sh: use the generic get_user_pages_fast code
authorChristoph Hellwig <hch@lst.de>
Fri, 12 Jul 2019 03:57:00 +0000 (20:57 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 12 Jul 2019 18:05:44 +0000 (11:05 -0700)
The sh code is mostly equivalent to the generic one, minus various
bugfixes and two arch overrides that this patch adds to pgtable.h.

Link: http://lkml.kernel.org/r/20190625143715.1689-7-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Miller <davem@davemloft.net>
Cc: James Hogan <jhogan@kernel.org>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Rich Felker <dalias@libc.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
arch/sh/Kconfig
arch/sh/include/asm/pgtable.h
arch/sh/mm/Makefile
arch/sh/mm/gup.c [deleted file]

index c7c99e1..02ae813 100644 (file)
@@ -15,6 +15,7 @@ config SUPERH
        select HAVE_ARCH_TRACEHOOK
        select HAVE_PERF_EVENTS
        select HAVE_DEBUG_BUGVERBOSE
+       select HAVE_GENERIC_GUP
        select ARCH_HAVE_CUSTOM_GPIO_H
        select ARCH_HAVE_NMI_SAFE_CMPXCHG if (GUSA_RB || CPU_SH4A)
        select ARCH_HAS_GCOV_PROFILE_ALL
@@ -64,6 +65,7 @@ config SUPERH
 config SUPERH32
        def_bool "$(ARCH)" = "sh"
        select ARCH_32BIT_OFF_T
+       select GUP_GET_PTE_LOW_HIGH if X2TLB
        select HAVE_KPROBES
        select HAVE_KRETPROBES
        select HAVE_IOREMAP_PROT if MMU && !X2TLB
index 3587103..9085d11 100644 (file)
@@ -149,6 +149,43 @@ extern void paging_init(void);
 extern void page_table_range_init(unsigned long start, unsigned long end,
                                  pgd_t *pgd);
 
+static inline bool __pte_access_permitted(pte_t pte, u64 prot)
+{
+       return (pte_val(pte) & (prot | _PAGE_SPECIAL)) == prot;
+}
+
+#ifdef CONFIG_X2TLB
+static inline bool pte_access_permitted(pte_t pte, bool write)
+{
+       u64 prot = _PAGE_PRESENT;
+
+       prot |= _PAGE_EXT(_PAGE_EXT_KERN_READ | _PAGE_EXT_USER_READ);
+       if (write)
+               prot |= _PAGE_EXT(_PAGE_EXT_KERN_WRITE | _PAGE_EXT_USER_WRITE);
+       return __pte_access_permitted(pte, prot);
+}
+#elif defined(CONFIG_SUPERH64)
+static inline bool pte_access_permitted(pte_t pte, bool write)
+{
+       u64 prot = _PAGE_PRESENT | _PAGE_USER | _PAGE_READ;
+
+       if (write)
+               prot |= _PAGE_WRITE;
+       return __pte_access_permitted(pte, prot);
+}
+#else
+static inline bool pte_access_permitted(pte_t pte, bool write)
+{
+       u64 prot = _PAGE_PRESENT | _PAGE_USER;
+
+       if (write)
+               prot |= _PAGE_RW;
+       return __pte_access_permitted(pte, prot);
+}
+#endif
+
+#define pte_access_permitted pte_access_permitted
+
 /* arch/sh/mm/mmap.c */
 #define HAVE_ARCH_UNMAPPED_AREA
 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
index fbe5e79..5051b38 100644 (file)
@@ -17,7 +17,7 @@ cacheops-$(CONFIG_CPU_SHX3)           += cache-shx3.o
 obj-y                  += $(cacheops-y)
 
 mmu-y                  := nommu.o extable_32.o
-mmu-$(CONFIG_MMU)      := extable_$(BITS).o fault.o gup.o ioremap.o kmap.o \
+mmu-$(CONFIG_MMU)      := extable_$(BITS).o fault.o ioremap.o kmap.o \
                           pgtable.o tlbex_$(BITS).o tlbflush_$(BITS).o
 
 obj-y                  += $(mmu-y)
diff --git a/arch/sh/mm/gup.c b/arch/sh/mm/gup.c
deleted file mode 100644 (file)
index 277c882..0000000
+++ /dev/null
@@ -1,277 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0
-/*
- * Lockless get_user_pages_fast for SuperH
- *
- * Copyright (C) 2009 - 2010  Paul Mundt
- *
- * Cloned from the x86 and PowerPC versions, by:
- *
- *     Copyright (C) 2008 Nick Piggin
- *     Copyright (C) 2008 Novell Inc.
- */
-#include <linux/sched.h>
-#include <linux/mm.h>
-#include <linux/vmstat.h>
-#include <linux/highmem.h>
-#include <asm/pgtable.h>
-
-static inline pte_t gup_get_pte(pte_t *ptep)
-{
-#ifndef CONFIG_X2TLB
-       return READ_ONCE(*ptep);
-#else
-       /*
-        * With get_user_pages_fast, we walk down the pagetables without
-        * taking any locks.  For this we would like to load the pointers
-        * atomically, but that is not possible with 64-bit PTEs.  What
-        * we do have is the guarantee that a pte will only either go
-        * from not present to present, or present to not present or both
-        * -- it will not switch to a completely different present page
-        * without a TLB flush in between; something that we are blocking
-        * by holding interrupts off.
-        *
-        * Setting ptes from not present to present goes:
-        * ptep->pte_high = h;
-        * smp_wmb();
-        * ptep->pte_low = l;
-        *
-        * And present to not present goes:
-        * ptep->pte_low = 0;
-        * smp_wmb();
-        * ptep->pte_high = 0;
-        *
-        * We must ensure here that the load of pte_low sees l iff pte_high
-        * sees h. We load pte_high *after* loading pte_low, which ensures we
-        * don't see an older value of pte_high.  *Then* we recheck pte_low,
-        * which ensures that we haven't picked up a changed pte high. We might
-        * have got rubbish values from pte_low and pte_high, but we are
-        * guaranteed that pte_low will not have the present bit set *unless*
-        * it is 'l'. And get_user_pages_fast only operates on present ptes, so
-        * we're safe.
-        *
-        * gup_get_pte should not be used or copied outside gup.c without being
-        * very careful -- it does not atomically load the pte or anything that
-        * is likely to be useful for you.
-        */
-       pte_t pte;
-
-retry:
-       pte.pte_low = ptep->pte_low;
-       smp_rmb();
-       pte.pte_high = ptep->pte_high;
-       smp_rmb();
-       if (unlikely(pte.pte_low != ptep->pte_low))
-               goto retry;
-
-       return pte;
-#endif
-}
-
-/*
- * The performance critical leaf functions are made noinline otherwise gcc
- * inlines everything into a single function which results in too much
- * register pressure.
- */
-static noinline int gup_pte_range(pmd_t pmd, unsigned long addr,
-               unsigned long end, int write, struct page **pages, int *nr)
-{
-       u64 mask, result;
-       pte_t *ptep;
-
-#ifdef CONFIG_X2TLB
-       result = _PAGE_PRESENT | _PAGE_EXT(_PAGE_EXT_KERN_READ | _PAGE_EXT_USER_READ);
-       if (write)
-               result |= _PAGE_EXT(_PAGE_EXT_KERN_WRITE | _PAGE_EXT_USER_WRITE);
-#elif defined(CONFIG_SUPERH64)
-       result = _PAGE_PRESENT | _PAGE_USER | _PAGE_READ;
-       if (write)
-               result |= _PAGE_WRITE;
-#else
-       result = _PAGE_PRESENT | _PAGE_USER;
-       if (write)
-               result |= _PAGE_RW;
-#endif
-
-       mask = result | _PAGE_SPECIAL;
-
-       ptep = pte_offset_map(&pmd, addr);
-       do {
-               pte_t pte = gup_get_pte(ptep);
-               struct page *page;
-
-               if ((pte_val(pte) & mask) != result) {
-                       pte_unmap(ptep);
-                       return 0;
-               }
-               VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
-               page = pte_page(pte);
-               get_page(page);
-               __flush_anon_page(page, addr);
-               flush_dcache_page(page);
-               pages[*nr] = page;
-               (*nr)++;
-
-       } while (ptep++, addr += PAGE_SIZE, addr != end);
-       pte_unmap(ptep - 1);
-
-       return 1;
-}
-
-static int gup_pmd_range(pud_t pud, unsigned long addr, unsigned long end,
-               int write, struct page **pages, int *nr)
-{
-       unsigned long next;
-       pmd_t *pmdp;
-
-       pmdp = pmd_offset(&pud, addr);
-       do {
-               pmd_t pmd = *pmdp;
-
-               next = pmd_addr_end(addr, end);
-               if (pmd_none(pmd))
-                       return 0;
-               if (!gup_pte_range(pmd, addr, next, write, pages, nr))
-                       return 0;
-       } while (pmdp++, addr = next, addr != end);
-
-       return 1;
-}
-
-static int gup_pud_range(pgd_t pgd, unsigned long addr, unsigned long end,
-                       int write, struct page **pages, int *nr)
-{
-       unsigned long next;
-       pud_t *pudp;
-
-       pudp = pud_offset(&pgd, addr);
-       do {
-               pud_t pud = *pudp;
-
-               next = pud_addr_end(addr, end);
-               if (pud_none(pud))
-                       return 0;
-               if (!gup_pmd_range(pud, addr, next, write, pages, nr))
-                       return 0;
-       } while (pudp++, addr = next, addr != end);
-
-       return 1;
-}
-
-/*
- * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
- * back to the regular GUP.
- * Note a difference with get_user_pages_fast: this always returns the
- * number of pages pinned, 0 if no pages were pinned.
- */
-int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
-                         struct page **pages)
-{
-       struct mm_struct *mm = current->mm;
-       unsigned long addr, len, end;
-       unsigned long next;
-       unsigned long flags;
-       pgd_t *pgdp;
-       int nr = 0;
-
-       start &= PAGE_MASK;
-       addr = start;
-       len = (unsigned long) nr_pages << PAGE_SHIFT;
-       end = start + len;
-       if (unlikely(!access_ok((void __user *)start, len)))
-               return 0;
-
-       /*
-        * This doesn't prevent pagetable teardown, but does prevent
-        * the pagetables and pages from being freed.
-        */
-       local_irq_save(flags);
-       pgdp = pgd_offset(mm, addr);
-       do {
-               pgd_t pgd = *pgdp;
-
-               next = pgd_addr_end(addr, end);
-               if (pgd_none(pgd))
-                       break;
-               if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
-                       break;
-       } while (pgdp++, addr = next, addr != end);
-       local_irq_restore(flags);
-
-       return nr;
-}
-
-/**
- * get_user_pages_fast() - pin user pages in memory
- * @start:     starting user address
- * @nr_pages:  number of pages from start to pin
- * @gup_flags: flags modifying pin behaviour
- * @pages:     array that receives pointers to the pages pinned.
- *             Should be at least nr_pages long.
- *
- * Attempt to pin user pages in memory without taking mm->mmap_sem.
- * If not successful, it will fall back to taking the lock and
- * calling get_user_pages().
- *
- * Returns number of pages pinned. This may be fewer than the number
- * requested. If nr_pages is 0 or negative, returns 0. If no pages
- * were pinned, returns -errno.
- */
-int get_user_pages_fast(unsigned long start, int nr_pages,
-                       unsigned int gup_flags, struct page **pages)
-{
-       struct mm_struct *mm = current->mm;
-       unsigned long addr, len, end;
-       unsigned long next;
-       pgd_t *pgdp;
-       int nr = 0;
-
-       start &= PAGE_MASK;
-       addr = start;
-       len = (unsigned long) nr_pages << PAGE_SHIFT;
-
-       end = start + len;
-       if (end < start)
-               goto slow_irqon;
-
-       local_irq_disable();
-       pgdp = pgd_offset(mm, addr);
-       do {
-               pgd_t pgd = *pgdp;
-
-               next = pgd_addr_end(addr, end);
-               if (pgd_none(pgd))
-                       goto slow;
-               if (!gup_pud_range(pgd, addr, next, gup_flags & FOLL_WRITE,
-                                  pages, &nr))
-                       goto slow;
-       } while (pgdp++, addr = next, addr != end);
-       local_irq_enable();
-
-       VM_BUG_ON(nr != (end - start) >> PAGE_SHIFT);
-       return nr;
-
-       {
-               int ret;
-
-slow:
-               local_irq_enable();
-slow_irqon:
-               /* Try to get the remaining pages with get_user_pages */
-               start += nr << PAGE_SHIFT;
-               pages += nr;
-
-               ret = get_user_pages_unlocked(start,
-                       (end - start) >> PAGE_SHIFT, pages,
-                       gup_flags);
-
-               /* Have to be a bit careful with return values */
-               if (nr > 0) {
-                       if (ret < 0)
-                               ret = nr;
-                       else
-                               ret += nr;
-               }
-
-               return ret;
-       }
-}