diff options
Diffstat (limited to 'arch/powerpc/mm')
| -rw-r--r-- | arch/powerpc/mm/hash_native_64.c | 40 | ||||
| -rw-r--r-- | arch/powerpc/mm/hugepage-hash64.c | 88 | ||||
| -rw-r--r-- | arch/powerpc/mm/numa.c | 13 | ||||
| -rw-r--r-- | arch/powerpc/mm/pgtable_64.c | 44 | ||||
| -rw-r--r-- | arch/powerpc/mm/tlb_hash64.c | 6 | ||||
| -rw-r--r-- | arch/powerpc/mm/tlb_nohash.c | 111 |
6 files changed, 198 insertions, 104 deletions
diff --git a/arch/powerpc/mm/hash_native_64.c b/arch/powerpc/mm/hash_native_64.c index cf1d325eae8b..afc0a8295f84 100644 --- a/arch/powerpc/mm/hash_native_64.c +++ b/arch/powerpc/mm/hash_native_64.c @@ -412,18 +412,18 @@ static void native_hpte_invalidate(unsigned long slot, unsigned long vpn, local_irq_restore(flags); } -static void native_hugepage_invalidate(struct mm_struct *mm, +static void native_hugepage_invalidate(unsigned long vsid, + unsigned long addr, unsigned char *hpte_slot_array, - unsigned long addr, int psize) + int psize, int ssize) { - int ssize = 0, i; - int lock_tlbie; + int i; struct hash_pte *hptep; int actual_psize = MMU_PAGE_16M; unsigned int max_hpte_count, valid; unsigned long flags, s_addr = addr; unsigned long hpte_v, want_v, shift; - unsigned long hidx, vpn = 0, vsid, hash, slot; + unsigned long hidx, vpn = 0, hash, slot; shift = mmu_psize_defs[psize].shift; max_hpte_count = 1U << (PMD_SHIFT - shift); @@ -437,15 +437,6 @@ static void native_hugepage_invalidate(struct mm_struct *mm, /* get the vpn */ addr = s_addr + (i * (1ul << shift)); - if (!is_kernel_addr(addr)) { - ssize = user_segment_size(addr); - vsid = get_vsid(mm->context.id, addr, ssize); - WARN_ON(vsid == 0); - } else { - vsid = get_kernel_vsid(addr, mmu_kernel_ssize); - ssize = mmu_kernel_ssize; - } - vpn = hpt_vpn(addr, vsid, ssize); hash = hpt_hash(vpn, shift, ssize); if (hidx & _PTEIDX_SECONDARY) @@ -465,22 +456,13 @@ static void native_hugepage_invalidate(struct mm_struct *mm, else /* Invalidate the hpte. NOTE: this also unlocks it */ hptep->v = 0; + /* + * We need to do tlb invalidate for all the address, tlbie + * instruction compares entry_VA in tlb with the VA specified + * here + */ + tlbie(vpn, psize, actual_psize, ssize, 0); } - /* - * Since this is a hugepage, we just need a single tlbie. - * use the last vpn. - */ - lock_tlbie = !mmu_has_feature(MMU_FTR_LOCKLESS_TLBIE); - if (lock_tlbie) - raw_spin_lock(&native_tlbie_lock); - - asm volatile("ptesync":::"memory"); - __tlbie(vpn, psize, actual_psize, ssize); - asm volatile("eieio; tlbsync; ptesync":::"memory"); - - if (lock_tlbie) - raw_spin_unlock(&native_tlbie_lock); - local_irq_restore(flags); } diff --git a/arch/powerpc/mm/hugepage-hash64.c b/arch/powerpc/mm/hugepage-hash64.c index 826893fcb3a7..5f5e6328c21c 100644 --- a/arch/powerpc/mm/hugepage-hash64.c +++ b/arch/powerpc/mm/hugepage-hash64.c @@ -18,6 +18,57 @@ #include <linux/mm.h> #include <asm/machdep.h> +static void invalidate_old_hpte(unsigned long vsid, unsigned long addr, + pmd_t *pmdp, unsigned int psize, int ssize) +{ + int i, max_hpte_count, valid; + unsigned long s_addr; + unsigned char *hpte_slot_array; + unsigned long hidx, shift, vpn, hash, slot; + + s_addr = addr & HPAGE_PMD_MASK; + hpte_slot_array = get_hpte_slot_array(pmdp); + /* + * IF we try to do a HUGE PTE update after a withdraw is done. + * we will find the below NULL. This happens when we do + * split_huge_page_pmd + */ + if (!hpte_slot_array) + return; + + if (ppc_md.hugepage_invalidate) + return ppc_md.hugepage_invalidate(vsid, s_addr, hpte_slot_array, + psize, ssize); + /* + * No bluk hpte removal support, invalidate each entry + */ + shift = mmu_psize_defs[psize].shift; + max_hpte_count = HPAGE_PMD_SIZE >> shift; + for (i = 0; i < max_hpte_count; i++) { + /* + * 8 bits per each hpte entries + * 000| [ secondary group (one bit) | hidx (3 bits) | valid bit] + */ + valid = hpte_valid(hpte_slot_array, i); + if (!valid) + continue; + hidx = hpte_hash_index(hpte_slot_array, i); + + /* get the vpn */ + addr = s_addr + (i * (1ul << shift)); + vpn = hpt_vpn(addr, vsid, ssize); + hash = hpt_hash(vpn, shift, ssize); + if (hidx & _PTEIDX_SECONDARY) + hash = ~hash; + + slot = (hash & htab_hash_mask) * HPTES_PER_GROUP; + slot += hidx & _PTEIDX_GROUP_IX; + ppc_md.hpte_invalidate(slot, vpn, psize, + MMU_PAGE_16M, ssize, 0); + } +} + + int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid, pmd_t *pmdp, unsigned long trap, int local, int ssize, unsigned int psize) @@ -33,7 +84,9 @@ int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid, * atomically mark the linux large page PMD busy and dirty */ do { - old_pmd = pmd_val(*pmdp); + pmd_t pmd = ACCESS_ONCE(*pmdp); + + old_pmd = pmd_val(pmd); /* If PMD busy, retry the access */ if (unlikely(old_pmd & _PAGE_BUSY)) return 0; @@ -85,6 +138,15 @@ int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid, vpn = hpt_vpn(ea, vsid, ssize); hash = hpt_hash(vpn, shift, ssize); hpte_slot_array = get_hpte_slot_array(pmdp); + if (psize == MMU_PAGE_4K) { + /* + * invalidate the old hpte entry if we have that mapped via 64K + * base page size. This is because demote_segment won't flush + * hash page table entries. + */ + if ((old_pmd & _PAGE_HASHPTE) && !(old_pmd & _PAGE_COMBO)) + invalidate_old_hpte(vsid, ea, pmdp, MMU_PAGE_64K, ssize); + } valid = hpte_valid(hpte_slot_array, index); if (valid) { @@ -107,11 +169,8 @@ int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid, * safely update this here. */ valid = 0; - new_pmd &= ~_PAGE_HPTEFLAGS; hpte_slot_array[index] = 0; - } else - /* clear the busy bits and set the hash pte bits */ - new_pmd = (new_pmd & ~_PAGE_HPTEFLAGS) | _PAGE_HASHPTE; + } } if (!valid) { @@ -119,11 +178,7 @@ int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid, /* insert new entry */ pa = pmd_pfn(__pmd(old_pmd)) << PAGE_SHIFT; -repeat: - hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL; - - /* clear the busy bits and set the hash pte bits */ - new_pmd = (new_pmd & ~_PAGE_HPTEFLAGS) | _PAGE_HASHPTE; + new_pmd |= _PAGE_HASHPTE; /* Add in WIMG bits */ rflags |= (new_pmd & (_PAGE_WRITETHRU | _PAGE_NO_CACHE | @@ -132,6 +187,8 @@ repeat: * enable the memory coherence always */ rflags |= HPTE_R_M; +repeat: + hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL; /* Insert into the hash table, primary slot */ slot = ppc_md.hpte_insert(hpte_group, vpn, pa, rflags, 0, @@ -172,8 +229,17 @@ repeat: mark_hpte_slot_valid(hpte_slot_array, index, slot); } /* - * No need to use ldarx/stdcx here + * Mark the pte with _PAGE_COMBO, if we are trying to hash it with + * base page size 4k. + */ + if (psize == MMU_PAGE_4K) + new_pmd |= _PAGE_COMBO; + /* + * The hpte valid is stored in the pgtable whose address is in the + * second half of the PMD. Order this against clearing of the busy bit in + * huge pmd. */ + smp_wmb(); *pmdp = __pmd(new_pmd & ~_PAGE_BUSY); return 0; } diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c index d3e9a78eaed3..d7737a542fd7 100644 --- a/arch/powerpc/mm/numa.c +++ b/arch/powerpc/mm/numa.c @@ -1049,7 +1049,7 @@ static void __init mark_reserved_regions_for_nid(int nid) void __init do_init_bootmem(void) { - int nid; + int nid, cpu; min_low_pfn = 0; max_low_pfn = memblock_end_of_DRAM() >> PAGE_SHIFT; @@ -1122,8 +1122,15 @@ void __init do_init_bootmem(void) reset_numa_cpu_lookup_table(); register_cpu_notifier(&ppc64_numa_nb); - cpu_numa_callback(&ppc64_numa_nb, CPU_UP_PREPARE, - (void *)(unsigned long)boot_cpuid); + /* + * We need the numa_cpu_lookup_table to be accurate for all CPUs, + * even before we online them, so that we can use cpu_to_{node,mem} + * early in boot, cf. smp_prepare_cpus(). + */ + for_each_possible_cpu(cpu) { + cpu_numa_callback(&ppc64_numa_nb, CPU_UP_PREPARE, + (void *)(unsigned long)cpu); + } } void __init paging_init(void) diff --git a/arch/powerpc/mm/pgtable_64.c b/arch/powerpc/mm/pgtable_64.c index 3b3c4d34c7a0..c8d709ab489d 100644 --- a/arch/powerpc/mm/pgtable_64.c +++ b/arch/powerpc/mm/pgtable_64.c @@ -54,6 +54,9 @@ #include "mmu_decl.h" +#define CREATE_TRACE_POINTS +#include <trace/events/thp.h> + /* Some sanity checking */ #if TASK_SIZE_USER64 > PGTABLE_RANGE #error TASK_SIZE_USER64 exceeds pagetable range @@ -537,8 +540,9 @@ unsigned long pmd_hugepage_update(struct mm_struct *mm, unsigned long addr, old = pmd_val(*pmdp); *pmdp = __pmd((old & ~clr) | set); #endif + trace_hugepage_update(addr, old, clr, set); if (old & _PAGE_HASHPTE) - hpte_do_hugepage_flush(mm, addr, pmdp); + hpte_do_hugepage_flush(mm, addr, pmdp, old); return old; } @@ -642,10 +646,11 @@ void pmdp_splitting_flush(struct vm_area_struct *vma, * If we didn't had the splitting flag set, go and flush the * HPTE entries. */ + trace_hugepage_splitting(address, old); if (!(old & _PAGE_SPLITTING)) { /* We need to flush the hpte */ if (old & _PAGE_HASHPTE) - hpte_do_hugepage_flush(vma->vm_mm, address, pmdp); + hpte_do_hugepage_flush(vma->vm_mm, address, pmdp, old); } /* * This ensures that generic code that rely on IRQ disabling @@ -709,6 +714,7 @@ void set_pmd_at(struct mm_struct *mm, unsigned long addr, assert_spin_locked(&mm->page_table_lock); WARN_ON(!pmd_trans_huge(pmd)); #endif + trace_hugepage_set_pmd(addr, pmd); return set_pte_at(mm, addr, pmdp_ptep(pmdp), pmd_pte(pmd)); } @@ -723,7 +729,7 @@ void pmdp_invalidate(struct vm_area_struct *vma, unsigned long address, * neesd to be flushed. */ void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr, - pmd_t *pmdp) + pmd_t *pmdp, unsigned long old_pmd) { int ssize, i; unsigned long s_addr; @@ -745,12 +751,29 @@ void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr, if (!hpte_slot_array) return; - /* get the base page size */ + /* get the base page size,vsid and segment size */ +#ifdef CONFIG_DEBUG_VM psize = get_slice_psize(mm, s_addr); + BUG_ON(psize == MMU_PAGE_16M); +#endif + if (old_pmd & _PAGE_COMBO) + psize = MMU_PAGE_4K; + else + psize = MMU_PAGE_64K; + + if (!is_kernel_addr(s_addr)) { + ssize = user_segment_size(s_addr); + vsid = get_vsid(mm->context.id, s_addr, ssize); + WARN_ON(vsid == 0); + } else { + vsid = get_kernel_vsid(s_addr, mmu_kernel_ssize); + ssize = mmu_kernel_ssize; + } if (ppc_md.hugepage_invalidate) - return ppc_md.hugepage_invalidate(mm, hpte_slot_array, - s_addr, psize); + return ppc_md.hugepage_invalidate(vsid, s_addr, + hpte_slot_array, + psize, ssize); /* * No bluk hpte removal support, invalidate each entry */ @@ -768,15 +791,6 @@ void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr, /* get the vpn */ addr = s_addr + (i * (1ul << shift)); - if (!is_kernel_addr(addr)) { - ssize = user_segment_size(addr); - vsid = get_vsid(mm->context.id, addr, ssize); - WARN_ON(vsid == 0); - } else { - vsid = get_kernel_vsid(addr, mmu_kernel_ssize); - ssize = mmu_kernel_ssize; - } - vpn = hpt_vpn(addr, vsid, ssize); hash = hpt_hash(vpn, shift, ssize); if (hidx & _PTEIDX_SECONDARY) diff --git a/arch/powerpc/mm/tlb_hash64.c b/arch/powerpc/mm/tlb_hash64.c index c99f6510a0b2..d2a94b85dbc2 100644 --- a/arch/powerpc/mm/tlb_hash64.c +++ b/arch/powerpc/mm/tlb_hash64.c @@ -30,6 +30,8 @@ #include <asm/tlb.h> #include <asm/bug.h> +#include <trace/events/thp.h> + DEFINE_PER_CPU(struct ppc64_tlb_batch, ppc64_tlb_batch); /* @@ -213,10 +215,12 @@ void __flush_hash_table_range(struct mm_struct *mm, unsigned long start, if (ptep == NULL) continue; pte = pte_val(*ptep); + if (hugepage_shift) + trace_hugepage_invalidate(start, pte_val(pte)); if (!(pte & _PAGE_HASHPTE)) continue; if (unlikely(hugepage_shift && pmd_trans_huge(*(pmd_t *)pte))) - hpte_do_hugepage_flush(mm, start, (pmd_t *)pte); + hpte_do_hugepage_flush(mm, start, (pmd_t *)ptep, pte); else hpte_need_flush(mm, start, ptep, pte, 0); } diff --git a/arch/powerpc/mm/tlb_nohash.c b/arch/powerpc/mm/tlb_nohash.c index 92cb18d52ea8..f38ea4df6a85 100644 --- a/arch/powerpc/mm/tlb_nohash.c +++ b/arch/powerpc/mm/tlb_nohash.c @@ -581,42 +581,10 @@ static void setup_mmu_htw(void) /* * Early initialization of the MMU TLB code */ -static void __early_init_mmu(int boot_cpu) +static void early_init_this_mmu(void) { unsigned int mas4; - /* XXX This will have to be decided at runtime, but right - * now our boot and TLB miss code hard wires it. Ideally - * we should find out a suitable page size and patch the - * TLB miss code (either that or use the PACA to store - * the value we want) - */ - mmu_linear_psize = MMU_PAGE_1G; - - /* XXX This should be decided at runtime based on supported - * page sizes in the TLB, but for now let's assume 16M is - * always there and a good fit (which it probably is) - * - * Freescale booke only supports 4K pages in TLB0, so use that. - */ - if (mmu_has_feature(MMU_FTR_TYPE_FSL_E)) - mmu_vmemmap_psize = MMU_PAGE_4K; - else - mmu_vmemmap_psize = MMU_PAGE_16M; - - /* XXX This code only checks for TLB 0 capabilities and doesn't - * check what page size combos are supported by the HW. It - * also doesn't handle the case where a separate array holds - * the IND entries from the array loaded by the PT. - */ - if (boot_cpu) { - /* Look for supported page sizes */ - setup_page_sizes(); - - /* Look for HW tablewalk support */ - setup_mmu_htw(); - } - /* Set MAS4 based on page table setting */ mas4 = 0x4 << MAS4_WIMGED_SHIFT; @@ -650,11 +618,6 @@ static void __early_init_mmu(int boot_cpu) } mtspr(SPRN_MAS4, mas4); - /* Set the global containing the top of the linear mapping - * for use by the TLB miss code - */ - linear_map_top = memblock_end_of_DRAM(); - #ifdef CONFIG_PPC_FSL_BOOK3E if (mmu_has_feature(MMU_FTR_TYPE_FSL_E)) { unsigned int num_cams; @@ -662,10 +625,49 @@ static void __early_init_mmu(int boot_cpu) /* use a quarter of the TLBCAM for bolted linear map */ num_cams = (mfspr(SPRN_TLB1CFG) & TLBnCFG_N_ENTRY) / 4; linear_map_top = map_mem_in_cams(linear_map_top, num_cams); + } +#endif - /* limit memory so we dont have linear faults */ - memblock_enforce_memory_limit(linear_map_top); + /* A sync won't hurt us after mucking around with + * the MMU configuration + */ + mb(); +} +static void __init early_init_mmu_global(void) +{ + /* XXX This will have to be decided at runtime, but right + * now our boot and TLB miss code hard wires it. Ideally + * we should find out a suitable page size and patch the + * TLB miss code (either that or use the PACA to store + * the value we want) + */ + mmu_linear_psize = MMU_PAGE_1G; + + /* XXX This should be decided at runtime based on supported + * page sizes in the TLB, but for now let's assume 16M is + * always there and a good fit (which it probably is) + * + * Freescale booke only supports 4K pages in TLB0, so use that. + */ + if (mmu_has_feature(MMU_FTR_TYPE_FSL_E)) + mmu_vmemmap_psize = MMU_PAGE_4K; + else + mmu_vmemmap_psize = MMU_PAGE_16M; + + /* XXX This code only checks for TLB 0 capabilities and doesn't + * check what page size combos are supported by the HW. It + * also doesn't handle the case where a separate array holds + * the IND entries from the array loaded by the PT. + */ + /* Look for supported page sizes */ + setup_page_sizes(); + + /* Look for HW tablewalk support */ + setup_mmu_htw(); + +#ifdef CONFIG_PPC_FSL_BOOK3E + if (mmu_has_feature(MMU_FTR_TYPE_FSL_E)) { if (book3e_htw_mode == PPC_HTW_NONE) { extlb_level_exc = EX_TLB_SIZE; patch_exception(0x1c0, exc_data_tlb_miss_bolted_book3e); @@ -675,22 +677,41 @@ static void __early_init_mmu(int boot_cpu) } #endif - /* A sync won't hurt us after mucking around with - * the MMU configuration + /* Set the global containing the top of the linear mapping + * for use by the TLB miss code */ - mb(); + linear_map_top = memblock_end_of_DRAM(); +} + +static void __init early_mmu_set_memory_limit(void) +{ +#ifdef CONFIG_PPC_FSL_BOOK3E + if (mmu_has_feature(MMU_FTR_TYPE_FSL_E)) { + /* + * Limit memory so we dont have linear faults. + * Unlike memblock_set_current_limit, which limits + * memory available during early boot, this permanently + * reduces the memory available to Linux. We need to + * do this because highmem is not supported on 64-bit. + */ + memblock_enforce_memory_limit(linear_map_top); + } +#endif memblock_set_current_limit(linear_map_top); } +/* boot cpu only */ void __init early_init_mmu(void) { - __early_init_mmu(1); + early_init_mmu_global(); + early_init_this_mmu(); + early_mmu_set_memory_limit(); } void early_init_mmu_secondary(void) { - __early_init_mmu(0); + early_init_this_mmu(); } void setup_initial_memory_limit(phys_addr_t first_memblock_base, |