* [PATCH 1/5] MIPS: Add support files for hugeTLBfs.
2009-05-28 0:46 [PATCH 0/5] MIPS: Add hugeTLBfs support David Daney
@ 2009-05-28 0:47 ` David Daney
2009-05-28 0:47 ` [PATCH 2/5] MIPS: Add hugeTLBfs page defines David Daney
` (3 subsequent siblings)
4 siblings, 0 replies; 8+ messages in thread
From: David Daney @ 2009-05-28 0:47 UTC (permalink / raw)
To: linux-mips, ralf; +Cc: wli, David Daney
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
---
arch/mips/include/asm/hugetlb.h | 114 +++++++++++++++++++++++++++++++++++++++
arch/mips/mm/Makefile | 1 +
arch/mips/mm/hugetlbpage.c | 101 ++++++++++++++++++++++++++++++++++
3 files changed, 216 insertions(+), 0 deletions(-)
create mode 100644 arch/mips/include/asm/hugetlb.h
create mode 100644 arch/mips/mm/hugetlbpage.c
diff --git a/arch/mips/include/asm/hugetlb.h b/arch/mips/include/asm/hugetlb.h
new file mode 100644
index 0000000..f5e8560
--- /dev/null
+++ b/arch/mips/include/asm/hugetlb.h
@@ -0,0 +1,114 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2008, 2009 Cavium Networks, Inc.
+ */
+
+#ifndef __ASM_HUGETLB_H
+#define __ASM_HUGETLB_H
+
+#include <asm/page.h>
+
+
+static inline int is_hugepage_only_range(struct mm_struct *mm,
+ unsigned long addr,
+ unsigned long len)
+{
+ return 0;
+}
+
+static inline int prepare_hugepage_range(struct file *file,
+ unsigned long addr,
+ unsigned long len)
+{
+ unsigned long task_size = STACK_TOP;
+ struct hstate *h = hstate_file(file);
+
+ if (len & ~huge_page_mask(h))
+ return -EINVAL;
+ if (addr & ~huge_page_mask(h))
+ return -EINVAL;
+ if (len > task_size)
+ return -ENOMEM;
+ if (task_size - len < addr)
+ return -EINVAL;
+ return 0;
+}
+
+static inline void hugetlb_prefault_arch_hook(struct mm_struct *mm)
+{
+}
+
+static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
+ unsigned long addr,
+ unsigned long end,
+ unsigned long floor,
+ unsigned long ceiling)
+{
+ free_pgd_range(tlb, addr, end, floor, ceiling);
+}
+
+static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte)
+{
+ set_pte_at(mm, addr, ptep, pte);
+}
+
+static inline pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep)
+{
+ pte_t clear;
+ pte_t pte = *ptep;
+
+ pte_val(clear) = (unsigned long)invalid_pte_table;
+ set_pte_at(mm, addr, ptep, clear);
+ return pte;
+}
+
+static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep)
+{
+}
+
+static inline int huge_pte_none(pte_t pte)
+{
+ unsigned long val = pte_val(pte) & ~_PAGE_GLOBAL;
+ return !val || (val == (unsigned long)invalid_pte_table);
+}
+
+static inline pte_t huge_pte_wrprotect(pte_t pte)
+{
+ return pte_wrprotect(pte);
+}
+
+static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep)
+{
+ ptep_set_wrprotect(mm, addr, ptep);
+}
+
+static inline int huge_ptep_set_access_flags(struct vm_area_struct *vma,
+ unsigned long addr,
+ pte_t *ptep, pte_t pte,
+ int dirty)
+{
+ return ptep_set_access_flags(vma, addr, ptep, pte, dirty);
+}
+
+static inline pte_t huge_ptep_get(pte_t *ptep)
+{
+ return *ptep;
+}
+
+static inline int arch_prepare_hugepage(struct page *page)
+{
+ return 0;
+}
+
+static inline void arch_release_hugepage(struct page *page)
+{
+}
+
+#endif /* __ASM_HUGETLB_H */
diff --git a/arch/mips/mm/Makefile b/arch/mips/mm/Makefile
index d7ec955..f0e4355 100644
--- a/arch/mips/mm/Makefile
+++ b/arch/mips/mm/Makefile
@@ -8,6 +8,7 @@ obj-y += cache.o dma-default.o extable.o fault.o \
obj-$(CONFIG_32BIT) += ioremap.o pgtable-32.o
obj-$(CONFIG_64BIT) += pgtable-64.o
obj-$(CONFIG_HIGHMEM) += highmem.o
+obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
obj-$(CONFIG_CPU_LOONGSON2) += c-r4k.o cex-gen.o tlb-r4k.o
obj-$(CONFIG_CPU_MIPS32) += c-r4k.o cex-gen.o tlb-r4k.o
diff --git a/arch/mips/mm/hugetlbpage.c b/arch/mips/mm/hugetlbpage.c
new file mode 100644
index 0000000..471c09a
--- /dev/null
+++ b/arch/mips/mm/hugetlbpage.c
@@ -0,0 +1,101 @@
+/*
+ * MIPS Huge TLB Page Support for Kernel.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2002, Rohit Seth <rohit.seth@intel.com>
+ * Copyright 2005, Embedded Alley Solutions, Inc.
+ * Matt Porter <mporter@embeddedalley.com>
+ * Copyright (C) 2008, 2009 Cavium Networks, Inc.
+ */
+
+#include <linux/init.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/hugetlb.h>
+#include <linux/pagemap.h>
+#include <linux/smp_lock.h>
+#include <linux/slab.h>
+#include <linux/err.h>
+#include <linux/sysctl.h>
+#include <asm/mman.h>
+#include <asm/tlb.h>
+#include <asm/tlbflush.h>
+
+pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr,
+ unsigned long sz)
+{
+ pgd_t *pgd;
+ pud_t *pud;
+ pte_t *pte = NULL;
+
+ pgd = pgd_offset(mm, addr);
+ pud = pud_alloc(mm, pgd, addr);
+ if (pud)
+ pte = (pte_t *)pmd_alloc(mm, pud, addr);
+
+ return pte;
+}
+
+pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
+{
+ pgd_t *pgd;
+ pud_t *pud;
+ pmd_t *pmd = NULL;
+
+ pgd = pgd_offset(mm, addr);
+ if (pgd_present(*pgd)) {
+ pud = pud_offset(pgd, addr);
+ if (pud_present(*pud))
+ pmd = pmd_offset(pud, addr);
+ }
+ return (pte_t *) pmd;
+}
+
+int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
+{
+ return 0;
+}
+
+/*
+ * This function checks for proper alignment of input addr and len parameters.
+ */
+int is_aligned_hugepage_range(unsigned long addr, unsigned long len)
+{
+ if (len & ~HPAGE_MASK)
+ return -EINVAL;
+ if (addr & ~HPAGE_MASK)
+ return -EINVAL;
+ return 0;
+}
+
+struct page *
+follow_huge_addr(struct mm_struct *mm, unsigned long address, int write)
+{
+ return ERR_PTR(-EINVAL);
+}
+
+int pmd_huge(pmd_t pmd)
+{
+ return (pmd_val(pmd) & _PAGE_HUGE) != 0;
+}
+
+int pud_huge(pud_t pud)
+{
+ return (pud_val(pud) & _PAGE_HUGE) != 0;
+}
+
+struct page *
+follow_huge_pmd(struct mm_struct *mm, unsigned long address,
+ pmd_t *pmd, int write)
+{
+ struct page *page;
+
+ page = pte_page(*(pte_t *)pmd);
+ if (page)
+ page += ((address & ~HPAGE_MASK) >> PAGE_SHIFT);
+ return page;
+}
+
--
1.6.0.6
^ permalink raw reply related [flat|nested] 8+ messages in thread* [PATCH 3/5] MIPS: TLB support for hugeTLBfs.
2009-05-28 0:46 [PATCH 0/5] MIPS: Add hugeTLBfs support David Daney
2009-05-28 0:47 ` [PATCH 1/5] MIPS: Add support files for hugeTLBfs David Daney
2009-05-28 0:47 ` [PATCH 2/5] MIPS: Add hugeTLBfs page defines David Daney
@ 2009-05-28 0:47 ` David Daney
2009-05-28 0:47 ` [PATCH 4/5] Enable hugetlbfs for more systems David Daney
2009-05-28 0:47 ` [PATCH 5/5] MIPS: Add SYS_SUPPORTS_HUGETLBFS Kconfig variable and enable it for some systems David Daney
4 siblings, 0 replies; 8+ messages in thread
From: David Daney @ 2009-05-28 0:47 UTC (permalink / raw)
To: linux-mips, ralf; +Cc: wli, David Daney
The TLB handlers need to check for huge pages and give them special
handling. Huge pages consist of two contiguous sub-pages of physical
memory.
* Loading entrylo0 and entrylo1 need to be handled specially.
* The page mask must be set for huge pages and then restored after
writing the TLB entries.
* The PTE for huge pages resides in the PMD, we halt traversal of the
tables there.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
---
arch/mips/mm/tlb-r4k.c | 43 +++++++++---
arch/mips/mm/tlbex.c | 165 +++++++++++++++++++++++++++++++++++++++++++++++-
2 files changed, 196 insertions(+), 12 deletions(-)
diff --git a/arch/mips/mm/tlb-r4k.c b/arch/mips/mm/tlb-r4k.c
index 9619f66..b8a73c4 100644
--- a/arch/mips/mm/tlb-r4k.c
+++ b/arch/mips/mm/tlb-r4k.c
@@ -11,6 +11,7 @@
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/mm.h>
+#include <linux/hugetlb.h>
#include <asm/cpu.h>
#include <asm/bootinfo.h>
@@ -297,21 +298,41 @@ void __update_tlb(struct vm_area_struct * vma, unsigned long address, pte_t pte)
pudp = pud_offset(pgdp, address);
pmdp = pmd_offset(pudp, address);
idx = read_c0_index();
- ptep = pte_offset_map(pmdp, address);
+#ifdef CONFIG_HUGETLB_PAGE
+ /* this could be a huge page */
+ if (pmd_huge(*pmdp)) {
+ unsigned long lo;
+ write_c0_pagemask(PM_HUGE_MASK);
+ ptep = (pte_t *)pmdp;
+ lo = pte_val(*ptep) >> 6;
+ write_c0_entrylo0(lo);
+ write_c0_entrylo1(lo + (HPAGE_SIZE >> 7));
+
+ mtc0_tlbw_hazard();
+ if (idx < 0)
+ tlb_write_random();
+ else
+ tlb_write_indexed();
+ write_c0_pagemask(PM_DEFAULT_MASK);
+ } else
+#endif
+ {
+ ptep = pte_offset_map(pmdp, address);
#if defined(CONFIG_64BIT_PHYS_ADDR) && defined(CONFIG_CPU_MIPS32)
- write_c0_entrylo0(ptep->pte_high);
- ptep++;
- write_c0_entrylo1(ptep->pte_high);
+ write_c0_entrylo0(ptep->pte_high);
+ ptep++;
+ write_c0_entrylo1(ptep->pte_high);
#else
- write_c0_entrylo0(pte_val(*ptep++) >> 6);
- write_c0_entrylo1(pte_val(*ptep) >> 6);
+ write_c0_entrylo0(pte_val(*ptep++) >> 6);
+ write_c0_entrylo1(pte_val(*ptep) >> 6);
#endif
- mtc0_tlbw_hazard();
- if (idx < 0)
- tlb_write_random();
- else
- tlb_write_indexed();
+ mtc0_tlbw_hazard();
+ if (idx < 0)
+ tlb_write_random();
+ else
+ tlb_write_indexed();
+ }
tlbw_use_hazard();
FLUSH_ITLB_VM(vma);
EXIT_CRITICAL(flags);
diff --git a/arch/mips/mm/tlbex.c b/arch/mips/mm/tlbex.c
index 62fbd0d..8f606ea 100644
--- a/arch/mips/mm/tlbex.c
+++ b/arch/mips/mm/tlbex.c
@@ -8,6 +8,7 @@
* Copyright (C) 2004, 2005, 2006, 2008 Thiemo Seufer
* Copyright (C) 2005, 2007, 2008, 2009 Maciej W. Rozycki
* Copyright (C) 2006 Ralf Baechle (ralf@linux-mips.org)
+ * Copyright (C) 2008, 2009 Cavium Networks, Inc.
*
* ... and the days got worse and worse and now you see
* I've gone completly out of my mind.
@@ -83,6 +84,9 @@ enum label_id {
label_nopage_tlbm,
label_smp_pgtable_change,
label_r3000_write_probe_fail,
+#ifdef CONFIG_HUGETLB_PAGE
+ label_tlb_huge_update,
+#endif
};
UASM_L_LA(_second_part)
@@ -99,6 +103,9 @@ UASM_L_LA(_nopage_tlbs)
UASM_L_LA(_nopage_tlbm)
UASM_L_LA(_smp_pgtable_change)
UASM_L_LA(_r3000_write_probe_fail)
+#ifdef CONFIG_HUGETLB_PAGE
+UASM_L_LA(_tlb_huge_update)
+#endif
/*
* For debug purposes.
@@ -126,6 +133,7 @@ static inline void dump_handler(const u32 *handler, int count)
#define C0_TCBIND 2, 2
#define C0_ENTRYLO1 3, 0
#define C0_CONTEXT 4, 0
+#define C0_PAGEMASK 5, 0
#define C0_BADVADDR 8, 0
#define C0_ENTRYHI 10, 0
#define C0_EPC 14, 0
@@ -383,6 +391,98 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l,
}
}
+#ifdef CONFIG_HUGETLB_PAGE
+static __cpuinit void build_huge_tlb_write_entry(u32 **p,
+ struct uasm_label **l,
+ struct uasm_reloc **r,
+ unsigned int tmp,
+ enum tlb_write_entry wmode)
+{
+ /* Set huge page tlb entry size */
+ uasm_i_lui(p, tmp, PM_HUGE_MASK >> 16);
+ uasm_i_ori(p, tmp, tmp, PM_HUGE_MASK & 0xffff);
+ uasm_i_mtc0(p, tmp, C0_PAGEMASK);
+
+ build_tlb_write_entry(p, l, r, wmode);
+
+ /* Reset default page size */
+ if (PM_DEFAULT_MASK >> 16) {
+ uasm_i_lui(p, tmp, PM_DEFAULT_MASK >> 16);
+ uasm_i_ori(p, tmp, tmp, PM_DEFAULT_MASK & 0xffff);
+ uasm_il_b(p, r, label_leave);
+ uasm_i_mtc0(p, tmp, C0_PAGEMASK);
+ } else if (PM_DEFAULT_MASK) {
+ uasm_i_ori(p, tmp, 0, PM_DEFAULT_MASK);
+ uasm_il_b(p, r, label_leave);
+ uasm_i_mtc0(p, tmp, C0_PAGEMASK);
+ } else {
+ uasm_il_b(p, r, label_leave);
+ uasm_i_mtc0(p, 0, C0_PAGEMASK);
+ }
+}
+
+/*
+ * Check if Huge PTE is present, if so then jump to LABEL.
+ */
+static void __cpuinit
+build_is_huge_pte(u32 **p, struct uasm_reloc **r, unsigned int tmp,
+ unsigned int pmd, int lid)
+{
+ UASM_i_LW(p, tmp, 0, pmd);
+ uasm_i_andi(p, tmp, tmp, _PAGE_HUGE);
+ uasm_il_bnez(p, r, tmp, lid);
+}
+
+static __cpuinit void build_huge_update_entries(u32 **p,
+ unsigned int pte,
+ unsigned int tmp)
+{
+ int small_sequence;
+
+ /*
+ * A huge PTE describes an area the size of the
+ * configured huge page size. This is twice the
+ * of the large TLB entry size we intend to use.
+ * A TLB entry half the size of the configured
+ * huge page size is configured into entrylo0
+ * and entrylo1 to cover the contiguous huge PTE
+ * address space.
+ */
+ small_sequence = (HPAGE_SIZE >> 7) < 0x10000;
+
+ /* We can clobber tmp. It isn't used after this.*/
+ if (!small_sequence)
+ uasm_i_lui(p, tmp, HPAGE_SIZE >> (7 + 16));
+
+ UASM_i_SRL(p, pte, pte, 6); /* convert to entrylo */
+ uasm_i_mtc0(p, pte, C0_ENTRYLO0); /* load it */
+ /* convert to entrylo1 */
+ if (small_sequence)
+ UASM_i_ADDIU(p, pte, pte, HPAGE_SIZE >> 7);
+ else
+ UASM_i_ADDU(p, pte, pte, tmp);
+
+ uasm_i_mtc0(p, pte, C0_ENTRYLO1); /* load it */
+}
+
+static __cpuinit void build_huge_handler_tail(u32 **p,
+ struct uasm_reloc **r,
+ struct uasm_label **l,
+ unsigned int pte,
+ unsigned int ptr)
+{
+#ifdef CONFIG_SMP
+ UASM_i_SC(p, pte, 0, ptr);
+ uasm_il_beqz(p, r, pte, label_tlb_huge_update);
+ UASM_i_LW(p, pte, 0, ptr); /* Needed because SC killed our PTE */
+#else
+ UASM_i_SW(p, pte, 0, ptr);
+#endif
+ build_huge_update_entries(p, pte, ptr);
+ build_huge_tlb_write_entry(p, l, r, pte, tlb_indexed);
+}
+#endif /* CONFIG_HUGETLB_PAGE */
+
#ifdef CONFIG_64BIT
/*
* TMP and PTR are scratch.
@@ -689,12 +789,23 @@ static void __cpuinit build_r4000_tlb_refill_handler(void)
build_get_pgde32(&p, K0, K1); /* get pgd in K1 */
#endif
+#ifdef CONFIG_HUGETLB_PAGE
+ build_is_huge_pte(&p, &r, K0, K1, label_tlb_huge_update);
+#endif
+
build_get_ptep(&p, K0, K1);
build_update_entries(&p, K0, K1);
build_tlb_write_entry(&p, &l, &r, tlb_random);
uasm_l_leave(&l, p);
uasm_i_eret(&p); /* return from trap */
+#ifdef CONFIG_HUGETLB_PAGE
+ uasm_l_tlb_huge_update(&l, p);
+ UASM_i_LW(&p, K0, 0, K1);
+ build_huge_update_entries(&p, K0, K1);
+ build_huge_tlb_write_entry(&p, &l, &r, K0, tlb_random);
+#endif
+
#ifdef CONFIG_64BIT
build_get_pgd_vmalloc64(&p, &l, &r, K0, K1);
#endif
@@ -733,7 +844,9 @@ static void __cpuinit build_r4000_tlb_refill_handler(void)
uasm_copy_handler(relocs, labels, tlb_handler, p, f);
final_len = p - tlb_handler;
} else {
-#ifdef MODULE_START
+#if defined(CONFIG_HUGETLB_PAGE)
+ const enum label_id ls = label_tlb_huge_update;
+#elif defined(MODULE_START)
const enum label_id ls = label_module_alloc;
#else
const enum label_id ls = label_vmalloc;
@@ -1130,6 +1243,15 @@ build_r4000_tlbchange_handler_head(u32 **p, struct uasm_label **l,
build_get_pgde32(p, pte, ptr); /* get pgd in ptr */
#endif
+#ifdef CONFIG_HUGETLB_PAGE
+ /*
+ * For huge tlb entries, pmd doesn't contain an address but
+ * instead contains the tlb pte. Check the PAGE_HUGE bit and
+ * see if we need to jump to huge tlb processing.
+ */
+ build_is_huge_pte(p, r, pte, ptr, label_tlb_huge_update);
+#endif
+
UASM_i_MFC0(p, pte, C0_BADVADDR);
UASM_i_LW(p, ptr, 0, ptr);
UASM_i_SRL(p, pte, pte, PAGE_SHIFT + PTE_ORDER - PTE_T_LOG2);
@@ -1187,6 +1309,19 @@ static void __cpuinit build_r4000_tlb_load_handler(void)
build_make_valid(&p, &r, K0, K1);
build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1);
+#ifdef CONFIG_HUGETLB_PAGE
+ /*
+ * This is the entry point when build_r4000_tlbchange_handler_head
+ * spots a huge page.
+ */
+ uasm_l_tlb_huge_update(&l, p);
+ iPTE_LW(&p, K0, K1);
+ build_pte_present(&p, &r, K0, K1, label_nopage_tlbl);
+ build_tlb_probe_entry(&p);
+ uasm_i_ori(&p, K0, K0, (_PAGE_ACCESSED | _PAGE_VALID));
+ build_huge_handler_tail(&p, &r, &l, K0, K1);
+#endif
+
uasm_l_nopage_tlbl(&l, p);
uasm_i_j(&p, (unsigned long)tlb_do_page_fault_0 & 0x0fffffff);
uasm_i_nop(&p);
@@ -1218,6 +1353,20 @@ static void __cpuinit build_r4000_tlb_store_handler(void)
build_make_write(&p, &r, K0, K1);
build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1);
+#ifdef CONFIG_HUGETLB_PAGE
+ /*
+ * This is the entry point when
+ * build_r4000_tlbchange_handler_head spots a huge page.
+ */
+ uasm_l_tlb_huge_update(&l, p);
+ iPTE_LW(&p, K0, K1);
+ build_pte_writable(&p, &r, K0, K1, label_nopage_tlbs);
+ build_tlb_probe_entry(&p);
+ uasm_i_ori(&p, K0, K0,
+ _PAGE_ACCESSED | _PAGE_MODIFIED | _PAGE_VALID | _PAGE_DIRTY);
+ build_huge_handler_tail(&p, &r, &l, K0, K1);
+#endif
+
uasm_l_nopage_tlbs(&l, p);
uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff);
uasm_i_nop(&p);
@@ -1250,6 +1399,20 @@ static void __cpuinit build_r4000_tlb_modify_handler(void)
build_make_write(&p, &r, K0, K1);
build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1);
+#ifdef CONFIG_HUGETLB_PAGE
+ /*
+ * This is the entry point when
+ * build_r4000_tlbchange_handler_head spots a huge page.
+ */
+ uasm_l_tlb_huge_update(&l, p);
+ iPTE_LW(&p, K0, K1);
+ build_pte_modifiable(&p, &r, K0, K1, label_nopage_tlbm);
+ build_tlb_probe_entry(&p);
+ uasm_i_ori(&p, K0, K0,
+ _PAGE_ACCESSED | _PAGE_MODIFIED | _PAGE_VALID | _PAGE_DIRTY);
+ build_huge_handler_tail(&p, &r, &l, K0, K1);
+#endif
+
uasm_l_nopage_tlbm(&l, p);
uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff);
uasm_i_nop(&p);
--
1.6.0.6
^ permalink raw reply related [flat|nested] 8+ messages in thread