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/memblock.h>
45 #include <linux/swiotlb.h>
48 #include <asm/processor.h>
50 #include <asm/pgtable.h>
53 #include <asm/pci-bridge.h>
54 #include <asm/iommu.h>
56 #include <asm/machdep.h>
59 #include <asm/nvram.h>
63 #include <asm/ppc-pci.h>
64 #include <asm/i8259.h>
67 #include <asm/firmware.h>
70 #include <asm/plpar_wrappers.h>
71 #include <asm/kexec.h>
72 #include <asm/isa-bridge.h>
73 #include <asm/security_features.h>
74 #include <asm/asm-const.h>
75 #include <asm/swiotlb.h>
78 #include "../../../../drivers/pci/pci.h"
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 /* Allocate per cpu slb area to save old slb contents during MCE */
149 size = sizeof(struct slb_entry) * mmu_slb_size * nr_cpus;
150 slb_ptr = memblock_alloc_try_nid_raw(size, sizeof(struct slb_entry),
151 MEMBLOCK_LOW_LIMIT, ppc64_rma_size,
154 panic("Failed to allocate %zu bytes below %pa for slb area\n",
155 size, &ppc64_rma_size);
157 for_each_possible_cpu(i)
158 paca_ptrs[i]->mce_faulty_slbs = slb_ptr + (mmu_slb_size * i);
162 static void pseries_8259_cascade(struct irq_desc *desc)
164 struct irq_chip *chip = irq_desc_get_chip(desc);
165 unsigned int cascade_irq = i8259_irq();
168 generic_handle_irq(cascade_irq);
170 chip->irq_eoi(&desc->irq_data);
173 static void __init pseries_setup_i8259_cascade(void)
175 struct device_node *np, *old, *found = NULL;
176 unsigned int cascade;
178 unsigned long intack = 0;
181 for_each_node_by_type(np, "interrupt-controller") {
182 if (of_device_is_compatible(np, "chrp,iic")) {
189 printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
193 cascade = irq_of_parse_and_map(found, 0);
195 printk(KERN_ERR "pic: failed to map cascade interrupt");
198 pr_debug("pic: cascade mapped to irq %d\n", cascade);
200 for (old = of_node_get(found); old != NULL ; old = np) {
201 np = of_get_parent(old);
205 if (!of_node_name_eq(np, "pci"))
207 addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
210 naddr = of_n_addr_cells(np);
211 intack = addrp[naddr-1];
213 intack |= ((unsigned long)addrp[naddr-2]) << 32;
216 printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
217 i8259_init(found, intack);
219 irq_set_chained_handler(cascade, pseries_8259_cascade);
222 static void __init pseries_init_irq(void)
224 /* Try using a XIVE if available, otherwise use a XICS */
225 if (!xive_spapr_init()) {
227 pseries_setup_i8259_cascade();
231 static void pseries_lpar_enable_pmcs(void)
233 unsigned long set, reset;
237 plpar_hcall_norets(H_PERFMON, set, reset);
240 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
242 struct of_reconfig_data *rd = data;
243 struct device_node *parent, *np = rd->dn;
248 case OF_RECONFIG_ATTACH_NODE:
249 parent = of_get_parent(np);
250 pdn = parent ? PCI_DN(parent) : NULL;
252 pci_add_device_node_info(pdn->phb, np);
256 case OF_RECONFIG_DETACH_NODE:
259 list_del(&pdn->list);
268 static struct notifier_block pci_dn_reconfig_nb = {
269 .notifier_call = pci_dn_reconfig_notifier,
272 struct kmem_cache *dtl_cache;
274 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
276 * Allocate space for the dispatch trace log for all possible cpus
277 * and register the buffers with the hypervisor. This is used for
278 * computing time stolen by the hypervisor.
280 static int alloc_dispatch_logs(void)
282 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
288 alloc_dtl_buffers(0);
290 /* Register the DTL for the current (boot) cpu */
291 register_dtl_buffer(smp_processor_id());
295 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
296 static inline int alloc_dispatch_logs(void)
300 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
302 static int alloc_dispatch_log_kmem_cache(void)
304 dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
305 DISPATCH_LOG_BYTES, 0, NULL);
307 pr_warn("Failed to create dispatch trace log buffer cache\n");
308 pr_warn("Stolen time statistics will be unreliable\n");
312 return alloc_dispatch_logs();
314 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
316 static void pseries_lpar_idle(void)
319 * Default handler to go into low thread priority and possibly
320 * low power mode by ceding processor to hypervisor
323 /* Indicate to hypervisor that we are idle. */
324 get_lppaca()->idle = 1;
327 * Yield the processor to the hypervisor. We return if
328 * an external interrupt occurs (which are driven prior
329 * to returning here) or if a prod occurs from another
330 * processor. When returning here, external interrupts
335 get_lppaca()->idle = 0;
339 * Enable relocation on during exceptions. This has partition wide scope and
340 * may take a while to complete, if it takes longer than one second we will
341 * just give up rather than wasting any more time on this - if that turns out
342 * to ever be a problem in practice we can move this into a kernel thread to
343 * finish off the process later in boot.
345 void pseries_enable_reloc_on_exc(void)
348 unsigned int delay, total_delay = 0;
351 rc = enable_reloc_on_exceptions();
352 if (!H_IS_LONG_BUSY(rc)) {
354 pr_info("Relocation on exceptions not"
356 } else if (rc != H_SUCCESS) {
357 pr_warn("Unable to enable relocation"
358 " on exceptions: %ld\n", rc);
363 delay = get_longbusy_msecs(rc);
364 total_delay += delay;
365 if (total_delay > 1000) {
366 pr_warn("Warning: Giving up waiting to enable "
367 "relocation on exceptions (%u msec)!\n",
375 EXPORT_SYMBOL(pseries_enable_reloc_on_exc);
377 void pseries_disable_reloc_on_exc(void)
382 rc = disable_reloc_on_exceptions();
383 if (!H_IS_LONG_BUSY(rc))
385 mdelay(get_longbusy_msecs(rc));
388 pr_warn("Warning: Failed to disable relocation on exceptions: %ld\n",
391 EXPORT_SYMBOL(pseries_disable_reloc_on_exc);
393 #ifdef CONFIG_KEXEC_CORE
394 static void pSeries_machine_kexec(struct kimage *image)
396 if (firmware_has_feature(FW_FEATURE_SET_MODE))
397 pseries_disable_reloc_on_exc();
399 default_machine_kexec(image);
403 #ifdef __LITTLE_ENDIAN__
404 void pseries_big_endian_exceptions(void)
409 rc = enable_big_endian_exceptions();
410 if (!H_IS_LONG_BUSY(rc))
412 mdelay(get_longbusy_msecs(rc));
416 * At this point it is unlikely panic() will get anything
417 * out to the user, since this is called very late in kexec
418 * but at least this will stop us from continuing on further
419 * and creating an even more difficult to debug situation.
421 * There is a known problem when kdump'ing, if cpus are offline
422 * the above call will fail. Rather than panicking again, keep
423 * going and hope the kdump kernel is also little endian, which
426 if (rc && !kdump_in_progress())
427 panic("Could not enable big endian exceptions");
430 void pseries_little_endian_exceptions(void)
435 rc = enable_little_endian_exceptions();
436 if (!H_IS_LONG_BUSY(rc))
438 mdelay(get_longbusy_msecs(rc));
441 ppc_md.progress("H_SET_MODE LE exception fail", 0);
442 panic("Could not enable little endian exceptions");
447 static void __init find_and_init_phbs(void)
449 struct device_node *node;
450 struct pci_controller *phb;
451 struct device_node *root = of_find_node_by_path("/");
453 for_each_child_of_node(root, node) {
454 if (!of_node_is_type(node, "pci") &&
455 !of_node_is_type(node, "pciex"))
458 phb = pcibios_alloc_controller(node);
462 pci_process_bridge_OF_ranges(phb, node, 0);
463 isa_bridge_find_early(phb);
464 phb->controller_ops = pseries_pci_controller_ops;
470 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
473 of_pci_check_probe_only();
476 static void init_cpu_char_feature_flags(struct h_cpu_char_result *result)
479 * The features below are disabled by default, so we instead look to see
480 * if firmware has *enabled* them, and set them if so.
482 if (result->character & H_CPU_CHAR_SPEC_BAR_ORI31)
483 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31);
485 if (result->character & H_CPU_CHAR_BCCTRL_SERIALISED)
486 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED);
488 if (result->character & H_CPU_CHAR_L1D_FLUSH_ORI30)
489 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30);
491 if (result->character & H_CPU_CHAR_L1D_FLUSH_TRIG2)
492 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2);
494 if (result->character & H_CPU_CHAR_L1D_THREAD_PRIV)
495 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV);
497 if (result->character & H_CPU_CHAR_COUNT_CACHE_DISABLED)
498 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED);
500 if (result->character & H_CPU_CHAR_BCCTR_FLUSH_ASSIST)
501 security_ftr_set(SEC_FTR_BCCTR_FLUSH_ASSIST);
503 if (result->behaviour & H_CPU_BEHAV_FLUSH_COUNT_CACHE)
504 security_ftr_set(SEC_FTR_FLUSH_COUNT_CACHE);
507 * The features below are enabled by default, so we instead look to see
508 * if firmware has *disabled* them, and clear them if so.
510 if (!(result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY))
511 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY);
513 if (!(result->behaviour & H_CPU_BEHAV_L1D_FLUSH_PR))
514 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR);
516 if (!(result->behaviour & H_CPU_BEHAV_BNDS_CHK_SPEC_BAR))
517 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR);
520 void pseries_setup_rfi_flush(void)
522 struct h_cpu_char_result result;
523 enum l1d_flush_type types;
528 * Set features to the defaults assumed by init_cpu_char_feature_flags()
529 * so it can set/clear again any features that might have changed after
530 * migration, and in case the hypercall fails and it is not even called.
532 powerpc_security_features = SEC_FTR_DEFAULT;
534 rc = plpar_get_cpu_characteristics(&result);
536 init_cpu_char_feature_flags(&result);
539 * We're the guest so this doesn't apply to us, clear it to simplify
540 * handling of it elsewhere.
542 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV);
544 types = L1D_FLUSH_FALLBACK;
546 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2))
547 types |= L1D_FLUSH_MTTRIG;
549 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30))
550 types |= L1D_FLUSH_ORI;
552 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) && \
553 security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR);
555 setup_rfi_flush(types, enable);
556 setup_count_cache_flush();
559 #ifdef CONFIG_PCI_IOV
560 enum rtas_iov_fw_value_map {
561 NUM_RES_PROPERTY = 0, /* Number of Resources */
562 LOW_INT = 1, /* Lowest 32 bits of Address */
563 START_OF_ENTRIES = 2, /* Always start of entry */
564 APERTURE_PROPERTY = 2, /* Start of entry+ to Aperture Size */
565 WDW_SIZE_PROPERTY = 4, /* Start of entry+ to Window Size */
566 NEXT_ENTRY = 7 /* Go to next entry on array */
569 enum get_iov_fw_value_index {
570 BAR_ADDRS = 1, /* Get Bar Address */
571 APERTURE_SIZE = 2, /* Get Aperture Size */
572 WDW_SIZE = 3 /* Get Window Size */
575 resource_size_t pseries_get_iov_fw_value(struct pci_dev *dev, int resno,
576 enum get_iov_fw_value_index value)
579 struct device_node *dn = pci_device_to_OF_node(dev);
580 int i, num_res, ret = 0;
582 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
587 * First element in the array is the number of Bars
588 * returned. Search through the list to find the matching
591 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
592 if (resno >= num_res)
593 return 0; /* or an errror */
595 i = START_OF_ENTRIES + NEXT_ENTRY * resno;
598 ret = of_read_number(&indexes[i], 2);
601 ret = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
604 ret = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
611 void of_pci_set_vf_bar_size(struct pci_dev *dev, const int *indexes)
613 struct resource *res;
614 resource_size_t base, size;
617 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
618 num_res = min_t(int, num_res, PCI_SRIOV_NUM_BARS);
619 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
620 i += NEXT_ENTRY, r++) {
621 res = &dev->resource[r + PCI_IOV_RESOURCES];
622 base = of_read_number(&indexes[i], 2);
623 size = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
624 res->flags = pci_parse_of_flags(of_read_number
625 (&indexes[i + LOW_INT], 1), 0);
626 res->flags |= (IORESOURCE_MEM_64 | IORESOURCE_PCI_FIXED);
627 res->name = pci_name(dev);
629 res->end = base + size - 1;
633 void of_pci_parse_iov_addrs(struct pci_dev *dev, const int *indexes)
635 struct resource *res, *root, *conflict;
636 resource_size_t base, size;
640 * First element in the array is the number of Bars
641 * returned. Search through the list to find the matching
642 * bars assign them from firmware into resources structure.
644 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
645 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
646 i += NEXT_ENTRY, r++) {
647 res = &dev->resource[r + PCI_IOV_RESOURCES];
648 base = of_read_number(&indexes[i], 2);
649 size = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
650 res->name = pci_name(dev);
652 res->end = base + size - 1;
653 root = &iomem_resource;
655 "pSeries IOV BAR %d: trying firmware assignment %pR\n",
656 r + PCI_IOV_RESOURCES, res);
657 conflict = request_resource_conflict(root, res);
660 "BAR %d: %pR conflicts with %s %pR\n",
661 r + PCI_IOV_RESOURCES, res,
662 conflict->name, conflict);
663 res->flags |= IORESOURCE_UNSET;
668 static void pseries_disable_sriov_resources(struct pci_dev *pdev)
672 pci_warn(pdev, "No hypervisor support for SR-IOV on this device, IOV BARs disabled.\n");
673 for (i = 0; i < PCI_SRIOV_NUM_BARS; i++)
674 pdev->resource[i + PCI_IOV_RESOURCES].flags = 0;
677 static void pseries_pci_fixup_resources(struct pci_dev *pdev)
680 struct device_node *dn = pci_device_to_OF_node(pdev);
682 /*Firmware must support open sriov otherwise dont configure*/
683 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
685 of_pci_set_vf_bar_size(pdev, indexes);
687 pseries_disable_sriov_resources(pdev);
690 static void pseries_pci_fixup_iov_resources(struct pci_dev *pdev)
693 struct device_node *dn = pci_device_to_OF_node(pdev);
695 if (!pdev->is_physfn || pci_dev_is_added(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_parse_iov_addrs(pdev, indexes);
702 pseries_disable_sriov_resources(pdev);
705 static resource_size_t pseries_pci_iov_resource_alignment(struct pci_dev *pdev,
709 struct device_node *dn = pci_device_to_OF_node(pdev);
711 /*Firmware must support open sriov otherwise report regular alignment*/
712 reg = of_get_property(dn, "ibm,is-open-sriov-pf", NULL);
714 return pci_iov_resource_size(pdev, resno);
716 if (!pdev->is_physfn)
718 return pseries_get_iov_fw_value(pdev,
719 resno - PCI_IOV_RESOURCES,
724 static void __init pSeries_setup_arch(void)
726 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
728 /* Discover PIC type and setup ppc_md accordingly */
732 /* openpic global configuration register (64-bit format). */
733 /* openpic Interrupt Source Unit pointer (64-bit format). */
734 /* python0 facility area (mmio) (64-bit format) REAL address. */
736 /* init to some ~sane value until calibrate_delay() runs */
737 loops_per_jiffy = 50000000;
741 pseries_setup_rfi_flush();
744 /* By default, only probe PCI (can be overridden by rtas_pci) */
745 pci_add_flags(PCI_PROBE_ONLY);
747 /* Find and initialize PCI host bridges */
748 init_pci_config_tokens();
749 find_and_init_phbs();
750 of_reconfig_notifier_register(&pci_dn_reconfig_nb);
752 pSeries_nvram_init();
754 if (firmware_has_feature(FW_FEATURE_LPAR)) {
755 vpa_init(boot_cpuid);
756 ppc_md.power_save = pseries_lpar_idle;
757 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
758 #ifdef CONFIG_PCI_IOV
759 ppc_md.pcibios_fixup_resources =
760 pseries_pci_fixup_resources;
761 ppc_md.pcibios_fixup_sriov =
762 pseries_pci_fixup_iov_resources;
763 ppc_md.pcibios_iov_resource_alignment =
764 pseries_pci_iov_resource_alignment;
767 /* No special idle routine */
768 ppc_md.enable_pmcs = power4_enable_pmcs;
771 ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
773 if (swiotlb_force == SWIOTLB_FORCE)
774 ppc_swiotlb_enable = 1;
777 static void pseries_panic(char *str)
779 panic_flush_kmsg_end();
783 static int __init pSeries_init_panel(void)
785 /* Manually leave the kernel version on the panel. */
786 #ifdef __BIG_ENDIAN__
787 ppc_md.progress("Linux ppc64\n", 0);
789 ppc_md.progress("Linux ppc64le\n", 0);
791 ppc_md.progress(init_utsname()->version, 0);
795 machine_arch_initcall(pseries, pSeries_init_panel);
797 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
799 return plpar_hcall_norets(H_SET_DABR, dabr);
802 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
804 /* Have to set at least one bit in the DABRX according to PAPR */
805 if (dabrx == 0 && dabr == 0)
807 /* PAPR says we can only set kernel and user bits */
808 dabrx &= DABRX_KERNEL | DABRX_USER;
810 return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
813 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
815 /* PAPR says we can't set HYP */
818 return plpar_set_watchpoint0(dawr, dawrx);
821 #define CMO_CHARACTERISTICS_TOKEN 44
822 #define CMO_MAXLENGTH 1026
824 void pSeries_coalesce_init(void)
826 struct hvcall_mpp_x_data mpp_x_data;
828 if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
829 powerpc_firmware_features |= FW_FEATURE_XCMO;
831 powerpc_firmware_features &= ~FW_FEATURE_XCMO;
835 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
836 * handle that here. (Stolen from parse_system_parameter_string)
838 static void pSeries_cmo_feature_init(void)
840 char *ptr, *key, *value, *end;
842 int page_order = IOMMU_PAGE_SHIFT_4K;
844 pr_debug(" -> fw_cmo_feature_init()\n");
845 spin_lock(&rtas_data_buf_lock);
846 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
847 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
849 CMO_CHARACTERISTICS_TOKEN,
853 if (call_status != 0) {
854 spin_unlock(&rtas_data_buf_lock);
855 pr_debug("CMO not available\n");
856 pr_debug(" <- fw_cmo_feature_init()\n");
860 end = rtas_data_buf + CMO_MAXLENGTH - 2;
861 ptr = rtas_data_buf + 2; /* step over strlen value */
864 while (*ptr && (ptr <= end)) {
865 /* Separate the key and value by replacing '=' with '\0' and
866 * point the value at the string after the '='
871 } else if (ptr[0] == '\0' || ptr[0] == ',') {
872 /* Terminate the string containing the key/value pair */
876 pr_debug("Malformed key/value pair\n");
877 /* Never found a '=', end processing */
881 if (0 == strcmp(key, "CMOPageSize"))
882 page_order = simple_strtol(value, NULL, 10);
883 else if (0 == strcmp(key, "PrPSP"))
884 CMO_PrPSP = simple_strtol(value, NULL, 10);
885 else if (0 == strcmp(key, "SecPSP"))
886 CMO_SecPSP = simple_strtol(value, NULL, 10);
887 value = key = ptr + 1;
892 /* Page size is returned as the power of 2 of the page size,
893 * convert to the page size in bytes before returning
895 CMO_PageSize = 1 << page_order;
896 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
898 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
899 pr_info("CMO enabled\n");
900 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
902 powerpc_firmware_features |= FW_FEATURE_CMO;
903 pSeries_coalesce_init();
905 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
907 spin_unlock(&rtas_data_buf_lock);
908 pr_debug(" <- fw_cmo_feature_init()\n");
912 * Early initialization. Relocation is on but do not reference unbolted pages
914 static void __init pseries_init(void)
916 pr_debug(" -> pseries_init()\n");
918 #ifdef CONFIG_HVC_CONSOLE
919 if (firmware_has_feature(FW_FEATURE_LPAR))
920 hvc_vio_init_early();
922 if (firmware_has_feature(FW_FEATURE_XDABR))
923 ppc_md.set_dabr = pseries_set_xdabr;
924 else if (firmware_has_feature(FW_FEATURE_DABR))
925 ppc_md.set_dabr = pseries_set_dabr;
927 if (firmware_has_feature(FW_FEATURE_SET_MODE))
928 ppc_md.set_dawr = pseries_set_dawr;
930 pSeries_cmo_feature_init();
931 iommu_init_early_pSeries();
933 pr_debug(" <- pseries_init()\n");
937 * pseries_power_off - tell firmware about how to power off the system.
939 * This function calls either the power-off rtas token in normal cases
940 * or the ibm,power-off-ups token (if present & requested) in case of
941 * a power failure. If power-off token is used, power on will only be
942 * possible with power button press. If ibm,power-off-ups token is used
943 * it will allow auto poweron after power is restored.
945 static void pseries_power_off(void)
948 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
950 if (rtas_flash_term_hook)
951 rtas_flash_term_hook(SYS_POWER_OFF);
953 if (rtas_poweron_auto == 0 ||
954 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
955 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
956 printk(KERN_INFO "RTAS power-off returned %d\n", rc);
958 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
959 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
964 static int __init pSeries_probe(void)
966 if (!of_node_is_type(of_root, "chrp"))
969 /* Cell blades firmware claims to be chrp while it's not. Until this
970 * is fixed, we need to avoid those here.
972 if (of_machine_is_compatible("IBM,CPBW-1.0") ||
973 of_machine_is_compatible("IBM,CBEA"))
976 pm_power_off = pseries_power_off;
978 pr_debug("Machine is%s LPAR !\n",
979 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
986 static int pSeries_pci_probe_mode(struct pci_bus *bus)
988 if (firmware_has_feature(FW_FEATURE_LPAR))
989 return PCI_PROBE_DEVTREE;
990 return PCI_PROBE_NORMAL;
993 struct pci_controller_ops pseries_pci_controller_ops = {
994 .probe_mode = pSeries_pci_probe_mode,
997 define_machine(pseries) {
999 .probe = pSeries_probe,
1000 .setup_arch = pSeries_setup_arch,
1001 .init_IRQ = pseries_init_irq,
1002 .show_cpuinfo = pSeries_show_cpuinfo,
1003 .log_error = pSeries_log_error,
1004 .pcibios_fixup = pSeries_final_fixup,
1005 .restart = rtas_restart,
1007 .panic = pseries_panic,
1008 .get_boot_time = rtas_get_boot_time,
1009 .get_rtc_time = rtas_get_rtc_time,
1010 .set_rtc_time = rtas_set_rtc_time,
1011 .calibrate_decr = generic_calibrate_decr,
1012 .progress = rtas_progress,
1013 .system_reset_exception = pSeries_system_reset_exception,
1014 .machine_check_early = pseries_machine_check_realmode,
1015 .machine_check_exception = pSeries_machine_check_exception,
1016 #ifdef CONFIG_KEXEC_CORE
1017 .machine_kexec = pSeries_machine_kexec,
1018 .kexec_cpu_down = pseries_kexec_cpu_down,
1020 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
1021 .memory_block_size = pseries_memory_block_size,