little-endian - now that interrupts
367 * are disabled, reset the HILE bit to big-endian so we don't
368 * take interrupts in the wrong endian later
370 * We reinit to enable both radix and hash on P9 to ensure
371 * the mode used by the next kernel is always supported.
373 reinit_flags
= OPAL_REINIT_CPUS_HILE_BE
;
374 if (cpu_has_feature(CPU_FTR_ARCH_300
))
375 reinit_flags
|= OPAL_REINIT_CPUS_MMU_RADIX
|
376 OPAL_REINIT_CPUS_MMU_HASH
;
377 opal_reinit_cpus(reinit_flags
);
380 #endif /* CONFIG_KEXEC_CORE */
382 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
383 static unsigned long pnv_memory_block_size(void)
386 * We map the kernel linear region with 1GB large pages on radix. For
387 * memory hot unplug to work our memory block size must be at least
391 return 1UL * 1024 * 1024 * 1024;
393 return 256UL * 1024 * 1024;
397 static void __init
pnv_setup_machdep_opal(void)
399 ppc_md
.get_boot_time
= opal_get_boot_time
;
400 ppc_md
.restart
= pnv_restart
;
401 pm_power_off
= pnv_power_off
;
402 ppc_md
.halt
= pnv_halt
;
403 /* ppc_md.system_reset_exception gets filled in by pnv_smp_init() */
404 ppc_md
.machine_check_exception
= opal_machine_check
;
405 ppc_md
.mce_check_early_recovery
= opal_mce_check_early_recovery
;
406 ppc_md
.hmi_exception_early
= opal_hmi_exception_early
;
407 ppc_md
.handle_hmi_exception
= opal_handle_hmi_exception
;
410 static int __init
pnv_probe(void)
412 if (!of_machine_is_compatible("ibm,powernv"))
415 if (firmware_has_feature(FW_FEATURE_OPAL
))
416 pnv_setup_machdep_opal();
418 pr_debug("PowerNV detected !\n");
425 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
426 void __init
pnv_tm_init(void)
428 if (!firmware_has_feature(FW_FEATURE_OPAL
) ||
429 !pvr_version_is(PVR_POWER9
) ||
430 early_cpu_has_feature(CPU_FTR_TM
))
433 if (opal_reinit_cpus(OPAL_REINIT_CPUS_TM_SUSPEND_DISABLED
) != OPAL_SUCCESS
)
436 pr_info("Enabling TM (Transactional Memory) with Suspend Disabled\n");
437 cur_cpu_spec
->cpu_features
|= CPU_FTR_TM
;
438 /* Make sure "normal" HTM is off (it should be) */
439 cur_cpu_spec
->cpu_user_features2
&= ~PPC_FEATURE2_HTM
;
440 /* Turn on no suspend mode, and HTM no SC */
441 cur_cpu_spec
->cpu_user_features2
|= PPC_FEATURE2_HTM_NO_SUSPEND
| \
442 PPC_FEATURE2_HTM_NOSC
;
443 tm_suspend_disabled
= true;
445 #endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
448 * Returns the cpu frequency for 'cpu' in Hz. This is used by
451 static unsigned long pnv_get_proc_freq(unsigned int cpu
)
453 unsigned long ret_freq
;
455 ret_freq
= cpufreq_get(cpu
) * 1000ul;
458 * If the backend cpufreq driver does not exist,
459 * then fallback to old way of reporting the clockrate.
462 ret_freq
= ppc_proc_freq
;
466 define_machine(powernv
) {
469 .setup_arch
= pnv_setup_arch
,
470 .init_IRQ
= pnv_init_IRQ
,
471 .show_cpuinfo
= pnv_show_cpuinfo
,
472 .get_proc_freq
= pnv_get_proc_freq
,
473 .progress
= pnv_progress
,
474 .machine_shutdown
= pnv_shutdown
,
476 .calibrate_decr
= generic_calibrate_decr
,
477 #ifdef CONFIG_KEXEC_CORE
478 .kexec_cpu_down
= pnv_kexec_cpu_down
,
480 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
481 .memory_block_size
= pnv_memory_block_size
,