From 6e6e41879e07daccb967bc75a31f29689354d11b Mon Sep 17 00:00:00 2001 From: Sam Ravnborg Date: Fri, 22 Apr 2016 19:41:06 +0200 Subject: sparc32: fix build with STRICT_MM_TYPECHECKS Based on recent thread on linux-arch (some weeks ago) I decided to check how much work was required to build sparc32 with STRICT_MM_TYPECHECKS enabled. The resulting binary (checked srmmu.o) was to my suprise smaller with STRICT_MM_TYPECHECKS defined, than without. As I have no working gear to test sparc32 bits at for the moment, I did not enable STRICT_MM_TYPECHECKS - but was tempeted to do so. Signed-off-by: Sam Ravnborg Cc: Arnd Bergmann Signed-off-by: David S. Miller --- arch/sparc/mm/io-unit.c | 4 ++-- arch/sparc/mm/srmmu.c | 15 ++++++++++----- 2 files changed, 12 insertions(+), 7 deletions(-) (limited to 'arch/sparc/mm') diff --git a/arch/sparc/mm/io-unit.c b/arch/sparc/mm/io-unit.c index f311bf219016..338fb71535de 100644 --- a/arch/sparc/mm/io-unit.c +++ b/arch/sparc/mm/io-unit.c @@ -133,7 +133,7 @@ nexti: scan = find_next_zero_bit(iounit->bmap, limit, scan); vaddr = IOUNIT_DMA_BASE + (scan << PAGE_SHIFT) + (vaddr & ~PAGE_MASK); for (k = 0; k < npages; k++, iopte = __iopte(iopte_val(iopte) + 0x100), scan++) { set_bit(scan, iounit->bmap); - sbus_writel(iopte, &iounit->page_table[scan]); + sbus_writel(iopte_val(iopte), &iounit->page_table[scan]); } IOD(("%08lx\n", vaddr)); return vaddr; @@ -228,7 +228,7 @@ static int iounit_map_dma_area(struct device *dev, dma_addr_t *pba, unsigned lon i = ((addr - IOUNIT_DMA_BASE) >> PAGE_SHIFT); iopte = iounit->page_table + i; - sbus_writel(MKIOPTE(__pa(page)), iopte); + sbus_writel(iopte_val(MKIOPTE(__pa(page))), iopte); } addr += PAGE_SIZE; va += PAGE_SIZE; diff --git a/arch/sparc/mm/srmmu.c b/arch/sparc/mm/srmmu.c index 5cbc96d801ff..3b1c047e7cf0 100644 --- a/arch/sparc/mm/srmmu.c +++ b/arch/sparc/mm/srmmu.c @@ -107,7 +107,12 @@ static inline int srmmu_pmd_none(pmd_t pmd) /* XXX should we hyper_flush_whole_icache here - Anton */ static inline void srmmu_ctxd_set(ctxd_t *ctxp, pgd_t *pgdp) -{ set_pte((pte_t *)ctxp, (SRMMU_ET_PTD | (__nocache_pa((unsigned long) pgdp) >> 4))); } +{ + pte_t pte; + + pte = __pte((SRMMU_ET_PTD | (__nocache_pa(pgdp) >> 4))); + set_pte((pte_t *)ctxp, pte); +} void pmd_set(pmd_t *pmdp, pte_t *ptep) { @@ -116,8 +121,8 @@ void pmd_set(pmd_t *pmdp, pte_t *ptep) ptp = __nocache_pa((unsigned long) ptep) >> 4; for (i = 0; i < PTRS_PER_PTE/SRMMU_REAL_PTRS_PER_PTE; i++) { - set_pte((pte_t *)&pmdp->pmdv[i], SRMMU_ET_PTD | ptp); - ptp += (SRMMU_REAL_PTRS_PER_PTE*sizeof(pte_t) >> 4); + set_pte((pte_t *)&pmdp->pmdv[i], __pte(SRMMU_ET_PTD | ptp)); + ptp += (SRMMU_REAL_PTRS_PER_PTE * sizeof(pte_t) >> 4); } } @@ -128,8 +133,8 @@ void pmd_populate(struct mm_struct *mm, pmd_t *pmdp, struct page *ptep) ptp = page_to_pfn(ptep) << (PAGE_SHIFT-4); /* watch for overflow */ for (i = 0; i < PTRS_PER_PTE/SRMMU_REAL_PTRS_PER_PTE; i++) { - set_pte((pte_t *)&pmdp->pmdv[i], SRMMU_ET_PTD | ptp); - ptp += (SRMMU_REAL_PTRS_PER_PTE*sizeof(pte_t) >> 4); + set_pte((pte_t *)&pmdp->pmdv[i], __pte(SRMMU_ET_PTD | ptp)); + ptp += (SRMMU_REAL_PTRS_PER_PTE * sizeof(pte_t) >> 4); } } -- cgit v1.2.3 From 6b1cabe8992441afb542503781f11b9cce91e285 Mon Sep 17 00:00:00 2001 From: Sam Ravnborg Date: Fri, 22 Apr 2016 19:41:07 +0200 Subject: sparc32: drop superfluous cast in calls to __nocache_pa() Signed-off-by: Sam Ravnborg Signed-off-by: David S. Miller --- arch/sparc/mm/srmmu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/sparc/mm') diff --git a/arch/sparc/mm/srmmu.c b/arch/sparc/mm/srmmu.c index 3b1c047e7cf0..c7f2a5295b3a 100644 --- a/arch/sparc/mm/srmmu.c +++ b/arch/sparc/mm/srmmu.c @@ -119,7 +119,7 @@ void pmd_set(pmd_t *pmdp, pte_t *ptep) unsigned long ptp; /* Physical address, shifted right by 4 */ int i; - ptp = __nocache_pa((unsigned long) ptep) >> 4; + ptp = __nocache_pa(ptep) >> 4; for (i = 0; i < PTRS_PER_PTE/SRMMU_REAL_PTRS_PER_PTE; i++) { set_pte((pte_t *)&pmdp->pmdv[i], __pte(SRMMU_ET_PTD | ptp)); ptp += (SRMMU_REAL_PTRS_PER_PTE * sizeof(pte_t) >> 4); @@ -916,7 +916,7 @@ void __init srmmu_paging_init(void) /* ctx table has to be physically aligned to its size */ srmmu_context_table = __srmmu_get_nocache(num_contexts * sizeof(ctxd_t), num_contexts * sizeof(ctxd_t)); - srmmu_ctx_table_phys = (ctxd_t *)__nocache_pa((unsigned long)srmmu_context_table); + srmmu_ctx_table_phys = (ctxd_t *)__nocache_pa(srmmu_context_table); for (i = 0; i < num_contexts; i++) srmmu_ctxd_set((ctxd_t *)__nocache_fix(&srmmu_context_table[i]), srmmu_swapper_pg_dir); -- cgit v1.2.3 From 24e49ee3d76b70853a96520e46b8837e5eae65b2 Mon Sep 17 00:00:00 2001 From: Nitin Gupta Date: Wed, 30 Mar 2016 11:17:13 -0700 Subject: sparc64: Reduce TLB flushes during hugepte changes During hugepage map/unmap, TSB and TLB flushes are currently issued at every PAGE_SIZE'd boundary which is unnecessary. We now issue the flush at REAL_HPAGE_SIZE boundaries only. Without this patch workloads which unmap a large hugepage backed VMA region get CPU lockups due to excessive TLB flush calls. Orabug: 22365539, 22643230, 22995196 Signed-off-by: Nitin Gupta Signed-off-by: David S. Miller --- arch/sparc/mm/hugetlbpage.c | 33 ++++++++++++++++++++++++++++----- arch/sparc/mm/init_64.c | 12 ------------ arch/sparc/mm/tlb.c | 25 ++++++++++++++++++------- arch/sparc/mm/tsb.c | 32 +++++++++++++++++--------------- 4 files changed, 63 insertions(+), 39 deletions(-) (limited to 'arch/sparc/mm') diff --git a/arch/sparc/mm/hugetlbpage.c b/arch/sparc/mm/hugetlbpage.c index 4977800e9770..ba52e6466a82 100644 --- a/arch/sparc/mm/hugetlbpage.c +++ b/arch/sparc/mm/hugetlbpage.c @@ -176,17 +176,31 @@ void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t entry) { int i; + pte_t orig[2]; + unsigned long nptes; if (!pte_present(*ptep) && pte_present(entry)) mm->context.huge_pte_count++; addr &= HPAGE_MASK; - for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) { - set_pte_at(mm, addr, ptep, entry); + + nptes = 1 << HUGETLB_PAGE_ORDER; + orig[0] = *ptep; + orig[1] = *(ptep + nptes / 2); + for (i = 0; i < nptes; i++) { + *ptep = entry; ptep++; addr += PAGE_SIZE; pte_val(entry) += PAGE_SIZE; } + + /* Issue TLB flush at REAL_HPAGE_SIZE boundaries */ + addr -= REAL_HPAGE_SIZE; + ptep -= nptes / 2; + maybe_tlb_batch_add(mm, addr, ptep, orig[1], 0); + addr -= REAL_HPAGE_SIZE; + ptep -= nptes / 2; + maybe_tlb_batch_add(mm, addr, ptep, orig[0], 0); } pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, @@ -194,19 +208,28 @@ pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, { pte_t entry; int i; + unsigned long nptes; entry = *ptep; if (pte_present(entry)) mm->context.huge_pte_count--; addr &= HPAGE_MASK; - - for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) { - pte_clear(mm, addr, ptep); + nptes = 1 << HUGETLB_PAGE_ORDER; + for (i = 0; i < nptes; i++) { + *ptep = __pte(0UL); addr += PAGE_SIZE; ptep++; } + /* Issue TLB flush at REAL_HPAGE_SIZE boundaries */ + addr -= REAL_HPAGE_SIZE; + ptep -= nptes / 2; + maybe_tlb_batch_add(mm, addr, ptep, entry, 0); + addr -= REAL_HPAGE_SIZE; + ptep -= nptes / 2; + maybe_tlb_batch_add(mm, addr, ptep, entry, 0); + return entry; } diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c index 09e838801e39..652683cb4b4b 100644 --- a/arch/sparc/mm/init_64.c +++ b/arch/sparc/mm/init_64.c @@ -324,18 +324,6 @@ static void __update_mmu_tsb_insert(struct mm_struct *mm, unsigned long tsb_inde tsb_insert(tsb, tag, tte); } -#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE) -static inline bool is_hugetlb_pte(pte_t pte) -{ - if ((tlb_type == hypervisor && - (pte_val(pte) & _PAGE_SZALL_4V) == _PAGE_SZHUGE_4V) || - (tlb_type != hypervisor && - (pte_val(pte) & _PAGE_SZALL_4U) == _PAGE_SZHUGE_4U)) - return true; - return false; -} -#endif - void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t *ptep) { struct mm_struct *mm; diff --git a/arch/sparc/mm/tlb.c b/arch/sparc/mm/tlb.c index 9df2190c097e..f81cd9736700 100644 --- a/arch/sparc/mm/tlb.c +++ b/arch/sparc/mm/tlb.c @@ -67,7 +67,7 @@ void arch_leave_lazy_mmu_mode(void) } static void tlb_batch_add_one(struct mm_struct *mm, unsigned long vaddr, - bool exec) + bool exec, bool huge) { struct tlb_batch *tb = &get_cpu_var(tlb_batch); unsigned long nr; @@ -84,13 +84,21 @@ static void tlb_batch_add_one(struct mm_struct *mm, unsigned long vaddr, } if (!tb->active) { - flush_tsb_user_page(mm, vaddr); + flush_tsb_user_page(mm, vaddr, huge); global_flush_tlb_page(mm, vaddr); goto out; } - if (nr == 0) + if (nr == 0) { tb->mm = mm; + tb->huge = huge; + } + + if (tb->huge != huge) { + flush_tlb_pending(); + tb->huge = huge; + nr = 0; + } tb->vaddrs[nr] = vaddr; tb->tlb_nr = ++nr; @@ -104,6 +112,8 @@ out: void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig, int fullmm) { + bool huge = is_hugetlb_pte(orig); + if (tlb_type != hypervisor && pte_dirty(orig)) { unsigned long paddr, pfn = pte_pfn(orig); @@ -129,7 +139,7 @@ void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, no_cache_flush: if (!fullmm) - tlb_batch_add_one(mm, vaddr, pte_exec(orig)); + tlb_batch_add_one(mm, vaddr, pte_exec(orig), huge); } #ifdef CONFIG_TRANSPARENT_HUGEPAGE @@ -145,7 +155,7 @@ static void tlb_batch_pmd_scan(struct mm_struct *mm, unsigned long vaddr, if (pte_val(*pte) & _PAGE_VALID) { bool exec = pte_exec(*pte); - tlb_batch_add_one(mm, vaddr, exec); + tlb_batch_add_one(mm, vaddr, exec, false); } pte++; vaddr += PAGE_SIZE; @@ -185,8 +195,9 @@ void set_pmd_at(struct mm_struct *mm, unsigned long addr, pte_t orig_pte = __pte(pmd_val(orig)); bool exec = pte_exec(orig_pte); - tlb_batch_add_one(mm, addr, exec); - tlb_batch_add_one(mm, addr + REAL_HPAGE_SIZE, exec); + tlb_batch_add_one(mm, addr, exec, true); + tlb_batch_add_one(mm, addr + REAL_HPAGE_SIZE, exec, + true); } else { tlb_batch_pmd_scan(mm, addr, orig); } diff --git a/arch/sparc/mm/tsb.c b/arch/sparc/mm/tsb.c index a06576683c38..a0604a493a36 100644 --- a/arch/sparc/mm/tsb.c +++ b/arch/sparc/mm/tsb.c @@ -76,14 +76,15 @@ void flush_tsb_user(struct tlb_batch *tb) spin_lock_irqsave(&mm->context.lock, flags); - base = (unsigned long) mm->context.tsb_block[MM_TSB_BASE].tsb; - nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries; - if (tlb_type == cheetah_plus || tlb_type == hypervisor) - base = __pa(base); - __flush_tsb_one(tb, PAGE_SHIFT, base, nentries); - + if (!tb->huge) { + base = (unsigned long) mm->context.tsb_block[MM_TSB_BASE].tsb; + nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries; + if (tlb_type == cheetah_plus || tlb_type == hypervisor) + base = __pa(base); + __flush_tsb_one(tb, PAGE_SHIFT, base, nentries); + } #if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE) - if (mm->context.tsb_block[MM_TSB_HUGE].tsb) { + if (tb->huge && mm->context.tsb_block[MM_TSB_HUGE].tsb) { base = (unsigned long) mm->context.tsb_block[MM_TSB_HUGE].tsb; nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries; if (tlb_type == cheetah_plus || tlb_type == hypervisor) @@ -94,20 +95,21 @@ void flush_tsb_user(struct tlb_batch *tb) spin_unlock_irqrestore(&mm->context.lock, flags); } -void flush_tsb_user_page(struct mm_struct *mm, unsigned long vaddr) +void flush_tsb_user_page(struct mm_struct *mm, unsigned long vaddr, bool huge) { unsigned long nentries, base, flags; spin_lock_irqsave(&mm->context.lock, flags); - base = (unsigned long) mm->context.tsb_block[MM_TSB_BASE].tsb; - nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries; - if (tlb_type == cheetah_plus || tlb_type == hypervisor) - base = __pa(base); - __flush_tsb_one_entry(base, vaddr, PAGE_SHIFT, nentries); - + if (!huge) { + base = (unsigned long) mm->context.tsb_block[MM_TSB_BASE].tsb; + nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries; + if (tlb_type == cheetah_plus || tlb_type == hypervisor) + base = __pa(base); + __flush_tsb_one_entry(base, vaddr, PAGE_SHIFT, nentries); + } #if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE) - if (mm->context.tsb_block[MM_TSB_HUGE].tsb) { + if (huge && mm->context.tsb_block[MM_TSB_HUGE].tsb) { base = (unsigned long) mm->context.tsb_block[MM_TSB_HUGE].tsb; nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries; if (tlb_type == cheetah_plus || tlb_type == hypervisor) -- cgit v1.2.3