/*
|
/*
|
* linux/mm/filemap.c
|
* linux/mm/filemap.c
|
*
|
*
|
* Copyright (C) 1994, 1995 Linus Torvalds
|
* Copyright (C) 1994, 1995 Linus Torvalds
|
*/
|
*/
|
|
|
/*
|
/*
|
* This file handles the generic file mmap semantics used by
|
* This file handles the generic file mmap semantics used by
|
* most "normal" filesystems (but you don't /have/ to use this:
|
* most "normal" filesystems (but you don't /have/ to use this:
|
* the NFS filesystem does this differently, for example)
|
* the NFS filesystem does this differently, for example)
|
*/
|
*/
|
#include <linux/config.h> /* CONFIG_READA_SMALL */
|
#include <linux/config.h> /* CONFIG_READA_SMALL */
|
#include <linux/stat.h>
|
#include <linux/stat.h>
|
#include <linux/sched.h>
|
#include <linux/sched.h>
|
#include <linux/kernel.h>
|
#include <linux/kernel.h>
|
#include <linux/mm.h>
|
#include <linux/mm.h>
|
#include <linux/shm.h>
|
#include <linux/shm.h>
|
#include <linux/errno.h>
|
#include <linux/errno.h>
|
#include <linux/mman.h>
|
#include <linux/mman.h>
|
#include <linux/string.h>
|
#include <linux/string.h>
|
#include <linux/malloc.h>
|
#include <linux/malloc.h>
|
#include <linux/fs.h>
|
#include <linux/fs.h>
|
#include <linux/locks.h>
|
#include <linux/locks.h>
|
#include <linux/pagemap.h>
|
#include <linux/pagemap.h>
|
#include <linux/swap.h>
|
#include <linux/swap.h>
|
|
|
#include <asm/segment.h>
|
#include <asm/segment.h>
|
#include <asm/system.h>
|
#include <asm/system.h>
|
#include <asm/pgtable.h>
|
#include <asm/pgtable.h>
|
|
|
/*
|
/*
|
* Shared mappings implemented 30.11.1994. It's not fully working yet,
|
* Shared mappings implemented 30.11.1994. It's not fully working yet,
|
* though.
|
* though.
|
*
|
*
|
* Shared mappings now work. 15.8.1995 Bruno.
|
* Shared mappings now work. 15.8.1995 Bruno.
|
*/
|
*/
|
|
|
unsigned long page_cache_size = 0;
|
unsigned long page_cache_size = 0;
|
struct page * page_hash_table[PAGE_HASH_SIZE];
|
struct page * page_hash_table[PAGE_HASH_SIZE];
|
|
|
/*
|
/*
|
* Simple routines for both non-shared and shared mappings.
|
* Simple routines for both non-shared and shared mappings.
|
*/
|
*/
|
|
|
#define release_page(page) __free_page((page))
|
#define release_page(page) __free_page((page))
|
|
|
/*
|
/*
|
* Invalidate the pages of an inode, removing all pages that aren't
|
* Invalidate the pages of an inode, removing all pages that aren't
|
* locked down (those are sure to be up-to-date anyway, so we shouldn't
|
* locked down (those are sure to be up-to-date anyway, so we shouldn't
|
* invalidate them).
|
* invalidate them).
|
*/
|
*/
|
void invalidate_inode_pages(struct inode * inode)
|
void invalidate_inode_pages(struct inode * inode)
|
{
|
{
|
struct page ** p;
|
struct page ** p;
|
struct page * page;
|
struct page * page;
|
|
|
p = &inode->i_pages;
|
p = &inode->i_pages;
|
while ((page = *p) != NULL) {
|
while ((page = *p) != NULL) {
|
if (PageLocked(page)) {
|
if (PageLocked(page)) {
|
p = &page->next;
|
p = &page->next;
|
continue;
|
continue;
|
}
|
}
|
inode->i_nrpages--;
|
inode->i_nrpages--;
|
if ((*p = page->next) != NULL)
|
if ((*p = page->next) != NULL)
|
(*p)->prev = page->prev;
|
(*p)->prev = page->prev;
|
page->dirty = 0;
|
page->dirty = 0;
|
page->next = NULL;
|
page->next = NULL;
|
page->prev = NULL;
|
page->prev = NULL;
|
remove_page_from_hash_queue(page);
|
remove_page_from_hash_queue(page);
|
page->inode = NULL;
|
page->inode = NULL;
|
__free_page(page);
|
__free_page(page);
|
continue;
|
continue;
|
}
|
}
|
}
|
}
|
|
|
/*
|
/*
|
* Truncate the page cache at a set offset, removing the pages
|
* Truncate the page cache at a set offset, removing the pages
|
* that are beyond that offset (and zeroing out partial pages).
|
* that are beyond that offset (and zeroing out partial pages).
|
*/
|
*/
|
void truncate_inode_pages(struct inode * inode, unsigned long start)
|
void truncate_inode_pages(struct inode * inode, unsigned long start)
|
{
|
{
|
struct page ** p;
|
struct page ** p;
|
struct page * page;
|
struct page * page;
|
|
|
repeat:
|
repeat:
|
p = &inode->i_pages;
|
p = &inode->i_pages;
|
while ((page = *p) != NULL) {
|
while ((page = *p) != NULL) {
|
unsigned long offset = page->offset;
|
unsigned long offset = page->offset;
|
|
|
/* page wholly truncated - free it */
|
/* page wholly truncated - free it */
|
if (offset >= start) {
|
if (offset >= start) {
|
if (PageLocked(page)) {
|
if (PageLocked(page)) {
|
__wait_on_page(page);
|
__wait_on_page(page);
|
goto repeat;
|
goto repeat;
|
}
|
}
|
inode->i_nrpages--;
|
inode->i_nrpages--;
|
if ((*p = page->next) != NULL)
|
if ((*p = page->next) != NULL)
|
(*p)->prev = page->prev;
|
(*p)->prev = page->prev;
|
page->dirty = 0;
|
page->dirty = 0;
|
page->next = NULL;
|
page->next = NULL;
|
page->prev = NULL;
|
page->prev = NULL;
|
remove_page_from_hash_queue(page);
|
remove_page_from_hash_queue(page);
|
page->inode = NULL;
|
page->inode = NULL;
|
__free_page(page);
|
__free_page(page);
|
continue;
|
continue;
|
}
|
}
|
p = &page->next;
|
p = &page->next;
|
offset = start - offset;
|
offset = start - offset;
|
/* partial truncate, clear end of page */
|
/* partial truncate, clear end of page */
|
if (offset < PAGE_SIZE) {
|
if (offset < PAGE_SIZE) {
|
unsigned long address = page_address(page);
|
unsigned long address = page_address(page);
|
memset((void *) (offset + address), 0, PAGE_SIZE - offset);
|
memset((void *) (offset + address), 0, PAGE_SIZE - offset);
|
flush_page_to_ram(address);
|
flush_page_to_ram(address);
|
}
|
}
|
}
|
}
|
}
|
}
|
|
|
int shrink_mmap(int priority, int dma, int free_buf)
|
int shrink_mmap(int priority, int dma, int free_buf)
|
{
|
{
|
static int clock = 0;
|
static int clock = 0;
|
struct page * page;
|
struct page * page;
|
unsigned long limit = MAP_NR(high_memory);
|
unsigned long limit = MAP_NR(high_memory);
|
struct buffer_head *tmp, *bh;
|
struct buffer_head *tmp, *bh;
|
int count_max, count_min;
|
int count_max, count_min;
|
|
|
count_max = (limit<<1) >> (priority>>1);
|
count_max = (limit<<1) >> (priority>>1);
|
count_min = (limit<<1) >> (priority);
|
count_min = (limit<<1) >> (priority);
|
|
|
page = mem_map + clock;
|
page = mem_map + clock;
|
do {
|
do {
|
count_max--;
|
count_max--;
|
if (page->inode || page->buffers)
|
if (page->inode || page->buffers)
|
count_min--;
|
count_min--;
|
|
|
if (PageLocked(page))
|
if (PageLocked(page))
|
goto next;
|
goto next;
|
if (dma && !PageDMA(page))
|
if (dma && !PageDMA(page))
|
goto next;
|
goto next;
|
/* First of all, regenerate the page's referenced bit
|
/* First of all, regenerate the page's referenced bit
|
from any buffers in the page */
|
from any buffers in the page */
|
bh = page->buffers;
|
bh = page->buffers;
|
if (bh) {
|
if (bh) {
|
tmp = bh;
|
tmp = bh;
|
do {
|
do {
|
if (buffer_touched(tmp)) {
|
if (buffer_touched(tmp)) {
|
clear_bit(BH_Touched, &tmp->b_state);
|
clear_bit(BH_Touched, &tmp->b_state);
|
set_bit(PG_referenced, &page->flags);
|
set_bit(PG_referenced, &page->flags);
|
}
|
}
|
tmp = tmp->b_this_page;
|
tmp = tmp->b_this_page;
|
} while (tmp != bh);
|
} while (tmp != bh);
|
}
|
}
|
|
|
/* We can't throw away shared pages, but we do mark
|
/* We can't throw away shared pages, but we do mark
|
them as referenced. This relies on the fact that
|
them as referenced. This relies on the fact that
|
no page is currently in both the page cache and the
|
no page is currently in both the page cache and the
|
buffer cache; we'd have to modify the following
|
buffer cache; we'd have to modify the following
|
test to allow for that case. */
|
test to allow for that case. */
|
|
|
switch (page->count) {
|
switch (page->count) {
|
case 1:
|
case 1:
|
/* If it has been referenced recently, don't free it */
|
/* If it has been referenced recently, don't free it */
|
if (clear_bit(PG_referenced, &page->flags)) {
|
if (clear_bit(PG_referenced, &page->flags)) {
|
/* age this page potential used */
|
/* age this page potential used */
|
if (priority < 4)
|
if (priority < 4)
|
age_page(page);
|
age_page(page);
|
break;
|
break;
|
}
|
}
|
|
|
/* is it a page cache page? */
|
/* is it a page cache page? */
|
if (page->inode) {
|
if (page->inode) {
|
remove_page_from_hash_queue(page);
|
remove_page_from_hash_queue(page);
|
remove_page_from_inode_queue(page);
|
remove_page_from_inode_queue(page);
|
__free_page(page);
|
__free_page(page);
|
return 1;
|
return 1;
|
}
|
}
|
|
|
/* is it a buffer cache page? */
|
/* is it a buffer cache page? */
|
if (free_buf && bh && try_to_free_buffer(bh, &bh, 6))
|
if (free_buf && bh && try_to_free_buffer(bh, &bh, 6))
|
return 1;
|
return 1;
|
break;
|
break;
|
|
|
default:
|
default:
|
/* more than one users: we can't throw it away */
|
/* more than one users: we can't throw it away */
|
set_bit(PG_referenced, &page->flags);
|
set_bit(PG_referenced, &page->flags);
|
/* fall through */
|
/* fall through */
|
case 0:
|
case 0:
|
/* nothing */
|
/* nothing */
|
}
|
}
|
next:
|
next:
|
page++;
|
page++;
|
clock++;
|
clock++;
|
if (clock >= limit) {
|
if (clock >= limit) {
|
clock = 0;
|
clock = 0;
|
page = mem_map;
|
page = mem_map;
|
}
|
}
|
} while (count_max > 0 && count_min > 0);
|
} while (count_max > 0 && count_min > 0);
|
return 0;
|
return 0;
|
}
|
}
|
|
|
/*
|
/*
|
* This is called from try_to_swap_out() when we try to get rid of some
|
* This is called from try_to_swap_out() when we try to get rid of some
|
* pages.. If we're unmapping the last occurrence of this page, we also
|
* pages.. If we're unmapping the last occurrence of this page, we also
|
* free it from the page hash-queues etc, as we don't want to keep it
|
* free it from the page hash-queues etc, as we don't want to keep it
|
* in-core unnecessarily.
|
* in-core unnecessarily.
|
*/
|
*/
|
unsigned long page_unuse(unsigned long page)
|
unsigned long page_unuse(unsigned long page)
|
{
|
{
|
struct page * p = mem_map + MAP_NR(page);
|
struct page * p = mem_map + MAP_NR(page);
|
int count = p->count;
|
int count = p->count;
|
|
|
if (count != 2)
|
if (count != 2)
|
return count;
|
return count;
|
if (!p->inode)
|
if (!p->inode)
|
return count;
|
return count;
|
remove_page_from_hash_queue(p);
|
remove_page_from_hash_queue(p);
|
remove_page_from_inode_queue(p);
|
remove_page_from_inode_queue(p);
|
free_page(page);
|
free_page(page);
|
return 1;
|
return 1;
|
}
|
}
|
|
|
/*
|
/*
|
* Update a page cache copy, when we're doing a "write()" system call
|
* Update a page cache copy, when we're doing a "write()" system call
|
* See also "update_vm_cache()".
|
* See also "update_vm_cache()".
|
*/
|
*/
|
void update_vm_cache(struct inode * inode, unsigned long pos, const char * buf, int count)
|
void update_vm_cache(struct inode * inode, unsigned long pos, const char * buf, int count)
|
{
|
{
|
unsigned long offset, len;
|
unsigned long offset, len;
|
|
|
offset = (pos & ~PAGE_MASK);
|
offset = (pos & ~PAGE_MASK);
|
pos = pos & PAGE_MASK;
|
pos = pos & PAGE_MASK;
|
len = PAGE_SIZE - offset;
|
len = PAGE_SIZE - offset;
|
do {
|
do {
|
struct page * page;
|
struct page * page;
|
|
|
if (len > count)
|
if (len > count)
|
len = count;
|
len = count;
|
page = find_page(inode, pos);
|
page = find_page(inode, pos);
|
if (page) {
|
if (page) {
|
wait_on_page(page);
|
wait_on_page(page);
|
memcpy((void *) (offset + page_address(page)), buf, len);
|
memcpy((void *) (offset + page_address(page)), buf, len);
|
release_page(page);
|
release_page(page);
|
}
|
}
|
count -= len;
|
count -= len;
|
buf += len;
|
buf += len;
|
len = PAGE_SIZE;
|
len = PAGE_SIZE;
|
offset = 0;
|
offset = 0;
|
pos += PAGE_SIZE;
|
pos += PAGE_SIZE;
|
} while (count);
|
} while (count);
|
}
|
}
|
|
|
static inline void add_to_page_cache(struct page * page,
|
static inline void add_to_page_cache(struct page * page,
|
struct inode * inode, unsigned long offset,
|
struct inode * inode, unsigned long offset,
|
struct page **hash)
|
struct page **hash)
|
{
|
{
|
page->count++;
|
page->count++;
|
page->flags &= ~((1 << PG_uptodate) | (1 << PG_error));
|
page->flags &= ~((1 << PG_uptodate) | (1 << PG_error));
|
page->offset = offset;
|
page->offset = offset;
|
add_page_to_inode_queue(inode, page);
|
add_page_to_inode_queue(inode, page);
|
__add_page_to_hash_queue(page, hash);
|
__add_page_to_hash_queue(page, hash);
|
}
|
}
|
|
|
/*
|
/*
|
* Try to read ahead in the file. "page_cache" is a potentially free page
|
* Try to read ahead in the file. "page_cache" is a potentially free page
|
* that we could use for the cache (if it is 0 we can try to create one,
|
* that we could use for the cache (if it is 0 we can try to create one,
|
* this is all overlapped with the IO on the previous page finishing anyway)
|
* this is all overlapped with the IO on the previous page finishing anyway)
|
*/
|
*/
|
static unsigned long try_to_read_ahead(struct inode * inode, unsigned long offset, unsigned long page_cache)
|
static unsigned long try_to_read_ahead(struct inode * inode, unsigned long offset, unsigned long page_cache)
|
{
|
{
|
struct page * page;
|
struct page * page;
|
struct page ** hash;
|
struct page ** hash;
|
|
|
offset &= PAGE_MASK;
|
offset &= PAGE_MASK;
|
switch (page_cache) {
|
switch (page_cache) {
|
case 0:
|
case 0:
|
page_cache = __get_free_page(GFP_KERNEL);
|
page_cache = __get_free_page(GFP_KERNEL);
|
if (!page_cache)
|
if (!page_cache)
|
break;
|
break;
|
default:
|
default:
|
if (offset >= inode->i_size)
|
if (offset >= inode->i_size)
|
break;
|
break;
|
hash = page_hash(inode, offset);
|
hash = page_hash(inode, offset);
|
page = __find_page(inode, offset, *hash);
|
page = __find_page(inode, offset, *hash);
|
if (!page) {
|
if (!page) {
|
/*
|
/*
|
* Ok, add the new page to the hash-queues...
|
* Ok, add the new page to the hash-queues...
|
*/
|
*/
|
page = mem_map + MAP_NR(page_cache);
|
page = mem_map + MAP_NR(page_cache);
|
add_to_page_cache(page, inode, offset, hash);
|
add_to_page_cache(page, inode, offset, hash);
|
inode->i_op->readpage(inode, page);
|
inode->i_op->readpage(inode, page);
|
page_cache = 0;
|
page_cache = 0;
|
}
|
}
|
release_page(page);
|
release_page(page);
|
}
|
}
|
return page_cache;
|
return page_cache;
|
}
|
}
|
|
|
/*
|
/*
|
* Wait for IO to complete on a locked page.
|
* Wait for IO to complete on a locked page.
|
*
|
*
|
* This must be called with the caller "holding" the page,
|
* This must be called with the caller "holding" the page,
|
* ie with increased "page->count" so that the page won't
|
* ie with increased "page->count" so that the page won't
|
* go away during the wait..
|
* go away during the wait..
|
*/
|
*/
|
void __wait_on_page(struct page *page)
|
void __wait_on_page(struct page *page)
|
{
|
{
|
struct wait_queue wait = { current, NULL };
|
struct wait_queue wait = { current, NULL };
|
|
|
add_wait_queue(&page->wait, &wait);
|
add_wait_queue(&page->wait, &wait);
|
repeat:
|
repeat:
|
run_task_queue(&tq_disk);
|
run_task_queue(&tq_disk);
|
current->state = TASK_UNINTERRUPTIBLE;
|
current->state = TASK_UNINTERRUPTIBLE;
|
if (PageLocked(page)) {
|
if (PageLocked(page)) {
|
schedule();
|
schedule();
|
goto repeat;
|
goto repeat;
|
}
|
}
|
remove_wait_queue(&page->wait, &wait);
|
remove_wait_queue(&page->wait, &wait);
|
current->state = TASK_RUNNING;
|
current->state = TASK_RUNNING;
|
}
|
}
|
|
|
#if 0
|
#if 0
|
#define PROFILE_READAHEAD
|
#define PROFILE_READAHEAD
|
#define DEBUG_READAHEAD
|
#define DEBUG_READAHEAD
|
#endif
|
#endif
|
|
|
/*
|
/*
|
* Read-ahead profiling information
|
* Read-ahead profiling information
|
* --------------------------------
|
* --------------------------------
|
* Every PROFILE_MAXREADCOUNT, the following information is written
|
* Every PROFILE_MAXREADCOUNT, the following information is written
|
* to the syslog:
|
* to the syslog:
|
* Percentage of asynchronous read-ahead.
|
* Percentage of asynchronous read-ahead.
|
* Average of read-ahead fields context value.
|
* Average of read-ahead fields context value.
|
* If DEBUG_READAHEAD is defined, a snapshot of these fields is written
|
* If DEBUG_READAHEAD is defined, a snapshot of these fields is written
|
* to the syslog.
|
* to the syslog.
|
*/
|
*/
|
|
|
#ifdef PROFILE_READAHEAD
|
#ifdef PROFILE_READAHEAD
|
|
|
#define PROFILE_MAXREADCOUNT 1000
|
#define PROFILE_MAXREADCOUNT 1000
|
|
|
static unsigned long total_reada;
|
static unsigned long total_reada;
|
static unsigned long total_async;
|
static unsigned long total_async;
|
static unsigned long total_ramax;
|
static unsigned long total_ramax;
|
static unsigned long total_ralen;
|
static unsigned long total_ralen;
|
static unsigned long total_rawin;
|
static unsigned long total_rawin;
|
|
|
static void profile_readahead(int async, struct file *filp)
|
static void profile_readahead(int async, struct file *filp)
|
{
|
{
|
unsigned long flags;
|
unsigned long flags;
|
|
|
++total_reada;
|
++total_reada;
|
if (async)
|
if (async)
|
++total_async;
|
++total_async;
|
|
|
total_ramax += filp->f_ramax;
|
total_ramax += filp->f_ramax;
|
total_ralen += filp->f_ralen;
|
total_ralen += filp->f_ralen;
|
total_rawin += filp->f_rawin;
|
total_rawin += filp->f_rawin;
|
|
|
if (total_reada > PROFILE_MAXREADCOUNT) {
|
if (total_reada > PROFILE_MAXREADCOUNT) {
|
save_flags(flags);
|
save_flags(flags);
|
cli();
|
cli();
|
if (!(total_reada > PROFILE_MAXREADCOUNT)) {
|
if (!(total_reada > PROFILE_MAXREADCOUNT)) {
|
restore_flags(flags);
|
restore_flags(flags);
|
return;
|
return;
|
}
|
}
|
|
|
printk("Readahead average: max=%ld, len=%ld, win=%ld, async=%ld%%\n",
|
printk("Readahead average: max=%ld, len=%ld, win=%ld, async=%ld%%\n",
|
total_ramax/total_reada,
|
total_ramax/total_reada,
|
total_ralen/total_reada,
|
total_ralen/total_reada,
|
total_rawin/total_reada,
|
total_rawin/total_reada,
|
(total_async*100)/total_reada);
|
(total_async*100)/total_reada);
|
#ifdef DEBUG_READAHEAD
|
#ifdef DEBUG_READAHEAD
|
printk("Readahead snapshot: max=%ld, len=%ld, win=%ld, raend=%ld\n",
|
printk("Readahead snapshot: max=%ld, len=%ld, win=%ld, raend=%ld\n",
|
filp->f_ramax, filp->f_ralen, filp->f_rawin, filp->f_raend);
|
filp->f_ramax, filp->f_ralen, filp->f_rawin, filp->f_raend);
|
#endif
|
#endif
|
|
|
total_reada = 0;
|
total_reada = 0;
|
total_async = 0;
|
total_async = 0;
|
total_ramax = 0;
|
total_ramax = 0;
|
total_ralen = 0;
|
total_ralen = 0;
|
total_rawin = 0;
|
total_rawin = 0;
|
|
|
restore_flags(flags);
|
restore_flags(flags);
|
}
|
}
|
}
|
}
|
#endif /* defined PROFILE_READAHEAD */
|
#endif /* defined PROFILE_READAHEAD */
|
|
|
/*
|
/*
|
* Read-ahead context:
|
* Read-ahead context:
|
* -------------------
|
* -------------------
|
* The read ahead context fields of the "struct file" are the following:
|
* The read ahead context fields of the "struct file" are the following:
|
* - f_raend : position of the first byte after the last page we tried to
|
* - f_raend : position of the first byte after the last page we tried to
|
* read ahead.
|
* read ahead.
|
* - f_ramax : current read-ahead maximum size.
|
* - f_ramax : current read-ahead maximum size.
|
* - f_ralen : length of the current IO read block we tried to read-ahead.
|
* - f_ralen : length of the current IO read block we tried to read-ahead.
|
* - f_rawin : length of the current read-ahead window.
|
* - f_rawin : length of the current read-ahead window.
|
* if last read-ahead was synchronous then
|
* if last read-ahead was synchronous then
|
* f_rawin = f_ralen
|
* f_rawin = f_ralen
|
* otherwise (was asynchronous)
|
* otherwise (was asynchronous)
|
* f_rawin = previous value of f_ralen + f_ralen
|
* f_rawin = previous value of f_ralen + f_ralen
|
*
|
*
|
* Read-ahead limits:
|
* Read-ahead limits:
|
* ------------------
|
* ------------------
|
* MIN_READAHEAD : minimum read-ahead size when read-ahead.
|
* MIN_READAHEAD : minimum read-ahead size when read-ahead.
|
* MAX_READAHEAD : maximum read-ahead size when read-ahead.
|
* MAX_READAHEAD : maximum read-ahead size when read-ahead.
|
*
|
*
|
* Synchronous read-ahead benefits:
|
* Synchronous read-ahead benefits:
|
* --------------------------------
|
* --------------------------------
|
* Using reasonable IO xfer length from peripheral devices increase system
|
* Using reasonable IO xfer length from peripheral devices increase system
|
* performances.
|
* performances.
|
* Reasonable means, in this context, not too large but not too small.
|
* Reasonable means, in this context, not too large but not too small.
|
* The actual maximum value is:
|
* The actual maximum value is:
|
* MAX_READAHEAD + PAGE_SIZE = 76k is CONFIG_READA_SMALL is undefined
|
* MAX_READAHEAD + PAGE_SIZE = 76k is CONFIG_READA_SMALL is undefined
|
* and 32K if defined (4K page size assumed).
|
* and 32K if defined (4K page size assumed).
|
*
|
*
|
* Asynchronous read-ahead benefits:
|
* Asynchronous read-ahead benefits:
|
* ---------------------------------
|
* ---------------------------------
|
* Overlapping next read request and user process execution increase system
|
* Overlapping next read request and user process execution increase system
|
* performance.
|
* performance.
|
*
|
*
|
* Read-ahead risks:
|
* Read-ahead risks:
|
* -----------------
|
* -----------------
|
* We have to guess which further data are needed by the user process.
|
* We have to guess which further data are needed by the user process.
|
* If these data are often not really needed, it's bad for system
|
* If these data are often not really needed, it's bad for system
|
* performances.
|
* performances.
|
* However, we know that files are often accessed sequentially by
|
* However, we know that files are often accessed sequentially by
|
* application programs and it seems that it is possible to have some good
|
* application programs and it seems that it is possible to have some good
|
* strategy in that guessing.
|
* strategy in that guessing.
|
* We only try to read-ahead files that seems to be read sequentially.
|
* We only try to read-ahead files that seems to be read sequentially.
|
*
|
*
|
* Asynchronous read-ahead risks:
|
* Asynchronous read-ahead risks:
|
* ------------------------------
|
* ------------------------------
|
* In order to maximize overlapping, we must start some asynchronous read
|
* In order to maximize overlapping, we must start some asynchronous read
|
* request from the device, as soon as possible.
|
* request from the device, as soon as possible.
|
* We must be very careful about:
|
* We must be very careful about:
|
* - The number of effective pending IO read requests.
|
* - The number of effective pending IO read requests.
|
* ONE seems to be the only reasonable value.
|
* ONE seems to be the only reasonable value.
|
* - The total memory pool usage for the file access stream.
|
* - The total memory pool usage for the file access stream.
|
* This maximum memory usage is implicitly 2 IO read chunks:
|
* This maximum memory usage is implicitly 2 IO read chunks:
|
* 2*(MAX_READAHEAD + PAGE_SIZE) = 156K if CONFIG_READA_SMALL is undefined,
|
* 2*(MAX_READAHEAD + PAGE_SIZE) = 156K if CONFIG_READA_SMALL is undefined,
|
* 64k if defined (4K page size assumed).
|
* 64k if defined (4K page size assumed).
|
*/
|
*/
|
|
|
#define PageAlignSize(size) (((size) + PAGE_SIZE -1) & PAGE_MASK)
|
#define PageAlignSize(size) (((size) + PAGE_SIZE -1) & PAGE_MASK)
|
|
|
#ifdef CONFIG_READA_SMALL /* small readahead */
|
#ifdef CONFIG_READA_SMALL /* small readahead */
|
#define MAX_READAHEAD PageAlignSize(4096*7)
|
#define MAX_READAHEAD PageAlignSize(4096*7)
|
#define MIN_READAHEAD PageAlignSize(4096*2)
|
#define MIN_READAHEAD PageAlignSize(4096*2)
|
#else /* large readahead */
|
#else /* large readahead */
|
#define MAX_READAHEAD PageAlignSize(4096*18)
|
#define MAX_READAHEAD PageAlignSize(4096*18)
|
#define MIN_READAHEAD PageAlignSize(4096*3)
|
#define MIN_READAHEAD PageAlignSize(4096*3)
|
#endif
|
#endif
|
|
|
static inline unsigned long generic_file_readahead(int reada_ok, struct file * filp, struct inode * inode,
|
static inline unsigned long generic_file_readahead(int reada_ok, struct file * filp, struct inode * inode,
|
unsigned long ppos, struct page * page,
|
unsigned long ppos, struct page * page,
|
unsigned long page_cache)
|
unsigned long page_cache)
|
{
|
{
|
unsigned long max_ahead, ahead;
|
unsigned long max_ahead, ahead;
|
unsigned long raend;
|
unsigned long raend;
|
|
|
raend = filp->f_raend & PAGE_MASK;
|
raend = filp->f_raend & PAGE_MASK;
|
max_ahead = 0;
|
max_ahead = 0;
|
|
|
/*
|
/*
|
* The current page is locked.
|
* The current page is locked.
|
* If the current position is inside the previous read IO request, do not
|
* If the current position is inside the previous read IO request, do not
|
* try to reread previously read ahead pages.
|
* try to reread previously read ahead pages.
|
* Otherwise decide or not to read ahead some pages synchronously.
|
* Otherwise decide or not to read ahead some pages synchronously.
|
* If we are not going to read ahead, set the read ahead context for this
|
* If we are not going to read ahead, set the read ahead context for this
|
* page only.
|
* page only.
|
*/
|
*/
|
if (PageLocked(page)) {
|
if (PageLocked(page)) {
|
if (!filp->f_ralen || ppos >= raend || ppos + filp->f_ralen < raend) {
|
if (!filp->f_ralen || ppos >= raend || ppos + filp->f_ralen < raend) {
|
raend = ppos;
|
raend = ppos;
|
if (raend < inode->i_size)
|
if (raend < inode->i_size)
|
max_ahead = filp->f_ramax;
|
max_ahead = filp->f_ramax;
|
filp->f_rawin = 0;
|
filp->f_rawin = 0;
|
filp->f_ralen = PAGE_SIZE;
|
filp->f_ralen = PAGE_SIZE;
|
if (!max_ahead) {
|
if (!max_ahead) {
|
filp->f_raend = ppos + filp->f_ralen;
|
filp->f_raend = ppos + filp->f_ralen;
|
filp->f_rawin += filp->f_ralen;
|
filp->f_rawin += filp->f_ralen;
|
}
|
}
|
}
|
}
|
}
|
}
|
/*
|
/*
|
* The current page is not locked.
|
* The current page is not locked.
|
* If we were reading ahead and,
|
* If we were reading ahead and,
|
* if the current max read ahead size is not zero and,
|
* if the current max read ahead size is not zero and,
|
* if the current position is inside the last read-ahead IO request,
|
* if the current position is inside the last read-ahead IO request,
|
* it is the moment to try to read ahead asynchronously.
|
* it is the moment to try to read ahead asynchronously.
|
* We will later force unplug device in order to force asynchronous read IO.
|
* We will later force unplug device in order to force asynchronous read IO.
|
*/
|
*/
|
else if (reada_ok && filp->f_ramax && raend >= PAGE_SIZE &&
|
else if (reada_ok && filp->f_ramax && raend >= PAGE_SIZE &&
|
ppos <= raend && ppos + filp->f_ralen >= raend) {
|
ppos <= raend && ppos + filp->f_ralen >= raend) {
|
/*
|
/*
|
* Add ONE page to max_ahead in order to try to have about the same IO max size
|
* Add ONE page to max_ahead in order to try to have about the same IO max size
|
* as synchronous read-ahead (MAX_READAHEAD + 1)*PAGE_SIZE.
|
* as synchronous read-ahead (MAX_READAHEAD + 1)*PAGE_SIZE.
|
* Compute the position of the last page we have tried to read in order to
|
* Compute the position of the last page we have tried to read in order to
|
* begin to read ahead just at the next page.
|
* begin to read ahead just at the next page.
|
*/
|
*/
|
raend -= PAGE_SIZE;
|
raend -= PAGE_SIZE;
|
if (raend < inode->i_size)
|
if (raend < inode->i_size)
|
max_ahead = filp->f_ramax + PAGE_SIZE;
|
max_ahead = filp->f_ramax + PAGE_SIZE;
|
|
|
if (max_ahead) {
|
if (max_ahead) {
|
filp->f_rawin = filp->f_ralen;
|
filp->f_rawin = filp->f_ralen;
|
filp->f_ralen = 0;
|
filp->f_ralen = 0;
|
reada_ok = 2;
|
reada_ok = 2;
|
}
|
}
|
}
|
}
|
/*
|
/*
|
* Try to read ahead pages.
|
* Try to read ahead pages.
|
* We hope that ll_rw_blk() plug/unplug, coalescence, requests sort and the
|
* We hope that ll_rw_blk() plug/unplug, coalescence, requests sort and the
|
* scheduler, will work enough for us to avoid too bad actuals IO requests.
|
* scheduler, will work enough for us to avoid too bad actuals IO requests.
|
*/
|
*/
|
ahead = 0;
|
ahead = 0;
|
while (ahead < max_ahead) {
|
while (ahead < max_ahead) {
|
ahead += PAGE_SIZE;
|
ahead += PAGE_SIZE;
|
page_cache = try_to_read_ahead(inode, raend + ahead, page_cache);
|
page_cache = try_to_read_ahead(inode, raend + ahead, page_cache);
|
}
|
}
|
/*
|
/*
|
* If we tried to read ahead some pages,
|
* If we tried to read ahead some pages,
|
* If we tried to read ahead asynchronously,
|
* If we tried to read ahead asynchronously,
|
* Try to force unplug of the device in order to start an asynchronous
|
* Try to force unplug of the device in order to start an asynchronous
|
* read IO request.
|
* read IO request.
|
* Update the read-ahead context.
|
* Update the read-ahead context.
|
* Store the length of the current read-ahead window.
|
* Store the length of the current read-ahead window.
|
* Double the current max read ahead size.
|
* Double the current max read ahead size.
|
* That heuristic avoid to do some large IO for files that are not really
|
* That heuristic avoid to do some large IO for files that are not really
|
* accessed sequentially.
|
* accessed sequentially.
|
*/
|
*/
|
if (ahead) {
|
if (ahead) {
|
if (reada_ok == 2) {
|
if (reada_ok == 2) {
|
run_task_queue(&tq_disk);
|
run_task_queue(&tq_disk);
|
}
|
}
|
|
|
filp->f_ralen += ahead;
|
filp->f_ralen += ahead;
|
filp->f_rawin += filp->f_ralen;
|
filp->f_rawin += filp->f_ralen;
|
filp->f_raend = raend + ahead + PAGE_SIZE;
|
filp->f_raend = raend + ahead + PAGE_SIZE;
|
|
|
filp->f_ramax += filp->f_ramax;
|
filp->f_ramax += filp->f_ramax;
|
|
|
if (filp->f_ramax > MAX_READAHEAD)
|
if (filp->f_ramax > MAX_READAHEAD)
|
filp->f_ramax = MAX_READAHEAD;
|
filp->f_ramax = MAX_READAHEAD;
|
|
|
#ifdef PROFILE_READAHEAD
|
#ifdef PROFILE_READAHEAD
|
profile_readahead((reada_ok == 2), filp);
|
profile_readahead((reada_ok == 2), filp);
|
#endif
|
#endif
|
}
|
}
|
|
|
return page_cache;
|
return page_cache;
|
}
|
}
|
|
|
|
|
/*
|
/*
|
* This is a generic file read routine, and uses the
|
* This is a generic file read routine, and uses the
|
* inode->i_op->readpage() function for the actual low-level
|
* inode->i_op->readpage() function for the actual low-level
|
* stuff.
|
* stuff.
|
*
|
*
|
* This is really ugly. But the goto's actually try to clarify some
|
* This is really ugly. But the goto's actually try to clarify some
|
* of the logic when it comes to error handling etc.
|
* of the logic when it comes to error handling etc.
|
*/
|
*/
|
|
|
int generic_file_read(struct inode * inode, struct file * filp, char * buf, int count)
|
int generic_file_read(struct inode * inode, struct file * filp, char * buf, int count)
|
{
|
{
|
int error, read;
|
int error, read;
|
unsigned long pos, ppos, page_cache;
|
unsigned long pos, ppos, page_cache;
|
int reada_ok;
|
int reada_ok;
|
|
|
error = 0;
|
error = 0;
|
read = 0;
|
read = 0;
|
page_cache = 0;
|
page_cache = 0;
|
|
|
pos = filp->f_pos;
|
pos = filp->f_pos;
|
ppos = pos & PAGE_MASK;
|
ppos = pos & PAGE_MASK;
|
/*
|
/*
|
* If the current position is outside the previous read-ahead window,
|
* If the current position is outside the previous read-ahead window,
|
* we reset the current read-ahead context and set read ahead max to zero
|
* we reset the current read-ahead context and set read ahead max to zero
|
* (will be set to just needed value later),
|
* (will be set to just needed value later),
|
* otherwise, we assume that the file accesses are sequential enough to
|
* otherwise, we assume that the file accesses are sequential enough to
|
* continue read-ahead.
|
* continue read-ahead.
|
*/
|
*/
|
if (ppos > filp->f_raend || ppos + filp->f_rawin < filp->f_raend) {
|
if (ppos > filp->f_raend || ppos + filp->f_rawin < filp->f_raend) {
|
reada_ok = 0;
|
reada_ok = 0;
|
filp->f_raend = 0;
|
filp->f_raend = 0;
|
filp->f_ralen = 0;
|
filp->f_ralen = 0;
|
filp->f_ramax = 0;
|
filp->f_ramax = 0;
|
filp->f_rawin = 0;
|
filp->f_rawin = 0;
|
} else {
|
} else {
|
reada_ok = 1;
|
reada_ok = 1;
|
}
|
}
|
/*
|
/*
|
* Adjust the current value of read-ahead max.
|
* Adjust the current value of read-ahead max.
|
* If the read operation stay in the first half page, force no readahead.
|
* If the read operation stay in the first half page, force no readahead.
|
* Otherwise try to increase read ahead max just enough to do the read request.
|
* Otherwise try to increase read ahead max just enough to do the read request.
|
* Then, at least MIN_READAHEAD if read ahead is ok,
|
* Then, at least MIN_READAHEAD if read ahead is ok,
|
* and at most MAX_READAHEAD in all cases.
|
* and at most MAX_READAHEAD in all cases.
|
*/
|
*/
|
if (pos + count <= (PAGE_SIZE >> 1)) {
|
if (pos + count <= (PAGE_SIZE >> 1)) {
|
filp->f_ramax = 0;
|
filp->f_ramax = 0;
|
} else {
|
} else {
|
unsigned long needed;
|
unsigned long needed;
|
|
|
needed = ((pos + count) & PAGE_MASK) - ppos;
|
needed = ((pos + count) & PAGE_MASK) - ppos;
|
|
|
if (filp->f_ramax < needed)
|
if (filp->f_ramax < needed)
|
filp->f_ramax = needed;
|
filp->f_ramax = needed;
|
|
|
if (reada_ok && filp->f_ramax < MIN_READAHEAD)
|
if (reada_ok && filp->f_ramax < MIN_READAHEAD)
|
filp->f_ramax = MIN_READAHEAD;
|
filp->f_ramax = MIN_READAHEAD;
|
if (filp->f_ramax > MAX_READAHEAD)
|
if (filp->f_ramax > MAX_READAHEAD)
|
filp->f_ramax = MAX_READAHEAD;
|
filp->f_ramax = MAX_READAHEAD;
|
}
|
}
|
|
|
for (;;) {
|
for (;;) {
|
struct page *page, **hash;
|
struct page *page, **hash;
|
|
|
if (pos >= inode->i_size)
|
if (pos >= inode->i_size)
|
break;
|
break;
|
|
|
/*
|
/*
|
* Try to find the data in the page cache..
|
* Try to find the data in the page cache..
|
*/
|
*/
|
hash = page_hash(inode, pos & PAGE_MASK);
|
hash = page_hash(inode, pos & PAGE_MASK);
|
page = __find_page(inode, pos & PAGE_MASK, *hash);
|
page = __find_page(inode, pos & PAGE_MASK, *hash);
|
if (!page)
|
if (!page)
|
goto no_cached_page;
|
goto no_cached_page;
|
|
|
found_page:
|
found_page:
|
/*
|
/*
|
* Try to read ahead only if the current page is filled or being filled.
|
* Try to read ahead only if the current page is filled or being filled.
|
* Otherwise, if we were reading ahead, decrease max read ahead size to
|
* Otherwise, if we were reading ahead, decrease max read ahead size to
|
* the minimum value.
|
* the minimum value.
|
* In this context, that seems to may happen only on some read error or if
|
* In this context, that seems to may happen only on some read error or if
|
* the page has been rewritten.
|
* the page has been rewritten.
|
*/
|
*/
|
if (PageUptodate(page) || PageLocked(page))
|
if (PageUptodate(page) || PageLocked(page))
|
page_cache = generic_file_readahead(reada_ok, filp, inode, pos & PAGE_MASK, page, page_cache);
|
page_cache = generic_file_readahead(reada_ok, filp, inode, pos & PAGE_MASK, page, page_cache);
|
else if (reada_ok && filp->f_ramax > MIN_READAHEAD)
|
else if (reada_ok && filp->f_ramax > MIN_READAHEAD)
|
filp->f_ramax = MIN_READAHEAD;
|
filp->f_ramax = MIN_READAHEAD;
|
|
|
wait_on_page(page);
|
wait_on_page(page);
|
|
|
if (!PageUptodate(page))
|
if (!PageUptodate(page))
|
goto page_read_error;
|
goto page_read_error;
|
|
|
success:
|
success:
|
/*
|
/*
|
* Ok, we have the page, it's up-to-date and ok,
|
* Ok, we have the page, it's up-to-date and ok,
|
* so now we can finally copy it to user space...
|
* so now we can finally copy it to user space...
|
*/
|
*/
|
{
|
{
|
unsigned long offset, nr;
|
unsigned long offset, nr;
|
offset = pos & ~PAGE_MASK;
|
offset = pos & ~PAGE_MASK;
|
nr = PAGE_SIZE - offset;
|
nr = PAGE_SIZE - offset;
|
if (nr > count)
|
if (nr > count)
|
nr = count;
|
nr = count;
|
|
|
if (nr > inode->i_size - pos)
|
if (nr > inode->i_size - pos)
|
nr = inode->i_size - pos;
|
nr = inode->i_size - pos;
|
memcpy_tofs(buf, (void *) (page_address(page) + offset), nr);
|
memcpy_tofs(buf, (void *) (page_address(page) + offset), nr);
|
release_page(page);
|
release_page(page);
|
buf += nr;
|
buf += nr;
|
pos += nr;
|
pos += nr;
|
read += nr;
|
read += nr;
|
count -= nr;
|
count -= nr;
|
if (count) {
|
if (count) {
|
/*
|
/*
|
* to prevent hogging the CPU on well-cached systems,
|
* to prevent hogging the CPU on well-cached systems,
|
* schedule if needed, it's safe to do it here:
|
* schedule if needed, it's safe to do it here:
|
*/
|
*/
|
if (need_resched)
|
if (need_resched)
|
schedule();
|
schedule();
|
continue;
|
continue;
|
}
|
}
|
break;
|
break;
|
}
|
}
|
|
|
no_cached_page:
|
no_cached_page:
|
/*
|
/*
|
* Ok, it wasn't cached, so we need to create a new
|
* Ok, it wasn't cached, so we need to create a new
|
* page..
|
* page..
|
*/
|
*/
|
if (!page_cache) {
|
if (!page_cache) {
|
page_cache = __get_free_page(GFP_KERNEL);
|
page_cache = __get_free_page(GFP_KERNEL);
|
/*
|
/*
|
* That could have slept, so go around to the
|
* That could have slept, so go around to the
|
* very beginning..
|
* very beginning..
|
*/
|
*/
|
if (page_cache)
|
if (page_cache)
|
continue;
|
continue;
|
error = -ENOMEM;
|
error = -ENOMEM;
|
break;
|
break;
|
}
|
}
|
|
|
/*
|
/*
|
* Ok, add the new page to the hash-queues...
|
* Ok, add the new page to the hash-queues...
|
*/
|
*/
|
page = mem_map + MAP_NR(page_cache);
|
page = mem_map + MAP_NR(page_cache);
|
page_cache = 0;
|
page_cache = 0;
|
add_to_page_cache(page, inode, pos & PAGE_MASK, hash);
|
add_to_page_cache(page, inode, pos & PAGE_MASK, hash);
|
|
|
/*
|
/*
|
* Error handling is tricky. If we get a read error,
|
* Error handling is tricky. If we get a read error,
|
* the cached page stays in the cache (but uptodate=0),
|
* the cached page stays in the cache (but uptodate=0),
|
* and the next process that accesses it will try to
|
* and the next process that accesses it will try to
|
* re-read it. This is needed for NFS etc, where the
|
* re-read it. This is needed for NFS etc, where the
|
* identity of the reader can decide if we can read the
|
* identity of the reader can decide if we can read the
|
* page or not..
|
* page or not..
|
*/
|
*/
|
/*
|
/*
|
* We have to read the page.
|
* We have to read the page.
|
* If we were reading ahead, we had previously tried to read this page,
|
* If we were reading ahead, we had previously tried to read this page,
|
* That means that the page has probably been removed from the cache before
|
* That means that the page has probably been removed from the cache before
|
* the application process needs it, or has been rewritten.
|
* the application process needs it, or has been rewritten.
|
* Decrease max readahead size to the minimum value in that situation.
|
* Decrease max readahead size to the minimum value in that situation.
|
*/
|
*/
|
if (reada_ok && filp->f_ramax > MIN_READAHEAD)
|
if (reada_ok && filp->f_ramax > MIN_READAHEAD)
|
filp->f_ramax = MIN_READAHEAD;
|
filp->f_ramax = MIN_READAHEAD;
|
|
|
error = inode->i_op->readpage(inode, page);
|
error = inode->i_op->readpage(inode, page);
|
if (!error)
|
if (!error)
|
goto found_page;
|
goto found_page;
|
release_page(page);
|
release_page(page);
|
break;
|
break;
|
|
|
page_read_error:
|
page_read_error:
|
/*
|
/*
|
* We found the page, but it wasn't up-to-date.
|
* We found the page, but it wasn't up-to-date.
|
* Try to re-read it _once_. We do this synchronously,
|
* Try to re-read it _once_. We do this synchronously,
|
* because this happens only if there were errors.
|
* because this happens only if there were errors.
|
*/
|
*/
|
error = inode->i_op->readpage(inode, page);
|
error = inode->i_op->readpage(inode, page);
|
if (!error) {
|
if (!error) {
|
wait_on_page(page);
|
wait_on_page(page);
|
if (PageUptodate(page) && !PageError(page))
|
if (PageUptodate(page) && !PageError(page))
|
goto success;
|
goto success;
|
error = -EIO; /* Some unspecified error occurred.. */
|
error = -EIO; /* Some unspecified error occurred.. */
|
}
|
}
|
release_page(page);
|
release_page(page);
|
break;
|
break;
|
}
|
}
|
|
|
filp->f_pos = pos;
|
filp->f_pos = pos;
|
filp->f_reada = 1;
|
filp->f_reada = 1;
|
if (page_cache)
|
if (page_cache)
|
free_page(page_cache);
|
free_page(page_cache);
|
UPDATE_ATIME(inode)
|
UPDATE_ATIME(inode)
|
if (!read)
|
if (!read)
|
read = error;
|
read = error;
|
return read;
|
return read;
|
}
|
}
|
|
|
/*
|
/*
|
* Semantics for shared and private memory areas are different past the end
|
* Semantics for shared and private memory areas are different past the end
|
* of the file. A shared mapping past the last page of the file is an error
|
* of the file. A shared mapping past the last page of the file is an error
|
* and results in a SIGBUS, while a private mapping just maps in a zero page.
|
* and results in a SIGBUS, while a private mapping just maps in a zero page.
|
*
|
*
|
* The goto's are kind of ugly, but this streamlines the normal case of having
|
* The goto's are kind of ugly, but this streamlines the normal case of having
|
* it in the page cache, and handles the special cases reasonably without
|
* it in the page cache, and handles the special cases reasonably without
|
* having a lot of duplicated code.
|
* having a lot of duplicated code.
|
*/
|
*/
|
static unsigned long filemap_nopage(struct vm_area_struct * area, unsigned long address, int no_share)
|
static unsigned long filemap_nopage(struct vm_area_struct * area, unsigned long address, int no_share)
|
{
|
{
|
unsigned long offset;
|
unsigned long offset;
|
struct page * page, **hash;
|
struct page * page, **hash;
|
struct inode * inode = area->vm_inode;
|
struct inode * inode = area->vm_inode;
|
unsigned long old_page, new_page;
|
unsigned long old_page, new_page;
|
|
|
new_page = 0;
|
new_page = 0;
|
offset = (address & PAGE_MASK) - area->vm_start + area->vm_offset;
|
offset = (address & PAGE_MASK) - area->vm_start + area->vm_offset;
|
if (offset >= inode->i_size && (area->vm_flags & VM_SHARED) && area->vm_mm == current->mm)
|
if (offset >= inode->i_size && (area->vm_flags & VM_SHARED) && area->vm_mm == current->mm)
|
goto no_page;
|
goto no_page;
|
|
|
/*
|
/*
|
* Do we have something in the page cache already?
|
* Do we have something in the page cache already?
|
*/
|
*/
|
hash = page_hash(inode, offset);
|
hash = page_hash(inode, offset);
|
page = __find_page(inode, offset, *hash);
|
page = __find_page(inode, offset, *hash);
|
if (!page)
|
if (!page)
|
goto no_cached_page;
|
goto no_cached_page;
|
|
|
found_page:
|
found_page:
|
/*
|
/*
|
* Ok, found a page in the page cache, now we need to check
|
* Ok, found a page in the page cache, now we need to check
|
* that it's up-to-date. First check whether we'll need an
|
* that it's up-to-date. First check whether we'll need an
|
* extra page -- better to overlap the allocation with the I/O.
|
* extra page -- better to overlap the allocation with the I/O.
|
*/
|
*/
|
if (no_share && !new_page) {
|
if (no_share && !new_page) {
|
new_page = __get_free_page(GFP_KERNEL);
|
new_page = __get_free_page(GFP_KERNEL);
|
if (!new_page)
|
if (!new_page)
|
goto failure;
|
goto failure;
|
}
|
}
|
|
|
if (PageLocked(page))
|
if (PageLocked(page))
|
goto page_locked_wait;
|
goto page_locked_wait;
|
if (!PageUptodate(page))
|
if (!PageUptodate(page))
|
goto page_read_error;
|
goto page_read_error;
|
|
|
success:
|
success:
|
/*
|
/*
|
* Found the page, need to check sharing and possibly
|
* Found the page, need to check sharing and possibly
|
* copy it over to another page..
|
* copy it over to another page..
|
*/
|
*/
|
old_page = page_address(page);
|
old_page = page_address(page);
|
if (!no_share) {
|
if (!no_share) {
|
/*
|
/*
|
* Ok, we can share the cached page directly.. Get rid
|
* Ok, we can share the cached page directly.. Get rid
|
* of any potential extra pages.
|
* of any potential extra pages.
|
*/
|
*/
|
if (new_page)
|
if (new_page)
|
free_page(new_page);
|
free_page(new_page);
|
|
|
flush_page_to_ram(old_page);
|
flush_page_to_ram(old_page);
|
return old_page;
|
return old_page;
|
}
|
}
|
|
|
/*
|
/*
|
* No sharing ... copy to the new page.
|
* No sharing ... copy to the new page.
|
*/
|
*/
|
memcpy((void *) new_page, (void *) old_page, PAGE_SIZE);
|
memcpy((void *) new_page, (void *) old_page, PAGE_SIZE);
|
flush_page_to_ram(new_page);
|
flush_page_to_ram(new_page);
|
release_page(page);
|
release_page(page);
|
return new_page;
|
return new_page;
|
|
|
no_cached_page:
|
no_cached_page:
|
new_page = __get_free_page(GFP_KERNEL);
|
new_page = __get_free_page(GFP_KERNEL);
|
if (!new_page)
|
if (!new_page)
|
goto no_page;
|
goto no_page;
|
|
|
/*
|
/*
|
* During getting the above page we might have slept,
|
* During getting the above page we might have slept,
|
* so we need to re-check the situation with the page
|
* so we need to re-check the situation with the page
|
* cache.. The page we just got may be useful if we
|
* cache.. The page we just got may be useful if we
|
* can't share, so don't get rid of it here.
|
* can't share, so don't get rid of it here.
|
*/
|
*/
|
page = find_page(inode, offset);
|
page = find_page(inode, offset);
|
if (page)
|
if (page)
|
goto found_page;
|
goto found_page;
|
|
|
/*
|
/*
|
* Now, create a new page-cache page from the page we got
|
* Now, create a new page-cache page from the page we got
|
*/
|
*/
|
page = mem_map + MAP_NR(new_page);
|
page = mem_map + MAP_NR(new_page);
|
new_page = 0;
|
new_page = 0;
|
add_to_page_cache(page, inode, offset, hash);
|
add_to_page_cache(page, inode, offset, hash);
|
|
|
if (inode->i_op->readpage(inode, page) != 0)
|
if (inode->i_op->readpage(inode, page) != 0)
|
goto failure;
|
goto failure;
|
|
|
/*
|
/*
|
* Do a very limited read-ahead if appropriate
|
* Do a very limited read-ahead if appropriate
|
*/
|
*/
|
if (PageLocked(page))
|
if (PageLocked(page))
|
new_page = try_to_read_ahead(inode, offset + PAGE_SIZE, 0);
|
new_page = try_to_read_ahead(inode, offset + PAGE_SIZE, 0);
|
goto found_page;
|
goto found_page;
|
|
|
page_locked_wait:
|
page_locked_wait:
|
__wait_on_page(page);
|
__wait_on_page(page);
|
if (PageUptodate(page))
|
if (PageUptodate(page))
|
goto success;
|
goto success;
|
|
|
page_read_error:
|
page_read_error:
|
/*
|
/*
|
* Umm, take care of errors if the page isn't up-to-date.
|
* Umm, take care of errors if the page isn't up-to-date.
|
* Try to re-read it _once_. We do this synchronously,
|
* Try to re-read it _once_. We do this synchronously,
|
* because there really aren't any performance issues here
|
* because there really aren't any performance issues here
|
* and we need to check for errors.
|
* and we need to check for errors.
|
*/
|
*/
|
if (inode->i_op->readpage(inode, page) != 0)
|
if (inode->i_op->readpage(inode, page) != 0)
|
goto failure;
|
goto failure;
|
wait_on_page(page);
|
wait_on_page(page);
|
if (PageError(page))
|
if (PageError(page))
|
goto failure;
|
goto failure;
|
if (PageUptodate(page))
|
if (PageUptodate(page))
|
goto success;
|
goto success;
|
|
|
/*
|
/*
|
* Uhhuh.. Things didn't work out. Return zero to tell the
|
* Uhhuh.. Things didn't work out. Return zero to tell the
|
* mm layer so, possibly freeing the page cache page first.
|
* mm layer so, possibly freeing the page cache page first.
|
*/
|
*/
|
failure:
|
failure:
|
release_page(page);
|
release_page(page);
|
if (new_page)
|
if (new_page)
|
free_page(new_page);
|
free_page(new_page);
|
no_page:
|
no_page:
|
return 0;
|
return 0;
|
}
|
}
|
|
|
/*
|
/*
|
* Tries to write a shared mapped page to its backing store. May return -EIO
|
* Tries to write a shared mapped page to its backing store. May return -EIO
|
* if the disk is full.
|
* if the disk is full.
|
*/
|
*/
|
static inline int do_write_page(struct inode * inode, struct file * file,
|
static inline int do_write_page(struct inode * inode, struct file * file,
|
const char * page, unsigned long offset)
|
const char * page, unsigned long offset)
|
{
|
{
|
int old_fs, retval;
|
int old_fs, retval;
|
unsigned long size;
|
unsigned long size;
|
|
|
size = offset + PAGE_SIZE;
|
size = offset + PAGE_SIZE;
|
/* refuse to extend file size.. */
|
/* refuse to extend file size.. */
|
if (S_ISREG(inode->i_mode)) {
|
if (S_ISREG(inode->i_mode)) {
|
if (size > inode->i_size)
|
if (size > inode->i_size)
|
size = inode->i_size;
|
size = inode->i_size;
|
/* Ho humm.. We should have tested for this earlier */
|
/* Ho humm.. We should have tested for this earlier */
|
if (size < offset)
|
if (size < offset)
|
return -EIO;
|
return -EIO;
|
}
|
}
|
size -= offset;
|
size -= offset;
|
old_fs = get_fs();
|
old_fs = get_fs();
|
set_fs(KERNEL_DS);
|
set_fs(KERNEL_DS);
|
retval = -EIO;
|
retval = -EIO;
|
if (size == file->f_op->write(inode, file, (const char *) page, size))
|
if (size == file->f_op->write(inode, file, (const char *) page, size))
|
retval = 0;
|
retval = 0;
|
set_fs(old_fs);
|
set_fs(old_fs);
|
return retval;
|
return retval;
|
}
|
}
|
|
|
static int filemap_write_page(struct vm_area_struct * vma,
|
static int filemap_write_page(struct vm_area_struct * vma,
|
unsigned long offset,
|
unsigned long offset,
|
unsigned long page)
|
unsigned long page)
|
{
|
{
|
int result;
|
int result;
|
struct file file;
|
struct file file;
|
struct inode * inode;
|
struct inode * inode;
|
struct buffer_head * bh;
|
struct buffer_head * bh;
|
|
|
bh = mem_map[MAP_NR(page)].buffers;
|
bh = mem_map[MAP_NR(page)].buffers;
|
if (bh) {
|
if (bh) {
|
/* whee.. just mark the buffer heads dirty */
|
/* whee.. just mark the buffer heads dirty */
|
struct buffer_head * tmp = bh;
|
struct buffer_head * tmp = bh;
|
do {
|
do {
|
mark_buffer_dirty(tmp, 0);
|
mark_buffer_dirty(tmp, 0);
|
tmp = tmp->b_this_page;
|
tmp = tmp->b_this_page;
|
} while (tmp != bh);
|
} while (tmp != bh);
|
return 0;
|
return 0;
|
}
|
}
|
|
|
inode = vma->vm_inode;
|
inode = vma->vm_inode;
|
file.f_op = inode->i_op->default_file_ops;
|
file.f_op = inode->i_op->default_file_ops;
|
if (!file.f_op->write)
|
if (!file.f_op->write)
|
return -EIO;
|
return -EIO;
|
file.f_mode = 3;
|
file.f_mode = 3;
|
file.f_flags = 0;
|
file.f_flags = 0;
|
file.f_count = 1;
|
file.f_count = 1;
|
file.f_inode = inode;
|
file.f_inode = inode;
|
file.f_pos = offset;
|
file.f_pos = offset;
|
file.f_reada = 0;
|
file.f_reada = 0;
|
|
|
down(&inode->i_sem);
|
down(&inode->i_sem);
|
result = do_write_page(inode, &file, (const char *) page, offset);
|
result = do_write_page(inode, &file, (const char *) page, offset);
|
up(&inode->i_sem);
|
up(&inode->i_sem);
|
return result;
|
return result;
|
}
|
}
|
|
|
|
|
/*
|
/*
|
* Swapping to a shared file: while we're busy writing out the page
|
* Swapping to a shared file: while we're busy writing out the page
|
* (and the page still exists in memory), we save the page information
|
* (and the page still exists in memory), we save the page information
|
* in the page table, so that "filemap_swapin()" can re-use the page
|
* in the page table, so that "filemap_swapin()" can re-use the page
|
* immediately if it is called while we're busy swapping it out..
|
* immediately if it is called while we're busy swapping it out..
|
*
|
*
|
* Once we've written it all out, we mark the page entry "empty", which
|
* Once we've written it all out, we mark the page entry "empty", which
|
* will result in a normal page-in (instead of a swap-in) from the now
|
* will result in a normal page-in (instead of a swap-in) from the now
|
* up-to-date disk file.
|
* up-to-date disk file.
|
*/
|
*/
|
int filemap_swapout(struct vm_area_struct * vma,
|
int filemap_swapout(struct vm_area_struct * vma,
|
unsigned long offset,
|
unsigned long offset,
|
pte_t *page_table)
|
pte_t *page_table)
|
{
|
{
|
int error;
|
int error;
|
unsigned long page = pte_page(*page_table);
|
unsigned long page = pte_page(*page_table);
|
unsigned long entry = SWP_ENTRY(SHM_SWP_TYPE, MAP_NR(page));
|
unsigned long entry = SWP_ENTRY(SHM_SWP_TYPE, MAP_NR(page));
|
|
|
flush_cache_page(vma, (offset + vma->vm_start - vma->vm_offset));
|
flush_cache_page(vma, (offset + vma->vm_start - vma->vm_offset));
|
set_pte(page_table, __pte(entry));
|
set_pte(page_table, __pte(entry));
|
flush_tlb_page(vma, (offset + vma->vm_start - vma->vm_offset));
|
flush_tlb_page(vma, (offset + vma->vm_start - vma->vm_offset));
|
error = filemap_write_page(vma, offset, page);
|
error = filemap_write_page(vma, offset, page);
|
if (pte_val(*page_table) == entry)
|
if (pte_val(*page_table) == entry)
|
pte_clear(page_table);
|
pte_clear(page_table);
|
return error;
|
return error;
|
}
|
}
|
|
|
/*
|
/*
|
* filemap_swapin() is called only if we have something in the page
|
* filemap_swapin() is called only if we have something in the page
|
* tables that is non-zero (but not present), which we know to be the
|
* tables that is non-zero (but not present), which we know to be the
|
* page index of a page that is busy being swapped out (see above).
|
* page index of a page that is busy being swapped out (see above).
|
* So we just use it directly..
|
* So we just use it directly..
|
*/
|
*/
|
static pte_t filemap_swapin(struct vm_area_struct * vma,
|
static pte_t filemap_swapin(struct vm_area_struct * vma,
|
unsigned long offset,
|
unsigned long offset,
|
unsigned long entry)
|
unsigned long entry)
|
{
|
{
|
unsigned long page = SWP_OFFSET(entry);
|
unsigned long page = SWP_OFFSET(entry);
|
|
|
mem_map[page].count++;
|
mem_map[page].count++;
|
page = (page << PAGE_SHIFT) + PAGE_OFFSET;
|
page = (page << PAGE_SHIFT) + PAGE_OFFSET;
|
return mk_pte(page,vma->vm_page_prot);
|
return mk_pte(page,vma->vm_page_prot);
|
}
|
}
|
|
|
|
|
static inline int filemap_sync_pte(pte_t * ptep, struct vm_area_struct *vma,
|
static inline int filemap_sync_pte(pte_t * ptep, struct vm_area_struct *vma,
|
unsigned long address, unsigned int flags)
|
unsigned long address, unsigned int flags)
|
{
|
{
|
pte_t pte = *ptep;
|
pte_t pte = *ptep;
|
unsigned long page;
|
unsigned long page;
|
int error;
|
int error;
|
|
|
if (pte_none(pte))
|
if (pte_none(pte))
|
return 0;
|
return 0;
|
if (!(flags & MS_INVALIDATE)) {
|
if (!(flags & MS_INVALIDATE)) {
|
if (!pte_present(pte))
|
if (!pte_present(pte))
|
return 0;
|
return 0;
|
if (!pte_dirty(pte))
|
if (!pte_dirty(pte))
|
return 0;
|
return 0;
|
flush_page_to_ram(pte_page(pte));
|
flush_page_to_ram(pte_page(pte));
|
flush_cache_page(vma, address);
|
flush_cache_page(vma, address);
|
set_pte(ptep, pte_mkclean(pte));
|
set_pte(ptep, pte_mkclean(pte));
|
flush_tlb_page(vma, address);
|
flush_tlb_page(vma, address);
|
page = pte_page(pte);
|
page = pte_page(pte);
|
mem_map[MAP_NR(page)].count++;
|
mem_map[MAP_NR(page)].count++;
|
} else {
|
} else {
|
flush_cache_page(vma, address);
|
flush_cache_page(vma, address);
|
pte_clear(ptep);
|
pte_clear(ptep);
|
flush_tlb_page(vma, address);
|
flush_tlb_page(vma, address);
|
if (!pte_present(pte)) {
|
if (!pte_present(pte)) {
|
swap_free(pte_val(pte));
|
swap_free(pte_val(pte));
|
return 0;
|
return 0;
|
}
|
}
|
page = pte_page(pte);
|
page = pte_page(pte);
|
if (!pte_dirty(pte) || flags == MS_INVALIDATE) {
|
if (!pte_dirty(pte) || flags == MS_INVALIDATE) {
|
free_page(page);
|
free_page(page);
|
return 0;
|
return 0;
|
}
|
}
|
}
|
}
|
error = filemap_write_page(vma, address - vma->vm_start + vma->vm_offset, page);
|
error = filemap_write_page(vma, address - vma->vm_start + vma->vm_offset, page);
|
free_page(page);
|
free_page(page);
|
return error;
|
return error;
|
}
|
}
|
|
|
static inline int filemap_sync_pte_range(pmd_t * pmd,
|
static inline int filemap_sync_pte_range(pmd_t * pmd,
|
unsigned long address, unsigned long size,
|
unsigned long address, unsigned long size,
|
struct vm_area_struct *vma, unsigned long offset, unsigned int flags)
|
struct vm_area_struct *vma, unsigned long offset, unsigned int flags)
|
{
|
{
|
pte_t * pte;
|
pte_t * pte;
|
unsigned long end;
|
unsigned long end;
|
int error;
|
int error;
|
|
|
if (pmd_none(*pmd))
|
if (pmd_none(*pmd))
|
return 0;
|
return 0;
|
if (pmd_bad(*pmd)) {
|
if (pmd_bad(*pmd)) {
|
printk("filemap_sync_pte_range: bad pmd (%08lx)\n", pmd_val(*pmd));
|
printk("filemap_sync_pte_range: bad pmd (%08lx)\n", pmd_val(*pmd));
|
pmd_clear(pmd);
|
pmd_clear(pmd);
|
return 0;
|
return 0;
|
}
|
}
|
pte = pte_offset(pmd, address);
|
pte = pte_offset(pmd, address);
|
offset += address & PMD_MASK;
|
offset += address & PMD_MASK;
|
address &= ~PMD_MASK;
|
address &= ~PMD_MASK;
|
end = address + size;
|
end = address + size;
|
if (end > PMD_SIZE)
|
if (end > PMD_SIZE)
|
end = PMD_SIZE;
|
end = PMD_SIZE;
|
error = 0;
|
error = 0;
|
do {
|
do {
|
error |= filemap_sync_pte(pte, vma, address + offset, flags);
|
error |= filemap_sync_pte(pte, vma, address + offset, flags);
|
address += PAGE_SIZE;
|
address += PAGE_SIZE;
|
pte++;
|
pte++;
|
} while (address < end);
|
} while (address < end);
|
return error;
|
return error;
|
}
|
}
|
|
|
static inline int filemap_sync_pmd_range(pgd_t * pgd,
|
static inline int filemap_sync_pmd_range(pgd_t * pgd,
|
unsigned long address, unsigned long size,
|
unsigned long address, unsigned long size,
|
struct vm_area_struct *vma, unsigned int flags)
|
struct vm_area_struct *vma, unsigned int flags)
|
{
|
{
|
pmd_t * pmd;
|
pmd_t * pmd;
|
unsigned long offset, end;
|
unsigned long offset, end;
|
int error;
|
int error;
|
|
|
if (pgd_none(*pgd))
|
if (pgd_none(*pgd))
|
return 0;
|
return 0;
|
if (pgd_bad(*pgd)) {
|
if (pgd_bad(*pgd)) {
|
printk("filemap_sync_pmd_range: bad pgd (%08lx)\n", pgd_val(*pgd));
|
printk("filemap_sync_pmd_range: bad pgd (%08lx)\n", pgd_val(*pgd));
|
pgd_clear(pgd);
|
pgd_clear(pgd);
|
return 0;
|
return 0;
|
}
|
}
|
pmd = pmd_offset(pgd, address);
|
pmd = pmd_offset(pgd, address);
|
offset = address & PGDIR_MASK;
|
offset = address & PGDIR_MASK;
|
address &= ~PGDIR_MASK;
|
address &= ~PGDIR_MASK;
|
end = address + size;
|
end = address + size;
|
if (end > PGDIR_SIZE)
|
if (end > PGDIR_SIZE)
|
end = PGDIR_SIZE;
|
end = PGDIR_SIZE;
|
error = 0;
|
error = 0;
|
do {
|
do {
|
error |= filemap_sync_pte_range(pmd, address, end - address, vma, offset, flags);
|
error |= filemap_sync_pte_range(pmd, address, end - address, vma, offset, flags);
|
address = (address + PMD_SIZE) & PMD_MASK;
|
address = (address + PMD_SIZE) & PMD_MASK;
|
pmd++;
|
pmd++;
|
} while (address < end);
|
} while (address < end);
|
return error;
|
return error;
|
}
|
}
|
|
|
static int filemap_sync(struct vm_area_struct * vma, unsigned long address,
|
static int filemap_sync(struct vm_area_struct * vma, unsigned long address,
|
size_t size, unsigned int flags)
|
size_t size, unsigned int flags)
|
{
|
{
|
pgd_t * dir;
|
pgd_t * dir;
|
unsigned long end = address + size;
|
unsigned long end = address + size;
|
int error = 0;
|
int error = 0;
|
|
|
dir = pgd_offset(vma->vm_mm, address);
|
dir = pgd_offset(vma->vm_mm, address);
|
flush_cache_range(vma->vm_mm, end - size, end);
|
flush_cache_range(vma->vm_mm, end - size, end);
|
while (address < end) {
|
while (address < end) {
|
error |= filemap_sync_pmd_range(dir, address, end - address, vma, flags);
|
error |= filemap_sync_pmd_range(dir, address, end - address, vma, flags);
|
address = (address + PGDIR_SIZE) & PGDIR_MASK;
|
address = (address + PGDIR_SIZE) & PGDIR_MASK;
|
dir++;
|
dir++;
|
}
|
}
|
flush_tlb_range(vma->vm_mm, end - size, end);
|
flush_tlb_range(vma->vm_mm, end - size, end);
|
return error;
|
return error;
|
}
|
}
|
|
|
/*
|
/*
|
* This handles (potentially partial) area unmaps..
|
* This handles (potentially partial) area unmaps..
|
*/
|
*/
|
static void filemap_unmap(struct vm_area_struct *vma, unsigned long start, size_t len)
|
static void filemap_unmap(struct vm_area_struct *vma, unsigned long start, size_t len)
|
{
|
{
|
filemap_sync(vma, start, len, MS_ASYNC);
|
filemap_sync(vma, start, len, MS_ASYNC);
|
}
|
}
|
|
|
/*
|
/*
|
* Shared mappings need to be able to do the right thing at
|
* Shared mappings need to be able to do the right thing at
|
* close/unmap/sync. They will also use the private file as
|
* close/unmap/sync. They will also use the private file as
|
* backing-store for swapping..
|
* backing-store for swapping..
|
*/
|
*/
|
static struct vm_operations_struct file_shared_mmap = {
|
static struct vm_operations_struct file_shared_mmap = {
|
NULL, /* no special open */
|
NULL, /* no special open */
|
NULL, /* no special close */
|
NULL, /* no special close */
|
filemap_unmap, /* unmap - we need to sync the pages */
|
filemap_unmap, /* unmap - we need to sync the pages */
|
NULL, /* no special protect */
|
NULL, /* no special protect */
|
filemap_sync, /* sync */
|
filemap_sync, /* sync */
|
NULL, /* advise */
|
NULL, /* advise */
|
filemap_nopage, /* nopage */
|
filemap_nopage, /* nopage */
|
NULL, /* wppage */
|
NULL, /* wppage */
|
filemap_swapout, /* swapout */
|
filemap_swapout, /* swapout */
|
filemap_swapin, /* swapin */
|
filemap_swapin, /* swapin */
|
};
|
};
|
|
|
/*
|
/*
|
* Private mappings just need to be able to load in the map.
|
* Private mappings just need to be able to load in the map.
|
*
|
*
|
* (This is actually used for shared mappings as well, if we
|
* (This is actually used for shared mappings as well, if we
|
* know they can't ever get write permissions..)
|
* know they can't ever get write permissions..)
|
*/
|
*/
|
static struct vm_operations_struct file_private_mmap = {
|
static struct vm_operations_struct file_private_mmap = {
|
NULL, /* open */
|
NULL, /* open */
|
NULL, /* close */
|
NULL, /* close */
|
NULL, /* unmap */
|
NULL, /* unmap */
|
NULL, /* protect */
|
NULL, /* protect */
|
NULL, /* sync */
|
NULL, /* sync */
|
NULL, /* advise */
|
NULL, /* advise */
|
filemap_nopage, /* nopage */
|
filemap_nopage, /* nopage */
|
NULL, /* wppage */
|
NULL, /* wppage */
|
NULL, /* swapout */
|
NULL, /* swapout */
|
NULL, /* swapin */
|
NULL, /* swapin */
|
};
|
};
|
|
|
/* This is used for a general mmap of a disk file */
|
/* This is used for a general mmap of a disk file */
|
int generic_file_mmap(struct inode * inode, struct file * file, struct vm_area_struct * vma)
|
int generic_file_mmap(struct inode * inode, struct file * file, struct vm_area_struct * vma)
|
{
|
{
|
struct vm_operations_struct * ops;
|
struct vm_operations_struct * ops;
|
|
|
if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE)) {
|
if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE)) {
|
ops = &file_shared_mmap;
|
ops = &file_shared_mmap;
|
/* share_page() can only guarantee proper page sharing if
|
/* share_page() can only guarantee proper page sharing if
|
* the offsets are all page aligned. */
|
* the offsets are all page aligned. */
|
if (vma->vm_offset & (PAGE_SIZE - 1))
|
if (vma->vm_offset & (PAGE_SIZE - 1))
|
return -EINVAL;
|
return -EINVAL;
|
} else {
|
} else {
|
ops = &file_private_mmap;
|
ops = &file_private_mmap;
|
if (vma->vm_offset & (inode->i_sb->s_blocksize - 1))
|
if (vma->vm_offset & (inode->i_sb->s_blocksize - 1))
|
return -EINVAL;
|
return -EINVAL;
|
}
|
}
|
if (!inode->i_sb || !S_ISREG(inode->i_mode))
|
if (!inode->i_sb || !S_ISREG(inode->i_mode))
|
return -EACCES;
|
return -EACCES;
|
if (!inode->i_op || !inode->i_op->readpage)
|
if (!inode->i_op || !inode->i_op->readpage)
|
return -ENOEXEC;
|
return -ENOEXEC;
|
UPDATE_ATIME(inode)
|
UPDATE_ATIME(inode)
|
vma->vm_inode = inode;
|
vma->vm_inode = inode;
|
inode->i_count++;
|
inode->i_count++;
|
vma->vm_ops = ops;
|
vma->vm_ops = ops;
|
return 0;
|
return 0;
|
}
|
}
|
|
|
|
|
/*
|
/*
|
* The msync() system call.
|
* The msync() system call.
|
*/
|
*/
|
|
|
static int msync_interval(struct vm_area_struct * vma,
|
static int msync_interval(struct vm_area_struct * vma,
|
unsigned long start, unsigned long end, int flags)
|
unsigned long start, unsigned long end, int flags)
|
{
|
{
|
if (vma->vm_inode && vma->vm_ops && vma->vm_ops->sync) {
|
if (vma->vm_inode && vma->vm_ops && vma->vm_ops->sync) {
|
int error;
|
int error;
|
error = vma->vm_ops->sync(vma, start, end-start, flags);
|
error = vma->vm_ops->sync(vma, start, end-start, flags);
|
if (error)
|
if (error)
|
return error;
|
return error;
|
if (flags & MS_SYNC)
|
if (flags & MS_SYNC)
|
return file_fsync(vma->vm_inode, NULL);
|
return file_fsync(vma->vm_inode, NULL);
|
return 0;
|
return 0;
|
}
|
}
|
return 0;
|
return 0;
|
}
|
}
|
|
|
asmlinkage int sys_msync(unsigned long start, size_t len, int flags)
|
asmlinkage int sys_msync(unsigned long start, size_t len, int flags)
|
{
|
{
|
unsigned long end;
|
unsigned long end;
|
struct vm_area_struct * vma;
|
struct vm_area_struct * vma;
|
int unmapped_error, error;
|
int unmapped_error, error;
|
|
|
if (start & ~PAGE_MASK)
|
if (start & ~PAGE_MASK)
|
return -EINVAL;
|
return -EINVAL;
|
len = (len + ~PAGE_MASK) & PAGE_MASK;
|
len = (len + ~PAGE_MASK) & PAGE_MASK;
|
end = start + len;
|
end = start + len;
|
if (end < start)
|
if (end < start)
|
return -EINVAL;
|
return -EINVAL;
|
if (flags & ~(MS_ASYNC | MS_INVALIDATE | MS_SYNC))
|
if (flags & ~(MS_ASYNC | MS_INVALIDATE | MS_SYNC))
|
return -EINVAL;
|
return -EINVAL;
|
if (end == start)
|
if (end == start)
|
return 0;
|
return 0;
|
/*
|
/*
|
* If the interval [start,end) covers some unmapped address ranges,
|
* If the interval [start,end) covers some unmapped address ranges,
|
* just ignore them, but return -EFAULT at the end.
|
* just ignore them, but return -EFAULT at the end.
|
*/
|
*/
|
vma = find_vma(current->mm, start);
|
vma = find_vma(current->mm, start);
|
unmapped_error = 0;
|
unmapped_error = 0;
|
for (;;) {
|
for (;;) {
|
/* Still start < end. */
|
/* Still start < end. */
|
if (!vma)
|
if (!vma)
|
return -EFAULT;
|
return -EFAULT;
|
/* Here start < vma->vm_end. */
|
/* Here start < vma->vm_end. */
|
if (start < vma->vm_start) {
|
if (start < vma->vm_start) {
|
unmapped_error = -EFAULT;
|
unmapped_error = -EFAULT;
|
start = vma->vm_start;
|
start = vma->vm_start;
|
}
|
}
|
/* Here vma->vm_start <= start < vma->vm_end. */
|
/* Here vma->vm_start <= start < vma->vm_end. */
|
if (end <= vma->vm_end) {
|
if (end <= vma->vm_end) {
|
if (start < end) {
|
if (start < end) {
|
error = msync_interval(vma, start, end, flags);
|
error = msync_interval(vma, start, end, flags);
|
if (error)
|
if (error)
|
return error;
|
return error;
|
}
|
}
|
return unmapped_error;
|
return unmapped_error;
|
}
|
}
|
/* Here vma->vm_start <= start < vma->vm_end < end. */
|
/* Here vma->vm_start <= start < vma->vm_end < end. */
|
error = msync_interval(vma, start, vma->vm_end, flags);
|
error = msync_interval(vma, start, vma->vm_end, flags);
|
if (error)
|
if (error)
|
return error;
|
return error;
|
start = vma->vm_end;
|
start = vma->vm_end;
|
vma = vma->vm_next;
|
vma = vma->vm_next;
|
}
|
}
|
}
|
}
|
|
|