mirror of
https://github.com/Maschell/libntfs-wiiu.git
synced 2024-11-16 09:19:16 +01:00
375 lines
11 KiB
C
375 lines
11 KiB
C
/*
|
|
cache.c
|
|
The cache is not visible to the user. It should be flushed
|
|
when any file is closed or changes are made to the filesystem.
|
|
|
|
This cache implements a least-used-page replacement policy. This will
|
|
distribute sectors evenly over the pages, so if less than the maximum
|
|
pages are used at once, they should all eventually remain in the cache.
|
|
This also has the benefit of throwing out old sectors, so as not to keep
|
|
too many stale pages around.
|
|
|
|
Copyright (c) 2006 Michael "Chishm" Chisholm
|
|
Copyright (c) 2009 shareese, rodries
|
|
Copyright (c) 2010 Dimok
|
|
|
|
Redistribution and use in source and binary forms, with or without modification,
|
|
are permitted provided that the following conditions are met:
|
|
|
|
1. Redistributions of source code must retain the above copyright notice,
|
|
this list of conditions and the following disclaimer.
|
|
2. Redistributions in binary form must reproduce the above copyright notice,
|
|
this list of conditions and the following disclaimer in the documentation and/or
|
|
other materials provided with the distribution.
|
|
3. The name of the author may not be used to endorse or promote products derived
|
|
from this software without specific prior written permission.
|
|
|
|
THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
|
|
WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
|
|
AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE
|
|
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
|
|
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
|
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
|
|
EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
*/
|
|
|
|
#include <ogc/lwp_watchdog.h>
|
|
#include <string.h>
|
|
#include <limits.h>
|
|
|
|
#include "cache2.h"
|
|
#include "bit_ops.h"
|
|
#include "mem_allocate.h"
|
|
|
|
#define CACHE_FREE UINT_MAX
|
|
|
|
NTFS_CACHE* _NTFS_cache_constructor (unsigned int numberOfPages, unsigned int sectorsPerPage, const DISC_INTERFACE* discInterface, sec_t endOfPartition, sec_t sectorSize) {
|
|
NTFS_CACHE* cache;
|
|
unsigned int i;
|
|
NTFS_CACHE_ENTRY* cacheEntries;
|
|
|
|
if(numberOfPages==0 || sectorsPerPage==0) return NULL;
|
|
|
|
if (numberOfPages < 4) {
|
|
numberOfPages = 4;
|
|
}
|
|
|
|
if (sectorsPerPage < 32) {
|
|
sectorsPerPage = 32;
|
|
}
|
|
|
|
cache = (NTFS_CACHE*) ntfs_alloc (sizeof(NTFS_CACHE));
|
|
if (cache == NULL) {
|
|
return NULL;
|
|
}
|
|
|
|
cache->disc = discInterface;
|
|
cache->endOfPartition = endOfPartition;
|
|
cache->numberOfPages = numberOfPages;
|
|
cache->sectorsPerPage = sectorsPerPage;
|
|
cache->sectorSize = sectorSize;
|
|
|
|
|
|
cacheEntries = (NTFS_CACHE_ENTRY*) ntfs_alloc ( sizeof(NTFS_CACHE_ENTRY) * numberOfPages);
|
|
if (cacheEntries == NULL) {
|
|
ntfs_free (cache);
|
|
return NULL;
|
|
}
|
|
|
|
for (i = 0; i < numberOfPages; i++) {
|
|
cacheEntries[i].sector = CACHE_FREE;
|
|
cacheEntries[i].count = 0;
|
|
cacheEntries[i].last_access = 0;
|
|
cacheEntries[i].dirty = false;
|
|
cacheEntries[i].cache = (uint8_t*) ntfs_align ( sectorsPerPage * cache->sectorSize );
|
|
}
|
|
|
|
cache->cacheEntries = cacheEntries;
|
|
|
|
return cache;
|
|
}
|
|
|
|
void _NTFS_cache_destructor (NTFS_CACHE* cache) {
|
|
unsigned int i;
|
|
|
|
if(cache==NULL) return;
|
|
|
|
// Clear out cache before destroying it
|
|
_NTFS_cache_flush(cache);
|
|
|
|
// Free memory in reverse allocation order
|
|
for (i = 0; i < cache->numberOfPages; i++) {
|
|
ntfs_free (cache->cacheEntries[i].cache);
|
|
}
|
|
ntfs_free (cache->cacheEntries);
|
|
ntfs_free (cache);
|
|
}
|
|
|
|
static u32 accessCounter = 0;
|
|
|
|
static u32 accessTime(){
|
|
accessCounter++;
|
|
return accessCounter;
|
|
}
|
|
|
|
static NTFS_CACHE_ENTRY* _NTFS_cache_getPage(NTFS_CACHE *cache,sec_t sector)
|
|
{
|
|
unsigned int i;
|
|
NTFS_CACHE_ENTRY* cacheEntries = cache->cacheEntries;
|
|
unsigned int numberOfPages = cache->numberOfPages;
|
|
unsigned int sectorsPerPage = cache->sectorsPerPage;
|
|
|
|
bool foundFree = false;
|
|
unsigned int oldUsed = 0;
|
|
unsigned int oldAccess = UINT_MAX;
|
|
|
|
for(i=0;i<numberOfPages;i++) {
|
|
if(sector>=cacheEntries[i].sector && sector<(cacheEntries[i].sector + cacheEntries[i].count)) {
|
|
cacheEntries[i].last_access = accessTime();
|
|
return &(cacheEntries[i]);
|
|
}
|
|
|
|
if(foundFree==false && (cacheEntries[i].sector==CACHE_FREE || cacheEntries[i].last_access<oldAccess)) {
|
|
if(cacheEntries[i].sector==CACHE_FREE) foundFree = true;
|
|
oldUsed = i;
|
|
oldAccess = cacheEntries[i].last_access;
|
|
}
|
|
}
|
|
|
|
if(foundFree==false && cacheEntries[oldUsed].dirty==true) {
|
|
if(!cache->disc->writeSectors(cacheEntries[oldUsed].sector,cacheEntries[oldUsed].count,cacheEntries[oldUsed].cache)) return NULL;
|
|
cacheEntries[oldUsed].dirty = false;
|
|
}
|
|
sector = (sector/sectorsPerPage)*sectorsPerPage; // align base sector to page size
|
|
sec_t next_page = sector + sectorsPerPage;
|
|
if(next_page > cache->endOfPartition) next_page = cache->endOfPartition;
|
|
|
|
if(!cache->disc->readSectors(sector,next_page-sector,cacheEntries[oldUsed].cache)) return NULL;
|
|
|
|
cacheEntries[oldUsed].sector = sector;
|
|
cacheEntries[oldUsed].count = next_page-sector;
|
|
cacheEntries[oldUsed].last_access = accessTime();
|
|
|
|
return &(cacheEntries[oldUsed]);
|
|
}
|
|
|
|
static NTFS_CACHE_ENTRY* _NTFS_cache_findPage(NTFS_CACHE *cache, sec_t sector, sec_t count) {
|
|
|
|
unsigned int i;
|
|
NTFS_CACHE_ENTRY* cacheEntries = cache->cacheEntries;
|
|
unsigned int numberOfPages = cache->numberOfPages;
|
|
NTFS_CACHE_ENTRY *entry = NULL;
|
|
sec_t lowest = UINT_MAX;
|
|
|
|
for(i=0;i<numberOfPages;i++) {
|
|
if (cacheEntries[i].sector != CACHE_FREE) {
|
|
bool intersect;
|
|
if (sector > cacheEntries[i].sector) {
|
|
intersect = sector - cacheEntries[i].sector < cacheEntries[i].count;
|
|
} else {
|
|
intersect = cacheEntries[i].sector - sector < count;
|
|
}
|
|
|
|
if ( intersect && (cacheEntries[i].sector < lowest)) {
|
|
lowest = cacheEntries[i].sector;
|
|
entry = &cacheEntries[i];
|
|
}
|
|
}
|
|
}
|
|
|
|
return entry;
|
|
}
|
|
|
|
bool _NTFS_cache_readSectors(NTFS_CACHE *cache,sec_t sector,sec_t numSectors,void *buffer)
|
|
{
|
|
sec_t sec;
|
|
sec_t secs_to_read;
|
|
NTFS_CACHE_ENTRY *entry;
|
|
uint8_t *dest = buffer;
|
|
|
|
while(numSectors>0) {
|
|
entry = _NTFS_cache_getPage(cache,sector);
|
|
if(entry==NULL) return false;
|
|
|
|
sec = sector - entry->sector;
|
|
secs_to_read = entry->count - sec;
|
|
if(secs_to_read>numSectors) secs_to_read = numSectors;
|
|
|
|
memcpy(dest,entry->cache + (sec*cache->sectorSize),(secs_to_read*cache->sectorSize));
|
|
|
|
dest += (secs_to_read*cache->sectorSize);
|
|
sector += secs_to_read;
|
|
numSectors -= secs_to_read;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
/*
|
|
Reads some data from a cache page, determined by the sector number
|
|
*/
|
|
|
|
bool _NTFS_cache_readPartialSector (NTFS_CACHE* cache, void* buffer, sec_t sector, unsigned int offset, size_t size)
|
|
{
|
|
sec_t sec;
|
|
NTFS_CACHE_ENTRY *entry;
|
|
|
|
if (offset + size > cache->sectorSize) return false;
|
|
|
|
entry = _NTFS_cache_getPage(cache,sector);
|
|
if(entry==NULL) return false;
|
|
|
|
sec = sector - entry->sector;
|
|
memcpy(buffer,entry->cache + ((sec*cache->sectorSize) + offset),size);
|
|
|
|
return true;
|
|
}
|
|
|
|
bool _NTFS_cache_readLittleEndianValue (NTFS_CACHE* cache, uint32_t *value, sec_t sector, unsigned int offset, int num_bytes) {
|
|
uint8_t buf[4];
|
|
if (!_NTFS_cache_readPartialSector(cache, buf, sector, offset, num_bytes)) return false;
|
|
|
|
switch(num_bytes) {
|
|
case 1: *value = buf[0]; break;
|
|
case 2: *value = u8array_to_u16(buf,0); break;
|
|
case 4: *value = u8array_to_u32(buf,0); break;
|
|
default: return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
/*
|
|
Writes some data to a cache page, making sure it is loaded into memory first.
|
|
*/
|
|
|
|
bool _NTFS_cache_writePartialSector (NTFS_CACHE* cache, const void* buffer, sec_t sector, unsigned int offset, size_t size)
|
|
{
|
|
sec_t sec;
|
|
NTFS_CACHE_ENTRY *entry;
|
|
|
|
if (offset + size > cache->sectorSize) return false;
|
|
|
|
entry = _NTFS_cache_getPage(cache,sector);
|
|
if(entry==NULL) return false;
|
|
|
|
sec = sector - entry->sector;
|
|
memcpy(entry->cache + ((sec*cache->sectorSize) + offset),buffer,size);
|
|
|
|
entry->dirty = true;
|
|
return true;
|
|
}
|
|
|
|
bool _NTFS_cache_writeLittleEndianValue (NTFS_CACHE* cache, const uint32_t value, sec_t sector, unsigned int offset, int size) {
|
|
uint8_t buf[4] = {0, 0, 0, 0};
|
|
|
|
switch(size) {
|
|
case 1: buf[0] = value; break;
|
|
case 2: u16_to_u8array(buf, 0, value); break;
|
|
case 4: u32_to_u8array(buf, 0, value); break;
|
|
default: return false;
|
|
}
|
|
|
|
return _NTFS_cache_writePartialSector(cache, buf, sector, offset, size);
|
|
}
|
|
|
|
/*
|
|
Writes some data to a cache page, zeroing out the page first
|
|
*/
|
|
|
|
bool _NTFS_cache_eraseWritePartialSector (NTFS_CACHE* cache, const void* buffer, sec_t sector, unsigned int offset, size_t size)
|
|
{
|
|
sec_t sec;
|
|
NTFS_CACHE_ENTRY *entry;
|
|
|
|
if (offset + size > cache->sectorSize) return false;
|
|
|
|
entry = _NTFS_cache_getPage(cache,sector);
|
|
if(entry==NULL) return false;
|
|
|
|
sec = sector - entry->sector;
|
|
memset(entry->cache + (sec*cache->sectorSize),0,cache->sectorSize);
|
|
memcpy(entry->cache + ((sec*cache->sectorSize) + offset),buffer,size);
|
|
|
|
entry->dirty = true;
|
|
return true;
|
|
}
|
|
|
|
bool _NTFS_cache_writeSectors (NTFS_CACHE* cache, sec_t sector, sec_t numSectors, const void* buffer)
|
|
{
|
|
sec_t sec;
|
|
sec_t secs_to_write;
|
|
NTFS_CACHE_ENTRY* entry;
|
|
const uint8_t *src = buffer;
|
|
|
|
while(numSectors>0)
|
|
{
|
|
entry = _NTFS_cache_findPage(cache,sector,numSectors);
|
|
|
|
if(entry!=NULL) {
|
|
|
|
if ( entry->sector > sector) {
|
|
|
|
secs_to_write = entry->sector - sector;
|
|
|
|
cache->disc->writeSectors(sector,secs_to_write,src);
|
|
src += (secs_to_write*cache->sectorSize);
|
|
sector += secs_to_write;
|
|
numSectors -= secs_to_write;
|
|
}
|
|
|
|
sec = sector - entry->sector;
|
|
secs_to_write = entry->count - sec;
|
|
|
|
if(secs_to_write>numSectors) secs_to_write = numSectors;
|
|
|
|
memcpy(entry->cache + (sec*cache->sectorSize),src,(secs_to_write*cache->sectorSize));
|
|
|
|
src += (secs_to_write*cache->sectorSize);
|
|
sector += secs_to_write;
|
|
numSectors -= secs_to_write;
|
|
|
|
entry->dirty = true;
|
|
|
|
} else {
|
|
cache->disc->writeSectors(sector,numSectors,src);
|
|
numSectors=0;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
/*
|
|
Flushes all dirty pages to disc, clearing the dirty flag.
|
|
*/
|
|
bool _NTFS_cache_flush (NTFS_CACHE* cache) {
|
|
unsigned int i;
|
|
if(cache==NULL) return true;
|
|
|
|
for (i = 0; i < cache->numberOfPages; i++) {
|
|
if (cache->cacheEntries[i].dirty) {
|
|
if (!cache->disc->writeSectors (cache->cacheEntries[i].sector, cache->cacheEntries[i].count, cache->cacheEntries[i].cache)) {
|
|
return false;
|
|
}
|
|
}
|
|
cache->cacheEntries[i].dirty = false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
void _NTFS_cache_invalidate (NTFS_CACHE* cache) {
|
|
unsigned int i;
|
|
if(cache==NULL)
|
|
return;
|
|
|
|
_NTFS_cache_flush(cache);
|
|
for (i = 0; i < cache->numberOfPages; i++) {
|
|
cache->cacheEntries[i].sector = CACHE_FREE;
|
|
cache->cacheEntries[i].last_access = 0;
|
|
cache->cacheEntries[i].count = 0;
|
|
cache->cacheEntries[i].dirty = false;
|
|
}
|
|
}
|