(1) fix for fault_in_page_*() not faulting in enough mmupages. (2) fix for bogus open-coded ptep_to_address() (3) fix for iounmap() missing its targets Unfortunately, none of these are the bug we're looking for. -- wli diff -prauN linux-2.5.70-bk4/arch/i386/Kconfig pgcl-2.5.70-bk4-1/arch/i386/Kconfig --- linux-2.5.70-bk4/arch/i386/Kconfig 2003-05-26 18:00:24.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/Kconfig 2003-05-30 21:03:23.000000000 -0700 @@ -676,6 +676,18 @@ config X86_PAE depends on HIGHMEM64G default y +config PAGE_CLUSTER + int "Page clustering factor" + default 3 if HIGHMEM64G + default 2 if HIGHMEM4G + default 1 + help + Select page clustering factor as a power of 2. + Defaults and examples: + 3 => 32KB PAGE_SIZE + 2 => 16KB PAGE_SIZE + 1 => 8KB PAGE_SIZE + # Common NUMA Features config NUMA bool "Numa Memory Allocation Support" diff -prauN linux-2.5.70-bk4/arch/i386/boot/setup.S pgcl-2.5.70-bk4-1/arch/i386/boot/setup.S --- linux-2.5.70-bk4/arch/i386/boot/setup.S 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/boot/setup.S 2003-05-30 21:03:23.000000000 -0700 @@ -58,6 +58,9 @@ #include #include #include + +#define VMALLOC_START (-0xC0000000 - 128*1024*1024) +#include #include /* Signature words to ensure LILO loaded us right */ @@ -162,7 +165,7 @@ cmd_line_ptr: .long 0 # (Header versio # can be located anywhere in # low memory 0x10000 or higher. -ramdisk_max: .long MAXMEM-1 # (Header version 0x0203 or later) +ramdisk_max: .long __MAXMEM-1 # (Header version 0x0203 or later) # The highest safe address for # the contents of an initrd diff -prauN linux-2.5.70-bk4/arch/i386/kernel/apic.c pgcl-2.5.70-bk4-1/arch/i386/kernel/apic.c --- linux-2.5.70-bk4/arch/i386/kernel/apic.c 2003-05-26 18:01:02.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/apic.c 2003-05-30 21:03:23.000000000 -0700 @@ -679,7 +679,7 @@ void __init init_apic_mappings(void) * one for the IO-APIC. */ if (!smp_found_config && detect_init_APIC()) { - apic_phys = (unsigned long) alloc_bootmem_pages(PAGE_SIZE); + apic_phys = (unsigned long) alloc_bootmem_pages(MMUPAGE_SIZE); apic_phys = __pa(apic_phys); } else apic_phys = mp_lapic_addr; @@ -711,7 +711,7 @@ void __init init_apic_mappings(void) } } else { fake_ioapic_page: - ioapic_phys = (unsigned long) alloc_bootmem_pages(PAGE_SIZE); + ioapic_phys = (unsigned long) alloc_bootmem_pages(MMUPAGE_SIZE); ioapic_phys = __pa(ioapic_phys); } set_fixmap_nocache(idx, ioapic_phys); diff -prauN linux-2.5.70-bk4/arch/i386/kernel/cpu/amd.c pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/amd.c --- linux-2.5.70-bk4/arch/i386/kernel/cpu/amd.c 2003-05-26 18:00:59.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/amd.c 2003-05-30 21:03:23.000000000 -0700 @@ -25,7 +25,7 @@ __asm__(".align 4\nvide: ret"); static void __init init_amd(struct cpuinfo_x86 *c) { u32 l, h; - int mbytes = num_physpages >> (20-PAGE_SHIFT); + int mbytes = num_physpages >> (20-MMUPAGE_SHIFT); int r; /* diff -prauN linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/amd.c pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/amd.c --- linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/amd.c 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/amd.c 2003-05-30 21:03:23.000000000 -0700 @@ -16,7 +16,7 @@ amd_get_mtrr(unsigned int reg, unsigned if (reg == 1) low = high; /* The base masks off on the right alignment */ - *base = (low & 0xFFFE0000) >> PAGE_SHIFT; + *base = (low & 0xFFFE0000) >> MMUPAGE_SHIFT; *type = 0; if (low & 1) *type = MTRR_TYPE_UNCACHABLE; @@ -42,7 +42,7 @@ amd_get_mtrr(unsigned int reg, unsigned * *128K ... */ low = (~low) & 0x1FFFC; - *size = (low + 4) << (15 - PAGE_SHIFT); + *size = (low + 4) << (15 - MMUPAGE_SHIFT); return; } @@ -77,8 +77,8 @@ static void amd_set_mtrr(unsigned int re desired 111 1111 1111 1100 mask But ~(x - 1) == ~x + 1 == -x. Two's complement rocks! */ - regs[reg] = (-size >> (15 - PAGE_SHIFT) & 0x0001FFFC) - | (base << PAGE_SHIFT) | (type + 1); + regs[reg] = (-size >> (15 - MMUPAGE_SHIFT) & 0x0001FFFC) + | (base << MMUPAGE_SHIFT) | (type + 1); /* * The writeback rule is quite specific. See the manual. Its @@ -97,7 +97,7 @@ static int amd_validate_add_page(unsigne o Power of 2 block o base suitably aligned to the power */ - if (type > MTRR_TYPE_WRCOMB || size < (1 << (17 - PAGE_SHIFT)) + if (type > MTRR_TYPE_WRCOMB || size < (1 << (17 - MMUPAGE_SHIFT)) || (size & ~(size - 1)) - size || (base & (size - 1))) return -EINVAL; return 0; diff -prauN linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/centaur.c pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/centaur.c --- linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/centaur.c 2003-05-26 18:00:27.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/centaur.c 2003-05-30 21:03:23.000000000 -0700 @@ -51,8 +51,8 @@ static void centaur_get_mcr(unsigned int reg, unsigned long *base, unsigned int *size, mtrr_type * type) { - *base = centaur_mcr[reg].high >> PAGE_SHIFT; - *size = -(centaur_mcr[reg].low & 0xfffff000) >> PAGE_SHIFT; + *base = centaur_mcr[reg].high >> MMUPAGE_SHIFT; + *size = -(centaur_mcr[reg].low & 0xfffff000) >> MMUPAGE_SHIFT; *type = MTRR_TYPE_WRCOMB; /* If it is there, it is write-combining */ if (centaur_mcr_type == 1 && ((centaur_mcr[reg].low & 31) & 2)) *type = MTRR_TYPE_UNCACHABLE; @@ -72,14 +72,14 @@ static void centaur_set_mcr(unsigned int /* Disable */ high = low = 0; } else { - high = base << PAGE_SHIFT; + high = base << MMUPAGE_SHIFT; if (centaur_mcr_type == 0) - low = -size << PAGE_SHIFT | 0x1f; /* only support write-combining... */ + low = -size << MMUPAGE_SHIFT | 0x1f; /* only support write-combining... */ else { if (type == MTRR_TYPE_UNCACHABLE) - low = -size << PAGE_SHIFT | 0x02; /* NC */ + low = -size << MMUPAGE_SHIFT | 0x02; /* NC */ else - low = -size << PAGE_SHIFT | 0x09; /* WWO,WC */ + low = -size << MMUPAGE_SHIFT | 0x09; /* WWO,WC */ } } centaur_mcr[reg].high = high; diff -prauN linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/cyrix.c pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/cyrix.c --- linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/cyrix.c 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/cyrix.c 2003-05-30 21:03:23.000000000 -0700 @@ -30,7 +30,7 @@ cyrix_get_arr(unsigned int reg, unsigned /* Enable interrupts if it was enabled previously */ local_irq_restore(flags); shift = ((unsigned char *) base)[1] & 0x0f; - *base >>= PAGE_SHIFT; + *base >>= MMUPAGE_SHIFT; /* Power of two, at least 4K on ARR0-ARR6, 256K on ARR7 * Note: shift==0xf means 4G, this is unsupported. @@ -203,7 +203,7 @@ static void cyrix_set_arr(unsigned int r prepare_set(); - base <<= PAGE_SHIFT; + base <<= MMUPAGE_SHIFT; setCx86(arr, ((unsigned char *) &base)[3]); setCx86(arr + 1, ((unsigned char *) &base)[2]); setCx86(arr + 2, (((unsigned char *) &base)[1]) | arr_size); diff -prauN linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/generic.c pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/generic.c --- linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/generic.c 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/generic.c 2003-05-30 21:03:23.000000000 -0700 @@ -131,13 +131,13 @@ void generic_get_mtrr(unsigned int reg, rdmsr(MTRRphysBase_MSR(reg), base_lo, base_hi); /* Work out the shifted address mask. */ - mask_lo = size_or_mask | mask_hi << (32 - PAGE_SHIFT) - | mask_lo >> PAGE_SHIFT; + mask_lo = size_or_mask | mask_hi << (32 - MMUPAGE_SHIFT) + | mask_lo >> MMUPAGE_SHIFT; /* This works correctly if size is a power of two, i.e. a contiguous range. */ *size = -mask_lo; - *base = base_hi << (32 - PAGE_SHIFT) | base_lo >> PAGE_SHIFT; + *base = base_hi << (32 - MMUPAGE_SHIFT) | base_lo >> MMUPAGE_SHIFT; *type = base_lo & 0xff; } @@ -317,10 +317,10 @@ static void generic_set_mtrr(unsigned in relevant mask register to disable a range. */ wrmsr(MTRRphysMask_MSR(reg), 0, 0); } else { - wrmsr(MTRRphysBase_MSR(reg), base << PAGE_SHIFT | type, - (base & size_and_mask) >> (32 - PAGE_SHIFT)); - wrmsr(MTRRphysMask_MSR(reg), -size << PAGE_SHIFT | 0x800, - (-size & size_and_mask) >> (32 - PAGE_SHIFT)); + wrmsr(MTRRphysBase_MSR(reg), base << MMUPAGE_SHIFT | type, + (base & size_and_mask) >> (32 - MMUPAGE_SHIFT)); + wrmsr(MTRRphysMask_MSR(reg), -size << MMUPAGE_SHIFT | 0x800, + (-size & size_and_mask) >> (32 - MMUPAGE_SHIFT)); } post_set(); @@ -335,7 +335,7 @@ int generic_validate_add_page(unsigned l if (is_cpu(INTEL) && boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 1 && boot_cpu_data.x86_mask <= 7) { - if (base & ((1 << (22 - PAGE_SHIFT)) - 1)) { + if (base & ((1 << (22 - MMUPAGE_SHIFT)) - 1)) { printk(KERN_WARNING "mtrr: base(0x%lx000) is not 4 MiB aligned\n", base); return -EINVAL; } diff -prauN linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/if.c pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/if.c --- linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/if.c 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/if.c 2003-05-30 21:03:23.000000000 -0700 @@ -33,10 +33,10 @@ mtrr_file_add(unsigned long base, unsign FILE_FCOUNT(file) = fcount; } if (!page) { - if ((base & (PAGE_SIZE - 1)) || (size & (PAGE_SIZE - 1))) + if ((base & (MMUPAGE_SIZE - 1)) || (size & (MMUPAGE_SIZE - 1))) return -EINVAL; - base >>= PAGE_SHIFT; - size >>= PAGE_SHIFT; + base >>= MMUPAGE_SHIFT; + size >>= MMUPAGE_SHIFT; } reg = mtrr_add_page(base, size, type, 1); if (reg >= 0) @@ -52,10 +52,10 @@ mtrr_file_del(unsigned long base, unsign unsigned int *fcount = FILE_FCOUNT(file); if (!page) { - if ((base & (PAGE_SIZE - 1)) || (size & (PAGE_SIZE - 1))) + if ((base & (MMUPAGE_SIZE - 1)) || (size & (MMUPAGE_SIZE - 1))) return -EINVAL; - base >>= PAGE_SHIFT; - size >>= PAGE_SHIFT; + base >>= MMUPAGE_SHIFT; + size >>= MMUPAGE_SHIFT; } reg = mtrr_del_page(-1, base, size); if (reg < 0) @@ -119,8 +119,8 @@ mtrr_write(struct file *file, const char for (i = 0; i < MTRR_NUM_TYPES; ++i) { if (strcmp(ptr, mtrr_strings[i])) continue; - base >>= PAGE_SHIFT; - size >>= PAGE_SHIFT; + base >>= MMUPAGE_SHIFT; + size >>= MMUPAGE_SHIFT; err = mtrr_add_page((unsigned long) base, (unsigned long) size, i, 1); @@ -194,8 +194,8 @@ mtrr_ioctl(struct inode *inode, struct f || gentry.size == 0x100000) gentry.base = gentry.size = gentry.type = 0; else { - gentry.base <<= PAGE_SHIFT; - gentry.size <<= PAGE_SHIFT; + gentry.base <<= MMUPAGE_SHIFT; + gentry.size <<= MMUPAGE_SHIFT; gentry.type = type; } @@ -320,18 +320,18 @@ static int mtrr_seq_show(struct seq_file if (size == 0) usage_table[i] = 0; else { - if (size < (0x100000 >> PAGE_SHIFT)) { + if (size < (0x100000 >> MMUPAGE_SHIFT)) { /* less than 1MB */ factor = 'K'; - size <<= PAGE_SHIFT - 10; + size <<= MMUPAGE_SHIFT - 10; } else { factor = 'M'; - size >>= 20 - PAGE_SHIFT; + size >>= 20 - MMUPAGE_SHIFT; } /* RED-PEN: base can be > 32bit */ len += seq_printf(seq, "reg%02i: base=0x%05lx000 (%4liMB), size=%4i%cB: %s, count=%d\n", - i, base, base >> (20 - PAGE_SHIFT), size, factor, + i, base, base >> (20 - MMUPAGE_SHIFT), size, factor, attrib_to_str(type), usage_table[i]); } } diff -prauN linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/main.c pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/main.c --- linux-2.5.70-bk4/arch/i386/kernel/cpu/mtrr/main.c 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/cpu/mtrr/main.c 2003-05-30 21:03:23.000000000 -0700 @@ -410,12 +410,12 @@ int mtrr_add(unsigned long base, unsigned long size, unsigned int type, char increment) { - if ((base & (PAGE_SIZE - 1)) || (size & (PAGE_SIZE - 1))) { + if ((base & (MMUPAGE_SIZE - 1)) || (size & (MMUPAGE_SIZE - 1))) { printk("mtrr: size and base must be multiples of 4 kiB\n"); printk("mtrr: size: 0x%lx base: 0x%lx\n", size, base); return -EINVAL; } - return mtrr_add_page(base >> PAGE_SHIFT, size >> PAGE_SHIFT, type, + return mtrr_add_page(base >> MMUPAGE_SHIFT, size >> MMUPAGE_SHIFT, type, increment); } @@ -506,12 +506,12 @@ int mtrr_del_page(int reg, unsigned long int mtrr_del(int reg, unsigned long base, unsigned long size) { - if ((base & (PAGE_SIZE - 1)) || (size & (PAGE_SIZE - 1))) { + if ((base & (MMUPAGE_SIZE - 1)) || (size & (MMUPAGE_SIZE - 1))) { printk("mtrr: size and base must be multiples of 4 kiB\n"); printk("mtrr: size: 0x%lx base: 0x%lx\n", size, base); return -EINVAL; } - return mtrr_del_page(reg, base >> PAGE_SHIFT, size >> PAGE_SHIFT); + return mtrr_del_page(reg, base >> MMUPAGE_SHIFT, size >> MMUPAGE_SHIFT); } EXPORT_SYMBOL(mtrr_add); @@ -579,7 +579,7 @@ static int __init mtrr_init(void) u32 phys_addr; phys_addr = cpuid_eax(0x80000008) & 0xff; size_or_mask = - ~((1 << (phys_addr - PAGE_SHIFT)) - 1); + ~((1 << (phys_addr - MMUPAGE_SHIFT)) - 1); size_and_mask = ~size_or_mask & 0xfff00000; } /* Athlon MTRRs use an Intel-compatible interface for diff -prauN linux-2.5.70-bk4/arch/i386/kernel/entry.S pgcl-2.5.70-bk4-1/arch/i386/kernel/entry.S --- linux-2.5.70-bk4/arch/i386/kernel/entry.S 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/entry.S 2003-05-30 21:03:23.000000000 -0700 @@ -160,7 +160,7 @@ do_lcall: movl %eax,EFLAGS(%ebp) # movl %edx,EIP(%ebp) # Now we move them to their "normal" places movl %ecx,CS(%ebp) # - andl $-8192, %ebp # GET_THREAD_INFO + andl $~(THREAD_SIZE-1), %ebp # GET_THREAD_INFO movl TI_EXEC_DOMAIN(%ebp), %edx # Get the execution domain call *4(%edx) # Call the lcall7 handler for the domain addl $4, %esp diff -prauN linux-2.5.70-bk4/arch/i386/kernel/head.S pgcl-2.5.70-bk4-1/arch/i386/kernel/head.S --- linux-2.5.70-bk4/arch/i386/kernel/head.S 2003-05-26 18:00:23.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/head.S 2003-05-30 21:03:23.000000000 -0700 @@ -16,6 +16,7 @@ #include #include #include +#include #define OLD_CL_MAGIC_ADDR 0x90020 #define OLD_CL_MAGIC 0xA33F @@ -325,7 +326,7 @@ rp_sidt: ret ENTRY(stack_start) - .long init_thread_union+8192 + .long init_thread_union+THREAD_SIZE .long __BOOT_DS /* This is the default interrupt "handler" :-) */ diff -prauN linux-2.5.70-bk4/arch/i386/kernel/microcode.c pgcl-2.5.70-bk4-1/arch/i386/kernel/microcode.c --- linux-2.5.70-bk4/arch/i386/kernel/microcode.c 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/microcode.c 2003-05-30 21:03:23.000000000 -0700 @@ -319,7 +319,7 @@ static ssize_t microcode_write(struct fi sizeof(struct microcode)); return -EINVAL; } - if ((len >> PAGE_SHIFT) > num_physpages) { + if ((len >> MMUPAGE_SHIFT) > num_physpages) { printk(KERN_ERR "microcode: too much data (max %ld pages)\n", num_physpages); return -EINVAL; } diff -prauN linux-2.5.70-bk4/arch/i386/kernel/mpparse.c pgcl-2.5.70-bk4-1/arch/i386/kernel/mpparse.c --- linux-2.5.70-bk4/arch/i386/kernel/mpparse.c 2003-05-26 18:00:27.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/mpparse.c 2003-05-30 21:03:23.000000000 -0700 @@ -710,7 +710,7 @@ static int __init smp_scan_config (unsig smp_found_config = 1; printk(KERN_INFO "found SMP MP-table at %08lx\n", virt_to_phys(mpf)); - reserve_bootmem(virt_to_phys(mpf), PAGE_SIZE); + reserve_bootmem(virt_to_phys(mpf), MMUPAGE_SIZE); if (mpf->mpf_physptr) { /* * We cannot access to MPC table to compute @@ -721,8 +721,8 @@ static int __init smp_scan_config (unsig * PAGE_SIZE from mpg->mpf_physptr yields BUG() * in reserve_bootmem. */ - unsigned long size = PAGE_SIZE; - unsigned long end = max_low_pfn * PAGE_SIZE; + unsigned long size = MMUPAGE_SIZE; + unsigned long end = max_low_pfn * MMUPAGE_SIZE; if (mpf->mpf_physptr + size > end) size = end - mpf->mpf_physptr; reserve_bootmem(mpf->mpf_physptr, size); diff -prauN linux-2.5.70-bk4/arch/i386/kernel/numaq.c pgcl-2.5.70-bk4-1/arch/i386/kernel/numaq.c --- linux-2.5.70-bk4/arch/i386/kernel/numaq.c 2003-05-26 18:00:45.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/numaq.c 2003-05-30 21:03:23.000000000 -0700 @@ -33,7 +33,7 @@ /* These are needed before the pgdat's are created */ extern long node_start_pfn[], node_end_pfn[]; -#define MB_TO_PAGES(addr) ((addr) << (20 - PAGE_SHIFT)) +#define MB_TO_PAGES(addr) ((addr) << (20 - MMUPAGE_SHIFT)) /* * Function: smp_dump_qct() diff -prauN linux-2.5.70-bk4/arch/i386/kernel/setup.c pgcl-2.5.70-bk4-1/arch/i386/kernel/setup.c --- linux-2.5.70-bk4/arch/i386/kernel/setup.c 2003-05-30 21:02:23.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/setup.c 2003-05-30 21:03:23.000000000 -0700 @@ -557,6 +557,8 @@ void __init find_max_pfn(void) continue; if (end > max_pfn) max_pfn = end; + + max_pfn &= ~(PAGE_MMUCOUNT - 1); } } @@ -567,6 +569,8 @@ unsigned long __init find_max_low_pfn(vo { unsigned long max_low_pfn; + printk("MAXMEM = %p\n", (void *)MAXMEM); + max_low_pfn = max_pfn; if (max_low_pfn > MAXMEM_PFN) { if (highmem_pages == -1) @@ -680,10 +684,10 @@ static unsigned long __init setup_memory highstart_pfn = max_low_pfn; } printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", - pages_to_mb(highend_pfn - highstart_pfn)); + (highend_pfn - highstart_pfn) >> (20 - MMUPAGE_SHIFT)); #endif printk(KERN_NOTICE "%ldMB LOWMEM available.\n", - pages_to_mb(max_low_pfn)); + max_low_pfn >> (20 - MMUPAGE_SHIFT)); /* * Initialize the boot-time allocator (with low memory only): */ @@ -704,7 +708,7 @@ static unsigned long __init setup_memory * reserve physical page 0 - it's a special BIOS page on many boxes, * enabling clean reboots, SMP operation, laptop functions. */ - reserve_bootmem(0, PAGE_SIZE); + reserve_bootmem(0, MMUPAGE_SIZE); #ifdef CONFIG_SMP /* @@ -712,7 +716,7 @@ static unsigned long __init setup_memory * FIXME: Don't need the extra page at 4K, but need to fix * trampoline before removing it. (see the GDT stuff) */ - reserve_bootmem(PAGE_SIZE, PAGE_SIZE); + reserve_bootmem(MMUPAGE_SIZE, MMUPAGE_SIZE); #endif #ifdef CONFIG_ACPI_SLEEP /* @@ -729,7 +733,7 @@ static unsigned long __init setup_memory #ifdef CONFIG_BLK_DEV_INITRD if (LOADER_TYPE && INITRD_START) { - if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) { + if (INITRD_START + INITRD_SIZE <= PFN_PHYS(max_low_pfn)) { reserve_bootmem(INITRD_START, INITRD_SIZE); initrd_start = INITRD_START ? INITRD_START + PAGE_OFFSET : 0; @@ -739,7 +743,7 @@ static unsigned long __init setup_memory printk(KERN_ERR "initrd extends beyond end of memory " "(0x%08lx > 0x%08lx)\ndisabling initrd\n", INITRD_START + INITRD_SIZE, - max_low_pfn << PAGE_SHIFT); + PFN_PHYS(max_low_pfn)); initrd_start = 0; } } @@ -793,7 +797,7 @@ static void __init register_memory(unsig request_resource(&ioport_resource, standard_io_resources+i); /* Tell the PCI layer not to allocate too close to the RAM area.. */ - low_mem_size = ((max_low_pfn << PAGE_SHIFT) + 0xfffff) & ~0xfffff; + low_mem_size = ((max_low_pfn << MMUPAGE_SHIFT) + 0xfffff) & ~0xfffff; if (low_mem_size > pci_mem_start) pci_mem_start = low_mem_size; } diff -prauN linux-2.5.70-bk4/arch/i386/kernel/smpboot.c pgcl-2.5.70-bk4-1/arch/i386/kernel/smpboot.c --- linux-2.5.70-bk4/arch/i386/kernel/smpboot.c 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/smpboot.c 2003-05-30 21:03:23.000000000 -0700 @@ -100,7 +100,7 @@ static unsigned long __init setup_trampo */ void __init smp_alloc_memory(void) { - trampoline_base = (void *) alloc_bootmem_low_pages(PAGE_SIZE); + trampoline_base = (void *) alloc_bootmem_low_pages(MMUPAGE_SIZE); /* * Has to be in very low memory so we can execute * real-mode AP code. diff -prauN linux-2.5.70-bk4/arch/i386/kernel/sys_i386.c pgcl-2.5.70-bk4-1/arch/i386/kernel/sys_i386.c --- linux-2.5.70-bk4/arch/i386/kernel/sys_i386.c 2003-05-26 18:00:59.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/sys_i386.c 2003-05-30 21:03:23.000000000 -0700 @@ -97,10 +97,10 @@ asmlinkage int old_mmap(struct mmap_arg_ goto out; err = -EINVAL; - if (a.offset & ~PAGE_MASK) + if (a.offset & ~MMUPAGE_MASK) goto out; - err = do_mmap2(a.addr, a.len, a.prot, a.flags, a.fd, a.offset >> PAGE_SHIFT); + err = do_mmap2(a.addr, a.len, a.prot, a.flags, a.fd, a.offset >> MMUPAGE_SHIFT); out: return err; } diff -prauN linux-2.5.70-bk4/arch/i386/kernel/traps.c pgcl-2.5.70-bk4-1/arch/i386/kernel/traps.c --- linux-2.5.70-bk4/arch/i386/kernel/traps.c 2003-05-26 18:00:25.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/kernel/traps.c 2003-05-30 21:03:23.000000000 -0700 @@ -120,7 +120,7 @@ void show_trace_task(struct task_struct unsigned long esp = tsk->thread.esp; /* User space on another CPU? */ - if ((esp ^ (unsigned long)tsk->thread_info) & (PAGE_MASK<<1)) + if ((esp ^ (unsigned long)tsk->thread_info) & ~(THREAD_SIZE-1)) return; show_trace((unsigned long *)esp); } @@ -431,6 +431,7 @@ static void unknown_nmi_error(unsigned c reason, smp_processor_id()); printk("Dazed and confused, but trying to continue\n"); printk("Do you have a strange power saving mode enabled?\n"); + dump_stack(); } static void default_do_nmi(struct pt_regs * regs) diff -prauN linux-2.5.70-bk4/arch/i386/lib/getuser.S pgcl-2.5.70-bk4-1/arch/i386/lib/getuser.S --- linux-2.5.70-bk4/arch/i386/lib/getuser.S 2003-05-26 18:00:20.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/lib/getuser.S 2003-05-30 21:03:23.000000000 -0700 @@ -8,9 +8,9 @@ * return an error value in addition to the "real" * return value. */ +#include #include - /* * __get_user_X * diff -prauN linux-2.5.70-bk4/arch/i386/mm/discontig.c pgcl-2.5.70-bk4-1/arch/i386/mm/discontig.c --- linux-2.5.70-bk4/arch/i386/mm/discontig.c 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/mm/discontig.c 2003-05-30 21:03:23.000000000 -0700 @@ -71,8 +71,6 @@ extern unsigned long max_low_pfn; extern unsigned long totalram_pages; extern unsigned long totalhigh_pages; -#define LARGE_PAGE_BYTES (PTRS_PER_PTE * PAGE_SIZE) - unsigned long node_remap_start_pfn[MAX_NUMNODES]; unsigned long node_remap_size[MAX_NUMNODES]; unsigned long node_remap_offset[MAX_NUMNODES]; @@ -129,7 +127,7 @@ static void __init allocate_pgdat(int ni if (nid) NODE_DATA(nid) = (pg_data_t *)node_remap_start_vaddr[nid]; else { - NODE_DATA(nid) = (pg_data_t *)(__va(min_low_pfn << PAGE_SHIFT)); + NODE_DATA(nid) = (pg_data_t *)(__va(min_low_pfn*MMUPAGE_SIZE)); min_low_pfn += PFN_UP(sizeof(pg_data_t)); memset(NODE_DATA(nid), 0, sizeof(pg_data_t)); } @@ -182,8 +180,8 @@ void __init remap_numa_kva(void) int node; for (node = 1; node < numnodes; ++node) { - for (pfn=0; pfn < node_remap_size[node]; pfn += PTRS_PER_PTE) { - vaddr = node_remap_start_vaddr[node]+(pfn< system_max_low_pfn) highstart_pfn = system_max_low_pfn; printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", - pages_to_mb(highend_pfn - highstart_pfn)); + (highend_pfn - highstart_pfn) >> (20 - MMUPAGE_SHIFT)); #endif system_max_low_pfn = max_low_pfn = max_low_pfn - reserve_pages; printk(KERN_NOTICE "%ldMB LOWMEM available.\n", - pages_to_mb(system_max_low_pfn)); + system_max_low_pfn >> (20 - MMUPAGE_SHIFT)); printk("min_low_pfn = %ld, max_low_pfn = %ld, highstart_pfn = %ld\n", min_low_pfn, max_low_pfn, highstart_pfn); @@ -261,6 +260,11 @@ unsigned long __init setup_memory(void) (ulong) pfn_to_kaddr(highstart_pfn)); for (nid = 0; nid < numnodes; nid++) find_max_pfn_node(nid); + printk("vmallocspace = [0x%lx, 0x%lx)\n", + VMALLOC_START, VMALLOC_END); + printk("fixmapspace = [0x%lx, 0x%lx)\n", + FIXADDR_START, FIXADDR_TOP); + printk("MAXMEM = 0x%lx\n", MAXMEM); NODE_DATA(0)->bdata = &node0_bdata; @@ -277,21 +281,21 @@ unsigned long __init setup_memory(void) * the (very unlikely) case of us accidentally initializing the * bootmem allocator with an invalid RAM area. */ - reserve_bootmem_node(NODE_DATA(0), HIGH_MEMORY, (PFN_PHYS(min_low_pfn) + - bootmap_size + PAGE_SIZE-1) - (HIGH_MEMORY)); + reserve_bootmem_node(NODE_DATA(0), HIGH_MEMORY, PFN_PHYS(min_low_pfn) + + bootmap_size - HIGH_MEMORY); /* * reserve physical page 0 - it's a special BIOS page on many boxes, * enabling clean reboots, SMP operation, laptop functions. */ - reserve_bootmem_node(NODE_DATA(0), 0, PAGE_SIZE); + reserve_bootmem_node(NODE_DATA(0), 0, MMUPAGE_SIZE); /* * But first pinch a few for the stack/trampoline stuff * FIXME: Don't need the extra page at 4K, but need to fix * trampoline before removing it. (see the GDT stuff) */ - reserve_bootmem_node(NODE_DATA(0), PAGE_SIZE, PAGE_SIZE); + reserve_bootmem_node(NODE_DATA(0), MMUPAGE_SIZE, MMUPAGE_SIZE); #ifdef CONFIG_ACPI_SLEEP /* @@ -307,7 +311,7 @@ unsigned long __init setup_memory(void) #ifdef CONFIG_BLK_DEV_INITRD if (LOADER_TYPE && INITRD_START) { - if (INITRD_START + INITRD_SIZE <= (system_max_low_pfn << PAGE_SHIFT)) { + if (INITRD_START + INITRD_SIZE <= (system_max_low_pfn << MMUPAGE_SHIFT)) { reserve_bootmem_node(NODE_DATA(0), INITRD_START, INITRD_SIZE); initrd_start = INITRD_START ? INITRD_START + PAGE_OFFSET : 0; @@ -317,7 +321,7 @@ unsigned long __init setup_memory(void) printk(KERN_ERR "initrd extends beyond end of memory " "(0x%08lx > 0x%08lx)\ndisabling initrd\n", INITRD_START + INITRD_SIZE, - system_max_low_pfn << PAGE_SHIFT); + system_max_low_pfn << MMUPAGE_SHIFT); initrd_start = 0; } } @@ -350,20 +354,20 @@ void __init zone_sizes_init(void) unsigned long start = node_start_pfn[nid]; unsigned long high = node_end_pfn[nid]; - max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; + max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> MMUPAGE_SHIFT; if (start > low) { #ifdef CONFIG_HIGHMEM - zones_size[ZONE_HIGHMEM] = high - start; + zones_size[ZONE_HIGHMEM] = (high - start) >> PAGE_MMUSHIFT; #endif } else { if (low < max_dma) - zones_size[ZONE_DMA] = low; + zones_size[ZONE_DMA] = low >> PAGE_MMUSHIFT; else { - zones_size[ZONE_DMA] = max_dma; - zones_size[ZONE_NORMAL] = low - max_dma; + zones_size[ZONE_DMA] = max_dma >> PAGE_MMUSHIFT; + zones_size[ZONE_NORMAL] = (low - max_dma) >> PAGE_MMUSHIFT; #ifdef CONFIG_HIGHMEM - zones_size[ZONE_HIGHMEM] = high - low; + zones_size[ZONE_HIGHMEM] = (high - low) >> PAGE_MMUSHIFT; #endif } } @@ -403,10 +407,14 @@ void __init set_highmem_pages_init(int b zone_start_pfn = NODE_DATA(nid)->node_zones[ZONE_HIGHMEM].zone_start_pfn; printk("Initializing highpages for node %d\n", nid); - for (node_pfn = 0; node_pfn < node_high_size; node_pfn++) { - one_highpage_init((struct page *)(zone_mem_map + node_pfn), - zone_start_pfn + node_pfn, bad_ppro); - } + + /* + * Note: zone->spanned_pages is in PAGE_SIZE units. + */ + for (node_pfn = 0; node_pfn < node_high_size; node_pfn++) + one_highpage_init(&zone_mem_map[node_pfn], + zone_start_pfn + node_pfn*PAGE_MMUCOUNT, + bad_ppro); } totalram_pages += totalhigh_pages; #endif diff -prauN linux-2.5.70-bk4/arch/i386/mm/fault.c pgcl-2.5.70-bk4-1/arch/i386/mm/fault.c --- linux-2.5.70-bk4/arch/i386/mm/fault.c 2003-05-26 18:00:20.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/mm/fault.c 2003-05-30 21:03:23.000000000 -0700 @@ -20,6 +20,8 @@ #include #include /* For unblank_screen() */ #include +#include /* for max_low_pfn */ +#include #include #include @@ -73,19 +75,22 @@ asmlinkage void do_page_fault(struct pt_ struct mm_struct *mm; struct vm_area_struct * vma; unsigned long address; - unsigned long page; int write; siginfo_t info; /* get the address */ __asm__("movl %%cr2,%0":"=r" (address)); + pr_debug("faulted on %lx,", address); + /* It's safe to allow irq's after cr2 has been saved */ if (regs->eflags & X86_EFLAGS_IF) local_irq_enable(); tsk = current; + pr_debug(" pid = %d\n", current->pid); + /* * We fault-in kernel-space virtual memory on-demand. The * 'reference' page table is init_mm.pgd. @@ -104,7 +109,20 @@ asmlinkage void do_page_fault(struct pt_ mm = tsk->mm; info.si_code = SEGV_MAPERR; - + if (1) { + pgd_t *pgd = pgd_offset(mm, address); + pmd_t *pmd = pmd_offset(pgd, address); + pr_debug("fault handled by PGD at vaddr %p, %Lx\n", + pgd, (u64)pgd_val(*pgd)); + pr_debug("fault handled by PMD at vaddr %p, %Lx\n", + pmd, (u64)pmd_val(*pmd)); + if (pmd_present(*pmd)) { + pr_debug("fault will be handled by PTE at paddr %Lx\n", + (u64)(pmd_val(*pmd) & MMUPAGE_MASK) + + pte_index(address)*sizeof(pte_t)); + } else + pr_debug("pmd not present\n"); + } /* * If we're in an interrupt, have no user context or are running in an * atomic region then we must not take the fault.. @@ -115,12 +133,16 @@ asmlinkage void do_page_fault(struct pt_ down_read(&mm->mmap_sem); vma = find_vma(mm, address); - if (!vma) + if (!vma) { + pr_debug("no vma, goto bad_area\n"); goto bad_area; + } if (vma->vm_start <= address) goto good_area; - if (!(vma->vm_flags & VM_GROWSDOWN)) + if (!(vma->vm_flags & VM_GROWSDOWN)) { + pr_debug("VM_GROWSDOWN not in vma->vm_flags, goto bad_area\n"); goto bad_area; + } if (error_code & 4) { /* * accessing the stack below %esp is always a bug. @@ -128,11 +150,15 @@ asmlinkage void do_page_fault(struct pt_ * pusha) doing post-decrement on the stack and that * doesn't show up until later.. */ - if (address + 32 < regs->esp) + if (address + 32 < regs->esp) { + pr_debug("postdecrement on stack, goto bad_area\n"); goto bad_area; } - if (expand_stack(vma, address)) + } + if (expand_stack(vma, address)) { + pr_debug("expand_stack() failed, goto bad_area\n"); goto bad_area; + } /* * Ok, we have a good vm_area for this memory access, so * we can handle it.. @@ -144,20 +170,25 @@ good_area: default: /* 3: write, present */ #ifdef TEST_VERIFY_AREA if (regs->cs == KERNEL_CS) - printk("WP fault at %08lx\n", regs->eip); + pr_debug("WP fault at %08lx\n", regs->eip); #endif /* fall through */ case 2: /* write, not present */ - if (!(vma->vm_flags & VM_WRITE)) + if (!(vma->vm_flags & VM_WRITE)) { + pr_debug("vma not writable, goto bad_area\n"); goto bad_area; + } write++; break; case 1: /* read, present */ + pr_debug("NFI what happened, goto bad_area\n"); goto bad_area; case 0: /* read, not present */ - if (!(vma->vm_flags & (VM_READ | VM_EXEC))) + if (!(vma->vm_flags & (VM_READ | VM_EXEC))) { + pr_debug("vma not read/exec, goto bad_area\n"); goto bad_area; } + } survive: /* @@ -184,7 +215,7 @@ good_area: * Did it hit the DOS screen memory VA from vm86 mode? */ if (regs->eflags & VM_MASK) { - unsigned long bit = (address - 0xA0000) >> PAGE_SHIFT; + unsigned long bit = (address - 0xA0000) >> MMUPAGE_SHIFT; if (bit < 32) tsk->thread.screen_bitmap |= 1 << bit; } @@ -200,6 +231,45 @@ bad_area: /* User mode accesses just cause a SIGSEGV */ if (error_code & 4) { + printk("user mode SIGSEGV, pid = %d, comm = %16s, EIP = %p, ESP = %p, CR2 = %p\n", + current->pid, current->comm, (void *)regs->eip, (void *)regs->esp, (void *)address); + spin_lock(&mm->page_table_lock); + for (vma = mm->mmap; vma; vma = vma->vm_next) { + unsigned long addr; + + printk("vma = [%lx,%lx) prot=%lx flags=%lx\n", + vma->vm_start, vma->vm_end, + vma->vm_page_prot.pgprot, vma->vm_flags); + + for (addr = vma->vm_start; addr < vma->vm_end; addr += MMUPAGE_SIZE) { + pgd_t *pgd = pgd_offset(mm, addr); + pmd_t *pmd; + pte_t *pte; + struct page *page; + void *mem; + + if (pgd_none(*pgd) || pgd_bad(*pgd)) + continue; + + pmd = pmd_offset(pgd, addr); + if (pmd_none(*pmd) || pmd_bad(*pmd)) + continue; + + pte = pte_offset_map(pmd, addr); + if (pte_none(*pte) || !pte_present(*pte) || + !pfn_valid(pte_pfn(*pte))) { + pte_unmap(pte); + continue; + } + page = pte_page(*pte); + mem = kmap_atomic(page, KM_USER0); + if (!memcmp(mem, page_address(ZERO_PAGE(0)), PAGE_SIZE)) + printk("page at 0x%lx zero!\n", addr); + kunmap_atomic(mem, KM_USER0); + pte_unmap(pte); + } + } + spin_unlock(&mm->page_table_lock); tsk->thread.cr2 = address; tsk->thread.error_code = error_code; tsk->thread.trap_no = 14; @@ -207,6 +277,13 @@ bad_area: info.si_errno = 0; /* info.si_code has been set above */ info.si_addr = (void *)address; +#if 0 + if (current->pid >= 1024) { + while (1) { + schedule_timeout(HZ); + } + } +#endif force_sig_info(SIGSEGV, &info, tsk); return; } @@ -239,30 +316,53 @@ no_context: bust_spinlocks(1); - if (address < PAGE_SIZE) + if (address < MMUPAGE_SIZE) printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference"); else printk(KERN_ALERT "Unable to handle kernel paging request"); printk(" at virtual address %08lx\n",address); printk(" printing eip:\n"); printk("%08lx\n", regs->eip); - asm("movl %%cr3,%0":"=r" (page)); - page = ((unsigned long *) __va(page))[address >> 22]; - printk(KERN_ALERT "*pde = %08lx\n", page); - /* - * We must not directly access the pte in the highpte - * case, the page table might be allocated in highmem. - * And lets rather not kmap-atomic the pte, just in case - * it's allocated already. - */ -#ifndef CONFIG_HIGHPTE - if (page & 1) { - page &= PAGE_MASK; - address &= 0x003ff000; - page = ((unsigned long *) __va(page))[address >> PAGE_SHIFT]; - printk(KERN_ALERT "*pte = %08lx\n", page); + { + unsigned long cr3; + pgd_t *pgd; + pmd_t *pmd; + char *fmt; + + asm("movl %%cr3,%0":"=r" (cr3)); + cr3 &= ~0x1f; /* lower 5 bits of %cr3 are flags */ + /* pgd's in lowmem, but only need to be < 4G (32-bit %cr3) */ + pgd = (pgd_t *)__va(cr3); + fmt = PTRS_PER_PMD > 1 ? KERN_ALERT "*pdpte = %Lx\n" : NULL; + if (PTRS_PER_PMD > 1) + printk(fmt, pgd_val(*pgd)); + + /* pmd's in lowmem, but can be anywhere (64-bit PDPTE) */ + pmd = pmd_offset(pgd, address); + if (PTRS_PER_PMD > 1) + fmt = KERN_ALERT "*pde = %Lx\n"; + else + fmt = KERN_ALERT "*pde = %08lx\n"; + printk(fmt, pmd_val(*pmd)); + + /* + * this is getting at what are potentially user + * PTE's with pte_offset_kernel(); it's mostly + * unsafe to try editing kernel PTE's at this + * point for kmap_atomic() so just drop out of it + * if pmd_val(*pmd)/MMUPAGE_SIZE > max_low_pfn + */ + + if (pmd_present(*pmd) && !pmd_large(*pmd) + && pmd_val(*pmd)/MMUPAGE_SIZE <= max_low_pfn) { + pte_t *pte = pte_offset_kernel(pmd, address); + if (PTRS_PER_PMD > 1) + fmt = KERN_ALERT "*pte = %Lx\n"; + else + fmt = KERN_ALERT "*pte = %08lx\n"; + printk(fmt, pte_val(*pte)); + } } -#endif die("Oops", regs, error_code); bust_spinlocks(0); do_exit(SIGKILL); @@ -290,6 +390,7 @@ do_sigbus: * Send a sigbus, regardless of whether we were in kernel * or user mode. */ + pr_debug("sending SIGBUS\n"); tsk->thread.cr2 = address; tsk->thread.error_code = error_code; tsk->thread.trap_no = 14; @@ -318,12 +419,16 @@ vmalloc_fault: pmd_t *pmd, *pmd_k; pte_t *pte_k; + pr_debug("took vmalloc fault on address %lx\n", address); + asm("movl %%cr3,%0":"=r" (pgd)); pgd = index + (pgd_t *)__va(pgd); pgd_k = init_mm.pgd + index; - if (!pgd_present(*pgd_k)) + if (!pgd_present(*pgd_k)) { + printk("missing pgd in vmalloc fault!\n"); goto no_context; + } /* * set_pgd(pgd, *pgd_k); here would be useless on PAE @@ -332,13 +437,17 @@ vmalloc_fault: pmd = pmd_offset(pgd, address); pmd_k = pmd_offset(pgd_k, address); - if (!pmd_present(*pmd_k)) + if (!pmd_present(*pmd_k)) { + printk("missing pmd in vmalloc fault!\n"); goto no_context; + } set_pmd(pmd, *pmd_k); pte_k = pte_offset_kernel(pmd_k, address); - if (!pte_present(*pte_k)) + if (!pte_present(*pte_k)) { + printk("missing pte in vmalloc fault!\n"); goto no_context; + } return; } } diff -prauN linux-2.5.70-bk4/arch/i386/mm/highmem.c pgcl-2.5.70-bk4-1/arch/i386/mm/highmem.c --- linux-2.5.70-bk4/arch/i386/mm/highmem.c 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/mm/highmem.c 2003-05-30 21:03:23.000000000 -0700 @@ -1,5 +1,9 @@ #include +/* + * XXX: resurrect kmap_pte + */ + void *kmap(struct page *page) { might_sleep(); @@ -28,44 +32,67 @@ void kunmap(struct page *page) void *kmap_atomic(struct page *page, enum km_type type) { enum fixed_addresses idx; - unsigned long vaddr; + unsigned long addr, vaddr, pfn; + int k; + pgd_t *pgd; + pmd_t *pmd; + pte_t *pte; inc_preempt_count(); if (page < highmem_start_page) return page_address(page); idx = type + KM_TYPE_NR*smp_processor_id(); - vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx); + vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx*PAGE_SIZE); + WARN_ON(vaddr > __fix_to_virt(FIX_KMAP_BEGIN)); + WARN_ON(vaddr < __fix_to_virt(FIX_KMAP_END)); + pfn = page_to_pfn(page); + addr = vaddr; + pgd_t *pgd = pgd_offset_k(addr); + pmd_t *pmd = pmd_offset(pgd, addr); + pte_t *pte = pte_offset_kernel(pmd, addr); + for (k = 0; k < PAGE_MMUCOUNT, ++k, addr += MMUPAGE_SIZE) { #ifdef CONFIG_DEBUG_HIGHMEM - if (!pte_none(*(kmap_pte-idx))) - BUG(); + BUG_ON(!pte_none(pte[k])); #endif - set_pte(kmap_pte-idx, mk_pte(page, kmap_prot)); - __flush_tlb_one(vaddr); - - return (void*) vaddr; + set_pte(&pte[k], pfn_pte(pfn + k, kmap_prot)); + __flush_tlb_one(addr); + } + return (void *)vaddr; } void kunmap_atomic(void *kvaddr, enum km_type type) { #ifdef CONFIG_DEBUG_HIGHMEM - unsigned long vaddr = (unsigned long) kvaddr & PAGE_MASK; + unsigned long vaddr = (unsigned long)kvaddr & PAGE_MASK; enum fixed_addresses idx = type + KM_TYPE_NR*smp_processor_id(); + unsigned long lo, hi; + int k; + pgd_t *pgd; + pmd_t *pmd; + pte_t *pte; if (vaddr < FIXADDR_START) { // FIXME dec_preempt_count(); return; } - if (vaddr != __fix_to_virt(FIX_KMAP_BEGIN+idx)) - BUG(); + lo = __fix_to_virt(FIX_KMAP_END) + PAGE_SIZE * idx; + hi = lo + PAGE_SIZE; + + BUG_ON(vaddr < lo || vaddr > hi);; /* * force other mappings to Oops if they'll try to access * this pte without first remap it */ - pte_clear(kmap_pte-idx); - __flush_tlb_one(vaddr); + pgd = pgd_offset_k(vaddr); + pmd = pmd_offset(pgd, vaddr); + pte = pte_offset_kernel(pmd, vaddr); + for (k = 0; k < PAGE_MMUCOUNT; ++k, vaddr += MMUPAGE_SIZE) { + pte_clear(&pte[k]); + __flush_tlb_one(vaddr); + } #endif dec_preempt_count(); @@ -73,14 +100,22 @@ void kunmap_atomic(void *kvaddr, enum km struct page *kmap_atomic_to_page(void *ptr) { - unsigned long idx, vaddr = (unsigned long)ptr; + unsigned long vaddr = (unsigned long)ptr; + pgd_t *pgd; + pmd_t *pmd; pte_t *pte; if (vaddr < FIXADDR_START) return virt_to_page(ptr); - idx = virt_to_fix(vaddr); - pte = kmap_pte - (idx - FIX_KMAP_BEGIN); + pgd = pgd_offset_k(vaddr); + pmd = pmd_offset(pgd, vaddr); + pte = pte_offset_kernel(pmd, vaddr); + + /* + * unsigned long idx = virt_to_fix(vaddr); + * pte = &kmap_pte[idx*PAGE_MMUCOUNT]; + */ return pte_page(*pte); } diff -prauN linux-2.5.70-bk4/arch/i386/mm/init.c pgcl-2.5.70-bk4-1/arch/i386/mm/init.c --- linux-2.5.70-bk4/arch/i386/mm/init.c 2003-05-26 18:00:45.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/mm/init.c 2003-05-30 21:03:23.000000000 -0700 @@ -43,6 +43,7 @@ DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); unsigned long highstart_pfn, highend_pfn; +struct page *zero_page; static int do_test_wp_bit(void); @@ -56,7 +57,7 @@ static pmd_t * __init one_md_table_init( pmd_t *pmd_table; #ifdef CONFIG_X86_PAE - pmd_table = (pmd_t *) alloc_bootmem_low_pages(PAGE_SIZE); + pmd_table = (pmd_t *) alloc_bootmem_low_pages(MMUPAGE_SIZE); set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); if (pmd_table != pmd_offset(pgd, 0)) BUG(); @@ -74,7 +75,7 @@ static pmd_t * __init one_md_table_init( static pte_t * __init one_page_table_init(pmd_t *pmd) { if (pmd_none(*pmd)) { - pte_t *page_table = (pte_t *) alloc_bootmem_low_pages(PAGE_SIZE); + pte_t *page_table = (pte_t *) alloc_bootmem_low_pages(MMUPAGE_SIZE); set_pmd(pmd, __pmd(__pa(page_table) | _PAGE_TABLE)); if (page_table != pte_offset_kernel(pmd, 0)) BUG(); @@ -95,6 +96,12 @@ static pte_t * __init one_page_table_ini * NOTE: The pagetables are allocated contiguous on the physical space * so we can cache the place of the first one and move around without * checking the pgd every time. + * + * Something happened here and I'm not sure what. This might back the + * thing out (I think). I think it was just a rename so I won't care + * unless it burns me. + * + * -- wli */ static void __init page_table_range_init (unsigned long start, unsigned long end, pgd_t *pgd_base) { @@ -111,7 +118,14 @@ static void __init page_table_range_init for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) { if (pgd_none(*pgd)) one_md_table_init(pgd); + } + + vaddr = start; + pgd_idx = pgd_index(vaddr); + pmd_idx = pmd_index(vaddr); + pgd = pgd_base + pgd_idx; + for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) { pmd = pmd_offset(pgd, vaddr); for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); pmd++, pmd_idx++) { if (pmd_none(*pmd)) @@ -180,8 +194,8 @@ static inline int page_is_ram(unsigned l * are not. Notably the 640->1Mb area. We need a sanity * check here. */ - addr = (e820.map[i].addr+PAGE_SIZE-1) >> PAGE_SHIFT; - end = (e820.map[i].addr+e820.map[i].size) >> PAGE_SHIFT; + addr = (e820.map[i].addr+MMUPAGE_SIZE-1) >> MMUPAGE_SHIFT; + end = (e820.map[i].addr+e820.map[i].size) >> MMUPAGE_SHIFT; if ((pagenr >= addr) && (pagenr < end)) return 1; } @@ -189,37 +203,13 @@ static inline int page_is_ram(unsigned l } #ifdef CONFIG_HIGHMEM -pte_t *kmap_pte; pgprot_t kmap_prot; -#define kmap_get_fixmap_pte(vaddr) \ - pte_offset_kernel(pmd_offset(pgd_offset_k(vaddr), (vaddr)), (vaddr)) - -void __init kmap_init(void) -{ - unsigned long kmap_vstart; - - /* cache the first kmap pte */ - kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN); - kmap_pte = kmap_get_fixmap_pte(kmap_vstart); - - kmap_prot = PAGE_KERNEL; -} +#define kmap_init() do { kmap_prot = PAGE_KERNEL; } while (0) void __init permanent_kmaps_init(pgd_t *pgd_base) { - pgd_t *pgd; - pmd_t *pmd; - pte_t *pte; - unsigned long vaddr; - - vaddr = PKMAP_BASE; - page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base); - - pgd = swapper_pg_dir + pgd_index(vaddr); - pmd = pmd_offset(pgd, vaddr); - pte = pte_offset_kernel(pmd, vaddr); - pkmap_page_table = pte; + page_table_range_init(PKMAP_BASE, PKMAP_BASE + PAGE_SIZE*LAST_PKMAP, pgd_base); } void __init one_highpage_init(struct page *page, int pfn, int bad_ppro) @@ -238,7 +228,7 @@ void __init one_highpage_init(struct pag void __init set_highmem_pages_init(int bad_ppro) { int pfn; - for (pfn = highstart_pfn; pfn < highend_pfn; pfn++) + for (pfn = highstart_pfn; pfn < highend_pfn; pfn += PAGE_MMUCOUNT) one_highpage_init(pfn_to_page(pfn), pfn, bad_ppro); totalram_pages += totalhigh_pages; } @@ -305,6 +295,34 @@ static void __init pagetable_init (void) */ pgd_base[0] = pgd_base[USER_PTRS_PER_PGD]; #endif + { + pgd_t *pgd; + pmd_t *pmd; + pte_t *pte; + unsigned long addr = VMALLOC_START; + + do { + pgd = pgd_offset_k(addr); + if (pgd_none(*pgd) || pgd_bad(*pgd)) { + addr += MMUPAGE_SIZE; + continue; + } + do { + pmd = pmd_offset(pgd, addr); + if (pmd_none(*pmd) || pmd_bad(*pmd)) { + addr += MMUPAGE_SIZE; + continue; + } + do { + pte = pte_offset_kernel(pmd, addr); + if (!pte_none(*pte) || pte_present(*pte)) { + printk("bad vmallocspace PTE at vaddr 0x%lx\n", addr); + } + addr += MMUPAGE_SIZE; + } while (addr < VMALLOC_END); + } while (addr < VMALLOC_END); + } while (addr < VMALLOC_END); + } } void zap_low_mappings (void) @@ -331,17 +349,17 @@ void __init zone_sizes_init(void) unsigned long zones_size[MAX_NR_ZONES] = {0, 0, 0}; unsigned int max_dma, high, low; - max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; + max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> MMUPAGE_SHIFT; low = max_low_pfn; high = highend_pfn; if (low < max_dma) - zones_size[ZONE_DMA] = low; + zones_size[ZONE_DMA] = low >> PAGE_MMUSHIFT; else { - zones_size[ZONE_DMA] = max_dma; - zones_size[ZONE_NORMAL] = low - max_dma; + zones_size[ZONE_DMA] = max_dma >> PAGE_MMUSHIFT; + zones_size[ZONE_NORMAL] = (low - max_dma) >> PAGE_MMUSHIFT; #ifdef CONFIG_HIGHMEM - zones_size[ZONE_HIGHMEM] = high - low; + zones_size[ZONE_HIGHMEM] = (high - low) >> PAGE_MMUSHIFT; #endif } free_area_init(zones_size); @@ -372,7 +390,6 @@ void __init paging_init(void) set_in_cr4(X86_CR4_PAE); #endif __flush_tlb_all(); - kmap_init(); zone_sizes_init(); } @@ -418,6 +435,7 @@ static void __init set_max_mapnr_init(vo #else max_mapnr = num_physpages = max_low_pfn; #endif + max_mapnr /= PAGE_MMUCOUNT; } #define __free_all_bootmem() free_all_bootmem() #else @@ -425,11 +443,14 @@ static void __init set_max_mapnr_init(vo extern void set_max_mapnr_init(void); #endif /* !CONFIG_DISCONTIGMEM */ +/* + * Most of the reporting here needs doublechecking. + */ void __init mem_init(void) { extern int ppro_with_ram_bug(void); int codesize, reservedpages, datasize, initsize; - int tmp; + int pfn; int bad_ppro; #ifndef CONFIG_DISCONTIGMEM @@ -439,36 +460,31 @@ void __init mem_init(void) bad_ppro = ppro_with_ram_bug(); -#ifdef CONFIG_HIGHMEM - /* check that fixmap and pkmap do not overlap */ - if (PKMAP_BASE+LAST_PKMAP*PAGE_SIZE >= FIXADDR_START) { - printk(KERN_ERR "fixmap and kmap areas overlap - this will crash\n"); - printk(KERN_ERR "pkstart: %lxh pkend: %lxh fixstart %lxh\n", - PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, FIXADDR_START); - BUG(); - } -#endif - set_max_mapnr_init(); #ifdef CONFIG_HIGHMEM - high_memory = (void *) __va(highstart_pfn * PAGE_SIZE); + high_memory = (void *) __va(highstart_pfn * MMUPAGE_SIZE); #else - high_memory = (void *) __va(max_low_pfn * PAGE_SIZE); + high_memory = (void *) __va(max_low_pfn * MMUPAGE_SIZE); #endif /* clear the zero-page */ - memset(empty_zero_page, 0, PAGE_SIZE); + memset(empty_zero_page, 0, MMUPAGE_SIZE); /* this will put all low memory onto the freelists */ totalram_pages += __free_all_bootmem(); + zero_page = alloc_page(GFP_ATOMIC|GFP_DMA); + clear_page(page_address(zero_page)); + SetPageReserved(zero_page); + totalram_pages--; + reservedpages = 0; - for (tmp = 0; tmp < max_low_pfn; tmp++) + for (pfn = 0; pfn < max_low_pfn; pfn += PAGE_MMUCOUNT) /* * Only count reserved RAM pages */ - if (page_is_ram(tmp) && PageReserved(pfn_to_page(tmp))) + if (page_is_ram(pfn) && PageReserved(pfn_to_page(pfn))) reservedpages++; set_highmem_pages_init(bad_ppro); @@ -479,13 +495,18 @@ void __init mem_init(void) printk(KERN_INFO "Memory: %luk/%luk available (%dk kernel code, %dk reserved, %dk data, %dk init, %ldk highmem)\n", (unsigned long) nr_free_pages() << (PAGE_SHIFT-10), - num_physpages << (PAGE_SHIFT-10), + num_physpages << (MMUPAGE_SHIFT-10), codesize >> 10, reservedpages << (PAGE_SHIFT-10), datasize >> 10, initsize >> 10, (unsigned long) (totalhigh_pages << (PAGE_SHIFT-10)) ); + printk("MAXMEM=0x%lx\n", MAXMEM); + printk("vmalloc: start = 0x%lx, end = 0x%lx\n", + VMALLOC_START, VMALLOC_END); + printk("fixaddr: start = 0x%lx, end = 0x%lx\n", + FIXADDR_START, FIXADDR_TOP); #ifdef CONFIG_X86_PAE if (!cpu_has_pae) @@ -505,20 +526,30 @@ void __init mem_init(void) #endif } -#ifdef CONFIG_X86_PAE -struct kmem_cache_s *pae_pgd_cachep; +kmem_cache_t *pgd_cache; +kmem_cache_t *pmd_cache; void __init pgtable_cache_init(void) { - /* - * PAE pgds must be 16-byte aligned: - */ - pae_pgd_cachep = kmem_cache_create("pae_pgd", 32, 0, - SLAB_HWCACHE_ALIGN | SLAB_MUST_HWCACHE_ALIGN, NULL, NULL); - if (!pae_pgd_cachep) - panic("init_pae(): Cannot alloc pae_pgd SLAB cache"); + if (PTRS_PER_PMD > 1) { + pmd_cache = kmem_cache_create("pmd", + PTRS_PER_PMD*sizeof(pmd_t), + 0, + SLAB_HWCACHE_ALIGN | SLAB_MUST_HWCACHE_ALIGN, + pmd_ctor, + NULL); + if (!pmd_cache) + panic("pgtable_cache_init(): cannot create pmd cache"); + } + pgd_cache = kmem_cache_create("pgd", + PTRS_PER_PGD*sizeof(pgd_t), + 0, + SLAB_HWCACHE_ALIGN | SLAB_MUST_HWCACHE_ALIGN, + pgd_ctor, + PTRS_PER_PMD == 1 ? pgd_dtor : NULL); + if (!pgd_cache) + panic("pgtable_cache_init(): Cannot create pgd cache"); } -#endif /* * This function cannot be __init, since exceptions don't work in that @@ -549,28 +580,43 @@ static int do_test_wp_bit(void) void free_initmem(void) { - unsigned long addr; + unsigned long addr, freed = 0;; addr = (unsigned long)(&__init_begin); - for (; addr < (unsigned long)(&__init_end); addr += PAGE_SIZE) { + addr = (addr + PAGE_SIZE - 1) & PAGE_MASK; + while(addr < (((unsigned long)(&__init_end)) & PAGE_MASK)) { ClearPageReserved(virt_to_page(addr)); set_page_count(virt_to_page(addr), 1); free_page(addr); totalram_pages++; + freed++; + addr += PAGE_SIZE; } - printk (KERN_INFO "Freeing unused kernel memory: %dk freed\n", (&__init_end - &__init_begin) >> 10); + printk(KERN_INFO "Freeing unused kernel memory: %ldk freed\n", + freed*(PAGE_SIZE/1024)); } #ifdef CONFIG_BLK_DEV_INITRD void free_initrd_mem(unsigned long start, unsigned long end) { - if (start < end) - printk (KERN_INFO "Freeing initrd memory: %ldk freed\n", (end - start) >> 10); - for (; start < end; start += PAGE_SIZE) { + unsigned long freed = 0; + + start = (start + PAGE_SIZE - 1) & PAGE_MASK; + end &= PAGE_MASK; + + if (start >= end) + return; + + while (start < end) { ClearPageReserved(virt_to_page(start)); set_page_count(virt_to_page(start), 1); free_page(start); totalram_pages++; + freed++; + start += PAGE_SIZE; } + + printk(KERN_INFO "Freeing initrd memory: %ldk freed\n", + freed*(PAGE_SIZE/1024)); } #endif diff -prauN linux-2.5.70-bk4/arch/i386/mm/ioremap.c pgcl-2.5.70-bk4-1/arch/i386/mm/ioremap.c --- linux-2.5.70-bk4/arch/i386/mm/ioremap.c 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/mm/ioremap.c 2003-05-30 21:03:23.000000000 -0700 @@ -30,7 +30,7 @@ static inline void remap_area_pte(pte_t end = PMD_SIZE; if (address >= end) BUG(); - pfn = phys_addr >> PAGE_SHIFT; + pfn = phys_addr >> MMUPAGE_SHIFT; do { if (!pte_none(*pte)) { printk("remap_area_pte: page already exists\n"); @@ -38,7 +38,7 @@ static inline void remap_area_pte(pte_t } set_pte(pte, pfn_pte(pfn, __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | flags))); - address += PAGE_SIZE; + address += MMUPAGE_SIZE; pfn++; pte++; } while (address && (address < end)); @@ -196,7 +196,7 @@ void *ioremap_nocache (unsigned long phy if (phys_addr + size < virt_to_phys(high_memory)) { struct page *ppage = virt_to_page(__va(phys_addr)); - unsigned long npages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; + unsigned long npages = (size + MMUPAGE_SIZE - 1) >> MMUPAGE_SHIFT; BUG_ON(phys_addr+size > (unsigned long)high_memory); BUG_ON(phys_addr + size < phys_addr); @@ -216,7 +216,7 @@ void iounmap(void *addr) struct vm_struct *p; if (addr <= high_memory) return; - p = remove_vm_area((void *) (PAGE_MASK & (unsigned long) addr)); + p = remove_vm_area((void *) (MMUPAGE_MASK & (unsigned long) addr)); if (!p) { printk("__iounmap: bad address %p\n", addr); return; @@ -224,7 +224,7 @@ void iounmap(void *addr) if (p->flags && p->phys_addr < virt_to_phys(high_memory)) { change_page_attr(virt_to_page(__va(p->phys_addr)), - p->size >> PAGE_SHIFT, + p->size >> MMUPAGE_SHIFT, PAGE_KERNEL); global_flush_tlb(); } @@ -251,14 +251,14 @@ void __init *bt_ioremap(unsigned long ph /* * Mappings have to be page-aligned */ - offset = phys_addr & ~PAGE_MASK; - phys_addr &= PAGE_MASK; - size = PAGE_ALIGN(last_addr) - phys_addr; + offset = phys_addr & ~MMUPAGE_MASK; + phys_addr &= MMUPAGE_MASK; + size = MMUPAGE_ALIGN(last_addr) - phys_addr; /* * Mappings have to fit in the FIX_BTMAP area. */ - nrpages = size >> PAGE_SHIFT; + nrpages = size >> MMUPAGE_SHIFT; if (nrpages > NR_FIX_BTMAPS) return NULL; @@ -268,7 +268,7 @@ void __init *bt_ioremap(unsigned long ph idx = FIX_BTMAP_BEGIN; while (nrpages > 0) { set_fixmap(idx, phys_addr); - phys_addr += PAGE_SIZE; + phys_addr += MMUPAGE_SIZE; --idx; --nrpages; } @@ -285,8 +285,8 @@ void __init bt_iounmap(void *addr, unsig virt_addr = (unsigned long)addr; if (virt_addr < fix_to_virt(FIX_BTMAP_BEGIN)) return; - offset = virt_addr & ~PAGE_MASK; - nrpages = PAGE_ALIGN(offset + size - 1) >> PAGE_SHIFT; + offset = virt_addr & ~MMUPAGE_MASK; + nrpages = MMUPAGE_ALIGN(offset + size - 1) >> MMUPAGE_SHIFT; idx = FIX_BTMAP_BEGIN; while (nrpages > 0) { diff -prauN linux-2.5.70-bk4/arch/i386/mm/pageattr.c pgcl-2.5.70-bk4-1/arch/i386/mm/pageattr.c --- linux-2.5.70-bk4/arch/i386/mm/pageattr.c 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/mm/pageattr.c 2003-05-30 21:03:23.000000000 -0700 @@ -38,8 +38,8 @@ static struct page *split_large_page(uns address = __pa(address); addr = address & LARGE_PAGE_MASK; pbase = (pte_t *)page_address(base); - for (i = 0; i < PTRS_PER_PTE; i++, addr += PAGE_SIZE) { - pbase[i] = pfn_pte(addr >> PAGE_SHIFT, + for (i = 0; i < PTRS_PER_PTE; i++, addr += MMUPAGE_SIZE) { + pbase[i] = pfn_pte(addr/MMUPAGE_SIZE, addr == address ? prot : PAGE_KERNEL); } return base; @@ -58,19 +58,22 @@ static void flush_kernel_map(void *dummy static void set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte) { + struct page *page; + unsigned long flags; + set_pte_atomic(kpte, pte); /* change init_mm */ -#ifndef CONFIG_X86_PAE - { - struct list_head *l; - spin_lock(&mmlist_lock); - list_for_each(l, &init_mm.mmlist) { - struct mm_struct *mm = list_entry(l, struct mm_struct, mmlist); - pmd_t *pmd = pmd_offset(pgd_offset(mm, address), address); + if (PTRS_PER_PMD > 1) + return; + + spin_lock_irqsave(&pgd_lock, flags); + list_for_each_entry(page, &pgd_list, lru) { + pgd_t *pgd; + pmd_t *pmd; + pgd = (pgd_t *)page_address(page) + pgd_index(address); + pmd = pmd_offset(pgd, address); set_pte_atomic((pte_t *)pmd, pte); } - spin_unlock(&mmlist_lock); - } -#endif + spin_unlock_irqrestore(&pgd_lock, flags); } /* @@ -82,7 +85,7 @@ static inline void revert_page(struct pa pte_t *linear = (pte_t *) pmd_offset(pgd_offset(&init_mm, address), address); set_pmd_pte(linear, address, - pfn_pte((__pa(address) & LARGE_PAGE_MASK) >> PAGE_SHIFT, + pfn_pte((__pa(address) & LARGE_PAGE_MASK)/MMUPAGE_SIZE, PAGE_KERNEL_LARGE)); } @@ -94,15 +97,14 @@ __change_page_attr(struct page *page, pg struct page *kpte_page; #ifdef CONFIG_HIGHMEM - if (page >= highmem_start_page) - BUG(); + BUG_ON(page >= highmem_start_page); #endif address = (unsigned long)page_address(page); kpte = lookup_address(address); if (!kpte) return -EINVAL; - kpte_page = virt_to_page(((unsigned long)kpte) & PAGE_MASK); + kpte_page = virt_to_page(((unsigned long)kpte) & MMUPAGE_MASK); if (pgprot_val(prot) != pgprot_val(PAGE_KERNEL)) { if ((pte_val(*kpte) & _PAGE_PSE) == 0) { pte_t old = *kpte; @@ -159,6 +161,8 @@ int change_page_attr(struct page *page, struct page *fpage; int i; + numpages = (numpages + PAGE_MMUCOUNT - 1)& ~(PAGE_MMUCOUNT-1); + down_write(&init_mm.mmap_sem); for (i = 0; i < numpages; i++, page++) { fpage = NULL; diff -prauN linux-2.5.70-bk4/arch/i386/mm/pgtable.c pgcl-2.5.70-bk4-1/arch/i386/mm/pgtable.c --- linux-2.5.70-bk4/arch/i386/mm/pgtable.c 2003-05-26 18:01:03.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/mm/pgtable.c 2003-05-30 21:03:23.000000000 -0700 @@ -12,6 +12,7 @@ #include #include #include +#include #include #include @@ -98,10 +99,12 @@ void set_pmd_pfn(unsigned long vaddr, un if (vaddr & (PMD_SIZE-1)) { /* vaddr is misaligned */ printk ("set_pmd_pfn: vaddr misaligned\n"); + printk ("vaddr = %lx, pfn = %lx\n", vaddr, pfn); return; /* BUG(); */ } - if (pfn & (PTRS_PER_PTE-1)) { /* pfn is misaligned */ + if (pfn & (PMD_SIZE/MMUPAGE_SIZE-1)) { /* pfn is misaligned */ printk ("set_pmd_pfn: pfn misaligned\n"); + printk ("vaddr = %lx, pfn = %lx\n", vaddr, pfn); return; /* BUG(); */ } pgd = swapper_pg_dir + pgd_index(vaddr); @@ -122,11 +125,13 @@ void __set_fixmap (enum fixed_addresses { unsigned long address = __fix_to_virt(idx); + printk("__set_fixmap(%d,%lx)\n", idx, phys); + if (idx >= __end_of_fixed_addresses) { BUG(); return; } - set_pte_pfn(address, phys >> PAGE_SHIFT, flags); + set_pte_pfn(address, phys >> MMUPAGE_SHIFT, flags); } pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address) @@ -151,61 +156,89 @@ struct page *pte_alloc_one(struct mm_str return pte; } -#ifdef CONFIG_X86_PAE +void pmd_ctor(void *pmd, kmem_cache_t *cache, unsigned long flags) +{ + memset(pmd, 0, PTRS_PER_PMD*sizeof(pmd_t)); +} -pgd_t *pgd_alloc(struct mm_struct *mm) +/* + * List of all pgd's needed for non-PAE so it can invalidate entries + * in both cached and uncached pgd's; not needed for PAE since the + * kernel pmd is shared. If PAE were not to share the pmd a similar + * tactic would be needed. This is essentially codepath-based locking + * against pageattr.c; it is the unique case in which a valid change + * of kernel pagetables can't be lazily synchronized by vmalloc faults. + * vmalloc faults work because attached pagetables are never freed. + * If the locking proves to be non-performant, a ticketing scheme with + * checks at dup_mmap(), exec(), and other mmlist addition points + * could be used. The locking scheme was chosen on the basis of + * manfred's recommendations and having no core impact whatsoever. + * -- wli + */ +spinlock_t pgd_lock = SPIN_LOCK_UNLOCKED; +LIST_HEAD(pgd_list); + +void pgd_ctor(void *pgd, kmem_cache_t *cache, unsigned long unused) { - int i; - pgd_t *pgd = kmem_cache_alloc(pae_pgd_cachep, GFP_KERNEL); + unsigned long flags; - if (pgd) { - for (i = 0; i < USER_PTRS_PER_PGD; i++) { - unsigned long pmd = __get_free_page(GFP_KERNEL); - if (!pmd) - goto out_oom; - clear_page(pmd); - set_pgd(pgd + i, __pgd(1 + __pa(pmd))); - } - memcpy(pgd + USER_PTRS_PER_PGD, + if (PTRS_PER_PMD == 1) + spin_lock_irqsave(&pgd_lock, flags); + + memcpy((pgd_t *)pgd + USER_PTRS_PER_PGD, swapper_pg_dir + USER_PTRS_PER_PGD, - (PTRS_PER_PGD - USER_PTRS_PER_PGD) * sizeof(pgd_t)); - } - return pgd; -out_oom: - for (i--; i >= 0; i--) - free_page((unsigned long)__va(pgd_val(pgd[i])-1)); - kmem_cache_free(pae_pgd_cachep, pgd); - return NULL; + (PTRS_PER_PGD - USER_PTRS_PER_PGD)*sizeof(pgd_t)); + + if (PTRS_PER_PMD > 1) + return; + + list_add(&virt_to_page(pgd)->lru, &pgd_list); + spin_unlock_irqrestore(&pgd_lock, flags); + memset(pgd, 0, USER_PTRS_PER_PGD*sizeof(pgd_t)); } -void pgd_free(pgd_t *pgd) +/* never called when PTRS_PER_PMD > 1 */ +void pgd_dtor(void *pgd, kmem_cache_t *cache, unsigned long unused) { - int i; + unsigned long flags; /* can be called from interrupt context */ - for (i = 0; i < USER_PTRS_PER_PGD; i++) - free_page((unsigned long)__va(pgd_val(pgd[i])-1)); - kmem_cache_free(pae_pgd_cachep, pgd); + spin_lock_irqsave(&pgd_lock, flags); + list_del(&virt_to_page(pgd)->lru); + spin_unlock_irqrestore(&pgd_lock, flags); } -#else - pgd_t *pgd_alloc(struct mm_struct *mm) { - pgd_t *pgd = (pgd_t *)__get_free_page(GFP_KERNEL); + int i; + pgd_t *pgd = kmem_cache_alloc(pgd_cache, GFP_KERNEL); - if (pgd) { - memset(pgd, 0, USER_PTRS_PER_PGD * sizeof(pgd_t)); - memcpy(pgd + USER_PTRS_PER_PGD, - swapper_pg_dir + USER_PTRS_PER_PGD, - (PTRS_PER_PGD - USER_PTRS_PER_PGD) * sizeof(pgd_t)); + if (PTRS_PER_PMD == 1 || !pgd) + return pgd; + + for (i = 0; i < USER_PTRS_PER_PGD; ++i) { + pmd_t *pmd = kmem_cache_alloc(pmd_cache, GFP_KERNEL); + if (!pmd) + goto out_oom; + set_pgd(&pgd[i], __pgd(1 + __pa((u64)(u32)__va(pgd_val(pgd[i])-1)))); } return pgd; + +out_oom: + for (i--; i >= 0; i--) + kmem_cache_free(pmd_cache, __va(pgd_val(pgd[i])-1)); + kmem_cache_free(pgd_cache, pgd); + return NULL; } void pgd_free(pgd_t *pgd) { - free_page((unsigned long)pgd); -} + int i; -#endif /* CONFIG_X86_PAE */ + /* in the PAE case user pgd entries are overwritten before usage */ + if (PTRS_PER_PMD > 1) + for (i = 0; i < USER_PTRS_PER_PGD; ++i) + kmem_cache_free(pmd_cache, __va(pgd_val(pgd[i])-1)); + /* in the non-PAE case, clear_page_tables() clears user pgd entries */ + kmem_cache_free(pgd_cache, pgd); +} diff -prauN linux-2.5.70-bk4/arch/i386/pci/i386.c pgcl-2.5.70-bk4-1/arch/i386/pci/i386.c --- linux-2.5.70-bk4/arch/i386/pci/i386.c 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/arch/i386/pci/i386.c 2003-05-30 21:03:23.000000000 -0700 @@ -291,7 +291,7 @@ int pci_mmap_page_range(struct pci_dev * /* Write-combine setting is ignored, it is changed via the mtrr * interfaces on this platform. */ - if (remap_page_range(vma, vma->vm_start, vma->vm_pgoff << PAGE_SHIFT, + if (remap_page_range(vma, vma->vm_start, vma->vm_pgoff << MMUPAGE_SHIFT, vma->vm_end - vma->vm_start, vma->vm_page_prot)) return -EAGAIN; diff -prauN linux-2.5.70-bk4/drivers/block/ll_rw_blk.c pgcl-2.5.70-bk4-1/drivers/block/ll_rw_blk.c --- linux-2.5.70-bk4/drivers/block/ll_rw_blk.c 2003-05-30 21:02:25.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/block/ll_rw_blk.c 2003-05-30 21:03:23.000000000 -0700 @@ -234,7 +234,7 @@ void blk_queue_make_request(request_queu **/ void blk_queue_bounce_limit(request_queue_t *q, u64 dma_addr) { - unsigned long bounce_pfn = dma_addr >> PAGE_SHIFT; + unsigned long bounce_pfn = dma_addr >> MMUPAGE_SHIFT; unsigned long mb = dma_addr >> 20; static request_queue_t *last_q; diff -prauN linux-2.5.70-bk4/drivers/char/agp/backend.c pgcl-2.5.70-bk4-1/drivers/char/agp/backend.c --- linux-2.5.70-bk4/drivers/char/agp/backend.c 2003-05-26 18:00:43.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/char/agp/backend.c 2003-05-30 21:03:23.000000000 -0700 @@ -106,7 +106,7 @@ static int agp_find_max(void) { long memory, index, result; - memory = (num_physpages << PAGE_SHIFT) >> 20; + memory = (num_physpages << MMUPAGE_SHIFT) >> 20; index = 1; while ((memory > maxes_table[index].mem) && (index < 8)) @@ -118,7 +118,7 @@ static int agp_find_max(void) (maxes_table[index].mem - maxes_table[index - 1].mem); printk(KERN_INFO PFX "Maximum main memory to use for agp memory: %ldM\n", result); - result = result << (20 - PAGE_SHIFT); + result = result << (20 - MMUPAGE_SHIFT); return result; } @@ -157,7 +157,7 @@ static int agp_backend_initialize(struct } got_gatt = 1; - bridge->key_list = vmalloc(PAGE_SIZE * 4); + bridge->key_list = vmalloc(4*MMUPAGE_SIZE); if (bridge->key_list == NULL) { printk(KERN_ERR PFX "error allocating memory for key lists.\n"); rc = -ENOMEM; @@ -166,7 +166,7 @@ static int agp_backend_initialize(struct got_keylist = 1; /* FIXME vmalloc'd memory not guaranteed contiguous */ - memset(bridge->key_list, 0, PAGE_SIZE * 4); + memset(bridge->key_list, 0, 4*MMUPAGE_SIZE); if (bridge->driver->configure()) { printk(KERN_ERR PFX "error configuring host chipset.\n"); diff -prauN linux-2.5.70-bk4/drivers/char/agp/generic.c pgcl-2.5.70-bk4-1/drivers/char/agp/generic.c --- linux-2.5.70-bk4/drivers/char/agp/generic.c 2003-05-26 18:00:37.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/char/agp/generic.c 2003-05-30 21:03:23.000000000 -0700 @@ -85,7 +85,7 @@ struct agp_memory *agp_create_memory(int kfree(new); return NULL; } - new->memory = vmalloc(PAGE_SIZE * scratch_pages); + new->memory = vmalloc(MMUPAGE_SIZE * scratch_pages); if (new->memory == NULL) { agp_free_key(new->key); @@ -130,7 +130,7 @@ void agp_free_memory(struct agp_memory * } EXPORT_SYMBOL(agp_free_memory); -#define ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(unsigned long)) +#define ENTRIES_PER_PAGE (MMUPAGE_SIZE / sizeof(unsigned long)) /** * agp_allocate_memory - allocate a group of pages of a certain type. @@ -668,7 +668,7 @@ int agp_generic_create_gatt_table(void) if (table == NULL) return -ENOMEM; - table_end = table + ((PAGE_SIZE * (1 << page_order)) - 1); + table_end = table + ((MMUPAGE_SIZE * (1 << page_order)) - 1); for (page = virt_to_page(table); page <= virt_to_page(table_end); page++) SetPageReserved(page); @@ -678,7 +678,7 @@ int agp_generic_create_gatt_table(void) agp_bridge->driver->cache_flush(); agp_bridge->gatt_table = ioremap_nocache(virt_to_phys(table), - (PAGE_SIZE * (1 << page_order))); + MMUPAGE_SIZE << page_order); agp_bridge->driver->cache_flush(); if (agp_bridge->gatt_table == NULL) { @@ -736,7 +736,7 @@ int agp_generic_free_gatt_table(void) iounmap(agp_bridge->gatt_table); table = (char *) agp_bridge->gatt_table_real; - table_end = table + ((PAGE_SIZE * (1 << page_order)) - 1); + table_end = table + ((MMUPAGE_SIZE * (1 << page_order)) - 1); for (page = virt_to_page(table); page <= virt_to_page(table_end); page++) ClearPageReserved(page); @@ -784,7 +784,7 @@ int agp_generic_insert_memory(struct agp break; } - num_entries -= agp_memory_reserved/PAGE_SIZE; + num_entries -= agp_memory_reserved/MMUPAGE_SIZE; if (num_entries < 0) num_entries = 0; if (type != 0 || mem->type != 0) { diff -prauN linux-2.5.70-bk4/drivers/char/mem.c pgcl-2.5.70-bk4-1/drivers/char/mem.c --- linux-2.5.70-bk4/drivers/char/mem.c 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/char/mem.c 2003-05-30 21:03:23.000000000 -0700 @@ -43,8 +43,8 @@ static ssize_t do_write_mem(struct file written = 0; #if defined(__sparc__) || (defined(__mc68000__) && defined(CONFIG_MMU)) /* we don't have page 0 mapped on sparc and m68k.. */ - if (realp < PAGE_SIZE) { - unsigned long sz = PAGE_SIZE-realp; + if (realp < MMUPAGE_SIZE) { + unsigned long sz = MMUPAGE_SIZE-realp; if (sz > count) sz = count; /* Hmm. Do something? */ buf+=sz; @@ -80,8 +80,8 @@ static ssize_t read_mem(struct file * fi read = 0; #if defined(__sparc__) || (defined(__mc68000__) && defined(CONFIG_MMU)) /* we don't have page 0 mapped on sparc and m68k.. */ - if (p < PAGE_SIZE) { - unsigned long sz = PAGE_SIZE-p; + if (p < MMUPAGE_SIZE) { + unsigned long sz = MMUPAGE_SIZE-p; if (sz > count) sz = count; if (sz > 0) { @@ -177,7 +177,7 @@ static inline int noncached_address(unsi static int mmap_mem(struct file * file, struct vm_area_struct * vma) { - unsigned long offset = vma->vm_pgoff << PAGE_SHIFT; + unsigned long offset = vma->vm_pgoff << MMUPAGE_SHIFT; /* * Accessing memory above the top the kernel knows about or @@ -223,8 +223,8 @@ static ssize_t read_kmem(struct file *fi #if defined(__sparc__) || (defined(__mc68000__) && defined(CONFIG_MMU)) /* we don't have page 0 mapped on sparc and m68k.. */ - if (p < PAGE_SIZE && read > 0) { - size_t tmp = PAGE_SIZE - p; + if (p < MMUPAGE_SIZE && read > 0) { + size_t tmp = MMUPAGE_SIZE - p; if (tmp > read) tmp = read; if (clear_user(buf, tmp)) return -EFAULT; @@ -248,8 +248,8 @@ static ssize_t read_kmem(struct file *fi while (count > 0) { int len = count; - if (len > PAGE_SIZE) - len = PAGE_SIZE; + if (len > MMUPAGE_SIZE) + len = MMUPAGE_SIZE; len = vread(kbuf, (char *)p, len); if (!len) break; @@ -298,8 +298,8 @@ static ssize_t write_kmem(struct file * while (count > 0) { int len = count; - if (len > PAGE_SIZE) - len = PAGE_SIZE; + if (len > MMUPAGE_SIZE) + len = MMUPAGE_SIZE; if (len && copy_from_user(kbuf, buf, len)) { free_page((unsigned long)kbuf); return -EFAULT; @@ -409,12 +409,12 @@ static inline size_t read_zero_pagealign /* The shared case is hard. Let's do the conventional zeroing. */ do { - unsigned long unwritten = clear_user(buf, PAGE_SIZE); + unsigned long unwritten = clear_user(buf, MMUPAGE_SIZE); if (unwritten) - return size + unwritten - PAGE_SIZE; + return size + unwritten - MMUPAGE_SIZE; cond_resched(); - buf += PAGE_SIZE; - size -= PAGE_SIZE; + buf += MMUPAGE_SIZE; + size -= MMUPAGE_SIZE; } while (size); return size; @@ -437,23 +437,23 @@ static ssize_t read_zero(struct file * f left = count; /* do we want to be clever? Arbitrary cut-off */ - if (count >= PAGE_SIZE*4) { + if (count >= MMUPAGE_SIZE*4) { unsigned long partial; /* How much left of the page? */ - partial = (PAGE_SIZE-1) & -(unsigned long) buf; + partial = (MMUPAGE_SIZE-1) & -(unsigned long) buf; unwritten = clear_user(buf, partial); written = partial - unwritten; if (unwritten) goto out; left -= partial; buf += partial; - unwritten = read_zero_pagealigned(buf, left & PAGE_MASK); - written += (left & PAGE_MASK) - unwritten; + unwritten = read_zero_pagealigned(buf, left & MMUPAGE_MASK); + written += (left & MMUPAGE_MASK) - unwritten; if (unwritten) goto out; - buf += left & PAGE_MASK; - left &= ~PAGE_MASK; + buf += left & MMUPAGE_MASK; + left &= ~MMUPAGE_MASK; } unwritten = clear_user(buf, left); written += left - unwritten; diff -prauN linux-2.5.70-bk4/drivers/oprofile/buffer_sync.c pgcl-2.5.70-bk4-1/drivers/oprofile/buffer_sync.c --- linux-2.5.70-bk4/drivers/oprofile/buffer_sync.c 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/oprofile/buffer_sync.c 2003-05-30 21:03:23.000000000 -0700 @@ -247,7 +247,7 @@ static unsigned long lookup_dcookie(stru cookie = fast_get_dcookie(vma->vm_file->f_dentry, vma->vm_file->f_vfsmnt); - *offset = (vma->vm_pgoff << PAGE_SHIFT) + addr - vma->vm_start; + *offset = MMUPAGE_SIZE*vma->vm_pgoff + addr - vma->vm_start; break; } out: diff -prauN linux-2.5.70-bk4/drivers/scsi/qlogicisp.c pgcl-2.5.70-bk4-1/drivers/scsi/qlogicisp.c --- linux-2.5.70-bk4/drivers/scsi/qlogicisp.c 2003-05-26 18:01:00.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/scsi/qlogicisp.c 2003-05-30 21:03:23.000000000 -0700 @@ -1415,7 +1415,7 @@ static int isp1020_init(struct Scsi_Host if ((command & PCI_COMMAND_MEMORY) && ((mem_flags & 1) == 0)) { - mem_base = (u_long) ioremap(mem_base, PAGE_SIZE); + mem_base = (u_long) ioremap(mem_base, MMUPAGE_SIZE); if (!mem_base) { printk("qlogicisp : i/o remapping failed.\n"); goto out_release; diff -prauN linux-2.5.70-bk4/drivers/scsi/sym53c8xx.c pgcl-2.5.70-bk4-1/drivers/scsi/sym53c8xx.c --- linux-2.5.70-bk4/drivers/scsi/sym53c8xx.c 2003-05-30 21:02:35.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/scsi/sym53c8xx.c 2003-05-30 21:03:23.000000000 -0700 @@ -686,7 +686,8 @@ spinlock_t sym53c8xx_lock = SPIN_LOCK_UN #ifndef SCSI_NCR_PCI_MEM_NOT_SUPPORTED static u_long __init remap_pci_mem(u_long base, u_long size) { - u_long page_base = ((u_long) base) & PAGE_MASK; + /* ioremap()/vmalloc() have MMUPAGE_SIZE granularity */ + u_long page_base = ((u_long) base) & MMUPAGE_MASK; u_long page_offs = ((u_long) base) - page_base; u_long page_remapped = (u_long) ioremap(page_base, page_offs+size); @@ -695,8 +696,9 @@ static u_long __init remap_pci_mem(u_lon static void __init unmap_pci_mem(u_long vaddr, u_long size) { + /* iounmap()/vfree() have MMUPAGE_SIZE granularity */ if (vaddr) - iounmap((void *) (vaddr & PAGE_MASK)); + iounmap((void *) (vaddr & MMUPAGE_MASK)); } #endif /* not def SCSI_NCR_PCI_MEM_NOT_SUPPORTED */ diff -prauN linux-2.5.70-bk4/drivers/scsi/sym53c8xx_2/sym_glue.c pgcl-2.5.70-bk4-1/drivers/scsi/sym53c8xx_2/sym_glue.c --- linux-2.5.70-bk4/drivers/scsi/sym53c8xx_2/sym_glue.c 2003-05-30 21:02:36.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/scsi/sym53c8xx_2/sym_glue.c 2003-05-30 21:03:23.000000000 -0700 @@ -215,7 +215,7 @@ m_addr_t __vtobus(m_pool_ident_t dev_dma #ifndef SYM_OPT_NO_BUS_MEMORY_MAPPING static u_long __init pci_map_mem(u_long base, u_long size) { - u_long page_base = ((u_long) base) & PAGE_MASK; + u_long page_base = ((u_long) base) & MMUPAGE_MASK; u_long page_offs = ((u_long) base) - page_base; u_long page_remapped = (u_long) ioremap(page_base, page_offs+size); @@ -225,7 +225,7 @@ static u_long __init pci_map_mem(u_long static void __init pci_unmap_mem(u_long vaddr, u_long size) { if (vaddr) - iounmap((void *) (vaddr & PAGE_MASK)); + iounmap((void *) (vaddr & MMUPAGE_MASK)); } #endif diff -prauN linux-2.5.70-bk4/drivers/scsi/sym53c8xx_comm.h pgcl-2.5.70-bk4-1/drivers/scsi/sym53c8xx_comm.h --- linux-2.5.70-bk4/drivers/scsi/sym53c8xx_comm.h 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/drivers/scsi/sym53c8xx_comm.h 2003-05-30 21:03:23.000000000 -0700 @@ -491,7 +491,7 @@ spinlock_t DRIVER_SMP_LOCK = SPIN_LOCK_U #ifndef SCSI_NCR_PCI_MEM_NOT_SUPPORTED static u_long __init remap_pci_mem(u_long base, u_long size) { - u_long page_base = ((u_long) base) & PAGE_MASK; + u_long page_base = ((u_long) base) & MMUPAGE_MASK; u_long page_offs = ((u_long) base) - page_base; u_long page_remapped = (u_long) ioremap(page_base, page_offs+size); @@ -501,7 +501,7 @@ static u_long __init remap_pci_mem(u_lon static void __init unmap_pci_mem(u_long vaddr, u_long size) { if (vaddr) - iounmap((void *) (vaddr & PAGE_MASK)); + iounmap((void *) (vaddr & MMUPAGE_MASK)); } #endif /* not def SCSI_NCR_PCI_MEM_NOT_SUPPORTED */ diff -prauN linux-2.5.70-bk4/fs/aio.c pgcl-2.5.70-bk4-1/fs/aio.c --- linux-2.5.70-bk4/fs/aio.c 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/aio.c 2003-05-30 21:03:23.000000000 -0700 @@ -87,7 +87,7 @@ static void aio_free_ring(struct kioctx long i; for (i=0; inr_pages; i++) - put_page(info->ring_pages[i]); + put_page(pfn_to_page(info->ring_pages[i])); if (info->mmap_size) { down_write(&ctx->mm->mmap_sem); @@ -114,25 +114,25 @@ static int aio_setup_ring(struct kioctx size = sizeof(struct aio_ring); size += sizeof(struct io_event) * nr_events; - nr_pages = (size + PAGE_SIZE-1) >> PAGE_SHIFT; + nr_pages = (size + MMUPAGE_SIZE-1) >> MMUPAGE_SHIFT; if (nr_pages < 0) return -EINVAL; info->nr_pages = nr_pages; - nr_events = (PAGE_SIZE * nr_pages - sizeof(struct aio_ring)) / sizeof(struct io_event); + nr_events = (MMUPAGE_SIZE*nr_pages - sizeof(struct aio_ring))/sizeof(struct io_event); info->nr = 0; info->ring_pages = info->internal_pages; if (nr_pages > AIO_RING_PAGES) { - info->ring_pages = kmalloc(sizeof(struct page *) * nr_pages, GFP_KERNEL); + info->ring_pages = kmalloc(sizeof(unsigned long)*nr_pages, GFP_KERNEL); if (!info->ring_pages) return -ENOMEM; - memset(info->ring_pages, 0, sizeof(struct page *) * nr_pages); + memset(info->ring_pages, 0, sizeof(unsigned long)*nr_pages); } - info->mmap_size = nr_pages * PAGE_SIZE; + info->mmap_size = nr_pages*MMUPAGE_SIZE; dprintk("attempting mmap of %lu bytes\n", info->mmap_size); down_write(&ctx->mm->mmap_sem); info->mmap_base = do_mmap(NULL, 0, info->mmap_size, @@ -161,7 +161,8 @@ static int aio_setup_ring(struct kioctx info->nr = nr_events; /* trusted copy */ - ring = kmap_atomic(info->ring_pages[0], KM_USER0); + ring = kmap_atomic(pfn_to_page(info->ring_pages[0]), KM_USER0) + + (info->ring_pages[0] % PAGE_MMUCOUNT)*MMUPAGE_SIZE; ring->nr = nr_events; /* user copy */ ring->id = ctx->user_id; ring->head = ring->tail = 0; @@ -178,15 +179,17 @@ static int aio_setup_ring(struct kioctx /* aio_ring_event: returns a pointer to the event at the given index from * kmap_atomic(, km). Release the pointer with put_aio_ring_event(); */ -#define AIO_EVENTS_PER_PAGE (PAGE_SIZE / sizeof(struct io_event)) -#define AIO_EVENTS_FIRST_PAGE ((PAGE_SIZE - sizeof(struct aio_ring)) / sizeof(struct io_event)) +#define AIO_EVENTS_PER_PAGE (MMUPAGE_SIZE/sizeof(struct io_event)) +#define AIO_EVENTS_FIRST_PAGE ((MMUPAGE_SIZE-sizeof(struct aio_ring))/sizeof(struct io_event)) #define AIO_EVENTS_OFFSET (AIO_EVENTS_PER_PAGE - AIO_EVENTS_FIRST_PAGE) #define aio_ring_event(info, nr, km) ({ \ unsigned pos = (nr) + AIO_EVENTS_OFFSET; \ struct io_event *__event; \ - __event = kmap_atomic( \ - (info)->ring_pages[pos / AIO_EVENTS_PER_PAGE], km); \ + unsigned long pfn; \ + pfn = (info)->ring_pages[pos/AIO_EVENTS_PER_PAGE]; \ + __event = kmap_atomic(pfn_to_page(pfn), km); \ + __event += (pfn % PAGE_MMUCOUNT) * MMUPAGE_SIZE; \ __event += pos % AIO_EVENTS_PER_PAGE; \ __event; \ }) @@ -194,7 +197,7 @@ static int aio_setup_ring(struct kioctx #define put_aio_ring_event(event, km) do { \ struct io_event *__event = (event); \ (void)__event; \ - kunmap_atomic((void *)((unsigned long)__event & PAGE_MASK), km); \ + kunmap_atomic((void *)((unsigned long)__event & MMUPAGE_MASK), km); \ } while(0) /* ioctx_alloc @@ -400,7 +403,8 @@ static struct kiocb *__aio_get_req(struc * accept an event from this io. */ spin_lock_irq(&ctx->ctx_lock); - ring = kmap_atomic(ctx->ring_info.ring_pages[0], KM_USER0); + ring = kmap_atomic(pfn_to_page(ctx->ring_info.ring_pages[0]), KM_USER0) + + (ctx->ring_info.ring_pages[0] % PAGE_MMUCOUNT)*MMUPAGE_SIZE; if (ctx->reqs_active < aio_ring_avail(&ctx->ring_info, ring)) { list_add(&req->ki_list, &ctx->active_reqs); get_ioctx(ctx); @@ -664,8 +668,8 @@ int aio_complete(struct kiocb *iocb, lon */ spin_lock_irqsave(&ctx->ctx_lock, flags); - ring = kmap_atomic(info->ring_pages[0], KM_IRQ1); - + ring = kmap_atomic(pfn_to_page(info->ring_pages[0]), KM_IRQ1) + + (info->ring_pages[0] % PAGE_MMUCOUNT)*MMUPAGE_SIZE; tail = info->tail; event = aio_ring_event(info, tail, KM_IRQ0); tail = (tail + 1) % info->nr; @@ -720,7 +724,8 @@ static int aio_read_evt(struct kioctx *i unsigned long head; int ret = 0; - ring = kmap_atomic(info->ring_pages[0], KM_USER0); + ring = kmap_atomic(pfn_to_page(info->ring_pages[0]), KM_USER0) + + (info->ring_pages[0] % PAGE_MMUCOUNT)*MMUPAGE_SIZE; dprintk("in aio_read_evt h%lu t%lu m%lu\n", (unsigned long)ring->head, (unsigned long)ring->tail, (unsigned long)ring->nr); diff -prauN linux-2.5.70-bk4/fs/binfmt_elf.c pgcl-2.5.70-bk4-1/fs/binfmt_elf.c --- linux-2.5.70-bk4/fs/binfmt_elf.c 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/binfmt_elf.c 2003-05-30 21:03:23.000000000 -0700 @@ -61,10 +61,10 @@ static int elf_core_dump(long signr, str #define elf_core_dump NULL #endif -#if ELF_EXEC_PAGESIZE > PAGE_SIZE +#if ELF_EXEC_PAGESIZE > MMUPAGE_SIZE # define ELF_MIN_ALIGN ELF_EXEC_PAGESIZE #else -# define ELF_MIN_ALIGN PAGE_SIZE +# define ELF_MIN_ALIGN MMUPAGE_SIZE #endif #define ELF_PAGESTART(_v) ((_v) & ~(unsigned long)(ELF_MIN_ALIGN-1)) @@ -781,9 +781,8 @@ static int load_elf_binary(struct linux_ and some applications "depend" upon this behavior. Since we do not have the power to recompile these, we emulate the SVr4 behavior. Sigh. */ - /* N.B. Shouldn't the size here be PAGE_SIZE?? */ down_write(¤t->mm->mmap_sem); - error = do_mmap(NULL, 0, 4096, PROT_READ | PROT_EXEC, + error = do_mmap(NULL, 0, MMUPAGE_SIZE, PROT_READ | PROT_EXEC, MAP_FIXED | MAP_PRIVATE, 0); up_write(¤t->mm->mmap_sem); } @@ -1370,21 +1369,26 @@ static int elf_core_dump(long signr, str for (addr = vma->vm_start; addr < vma->vm_end; - addr += PAGE_SIZE) { + addr += MMUPAGE_SIZE) { struct page* page; + unsigned long pfn = 0; struct vm_area_struct *vma; if (get_user_pages(current, current->mm, addr, 1, 0, 1, - &page, &vma) <= 0) { - DUMP_SEEK (file->f_pos + PAGE_SIZE); + &pfn, &vma) <= 0) { + DUMP_SEEK (file->f_pos + MMUPAGE_SIZE); } else { + page = pfn_to_page(pfn); if (page == ZERO_PAGE(addr)) { - DUMP_SEEK (file->f_pos + PAGE_SIZE); + DUMP_SEEK (file->f_pos + MMUPAGE_SIZE); } else { void *kaddr; + unsigned long subpfn; + subpfn = pfn % PAGE_MMUCOUNT; flush_cache_page(vma, addr); kaddr = kmap(page); - DUMP_WRITE(kaddr, PAGE_SIZE); + kaddr += subpfn * MMUPAGE_SIZE; + DUMP_WRITE(kaddr, MMUPAGE_SIZE); kunmap(page); } page_cache_release(page); diff -prauN linux-2.5.70-bk4/fs/bio.c pgcl-2.5.70-bk4-1/fs/bio.c --- linux-2.5.70-bk4/fs/bio.c 2003-05-30 21:02:36.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/bio.c 2003-05-30 21:03:23.000000000 -0700 @@ -445,12 +445,12 @@ static struct bio *__bio_map_user(struct unsigned long uaddr, unsigned int len, int write_to_vm) { - unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; - unsigned long start = uaddr >> PAGE_SHIFT; + unsigned long end = (uaddr + len + MMUPAGE_SIZE - 1) >> MMUPAGE_SHIFT; + unsigned long start = uaddr >> MMUPAGE_SHIFT; const int nr_pages = end - start; request_queue_t *q = bdev_get_queue(bdev); int ret, offset, i; - struct page **pages; + unsigned long *pages; struct bio *bio; /* @@ -464,7 +464,7 @@ static struct bio *__bio_map_user(struct if (!bio) return NULL; - pages = kmalloc(nr_pages * sizeof(struct page *), GFP_KERNEL); + pages = kmalloc(nr_pages * sizeof(unsigned long), GFP_KERNEL); if (!pages) goto out; @@ -478,9 +478,11 @@ static struct bio *__bio_map_user(struct bio->bi_bdev = bdev; - offset = uaddr & ~PAGE_MASK; + offset = uaddr & ~MMUPAGE_MASK; for (i = 0; i < nr_pages; i++) { - unsigned int bytes = PAGE_SIZE - offset; + unsigned int bytes = MMUPAGE_SIZE - offset; + int suboff = (pages[i] % PAGE_MMUCOUNT)*MMUPAGE_SIZE; + struct page *pg = pfn_to_page(pages[i]); if (len <= 0) break; @@ -491,7 +493,7 @@ static struct bio *__bio_map_user(struct /* * sorry... */ - if (bio_add_page(bio, pages[i], bytes, offset) < bytes) + if (bio_add_page(bio, pg, bytes, offset + suboff) < bytes) break; len -= bytes; @@ -502,7 +504,7 @@ static struct bio *__bio_map_user(struct * release the pages we didn't map into the bio, if any */ while (i < nr_pages) - page_cache_release(pages[i++]); + page_cache_release(pfn_to_page(pages[i++])); kfree(pages); diff -prauN linux-2.5.70-bk4/fs/direct-io.c pgcl-2.5.70-bk4-1/fs/direct-io.c --- linux-2.5.70-bk4/fs/direct-io.c 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/direct-io.c 2003-05-30 21:03:23.000000000 -0700 @@ -35,7 +35,9 @@ /* * How many user pages to map in one call to get_user_pages(). This determines - * the size of a structure on the stack. + * the size of a structure on the stack. But these are mmupages; this + * will _not_ even be able to see a whole PAGE_SIZE area if you make + * PAGE_MMUCOUNT > DIO_PAGES. */ #define DIO_PAGES 64 @@ -49,6 +51,20 @@ * * If blkfactor is zero then the user's request was aligned to the filesystem's * blocksize. + * + * XXX: + * Okay, I just broke this and I'm not sure how to put it back together. + * Basically the issue is that we're pointed at _pfn's_ only by + * get_user_pages() so the assumption of virtual contiguity doesn't even + * guarantee PAGE_SIZE -aligned physical contiguity. + * + * AFAICT the fixup is to "opportunistically" merge all this stuff together + * into PAGE_SIZE-aligned contiguous bits and either special-case or be + * able to handle the rest as they come. I've left this broken for now. + * I'm relatively fearful of eating stackspace to keep count of the number + * mmupages starting at a given pfn there are while merging. + * + * -- wli */ struct dio { @@ -100,7 +116,7 @@ struct dio { * Page queue. These variables belong to dio_refill_pages() and * dio_get_page(). */ - struct page *pages[DIO_PAGES]; /* page buffer */ + unsigned long pages[DIO_PAGES]; /* page buffer */ unsigned head; /* next page to process */ unsigned tail; /* last valid page + 1 */ int page_errors; /* errno from get_user_pages() */ @@ -155,7 +171,7 @@ static int dio_refill_pages(struct dio * */ if (dio->page_errors == 0) dio->page_errors = ret; - dio->pages[0] = ZERO_PAGE(dio->curr_user_address); + dio->pages[0] = page_to_pfn(ZERO_PAGE(dio->curr_user_address)); dio->head = 0; dio->tail = 1; ret = 0; @@ -163,7 +179,7 @@ static int dio_refill_pages(struct dio * } if (ret >= 0) { - dio->curr_user_address += ret * PAGE_SIZE; + dio->curr_user_address += ret * MMUPAGE_SIZE; dio->curr_page += ret; dio->head = 0; dio->tail = ret; @@ -179,8 +195,13 @@ out: * decent number of pages, less frequently. To provide nicer use of the * L1 cache. */ -static struct page *dio_get_page(struct dio *dio) +static struct page *dio_get_page(struct dio *dio, int *pfoff_in_page, int *page_size) { + int pg_size = MMUPAGE_SIZE; + int pfn, tpfn; + struct page *page; + int i = 0; + if (dio_pages_present(dio) == 0) { int ret; @@ -189,7 +210,33 @@ static struct page *dio_get_page(struct return ERR_PTR(ret); BUG_ON(dio_pages_present(dio) == 0); } - return dio->pages[dio->head++]; + + pfn = dio->pages[dio->head++]; + *pfoff_in_page = (pfn % PAGE_MMUCOUNT) * MMUPAGE_SIZE; + + /* Try to cluster all pfns that belongs to this page together */ + tpfn = pfn + 1; + while (pg_size + *pfoff_in_page < PAGE_SIZE) { + if (dio->head == dio->tail) break; + if (tpfn != dio->pages[dio->head]) break; + tpfn++; + dio->head++; + pg_size += MMUPAGE_SIZE; + i++; + } + + page = pfn_to_page(pfn); + *page_size = pg_size; + + /* + * FIXME - get_user_pages got ref for each pfn, we need to drop + * the extra refs for this page + */ + while (i--) { + page_cache_release(page); + } + + return page; } /* @@ -293,8 +340,9 @@ static void dio_bio_submit(struct dio *d */ static void dio_cleanup(struct dio *dio) { + int a, b; while (dio_pages_present(dio)) - page_cache_release(dio_get_page(dio)); + page_cache_release(dio_get_page(dio, &a, &b)); } /* @@ -686,22 +734,26 @@ static void dio_zero_block(struct dio *d static int do_direct_IO(struct dio *dio) { const unsigned blkbits = dio->blkbits; - const unsigned blocks_per_page = PAGE_SIZE >> blkbits; + unsigned blocks_per_page = PAGE_SIZE >> blkbits; struct page *page; unsigned block_in_page; struct buffer_head *map_bh = &dio->map_bh; int ret = 0; + int page_size; + int pf_pgoff; /* The I/O can start at any block offset within the first page */ block_in_page = dio->first_block_in_page; while (dio->block_in_file < dio->final_block_in_request) { - page = dio_get_page(dio); + page = dio_get_page(dio, &pf_pgoff, &page_size); + if (IS_ERR(page)) { ret = PTR_ERR(page); goto out; } + blocks_per_page = page_size >> blkbits; while (block_in_page < blocks_per_page) { unsigned offset_in_page = block_in_page << blkbits; unsigned this_chunk_bytes; /* # of bytes mapped */ @@ -785,7 +837,7 @@ do_holes: * can add to this page */ this_chunk_blocks = dio->blocks_available; - u = (PAGE_SIZE - offset_in_page) >> blkbits; + u = (page_size - offset_in_page) >> blkbits; if (this_chunk_blocks > u) this_chunk_blocks = u; u = dio->final_block_in_request - dio->block_in_file; @@ -795,7 +847,7 @@ do_holes: BUG_ON(this_chunk_bytes == 0); dio->boundary = buffer_boundary(map_bh); - ret = submit_page_section(dio, page, offset_in_page, + ret = submit_page_section(dio, page, pf_pgoff + offset_in_page, this_chunk_bytes, dio->next_block_for_io); if (ret) { page_cache_release(page); @@ -882,7 +934,7 @@ direct_io_worker(int rw, struct kiocb *i bytes = iov[seg].iov_len; /* Index into the first page of the first block */ - dio->first_block_in_page = (user_addr & ~PAGE_MASK) >> blkbits; + dio->first_block_in_page = (user_addr & ~MMUPAGE_MASK) >> blkbits; dio->final_block_in_request = dio->block_in_file + (bytes >> blkbits); /* Page fetching state */ @@ -891,11 +943,11 @@ direct_io_worker(int rw, struct kiocb *i dio->curr_page = 0; dio->total_pages = 0; - if (user_addr & (PAGE_SIZE-1)) { + if (user_addr & (MMUPAGE_SIZE-1)) { dio->total_pages++; - bytes -= PAGE_SIZE - (user_addr & (PAGE_SIZE - 1)); + bytes -= MMUPAGE_SIZE - (user_addr & (MMUPAGE_SIZE - 1)); } - dio->total_pages += (bytes + PAGE_SIZE - 1) / PAGE_SIZE; + dio->total_pages += (bytes + MMUPAGE_SIZE - 1) / MMUPAGE_SIZE; dio->curr_user_address = user_addr; ret = do_direct_IO(dio); diff -prauN linux-2.5.70-bk4/fs/exec.c pgcl-2.5.70-bk4-1/fs/exec.c --- linux-2.5.70-bk4/fs/exec.c 2003-05-26 18:00:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/exec.c 2003-05-30 21:03:23.000000000 -0700 @@ -285,51 +285,64 @@ int copy_strings_kernel(int argc,char ** * This routine is used to map in a page into an address space: needed by * execve() for the initial stack and environment pages. * - * tsk->mmap_sem is held for writing. + * task->mm->mmap_sem is held for writing. */ -void put_dirty_page(struct task_struct *tsk, struct page *page, - unsigned long address, pgprot_t prot) +void put_dirty_page(struct task_struct *task, struct page *page, int min_subpfn, + unsigned long addr, pgprot_t prot) { - pgd_t * pgd; - pmd_t * pmd; - pte_t * pte; - struct pte_chain *pte_chain; - - if (page_count(page) != 1) - printk(KERN_ERR "mem_map disagrees with %p at %08lx\n", - page, address); - - pgd = pgd_offset(tsk->mm, address); - pte_chain = pte_chain_alloc(GFP_KERNEL); - if (!pte_chain) - goto out_sig; - spin_lock(&tsk->mm->page_table_lock); - pmd = pmd_alloc(tsk->mm, pgd, address); - if (!pmd) - goto out; - pte = pte_alloc_map(tsk->mm, pmd, address); - if (!pte) - goto out; - if (!pte_none(*pte)) { + unsigned long page_pfn, subpfn; + struct pte_chain *pte_chain = NULL; + struct mm_struct *mm = task->mm; + + page_pfn = page_to_pfn(page); + spin_lock(&mm->page_table_lock); + + addr += MMUPAGE_SIZE * min_subpfn; + for (subpfn = min_subpfn; subpfn < PAGE_MMUCOUNT; ++subpfn, addr += MMUPAGE_SIZE) { + pgd_t *pgd; + pmd_t *pmd; + pte_t *pte; + unsigned long pfn; + + pgd = pgd_offset(mm, addr); + if (!pte_chain) + pte_chain = pte_chain_alloc(GFP_ATOMIC); + if (!pte_chain) { + spin_unlock(&mm->page_table_lock); + pte_chain = pte_chain_alloc(GFP_KERNEL); + if (!pte_chain) + goto out_nolock; + spin_lock(&mm->page_table_lock); + } + pmd = pmd_alloc(mm, pgd, addr); + if (!pmd) + goto out; + pte = pte_alloc_map(mm, pmd, addr); + if (!pte) + goto out; + if (!pte_none(*pte)) { + pte_unmap(pte); + continue; + } + pfn = page_pfn + subpfn; + set_pte(pte, pte_mkdirty(pte_mkwrite(pfn_pte(pfn, prot)))); + pte_chain = page_add_rmap(page, pte, pte_chain); + page_cache_get(page); pte_unmap(pte); - goto out; + task->mm->rss++; } + spin_unlock(&mm->page_table_lock); + pte_chain_free(pte_chain); lru_cache_add_active(page); flush_dcache_page(page); - set_pte(pte, pte_mkdirty(pte_mkwrite(mk_pte(page, prot)))); - pte_chain = page_add_rmap(page, pte, pte_chain); - pte_unmap(pte); - tsk->mm->rss++; - spin_unlock(&tsk->mm->page_table_lock); - /* no need for flush_tlb */ - pte_chain_free(pte_chain); + page_cache_release(page); /* want to add PAGE_MMUCOUNT-1 */ return; out: - spin_unlock(&tsk->mm->page_table_lock); -out_sig: - __free_page(page); - force_sig(SIGKILL, tsk); + spin_unlock(&mm->page_table_lock); +out_nolock: + page_cache_release(page); + force_sig(SIGKILL, task); pte_chain_free(pte_chain); return; } @@ -392,7 +405,8 @@ int setup_arg_pages(struct linux_binprm if (!mpnt) return -ENOMEM; - if (!vm_enough_memory((STACK_TOP - (PAGE_MASK & (unsigned long) bprm->p))>>PAGE_SHIFT)) { + /* must match the length of mpnt below */ + if (!vm_enough_memory((STACK_TOP-(MMUPAGE_MASK&(unsigned long)bprm->p))/MMUPAGE_SIZE)) { kmem_cache_free(vm_area_cachep, mpnt); return -ENOMEM; } @@ -402,10 +416,9 @@ int setup_arg_pages(struct linux_binprm mpnt->vm_mm = mm; #ifdef CONFIG_STACK_GROWSUP mpnt->vm_start = stack_base; - mpnt->vm_end = PAGE_MASK & - (PAGE_SIZE - 1 + (unsigned long) bprm->p); + mpnt->vm_end = MMUPAGE_ALIGN((unsigned long)bprm->p); #else - mpnt->vm_start = PAGE_MASK & (unsigned long) bprm->p; + mpnt->vm_start = MMUPAGE_MASK & (unsigned long)bprm->p; mpnt->vm_end = STACK_TOP; #endif mpnt->vm_page_prot = protection_map[VM_STACK_FLAGS & 0x7]; @@ -414,17 +427,23 @@ int setup_arg_pages(struct linux_binprm mpnt->vm_pgoff = 0; mpnt->vm_file = NULL; INIT_LIST_HEAD(&mpnt->shared); - mpnt->vm_private_data = (void *) 0; + mpnt->vm_private_data = NULL; insert_vm_struct(mm, mpnt); - mm->total_vm = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; + mm->total_vm = (mpnt->vm_end - mpnt->vm_start) >> MMUPAGE_SHIFT; } for (i = 0 ; i < MAX_ARG_PAGES ; i++) { struct page *page = bprm->page[i]; if (page) { + int min_subpfn; + + if (mpnt->vm_start <= stack_base) + min_subpfn = 0; + else + min_subpfn = (mpnt->vm_start - stack_base)/MMUPAGE_SIZE; bprm->page[i] = NULL; - put_dirty_page(current, page, stack_base, - mpnt->vm_page_prot); + put_dirty_page(current, page, min_subpfn, stack_base, + mpnt->vm_page_prot); } stack_base += PAGE_SIZE; } @@ -1122,6 +1141,7 @@ out_file: allow_write_access(bprm.file); fput(bprm.file); } + return retval; } diff -prauN linux-2.5.70-bk4/fs/ext2/dir.c pgcl-2.5.70-bk4-1/fs/ext2/dir.c --- linux-2.5.70-bk4/fs/ext2/dir.c 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/ext2/dir.c 2003-05-30 21:03:23.000000000 -0700 @@ -432,15 +432,15 @@ int ext2_add_link (struct dentry *dentry struct inode *dir = dentry->d_parent->d_inode; const char *name = dentry->d_name.name; int namelen = dentry->d_name.len; - unsigned chunk_size = ext2_chunk_size(dir); - unsigned reclen = EXT2_DIR_REC_LEN(namelen); - unsigned short rec_len, name_len; + unsigned long chunk_size = ext2_chunk_size(dir); + unsigned long reclen = EXT2_DIR_REC_LEN(namelen); + unsigned long rec_len, name_len; struct page *page = NULL; ext2_dirent * de; unsigned long npages = dir_pages(dir); unsigned long n; char *kaddr; - unsigned from, to; + unsigned long from, to; int err; /* diff -prauN linux-2.5.70-bk4/fs/file_table.c pgcl-2.5.70-bk4-1/fs/file_table.c --- linux-2.5.70-bk4/fs/file_table.c 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/file_table.c 2003-05-30 21:03:23.000000000 -0700 @@ -279,7 +279,7 @@ void __init files_init(unsigned long mem * Per default don't use more than 10% of our memory for files. */ - n = (mempages * (PAGE_SIZE / 1024)) / 10; + n = (mempages * (MMUPAGE_SIZE / 1024)) / 10; files_stat.max_files = n; if (files_stat.max_files < NR_FILE) files_stat.max_files = NR_FILE; diff -prauN linux-2.5.70-bk4/fs/inode.c pgcl-2.5.70-bk4-1/fs/inode.c --- linux-2.5.70-bk4/fs/inode.c 2003-05-26 18:01:00.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/inode.c 2003-05-30 21:03:23.000000000 -0700 @@ -1310,7 +1310,11 @@ void __init inode_init(unsigned long mem for (i = 0; i < ARRAY_SIZE(i_wait_queue_heads); i++) init_waitqueue_head(&i_wait_queue_heads[i].wqh); +#if PAGE_SHIFT <= 14 mempages >>= (14 - PAGE_SHIFT); +#else + mempages <<= PAGE_SHIFT - 14; +#endif mempages *= sizeof(struct hlist_head); for (order = 0; ((1UL << order) << PAGE_SHIFT) < mempages; order++) ; diff -prauN linux-2.5.70-bk4/fs/proc/base.c pgcl-2.5.70-bk4-1/fs/proc/base.c --- linux-2.5.70-bk4/fs/proc/base.c 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/proc/base.c 2003-05-30 21:03:23.000000000 -0700 @@ -32,6 +32,7 @@ #include #include #include +#include /* * For hysterical raisins we keep the same inumbers as in the old procfs. @@ -456,29 +457,37 @@ static ssize_t mem_read(struct file * fi size_t count, loff_t *ppos) { struct task_struct *task = proc_task(file->f_dentry->d_inode); - char *page; + char *kbuf; + struct page *page; unsigned long src = *ppos; int ret = -ESRCH; struct mm_struct *mm; - if (!MAY_PTRACE(task)) + if (0 && !MAY_PTRACE(task)) goto out; ret = -ENOMEM; - page = (char *)__get_free_page(GFP_USER); - if (!page) + page = alloc_page(GFP_HIGHUSER); + if (!page) { + printk("alloc_page() failed in mem_read()\n"); goto out; + } + kbuf = kmap(page); ret = 0; mm = get_task_mm(task); - if (!mm) + if (!mm) { + printk("get_task_mm() failed in mem_read()\n"); goto out_free; + } +#if 0 ret = -EIO; if (file->private_data != (void*)((long)current->self_exec_id)) goto out_put; +#endif ret = 0; @@ -486,14 +495,16 @@ static ssize_t mem_read(struct file * fi int this_len, retval; this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; - retval = access_process_vm(task, src, page, this_len, 0); + retval = access_process_vm(task, src, kbuf, this_len, 0); if (!retval) { + printk("access_process_vm() failed in mem_read()\n"); if (!ret) ret = -EIO; break; } - if (copy_to_user(buf, page, retval)) { + if (copy_to_user(buf, kbuf, retval)) { + printk("copy_to_user() failed in mem_read()\n"); ret = -EFAULT; break; } @@ -505,15 +516,17 @@ static ssize_t mem_read(struct file * fi } *ppos = src; -out_put: mmput(mm); out_free: - free_page((unsigned long) page); + kunmap(page); + __free_page(page); out: return ret; } +#if 0 #define mem_write NULL +#endif #ifndef mem_write /* This is a security hazard */ @@ -521,26 +534,28 @@ static ssize_t mem_write(struct file * f size_t count, loff_t *ppos) { int copied = 0; - char *page; + char *kbuf; + struct page *page; struct task_struct *task = proc_task(file->f_dentry->d_inode); unsigned long dst = *ppos; - if (!MAY_PTRACE(task)) + if (0 && !MAY_PTRACE(task)) return -ESRCH; - page = (char *)__get_free_page(GFP_USER); + page = alloc_page(GFP_HIGHUSER); if (!page) return -ENOMEM; + kbuf = kmap(page); while (count > 0) { int this_len, retval; this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; - if (copy_from_user(page, buf, this_len)) { + if (copy_from_user(kbuf, buf, this_len)) { copied = -EFAULT; break; } - retval = access_process_vm(task, dst, page, this_len, 1); + retval = access_process_vm(task, dst, kbuf, this_len, 1); if (!retval) { if (!copied) copied = -EIO; @@ -552,7 +567,8 @@ static ssize_t mem_write(struct file * f count -= retval; } *ppos = dst; - free_page((unsigned long) page); + kunmap(page); + __free_page(page); return copied; } #endif diff -prauN linux-2.5.70-bk4/fs/proc/proc_misc.c pgcl-2.5.70-bk4-1/fs/proc/proc_misc.c --- linux-2.5.70-bk4/fs/proc/proc_misc.c 2003-05-26 18:00:24.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/proc/proc_misc.c 2003-05-30 21:03:23.000000000 -0700 @@ -241,7 +241,7 @@ static int meminfo_read_proc(char *page, K(ps.nr_writeback), K(ps.nr_mapped), K(ps.nr_slab), - K(committed), + committed << (MMUPAGE_SHIFT - 10), K(ps.nr_page_table_pages), vmtot, vmi.used, diff -prauN linux-2.5.70-bk4/fs/proc/task_mmu.c pgcl-2.5.70-bk4-1/fs/proc/task_mmu.c --- linux-2.5.70-bk4/fs/proc/task_mmu.c 2003-05-26 18:00:24.000000000 -0700 +++ pgcl-2.5.70-bk4-1/fs/proc/task_mmu.c 2003-05-30 21:03:23.000000000 -0700 @@ -56,7 +56,7 @@ int task_statm(struct mm_struct *mm, int *resident = mm->rss; for (vma = mm->mmap; vma; vma = vma->vm_next) { - int pages = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; + int pages = (vma->vm_end - vma->vm_start) >> MMUPAGE_SHIFT; size += pages; if (is_vm_hugetlb_page(vma)) { diff -prauN linux-2.5.70-bk4/include/asm-alpha/page.h pgcl-2.5.70-bk4-1/include/asm-alpha/page.h --- linux-2.5.70-bk4/include/asm-alpha/page.h 2003-05-26 18:00:45.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-alpha/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -98,6 +98,8 @@ extern __inline__ int get_order(unsigned #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif /* _ALPHA_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-arm/page.h pgcl-2.5.70-bk4-1/include/asm-arm/page.h --- linux-2.5.70-bk4/include/asm-arm/page.h 2003-05-26 18:00:25.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-arm/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -181,6 +181,8 @@ static inline int get_order(unsigned lon #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif diff -prauN linux-2.5.70-bk4/include/asm-cris/page.h pgcl-2.5.70-bk4-1/include/asm-cris/page.h --- linux-2.5.70-bk4/include/asm-cris/page.h 2003-05-26 18:00:27.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-cris/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -109,6 +109,8 @@ extern unsigned long dram_start, dram_en #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif /* _CRIS_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-generic/page.h pgcl-2.5.70-bk4-1/include/asm-generic/page.h --- linux-2.5.70-bk4/include/asm-generic/page.h 1969-12-31 16:00:00.000000000 -0800 +++ pgcl-2.5.70-bk4-1/include/asm-generic/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -0,0 +1,11 @@ +#ifndef _ASM_GENERIC_PAGE_H +#define _ASM_GENERIC_PAGE_H + +#define MMUPAGE_SHIFT PAGE_SHIFT +#define MMUPAGE_SIZE PAGE_SIZE +#define MMUPAGE_MASK PAGE_MASK +#define MMUPAGE_ALIGN(x) PAGE_ALIGN(x) +#define PAGE_MMUSHIFT 0 +#define PAGE_MMUCOUNT 1 + +#endif /* _ASM_GENERIC_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-generic/rmap.h pgcl-2.5.70-bk4-1/include/asm-generic/rmap.h --- linux-2.5.70-bk4/include/asm-generic/rmap.h 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-generic/rmap.h 2003-05-30 21:03:23.000000000 -0700 @@ -15,7 +15,7 @@ * offset of the page table entry within the page table page * * For CONFIG_HIGHPTE, we need to represent the address of a pte in a - * scalar pte_addr_t. The pfn of the pte's page is shifted left by PAGE_SIZE + * scalar pte_addr_t. The pfn of the pte's page is shifted left by MMUPAGE_SIZE * bits and is then ORed with the byte offset of the pte within its page. * * For CONFIG_HIGHMEM4G, the pte_addr_t is 32 bits. 20 for the pfn, 12 for @@ -26,7 +26,15 @@ */ #include -static inline void pgtable_add_rmap(struct page * page, struct mm_struct * mm, unsigned long address) +/* + * This looks bizarre, but it's actually meaningful. + */ +#define MMUPAGES_MAPPED_PER_PTE_PAGE (PTRS_PER_PTE * PAGE_MMUCOUNT) +#define VIRT_AREA_MAPPED_PER_PTE_PAGE \ + (MMUPAGES_MAPPED_PER_PTE_PAGE*MMUPAGE_SIZE) + +static inline void pgtable_add_rmap(struct page *page, struct mm_struct *mm, + unsigned long address) { #ifdef BROKEN_PPC_PTE_ALLOC_ONE /* OK, so PPC calls pte_alloc() before mem_map[] is setup ... ;( */ @@ -35,13 +43,38 @@ static inline void pgtable_add_rmap(stru if (!mem_init_done) return; #endif + + /* rmap's accounting is already set up */ + if (page->mapping) { + /* + * address is presumably large. if smaller, overflow traps + * the error; if larger, check the distance + */ + WARN_ON(address - page->index >= VIRT_AREA_MAPPED_PER_PTE_PAGE); + return; + } + page->mapping = (void *)mm; - page->index = address & ~((PTRS_PER_PTE * PAGE_SIZE) - 1); + page->index = address & ~(VIRT_AREA_MAPPED_PER_PTE_PAGE - 1); inc_page_state(nr_page_table_pages); } static inline void pgtable_remove_rmap(struct page * page) { + /* we're not down to a unique reference */ + if (PAGE_MMUCOUNT > 1) { + if (atomic_read(&page->count) > 1) + return; + + /* + * A zero reference count should not be possible; + * put_page() should have freed the things outright + * so this essentially means use-after-free is happening. + */ + else + BUG_ON(atomic_read(&page->count) <= 0); + } + page->mapping = NULL; page->index = 0; dec_page_state(nr_page_table_pages); @@ -55,18 +88,19 @@ static inline struct mm_struct * ptep_to static inline unsigned long ptep_to_address(pte_t * ptep) { - struct page * page = kmap_atomic_to_page(ptep); - unsigned long low_bits; - low_bits = ((unsigned long)ptep & ~PAGE_MASK) * PTRS_PER_PTE; - return page->index + low_bits; + struct page *page = kmap_atomic_to_page(ptep); + unsigned long swpage_voff = ((unsigned long)ptep)/sizeof(pte_t); + swpage_voff %= MMUPAGES_MAPPED_PER_PTE_PAGE; + return page->index + MMUPAGE_SIZE*swpage_voff; } #ifdef CONFIG_HIGHPTE static inline pte_addr_t ptep_to_paddr(pte_t *ptep) { - pte_addr_t paddr; - paddr = ((pte_addr_t)page_to_pfn(kmap_atomic_to_page(ptep))) << PAGE_SHIFT; - return paddr + (pte_addr_t)((unsigned long)ptep & ~PAGE_MASK); + unsigned long pfn, subpfn, vaddr = (unsigned long)ptep; + subpfn = (vaddr/MMUPAGE_SIZE) % PAGE_MMUCOUNT; + pfn = page_to_pfn(kmap_atomic_to_page(ptep)) + subpfn; + return MMUPAGE_SIZE*((pte_addr_t)pfn) + (vaddr & ~MMUPAGE_MASK); } #else static inline pte_addr_t ptep_to_paddr(pte_t *ptep) diff -prauN linux-2.5.70-bk4/include/asm-i386/dma-mapping.h pgcl-2.5.70-bk4-1/include/asm-i386/dma-mapping.h --- linux-2.5.70-bk4/include/asm-i386/dma-mapping.h 2003-05-26 18:01:03.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/dma-mapping.h 2003-05-30 21:03:23.000000000 -0700 @@ -51,7 +51,7 @@ dma_map_page(struct device *dev, struct size_t size, enum dma_data_direction direction) { BUG_ON(direction == DMA_NONE); - return (dma_addr_t)(page_to_pfn(page)) * PAGE_SIZE + offset; + return (dma_addr_t)(page_to_pfn(page)) * MMUPAGE_SIZE + offset; } static inline void diff -prauN linux-2.5.70-bk4/include/asm-i386/fixmap.h pgcl-2.5.70-bk4-1/include/asm-i386/fixmap.h --- linux-2.5.70-bk4/include/asm-i386/fixmap.h 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/fixmap.h 2003-05-30 21:03:23.000000000 -0700 @@ -41,8 +41,29 @@ * TLB entries of such buffers will not be flushed across * task switches. */ + +/* + * Right now we initialize only a single pte table. It can be extended + * easily, subsequent pte tables have to be allocated in one physical + * chunk of RAM. + */ +#define PKMAP_NR(virt) (((virt) - PKMAP_BASE) >> PAGE_SHIFT) +#define PKMAP_ADDR(nr) (PKMAP_BASE + ((nr) << PAGE_SHIFT)) +#define LAST_PKMAP 1024 +#define LAST_PKMAP_MASK (LAST_PKMAP-1) + enum fixed_addresses { - FIX_HOLE, + /* + * leave a hole of exactly PAGE_SIZE at the top for CONFIG_HIGHMEM + * this makes things easier on core code; the math works out funny + */ + FIX_HOLE = PAGE_MMUCOUNT > 1 ? PAGE_MMUCOUNT - 1 : 0, +#ifdef CONFIG_HIGHMEM + FIX_KMAP_BEGIN, /* reserved pte's for temporary kernel mappings */ + FIX_KMAP_END = FIX_KMAP_BEGIN+((KM_TYPE_NR*NR_CPUS+1)*PAGE_MMUCOUNT)-1, + FIX_PKMAP_BEGIN, + FIX_PKMAP_END = FIX_PKMAP_BEGIN + (LAST_PKMAP+1)*PAGE_MMUCOUNT - 1, +#endif FIX_VSYSCALL, #ifdef CONFIG_X86_LOCAL_APIC FIX_APIC_BASE, /* local (CPU) APIC) -- required for SMP or not */ @@ -63,10 +84,6 @@ enum fixed_addresses { #ifdef CONFIG_X86_CYCLONE_TIMER FIX_CYCLONE_TIMER, /*cyclone timer register*/ #endif -#ifdef CONFIG_HIGHMEM - FIX_KMAP_BEGIN, /* reserved pte's for temporary kernel mappings */ - FIX_KMAP_END = FIX_KMAP_BEGIN+(KM_TYPE_NR*NR_CPUS)-1, -#endif #ifdef CONFIG_ACPI_BOOT FIX_ACPI_BEGIN, FIX_ACPI_END = FIX_ACPI_BEGIN + FIX_ACPI_PAGES - 1, @@ -101,11 +118,11 @@ extern void __set_fixmap (enum fixed_add * the start of the fixmap. */ #define FIXADDR_TOP (0xfffff000UL) -#define __FIXADDR_SIZE (__end_of_permanent_fixed_addresses << PAGE_SHIFT) +#define __FIXADDR_SIZE (__end_of_permanent_fixed_addresses << MMUPAGE_SHIFT) #define FIXADDR_START (FIXADDR_TOP - __FIXADDR_SIZE) -#define __fix_to_virt(x) (FIXADDR_TOP - ((x) << PAGE_SHIFT)) -#define __virt_to_fix(x) ((FIXADDR_TOP - ((x)&PAGE_MASK)) >> PAGE_SHIFT) +#define __fix_to_virt(x) (FIXADDR_TOP - ((x) << MMUPAGE_SHIFT)) +#define __virt_to_fix(x) ((FIXADDR_TOP - ((x) & MMUPAGE_MASK)) >> MMUPAGE_SHIFT) extern void __this_fixmap_does_not_exist(void); @@ -133,8 +150,13 @@ static inline unsigned long fix_to_virt( static inline unsigned long virt_to_fix(const unsigned long vaddr) { - BUG_ON(vaddr >= FIXADDR_TOP || vaddr < FIXADDR_START); + if (vaddr >= FIXADDR_TOP || vaddr < FIXADDR_START) { + printk("bad vaddr in virt_to_fix 0x%lx\n", vaddr); + BUG(); + } return __virt_to_fix(vaddr); } +#define PKMAP_BASE fix_to_virt(FIX_PKMAP_END) + #endif diff -prauN linux-2.5.70-bk4/include/asm-i386/highmem.h pgcl-2.5.70-bk4-1/include/asm-i386/highmem.h --- linux-2.5.70-bk4/include/asm-i386/highmem.h 2003-05-26 18:00:23.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/highmem.h 2003-05-30 21:03:23.000000000 -0700 @@ -34,23 +34,8 @@ extern pte_t *pkmap_page_table; extern void kmap_init(void); -/* - * Right now we initialize only a single pte table. It can be extended - * easily, subsequent pte tables have to be allocated in one physical - * chunk of RAM. - */ -#define PKMAP_BASE (0xff800000UL) -#ifdef CONFIG_X86_PAE -#define LAST_PKMAP 512 -#else -#define LAST_PKMAP 1024 -#endif -#define LAST_PKMAP_MASK (LAST_PKMAP-1) -#define PKMAP_NR(virt) ((virt-PKMAP_BASE) >> PAGE_SHIFT) -#define PKMAP_ADDR(nr) (PKMAP_BASE + ((nr) << PAGE_SHIFT)) - -extern void * FASTCALL(kmap_high(struct page *page)); -extern void FASTCALL(kunmap_high(struct page *page)); +void *FASTCALL(kmap_high(struct page *page)); +void FASTCALL(kunmap_high(struct page *page)); void *kmap(struct page *page); void kunmap(struct page *page); diff -prauN linux-2.5.70-bk4/include/asm-i386/io.h pgcl-2.5.70-bk4-1/include/asm-i386/io.h --- linux-2.5.70-bk4/include/asm-i386/io.h 2003-05-26 18:01:00.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/io.h 2003-05-30 21:03:23.000000000 -0700 @@ -95,7 +95,7 @@ static inline void * phys_to_virt(unsign /* * Change "struct page" to physical address. */ -#define page_to_phys(page) ((dma_addr_t)page_to_pfn(page) << PAGE_SHIFT) +#define page_to_phys(page) ((dma_addr_t)page_to_pfn(page) << MMUPAGE_SHIFT) extern void * __ioremap(unsigned long offset, unsigned long size, unsigned long flags); diff -prauN linux-2.5.70-bk4/include/asm-i386/io_apic.h pgcl-2.5.70-bk4-1/include/asm-i386/io_apic.h --- linux-2.5.70-bk4/include/asm-i386/io_apic.h 2003-05-26 18:00:42.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/io_apic.h 2003-05-30 21:03:23.000000000 -0700 @@ -17,7 +17,7 @@ #define IO_APIC_BASE(idx) \ ((volatile int *)(__fix_to_virt(FIX_IO_APIC_BASE_0 + idx) \ - + (mp_ioapics[idx].mpc_apicaddr & ~PAGE_MASK))) + + (mp_ioapics[idx].mpc_apicaddr & ~MMUPAGE_MASK))) /* * The structure of the IO-APIC: diff -prauN linux-2.5.70-bk4/include/asm-i386/mmzone.h pgcl-2.5.70-bk4-1/include/asm-i386/mmzone.h --- linux-2.5.70-bk4/include/asm-i386/mmzone.h 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/mmzone.h 2003-05-30 21:03:23.000000000 -0700 @@ -22,18 +22,18 @@ extern struct pglist_data *node_data[]; #define alloc_bootmem_low(x) \ __alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES, 0) #define alloc_bootmem_pages(x) \ - __alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, __pa(MAX_DMA_ADDRESS)) + __alloc_bootmem_node(NODE_DATA(0), (x), MMUPAGE_SIZE, __pa(MAX_DMA_ADDRESS)) #define alloc_bootmem_low_pages(x) \ - __alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, 0) + __alloc_bootmem_node(NODE_DATA(0), (x), MMUPAGE_SIZE, 0) #define alloc_bootmem_node(ignore, x) \ __alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES, __pa(MAX_DMA_ADDRESS)) #define alloc_bootmem_pages_node(ignore, x) \ - __alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, __pa(MAX_DMA_ADDRESS)) + __alloc_bootmem_node(NODE_DATA(0), (x), MMUPAGE_SIZE, __pa(MAX_DMA_ADDRESS)) #define alloc_bootmem_low_pages_node(ignore, x) \ - __alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, 0) + __alloc_bootmem_node(NODE_DATA(0), (x), MMUPAGE_SIZE, 0) #define node_size(nid) (node_data[nid]->node_size) -#define node_localnr(pfn, nid) ((pfn) - node_data[nid]->node_start_pfn) +#define node_localnr(pfn, nid) (((pfn) - node_data[nid]->node_start_pfn) / PAGE_MMUCOUNT) /* * Following are macros that each numa implmentation must define. @@ -42,25 +42,41 @@ extern struct pglist_data *node_data[]; /* * Given a kernel address, find the home node of the underlying memory. */ -#define kvaddr_to_nid(kaddr) pfn_to_nid(__pa(kaddr) >> PAGE_SHIFT) +#define kvaddr_to_nid(kaddr) pfn_to_nid(__pa(kaddr) >> MMUPAGE_SHIFT) /* * Return a pointer to the node data for node n. */ #define NODE_DATA(nid) (node_data[nid]) +/* + * These names clash. I blame mbligh. + */ #define node_mem_map(nid) (NODE_DATA(nid)->node_mem_map) #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn) + +/* + * pgdat->node_size is calculated from zone_sizes[], which is in + * units of PAGE_SIZE. I don't trust this. + */ #define node_end_pfn(nid) \ ({ \ pg_data_t *__pgdat = NODE_DATA(nid); \ - __pgdat->node_start_pfn + __pgdat->node_size; \ + __pgdat->node_start_pfn + __pgdat->node_size*PAGE_MMUCOUNT; \ }) #define local_mapnr(kvaddr) \ ({ \ - unsigned long __pfn = __pa(kvaddr) >> PAGE_SHIFT; \ - (__pfn - node_start_pfn(pfn_to_nid(__pfn))); \ + unsigned long __pfn = __pa(kvaddr) >> MMUPAGE_SHIFT; \ + (__pfn - node_start_pfn(pfn_to_nid(__pfn)))/PAGE_MMUCOUNT; \ +}) + +#define local_pfn(pg) \ +({ \ + struct page *__pg = pg; \ + unsigned long __nr; \ + __nr = (unsigned long)(__pg - page_zone(__pg)->zone_mem_map); \ + __nr*PAGE_MMUCOUNT; \ }) #define kern_addr_valid(kaddr) \ @@ -81,10 +97,9 @@ extern struct pglist_data *node_data[]; ({ \ struct page *__page = pg; \ struct zone *__zone = page_zone(__page); \ - (unsigned long)(__page - __zone->zone_mem_map) \ - + __zone->zone_start_pfn; \ + local_pfn(__page) + __zone->zone_start_pfn; \ }) -#define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)) +#define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> MMUPAGE_SHIFT)) /* * pfn_valid should be made as fast as possible, and the current definition * is valid for machines that are NUMA, but still contiguous, which is what diff -prauN linux-2.5.70-bk4/include/asm-i386/numaq.h pgcl-2.5.70-bk4-1/include/asm-i386/numaq.h --- linux-2.5.70-bk4/include/asm-i386/numaq.h 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/numaq.h 2003-05-30 21:03:23.000000000 -0700 @@ -28,7 +28,7 @@ #ifdef CONFIG_X86_NUMAQ -#define MAX_NUMNODES 8 +#define MAX_NUMNODES 16 extern void get_memcfg_numaq(void); #define get_memcfg_numa() get_memcfg_numaq() @@ -159,7 +159,7 @@ struct sys_cfg_data { static inline unsigned long *get_zholes_size(int nid) { - return 0; + return NULL; } #endif /* CONFIG_X86_NUMAQ */ #endif /* NUMAQ_H */ diff -prauN linux-2.5.70-bk4/include/asm-i386/page.h pgcl-2.5.70-bk4-1/include/asm-i386/page.h --- linux-2.5.70-bk4/include/asm-i386/page.h 2003-05-26 18:00:24.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -1,13 +1,36 @@ #ifndef _I386_PAGE_H #define _I386_PAGE_H -/* PAGE_SHIFT determines the page size */ -#define PAGE_SHIFT 12 +#include /* for CONFIG_PAGE_CLUSTER */ + +/* + * One mmupage is represented by one Page Table Entry at the MMU level, + * and corresponds to one page at the user process level: its size is + * the same as param.h EXEC_PAGESIZE (for getpagesize(2) and mmap(2)). + */ +#define MMUPAGE_SHIFT 12 +#define MMUPAGE_SIZE (1 << MMUPAGE_SHIFT) +#define MMUPAGE_MASK (~(MMUPAGE_SIZE-1)) + +/* + * 2**N adjacent mmupages may be clustered to make up one kernel page. + * Reasonable and tested values for PAGE_MMUSHIFT are 0 (4k page), + * 1 (8k page), 2 (16k page), 3 (32k page). Higher values will not + * work without further changes e.g. to unsigned short b_size. + */ +#define PAGE_MMUSHIFT CONFIG_PAGE_CLUSTER +#define PAGE_MMUCOUNT (1 << PAGE_MMUSHIFT) + +/* + * One kernel page is represented by one struct page (see mm.h), + * and is the kernel's principal unit of memory allocation. + */ +#define PAGE_SHIFT (PAGE_MMUSHIFT + MMUPAGE_SHIFT) #define PAGE_SIZE (1UL << PAGE_SHIFT) #define PAGE_MASK (~(PAGE_SIZE-1)) #define LARGE_PAGE_MASK (~(LARGE_PAGE_SIZE-1)) -#define LARGE_PAGE_SIZE (1UL << PMD_SHIFT) +#define LARGE_PAGE_SIZE (1 << PMD_SHIFT) #ifdef __KERNEL__ #ifndef __ASSEMBLY__ @@ -53,7 +76,7 @@ typedef struct { unsigned long pgd; } pg #define pte_val(x) ((x).pte_low) #define HPAGE_SHIFT 22 #endif -#define PTE_MASK PAGE_MASK +#define PTE_MASK MMUPAGE_MASK #ifdef CONFIG_HUGETLB_PAGE #define HPAGE_SIZE ((1UL) << HPAGE_SHIFT) @@ -76,6 +99,7 @@ typedef struct { unsigned long pgprot; } /* to align the pointer to the (next) page boundary */ #define PAGE_ALIGN(addr) (((addr)+PAGE_SIZE-1)&PAGE_MASK) +#define MMUPAGE_ALIGN(addr) (((addr)+MMUPAGE_SIZE-1)&MMUPAGE_MASK) /* * This handles the memory map.. We could make this a config @@ -123,18 +147,22 @@ static __inline__ int get_order(unsigned #define PAGE_OFFSET ((unsigned long)__PAGE_OFFSET) #define VMALLOC_RESERVE ((unsigned long)__VMALLOC_RESERVE) -#define MAXMEM (-__PAGE_OFFSET-__VMALLOC_RESERVE) +#define __MAXMEM \ + ((VMALLOC_START-2*MMUPAGE_SIZE-__PAGE_OFFSET) & LARGE_PAGE_MASK) +#define MAXMEM \ + __pa((VMALLOC_START-2*MMUPAGE_SIZE) & LARGE_PAGE_MASK) #define __pa(x) ((unsigned long)(x)-PAGE_OFFSET) #define __va(x) ((void *)((unsigned long)(x)+PAGE_OFFSET)) -#define pfn_to_kaddr(pfn) __va((pfn) << PAGE_SHIFT) +#define pfn_to_kaddr(pfn) __va(MMUPAGE_SIZE*(pfn)) #ifndef CONFIG_DISCONTIGMEM -#define pfn_to_page(pfn) (mem_map + (pfn)) -#define page_to_pfn(page) ((unsigned long)((page) - mem_map)) -#define pfn_valid(pfn) ((pfn) < max_mapnr) +#define pfn_to_page(pfn) (&mem_map[(pfn)/PAGE_MMUCOUNT]) +#define page_to_mapnr(page) ((unsigned long)((page) - mem_map)) +#define page_to_pfn(page) (PAGE_MMUCOUNT*page_to_mapnr(page)) +#define pfn_valid(pfn) ((pfn) < max_mapnr*PAGE_MMUCOUNT) #endif /* !CONFIG_DISCONTIGMEM */ -#define virt_to_page(kaddr) pfn_to_page(__pa(kaddr) >> PAGE_SHIFT) +#define virt_to_page(kaddr) pfn_to_page(__pa(kaddr)/MMUPAGE_SIZE) -#define virt_addr_valid(kaddr) pfn_valid(__pa(kaddr) >> PAGE_SHIFT) +#define virt_addr_valid(kaddr) pfn_valid(__pa(kaddr)/MMUPAGE_SIZE) #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) diff -prauN linux-2.5.70-bk4/include/asm-i386/pci.h pgcl-2.5.70-bk4-1/include/asm-i386/pci.h --- linux-2.5.70-bk4/include/asm-i386/pci.h 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/pci.h 2003-05-30 21:03:23.000000000 -0700 @@ -67,13 +67,13 @@ pci_dac_page_to_dma(struct pci_dev *pdev static __inline__ struct page * pci_dac_dma_to_page(struct pci_dev *pdev, dma64_addr_t dma_addr) { - return pfn_to_page(dma_addr >> PAGE_SHIFT); + return pfn_to_page(dma_addr >> MMUPAGE_SHIFT); } static __inline__ unsigned long pci_dac_dma_to_offset(struct pci_dev *pdev, dma64_addr_t dma_addr) { - return (dma_addr & ~PAGE_MASK); + return dma_addr & ~PAGE_MASK; } static __inline__ void diff -prauN linux-2.5.70-bk4/include/asm-i386/pgalloc.h pgcl-2.5.70-bk4-1/include/asm-i386/pgalloc.h --- linux-2.5.70-bk4/include/asm-i386/pgalloc.h 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/pgalloc.h 2003-05-30 21:03:23.000000000 -0700 @@ -2,19 +2,89 @@ #define _I386_PGALLOC_H #include +#include #include #include #include #include /* for struct page */ +#include /* to make asm-generic/rmap.h happy */ +#include /* for pgtable_remove_rmap() */ #define pmd_populate_kernel(mm, pmd, pte) \ set_pmd(pmd, __pmd(_PAGE_TABLE + __pa(pte))) -static inline void pmd_populate(struct mm_struct *mm, pmd_t *pmd, struct page *pte) +static inline void pgtable_remove_rmap(struct page *); +static inline void pmd_populate(struct mm_struct *mm, pmd_t *pmd, struct page *page) { - set_pmd(pmd, __pmd(_PAGE_TABLE + - ((unsigned long long)page_to_pfn(pte) << - (unsigned long long) PAGE_SHIFT))); + unsigned long pfn, pmd_off = (unsigned long)pmd; + int k; + + pmd_off = (pmd_off/sizeof(pmd_t)) % PAGE_MMUCOUNT; + pfn = page_to_pfn(page); + pmd -= pmd_off; + + if (PAGE_MMUCOUNT > 1) { + struct page *old_page = NULL; + + if (atomic_read(&page->count) != 1) { + WARN_ON(1); + printk(KERN_DEBUG "bad pte refcount = %d\n", + atomic_read(&page->count)); + } + + for (k = 0; k < PAGE_MMUCOUNT; ++k) { + if (pmd_present(pmd[k]) || !pmd_none(pmd[k])) { + if (old_page) + WARN_ON(old_page != pmd_page(pmd[k])); + else + old_page = pmd_page(pmd[k]); + } + } + + if (!old_page || old_page == page) + atomic_set(&page->count, PAGE_MMUCOUNT); + else { + /* + * old_page->index can legitimately be 0 + * but something's corrupt if it's mapping's wrong + */ + BUG_ON((struct mm_struct *)old_page->mapping != mm); + + /* + * errant callers can potentially do things + * out-of-order + */ + WARN_ON((struct mm_struct *)page->mapping != mm); + /* if (old_page->mapping != mm) + pgtable_add_rmap(page, mm, page->index); */ + pgtable_remove_rmap(page); + put_page(page); + atomic_set(&old_page->count, PAGE_MMUCOUNT); + for (k = 0; k < PAGE_MMUCOUNT; ++k) { + unsigned long long pmdval; + pmdval = page_to_pfn(old_page) + k; + pmdval <<= MMUPAGE_SHIFT; + if (pmd_present(pmd[k]) || !pmd_none(pmd[k])) { + WARN_ON(old_page != pmd_page(pmd[k])); + continue; + } else + set_pmd(&pmd[k], __pmd(_PAGE_TABLE + pmdval)); + } + return; + } + } + + for (k = 0; k < PAGE_MMUCOUNT; ++k) { + unsigned long long pmdval; + pmdval = (unsigned long long)(pfn + k) << MMUPAGE_SHIFT; + if (likely(pmd_none(pmd[k]) || !pmd_present(pmd[k]))) + set_pmd(&pmd[k], __pmd(_PAGE_TABLE + pmdval)); + else { + WARN_ON(1); + printk(KERN_DEBUG "pmdval=%Lx\n", (u64)pmd_val(pmd[k])); + put_page(page); /* a reference will be omitted */ + } + } } /* * Allocate and free page tables. @@ -31,13 +101,28 @@ static inline void pte_free_kernel(pte_t free_page((unsigned long)pte); } -static inline void pte_free(struct page *pte) +static inline void pte_free(struct page *page) { - __free_page(pte); + if (PAGE_MMUCOUNT == 1) + __free_page(page); + else + put_page(page); } -#define __pte_free_tlb(tlb,pte) tlb_remove_page((tlb),(pte)) +struct mmu_gather; +static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page); +static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *page) +{ + /* restore the reference count so later frees don't BUG() */ + if (PAGE_MMUCOUNT > 1) { + if (atomic_dec_and_test(&page->count)) + atomic_set(&page->count, 1); + else + return; + } + tlb_remove_page(tlb, page); +} /* * allocating and freeing a pmd is trivial: the 1-entry pmd is @@ -52,4 +137,6 @@ static inline void pte_free(struct page #define check_pgt_cache() do { } while (0) +#include + #endif /* _I386_PGALLOC_H */ diff -prauN linux-2.5.70-bk4/include/asm-i386/pgtable-2level.h pgcl-2.5.70-bk4-1/include/asm-i386/pgtable-2level.h --- linux-2.5.70-bk4/include/asm-i386/pgtable-2level.h 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/pgtable-2level.h 2003-05-30 21:03:23.000000000 -0700 @@ -17,6 +17,7 @@ #define PTRS_PER_PTE 1024 +#ifndef __ASSEMBLY__ #define pte_ERROR(e) \ printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, (e).pte_low) #define pmd_ERROR(e) \ @@ -49,7 +50,7 @@ static inline int pgd_present(pgd_t pgd) #define set_pgd(pgdptr, pgdval) (*(pgdptr) = pgdval) #define pgd_page(pgd) \ -((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +((unsigned long) __va(pgd_val(pgd) & MMUPAGE_MASK)) static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) { @@ -59,9 +60,11 @@ static inline pmd_t * pmd_offset(pgd_t * #define pte_same(a, b) ((a).pte_low == (b).pte_low) #define pte_page(x) pfn_to_page(pte_pfn(x)) #define pte_none(x) (!(x).pte_low) -#define pte_pfn(x) ((unsigned long)(((x).pte_low >> PAGE_SHIFT))) -#define pfn_pte(pfn, prot) __pte(((pfn) << PAGE_SHIFT) | pgprot_val(prot)) -#define pfn_pmd(pfn, prot) __pmd(((pfn) << PAGE_SHIFT) | pgprot_val(prot)) +#define pte_pfn(x) ((unsigned long)(((x).pte_low>>MMUPAGE_SHIFT))) +#define pfn_pte(pfn, prot) __pte(((pfn)<> PAGE_SHIFT) | - (pte.pte_high << (32 - PAGE_SHIFT)); + return (pte.pte_low >> MMUPAGE_SHIFT) | + (pte.pte_high << (32 - MMUPAGE_SHIFT)); } static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot) { pte_t pte; - pte.pte_high = page_nr >> (32 - PAGE_SHIFT); - pte.pte_low = (page_nr << PAGE_SHIFT) | pgprot_val(pgprot); + pte.pte_high = page_nr >> (32 - MMUPAGE_SHIFT); + pte.pte_low = (page_nr << MMUPAGE_SHIFT) | pgprot_val(pgprot); return pte; } static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot) { - return __pmd(((unsigned long long)page_nr << PAGE_SHIFT) | pgprot_val(pgprot)); + return __pmd(((unsigned long long)page_nr << MMUPAGE_SHIFT) | pgprot_val(pgprot)); } /* @@ -123,6 +124,6 @@ static inline pmd_t pfn_pmd(unsigned lon #define pgoff_to_pte(off) ((pte_t) { _PAGE_FILE, (off) }) #define PTE_FILE_MAX_BITS 32 -extern struct kmem_cache_s *pae_pgd_cachep; +#endif /* !__ASSEMBLY__ */ #endif /* _I386_PGTABLE_3LEVEL_H */ diff -prauN linux-2.5.70-bk4/include/asm-i386/pgtable.h pgcl-2.5.70-bk4-1/include/asm-i386/pgtable.h --- linux-2.5.70-bk4/include/asm-i386/pgtable.h 2003-05-26 18:00:45.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/pgtable.h 2003-05-30 21:03:23.000000000 -0700 @@ -21,15 +21,28 @@ #include #endif -extern pgd_t swapper_pg_dir[1024]; -extern void paging_init(void); +#include +#include +#include /* * ZERO_PAGE is a global shared page that is always zero: used * for zero-mapped memory areas etc.. */ extern unsigned long empty_zero_page[1024]; -#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page)) +extern struct page *zero_page; +#define ZERO_PAGE(vaddr) (zero_page) +extern pgd_t swapper_pg_dir[1024]; +extern kmem_cache_t *pgd_cache; +extern kmem_cache_t *pmd_cache; +extern spinlock_t pgd_lock; +extern struct list_head pgd_list; + +void pmd_ctor(void *, kmem_cache_t *, unsigned long); +void pgd_ctor(void *, kmem_cache_t *, unsigned long); +void pgd_dtor(void *, kmem_cache_t *, unsigned long); +void pgtable_cache_init(void); +void paging_init(void); #endif /* !__ASSEMBLY__ */ @@ -38,24 +51,10 @@ extern unsigned long empty_zero_page[102 * implements both the traditional 2-level x86 page tables and the * newer 3-level PAE-mode page tables. */ -#ifndef __ASSEMBLY__ #ifdef CONFIG_X86_PAE # include - -/* - * Need to initialise the X86 PAE caches - */ -extern void pgtable_cache_init(void); - #else # include - -/* - * No page table caches to initialise - */ -#define pgtable_cache_init() do { } while (0) - -#endif #endif #define PMD_SIZE (1UL << PMD_SHIFT) @@ -82,15 +81,15 @@ extern void pgtable_cache_init(void); * The vmalloc() routines leaves a hole of 4kB between each vmalloced * area for the same reason. ;) */ -#define VMALLOC_OFFSET (8*1024*1024) -#define VMALLOC_START (((unsigned long) high_memory + 2*VMALLOC_OFFSET-1) & \ - ~(VMALLOC_OFFSET-1)) #define VMALLOC_VMADDR(x) ((unsigned long)(x)) -#ifdef CONFIG_HIGHMEM -# define VMALLOC_END (PKMAP_BASE-2*PAGE_SIZE) -#else -# define VMALLOC_END (FIXADDR_START-2*PAGE_SIZE) -#endif +#define VMALLOC_END (FIXADDR_START-2*MMUPAGE_SIZE) + +#define __VMALLOC_START (VMALLOC_END - VMALLOC_RESERVE - 2*MMUPAGE_SIZE) +#define VMALLOC_START \ + (high_memory \ + ? max(__VMALLOC_START, (unsigned long)high_memory) \ + : __VMALLOC_START \ + ) /* * The 4MB page is guessing.. Detailed in the infamous "Chapter H" @@ -183,7 +182,7 @@ extern unsigned long pg0[1024]; #define pmd_none(x) (!pmd_val(x)) #define pmd_present(x) (pmd_val(x) & _PAGE_PRESENT) #define pmd_clear(xp) do { set_pmd(xp, __pmd(0)); } while (0) -#define pmd_bad(x) ((pmd_val(x) & (~PAGE_MASK & ~_PAGE_USER)) != _KERNPG_TABLE) +#define pmd_bad(x) ((pmd_val(x) & (~MMUPAGE_MASK & ~_PAGE_USER)) != _KERNPG_TABLE) #define pages_to_mb(x) ((x) >> (20-PAGE_SHIFT)) @@ -238,10 +237,10 @@ static inline pte_t pte_modify(pte_t pte #define page_pte(page) page_pte_prot(page, __pgprot(0)) #define pmd_page_kernel(pmd) \ -((unsigned long) __va(pmd_val(pmd) & PAGE_MASK)) +((unsigned long) __va(pmd_val(pmd) & MMUPAGE_MASK)) #ifndef CONFIG_DISCONTIGMEM -#define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)) +#define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> MMUPAGE_SHIFT)) #endif /* !CONFIG_DISCONTIGMEM */ #define pmd_large(pmd) \ @@ -283,20 +282,32 @@ static inline pte_t pte_modify(pte_t pte * control the given virtual address */ #define pte_index(address) \ - (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) + (((address) >> MMUPAGE_SHIFT) & (PTRS_PER_PTE - 1)) #define pte_offset_kernel(dir, address) \ ((pte_t *) pmd_page_kernel(*(dir)) + pte_index(address)) #if defined(CONFIG_HIGHPTE) #define pte_offset_map(dir, address) \ - ((pte_t *)kmap_atomic(pmd_page(*(dir)),KM_PTE0) + pte_index(address)) +( \ + (pte_t *)kmap_atomic(pmd_page(*(dir)),KM_PTE0) \ + + (PTRS_PER_PTE*((pmd_val(*(dir))/MMUPAGE_SIZE)%PAGE_MMUCOUNT)\ + + pte_index(address)) \ +) #define pte_offset_map_nested(dir, address) \ - ((pte_t *)kmap_atomic(pmd_page(*(dir)),KM_PTE1) + pte_index(address)) +( \ + (pte_t *)kmap_atomic(pmd_page(*(dir)),KM_PTE1) \ + + (PTRS_PER_PTE*((pmd_val(*(dir))/MMUPAGE_SIZE)%PAGE_MMUCOUNT)\ + + pte_index(address)) \ +) #define pte_unmap(pte) kunmap_atomic(pte, KM_PTE0) #define pte_unmap_nested(pte) kunmap_atomic(pte, KM_PTE1) #else #define pte_offset_map(dir, address) \ - ((pte_t *)page_address(pmd_page(*(dir))) + pte_index(address)) +( \ + (pte_t *)page_address(pmd_page(*(dir))) \ + + (PTRS_PER_PTE*((pmd_val(*(dir))/MMUPAGE_SIZE)%PAGE_MMUCOUNT)\ + + pte_index(address)) \ +) #define pte_offset_map_nested(dir, address) pte_offset_map(dir, address) #define pte_unmap(pte) do { } while (0) #define pte_unmap_nested(pte) do { } while (0) diff -prauN linux-2.5.70-bk4/include/asm-i386/rmap.h pgcl-2.5.70-bk4-1/include/asm-i386/rmap.h --- linux-2.5.70-bk4/include/asm-i386/rmap.h 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/rmap.h 2003-05-30 21:03:23.000000000 -0700 @@ -5,10 +5,17 @@ #include #ifdef CONFIG_HIGHPTE +/* + * The byte offset needs to be relative to PAGE_SIZE, the pfn will be + * implicitly truncated to a PAGE_SIZE boundary, the mapping will be + * returned rounded downward, and will need compensation by adding in + * the paddr's offset within the PAGE_SIZE-aligned region to the vaddr + * returned from kmap_atomic(). + */ static inline pte_t *rmap_ptep_map(pte_addr_t pte_paddr) { - unsigned long pfn = (unsigned long)(pte_paddr >> PAGE_SHIFT); - unsigned long off = ((unsigned long)pte_paddr) & ~PAGE_MASK; + unsigned long pfn = (unsigned long)(pte_paddr/MMUPAGE_SIZE); + unsigned long off = (unsigned long)pte_paddr & ~PAGE_MASK; return (pte_t *)((char *)kmap_atomic(pfn_to_page(pfn), KM_PTE2) + off); } diff -prauN linux-2.5.70-bk4/include/asm-i386/setup.h pgcl-2.5.70-bk4-1/include/asm-i386/setup.h --- linux-2.5.70-bk4/include/asm-i386/setup.h 2003-05-26 18:00:42.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/setup.h 2003-05-30 21:03:23.000000000 -0700 @@ -6,15 +6,15 @@ #ifndef _i386_SETUP_H #define _i386_SETUP_H -#define PFN_UP(x) (((x) + PAGE_SIZE-1) >> PAGE_SHIFT) -#define PFN_DOWN(x) ((x) >> PAGE_SHIFT) -#define PFN_PHYS(x) ((x) << PAGE_SHIFT) +#define PFN_UP(x) (((x) + MMUPAGE_SIZE-1) >> MMUPAGE_SHIFT) +#define PFN_DOWN(x) ((x) >> MMUPAGE_SHIFT) +#define PFN_PHYS(x) ((x) << MMUPAGE_SHIFT) /* * Reserved space for vmalloc and iomap - defined in asm/page.h */ #define MAXMEM_PFN PFN_DOWN(MAXMEM) -#define MAX_NONPAE_PFN (1 << 20) +#define MAX_NONPAE_PFN (1 << (32 - MMUPAGE_SHIFT)) /* * This is set up by the setup-routine at boot-time diff -prauN linux-2.5.70-bk4/include/asm-i386/shmparam.h pgcl-2.5.70-bk4-1/include/asm-i386/shmparam.h --- linux-2.5.70-bk4/include/asm-i386/shmparam.h 2003-05-26 18:00:37.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/shmparam.h 2003-05-30 21:03:23.000000000 -0700 @@ -1,6 +1,6 @@ #ifndef _ASMI386_SHMPARAM_H #define _ASMI386_SHMPARAM_H -#define SHMLBA PAGE_SIZE /* attach addr a multiple of this */ +#define SHMLBA MMUPAGE_SIZE /* attach addr a multiple of this */ #endif /* _ASMI386_SHMPARAM_H */ diff -prauN linux-2.5.70-bk4/include/asm-i386/thread_info.h pgcl-2.5.70-bk4-1/include/asm-i386/thread_info.h --- linux-2.5.70-bk4/include/asm-i386/thread_info.h 2003-05-26 18:00:24.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/thread_info.h 2003-05-30 21:03:23.000000000 -0700 @@ -53,6 +53,7 @@ struct thread_info { #endif #define PREEMPT_ACTIVE 0x4000000 +#define THREAD_SIZE (2*MMUPAGE_SIZE) /* * macros/functions for gaining access to the thread information structure @@ -81,14 +82,13 @@ struct thread_info { static inline struct thread_info *current_thread_info(void) { struct thread_info *ti; - __asm__("andl %%esp,%0; ":"=r" (ti) : "0" (~8191UL)); + __asm__("andl %%esp,%0; ":"=r" (ti) : "0" (~(THREAD_SIZE - 1))); return ti; } /* thread information allocation */ -#define THREAD_SIZE (2*PAGE_SIZE) -#define alloc_thread_info() ((struct thread_info *) __get_free_pages(GFP_KERNEL,1)) -#define free_thread_info(ti) free_pages((unsigned long) (ti), 1) +#define alloc_thread_info() ((struct thread_info *) kmalloc(THREAD_SIZE, SLAB_KERNEL)) +#define free_thread_info(ti) kfree(ti) #define get_thread_info(ti) get_task_struct((ti)->task) #define put_thread_info(ti) put_task_struct((ti)->task) @@ -96,7 +96,7 @@ static inline struct thread_info *curren /* how to get the thread information struct from ASM */ #define GET_THREAD_INFO(reg) \ - movl $-8192, reg; \ + movl $~(THREAD_SIZE-1), reg; \ andl %esp, reg #endif diff -prauN linux-2.5.70-bk4/include/asm-i386/tlbflush.h pgcl-2.5.70-bk4-1/include/asm-i386/tlbflush.h --- linux-2.5.70-bk4/include/asm-i386/tlbflush.h 2003-05-26 18:00:59.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-i386/tlbflush.h 2003-05-30 21:03:23.000000000 -0700 @@ -92,8 +92,17 @@ static inline void flush_tlb_mm(struct m static inline void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr) { + int k; if (vma->vm_mm == current->active_mm) - __flush_tlb_one(addr); + for (k = 0; k < PAGE_MMUCOUNT; ++k) { + unsigned long vaddr = addr + k*MMUPAGE_SIZE; + if (vaddr < vma->vm_start) + continue; + else if (vaddr >= vma->vm_end) + break; + else + __flush_tlb_one(vaddr); + } } static inline void flush_tlb_range(struct vm_area_struct *vma, diff -prauN linux-2.5.70-bk4/include/asm-ia64/page.h pgcl-2.5.70-bk4-1/include/asm-ia64/page.h --- linux-2.5.70-bk4/include/asm-ia64/page.h 2003-05-26 18:00:24.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-ia64/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -197,4 +197,6 @@ get_order (unsigned long size) (((current->thread.flags & IA64_THREAD_XSTACK) != 0) \ ? VM_EXEC : 0)) +#include + #endif /* _ASM_IA64_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-m68k/page.h pgcl-2.5.70-bk4-1/include/asm-m68k/page.h --- linux-2.5.70-bk4/include/asm-m68k/page.h 2003-05-26 18:01:03.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-m68k/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -192,6 +192,8 @@ static inline void *__va(unsigned long x #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif /* _M68K_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-m68knommu/page.h pgcl-2.5.70-bk4-1/include/asm-m68knommu/page.h --- linux-2.5.70-bk4/include/asm-m68knommu/page.h 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-m68knommu/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -94,6 +94,8 @@ extern unsigned long memory_end; #define virt_addr_valid(kaddr) (((void *)(kaddr) >= (void *)PAGE_OFFSET) && \ ((void *)(kaddr) < (void *)memory_end)) +#include + #endif /* __ASSEMBLY__ */ #endif /* __KERNEL__ */ diff -prauN linux-2.5.70-bk4/include/asm-mips/page.h pgcl-2.5.70-bk4-1/include/asm-mips/page.h --- linux-2.5.70-bk4/include/asm-mips/page.h 2003-05-26 18:00:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-mips/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -85,6 +85,8 @@ extern __inline__ int get_order(unsigned #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* defined (__KERNEL__) */ #endif /* __ASM_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-mips64/page.h pgcl-2.5.70-bk4-1/include/asm-mips64/page.h --- linux-2.5.70-bk4/include/asm-mips64/page.h 2003-05-26 18:00:20.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-mips64/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -80,6 +80,8 @@ typedef struct { unsigned long pgprot; } #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* defined (__KERNEL__) */ #endif /* _ASM_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-parisc/page.h pgcl-2.5.70-bk4-1/include/asm-parisc/page.h --- linux-2.5.70-bk4/include/asm-parisc/page.h 2003-05-26 18:00:25.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-parisc/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -112,6 +112,8 @@ extern int npmem_ranges; #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif /* _PARISC_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-ppc/page.h pgcl-2.5.70-bk4-1/include/asm-ppc/page.h --- linux-2.5.70-bk4/include/asm-ppc/page.h 2003-05-26 18:00:27.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-ppc/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -142,5 +142,7 @@ extern __inline__ int get_order(unsigned #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif /* _PPC_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-ppc64/page.h pgcl-2.5.70-bk4-1/include/asm-ppc64/page.h --- linux-2.5.70-bk4/include/asm-ppc64/page.h 2003-05-26 18:00:20.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-ppc64/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -204,5 +204,7 @@ static inline int get_order(unsigned lon #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif /* _PPC64_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-s390/page.h pgcl-2.5.70-bk4-1/include/asm-s390/page.h --- linux-2.5.70-bk4/include/asm-s390/page.h 2003-05-26 18:00:24.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-s390/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -181,6 +181,8 @@ typedef struct { unsigned long pgd; } pg #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif /* _S390_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-sh/page.h pgcl-2.5.70-bk4-1/include/asm-sh/page.h --- linux-2.5.70-bk4/include/asm-sh/page.h 2003-05-26 18:00:57.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-sh/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -106,6 +106,8 @@ static __inline__ int get_order(unsigned #endif +#include + #endif /* __KERNEL__ */ #endif /* __ASM_SH_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-sparc/page.h pgcl-2.5.70-bk4-1/include/asm-sparc/page.h --- linux-2.5.70-bk4/include/asm-sparc/page.h 2003-05-26 18:00:27.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-sparc/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -171,6 +171,8 @@ extern __inline__ int get_order(unsigned #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* __KERNEL__ */ #endif /* _SPARC_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/asm-sparc64/page.h pgcl-2.5.70-bk4-1/include/asm-sparc64/page.h --- linux-2.5.70-bk4/include/asm-sparc64/page.h 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-sparc64/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -168,6 +168,8 @@ static __inline__ int get_order(unsigned #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) +#include + #endif /* !(__KERNEL__) */ #endif /* !(_SPARC64_PAGE_H) */ diff -prauN linux-2.5.70-bk4/include/asm-v850/page.h pgcl-2.5.70-bk4-1/include/asm-v850/page.h --- linux-2.5.70-bk4/include/asm-v850/page.h 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-v850/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -140,6 +140,7 @@ extern __inline__ int get_order (unsigne #define __pa(x) __virt_to_phys ((unsigned long)(x)) #define __va(x) ((void *)__phys_to_virt ((unsigned long)(x))) +#include #endif /* KERNEL */ diff -prauN linux-2.5.70-bk4/include/asm-x86_64/page.h pgcl-2.5.70-bk4-1/include/asm-x86_64/page.h --- linux-2.5.70-bk4/include/asm-x86_64/page.h 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/asm-x86_64/page.h 2003-05-30 21:03:23.000000000 -0700 @@ -136,6 +136,8 @@ extern __inline__ int get_order(unsigned (test_thread_flag(TIF_IA32) ? vm_stack_flags32 : vm_stack_flags) +#include + #endif /* __KERNEL__ */ #endif /* _X86_64_PAGE_H */ diff -prauN linux-2.5.70-bk4/include/linux/aio.h pgcl-2.5.70-bk4-1/include/linux/aio.h --- linux-2.5.70-bk4/include/linux/aio.h 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/aio.h 2003-05-30 21:03:23.000000000 -0700 @@ -104,13 +104,13 @@ struct aio_ring_info { unsigned long mmap_base; unsigned long mmap_size; - struct page **ring_pages; + unsigned long *ring_pages; spinlock_t ring_lock; long nr_pages; unsigned nr, tail; - struct page *internal_pages[AIO_RING_PAGES]; + unsigned long internal_pages[AIO_RING_PAGES]; /* pfn's */ }; struct kioctx { diff -prauN linux-2.5.70-bk4/include/linux/binfmts.h pgcl-2.5.70-bk4-1/include/linux/binfmts.h --- linux-2.5.70-bk4/include/linux/binfmts.h 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/binfmts.h 2003-05-30 21:03:23.000000000 -0700 @@ -2,6 +2,7 @@ #define _LINUX_BINFMTS_H #include +#include /* for PAGE_MMUCOUNT */ struct pt_regs; @@ -9,8 +10,15 @@ struct pt_regs; * MAX_ARG_PAGES defines the number of pages allocated for arguments * and envelope for the new program. 32 should suffice, this gives * a maximum env+arg of 128kB w/4KB pages! + * Now that PAGE_SIZE is a software construct and varies wildly, + * MAX_ARG_PAGES should represent a constant size of 128KB. When + * PAGE_SIZE exceeds that, we're in trouble. */ -#define MAX_ARG_PAGES 32 +#if PAGE_MMUCOUNT <= 32 +#define MAX_ARG_PAGES (32/PAGE_MMUCOUNT) +#else +#error PAGE_SIZE too large to enforce MAX_ARG_PAGES! +#endif /* sizeof(linux_binprm->buf) */ #define BINPRM_BUF_SIZE 128 diff -prauN linux-2.5.70-bk4/include/linux/bio.h pgcl-2.5.70-bk4-1/include/linux/bio.h --- linux-2.5.70-bk4/include/linux/bio.h 2003-05-30 21:02:37.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/bio.h 2003-05-30 21:03:23.000000000 -0700 @@ -257,26 +257,20 @@ extern void bio_check_pages_dirty(struct */ extern inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags) { - unsigned long addr; + char *addr; /* * might not be a highmem page, but the preempt/irq count * balancing is a lot nicer this way */ local_irq_save(*flags); - addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ); - - if (addr & ~PAGE_MASK) - BUG(); - - return (char *) addr + bvec->bv_offset; + addr = (char *)kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ); + return addr + bvec->bv_offset; } extern inline void bvec_kunmap_irq(char *buffer, unsigned long *flags) { - unsigned long ptr = (unsigned long) buffer & PAGE_MASK; - - kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ); + kunmap_atomic(buffer, KM_BIO_SRC_IRQ); local_irq_restore(*flags); } diff -prauN linux-2.5.70-bk4/include/linux/highmem.h pgcl-2.5.70-bk4-1/include/linux/highmem.h --- linux-2.5.70-bk4/include/linux/highmem.h 2003-05-26 18:00:45.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/highmem.h 2003-05-30 21:03:23.000000000 -0700 @@ -81,6 +81,17 @@ static inline void copy_user_highpage(st kunmap_atomic(vto, KM_USER1); } +static inline void copy_user_mmupages(struct page *dst, struct page *src, int offset, int size) +{ + char *vfrom, *vto; + + vfrom = kmap_atomic(src, KM_USER0); + vto = kmap_atomic(dst, KM_USER1); + memcpy(&vto[offset], &vfrom[offset], size); + kunmap_atomic(src, KM_USER0); + kunmap_atomic(dst, KM_USER1); +} + static inline void copy_highpage(struct page *to, struct page *from) { char *vfrom, *vto; diff -prauN linux-2.5.70-bk4/include/linux/ide.h pgcl-2.5.70-bk4-1/include/linux/ide.h --- linux-2.5.70-bk4/include/linux/ide.h 2003-05-26 18:00:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/ide.h 2003-05-30 21:03:23.000000000 -0700 @@ -225,7 +225,7 @@ typedef unsigned char byte; /* used ever * allowing each to have about 256 entries (8 bytes each) from this. */ #define PRD_BYTES 8 -#define PRD_ENTRIES (PAGE_SIZE / (2 * PRD_BYTES)) +#define PRD_ENTRIES (MMUPAGE_SIZE / (2 * PRD_BYTES)) /* * Some more useful definitions diff -prauN linux-2.5.70-bk4/include/linux/mm.h pgcl-2.5.70-bk4-1/include/linux/mm.h --- linux-2.5.70-bk4/include/linux/mm.h 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/mm.h 2003-05-30 21:03:23.000000000 -0700 @@ -73,7 +73,7 @@ struct vm_area_struct { struct vm_operations_struct * vm_ops; /* Information about our backing store: */ - unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE + unsigned long vm_pgoff; /* Offset (within vm_file) in MMUPAGE_SIZE units, *not* PAGE_CACHE_SIZE */ struct file * vm_file; /* File we map to (can be NULL). */ void * vm_private_data; /* was vm_pte (shared mem) */ @@ -172,7 +172,8 @@ struct page { atomic_t count; /* Usage count, see below. */ struct list_head list; /* ->mapping has some page lists. */ struct address_space *mapping; /* The inode (or ...) we belong to. */ - unsigned long index; /* Our offset within mapping. */ + unsigned long index; /* Our offset within mapping. + * in PAGE_CACHE_SIZE units. */ struct list_head lru; /* Pageout list, eg. active_list; protected by zone->lru_lock !! */ union { @@ -339,10 +340,19 @@ static inline void set_page_zone(struct page->flags |= zone_num << ZONE_SHIFT; } -static inline void * lowmem_page_address(struct page *page) -{ - return __va( ( (page - page_zone(page)->zone_mem_map) + page_zone(page)->zone_start_pfn) << PAGE_SHIFT); -} + +#if 1 +#define lowmem_page_address(page) __va(page_to_pfn(page)*MMUPAGE_SIZE) +#else + #define lowmem_page_address(page) \ +({ \ + extern unsigned long max_low_pfn; \ + const unsigned long __lpa_pfn = page_to_pfn(page); \ + BUG_ON(max_low_pfn && __lpa_pfn > max_low_pfn); \ + BUG_ON(__lpa_pfn >= (~PAGE_OFFSET+1)/MMUPAGE_SIZE); \ + __va(__lpa_pfn*MMUPAGE_SIZE); \ +}) +#endif #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) #define HASHED_PAGE_VIRTUAL @@ -425,16 +435,16 @@ extern int vmtruncate(struct inode * ino extern pmd_t *FASTCALL(__pmd_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)); extern pte_t *FASTCALL(pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); extern pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); -extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot); +extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot, int subpfn); extern int handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, unsigned long address, int write_access); extern int make_pages_present(unsigned long addr, unsigned long end); extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); extern long sys_remap_file_pages(unsigned long start, unsigned long size, unsigned long prot, unsigned long pgoff, unsigned long nonblock); -void put_dirty_page(struct task_struct *tsk, struct page *page, - unsigned long address, pgprot_t prot); +void put_dirty_page(struct task_struct *task, struct page *page, + int min_subpfn, unsigned long address, pgprot_t prot); int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start, - int len, int write, int force, struct page **pages, struct vm_area_struct **vmas); + int len, int write, int force, unsigned long *pages, struct vm_area_struct **vmas); int __set_page_dirty_buffers(struct page *page); int __set_page_dirty_nobuffers(struct page *page); @@ -519,10 +529,10 @@ static inline unsigned long do_mmap(stru unsigned long flag, unsigned long offset) { unsigned long ret = -EINVAL; - if ((offset + PAGE_ALIGN(len)) < offset) + if ((offset + MMUPAGE_ALIGN(len)) < offset) goto out; - if (!(offset & ~PAGE_MASK)) - ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); + if (!(offset & ~MMUPAGE_MASK)) + ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> MMUPAGE_SHIFT); out: return ret; } @@ -604,8 +614,18 @@ extern struct vm_area_struct *find_exten extern unsigned int nr_used_zone_pages(void); +/* + * Return byte offset from start of page containing virtual address in + * vma, to start of mmupage containing it: 0 if PAGE_MMUSHIFT 0. + */ +static inline unsigned long vma_suboffset(struct vm_area_struct *vma, unsigned long address) +{ + return (address - vma->vm_start + MMUPAGE_SIZE * vma->vm_pgoff) + & (MMUPAGE_MASK - PAGE_MASK); +} + extern struct page * vmalloc_to_page(void *addr); -extern struct page * follow_page(struct mm_struct *mm, unsigned long address, +unsigned long follow_page(struct mm_struct *mm, unsigned long address, int write); extern int remap_page_range(struct vm_area_struct *vma, unsigned long from, unsigned long to, unsigned long size, pgprot_t prot); diff -prauN linux-2.5.70-bk4/include/linux/mmzone.h pgcl-2.5.70-bk4-1/include/linux/mmzone.h --- linux-2.5.70-bk4/include/linux/mmzone.h 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/mmzone.h 2003-05-30 21:03:23.000000000 -0700 @@ -20,7 +20,7 @@ /* Free memory management - zoned buddy allocator. */ #ifndef CONFIG_FORCE_MAX_ZONEORDER -#define MAX_ORDER 11 +#define MAX_ORDER (11 - PAGE_MMUSHIFT) #else #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER #endif diff -prauN linux-2.5.70-bk4/include/linux/pagemap.h pgcl-2.5.70-bk4-1/include/linux/pagemap.h --- linux-2.5.70-bk4/include/linux/pagemap.h 2003-05-26 18:00:20.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/pagemap.h 2003-05-30 21:17:20.000000000 -0700 @@ -23,6 +23,9 @@ #define PAGE_CACHE_MASK PAGE_MASK #define PAGE_CACHE_ALIGN(addr) (((addr)+PAGE_CACHE_SIZE-1)&PAGE_CACHE_MASK) +#define PAGE_CACHE_MMUSHIFT (PAGE_CACHE_SHIFT - MMUPAGE_SHIFT) +#define PAGE_CACHE_MMUCOUNT (PAGE_CACHE_SIZE/MMUPAGE_SIZE) + #define page_cache_get(page) get_page(page) #define page_cache_release(page) put_page(page) void release_pages(struct page **pages, int nr, int cold); @@ -174,40 +177,27 @@ extern void end_page_writeback(struct pa */ static inline int fault_in_pages_writeable(char __user *uaddr, int size) { - int ret; + int ret = 0; + unsigned long addr, end = (unsigned long)uaddr + size - 1; /* * Writing zeroes into userspace here is OK, because we know that if * the zero gets there, we'll be overwriting it. */ - ret = __put_user(0, uaddr); - if (ret == 0) { - char __user *end = uaddr + size - 1; - - /* - * If the page was already mapped, this will get a cache miss - * for sure, so try to avoid doing it. - */ - if (((unsigned long)uaddr & PAGE_MASK) != - ((unsigned long)end & PAGE_MASK)) - ret = __put_user(0, end); - } + for (addr = (unsigned long)uaddr; addr <= MMUPAGE_ALIGN(end); addr += MMUPAGE_SIZE) + if ((ret = __put_user(0, (char *)min(addr, end))) != 0) + break; + return ret; } static inline void fault_in_pages_readable(const char __user *uaddr, int size) { volatile char c; - int ret; + unsigned long addr, end = (unsigned long)uaddr + size - 1; - ret = __get_user(c, (char *)uaddr); - if (ret == 0) { - const char __user *end = uaddr + size - 1; - - if (((unsigned long)uaddr & PAGE_MASK) != - ((unsigned long)end & PAGE_MASK)) - __get_user(c, (char *)end); - } + for (addr = (unsigned long)uaddr; addr <= MMUPAGE_ALIGN(end); addr += MMUPAGE_SIZE) + __get_user(c, (char *)min(addr, end)); } #endif /* _LINUX_PAGEMAP_H */ diff -prauN linux-2.5.70-bk4/include/linux/sched.h pgcl-2.5.70-bk4-1/include/linux/sched.h --- linux-2.5.70-bk4/include/linux/sched.h 2003-05-26 18:00:23.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/sched.h 2003-05-30 21:03:24.000000000 -0700 @@ -195,7 +195,7 @@ struct mm_struct { unsigned long start_code, end_code, start_data, end_data; unsigned long start_brk, brk, start_stack; unsigned long arg_start, arg_end, env_start, env_end; - unsigned long rss, total_vm, locked_vm; + unsigned long rss, total_vm, locked_vm; /* in MMUPAGE_SIZE units */ unsigned long def_flags; unsigned long cpu_vm_mask; unsigned long swap_address; @@ -613,12 +613,7 @@ static inline int capable(int cap) extern struct mm_struct * mm_alloc(void); /* mmdrop drops the mm and the page tables */ -extern inline void FASTCALL(__mmdrop(struct mm_struct *)); -static inline void mmdrop(struct mm_struct * mm) -{ - if (atomic_dec_and_test(&mm->mm_count)) - __mmdrop(mm); -} +void mmdrop(struct mm_struct * mm); /* mmput gets rid of the mappings and all user-space */ extern void mmput(struct mm_struct *); diff -prauN linux-2.5.70-bk4/include/linux/shm.h pgcl-2.5.70-bk4-1/include/linux/shm.h --- linux-2.5.70-bk4/include/linux/shm.h 2003-05-26 18:00:45.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/shm.h 2003-05-30 21:03:24.000000000 -0700 @@ -12,7 +12,7 @@ #define SHMMAX 0x2000000 /* max shared seg size (bytes) */ #define SHMMIN 1 /* min shared seg size (bytes) */ #define SHMMNI 4096 /* max num of segs system wide */ -#define SHMALL (SHMMAX/PAGE_SIZE*(SHMMNI/16)) /* max shm system wide (pages) */ +#define SHMALL (SHMMAX/MMUPAGE_SIZE*(SHMMNI/16)) /* max shm system wide (mmupages) */ #define SHMSEG SHMMNI /* max shared segs per process */ #include diff -prauN linux-2.5.70-bk4/include/linux/sunrpc/svc.h pgcl-2.5.70-bk4-1/include/linux/sunrpc/svc.h --- linux-2.5.70-bk4/include/linux/sunrpc/svc.h 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/sunrpc/svc.h 2003-05-30 21:03:24.000000000 -0700 @@ -73,7 +73,8 @@ struct svc_serv { * This assumes that the non-page part of an rpc reply will fit * in a page - NFSd ensures this. lockd also has no trouble. */ -#define RPCSVC_MAXPAGES ((RPCSVC_MAXPAYLOAD+PAGE_SIZE-1)/PAGE_SIZE + 1) + +#define RPCSVC_MAXPAGES (2+((RPCSVC_MAXPAYLOAD+PAGE_SIZE-1)/PAGE_SIZE+1)) static inline u32 svc_getu32(struct iovec *iov) { diff -prauN linux-2.5.70-bk4/include/linux/swap.h pgcl-2.5.70-bk4-1/include/linux/swap.h --- linux-2.5.70-bk4/include/linux/swap.h 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/include/linux/swap.h 2003-05-30 21:03:24.000000000 -0700 @@ -45,7 +45,7 @@ static inline int current_is_kswapd(void */ union swap_header { struct { - char reserved[PAGE_SIZE - 10]; + char reserved[MMUPAGE_SIZE - 10]; char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */ } magic; struct { @@ -112,8 +112,8 @@ enum { #define SWAP_CLUSTER_MAX 32 -#define SWAP_MAP_MAX 0x7fff -#define SWAP_MAP_BAD 0x8000 +#define SWAP_MAP_MAX 0xfffe +#define SWAP_MAP_BAD 0xffff /* * The in-memory structure used to track swap areas. diff -prauN linux-2.5.70-bk4/init/main.c pgcl-2.5.70-bk4-1/init/main.c --- linux-2.5.70-bk4/init/main.c 2003-05-26 18:00:25.000000000 -0700 +++ pgcl-2.5.70-bk4-1/init/main.c 2003-05-30 21:03:24.000000000 -0700 @@ -372,6 +372,7 @@ static void rest_init(void) cpu_idle(); } + /* * Activate the first processor. */ @@ -421,9 +422,9 @@ asmlinkage void __init start_kernel(void calibrate_delay(); #ifdef CONFIG_BLK_DEV_INITRD if (initrd_start && !initrd_below_start_ok && - initrd_start < min_low_pfn << PAGE_SHIFT) { + initrd_start < min_low_pfn << MMUPAGE_SHIFT) { printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " - "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT); + "disabling it.\n",initrd_start,min_low_pfn << MMUPAGE_SHIFT); initrd_start = 0; } #endif diff -prauN linux-2.5.70-bk4/ipc/shm.c pgcl-2.5.70-bk4-1/ipc/shm.c --- linux-2.5.70-bk4/ipc/shm.c 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/ipc/shm.c 2003-05-30 21:03:24.000000000 -0700 @@ -110,7 +110,7 @@ static void shm_open (struct vm_area_str */ static void shm_destroy (struct shmid_kernel *shp) { - shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; + shm_tot -= (shp->shm_segsz + MMUPAGE_SIZE - 1) >> MMUPAGE_SHIFT; shm_rmid (shp->id); shm_unlock(shp); if (!is_file_hugepages(shp->shm_file)) @@ -169,7 +169,7 @@ static int newseg (key_t key, int shmflg { int error; struct shmid_kernel *shp; - int numpages = (size + PAGE_SIZE -1) >> PAGE_SHIFT; + int numpages = (size + MMUPAGE_SIZE -1) >> MMUPAGE_SHIFT; struct file * file; char name[13]; int id; @@ -717,7 +717,7 @@ long sys_shmat(int shmid, char __user *s * space left for the stack to grow (at least 4 pages). */ if (addr < current->mm->start_stack && - addr > current->mm->start_stack - size - PAGE_SIZE * 5) + addr > current->mm->start_stack - size - MMUPAGE_SIZE * 5) goto invalid; } @@ -775,7 +775,7 @@ asmlinkage long sys_shmdt(char __user *s * otherwise it starts at this address with no hassles. */ if ((vma->vm_ops == &shm_vm_ops || is_vm_hugetlb_page(vma)) && - (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { + (vma->vm_start - addr)/MMUPAGE_SIZE == vma->vm_pgoff) { size = vma->vm_file->f_dentry->d_inode->i_size; @@ -803,7 +803,7 @@ asmlinkage long sys_shmdt(char __user *s /* finding a matching vma now does not alter retval */ if ((vma->vm_ops == &shm_vm_ops || is_vm_hugetlb_page(vma)) && - (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) + (vma->vm_start - addr)/MMUPAGE_SIZE == vma->vm_pgoff) do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start); vma = next; diff -prauN linux-2.5.70-bk4/kernel/fork.c pgcl-2.5.70-bk4-1/kernel/fork.c --- linux-2.5.70-bk4/kernel/fork.c 2003-05-26 18:00:23.000000000 -0700 +++ pgcl-2.5.70-bk4-1/kernel/fork.c 2003-05-30 21:03:24.000000000 -0700 @@ -199,7 +199,7 @@ void __init fork_init(unsigned long memp * value: the thread structures can take up at most half * of memory. */ - max_threads = mempages / (THREAD_SIZE/PAGE_SIZE) / 8; + max_threads = mempages / 8; /* * we need to allow at least 20 threads to boot a system */ @@ -279,7 +279,7 @@ static inline int dup_mmap(struct mm_str if(mpnt->vm_flags & VM_DONTCOPY) continue; if (mpnt->vm_flags & VM_ACCOUNT) { - unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; + unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> MMUPAGE_SHIFT; if (!vm_enough_memory(len)) goto fail_nomem; charge += len; @@ -400,8 +400,11 @@ struct mm_struct * mm_alloc(void) * is dropped: either by a lazy thread or by * mmput. Free the page directory and the mm. */ -inline void __mmdrop(struct mm_struct *mm) +void mmdrop(struct mm_struct *mm) { + if (!atomic_dec_and_test(&mm->mm_count)) + return; + BUG_ON(mm == &init_mm); mm_free_pgd(mm); destroy_context(mm); diff -prauN linux-2.5.70-bk4/kernel/futex.c pgcl-2.5.70-bk4-1/kernel/futex.c --- linux-2.5.70-bk4/kernel/futex.c 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/kernel/futex.c 2003-05-30 21:03:24.000000000 -0700 @@ -108,15 +108,15 @@ static inline struct page *__pin_page_at static struct page *__pin_page(unsigned long addr) { struct mm_struct *mm = current->mm; - struct page *page, *tmp; + unsigned long pfn, tmp; int err; /* * Do a quick atomic lookup first - this is the fastpath. */ - page = follow_page(mm, addr, 0); - if (likely(page != NULL)) - return __pin_page_atomic(page); + pfn = follow_page(mm, addr, 0); + if (likely(pfn_valid(pfn))) + return __pin_page_atomic(pfn_to_page(pfn)); /* * No luck - need to fault in the page: @@ -126,7 +126,7 @@ repeat_lookup: unlock_futex_mm(); down_read(&mm->mmap_sem); - err = get_user_pages(current, mm, addr, 1, 0, 0, &page, NULL); + err = get_user_pages(current, mm, addr, 1, 0, 0, &pfn, NULL); up_read(&mm->mmap_sem); lock_futex_mm(); @@ -138,12 +138,18 @@ repeat_lookup: * check for races: */ tmp = follow_page(mm, addr, 0); - if (tmp != page) { + + /* + * XXX: this is weird, it can refer to a different pfn in the + * same page. Counts as a race in my book. + */ + if (tmp != pfn) { + struct page *page = pfn_to_page(pfn); put_page(page); goto repeat_lookup; } - return page; + return pfn ? pfn_to_page(pfn) : NULL; } /* @@ -164,6 +170,11 @@ static inline int futex_wake(unsigned lo return -EFAULT; } + /* + * XXX: I broke this. + * This needs to include a suboffset w/in the struct page's area. + * -- wli + */ head = hash_futex(page, offset); list_for_each_safe(i, next, head) { @@ -485,7 +496,7 @@ long do_futex(unsigned long uaddr, int o unsigned long pos_in_page; int ret; - pos_in_page = uaddr % PAGE_SIZE; + pos_in_page = uaddr % MMUPAGE_SIZE; /* Must be "naturally" aligned */ if (pos_in_page % sizeof(u32)) diff -prauN linux-2.5.70-bk4/kernel/ksyms.c pgcl-2.5.70-bk4-1/kernel/ksyms.c --- linux-2.5.70-bk4/kernel/ksyms.c 2003-05-26 18:00:20.000000000 -0700 +++ pgcl-2.5.70-bk4-1/kernel/ksyms.c 2003-05-30 21:03:24.000000000 -0700 @@ -127,7 +127,6 @@ EXPORT_SYMBOL(kmap_high); EXPORT_SYMBOL(kunmap_high); EXPORT_SYMBOL(highmem_start_page); EXPORT_SYMBOL(kmap_prot); -EXPORT_SYMBOL(kmap_pte); #endif #ifdef HASHED_PAGE_VIRTUAL EXPORT_SYMBOL(page_address); diff -prauN linux-2.5.70-bk4/kernel/ptrace.c pgcl-2.5.70-bk4-1/kernel/ptrace.c --- linux-2.5.70-bk4/kernel/ptrace.c 2003-05-26 18:01:03.000000000 -0700 +++ pgcl-2.5.70-bk4-1/kernel/ptrace.c 2003-05-30 21:03:24.000000000 -0700 @@ -155,27 +155,32 @@ int access_process_vm(struct task_struct struct mm_struct *mm; struct vm_area_struct *vma; struct page *page; + unsigned long pfn = 0; void *old_buf = buf; mm = get_task_mm(tsk); - if (!mm) + if (!mm) { + printk("get_task_mm() failed in access_process_vm()\n"); return 0; + } down_read(&mm->mmap_sem); /* ignore errors, just check how much was sucessfully transfered */ while (len) { int bytes, ret, offset; + unsigned long dst_off; void *maddr; - ret = get_user_pages(current, mm, addr, 1, - write, 1, &page, &vma); - if (ret <= 0) + ret = get_user_pages(current, mm, addr, 1, write, 1, &pfn, &vma); + if (ret <= 0) { + printk("get_uesr_pages() failed in access_process_vm()\n"); break; + } bytes = len; - offset = addr & (PAGE_SIZE-1); - if (bytes > PAGE_SIZE-offset) - bytes = PAGE_SIZE-offset; + offset = addr & ~MMUPAGE_MASK; + if (bytes > MMUPAGE_SIZE-offset) + bytes = MMUPAGE_SIZE-offset; flush_cache_page(vma, addr); @@ -185,12 +190,14 @@ int access_process_vm(struct task_struct * to handle this correctly. */ + page = pfn_to_page(pfn); maddr = kmap(page); + dst_off = (pfn % PAGE_MMUCOUNT)*MMUPAGE_SIZE; if (write) { - memcpy(maddr + offset, buf, bytes); + memcpy(maddr + offset + dst_off, buf, bytes); flush_icache_user_range(vma, page, addr, bytes); } else { - memcpy(buf, maddr + offset, bytes); + memcpy(buf, maddr + offset + dst_off, bytes); } kunmap(page); page_cache_release(page); diff -prauN linux-2.5.70-bk4/mm/bootmem.c pgcl-2.5.70-bk4-1/mm/bootmem.c --- linux-2.5.70-bk4/mm/bootmem.c 2003-05-26 18:00:27.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/bootmem.c 2003-05-30 21:03:24.000000000 -0700 @@ -33,10 +33,7 @@ unsigned long __init bootmem_bootmap_pag unsigned long mapsize; mapsize = (pages+7)/8; - mapsize = (mapsize + ~PAGE_MASK) & PAGE_MASK; - mapsize >>= PAGE_SHIFT; - - return mapsize; + return (mapsize + MMUPAGE_SIZE - 1) >> MMUPAGE_SHIFT; } /* @@ -46,14 +43,17 @@ static unsigned long __init init_bootmem unsigned long mapstart, unsigned long start, unsigned long end) { bootmem_data_t *bdata = pgdat->bdata; - unsigned long mapsize = ((end - start)+7)/8; + unsigned long mapsize; pgdat->pgdat_next = pgdat_list; pgdat_list = pgdat; + /* round start down to simplify free_all_bootmem_core() */ + start &= ~(PAGE_MMUCOUNT - 1); + mapsize = ((end - start)+7)/8; mapsize = (mapsize + (sizeof(long) - 1UL)) & ~(sizeof(long) - 1UL); - bdata->node_bootmem_map = phys_to_virt(mapstart << PAGE_SHIFT); - bdata->node_boot_start = (start << PAGE_SHIFT); + bdata->node_bootmem_map = phys_to_virt(mapstart << MMUPAGE_SHIFT); + bdata->node_boot_start = (start << MMUPAGE_SHIFT); bdata->node_low_pfn = end; /* @@ -77,10 +77,10 @@ static void __init reserve_bootmem_core( * round up, partially reserved pages are considered * fully reserved. */ - unsigned long sidx = (addr - bdata->node_boot_start)/PAGE_SIZE; + unsigned long sidx = (addr - bdata->node_boot_start)/MMUPAGE_SIZE; unsigned long eidx = (addr + size - bdata->node_boot_start + - PAGE_SIZE-1)/PAGE_SIZE; - unsigned long end = (addr + size + PAGE_SIZE-1)/PAGE_SIZE; + MMUPAGE_SIZE-1)/MMUPAGE_SIZE; + unsigned long end_pfn = (addr + size + MMUPAGE_SIZE-1)/MMUPAGE_SIZE; if (!size) BUG(); @@ -90,13 +90,11 @@ static void __init reserve_bootmem_core( BUG(); if (sidx >= eidx) BUG(); - if ((addr >> PAGE_SHIFT) >= bdata->node_low_pfn) - BUG(); - if (end > bdata->node_low_pfn) + if (end_pfn > bdata->node_low_pfn) BUG(); for (i = sidx; i < eidx; i++) if (test_and_set_bit(i, bdata->node_bootmem_map)) - printk("hm, page %08lx reserved twice.\n", i*PAGE_SIZE); + printk("hm, page %08lx reserved twice.\n", i*MMUPAGE_SIZE); } static void __init free_bootmem_core(bootmem_data_t *bdata, unsigned long addr, unsigned long size) @@ -108,11 +106,11 @@ static void __init free_bootmem_core(boo * considered reserved. */ unsigned long sidx; - unsigned long eidx = (addr + size - bdata->node_boot_start)/PAGE_SIZE; - unsigned long end = (addr + size)/PAGE_SIZE; + unsigned long eidx = (addr + size - bdata->node_boot_start)/MMUPAGE_SIZE; + unsigned long end_pfn = (addr + size)/MMUPAGE_SIZE; if (!size) BUG(); - if (end > bdata->node_low_pfn) + if (end_pfn > bdata->node_low_pfn) BUG(); if (addr < bdata->last_success) @@ -121,8 +119,8 @@ static void __init free_bootmem_core(boo /* * Round up the beginning of the address. */ - start = (addr + PAGE_SIZE-1) / PAGE_SIZE; - sidx = start - (bdata->node_boot_start/PAGE_SIZE); + start = (addr + MMUPAGE_SIZE-1) / MMUPAGE_SIZE; + sidx = start - (bdata->node_boot_start/MMUPAGE_SIZE); for (i = sidx; i < eidx; i++) { if (!test_and_clear_bit(i, bdata->node_bootmem_map)) @@ -154,19 +152,19 @@ __alloc_bootmem_core(struct bootmem_data BUG_ON(!size); BUG_ON(align & (align-1)); - eidx = bdata->node_low_pfn - (bdata->node_boot_start >> PAGE_SHIFT); + eidx = bdata->node_low_pfn - (bdata->node_boot_start >> MMUPAGE_SHIFT); offset = 0; if (align && (bdata->node_boot_start & (align - 1UL)) != 0) offset = (align - (bdata->node_boot_start & (align - 1UL))); - offset >>= PAGE_SHIFT; + offset >>= MMUPAGE_SHIFT; /* * We try to allocate bootmem pages above 'goal' * first, then we try to allocate lower pages. */ if (goal && (goal >= bdata->node_boot_start) && - ((goal >> PAGE_SHIFT) < bdata->node_low_pfn)) { + ((goal >> MMUPAGE_SHIFT) < bdata->node_low_pfn)) { preferred = goal - bdata->node_boot_start; if (bdata->last_success >= preferred) @@ -174,10 +172,10 @@ __alloc_bootmem_core(struct bootmem_data } else preferred = 0; - preferred = ((preferred + align - 1) & ~(align - 1)) >> PAGE_SHIFT; + preferred = ((preferred + align - 1) & ~(align - 1)) >> MMUPAGE_SHIFT; preferred += offset; - areasize = (size+PAGE_SIZE-1)/PAGE_SIZE; - incr = align >> PAGE_SHIFT ? : 1; + areasize = (size+MMUPAGE_SIZE-1)/MMUPAGE_SIZE; + incr = align >> MMUPAGE_SHIFT ? : 1; restart_scan: for (i = preferred; i < eidx; i += incr) { @@ -205,7 +203,7 @@ restart_scan: return NULL; found: - bdata->last_success = start << PAGE_SHIFT; + bdata->last_success = start << MMUPAGE_SHIFT; BUG_ON(start >= eidx); /* @@ -213,30 +211,30 @@ found: * of this allocation's buffer? If yes then we can 'merge' * the previous partial page with this allocation. */ - if (align < PAGE_SIZE && + if (align < MMUPAGE_SIZE && bdata->last_offset && bdata->last_pos+1 == start) { offset = (bdata->last_offset+align-1) & ~(align-1); - BUG_ON(offset > PAGE_SIZE); - remaining_size = PAGE_SIZE-offset; + BUG_ON(offset > MMUPAGE_SIZE); + remaining_size = MMUPAGE_SIZE-offset; if (size < remaining_size) { areasize = 0; /* last_pos unchanged */ bdata->last_offset = offset+size; - ret = phys_to_virt(bdata->last_pos*PAGE_SIZE + offset + + ret = phys_to_virt(bdata->last_pos*MMUPAGE_SIZE + offset + bdata->node_boot_start); } else { remaining_size = size - remaining_size; - areasize = (remaining_size+PAGE_SIZE-1)/PAGE_SIZE; - ret = phys_to_virt(bdata->last_pos*PAGE_SIZE + offset + + areasize = (remaining_size+MMUPAGE_SIZE-1)/MMUPAGE_SIZE; + ret = phys_to_virt(bdata->last_pos*MMUPAGE_SIZE + offset + bdata->node_boot_start); bdata->last_pos = start+areasize-1; bdata->last_offset = remaining_size; } - bdata->last_offset &= ~PAGE_MASK; + bdata->last_offset &= ~MMUPAGE_MASK; } else { bdata->last_pos = start + areasize - 1; - bdata->last_offset = size & ~PAGE_MASK; - ret = phys_to_virt(start * PAGE_SIZE + bdata->node_boot_start); + bdata->last_offset = size & ~MMUPAGE_MASK; + ret = phys_to_virt(start * MMUPAGE_SIZE + bdata->node_boot_start); } /* @@ -253,49 +251,37 @@ static unsigned long __init free_all_boo { struct page *page = pgdat->node_mem_map; bootmem_data_t *bdata = pgdat->bdata; - unsigned long i, count, total = 0; - unsigned long idx; + unsigned long i, total = 0; + unsigned long idx, mapnr, node_low_mapnr; unsigned long *map; - if (!bdata->node_bootmem_map) BUG(); - - count = 0; - idx = bdata->node_low_pfn - (bdata->node_boot_start >> PAGE_SHIFT); + BUG_ON(!bdata->node_bootmem_map); map = bdata->node_bootmem_map; - for (i = 0; i < idx; ) { - unsigned long v = ~map[i / BITS_PER_LONG]; - if (v) { - unsigned long m; - for (m = 1; m && i < idx; m<<=1, page++, i++) { - if (v & m) { - count++; + i = 0; + idx = bdata->node_low_pfn - (bdata->node_boot_start >> MMUPAGE_SHIFT); + + node_low_mapnr = (bdata->node_low_pfn - bdata->node_boot_start/MMUPAGE_SIZE)/PAGE_MMUCOUNT; + for (mapnr = 0; mapnr < node_low_mapnr; ++mapnr) { + int k, should_free = 1; + for (k = 0; k < PAGE_MMUCOUNT; ++k) + if (test_bit(mapnr*PAGE_MMUCOUNT + k, map)) + should_free = 0; + if (should_free) { + page = &pgdat->node_mem_map[mapnr]; ClearPageReserved(page); set_page_count(page, 1); __free_page(page); + ++total; } } - } else { - i+=BITS_PER_LONG; - page += BITS_PER_LONG; - } - } - total += count; /* - * Now free the allocator bitmap itself, it's not - * needed anymore: - */ - page = virt_to_page(bdata->node_bootmem_map); - count = 0; - for (i = 0; i < ((bdata->node_low_pfn-(bdata->node_boot_start >> PAGE_SHIFT))/8 + PAGE_SIZE-1)/PAGE_SIZE; i++,page++) { - count++; - ClearPageReserved(page); - set_page_count(page, 1); - __free_page(page); } - total += count; - bdata->node_bootmem_map = NULL; + * Leak the allocator bitmap; it's not worth saving. + */ + bdata->node_bootmem_map = NULL; + printk("bootmem: freed %lx pages in node %d\n", total, pgdat->node_id); return total; } diff -prauN linux-2.5.70-bk4/mm/filemap.c pgcl-2.5.70-bk4-1/mm/filemap.c --- linux-2.5.70-bk4/mm/filemap.c 2003-05-30 21:02:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/filemap.c 2003-05-30 21:03:24.000000000 -0700 @@ -946,8 +946,8 @@ struct page * filemap_nopage(struct vm_a unsigned long size, pgoff, endoff; int did_readahead; - pgoff = ((address - area->vm_start) >> PAGE_CACHE_SHIFT) + area->vm_pgoff; - endoff = ((area->vm_end - area->vm_start) >> PAGE_CACHE_SHIFT) + area->vm_pgoff; + pgoff = (address - area->vm_start)/MMUPAGE_SIZE + area->vm_pgoff; + endoff = (area->vm_end - area->vm_start + MMUPAGE_SIZE - 1)/MMUPAGE_SIZE + area->vm_pgoff; retry_all: /* @@ -955,15 +955,15 @@ retry_all: * accessible.. */ size = (inode->i_size + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; - if ((pgoff >= size) && (area->vm_mm == current->mm)) + if ((pgoff/PAGE_CACHE_MMUCOUNT >= size) && (area->vm_mm == current->mm)) return NULL; /* * The "size" of the file, as far as mmap is concerned, isn't bigger * than the mapping */ - if (size > endoff) - size = endoff; + if (size > endoff/PAGE_CACHE_MMUCOUNT) + size = endoff/PAGE_CACHE_MMUCOUNT; did_readahead = 0; @@ -973,23 +973,23 @@ retry_all: */ if (VM_SequentialReadHint(area)) { did_readahead = 1; - page_cache_readahead(mapping, ra, file, pgoff); + page_cache_readahead(mapping, ra, file, pgoff/PAGE_CACHE_MMUCOUNT); } /* * If the offset is outside the mapping size we're off the end * of a privately mapped file, so we need to map a zero page. */ - if ((pgoff < size) && !VM_RandomReadHint(area)) { + if ((pgoff/PAGE_CACHE_MMUCOUNT < size) && !VM_RandomReadHint(area)) { did_readahead = 1; - page_cache_readaround(mapping, ra, file, pgoff); + page_cache_readaround(mapping, ra, file, pgoff/PAGE_CACHE_MMUCOUNT); } /* * Do we have something in the page cache already? */ retry_find: - page = find_get_page(mapping, pgoff); + page = find_get_page(mapping, pgoff/PAGE_CACHE_MMUCOUNT); if (!page) { if (did_readahead) { handle_ra_miss(mapping, ra, pgoff); @@ -1017,7 +1017,7 @@ no_cached_page: * We're only likely to ever get here if MADV_RANDOM is in * effect. */ - error = page_cache_read(file, pgoff); + error = page_cache_read(file, pgoff/PAGE_CACHE_MMUCOUNT); /* * The page we want has now been added to the page cache. @@ -1105,7 +1105,7 @@ static struct page * filemap_getpage(str * Do we have something in the page cache already? */ retry_find: - page = find_get_page(mapping, pgoff); + page = find_get_page(mapping, pgoff/PAGE_CACHE_MMUCOUNT); if (!page) { if (nonblock) return NULL; @@ -1127,7 +1127,7 @@ success: return page; no_cached_page: - error = page_cache_read(file, pgoff); + error = page_cache_read(file, pgoff/PAGE_CACHE_MMUCOUNT); /* * The page we want has now been added to the page cache. @@ -1218,26 +1218,30 @@ static int filemap_populate(struct vm_ar if (!nonblock) do_page_cache_readahead(mapping, vma->vm_file, - pgoff, len >> PAGE_CACHE_SHIFT); + pgoff/PAGE_MMUCOUNT, len >> PAGE_CACHE_SHIFT); repeat: size = (inode->i_size + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; - if (pgoff + (len >> PAGE_CACHE_SHIFT) > size) + if (pgoff + len/MMUPAGE_SIZE > size) return -EINVAL; page = filemap_getpage(file, pgoff, nonblock); if (!page && !nonblock) return -ENOMEM; if (page) { - err = install_page(mm, vma, addr, page, prot); + /* + * page caches bytes index*PAGE_SIZE to index*(PAGE_SIZE+1)-1 + * pgoff % PAGE_MMUCOUNT is the subpfn w/in the page + */ + err = install_page(mm, vma, addr, page, prot, pgoff % PAGE_MMUCOUNT); if (err) { page_cache_release(page); return err; } } - len -= PAGE_SIZE; - addr += PAGE_SIZE; + len -= MMUPAGE_SIZE; + addr += MMUPAGE_SIZE; pgoff++; if (len) goto repeat; diff -prauN linux-2.5.70-bk4/mm/fremap.c pgcl-2.5.70-bk4-1/mm/fremap.c --- linux-2.5.70-bk4/mm/fremap.c 2003-05-26 18:00:20.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/fremap.c 2003-05-30 21:03:24.000000000 -0700 @@ -53,13 +53,14 @@ static inline int zap_pte(struct mm_stru * previously existing mapping. */ int install_page(struct mm_struct *mm, struct vm_area_struct *vma, - unsigned long addr, struct page *page, pgprot_t prot) + unsigned long addr, struct page *page, pgprot_t prot, int subpfn) { int err = -ENOMEM, flush; pte_t *pte; pgd_t *pgd; pmd_t *pmd; struct pte_chain *pte_chain; + unsigned long pfn = page_to_pfn(page) + subpfn; pte_chain = pte_chain_alloc(GFP_KERNEL); if (!pte_chain) @@ -79,7 +80,7 @@ int install_page(struct mm_struct *mm, s mm->rss++; flush_icache_page(vma, page); - set_pte(pte, mk_pte(page, prot)); + set_pte(pte, pfn_pte(pfn, prot)); pte_chain = page_add_rmap(page, pte, pte_chain); pte_unmap(pte); if (flush) @@ -127,8 +128,8 @@ long sys_remap_file_pages(unsigned long /* * Sanitize the syscall parameters: */ - start = start & PAGE_MASK; - size = size & PAGE_MASK; + start = start & MMUPAGE_MASK; + size = size & MMUPAGE_MASK; /* Does the address range wrap, or is the span zero-sized? */ if (start + size <= start) @@ -136,7 +137,7 @@ long sys_remap_file_pages(unsigned long /* Can we represent this offset inside this architecture's pte's? */ #if PTE_FILE_MAX_BITS < BITS_PER_LONG - if (pgoff + (size >> PAGE_SHIFT) >= (1UL << PTE_FILE_MAX_BITS)) + if (pgoff + (size >> MMUPAGE_SHIFT) >= (1UL << PTE_FILE_MAX_BITS)) return err; #endif diff -prauN linux-2.5.70-bk4/mm/highmem.c pgcl-2.5.70-bk4-1/mm/highmem.c --- linux-2.5.70-bk4/mm/highmem.c 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/highmem.c 2003-05-30 21:03:24.000000000 -0700 @@ -54,8 +54,6 @@ static int pkmap_count[LAST_PKMAP]; static unsigned int last_pkmap_nr; static spinlock_t kmap_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED; -pte_t * pkmap_page_table; - static DECLARE_WAIT_QUEUE_HEAD(pkmap_map_wait); static void flush_all_zero_pkmaps(void) @@ -65,6 +63,8 @@ static void flush_all_zero_pkmaps(void) flush_cache_all(); for (i = 0; i < LAST_PKMAP; i++) { + int j; + unsigned long vaddr = PKMAP_ADDR(i); struct page *page; /* @@ -78,8 +78,14 @@ static void flush_all_zero_pkmaps(void) pkmap_count[i] = 0; /* sanity check */ - if (pte_none(pkmap_page_table[i])) - BUG(); + for (j = 0; j < PAGE_MMUCOUNT; ++j) { + unsigned long addr = vaddr + j*MMUPAGE_SIZE; + pgd_t *pgd = pgd_offset_k(addr); + pmd_t *pmd = pmd_offset(pgd, addr); + pte_t *pte = pte_offset_kernel(pmd, addr); + + BUG_ON(pte_none(*pte)); + } /* * Don't need an atomic fetch-and-clear op here; @@ -88,8 +94,20 @@ static void flush_all_zero_pkmaps(void) * getting the kmap_lock (which is held here). * So no dangers, even with speculative execution. */ - page = pte_page(pkmap_page_table[i]); - pte_clear(&pkmap_page_table[i]); + { + pgd_t *pgd = pgd_offset_k(vaddr); + pmd_t *pmd = pmd_offset(pgd, vaddr); + pte_t *pte = pte_offset_kernel(pmd, vaddr); + page = pte_page(*pte); + } + + for (j = 0; j < PAGE_MMUCOUNT; ++j) { + unsigned long addr = vaddr + j*MMUPAGE_SIZE; + pgd_t *pgd = pgd_offset_k(addr); + pmd_t *pmd = pmd_offset(pgd, addr); + pte_t *pte = pte_offset_kernel(pmd, addr); + pte_clear(pte); + } set_page_address(page, NULL); } @@ -99,7 +117,7 @@ static void flush_all_zero_pkmaps(void) static inline unsigned long map_new_virtual(struct page *page) { unsigned long vaddr; - int count; + int k, count; start: count = LAST_PKMAP; @@ -137,7 +155,15 @@ start: } } vaddr = PKMAP_ADDR(last_pkmap_nr); - set_pte(&(pkmap_page_table[last_pkmap_nr]), mk_pte(page, kmap_prot)); + WARN_ON(vaddr > __fix_to_virt(FIX_PKMAP_BEGIN)); + WARN_ON(vaddr < __fix_to_virt(FIX_PKMAP_END)); + for (k = 0; k < PAGE_MMUCOUNT; ++k) { + unsigned long addr = vaddr + k * MMUPAGE_SIZE; + pgd_t *pgd = pgd_offset_k(addr); + pmd_t *pmd = pmd_offset(pgd, addr); + pte_t *pte = pte_offset_kernel(pmd, addr); + set_pte(pte, pfn_pte(page_to_pfn(page) + k, kmap_prot)); + } pkmap_count[last_pkmap_nr] = 1; set_page_address(page, (void *)vaddr); @@ -479,12 +505,19 @@ void check_highmem_ptes(void) preempt_disable(); for (type = 0; type < KM_TYPE_NR; type++) { + int k; idx = type + KM_TYPE_NR*smp_processor_id(); - if (!pte_none(*(kmap_pte-idx))) { + for (k = 0; k < PAGE_MMUCOUNT; ++k) { + unsigned long addr = __fix_to_virt(FIX_KMAP_END) + idx*PAGE_SIZE + k*MMUPAGE_SIZE; + pgd_t *pgd = pgd_offset_k(addr); + pmd_t *pmd = pmd_offset(pgd, addr); + pte_t *pte = pte_offset_kernel(pmd, addr); + if (!pte_none(*pte)) { printk("scheduling with KM_TYPE %d held!\n", type); BUG(); } } + } preempt_enable(); } #endif diff -prauN linux-2.5.70-bk4/mm/madvise.c pgcl-2.5.70-bk4-1/mm/madvise.c --- linux-2.5.70-bk4/mm/madvise.c 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/madvise.c 2003-05-30 21:03:24.000000000 -0700 @@ -60,10 +60,12 @@ static long madvise_willneed(struct vm_a if (!file) return -EBADF; - start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; + start = ((start - vma->vm_start) >> MMUPAGE_SHIFT) + vma->vm_pgoff; if (end > vma->vm_end) end = vma->vm_end; - end = ((end - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; + end = ((end - vma->vm_start) >> MMUPAGE_SHIFT) + vma->vm_pgoff; + start /= PAGE_MMUCOUNT; + end /= PAGE_MMUCOUNT; do_page_cache_readahead(file->f_dentry->d_inode->i_mapping, file, start, max_sane_readahead(end - start)); @@ -170,9 +172,9 @@ asmlinkage long sys_madvise(unsigned lon down_write(¤t->mm->mmap_sem); - if (start & ~PAGE_MASK) + if (start & ~MMUPAGE_MASK) goto out; - len = (len + ~PAGE_MASK) & PAGE_MASK; + len = (len + ~MMUPAGE_MASK) & MMUPAGE_MASK; end = start + len; if (end < start) goto out; diff -prauN linux-2.5.70-bk4/mm/memory.c pgcl-2.5.70-bk4-1/mm/memory.c --- linux-2.5.70-bk4/mm/memory.c 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/memory.c 2003-05-30 21:17:20.000000000 -0700 @@ -150,11 +150,32 @@ pte_t * pte_alloc_map(struct mm_struct * /* * Because we dropped the lock, we should re-check the * entry, as somebody else could have populated it.. + * If we raced, we also need to drop all the reference + * counts originally taken with the intent of conferring + * them to all the pte entries spanned by the pte page. */ if (pmd_present(*pmd)) { + if (PAGE_MMUCOUNT > 1) + atomic_sub(PAGE_MMUCOUNT-1, &new->count); pte_free(new); goto out; } +#if 0 + { + int k; + pmd_t *base; + unsigned long addr, __pmd = (unsigned long)pmd; + addr = address & ~(PAGE_MMUCOUNT*PMD_SIZE - 1); + base = pmd - ((__pmd/sizeof(pmd_t)) % PAGE_MMUCOUNT); + for (k = 0; k < PAGE_MMUCOUNT; ++k) + if (!pmd_none(base[k]) || pmd_present(base[k])) + printk(KERN_DEBUG + "redundant pmd instantiation " + "at vaddr 0x%lx pmd = 0x%p\n", + addr + PMD_SIZE*k, + &base[k]); + } +#endif pgtable_add_rmap(new, mm, address); pmd_populate(mm, pmd, new); } @@ -347,7 +368,7 @@ skip_copy_pte_range: src_pte = pte_offset_map_nested(src_pmd, address); cont_copy_pte_range_noset: - address += PAGE_SIZE; + address += MMUPAGE_SIZE; if (address >= end) { pte_unmap_nested(src_pte); pte_unmap(dst_pte); @@ -393,8 +414,8 @@ zap_pte_range(struct mmu_gather *tlb, pm offset = address & ~PMD_MASK; if (offset + size > PMD_SIZE) size = PMD_SIZE - offset; - size &= PAGE_MASK; - for (offset=0; offset < size; ptep++, offset += PAGE_SIZE) { + size &= MMUPAGE_MASK; + for (offset=0; offset < size; ptep++, offset += MMUPAGE_SIZE) { pte_t pte = *ptep; if (pte_none(pte)) continue; @@ -417,8 +438,14 @@ zap_pte_range(struct mmu_gather *tlb, pm } } } else { - if (!pte_file(pte)) + if (!pte_file(pte)) { + if (pte_to_swp_entry(pte).val == 0x8073756dUL) + printk(KERN_DEBUG "detected fsckup " + "early, leaking stuff to " + "work around it\n"); + else free_swap_and_cache(pte_to_swp_entry(pte)); + } pte_clear(ptep); } } @@ -474,12 +501,12 @@ void unmap_page_range(struct mmu_gather /* Dispose of an entire struct mmu_gather per rescheduling point */ #if defined(CONFIG_SMP) && defined(CONFIG_PREEMPT) -#define ZAP_BLOCK_SIZE (FREE_PTE_NR * PAGE_SIZE) +#define ZAP_BLOCK_SIZE (FREE_PTE_NR * MMUPAGE_SIZE) #endif /* For UP, 256 pages at a time gives nice low latency */ #if !defined(CONFIG_SMP) && defined(CONFIG_PREEMPT) -#define ZAP_BLOCK_SIZE (256 * PAGE_SIZE) +#define ZAP_BLOCK_SIZE (256 * MMUPAGE_SIZE) #endif /* No preempt: go for the best straight-line efficiency */ @@ -524,10 +551,10 @@ int unmap_vmas(struct mmu_gather **tlbp, if (vma) { /* debug. killme. */ if (end_addr <= vma->vm_start) - printk("%s: end_addr(0x%08lx) <= vm_start(0x%08lx)\n", + pr_debug("%s: end_addr(0x%08lx) <= vm_start(0x%08lx)\n", __FUNCTION__, end_addr, vma->vm_start); if (start_addr >= vma->vm_end) - printk("%s: start_addr(0x%08lx) <= vm_end(0x%08lx)\n", + pr_debug("%s: start_addr(0x%08lx) <= vm_end(0x%08lx)\n", __FUNCTION__, start_addr, vma->vm_end); } @@ -543,7 +570,7 @@ int unmap_vmas(struct mmu_gather **tlbp, continue; if (vma->vm_flags & VM_ACCOUNT) - *nr_accounted += (end - start) >> PAGE_SHIFT; + *nr_accounted += (end - start) >> MMUPAGE_SHIFT; ret++; while (start != end) { @@ -573,7 +600,7 @@ int unmap_vmas(struct mmu_gather **tlbp, zap_bytes = ZAP_BLOCK_SIZE; } if (vma->vm_next && vma->vm_next->vm_start < vma->vm_end) - printk("%s: VMA list is not sorted correctly!\n", + pr_debug("%s: VMA list is not sorted correctly!\n", __FUNCTION__); } return ret; @@ -612,18 +639,19 @@ void zap_page_range(struct vm_area_struc * Do a quick page-table lookup for a single page. * mm->page_table_lock must be held. */ -struct page * -follow_page(struct mm_struct *mm, unsigned long address, int write) +unsigned long follow_page(struct mm_struct *mm, unsigned long address, int write) { pgd_t *pgd; pmd_t *pmd; pte_t *ptep, pte; unsigned long pfn; - struct vm_area_struct *vma; +#if 0 + struct vm_area_struct *vma; vma = hugepage_vma(mm, address); if (vma) return follow_huge_addr(mm, vma, address, write); +#endif pgd = pgd_offset(mm, address); if (pgd_none(*pgd) || pgd_bad(*pgd)) @@ -632,8 +660,15 @@ follow_page(struct mm_struct *mm, unsign pmd = pmd_offset(pgd, address); if (pmd_none(*pmd)) goto out; + + /* + * hugetlb's still broken in pgcl; not difficult to fix, + * but an unnecessary distraction while it's in flux + */ +#if 0 if (pmd_huge(*pmd)) return follow_huge_pmd(mm, address, pmd, write); +#endif if (pmd_bad(*pmd)) goto out; @@ -647,12 +682,12 @@ follow_page(struct mm_struct *mm, unsign if (!write || (pte_write(pte) && pte_dirty(pte))) { pfn = pte_pfn(pte); if (pfn_valid(pfn)) - return pfn_to_page(pfn); + return pfn; /* pfn_to_page(pfn) */ } } out: - return NULL; + return 0; /* NULL */ } /* @@ -664,14 +699,26 @@ out: static inline struct page *get_page_map(struct page *page) { if (!pfn_valid(page_to_pfn(page))) - return 0; + return NULL; return page; } +static inline unsigned long get_pfn_map(unsigned long pfn) +{ + return pfn_valid(pfn) ? pfn : 0; +} + +/* + * This puppy is handing back MMUPAGE_SIZE -sized slots. + * Callers need auditing. + * This function is a goddamn train wreck. Someone needs to + * janitor the idiot thing for mainline to at very least kill + * the #ifdef FIXADDR_START bullcrap. + */ int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start, int len, int write, int force, - struct page **pages, struct vm_area_struct **vmas) + unsigned long *pfns, struct vm_area_struct **vmas) { int i; unsigned int flags; @@ -701,46 +748,48 @@ int get_user_pages(struct task_struct *t .vm_page_prot = PAGE_READONLY, .vm_flags = VM_READ | VM_EXEC, }; - unsigned long pg = start & PAGE_MASK; + unsigned long pg = start & MMUPAGE_MASK; pgd_t *pgd; pmd_t *pmd; pte_t *pte; pgd = pgd_offset_k(pg); if (!pgd) - return i ? : -EFAULT; + return i ? i : -EFAULT; pmd = pmd_offset(pgd, pg); if (!pmd) - return i ? : -EFAULT; + return i ? i : -EFAULT; pte = pte_offset_kernel(pmd, pg); if (!pte || !pte_present(*pte) || !pte_user(*pte) || !(write ? pte_write(*pte) : pte_read(*pte))) - return i ? : -EFAULT; - if (pages) { - pages[i] = pte_page(*pte); - get_page(pages[i]); + return i ? i : -EFAULT; + if (pfns) { + pfns[i] = pte_pfn(*pte); + get_page(pfn_to_page(pfns[i])); } if (vmas) vmas[i] = &fixmap_vma; i++; - start += PAGE_SIZE; + start += MMUPAGE_SIZE; len--; continue; } #endif - if (!vma || (pages && (vma->vm_flags & VM_IO)) + if (!vma || (pfns && (vma->vm_flags & VM_IO)) || !(flags & vma->vm_flags)) - return i ? : -EFAULT; + return i ? i : -EFAULT; +#if 0 if (is_vm_hugetlb_page(vma)) { i = follow_hugetlb_page(mm, vma, pages, vmas, &start, &len, i); continue; } +#endif spin_lock(&mm->page_table_lock); do { - struct page *map; - while (!(map = follow_page(mm, start, write))) { + unsigned long map_pfn; + while (!(map_pfn = follow_page(mm, start, write))) { spin_unlock(&mm->page_table_lock); switch (handle_mm_fault(mm,vma,start,write)) { case VM_FAULT_MINOR: @@ -750,36 +799,50 @@ int get_user_pages(struct task_struct *t tsk->maj_flt++; break; case VM_FAULT_SIGBUS: + if (!i) + printk("get_user_pages(): VM_FAULT_SIGBUS\n"); return i ? i : -EFAULT; case VM_FAULT_OOM: + if (!i) + printk("get_user_pages(): VM_FAULT_OOM\n"); return i ? i : -ENOMEM; default: BUG(); } spin_lock(&mm->page_table_lock); } - if (pages) { - pages[i] = get_page_map(map); - if (!pages[i]) { + if (pfns) { + pfns[i] = get_pfn_map(map_pfn); + if (!pfns[i]) { spin_unlock(&mm->page_table_lock); - while (i--) - page_cache_release(pages[i]); + while (i--) { + struct page *map; + map = pfn_to_page(pfns[i]); + page_cache_release(map); + } i = -EFAULT; + printk("get_user_pages(): saw a zero pfn\n"); goto out; } - flush_dcache_page(pages[i]); - if (!PageReserved(pages[i])) - page_cache_get(pages[i]); + if (1) { + struct page *map; + map = pfn_to_page(pfns[i]); + flush_dcache_page(map); + if (!PageReserved(map)) + page_cache_get(map); + } } if (vmas) vmas[i] = vma; i++; - start += PAGE_SIZE; + start += MMUPAGE_SIZE; len--; } while(len && start < vma->vm_end); spin_unlock(&mm->page_table_lock); } while(len); out: + if (i < 0) + pr_debug("get_user_pages() returning an error\n"); return i; } @@ -796,7 +859,7 @@ static void zeromap_pte_range(pte_t * pt pte_t zero_pte = pte_wrprotect(mk_pte(ZERO_PAGE(address), prot)); BUG_ON(!pte_none(*pte)); set_pte(pte, zero_pte); - address += PAGE_SIZE; + address += MMUPAGE_SIZE; pte++; } while (address && (address < end)); } @@ -832,8 +895,7 @@ int zeromap_page_range(struct vm_area_st dir = pgd_offset(mm, address); flush_cache_range(vma, beg, end); - if (address >= end) - BUG(); + BUG_ON(address >= end); spin_lock(&mm->page_table_lock); do { @@ -867,12 +929,12 @@ static inline void remap_pte_range(pte_t end = address + size; if (end > PMD_SIZE) end = PMD_SIZE; - pfn = phys_addr >> PAGE_SHIFT; + pfn = phys_addr >> MMUPAGE_SHIFT; do { BUG_ON(!pte_none(*pte)); if (!pfn_valid(pfn) || PageReserved(pfn_to_page(pfn))) set_pte(pte, pfn_pte(pfn, prot)); - address += PAGE_SIZE; + address += MMUPAGE_SIZE; pfn++; pte++; } while (address && (address < end)); @@ -913,8 +975,7 @@ int remap_page_range(struct vm_area_stru phys_addr -= from; dir = pgd_offset(mm, from); flush_cache_range(vma, beg, end); - if (from >= end) - BUG(); + BUG_ON(from >= end); spin_lock(&mm->page_table_lock); do { @@ -951,12 +1012,12 @@ static inline void establish_pte(struct /* * We hold the mm semaphore for reading and vma->vm_mm->page_table_lock */ -static inline void break_cow(struct vm_area_struct * vma, struct page * new_page, unsigned long address, - pte_t *page_table) +static inline void break_cow(struct vm_area_struct * vma, struct page * new_page, unsigned long address, pte_t *page_table, unsigned long subpfn) { + pte_t pte = pfn_pte(page_to_pfn(new_page) + subpfn, vma->vm_page_prot); invalidate_vcache(address, vma->vm_mm, new_page); flush_cache_page(vma, address); - establish_pte(vma, address, page_table, pte_mkwrite(pte_mkdirty(mk_pte(new_page, vma->vm_page_prot)))); + establish_pte(vma, address, page_table, pte_mkwrite(pte_mkdirty(pte))); } /* @@ -1004,6 +1065,10 @@ static int do_wp_page(struct mm_struct * int reuse = can_share_swap_page(old_page); unlock_page(old_page); if (reuse) { + /* + * XXX: this should sweep the pagetables to + * prefault all the pte's. This is free, take it. + */ flush_cache_page(vma, address); establish_pte(vma, address, page_table, pte_mkyoung(pte_mkdirty(pte_mkwrite(pte)))); @@ -1016,6 +1081,8 @@ static int do_wp_page(struct mm_struct * /* * Ok, we need to copy. Oh, well.. + * XXX: This needs to sweep the pagetables in an analogous + * manner to do_anonymous_page(). */ page_cache_get(old_page); spin_unlock(&mm->page_table_lock); @@ -1034,10 +1101,11 @@ static int do_wp_page(struct mm_struct * spin_lock(&mm->page_table_lock); page_table = pte_offset_map(pmd, address); if (pte_same(*page_table, pte)) { + unsigned long subpfn = pfn & (PAGE_MMUCOUNT-1); if (PageReserved(old_page)) ++mm->rss; page_remove_rmap(old_page, page_table); - break_cow(vma, new_page, address, page_table); + break_cow(vma, new_page, address, page_table, subpfn); pte_chain = page_add_rmap(new_page, page_table, pte_chain); lru_cache_add_active(new_page); @@ -1079,14 +1147,14 @@ static void vmtruncate_list(struct list_ } /* mapping wholly unaffected? */ - len = len >> PAGE_SHIFT; + len = len >> MMUPAGE_SHIFT; diff = pgoff - vma->vm_pgoff; if (diff >= len) continue; /* Ok, partially affected.. */ - start += diff << PAGE_SHIFT; - len = (len - diff) << PAGE_SHIFT; + start += diff << MMUPAGE_SHIFT; + len = (len - diff) << MMUPAGE_SHIFT; zap_page_range(vma, start, len); } } @@ -1108,7 +1176,7 @@ int vmtruncate(struct inode * inode, lof if (inode->i_size < offset) goto do_expand; inode->i_size = offset; - pgoff = (offset + PAGE_SIZE - 1) >> PAGE_SHIFT; + pgoff = (offset + MMUPAGE_SIZE - 1) / MMUPAGE_SIZE; down(&mapping->i_shared_sem); if (unlikely(!list_empty(&mapping->i_mmap))) vmtruncate_list(&mapping->i_mmap, pgoff); @@ -1174,8 +1242,13 @@ static int do_swap_page(struct mm_struct struct page *page; swp_entry_t entry = pte_to_swp_entry(orig_pte); pte_t pte; - int ret = VM_FAULT_MINOR; + int rss, ret = VM_FAULT_MINOR; struct pte_chain *pte_chain = NULL; + unsigned long subpfn, flt_subpfn = swp_offset(entry) % PAGE_MMUCOUNT; + unsigned long pfn, lo_vaddr, hi_vaddr, vaddr; + + lo_vaddr = max(address & PAGE_MASK, vma->vm_start); + hi_vaddr = min(PAGE_ALIGN(address), vma->vm_end); pte_unmap(page_table); spin_unlock(&mm->page_table_lock); @@ -1207,7 +1280,7 @@ static int do_swap_page(struct mm_struct mark_page_accessed(page); pte_chain = pte_chain_alloc(GFP_KERNEL); if (!pte_chain) { - ret = -ENOMEM; + ret = VM_FAULT_OOM; goto out; } lock_page(page); @@ -1229,24 +1302,90 @@ static int do_swap_page(struct mm_struct /* The page isn't present yet, go ahead with the fault. */ + /* swap_free(entry); if (vm_swap_full()) remove_exclusive_swap_page(page); + */ + + rss = 0; + vaddr = lo_vaddr; + page_table -= (address - lo_vaddr)/MMUPAGE_SIZE; + + flush_icache_page(vma, page); + + pfn = page_to_pfn(page); + + do { + /* already faulted in? less work for me */ + if (pte_present(*page_table)) + goto next; + + entry = pte_to_swp_entry(*page_table); + + if (!pte_none(*page_table) && + swp_offset(entry)/PAGE_MMUCOUNT == page->index) { + + swap_free(entry); + if (vm_swap_full()) + remove_exclusive_swap_page(page); + subpfn = swp_offset(entry) % PAGE_MMUCOUNT; + pte = pfn_pte(pfn + subpfn, vma->vm_page_prot); + + } else if (pte_none(*page_table)) { + + subpfn = flt_subpfn + (vaddr - address)/MMUPAGE_SHIFT; + + /* it'd fall outside the page */ + if (subpfn >= PAGE_MMUCOUNT) + goto next; + + pte = pfn_pte(pfn + subpfn, vma->vm_page_prot); + + /* !pte_none() && swp_offset()/PAGE_MMUCOUNT != page->index */ + } else + goto next; - mm->rss++; - pte = mk_pte(page, vma->vm_page_prot); if (write_access && can_share_swap_page(page)) pte = pte_mkdirty(pte_mkwrite(pte)); - unlock_page(page); - flush_icache_page(vma, page); + if (!pte_chain) + pte_chain = pte_chain_alloc(GFP_ATOMIC); + if (!pte_chain) { + pte_unmap(page_table); + spin_unlock(&mm->page_table_lock); + pte_chain = pte_chain_alloc(GFP_KERNEL); + if (!pte_chain) { + ret = VM_FAULT_OOM; + spin_lock(&mm->page_table_lock); + mm->rss += rss; + spin_unlock(&mm->page_table_lock); + goto no_mem; + } + spin_lock(&mm->page_table_lock); + page_table = pte_offset_map(pmd, vaddr); + } + set_pte(page_table, pte); + ++rss; pte_chain = page_add_rmap(page, page_table, pte_chain); +next: + vaddr += MMUPAGE_SIZE; + page_table++; + } while (vaddr < hi_vaddr); - /* No need to invalidate - it was non-present before */ + unlock_page(page); update_mmu_cache(vma, address, pte); - pte_unmap(page_table); + mm->rss += rss; + pte_unmap(page_table-1); spin_unlock(&mm->page_table_lock); +no_mem: + if (!page) + goto out; + if (!rss) + page_cache_release(page); + else if (rss > 1) + atomic_add(rss - 1, &page->count); out: pte_chain_free(pte_chain); return ret; @@ -1262,66 +1401,200 @@ do_anonymous_page(struct mm_struct *mm, pte_t *page_table, pmd_t *pmd, int write_access, unsigned long addr) { - pte_t entry; - struct page * page = ZERO_PAGE(addr); - struct pte_chain *pte_chain; - int ret; - + struct page *page = NULL; + struct pte_chain *pte_chain = NULL; + unsigned long up_vaddr, dn_vaddr, lo_vaddr, hi_vaddr; + unsigned long pfn, subpfn, dn_subpfn, up_subpfn; + pte_t *ptes[PAGE_MMUCOUNT] = { [0 ... PAGE_MMUCOUNT-1] = NULL }; + pte_t *up_pte, *dn_pte; + int rss, ret = VM_FAULT_MINOR; + + if (write_access) + pr_debug("write fault on 0x%lx\n", addr); + else + pr_debug("read fault on 0x%lx\n", addr); + pr_debug("page_table = 0x%p\n", page_table); + + if (!write_access) + page = ZERO_PAGE(addr); + else { + if (!pte_chain) pte_chain = pte_chain_alloc(GFP_ATOMIC); - if (!pte_chain) { pte_unmap(page_table); spin_unlock(&mm->page_table_lock); + if (!pte_chain) { pte_chain = pte_chain_alloc(GFP_KERNEL); if (!pte_chain) - goto no_mem; - spin_lock(&mm->page_table_lock); - page_table = pte_offset_map(pmd, addr); + return VM_FAULT_OOM; + } + page = alloc_page(GFP_HIGHUSER); + if (!page) { + pte_chain_free(pte_chain); + return VM_FAULT_OOM; + } + clear_user_highpage(page, addr); } - /* Read-only mapping of ZERO_PAGE. */ - entry = pte_wrprotect(mk_pte(ZERO_PAGE(addr), vma->vm_page_prot)); + lo_vaddr = max(addr & ~(PAGE_MMUCOUNT*PMD_SIZE - 1), vma->vm_start); + hi_vaddr = min(vma->vm_end, (addr + PAGE_MMUCOUNT*PMD_SIZE - 1) + & ~(PAGE_MMUCOUNT*PMD_SIZE - 1)); + dn_subpfn = 0; + up_subpfn = PAGE_MMUCOUNT - 1; + dn_vaddr = addr & MMUPAGE_MASK; + up_vaddr = MMUPAGE_ALIGN(addr + 1); + + pr_debug("vma->vm_start = 0x%lx, vma->vm_end = 0x%lx\n", + vma->vm_start, vma->vm_end); + pr_debug("lo_vaddr = 0x%lx, hi_vaddr = 0x%lx\n", lo_vaddr, hi_vaddr); + pr_debug("dn_vaddr = 0x%lx, up_vaddr = 0x%lx\n", dn_vaddr, up_vaddr); - /* ..except if it's a write access */ if (write_access) { - /* Allocate our own private page. */ - pte_unmap(page_table); - spin_unlock(&mm->page_table_lock); + pr_debug("about to take mm->page_table_lock\n"); + if (spin_is_locked(&mm->page_table_lock)) + printk("hmm, I see a deadlock coming\n"); + spin_lock(&mm->page_table_lock); + } - page = alloc_page(GFP_HIGHUSER); - if (!page) - goto no_mem; - clear_user_highpage(page, addr); + pr_debug("starting PTE search loop\n"); + if (write_access) + page_table = dn_pte = pte_offset_map(pmd, dn_vaddr); + else + dn_pte = page_table; + up_pte = dn_pte + 1; - spin_lock(&mm->page_table_lock); - page_table = pte_offset_map(pmd, addr); + do { + if (up_vaddr < hi_vaddr && up_subpfn > dn_subpfn) { + if (pte_none(*up_pte)) { + ptes[up_subpfn] = up_pte; + up_subpfn--; + } + up_vaddr += MMUPAGE_SIZE; + up_pte++; + } - if (!pte_none(*page_table)) { - pte_unmap(page_table); - page_cache_release(page); - spin_unlock(&mm->page_table_lock); - ret = VM_FAULT_MINOR; - goto out; + if (dn_vaddr >= lo_vaddr && dn_subpfn < up_subpfn) { + if (pte_none(*dn_pte)) { + ptes[dn_subpfn] = dn_pte; + dn_subpfn++; } - mm->rss++; - entry = pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot))); - lru_cache_add_active(page); - mark_page_accessed(page); + dn_vaddr -= MMUPAGE_SIZE; + dn_pte--; } + pr_debug("dn_vaddr = 0x%lx, up_vaddr = 0x%lx\n", + dn_vaddr, up_vaddr); + pr_debug("dn_subpfn = 0x%lx, up_subpfn = 0x%lx\n", + dn_subpfn, up_subpfn); + } while ((up_vaddr < hi_vaddr || dn_vaddr >= lo_vaddr) && + up_subpfn > dn_subpfn); + + pr_debug("finishing PTE search loop\n"); + pr_debug("starting PTE instantiation loop\n"); + + pfn = page_to_pfn(page); + rss = 0; + for (subpfn = 0; subpfn < PAGE_MMUCOUNT; ++subpfn) { + pte_t pte; - set_pte(page_table, entry); - /* ignores ZERO_PAGE */ - pte_chain = page_add_rmap(page, page_table, pte_chain); - pte_unmap(page_table); + pr_debug("subpfn = 0x%lx, ptep = 0x%p\n", subpfn, ptes[subpfn]); - /* No need to invalidate - it was non-present before */ - update_mmu_cache(vma, addr, entry); - spin_unlock(&mm->page_table_lock); - ret = VM_FAULT_MINOR; - goto out; + if (!ptes[subpfn]) { + pr_debug("pte empty\n"); + continue; + } else if (!pte_none(*ptes[subpfn])) { + pr_debug("pte non-none\n"); + continue; + } -no_mem: - ret = VM_FAULT_OOM; -out: + pte = pfn_pte(pfn + subpfn, vma->vm_page_prot); + if (!write_access) { + pr_debug("setting pte to zero page\n"); + set_pte(ptes[subpfn], pte_wrprotect(pte)); + } else { + pr_debug("setting pte to newly zeroed anonymous page\n"); + if (!pte_chain) + pte_chain = pte_chain_alloc(GFP_ATOMIC); + if (!pte_chain) { + unsigned long vaddr, offset; + int k; + + pr_debug("doing sleeping alloc of pte_chain" + " for non-anonymous page\n"); + + vaddr = ptep_to_address(ptes[subpfn]); + + pr_debug("vaddr = 0x%lx\n", vaddr); + + pte_unmap(ptes[subpfn]); + spin_unlock(&mm->page_table_lock); + pte_chain = pte_chain_alloc(GFP_KERNEL); + if (!pte_chain) { + pr_debug("going to out_oom\n"); + ret = VM_FAULT_OOM; + goto out_oom; + } + spin_lock(&mm->page_table_lock); + page_table = pte_offset_map(pmd, vaddr); + + /* is this safe from gcc? NFI */ + if (page_table != ptes[subpfn]) { + pr_debug("(page_table) 0x%p != 0x%p" + " (ptes[subpfn])\n", + page_table, + ptes[subpfn]); + offset = (unsigned long) + (page_table - ptes[subpfn]); + pr_debug("adjusting all ptes by" + " offset 0x%lx\n", + offset); + for (k = subpfn; k < PAGE_MMUCOUNT; ++k) { + pr_debug("pte before 0x%p\n", + ptes[k]); + if (ptes[k]) + ptes[k] += offset; + pr_debug("pte after 0x%p\n", + ptes[k]); + } + } + + /* check for races */ + if (!pte_none(*ptes[subpfn])) { + pr_debug("raced, skipping PTE\n"); + continue; + } + } + pr_debug("setting pte for anonymous zeroed page\n"); + pr_debug("ptep = 0x%p, pte = 0x%Lx\n", + ptes[subpfn], (u64)pte_val(pte)); + set_pte(ptes[subpfn], pte_mkwrite(pte_mkdirty(pte))); + pr_debug("about to page_add_rmap()\n"); + pte_chain = page_add_rmap(page, ptes[subpfn], + pte_chain); + pr_debug("about to update_mmu_cache()\n"); + update_mmu_cache(vma, addr, pte); + rss++; + pr_debug("about to page_cache_get()\n"); + page_cache_get(page); + } + pr_debug("falling through to next subpfn\n"); + } + pr_debug("doing pte_unmap(0x%p)\n", page_table); + pte_unmap(page_table); + pr_debug("adding %d to rss\n", rss); + mm->rss += rss; + spin_unlock(&mm->page_table_lock); + pr_debug("broke out of PTE instantiation loop\n"); +out_oom: + pr_debug("at out_oom\n"); + if (write_access) { + if (rss) { + pr_debug("adding page to LRU\n"); + lru_cache_add_active(page); + mark_page_accessed(page); + } + pr_debug("releasing page\n"); + page_cache_release(page); + } + pr_debug("doing pte_chain_free()\n"); pte_chain_free(pte_chain); return ret; } @@ -1353,12 +1626,12 @@ do_no_page(struct mm_struct *mm, struct pte_unmap(page_table); spin_unlock(&mm->page_table_lock); - new_page = vma->vm_ops->nopage(vma, address & PAGE_MASK, 0); + new_page = vma->vm_ops->nopage(vma, address & MMUPAGE_MASK, 0); /* no page was available -- either SIGBUS or OOM */ if (new_page == NOPAGE_SIGBUS) return VM_FAULT_SIGBUS; - if (new_page == NOPAGE_OOM) + else if (new_page == NOPAGE_OOM) return VM_FAULT_OOM; pte_chain = pte_chain_alloc(GFP_KERNEL); @@ -1392,12 +1665,17 @@ do_no_page(struct mm_struct *mm, struct * an exclusive copy of the page, or this is a shared mapping, * so we can make it writable and dirty to avoid having to * handle that later. + * + * XXX: this should sweep pagetables and prefault */ /* Only go through if we didn't race with anybody else... */ if (pte_none(*page_table)) { + unsigned long pfn; ++mm->rss; flush_icache_page(vma, new_page); - entry = mk_pte(new_page, vma->vm_page_prot); + pfn = page_to_pfn(new_page) + + vma_suboffset(vma, address)/MMUPAGE_SIZE; + entry = pfn_pte(pfn, vma->vm_page_prot); if (write_access) entry = pte_mkwrite(pte_mkdirty(entry)); set_pte(page_table, entry); @@ -1451,7 +1729,7 @@ static int do_file_page(struct mm_struct pte_unmap(pte); spin_unlock(&mm->page_table_lock); - err = vma->vm_ops->populate(vma, address & PAGE_MASK, PAGE_SIZE, vma->vm_page_prot, pgoff, 0); + err = vma->vm_ops->populate(vma, address & MMUPAGE_MASK, MMUPAGE_SIZE, vma->vm_page_prot, pgoff, 0); if (err == -ENOMEM) return VM_FAULT_OOM; if (err) @@ -1585,11 +1863,9 @@ int make_pages_present(unsigned long add vma = find_vma(current->mm, addr); write = (vma->vm_flags & VM_WRITE) != 0; - if (addr >= end) - BUG(); - if (end > vma->vm_end) - BUG(); - len = (end+PAGE_SIZE-1)/PAGE_SIZE-addr/PAGE_SIZE; + BUG_ON(addr >= end); + BUG_ON(end > vma->vm_end); + len = (end+MMUPAGE_SIZE-1)/MMUPAGE_SIZE-addr/MMUPAGE_SIZE; ret = get_user_pages(current, current->mm, addr, len, write, 0, NULL, NULL); return ret == len ? 0 : -1; diff -prauN linux-2.5.70-bk4/mm/mincore.c pgcl-2.5.70-bk4-1/mm/mincore.c --- linux-2.5.70-bk4/mm/mincore.c 2003-05-26 18:01:00.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/mincore.c 2003-05-30 21:03:24.000000000 -0700 @@ -29,7 +29,7 @@ static unsigned char mincore_page(struct struct address_space * as = vma->vm_file->f_dentry->d_inode->i_mapping; struct page * page; - page = find_get_page(as, pgoff); + page = find_get_page(as, pgoff/PAGE_CACHE_MMUCOUNT); if (page) { present = PageUptodate(page); page_cache_release(page); @@ -42,41 +42,43 @@ static long mincore_vma(struct vm_area_s unsigned long start, unsigned long end, unsigned char __user * vec) { long error, i, remaining; - unsigned char * tmp; + unsigned char *kaddr; + struct page *page; error = -ENOMEM; if (!vma->vm_file) return error; - start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; + start = ((start - vma->vm_start) >> MMUPAGE_SHIFT) + vma->vm_pgoff; if (end > vma->vm_end) end = vma->vm_end; - end = ((end - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; + end = ((end - vma->vm_start) >> MMUPAGE_SHIFT) + vma->vm_pgoff; error = -EAGAIN; - tmp = (unsigned char *) __get_free_page(GFP_KERNEL); - if (!tmp) + page = alloc_page(GFP_HIGHUSER); + if (!page) return error; /* (end - start) is # of pages, and also # of bytes in "vec */ - remaining = (end - start), + remaining = end - start; error = 0; + kaddr = kmap_atomic(page, KM_USER0); for (i = 0; remaining > 0; remaining -= PAGE_SIZE, i++) { int j = 0; long thispiece = (remaining < PAGE_SIZE) ? remaining : PAGE_SIZE; while (j < thispiece) - tmp[j++] = mincore_page(vma, start++); + kaddr[j++] = mincore_page(vma, start++); - if (copy_to_user(vec + PAGE_SIZE * i, tmp, thispiece)) { + if (copy_to_user(vec + PAGE_SIZE * i, kaddr, thispiece)) { error = -EFAULT; break; } } - - free_page((unsigned long) tmp); + kunmap_atomic(kaddr, KM_USER0); + __free_page(page); return error; } @@ -116,15 +118,15 @@ asmlinkage long sys_mincore(unsigned lon down_read(¤t->mm->mmap_sem); - if (start & ~PAGE_CACHE_MASK) + if (start & ~MMUPAGE_MASK) goto out; - len = (len + ~PAGE_CACHE_MASK) & PAGE_CACHE_MASK; + len = (len + ~MMUPAGE_MASK) & MMUPAGE_MASK; end = start + len; if (end < start) goto out; error = -EFAULT; - if (!access_ok(VERIFY_WRITE, (unsigned long) vec, len >> PAGE_SHIFT)) + if (!access_ok(VERIFY_WRITE, (unsigned long) vec, len >> MMUPAGE_SHIFT)) goto out; error = 0; @@ -164,7 +166,7 @@ asmlinkage long sys_mincore(unsigned lon error = mincore_vma(vma, start, vma->vm_end, &vec[index]); if (error) goto out; - index += (vma->vm_end - start) >> PAGE_CACHE_SHIFT; + index += (vma->vm_end - start)/MMUPAGE_SIZE; start = vma->vm_end; vma = vma->vm_next; } diff -prauN linux-2.5.70-bk4/mm/mlock.c pgcl-2.5.70-bk4-1/mm/mlock.c --- linux-2.5.70-bk4/mm/mlock.c 2003-05-26 18:00:21.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/mlock.c 2003-05-30 21:03:24.000000000 -0700 @@ -37,7 +37,7 @@ static int mlock_fixup(struct vm_area_st /* * Keep track of amount of locked VM. */ - pages = (end - start) >> PAGE_SHIFT; + pages = (end - start) >> MMUPAGE_SHIFT; if (newflags & VM_LOCKED) { pages = -pages; make_pages_present(start, end); @@ -55,7 +55,7 @@ static int do_mlock(unsigned long start, if (on && !capable(CAP_IPC_LOCK)) return -EPERM; - len = PAGE_ALIGN(len); + len = MMUPAGE_ALIGN(len); end = start + len; if (end < start) return -EINVAL; @@ -101,14 +101,14 @@ asmlinkage long sys_mlock(unsigned long int error = -ENOMEM; down_write(¤t->mm->mmap_sem); - len = PAGE_ALIGN(len + (start & ~PAGE_MASK)); - start &= PAGE_MASK; + len = MMUPAGE_ALIGN(len + (start & ~MMUPAGE_MASK)); + start &= MMUPAGE_MASK; - locked = len >> PAGE_SHIFT; + locked = len >> MMUPAGE_SHIFT; locked += current->mm->locked_vm; lock_limit = current->rlim[RLIMIT_MEMLOCK].rlim_cur; - lock_limit >>= PAGE_SHIFT; + lock_limit >>= MMUPAGE_SHIFT; /* check against resource limits */ if (locked <= lock_limit) @@ -122,8 +122,8 @@ asmlinkage long sys_munlock(unsigned lon int ret; down_write(¤t->mm->mmap_sem); - len = PAGE_ALIGN(len + (start & ~PAGE_MASK)); - start &= PAGE_MASK; + len = MMUPAGE_ALIGN(len + (start & ~MMUPAGE_MASK)); + start &= MMUPAGE_MASK; ret = do_mlock(start, len, 0); up_write(¤t->mm->mmap_sem); return ret; @@ -167,7 +167,7 @@ asmlinkage long sys_mlockall(int flags) goto out; lock_limit = current->rlim[RLIMIT_MEMLOCK].rlim_cur; - lock_limit >>= PAGE_SHIFT; + lock_limit >>= MMUPAGE_SHIFT; ret = -ENOMEM; if (current->mm->total_vm <= lock_limit) diff -prauN linux-2.5.70-bk4/mm/mmap.c pgcl-2.5.70-bk4-1/mm/mmap.c --- linux-2.5.70-bk4/mm/mmap.c 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/mmap.c 2003-05-30 21:03:24.000000000 -0700 @@ -56,7 +56,8 @@ atomic_t vm_committed_space = ATOMIC_INI /* * Check that a process has enough memory to allocate a new virtual * mapping. 1 means there is enough memory for the allocation to - * succeed and 0 implies there is not. + * succeed and 0 implies there is not. the "pages" argument is in + * mmupages. * * We currently support three overcommit policies, which are set via the * vm.overcommit_memory sysctl. See Documentation/vm/overcommit-acounting @@ -78,7 +79,7 @@ int vm_enough_memory(long pages) return 1; if (sysctl_overcommit_memory == 0) { - free = get_page_cache_size(); + free = get_page_cache_size(); free += nr_free_pages(); free += nr_swap_pages; @@ -89,6 +90,7 @@ int vm_enough_memory(long pages) * cache and most inode caches should fall into this */ free += atomic_read(&slab_reclaim_pages); + free *= PAGE_MMUCOUNT; /* * Leave the last 3% for root @@ -102,14 +104,14 @@ int vm_enough_memory(long pages) return 0; } - allowed = totalram_pages * sysctl_overcommit_ratio / 100; + allowed = totalram_pages*(sysctl_overcommit_ratio/100); allowed += total_swap_pages; + allowed *= PAGE_MMUCOUNT; if (atomic_read(&vm_committed_space) < allowed) return 1; vm_unacct_memory(pages); - return 0; } @@ -159,8 +161,8 @@ asmlinkage unsigned long sys_brk(unsigne if (brk < mm->end_code) goto out; - newbrk = PAGE_ALIGN(brk); - oldbrk = PAGE_ALIGN(mm->brk); + newbrk = MMUPAGE_ALIGN(brk); + oldbrk = MMUPAGE_ALIGN(mm->brk); if (oldbrk == newbrk) goto set_brk; @@ -177,7 +179,7 @@ asmlinkage unsigned long sys_brk(unsigne goto out; /* Check against existing mmap mappings. */ - if (find_vma_intersection(mm, oldbrk, newbrk+PAGE_SIZE)) + if (find_vma_intersection(mm, oldbrk, newbrk+MMUPAGE_SIZE)) goto out; /* Ok, looks good - let it rip. */ @@ -534,10 +536,10 @@ unsigned long do_mmap_pgoff(struct file if (len > TASK_SIZE) return -EINVAL; - len = PAGE_ALIGN(len); + len = MMUPAGE_ALIGN(len); /* offset overflow? */ - if ((pgoff + (len >> PAGE_SHIFT)) < pgoff) + if ((pgoff + (len >> MMUPAGE_SHIFT)) < pgoff) return -EINVAL; /* Too many mappings? */ @@ -548,7 +550,7 @@ unsigned long do_mmap_pgoff(struct file * that it represents a valid section of the address space. */ addr = get_unmapped_area(file, addr, len, pgoff, flags); - if (addr & ~PAGE_MASK) + if (addr & ~MMUPAGE_MASK) return addr; /* Do simple checking here so the lower-level routines won't have @@ -565,7 +567,7 @@ unsigned long do_mmap_pgoff(struct file } /* mlock MCL_FUTURE? */ if (vm_flags & VM_LOCKED) { - unsigned long locked = mm->locked_vm << PAGE_SHIFT; + unsigned long locked = mm->locked_vm << MMUPAGE_SHIFT; locked += len; if (locked > current->rlim[RLIMIT_MEMLOCK].rlim_cur) return -EAGAIN; @@ -633,7 +635,7 @@ munmap_back: } /* Check against address space limit. */ - if ((mm->total_vm << PAGE_SHIFT) + len + if ((mm->total_vm << MMUPAGE_SHIFT) + len > current->rlim[RLIMIT_AS].rlim_cur) return -ENOMEM; @@ -645,7 +647,7 @@ munmap_back: /* * Private writable mapping: check memory availability */ - charged = len >> PAGE_SHIFT; + charged = len >> MMUPAGE_SHIFT; if (!vm_enough_memory(charged)) return -ENOMEM; vm_flags |= VM_ACCOUNT; @@ -729,9 +731,9 @@ munmap_back: kmem_cache_free(vm_area_cachep, vma); } out: - mm->total_vm += len >> PAGE_SHIFT; + mm->total_vm += len >> MMUPAGE_SHIFT; if (vm_flags & VM_LOCKED) { - mm->locked_vm += len >> PAGE_SHIFT; + mm->locked_vm += len >> MMUPAGE_SHIFT; make_pages_present(addr, addr + len); } if (flags & MAP_POPULATE) { @@ -764,7 +766,7 @@ unacct_error: * Ugly calling convention alert: * Return value with the low bits set means error value, * ie - * if (ret & ~PAGE_MASK) + * if (ret & ~MMUPAGE_MASK) * error = ret; * * This function "knows" that -ENOMEM has the bits set. @@ -782,7 +784,7 @@ arch_get_unmapped_area(struct file *filp return -ENOMEM; if (addr) { - addr = PAGE_ALIGN(addr); + addr = MMUPAGE_ALIGN(addr); vma = find_vma(mm, addr); if (TASK_SIZE - len >= addr && (!vma || addr + len <= vma->vm_start)) @@ -821,7 +823,7 @@ get_unmapped_area(struct file *file, uns if (addr > TASK_SIZE - len) return -ENOMEM; - if (addr & ~PAGE_MASK) + if (addr & ~MMUPAGE_MASK) return -EINVAL; if (file && is_file_hugepages(file)) { /* @@ -927,18 +929,20 @@ int expand_stack(struct vm_area_struct * { unsigned long grow; - if (!(vma->vm_flags & VM_GROWSUP)) + if (!(vma->vm_flags & VM_GROWSUP)) { + printk("bad vma flags in expand_stack()\n"); return -EFAULT; + } /* * vma->vm_start/vm_end cannot change under us because the caller * is required to hold the mmap_sem in read mode. We need to get * the spinlock only before relocating the vma range ourself. */ - address += 4 + PAGE_SIZE - 1; - address &= PAGE_MASK; + address += 4 + MMUPAGE_SIZE - 1; + address &= MMUPAGE_MASK; spin_lock(&vma->vm_mm->page_table_lock); - grow = (address - vma->vm_end) >> PAGE_SHIFT; + grow = (address - vma->vm_end) >> MMUPAGE_SHIFT; /* Overcommit.. */ if (!vm_enough_memory(grow)) { @@ -947,7 +951,7 @@ int expand_stack(struct vm_area_struct * } if (address - vma->vm_start > current->rlim[RLIMIT_STACK].rlim_cur || - ((vma->vm_mm->total_vm + grow) << PAGE_SHIFT) > + ((vma->vm_mm->total_vm + grow) << MMUPAGE_SHIFT) > current->rlim[RLIMIT_AS].rlim_cur) { spin_unlock(&vma->vm_mm->page_table_lock); vm_unacct_memory(grow); @@ -966,7 +970,7 @@ find_extend_vma(struct mm_struct *mm, un { struct vm_area_struct *vma, *prev; - addr &= PAGE_MASK; + addr &= MMUPAGE_MASK; vma = find_vma_prev(mm, addr, &prev); if (vma && (vma->vm_start <= addr)) return vma; @@ -990,9 +994,9 @@ int expand_stack(struct vm_area_struct * * is required to hold the mmap_sem in read mode. We need to get * the spinlock only before relocating the vma range ourself. */ - address &= PAGE_MASK; + address &= MMUPAGE_MASK; spin_lock(&vma->vm_mm->page_table_lock); - grow = (vma->vm_start - address) >> PAGE_SHIFT; + grow = (vma->vm_start - address) >> MMUPAGE_SHIFT; /* Overcommit.. */ if (!vm_enough_memory(grow)) { @@ -1001,7 +1005,7 @@ int expand_stack(struct vm_area_struct * } if (vma->vm_end - address > current->rlim[RLIMIT_STACK].rlim_cur || - ((vma->vm_mm->total_vm + grow) << PAGE_SHIFT) > + ((vma->vm_mm->total_vm + grow) << MMUPAGE_SHIFT) > current->rlim[RLIMIT_AS].rlim_cur) { spin_unlock(&vma->vm_mm->page_table_lock); vm_unacct_memory(grow); @@ -1022,7 +1026,7 @@ find_extend_vma(struct mm_struct * mm, u struct vm_area_struct * vma; unsigned long start; - addr &= PAGE_MASK; + addr &= MMUPAGE_MASK; vma = find_vma(mm,addr); if (!vma) return NULL; @@ -1114,9 +1118,9 @@ static void unmap_vma(struct mm_struct * { size_t len = area->vm_end - area->vm_start; - area->vm_mm->total_vm -= len >> PAGE_SHIFT; + area->vm_mm->total_vm -= len >> MMUPAGE_SHIFT; if (area->vm_flags & VM_LOCKED) - area->vm_mm->locked_vm -= len >> PAGE_SHIFT; + area->vm_mm->locked_vm -= len >> MMUPAGE_SHIFT; /* * Is this a new hole at the lowest possible address? */ @@ -1221,11 +1225,11 @@ int split_vma(struct mm_struct * mm, str if (new_below) { new->vm_end = addr; vma->vm_start = addr; - vma->vm_pgoff += ((addr - new->vm_start) >> PAGE_SHIFT); + vma->vm_pgoff += ((addr - new->vm_start) >> MMUPAGE_SHIFT); } else { vma->vm_end = addr; new->vm_start = addr; - new->vm_pgoff += ((addr - vma->vm_start) >> PAGE_SHIFT); + new->vm_pgoff += ((addr - vma->vm_start) >> MMUPAGE_SHIFT); } if (new->vm_file) @@ -1248,10 +1252,10 @@ int do_munmap(struct mm_struct *mm, unsi unsigned long end; struct vm_area_struct *mpnt, *prev, *last; - if ((start & ~PAGE_MASK) || start > TASK_SIZE || len > TASK_SIZE-start) + if ((start & ~MMUPAGE_MASK) || start > TASK_SIZE || len > TASK_SIZE-start) return -EINVAL; - if ((len = PAGE_ALIGN(len)) == 0) + if ((len = MMUPAGE_ALIGN(len)) == 0) return -EINVAL; /* Find the first overlapping VMA */ @@ -1334,7 +1338,7 @@ unsigned long do_brk(unsigned long addr, unsigned long flags; struct rb_node ** rb_link, * rb_parent; - len = PAGE_ALIGN(len); + len = MMUPAGE_ALIGN(len); if (!len) return addr; @@ -1342,7 +1346,7 @@ unsigned long do_brk(unsigned long addr, * mlock MCL_FUTURE? */ if (mm->def_flags & VM_LOCKED) { - unsigned long locked = mm->locked_vm << PAGE_SHIFT; + unsigned long locked = mm->locked_vm << MMUPAGE_SHIFT; locked += len; if (locked > current->rlim[RLIMIT_MEMLOCK].rlim_cur) return -EAGAIN; @@ -1360,14 +1364,14 @@ unsigned long do_brk(unsigned long addr, } /* Check against address space limits *after* clearing old maps... */ - if ((mm->total_vm << PAGE_SHIFT) + len + if ((mm->total_vm << MMUPAGE_SHIFT) + len > current->rlim[RLIMIT_AS].rlim_cur) return -ENOMEM; if (mm->map_count > MAX_MAP_COUNT) return -ENOMEM; - if (!vm_enough_memory(len >> PAGE_SHIFT)) + if (!vm_enough_memory(len >> MMUPAGE_SHIFT)) return -ENOMEM; flags = VM_DATA_DEFAULT_FLAGS | VM_ACCOUNT | mm->def_flags; @@ -1382,7 +1386,7 @@ unsigned long do_brk(unsigned long addr, */ vma = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL); if (!vma) { - vm_unacct_memory(len >> PAGE_SHIFT); + vm_unacct_memory(len >> MMUPAGE_SHIFT); return -ENOMEM; } @@ -1400,9 +1404,9 @@ unsigned long do_brk(unsigned long addr, vma_link(mm, vma, prev, rb_link, rb_parent); out: - mm->total_vm += len >> PAGE_SHIFT; + mm->total_vm += len >> MMUPAGE_SHIFT; if (flags & VM_LOCKED) { - mm->locked_vm += len >> PAGE_SHIFT; + mm->locked_vm += len >> MMUPAGE_SHIFT; make_pages_present(addr, addr + len); } return addr; diff -prauN linux-2.5.70-bk4/mm/mprotect.c pgcl-2.5.70-bk4-1/mm/mprotect.c --- linux-2.5.70-bk4/mm/mprotect.c 2003-05-26 18:00:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/mprotect.c 2003-05-30 21:03:24.000000000 -0700 @@ -53,7 +53,7 @@ change_pte_range(pmd_t *pmd, unsigned lo entry = ptep_get_and_clear(pte); set_pte(pte, pte_modify(entry, newprot)); } - address += PAGE_SIZE; + address += MMUPAGE_SIZE; pte++; } while (address && (address < end)); pte_unmap(pte - 1); @@ -174,9 +174,11 @@ mprotect_fixup(struct vm_area_struct *vm */ if (newflags & VM_WRITE) { if (!(vma->vm_flags & (VM_ACCOUNT|VM_WRITE|VM_SHARED))) { - charged = (end - start) >> PAGE_SHIFT; - if (!vm_enough_memory(charged)) + charged = (end - start) >> MMUPAGE_SHIFT; + if (!vm_enough_memory(charged)) { + printk("mprotect_fixup(): OOM\n"); return -ENOMEM; + } newflags |= VM_ACCOUNT; } } @@ -228,9 +230,9 @@ sys_mprotect(unsigned long start, size_t struct vm_area_struct * vma, * next, * prev; int error = -EINVAL; - if (start & ~PAGE_MASK) + if (start & ~MMUPAGE_MASK) return -EINVAL; - len = PAGE_ALIGN(len); + len = MMUPAGE_ALIGN(len); end = start + len; if (end < start) return -EINVAL; diff -prauN linux-2.5.70-bk4/mm/mremap.c pgcl-2.5.70-bk4-1/mm/mremap.c --- linux-2.5.70-bk4/mm/mremap.c 2003-05-26 18:00:40.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/mremap.c 2003-05-30 21:03:24.000000000 -0700 @@ -162,7 +162,7 @@ static int move_page_tables(struct vm_ar * only a few pages.. This also makes error recovery easier. */ while (offset) { - offset -= PAGE_SIZE; + offset -= MMUPAGE_SIZE; if (move_one_page(vma, old_addr + offset, new_addr + offset)) goto oops_we_failed; } @@ -177,7 +177,7 @@ static int move_page_tables(struct vm_ar */ oops_we_failed: flush_cache_range(vma, new_addr, new_addr + len); - while ((offset += PAGE_SIZE) < len) + while ((offset += MMUPAGE_SIZE) < len) move_one_page(vma, new_addr + offset, old_addr + offset); zap_page_range(vma, new_addr, len); return -1; @@ -253,7 +253,7 @@ static unsigned long move_vma(struct vm_ INIT_LIST_HEAD(&new_vma->shared); new_vma->vm_start = new_addr; new_vma->vm_end = new_addr+new_len; - new_vma->vm_pgoff += (addr-vma->vm_start) >> PAGE_SHIFT; + new_vma->vm_pgoff += (addr - vma->vm_start) >> MMUPAGE_SHIFT; if (new_vma->vm_file) get_file(new_vma->vm_file); if (new_vma->vm_ops && new_vma->vm_ops->open) @@ -287,9 +287,9 @@ static unsigned long move_vma(struct vm_ vma->vm_next->vm_flags |= VM_ACCOUNT; } - current->mm->total_vm += new_len >> PAGE_SHIFT; + current->mm->total_vm += new_len >> MMUPAGE_SHIFT; if (must_fault_in) { - current->mm->locked_vm += new_len >> PAGE_SHIFT; + current->mm->locked_vm += new_len >> MMUPAGE_SHIFT; make_pages_present(fault_in_start, fault_in_end); } return new_addr; @@ -318,15 +318,15 @@ unsigned long do_mremap(unsigned long ad if (flags & ~(MREMAP_FIXED | MREMAP_MAYMOVE)) goto out; - if (addr & ~PAGE_MASK) + if (addr & ~MMUPAGE_MASK) goto out; - old_len = PAGE_ALIGN(old_len); - new_len = PAGE_ALIGN(new_len); + old_len = MMUPAGE_ALIGN(old_len); + new_len = MMUPAGE_ALIGN(new_len); /* new_addr is only valid if MREMAP_FIXED is specified */ if (flags & MREMAP_FIXED) { - if (new_addr & ~PAGE_MASK) + if (new_addr & ~MMUPAGE_MASK) goto out; if (!(flags & MREMAP_MAYMOVE)) goto out; @@ -378,19 +378,19 @@ unsigned long do_mremap(unsigned long ad goto out; } if (vma->vm_flags & VM_LOCKED) { - unsigned long locked = current->mm->locked_vm << PAGE_SHIFT; + unsigned long locked = current->mm->locked_vm << MMUPAGE_SHIFT; locked += new_len - old_len; ret = -EAGAIN; if (locked > current->rlim[RLIMIT_MEMLOCK].rlim_cur) goto out; } ret = -ENOMEM; - if ((current->mm->total_vm << PAGE_SHIFT) + (new_len - old_len) + if ((current->mm->total_vm << MMUPAGE_SHIFT) + (new_len - old_len) > current->rlim[RLIMIT_AS].rlim_cur) goto out; if (vma->vm_flags & VM_ACCOUNT) { - charged = (new_len - old_len) >> PAGE_SHIFT; + charged = (new_len - old_len) >> MMUPAGE_SHIFT; if (!vm_enough_memory(charged)) goto out_nc; } @@ -406,7 +406,7 @@ unsigned long do_mremap(unsigned long ad max_addr = vma->vm_next->vm_start; /* can we just expand the current mapping? */ if (max_addr - addr >= new_len) { - int pages = (new_len - old_len) >> PAGE_SHIFT; + int pages = (new_len - old_len) >> MMUPAGE_SHIFT; spin_lock(&vma->vm_mm->page_table_lock); vma->vm_end = addr + new_len; spin_unlock(&vma->vm_mm->page_table_lock); @@ -435,13 +435,13 @@ unsigned long do_mremap(unsigned long ad new_addr = get_unmapped_area(vma->vm_file, 0, new_len, vma->vm_pgoff, map_flags); ret = new_addr; - if (new_addr & ~PAGE_MASK) + if (new_addr & ~MMUPAGE_MASK) goto out; } ret = move_vma(vma, addr, old_len, new_len, new_addr); } out: - if (ret & ~PAGE_MASK) + if (ret & ~MMUPAGE_MASK) vm_unacct_memory(charged); out_nc: return ret; diff -prauN linux-2.5.70-bk4/mm/msync.c pgcl-2.5.70-bk4-1/mm/msync.c --- linux-2.5.70-bk4/mm/msync.c 2003-05-26 18:00:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/msync.c 2003-05-30 21:03:24.000000000 -0700 @@ -59,7 +59,7 @@ static int filemap_sync_pte_range(pmd_t error = 0; do { error |= filemap_sync_pte(pte, vma, address, flags); - address += PAGE_SIZE; + address += MMUPAGE_SIZE; pte++; } while (address && (address < end)); @@ -174,12 +174,12 @@ asmlinkage long sys_msync(unsigned long down_read(¤t->mm->mmap_sem); if (flags & ~(MS_ASYNC | MS_INVALIDATE | MS_SYNC)) goto out; - if (start & ~PAGE_MASK) + if (start & ~MMUPAGE_MASK) goto out; if ((flags & MS_ASYNC) && (flags & MS_SYNC)) goto out; error = -ENOMEM; - len = (len + ~PAGE_MASK) & PAGE_MASK; + len = (len + ~MMUPAGE_MASK) & MMUPAGE_MASK; end = start + len; if (end < start) goto out; diff -prauN linux-2.5.70-bk4/mm/page-writeback.c pgcl-2.5.70-bk4-1/mm/page-writeback.c --- linux-2.5.70-bk4/mm/page-writeback.c 2003-05-30 21:02:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/page-writeback.c 2003-05-30 21:03:24.000000000 -0700 @@ -361,8 +361,8 @@ static void set_ratelimit(void) ratelimit_pages = total_pages / (num_online_cpus() * 32); if (ratelimit_pages < 16) ratelimit_pages = 16; - if (ratelimit_pages * PAGE_CACHE_SIZE > 4096 * 1024) - ratelimit_pages = (4096 * 1024) / PAGE_CACHE_SIZE; + if (ratelimit_pages * PAGE_CACHE_SIZE > PAGE_SIZE * 1024) + ratelimit_pages = (PAGE_SIZE * 1024) / PAGE_CACHE_SIZE; } static int diff -prauN linux-2.5.70-bk4/mm/page_alloc.c pgcl-2.5.70-bk4-1/mm/page_alloc.c --- linux-2.5.70-bk4/mm/page_alloc.c 2003-05-26 18:00:22.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/page_alloc.c 2003-05-30 21:03:24.000000000 -0700 @@ -57,7 +57,7 @@ static int zone_balance_max[MAX_NR_ZONES */ static int bad_range(struct zone *zone, struct page *page) { - if (page_to_pfn(page) >= zone->zone_start_pfn + zone->spanned_pages) + if (page_to_pfn(page) >= zone->zone_start_pfn + zone->spanned_pages*PAGE_MMUCOUNT) return 1; if (page_to_pfn(page) < zone->zone_start_pfn) return 1; @@ -1175,9 +1175,9 @@ void __init memmap_init_zone(struct page #ifdef WANT_PAGE_VIRTUAL /* The shift won't overflow because ZONE_NORMAL is below 4G. */ if (zone != ZONE_HIGHMEM) - set_page_address(page, __va(start_pfn << PAGE_SHIFT)); + set_page_address(page, __va(start_pfn << MMUPAGE_SHIFT)); #endif - start_pfn++; + start_pfn += PAGE_MMUCOUNT; } } @@ -1196,7 +1196,7 @@ static void __init free_area_init_core(s unsigned long *zones_size, unsigned long *zholes_size) { unsigned long i, j; - const unsigned long zone_required_alignment = 1UL << (MAX_ORDER-1); + const unsigned long zone_required_alignment = 1UL << (MAX_ORDER-PAGE_MMUSHIFT-1); int cpu, nid = pgdat->node_id; struct page *lmem_map = pgdat->node_mem_map; unsigned long zone_start_pfn = pgdat->node_start_pfn; @@ -1255,7 +1255,7 @@ static void __init free_area_init_core(s INIT_LIST_HEAD(&pcp->list); } printk(" %s zone: %lu pages, LIFO batch:%lu\n", - zone_names[j], realsize, batch); + zone_names[j], realsize*PAGE_MMUCOUNT, batch); INIT_LIST_HEAD(&zone->active_list); INIT_LIST_HEAD(&zone->inactive_list); atomic_set(&zone->refill_counter, 0); @@ -1297,7 +1297,7 @@ static void __init free_area_init_core(s memmap_init(lmem_map, size, nid, j, zone_start_pfn); - zone_start_pfn += size; + zone_start_pfn += PAGE_MMUCOUNT*size; lmem_map += size; for (i = 0; ; i++) { @@ -1368,7 +1368,7 @@ struct pglist_data contig_page_data = { void __init free_area_init(unsigned long *zones_size) { free_area_init_node(0, &contig_page_data, NULL, zones_size, - __pa(PAGE_OFFSET) >> PAGE_SHIFT, NULL); + __pa(PAGE_OFFSET) >> MMUPAGE_SHIFT, NULL); mem_map = contig_page_data.node_mem_map; } #endif diff -prauN linux-2.5.70-bk4/mm/page_io.c pgcl-2.5.70-bk4-1/mm/page_io.c --- linux-2.5.70-bk4/mm/page_io.c 2003-05-26 18:00:23.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/page_io.c 2003-05-30 21:03:24.000000000 -0700 @@ -32,7 +32,7 @@ get_swap_bio(int gfp_flags, struct page swp_entry_t entry; BUG_ON(!PageSwapCache(page)); - entry.val = page->index; + entry.val = page->index*PAGE_MMUCOUNT; sis = get_swap_info_struct(swp_type(entry)); bio->bi_sector = map_swap_page(sis, swp_offset(entry)) * @@ -152,7 +152,7 @@ int rw_swap_page_sync(int rw, swp_entry_ BUG_ON(page->mapping); page->mapping = &swapper_space; - page->index = entry.val; + page->index = entry.val/PAGE_MMUCOUNT; if (rw == READ) { ret = swap_readpage(NULL, page); diff -prauN linux-2.5.70-bk4/mm/rmap.c pgcl-2.5.70-bk4-1/mm/rmap.c --- linux-2.5.70-bk4/mm/rmap.c 2003-05-26 18:00:58.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/rmap.c 2003-05-30 21:03:24.000000000 -0700 @@ -337,7 +337,8 @@ static int try_to_unmap_one(struct page * Store the swap location in the pte. * See handle_pte_fault() ... */ - swp_entry_t entry = { .val = page->index }; + swp_entry_t entry = { .val = page->index*PAGE_MMUCOUNT + + (pte_pfn(pte) % PAGE_MMUCOUNT) }; swap_duplicate(entry); set_pte(ptep, swp_entry_to_pte(entry)); BUG_ON(pte_file(*ptep)); @@ -347,11 +348,11 @@ static int try_to_unmap_one(struct page * If a nonlinear mapping then store the file page offset * in the pte. */ - pgidx = (address - vma->vm_start) >> PAGE_SHIFT; + pgidx = (address - vma->vm_start) >> MMUPAGE_SHIFT; pgidx += vma->vm_pgoff; - pgidx >>= PAGE_CACHE_SHIFT - PAGE_SHIFT; + pgidx >>= PAGE_CACHE_SHIFT - MMUPAGE_SHIFT; if (page->index != pgidx) { - set_pte(ptep, pgoff_to_pte(page->index)); + set_pte(ptep, pgoff_to_pte(page->index*PAGE_MMUCOUNT)); BUG_ON(!pte_file(*ptep)); } } diff -prauN linux-2.5.70-bk4/mm/shmem.c pgcl-2.5.70-bk4-1/mm/shmem.c --- linux-2.5.70-bk4/mm/shmem.c 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/shmem.c 2003-05-30 21:03:24.000000000 -0700 @@ -47,7 +47,7 @@ #define SHMEM_MAX_INDEX (SHMEM_NR_DIRECT + (ENTRIES_PER_PAGEPAGE/2) * (ENTRIES_PER_PAGE+1)) #define SHMEM_MAX_BYTES ((unsigned long long)SHMEM_MAX_INDEX << PAGE_CACHE_SHIFT) -#define VM_ACCT(size) (PAGE_CACHE_ALIGN(size) >> PAGE_SHIFT) +#define VM_ACCT(size) (MMUPAGE_ALIGN(size)/MMUPAGE_SIZE) /* Pretend that each entry is of this size in directory's i_size */ #define BOGO_DIRENT_SIZE 20 @@ -71,14 +71,14 @@ static inline struct page *shmem_dir_all /* * The above definition of ENTRIES_PER_PAGE, and the use of * BLOCKS_PER_PAGE on indirect pages, assume PAGE_CACHE_SIZE: - * might be reconsidered if it ever diverges from PAGE_SIZE. + * might be reconsidered if it ever diverges from MMUPAGE_SIZE. */ - return alloc_pages(gfp_mask, PAGE_CACHE_SHIFT-PAGE_SHIFT); + return alloc_pages(gfp_mask, PAGE_CACHE_MMUSHIFT); } static inline void shmem_dir_free(struct page *page) { - __free_pages(page, PAGE_CACHE_SHIFT-PAGE_SHIFT); + __free_pages(page, PAGE_CACHE_MMUSHIFT); } static struct page **shmem_dir_map(struct page *page) @@ -297,7 +297,7 @@ static swp_entry_t *shmem_swp_alloc(stru static const swp_entry_t unswapped = {0}; if (sgp != SGP_WRITE && - ((loff_t) index << PAGE_CACHE_SHIFT) >= inode->i_size) + (loff_t)index*PAGE_CACHE_SIZE >= inode->i_size) return ERR_PTR(-EINVAL); while (!(entry = shmem_swp_entry(info, index, &page))) { @@ -330,7 +330,7 @@ static swp_entry_t *shmem_swp_alloc(stru return ERR_PTR(-ENOMEM); } if (sgp != SGP_WRITE && - ((loff_t) index << PAGE_CACHE_SHIFT) >= inode->i_size) { + (loff_t)index*PAGE_CACHE_SIZE >= inode->i_size) { entry = ERR_PTR(-EINVAL); break; } @@ -383,7 +383,7 @@ static void shmem_truncate(struct inode int freed; inode->i_ctime = inode->i_mtime = CURRENT_TIME; - idx = (inode->i_size + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; + idx = (inode->i_size + PAGE_CACHE_SIZE - 1)/PAGE_CACHE_SIZE; if (idx >= info->next_index) return; @@ -509,7 +509,7 @@ static int shmem_notify_change(struct de long change = 0; int error; - if ((attr->ia_valid & ATTR_SIZE) && (attr->ia_size <= SHMEM_MAX_BYTES)) { + if ((attr->ia_valid & ATTR_SIZE) && attr->ia_size <= SHMEM_MAX_BYTES) { /* * Account swap file usage based on new file size, * but just let vmtruncate fail on out-of-range sizes. @@ -527,9 +527,9 @@ static int shmem_notify_change(struct de * truncate_partial_page cannnot miss it were * it assigned to swap. */ - if (attr->ia_size & (PAGE_CACHE_SIZE-1)) { + if (attr->ia_size % PAGE_CACHE_SIZE) { (void) shmem_getpage(inode, - attr->ia_size>>PAGE_CACHE_SHIFT, + attr->ia_size/PAGE_CACHE_SIZE, &page, SGP_READ); } } @@ -940,16 +940,14 @@ struct page *shmem_nopage(struct vm_area { struct inode *inode = vma->vm_file->f_dentry->d_inode; struct page *page = NULL; - unsigned long idx; + unsigned long pgoff; int error; - idx = (address - vma->vm_start) >> PAGE_SHIFT; - idx += vma->vm_pgoff; - idx >>= PAGE_CACHE_SHIFT - PAGE_SHIFT; + pgoff = (address - vma->vm_start)/MMUPAGE_SIZE + vma->vm_pgoff; - error = shmem_getpage(inode, idx, &page, SGP_CACHE); + error = shmem_getpage(inode, pgoff/PAGE_CACHE_MMUCOUNT, &page, SGP_CACHE); if (error) - return (error == -ENOMEM)? NOPAGE_OOM: NOPAGE_SIGBUS; + return error == -ENOMEM ? NOPAGE_OOM : NOPAGE_SIGBUS; mark_page_accessed(page); return page; @@ -964,8 +962,8 @@ static int shmem_populate(struct vm_area enum sgp_type sgp = nonblock? SGP_QUICK: SGP_CACHE; unsigned long size; - size = (inode->i_size + PAGE_SIZE - 1) >> PAGE_SHIFT; - if (pgoff >= size || pgoff + (len >> PAGE_SHIFT) > size) + size = (inode->i_size + MMUPAGE_SIZE - 1)/MMUPAGE_SIZE; + if (pgoff >= size || pgoff + len/MMUPAGE_SIZE > size) return -EINVAL; while ((long) len > 0) { @@ -974,19 +972,19 @@ static int shmem_populate(struct vm_area /* * Will need changing if PAGE_CACHE_SIZE != PAGE_SIZE */ - err = shmem_getpage(inode, pgoff, &page, sgp); + err = shmem_getpage(inode, pgoff/PAGE_CACHE_MMUCOUNT, &page, sgp); if (err) return err; if (page) { mark_page_accessed(page); - err = install_page(mm, vma, addr, page, prot); + err = install_page(mm, vma, addr, page, prot, pgoff % PAGE_CACHE_MMUCOUNT); if (err) { page_cache_release(page); return err; } } - len -= PAGE_SIZE; - addr += PAGE_SIZE; + len -= MMUPAGE_SIZE; + addr += MMUPAGE_SIZE; pgoff++; } return 0; @@ -1155,8 +1153,8 @@ shmem_file_write(struct file *file, cons char *kaddr; int left; - offset = (pos & (PAGE_CACHE_SIZE -1)); /* Within page */ - index = pos >> PAGE_CACHE_SHIFT; + offset = pos % PAGE_CACHE_SIZE; /* Within page */ + index = pos/PAGE_CACHE_SIZE; bytes = PAGE_CACHE_SIZE - offset; if (bytes > count) bytes = count; @@ -1231,18 +1229,18 @@ static void do_shmem_file_read(struct fi struct address_space *mapping = inode->i_mapping; unsigned long index, offset; - index = *ppos >> PAGE_CACHE_SHIFT; - offset = *ppos & ~PAGE_CACHE_MASK; + index = *ppos/PAGE_CACHE_SIZE; + offset = *ppos % PAGE_CACHE_SIZE; for (;;) { struct page *page = NULL; unsigned long end_index, nr, ret; - end_index = inode->i_size >> PAGE_CACHE_SHIFT; + end_index = inode->i_size/PAGE_CACHE_SIZE; if (index > end_index) break; if (index == end_index) { - nr = inode->i_size & ~PAGE_CACHE_MASK; + nr = inode->i_size % PAGE_CACHE_SIZE; if (nr <= offset) break; } @@ -1259,9 +1257,9 @@ static void do_shmem_file_read(struct fi * are called without i_sem protection against truncate */ nr = PAGE_CACHE_SIZE; - end_index = inode->i_size >> PAGE_CACHE_SHIFT; + end_index = inode->i_size/PAGE_CACHE_SIZE; if (index == end_index) { - nr = inode->i_size & ~PAGE_CACHE_MASK; + nr = inode->i_size % PAGE_CACHE_SIZE; if (nr <= offset) { page_cache_release(page); break; diff -prauN linux-2.5.70-bk4/mm/slab.c pgcl-2.5.70-bk4-1/mm/slab.c --- linux-2.5.70-bk4/mm/slab.c 2003-05-26 18:00:46.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/slab.c 2003-05-30 21:03:24.000000000 -0700 @@ -591,7 +591,7 @@ void __init kmem_cache_init(void) * Fragmentation resistance on low memory - only use bigger * page orders on machines with more than 32MB of memory. */ - if (num_physpages > (32 << 20) >> PAGE_SHIFT) + if (num_physpages > (32 << 20) >> MMUPAGE_SHIFT) slab_break_gfp_order = BREAK_GFP_ORDER_HI; @@ -1013,7 +1013,7 @@ kmem_cache_create (const char *name, siz align = L1_CACHE_BYTES; /* Determine if the slab management is 'on' or 'off' slab. */ - if (size >= (PAGE_SIZE>>3)) + if (size >= PAGE_SIZE/8 || ((flags & SLAB_MUST_HWCACHE_ALIGN) && size >= MMUPAGE_SIZE)) /* * Size is large, assume best to place the slab management obj * off-slab (should allow better packing of objs). diff -prauN linux-2.5.70-bk4/mm/swap.c pgcl-2.5.70-bk4-1/mm/swap.c --- linux-2.5.70-bk4/mm/swap.c 2003-05-26 18:00:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/swap.c 2003-05-30 21:03:24.000000000 -0700 @@ -379,7 +379,7 @@ void vm_acct_memory(long pages) */ void __init swap_setup(void) { - unsigned long megs = num_physpages >> (20 - PAGE_SHIFT); + unsigned long megs = num_physpages >> (20 - MMUPAGE_SHIFT); /* Use a smaller cluster for small-memory machines */ if (megs < 16) diff -prauN linux-2.5.70-bk4/mm/swap_state.c pgcl-2.5.70-bk4-1/mm/swap_state.c --- linux-2.5.70-bk4/mm/swap_state.c 2003-05-26 18:00:39.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/swap_state.c 2003-05-30 21:03:24.000000000 -0700 @@ -77,7 +77,7 @@ static int add_to_swap_cache(struct page INC_CACHE_INFO(noent_race); return -ENOENT; } - error = add_to_page_cache(page, &swapper_space, entry.val, GFP_KERNEL); + error = add_to_page_cache(page, &swapper_space, entry.val/PAGE_MMUCOUNT, GFP_KERNEL); /* * Anon pages are already on the LRU, we don't run lru_cache_add here. */ @@ -149,7 +149,7 @@ int add_to_swap(struct page * page) * Add it to the swap cache and mark it dirty */ err = add_to_page_cache(page, &swapper_space, - entry.val, GFP_ATOMIC); + entry.val/PAGE_MMUCOUNT, GFP_ATOMIC); if (pf_flags & PF_MEMALLOC) current->flags |= PF_MEMALLOC; @@ -188,7 +188,7 @@ void delete_from_swap_cache(struct page BUG_ON(PageWriteback(page)); BUG_ON(PagePrivate(page)); - entry.val = page->index; + entry.val = page->index*PAGE_MMUCOUNT; spin_lock(&swapper_space.page_lock); __delete_from_swap_cache(page); @@ -206,10 +206,10 @@ int move_to_swap_cache(struct page *page spin_lock(&swapper_space.page_lock); spin_lock(&mapping->page_lock); - err = radix_tree_insert(&swapper_space.page_tree, entry.val, page); + err = radix_tree_insert(&swapper_space.page_tree, entry.val/PAGE_MMUCOUNT, page); if (!err) { __remove_from_page_cache(page); - ___add_to_page_cache(page, &swapper_space, entry.val); + ___add_to_page_cache(page, &swapper_space, entry.val/PAGE_MMUCOUNT); } spin_unlock(&mapping->page_lock); @@ -237,7 +237,7 @@ int move_from_swap_cache(struct page *pa BUG_ON(PageWriteback(page)); BUG_ON(PagePrivate(page)); - entry.val = page->index; + entry.val = page->index*PAGE_MMUCOUNT; spin_lock(&swapper_space.page_lock); spin_lock(&mapping->page_lock); @@ -320,7 +320,7 @@ struct page * lookup_swap_cache(swp_entr { struct page *found; - found = find_get_page(&swapper_space, entry.val); + found = find_get_page(&swapper_space, entry.val/PAGE_MMUCOUNT); /* * Unsafe to assert PageSwapCache and mapping on page found: * if SMP nothing prevents swapoff from deleting this page from @@ -351,7 +351,7 @@ struct page * read_swap_cache_async(swp_ * that would confuse statistics: use find_get_page() * directly. */ - found_page = find_get_page(&swapper_space, entry.val); + found_page = find_get_page(&swapper_space, entry.val/PAGE_MMUCOUNT); if (found_page) break; diff -prauN linux-2.5.70-bk4/mm/swapfile.c pgcl-2.5.70-bk4-1/mm/swapfile.c --- linux-2.5.70-bk4/mm/swapfile.c 2003-05-26 18:00:25.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/swapfile.c 2003-05-30 21:03:24.000000000 -0700 @@ -41,6 +41,10 @@ struct swap_info_struct swap_info[MAX_SW #define SWAPFILE_CLUSTER 256 +/* + * returns offset into ->swap_map[] array, each entry of which + * tracks PAGE_SIZE (not MMUPAGE_SIZE) + */ static inline int scan_swap_map(struct swap_info_struct *si) { unsigned long offset; @@ -127,7 +131,7 @@ swp_entry_t get_swap_page(void) offset = scan_swap_map(p); swap_device_unlock(p); if (offset) { - entry = swp_entry(type,offset); + entry = swp_entry(type, offset*PAGE_MMUCOUNT); type = swap_info[type].next; if (type < 0 || p->prio != swap_info[type].prio) { @@ -161,15 +165,23 @@ static struct swap_info_struct * swap_in if (!entry.val) goto out; type = swp_type(entry); - if (type >= nr_swapfiles) + if (type >= nr_swapfiles) { + printk(KERN_ERR "bad type %lu beyond nr_swapfiles %u " + "in swap_info_get()\n", type, nr_swapfiles); goto bad_nofile; + } p = & swap_info[type]; if (!(p->flags & SWP_USED)) goto bad_device; offset = swp_offset(entry); - if (offset >= p->max) + + /* + * offset returned by swp_offset() is in MMUPAGE_SIZE units, + * p->max is in PAGE_SIZE units + */ + if (offset >= p->max*PAGE_MMUCOUNT) goto bad_offset; - if (!p->swap_map[offset]) + if (!p->swap_map[offset/PAGE_MMUCOUNT]) goto bad_free; swap_list_lock(); if (p->prio > swap_info[swap_list.next].prio) @@ -179,15 +191,70 @@ static struct swap_info_struct * swap_in bad_free: printk(KERN_ERR "swap_free: %s%08lx\n", Unused_offset, entry.val); + WARN_ON(1); goto out; bad_offset: printk(KERN_ERR "swap_free: %s%08lx\n", Bad_offset, entry.val); + WARN_ON(1); goto out; bad_device: printk(KERN_ERR "swap_free: %s%08lx\n", Unused_file, entry.val); + WARN_ON(1); goto out; bad_nofile: printk(KERN_ERR "swap_free: %s%08lx\n", Bad_file, entry.val); + WARN_ON(1); + +/* dump pagetables */ +#if 1 + { + struct mm_struct *mm = current->mm; + pgd_t *pgd; + pmd_t *pmd; + pte_t *pte; + unsigned long vaddr; + + if (!mm) { + /* we're dead here anyway, but... */ + printk(KERN_ERR "bug in free_swap_and_cache() " + "with no mm!\n"); + goto out_noscan; + } + + for (vaddr = 0; vaddr < TASK_SIZE; vaddr += PGDIR_SIZE) { + pgd = pgd_offset(mm, vaddr); + printk(KERN_DEBUG "pgd for 0x%lx = 0x%Lx\n", + vaddr, (u64)pgd_val(*pgd)); + } + + if (PTRS_PER_PMD > 1) { + for (vaddr = 0; vaddr < TASK_SIZE; vaddr += PMD_SIZE) { + pgd = pgd_offset(mm, vaddr); + if (pgd_none(*pgd) || !pgd_present(*pgd)) + continue; + pmd = pmd_offset(pgd, vaddr); + printk(KERN_DEBUG "pmd for 0x%lx = 0x%Lx\n", + vaddr, (u64)pmd_val(*pmd)); + } + } + + for (vaddr = 0; vaddr < TASK_SIZE; vaddr += MMUPAGE_SIZE) { + pgd = pgd_offset(mm, vaddr); + if (pgd_none(*pgd) || !pgd_present(*pgd)) + continue; + pmd = pmd_offset(pgd, vaddr); + if (pmd_none(*pmd) || !pmd_present(*pmd)) + continue; + pte = pte_offset_map_nested(pmd, vaddr); + if (!pte_none(*pte) && pte_present(*pte)) + printk(KERN_DEBUG "pte for 0x%lx = 0x%Lx\n", + vaddr, (u64)pte_val(*pte)); + pte_unmap_nested(pte); + } +out_noscan: + ; + } +#endif out: return NULL; } @@ -198,6 +265,9 @@ static void swap_info_put(struct swap_in swap_list_unlock(); } +/* + * offset is entry.val/PAGE_MMUCOUNT + */ static int swap_entry_free(struct swap_info_struct *p, unsigned long offset) { int count = p->swap_map[offset]; @@ -226,11 +296,12 @@ void swap_free(swp_entry_t entry) struct swap_info_struct * p; p = swap_info_get(entry); - if (p) { - swap_entry_free(p, swp_offset(entry)); + if (!p) + return; + + swap_entry_free(p, swp_offset(entry)/PAGE_MMUCOUNT); swap_info_put(p); } -} /* * Check if we're the only user of a swap page, @@ -242,11 +313,11 @@ static int exclusive_swap_page(struct pa struct swap_info_struct * p; swp_entry_t entry; - entry.val = page->index; + entry.val = page->index*PAGE_MMUCOUNT; p = swap_info_get(entry); if (p) { /* Is the only swap cache user the cache itself? */ - if (p->swap_map[swp_offset(entry)] == 1) { + if (p->swap_map[swp_offset(entry)/PAGE_MMUCOUNT] == 1) { /* Recheck the page count with the pagecache lock held.. */ spin_lock(&swapper_space.page_lock); if (page_count(page) - !!PagePrivate(page) == 2) @@ -310,14 +381,14 @@ int remove_exclusive_swap_page(struct pa if (page_count(page) != 2) /* 2: us + cache */ return 0; - entry.val = page->index; + entry.val = page->index*PAGE_MMUCOUNT; p = swap_info_get(entry); if (!p) return 0; /* Is the only swap cache user the cache itself? */ retval = 0; - if (p->swap_map[swp_offset(entry)] == 1) { + if (p->swap_map[swp_offset(entry)/PAGE_MMUCOUNT] == 1) { /* Recheck the page count with the pagecache lock held.. */ spin_lock(&swapper_space.page_lock); if ((page_count(page) == 2) && !PageWriteback(page)) { @@ -348,8 +419,8 @@ void free_swap_and_cache(swp_entry_t ent p = swap_info_get(entry); if (p) { - if (swap_entry_free(p, swp_offset(entry)) == 1) - page = find_trylock_page(&swapper_space, entry.val); + if (swap_entry_free(p, swp_offset(entry)/PAGE_MMUCOUNT) == 1) + page = find_trylock_page(&swapper_space, entry.val/PAGE_MMUCOUNT); swap_info_put(p); } if (page) { @@ -382,9 +453,12 @@ static void unuse_pte(struct vm_area_struct *vma, unsigned long address, pte_t *dir, swp_entry_t entry, struct page *page, struct pte_chain **pte_chainp) { + unsigned long pfn; + /* vma_suboffset() would be meaningless; these are anonymous */ + pfn = page_to_pfn(page) + (entry.val % PAGE_MMUCOUNT); vma->vm_mm->rss++; get_page(page); - set_pte(dir, pte_mkold(mk_pte(page, vma->vm_page_prot))); + set_pte(dir, pte_mkold(pfn_pte(pfn, vma->vm_page_prot))); *pte_chainp = page_add_rmap(page, dir, *pte_chainp); swap_free(entry); } @@ -422,7 +496,7 @@ static int unuse_pmd(struct vm_area_stru pte_unmap(pte); return 1; } - address += PAGE_SIZE; + address += MMUPAGE_SIZE; pte++; } while (address && (address < end)); pte_unmap(pte - 1); @@ -572,6 +646,9 @@ static int try_to_unuse(unsigned int typ * child immediately after parent. If we race with dup_mmap(), * we very much want to resolve parent before child, otherwise * we may miss some entries: using last mm would invert that. + * + * The whole of the preceding discussion is bogus now that + * physical scanning is in place. */ start_mm = &init_mm; atomic_inc(&init_mm.mm_users); @@ -601,7 +678,7 @@ static int try_to_unuse(unsigned int typ * page and read the swap into it. */ swap_map = &si->swap_map[i]; - entry = swp_entry(type, i); + entry = swp_entry(type, i*PAGE_MMUCOUNT); page = read_swap_cache_async(entry); if (!page) { /* @@ -710,6 +787,10 @@ static int try_to_unuse(unsigned int typ * we might be resetting SWAP_MAP_MAX too early here. * We know "Undead"s can happen, they're okay, so don't * report them; but do report if we reset SWAP_MAP_MAX. + * + * The whole of the preceding discussion is bogus given + * the new process capacities and there are probably + * resource leaks to fix up here. */ if (*swap_map == SWAP_MAP_MAX) { swap_device_lock(si); @@ -786,6 +867,8 @@ sector_t map_swap_page(struct swap_info_ struct swap_extent *se = sis->curr_swap_extent; struct swap_extent *start_se = se; + offset /= PAGE_MMUCOUNT; + for ( ; ; ) { struct list_head *lh; @@ -992,7 +1075,7 @@ int page_queue_congested(struct page *pa bdi = page->mapping->backing_dev_info; if (PageSwapCache(page)) { - swp_entry_t entry = { .val = page->index }; + swp_entry_t entry = { .val = page->index*PAGE_MMUCOUNT }; struct swap_info_struct *sis; sis = get_swap_info_struct(swp_type(entry)); @@ -1356,20 +1439,20 @@ asmlinkage long sys_swapon(const char __ maxpages = swp_offset(swp_entry(0,~0UL)) - 1; if (maxpages > swap_header->info.last_page) maxpages = swap_header->info.last_page; - p->highest_bit = maxpages - 1; + p->highest_bit = maxpages/PAGE_MMUCOUNT - 1; error = -EINVAL; if (swap_header->info.nr_badpages > MAX_SWAP_BADPAGES) goto bad_swap; /* OK, set up the swap map and apply the bad block list */ - if (!(p->swap_map = vmalloc(maxpages * sizeof(short)))) { + if (!(p->swap_map = vmalloc(maxpages/PAGE_MMUCOUNT * sizeof(short)))) { error = -ENOMEM; goto bad_swap; } error = 0; - memset(p->swap_map, 0, maxpages * sizeof(short)); + memset(p->swap_map, 0, maxpages/PAGE_MMUCOUNT * sizeof(short)); for (i=0; iinfo.nr_badpages; i++) { int page = swap_header->info.badpages[i]; if (page <= 0 || page >= swap_header->info.last_page) @@ -1377,14 +1460,14 @@ asmlinkage long sys_swapon(const char __ else p->swap_map[page] = SWAP_MAP_BAD; } - nr_good_pages = swap_header->info.last_page - + nr_good_pages = (swap_header->info.last_page - swap_header->info.nr_badpages - - 1 /* header page */; + 1)/PAGE_MMUCOUNT /* header page */; if (error) goto bad_swap; } - if (swapfilesize && maxpages > swapfilesize) { + if (swapfilesize && maxpages/PAGE_MMUCOUNT > swapfilesize) { printk(KERN_WARNING "Swap area shorter than signature indicates\n"); error = -EINVAL; @@ -1396,7 +1479,7 @@ asmlinkage long sys_swapon(const char __ goto bad_swap; } p->swap_map[0] = SWAP_MAP_BAD; - p->max = maxpages; + p->max = maxpages/PAGE_MMUCOUNT; p->pages = nr_good_pages; if (setup_swap_extents(p)) @@ -1493,7 +1576,7 @@ int swap_duplicate(swp_entry_t entry) if (type >= nr_swapfiles) goto bad_file; p = type + swap_info; - offset = swp_offset(entry); + offset = swp_offset(entry)/PAGE_MMUCOUNT; swap_device_lock(p); if (offset < p->max && p->swap_map[offset]) { @@ -1513,6 +1596,7 @@ out: bad_file: printk(KERN_ERR "swap_dup: %s%08lx\n", Bad_file, entry.val); + BUG(); goto out; } @@ -1534,7 +1618,7 @@ int valid_swaphandles(swp_entry_t entry, if (!page_cluster) /* no readahead */ return 0; - toff = (swp_offset(entry) >> page_cluster) << page_cluster; + toff = (swp_offset(entry)/PAGE_MMUCOUNT) & ~((1UL << page_cluster)-1); if (!toff) /* first page is swap header */ toff++, i--; *offset = toff; diff -prauN linux-2.5.70-bk4/mm/vcache.c pgcl-2.5.70-bk4-1/mm/vcache.c --- linux-2.5.70-bk4/mm/vcache.c 2003-05-26 18:00:26.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/vcache.c 2003-05-30 21:03:24.000000000 -0700 @@ -34,7 +34,7 @@ void __attach_vcache(vcache_t *vcache, { struct list_head *hash_head; - address &= PAGE_MASK; + address &= MMUPAGE_MASK; vcache->address = address; vcache->mm = mm; vcache->callback = callback; diff -prauN linux-2.5.70-bk4/mm/vmalloc.c pgcl-2.5.70-bk4-1/mm/vmalloc.c --- linux-2.5.70-bk4/mm/vmalloc.c 2003-05-26 18:00:41.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/vmalloc.c 2003-05-30 21:17:20.000000000 -0700 @@ -44,15 +44,12 @@ static void unmap_area_pte(pmd_t *pmd, u end = PMD_SIZE; do { - pte_t page; - page = ptep_get_and_clear(pte); - address += PAGE_SIZE; - pte++; - if (pte_none(page)) - continue; - if (pte_present(page)) - continue; + if (pte_present(*pte)) + pte_clear(pte); + else if (!pte_none(*pte)) printk(KERN_CRIT "Whee.. Swapped out page in kernel page table\n"); + pte++; + address += MMUPAGE_SIZE; } while (address < end); } @@ -83,56 +80,6 @@ static void unmap_area_pmd(pgd_t *dir, u } while (address < end); } -static int map_area_pte(pte_t *pte, unsigned long address, - unsigned long size, pgprot_t prot, - struct page ***pages) -{ - unsigned long end; - - address &= ~PMD_MASK; - end = address + size; - if (end > PMD_SIZE) - end = PMD_SIZE; - - do { - struct page *page = **pages; - - WARN_ON(!pte_none(*pte)); - if (!page) - return -ENOMEM; - - set_pte(pte, mk_pte(page, prot)); - address += PAGE_SIZE; - pte++; - (*pages)++; - } while (address < end); - return 0; -} - -static int map_area_pmd(pmd_t *pmd, unsigned long address, - unsigned long size, pgprot_t prot, - struct page ***pages) -{ - unsigned long end; - - address &= ~PGDIR_MASK; - end = address + size; - if (end > PGDIR_SIZE) - end = PGDIR_SIZE; - - do { - pte_t * pte = pte_alloc_kernel(&init_mm, pmd, address); - if (!pte) - return -ENOMEM; - if (map_area_pte(pte, address, end - address, prot, pages)) - return -ENOMEM; - address = (address + PMD_SIZE) & PMD_MASK; - pmd++; - } while (address < end); - - return 0; -} - void unmap_vm_area(struct vm_struct *area) { unsigned long address = VMALLOC_VMADDR(area->addr); @@ -149,30 +96,48 @@ void unmap_vm_area(struct vm_struct *are flush_tlb_kernel_range(VMALLOC_VMADDR(area->addr), end); } +#define PTE_TABLE_MASK ((PTRS_PER_PTE-1) * sizeof(pte_t)) +#define PMD_TABLE_MASK ((PTRS_PER_PMD-1) * sizeof(pmd_t)) + int map_vm_area(struct vm_struct *area, pgprot_t prot, struct page ***pages) { unsigned long address = VMALLOC_VMADDR(area->addr); - unsigned long end = address + (area->size-PAGE_SIZE); - pgd_t *dir; + /* don't instantiate PTE's for the guard page */ + unsigned long end = address + area->size - MMUPAGE_SIZE; + unsigned long voffset = 0; + pgd_t *pgd; int err = 0; - dir = pgd_offset_k(address); + pgd = pgd_offset_k(address); spin_lock(&init_mm.page_table_lock); do { - pmd_t *pmd = pmd_alloc(&init_mm, dir, address); + pmd_t *pmd = pmd_alloc(&init_mm, pgd, address); if (!pmd) { err = -ENOMEM; - break; + goto out; } - if (map_area_pmd(pmd, address, end - address, prot, pages)) { + + do { + pte_t *pte = pte_alloc_kernel(&init_mm, pmd, address); + if (!pte) { err = -ENOMEM; - break; + goto out; } - - address = (address + PGDIR_SIZE) & PGDIR_MASK; - dir++; - } while (address && (address < end)); - + do { + unsigned long pfn; + pfn = page_to_pfn((*pages)[voffset/PAGE_SIZE]); + pfn += (voffset/MMUPAGE_SIZE) % PAGE_MMUCOUNT; + set_pte(pte, pfn_pte(pfn, prot)); + ++pte; + address += MMUPAGE_SIZE; + voffset += MMUPAGE_SIZE; + } while (((unsigned long)pte & PTE_TABLE_MASK) && address < end); + ++pmd; + } while (((unsigned long)pmd & PMD_TABLE_MASK) && address < end); + ++pgd; + /* presumably address could wrap to 0, but I doubt it */ + } while (address && address < end); +out: spin_unlock(&init_mm.page_table_lock); flush_cache_all(); return err; @@ -201,7 +166,7 @@ struct vm_struct *get_vm_area(unsigned l /* * We always allocate a guard page. */ - size += PAGE_SIZE; + size += MMUPAGE_SIZE; if (unlikely(!size)) { kfree (area); return NULL; @@ -230,6 +195,9 @@ found: area->phys_addr = 0; write_unlock(&vmlist_lock); + printk("vmalloc, returning [0x%p, 0x%p)\n", + area->addr, ((char *)area->addr) + area->size); + return area; out: @@ -243,17 +211,20 @@ out: * * @addr: base address * - * Search for the kernel VM area starting at @addr, and remove it. + * Search for the kernel VM area containing @addr, and remove it. * This function returns the found VM area, but using it is NOT safe - * on SMP machines. + * on SMP machines; the final removal of an area must be serialized + * externally, and those who allocated the area own it. */ -struct vm_struct *remove_vm_area(void *addr) +struct vm_struct *remove_vm_area(void *__addr) { struct vm_struct **p, *tmp; + unsigned long addr = (unsigned long)__addr; write_lock(&vmlist_lock); - for (p = &vmlist ; (tmp = *p) ;p = &tmp->next) { - if (tmp->addr == addr) + for (p = &vmlist; (tmp = *p); p = &tmp->next) { + unsigned long tmp_addr = (unsigned long)tmp->addr; + if (addr >= tmp_addr && addr - tmp_addr < tmp->size) goto found; } write_unlock(&vmlist_lock); @@ -273,7 +244,7 @@ void __vunmap(void *addr, int deallocate if (!addr) return; - if ((PAGE_SIZE-1) & (unsigned long)addr) { + if ((MMUPAGE_SIZE-1) & (unsigned long)addr) { printk(KERN_ERR "Trying to vfree() bad address (%p)\n", addr); return; } @@ -289,8 +260,7 @@ void __vunmap(void *addr, int deallocate int i; for (i = 0; i < area->nr_pages; i++) { - if (unlikely(!area->pages[i])) - BUG(); + BUG_ON(unlikely(!area->pages[i])); __free_page(area->pages[i]); } @@ -349,10 +319,10 @@ void *vmap(struct page **pages, unsigned { struct vm_struct *area; - if (count > num_physpages) + if (PAGE_MMUCOUNT*count > num_physpages) return NULL; - area = get_vm_area((count << PAGE_SHIFT), flags); + area = get_vm_area(PAGE_SIZE*count, flags); if (!area) return NULL; if (map_vm_area(area, prot, &pages)) { @@ -380,16 +350,16 @@ void *__vmalloc(unsigned long size, int struct page **pages; unsigned int nr_pages, array_size, i; - size = PAGE_ALIGN(size); - if (!size || (size >> PAGE_SHIFT) > num_physpages) + size = MMUPAGE_ALIGN(size); + if (!size || (size >> MMUPAGE_SHIFT) > num_physpages) return NULL; area = get_vm_area(size, VM_ALLOC); if (!area) return NULL; - nr_pages = size >> PAGE_SHIFT; - array_size = (nr_pages * sizeof(struct page *)); + nr_pages = PAGE_ALIGN(size)/PAGE_SIZE; + array_size = nr_pages * sizeof(struct page *); area->nr_pages = nr_pages; area->pages = pages = kmalloc(array_size, (gfp_mask & ~__GFP_HIGHMEM)); @@ -460,7 +430,7 @@ long vread(char *buf, char *addr, unsign read_lock(&vmlist_lock); for (tmp = vmlist; tmp; tmp = tmp->next) { vaddr = (char *) tmp->addr; - if (addr >= vaddr + tmp->size - PAGE_SIZE) + if (addr >= vaddr + tmp->size - MMUPAGE_SIZE) continue; while (addr < vaddr) { if (count == 0) @@ -470,7 +440,7 @@ long vread(char *buf, char *addr, unsign addr++; count--; } - n = vaddr + tmp->size - PAGE_SIZE - addr; + n = vaddr + tmp->size - MMUPAGE_SIZE - addr; do { if (count == 0) goto finished; @@ -498,7 +468,7 @@ long vwrite(char *buf, char *addr, unsig read_lock(&vmlist_lock); for (tmp = vmlist; tmp; tmp = tmp->next) { vaddr = (char *) tmp->addr; - if (addr >= vaddr + tmp->size - PAGE_SIZE) + if (addr >= vaddr + tmp->size - MMUPAGE_SIZE) continue; while (addr < vaddr) { if (count == 0) @@ -507,7 +477,7 @@ long vwrite(char *buf, char *addr, unsig addr++; count--; } - n = vaddr + tmp->size - PAGE_SIZE - addr; + n = vaddr + tmp->size - MMUPAGE_SIZE - addr; do { if (count == 0) goto finished; diff -prauN linux-2.5.70-bk4/mm/vmscan.c pgcl-2.5.70-bk4-1/mm/vmscan.c --- linux-2.5.70-bk4/mm/vmscan.c 2003-05-26 18:00:24.000000000 -0700 +++ pgcl-2.5.70-bk4-1/mm/vmscan.c 2003-05-30 21:03:24.000000000 -0700 @@ -398,7 +398,7 @@ shrink_list(struct list_head *page_list, #ifdef CONFIG_SWAP if (PageSwapCache(page)) { - swp_entry_t swap = { .val = page->index }; + swp_entry_t swap = { .val = page->index*PAGE_MMUCOUNT }; __delete_from_swap_cache(page); spin_unlock(&mapping->page_lock); swap_free(swap); diff -prauN linux-2.5.70-bk4/net/ipv4/netfilter/ip_conntrack_core.c pgcl-2.5.70-bk4-1/net/ipv4/netfilter/ip_conntrack_core.c --- linux-2.5.70-bk4/net/ipv4/netfilter/ip_conntrack_core.c 2003-05-26 18:00:19.000000000 -0700 +++ pgcl-2.5.70-bk4-1/net/ipv4/netfilter/ip_conntrack_core.c 2003-05-30 21:03:24.000000000 -0700 @@ -1413,9 +1413,9 @@ int __init ip_conntrack_init(void) ip_conntrack_htable_size = hashsize; } else { ip_conntrack_htable_size - = (((num_physpages << PAGE_SHIFT) / 16384) + = (((num_physpages << MMUPAGE_SHIFT) / 16384) / sizeof(struct list_head)); - if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE)) + if (num_physpages > (1024 * 1024 * 1024 / MMUPAGE_SIZE)) ip_conntrack_htable_size = 8192; if (ip_conntrack_htable_size < 16) ip_conntrack_htable_size = 16; diff -prauN linux-2.5.70-bk4/net/ipv4/tcp.c pgcl-2.5.70-bk4-1/net/ipv4/tcp.c --- linux-2.5.70-bk4/net/ipv4/tcp.c 2003-05-30 21:02:38.000000000 -0700 +++ pgcl-2.5.70-bk4-1/net/ipv4/tcp.c 2003-05-30 21:03:24.000000000 -0700 @@ -2599,9 +2599,9 @@ void __init tcp_init(void) * The methodology is similar to that of the buffer cache. */ if (num_physpages >= (128 * 1024)) - goal = num_physpages >> (21 - PAGE_SHIFT); + goal = num_physpages >> (21 - MMUPAGE_SHIFT); else - goal = num_physpages >> (23 - PAGE_SHIFT); + goal = num_physpages >> (23 - MMUPAGE_SHIFT); for (order = 0; (1UL << order) < goal; order++) ;