From 3bbf7157ac66a88d94b291d4d5e2b2a9319a0f90 Mon Sep 17 00:00:00 2001 From: Catalin Marinas Date: Mon, 26 Jun 2017 14:27:36 +0100 Subject: arm64: Convert pte handling from inline asm to using (cmp)xchg With the support for hardware updates of the access and dirty states, the following pte handling functions had to be implemented using exclusives: __ptep_test_and_clear_young(), ptep_get_and_clear(), ptep_set_wrprotect() and ptep_set_access_flags(). To take advantage of the LSE atomic instructions and also make the code cleaner, convert these pte functions to use the more generic cmpxchg()/xchg(). Reviewed-by: Will Deacon Acked-by: Mark Rutland Acked-by: Steve Capper Signed-off-by: Catalin Marinas --- arch/arm64/include/asm/pgtable.h | 71 +++++++++++++++++++--------------------- arch/arm64/mm/fault.c | 24 +++++++------- 2 files changed, 44 insertions(+), 51 deletions(-) (limited to 'arch') diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h index 6eae342ced6b..9127688ae775 100644 --- a/arch/arm64/include/asm/pgtable.h +++ b/arch/arm64/include/asm/pgtable.h @@ -39,6 +39,7 @@ #ifndef __ASSEMBLY__ +#include #include #include @@ -173,6 +174,11 @@ static inline pte_t pte_clear_rdonly(pte_t pte) return clear_pte_bit(pte, __pgprot(PTE_RDONLY)); } +static inline pte_t pte_set_rdonly(pte_t pte) +{ + return set_pte_bit(pte, __pgprot(PTE_RDONLY)); +} + static inline pte_t pte_mkpresent(pte_t pte) { return set_pte_bit(pte, __pgprot(PTE_VALID)); @@ -593,20 +599,17 @@ static inline int pmdp_set_access_flags(struct vm_area_struct *vma, #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG static inline int __ptep_test_and_clear_young(pte_t *ptep) { - pteval_t pteval; - unsigned int tmp, res; + pte_t old_pte, pte; - asm volatile("// __ptep_test_and_clear_young\n" - " prfm pstl1strm, %2\n" - "1: ldxr %0, %2\n" - " ubfx %w3, %w0, %5, #1 // extract PTE_AF (young)\n" - " and %0, %0, %4 // clear PTE_AF\n" - " stxr %w1, %0, %2\n" - " cbnz %w1, 1b\n" - : "=&r" (pteval), "=&r" (tmp), "+Q" (pte_val(*ptep)), "=&r" (res) - : "L" (~PTE_AF), "I" (ilog2(PTE_AF))); + pte = READ_ONCE(*ptep); + do { + old_pte = pte; + pte = pte_mkold(pte); + pte_val(pte) = cmpxchg_relaxed(&pte_val(*ptep), + pte_val(old_pte), pte_val(pte)); + } while (pte_val(pte) != pte_val(old_pte)); - return res; + return pte_young(pte); } static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, @@ -630,17 +633,7 @@ static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma, static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long address, pte_t *ptep) { - pteval_t old_pteval; - unsigned int tmp; - - asm volatile("// ptep_get_and_clear\n" - " prfm pstl1strm, %2\n" - "1: ldxr %0, %2\n" - " stxr %w1, xzr, %2\n" - " cbnz %w1, 1b\n" - : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep))); - - return __pte(old_pteval); + return __pte(xchg_relaxed(&pte_val(*ptep), 0)); } #ifdef CONFIG_TRANSPARENT_HUGEPAGE @@ -659,21 +652,23 @@ static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm, #define __HAVE_ARCH_PTEP_SET_WRPROTECT static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long address, pte_t *ptep) { - pteval_t pteval; - unsigned long tmp; - - asm volatile("// ptep_set_wrprotect\n" - " prfm pstl1strm, %2\n" - "1: ldxr %0, %2\n" - " tst %0, %4 // check for hw dirty (!PTE_RDONLY)\n" - " csel %1, %3, xzr, eq // set PTE_DIRTY|PTE_RDONLY if dirty\n" - " orr %0, %0, %1 // if !dirty, PTE_RDONLY is already set\n" - " and %0, %0, %5 // clear PTE_WRITE/PTE_DBM\n" - " stxr %w1, %0, %2\n" - " cbnz %w1, 1b\n" - : "=&r" (pteval), "=&r" (tmp), "+Q" (pte_val(*ptep)) - : "r" (PTE_DIRTY|PTE_RDONLY), "L" (PTE_RDONLY), "L" (~PTE_WRITE) - : "cc"); + pte_t old_pte, pte; + + pte = READ_ONCE(*ptep); + do { + old_pte = pte; + /* + * If hardware-dirty (PTE_WRITE/DBM bit set and PTE_RDONLY + * clear), set the PTE_DIRTY and PTE_RDONLY bits. + */ + if (pte_hw_dirty(pte)) { + pte = pte_mkdirty(pte); + pte = pte_set_rdonly(pte); + } + pte = pte_wrprotect(pte); + pte_val(pte) = cmpxchg_relaxed(&pte_val(*ptep), + pte_val(old_pte), pte_val(pte)); + } while (pte_val(pte) != pte_val(old_pte)); } #ifdef CONFIG_TRANSPARENT_HUGEPAGE diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c index 52ee273afeec..430eaf82da49 100644 --- a/arch/arm64/mm/fault.c +++ b/arch/arm64/mm/fault.c @@ -34,6 +34,7 @@ #include #include +#include #include #include #include @@ -197,8 +198,7 @@ int ptep_set_access_flags(struct vm_area_struct *vma, unsigned long address, pte_t *ptep, pte_t entry, int dirty) { - pteval_t old_pteval; - unsigned int tmp; + pteval_t old_pteval, pteval; if (pte_same(*ptep, entry)) return 0; @@ -208,7 +208,7 @@ int ptep_set_access_flags(struct vm_area_struct *vma, /* set PTE_RDONLY if actual read-only or clean PTE */ if (!pte_write(entry) || !pte_sw_dirty(entry)) - pte_val(entry) |= PTE_RDONLY; + entry = pte_set_rdonly(entry); /* * Setting the flags must be done atomically to avoid racing with the @@ -217,16 +217,14 @@ int ptep_set_access_flags(struct vm_area_struct *vma, * (calculated as: a & b == ~(~a | ~b)). */ pte_val(entry) ^= PTE_RDONLY; - asm volatile("// ptep_set_access_flags\n" - " prfm pstl1strm, %2\n" - "1: ldxr %0, %2\n" - " eor %0, %0, %3 // negate PTE_RDONLY in *ptep\n" - " orr %0, %0, %4 // set flags\n" - " eor %0, %0, %3 // negate final PTE_RDONLY\n" - " stxr %w1, %0, %2\n" - " cbnz %w1, 1b\n" - : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep)) - : "L" (PTE_RDONLY), "r" (pte_val(entry))); + pteval = READ_ONCE(pte_val(*ptep)); + do { + old_pteval = pteval; + pteval ^= PTE_RDONLY; + pteval |= pte_val(entry); + pteval ^= PTE_RDONLY; + pteval = cmpxchg_relaxed(&pte_val(*ptep), old_pteval, pteval); + } while (pteval != old_pteval); flush_tlb_fix_spurious_fault(vma, address); return 1; -- cgit v1.2.3-59-g8ed1b