1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * 64-bit pSeries and RS/6000 setup code.
5 * Copyright (C) 1995 Linus Torvalds
6 * Adapted from 'alpha' version by Gary Thomas
7 * Modified by Cort Dougan (cort@cs.nmt.edu)
8 * Modified by PPC64 Team, IBM Corp
12 * bootup setup stuff..
15 #include <linux/cpu.h>
16 #include <linux/errno.h>
17 #include <linux/sched.h>
18 #include <linux/kernel.h>
20 #include <linux/stddef.h>
21 #include <linux/unistd.h>
22 #include <linux/user.h>
23 #include <linux/tty.h>
24 #include <linux/major.h>
25 #include <linux/interrupt.h>
26 #include <linux/reboot.h>
27 #include <linux/init.h>
28 #include <linux/ioport.h>
29 #include <linux/console.h>
30 #include <linux/pci.h>
31 #include <linux/utsname.h>
32 #include <linux/adb.h>
33 #include <linux/export.h>
34 #include <linux/delay.h>
35 #include <linux/irq.h>
36 #include <linux/seq_file.h>
37 #include <linux/root_dev.h>
39 #include <linux/of_pci.h>
40 #include <linux/memblock.h>
41 #include <linux/swiotlb.h>
44 #include <asm/processor.h>
48 #include <asm/pci-bridge.h>
49 #include <asm/iommu.h>
51 #include <asm/machdep.h>
54 #include <asm/nvram.h>
58 #include <asm/ppc-pci.h>
59 #include <asm/i8259.h>
62 #include <asm/firmware.h>
65 #include <asm/plpar_wrappers.h>
66 #include <asm/kexec.h>
67 #include <asm/isa-bridge.h>
68 #include <asm/security_features.h>
69 #include <asm/asm-const.h>
71 #include <asm/swiotlb.h>
76 #include "../../../../drivers/pci/pci.h"
78 DEFINE_STATIC_KEY_FALSE(shared_processor);
79 EXPORT_SYMBOL_GPL(shared_processor);
83 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);
84 EXPORT_SYMBOL(CMO_PageSize);
86 int fwnmi_active; /* TRUE if an FWNMI handler is present */
87 int ibm_nmi_interlock_token;
89 static void pSeries_show_cpuinfo(struct seq_file *m)
91 struct device_node *root;
92 const char *model = "";
94 root = of_find_node_by_path("/");
96 model = of_get_property(root, "model", NULL);
97 seq_printf(m, "machine\t\t: CHRP %s\n", model);
100 seq_printf(m, "MMU\t\t: Radix\n");
102 seq_printf(m, "MMU\t\t: Hash\n");
105 /* Initialize firmware assisted non-maskable interrupts if
106 * the firmware supports this feature.
108 static void __init fwnmi_init(void)
110 unsigned long system_reset_addr, machine_check_addr;
113 int nr_cpus = num_possible_cpus();
114 #ifdef CONFIG_PPC_BOOK3S_64
115 struct slb_entry *slb_ptr;
118 int ibm_nmi_register_token;
120 ibm_nmi_register_token = rtas_token("ibm,nmi-register");
121 if (ibm_nmi_register_token == RTAS_UNKNOWN_SERVICE)
124 ibm_nmi_interlock_token = rtas_token("ibm,nmi-interlock");
125 if (WARN_ON(ibm_nmi_interlock_token == RTAS_UNKNOWN_SERVICE))
128 /* If the kernel's not linked at zero we point the firmware at low
129 * addresses anyway, and use a trampoline to get to the real code. */
130 system_reset_addr = __pa(system_reset_fwnmi) - PHYSICAL_START;
131 machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START;
133 if (0 == rtas_call(ibm_nmi_register_token, 2, 1, NULL,
134 system_reset_addr, machine_check_addr))
138 * Allocate a chunk for per cpu buffer to hold rtas errorlog.
139 * It will be used in real mode mce handler, hence it needs to be
142 mce_data_buf = memblock_alloc_try_nid_raw(RTAS_ERROR_LOG_MAX * nr_cpus,
143 RTAS_ERROR_LOG_MAX, MEMBLOCK_LOW_LIMIT,
144 ppc64_rma_size, NUMA_NO_NODE);
146 panic("Failed to allocate %d bytes below %pa for MCE buffer\n",
147 RTAS_ERROR_LOG_MAX * nr_cpus, &ppc64_rma_size);
149 for_each_possible_cpu(i) {
150 paca_ptrs[i]->mce_data_buf = mce_data_buf +
151 (RTAS_ERROR_LOG_MAX * i);
154 #ifdef CONFIG_PPC_BOOK3S_64
155 if (!radix_enabled()) {
156 /* Allocate per cpu area to save old slb contents during MCE */
157 size = sizeof(struct slb_entry) * mmu_slb_size * nr_cpus;
158 slb_ptr = memblock_alloc_try_nid_raw(size,
159 sizeof(struct slb_entry), MEMBLOCK_LOW_LIMIT,
160 ppc64_rma_size, NUMA_NO_NODE);
162 panic("Failed to allocate %zu bytes below %pa for slb area\n",
163 size, &ppc64_rma_size);
165 for_each_possible_cpu(i)
166 paca_ptrs[i]->mce_faulty_slbs = slb_ptr + (mmu_slb_size * i);
171 static void pseries_8259_cascade(struct irq_desc *desc)
173 struct irq_chip *chip = irq_desc_get_chip(desc);
174 unsigned int cascade_irq = i8259_irq();
177 generic_handle_irq(cascade_irq);
179 chip->irq_eoi(&desc->irq_data);
182 static void __init pseries_setup_i8259_cascade(void)
184 struct device_node *np, *old, *found = NULL;
185 unsigned int cascade;
187 unsigned long intack = 0;
190 for_each_node_by_type(np, "interrupt-controller") {
191 if (of_device_is_compatible(np, "chrp,iic")) {
198 printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
202 cascade = irq_of_parse_and_map(found, 0);
204 printk(KERN_ERR "pic: failed to map cascade interrupt");
207 pr_debug("pic: cascade mapped to irq %d\n", cascade);
209 for (old = of_node_get(found); old != NULL ; old = np) {
210 np = of_get_parent(old);
214 if (!of_node_name_eq(np, "pci"))
216 addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
219 naddr = of_n_addr_cells(np);
220 intack = addrp[naddr-1];
222 intack |= ((unsigned long)addrp[naddr-2]) << 32;
225 printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
226 i8259_init(found, intack);
228 irq_set_chained_handler(cascade, pseries_8259_cascade);
231 static void __init pseries_init_irq(void)
233 /* Try using a XIVE if available, otherwise use a XICS */
234 if (!xive_spapr_init()) {
236 pseries_setup_i8259_cascade();
240 static void pseries_lpar_enable_pmcs(void)
242 unsigned long set, reset;
246 plpar_hcall_norets(H_PERFMON, set, reset);
249 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
251 struct of_reconfig_data *rd = data;
252 struct device_node *parent, *np = rd->dn;
257 case OF_RECONFIG_ATTACH_NODE:
258 parent = of_get_parent(np);
259 pdn = parent ? PCI_DN(parent) : NULL;
261 pci_add_device_node_info(pdn->phb, np);
265 case OF_RECONFIG_DETACH_NODE:
268 list_del(&pdn->list);
277 static struct notifier_block pci_dn_reconfig_nb = {
278 .notifier_call = pci_dn_reconfig_notifier,
281 struct kmem_cache *dtl_cache;
283 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
285 * Allocate space for the dispatch trace log for all possible cpus
286 * and register the buffers with the hypervisor. This is used for
287 * computing time stolen by the hypervisor.
289 static int alloc_dispatch_logs(void)
291 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
297 alloc_dtl_buffers(0);
299 /* Register the DTL for the current (boot) cpu */
300 register_dtl_buffer(smp_processor_id());
304 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
305 static inline int alloc_dispatch_logs(void)
309 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
311 static int alloc_dispatch_log_kmem_cache(void)
313 void (*ctor)(void *) = get_dtl_cache_ctor();
315 dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
316 DISPATCH_LOG_BYTES, 0, ctor);
318 pr_warn("Failed to create dispatch trace log buffer cache\n");
319 pr_warn("Stolen time statistics will be unreliable\n");
323 return alloc_dispatch_logs();
325 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
327 DEFINE_PER_CPU(u64, idle_spurr_cycles);
328 DEFINE_PER_CPU(u64, idle_entry_purr_snap);
329 DEFINE_PER_CPU(u64, idle_entry_spurr_snap);
330 static void pseries_lpar_idle(void)
333 * Default handler to go into low thread priority and possibly
334 * low power mode by ceding processor to hypervisor
337 if (!prep_irq_for_idle())
340 /* Indicate to hypervisor that we are idle. */
341 pseries_idle_prolog();
344 * Yield the processor to the hypervisor. We return if
345 * an external interrupt occurs (which are driven prior
346 * to returning here) or if a prod occurs from another
347 * processor. When returning here, external interrupts
352 pseries_idle_epilog();
356 * Enable relocation on during exceptions. This has partition wide scope and
357 * may take a while to complete, if it takes longer than one second we will
358 * just give up rather than wasting any more time on this - if that turns out
359 * to ever be a problem in practice we can move this into a kernel thread to
360 * finish off the process later in boot.
362 bool pseries_enable_reloc_on_exc(void)
365 unsigned int delay, total_delay = 0;
368 rc = enable_reloc_on_exceptions();
369 if (!H_IS_LONG_BUSY(rc)) {
371 pr_info("Relocation on exceptions not"
374 } else if (rc != H_SUCCESS) {
375 pr_warn("Unable to enable relocation"
376 " on exceptions: %ld\n", rc);
382 delay = get_longbusy_msecs(rc);
383 total_delay += delay;
384 if (total_delay > 1000) {
385 pr_warn("Warning: Giving up waiting to enable "
386 "relocation on exceptions (%u msec)!\n",
394 EXPORT_SYMBOL(pseries_enable_reloc_on_exc);
396 void pseries_disable_reloc_on_exc(void)
401 rc = disable_reloc_on_exceptions();
402 if (!H_IS_LONG_BUSY(rc))
404 mdelay(get_longbusy_msecs(rc));
407 pr_warn("Warning: Failed to disable relocation on exceptions: %ld\n",
410 EXPORT_SYMBOL(pseries_disable_reloc_on_exc);
412 #ifdef CONFIG_KEXEC_CORE
413 static void pSeries_machine_kexec(struct kimage *image)
415 if (firmware_has_feature(FW_FEATURE_SET_MODE))
416 pseries_disable_reloc_on_exc();
418 default_machine_kexec(image);
422 #ifdef __LITTLE_ENDIAN__
423 void pseries_big_endian_exceptions(void)
428 rc = enable_big_endian_exceptions();
429 if (!H_IS_LONG_BUSY(rc))
431 mdelay(get_longbusy_msecs(rc));
435 * At this point it is unlikely panic() will get anything
436 * out to the user, since this is called very late in kexec
437 * but at least this will stop us from continuing on further
438 * and creating an even more difficult to debug situation.
440 * There is a known problem when kdump'ing, if cpus are offline
441 * the above call will fail. Rather than panicking again, keep
442 * going and hope the kdump kernel is also little endian, which
445 if (rc && !kdump_in_progress())
446 panic("Could not enable big endian exceptions");
449 void pseries_little_endian_exceptions(void)
454 rc = enable_little_endian_exceptions();
455 if (!H_IS_LONG_BUSY(rc))
457 mdelay(get_longbusy_msecs(rc));
460 ppc_md.progress("H_SET_MODE LE exception fail", 0);
461 panic("Could not enable little endian exceptions");
466 static void __init find_and_init_phbs(void)
468 struct device_node *node;
469 struct pci_controller *phb;
470 struct device_node *root = of_find_node_by_path("/");
472 for_each_child_of_node(root, node) {
473 if (!of_node_is_type(node, "pci") &&
474 !of_node_is_type(node, "pciex"))
477 phb = pcibios_alloc_controller(node);
481 pci_process_bridge_OF_ranges(phb, node, 0);
482 isa_bridge_find_early(phb);
483 phb->controller_ops = pseries_pci_controller_ops;
489 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
492 of_pci_check_probe_only();
495 static void init_cpu_char_feature_flags(struct h_cpu_char_result *result)
498 * The features below are disabled by default, so we instead look to see
499 * if firmware has *enabled* them, and set them if so.
501 if (result->character & H_CPU_CHAR_SPEC_BAR_ORI31)
502 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31);
504 if (result->character & H_CPU_CHAR_BCCTRL_SERIALISED)
505 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED);
507 if (result->character & H_CPU_CHAR_L1D_FLUSH_ORI30)
508 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30);
510 if (result->character & H_CPU_CHAR_L1D_FLUSH_TRIG2)
511 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2);
513 if (result->character & H_CPU_CHAR_L1D_THREAD_PRIV)
514 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV);
516 if (result->character & H_CPU_CHAR_COUNT_CACHE_DISABLED)
517 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED);
519 if (result->character & H_CPU_CHAR_BCCTR_FLUSH_ASSIST)
520 security_ftr_set(SEC_FTR_BCCTR_FLUSH_ASSIST);
522 if (result->character & H_CPU_CHAR_BCCTR_LINK_FLUSH_ASSIST)
523 security_ftr_set(SEC_FTR_BCCTR_LINK_FLUSH_ASSIST);
525 if (result->behaviour & H_CPU_BEHAV_FLUSH_COUNT_CACHE)
526 security_ftr_set(SEC_FTR_FLUSH_COUNT_CACHE);
528 if (result->behaviour & H_CPU_BEHAV_FLUSH_LINK_STACK)
529 security_ftr_set(SEC_FTR_FLUSH_LINK_STACK);
532 * The features below are enabled by default, so we instead look to see
533 * if firmware has *disabled* them, and clear them if so.
535 if (!(result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY))
536 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY);
538 if (!(result->behaviour & H_CPU_BEHAV_L1D_FLUSH_PR))
539 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR);
541 if (!(result->behaviour & H_CPU_BEHAV_BNDS_CHK_SPEC_BAR))
542 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR);
545 void pseries_setup_rfi_flush(void)
547 struct h_cpu_char_result result;
548 enum l1d_flush_type types;
553 * Set features to the defaults assumed by init_cpu_char_feature_flags()
554 * so it can set/clear again any features that might have changed after
555 * migration, and in case the hypercall fails and it is not even called.
557 powerpc_security_features = SEC_FTR_DEFAULT;
559 rc = plpar_get_cpu_characteristics(&result);
561 init_cpu_char_feature_flags(&result);
564 * We're the guest so this doesn't apply to us, clear it to simplify
565 * handling of it elsewhere.
567 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV);
569 types = L1D_FLUSH_FALLBACK;
571 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2))
572 types |= L1D_FLUSH_MTTRIG;
574 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30))
575 types |= L1D_FLUSH_ORI;
577 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) && \
578 security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR);
580 setup_rfi_flush(types, enable);
581 setup_count_cache_flush();
583 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) &&
584 security_ftr_enabled(SEC_FTR_L1D_FLUSH_ENTRY);
585 setup_entry_flush(enable);
587 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) &&
588 security_ftr_enabled(SEC_FTR_L1D_FLUSH_UACCESS);
589 setup_uaccess_flush(enable);
592 #ifdef CONFIG_PCI_IOV
593 enum rtas_iov_fw_value_map {
594 NUM_RES_PROPERTY = 0, /* Number of Resources */
595 LOW_INT = 1, /* Lowest 32 bits of Address */
596 START_OF_ENTRIES = 2, /* Always start of entry */
597 APERTURE_PROPERTY = 2, /* Start of entry+ to Aperture Size */
598 WDW_SIZE_PROPERTY = 4, /* Start of entry+ to Window Size */
599 NEXT_ENTRY = 7 /* Go to next entry on array */
602 enum get_iov_fw_value_index {
603 BAR_ADDRS = 1, /* Get Bar Address */
604 APERTURE_SIZE = 2, /* Get Aperture Size */
605 WDW_SIZE = 3 /* Get Window Size */
608 resource_size_t pseries_get_iov_fw_value(struct pci_dev *dev, int resno,
609 enum get_iov_fw_value_index value)
612 struct device_node *dn = pci_device_to_OF_node(dev);
613 int i, num_res, ret = 0;
615 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
620 * First element in the array is the number of Bars
621 * returned. Search through the list to find the matching
624 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
625 if (resno >= num_res)
626 return 0; /* or an errror */
628 i = START_OF_ENTRIES + NEXT_ENTRY * resno;
631 ret = of_read_number(&indexes[i], 2);
634 ret = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
637 ret = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
644 void of_pci_set_vf_bar_size(struct pci_dev *dev, const int *indexes)
646 struct resource *res;
647 resource_size_t base, size;
650 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
651 num_res = min_t(int, num_res, PCI_SRIOV_NUM_BARS);
652 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
653 i += NEXT_ENTRY, r++) {
654 res = &dev->resource[r + PCI_IOV_RESOURCES];
655 base = of_read_number(&indexes[i], 2);
656 size = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
657 res->flags = pci_parse_of_flags(of_read_number
658 (&indexes[i + LOW_INT], 1), 0);
659 res->flags |= (IORESOURCE_MEM_64 | IORESOURCE_PCI_FIXED);
660 res->name = pci_name(dev);
662 res->end = base + size - 1;
666 void of_pci_parse_iov_addrs(struct pci_dev *dev, const int *indexes)
668 struct resource *res, *root, *conflict;
669 resource_size_t base, size;
673 * First element in the array is the number of Bars
674 * returned. Search through the list to find the matching
675 * bars assign them from firmware into resources structure.
677 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
678 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
679 i += NEXT_ENTRY, r++) {
680 res = &dev->resource[r + PCI_IOV_RESOURCES];
681 base = of_read_number(&indexes[i], 2);
682 size = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
683 res->name = pci_name(dev);
685 res->end = base + size - 1;
686 root = &iomem_resource;
688 "pSeries IOV BAR %d: trying firmware assignment %pR\n",
689 r + PCI_IOV_RESOURCES, res);
690 conflict = request_resource_conflict(root, res);
693 "BAR %d: %pR conflicts with %s %pR\n",
694 r + PCI_IOV_RESOURCES, res,
695 conflict->name, conflict);
696 res->flags |= IORESOURCE_UNSET;
701 static void pseries_disable_sriov_resources(struct pci_dev *pdev)
705 pci_warn(pdev, "No hypervisor support for SR-IOV on this device, IOV BARs disabled.\n");
706 for (i = 0; i < PCI_SRIOV_NUM_BARS; i++)
707 pdev->resource[i + PCI_IOV_RESOURCES].flags = 0;
710 static void pseries_pci_fixup_resources(struct pci_dev *pdev)
713 struct device_node *dn = pci_device_to_OF_node(pdev);
715 /*Firmware must support open sriov otherwise dont configure*/
716 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
718 of_pci_set_vf_bar_size(pdev, indexes);
720 pseries_disable_sriov_resources(pdev);
723 static void pseries_pci_fixup_iov_resources(struct pci_dev *pdev)
726 struct device_node *dn = pci_device_to_OF_node(pdev);
728 if (!pdev->is_physfn || pci_dev_is_added(pdev))
730 /*Firmware must support open sriov otherwise dont configure*/
731 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
733 of_pci_parse_iov_addrs(pdev, indexes);
735 pseries_disable_sriov_resources(pdev);
738 static resource_size_t pseries_pci_iov_resource_alignment(struct pci_dev *pdev,
742 struct device_node *dn = pci_device_to_OF_node(pdev);
744 /*Firmware must support open sriov otherwise report regular alignment*/
745 reg = of_get_property(dn, "ibm,is-open-sriov-pf", NULL);
747 return pci_iov_resource_size(pdev, resno);
749 if (!pdev->is_physfn)
751 return pseries_get_iov_fw_value(pdev,
752 resno - PCI_IOV_RESOURCES,
757 static void __init pSeries_setup_arch(void)
759 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
761 /* Discover PIC type and setup ppc_md accordingly */
765 if (radix_enabled() && !mmu_has_feature(MMU_FTR_GTSE))
766 if (!firmware_has_feature(FW_FEATURE_RPT_INVALIDATE))
767 panic("BUG: Radix support requires either GTSE or RPT_INVALIDATE\n");
770 /* openpic global configuration register (64-bit format). */
771 /* openpic Interrupt Source Unit pointer (64-bit format). */
772 /* python0 facility area (mmio) (64-bit format) REAL address. */
774 /* init to some ~sane value until calibrate_delay() runs */
775 loops_per_jiffy = 50000000;
779 pseries_setup_rfi_flush();
781 pseries_lpar_read_hblkrm_characteristics();
783 /* By default, only probe PCI (can be overridden by rtas_pci) */
784 pci_add_flags(PCI_PROBE_ONLY);
786 /* Find and initialize PCI host bridges */
787 init_pci_config_tokens();
788 find_and_init_phbs();
789 of_reconfig_notifier_register(&pci_dn_reconfig_nb);
791 pSeries_nvram_init();
793 if (firmware_has_feature(FW_FEATURE_LPAR)) {
794 vpa_init(boot_cpuid);
796 if (lppaca_shared_proc(get_lppaca())) {
797 static_branch_enable(&shared_processor);
801 ppc_md.power_save = pseries_lpar_idle;
802 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
803 #ifdef CONFIG_PCI_IOV
804 ppc_md.pcibios_fixup_resources =
805 pseries_pci_fixup_resources;
806 ppc_md.pcibios_fixup_sriov =
807 pseries_pci_fixup_iov_resources;
808 ppc_md.pcibios_iov_resource_alignment =
809 pseries_pci_iov_resource_alignment;
812 /* No special idle routine */
813 ppc_md.enable_pmcs = power4_enable_pmcs;
816 ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
818 if (swiotlb_force == SWIOTLB_FORCE)
819 ppc_swiotlb_enable = 1;
822 static void pseries_panic(char *str)
824 panic_flush_kmsg_end();
828 static int __init pSeries_init_panel(void)
830 /* Manually leave the kernel version on the panel. */
831 #ifdef __BIG_ENDIAN__
832 ppc_md.progress("Linux ppc64\n", 0);
834 ppc_md.progress("Linux ppc64le\n", 0);
836 ppc_md.progress(init_utsname()->version, 0);
840 machine_arch_initcall(pseries, pSeries_init_panel);
842 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
844 return plpar_hcall_norets(H_SET_DABR, dabr);
847 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
849 /* Have to set at least one bit in the DABRX according to PAPR */
850 if (dabrx == 0 && dabr == 0)
852 /* PAPR says we can only set kernel and user bits */
853 dabrx &= DABRX_KERNEL | DABRX_USER;
855 return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
858 static int pseries_set_dawr(int nr, unsigned long dawr, unsigned long dawrx)
860 /* PAPR says we can't set HYP */
864 return plpar_set_watchpoint0(dawr, dawrx);
866 return plpar_set_watchpoint1(dawr, dawrx);
869 #define CMO_CHARACTERISTICS_TOKEN 44
870 #define CMO_MAXLENGTH 1026
872 void pSeries_coalesce_init(void)
874 struct hvcall_mpp_x_data mpp_x_data;
876 if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
877 powerpc_firmware_features |= FW_FEATURE_XCMO;
879 powerpc_firmware_features &= ~FW_FEATURE_XCMO;
883 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
884 * handle that here. (Stolen from parse_system_parameter_string)
886 static void pSeries_cmo_feature_init(void)
888 char *ptr, *key, *value, *end;
890 int page_order = IOMMU_PAGE_SHIFT_4K;
892 pr_debug(" -> fw_cmo_feature_init()\n");
893 spin_lock(&rtas_data_buf_lock);
894 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
895 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
897 CMO_CHARACTERISTICS_TOKEN,
901 if (call_status != 0) {
902 spin_unlock(&rtas_data_buf_lock);
903 pr_debug("CMO not available\n");
904 pr_debug(" <- fw_cmo_feature_init()\n");
908 end = rtas_data_buf + CMO_MAXLENGTH - 2;
909 ptr = rtas_data_buf + 2; /* step over strlen value */
912 while (*ptr && (ptr <= end)) {
913 /* Separate the key and value by replacing '=' with '\0' and
914 * point the value at the string after the '='
919 } else if (ptr[0] == '\0' || ptr[0] == ',') {
920 /* Terminate the string containing the key/value pair */
924 pr_debug("Malformed key/value pair\n");
925 /* Never found a '=', end processing */
929 if (0 == strcmp(key, "CMOPageSize"))
930 page_order = simple_strtol(value, NULL, 10);
931 else if (0 == strcmp(key, "PrPSP"))
932 CMO_PrPSP = simple_strtol(value, NULL, 10);
933 else if (0 == strcmp(key, "SecPSP"))
934 CMO_SecPSP = simple_strtol(value, NULL, 10);
935 value = key = ptr + 1;
940 /* Page size is returned as the power of 2 of the page size,
941 * convert to the page size in bytes before returning
943 CMO_PageSize = 1 << page_order;
944 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
946 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
947 pr_info("CMO enabled\n");
948 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
950 powerpc_firmware_features |= FW_FEATURE_CMO;
951 pSeries_coalesce_init();
953 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
955 spin_unlock(&rtas_data_buf_lock);
956 pr_debug(" <- fw_cmo_feature_init()\n");
960 * Early initialization. Relocation is on but do not reference unbolted pages
962 static void __init pseries_init(void)
964 pr_debug(" -> pseries_init()\n");
966 #ifdef CONFIG_HVC_CONSOLE
967 if (firmware_has_feature(FW_FEATURE_LPAR))
968 hvc_vio_init_early();
970 if (firmware_has_feature(FW_FEATURE_XDABR))
971 ppc_md.set_dabr = pseries_set_xdabr;
972 else if (firmware_has_feature(FW_FEATURE_DABR))
973 ppc_md.set_dabr = pseries_set_dabr;
975 if (firmware_has_feature(FW_FEATURE_SET_MODE))
976 ppc_md.set_dawr = pseries_set_dawr;
978 pSeries_cmo_feature_init();
979 iommu_init_early_pSeries();
981 pr_debug(" <- pseries_init()\n");
985 * pseries_power_off - tell firmware about how to power off the system.
987 * This function calls either the power-off rtas token in normal cases
988 * or the ibm,power-off-ups token (if present & requested) in case of
989 * a power failure. If power-off token is used, power on will only be
990 * possible with power button press. If ibm,power-off-ups token is used
991 * it will allow auto poweron after power is restored.
993 static void pseries_power_off(void)
996 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
998 if (rtas_flash_term_hook)
999 rtas_flash_term_hook(SYS_POWER_OFF);
1001 if (rtas_poweron_auto == 0 ||
1002 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
1003 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
1004 printk(KERN_INFO "RTAS power-off returned %d\n", rc);
1006 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
1007 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
1012 static int __init pSeries_probe(void)
1014 if (!of_node_is_type(of_root, "chrp"))
1017 /* Cell blades firmware claims to be chrp while it's not. Until this
1018 * is fixed, we need to avoid those here.
1020 if (of_machine_is_compatible("IBM,CPBW-1.0") ||
1021 of_machine_is_compatible("IBM,CBEA"))
1024 pm_power_off = pseries_power_off;
1026 pr_debug("Machine is%s LPAR !\n",
1027 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
1034 static int pSeries_pci_probe_mode(struct pci_bus *bus)
1036 if (firmware_has_feature(FW_FEATURE_LPAR))
1037 return PCI_PROBE_DEVTREE;
1038 return PCI_PROBE_NORMAL;
1041 struct pci_controller_ops pseries_pci_controller_ops = {
1042 .probe_mode = pSeries_pci_probe_mode,
1045 define_machine(pseries) {
1047 .probe = pSeries_probe,
1048 .setup_arch = pSeries_setup_arch,
1049 .init_IRQ = pseries_init_irq,
1050 .show_cpuinfo = pSeries_show_cpuinfo,
1051 .log_error = pSeries_log_error,
1052 .pcibios_fixup = pSeries_final_fixup,
1053 .restart = rtas_restart,
1055 .panic = pseries_panic,
1056 .get_boot_time = rtas_get_boot_time,
1057 .get_rtc_time = rtas_get_rtc_time,
1058 .set_rtc_time = rtas_set_rtc_time,
1059 .calibrate_decr = generic_calibrate_decr,
1060 .progress = rtas_progress,
1061 .system_reset_exception = pSeries_system_reset_exception,
1062 .machine_check_early = pseries_machine_check_realmode,
1063 .machine_check_exception = pSeries_machine_check_exception,
1064 #ifdef CONFIG_KEXEC_CORE
1065 .machine_kexec = pSeries_machine_kexec,
1066 .kexec_cpu_down = pseries_kexec_cpu_down,
1068 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
1069 .memory_block_size = pseries_memory_block_size,