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)
283 struct paca_struct *pp;
284 struct dtl_entry *dtl;
286 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
292 for_each_possible_cpu(cpu) {
294 dtl = kmem_cache_alloc(dtl_cache, GFP_KERNEL);
296 pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
298 pr_warn("Stolen time statistics will be unreliable\n");
303 pp->dispatch_log = dtl;
304 pp->dispatch_log_end = dtl + N_DISPATCH_LOG;
308 /* Register the DTL for the current (boot) cpu */
309 dtl = get_paca()->dispatch_log;
310 get_paca()->dtl_ridx = 0;
311 get_paca()->dtl_curr = dtl;
312 get_paca()->lppaca_ptr->dtl_idx = 0;
314 /* hypervisor reads buffer length from this field */
315 dtl->enqueue_to_dispatch_time = cpu_to_be32(DISPATCH_LOG_BYTES);
316 ret = register_dtl(hard_smp_processor_id(), __pa(dtl));
318 pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
319 "with %d\n", smp_processor_id(),
320 hard_smp_processor_id(), ret);
321 get_paca()->lppaca_ptr->dtl_enable_mask = 2;
325 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
326 static inline int alloc_dispatch_logs(void)
330 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
332 static int alloc_dispatch_log_kmem_cache(void)
334 dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
335 DISPATCH_LOG_BYTES, 0, NULL);
337 pr_warn("Failed to create dispatch trace log buffer cache\n");
338 pr_warn("Stolen time statistics will be unreliable\n");
342 return alloc_dispatch_logs();
344 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
346 static void pseries_lpar_idle(void)
349 * Default handler to go into low thread priority and possibly
350 * low power mode by ceding processor to hypervisor
353 /* Indicate to hypervisor that we are idle. */
354 get_lppaca()->idle = 1;
357 * Yield the processor to the hypervisor. We return if
358 * an external interrupt occurs (which are driven prior
359 * to returning here) or if a prod occurs from another
360 * processor. When returning here, external interrupts
365 get_lppaca()->idle = 0;
369 * Enable relocation on during exceptions. This has partition wide scope and
370 * may take a while to complete, if it takes longer than one second we will
371 * just give up rather than wasting any more time on this - if that turns out
372 * to ever be a problem in practice we can move this into a kernel thread to
373 * finish off the process later in boot.
375 void pseries_enable_reloc_on_exc(void)
378 unsigned int delay, total_delay = 0;
381 rc = enable_reloc_on_exceptions();
382 if (!H_IS_LONG_BUSY(rc)) {
384 pr_info("Relocation on exceptions not"
386 } else if (rc != H_SUCCESS) {
387 pr_warn("Unable to enable relocation"
388 " on exceptions: %ld\n", rc);
393 delay = get_longbusy_msecs(rc);
394 total_delay += delay;
395 if (total_delay > 1000) {
396 pr_warn("Warning: Giving up waiting to enable "
397 "relocation on exceptions (%u msec)!\n",
405 EXPORT_SYMBOL(pseries_enable_reloc_on_exc);
407 void pseries_disable_reloc_on_exc(void)
412 rc = disable_reloc_on_exceptions();
413 if (!H_IS_LONG_BUSY(rc))
415 mdelay(get_longbusy_msecs(rc));
418 pr_warn("Warning: Failed to disable relocation on exceptions: %ld\n",
421 EXPORT_SYMBOL(pseries_disable_reloc_on_exc);
423 #ifdef CONFIG_KEXEC_CORE
424 static void pSeries_machine_kexec(struct kimage *image)
426 if (firmware_has_feature(FW_FEATURE_SET_MODE))
427 pseries_disable_reloc_on_exc();
429 default_machine_kexec(image);
433 #ifdef __LITTLE_ENDIAN__
434 void pseries_big_endian_exceptions(void)
439 rc = enable_big_endian_exceptions();
440 if (!H_IS_LONG_BUSY(rc))
442 mdelay(get_longbusy_msecs(rc));
446 * At this point it is unlikely panic() will get anything
447 * out to the user, since this is called very late in kexec
448 * but at least this will stop us from continuing on further
449 * and creating an even more difficult to debug situation.
451 * There is a known problem when kdump'ing, if cpus are offline
452 * the above call will fail. Rather than panicking again, keep
453 * going and hope the kdump kernel is also little endian, which
456 if (rc && !kdump_in_progress())
457 panic("Could not enable big endian exceptions");
460 void pseries_little_endian_exceptions(void)
465 rc = enable_little_endian_exceptions();
466 if (!H_IS_LONG_BUSY(rc))
468 mdelay(get_longbusy_msecs(rc));
471 ppc_md.progress("H_SET_MODE LE exception fail", 0);
472 panic("Could not enable little endian exceptions");
477 static void __init find_and_init_phbs(void)
479 struct device_node *node;
480 struct pci_controller *phb;
481 struct device_node *root = of_find_node_by_path("/");
483 for_each_child_of_node(root, node) {
484 if (!of_node_is_type(node, "pci") &&
485 !of_node_is_type(node, "pciex"))
488 phb = pcibios_alloc_controller(node);
492 pci_process_bridge_OF_ranges(phb, node, 0);
493 isa_bridge_find_early(phb);
494 phb->controller_ops = pseries_pci_controller_ops;
500 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
503 of_pci_check_probe_only();
506 static void init_cpu_char_feature_flags(struct h_cpu_char_result *result)
509 * The features below are disabled by default, so we instead look to see
510 * if firmware has *enabled* them, and set them if so.
512 if (result->character & H_CPU_CHAR_SPEC_BAR_ORI31)
513 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31);
515 if (result->character & H_CPU_CHAR_BCCTRL_SERIALISED)
516 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED);
518 if (result->character & H_CPU_CHAR_L1D_FLUSH_ORI30)
519 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30);
521 if (result->character & H_CPU_CHAR_L1D_FLUSH_TRIG2)
522 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2);
524 if (result->character & H_CPU_CHAR_L1D_THREAD_PRIV)
525 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV);
527 if (result->character & H_CPU_CHAR_COUNT_CACHE_DISABLED)
528 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED);
530 if (result->character & H_CPU_CHAR_BCCTR_FLUSH_ASSIST)
531 security_ftr_set(SEC_FTR_BCCTR_FLUSH_ASSIST);
533 if (result->behaviour & H_CPU_BEHAV_FLUSH_COUNT_CACHE)
534 security_ftr_set(SEC_FTR_FLUSH_COUNT_CACHE);
537 * The features below are enabled by default, so we instead look to see
538 * if firmware has *disabled* them, and clear them if so.
540 if (!(result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY))
541 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY);
543 if (!(result->behaviour & H_CPU_BEHAV_L1D_FLUSH_PR))
544 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR);
546 if (!(result->behaviour & H_CPU_BEHAV_BNDS_CHK_SPEC_BAR))
547 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR);
550 void pseries_setup_rfi_flush(void)
552 struct h_cpu_char_result result;
553 enum l1d_flush_type types;
558 * Set features to the defaults assumed by init_cpu_char_feature_flags()
559 * so it can set/clear again any features that might have changed after
560 * migration, and in case the hypercall fails and it is not even called.
562 powerpc_security_features = SEC_FTR_DEFAULT;
564 rc = plpar_get_cpu_characteristics(&result);
566 init_cpu_char_feature_flags(&result);
569 * We're the guest so this doesn't apply to us, clear it to simplify
570 * handling of it elsewhere.
572 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV);
574 types = L1D_FLUSH_FALLBACK;
576 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2))
577 types |= L1D_FLUSH_MTTRIG;
579 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30))
580 types |= L1D_FLUSH_ORI;
582 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) && \
583 security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR);
585 setup_rfi_flush(types, enable);
586 setup_count_cache_flush();
589 #ifdef CONFIG_PCI_IOV
590 enum rtas_iov_fw_value_map {
591 NUM_RES_PROPERTY = 0, /* Number of Resources */
592 LOW_INT = 1, /* Lowest 32 bits of Address */
593 START_OF_ENTRIES = 2, /* Always start of entry */
594 APERTURE_PROPERTY = 2, /* Start of entry+ to Aperture Size */
595 WDW_SIZE_PROPERTY = 4, /* Start of entry+ to Window Size */
596 NEXT_ENTRY = 7 /* Go to next entry on array */
599 enum get_iov_fw_value_index {
600 BAR_ADDRS = 1, /* Get Bar Address */
601 APERTURE_SIZE = 2, /* Get Aperture Size */
602 WDW_SIZE = 3 /* Get Window Size */
605 resource_size_t pseries_get_iov_fw_value(struct pci_dev *dev, int resno,
606 enum get_iov_fw_value_index value)
609 struct device_node *dn = pci_device_to_OF_node(dev);
610 int i, num_res, ret = 0;
612 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
617 * First element in the array is the number of Bars
618 * returned. Search through the list to find the matching
621 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
622 if (resno >= num_res)
623 return 0; /* or an errror */
625 i = START_OF_ENTRIES + NEXT_ENTRY * resno;
628 ret = of_read_number(&indexes[i], 2);
631 ret = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
634 ret = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
641 void of_pci_set_vf_bar_size(struct pci_dev *dev, const int *indexes)
643 struct resource *res;
644 resource_size_t base, size;
647 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
648 num_res = min_t(int, num_res, PCI_SRIOV_NUM_BARS);
649 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
650 i += NEXT_ENTRY, r++) {
651 res = &dev->resource[r + PCI_IOV_RESOURCES];
652 base = of_read_number(&indexes[i], 2);
653 size = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
654 res->flags = pci_parse_of_flags(of_read_number
655 (&indexes[i + LOW_INT], 1), 0);
656 res->flags |= (IORESOURCE_MEM_64 | IORESOURCE_PCI_FIXED);
657 res->name = pci_name(dev);
659 res->end = base + size - 1;
663 void of_pci_parse_iov_addrs(struct pci_dev *dev, const int *indexes)
665 struct resource *res, *root, *conflict;
666 resource_size_t base, size;
670 * First element in the array is the number of Bars
671 * returned. Search through the list to find the matching
672 * bars assign them from firmware into resources structure.
674 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
675 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
676 i += NEXT_ENTRY, r++) {
677 res = &dev->resource[r + PCI_IOV_RESOURCES];
678 base = of_read_number(&indexes[i], 2);
679 size = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
680 res->name = pci_name(dev);
682 res->end = base + size - 1;
683 root = &iomem_resource;
685 "pSeries IOV BAR %d: trying firmware assignment %pR\n",
686 r + PCI_IOV_RESOURCES, res);
687 conflict = request_resource_conflict(root, res);
690 "BAR %d: %pR conflicts with %s %pR\n",
691 r + PCI_IOV_RESOURCES, res,
692 conflict->name, conflict);
693 res->flags |= IORESOURCE_UNSET;
698 static void pseries_disable_sriov_resources(struct pci_dev *pdev)
702 pci_warn(pdev, "No hypervisor support for SR-IOV on this device, IOV BARs disabled.\n");
703 for (i = 0; i < PCI_SRIOV_NUM_BARS; i++)
704 pdev->resource[i + PCI_IOV_RESOURCES].flags = 0;
707 static void pseries_pci_fixup_resources(struct pci_dev *pdev)
710 struct device_node *dn = pci_device_to_OF_node(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_set_vf_bar_size(pdev, indexes);
717 pseries_disable_sriov_resources(pdev);
720 static void pseries_pci_fixup_iov_resources(struct pci_dev *pdev)
723 struct device_node *dn = pci_device_to_OF_node(pdev);
725 if (!pdev->is_physfn || pci_dev_is_added(pdev))
727 /*Firmware must support open sriov otherwise dont configure*/
728 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
730 of_pci_parse_iov_addrs(pdev, indexes);
732 pseries_disable_sriov_resources(pdev);
735 static resource_size_t pseries_pci_iov_resource_alignment(struct pci_dev *pdev,
739 struct device_node *dn = pci_device_to_OF_node(pdev);
741 /*Firmware must support open sriov otherwise report regular alignment*/
742 reg = of_get_property(dn, "ibm,is-open-sriov-pf", NULL);
744 return pci_iov_resource_size(pdev, resno);
746 if (!pdev->is_physfn)
748 return pseries_get_iov_fw_value(pdev,
749 resno - PCI_IOV_RESOURCES,
754 static void __init pSeries_setup_arch(void)
756 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
758 /* Discover PIC type and setup ppc_md accordingly */
762 /* openpic global configuration register (64-bit format). */
763 /* openpic Interrupt Source Unit pointer (64-bit format). */
764 /* python0 facility area (mmio) (64-bit format) REAL address. */
766 /* init to some ~sane value until calibrate_delay() runs */
767 loops_per_jiffy = 50000000;
771 pseries_setup_rfi_flush();
774 /* By default, only probe PCI (can be overridden by rtas_pci) */
775 pci_add_flags(PCI_PROBE_ONLY);
777 /* Find and initialize PCI host bridges */
778 init_pci_config_tokens();
779 find_and_init_phbs();
780 of_reconfig_notifier_register(&pci_dn_reconfig_nb);
782 pSeries_nvram_init();
784 if (firmware_has_feature(FW_FEATURE_LPAR)) {
785 vpa_init(boot_cpuid);
786 ppc_md.power_save = pseries_lpar_idle;
787 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
788 #ifdef CONFIG_PCI_IOV
789 ppc_md.pcibios_fixup_resources =
790 pseries_pci_fixup_resources;
791 ppc_md.pcibios_fixup_sriov =
792 pseries_pci_fixup_iov_resources;
793 ppc_md.pcibios_iov_resource_alignment =
794 pseries_pci_iov_resource_alignment;
797 /* No special idle routine */
798 ppc_md.enable_pmcs = power4_enable_pmcs;
801 ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
803 if (swiotlb_force == SWIOTLB_FORCE)
804 ppc_swiotlb_enable = 1;
807 static void pseries_panic(char *str)
809 panic_flush_kmsg_end();
813 static int __init pSeries_init_panel(void)
815 /* Manually leave the kernel version on the panel. */
816 #ifdef __BIG_ENDIAN__
817 ppc_md.progress("Linux ppc64\n", 0);
819 ppc_md.progress("Linux ppc64le\n", 0);
821 ppc_md.progress(init_utsname()->version, 0);
825 machine_arch_initcall(pseries, pSeries_init_panel);
827 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
829 return plpar_hcall_norets(H_SET_DABR, dabr);
832 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
834 /* Have to set at least one bit in the DABRX according to PAPR */
835 if (dabrx == 0 && dabr == 0)
837 /* PAPR says we can only set kernel and user bits */
838 dabrx &= DABRX_KERNEL | DABRX_USER;
840 return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
843 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
845 /* PAPR says we can't set HYP */
848 return plpar_set_watchpoint0(dawr, dawrx);
851 #define CMO_CHARACTERISTICS_TOKEN 44
852 #define CMO_MAXLENGTH 1026
854 void pSeries_coalesce_init(void)
856 struct hvcall_mpp_x_data mpp_x_data;
858 if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
859 powerpc_firmware_features |= FW_FEATURE_XCMO;
861 powerpc_firmware_features &= ~FW_FEATURE_XCMO;
865 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
866 * handle that here. (Stolen from parse_system_parameter_string)
868 static void pSeries_cmo_feature_init(void)
870 char *ptr, *key, *value, *end;
872 int page_order = IOMMU_PAGE_SHIFT_4K;
874 pr_debug(" -> fw_cmo_feature_init()\n");
875 spin_lock(&rtas_data_buf_lock);
876 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
877 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
879 CMO_CHARACTERISTICS_TOKEN,
883 if (call_status != 0) {
884 spin_unlock(&rtas_data_buf_lock);
885 pr_debug("CMO not available\n");
886 pr_debug(" <- fw_cmo_feature_init()\n");
890 end = rtas_data_buf + CMO_MAXLENGTH - 2;
891 ptr = rtas_data_buf + 2; /* step over strlen value */
894 while (*ptr && (ptr <= end)) {
895 /* Separate the key and value by replacing '=' with '\0' and
896 * point the value at the string after the '='
901 } else if (ptr[0] == '\0' || ptr[0] == ',') {
902 /* Terminate the string containing the key/value pair */
906 pr_debug("Malformed key/value pair\n");
907 /* Never found a '=', end processing */
911 if (0 == strcmp(key, "CMOPageSize"))
912 page_order = simple_strtol(value, NULL, 10);
913 else if (0 == strcmp(key, "PrPSP"))
914 CMO_PrPSP = simple_strtol(value, NULL, 10);
915 else if (0 == strcmp(key, "SecPSP"))
916 CMO_SecPSP = simple_strtol(value, NULL, 10);
917 value = key = ptr + 1;
922 /* Page size is returned as the power of 2 of the page size,
923 * convert to the page size in bytes before returning
925 CMO_PageSize = 1 << page_order;
926 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
928 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
929 pr_info("CMO enabled\n");
930 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
932 powerpc_firmware_features |= FW_FEATURE_CMO;
933 pSeries_coalesce_init();
935 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
937 spin_unlock(&rtas_data_buf_lock);
938 pr_debug(" <- fw_cmo_feature_init()\n");
942 * Early initialization. Relocation is on but do not reference unbolted pages
944 static void __init pseries_init(void)
946 pr_debug(" -> pseries_init()\n");
948 #ifdef CONFIG_HVC_CONSOLE
949 if (firmware_has_feature(FW_FEATURE_LPAR))
950 hvc_vio_init_early();
952 if (firmware_has_feature(FW_FEATURE_XDABR))
953 ppc_md.set_dabr = pseries_set_xdabr;
954 else if (firmware_has_feature(FW_FEATURE_DABR))
955 ppc_md.set_dabr = pseries_set_dabr;
957 if (firmware_has_feature(FW_FEATURE_SET_MODE))
958 ppc_md.set_dawr = pseries_set_dawr;
960 pSeries_cmo_feature_init();
961 iommu_init_early_pSeries();
963 pr_debug(" <- pseries_init()\n");
967 * pseries_power_off - tell firmware about how to power off the system.
969 * This function calls either the power-off rtas token in normal cases
970 * or the ibm,power-off-ups token (if present & requested) in case of
971 * a power failure. If power-off token is used, power on will only be
972 * possible with power button press. If ibm,power-off-ups token is used
973 * it will allow auto poweron after power is restored.
975 static void pseries_power_off(void)
978 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
980 if (rtas_flash_term_hook)
981 rtas_flash_term_hook(SYS_POWER_OFF);
983 if (rtas_poweron_auto == 0 ||
984 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
985 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
986 printk(KERN_INFO "RTAS power-off returned %d\n", rc);
988 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
989 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
994 static int __init pSeries_probe(void)
996 if (!of_node_is_type(of_root, "chrp"))
999 /* Cell blades firmware claims to be chrp while it's not. Until this
1000 * is fixed, we need to avoid those here.
1002 if (of_machine_is_compatible("IBM,CPBW-1.0") ||
1003 of_machine_is_compatible("IBM,CBEA"))
1006 pm_power_off = pseries_power_off;
1008 pr_debug("Machine is%s LPAR !\n",
1009 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
1016 static int pSeries_pci_probe_mode(struct pci_bus *bus)
1018 if (firmware_has_feature(FW_FEATURE_LPAR))
1019 return PCI_PROBE_DEVTREE;
1020 return PCI_PROBE_NORMAL;
1023 struct pci_controller_ops pseries_pci_controller_ops = {
1024 .probe_mode = pSeries_pci_probe_mode,
1027 define_machine(pseries) {
1029 .probe = pSeries_probe,
1030 .setup_arch = pSeries_setup_arch,
1031 .init_IRQ = pseries_init_irq,
1032 .show_cpuinfo = pSeries_show_cpuinfo,
1033 .log_error = pSeries_log_error,
1034 .pcibios_fixup = pSeries_final_fixup,
1035 .restart = rtas_restart,
1037 .panic = pseries_panic,
1038 .get_boot_time = rtas_get_boot_time,
1039 .get_rtc_time = rtas_get_rtc_time,
1040 .set_rtc_time = rtas_set_rtc_time,
1041 .calibrate_decr = generic_calibrate_decr,
1042 .progress = rtas_progress,
1043 .system_reset_exception = pSeries_system_reset_exception,
1044 .machine_check_early = pseries_machine_check_realmode,
1045 .machine_check_exception = pSeries_machine_check_exception,
1046 #ifdef CONFIG_KEXEC_CORE
1047 .machine_kexec = pSeries_machine_kexec,
1048 .kexec_cpu_down = pseries_kexec_cpu_down,
1050 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
1051 .memory_block_size = pseries_memory_block_size,