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>
46 #include <asm/processor.h>
48 #include <asm/pgtable.h>
51 #include <asm/pci-bridge.h>
52 #include <asm/iommu.h>
54 #include <asm/machdep.h>
57 #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>
69 #include <asm/kexec.h>
70 #include <asm/isa-bridge.h>
71 #include <asm/security_features.h>
74 #include "../../../../drivers/pci/pci.h"
78 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);
79 EXPORT_SYMBOL(CMO_PageSize);
81 int fwnmi_active; /* TRUE if an FWNMI handler is present */
83 static void pSeries_show_cpuinfo(struct seq_file *m)
85 struct device_node *root;
86 const char *model = "";
88 root = of_find_node_by_path("/");
90 model = of_get_property(root, "model", NULL);
91 seq_printf(m, "machine\t\t: CHRP %s\n", model);
94 seq_printf(m, "MMU\t\t: Radix\n");
96 seq_printf(m, "MMU\t\t: Hash\n");
99 /* Initialize firmware assisted non-maskable interrupts if
100 * the firmware supports this feature.
102 static void __init fwnmi_init(void)
104 unsigned long system_reset_addr, machine_check_addr;
106 int ibm_nmi_register = rtas_token("ibm,nmi-register");
107 if (ibm_nmi_register == RTAS_UNKNOWN_SERVICE)
110 /* If the kernel's not linked at zero we point the firmware at low
111 * addresses anyway, and use a trampoline to get to the real code. */
112 system_reset_addr = __pa(system_reset_fwnmi) - PHYSICAL_START;
113 machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START;
115 if (0 == rtas_call(ibm_nmi_register, 2, 1, NULL, system_reset_addr,
120 static void pseries_8259_cascade(struct irq_desc *desc)
122 struct irq_chip *chip = irq_desc_get_chip(desc);
123 unsigned int cascade_irq = i8259_irq();
126 generic_handle_irq(cascade_irq);
128 chip->irq_eoi(&desc->irq_data);
131 static void __init pseries_setup_i8259_cascade(void)
133 struct device_node *np, *old, *found = NULL;
134 unsigned int cascade;
136 unsigned long intack = 0;
139 for_each_node_by_type(np, "interrupt-controller") {
140 if (of_device_is_compatible(np, "chrp,iic")) {
147 printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
151 cascade = irq_of_parse_and_map(found, 0);
153 printk(KERN_ERR "pic: failed to map cascade interrupt");
156 pr_debug("pic: cascade mapped to irq %d\n", cascade);
158 for (old = of_node_get(found); old != NULL ; old = np) {
159 np = of_get_parent(old);
163 if (strcmp(np->name, "pci") != 0)
165 addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
168 naddr = of_n_addr_cells(np);
169 intack = addrp[naddr-1];
171 intack |= ((unsigned long)addrp[naddr-2]) << 32;
174 printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
175 i8259_init(found, intack);
177 irq_set_chained_handler(cascade, pseries_8259_cascade);
180 static void __init pseries_init_irq(void)
182 /* Try using a XIVE if available, otherwise use a XICS */
183 if (!xive_spapr_init()) {
185 pseries_setup_i8259_cascade();
189 static void pseries_lpar_enable_pmcs(void)
191 unsigned long set, reset;
195 plpar_hcall_norets(H_PERFMON, set, reset);
198 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
200 struct of_reconfig_data *rd = data;
201 struct device_node *parent, *np = rd->dn;
206 case OF_RECONFIG_ATTACH_NODE:
207 parent = of_get_parent(np);
208 pdn = parent ? PCI_DN(parent) : NULL;
210 pci_add_device_node_info(pdn->phb, np);
214 case OF_RECONFIG_DETACH_NODE:
217 list_del(&pdn->list);
226 static struct notifier_block pci_dn_reconfig_nb = {
227 .notifier_call = pci_dn_reconfig_notifier,
230 struct kmem_cache *dtl_cache;
232 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
234 * Allocate space for the dispatch trace log for all possible cpus
235 * and register the buffers with the hypervisor. This is used for
236 * computing time stolen by the hypervisor.
238 static int alloc_dispatch_logs(void)
241 struct paca_struct *pp;
242 struct dtl_entry *dtl;
244 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
250 for_each_possible_cpu(cpu) {
252 dtl = kmem_cache_alloc(dtl_cache, GFP_KERNEL);
254 pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
256 pr_warn("Stolen time statistics will be unreliable\n");
261 pp->dispatch_log = dtl;
262 pp->dispatch_log_end = dtl + N_DISPATCH_LOG;
266 /* Register the DTL for the current (boot) cpu */
267 dtl = get_paca()->dispatch_log;
268 get_paca()->dtl_ridx = 0;
269 get_paca()->dtl_curr = dtl;
270 get_paca()->lppaca_ptr->dtl_idx = 0;
272 /* hypervisor reads buffer length from this field */
273 dtl->enqueue_to_dispatch_time = cpu_to_be32(DISPATCH_LOG_BYTES);
274 ret = register_dtl(hard_smp_processor_id(), __pa(dtl));
276 pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
277 "with %d\n", smp_processor_id(),
278 hard_smp_processor_id(), ret);
279 get_paca()->lppaca_ptr->dtl_enable_mask = 2;
283 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
284 static inline int alloc_dispatch_logs(void)
288 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
290 static int alloc_dispatch_log_kmem_cache(void)
292 dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
293 DISPATCH_LOG_BYTES, 0, NULL);
295 pr_warn("Failed to create dispatch trace log buffer cache\n");
296 pr_warn("Stolen time statistics will be unreliable\n");
300 return alloc_dispatch_logs();
302 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
304 static void pseries_lpar_idle(void)
307 * Default handler to go into low thread priority and possibly
308 * low power mode by ceding processor to hypervisor
311 /* Indicate to hypervisor that we are idle. */
312 get_lppaca()->idle = 1;
315 * Yield the processor to the hypervisor. We return if
316 * an external interrupt occurs (which are driven prior
317 * to returning here) or if a prod occurs from another
318 * processor. When returning here, external interrupts
323 get_lppaca()->idle = 0;
327 * Enable relocation on during exceptions. This has partition wide scope and
328 * may take a while to complete, if it takes longer than one second we will
329 * just give up rather than wasting any more time on this - if that turns out
330 * to ever be a problem in practice we can move this into a kernel thread to
331 * finish off the process later in boot.
333 void pseries_enable_reloc_on_exc(void)
336 unsigned int delay, total_delay = 0;
339 rc = enable_reloc_on_exceptions();
340 if (!H_IS_LONG_BUSY(rc)) {
342 pr_info("Relocation on exceptions not"
344 } else if (rc != H_SUCCESS) {
345 pr_warn("Unable to enable relocation"
346 " on exceptions: %ld\n", rc);
351 delay = get_longbusy_msecs(rc);
352 total_delay += delay;
353 if (total_delay > 1000) {
354 pr_warn("Warning: Giving up waiting to enable "
355 "relocation on exceptions (%u msec)!\n",
363 EXPORT_SYMBOL(pseries_enable_reloc_on_exc);
365 void pseries_disable_reloc_on_exc(void)
370 rc = disable_reloc_on_exceptions();
371 if (!H_IS_LONG_BUSY(rc))
373 mdelay(get_longbusy_msecs(rc));
376 pr_warn("Warning: Failed to disable relocation on exceptions: %ld\n",
379 EXPORT_SYMBOL(pseries_disable_reloc_on_exc);
381 #ifdef CONFIG_KEXEC_CORE
382 static void pSeries_machine_kexec(struct kimage *image)
384 if (firmware_has_feature(FW_FEATURE_SET_MODE))
385 pseries_disable_reloc_on_exc();
387 default_machine_kexec(image);
391 #ifdef __LITTLE_ENDIAN__
392 void pseries_big_endian_exceptions(void)
397 rc = enable_big_endian_exceptions();
398 if (!H_IS_LONG_BUSY(rc))
400 mdelay(get_longbusy_msecs(rc));
404 * At this point it is unlikely panic() will get anything
405 * out to the user, since this is called very late in kexec
406 * but at least this will stop us from continuing on further
407 * and creating an even more difficult to debug situation.
409 * There is a known problem when kdump'ing, if cpus are offline
410 * the above call will fail. Rather than panicking again, keep
411 * going and hope the kdump kernel is also little endian, which
414 if (rc && !kdump_in_progress())
415 panic("Could not enable big endian exceptions");
418 void pseries_little_endian_exceptions(void)
423 rc = enable_little_endian_exceptions();
424 if (!H_IS_LONG_BUSY(rc))
426 mdelay(get_longbusy_msecs(rc));
429 ppc_md.progress("H_SET_MODE LE exception fail", 0);
430 panic("Could not enable little endian exceptions");
435 static void __init find_and_init_phbs(void)
437 struct device_node *node;
438 struct pci_controller *phb;
439 struct device_node *root = of_find_node_by_path("/");
441 for_each_child_of_node(root, node) {
442 if (node->type == NULL || (strcmp(node->type, "pci") != 0 &&
443 strcmp(node->type, "pciex") != 0))
446 phb = pcibios_alloc_controller(node);
450 pci_process_bridge_OF_ranges(phb, node, 0);
451 isa_bridge_find_early(phb);
452 phb->controller_ops = pseries_pci_controller_ops;
458 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
461 of_pci_check_probe_only();
464 static void init_cpu_char_feature_flags(struct h_cpu_char_result *result)
467 * The features below are disabled by default, so we instead look to see
468 * if firmware has *enabled* them, and set them if so.
470 if (result->character & H_CPU_CHAR_SPEC_BAR_ORI31)
471 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31);
473 if (result->character & H_CPU_CHAR_BCCTRL_SERIALISED)
474 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED);
476 if (result->character & H_CPU_CHAR_L1D_FLUSH_ORI30)
477 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30);
479 if (result->character & H_CPU_CHAR_L1D_FLUSH_TRIG2)
480 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2);
482 if (result->character & H_CPU_CHAR_L1D_THREAD_PRIV)
483 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV);
485 if (result->character & H_CPU_CHAR_COUNT_CACHE_DISABLED)
486 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED);
489 * The features below are enabled by default, so we instead look to see
490 * if firmware has *disabled* them, and clear them if so.
492 if (!(result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY))
493 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY);
495 if (!(result->behaviour & H_CPU_BEHAV_L1D_FLUSH_PR))
496 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR);
498 if (!(result->behaviour & H_CPU_BEHAV_BNDS_CHK_SPEC_BAR))
499 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR);
502 void pseries_setup_rfi_flush(void)
504 struct h_cpu_char_result result;
505 enum l1d_flush_type types;
510 * Set features to the defaults assumed by init_cpu_char_feature_flags()
511 * so it can set/clear again any features that might have changed after
512 * migration, and in case the hypercall fails and it is not even called.
514 powerpc_security_features = SEC_FTR_DEFAULT;
516 rc = plpar_get_cpu_characteristics(&result);
518 init_cpu_char_feature_flags(&result);
521 * We're the guest so this doesn't apply to us, clear it to simplify
522 * handling of it elsewhere.
524 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV);
526 types = L1D_FLUSH_FALLBACK;
528 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2))
529 types |= L1D_FLUSH_MTTRIG;
531 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30))
532 types |= L1D_FLUSH_ORI;
534 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) && \
535 security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR);
537 setup_rfi_flush(types, enable);
538 setup_barrier_nospec();
541 #ifdef CONFIG_PCI_IOV
542 enum rtas_iov_fw_value_map {
543 NUM_RES_PROPERTY = 0, /* Number of Resources */
544 LOW_INT = 1, /* Lowest 32 bits of Address */
545 START_OF_ENTRIES = 2, /* Always start of entry */
546 APERTURE_PROPERTY = 2, /* Start of entry+ to Aperture Size */
547 WDW_SIZE_PROPERTY = 4, /* Start of entry+ to Window Size */
548 NEXT_ENTRY = 7 /* Go to next entry on array */
551 enum get_iov_fw_value_index {
552 BAR_ADDRS = 1, /* Get Bar Address */
553 APERTURE_SIZE = 2, /* Get Aperture Size */
554 WDW_SIZE = 3 /* Get Window Size */
557 resource_size_t pseries_get_iov_fw_value(struct pci_dev *dev, int resno,
558 enum get_iov_fw_value_index value)
561 struct device_node *dn = pci_device_to_OF_node(dev);
562 int i, num_res, ret = 0;
564 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
569 * First element in the array is the number of Bars
570 * returned. Search through the list to find the matching
573 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
574 if (resno >= num_res)
575 return 0; /* or an errror */
577 i = START_OF_ENTRIES + NEXT_ENTRY * resno;
580 ret = of_read_number(&indexes[i], 2);
583 ret = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
586 ret = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
593 void of_pci_set_vf_bar_size(struct pci_dev *dev, const int *indexes)
595 struct resource *res;
596 resource_size_t base, size;
599 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
600 num_res = min_t(int, num_res, PCI_SRIOV_NUM_BARS);
601 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
602 i += NEXT_ENTRY, r++) {
603 res = &dev->resource[r + PCI_IOV_RESOURCES];
604 base = of_read_number(&indexes[i], 2);
605 size = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
606 res->flags = pci_parse_of_flags(of_read_number
607 (&indexes[i + LOW_INT], 1), 0);
608 res->flags |= (IORESOURCE_MEM_64 | IORESOURCE_PCI_FIXED);
609 res->name = pci_name(dev);
611 res->end = base + size - 1;
615 void of_pci_parse_iov_addrs(struct pci_dev *dev, const int *indexes)
617 struct resource *res, *root, *conflict;
618 resource_size_t base, size;
622 * First element in the array is the number of Bars
623 * returned. Search through the list to find the matching
624 * bars assign them from firmware into resources structure.
626 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
627 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
628 i += NEXT_ENTRY, r++) {
629 res = &dev->resource[r + PCI_IOV_RESOURCES];
630 base = of_read_number(&indexes[i], 2);
631 size = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
632 res->name = pci_name(dev);
634 res->end = base + size - 1;
635 root = &iomem_resource;
637 "pSeries IOV BAR %d: trying firmware assignment %pR\n",
638 r + PCI_IOV_RESOURCES, res);
639 conflict = request_resource_conflict(root, res);
642 "BAR %d: %pR conflicts with %s %pR\n",
643 r + PCI_IOV_RESOURCES, res,
644 conflict->name, conflict);
645 res->flags |= IORESOURCE_UNSET;
650 static void pseries_pci_fixup_resources(struct pci_dev *pdev)
653 struct device_node *dn = pci_device_to_OF_node(pdev);
655 /*Firmware must support open sriov otherwise dont configure*/
656 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
659 /* Assign the addresses from device tree*/
660 of_pci_set_vf_bar_size(pdev, indexes);
663 static void pseries_pci_fixup_iov_resources(struct pci_dev *pdev)
666 struct device_node *dn = pci_device_to_OF_node(pdev);
668 if (!pdev->is_physfn || pci_dev_is_added(pdev))
670 /*Firmware must support open sriov otherwise dont configure*/
671 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
674 /* Assign the addresses from device tree*/
675 of_pci_parse_iov_addrs(pdev, indexes);
678 static resource_size_t pseries_pci_iov_resource_alignment(struct pci_dev *pdev,
682 struct device_node *dn = pci_device_to_OF_node(pdev);
684 /*Firmware must support open sriov otherwise report regular alignment*/
685 reg = of_get_property(dn, "ibm,is-open-sriov-pf", NULL);
687 return pci_iov_resource_size(pdev, resno);
689 if (!pdev->is_physfn)
691 return pseries_get_iov_fw_value(pdev,
692 resno - PCI_IOV_RESOURCES,
697 static void __init pSeries_setup_arch(void)
699 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
701 /* Discover PIC type and setup ppc_md accordingly */
705 /* openpic global configuration register (64-bit format). */
706 /* openpic Interrupt Source Unit pointer (64-bit format). */
707 /* python0 facility area (mmio) (64-bit format) REAL address. */
709 /* init to some ~sane value until calibrate_delay() runs */
710 loops_per_jiffy = 50000000;
714 pseries_setup_rfi_flush();
717 /* By default, only probe PCI (can be overridden by rtas_pci) */
718 pci_add_flags(PCI_PROBE_ONLY);
720 /* Find and initialize PCI host bridges */
721 init_pci_config_tokens();
722 find_and_init_phbs();
723 of_reconfig_notifier_register(&pci_dn_reconfig_nb);
725 pSeries_nvram_init();
727 if (firmware_has_feature(FW_FEATURE_LPAR)) {
728 vpa_init(boot_cpuid);
729 ppc_md.power_save = pseries_lpar_idle;
730 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
731 #ifdef CONFIG_PCI_IOV
732 ppc_md.pcibios_fixup_resources =
733 pseries_pci_fixup_resources;
734 ppc_md.pcibios_fixup_sriov =
735 pseries_pci_fixup_iov_resources;
736 ppc_md.pcibios_iov_resource_alignment =
737 pseries_pci_iov_resource_alignment;
740 /* No special idle routine */
741 ppc_md.enable_pmcs = power4_enable_pmcs;
744 ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
747 static void pseries_panic(char *str)
749 panic_flush_kmsg_end();
753 static int __init pSeries_init_panel(void)
755 /* Manually leave the kernel version on the panel. */
756 #ifdef __BIG_ENDIAN__
757 ppc_md.progress("Linux ppc64\n", 0);
759 ppc_md.progress("Linux ppc64le\n", 0);
761 ppc_md.progress(init_utsname()->version, 0);
765 machine_arch_initcall(pseries, pSeries_init_panel);
767 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
769 return plpar_hcall_norets(H_SET_DABR, dabr);
772 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
774 /* Have to set at least one bit in the DABRX according to PAPR */
775 if (dabrx == 0 && dabr == 0)
777 /* PAPR says we can only set kernel and user bits */
778 dabrx &= DABRX_KERNEL | DABRX_USER;
780 return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
783 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
785 /* PAPR says we can't set HYP */
788 return plpar_set_watchpoint0(dawr, dawrx);
791 #define CMO_CHARACTERISTICS_TOKEN 44
792 #define CMO_MAXLENGTH 1026
794 void pSeries_coalesce_init(void)
796 struct hvcall_mpp_x_data mpp_x_data;
798 if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
799 powerpc_firmware_features |= FW_FEATURE_XCMO;
801 powerpc_firmware_features &= ~FW_FEATURE_XCMO;
805 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
806 * handle that here. (Stolen from parse_system_parameter_string)
808 static void pSeries_cmo_feature_init(void)
810 char *ptr, *key, *value, *end;
812 int page_order = IOMMU_PAGE_SHIFT_4K;
814 pr_debug(" -> fw_cmo_feature_init()\n");
815 spin_lock(&rtas_data_buf_lock);
816 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
817 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
819 CMO_CHARACTERISTICS_TOKEN,
823 if (call_status != 0) {
824 spin_unlock(&rtas_data_buf_lock);
825 pr_debug("CMO not available\n");
826 pr_debug(" <- fw_cmo_feature_init()\n");
830 end = rtas_data_buf + CMO_MAXLENGTH - 2;
831 ptr = rtas_data_buf + 2; /* step over strlen value */
834 while (*ptr && (ptr <= end)) {
835 /* Separate the key and value by replacing '=' with '\0' and
836 * point the value at the string after the '='
841 } else if (ptr[0] == '\0' || ptr[0] == ',') {
842 /* Terminate the string containing the key/value pair */
846 pr_debug("Malformed key/value pair\n");
847 /* Never found a '=', end processing */
851 if (0 == strcmp(key, "CMOPageSize"))
852 page_order = simple_strtol(value, NULL, 10);
853 else if (0 == strcmp(key, "PrPSP"))
854 CMO_PrPSP = simple_strtol(value, NULL, 10);
855 else if (0 == strcmp(key, "SecPSP"))
856 CMO_SecPSP = simple_strtol(value, NULL, 10);
857 value = key = ptr + 1;
862 /* Page size is returned as the power of 2 of the page size,
863 * convert to the page size in bytes before returning
865 CMO_PageSize = 1 << page_order;
866 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
868 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
869 pr_info("CMO enabled\n");
870 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
872 powerpc_firmware_features |= FW_FEATURE_CMO;
873 pSeries_coalesce_init();
875 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
877 spin_unlock(&rtas_data_buf_lock);
878 pr_debug(" <- fw_cmo_feature_init()\n");
882 * Early initialization. Relocation is on but do not reference unbolted pages
884 static void __init pseries_init(void)
886 pr_debug(" -> pseries_init()\n");
888 #ifdef CONFIG_HVC_CONSOLE
889 if (firmware_has_feature(FW_FEATURE_LPAR))
890 hvc_vio_init_early();
892 if (firmware_has_feature(FW_FEATURE_XDABR))
893 ppc_md.set_dabr = pseries_set_xdabr;
894 else if (firmware_has_feature(FW_FEATURE_DABR))
895 ppc_md.set_dabr = pseries_set_dabr;
897 if (firmware_has_feature(FW_FEATURE_SET_MODE))
898 ppc_md.set_dawr = pseries_set_dawr;
900 pSeries_cmo_feature_init();
901 iommu_init_early_pSeries();
903 pr_debug(" <- pseries_init()\n");
907 * pseries_power_off - tell firmware about how to power off the system.
909 * This function calls either the power-off rtas token in normal cases
910 * or the ibm,power-off-ups token (if present & requested) in case of
911 * a power failure. If power-off token is used, power on will only be
912 * possible with power button press. If ibm,power-off-ups token is used
913 * it will allow auto poweron after power is restored.
915 static void pseries_power_off(void)
918 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
920 if (rtas_flash_term_hook)
921 rtas_flash_term_hook(SYS_POWER_OFF);
923 if (rtas_poweron_auto == 0 ||
924 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
925 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
926 printk(KERN_INFO "RTAS power-off returned %d\n", rc);
928 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
929 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
934 static int __init pSeries_probe(void)
936 const char *dtype = of_get_property(of_root, "device_type", NULL);
940 if (strcmp(dtype, "chrp"))
943 /* Cell blades firmware claims to be chrp while it's not. Until this
944 * is fixed, we need to avoid those here.
946 if (of_machine_is_compatible("IBM,CPBW-1.0") ||
947 of_machine_is_compatible("IBM,CBEA"))
950 pm_power_off = pseries_power_off;
952 pr_debug("Machine is%s LPAR !\n",
953 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
960 static int pSeries_pci_probe_mode(struct pci_bus *bus)
962 if (firmware_has_feature(FW_FEATURE_LPAR))
963 return PCI_PROBE_DEVTREE;
964 return PCI_PROBE_NORMAL;
967 struct pci_controller_ops pseries_pci_controller_ops = {
968 .probe_mode = pSeries_pci_probe_mode,
971 define_machine(pseries) {
973 .probe = pSeries_probe,
974 .setup_arch = pSeries_setup_arch,
975 .init_IRQ = pseries_init_irq,
976 .show_cpuinfo = pSeries_show_cpuinfo,
977 .log_error = pSeries_log_error,
978 .pcibios_fixup = pSeries_final_fixup,
979 .restart = rtas_restart,
981 .panic = pseries_panic,
982 .get_boot_time = rtas_get_boot_time,
983 .get_rtc_time = rtas_get_rtc_time,
984 .set_rtc_time = rtas_set_rtc_time,
985 .calibrate_decr = generic_calibrate_decr,
986 .progress = rtas_progress,
987 .system_reset_exception = pSeries_system_reset_exception,
988 .machine_check_exception = pSeries_machine_check_exception,
989 #ifdef CONFIG_KEXEC_CORE
990 .machine_kexec = pSeries_machine_kexec,
991 .kexec_cpu_down = pseries_kexec_cpu_down,
993 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
994 .memory_block_size = pseries_memory_block_size,