* Copyright 1999 SuSE GmbH
* changed by Philipp Rumpf
* Copyright 1999 Philipp Rumpf (prumpf@tux.org)
+ * Copyright 2004 Randolph Chung (tausq@debian.org)
+ * Copyright 2006 Helge Deller (deller@gmx.de)
*
*/
-#include <linux/config.h>
#include <linux/module.h>
#include <linux/mm.h>
#include <linux/initrd.h>
#include <linux/swap.h>
#include <linux/unistd.h>
+#include <linux/nodemask.h> /* for node_online_map */
+#include <linux/pagemap.h> /* for release_pages and page_cache_release */
#include <asm/pgalloc.h>
#include <asm/tlb.h>
#include <asm/pdc_chassis.h>
+#include <asm/mmzone.h>
+#include <asm/sections.h>
DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-extern char _text; /* start of kernel code, defined by linker */
extern int data_start;
-extern char _end; /* end of BSS, defined by linker */
-extern char __init_begin, __init_end;
#ifdef CONFIG_DISCONTIGMEM
-struct node_map_data node_data[MAX_PHYSMEM_RANGES];
-bootmem_data_t bmem_data[MAX_PHYSMEM_RANGES];
-unsigned char *chunkmap;
-unsigned int maxchunkmap;
+struct node_map_data node_data[MAX_NUMNODES] __read_mostly;
+bootmem_data_t bmem_data[MAX_NUMNODES] __read_mostly;
+unsigned char pfnnid_map[PFNNID_MAP_MAX] __read_mostly;
#endif
static struct resource data_resource = {
.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
};
-static struct resource sysram_resources[MAX_PHYSMEM_RANGES];
-
-static unsigned long max_pfn;
+static struct resource sysram_resources[MAX_PHYSMEM_RANGES] __read_mostly;
/* The following array is initialized from the firmware specific
* information retrieved in kernel/inventory.c.
*/
-physmem_range_t pmem_ranges[MAX_PHYSMEM_RANGES];
-int npmem_ranges;
+physmem_range_t pmem_ranges[MAX_PHYSMEM_RANGES] __read_mostly;
+int npmem_ranges __read_mostly;
#ifdef __LP64__
#define MAX_MEM (~0UL)
#define MAX_MEM (3584U*1024U*1024U)
#endif /* !__LP64__ */
-static unsigned long mem_limit = MAX_MEM;
+static unsigned long mem_limit __read_mostly = MAX_MEM;
static void __init mem_limit_func(void)
{
disable_sr_hashing(); /* Turn off space register hashing */
-#ifdef CONFIG_DISCONTIGMEM
- /*
- * The below is still true as of 2.4.2. If this is ever fixed,
- * we can remove this warning!
- */
-
- printk(KERN_WARNING "\n\n");
- printk(KERN_WARNING "CONFIG_DISCONTIGMEM is enabled, which is probably a mistake. This\n");
- printk(KERN_WARNING "option can lead to heavy swapping, even when there are gigabytes\n");
- printk(KERN_WARNING "of free memory.\n\n");
-#endif
-
-#ifdef __LP64__
-
-#ifndef CONFIG_DISCONTIGMEM
/*
* Sort the ranges. Since the number of ranges is typically
* small, and performance is not an issue here, just do
}
}
+#ifndef CONFIG_DISCONTIGMEM
/*
* Throw out ranges that are too far apart (controlled by
- * MAX_GAP). If CONFIG_DISCONTIGMEM wasn't implemented so
- * poorly, we would recommend enabling that option, but,
- * until it is fixed, this is the best way to go.
+ * MAX_GAP).
*/
for (i = 1; i < npmem_ranges; i++) {
(pmem_ranges[i-1].start_pfn +
pmem_ranges[i-1].pages) > MAX_GAP) {
npmem_ranges = i;
+ printk("Large gap in memory detected (%ld pages). "
+ "Consider turning on CONFIG_DISCONTIGMEM\n",
+ pmem_ranges[i].start_pfn -
+ (pmem_ranges[i-1].start_pfn +
+ pmem_ranges[i-1].pages));
break;
}
}
size = (pmem_ranges[i].pages << PAGE_SHIFT);
start = (pmem_ranges[i].start_pfn << PAGE_SHIFT);
- printk(KERN_INFO "%2d) Start 0x%016lx End 0x%016lx Size %6ld Mb\n",
+ printk(KERN_INFO "%2d) Start 0x%016lx End 0x%016lx Size %6ld MB\n",
i,start, start + (size - 1), size >> 20);
}
}
-#endif /* __LP64__ */
-
sysram_resource_count = npmem_ranges;
for (i = 0; i < sysram_resource_count; i++) {
struct resource *res = &sysram_resources[i];
mem_limit_func(); /* check for "mem=" argument */
mem_max = 0;
+ num_physpages = 0;
for (i = 0; i < npmem_ranges; i++) {
unsigned long rsize;
rsize = pmem_ranges[i].pages << PAGE_SHIFT;
if ((mem_max + rsize) > mem_limit) {
- printk(KERN_WARNING "Memory truncated to %ld Mb\n", mem_limit >> 20);
+ printk(KERN_WARNING "Memory truncated to %ld MB\n", mem_limit >> 20);
if (mem_max == mem_limit)
npmem_ranges = i;
else {
npmem_ranges = i + 1;
mem_max = mem_limit;
}
+ num_physpages += pmem_ranges[i].pages;
break;
}
+ num_physpages += pmem_ranges[i].pages;
mem_max += rsize;
}
- printk(KERN_INFO "Total Memory: %ld Mb\n",mem_max >> 20);
+ printk(KERN_INFO "Total Memory: %ld MB\n",mem_max >> 20);
#ifndef CONFIG_DISCONTIGMEM
-
/* Merge the ranges, keeping track of the holes */
{
bootmap_start_pfn = PAGE_ALIGN(__pa((unsigned long) &_end)) >> PAGE_SHIFT;
#ifdef CONFIG_DISCONTIGMEM
+ for (i = 0; i < MAX_PHYSMEM_RANGES; i++) {
+ memset(NODE_DATA(i), 0, sizeof(pg_data_t));
+ NODE_DATA(i)->bdata = &bmem_data[i];
+ }
+ memset(pfnnid_map, 0xff, sizeof(pfnnid_map));
+
for (i = 0; i < npmem_ranges; i++)
- node_data[i].pg_data.bdata = &bmem_data[i];
+ node_set_online(i);
#endif
+
/*
* Initialize and free the full range of memory in each range.
* Note that the only writing these routines do are to the bootmap,
max_pfn = start_pfn + npages;
}
+ /* IOMMU is always used to access "high mem" on those boxes
+ * that can support enough mem that a PCI device couldn't
+ * directly DMA to any physical addresses.
+ * ISA DMA support will need to revisit this.
+ */
+ max_low_pfn = max_pfn;
+
if ((bootmap_pfn - bootmap_start_pfn) != bootmap_pages) {
printk(KERN_WARNING "WARNING! bootmap sizing is messed up!\n");
BUG();
reserve_bootmem_node(NODE_DATA(0), 0UL,
(unsigned long)(PAGE0->mem_free + PDC_CONSOLE_IO_IODC_SIZE));
- reserve_bootmem_node(NODE_DATA(0),__pa((unsigned long)&_text),
- (unsigned long)(&_end - &_text));
+ reserve_bootmem_node(NODE_DATA(0), __pa((unsigned long)_text),
+ (unsigned long)(_end - _text));
reserve_bootmem_node(NODE_DATA(0), (bootmap_start_pfn << PAGE_SHIFT),
((bootmap_pfn - bootmap_start_pfn) << PAGE_SHIFT));
#endif
data_resource.start = virt_to_phys(&data_start);
- data_resource.end = virt_to_phys(&_end)-1;
- code_resource.start = virt_to_phys(&_text);
+ data_resource.end = virt_to_phys(_end) - 1;
+ code_resource.start = virt_to_phys(_text);
code_resource.end = virt_to_phys(&data_start)-1;
/* We don't know which region the kernel will be in, so try
void free_initmem(void)
{
- /* FIXME: */
-#if 0
- printk(KERN_INFO "NOT FREEING INITMEM (%dk)\n",
- (&__init_end - &__init_begin) >> 10);
- return;
-#else
- unsigned long addr;
-
+ unsigned long addr, init_begin, init_end;
+
printk(KERN_INFO "Freeing unused kernel memory: ");
-#if 1
+#ifdef CONFIG_DEBUG_KERNEL
/* Attempt to catch anyone trying to execute code here
* by filling the page with BRK insns.
*
*/
local_irq_disable();
- memset(&__init_begin, 0x00,
- (unsigned long)&__init_end - (unsigned long)&__init_begin);
+ memset(__init_begin, 0x00,
+ (unsigned long)__init_end - (unsigned long)__init_begin);
flush_data_cache();
asm volatile("sync" : : );
- flush_icache_range((unsigned long)&__init_begin, (unsigned long)&__init_end);
+ flush_icache_range((unsigned long)__init_begin, (unsigned long)__init_end);
asm volatile("sync" : : );
local_irq_enable();
#endif
- addr = (unsigned long)(&__init_begin);
- for (; addr < (unsigned long)(&__init_end); addr += PAGE_SIZE) {
+ /* align __init_begin and __init_end to page size,
+ ignoring linker script where we might have tried to save RAM */
+ init_begin = PAGE_ALIGN((unsigned long)(__init_begin));
+ init_end = PAGE_ALIGN((unsigned long)(__init_end));
+ for (addr = init_begin; addr < init_end; addr += PAGE_SIZE) {
ClearPageReserved(virt_to_page(addr));
- set_page_count(virt_to_page(addr), 1);
+ init_page_count(virt_to_page(addr));
free_page(addr);
num_physpages++;
totalram_pages++;
/* set up a new led state on systems shipped LED State panel */
pdc_chassis_send_status(PDC_CHASSIS_DIRECT_BCOMPLETE);
- printk("%luk freed\n", (unsigned long)(&__init_end - &__init_begin) >> 10);
-#endif
+ printk("%luk freed\n", (init_end - init_begin) >> 10);
+}
+
+
+#ifdef CONFIG_DEBUG_RODATA
+void mark_rodata_ro(void)
+{
+ /* rodata memory was already mapped with KERNEL_RO access rights by
+ pagetable_init() and map_pages(). No need to do additional stuff here */
+ printk (KERN_INFO "Write protecting the kernel read-only data: %luk\n",
+ (unsigned long)(__end_rodata - __start_rodata) >> 10);
}
+#endif
+
/*
* Just an arbitrary offset to serve as a "hole" between mapping areas
#define SET_MAP_OFFSET(x) ((void *)(((unsigned long)(x) + VM_MAP_OFFSET) \
& ~(VM_MAP_OFFSET-1)))
-void *vmalloc_start;
+void *vmalloc_start __read_mostly;
EXPORT_SYMBOL(vmalloc_start);
#ifdef CONFIG_PA11
-unsigned long pcxl_dma_start;
+unsigned long pcxl_dma_start __read_mostly;
#endif
void __init mem_init(void)
{
- int i;
-
high_memory = __va((max_pfn << PAGE_SHIFT));
- max_mapnr = (virt_to_page(high_memory - 1) - mem_map) + 1;
- num_physpages = 0;
- mem_map = zone_table[0]->zone_mem_map;
- for (i = 0; i < npmem_ranges; i++)
- num_physpages += free_all_bootmem_node(NODE_DATA(i));
- totalram_pages = num_physpages;
+#ifndef CONFIG_DISCONTIGMEM
+ max_mapnr = page_to_pfn(virt_to_page(high_memory - 1)) + 1;
+ totalram_pages += free_all_bootmem();
+#else
+ {
+ int i;
+
+ for (i = 0; i < npmem_ranges; i++)
+ totalram_pages += free_all_bootmem_node(NODE_DATA(i));
+ }
+#endif
printk(KERN_INFO "Memory: %luk available\n", num_physpages << (PAGE_SHIFT-10));
}
-int do_check_pgt_cache(int low, int high)
-{
- return 0;
-}
-
-unsigned long *empty_zero_page;
+unsigned long *empty_zero_page __read_mostly;
void show_mem(void)
{
printk(KERN_INFO "Mem-info:\n");
show_free_areas();
- printk(KERN_INFO "Free swap: %6dkB\n",nr_swap_pages<<(PAGE_SHIFT-10));
+ printk(KERN_INFO "Free swap: %6ldkB\n",
+ nr_swap_pages<<(PAGE_SHIFT-10));
+#ifndef CONFIG_DISCONTIGMEM
i = max_mapnr;
while (i-- > 0) {
total++;
reserved++;
else if (PageSwapCache(mem_map+i))
cached++;
- else if (!atomic_read(&mem_map[i].count))
+ else if (!page_count(&mem_map[i]))
free++;
else
- shared += atomic_read(&mem_map[i].count) - 1;
+ shared += page_count(&mem_map[i]) - 1;
+ }
+#else
+ for (i = 0; i < npmem_ranges; i++) {
+ int j;
+
+ for (j = node_start_pfn(i); j < node_end_pfn(i); j++) {
+ struct page *p;
+ unsigned long flags;
+
+ pgdat_resize_lock(NODE_DATA(i), &flags);
+ p = nid_page_nr(i, j) - node_start_pfn(i);
+
+ total++;
+ if (PageReserved(p))
+ reserved++;
+ else if (PageSwapCache(p))
+ cached++;
+ else if (!page_count(p))
+ free++;
+ else
+ shared += page_count(p) - 1;
+ pgdat_resize_unlock(NODE_DATA(i), &flags);
+ }
}
+#endif
printk(KERN_INFO "%d pages of RAM\n", total);
printk(KERN_INFO "%d reserved pages\n", reserved);
printk(KERN_INFO "%d pages shared\n", shared);
printk(KERN_INFO "%d pages swap cached\n", cached);
+
+
+#ifdef CONFIG_DISCONTIGMEM
+ {
+ struct zonelist *zl;
+ int i, j, k;
+
+ for (i = 0; i < npmem_ranges; i++) {
+ for (j = 0; j < MAX_NR_ZONES; j++) {
+ zl = NODE_DATA(i)->node_zonelists + j;
+
+ printk("Zone list for zone %d on node %d: ", j, i);
+ for (k = 0; zl->zones[k] != NULL; k++)
+ printk("[%d/%s] ", zone_to_nid(zl->zones[k]), zl->zones[k]->name);
+ printk("\n");
+ }
+ }
+ }
+#endif
}
extern const unsigned long fault_vector_20;
extern void * const linux_gateway_page;
- ro_start = __pa((unsigned long)&_text);
+ ro_start = __pa((unsigned long)_text);
ro_end = __pa((unsigned long)&data_start);
fv_addr = __pa((unsigned long)&fault_vector_20) & PAGE_MASK;
gw_addr = __pa((unsigned long)&linux_gateway_page) & PAGE_MASK;
#if PTRS_PER_PMD == 1
pmd = (pmd_t *)__pa(pg_dir);
#else
- pmd = (pmd_t *) (PAGE_MASK & pgd_val(*pg_dir));
+ pmd = (pmd_t *)pgd_address(*pg_dir);
/*
* pmd is physical at this point
pmd = (pmd_t *) __pa(pmd);
}
- pgd_val(*pg_dir) = _PAGE_TABLE | (unsigned long) pmd;
+ pgd_populate(NULL, pg_dir, __va(pmd));
#endif
pg_dir++;
* pg_table is physical at this point
*/
- pg_table = (pte_t *) (PAGE_MASK & pmd_val(*pmd));
+ pg_table = (pte_t *)pmd_address(*pmd);
if (!pg_table) {
pg_table = (pte_t *)
alloc_bootmem_low_pages_node(NODE_DATA(0),PAGE_SIZE);
pg_table = (pte_t *) __pa(pg_table);
}
- pmd_val(*pmd) = _PAGE_TABLE |
- (unsigned long) pg_table;
+ pmd_populate_kernel(NULL, pmd, __va(pg_table));
/* now change pg_table to kernel virtual addresses */
for (tmp2 = start_pte; tmp2 < PTRS_PER_PTE; tmp2++,pg_table++) {
pte_t pte;
-#if !defined(CONFIG_STI_CONSOLE)
-#warning STI console should explicitly allocate executable pages but does not
/*
* Map the fault vector writable so we can
* write the HPMC checksum.
*/
+#if defined(CONFIG_PARISC_PAGE_SIZE_4KB)
if (address >= ro_start && address < ro_end
&& address != fv_addr
&& address != gw_addr)
#ifdef CONFIG_BLK_DEV_INITRD
if (initrd_end && initrd_end > mem_limit) {
- printk("initrd: mapping %08lx-%08lx\n", initrd_start, initrd_end);
+ printk(KERN_INFO "initrd: mapping %08lx-%08lx\n", initrd_start, initrd_end);
map_pages(initrd_start, __pa(initrd_start),
initrd_end - initrd_start, PAGE_KERNEL);
}
#if PTRS_PER_PMD == 1
pmd = (pmd_t *)__pa(pg_dir);
#else
- pmd = (pmd_t *) (PAGE_MASK & pgd_val(*pg_dir));
+ pmd = (pmd_t *) pgd_address(*pg_dir);
/*
* pmd is physical at this point
pmd = (pmd_t *) __pa(pmd);
}
- pgd_val(*pg_dir) = _PAGE_TABLE | (unsigned long) pmd;
+ __pgd_val_set(*pg_dir, PxD_FLAG_PRESENT | PxD_FLAG_VALID | (unsigned long) pmd);
#endif
/* now change pmd to kernel virtual addresses */
* pg_table is physical at this point
*/
- pg_table = (pte_t *) (PAGE_MASK & pmd_val(*pmd));
+ pg_table = (pte_t *) pmd_address(*pmd);
if (!pg_table)
pg_table = (pte_t *) __pa(get_zeroed_page(GFP_KERNEL));
- pmd_val(*pmd) = _PAGE_TABLE | (unsigned long) pg_table;
+ __pmd_val_set(*pmd, PxD_FLAG_PRESENT | PxD_FLAG_VALID | (unsigned long) pg_table);
/* now change pg_table to kernel virtual addresses */
EXPORT_SYMBOL(map_hpux_gateway_page);
#endif
-extern void flush_tlb_all_local(void);
-
void __init paging_init(void)
{
int i;
pagetable_init();
gateway_init();
flush_cache_all_local(); /* start with known state */
- flush_tlb_all_local();
+ flush_tlb_all_local(NULL);
for (i = 0; i < npmem_ranges; i++) {
- unsigned long zones_size[MAX_NR_ZONES] = { 0, 0, 0, };
+ unsigned long zones_size[MAX_NR_ZONES] = { 0, };
+ /* We have an IOMMU, so all memory can go into a single
+ ZONE_DMA zone. */
zones_size[ZONE_DMA] = pmem_ranges[i].pages;
- free_area_init_node(i,NODE_DATA(i),NULL,zones_size,
- (pmem_ranges[i].start_pfn << PAGE_SHIFT),0);
- }
#ifdef CONFIG_DISCONTIGMEM
- /*
- * Initialize support for virt_to_page() macro.
- *
- * Note that MAX_ADDRESS is the largest virtual address that
- * we can map. However, since we map all physical memory into
- * the kernel address space, it also has an effect on the maximum
- * physical address we can map (MAX_ADDRESS - PAGE_OFFSET).
- */
-
- maxchunkmap = MAX_ADDRESS >> CHUNKSHIFT;
- chunkmap = (unsigned char *)alloc_bootmem(maxchunkmap);
-
- for (i = 0; i < maxchunkmap; i++)
- chunkmap[i] = BADCHUNK;
-
- for (i = 0; i < npmem_ranges; i++) {
-
- ADJ_NODE_MEM_MAP(i) = NODE_MEM_MAP(i) - pmem_ranges[i].start_pfn;
+ /* Need to initialize the pfnnid_map before we can initialize
+ the zone */
{
- unsigned long chunk_paddr;
- unsigned long end_paddr;
- int chunknum;
-
- chunk_paddr = (pmem_ranges[i].start_pfn << PAGE_SHIFT);
- end_paddr = chunk_paddr + (pmem_ranges[i].pages << PAGE_SHIFT);
- chunk_paddr &= CHUNKMASK;
-
- chunknum = (int)CHUNKNUM(chunk_paddr);
- while (chunk_paddr < end_paddr) {
- if (chunknum >= maxchunkmap)
- goto badchunkmap1;
- if (chunkmap[chunknum] != BADCHUNK)
- goto badchunkmap2;
- chunkmap[chunknum] = (unsigned char)i;
- chunk_paddr += CHUNKSZ;
- chunknum++;
- }
+ int j;
+ for (j = (pmem_ranges[i].start_pfn >> PFNNID_SHIFT);
+ j <= ((pmem_ranges[i].start_pfn + pmem_ranges[i].pages) >> PFNNID_SHIFT);
+ j++) {
+ pfnnid_map[j] = i;
+ }
}
- }
-
- return;
-
-badchunkmap1:
- panic("paging_init: Physical address exceeds maximum address space!\n");
-badchunkmap2:
- panic("paging_init: Collision in chunk map array. CHUNKSZ needs to be smaller\n");
#endif
+
+ free_area_init_node(i, NODE_DATA(i), zones_size,
+ pmem_ranges[i].start_pfn, NULL);
+ }
}
#ifdef CONFIG_PA20
static unsigned long free_space_ids = NR_SPACE_IDS - 1;
static unsigned long dirty_space_ids = 0;
-static spinlock_t sid_lock = SPIN_LOCK_UNLOCKED;
+static DEFINE_SPINLOCK(sid_lock);
unsigned long alloc_sid(void)
{
flush_tlb_all(); /* flush_tlb_all() calls recycle_sids() */
spin_lock(&sid_lock);
}
- if (free_space_ids == 0)
- BUG();
+ BUG_ON(free_space_ids == 0);
}
free_space_ids--;
spin_lock(&sid_lock);
- if (*dirty_space_offset & (1L << index))
- BUG(); /* attempt to free space id twice */
+ BUG_ON(*dirty_space_offset & (1L << index)); /* attempt to free space id twice */
*dirty_space_offset |= (1L << index);
dirty_space_ids++;
static unsigned long recycle_ndirty;
static unsigned long recycle_dirty_array[SID_ARRAY_SIZE];
-static unsigned int recycle_inuse = 0;
+static unsigned int recycle_inuse;
void flush_tlb_all(void)
{
do_recycle = 0;
spin_lock(&sid_lock);
if (dirty_space_ids > RECYCLE_THRESHOLD) {
- if (recycle_inuse) {
- BUG(); /* FIXME: Use a semaphore/wait queue here */
- }
+ BUG_ON(recycle_inuse); /* FIXME: Use a semaphore/wait queue here */
get_dirty_sids(&recycle_ndirty,recycle_dirty_array);
recycle_inuse++;
do_recycle++;
}
spin_unlock(&sid_lock);
- on_each_cpu((void (*)(void *))flush_tlb_all_local, NULL, 1, 1);
+ on_each_cpu(flush_tlb_all_local, NULL, 1, 1);
if (do_recycle) {
spin_lock(&sid_lock);
recycle_sids(recycle_ndirty,recycle_dirty_array);
void flush_tlb_all(void)
{
spin_lock(&sid_lock);
- flush_tlb_all_local();
+ flush_tlb_all_local(NULL);
recycle_sids();
spin_unlock(&sid_lock);
}
#ifdef CONFIG_BLK_DEV_INITRD
void free_initrd_mem(unsigned long start, unsigned long end)
{
-#if 0
- if (start < end)
- printk(KERN_INFO "Freeing initrd memory: %ldk freed\n", (end - start) >> 10);
+ if (start >= end)
+ return;
+ printk(KERN_INFO "Freeing initrd memory: %ldk freed\n", (end - start) >> 10);
for (; start < end; start += PAGE_SIZE) {
ClearPageReserved(virt_to_page(start));
- set_page_count(virt_to_page(start), 1);
+ init_page_count(virt_to_page(start));
free_page(start);
num_physpages++;
totalram_pages++;
}
-#endif
}
#endif
git://git.onelab.eu / linux-2.6.git / blobdiff