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>
46 #include <asm/pgtable.h>
49 #include <asm/pci-bridge.h>
50 #include <asm/iommu.h>
52 #include <asm/machdep.h>
55 #include <asm/nvram.h>
59 #include <asm/ppc-pci.h>
60 #include <asm/i8259.h>
63 #include <asm/firmware.h>
66 #include <asm/plpar_wrappers.h>
67 #include <asm/kexec.h>
68 #include <asm/isa-bridge.h>
69 #include <asm/security_features.h>
70 #include <asm/asm-const.h>
71 #include <asm/swiotlb.h>
75 #include "../../../../drivers/pci/pci.h"
77 DEFINE_STATIC_KEY_FALSE(shared_processor);
78 EXPORT_SYMBOL(shared_processor);
82 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);
83 EXPORT_SYMBOL(CMO_PageSize);
85 int fwnmi_active; /* TRUE if an FWNMI handler is present */
87 static void pSeries_show_cpuinfo(struct seq_file *m)
89 struct device_node *root;
90 const char *model = "";
92 root = of_find_node_by_path("/");
94 model = of_get_property(root, "model", NULL);
95 seq_printf(m, "machine\t\t: CHRP %s\n", model);
98 seq_printf(m, "MMU\t\t: Radix\n");
100 seq_printf(m, "MMU\t\t: Hash\n");
103 /* Initialize firmware assisted non-maskable interrupts if
104 * the firmware supports this feature.
106 static void __init fwnmi_init(void)
108 unsigned long system_reset_addr, machine_check_addr;
111 int nr_cpus = num_possible_cpus();
112 #ifdef CONFIG_PPC_BOOK3S_64
113 struct slb_entry *slb_ptr;
117 int ibm_nmi_register = rtas_token("ibm,nmi-register");
118 if (ibm_nmi_register == RTAS_UNKNOWN_SERVICE)
121 /* If the kernel's not linked at zero we point the firmware at low
122 * addresses anyway, and use a trampoline to get to the real code. */
123 system_reset_addr = __pa(system_reset_fwnmi) - PHYSICAL_START;
124 machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START;
126 if (0 == rtas_call(ibm_nmi_register, 2, 1, NULL, system_reset_addr,
131 * Allocate a chunk for per cpu buffer to hold rtas errorlog.
132 * It will be used in real mode mce handler, hence it needs to be
135 mce_data_buf = memblock_alloc_try_nid_raw(RTAS_ERROR_LOG_MAX * nr_cpus,
136 RTAS_ERROR_LOG_MAX, MEMBLOCK_LOW_LIMIT,
137 ppc64_rma_size, NUMA_NO_NODE);
139 panic("Failed to allocate %d bytes below %pa for MCE buffer\n",
140 RTAS_ERROR_LOG_MAX * nr_cpus, &ppc64_rma_size);
142 for_each_possible_cpu(i) {
143 paca_ptrs[i]->mce_data_buf = mce_data_buf +
144 (RTAS_ERROR_LOG_MAX * i);
147 #ifdef CONFIG_PPC_BOOK3S_64
148 if (!radix_enabled()) {
149 /* Allocate per cpu area to save old slb contents during MCE */
150 size = sizeof(struct slb_entry) * mmu_slb_size * nr_cpus;
151 slb_ptr = memblock_alloc_try_nid_raw(size,
152 sizeof(struct slb_entry), MEMBLOCK_LOW_LIMIT,
153 ppc64_rma_size, NUMA_NO_NODE);
155 panic("Failed to allocate %zu bytes below %pa for slb area\n",
156 size, &ppc64_rma_size);
158 for_each_possible_cpu(i)
159 paca_ptrs[i]->mce_faulty_slbs = slb_ptr + (mmu_slb_size * i);
164 static void pseries_8259_cascade(struct irq_desc *desc)
166 struct irq_chip *chip = irq_desc_get_chip(desc);
167 unsigned int cascade_irq = i8259_irq();
170 generic_handle_irq(cascade_irq);
172 chip->irq_eoi(&desc->irq_data);
175 static void __init pseries_setup_i8259_cascade(void)
177 struct device_node *np, *old, *found = NULL;
178 unsigned int cascade;
180 unsigned long intack = 0;
183 for_each_node_by_type(np, "interrupt-controller") {
184 if (of_device_is_compatible(np, "chrp,iic")) {
191 printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
195 cascade = irq_of_parse_and_map(found, 0);
197 printk(KERN_ERR "pic: failed to map cascade interrupt");
200 pr_debug("pic: cascade mapped to irq %d\n", cascade);
202 for (old = of_node_get(found); old != NULL ; old = np) {
203 np = of_get_parent(old);
207 if (!of_node_name_eq(np, "pci"))
209 addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
212 naddr = of_n_addr_cells(np);
213 intack = addrp[naddr-1];
215 intack |= ((unsigned long)addrp[naddr-2]) << 32;
218 printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
219 i8259_init(found, intack);
221 irq_set_chained_handler(cascade, pseries_8259_cascade);
224 static void __init pseries_init_irq(void)
226 /* Try using a XIVE if available, otherwise use a XICS */
227 if (!xive_spapr_init()) {
229 pseries_setup_i8259_cascade();
233 static void pseries_lpar_enable_pmcs(void)
235 unsigned long set, reset;
239 plpar_hcall_norets(H_PERFMON, set, reset);
242 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
244 struct of_reconfig_data *rd = data;
245 struct device_node *parent, *np = rd->dn;
250 case OF_RECONFIG_ATTACH_NODE:
251 parent = of_get_parent(np);
252 pdn = parent ? PCI_DN(parent) : NULL;
254 pci_add_device_node_info(pdn->phb, np);
258 case OF_RECONFIG_DETACH_NODE:
261 list_del(&pdn->list);
270 static struct notifier_block pci_dn_reconfig_nb = {
271 .notifier_call = pci_dn_reconfig_notifier,
274 struct kmem_cache *dtl_cache;
276 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
278 * Allocate space for the dispatch trace log for all possible cpus
279 * and register the buffers with the hypervisor. This is used for
280 * computing time stolen by the hypervisor.
282 static int alloc_dispatch_logs(void)
284 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
290 alloc_dtl_buffers(0);
292 /* Register the DTL for the current (boot) cpu */
293 register_dtl_buffer(smp_processor_id());
297 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
298 static inline int alloc_dispatch_logs(void)
302 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
304 static int alloc_dispatch_log_kmem_cache(void)
306 void (*ctor)(void *) = get_dtl_cache_ctor();
308 dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
309 DISPATCH_LOG_BYTES, 0, ctor);
311 pr_warn("Failed to create dispatch trace log buffer cache\n");
312 pr_warn("Stolen time statistics will be unreliable\n");
316 return alloc_dispatch_logs();
318 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
320 static void pseries_lpar_idle(void)
323 * Default handler to go into low thread priority and possibly
324 * low power mode by ceding processor to hypervisor
327 if (!prep_irq_for_idle())
330 /* Indicate to hypervisor that we are idle. */
331 get_lppaca()->idle = 1;
334 * Yield the processor to the hypervisor. We return if
335 * an external interrupt occurs (which are driven prior
336 * to returning here) or if a prod occurs from another
337 * processor. When returning here, external interrupts
342 get_lppaca()->idle = 0;
346 * Enable relocation on during exceptions. This has partition wide scope and
347 * may take a while to complete, if it takes longer than one second we will
348 * just give up rather than wasting any more time on this - if that turns out
349 * to ever be a problem in practice we can move this into a kernel thread to
350 * finish off the process later in boot.
352 void pseries_enable_reloc_on_exc(void)
355 unsigned int delay, total_delay = 0;
358 rc = enable_reloc_on_exceptions();
359 if (!H_IS_LONG_BUSY(rc)) {
361 pr_info("Relocation on exceptions not"
363 } else if (rc != H_SUCCESS) {
364 pr_warn("Unable to enable relocation"
365 " on exceptions: %ld\n", rc);
370 delay = get_longbusy_msecs(rc);
371 total_delay += delay;
372 if (total_delay > 1000) {
373 pr_warn("Warning: Giving up waiting to enable "
374 "relocation on exceptions (%u msec)!\n",
382 EXPORT_SYMBOL(pseries_enable_reloc_on_exc);
384 void pseries_disable_reloc_on_exc(void)
389 rc = disable_reloc_on_exceptions();
390 if (!H_IS_LONG_BUSY(rc))
392 mdelay(get_longbusy_msecs(rc));
395 pr_warn("Warning: Failed to disable relocation on exceptions: %ld\n",
398 EXPORT_SYMBOL(pseries_disable_reloc_on_exc);
400 #ifdef CONFIG_KEXEC_CORE
401 static void pSeries_machine_kexec(struct kimage *image)
403 if (firmware_has_feature(FW_FEATURE_SET_MODE))
404 pseries_disable_reloc_on_exc();
406 default_machine_kexec(image);
410 #ifdef __LITTLE_ENDIAN__
411 void pseries_big_endian_exceptions(void)
416 rc = enable_big_endian_exceptions();
417 if (!H_IS_LONG_BUSY(rc))
419 mdelay(get_longbusy_msecs(rc));
423 * At this point it is unlikely panic() will get anything
424 * out to the user, since this is called very late in kexec
425 * but at least this will stop us from continuing on further
426 * and creating an even more difficult to debug situation.
428 * There is a known problem when kdump'ing, if cpus are offline
429 * the above call will fail. Rather than panicking again, keep
430 * going and hope the kdump kernel is also little endian, which
433 if (rc && !kdump_in_progress())
434 panic("Could not enable big endian exceptions");
437 void pseries_little_endian_exceptions(void)
442 rc = enable_little_endian_exceptions();
443 if (!H_IS_LONG_BUSY(rc))
445 mdelay(get_longbusy_msecs(rc));
448 ppc_md.progress("H_SET_MODE LE exception fail", 0);
449 panic("Could not enable little endian exceptions");
454 static void __init find_and_init_phbs(void)
456 struct device_node *node;
457 struct pci_controller *phb;
458 struct device_node *root = of_find_node_by_path("/");
460 for_each_child_of_node(root, node) {
461 if (!of_node_is_type(node, "pci") &&
462 !of_node_is_type(node, "pciex"))
465 phb = pcibios_alloc_controller(node);
469 pci_process_bridge_OF_ranges(phb, node, 0);
470 isa_bridge_find_early(phb);
471 phb->controller_ops = pseries_pci_controller_ops;
477 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
480 of_pci_check_probe_only();
483 static void init_cpu_char_feature_flags(struct h_cpu_char_result *result)
486 * The features below are disabled by default, so we instead look to see
487 * if firmware has *enabled* them, and set them if so.
489 if (result->character & H_CPU_CHAR_SPEC_BAR_ORI31)
490 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31);
492 if (result->character & H_CPU_CHAR_BCCTRL_SERIALISED)
493 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED);
495 if (result->character & H_CPU_CHAR_L1D_FLUSH_ORI30)
496 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30);
498 if (result->character & H_CPU_CHAR_L1D_FLUSH_TRIG2)
499 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2);
501 if (result->character & H_CPU_CHAR_L1D_THREAD_PRIV)
502 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV);
504 if (result->character & H_CPU_CHAR_COUNT_CACHE_DISABLED)
505 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED);
507 if (result->character & H_CPU_CHAR_BCCTR_FLUSH_ASSIST)
508 security_ftr_set(SEC_FTR_BCCTR_FLUSH_ASSIST);
510 if (result->behaviour & H_CPU_BEHAV_FLUSH_COUNT_CACHE)
511 security_ftr_set(SEC_FTR_FLUSH_COUNT_CACHE);
514 * The features below are enabled by default, so we instead look to see
515 * if firmware has *disabled* them, and clear them if so.
517 if (!(result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY))
518 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY);
520 if (!(result->behaviour & H_CPU_BEHAV_L1D_FLUSH_PR))
521 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR);
523 if (!(result->behaviour & H_CPU_BEHAV_BNDS_CHK_SPEC_BAR))
524 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR);
527 void pseries_setup_rfi_flush(void)
529 struct h_cpu_char_result result;
530 enum l1d_flush_type types;
535 * Set features to the defaults assumed by init_cpu_char_feature_flags()
536 * so it can set/clear again any features that might have changed after
537 * migration, and in case the hypercall fails and it is not even called.
539 powerpc_security_features = SEC_FTR_DEFAULT;
541 rc = plpar_get_cpu_characteristics(&result);
543 init_cpu_char_feature_flags(&result);
546 * We're the guest so this doesn't apply to us, clear it to simplify
547 * handling of it elsewhere.
549 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV);
551 types = L1D_FLUSH_FALLBACK;
553 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2))
554 types |= L1D_FLUSH_MTTRIG;
556 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30))
557 types |= L1D_FLUSH_ORI;
559 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) && \
560 security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR);
562 setup_rfi_flush(types, enable);
563 setup_count_cache_flush();
565 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) &&
566 security_ftr_enabled(SEC_FTR_L1D_FLUSH_ENTRY);
567 setup_entry_flush(enable);
569 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) &&
570 security_ftr_enabled(SEC_FTR_L1D_FLUSH_UACCESS);
571 setup_uaccess_flush(enable);
574 #ifdef CONFIG_PCI_IOV
575 enum rtas_iov_fw_value_map {
576 NUM_RES_PROPERTY = 0, /* Number of Resources */
577 LOW_INT = 1, /* Lowest 32 bits of Address */
578 START_OF_ENTRIES = 2, /* Always start of entry */
579 APERTURE_PROPERTY = 2, /* Start of entry+ to Aperture Size */
580 WDW_SIZE_PROPERTY = 4, /* Start of entry+ to Window Size */
581 NEXT_ENTRY = 7 /* Go to next entry on array */
584 enum get_iov_fw_value_index {
585 BAR_ADDRS = 1, /* Get Bar Address */
586 APERTURE_SIZE = 2, /* Get Aperture Size */
587 WDW_SIZE = 3 /* Get Window Size */
590 resource_size_t pseries_get_iov_fw_value(struct pci_dev *dev, int resno,
591 enum get_iov_fw_value_index value)
594 struct device_node *dn = pci_device_to_OF_node(dev);
595 int i, num_res, ret = 0;
597 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
602 * First element in the array is the number of Bars
603 * returned. Search through the list to find the matching
606 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
607 if (resno >= num_res)
608 return 0; /* or an errror */
610 i = START_OF_ENTRIES + NEXT_ENTRY * resno;
613 ret = of_read_number(&indexes[i], 2);
616 ret = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
619 ret = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
626 void of_pci_set_vf_bar_size(struct pci_dev *dev, const int *indexes)
628 struct resource *res;
629 resource_size_t base, size;
632 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
633 num_res = min_t(int, num_res, PCI_SRIOV_NUM_BARS);
634 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
635 i += NEXT_ENTRY, r++) {
636 res = &dev->resource[r + PCI_IOV_RESOURCES];
637 base = of_read_number(&indexes[i], 2);
638 size = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
639 res->flags = pci_parse_of_flags(of_read_number
640 (&indexes[i + LOW_INT], 1), 0);
641 res->flags |= (IORESOURCE_MEM_64 | IORESOURCE_PCI_FIXED);
642 res->name = pci_name(dev);
644 res->end = base + size - 1;
648 void of_pci_parse_iov_addrs(struct pci_dev *dev, const int *indexes)
650 struct resource *res, *root, *conflict;
651 resource_size_t base, size;
655 * First element in the array is the number of Bars
656 * returned. Search through the list to find the matching
657 * bars assign them from firmware into resources structure.
659 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
660 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
661 i += NEXT_ENTRY, r++) {
662 res = &dev->resource[r + PCI_IOV_RESOURCES];
663 base = of_read_number(&indexes[i], 2);
664 size = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
665 res->name = pci_name(dev);
667 res->end = base + size - 1;
668 root = &iomem_resource;
670 "pSeries IOV BAR %d: trying firmware assignment %pR\n",
671 r + PCI_IOV_RESOURCES, res);
672 conflict = request_resource_conflict(root, res);
675 "BAR %d: %pR conflicts with %s %pR\n",
676 r + PCI_IOV_RESOURCES, res,
677 conflict->name, conflict);
678 res->flags |= IORESOURCE_UNSET;
683 static void pseries_disable_sriov_resources(struct pci_dev *pdev)
687 pci_warn(pdev, "No hypervisor support for SR-IOV on this device, IOV BARs disabled.\n");
688 for (i = 0; i < PCI_SRIOV_NUM_BARS; i++)
689 pdev->resource[i + PCI_IOV_RESOURCES].flags = 0;
692 static void pseries_pci_fixup_resources(struct pci_dev *pdev)
695 struct device_node *dn = pci_device_to_OF_node(pdev);
697 /*Firmware must support open sriov otherwise dont configure*/
698 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
700 of_pci_set_vf_bar_size(pdev, indexes);
702 pseries_disable_sriov_resources(pdev);
705 static void pseries_pci_fixup_iov_resources(struct pci_dev *pdev)
708 struct device_node *dn = pci_device_to_OF_node(pdev);
710 if (!pdev->is_physfn || pci_dev_is_added(pdev))
712 /*Firmware must support open sriov otherwise dont configure*/
713 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
715 of_pci_parse_iov_addrs(pdev, indexes);
717 pseries_disable_sriov_resources(pdev);
720 static resource_size_t pseries_pci_iov_resource_alignment(struct pci_dev *pdev,
724 struct device_node *dn = pci_device_to_OF_node(pdev);
726 /*Firmware must support open sriov otherwise report regular alignment*/
727 reg = of_get_property(dn, "ibm,is-open-sriov-pf", NULL);
729 return pci_iov_resource_size(pdev, resno);
731 if (!pdev->is_physfn)
733 return pseries_get_iov_fw_value(pdev,
734 resno - PCI_IOV_RESOURCES,
739 static void __init pSeries_setup_arch(void)
741 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
743 /* Discover PIC type and setup ppc_md accordingly */
747 /* openpic global configuration register (64-bit format). */
748 /* openpic Interrupt Source Unit pointer (64-bit format). */
749 /* python0 facility area (mmio) (64-bit format) REAL address. */
751 /* init to some ~sane value until calibrate_delay() runs */
752 loops_per_jiffy = 50000000;
756 pseries_setup_rfi_flush();
758 pseries_lpar_read_hblkrm_characteristics();
760 /* By default, only probe PCI (can be overridden by rtas_pci) */
761 pci_add_flags(PCI_PROBE_ONLY);
763 /* Find and initialize PCI host bridges */
764 init_pci_config_tokens();
765 find_and_init_phbs();
766 of_reconfig_notifier_register(&pci_dn_reconfig_nb);
768 pSeries_nvram_init();
770 if (firmware_has_feature(FW_FEATURE_LPAR)) {
771 vpa_init(boot_cpuid);
773 if (lppaca_shared_proc(get_lppaca()))
774 static_branch_enable(&shared_processor);
776 ppc_md.power_save = pseries_lpar_idle;
777 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
778 #ifdef CONFIG_PCI_IOV
779 ppc_md.pcibios_fixup_resources =
780 pseries_pci_fixup_resources;
781 ppc_md.pcibios_fixup_sriov =
782 pseries_pci_fixup_iov_resources;
783 ppc_md.pcibios_iov_resource_alignment =
784 pseries_pci_iov_resource_alignment;
787 /* No special idle routine */
788 ppc_md.enable_pmcs = power4_enable_pmcs;
791 ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
793 if (swiotlb_force == SWIOTLB_FORCE)
794 ppc_swiotlb_enable = 1;
799 static void pseries_panic(char *str)
801 panic_flush_kmsg_end();
805 static int __init pSeries_init_panel(void)
807 /* Manually leave the kernel version on the panel. */
808 #ifdef __BIG_ENDIAN__
809 ppc_md.progress("Linux ppc64\n", 0);
811 ppc_md.progress("Linux ppc64le\n", 0);
813 ppc_md.progress(init_utsname()->version, 0);
817 machine_arch_initcall(pseries, pSeries_init_panel);
819 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
821 return plpar_hcall_norets(H_SET_DABR, dabr);
824 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
826 /* Have to set at least one bit in the DABRX according to PAPR */
827 if (dabrx == 0 && dabr == 0)
829 /* PAPR says we can only set kernel and user bits */
830 dabrx &= DABRX_KERNEL | DABRX_USER;
832 return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
835 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
837 /* PAPR says we can't set HYP */
840 return plpar_set_watchpoint0(dawr, dawrx);
843 #define CMO_CHARACTERISTICS_TOKEN 44
844 #define CMO_MAXLENGTH 1026
846 void pSeries_coalesce_init(void)
848 struct hvcall_mpp_x_data mpp_x_data;
850 if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
851 powerpc_firmware_features |= FW_FEATURE_XCMO;
853 powerpc_firmware_features &= ~FW_FEATURE_XCMO;
857 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
858 * handle that here. (Stolen from parse_system_parameter_string)
860 static void pSeries_cmo_feature_init(void)
862 char *ptr, *key, *value, *end;
864 int page_order = IOMMU_PAGE_SHIFT_4K;
866 pr_debug(" -> fw_cmo_feature_init()\n");
867 spin_lock(&rtas_data_buf_lock);
868 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
869 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
871 CMO_CHARACTERISTICS_TOKEN,
875 if (call_status != 0) {
876 spin_unlock(&rtas_data_buf_lock);
877 pr_debug("CMO not available\n");
878 pr_debug(" <- fw_cmo_feature_init()\n");
882 end = rtas_data_buf + CMO_MAXLENGTH - 2;
883 ptr = rtas_data_buf + 2; /* step over strlen value */
886 while (*ptr && (ptr <= end)) {
887 /* Separate the key and value by replacing '=' with '\0' and
888 * point the value at the string after the '='
893 } else if (ptr[0] == '\0' || ptr[0] == ',') {
894 /* Terminate the string containing the key/value pair */
898 pr_debug("Malformed key/value pair\n");
899 /* Never found a '=', end processing */
903 if (0 == strcmp(key, "CMOPageSize"))
904 page_order = simple_strtol(value, NULL, 10);
905 else if (0 == strcmp(key, "PrPSP"))
906 CMO_PrPSP = simple_strtol(value, NULL, 10);
907 else if (0 == strcmp(key, "SecPSP"))
908 CMO_SecPSP = simple_strtol(value, NULL, 10);
909 value = key = ptr + 1;
914 /* Page size is returned as the power of 2 of the page size,
915 * convert to the page size in bytes before returning
917 CMO_PageSize = 1 << page_order;
918 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
920 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
921 pr_info("CMO enabled\n");
922 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
924 powerpc_firmware_features |= FW_FEATURE_CMO;
925 pSeries_coalesce_init();
927 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
929 spin_unlock(&rtas_data_buf_lock);
930 pr_debug(" <- fw_cmo_feature_init()\n");
934 * Early initialization. Relocation is on but do not reference unbolted pages
936 static void __init pseries_init(void)
938 pr_debug(" -> pseries_init()\n");
940 #ifdef CONFIG_HVC_CONSOLE
941 if (firmware_has_feature(FW_FEATURE_LPAR))
942 hvc_vio_init_early();
944 if (firmware_has_feature(FW_FEATURE_XDABR))
945 ppc_md.set_dabr = pseries_set_xdabr;
946 else if (firmware_has_feature(FW_FEATURE_DABR))
947 ppc_md.set_dabr = pseries_set_dabr;
949 if (firmware_has_feature(FW_FEATURE_SET_MODE))
950 ppc_md.set_dawr = pseries_set_dawr;
952 pSeries_cmo_feature_init();
953 iommu_init_early_pSeries();
955 pr_debug(" <- pseries_init()\n");
959 * pseries_power_off - tell firmware about how to power off the system.
961 * This function calls either the power-off rtas token in normal cases
962 * or the ibm,power-off-ups token (if present & requested) in case of
963 * a power failure. If power-off token is used, power on will only be
964 * possible with power button press. If ibm,power-off-ups token is used
965 * it will allow auto poweron after power is restored.
967 static void pseries_power_off(void)
970 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
972 if (rtas_flash_term_hook)
973 rtas_flash_term_hook(SYS_POWER_OFF);
975 if (rtas_poweron_auto == 0 ||
976 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
977 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
978 printk(KERN_INFO "RTAS power-off returned %d\n", rc);
980 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
981 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
986 static int __init pSeries_probe(void)
988 if (!of_node_is_type(of_root, "chrp"))
991 /* Cell blades firmware claims to be chrp while it's not. Until this
992 * is fixed, we need to avoid those here.
994 if (of_machine_is_compatible("IBM,CPBW-1.0") ||
995 of_machine_is_compatible("IBM,CBEA"))
998 pm_power_off = pseries_power_off;
1000 pr_debug("Machine is%s LPAR !\n",
1001 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
1008 static int pSeries_pci_probe_mode(struct pci_bus *bus)
1010 if (firmware_has_feature(FW_FEATURE_LPAR))
1011 return PCI_PROBE_DEVTREE;
1012 return PCI_PROBE_NORMAL;
1015 struct pci_controller_ops pseries_pci_controller_ops = {
1016 .probe_mode = pSeries_pci_probe_mode,
1019 define_machine(pseries) {
1021 .probe = pSeries_probe,
1022 .setup_arch = pSeries_setup_arch,
1023 .init_IRQ = pseries_init_irq,
1024 .show_cpuinfo = pSeries_show_cpuinfo,
1025 .log_error = pSeries_log_error,
1026 .pcibios_fixup = pSeries_final_fixup,
1027 .restart = rtas_restart,
1029 .panic = pseries_panic,
1030 .get_boot_time = rtas_get_boot_time,
1031 .get_rtc_time = rtas_get_rtc_time,
1032 .set_rtc_time = rtas_set_rtc_time,
1033 .calibrate_decr = generic_calibrate_decr,
1034 .progress = rtas_progress,
1035 .system_reset_exception = pSeries_system_reset_exception,
1036 .machine_check_early = pseries_machine_check_realmode,
1037 .machine_check_exception = pSeries_machine_check_exception,
1038 #ifdef CONFIG_KEXEC_CORE
1039 .machine_kexec = pSeries_machine_kexec,
1040 .kexec_cpu_down = pseries_kexec_cpu_down,
1042 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
1043 .memory_block_size = pseries_memory_block_size,