1/*
2 * This file contains common routines for dealing with free of page tables
3 * Along with common page table handling code
4 *
5 *  Derived from arch/powerpc/mm/tlb_64.c:
6 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
7 *
8 *  Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au)
9 *  and Cort Dougan (PReP) (cort@cs.nmt.edu)
10 *    Copyright (C) 1996 Paul Mackerras
11 *
12 *  Derived from "arch/i386/mm/init.c"
13 *    Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
14 *
15 *  Dave Engebretsen <engebret@us.ibm.com>
16 *      Rework for PPC64 port.
17 *
18 *  This program is free software; you can redistribute it and/or
19 *  modify it under the terms of the GNU General Public License
20 *  as published by the Free Software Foundation; either version
21 *  2 of the License, or (at your option) any later version.
22 */
23
24#include <linux/kernel.h>
25#include <linux/gfp.h>
26#include <linux/mm.h>
27#include <linux/percpu.h>
28#include <linux/hardirq.h>
29#include <linux/hugetlb.h>
30#include <asm/pgalloc.h>
31#include <asm/tlbflush.h>
32#include <asm/tlb.h>
33
34static inline int is_exec_fault(void)
35{
36	return current->thread.regs && TRAP(current->thread.regs) == 0x400;
37}
38
39/* We only try to do i/d cache coherency on stuff that looks like
40 * reasonably "normal" PTEs. We currently require a PTE to be present
41 * and we avoid _PAGE_SPECIAL and _PAGE_NO_CACHE. We also only do that
42 * on userspace PTEs
43 */
44static inline int pte_looks_normal(pte_t pte)
45{
46	return (pte_val(pte) &
47	    (_PAGE_PRESENT | _PAGE_SPECIAL | _PAGE_NO_CACHE | _PAGE_USER)) ==
48	    (_PAGE_PRESENT | _PAGE_USER);
49}
50
51static struct page *maybe_pte_to_page(pte_t pte)
52{
53	unsigned long pfn = pte_pfn(pte);
54	struct page *page;
55
56	if (unlikely(!pfn_valid(pfn)))
57		return NULL;
58	page = pfn_to_page(pfn);
59	if (PageReserved(page))
60		return NULL;
61	return page;
62}
63
64#if defined(CONFIG_PPC_STD_MMU) || _PAGE_EXEC == 0
65
66/* Server-style MMU handles coherency when hashing if HW exec permission
67 * is supposed per page (currently 64-bit only). If not, then, we always
68 * flush the cache for valid PTEs in set_pte. Embedded CPU without HW exec
69 * support falls into the same category.
70 */
71
72static pte_t set_pte_filter(pte_t pte)
73{
74	pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
75	if (pte_looks_normal(pte) && !(cpu_has_feature(CPU_FTR_COHERENT_ICACHE) ||
76				       cpu_has_feature(CPU_FTR_NOEXECUTE))) {
77		struct page *pg = maybe_pte_to_page(pte);
78		if (!pg)
79			return pte;
80		if (!test_bit(PG_arch_1, &pg->flags)) {
81			flush_dcache_icache_page(pg);
82			set_bit(PG_arch_1, &pg->flags);
83		}
84	}
85	return pte;
86}
87
88static pte_t set_access_flags_filter(pte_t pte, struct vm_area_struct *vma,
89				     int dirty)
90{
91	return pte;
92}
93
94#else /* defined(CONFIG_PPC_STD_MMU) || _PAGE_EXEC == 0 */
95
96/* Embedded type MMU with HW exec support. This is a bit more complicated
97 * as we don't have two bits to spare for _PAGE_EXEC and _PAGE_HWEXEC so
98 * instead we "filter out" the exec permission for non clean pages.
99 */
100static pte_t set_pte_filter(pte_t pte)
101{
102	struct page *pg;
103
104	/* No exec permission in the first place, move on */
105	if (!(pte_val(pte) & _PAGE_EXEC) || !pte_looks_normal(pte))
106		return pte;
107
108	/* If you set _PAGE_EXEC on weird pages you're on your own */
109	pg = maybe_pte_to_page(pte);
110	if (unlikely(!pg))
111		return pte;
112
113	/* If the page clean, we move on */
114	if (test_bit(PG_arch_1, &pg->flags))
115		return pte;
116
117	/* If it's an exec fault, we flush the cache and make it clean */
118	if (is_exec_fault()) {
119		flush_dcache_icache_page(pg);
120		set_bit(PG_arch_1, &pg->flags);
121		return pte;
122	}
123
124	/* Else, we filter out _PAGE_EXEC */
125	return __pte(pte_val(pte) & ~_PAGE_EXEC);
126}
127
128static pte_t set_access_flags_filter(pte_t pte, struct vm_area_struct *vma,
129				     int dirty)
130{
131	struct page *pg;
132
133	/* So here, we only care about exec faults, as we use them
134	 * to recover lost _PAGE_EXEC and perform I$/D$ coherency
135	 * if necessary. Also if _PAGE_EXEC is already set, same deal,
136	 * we just bail out
137	 */
138	if (dirty || (pte_val(pte) & _PAGE_EXEC) || !is_exec_fault())
139		return pte;
140
141#ifdef CONFIG_DEBUG_VM
142	/* So this is an exec fault, _PAGE_EXEC is not set. If it was
143	 * an error we would have bailed out earlier in do_page_fault()
144	 * but let's make sure of it
145	 */
146	if (WARN_ON(!(vma->vm_flags & VM_EXEC)))
147		return pte;
148#endif /* CONFIG_DEBUG_VM */
149
150	/* If you set _PAGE_EXEC on weird pages you're on your own */
151	pg = maybe_pte_to_page(pte);
152	if (unlikely(!pg))
153		goto bail;
154
155	/* If the page is already clean, we move on */
156	if (test_bit(PG_arch_1, &pg->flags))
157		goto bail;
158
159	/* Clean the page and set PG_arch_1 */
160	flush_dcache_icache_page(pg);
161	set_bit(PG_arch_1, &pg->flags);
162
163 bail:
164	return __pte(pte_val(pte) | _PAGE_EXEC);
165}
166
167#endif /* !(defined(CONFIG_PPC_STD_MMU) || _PAGE_EXEC == 0) */
168
169/*
170 * set_pte stores a linux PTE into the linux page table.
171 */
172void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep,
173		pte_t pte)
174{
175	/*
176	 * When handling numa faults, we already have the pte marked
177	 * _PAGE_PRESENT, but we can be sure that it is not in hpte.
178	 * Hence we can use set_pte_at for them.
179	 */
180	VM_WARN_ON((pte_val(*ptep) & (_PAGE_PRESENT | _PAGE_USER)) ==
181		(_PAGE_PRESENT | _PAGE_USER));
182
183	/* Note: mm->context.id might not yet have been assigned as
184	 * this context might not have been activated yet when this
185	 * is called.
186	 */
187	pte = set_pte_filter(pte);
188
189	/* Perform the setting of the PTE */
190	__set_pte_at(mm, addr, ptep, pte, 0);
191}
192
193/*
194 * This is called when relaxing access to a PTE. It's also called in the page
195 * fault path when we don't hit any of the major fault cases, ie, a minor
196 * update of _PAGE_ACCESSED, _PAGE_DIRTY, etc... The generic code will have
197 * handled those two for us, we additionally deal with missing execute
198 * permission here on some processors
199 */
200int ptep_set_access_flags(struct vm_area_struct *vma, unsigned long address,
201			  pte_t *ptep, pte_t entry, int dirty)
202{
203	int changed;
204	entry = set_access_flags_filter(entry, vma, dirty);
205	changed = !pte_same(*(ptep), entry);
206	if (changed) {
207		if (!is_vm_hugetlb_page(vma))
208			assert_pte_locked(vma->vm_mm, address);
209		__ptep_set_access_flags(ptep, entry);
210		flush_tlb_page_nohash(vma, address);
211	}
212	return changed;
213}
214
215#ifdef CONFIG_DEBUG_VM
216void assert_pte_locked(struct mm_struct *mm, unsigned long addr)
217{
218	pgd_t *pgd;
219	pud_t *pud;
220	pmd_t *pmd;
221
222	if (mm == &init_mm)
223		return;
224	pgd = mm->pgd + pgd_index(addr);
225	BUG_ON(pgd_none(*pgd));
226	pud = pud_offset(pgd, addr);
227	BUG_ON(pud_none(*pud));
228	pmd = pmd_offset(pud, addr);
229	/*
230	 * khugepaged to collapse normal pages to hugepage, first set
231	 * pmd to none to force page fault/gup to take mmap_sem. After
232	 * pmd is set to none, we do a pte_clear which does this assertion
233	 * so if we find pmd none, return.
234	 */
235	if (pmd_none(*pmd))
236		return;
237	BUG_ON(!pmd_present(*pmd));
238	assert_spin_locked(pte_lockptr(mm, pmd));
239}
240#endif /* CONFIG_DEBUG_VM */
241
242