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>
43 #include <linux/of_pci.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>
61 #include <asm/ppc-pci.h>
62 #include <asm/i8259.h>
65 #include <asm/firmware.h>
68 #include <asm/plpar_wrappers.h>
74 unsigned long CMO_PageSize
= (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K
);
75 EXPORT_SYMBOL(CMO_PageSize
);
77 int fwnmi_active
; /* TRUE if an FWNMI handler is present */
79 static void pSeries_show_cpuinfo(struct seq_file
*m
)
81 struct device_node
*root
;
82 const char *model
= "";
84 root
= of_find_node_by_path("/");
86 model
= of_get_property(root
, "model", NULL
);
87 seq_printf(m
, "machine\t\t: CHRP %s\n", model
);
91 /* Initialize firmware assisted non-maskable interrupts if
92 * the firmware supports this feature.
94 static void __init
fwnmi_init(void)
96 unsigned long system_reset_addr
, machine_check_addr
;
98 int ibm_nmi_register
= rtas_token("ibm,nmi-register");
99 if (ibm_nmi_register
== RTAS_UNKNOWN_SERVICE
)
102 /* If the kernel's not linked at zero we point the firmware at low
103 * addresses anyway, and use a trampoline to get to the real code. */
104 system_reset_addr
= __pa(system_reset_fwnmi
) - PHYSICAL_START
;
105 machine_check_addr
= __pa(machine_check_fwnmi
) - PHYSICAL_START
;
107 if (0 == rtas_call(ibm_nmi_register
, 2, 1, NULL
, system_reset_addr
,
112 static void pseries_8259_cascade(struct irq_desc
*desc
)
114 struct irq_chip
*chip
= irq_desc_get_chip(desc
);
115 unsigned int cascade_irq
= i8259_irq();
117 if (cascade_irq
!= NO_IRQ
)
118 generic_handle_irq(cascade_irq
);
120 chip
->irq_eoi(&desc
->irq_data
);
123 static void __init
pseries_setup_i8259_cascade(void)
125 struct device_node
*np
, *old
, *found
= NULL
;
126 unsigned int cascade
;
128 unsigned long intack
= 0;
131 for_each_node_by_type(np
, "interrupt-controller") {
132 if (of_device_is_compatible(np
, "chrp,iic")) {
139 printk(KERN_DEBUG
"pic: no ISA interrupt controller\n");
143 cascade
= irq_of_parse_and_map(found
, 0);
144 if (cascade
== NO_IRQ
) {
145 printk(KERN_ERR
"pic: failed to map cascade interrupt");
148 pr_debug("pic: cascade mapped to irq %d\n", cascade
);
150 for (old
= of_node_get(found
); old
!= NULL
; old
= np
) {
151 np
= of_get_parent(old
);
155 if (strcmp(np
->name
, "pci") != 0)
157 addrp
= of_get_property(np
, "8259-interrupt-acknowledge", NULL
);
160 naddr
= of_n_addr_cells(np
);
161 intack
= addrp
[naddr
-1];
163 intack
|= ((unsigned long)addrp
[naddr
-2]) << 32;
166 printk(KERN_DEBUG
"pic: PCI 8259 intack at 0x%016lx\n", intack
);
167 i8259_init(found
, intack
);
169 irq_set_chained_handler(cascade
, pseries_8259_cascade
);
172 static void __init
pseries_init_irq(void)
175 pseries_setup_i8259_cascade();
178 static void pseries_lpar_enable_pmcs(void)
180 unsigned long set
, reset
;
184 plpar_hcall_norets(H_PERFMON
, set
, reset
);
187 static int pci_dn_reconfig_notifier(struct notifier_block
*nb
, unsigned long action
, void *data
)
189 struct of_reconfig_data
*rd
= data
;
190 struct device_node
*parent
, *np
= rd
->dn
;
195 case OF_RECONFIG_ATTACH_NODE
:
196 parent
= of_get_parent(np
);
197 pdn
= parent
? PCI_DN(parent
) : NULL
;
199 /* Create pdn and EEH device */
200 pci_add_device_node_info(pdn
->phb
, np
);
201 eeh_dev_init(PCI_DN(np
), pdn
->phb
);
206 case OF_RECONFIG_DETACH_NODE
:
209 list_del(&pdn
->list
);
218 static struct notifier_block pci_dn_reconfig_nb
= {
219 .notifier_call
= pci_dn_reconfig_notifier
,
222 struct kmem_cache
*dtl_cache
;
224 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
226 * Allocate space for the dispatch trace log for all possible cpus
227 * and register the buffers with the hypervisor. This is used for
228 * computing time stolen by the hypervisor.
230 static int alloc_dispatch_logs(void)
233 struct paca_struct
*pp
;
234 struct dtl_entry
*dtl
;
236 if (!firmware_has_feature(FW_FEATURE_SPLPAR
))
242 for_each_possible_cpu(cpu
) {
244 dtl
= kmem_cache_alloc(dtl_cache
, GFP_KERNEL
);
246 pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
248 pr_warn("Stolen time statistics will be unreliable\n");
253 pp
->dispatch_log
= dtl
;
254 pp
->dispatch_log_end
= dtl
+ N_DISPATCH_LOG
;
258 /* Register the DTL for the current (boot) cpu */
259 dtl
= get_paca()->dispatch_log
;
260 get_paca()->dtl_ridx
= 0;
261 get_paca()->dtl_curr
= dtl
;
262 get_paca()->lppaca_ptr
->dtl_idx
= 0;
264 /* hypervisor reads buffer length from this field */
265 dtl
->enqueue_to_dispatch_time
= cpu_to_be32(DISPATCH_LOG_BYTES
);
266 ret
= register_dtl(hard_smp_processor_id(), __pa(dtl
));
268 pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
269 "with %d\n", smp_processor_id(),
270 hard_smp_processor_id(), ret
);
271 get_paca()->lppaca_ptr
->dtl_enable_mask
= 2;
275 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
276 static inline int alloc_dispatch_logs(void)
280 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
282 static int alloc_dispatch_log_kmem_cache(void)
284 dtl_cache
= kmem_cache_create("dtl", DISPATCH_LOG_BYTES
,
285 DISPATCH_LOG_BYTES
, 0, NULL
);
287 pr_warn("Failed to create dispatch trace log buffer cache\n");
288 pr_warn("Stolen time statistics will be unreliable\n");
292 return alloc_dispatch_logs();
294 machine_early_initcall(pseries
, alloc_dispatch_log_kmem_cache
);
296 static void pseries_lpar_idle(void)
299 * Default handler to go into low thread priority and possibly
300 * low power mode by ceding processor to hypervisor
303 /* Indicate to hypervisor that we are idle. */
304 get_lppaca()->idle
= 1;
307 * Yield the processor to the hypervisor. We return if
308 * an external interrupt occurs (which are driven prior
309 * to returning here) or if a prod occurs from another
310 * processor. When returning here, external interrupts
315 get_lppaca()->idle
= 0;
319 * Enable relocation on during exceptions. This has partition wide scope and
320 * may take a while to complete, if it takes longer than one second we will
321 * just give up rather than wasting any more time on this - if that turns out
322 * to ever be a problem in practice we can move this into a kernel thread to
323 * finish off the process later in boot.
325 long pSeries_enable_reloc_on_exc(void)
328 unsigned int delay
, total_delay
= 0;
331 rc
= enable_reloc_on_exceptions();
332 if (!H_IS_LONG_BUSY(rc
))
335 delay
= get_longbusy_msecs(rc
);
336 total_delay
+= delay
;
337 if (total_delay
> 1000) {
338 pr_warn("Warning: Giving up waiting to enable "
339 "relocation on exceptions (%u msec)!\n",
347 EXPORT_SYMBOL(pSeries_enable_reloc_on_exc
);
349 long pSeries_disable_reloc_on_exc(void)
354 rc
= disable_reloc_on_exceptions();
355 if (!H_IS_LONG_BUSY(rc
))
357 mdelay(get_longbusy_msecs(rc
));
360 EXPORT_SYMBOL(pSeries_disable_reloc_on_exc
);
363 static void pSeries_machine_kexec(struct kimage
*image
)
367 if (firmware_has_feature(FW_FEATURE_SET_MODE
)) {
368 rc
= pSeries_disable_reloc_on_exc();
370 pr_warning("Warning: Failed to disable relocation on "
371 "exceptions: %ld\n", rc
);
374 default_machine_kexec(image
);
378 #ifdef __LITTLE_ENDIAN__
379 long pseries_big_endian_exceptions(void)
384 rc
= enable_big_endian_exceptions();
385 if (!H_IS_LONG_BUSY(rc
))
387 mdelay(get_longbusy_msecs(rc
));
391 static long pseries_little_endian_exceptions(void)
396 rc
= enable_little_endian_exceptions();
397 if (!H_IS_LONG_BUSY(rc
))
399 mdelay(get_longbusy_msecs(rc
));
404 static void __init
find_and_init_phbs(void)
406 struct device_node
*node
;
407 struct pci_controller
*phb
;
408 struct device_node
*root
= of_find_node_by_path("/");
410 for_each_child_of_node(root
, node
) {
411 if (node
->type
== NULL
|| (strcmp(node
->type
, "pci") != 0 &&
412 strcmp(node
->type
, "pciex") != 0))
415 phb
= pcibios_alloc_controller(node
);
419 pci_process_bridge_OF_ranges(phb
, node
, 0);
420 isa_bridge_find_early(phb
);
421 phb
->controller_ops
= pseries_pci_controller_ops
;
428 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
431 of_pci_check_probe_only();
434 static void __init
pSeries_setup_arch(void)
436 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT
);
438 /* Discover PIC type and setup ppc_md accordingly */
439 setup_kexec_cpu_down_xics();
440 smp_init_pseries_xics();
443 /* openpic global configuration register (64-bit format). */
444 /* openpic Interrupt Source Unit pointer (64-bit format). */
445 /* python0 facility area (mmio) (64-bit format) REAL address. */
447 /* init to some ~sane value until calibrate_delay() runs */
448 loops_per_jiffy
= 50000000;
452 /* By default, only probe PCI (can be overridden by rtas_pci) */
453 pci_add_flags(PCI_PROBE_ONLY
);
455 /* Find and initialize PCI host bridges */
456 init_pci_config_tokens();
457 find_and_init_phbs();
458 of_reconfig_notifier_register(&pci_dn_reconfig_nb
);
460 pSeries_nvram_init();
462 if (firmware_has_feature(FW_FEATURE_LPAR
)) {
463 vpa_init(boot_cpuid
);
464 ppc_md
.power_save
= pseries_lpar_idle
;
465 ppc_md
.enable_pmcs
= pseries_lpar_enable_pmcs
;
467 /* No special idle routine */
468 ppc_md
.enable_pmcs
= power4_enable_pmcs
;
471 ppc_md
.pcibios_root_bridge_prepare
= pseries_root_bridge_prepare
;
473 if (firmware_has_feature(FW_FEATURE_SET_MODE
)) {
476 rc
= pSeries_enable_reloc_on_exc();
478 pr_info("Relocation on exceptions not supported\n");
479 } else if (rc
!= H_SUCCESS
) {
480 pr_warn("Unable to enable relocation on exceptions: "
486 static int __init
pSeries_init_panel(void)
488 /* Manually leave the kernel version on the panel. */
489 #ifdef __BIG_ENDIAN__
490 ppc_md
.progress("Linux ppc64\n", 0);
492 ppc_md
.progress("Linux ppc64le\n", 0);
494 ppc_md
.progress(init_utsname()->version
, 0);
498 machine_arch_initcall(pseries
, pSeries_init_panel
);
500 static int pseries_set_dabr(unsigned long dabr
, unsigned long dabrx
)
502 return plpar_hcall_norets(H_SET_DABR
, dabr
);
505 static int pseries_set_xdabr(unsigned long dabr
, unsigned long dabrx
)
507 /* Have to set at least one bit in the DABRX according to PAPR */
508 if (dabrx
== 0 && dabr
== 0)
510 /* PAPR says we can only set kernel and user bits */
511 dabrx
&= DABRX_KERNEL
| DABRX_USER
;
513 return plpar_hcall_norets(H_SET_XDABR
, dabr
, dabrx
);
516 static int pseries_set_dawr(unsigned long dawr
, unsigned long dawrx
)
518 /* PAPR says we can't set HYP */
521 return plapr_set_watchpoint0(dawr
, dawrx
);
524 #define CMO_CHARACTERISTICS_TOKEN 44
525 #define CMO_MAXLENGTH 1026
527 void pSeries_coalesce_init(void)
529 struct hvcall_mpp_x_data mpp_x_data
;
531 if (firmware_has_feature(FW_FEATURE_CMO
) && !h_get_mpp_x(&mpp_x_data
))
532 powerpc_firmware_features
|= FW_FEATURE_XCMO
;
534 powerpc_firmware_features
&= ~FW_FEATURE_XCMO
;
538 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
539 * handle that here. (Stolen from parse_system_parameter_string)
541 static void pSeries_cmo_feature_init(void)
543 char *ptr
, *key
, *value
, *end
;
545 int page_order
= IOMMU_PAGE_SHIFT_4K
;
547 pr_debug(" -> fw_cmo_feature_init()\n");
548 spin_lock(&rtas_data_buf_lock
);
549 memset(rtas_data_buf
, 0, RTAS_DATA_BUF_SIZE
);
550 call_status
= rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
552 CMO_CHARACTERISTICS_TOKEN
,
556 if (call_status
!= 0) {
557 spin_unlock(&rtas_data_buf_lock
);
558 pr_debug("CMO not available\n");
559 pr_debug(" <- fw_cmo_feature_init()\n");
563 end
= rtas_data_buf
+ CMO_MAXLENGTH
- 2;
564 ptr
= rtas_data_buf
+ 2; /* step over strlen value */
567 while (*ptr
&& (ptr
<= end
)) {
568 /* Separate the key and value by replacing '=' with '\0' and
569 * point the value at the string after the '='
574 } else if (ptr
[0] == '\0' || ptr
[0] == ',') {
575 /* Terminate the string containing the key/value pair */
579 pr_debug("Malformed key/value pair\n");
580 /* Never found a '=', end processing */
584 if (0 == strcmp(key
, "CMOPageSize"))
585 page_order
= simple_strtol(value
, NULL
, 10);
586 else if (0 == strcmp(key
, "PrPSP"))
587 CMO_PrPSP
= simple_strtol(value
, NULL
, 10);
588 else if (0 == strcmp(key
, "SecPSP"))
589 CMO_SecPSP
= simple_strtol(value
, NULL
, 10);
590 value
= key
= ptr
+ 1;
595 /* Page size is returned as the power of 2 of the page size,
596 * convert to the page size in bytes before returning
598 CMO_PageSize
= 1 << page_order
;
599 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize
);
601 if (CMO_PrPSP
!= -1 || CMO_SecPSP
!= -1) {
602 pr_info("CMO enabled\n");
603 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP
,
605 powerpc_firmware_features
|= FW_FEATURE_CMO
;
606 pSeries_coalesce_init();
608 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP
,
610 spin_unlock(&rtas_data_buf_lock
);
611 pr_debug(" <- fw_cmo_feature_init()\n");
615 * Early initialization. Relocation is on but do not reference unbolted pages
617 static void __init
pSeries_init_early(void)
619 pr_debug(" -> pSeries_init_early()\n");
621 #ifdef CONFIG_HVC_CONSOLE
622 if (firmware_has_feature(FW_FEATURE_LPAR
))
623 hvc_vio_init_early();
625 if (firmware_has_feature(FW_FEATURE_XDABR
))
626 ppc_md
.set_dabr
= pseries_set_xdabr
;
627 else if (firmware_has_feature(FW_FEATURE_DABR
))
628 ppc_md
.set_dabr
= pseries_set_dabr
;
630 if (firmware_has_feature(FW_FEATURE_SET_MODE
))
631 ppc_md
.set_dawr
= pseries_set_dawr
;
633 pSeries_cmo_feature_init();
634 iommu_init_early_pSeries();
636 pr_debug(" <- pSeries_init_early()\n");
640 * pseries_power_off - tell firmware about how to power off the system.
642 * This function calls either the power-off rtas token in normal cases
643 * or the ibm,power-off-ups token (if present & requested) in case of
644 * a power failure. If power-off token is used, power on will only be
645 * possible with power button press. If ibm,power-off-ups token is used
646 * it will allow auto poweron after power is restored.
648 static void pseries_power_off(void)
651 int rtas_poweroff_ups_token
= rtas_token("ibm,power-off-ups");
653 if (rtas_flash_term_hook
)
654 rtas_flash_term_hook(SYS_POWER_OFF
);
656 if (rtas_poweron_auto
== 0 ||
657 rtas_poweroff_ups_token
== RTAS_UNKNOWN_SERVICE
) {
658 rc
= rtas_call(rtas_token("power-off"), 2, 1, NULL
, -1, -1);
659 printk(KERN_INFO
"RTAS power-off returned %d\n", rc
);
661 rc
= rtas_call(rtas_poweroff_ups_token
, 0, 1, NULL
);
662 printk(KERN_INFO
"RTAS ibm,power-off-ups returned %d\n", rc
);
668 * Called very early, MMU is off, device-tree isn't unflattened
671 static int __init
pseries_probe_fw_features(unsigned long node
,
672 const char *uname
, int depth
,
677 static int hypertas_found
;
678 static int vec5_found
;
683 if (!strcmp(uname
, "rtas") || !strcmp(uname
, "rtas@0")) {
684 prop
= of_get_flat_dt_prop(node
, "ibm,hypertas-functions",
687 powerpc_firmware_features
|= FW_FEATURE_LPAR
;
688 fw_hypertas_feature_init(prop
, len
);
694 if (!strcmp(uname
, "chosen")) {
695 prop
= of_get_flat_dt_prop(node
, "ibm,architecture-vec-5",
698 fw_vec5_feature_init(prop
, len
);
703 return hypertas_found
&& vec5_found
;
706 static int __init
pSeries_probe(void)
708 unsigned long root
= of_get_flat_dt_root();
709 const char *dtype
= of_get_flat_dt_prop(root
, "device_type", NULL
);
713 if (strcmp(dtype
, "chrp"))
716 /* Cell blades firmware claims to be chrp while it's not. Until this
717 * is fixed, we need to avoid those here.
719 if (of_flat_dt_is_compatible(root
, "IBM,CPBW-1.0") ||
720 of_flat_dt_is_compatible(root
, "IBM,CBEA"))
723 pr_debug("pSeries detected, looking for LPAR capability...\n");
725 /* Now try to figure out if we are running on LPAR */
726 of_scan_flat_dt(pseries_probe_fw_features
, NULL
);
728 #ifdef __LITTLE_ENDIAN__
729 if (firmware_has_feature(FW_FEATURE_SET_MODE
)) {
732 * Tell the hypervisor that we want our exceptions to
733 * be taken in little endian mode. If this fails we don't
734 * want to use BUG() because it will trigger an exception.
736 rc
= pseries_little_endian_exceptions();
738 ppc_md
.progress("H_SET_MODE LE exception fail", 0);
739 panic("Could not enable little endian exceptions");
744 if (firmware_has_feature(FW_FEATURE_LPAR
))
749 pm_power_off
= pseries_power_off
;
751 pr_debug("Machine is%s LPAR !\n",
752 (powerpc_firmware_features
& FW_FEATURE_LPAR
) ? "" : " not");
757 static int pSeries_pci_probe_mode(struct pci_bus
*bus
)
759 if (firmware_has_feature(FW_FEATURE_LPAR
))
760 return PCI_PROBE_DEVTREE
;
761 return PCI_PROBE_NORMAL
;
764 struct pci_controller_ops pseries_pci_controller_ops
= {
765 .probe_mode
= pSeries_pci_probe_mode
,
768 define_machine(pseries
) {
770 .probe
= pSeries_probe
,
771 .setup_arch
= pSeries_setup_arch
,
772 .init_early
= pSeries_init_early
,
773 .init_IRQ
= pseries_init_irq
,
774 .show_cpuinfo
= pSeries_show_cpuinfo
,
775 .log_error
= pSeries_log_error
,
776 .pcibios_fixup
= pSeries_final_fixup
,
777 .restart
= rtas_restart
,
779 .panic
= rtas_os_term
,
780 .get_boot_time
= rtas_get_boot_time
,
781 .get_rtc_time
= rtas_get_rtc_time
,
782 .set_rtc_time
= rtas_set_rtc_time
,
783 .calibrate_decr
= generic_calibrate_decr
,
784 .progress
= rtas_progress
,
785 .system_reset_exception
= pSeries_system_reset_exception
,
786 .machine_check_exception
= pSeries_machine_check_exception
,
788 .machine_kexec
= pSeries_machine_kexec
,
790 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
791 .memory_block_size
= pseries_memory_block_size
,