2 * 64-bit pSeries and RS/6000 setup code.
4 * Copyright (C) 1995 Linus Torvalds
5 * Adapted from 'alpha' version by Gary Thomas
6 * Modified by Cort Dougan (cort@cs.nmt.edu)
7 * Modified by PPC64 Team, IBM Corp
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * bootup setup stuff..
19 #include <linux/cpu.h>
20 #include <linux/errno.h>
21 #include <linux/sched.h>
22 #include <linux/kernel.h>
24 #include <linux/stddef.h>
25 #include <linux/unistd.h>
26 #include <linux/user.h>
27 #include <linux/tty.h>
28 #include <linux/major.h>
29 #include <linux/interrupt.h>
30 #include <linux/reboot.h>
31 #include <linux/init.h>
32 #include <linux/ioport.h>
33 #include <linux/console.h>
34 #include <linux/pci.h>
35 #include <linux/utsname.h>
36 #include <linux/adb.h>
37 #include <linux/export.h>
38 #include <linux/delay.h>
39 #include <linux/irq.h>
40 #include <linux/seq_file.h>
41 #include <linux/root_dev.h>
42 #include <linux/cpuidle.h>
44 #include <linux/kexec.h>
47 #include <asm/processor.h>
49 #include <asm/pgtable.h>
52 #include <asm/pci-bridge.h>
53 #include <asm/iommu.h>
55 #include <asm/machdep.h>
58 #include <asm/nvram.h>
62 #include <asm/ppc-pci.h>
63 #include <asm/i8259.h>
66 #include <asm/firmware.h>
69 #include <asm/plpar_wrappers.h>
75 unsigned long CMO_PageSize
= (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K
);
76 EXPORT_SYMBOL(CMO_PageSize
);
78 int fwnmi_active
; /* TRUE if an FWNMI handler is present */
80 static struct device_node
*pSeries_mpic_node
;
82 static void pSeries_show_cpuinfo(struct seq_file
*m
)
84 struct device_node
*root
;
85 const char *model
= "";
87 root
= of_find_node_by_path("/");
89 model
= of_get_property(root
, "model", NULL
);
90 seq_printf(m
, "machine\t\t: CHRP %s\n", model
);
94 /* Initialize firmware assisted non-maskable interrupts if
95 * the firmware supports this feature.
97 static void __init
fwnmi_init(void)
99 unsigned long system_reset_addr
, machine_check_addr
;
101 int ibm_nmi_register
= rtas_token("ibm,nmi-register");
102 if (ibm_nmi_register
== RTAS_UNKNOWN_SERVICE
)
105 /* If the kernel's not linked at zero we point the firmware at low
106 * addresses anyway, and use a trampoline to get to the real code. */
107 system_reset_addr
= __pa(system_reset_fwnmi
) - PHYSICAL_START
;
108 machine_check_addr
= __pa(machine_check_fwnmi
) - PHYSICAL_START
;
110 if (0 == rtas_call(ibm_nmi_register
, 2, 1, NULL
, system_reset_addr
,
115 static void pseries_8259_cascade(unsigned int irq
, struct irq_desc
*desc
)
117 struct irq_chip
*chip
= irq_desc_get_chip(desc
);
118 unsigned int cascade_irq
= i8259_irq();
120 if (cascade_irq
!= NO_IRQ
)
121 generic_handle_irq(cascade_irq
);
123 chip
->irq_eoi(&desc
->irq_data
);
126 static void __init
pseries_setup_i8259_cascade(void)
128 struct device_node
*np
, *old
, *found
= NULL
;
129 unsigned int cascade
;
131 unsigned long intack
= 0;
134 for_each_node_by_type(np
, "interrupt-controller") {
135 if (of_device_is_compatible(np
, "chrp,iic")) {
142 printk(KERN_DEBUG
"pic: no ISA interrupt controller\n");
146 cascade
= irq_of_parse_and_map(found
, 0);
147 if (cascade
== NO_IRQ
) {
148 printk(KERN_ERR
"pic: failed to map cascade interrupt");
151 pr_debug("pic: cascade mapped to irq %d\n", cascade
);
153 for (old
= of_node_get(found
); old
!= NULL
; old
= np
) {
154 np
= of_get_parent(old
);
158 if (strcmp(np
->name
, "pci") != 0)
160 addrp
= of_get_property(np
, "8259-interrupt-acknowledge", NULL
);
163 naddr
= of_n_addr_cells(np
);
164 intack
= addrp
[naddr
-1];
166 intack
|= ((unsigned long)addrp
[naddr
-2]) << 32;
169 printk(KERN_DEBUG
"pic: PCI 8259 intack at 0x%016lx\n", intack
);
170 i8259_init(found
, intack
);
172 irq_set_chained_handler(cascade
, pseries_8259_cascade
);
175 static void __init
pseries_mpic_init_IRQ(void)
177 struct device_node
*np
;
178 const unsigned int *opprop
;
179 unsigned long openpic_addr
= 0;
180 int naddr
, n
, i
, opplen
;
183 np
= of_find_node_by_path("/");
184 naddr
= of_n_addr_cells(np
);
185 opprop
= of_get_property(np
, "platform-open-pic", &opplen
);
186 if (opprop
!= NULL
) {
187 openpic_addr
= of_read_number(opprop
, naddr
);
188 printk(KERN_DEBUG
"OpenPIC addr: %lx\n", openpic_addr
);
192 BUG_ON(openpic_addr
== 0);
194 /* Setup the openpic driver */
195 mpic
= mpic_alloc(pSeries_mpic_node
, openpic_addr
,
196 MPIC_NO_RESET
, 16, 0, " MPIC ");
197 BUG_ON(mpic
== NULL
);
200 opplen
/= sizeof(u32
);
201 for (n
= 0, i
= naddr
; i
< opplen
; i
+= naddr
, n
++) {
202 unsigned long isuaddr
= of_read_number(opprop
+ i
, naddr
);
203 mpic_assign_isu(mpic
, n
, isuaddr
);
206 /* Setup top-level get_irq */
207 ppc_md
.get_irq
= mpic_get_irq
;
209 /* All ISUs are setup, complete initialization */
212 /* Look for cascade */
213 pseries_setup_i8259_cascade();
216 static void __init
pseries_xics_init_IRQ(void)
219 pseries_setup_i8259_cascade();
222 static void pseries_lpar_enable_pmcs(void)
224 unsigned long set
, reset
;
228 plpar_hcall_norets(H_PERFMON
, set
, reset
);
231 static void __init
pseries_discover_pic(void)
233 struct device_node
*np
;
236 for (np
= NULL
; (np
= of_find_node_by_name(np
,
237 "interrupt-controller"));) {
238 typep
= of_get_property(np
, "compatible", NULL
);
239 if (strstr(typep
, "open-pic")) {
240 pSeries_mpic_node
= of_node_get(np
);
241 ppc_md
.init_IRQ
= pseries_mpic_init_IRQ
;
242 setup_kexec_cpu_down_mpic();
243 smp_init_pseries_mpic();
245 } else if (strstr(typep
, "ppc-xicp")) {
246 ppc_md
.init_IRQ
= pseries_xics_init_IRQ
;
247 setup_kexec_cpu_down_xics();
248 smp_init_pseries_xics();
252 printk(KERN_ERR
"pSeries_discover_pic: failed to recognize"
253 " interrupt-controller\n");
256 static int pci_dn_reconfig_notifier(struct notifier_block
*nb
, unsigned long action
, void *node
)
258 struct device_node
*np
= node
;
259 struct pci_dn
*pci
= NULL
;
263 case OF_RECONFIG_ATTACH_NODE
:
264 pci
= np
->parent
->data
;
266 update_dn_pci_info(np
, pci
->phb
);
268 /* Create EEH device for the OF node */
269 eeh_dev_init(np
, pci
->phb
);
279 static struct notifier_block pci_dn_reconfig_nb
= {
280 .notifier_call
= pci_dn_reconfig_notifier
,
283 struct kmem_cache
*dtl_cache
;
285 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
287 * Allocate space for the dispatch trace log for all possible cpus
288 * and register the buffers with the hypervisor. This is used for
289 * computing time stolen by the hypervisor.
291 static int alloc_dispatch_logs(void)
294 struct paca_struct
*pp
;
295 struct dtl_entry
*dtl
;
297 if (!firmware_has_feature(FW_FEATURE_SPLPAR
))
303 for_each_possible_cpu(cpu
) {
305 dtl
= kmem_cache_alloc(dtl_cache
, GFP_KERNEL
);
307 pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
309 pr_warn("Stolen time statistics will be unreliable\n");
314 pp
->dispatch_log
= dtl
;
315 pp
->dispatch_log_end
= dtl
+ N_DISPATCH_LOG
;
319 /* Register the DTL for the current (boot) cpu */
320 dtl
= get_paca()->dispatch_log
;
321 get_paca()->dtl_ridx
= 0;
322 get_paca()->dtl_curr
= dtl
;
323 get_paca()->lppaca_ptr
->dtl_idx
= 0;
325 /* hypervisor reads buffer length from this field */
326 dtl
->enqueue_to_dispatch_time
= cpu_to_be32(DISPATCH_LOG_BYTES
);
327 ret
= register_dtl(hard_smp_processor_id(), __pa(dtl
));
329 pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
330 "with %d\n", smp_processor_id(),
331 hard_smp_processor_id(), ret
);
332 get_paca()->lppaca_ptr
->dtl_enable_mask
= 2;
336 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
337 static inline int alloc_dispatch_logs(void)
341 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
343 static int alloc_dispatch_log_kmem_cache(void)
345 dtl_cache
= kmem_cache_create("dtl", DISPATCH_LOG_BYTES
,
346 DISPATCH_LOG_BYTES
, 0, NULL
);
348 pr_warn("Failed to create dispatch trace log buffer cache\n");
349 pr_warn("Stolen time statistics will be unreliable\n");
353 return alloc_dispatch_logs();
355 early_initcall(alloc_dispatch_log_kmem_cache
);
357 static void pseries_lpar_idle(void)
359 /* This would call on the cpuidle framework, and the back-end pseries
360 * driver to go to idle states
362 if (cpuidle_idle_call()) {
363 /* On error, execute default handler
364 * to go into low thread priority and possibly
365 * low power mode by cedeing processor to hypervisor
368 /* Indicate to hypervisor that we are idle. */
369 get_lppaca()->idle
= 1;
372 * Yield the processor to the hypervisor. We return if
373 * an external interrupt occurs (which are driven prior
374 * to returning here) or if a prod occurs from another
375 * processor. When returning here, external interrupts
380 get_lppaca()->idle
= 0;
385 * Enable relocation on during exceptions. This has partition wide scope and
386 * may take a while to complete, if it takes longer than one second we will
387 * just give up rather than wasting any more time on this - if that turns out
388 * to ever be a problem in practice we can move this into a kernel thread to
389 * finish off the process later in boot.
391 long pSeries_enable_reloc_on_exc(void)
394 unsigned int delay
, total_delay
= 0;
397 rc
= enable_reloc_on_exceptions();
398 if (!H_IS_LONG_BUSY(rc
))
401 delay
= get_longbusy_msecs(rc
);
402 total_delay
+= delay
;
403 if (total_delay
> 1000) {
404 pr_warn("Warning: Giving up waiting to enable "
405 "relocation on exceptions (%u msec)!\n",
413 EXPORT_SYMBOL(pSeries_enable_reloc_on_exc
);
415 long pSeries_disable_reloc_on_exc(void)
420 rc
= disable_reloc_on_exceptions();
421 if (!H_IS_LONG_BUSY(rc
))
423 mdelay(get_longbusy_msecs(rc
));
426 EXPORT_SYMBOL(pSeries_disable_reloc_on_exc
);
429 static void pSeries_machine_kexec(struct kimage
*image
)
433 if (firmware_has_feature(FW_FEATURE_SET_MODE
) &&
434 (image
->type
!= KEXEC_TYPE_CRASH
)) {
435 rc
= pSeries_disable_reloc_on_exc();
437 pr_warning("Warning: Failed to disable relocation on "
438 "exceptions: %ld\n", rc
);
441 default_machine_kexec(image
);
445 #ifdef __LITTLE_ENDIAN__
446 long pseries_big_endian_exceptions(void)
451 rc
= enable_big_endian_exceptions();
452 if (!H_IS_LONG_BUSY(rc
))
454 mdelay(get_longbusy_msecs(rc
));
458 static long pseries_little_endian_exceptions(void)
463 rc
= enable_little_endian_exceptions();
464 if (!H_IS_LONG_BUSY(rc
))
466 mdelay(get_longbusy_msecs(rc
));
471 static void __init
pSeries_setup_arch(void)
473 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT
);
475 /* Discover PIC type and setup ppc_md accordingly */
476 pseries_discover_pic();
478 /* openpic global configuration register (64-bit format). */
479 /* openpic Interrupt Source Unit pointer (64-bit format). */
480 /* python0 facility area (mmio) (64-bit format) REAL address. */
482 /* init to some ~sane value until calibrate_delay() runs */
483 loops_per_jiffy
= 50000000;
487 /* By default, only probe PCI (can be overriden by rtas_pci) */
488 pci_add_flags(PCI_PROBE_ONLY
);
490 /* Find and initialize PCI host bridges */
491 init_pci_config_tokens();
492 find_and_init_phbs();
493 of_reconfig_notifier_register(&pci_dn_reconfig_nb
);
495 pSeries_nvram_init();
497 if (firmware_has_feature(FW_FEATURE_LPAR
)) {
498 vpa_init(boot_cpuid
);
499 ppc_md
.power_save
= pseries_lpar_idle
;
500 ppc_md
.enable_pmcs
= pseries_lpar_enable_pmcs
;
502 /* No special idle routine */
503 ppc_md
.enable_pmcs
= power4_enable_pmcs
;
506 ppc_md
.pcibios_root_bridge_prepare
= pseries_root_bridge_prepare
;
508 if (firmware_has_feature(FW_FEATURE_SET_MODE
)) {
510 if ((rc
= pSeries_enable_reloc_on_exc()) != H_SUCCESS
) {
511 pr_warn("Unable to enable relocation on exceptions: "
517 static int __init
pSeries_init_panel(void)
519 /* Manually leave the kernel version on the panel. */
520 ppc_md
.progress("Linux ppc64\n", 0);
521 ppc_md
.progress(init_utsname()->version
, 0);
525 machine_arch_initcall(pseries
, pSeries_init_panel
);
527 static int pseries_set_dabr(unsigned long dabr
, unsigned long dabrx
)
529 return plpar_hcall_norets(H_SET_DABR
, dabr
);
532 static int pseries_set_xdabr(unsigned long dabr
, unsigned long dabrx
)
534 /* Have to set at least one bit in the DABRX according to PAPR */
535 if (dabrx
== 0 && dabr
== 0)
537 /* PAPR says we can only set kernel and user bits */
538 dabrx
&= DABRX_KERNEL
| DABRX_USER
;
540 return plpar_hcall_norets(H_SET_XDABR
, dabr
, dabrx
);
543 static int pseries_set_dawr(unsigned long dawr
, unsigned long dawrx
)
545 /* PAPR says we can't set HYP */
548 return plapr_set_watchpoint0(dawr
, dawrx
);
551 #define CMO_CHARACTERISTICS_TOKEN 44
552 #define CMO_MAXLENGTH 1026
554 void pSeries_coalesce_init(void)
556 struct hvcall_mpp_x_data mpp_x_data
;
558 if (firmware_has_feature(FW_FEATURE_CMO
) && !h_get_mpp_x(&mpp_x_data
))
559 powerpc_firmware_features
|= FW_FEATURE_XCMO
;
561 powerpc_firmware_features
&= ~FW_FEATURE_XCMO
;
565 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
566 * handle that here. (Stolen from parse_system_parameter_string)
568 void pSeries_cmo_feature_init(void)
570 char *ptr
, *key
, *value
, *end
;
572 int page_order
= IOMMU_PAGE_SHIFT_4K
;
574 pr_debug(" -> fw_cmo_feature_init()\n");
575 spin_lock(&rtas_data_buf_lock
);
576 memset(rtas_data_buf
, 0, RTAS_DATA_BUF_SIZE
);
577 call_status
= rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
579 CMO_CHARACTERISTICS_TOKEN
,
583 if (call_status
!= 0) {
584 spin_unlock(&rtas_data_buf_lock
);
585 pr_debug("CMO not available\n");
586 pr_debug(" <- fw_cmo_feature_init()\n");
590 end
= rtas_data_buf
+ CMO_MAXLENGTH
- 2;
591 ptr
= rtas_data_buf
+ 2; /* step over strlen value */
594 while (*ptr
&& (ptr
<= end
)) {
595 /* Separate the key and value by replacing '=' with '\0' and
596 * point the value at the string after the '='
601 } else if (ptr
[0] == '\0' || ptr
[0] == ',') {
602 /* Terminate the string containing the key/value pair */
606 pr_debug("Malformed key/value pair\n");
607 /* Never found a '=', end processing */
611 if (0 == strcmp(key
, "CMOPageSize"))
612 page_order
= simple_strtol(value
, NULL
, 10);
613 else if (0 == strcmp(key
, "PrPSP"))
614 CMO_PrPSP
= simple_strtol(value
, NULL
, 10);
615 else if (0 == strcmp(key
, "SecPSP"))
616 CMO_SecPSP
= simple_strtol(value
, NULL
, 10);
617 value
= key
= ptr
+ 1;
622 /* Page size is returned as the power of 2 of the page size,
623 * convert to the page size in bytes before returning
625 CMO_PageSize
= 1 << page_order
;
626 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize
);
628 if (CMO_PrPSP
!= -1 || CMO_SecPSP
!= -1) {
629 pr_info("CMO enabled\n");
630 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP
,
632 powerpc_firmware_features
|= FW_FEATURE_CMO
;
633 pSeries_coalesce_init();
635 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP
,
637 spin_unlock(&rtas_data_buf_lock
);
638 pr_debug(" <- fw_cmo_feature_init()\n");
642 * Early initialization. Relocation is on but do not reference unbolted pages
644 static void __init
pSeries_init_early(void)
646 pr_debug(" -> pSeries_init_early()\n");
648 #ifdef CONFIG_HVC_CONSOLE
649 if (firmware_has_feature(FW_FEATURE_LPAR
))
650 hvc_vio_init_early();
652 if (firmware_has_feature(FW_FEATURE_XDABR
))
653 ppc_md
.set_dabr
= pseries_set_xdabr
;
654 else if (firmware_has_feature(FW_FEATURE_DABR
))
655 ppc_md
.set_dabr
= pseries_set_dabr
;
657 if (firmware_has_feature(FW_FEATURE_SET_MODE
))
658 ppc_md
.set_dawr
= pseries_set_dawr
;
660 pSeries_cmo_feature_init();
661 iommu_init_early_pSeries();
663 pr_debug(" <- pSeries_init_early()\n");
667 * Called very early, MMU is off, device-tree isn't unflattened
670 static int __init
pseries_probe_fw_features(unsigned long node
,
671 const char *uname
, int depth
,
676 static int hypertas_found
;
677 static int vec5_found
;
682 if (!strcmp(uname
, "rtas") || !strcmp(uname
, "rtas@0")) {
683 prop
= of_get_flat_dt_prop(node
, "ibm,hypertas-functions",
686 powerpc_firmware_features
|= FW_FEATURE_LPAR
;
687 fw_hypertas_feature_init(prop
, len
);
693 if (!strcmp(uname
, "chosen")) {
694 prop
= of_get_flat_dt_prop(node
, "ibm,architecture-vec-5",
697 fw_vec5_feature_init(prop
, len
);
702 return hypertas_found
&& vec5_found
;
705 static int __init
pSeries_probe(void)
707 unsigned long root
= of_get_flat_dt_root();
708 char *dtype
= of_get_flat_dt_prop(root
, "device_type", NULL
);
712 if (strcmp(dtype
, "chrp"))
715 /* Cell blades firmware claims to be chrp while it's not. Until this
716 * is fixed, we need to avoid those here.
718 if (of_flat_dt_is_compatible(root
, "IBM,CPBW-1.0") ||
719 of_flat_dt_is_compatible(root
, "IBM,CBEA"))
722 pr_debug("pSeries detected, looking for LPAR capability...\n");
724 /* Now try to figure out if we are running on LPAR */
725 of_scan_flat_dt(pseries_probe_fw_features
, NULL
);
727 #ifdef __LITTLE_ENDIAN__
728 if (firmware_has_feature(FW_FEATURE_SET_MODE
)) {
731 * Tell the hypervisor that we want our exceptions to
732 * be taken in little endian mode. If this fails we don't
733 * want to use BUG() because it will trigger an exception.
735 rc
= pseries_little_endian_exceptions();
737 ppc_md
.progress("H_SET_MODE LE exception fail", 0);
738 panic("Could not enable little endian exceptions");
743 if (firmware_has_feature(FW_FEATURE_LPAR
))
748 pr_debug("Machine is%s LPAR !\n",
749 (powerpc_firmware_features
& FW_FEATURE_LPAR
) ? "" : " not");
754 static int pSeries_pci_probe_mode(struct pci_bus
*bus
)
756 if (firmware_has_feature(FW_FEATURE_LPAR
))
757 return PCI_PROBE_DEVTREE
;
758 return PCI_PROBE_NORMAL
;
762 * pSeries_power_off - tell firmware about how to power off the system.
764 * This function calls either the power-off rtas token in normal cases
765 * or the ibm,power-off-ups token (if present & requested) in case of
766 * a power failure. If power-off token is used, power on will only be
767 * possible with power button press. If ibm,power-off-ups token is used
768 * it will allow auto poweron after power is restored.
770 static void pSeries_power_off(void)
773 int rtas_poweroff_ups_token
= rtas_token("ibm,power-off-ups");
775 if (rtas_flash_term_hook
)
776 rtas_flash_term_hook(SYS_POWER_OFF
);
778 if (rtas_poweron_auto
== 0 ||
779 rtas_poweroff_ups_token
== RTAS_UNKNOWN_SERVICE
) {
780 rc
= rtas_call(rtas_token("power-off"), 2, 1, NULL
, -1, -1);
781 printk(KERN_INFO
"RTAS power-off returned %d\n", rc
);
783 rc
= rtas_call(rtas_poweroff_ups_token
, 0, 1, NULL
);
784 printk(KERN_INFO
"RTAS ibm,power-off-ups returned %d\n", rc
);
790 void pSeries_final_fixup(void) { }
793 define_machine(pseries
) {
795 .probe
= pSeries_probe
,
796 .setup_arch
= pSeries_setup_arch
,
797 .init_early
= pSeries_init_early
,
798 .show_cpuinfo
= pSeries_show_cpuinfo
,
799 .log_error
= pSeries_log_error
,
800 .pcibios_fixup
= pSeries_final_fixup
,
801 .pci_probe_mode
= pSeries_pci_probe_mode
,
802 .restart
= rtas_restart
,
803 .power_off
= pSeries_power_off
,
805 .panic
= rtas_os_term
,
806 .get_boot_time
= rtas_get_boot_time
,
807 .get_rtc_time
= rtas_get_rtc_time
,
808 .set_rtc_time
= rtas_set_rtc_time
,
809 .calibrate_decr
= generic_calibrate_decr
,
810 .progress
= rtas_progress
,
811 .system_reset_exception
= pSeries_system_reset_exception
,
812 .machine_check_exception
= pSeries_machine_check_exception
,
814 .machine_kexec
= pSeries_machine_kexec
,