1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Page table handling routines for radix page table.
4 *
5 * Copyright 2015-2016, Aneesh Kumar K.V, IBM Corporation.
6 */
7
8#define pr_fmt(fmt) "radix-mmu: " fmt
9
10#include <linux/io.h>
11#include <linux/kernel.h>
12#include <linux/sched/mm.h>
13#include <linux/memblock.h>
14#include <linux/of_fdt.h>
15#include <linux/mm.h>
16#include <linux/hugetlb.h>
17#include <linux/string_helpers.h>
18#include <linux/memory.h>
19
20#include <asm/pgalloc.h>
21#include <asm/mmu_context.h>
22#include <asm/dma.h>
23#include <asm/machdep.h>
24#include <asm/mmu.h>
25#include <asm/firmware.h>
26#include <asm/powernv.h>
27#include <asm/sections.h>
28#include <asm/smp.h>
29#include <asm/trace.h>
30#include <asm/uaccess.h>
31#include <asm/ultravisor.h>
32
33#include <trace/events/thp.h>
34
35unsigned int mmu_pid_bits;
36unsigned int mmu_base_pid;
37unsigned long radix_mem_block_size __ro_after_init;
38
39static __ref void *early_alloc_pgtable(unsigned long size, int nid,
40			unsigned long region_start, unsigned long region_end)
41{
42	phys_addr_t min_addr = MEMBLOCK_LOW_LIMIT;
43	phys_addr_t max_addr = MEMBLOCK_ALLOC_ANYWHERE;
44	void *ptr;
45
46	if (region_start)
47		min_addr = region_start;
48	if (region_end)
49		max_addr = region_end;
50
51	ptr = memblock_alloc_try_nid(size, size, min_addr, max_addr, nid);
52
53	if (!ptr)
54		panic("%s: Failed to allocate %lu bytes align=0x%lx nid=%d from=%pa max_addr=%pa\n",
55		      __func__, size, size, nid, &min_addr, &max_addr);
56
57	return ptr;
58}
59
60/*
61 * When allocating pud or pmd pointers, we allocate a complete page
62 * of PAGE_SIZE rather than PUD_TABLE_SIZE or PMD_TABLE_SIZE. This
63 * is to ensure that the page obtained from the memblock allocator
64 * can be completely used as page table page and can be freed
65 * correctly when the page table entries are removed.
66 */
67static int early_map_kernel_page(unsigned long ea, unsigned long pa,
68			  pgprot_t flags,
69			  unsigned int map_page_size,
70			  int nid,
71			  unsigned long region_start, unsigned long region_end)
72{
73	unsigned long pfn = pa >> PAGE_SHIFT;
74	pgd_t *pgdp;
75	p4d_t *p4dp;
76	pud_t *pudp;
77	pmd_t *pmdp;
78	pte_t *ptep;
79
80	pgdp = pgd_offset_k(ea);
81	p4dp = p4d_offset(pgdp, ea);
82	if (p4d_none(*p4dp)) {
83		pudp = early_alloc_pgtable(PAGE_SIZE, nid,
84					   region_start, region_end);
85		p4d_populate(&init_mm, p4dp, pudp);
86	}
87	pudp = pud_offset(p4dp, ea);
88	if (map_page_size == PUD_SIZE) {
89		ptep = (pte_t *)pudp;
90		goto set_the_pte;
91	}
92	if (pud_none(*pudp)) {
93		pmdp = early_alloc_pgtable(PAGE_SIZE, nid, region_start,
94					   region_end);
95		pud_populate(&init_mm, pudp, pmdp);
96	}
97	pmdp = pmd_offset(pudp, ea);
98	if (map_page_size == PMD_SIZE) {
99		ptep = pmdp_ptep(pmdp);
100		goto set_the_pte;
101	}
102	if (!pmd_present(*pmdp)) {
103		ptep = early_alloc_pgtable(PAGE_SIZE, nid,
104						region_start, region_end);
105		pmd_populate_kernel(&init_mm, pmdp, ptep);
106	}
107	ptep = pte_offset_kernel(pmdp, ea);
108
109set_the_pte:
110	set_pte_at(&init_mm, ea, ptep, pfn_pte(pfn, flags));
111	asm volatile("ptesync": : :"memory");
112	return 0;
113}
114
115/*
116 * nid, region_start, and region_end are hints to try to place the page
117 * table memory in the same node or region.
118 */
119static int __map_kernel_page(unsigned long ea, unsigned long pa,
120			  pgprot_t flags,
121			  unsigned int map_page_size,
122			  int nid,
123			  unsigned long region_start, unsigned long region_end)
124{
125	unsigned long pfn = pa >> PAGE_SHIFT;
126	pgd_t *pgdp;
127	p4d_t *p4dp;
128	pud_t *pudp;
129	pmd_t *pmdp;
130	pte_t *ptep;
131	/*
132	 * Make sure task size is correct as per the max adddr
133	 */
134	BUILD_BUG_ON(TASK_SIZE_USER64 > RADIX_PGTABLE_RANGE);
135
136#ifdef CONFIG_PPC_64K_PAGES
137	BUILD_BUG_ON(RADIX_KERN_MAP_SIZE != (1UL << MAX_EA_BITS_PER_CONTEXT));
138#endif
139
140	if (unlikely(!slab_is_available()))
141		return early_map_kernel_page(ea, pa, flags, map_page_size,
142						nid, region_start, region_end);
143
144	/*
145	 * Should make page table allocation functions be able to take a
146	 * node, so we can place kernel page tables on the right nodes after
147	 * boot.
148	 */
149	pgdp = pgd_offset_k(ea);
150	p4dp = p4d_offset(pgdp, ea);
151	pudp = pud_alloc(&init_mm, p4dp, ea);
152	if (!pudp)
153		return -ENOMEM;
154	if (map_page_size == PUD_SIZE) {
155		ptep = (pte_t *)pudp;
156		goto set_the_pte;
157	}
158	pmdp = pmd_alloc(&init_mm, pudp, ea);
159	if (!pmdp)
160		return -ENOMEM;
161	if (map_page_size == PMD_SIZE) {
162		ptep = pmdp_ptep(pmdp);
163		goto set_the_pte;
164	}
165	ptep = pte_alloc_kernel(pmdp, ea);
166	if (!ptep)
167		return -ENOMEM;
168
169set_the_pte:
170	set_pte_at(&init_mm, ea, ptep, pfn_pte(pfn, flags));
171	asm volatile("ptesync": : :"memory");
172	return 0;
173}
174
175int radix__map_kernel_page(unsigned long ea, unsigned long pa,
176			  pgprot_t flags,
177			  unsigned int map_page_size)
178{
179	return __map_kernel_page(ea, pa, flags, map_page_size, -1, 0, 0);
180}
181
182#ifdef CONFIG_STRICT_KERNEL_RWX
183void radix__change_memory_range(unsigned long start, unsigned long end,
184				unsigned long clear)
185{
186	unsigned long idx;
187	pgd_t *pgdp;
188	p4d_t *p4dp;
189	pud_t *pudp;
190	pmd_t *pmdp;
191	pte_t *ptep;
192
193	start = ALIGN_DOWN(start, PAGE_SIZE);
194	end = PAGE_ALIGN(end); // aligns up
195
196	pr_debug("Changing flags on range %lx-%lx removing 0x%lx\n",
197		 start, end, clear);
198
199	for (idx = start; idx < end; idx += PAGE_SIZE) {
200		pgdp = pgd_offset_k(idx);
201		p4dp = p4d_offset(pgdp, idx);
202		pudp = pud_alloc(&init_mm, p4dp, idx);
203		if (!pudp)
204			continue;
205		if (pud_is_leaf(*pudp)) {
206			ptep = (pte_t *)pudp;
207			goto update_the_pte;
208		}
209		pmdp = pmd_alloc(&init_mm, pudp, idx);
210		if (!pmdp)
211			continue;
212		if (pmd_is_leaf(*pmdp)) {
213			ptep = pmdp_ptep(pmdp);
214			goto update_the_pte;
215		}
216		ptep = pte_alloc_kernel(pmdp, idx);
217		if (!ptep)
218			continue;
219update_the_pte:
220		radix__pte_update(&init_mm, idx, ptep, clear, 0, 0);
221	}
222
223	radix__flush_tlb_kernel_range(start, end);
224}
225
226void radix__mark_rodata_ro(void)
227{
228	unsigned long start, end;
229
230	start = (unsigned long)_stext;
231	end = (unsigned long)__init_begin;
232
233	radix__change_memory_range(start, end, _PAGE_WRITE);
234}
235
236void radix__mark_initmem_nx(void)
237{
238	unsigned long start = (unsigned long)__init_begin;
239	unsigned long end = (unsigned long)__init_end;
240
241	radix__change_memory_range(start, end, _PAGE_EXEC);
242}
243#endif /* CONFIG_STRICT_KERNEL_RWX */
244
245static inline void __meminit
246print_mapping(unsigned long start, unsigned long end, unsigned long size, bool exec)
247{
248	char buf[10];
249
250	if (end <= start)
251		return;
252
253	string_get_size(size, 1, STRING_UNITS_2, buf, sizeof(buf));
254
255	pr_info("Mapped 0x%016lx-0x%016lx with %s pages%s\n", start, end, buf,
256		exec ? " (exec)" : "");
257}
258
259static unsigned long next_boundary(unsigned long addr, unsigned long end)
260{
261#ifdef CONFIG_STRICT_KERNEL_RWX
262	if (addr < __pa_symbol(__init_begin))
263		return __pa_symbol(__init_begin);
264#endif
265	return end;
266}
267
268static int __meminit create_physical_mapping(unsigned long start,
269					     unsigned long end,
270					     unsigned long max_mapping_size,
271					     int nid, pgprot_t _prot)
272{
273	unsigned long vaddr, addr, mapping_size = 0;
274	bool prev_exec, exec = false;
275	pgprot_t prot;
276	int psize;
277
278	start = ALIGN(start, PAGE_SIZE);
279	end   = ALIGN_DOWN(end, PAGE_SIZE);
280	for (addr = start; addr < end; addr += mapping_size) {
281		unsigned long gap, previous_size;
282		int rc;
283
284		gap = next_boundary(addr, end) - addr;
285		if (gap > max_mapping_size)
286			gap = max_mapping_size;
287		previous_size = mapping_size;
288		prev_exec = exec;
289
290		if (IS_ALIGNED(addr, PUD_SIZE) && gap >= PUD_SIZE &&
291		    mmu_psize_defs[MMU_PAGE_1G].shift) {
292			mapping_size = PUD_SIZE;
293			psize = MMU_PAGE_1G;
294		} else if (IS_ALIGNED(addr, PMD_SIZE) && gap >= PMD_SIZE &&
295			   mmu_psize_defs[MMU_PAGE_2M].shift) {
296			mapping_size = PMD_SIZE;
297			psize = MMU_PAGE_2M;
298		} else {
299			mapping_size = PAGE_SIZE;
300			psize = mmu_virtual_psize;
301		}
302
303		vaddr = (unsigned long)__va(addr);
304
305		if (overlaps_kernel_text(vaddr, vaddr + mapping_size) ||
306		    overlaps_interrupt_vector_text(vaddr, vaddr + mapping_size)) {
307			prot = PAGE_KERNEL_X;
308			exec = true;
309		} else {
310			prot = _prot;
311			exec = false;
312		}
313
314		if (mapping_size != previous_size || exec != prev_exec) {
315			print_mapping(start, addr, previous_size, prev_exec);
316			start = addr;
317		}
318
319		rc = __map_kernel_page(vaddr, addr, prot, mapping_size, nid, start, end);
320		if (rc)
321			return rc;
322
323		update_page_count(psize, 1);
324	}
325
326	print_mapping(start, addr, mapping_size, exec);
327	return 0;
328}
329
330static void __init radix_init_pgtable(void)
331{
332	unsigned long rts_field;
333	phys_addr_t start, end;
334	u64 i;
335
336	/* We don't support slb for radix */
337	mmu_slb_size = 0;
338
339	/*
340	 * Create the linear mapping
341	 */
342	for_each_mem_range(i, &start, &end) {
343		/*
344		 * The memblock allocator  is up at this point, so the
345		 * page tables will be allocated within the range. No
346		 * need or a node (which we don't have yet).
347		 */
348
349		if (end >= RADIX_VMALLOC_START) {
350			pr_warn("Outside the supported range\n");
351			continue;
352		}
353
354		WARN_ON(create_physical_mapping(start, end,
355						radix_mem_block_size,
356						-1, PAGE_KERNEL));
357	}
358
359	/* Find out how many PID bits are supported */
360	if (!cpu_has_feature(CPU_FTR_P9_RADIX_PREFETCH_BUG)) {
361		if (!mmu_pid_bits)
362			mmu_pid_bits = 20;
363		mmu_base_pid = 1;
364	} else if (cpu_has_feature(CPU_FTR_HVMODE)) {
365		if (!mmu_pid_bits)
366			mmu_pid_bits = 20;
367#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
368		/*
369		 * When KVM is possible, we only use the top half of the
370		 * PID space to avoid collisions between host and guest PIDs
371		 * which can cause problems due to prefetch when exiting the
372		 * guest with AIL=3
373		 */
374		mmu_base_pid = 1 << (mmu_pid_bits - 1);
375#else
376		mmu_base_pid = 1;
377#endif
378	} else {
379		/* The guest uses the bottom half of the PID space */
380		if (!mmu_pid_bits)
381			mmu_pid_bits = 19;
382		mmu_base_pid = 1;
383	}
384
385	/*
386	 * Allocate Partition table and process table for the
387	 * host.
388	 */
389	BUG_ON(PRTB_SIZE_SHIFT > 36);
390	process_tb = early_alloc_pgtable(1UL << PRTB_SIZE_SHIFT, -1, 0, 0);
391	/*
392	 * Fill in the process table.
393	 */
394	rts_field = radix__get_tree_size();
395	process_tb->prtb0 = cpu_to_be64(rts_field | __pa(init_mm.pgd) | RADIX_PGD_INDEX_SIZE);
396
397	/*
398	 * The init_mm context is given the first available (non-zero) PID,
399	 * which is the "guard PID" and contains no page table. PIDR should
400	 * never be set to zero because that duplicates the kernel address
401	 * space at the 0x0... offset (quadrant 0)!
402	 *
403	 * An arbitrary PID that may later be allocated by the PID allocator
404	 * for userspace processes must not be used either, because that
405	 * would cause stale user mappings for that PID on CPUs outside of
406	 * the TLB invalidation scheme (because it won't be in mm_cpumask).
407	 *
408	 * So permanently carve out one PID for the purpose of a guard PID.
409	 */
410	init_mm.context.id = mmu_base_pid;
411	mmu_base_pid++;
412}
413
414static void __init radix_init_partition_table(void)
415{
416	unsigned long rts_field, dw0, dw1;
417
418	mmu_partition_table_init();
419	rts_field = radix__get_tree_size();
420	dw0 = rts_field | __pa(init_mm.pgd) | RADIX_PGD_INDEX_SIZE | PATB_HR;
421	dw1 = __pa(process_tb) | (PRTB_SIZE_SHIFT - 12) | PATB_GR;
422	mmu_partition_table_set_entry(0, dw0, dw1, false);
423
424	pr_info("Initializing Radix MMU\n");
425}
426
427static int __init get_idx_from_shift(unsigned int shift)
428{
429	int idx = -1;
430
431	switch (shift) {
432	case 0xc:
433		idx = MMU_PAGE_4K;
434		break;
435	case 0x10:
436		idx = MMU_PAGE_64K;
437		break;
438	case 0x15:
439		idx = MMU_PAGE_2M;
440		break;
441	case 0x1e:
442		idx = MMU_PAGE_1G;
443		break;
444	}
445	return idx;
446}
447
448static int __init radix_dt_scan_page_sizes(unsigned long node,
449					   const char *uname, int depth,
450					   void *data)
451{
452	int size = 0;
453	int shift, idx;
454	unsigned int ap;
455	const __be32 *prop;
456	const char *type = of_get_flat_dt_prop(node, "device_type", NULL);
457
458	/* We are scanning "cpu" nodes only */
459	if (type == NULL || strcmp(type, "cpu") != 0)
460		return 0;
461
462	/* Find MMU PID size */
463	prop = of_get_flat_dt_prop(node, "ibm,mmu-pid-bits", &size);
464	if (prop && size == 4)
465		mmu_pid_bits = be32_to_cpup(prop);
466
467	/* Grab page size encodings */
468	prop = of_get_flat_dt_prop(node, "ibm,processor-radix-AP-encodings", &size);
469	if (!prop)
470		return 0;
471
472	pr_info("Page sizes from device-tree:\n");
473	for (; size >= 4; size -= 4, ++prop) {
474
475		struct mmu_psize_def *def;
476
477		/* top 3 bit is AP encoding */
478		shift = be32_to_cpu(prop[0]) & ~(0xe << 28);
479		ap = be32_to_cpu(prop[0]) >> 29;
480		pr_info("Page size shift = %d AP=0x%x\n", shift, ap);
481
482		idx = get_idx_from_shift(shift);
483		if (idx < 0)
484			continue;
485
486		def = &mmu_psize_defs[idx];
487		def->shift = shift;
488		def->ap  = ap;
489	}
490
491	/* needed ? */
492	cur_cpu_spec->mmu_features &= ~MMU_FTR_NO_SLBIE_B;
493	return 1;
494}
495
496#ifdef CONFIG_MEMORY_HOTPLUG
497static int __init probe_memory_block_size(unsigned long node, const char *uname, int
498					  depth, void *data)
499{
500	unsigned long *mem_block_size = (unsigned long *)data;
501	const __be32 *prop;
502	int len;
503
504	if (depth != 1)
505		return 0;
506
507	if (strcmp(uname, "ibm,dynamic-reconfiguration-memory"))
508		return 0;
509
510	prop = of_get_flat_dt_prop(node, "ibm,lmb-size", &len);
511
512	if (!prop || len < dt_root_size_cells * sizeof(__be32))
513		/*
514		 * Nothing in the device tree
515		 */
516		*mem_block_size = MIN_MEMORY_BLOCK_SIZE;
517	else
518		*mem_block_size = of_read_number(prop, dt_root_size_cells);
519	return 1;
520}
521
522static unsigned long radix_memory_block_size(void)
523{
524	unsigned long mem_block_size = MIN_MEMORY_BLOCK_SIZE;
525
526	/*
527	 * OPAL firmware feature is set by now. Hence we are ok
528	 * to test OPAL feature.
529	 */
530	if (firmware_has_feature(FW_FEATURE_OPAL))
531		mem_block_size = 1UL * 1024 * 1024 * 1024;
532	else
533		of_scan_flat_dt(probe_memory_block_size, &mem_block_size);
534
535	return mem_block_size;
536}
537
538#else   /* CONFIG_MEMORY_HOTPLUG */
539
540static unsigned long radix_memory_block_size(void)
541{
542	return 1UL * 1024 * 1024 * 1024;
543}
544
545#endif /* CONFIG_MEMORY_HOTPLUG */
546
547
548void __init radix__early_init_devtree(void)
549{
550	int rc;
551
552	/*
553	 * Try to find the available page sizes in the device-tree
554	 */
555	rc = of_scan_flat_dt(radix_dt_scan_page_sizes, NULL);
556	if (!rc) {
557		/*
558		 * No page size details found in device tree.
559		 * Let's assume we have page 4k and 64k support
560		 */
561		mmu_psize_defs[MMU_PAGE_4K].shift = 12;
562		mmu_psize_defs[MMU_PAGE_4K].ap = 0x0;
563
564		mmu_psize_defs[MMU_PAGE_64K].shift = 16;
565		mmu_psize_defs[MMU_PAGE_64K].ap = 0x5;
566	}
567
568	/*
569	 * Max mapping size used when mapping pages. We don't use
570	 * ppc_md.memory_block_size() here because this get called
571	 * early and we don't have machine probe called yet. Also
572	 * the pseries implementation only check for ibm,lmb-size.
573	 * All hypervisor supporting radix do expose that device
574	 * tree node.
575	 */
576	radix_mem_block_size = radix_memory_block_size();
577	return;
578}
579
580static void radix_init_amor(void)
581{
582	/*
583	* In HV mode, we init AMOR (Authority Mask Override Register) so that
584	* the hypervisor and guest can setup IAMR (Instruction Authority Mask
585	* Register), enable key 0 and set it to 1.
586	*
587	* AMOR = 0b1100 .... 0000 (Mask for key 0 is 11)
588	*/
589	mtspr(SPRN_AMOR, (3ul << 62));
590}
591
592#ifdef CONFIG_PPC_KUEP
593void setup_kuep(bool disabled)
594{
595	if (disabled || !early_radix_enabled())
596		return;
597
598	if (smp_processor_id() == boot_cpuid) {
599		pr_info("Activating Kernel Userspace Execution Prevention\n");
600		cur_cpu_spec->mmu_features |= MMU_FTR_KUEP;
601	}
602
603	/*
604	 * Radix always uses key0 of the IAMR to determine if an access is
605	 * allowed. We set bit 0 (IBM bit 1) of key0, to prevent instruction
606	 * fetch.
607	 */
608	mtspr(SPRN_IAMR, (1ul << 62));
609}
610#endif
611
612#ifdef CONFIG_PPC_KUAP
613void setup_kuap(bool disabled)
614{
615	if (disabled || !early_radix_enabled())
616		return;
617
618	if (smp_processor_id() == boot_cpuid) {
619		pr_info("Activating Kernel Userspace Access Prevention\n");
620		cur_cpu_spec->mmu_features |= MMU_FTR_RADIX_KUAP;
621	}
622
623	/* Make sure userspace can't change the AMR */
624	mtspr(SPRN_UAMOR, 0);
625
626	/*
627	 * Set the default kernel AMR values on all cpus.
628	 */
629	mtspr(SPRN_AMR, AMR_KUAP_BLOCKED);
630	isync();
631}
632#endif
633
634void __init radix__early_init_mmu(void)
635{
636	unsigned long lpcr;
637
638#ifdef CONFIG_PPC_64K_PAGES
639	/* PAGE_SIZE mappings */
640	mmu_virtual_psize = MMU_PAGE_64K;
641#else
642	mmu_virtual_psize = MMU_PAGE_4K;
643#endif
644
645#ifdef CONFIG_SPARSEMEM_VMEMMAP
646	/* vmemmap mapping */
647	if (mmu_psize_defs[MMU_PAGE_2M].shift) {
648		/*
649		 * map vmemmap using 2M if available
650		 */
651		mmu_vmemmap_psize = MMU_PAGE_2M;
652	} else
653		mmu_vmemmap_psize = mmu_virtual_psize;
654#endif
655	/*
656	 * initialize page table size
657	 */
658	__pte_index_size = RADIX_PTE_INDEX_SIZE;
659	__pmd_index_size = RADIX_PMD_INDEX_SIZE;
660	__pud_index_size = RADIX_PUD_INDEX_SIZE;
661	__pgd_index_size = RADIX_PGD_INDEX_SIZE;
662	__pud_cache_index = RADIX_PUD_INDEX_SIZE;
663	__pte_table_size = RADIX_PTE_TABLE_SIZE;
664	__pmd_table_size = RADIX_PMD_TABLE_SIZE;
665	__pud_table_size = RADIX_PUD_TABLE_SIZE;
666	__pgd_table_size = RADIX_PGD_TABLE_SIZE;
667
668	__pmd_val_bits = RADIX_PMD_VAL_BITS;
669	__pud_val_bits = RADIX_PUD_VAL_BITS;
670	__pgd_val_bits = RADIX_PGD_VAL_BITS;
671
672	__kernel_virt_start = RADIX_KERN_VIRT_START;
673	__vmalloc_start = RADIX_VMALLOC_START;
674	__vmalloc_end = RADIX_VMALLOC_END;
675	__kernel_io_start = RADIX_KERN_IO_START;
676	__kernel_io_end = RADIX_KERN_IO_END;
677	vmemmap = (struct page *)RADIX_VMEMMAP_START;
678	ioremap_bot = IOREMAP_BASE;
679
680#ifdef CONFIG_PCI
681	pci_io_base = ISA_IO_BASE;
682#endif
683	__pte_frag_nr = RADIX_PTE_FRAG_NR;
684	__pte_frag_size_shift = RADIX_PTE_FRAG_SIZE_SHIFT;
685	__pmd_frag_nr = RADIX_PMD_FRAG_NR;
686	__pmd_frag_size_shift = RADIX_PMD_FRAG_SIZE_SHIFT;
687
688	radix_init_pgtable();
689
690	if (!firmware_has_feature(FW_FEATURE_LPAR)) {
691		lpcr = mfspr(SPRN_LPCR);
692		mtspr(SPRN_LPCR, lpcr | LPCR_UPRT | LPCR_HR);
693		radix_init_partition_table();
694		radix_init_amor();
695	} else {
696		radix_init_pseries();
697	}
698
699	memblock_set_current_limit(MEMBLOCK_ALLOC_ANYWHERE);
700
701	/* Switch to the guard PID before turning on MMU */
702	radix__switch_mmu_context(NULL, &init_mm);
703	tlbiel_all();
704}
705
706void radix__early_init_mmu_secondary(void)
707{
708	unsigned long lpcr;
709	/*
710	 * update partition table control register and UPRT
711	 */
712	if (!firmware_has_feature(FW_FEATURE_LPAR)) {
713		lpcr = mfspr(SPRN_LPCR);
714		mtspr(SPRN_LPCR, lpcr | LPCR_UPRT | LPCR_HR);
715
716		set_ptcr_when_no_uv(__pa(partition_tb) |
717				    (PATB_SIZE_SHIFT - 12));
718
719		radix_init_amor();
720	}
721
722	radix__switch_mmu_context(NULL, &init_mm);
723	tlbiel_all();
724}
725
726void radix__mmu_cleanup_all(void)
727{
728	unsigned long lpcr;
729
730	if (!firmware_has_feature(FW_FEATURE_LPAR)) {
731		lpcr = mfspr(SPRN_LPCR);
732		mtspr(SPRN_LPCR, lpcr & ~LPCR_UPRT);
733		set_ptcr_when_no_uv(0);
734		powernv_set_nmmu_ptcr(0);
735		radix__flush_tlb_all();
736	}
737}
738
739#ifdef CONFIG_MEMORY_HOTPLUG
740static void free_pte_table(pte_t *pte_start, pmd_t *pmd)
741{
742	pte_t *pte;
743	int i;
744
745	for (i = 0; i < PTRS_PER_PTE; i++) {
746		pte = pte_start + i;
747		if (!pte_none(*pte))
748			return;
749	}
750
751	pte_free_kernel(&init_mm, pte_start);
752	pmd_clear(pmd);
753}
754
755static void free_pmd_table(pmd_t *pmd_start, pud_t *pud)
756{
757	pmd_t *pmd;
758	int i;
759
760	for (i = 0; i < PTRS_PER_PMD; i++) {
761		pmd = pmd_start + i;
762		if (!pmd_none(*pmd))
763			return;
764	}
765
766	pmd_free(&init_mm, pmd_start);
767	pud_clear(pud);
768}
769
770static void free_pud_table(pud_t *pud_start, p4d_t *p4d)
771{
772	pud_t *pud;
773	int i;
774
775	for (i = 0; i < PTRS_PER_PUD; i++) {
776		pud = pud_start + i;
777		if (!pud_none(*pud))
778			return;
779	}
780
781	pud_free(&init_mm, pud_start);
782	p4d_clear(p4d);
783}
784
785static void remove_pte_table(pte_t *pte_start, unsigned long addr,
786			     unsigned long end, bool direct)
787{
788	unsigned long next, pages = 0;
789	pte_t *pte;
790
791	pte = pte_start + pte_index(addr);
792	for (; addr < end; addr = next, pte++) {
793		next = (addr + PAGE_SIZE) & PAGE_MASK;
794		if (next > end)
795			next = end;
796
797		if (!pte_present(*pte))
798			continue;
799
800		if (!PAGE_ALIGNED(addr) || !PAGE_ALIGNED(next)) {
801			/*
802			 * The vmemmap_free() and remove_section_mapping()
803			 * codepaths call us with aligned addresses.
804			 */
805			WARN_ONCE(1, "%s: unaligned range\n", __func__);
806			continue;
807		}
808
809		pte_clear(&init_mm, addr, pte);
810		pages++;
811	}
812	if (direct)
813		update_page_count(mmu_virtual_psize, -pages);
814}
815
816static void __meminit remove_pmd_table(pmd_t *pmd_start, unsigned long addr,
817				       unsigned long end, bool direct)
818{
819	unsigned long next, pages = 0;
820	pte_t *pte_base;
821	pmd_t *pmd;
822
823	pmd = pmd_start + pmd_index(addr);
824	for (; addr < end; addr = next, pmd++) {
825		next = pmd_addr_end(addr, end);
826
827		if (!pmd_present(*pmd))
828			continue;
829
830		if (pmd_is_leaf(*pmd)) {
831			if (!IS_ALIGNED(addr, PMD_SIZE) ||
832			    !IS_ALIGNED(next, PMD_SIZE)) {
833				WARN_ONCE(1, "%s: unaligned range\n", __func__);
834				continue;
835			}
836			pte_clear(&init_mm, addr, (pte_t *)pmd);
837			pages++;
838			continue;
839		}
840
841		pte_base = (pte_t *)pmd_page_vaddr(*pmd);
842		remove_pte_table(pte_base, addr, next, direct);
843		free_pte_table(pte_base, pmd);
844	}
845	if (direct)
846		update_page_count(MMU_PAGE_2M, -pages);
847}
848
849static void __meminit remove_pud_table(pud_t *pud_start, unsigned long addr,
850				       unsigned long end, bool direct)
851{
852	unsigned long next, pages = 0;
853	pmd_t *pmd_base;
854	pud_t *pud;
855
856	pud = pud_start + pud_index(addr);
857	for (; addr < end; addr = next, pud++) {
858		next = pud_addr_end(addr, end);
859
860		if (!pud_present(*pud))
861			continue;
862
863		if (pud_is_leaf(*pud)) {
864			if (!IS_ALIGNED(addr, PUD_SIZE) ||
865			    !IS_ALIGNED(next, PUD_SIZE)) {
866				WARN_ONCE(1, "%s: unaligned range\n", __func__);
867				continue;
868			}
869			pte_clear(&init_mm, addr, (pte_t *)pud);
870			pages++;
871			continue;
872		}
873
874		pmd_base = pud_pgtable(*pud);
875		remove_pmd_table(pmd_base, addr, next, direct);
876		free_pmd_table(pmd_base, pud);
877	}
878	if (direct)
879		update_page_count(MMU_PAGE_1G, -pages);
880}
881
882static void __meminit remove_pagetable(unsigned long start, unsigned long end,
883				       bool direct)
884{
885	unsigned long addr, next;
886	pud_t *pud_base;
887	pgd_t *pgd;
888	p4d_t *p4d;
889
890	spin_lock(&init_mm.page_table_lock);
891
892	for (addr = start; addr < end; addr = next) {
893		next = pgd_addr_end(addr, end);
894
895		pgd = pgd_offset_k(addr);
896		p4d = p4d_offset(pgd, addr);
897		if (!p4d_present(*p4d))
898			continue;
899
900		if (p4d_is_leaf(*p4d)) {
901			if (!IS_ALIGNED(addr, P4D_SIZE) ||
902			    !IS_ALIGNED(next, P4D_SIZE)) {
903				WARN_ONCE(1, "%s: unaligned range\n", __func__);
904				continue;
905			}
906
907			pte_clear(&init_mm, addr, (pte_t *)pgd);
908			continue;
909		}
910
911		pud_base = p4d_pgtable(*p4d);
912		remove_pud_table(pud_base, addr, next, direct);
913		free_pud_table(pud_base, p4d);
914	}
915
916	spin_unlock(&init_mm.page_table_lock);
917	radix__flush_tlb_kernel_range(start, end);
918}
919
920int __meminit radix__create_section_mapping(unsigned long start,
921					    unsigned long end, int nid,
922					    pgprot_t prot)
923{
924	if (end >= RADIX_VMALLOC_START) {
925		pr_warn("Outside the supported range\n");
926		return -1;
927	}
928
929	return create_physical_mapping(__pa(start), __pa(end),
930				       radix_mem_block_size, nid, prot);
931}
932
933int __meminit radix__remove_section_mapping(unsigned long start, unsigned long end)
934{
935	remove_pagetable(start, end, true);
936	return 0;
937}
938#endif /* CONFIG_MEMORY_HOTPLUG */
939
940#ifdef CONFIG_SPARSEMEM_VMEMMAP
941static int __map_kernel_page_nid(unsigned long ea, unsigned long pa,
942				 pgprot_t flags, unsigned int map_page_size,
943				 int nid)
944{
945	return __map_kernel_page(ea, pa, flags, map_page_size, nid, 0, 0);
946}
947
948int __meminit radix__vmemmap_create_mapping(unsigned long start,
949				      unsigned long page_size,
950				      unsigned long phys)
951{
952	/* Create a PTE encoding */
953	unsigned long flags = _PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_KERNEL_RW;
954	int nid = early_pfn_to_nid(phys >> PAGE_SHIFT);
955	int ret;
956
957	if ((start + page_size) >= RADIX_VMEMMAP_END) {
958		pr_warn("Outside the supported range\n");
959		return -1;
960	}
961
962	ret = __map_kernel_page_nid(start, phys, __pgprot(flags), page_size, nid);
963	BUG_ON(ret);
964
965	return 0;
966}
967
968#ifdef CONFIG_MEMORY_HOTPLUG
969void __meminit radix__vmemmap_remove_mapping(unsigned long start, unsigned long page_size)
970{
971	remove_pagetable(start, start + page_size, false);
972}
973#endif
974#endif
975
976#ifdef CONFIG_TRANSPARENT_HUGEPAGE
977
978unsigned long radix__pmd_hugepage_update(struct mm_struct *mm, unsigned long addr,
979				  pmd_t *pmdp, unsigned long clr,
980				  unsigned long set)
981{
982	unsigned long old;
983
984#ifdef CONFIG_DEBUG_VM
985	WARN_ON(!radix__pmd_trans_huge(*pmdp) && !pmd_devmap(*pmdp));
986	assert_spin_locked(pmd_lockptr(mm, pmdp));
987#endif
988
989	old = radix__pte_update(mm, addr, (pte_t *)pmdp, clr, set, 1);
990	trace_hugepage_update(addr, old, clr, set);
991
992	return old;
993}
994
995pmd_t radix__pmdp_collapse_flush(struct vm_area_struct *vma, unsigned long address,
996			pmd_t *pmdp)
997
998{
999	pmd_t pmd;
1000
1001	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
1002	VM_BUG_ON(radix__pmd_trans_huge(*pmdp));
1003	VM_BUG_ON(pmd_devmap(*pmdp));
1004	/*
1005	 * khugepaged calls this for normal pmd
1006	 */
1007	pmd = *pmdp;
1008	pmd_clear(pmdp);
1009
1010	radix__flush_tlb_collapsed_pmd(vma->vm_mm, address);
1011
1012	return pmd;
1013}
1014
1015/*
1016 * For us pgtable_t is pte_t *. Inorder to save the deposisted
1017 * page table, we consider the allocated page table as a list
1018 * head. On withdraw we need to make sure we zero out the used
1019 * list_head memory area.
1020 */
1021void radix__pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1022				 pgtable_t pgtable)
1023{
1024	struct list_head *lh = (struct list_head *) pgtable;
1025
1026	assert_spin_locked(pmd_lockptr(mm, pmdp));
1027
1028	/* FIFO */
1029	if (!pmd_huge_pte(mm, pmdp))
1030		INIT_LIST_HEAD(lh);
1031	else
1032		list_add(lh, (struct list_head *) pmd_huge_pte(mm, pmdp));
1033	pmd_huge_pte(mm, pmdp) = pgtable;
1034}
1035
1036pgtable_t radix__pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp)
1037{
1038	pte_t *ptep;
1039	pgtable_t pgtable;
1040	struct list_head *lh;
1041
1042	assert_spin_locked(pmd_lockptr(mm, pmdp));
1043
1044	/* FIFO */
1045	pgtable = pmd_huge_pte(mm, pmdp);
1046	lh = (struct list_head *) pgtable;
1047	if (list_empty(lh))
1048		pmd_huge_pte(mm, pmdp) = NULL;
1049	else {
1050		pmd_huge_pte(mm, pmdp) = (pgtable_t) lh->next;
1051		list_del(lh);
1052	}
1053	ptep = (pte_t *) pgtable;
1054	*ptep = __pte(0);
1055	ptep++;
1056	*ptep = __pte(0);
1057	return pgtable;
1058}
1059
1060pmd_t radix__pmdp_huge_get_and_clear(struct mm_struct *mm,
1061				     unsigned long addr, pmd_t *pmdp)
1062{
1063	pmd_t old_pmd;
1064	unsigned long old;
1065
1066	old = radix__pmd_hugepage_update(mm, addr, pmdp, ~0UL, 0);
1067	old_pmd = __pmd(old);
1068	return old_pmd;
1069}
1070
1071#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1072
1073void radix__ptep_set_access_flags(struct vm_area_struct *vma, pte_t *ptep,
1074				  pte_t entry, unsigned long address, int psize)
1075{
1076	struct mm_struct *mm = vma->vm_mm;
1077	unsigned long set = pte_val(entry) & (_PAGE_DIRTY | _PAGE_SOFT_DIRTY |
1078					      _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
1079
1080	unsigned long change = pte_val(entry) ^ pte_val(*ptep);
1081	/*
1082	 * To avoid NMMU hang while relaxing access, we need mark
1083	 * the pte invalid in between.
1084	 */
1085	if ((change & _PAGE_RW) && atomic_read(&mm->context.copros) > 0) {
1086		unsigned long old_pte, new_pte;
1087
1088		old_pte = __radix_pte_update(ptep, _PAGE_PRESENT, _PAGE_INVALID);
1089		/*
1090		 * new value of pte
1091		 */
1092		new_pte = old_pte | set;
1093		radix__flush_tlb_page_psize(mm, address, psize);
1094		__radix_pte_update(ptep, _PAGE_INVALID, new_pte);
1095	} else {
1096		__radix_pte_update(ptep, 0, set);
1097		/*
1098		 * Book3S does not require a TLB flush when relaxing access
1099		 * restrictions when the address space is not attached to a
1100		 * NMMU, because the core MMU will reload the pte after taking
1101		 * an access fault, which is defined by the architectue.
1102		 */
1103	}
1104	/* See ptesync comment in radix__set_pte_at */
1105}
1106
1107void radix__ptep_modify_prot_commit(struct vm_area_struct *vma,
1108				    unsigned long addr, pte_t *ptep,
1109				    pte_t old_pte, pte_t pte)
1110{
1111	struct mm_struct *mm = vma->vm_mm;
1112
1113	/*
1114	 * To avoid NMMU hang while relaxing access we need to flush the tlb before
1115	 * we set the new value. We need to do this only for radix, because hash
1116	 * translation does flush when updating the linux pte.
1117	 */
1118	if (is_pte_rw_upgrade(pte_val(old_pte), pte_val(pte)) &&
1119	    (atomic_read(&mm->context.copros) > 0))
1120		radix__flush_tlb_page(vma, addr);
1121
1122	set_pte_at(mm, addr, ptep, pte);
1123}
1124
1125int __init arch_ioremap_pud_supported(void)
1126{
1127	/* HPT does not cope with large pages in the vmalloc area */
1128	return radix_enabled();
1129}
1130
1131int __init arch_ioremap_pmd_supported(void)
1132{
1133	return radix_enabled();
1134}
1135
1136int p4d_free_pud_page(p4d_t *p4d, unsigned long addr)
1137{
1138	return 0;
1139}
1140
1141int pud_set_huge(pud_t *pud, phys_addr_t addr, pgprot_t prot)
1142{
1143	pte_t *ptep = (pte_t *)pud;
1144	pte_t new_pud = pfn_pte(__phys_to_pfn(addr), prot);
1145
1146	if (!radix_enabled())
1147		return 0;
1148
1149	set_pte_at(&init_mm, 0 /* radix unused */, ptep, new_pud);
1150
1151	return 1;
1152}
1153
1154int pud_clear_huge(pud_t *pud)
1155{
1156	if (pud_is_leaf(*pud)) {
1157		pud_clear(pud);
1158		return 1;
1159	}
1160
1161	return 0;
1162}
1163
1164int pud_free_pmd_page(pud_t *pud, unsigned long addr)
1165{
1166	pmd_t *pmd;
1167	int i;
1168
1169	pmd = pud_pgtable(*pud);
1170	pud_clear(pud);
1171
1172	flush_tlb_kernel_range(addr, addr + PUD_SIZE);
1173
1174	for (i = 0; i < PTRS_PER_PMD; i++) {
1175		if (!pmd_none(pmd[i])) {
1176			pte_t *pte;
1177			pte = (pte_t *)pmd_page_vaddr(pmd[i]);
1178
1179			pte_free_kernel(&init_mm, pte);
1180		}
1181	}
1182
1183	pmd_free(&init_mm, pmd);
1184
1185	return 1;
1186}
1187
1188int pmd_set_huge(pmd_t *pmd, phys_addr_t addr, pgprot_t prot)
1189{
1190	pte_t *ptep = (pte_t *)pmd;
1191	pte_t new_pmd = pfn_pte(__phys_to_pfn(addr), prot);
1192
1193	if (!radix_enabled())
1194		return 0;
1195
1196	set_pte_at(&init_mm, 0 /* radix unused */, ptep, new_pmd);
1197
1198	return 1;
1199}
1200
1201int pmd_clear_huge(pmd_t *pmd)
1202{
1203	if (pmd_is_leaf(*pmd)) {
1204		pmd_clear(pmd);
1205		return 1;
1206	}
1207
1208	return 0;
1209}
1210
1211int pmd_free_pte_page(pmd_t *pmd, unsigned long addr)
1212{
1213	pte_t *pte;
1214
1215	pte = (pte_t *)pmd_page_vaddr(*pmd);
1216	pmd_clear(pmd);
1217
1218	flush_tlb_kernel_range(addr, addr + PMD_SIZE);
1219
1220	pte_free_kernel(&init_mm, pte);
1221
1222	return 1;
1223}
1224
1225int __init arch_ioremap_p4d_supported(void)
1226{
1227	return 0;
1228}
1229