1// SPDX-License-Identifier: GPL-2.0-or-later 2/* 3 * This file contains common routines for dealing with free of page tables 4 * Along with common page table handling code 5 * 6 * Derived from arch/powerpc/mm/tlb_64.c: 7 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org) 8 * 9 * Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au) 10 * and Cort Dougan (PReP) (cort@cs.nmt.edu) 11 * Copyright (C) 1996 Paul Mackerras 12 * 13 * Derived from "arch/i386/mm/init.c" 14 * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds 15 * 16 * Dave Engebretsen <engebret@us.ibm.com> 17 * Rework for PPC64 port. 18 */ 19 20#include <linux/kernel.h> 21#include <linux/gfp.h> 22#include <linux/mm.h> 23#include <linux/percpu.h> 24#include <linux/hardirq.h> 25#include <linux/hugetlb.h> 26#include <asm/tlbflush.h> 27#include <asm/tlb.h> 28#include <asm/hugetlb.h> 29 30static inline int is_exec_fault(void) 31{ 32 return current->thread.regs && TRAP(current->thread.regs) == 0x400; 33} 34 35/* We only try to do i/d cache coherency on stuff that looks like 36 * reasonably "normal" PTEs. We currently require a PTE to be present 37 * and we avoid _PAGE_SPECIAL and cache inhibited pte. We also only do that 38 * on userspace PTEs 39 */ 40static inline int pte_looks_normal(pte_t pte) 41{ 42 43 if (pte_present(pte) && !pte_special(pte)) { 44 if (pte_ci(pte)) 45 return 0; 46 if (pte_user(pte)) 47 return 1; 48 } 49 return 0; 50} 51 52static struct page *maybe_pte_to_page(pte_t pte) 53{ 54 unsigned long pfn = pte_pfn(pte); 55 struct page *page; 56 57 if (unlikely(!pfn_valid(pfn))) 58 return NULL; 59 page = pfn_to_page(pfn); 60 if (PageReserved(page)) 61 return NULL; 62 return page; 63} 64 65#ifdef CONFIG_PPC_BOOK3S 66 67/* Server-style MMU handles coherency when hashing if HW exec permission 68 * is supposed per page (currently 64-bit only). If not, then, we always 69 * flush the cache for valid PTEs in set_pte. Embedded CPU without HW exec 70 * support falls into the same category. 71 */ 72 73static pte_t set_pte_filter_hash(pte_t pte) 74{ 75 if (radix_enabled()) 76 return pte; 77 78 pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS); 79 if (pte_looks_normal(pte) && !(cpu_has_feature(CPU_FTR_COHERENT_ICACHE) || 80 cpu_has_feature(CPU_FTR_NOEXECUTE))) { 81 struct page *pg = maybe_pte_to_page(pte); 82 if (!pg) 83 return pte; 84 if (!test_bit(PG_arch_1, &pg->flags)) { 85 flush_dcache_icache_page(pg); 86 set_bit(PG_arch_1, &pg->flags); 87 } 88 } 89 return pte; 90} 91 92#else /* CONFIG_PPC_BOOK3S */ 93 94static pte_t set_pte_filter_hash(pte_t pte) { return pte; } 95 96#endif /* CONFIG_PPC_BOOK3S */ 97 98/* Embedded type MMU with HW exec support. This is a bit more complicated 99 * as we don't have two bits to spare for _PAGE_EXEC and _PAGE_HWEXEC so 100 * instead we "filter out" the exec permission for non clean pages. 101 */ 102static inline pte_t set_pte_filter(pte_t pte) 103{ 104 struct page *pg; 105 106 if (mmu_has_feature(MMU_FTR_HPTE_TABLE)) 107 return set_pte_filter_hash(pte); 108 109 /* No exec permission in the first place, move on */ 110 if (!pte_exec(pte) || !pte_looks_normal(pte)) 111 return pte; 112 113 /* If you set _PAGE_EXEC on weird pages you're on your own */ 114 pg = maybe_pte_to_page(pte); 115 if (unlikely(!pg)) 116 return pte; 117 118 /* If the page clean, we move on */ 119 if (test_bit(PG_arch_1, &pg->flags)) 120 return pte; 121 122 /* If it's an exec fault, we flush the cache and make it clean */ 123 if (is_exec_fault()) { 124 flush_dcache_icache_page(pg); 125 set_bit(PG_arch_1, &pg->flags); 126 return pte; 127 } 128 129 /* Else, we filter out _PAGE_EXEC */ 130 return pte_exprotect(pte); 131} 132 133static pte_t set_access_flags_filter(pte_t pte, struct vm_area_struct *vma, 134 int dirty) 135{ 136 struct page *pg; 137 138 if (mmu_has_feature(MMU_FTR_HPTE_TABLE)) 139 return pte; 140 141 /* So here, we only care about exec faults, as we use them 142 * to recover lost _PAGE_EXEC and perform I$/D$ coherency 143 * if necessary. Also if _PAGE_EXEC is already set, same deal, 144 * we just bail out 145 */ 146 if (dirty || pte_exec(pte) || !is_exec_fault()) 147 return pte; 148 149#ifdef CONFIG_DEBUG_VM 150 /* So this is an exec fault, _PAGE_EXEC is not set. If it was 151 * an error we would have bailed out earlier in do_page_fault() 152 * but let's make sure of it 153 */ 154 if (WARN_ON(!(vma->vm_flags & VM_EXEC))) 155 return pte; 156#endif /* CONFIG_DEBUG_VM */ 157 158 /* If you set _PAGE_EXEC on weird pages you're on your own */ 159 pg = maybe_pte_to_page(pte); 160 if (unlikely(!pg)) 161 goto bail; 162 163 /* If the page is already clean, we move on */ 164 if (test_bit(PG_arch_1, &pg->flags)) 165 goto bail; 166 167 /* Clean the page and set PG_arch_1 */ 168 flush_dcache_icache_page(pg); 169 set_bit(PG_arch_1, &pg->flags); 170 171 bail: 172 return pte_mkexec(pte); 173} 174 175/* 176 * set_pte stores a linux PTE into the linux page table. 177 */ 178void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, 179 pte_t pte) 180{ 181 /* 182 * Make sure hardware valid bit is not set. We don't do 183 * tlb flush for this update. 184 */ 185 VM_WARN_ON(pte_hw_valid(*ptep) && !pte_protnone(*ptep)); 186 187 /* Note: mm->context.id might not yet have been assigned as 188 * this context might not have been activated yet when this 189 * is called. 190 */ 191 pte = set_pte_filter(pte); 192 193 /* Perform the setting of the PTE */ 194 __set_pte_at(mm, addr, ptep, pte, 0); 195} 196 197void unmap_kernel_page(unsigned long va) 198{ 199 pmd_t *pmdp = pmd_off_k(va); 200 pte_t *ptep = pte_offset_kernel(pmdp, va); 201 202 pte_clear(&init_mm, va, ptep); 203 flush_tlb_kernel_range(va, va + PAGE_SIZE); 204} 205 206/* 207 * This is called when relaxing access to a PTE. It's also called in the page 208 * fault path when we don't hit any of the major fault cases, ie, a minor 209 * update of _PAGE_ACCESSED, _PAGE_DIRTY, etc... The generic code will have 210 * handled those two for us, we additionally deal with missing execute 211 * permission here on some processors 212 */ 213int ptep_set_access_flags(struct vm_area_struct *vma, unsigned long address, 214 pte_t *ptep, pte_t entry, int dirty) 215{ 216 int changed; 217 entry = set_access_flags_filter(entry, vma, dirty); 218 changed = !pte_same(*(ptep), entry); 219 if (changed) { 220 assert_pte_locked(vma->vm_mm, address); 221 __ptep_set_access_flags(vma, ptep, entry, 222 address, mmu_virtual_psize); 223 } 224 return changed; 225} 226 227#ifdef CONFIG_HUGETLB_PAGE 228int huge_ptep_set_access_flags(struct vm_area_struct *vma, 229 unsigned long addr, pte_t *ptep, 230 pte_t pte, int dirty) 231{ 232#ifdef HUGETLB_NEED_PRELOAD 233 /* 234 * The "return 1" forces a call of update_mmu_cache, which will write a 235 * TLB entry. Without this, platforms that don't do a write of the TLB 236 * entry in the TLB miss handler asm will fault ad infinitum. 237 */ 238 ptep_set_access_flags(vma, addr, ptep, pte, dirty); 239 return 1; 240#else 241 int changed, psize; 242 243 pte = set_access_flags_filter(pte, vma, dirty); 244 changed = !pte_same(*(ptep), pte); 245 if (changed) { 246 247#ifdef CONFIG_PPC_BOOK3S_64 248 struct hstate *h = hstate_vma(vma); 249 250 psize = hstate_get_psize(h); 251#ifdef CONFIG_DEBUG_VM 252 assert_spin_locked(huge_pte_lockptr(h, vma->vm_mm, ptep)); 253#endif 254 255#else 256 /* 257 * Not used on non book3s64 platforms. 258 * 8xx compares it with mmu_virtual_psize to 259 * know if it is a huge page or not. 260 */ 261 psize = MMU_PAGE_COUNT; 262#endif 263 __ptep_set_access_flags(vma, ptep, pte, addr, psize); 264 } 265 return changed; 266#endif 267} 268 269#if defined(CONFIG_PPC_8xx) 270void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte) 271{ 272 pmd_t *pmd = pmd_off(mm, addr); 273 pte_basic_t val; 274 pte_basic_t *entry = &ptep->pte; 275 int num, i; 276 277 /* 278 * Make sure hardware valid bit is not set. We don't do 279 * tlb flush for this update. 280 */ 281 VM_WARN_ON(pte_hw_valid(*ptep) && !pte_protnone(*ptep)); 282 283 pte = set_pte_filter(pte); 284 285 val = pte_val(pte); 286 287 num = number_of_cells_per_pte(pmd, val, 1); 288 289 for (i = 0; i < num; i++, entry++, val += SZ_4K) 290 *entry = val; 291} 292#endif 293#endif /* CONFIG_HUGETLB_PAGE */ 294 295#ifdef CONFIG_DEBUG_VM 296void assert_pte_locked(struct mm_struct *mm, unsigned long addr) 297{ 298 pgd_t *pgd; 299 p4d_t *p4d; 300 pud_t *pud; 301 pmd_t *pmd; 302 303 if (mm == &init_mm) 304 return; 305 pgd = mm->pgd + pgd_index(addr); 306 BUG_ON(pgd_none(*pgd)); 307 p4d = p4d_offset(pgd, addr); 308 BUG_ON(p4d_none(*p4d)); 309 pud = pud_offset(p4d, addr); 310 BUG_ON(pud_none(*pud)); 311 pmd = pmd_offset(pud, addr); 312 /* 313 * khugepaged to collapse normal pages to hugepage, first set 314 * pmd to none to force page fault/gup to take mmap_lock. After 315 * pmd is set to none, we do a pte_clear which does this assertion 316 * so if we find pmd none, return. 317 */ 318 if (pmd_none(*pmd)) 319 return; 320 BUG_ON(!pmd_present(*pmd)); 321 assert_spin_locked(pte_lockptr(mm, pmd)); 322} 323#endif /* CONFIG_DEBUG_VM */ 324 325unsigned long vmalloc_to_phys(void *va) 326{ 327 unsigned long pfn = vmalloc_to_pfn(va); 328 329 BUG_ON(!pfn); 330 return __pa(pfn_to_kaddr(pfn)) + offset_in_page(va); 331} 332EXPORT_SYMBOL_GPL(vmalloc_to_phys); 333 334/* 335 * We have 4 cases for pgds and pmds: 336 * (1) invalid (all zeroes) 337 * (2) pointer to next table, as normal; bottom 6 bits == 0 338 * (3) leaf pte for huge page _PAGE_PTE set 339 * (4) hugepd pointer, _PAGE_PTE = 0 and bits [2..6] indicate size of table 340 * 341 * So long as we atomically load page table pointers we are safe against teardown, 342 * we can follow the address down to the the page and take a ref on it. 343 * This function need to be called with interrupts disabled. We use this variant 344 * when we have MSR[EE] = 0 but the paca->irq_soft_mask = IRQS_ENABLED 345 */ 346pte_t *__find_linux_pte(pgd_t *pgdir, unsigned long ea, 347 bool *is_thp, unsigned *hpage_shift) 348{ 349 pgd_t *pgdp; 350 p4d_t p4d, *p4dp; 351 pud_t pud, *pudp; 352 pmd_t pmd, *pmdp; 353 pte_t *ret_pte; 354 hugepd_t *hpdp = NULL; 355 unsigned pdshift; 356 357 if (hpage_shift) 358 *hpage_shift = 0; 359 360 if (is_thp) 361 *is_thp = false; 362 363 /* 364 * Always operate on the local stack value. This make sure the 365 * value don't get updated by a parallel THP split/collapse, 366 * page fault or a page unmap. The return pte_t * is still not 367 * stable. So should be checked there for above conditions. 368 * Top level is an exception because it is folded into p4d. 369 */ 370 pgdp = pgdir + pgd_index(ea); 371 p4dp = p4d_offset(pgdp, ea); 372 p4d = READ_ONCE(*p4dp); 373 pdshift = P4D_SHIFT; 374 375 if (p4d_none(p4d)) 376 return NULL; 377 378 if (p4d_is_leaf(p4d)) { 379 ret_pte = (pte_t *)p4dp; 380 goto out; 381 } 382 383 if (is_hugepd(__hugepd(p4d_val(p4d)))) { 384 hpdp = (hugepd_t *)&p4d; 385 goto out_huge; 386 } 387 388 /* 389 * Even if we end up with an unmap, the pgtable will not 390 * be freed, because we do an rcu free and here we are 391 * irq disabled 392 */ 393 pdshift = PUD_SHIFT; 394 pudp = pud_offset(&p4d, ea); 395 pud = READ_ONCE(*pudp); 396 397 if (pud_none(pud)) 398 return NULL; 399 400 if (pud_is_leaf(pud)) { 401 ret_pte = (pte_t *)pudp; 402 goto out; 403 } 404 405 if (is_hugepd(__hugepd(pud_val(pud)))) { 406 hpdp = (hugepd_t *)&pud; 407 goto out_huge; 408 } 409 410 pdshift = PMD_SHIFT; 411 pmdp = pmd_offset(&pud, ea); 412 pmd = READ_ONCE(*pmdp); 413 414 /* 415 * A hugepage collapse is captured by this condition, see 416 * pmdp_collapse_flush. 417 */ 418 if (pmd_none(pmd)) 419 return NULL; 420 421#ifdef CONFIG_PPC_BOOK3S_64 422 /* 423 * A hugepage split is captured by this condition, see 424 * pmdp_invalidate. 425 * 426 * Huge page modification can be caught here too. 427 */ 428 if (pmd_is_serializing(pmd)) 429 return NULL; 430#endif 431 432 if (pmd_trans_huge(pmd) || pmd_devmap(pmd)) { 433 if (is_thp) 434 *is_thp = true; 435 ret_pte = (pte_t *)pmdp; 436 goto out; 437 } 438 439 if (pmd_is_leaf(pmd)) { 440 ret_pte = (pte_t *)pmdp; 441 goto out; 442 } 443 444 if (is_hugepd(__hugepd(pmd_val(pmd)))) { 445 hpdp = (hugepd_t *)&pmd; 446 goto out_huge; 447 } 448 449 return pte_offset_kernel(&pmd, ea); 450 451out_huge: 452 if (!hpdp) 453 return NULL; 454 455 ret_pte = hugepte_offset(*hpdp, ea, pdshift); 456 pdshift = hugepd_shift(*hpdp); 457out: 458 if (hpage_shift) 459 *hpage_shift = pdshift; 460 return ret_pte; 461} 462EXPORT_SYMBOL_GPL(__find_linux_pte); 463