aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--arch/powerpc/mm/Makefile8
-rw-r--r--arch/powerpc/mm/hash_low_64.S (renamed from arch/ppc64/mm/hash_low.S)2
-rw-r--r--arch/powerpc/mm/hash_native_64.c (renamed from arch/ppc64/mm/hash_native.c)0
-rw-r--r--arch/powerpc/mm/hash_utils_64.c (renamed from arch/ppc64/mm/hash_utils.c)2
-rw-r--r--arch/powerpc/mm/hugetlbpage.c (renamed from arch/ppc64/mm/hugetlbpage.c)0
-rw-r--r--arch/powerpc/mm/imalloc.c (renamed from arch/ppc64/mm/imalloc.c)0
-rw-r--r--arch/powerpc/mm/init_64.c36
-rw-r--r--arch/powerpc/mm/mem.c11
-rw-r--r--arch/powerpc/mm/mmap.c (renamed from arch/ppc64/mm/mmap.c)0
-rw-r--r--arch/powerpc/mm/mmu_decl.h18
-rw-r--r--arch/powerpc/mm/numa.c (renamed from arch/ppc64/mm/numa.c)0
-rw-r--r--arch/powerpc/mm/pgtable_64.c34
-rw-r--r--arch/powerpc/mm/slb.c (renamed from arch/ppc64/mm/slb.c)0
-rw-r--r--arch/powerpc/mm/slb_low.S (renamed from arch/ppc64/mm/slb_low.S)0
-rw-r--r--arch/powerpc/mm/stab.c (renamed from arch/ppc64/mm/stab.c)0
-rw-r--r--arch/powerpc/mm/tlb_64.c (renamed from arch/ppc64/mm/tlb.c)0
-rw-r--r--arch/ppc64/Makefile2
-rw-r--r--arch/ppc64/mm/Makefile11
-rw-r--r--arch/ppc64/mm/fault.c333
-rw-r--r--arch/ppc64/mm/init.c870
20 files changed, 39 insertions, 1288 deletions
diff --git a/arch/powerpc/mm/Makefile b/arch/powerpc/mm/Makefile
index 35497deeb4b2..612bc4ec72b1 100644
--- a/arch/powerpc/mm/Makefile
+++ b/arch/powerpc/mm/Makefile
@@ -5,8 +5,14 @@
5obj-y := fault.o mem.o lmb.o 5obj-y := fault.o mem.o lmb.o
6obj-$(CONFIG_PPC32) += init_32.o pgtable_32.o mmu_context_32.o \ 6obj-$(CONFIG_PPC32) += init_32.o pgtable_32.o mmu_context_32.o \
7 tlb_32.o 7 tlb_32.o
8obj-$(CONFIG_PPC64) += init_64.o pgtable_64.o mmu_context_64.o 8hash-$(CONFIG_PPC_MULTIPLATFORM) := hash_native_64.o
9obj-$(CONFIG_PPC64) += init_64.o pgtable_64.o mmu_context_64.o \
10 hash_utils_64.o hash_low_64.o tlb_64.o \
11 slb_low.o slb.o stab.o mmap.o imalloc.o \
12 $(hash-y)
9obj-$(CONFIG_PPC_STD_MMU_32) += ppc_mmu_32.o hash_low_32.o 13obj-$(CONFIG_PPC_STD_MMU_32) += ppc_mmu_32.o hash_low_32.o
10obj-$(CONFIG_40x) += 4xx_mmu.o 14obj-$(CONFIG_40x) += 4xx_mmu.o
11obj-$(CONFIG_44x) += 44x_mmu.o 15obj-$(CONFIG_44x) += 44x_mmu.o
12obj-$(CONFIG_FSL_BOOKE) += fsl_booke_mmu.o 16obj-$(CONFIG_FSL_BOOKE) += fsl_booke_mmu.o
17obj-$(CONFIG_NEED_MULTIPLE_NODES) += numa.o
18obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
diff --git a/arch/ppc64/mm/hash_low.S b/arch/powerpc/mm/hash_low_64.S
index ee5a5d36bfa8..d6ed9102eeea 100644
--- a/arch/ppc64/mm/hash_low.S
+++ b/arch/powerpc/mm/hash_low_64.S
@@ -10,7 +10,7 @@
10 * described in the kernel's COPYING file. 10 * described in the kernel's COPYING file.
11 */ 11 */
12 12
13#include <asm/processor.h> 13#include <asm/reg.h>
14#include <asm/pgtable.h> 14#include <asm/pgtable.h>
15#include <asm/mmu.h> 15#include <asm/mmu.h>
16#include <asm/page.h> 16#include <asm/page.h>
diff --git a/arch/ppc64/mm/hash_native.c b/arch/powerpc/mm/hash_native_64.c
index 174d14576c28..174d14576c28 100644
--- a/arch/ppc64/mm/hash_native.c
+++ b/arch/powerpc/mm/hash_native_64.c
diff --git a/arch/ppc64/mm/hash_utils.c b/arch/powerpc/mm/hash_utils_64.c
index 83507438d6a0..35dd93eeaf4b 100644
--- a/arch/ppc64/mm/hash_utils.c
+++ b/arch/powerpc/mm/hash_utils_64.c
@@ -78,7 +78,7 @@ extern unsigned long dart_tablebase;
78hpte_t *htab_address; 78hpte_t *htab_address;
79unsigned long htab_hash_mask; 79unsigned long htab_hash_mask;
80 80
81extern unsigned long _SDR1; 81unsigned long _SDR1;
82 82
83#define KB (1024) 83#define KB (1024)
84#define MB (1024*KB) 84#define MB (1024*KB)
diff --git a/arch/ppc64/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
index 0ea0994ed974..0ea0994ed974 100644
--- a/arch/ppc64/mm/hugetlbpage.c
+++ b/arch/powerpc/mm/hugetlbpage.c
diff --git a/arch/ppc64/mm/imalloc.c b/arch/powerpc/mm/imalloc.c
index c65b87b92756..c65b87b92756 100644
--- a/arch/ppc64/mm/imalloc.c
+++ b/arch/powerpc/mm/imalloc.c
diff --git a/arch/powerpc/mm/init_64.c b/arch/powerpc/mm/init_64.c
index c0ce6a7af3c7..b0fc822ec29f 100644
--- a/arch/powerpc/mm/init_64.c
+++ b/arch/powerpc/mm/init_64.c
@@ -73,18 +73,8 @@
73#warning TASK_SIZE is smaller than it needs to be. 73#warning TASK_SIZE is smaller than it needs to be.
74#endif 74#endif
75 75
76int mem_init_done;
77unsigned long ioremap_bot = IMALLOC_BASE;
78static unsigned long phbs_io_bot = PHBS_IO_BASE;
79
80extern pgd_t swapper_pg_dir[];
81extern struct task_struct *current_set[NR_CPUS];
82
83unsigned long klimit = (unsigned long)_end; 76unsigned long klimit = (unsigned long)_end;
84 77
85unsigned long _SDR1=0;
86unsigned long _ASR=0;
87
88/* max amount of RAM to use */ 78/* max amount of RAM to use */
89unsigned long __max_memory; 79unsigned long __max_memory;
90 80
@@ -193,19 +183,6 @@ static int __init setup_kcore(void)
193} 183}
194module_init(setup_kcore); 184module_init(setup_kcore);
195 185
196void __iomem * reserve_phb_iospace(unsigned long size)
197{
198 void __iomem *virt_addr;
199
200 if (phbs_io_bot >= IMALLOC_BASE)
201 panic("reserve_phb_iospace(): phb io space overflow\n");
202
203 virt_addr = (void __iomem *) phbs_io_bot;
204 phbs_io_bot += size;
205
206 return virt_addr;
207}
208
209static void zero_ctor(void *addr, kmem_cache_t *cache, unsigned long flags) 186static void zero_ctor(void *addr, kmem_cache_t *cache, unsigned long flags)
210{ 187{
211 memset(addr, 0, kmem_cache_size(cache)); 188 memset(addr, 0, kmem_cache_size(cache));
@@ -244,16 +221,3 @@ void pgtable_cache_init(void)
244 name); 221 name);
245 } 222 }
246} 223}
247
248pgprot_t phys_mem_access_prot(struct file *file, unsigned long addr,
249 unsigned long size, pgprot_t vma_prot)
250{
251 if (ppc_md.phys_mem_access_prot)
252 return ppc_md.phys_mem_access_prot(file, addr, size, vma_prot);
253
254 if (!page_is_ram(addr >> PAGE_SHIFT))
255 vma_prot = __pgprot(pgprot_val(vma_prot)
256 | _PAGE_GUARDED | _PAGE_NO_CACHE);
257 return vma_prot;
258}
259EXPORT_SYMBOL(phys_mem_access_prot);
diff --git a/arch/powerpc/mm/mem.c b/arch/powerpc/mm/mem.c
index 0650de74d0b3..55b5860ed3c9 100644
--- a/arch/powerpc/mm/mem.c
+++ b/arch/powerpc/mm/mem.c
@@ -47,6 +47,9 @@
47#include <asm/prom.h> 47#include <asm/prom.h>
48#include <asm/lmb.h> 48#include <asm/lmb.h>
49#include <asm/sections.h> 49#include <asm/sections.h>
50#ifdef CONFIG_PPC64
51#include <asm/vdso.h>
52#endif
50 53
51#include "mmu_decl.h" 54#include "mmu_decl.h"
52 55
@@ -334,7 +337,7 @@ void flush_dcache_icache_page(struct page *page)
334 void *start = kmap_atomic(page, KM_PPC_SYNC_ICACHE); 337 void *start = kmap_atomic(page, KM_PPC_SYNC_ICACHE);
335 __flush_dcache_icache(start); 338 __flush_dcache_icache(start);
336 kunmap_atomic(start, KM_PPC_SYNC_ICACHE); 339 kunmap_atomic(start, KM_PPC_SYNC_ICACHE);
337#elif defined(CONFIG_8xx) 340#elif defined(CONFIG_8xx) || defined(CONFIG_PPC64)
338 /* On 8xx there is no need to kmap since highmem is not supported */ 341 /* On 8xx there is no need to kmap since highmem is not supported */
339 __flush_dcache_icache(page_address(page)); 342 __flush_dcache_icache(page_address(page));
340#else 343#else
@@ -463,18 +466,18 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long address,
463 if (pgdir == NULL) 466 if (pgdir == NULL)
464 return; 467 return;
465 468
466 ptep = find_linux_pte(pgdir, ea); 469 ptep = find_linux_pte(pgdir, address);
467 if (!ptep) 470 if (!ptep)
468 return; 471 return;
469 472
470 vsid = get_vsid(vma->vm_mm->context.id, ea); 473 vsid = get_vsid(vma->vm_mm->context.id, address);
471 474
472 local_irq_save(flags); 475 local_irq_save(flags);
473 tmp = cpumask_of_cpu(smp_processor_id()); 476 tmp = cpumask_of_cpu(smp_processor_id());
474 if (cpus_equal(vma->vm_mm->cpu_vm_mask, tmp)) 477 if (cpus_equal(vma->vm_mm->cpu_vm_mask, tmp))
475 local = 1; 478 local = 1;
476 479
477 __hash_page(ea, pte_val(pte) & (_PAGE_USER|_PAGE_RW), vsid, ptep, 480 __hash_page(address, pte_val(pte) & (_PAGE_USER|_PAGE_RW), vsid, ptep,
478 0x300, local); 481 0x300, local);
479 local_irq_restore(flags); 482 local_irq_restore(flags);
480#endif 483#endif
diff --git a/arch/ppc64/mm/mmap.c b/arch/powerpc/mm/mmap.c
index fe65f522aff3..fe65f522aff3 100644
--- a/arch/ppc64/mm/mmap.c
+++ b/arch/powerpc/mm/mmap.c
diff --git a/arch/powerpc/mm/mmu_decl.h b/arch/powerpc/mm/mmu_decl.h
index 06fe8af3af55..a4d7a327c0e5 100644
--- a/arch/powerpc/mm/mmu_decl.h
+++ b/arch/powerpc/mm/mmu_decl.h
@@ -22,11 +22,11 @@
22#include <asm/tlbflush.h> 22#include <asm/tlbflush.h>
23#include <asm/mmu.h> 23#include <asm/mmu.h>
24 24
25#ifdef CONFIG_PPC32
25extern void mapin_ram(void); 26extern void mapin_ram(void);
26extern int map_page(unsigned long va, phys_addr_t pa, int flags); 27extern int map_page(unsigned long va, phys_addr_t pa, int flags);
27extern void setbat(int index, unsigned long virt, unsigned long phys, 28extern void setbat(int index, unsigned long virt, unsigned long phys,
28 unsigned int size, int flags); 29 unsigned int size, int flags);
29extern void reserve_phys_mem(unsigned long start, unsigned long size);
30extern void settlbcam(int index, unsigned long virt, phys_addr_t phys, 30extern void settlbcam(int index, unsigned long virt, phys_addr_t phys,
31 unsigned int size, int flags, unsigned int pid); 31 unsigned int size, int flags, unsigned int pid);
32extern void invalidate_tlbcam_entry(int index); 32extern void invalidate_tlbcam_entry(int index);
@@ -36,16 +36,16 @@ extern unsigned long ioremap_base;
36extern unsigned long ioremap_bot; 36extern unsigned long ioremap_bot;
37extern unsigned int rtas_data, rtas_size; 37extern unsigned int rtas_data, rtas_size;
38 38
39extern unsigned long __max_low_memory;
40extern unsigned long __initial_memory_limit;
41extern unsigned long total_memory;
42extern unsigned long total_lowmem;
43extern int mem_init_done;
44
45extern PTE *Hash, *Hash_end; 39extern PTE *Hash, *Hash_end;
46extern unsigned long Hash_size, Hash_mask; 40extern unsigned long Hash_size, Hash_mask;
47 41
48extern unsigned int num_tlbcam_entries; 42extern unsigned int num_tlbcam_entries;
43#endif
44
45extern unsigned long __max_low_memory;
46extern unsigned long __initial_memory_limit;
47extern unsigned long total_memory;
48extern unsigned long total_lowmem;
49 49
50/* ...and now those things that may be slightly different between processor 50/* ...and now those things that may be slightly different between processor
51 * architectures. -- Dan 51 * architectures. -- Dan
@@ -66,8 +66,8 @@ extern void MMU_init_hw(void);
66extern unsigned long mmu_mapin_ram(void); 66extern unsigned long mmu_mapin_ram(void);
67extern void adjust_total_lowmem(void); 67extern void adjust_total_lowmem(void);
68 68
69#else 69#elif defined(CONFIG_PPC32)
70/* anything except 4xx or 8xx */ 70/* anything 32-bit except 4xx or 8xx */
71extern void MMU_init_hw(void); 71extern void MMU_init_hw(void);
72extern unsigned long mmu_mapin_ram(void); 72extern unsigned long mmu_mapin_ram(void);
73 73
diff --git a/arch/ppc64/mm/numa.c b/arch/powerpc/mm/numa.c
index cb864b8f2750..cb864b8f2750 100644
--- a/arch/ppc64/mm/numa.c
+++ b/arch/powerpc/mm/numa.c
diff --git a/arch/powerpc/mm/pgtable_64.c b/arch/powerpc/mm/pgtable_64.c
index 724f97e5dee5..484d24f9208b 100644
--- a/arch/powerpc/mm/pgtable_64.c
+++ b/arch/powerpc/mm/pgtable_64.c
@@ -67,30 +67,9 @@
67#include <asm/vdso.h> 67#include <asm/vdso.h>
68#include <asm/imalloc.h> 68#include <asm/imalloc.h>
69 69
70#if PGTABLE_RANGE > USER_VSID_RANGE
71#warning Limited user VSID range means pagetable space is wasted
72#endif
73
74#if (TASK_SIZE_USER64 < PGTABLE_RANGE) && (TASK_SIZE_USER64 < USER_VSID_RANGE)
75#warning TASK_SIZE is smaller than it needs to be.
76#endif
77
78int mem_init_done;
79unsigned long ioremap_bot = IMALLOC_BASE; 70unsigned long ioremap_bot = IMALLOC_BASE;
80static unsigned long phbs_io_bot = PHBS_IO_BASE; 71static unsigned long phbs_io_bot = PHBS_IO_BASE;
81 72
82extern pgd_t swapper_pg_dir[];
83extern struct task_struct *current_set[NR_CPUS];
84
85unsigned long klimit = (unsigned long)_end;
86
87/* max amount of RAM to use */
88unsigned long __max_memory;
89
90/* info on what we think the IO hole is */
91unsigned long io_hole_start;
92unsigned long io_hole_size;
93
94#ifdef CONFIG_PPC_ISERIES 73#ifdef CONFIG_PPC_ISERIES
95 74
96void __iomem *ioremap(unsigned long addr, unsigned long size) 75void __iomem *ioremap(unsigned long addr, unsigned long size)
@@ -355,3 +334,16 @@ int iounmap_explicit(volatile void __iomem *start, unsigned long size)
355EXPORT_SYMBOL(ioremap); 334EXPORT_SYMBOL(ioremap);
356EXPORT_SYMBOL(__ioremap); 335EXPORT_SYMBOL(__ioremap);
357EXPORT_SYMBOL(iounmap); 336EXPORT_SYMBOL(iounmap);
337
338void __iomem * reserve_phb_iospace(unsigned long size)
339{
340 void __iomem *virt_addr;
341
342 if (phbs_io_bot >= IMALLOC_BASE)
343 panic("reserve_phb_iospace(): phb io space overflow\n");
344
345 virt_addr = (void __iomem *) phbs_io_bot;
346 phbs_io_bot += size;
347
348 return virt_addr;
349}
diff --git a/arch/ppc64/mm/slb.c b/arch/powerpc/mm/slb.c
index 0473953f6a37..0473953f6a37 100644
--- a/arch/ppc64/mm/slb.c
+++ b/arch/powerpc/mm/slb.c
diff --git a/arch/ppc64/mm/slb_low.S b/arch/powerpc/mm/slb_low.S
index a3a03da503bc..a3a03da503bc 100644
--- a/arch/ppc64/mm/slb_low.S
+++ b/arch/powerpc/mm/slb_low.S
diff --git a/arch/ppc64/mm/stab.c b/arch/powerpc/mm/stab.c
index 1b83f002bf27..1b83f002bf27 100644
--- a/arch/ppc64/mm/stab.c
+++ b/arch/powerpc/mm/stab.c
diff --git a/arch/ppc64/mm/tlb.c b/arch/powerpc/mm/tlb_64.c
index 09ab81a10f4f..09ab81a10f4f 100644
--- a/arch/ppc64/mm/tlb.c
+++ b/arch/powerpc/mm/tlb_64.c
diff --git a/arch/ppc64/Makefile b/arch/ppc64/Makefile
index fa889204d6ae..4a9928ef3032 100644
--- a/arch/ppc64/Makefile
+++ b/arch/ppc64/Makefile
@@ -83,7 +83,7 @@ head-y := arch/ppc64/kernel/head.o
83 83
84libs-y += arch/ppc64/lib/ 84libs-y += arch/ppc64/lib/
85core-y += arch/ppc64/kernel/ arch/powerpc/kernel/ 85core-y += arch/ppc64/kernel/ arch/powerpc/kernel/
86core-y += arch/ppc64/mm/ 86core-y += arch/powerpc/mm/
87core-y += arch/powerpc/platforms/ 87core-y += arch/powerpc/platforms/
88core-$(CONFIG_XMON) += arch/ppc64/xmon/ 88core-$(CONFIG_XMON) += arch/ppc64/xmon/
89drivers-$(CONFIG_OPROFILE) += arch/powerpc/oprofile/ 89drivers-$(CONFIG_OPROFILE) += arch/powerpc/oprofile/
diff --git a/arch/ppc64/mm/Makefile b/arch/ppc64/mm/Makefile
deleted file mode 100644
index 3695d00d347f..000000000000
--- a/arch/ppc64/mm/Makefile
+++ /dev/null
@@ -1,11 +0,0 @@
1#
2# Makefile for the linux ppc-specific parts of the memory manager.
3#
4
5EXTRA_CFLAGS += -mno-minimal-toc
6
7obj-y := fault.o init.o imalloc.o hash_utils.o hash_low.o tlb.o \
8 slb_low.o slb.o stab.o mmap.o
9obj-$(CONFIG_NEED_MULTIPLE_NODES) += numa.o
10obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
11obj-$(CONFIG_PPC_MULTIPLATFORM) += hash_native.o
diff --git a/arch/ppc64/mm/fault.c b/arch/ppc64/mm/fault.c
deleted file mode 100644
index be3f25cf3e9f..000000000000
--- a/arch/ppc64/mm/fault.c
+++ /dev/null
@@ -1,333 +0,0 @@
1/*
2 * arch/ppc/mm/fault.c
3 *
4 * PowerPC version
5 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
6 *
7 * Derived from "arch/i386/mm/fault.c"
8 * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
9 *
10 * Modified by Cort Dougan and Paul Mackerras.
11 *
12 * Modified for PPC64 by Dave Engebretsen (engebret@ibm.com)
13 *
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
18 */
19
20#include <linux/config.h>
21#include <linux/signal.h>
22#include <linux/sched.h>
23#include <linux/kernel.h>
24#include <linux/errno.h>
25#include <linux/string.h>
26#include <linux/types.h>
27#include <linux/mman.h>
28#include <linux/mm.h>
29#include <linux/interrupt.h>
30#include <linux/smp_lock.h>
31#include <linux/module.h>
32#include <linux/kprobes.h>
33
34#include <asm/page.h>
35#include <asm/pgtable.h>
36#include <asm/mmu.h>
37#include <asm/mmu_context.h>
38#include <asm/system.h>
39#include <asm/uaccess.h>
40#include <asm/kdebug.h>
41#include <asm/siginfo.h>
42
43/*
44 * Check whether the instruction at regs->nip is a store using
45 * an update addressing form which will update r1.
46 */
47static int store_updates_sp(struct pt_regs *regs)
48{
49 unsigned int inst;
50
51 if (get_user(inst, (unsigned int __user *)regs->nip))
52 return 0;
53 /* check for 1 in the rA field */
54 if (((inst >> 16) & 0x1f) != 1)
55 return 0;
56 /* check major opcode */
57 switch (inst >> 26) {
58 case 37: /* stwu */
59 case 39: /* stbu */
60 case 45: /* sthu */
61 case 53: /* stfsu */
62 case 55: /* stfdu */
63 return 1;
64 case 62: /* std or stdu */
65 return (inst & 3) == 1;
66 case 31:
67 /* check minor opcode */
68 switch ((inst >> 1) & 0x3ff) {
69 case 181: /* stdux */
70 case 183: /* stwux */
71 case 247: /* stbux */
72 case 439: /* sthux */
73 case 695: /* stfsux */
74 case 759: /* stfdux */
75 return 1;
76 }
77 }
78 return 0;
79}
80
81static void do_dabr(struct pt_regs *regs, unsigned long error_code)
82{
83 siginfo_t info;
84
85 if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
86 11, SIGSEGV) == NOTIFY_STOP)
87 return;
88
89 if (debugger_dabr_match(regs))
90 return;
91
92 /* Clear the DABR */
93 set_dabr(0);
94
95 /* Deliver the signal to userspace */
96 info.si_signo = SIGTRAP;
97 info.si_errno = 0;
98 info.si_code = TRAP_HWBKPT;
99 info.si_addr = (void __user *)regs->nip;
100 force_sig_info(SIGTRAP, &info, current);
101}
102
103/*
104 * The error_code parameter is
105 * - DSISR for a non-SLB data access fault,
106 * - SRR1 & 0x08000000 for a non-SLB instruction access fault
107 * - 0 any SLB fault.
108 * The return value is 0 if the fault was handled, or the signal
109 * number if this is a kernel fault that can't be handled here.
110 */
111int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
112 unsigned long error_code)
113{
114 struct vm_area_struct * vma;
115 struct mm_struct *mm = current->mm;
116 siginfo_t info;
117 unsigned long code = SEGV_MAPERR;
118 unsigned long is_write = error_code & DSISR_ISSTORE;
119 unsigned long trap = TRAP(regs);
120 unsigned long is_exec = trap == 0x400;
121
122 BUG_ON((trap == 0x380) || (trap == 0x480));
123
124 if (notify_die(DIE_PAGE_FAULT, "page_fault", regs, error_code,
125 11, SIGSEGV) == NOTIFY_STOP)
126 return 0;
127
128 if (trap == 0x300) {
129 if (debugger_fault_handler(regs))
130 return 0;
131 }
132
133 /* On a kernel SLB miss we can only check for a valid exception entry */
134 if (!user_mode(regs) && (address >= TASK_SIZE))
135 return SIGSEGV;
136
137 if (error_code & DSISR_DABRMATCH) {
138 do_dabr(regs, error_code);
139 return 0;
140 }
141
142 if (in_atomic() || mm == NULL) {
143 if (!user_mode(regs))
144 return SIGSEGV;
145 /* in_atomic() in user mode is really bad,
146 as is current->mm == NULL. */
147 printk(KERN_EMERG "Page fault in user mode with"
148 "in_atomic() = %d mm = %p\n", in_atomic(), mm);
149 printk(KERN_EMERG "NIP = %lx MSR = %lx\n",
150 regs->nip, regs->msr);
151 die("Weird page fault", regs, SIGSEGV);
152 }
153
154 /* When running in the kernel we expect faults to occur only to
155 * addresses in user space. All other faults represent errors in the
156 * kernel and should generate an OOPS. Unfortunatly, in the case of an
157 * erroneous fault occuring in a code path which already holds mmap_sem
158 * we will deadlock attempting to validate the fault against the
159 * address space. Luckily the kernel only validly references user
160 * space from well defined areas of code, which are listed in the
161 * exceptions table.
162 *
163 * As the vast majority of faults will be valid we will only perform
164 * the source reference check when there is a possibilty of a deadlock.
165 * Attempt to lock the address space, if we cannot we then validate the
166 * source. If this is invalid we can skip the address space check,
167 * thus avoiding the deadlock.
168 */
169 if (!down_read_trylock(&mm->mmap_sem)) {
170 if (!user_mode(regs) && !search_exception_tables(regs->nip))
171 goto bad_area_nosemaphore;
172
173 down_read(&mm->mmap_sem);
174 }
175
176 vma = find_vma(mm, address);
177 if (!vma)
178 goto bad_area;
179
180 if (vma->vm_start <= address) {
181 goto good_area;
182 }
183 if (!(vma->vm_flags & VM_GROWSDOWN))
184 goto bad_area;
185
186 /*
187 * N.B. The POWER/Open ABI allows programs to access up to
188 * 288 bytes below the stack pointer.
189 * The kernel signal delivery code writes up to about 1.5kB
190 * below the stack pointer (r1) before decrementing it.
191 * The exec code can write slightly over 640kB to the stack
192 * before setting the user r1. Thus we allow the stack to
193 * expand to 1MB without further checks.
194 */
195 if (address + 0x100000 < vma->vm_end) {
196 /* get user regs even if this fault is in kernel mode */
197 struct pt_regs *uregs = current->thread.regs;
198 if (uregs == NULL)
199 goto bad_area;
200
201 /*
202 * A user-mode access to an address a long way below
203 * the stack pointer is only valid if the instruction
204 * is one which would update the stack pointer to the
205 * address accessed if the instruction completed,
206 * i.e. either stwu rs,n(r1) or stwux rs,r1,rb
207 * (or the byte, halfword, float or double forms).
208 *
209 * If we don't check this then any write to the area
210 * between the last mapped region and the stack will
211 * expand the stack rather than segfaulting.
212 */
213 if (address + 2048 < uregs->gpr[1]
214 && (!user_mode(regs) || !store_updates_sp(regs)))
215 goto bad_area;
216 }
217
218 if (expand_stack(vma, address))
219 goto bad_area;
220
221good_area:
222 code = SEGV_ACCERR;
223
224 if (is_exec) {
225 /* protection fault */
226 if (error_code & DSISR_PROTFAULT)
227 goto bad_area;
228 if (!(vma->vm_flags & VM_EXEC))
229 goto bad_area;
230 /* a write */
231 } else if (is_write) {
232 if (!(vma->vm_flags & VM_WRITE))
233 goto bad_area;
234 /* a read */
235 } else {
236 if (!(vma->vm_flags & VM_READ))
237 goto bad_area;
238 }
239
240 survive:
241 /*
242 * If for any reason at all we couldn't handle the fault,
243 * make sure we exit gracefully rather than endlessly redo
244 * the fault.
245 */
246 switch (handle_mm_fault(mm, vma, address, is_write)) {
247
248 case VM_FAULT_MINOR:
249 current->min_flt++;
250 break;
251 case VM_FAULT_MAJOR:
252 current->maj_flt++;
253 break;
254 case VM_FAULT_SIGBUS:
255 goto do_sigbus;
256 case VM_FAULT_OOM:
257 goto out_of_memory;
258 default:
259 BUG();
260 }
261
262 up_read(&mm->mmap_sem);
263 return 0;
264
265bad_area:
266 up_read(&mm->mmap_sem);
267
268bad_area_nosemaphore:
269 /* User mode accesses cause a SIGSEGV */
270 if (user_mode(regs)) {
271 info.si_signo = SIGSEGV;
272 info.si_errno = 0;
273 info.si_code = code;
274 info.si_addr = (void __user *) address;
275 force_sig_info(SIGSEGV, &info, current);
276 return 0;
277 }
278
279 if (trap == 0x400 && (error_code & DSISR_PROTFAULT)
280 && printk_ratelimit())
281 printk(KERN_CRIT "kernel tried to execute NX-protected"
282 " page (%lx) - exploit attempt? (uid: %d)\n",
283 address, current->uid);
284
285 return SIGSEGV;
286
287/*
288 * We ran out of memory, or some other thing happened to us that made
289 * us unable to handle the page fault gracefully.
290 */
291out_of_memory:
292 up_read(&mm->mmap_sem);
293 if (current->pid == 1) {
294 yield();
295 down_read(&mm->mmap_sem);
296 goto survive;
297 }
298 printk("VM: killing process %s\n", current->comm);
299 if (user_mode(regs))
300 do_exit(SIGKILL);
301 return SIGKILL;
302
303do_sigbus:
304 up_read(&mm->mmap_sem);
305 if (user_mode(regs)) {
306 info.si_signo = SIGBUS;
307 info.si_errno = 0;
308 info.si_code = BUS_ADRERR;
309 info.si_addr = (void __user *)address;
310 force_sig_info(SIGBUS, &info, current);
311 return 0;
312 }
313 return SIGBUS;
314}
315
316/*
317 * bad_page_fault is called when we have a bad access from the kernel.
318 * It is called from do_page_fault above and from some of the procedures
319 * in traps.c.
320 */
321void bad_page_fault(struct pt_regs *regs, unsigned long address, int sig)
322{
323 const struct exception_table_entry *entry;
324
325 /* Are we prepared to handle this fault? */
326 if ((entry = search_exception_tables(regs->nip)) != NULL) {
327 regs->nip = entry->fixup;
328 return;
329 }
330
331 /* kernel has accessed a bad area */
332 die("Kernel access of bad area", regs, sig);
333}
diff --git a/arch/ppc64/mm/init.c b/arch/ppc64/mm/init.c
deleted file mode 100644
index c2157c9c3acb..000000000000
--- a/arch/ppc64/mm/init.c
+++ /dev/null
@@ -1,870 +0,0 @@
1/*
2 * PowerPC version
3 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
4 *
5 * Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au)
6 * and Cort Dougan (PReP) (cort@cs.nmt.edu)
7 * Copyright (C) 1996 Paul Mackerras
8 * Amiga/APUS changes by Jesper Skov (jskov@cygnus.co.uk).
9 *
10 * Derived from "arch/i386/mm/init.c"
11 * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
12 *
13 * Dave Engebretsen <engebret@us.ibm.com>
14 * Rework for PPC64 port.
15 *
16 * This program is free software; you can redistribute it and/or
17 * modify it under the terms of the GNU General Public License
18 * as published by the Free Software Foundation; either version
19 * 2 of the License, or (at your option) any later version.
20 *
21 */
22
23#include <linux/config.h>
24#include <linux/signal.h>
25#include <linux/sched.h>
26#include <linux/kernel.h>
27#include <linux/errno.h>
28#include <linux/string.h>
29#include <linux/types.h>
30#include <linux/mman.h>
31#include <linux/mm.h>
32#include <linux/swap.h>
33#include <linux/stddef.h>
34#include <linux/vmalloc.h>
35#include <linux/init.h>
36#include <linux/delay.h>
37#include <linux/bootmem.h>
38#include <linux/highmem.h>
39#include <linux/idr.h>
40#include <linux/nodemask.h>
41#include <linux/module.h>
42
43#include <asm/pgalloc.h>
44#include <asm/page.h>
45#include <asm/prom.h>
46#include <asm/lmb.h>
47#include <asm/rtas.h>
48#include <asm/io.h>
49#include <asm/mmu_context.h>
50#include <asm/pgtable.h>
51#include <asm/mmu.h>
52#include <asm/uaccess.h>
53#include <asm/smp.h>
54#include <asm/machdep.h>
55#include <asm/tlb.h>
56#include <asm/eeh.h>
57#include <asm/processor.h>
58#include <asm/mmzone.h>
59#include <asm/cputable.h>
60#include <asm/ppcdebug.h>
61#include <asm/sections.h>
62#include <asm/system.h>
63#include <asm/iommu.h>
64#include <asm/abs_addr.h>
65#include <asm/vdso.h>
66#include <asm/imalloc.h>
67
68#if PGTABLE_RANGE > USER_VSID_RANGE
69#warning Limited user VSID range means pagetable space is wasted
70#endif
71
72#if (TASK_SIZE_USER64 < PGTABLE_RANGE) && (TASK_SIZE_USER64 < USER_VSID_RANGE)
73#warning TASK_SIZE is smaller than it needs to be.
74#endif
75
76int mem_init_done;
77unsigned long ioremap_bot = IMALLOC_BASE;
78static unsigned long phbs_io_bot = PHBS_IO_BASE;
79
80extern pgd_t swapper_pg_dir[];
81extern struct task_struct *current_set[NR_CPUS];
82
83unsigned long klimit = (unsigned long)_end;
84
85unsigned long _SDR1=0;
86unsigned long _ASR=0;
87
88/* max amount of RAM to use */
89unsigned long __max_memory;
90
91/* info on what we think the IO hole is */
92unsigned long io_hole_start;
93unsigned long io_hole_size;
94
95void show_mem(void)
96{
97 unsigned long total = 0, reserved = 0;
98 unsigned long shared = 0, cached = 0;
99 struct page *page;
100 pg_data_t *pgdat;
101 unsigned long i;
102
103 printk("Mem-info:\n");
104 show_free_areas();
105 printk("Free swap: %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
106 for_each_pgdat(pgdat) {
107 for (i = 0; i < pgdat->node_spanned_pages; i++) {
108 page = pgdat_page_nr(pgdat, i);
109 total++;
110 if (PageReserved(page))
111 reserved++;
112 else if (PageSwapCache(page))
113 cached++;
114 else if (page_count(page))
115 shared += page_count(page) - 1;
116 }
117 }
118 printk("%ld pages of RAM\n", total);
119 printk("%ld reserved pages\n", reserved);
120 printk("%ld pages shared\n", shared);
121 printk("%ld pages swap cached\n", cached);
122}
123
124#ifdef CONFIG_PPC_ISERIES
125
126void __iomem *ioremap(unsigned long addr, unsigned long size)
127{
128 return (void __iomem *)addr;
129}
130
131extern void __iomem *__ioremap(unsigned long addr, unsigned long size,
132 unsigned long flags)
133{
134 return (void __iomem *)addr;
135}
136
137void iounmap(volatile void __iomem *addr)
138{
139 return;
140}
141
142#else
143
144/*
145 * map_io_page currently only called by __ioremap
146 * map_io_page adds an entry to the ioremap page table
147 * and adds an entry to the HPT, possibly bolting it
148 */
149static int map_io_page(unsigned long ea, unsigned long pa, int flags)
150{
151 pgd_t *pgdp;
152 pud_t *pudp;
153 pmd_t *pmdp;
154 pte_t *ptep;
155 unsigned long vsid;
156
157 if (mem_init_done) {
158 spin_lock(&init_mm.page_table_lock);
159 pgdp = pgd_offset_k(ea);
160 pudp = pud_alloc(&init_mm, pgdp, ea);
161 if (!pudp)
162 return -ENOMEM;
163 pmdp = pmd_alloc(&init_mm, pudp, ea);
164 if (!pmdp)
165 return -ENOMEM;
166 ptep = pte_alloc_kernel(&init_mm, pmdp, ea);
167 if (!ptep)
168 return -ENOMEM;
169 set_pte_at(&init_mm, ea, ptep, pfn_pte(pa >> PAGE_SHIFT,
170 __pgprot(flags)));
171 spin_unlock(&init_mm.page_table_lock);
172 } else {
173 unsigned long va, vpn, hash, hpteg;
174
175 /*
176 * If the mm subsystem is not fully up, we cannot create a
177 * linux page table entry for this mapping. Simply bolt an
178 * entry in the hardware page table.
179 */
180 vsid = get_kernel_vsid(ea);
181 va = (vsid << 28) | (ea & 0xFFFFFFF);
182 vpn = va >> PAGE_SHIFT;
183
184 hash = hpt_hash(vpn, 0);
185
186 hpteg = ((hash & htab_hash_mask) * HPTES_PER_GROUP);
187
188 /* Panic if a pte grpup is full */
189 if (ppc_md.hpte_insert(hpteg, va, pa >> PAGE_SHIFT,
190 HPTE_V_BOLTED,
191 _PAGE_NO_CACHE|_PAGE_GUARDED|PP_RWXX)
192 == -1) {
193 panic("map_io_page: could not insert mapping");
194 }
195 }
196 return 0;
197}
198
199
200static void __iomem * __ioremap_com(unsigned long addr, unsigned long pa,
201 unsigned long ea, unsigned long size,
202 unsigned long flags)
203{
204 unsigned long i;
205
206 if ((flags & _PAGE_PRESENT) == 0)
207 flags |= pgprot_val(PAGE_KERNEL);
208
209 for (i = 0; i < size; i += PAGE_SIZE)
210 if (map_io_page(ea+i, pa+i, flags))
211 return NULL;
212
213 return (void __iomem *) (ea + (addr & ~PAGE_MASK));
214}
215
216
217void __iomem *
218ioremap(unsigned long addr, unsigned long size)
219{
220 return __ioremap(addr, size, _PAGE_NO_CACHE | _PAGE_GUARDED);
221}
222
223void __iomem * __ioremap(unsigned long addr, unsigned long size,
224 unsigned long flags)
225{
226 unsigned long pa, ea;
227 void __iomem *ret;
228
229 /*
230 * Choose an address to map it to.
231 * Once the imalloc system is running, we use it.
232 * Before that, we map using addresses going
233 * up from ioremap_bot. imalloc will use
234 * the addresses from ioremap_bot through
235 * IMALLOC_END
236 *
237 */
238 pa = addr & PAGE_MASK;
239 size = PAGE_ALIGN(addr + size) - pa;
240
241 if (size == 0)
242 return NULL;
243
244 if (mem_init_done) {
245 struct vm_struct *area;
246 area = im_get_free_area(size);
247 if (area == NULL)
248 return NULL;
249 ea = (unsigned long)(area->addr);
250 ret = __ioremap_com(addr, pa, ea, size, flags);
251 if (!ret)
252 im_free(area->addr);
253 } else {
254 ea = ioremap_bot;
255 ret = __ioremap_com(addr, pa, ea, size, flags);
256 if (ret)
257 ioremap_bot += size;
258 }
259 return ret;
260}
261
262#define IS_PAGE_ALIGNED(_val) ((_val) == ((_val) & PAGE_MASK))
263
264int __ioremap_explicit(unsigned long pa, unsigned long ea,
265 unsigned long size, unsigned long flags)
266{
267 struct vm_struct *area;
268 void __iomem *ret;
269
270 /* For now, require page-aligned values for pa, ea, and size */
271 if (!IS_PAGE_ALIGNED(pa) || !IS_PAGE_ALIGNED(ea) ||
272 !IS_PAGE_ALIGNED(size)) {
273 printk(KERN_ERR "unaligned value in %s\n", __FUNCTION__);
274 return 1;
275 }
276
277 if (!mem_init_done) {
278 /* Two things to consider in this case:
279 * 1) No records will be kept (imalloc, etc) that the region
280 * has been remapped
281 * 2) It won't be easy to iounmap() the region later (because
282 * of 1)
283 */
284 ;
285 } else {
286 area = im_get_area(ea, size,
287 IM_REGION_UNUSED|IM_REGION_SUBSET|IM_REGION_EXISTS);
288 if (area == NULL) {
289 /* Expected when PHB-dlpar is in play */
290 return 1;
291 }
292 if (ea != (unsigned long) area->addr) {
293 printk(KERN_ERR "unexpected addr return from "
294 "im_get_area\n");
295 return 1;
296 }
297 }
298
299 ret = __ioremap_com(pa, pa, ea, size, flags);
300 if (ret == NULL) {
301 printk(KERN_ERR "ioremap_explicit() allocation failure !\n");
302 return 1;
303 }
304 if (ret != (void *) ea) {
305 printk(KERN_ERR "__ioremap_com() returned unexpected addr\n");
306 return 1;
307 }
308
309 return 0;
310}
311
312/*
313 * Unmap an IO region and remove it from imalloc'd list.
314 * Access to IO memory should be serialized by driver.
315 * This code is modeled after vmalloc code - unmap_vm_area()
316 *
317 * XXX what about calls before mem_init_done (ie python_countermeasures())
318 */
319void iounmap(volatile void __iomem *token)
320{
321 void *addr;
322
323 if (!mem_init_done)
324 return;
325
326 addr = (void *) ((unsigned long __force) token & PAGE_MASK);
327
328 im_free(addr);
329}
330
331static int iounmap_subset_regions(unsigned long addr, unsigned long size)
332{
333 struct vm_struct *area;
334
335 /* Check whether subsets of this region exist */
336 area = im_get_area(addr, size, IM_REGION_SUPERSET);
337 if (area == NULL)
338 return 1;
339
340 while (area) {
341 iounmap((void __iomem *) area->addr);
342 area = im_get_area(addr, size,
343 IM_REGION_SUPERSET);
344 }
345
346 return 0;
347}
348
349int iounmap_explicit(volatile void __iomem *start, unsigned long size)
350{
351 struct vm_struct *area;
352 unsigned long addr;
353 int rc;
354
355 addr = (unsigned long __force) start & PAGE_MASK;
356
357 /* Verify that the region either exists or is a subset of an existing
358 * region. In the latter case, split the parent region to create
359 * the exact region
360 */
361 area = im_get_area(addr, size,
362 IM_REGION_EXISTS | IM_REGION_SUBSET);
363 if (area == NULL) {
364 /* Determine whether subset regions exist. If so, unmap */
365 rc = iounmap_subset_regions(addr, size);
366 if (rc) {
367 printk(KERN_ERR
368 "%s() cannot unmap nonexistent range 0x%lx\n",
369 __FUNCTION__, addr);
370 return 1;
371 }
372 } else {
373 iounmap((void __iomem *) area->addr);
374 }
375 /*
376 * FIXME! This can't be right:
377 iounmap(area->addr);
378 * Maybe it should be "iounmap(area);"
379 */
380 return 0;
381}
382
383#endif
384
385EXPORT_SYMBOL(ioremap);
386EXPORT_SYMBOL(__ioremap);
387EXPORT_SYMBOL(iounmap);
388
389void free_initmem(void)
390{
391 unsigned long addr;
392
393 addr = (unsigned long)__init_begin;
394 for (; addr < (unsigned long)__init_end; addr += PAGE_SIZE) {
395 memset((void *)addr, 0xcc, PAGE_SIZE);
396 ClearPageReserved(virt_to_page(addr));
397 set_page_count(virt_to_page(addr), 1);
398 free_page(addr);
399 totalram_pages++;
400 }
401 printk ("Freeing unused kernel memory: %luk freed\n",
402 ((unsigned long)__init_end - (unsigned long)__init_begin) >> 10);
403}
404
405#ifdef CONFIG_BLK_DEV_INITRD
406void free_initrd_mem(unsigned long start, unsigned long end)
407{
408 if (start < end)
409 printk ("Freeing initrd memory: %ldk freed\n", (end - start) >> 10);
410 for (; start < end; start += PAGE_SIZE) {
411 ClearPageReserved(virt_to_page(start));
412 set_page_count(virt_to_page(start), 1);
413 free_page(start);
414 totalram_pages++;
415 }
416}
417#endif
418
419static DEFINE_SPINLOCK(mmu_context_lock);
420static DEFINE_IDR(mmu_context_idr);
421
422int init_new_context(struct task_struct *tsk, struct mm_struct *mm)
423{
424 int index;
425 int err;
426
427again:
428 if (!idr_pre_get(&mmu_context_idr, GFP_KERNEL))
429 return -ENOMEM;
430
431 spin_lock(&mmu_context_lock);
432 err = idr_get_new_above(&mmu_context_idr, NULL, 1, &index);
433 spin_unlock(&mmu_context_lock);
434
435 if (err == -EAGAIN)
436 goto again;
437 else if (err)
438 return err;
439
440 if (index > MAX_CONTEXT) {
441 idr_remove(&mmu_context_idr, index);
442 return -ENOMEM;
443 }
444
445 mm->context.id = index;
446
447 return 0;
448}
449
450void destroy_context(struct mm_struct *mm)
451{
452 spin_lock(&mmu_context_lock);
453 idr_remove(&mmu_context_idr, mm->context.id);
454 spin_unlock(&mmu_context_lock);
455
456 mm->context.id = NO_CONTEXT;
457}
458
459/*
460 * Do very early mm setup.
461 */
462void __init mm_init_ppc64(void)
463{
464#ifndef CONFIG_PPC_ISERIES
465 unsigned long i;
466#endif
467
468 ppc64_boot_msg(0x100, "MM Init");
469
470 /* This is the story of the IO hole... please, keep seated,
471 * unfortunately, we are out of oxygen masks at the moment.
472 * So we need some rough way to tell where your big IO hole
473 * is. On pmac, it's between 2G and 4G, on POWER3, it's around
474 * that area as well, on POWER4 we don't have one, etc...
475 * We need that as a "hint" when sizing the TCE table on POWER3
476 * So far, the simplest way that seem work well enough for us it
477 * to just assume that the first discontinuity in our physical
478 * RAM layout is the IO hole. That may not be correct in the future
479 * (and isn't on iSeries but then we don't care ;)
480 */
481
482#ifndef CONFIG_PPC_ISERIES
483 for (i = 1; i < lmb.memory.cnt; i++) {
484 unsigned long base, prevbase, prevsize;
485
486 prevbase = lmb.memory.region[i-1].base;
487 prevsize = lmb.memory.region[i-1].size;
488 base = lmb.memory.region[i].base;
489 if (base > (prevbase + prevsize)) {
490 io_hole_start = prevbase + prevsize;
491 io_hole_size = base - (prevbase + prevsize);
492 break;
493 }
494 }
495#endif /* CONFIG_PPC_ISERIES */
496 if (io_hole_start)
497 printk("IO Hole assumed to be %lx -> %lx\n",
498 io_hole_start, io_hole_start + io_hole_size - 1);
499
500 ppc64_boot_msg(0x100, "MM Init Done");
501}
502
503/*
504 * This is called by /dev/mem to know if a given address has to
505 * be mapped non-cacheable or not
506 */
507int page_is_ram(unsigned long pfn)
508{
509 int i;
510 unsigned long paddr = (pfn << PAGE_SHIFT);
511
512 for (i=0; i < lmb.memory.cnt; i++) {
513 unsigned long base;
514
515 base = lmb.memory.region[i].base;
516
517 if ((paddr >= base) &&
518 (paddr < (base + lmb.memory.region[i].size))) {
519 return 1;
520 }
521 }
522
523 return 0;
524}
525EXPORT_SYMBOL(page_is_ram);
526
527/*
528 * Initialize the bootmem system and give it all the memory we
529 * have available.
530 */
531#ifndef CONFIG_NEED_MULTIPLE_NODES
532void __init do_init_bootmem(void)
533{
534 unsigned long i;
535 unsigned long start, bootmap_pages;
536 unsigned long total_pages = lmb_end_of_DRAM() >> PAGE_SHIFT;
537 int boot_mapsize;
538
539 /*
540 * Find an area to use for the bootmem bitmap. Calculate the size of
541 * bitmap required as (Total Memory) / PAGE_SIZE / BITS_PER_BYTE.
542 * Add 1 additional page in case the address isn't page-aligned.
543 */
544 bootmap_pages = bootmem_bootmap_pages(total_pages);
545
546 start = lmb_alloc(bootmap_pages<<PAGE_SHIFT, PAGE_SIZE);
547 BUG_ON(!start);
548
549 boot_mapsize = init_bootmem(start >> PAGE_SHIFT, total_pages);
550
551 max_pfn = max_low_pfn;
552
553 /* Add all physical memory to the bootmem map, mark each area
554 * present.
555 */
556 for (i=0; i < lmb.memory.cnt; i++)
557 free_bootmem(lmb.memory.region[i].base,
558 lmb_size_bytes(&lmb.memory, i));
559
560 /* reserve the sections we're already using */
561 for (i=0; i < lmb.reserved.cnt; i++)
562 reserve_bootmem(lmb.reserved.region[i].base,
563 lmb_size_bytes(&lmb.reserved, i));
564
565 for (i=0; i < lmb.memory.cnt; i++)
566 memory_present(0, lmb_start_pfn(&lmb.memory, i),
567 lmb_end_pfn(&lmb.memory, i));
568}
569
570/*
571 * paging_init() sets up the page tables - in fact we've already done this.
572 */
573void __init paging_init(void)
574{
575 unsigned long zones_size[MAX_NR_ZONES];
576 unsigned long zholes_size[MAX_NR_ZONES];
577 unsigned long total_ram = lmb_phys_mem_size();
578 unsigned long top_of_ram = lmb_end_of_DRAM();
579
580 printk(KERN_INFO "Top of RAM: 0x%lx, Total RAM: 0x%lx\n",
581 top_of_ram, total_ram);
582 printk(KERN_INFO "Memory hole size: %ldMB\n",
583 (top_of_ram - total_ram) >> 20);
584 /*
585 * All pages are DMA-able so we put them all in the DMA zone.
586 */
587 memset(zones_size, 0, sizeof(zones_size));
588 memset(zholes_size, 0, sizeof(zholes_size));
589
590 zones_size[ZONE_DMA] = top_of_ram >> PAGE_SHIFT;
591 zholes_size[ZONE_DMA] = (top_of_ram - total_ram) >> PAGE_SHIFT;
592
593 free_area_init_node(0, NODE_DATA(0), zones_size,
594 __pa(PAGE_OFFSET) >> PAGE_SHIFT, zholes_size);
595}
596#endif /* ! CONFIG_NEED_MULTIPLE_NODES */
597
598static struct kcore_list kcore_vmem;
599
600static int __init setup_kcore(void)
601{
602 int i;
603
604 for (i=0; i < lmb.memory.cnt; i++) {
605 unsigned long base, size;
606 struct kcore_list *kcore_mem;
607
608 base = lmb.memory.region[i].base;
609 size = lmb.memory.region[i].size;
610
611 /* GFP_ATOMIC to avoid might_sleep warnings during boot */
612 kcore_mem = kmalloc(sizeof(struct kcore_list), GFP_ATOMIC);
613 if (!kcore_mem)
614 panic("mem_init: kmalloc failed\n");
615
616 kclist_add(kcore_mem, __va(base), size);
617 }
618
619 kclist_add(&kcore_vmem, (void *)VMALLOC_START, VMALLOC_END-VMALLOC_START);
620
621 return 0;
622}
623module_init(setup_kcore);
624
625void __init mem_init(void)
626{
627#ifdef CONFIG_NEED_MULTIPLE_NODES
628 int nid;
629#endif
630 pg_data_t *pgdat;
631 unsigned long i;
632 struct page *page;
633 unsigned long reservedpages = 0, codesize, initsize, datasize, bsssize;
634
635 num_physpages = max_low_pfn; /* RAM is assumed contiguous */
636 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE);
637
638#ifdef CONFIG_NEED_MULTIPLE_NODES
639 for_each_online_node(nid) {
640 if (NODE_DATA(nid)->node_spanned_pages != 0) {
641 printk("freeing bootmem node %x\n", nid);
642 totalram_pages +=
643 free_all_bootmem_node(NODE_DATA(nid));
644 }
645 }
646#else
647 max_mapnr = num_physpages;
648 totalram_pages += free_all_bootmem();
649#endif
650
651 for_each_pgdat(pgdat) {
652 for (i = 0; i < pgdat->node_spanned_pages; i++) {
653 page = pgdat_page_nr(pgdat, i);
654 if (PageReserved(page))
655 reservedpages++;
656 }
657 }
658
659 codesize = (unsigned long)&_etext - (unsigned long)&_stext;
660 initsize = (unsigned long)&__init_end - (unsigned long)&__init_begin;
661 datasize = (unsigned long)&_edata - (unsigned long)&__init_end;
662 bsssize = (unsigned long)&__bss_stop - (unsigned long)&__bss_start;
663
664 printk(KERN_INFO "Memory: %luk/%luk available (%luk kernel code, "
665 "%luk reserved, %luk data, %luk bss, %luk init)\n",
666 (unsigned long)nr_free_pages() << (PAGE_SHIFT-10),
667 num_physpages << (PAGE_SHIFT-10),
668 codesize >> 10,
669 reservedpages << (PAGE_SHIFT-10),
670 datasize >> 10,
671 bsssize >> 10,
672 initsize >> 10);
673
674 mem_init_done = 1;
675
676 /* Initialize the vDSO */
677 vdso_init();
678}
679
680/*
681 * This is called when a page has been modified by the kernel.
682 * It just marks the page as not i-cache clean. We do the i-cache
683 * flush later when the page is given to a user process, if necessary.
684 */
685void flush_dcache_page(struct page *page)
686{
687 if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE))
688 return;
689 /* avoid an atomic op if possible */
690 if (test_bit(PG_arch_1, &page->flags))
691 clear_bit(PG_arch_1, &page->flags);
692}
693EXPORT_SYMBOL(flush_dcache_page);
694
695void clear_user_page(void *page, unsigned long vaddr, struct page *pg)
696{
697 clear_page(page);
698
699 if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE))
700 return;
701 /*
702 * We shouldnt have to do this, but some versions of glibc
703 * require it (ld.so assumes zero filled pages are icache clean)
704 * - Anton
705 */
706
707 /* avoid an atomic op if possible */
708 if (test_bit(PG_arch_1, &pg->flags))
709 clear_bit(PG_arch_1, &pg->flags);
710}
711EXPORT_SYMBOL(clear_user_page);
712
713void copy_user_page(void *vto, void *vfrom, unsigned long vaddr,
714 struct page *pg)
715{
716 copy_page(vto, vfrom);
717
718 /*
719 * We should be able to use the following optimisation, however
720 * there are two problems.
721 * Firstly a bug in some versions of binutils meant PLT sections
722 * were not marked executable.
723 * Secondly the first word in the GOT section is blrl, used
724 * to establish the GOT address. Until recently the GOT was
725 * not marked executable.
726 * - Anton
727 */
728#if 0
729 if (!vma->vm_file && ((vma->vm_flags & VM_EXEC) == 0))
730 return;
731#endif
732
733 if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE))
734 return;
735
736 /* avoid an atomic op if possible */
737 if (test_bit(PG_arch_1, &pg->flags))
738 clear_bit(PG_arch_1, &pg->flags);
739}
740
741void flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
742 unsigned long addr, int len)
743{
744 unsigned long maddr;
745
746 maddr = (unsigned long)page_address(page) + (addr & ~PAGE_MASK);
747 flush_icache_range(maddr, maddr + len);
748}
749EXPORT_SYMBOL(flush_icache_user_range);
750
751/*
752 * This is called at the end of handling a user page fault, when the
753 * fault has been handled by updating a PTE in the linux page tables.
754 * We use it to preload an HPTE into the hash table corresponding to
755 * the updated linux PTE.
756 *
757 * This must always be called with the mm->page_table_lock held
758 */
759void update_mmu_cache(struct vm_area_struct *vma, unsigned long ea,
760 pte_t pte)
761{
762 unsigned long vsid;
763 void *pgdir;
764 pte_t *ptep;
765 int local = 0;
766 cpumask_t tmp;
767 unsigned long flags;
768
769 /* handle i-cache coherency */
770 if (!cpu_has_feature(CPU_FTR_COHERENT_ICACHE) &&
771 !cpu_has_feature(CPU_FTR_NOEXECUTE)) {
772 unsigned long pfn = pte_pfn(pte);
773 if (pfn_valid(pfn)) {
774 struct page *page = pfn_to_page(pfn);
775 if (!PageReserved(page)
776 && !test_bit(PG_arch_1, &page->flags)) {
777 __flush_dcache_icache(page_address(page));
778 set_bit(PG_arch_1, &page->flags);
779 }
780 }
781 }
782
783 /* We only want HPTEs for linux PTEs that have _PAGE_ACCESSED set */
784 if (!pte_young(pte))
785 return;
786
787 pgdir = vma->vm_mm->pgd;
788 if (pgdir == NULL)
789 return;
790
791 ptep = find_linux_pte(pgdir, ea);
792 if (!ptep)
793 return;
794
795 vsid = get_vsid(vma->vm_mm->context.id, ea);
796
797 local_irq_save(flags);
798 tmp = cpumask_of_cpu(smp_processor_id());
799 if (cpus_equal(vma->vm_mm->cpu_vm_mask, tmp))
800 local = 1;
801
802 __hash_page(ea, pte_val(pte) & (_PAGE_USER|_PAGE_RW), vsid, ptep,
803 0x300, local);
804 local_irq_restore(flags);
805}
806
807void __iomem * reserve_phb_iospace(unsigned long size)
808{
809 void __iomem *virt_addr;
810
811 if (phbs_io_bot >= IMALLOC_BASE)
812 panic("reserve_phb_iospace(): phb io space overflow\n");
813
814 virt_addr = (void __iomem *) phbs_io_bot;
815 phbs_io_bot += size;
816
817 return virt_addr;
818}
819
820static void zero_ctor(void *addr, kmem_cache_t *cache, unsigned long flags)
821{
822 memset(addr, 0, kmem_cache_size(cache));
823}
824
825static const int pgtable_cache_size[2] = {
826 PTE_TABLE_SIZE, PMD_TABLE_SIZE
827};
828static const char *pgtable_cache_name[ARRAY_SIZE(pgtable_cache_size)] = {
829 "pgd_pte_cache", "pud_pmd_cache",
830};
831
832kmem_cache_t *pgtable_cache[ARRAY_SIZE(pgtable_cache_size)];
833
834void pgtable_cache_init(void)
835{
836 int i;
837
838 BUILD_BUG_ON(PTE_TABLE_SIZE != pgtable_cache_size[PTE_CACHE_NUM]);
839 BUILD_BUG_ON(PMD_TABLE_SIZE != pgtable_cache_size[PMD_CACHE_NUM]);
840 BUILD_BUG_ON(PUD_TABLE_SIZE != pgtable_cache_size[PUD_CACHE_NUM]);
841 BUILD_BUG_ON(PGD_TABLE_SIZE != pgtable_cache_size[PGD_CACHE_NUM]);
842
843 for (i = 0; i < ARRAY_SIZE(pgtable_cache_size); i++) {
844 int size = pgtable_cache_size[i];
845 const char *name = pgtable_cache_name[i];
846
847 pgtable_cache[i] = kmem_cache_create(name,
848 size, size,
849 SLAB_HWCACHE_ALIGN
850 | SLAB_MUST_HWCACHE_ALIGN,
851 zero_ctor,
852 NULL);
853 if (! pgtable_cache[i])
854 panic("pgtable_cache_init(): could not create %s!\n",
855 name);
856 }
857}
858
859pgprot_t phys_mem_access_prot(struct file *file, unsigned long addr,
860 unsigned long size, pgprot_t vma_prot)
861{
862 if (ppc_md.phys_mem_access_prot)
863 return ppc_md.phys_mem_access_prot(file, addr, size, vma_prot);
864
865 if (!page_is_ram(addr >> PAGE_SHIFT))
866 vma_prot = __pgprot(pgprot_val(vma_prot)
867 | _PAGE_GUARDED | _PAGE_NO_CACHE);
868 return vma_prot;
869}
870EXPORT_SYMBOL(phys_mem_access_prot);
generated by cgit v1.2.2 (git 2.25.0) at 2025-10-27 19:04:53 -0400