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>
39 #include <asm/setup.h>
40 #include <asm/security_features.h>
45 static bool fw_feature_is(const char *state
, const char *name
,
46 struct device_node
*fw_features
)
48 struct device_node
*np
;
51 np
= of_get_child_by_name(fw_features
, name
);
53 rc
= of_property_read_bool(np
, state
);
60 static void init_fw_feat_flags(struct device_node
*np
)
62 if (fw_feature_is("enabled", "inst-spec-barrier-ori31,31,0", np
))
63 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31
);
65 if (fw_feature_is("enabled", "fw-bcctrl-serialized", np
))
66 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED
);
68 if (fw_feature_is("enabled", "inst-spec-barrier-ori31,31,0", np
))
69 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30
);
71 if (fw_feature_is("enabled", "inst-l1d-flush-trig2", np
))
72 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2
);
74 if (fw_feature_is("enabled", "fw-l1d-thread-split", np
))
75 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV
);
77 if (fw_feature_is("enabled", "fw-count-cache-disabled", np
))
78 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED
);
81 * The features below are enabled by default, so we instead look to see
82 * if firmware has *disabled* them, and clear them if so.
84 if (fw_feature_is("disabled", "speculation-policy-favor-security", np
))
85 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY
);
87 if (fw_feature_is("disabled", "needs-l1d-flush-msr-pr-0-to-1", np
))
88 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR
);
90 if (fw_feature_is("disabled", "needs-l1d-flush-msr-hv-1-to-0", np
))
91 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV
);
93 if (fw_feature_is("disabled", "needs-spec-barrier-for-bound-checks", np
))
94 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR
);
97 static void pnv_setup_rfi_flush(void)
99 struct device_node
*np
, *fw_features
;
100 enum l1d_flush_type type
;
103 /* Default to fallback in case fw-features are not available */
104 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
);
114 np
= of_get_child_by_name(fw_features
, "inst-l1d-flush-trig2");
115 if (np
&& of_property_read_bool(np
, "enabled"))
116 type
= L1D_FLUSH_MTTRIG
;
120 np
= of_get_child_by_name(fw_features
, "inst-l1d-flush-ori30,30,0");
121 if (np
&& of_property_read_bool(np
, "enabled"))
122 type
= L1D_FLUSH_ORI
;
126 /* Enable unless firmware says NOT to */
128 np
= of_get_child_by_name(fw_features
, "needs-l1d-flush-msr-hv-1-to-0");
129 if (np
&& of_property_read_bool(np
, "disabled"))
134 np
= of_get_child_by_name(fw_features
, "needs-l1d-flush-msr-pr-0-to-1");
135 if (np
&& of_property_read_bool(np
, "disabled"))
138 np
= of_get_child_by_name(fw_features
, "speculation-policy-favor-security");
139 if (np
&& of_property_read_bool(np
, "disabled"))
143 of_node_put(fw_features
);
146 setup_rfi_flush(type
, enable
> 0);
149 static void __init
pnv_setup_arch(void)
151 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT
);
153 pnv_setup_rfi_flush();
161 /* Setup RTC and NVRAM callbacks */
162 if (firmware_has_feature(FW_FEATURE_OPAL
))
165 /* Enable NAP mode */
171 static void __init
pnv_init(void)
174 * Initialize the LPC bus now so that legacy serial
175 * ports can be found on it
179 #ifdef CONFIG_HVC_OPAL
180 if (firmware_has_feature(FW_FEATURE_OPAL
))
181 hvc_opal_init_early();
184 add_preferred_console("hvc", 0, NULL
);
187 static void __init
pnv_init_IRQ(void)
189 /* Try using a XIVE if available, otherwise use a XICS */
190 if (!xive_native_init())
193 WARN_ON(!ppc_md
.get_irq
);
196 static void pnv_show_cpuinfo(struct seq_file
*m
)
198 struct device_node
*root
;
199 const char *model
= "";
201 root
= of_find_node_by_path("/");
203 model
= of_get_property(root
, "model", NULL
);
204 seq_printf(m
, "machine\t\t: PowerNV %s\n", model
);
205 if (firmware_has_feature(FW_FEATURE_OPAL
))
206 seq_printf(m
, "firmware\t: OPAL\n");
208 seq_printf(m
, "firmware\t: BML\n");
211 seq_printf(m
, "MMU\t\t: Radix\n");
213 seq_printf(m
, "MMU\t\t: Hash\n");
216 static void pnv_prepare_going_down(void)
219 * Disable all notifiers from OPAL, we can't
220 * service interrupts anymore anyway
222 opal_event_shutdown();
224 /* Soft disable interrupts */
228 * Return secondary CPUs to firwmare if a flash update
229 * is pending otherwise we will get all sort of error
230 * messages about CPU being stuck etc.. This will also
231 * have the side effect of hard disabling interrupts so
232 * past this point, the kernel is effectively dead.
234 opal_flash_term_callback();
237 static void __noreturn
pnv_restart(char *cmd
)
241 pnv_prepare_going_down();
243 while (rc
== OPAL_BUSY
|| rc
== OPAL_BUSY_EVENT
) {
244 rc
= opal_cec_reboot();
245 if (rc
== OPAL_BUSY_EVENT
)
246 opal_poll_events(NULL
);
251 opal_poll_events(NULL
);
254 static void __noreturn
pnv_power_off(void)
258 pnv_prepare_going_down();
260 while (rc
== OPAL_BUSY
|| rc
== OPAL_BUSY_EVENT
) {
261 rc
= opal_cec_power_down(0);
262 if (rc
== OPAL_BUSY_EVENT
)
263 opal_poll_events(NULL
);
268 opal_poll_events(NULL
);
271 static void __noreturn
pnv_halt(void)
276 static void pnv_progress(char *s
, unsigned short hex
)
280 static void pnv_shutdown(void)
282 /* Let the PCI code clear up IODA tables */
286 * Stop OPAL activity: Unregister all OPAL interrupts so they
287 * don't fire up while we kexec and make sure all potentially
288 * DMA'ing ops are complete (such as dump retrieval).
293 #ifdef CONFIG_KEXEC_CORE
294 static void pnv_kexec_wait_secondaries_down(void)
296 int my_cpu
, i
, notified
= -1;
300 for_each_online_cpu(i
) {
302 int64_t rc
, timeout
= 1000;
308 rc
= opal_query_cpu_status(get_hard_smp_processor_id(i
),
310 if (rc
!= OPAL_SUCCESS
|| status
!= OPAL_THREAD_STARTED
)
314 printk(KERN_INFO
"kexec: waiting for cpu %d "
315 "(physical %d) to enter OPAL\n",
316 i
, paca
[i
].hw_cpu_id
);
321 * On crash secondaries might be unreachable or hung,
322 * so timeout if we've waited too long
325 if (timeout
-- == 0) {
326 printk(KERN_ERR
"kexec: timed out waiting for "
327 "cpu %d (physical %d) to enter OPAL\n",
328 i
, paca
[i
].hw_cpu_id
);
335 static void pnv_kexec_cpu_down(int crash_shutdown
, int secondary
)
340 xive_kexec_teardown_cpu(secondary
);
342 xics_kexec_teardown_cpu(secondary
);
344 /* On OPAL, we return all CPUs to firmware */
345 if (!firmware_has_feature(FW_FEATURE_OPAL
))
349 /* Return secondary CPUs to firmware on OPAL v3 */
351 get_paca()->kexec_state
= KEXEC_STATE_REAL_MODE
;
354 /* Return the CPU to OPAL */
357 /* Primary waits for the secondaries to have reached OPAL */
358 pnv_kexec_wait_secondaries_down();
360 /* Switch XIVE back to emulation mode */
365 * We might be running as little-endian - now that interrupts
366 * are disabled, reset the HILE bit to big-endian so we don't
367 * take interrupts in the wrong endian later
369 * We reinit to enable both radix and hash on P9 to ensure
370 * the mode used by the next kernel is always supported.
372 reinit_flags
= OPAL_REINIT_CPUS_HILE_BE
;
373 if (cpu_has_feature(CPU_FTR_ARCH_300
))
374 reinit_flags
|= OPAL_REINIT_CPUS_MMU_RADIX
|
375 OPAL_REINIT_CPUS_MMU_HASH
;
376 opal_reinit_cpus(reinit_flags
);
379 #endif /* CONFIG_KEXEC_CORE */
381 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
382 static unsigned long pnv_memory_block_size(void)
385 * We map the kernel linear region with 1GB large pages on radix. For
386 * memory hot unplug to work our memory block size must be at least
390 return 1UL * 1024 * 1024 * 1024;
392 return 256UL * 1024 * 1024;
396 static void __init
pnv_setup_machdep_opal(void)
398 ppc_md
.get_boot_time
= opal_get_boot_time
;
399 ppc_md
.restart
= pnv_restart
;
400 pm_power_off
= pnv_power_off
;
401 ppc_md
.halt
= pnv_halt
;
402 ppc_md
.machine_check_exception
= opal_machine_check
;
403 ppc_md
.mce_check_early_recovery
= opal_mce_check_early_recovery
;
404 ppc_md
.hmi_exception_early
= opal_hmi_exception_early
;
405 ppc_md
.handle_hmi_exception
= opal_handle_hmi_exception
;
408 static int __init
pnv_probe(void)
410 if (!of_machine_is_compatible("ibm,powernv"))
413 if (firmware_has_feature(FW_FEATURE_OPAL
))
414 pnv_setup_machdep_opal();
416 pr_debug("PowerNV detected !\n");
424 * Returns the cpu frequency for 'cpu' in Hz. This is used by
427 static unsigned long pnv_get_proc_freq(unsigned int cpu
)
429 unsigned long ret_freq
;
431 ret_freq
= cpufreq_quick_get(cpu
) * 1000ul;
434 * If the backend cpufreq driver does not exist,
435 * then fallback to old way of reporting the clockrate.
438 ret_freq
= ppc_proc_freq
;
442 define_machine(powernv
) {
445 .setup_arch
= pnv_setup_arch
,
446 .init_IRQ
= pnv_init_IRQ
,
447 .show_cpuinfo
= pnv_show_cpuinfo
,
448 .get_proc_freq
= pnv_get_proc_freq
,
449 .progress
= pnv_progress
,
450 .machine_shutdown
= pnv_shutdown
,
452 .calibrate_decr
= generic_calibrate_decr
,
453 #ifdef CONFIG_KEXEC_CORE
454 .kexec_cpu_down
= pnv_kexec_cpu_down
,
456 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
457 .memory_block_size
= pnv_memory_block_size
,