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>
47 #include <asm/processor.h>
49 #include <asm/pgtable.h>
52 #include <asm/pci-bridge.h>
53 #include <asm/iommu.h>
55 #include <asm/machdep.h>
58 #include <asm/nvram.h>
62 #include <asm/ppc-pci.h>
63 #include <asm/i8259.h>
66 #include <asm/firmware.h>
69 #include <asm/plpar_wrappers.h>
70 #include <asm/kexec.h>
71 #include <asm/isa-bridge.h>
72 #include <asm/security_features.h>
73 #include <asm/asm-const.h>
76 #include "../../../../drivers/pci/pci.h"
80 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);
81 EXPORT_SYMBOL(CMO_PageSize);
83 int fwnmi_active; /* TRUE if an FWNMI handler is present */
85 static void pSeries_show_cpuinfo(struct seq_file *m)
87 struct device_node *root;
88 const char *model = "";
90 root = of_find_node_by_path("/");
92 model = of_get_property(root, "model", NULL);
93 seq_printf(m, "machine\t\t: CHRP %s\n", model);
96 seq_printf(m, "MMU\t\t: Radix\n");
98 seq_printf(m, "MMU\t\t: Hash\n");
101 /* Initialize firmware assisted non-maskable interrupts if
102 * the firmware supports this feature.
104 static void __init fwnmi_init(void)
106 unsigned long system_reset_addr, machine_check_addr;
109 int nr_cpus = num_possible_cpus();
110 #ifdef CONFIG_PPC_BOOK3S_64
111 struct slb_entry *slb_ptr;
115 int ibm_nmi_register = rtas_token("ibm,nmi-register");
116 if (ibm_nmi_register == RTAS_UNKNOWN_SERVICE)
119 /* If the kernel's not linked at zero we point the firmware at low
120 * addresses anyway, and use a trampoline to get to the real code. */
121 system_reset_addr = __pa(system_reset_fwnmi) - PHYSICAL_START;
122 machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START;
124 if (0 == rtas_call(ibm_nmi_register, 2, 1, NULL, system_reset_addr,
129 * Allocate a chunk for per cpu buffer to hold rtas errorlog.
130 * It will be used in real mode mce handler, hence it needs to be
133 mce_data_buf = __va(memblock_alloc_base(RTAS_ERROR_LOG_MAX * nr_cpus,
134 RTAS_ERROR_LOG_MAX, ppc64_rma_size));
135 for_each_possible_cpu(i) {
136 paca_ptrs[i]->mce_data_buf = mce_data_buf +
137 (RTAS_ERROR_LOG_MAX * i);
140 #ifdef CONFIG_PPC_BOOK3S_64
141 /* Allocate per cpu slb area to save old slb contents during MCE */
142 size = sizeof(struct slb_entry) * mmu_slb_size * nr_cpus;
143 slb_ptr = __va(memblock_alloc_base(size, sizeof(struct slb_entry),
145 for_each_possible_cpu(i)
146 paca_ptrs[i]->mce_faulty_slbs = slb_ptr + (mmu_slb_size * i);
150 static void pseries_8259_cascade(struct irq_desc *desc)
152 struct irq_chip *chip = irq_desc_get_chip(desc);
153 unsigned int cascade_irq = i8259_irq();
156 generic_handle_irq(cascade_irq);
158 chip->irq_eoi(&desc->irq_data);
161 static void __init pseries_setup_i8259_cascade(void)
163 struct device_node *np, *old, *found = NULL;
164 unsigned int cascade;
166 unsigned long intack = 0;
169 for_each_node_by_type(np, "interrupt-controller") {
170 if (of_device_is_compatible(np, "chrp,iic")) {
177 printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
181 cascade = irq_of_parse_and_map(found, 0);
183 printk(KERN_ERR "pic: failed to map cascade interrupt");
186 pr_debug("pic: cascade mapped to irq %d\n", cascade);
188 for (old = of_node_get(found); old != NULL ; old = np) {
189 np = of_get_parent(old);
193 if (strcmp(np->name, "pci") != 0)
195 addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
198 naddr = of_n_addr_cells(np);
199 intack = addrp[naddr-1];
201 intack |= ((unsigned long)addrp[naddr-2]) << 32;
204 printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
205 i8259_init(found, intack);
207 irq_set_chained_handler(cascade, pseries_8259_cascade);
210 static void __init pseries_init_irq(void)
212 /* Try using a XIVE if available, otherwise use a XICS */
213 if (!xive_spapr_init()) {
215 pseries_setup_i8259_cascade();
219 static void pseries_lpar_enable_pmcs(void)
221 unsigned long set, reset;
225 plpar_hcall_norets(H_PERFMON, set, reset);
228 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
230 struct of_reconfig_data *rd = data;
231 struct device_node *parent, *np = rd->dn;
236 case OF_RECONFIG_ATTACH_NODE:
237 parent = of_get_parent(np);
238 pdn = parent ? PCI_DN(parent) : NULL;
240 pci_add_device_node_info(pdn->phb, np);
244 case OF_RECONFIG_DETACH_NODE:
247 list_del(&pdn->list);
256 static struct notifier_block pci_dn_reconfig_nb = {
257 .notifier_call = pci_dn_reconfig_notifier,
260 struct kmem_cache *dtl_cache;
262 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
264 * Allocate space for the dispatch trace log for all possible cpus
265 * and register the buffers with the hypervisor. This is used for
266 * computing time stolen by the hypervisor.
268 static int alloc_dispatch_logs(void)
271 struct paca_struct *pp;
272 struct dtl_entry *dtl;
274 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
280 for_each_possible_cpu(cpu) {
282 dtl = kmem_cache_alloc(dtl_cache, GFP_KERNEL);
284 pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
286 pr_warn("Stolen time statistics will be unreliable\n");
291 pp->dispatch_log = dtl;
292 pp->dispatch_log_end = dtl + N_DISPATCH_LOG;
296 /* Register the DTL for the current (boot) cpu */
297 dtl = get_paca()->dispatch_log;
298 get_paca()->dtl_ridx = 0;
299 get_paca()->dtl_curr = dtl;
300 get_paca()->lppaca_ptr->dtl_idx = 0;
302 /* hypervisor reads buffer length from this field */
303 dtl->enqueue_to_dispatch_time = cpu_to_be32(DISPATCH_LOG_BYTES);
304 ret = register_dtl(hard_smp_processor_id(), __pa(dtl));
306 pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
307 "with %d\n", smp_processor_id(),
308 hard_smp_processor_id(), ret);
309 get_paca()->lppaca_ptr->dtl_enable_mask = 2;
313 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
314 static inline int alloc_dispatch_logs(void)
318 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
320 static int alloc_dispatch_log_kmem_cache(void)
322 dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
323 DISPATCH_LOG_BYTES, 0, NULL);
325 pr_warn("Failed to create dispatch trace log buffer cache\n");
326 pr_warn("Stolen time statistics will be unreliable\n");
330 return alloc_dispatch_logs();
332 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
334 static void pseries_lpar_idle(void)
337 * Default handler to go into low thread priority and possibly
338 * low power mode by ceding processor to hypervisor
341 /* Indicate to hypervisor that we are idle. */
342 get_lppaca()->idle = 1;
345 * Yield the processor to the hypervisor. We return if
346 * an external interrupt occurs (which are driven prior
347 * to returning here) or if a prod occurs from another
348 * processor. When returning here, external interrupts
353 get_lppaca()->idle = 0;
357 * Enable relocation on during exceptions. This has partition wide scope and
358 * may take a while to complete, if it takes longer than one second we will
359 * just give up rather than wasting any more time on this - if that turns out
360 * to ever be a problem in practice we can move this into a kernel thread to
361 * finish off the process later in boot.
363 void pseries_enable_reloc_on_exc(void)
366 unsigned int delay, total_delay = 0;
369 rc = enable_reloc_on_exceptions();
370 if (!H_IS_LONG_BUSY(rc)) {
372 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);
381 delay = get_longbusy_msecs(rc);
382 total_delay += delay;
383 if (total_delay > 1000) {
384 pr_warn("Warning: Giving up waiting to enable "
385 "relocation on exceptions (%u msec)!\n",
393 EXPORT_SYMBOL(pseries_enable_reloc_on_exc);
395 void pseries_disable_reloc_on_exc(void)
400 rc = disable_reloc_on_exceptions();
401 if (!H_IS_LONG_BUSY(rc))
403 mdelay(get_longbusy_msecs(rc));
406 pr_warn("Warning: Failed to disable relocation on exceptions: %ld\n",
409 EXPORT_SYMBOL(pseries_disable_reloc_on_exc);
411 #ifdef CONFIG_KEXEC_CORE
412 static void pSeries_machine_kexec(struct kimage *image)
414 if (firmware_has_feature(FW_FEATURE_SET_MODE))
415 pseries_disable_reloc_on_exc();
417 default_machine_kexec(image);
421 #ifdef __LITTLE_ENDIAN__
422 void pseries_big_endian_exceptions(void)
427 rc = enable_big_endian_exceptions();
428 if (!H_IS_LONG_BUSY(rc))
430 mdelay(get_longbusy_msecs(rc));
434 * At this point it is unlikely panic() will get anything
435 * out to the user, since this is called very late in kexec
436 * but at least this will stop us from continuing on further
437 * and creating an even more difficult to debug situation.
439 * There is a known problem when kdump'ing, if cpus are offline
440 * the above call will fail. Rather than panicking again, keep
441 * going and hope the kdump kernel is also little endian, which
444 if (rc && !kdump_in_progress())
445 panic("Could not enable big endian exceptions");
448 void pseries_little_endian_exceptions(void)
453 rc = enable_little_endian_exceptions();
454 if (!H_IS_LONG_BUSY(rc))
456 mdelay(get_longbusy_msecs(rc));
459 ppc_md.progress("H_SET_MODE LE exception fail", 0);
460 panic("Could not enable little endian exceptions");
465 static void __init find_and_init_phbs(void)
467 struct device_node *node;
468 struct pci_controller *phb;
469 struct device_node *root = of_find_node_by_path("/");
471 for_each_child_of_node(root, node) {
472 if (node->type == NULL || (strcmp(node->type, "pci") != 0 &&
473 strcmp(node->type, "pciex") != 0))
476 phb = pcibios_alloc_controller(node);
480 pci_process_bridge_OF_ranges(phb, node, 0);
481 isa_bridge_find_early(phb);
482 phb->controller_ops = pseries_pci_controller_ops;
488 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
491 of_pci_check_probe_only();
494 static void init_cpu_char_feature_flags(struct h_cpu_char_result *result)
497 * The features below are disabled by default, so we instead look to see
498 * if firmware has *enabled* them, and set them if so.
500 if (result->character & H_CPU_CHAR_SPEC_BAR_ORI31)
501 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31);
503 if (result->character & H_CPU_CHAR_BCCTRL_SERIALISED)
504 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED);
506 if (result->character & H_CPU_CHAR_L1D_FLUSH_ORI30)
507 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30);
509 if (result->character & H_CPU_CHAR_L1D_FLUSH_TRIG2)
510 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2);
512 if (result->character & H_CPU_CHAR_L1D_THREAD_PRIV)
513 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV);
515 if (result->character & H_CPU_CHAR_COUNT_CACHE_DISABLED)
516 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED);
518 if (result->character & H_CPU_CHAR_BCCTR_FLUSH_ASSIST)
519 security_ftr_set(SEC_FTR_BCCTR_FLUSH_ASSIST);
521 if (result->behaviour & H_CPU_BEHAV_FLUSH_COUNT_CACHE)
522 security_ftr_set(SEC_FTR_FLUSH_COUNT_CACHE);
525 * The features below are enabled by default, so we instead look to see
526 * if firmware has *disabled* them, and clear them if so.
528 if (!(result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY))
529 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY);
531 if (!(result->behaviour & H_CPU_BEHAV_L1D_FLUSH_PR))
532 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR);
534 if (!(result->behaviour & H_CPU_BEHAV_BNDS_CHK_SPEC_BAR))
535 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR);
538 void pseries_setup_rfi_flush(void)
540 struct h_cpu_char_result result;
541 enum l1d_flush_type types;
546 * Set features to the defaults assumed by init_cpu_char_feature_flags()
547 * so it can set/clear again any features that might have changed after
548 * migration, and in case the hypercall fails and it is not even called.
550 powerpc_security_features = SEC_FTR_DEFAULT;
552 rc = plpar_get_cpu_characteristics(&result);
554 init_cpu_char_feature_flags(&result);
557 * We're the guest so this doesn't apply to us, clear it to simplify
558 * handling of it elsewhere.
560 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV);
562 types = L1D_FLUSH_FALLBACK;
564 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2))
565 types |= L1D_FLUSH_MTTRIG;
567 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30))
568 types |= L1D_FLUSH_ORI;
570 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) && \
571 security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR);
573 setup_rfi_flush(types, enable);
574 setup_count_cache_flush();
577 #ifdef CONFIG_PCI_IOV
578 enum rtas_iov_fw_value_map {
579 NUM_RES_PROPERTY = 0, /* Number of Resources */
580 LOW_INT = 1, /* Lowest 32 bits of Address */
581 START_OF_ENTRIES = 2, /* Always start of entry */
582 APERTURE_PROPERTY = 2, /* Start of entry+ to Aperture Size */
583 WDW_SIZE_PROPERTY = 4, /* Start of entry+ to Window Size */
584 NEXT_ENTRY = 7 /* Go to next entry on array */
587 enum get_iov_fw_value_index {
588 BAR_ADDRS = 1, /* Get Bar Address */
589 APERTURE_SIZE = 2, /* Get Aperture Size */
590 WDW_SIZE = 3 /* Get Window Size */
593 resource_size_t pseries_get_iov_fw_value(struct pci_dev *dev, int resno,
594 enum get_iov_fw_value_index value)
597 struct device_node *dn = pci_device_to_OF_node(dev);
598 int i, num_res, ret = 0;
600 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
605 * First element in the array is the number of Bars
606 * returned. Search through the list to find the matching
609 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
610 if (resno >= num_res)
611 return 0; /* or an errror */
613 i = START_OF_ENTRIES + NEXT_ENTRY * resno;
616 ret = of_read_number(&indexes[i], 2);
619 ret = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
622 ret = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
629 void of_pci_set_vf_bar_size(struct pci_dev *dev, const int *indexes)
631 struct resource *res;
632 resource_size_t base, size;
635 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
636 num_res = min_t(int, num_res, PCI_SRIOV_NUM_BARS);
637 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
638 i += NEXT_ENTRY, r++) {
639 res = &dev->resource[r + PCI_IOV_RESOURCES];
640 base = of_read_number(&indexes[i], 2);
641 size = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
642 res->flags = pci_parse_of_flags(of_read_number
643 (&indexes[i + LOW_INT], 1), 0);
644 res->flags |= (IORESOURCE_MEM_64 | IORESOURCE_PCI_FIXED);
645 res->name = pci_name(dev);
647 res->end = base + size - 1;
651 void of_pci_parse_iov_addrs(struct pci_dev *dev, const int *indexes)
653 struct resource *res, *root, *conflict;
654 resource_size_t base, size;
658 * First element in the array is the number of Bars
659 * returned. Search through the list to find the matching
660 * bars assign them from firmware into resources structure.
662 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
663 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
664 i += NEXT_ENTRY, r++) {
665 res = &dev->resource[r + PCI_IOV_RESOURCES];
666 base = of_read_number(&indexes[i], 2);
667 size = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
668 res->name = pci_name(dev);
670 res->end = base + size - 1;
671 root = &iomem_resource;
673 "pSeries IOV BAR %d: trying firmware assignment %pR\n",
674 r + PCI_IOV_RESOURCES, res);
675 conflict = request_resource_conflict(root, res);
678 "BAR %d: %pR conflicts with %s %pR\n",
679 r + PCI_IOV_RESOURCES, res,
680 conflict->name, conflict);
681 res->flags |= IORESOURCE_UNSET;
686 static void pseries_disable_sriov_resources(struct pci_dev *pdev)
690 pci_warn(pdev, "No hypervisor support for SR-IOV on this device, IOV BARs disabled.\n");
691 for (i = 0; i < PCI_SRIOV_NUM_BARS; i++)
692 pdev->resource[i + PCI_IOV_RESOURCES].flags = 0;
695 static void pseries_pci_fixup_resources(struct pci_dev *pdev)
698 struct device_node *dn = pci_device_to_OF_node(pdev);
700 /*Firmware must support open sriov otherwise dont configure*/
701 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
703 of_pci_set_vf_bar_size(pdev, indexes);
705 pseries_disable_sriov_resources(pdev);
708 static void pseries_pci_fixup_iov_resources(struct pci_dev *pdev)
711 struct device_node *dn = pci_device_to_OF_node(pdev);
713 if (!pdev->is_physfn || pci_dev_is_added(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_parse_iov_addrs(pdev, indexes);
720 pseries_disable_sriov_resources(pdev);
723 static resource_size_t pseries_pci_iov_resource_alignment(struct pci_dev *pdev,
727 struct device_node *dn = pci_device_to_OF_node(pdev);
729 /*Firmware must support open sriov otherwise report regular alignment*/
730 reg = of_get_property(dn, "ibm,is-open-sriov-pf", NULL);
732 return pci_iov_resource_size(pdev, resno);
734 if (!pdev->is_physfn)
736 return pseries_get_iov_fw_value(pdev,
737 resno - PCI_IOV_RESOURCES,
742 static void __init pSeries_setup_arch(void)
744 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
746 /* Discover PIC type and setup ppc_md accordingly */
750 /* openpic global configuration register (64-bit format). */
751 /* openpic Interrupt Source Unit pointer (64-bit format). */
752 /* python0 facility area (mmio) (64-bit format) REAL address. */
754 /* init to some ~sane value until calibrate_delay() runs */
755 loops_per_jiffy = 50000000;
759 pseries_setup_rfi_flush();
762 /* By default, only probe PCI (can be overridden by rtas_pci) */
763 pci_add_flags(PCI_PROBE_ONLY);
765 /* Find and initialize PCI host bridges */
766 init_pci_config_tokens();
767 find_and_init_phbs();
768 of_reconfig_notifier_register(&pci_dn_reconfig_nb);
770 pSeries_nvram_init();
772 if (firmware_has_feature(FW_FEATURE_LPAR)) {
773 vpa_init(boot_cpuid);
774 ppc_md.power_save = pseries_lpar_idle;
775 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
776 #ifdef CONFIG_PCI_IOV
777 ppc_md.pcibios_fixup_resources =
778 pseries_pci_fixup_resources;
779 ppc_md.pcibios_fixup_sriov =
780 pseries_pci_fixup_iov_resources;
781 ppc_md.pcibios_iov_resource_alignment =
782 pseries_pci_iov_resource_alignment;
785 /* No special idle routine */
786 ppc_md.enable_pmcs = power4_enable_pmcs;
789 ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
792 static void pseries_panic(char *str)
794 panic_flush_kmsg_end();
798 static int __init pSeries_init_panel(void)
800 /* Manually leave the kernel version on the panel. */
801 #ifdef __BIG_ENDIAN__
802 ppc_md.progress("Linux ppc64\n", 0);
804 ppc_md.progress("Linux ppc64le\n", 0);
806 ppc_md.progress(init_utsname()->version, 0);
810 machine_arch_initcall(pseries, pSeries_init_panel);
812 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
814 return plpar_hcall_norets(H_SET_DABR, dabr);
817 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
819 /* Have to set at least one bit in the DABRX according to PAPR */
820 if (dabrx == 0 && dabr == 0)
822 /* PAPR says we can only set kernel and user bits */
823 dabrx &= DABRX_KERNEL | DABRX_USER;
825 return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
828 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
830 /* PAPR says we can't set HYP */
833 return plpar_set_watchpoint0(dawr, dawrx);
836 #define CMO_CHARACTERISTICS_TOKEN 44
837 #define CMO_MAXLENGTH 1026
839 void pSeries_coalesce_init(void)
841 struct hvcall_mpp_x_data mpp_x_data;
843 if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
844 powerpc_firmware_features |= FW_FEATURE_XCMO;
846 powerpc_firmware_features &= ~FW_FEATURE_XCMO;
850 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
851 * handle that here. (Stolen from parse_system_parameter_string)
853 static void pSeries_cmo_feature_init(void)
855 char *ptr, *key, *value, *end;
857 int page_order = IOMMU_PAGE_SHIFT_4K;
859 pr_debug(" -> fw_cmo_feature_init()\n");
860 spin_lock(&rtas_data_buf_lock);
861 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
862 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
864 CMO_CHARACTERISTICS_TOKEN,
868 if (call_status != 0) {
869 spin_unlock(&rtas_data_buf_lock);
870 pr_debug("CMO not available\n");
871 pr_debug(" <- fw_cmo_feature_init()\n");
875 end = rtas_data_buf + CMO_MAXLENGTH - 2;
876 ptr = rtas_data_buf + 2; /* step over strlen value */
879 while (*ptr && (ptr <= end)) {
880 /* Separate the key and value by replacing '=' with '\0' and
881 * point the value at the string after the '='
886 } else if (ptr[0] == '\0' || ptr[0] == ',') {
887 /* Terminate the string containing the key/value pair */
891 pr_debug("Malformed key/value pair\n");
892 /* Never found a '=', end processing */
896 if (0 == strcmp(key, "CMOPageSize"))
897 page_order = simple_strtol(value, NULL, 10);
898 else if (0 == strcmp(key, "PrPSP"))
899 CMO_PrPSP = simple_strtol(value, NULL, 10);
900 else if (0 == strcmp(key, "SecPSP"))
901 CMO_SecPSP = simple_strtol(value, NULL, 10);
902 value = key = ptr + 1;
907 /* Page size is returned as the power of 2 of the page size,
908 * convert to the page size in bytes before returning
910 CMO_PageSize = 1 << page_order;
911 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
913 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
914 pr_info("CMO enabled\n");
915 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
917 powerpc_firmware_features |= FW_FEATURE_CMO;
918 pSeries_coalesce_init();
920 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
922 spin_unlock(&rtas_data_buf_lock);
923 pr_debug(" <- fw_cmo_feature_init()\n");
927 * Early initialization. Relocation is on but do not reference unbolted pages
929 static void __init pseries_init(void)
931 pr_debug(" -> pseries_init()\n");
933 #ifdef CONFIG_HVC_CONSOLE
934 if (firmware_has_feature(FW_FEATURE_LPAR))
935 hvc_vio_init_early();
937 if (firmware_has_feature(FW_FEATURE_XDABR))
938 ppc_md.set_dabr = pseries_set_xdabr;
939 else if (firmware_has_feature(FW_FEATURE_DABR))
940 ppc_md.set_dabr = pseries_set_dabr;
942 if (firmware_has_feature(FW_FEATURE_SET_MODE))
943 ppc_md.set_dawr = pseries_set_dawr;
945 pSeries_cmo_feature_init();
946 iommu_init_early_pSeries();
948 pr_debug(" <- pseries_init()\n");
952 * pseries_power_off - tell firmware about how to power off the system.
954 * This function calls either the power-off rtas token in normal cases
955 * or the ibm,power-off-ups token (if present & requested) in case of
956 * a power failure. If power-off token is used, power on will only be
957 * possible with power button press. If ibm,power-off-ups token is used
958 * it will allow auto poweron after power is restored.
960 static void pseries_power_off(void)
963 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
965 if (rtas_flash_term_hook)
966 rtas_flash_term_hook(SYS_POWER_OFF);
968 if (rtas_poweron_auto == 0 ||
969 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
970 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
971 printk(KERN_INFO "RTAS power-off returned %d\n", rc);
973 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
974 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
979 static int __init pSeries_probe(void)
981 const char *dtype = of_get_property(of_root, "device_type", NULL);
985 if (strcmp(dtype, "chrp"))
988 /* Cell blades firmware claims to be chrp while it's not. Until this
989 * is fixed, we need to avoid those here.
991 if (of_machine_is_compatible("IBM,CPBW-1.0") ||
992 of_machine_is_compatible("IBM,CBEA"))
995 pm_power_off = pseries_power_off;
997 pr_debug("Machine is%s LPAR !\n",
998 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
1005 static int pSeries_pci_probe_mode(struct pci_bus *bus)
1007 if (firmware_has_feature(FW_FEATURE_LPAR))
1008 return PCI_PROBE_DEVTREE;
1009 return PCI_PROBE_NORMAL;
1012 struct pci_controller_ops pseries_pci_controller_ops = {
1013 .probe_mode = pSeries_pci_probe_mode,
1016 define_machine(pseries) {
1018 .probe = pSeries_probe,
1019 .setup_arch = pSeries_setup_arch,
1020 .init_IRQ = pseries_init_irq,
1021 .show_cpuinfo = pSeries_show_cpuinfo,
1022 .log_error = pSeries_log_error,
1023 .pcibios_fixup = pSeries_final_fixup,
1024 .restart = rtas_restart,
1026 .panic = pseries_panic,
1027 .get_boot_time = rtas_get_boot_time,
1028 .get_rtc_time = rtas_get_rtc_time,
1029 .set_rtc_time = rtas_set_rtc_time,
1030 .calibrate_decr = generic_calibrate_decr,
1031 .progress = rtas_progress,
1032 .system_reset_exception = pSeries_system_reset_exception,
1033 .machine_check_early = pseries_machine_check_realmode,
1034 .machine_check_exception = pSeries_machine_check_exception,
1035 #ifdef CONFIG_KEXEC_CORE
1036 .machine_kexec = pSeries_machine_kexec,
1037 .kexec_cpu_down = pseries_kexec_cpu_down,
1039 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
1040 .memory_block_size = pseries_memory_block_size,