linux/fs/jfs/jfs_metapage.c

859 lines
20 KiB
C
Raw Normal View History

/*
* Copyright (C) International Business Machines Corp., 2000-2005
* Portions Copyright (C) Christoph Hellwig, 2001-2002
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
* the GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/bio.h>
#include <linux/init.h>
#include <linux/buffer_head.h>
#include <linux/mempool.h>
#include "jfs_incore.h"
#include "jfs_superblock.h"
#include "jfs_filsys.h"
#include "jfs_metapage.h"
#include "jfs_txnmgr.h"
#include "jfs_debug.h"
#ifdef CONFIG_JFS_STATISTICS
static struct {
uint pagealloc; /* # of page allocations */
uint pagefree; /* # of page frees */
uint lockwait; /* # of sleeping lock_metapage() calls */
} mpStat;
#endif
#define metapage_locked(mp) test_bit(META_locked, &(mp)->flag)
#define trylock_metapage(mp) test_and_set_bit(META_locked, &(mp)->flag)
static inline void unlock_metapage(struct metapage *mp)
{
clear_bit(META_locked, &mp->flag);
wake_up(&mp->wait);
}
static inline void __lock_metapage(struct metapage *mp)
{
DECLARE_WAITQUEUE(wait, current);
INCREMENT(mpStat.lockwait);
add_wait_queue_exclusive(&mp->wait, &wait);
do {
set_current_state(TASK_UNINTERRUPTIBLE);
if (metapage_locked(mp)) {
unlock_page(mp->page);
schedule();
lock_page(mp->page);
}
} while (trylock_metapage(mp));
__set_current_state(TASK_RUNNING);
remove_wait_queue(&mp->wait, &wait);
}
/*
* Must have mp->page locked
*/
static inline void lock_metapage(struct metapage *mp)
{
if (trylock_metapage(mp))
__lock_metapage(mp);
}
#define METAPOOL_MIN_PAGES 32
static kmem_cache_t *metapage_cache;
static mempool_t *metapage_mempool;
#define MPS_PER_PAGE (PAGE_CACHE_SIZE >> L2PSIZE)
#if MPS_PER_PAGE > 1
struct meta_anchor {
int mp_count;
atomic_t io_count;
struct metapage *mp[MPS_PER_PAGE];
};
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
#define mp_anchor(page) ((struct meta_anchor *)page_private(page))
static inline struct metapage *page_to_mp(struct page *page, uint offset)
{
if (!PagePrivate(page))
return NULL;
return mp_anchor(page)->mp[offset >> L2PSIZE];
}
static inline int insert_metapage(struct page *page, struct metapage *mp)
{
struct meta_anchor *a;
int index;
int l2mp_blocks; /* log2 blocks per metapage */
if (PagePrivate(page))
a = mp_anchor(page);
else {
a = kzalloc(sizeof(struct meta_anchor), GFP_NOFS);
if (!a)
return -ENOMEM;
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
set_page_private(page, (unsigned long)a);
SetPagePrivate(page);
kmap(page);
}
if (mp) {
l2mp_blocks = L2PSIZE - page->mapping->host->i_blkbits;
index = (mp->index >> l2mp_blocks) & (MPS_PER_PAGE - 1);
a->mp_count++;
a->mp[index] = mp;
}
return 0;
}
static inline void remove_metapage(struct page *page, struct metapage *mp)
{
struct meta_anchor *a = mp_anchor(page);
int l2mp_blocks = L2PSIZE - page->mapping->host->i_blkbits;
int index;
index = (mp->index >> l2mp_blocks) & (MPS_PER_PAGE - 1);
BUG_ON(a->mp[index] != mp);
a->mp[index] = NULL;
if (--a->mp_count == 0) {
kfree(a);
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
set_page_private(page, 0);
ClearPagePrivate(page);
kunmap(page);
}
}
static inline void inc_io(struct page *page)
{
atomic_inc(&mp_anchor(page)->io_count);
}
static inline void dec_io(struct page *page, void (*handler) (struct page *))
{
if (atomic_dec_and_test(&mp_anchor(page)->io_count))
handler(page);
}
#else
static inline struct metapage *page_to_mp(struct page *page, uint offset)
{
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
return PagePrivate(page) ? (struct metapage *)page_private(page) : NULL;
}
static inline int insert_metapage(struct page *page, struct metapage *mp)
{
if (mp) {
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
set_page_private(page, (unsigned long)mp);
SetPagePrivate(page);
kmap(page);
}
return 0;
}
static inline void remove_metapage(struct page *page, struct metapage *mp)
{
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
set_page_private(page, 0);
ClearPagePrivate(page);
kunmap(page);
}
#define inc_io(page) do {} while(0)
#define dec_io(page, handler) handler(page)
#endif
static void init_once(void *foo, kmem_cache_t *cachep, unsigned long flags)
{
struct metapage *mp = (struct metapage *)foo;
if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
SLAB_CTOR_CONSTRUCTOR) {
mp->lid = 0;
mp->lsn = 0;
mp->flag = 0;
mp->data = NULL;
mp->clsn = 0;
mp->log = NULL;
set_bit(META_free, &mp->flag);
init_waitqueue_head(&mp->wait);
}
}
static inline struct metapage *alloc_metapage(gfp_t gfp_mask)
{
return mempool_alloc(metapage_mempool, gfp_mask);
}
static inline void free_metapage(struct metapage *mp)
{
mp->flag = 0;
set_bit(META_free, &mp->flag);
mempool_free(mp, metapage_mempool);
}
int __init metapage_init(void)
{
/*
* Allocate the metapage structures
*/
metapage_cache = kmem_cache_create("jfs_mp", sizeof(struct metapage),
0, 0, init_once, NULL);
if (metapage_cache == NULL)
return -ENOMEM;
metapage_mempool = mempool_create_slab_pool(METAPOOL_MIN_PAGES,
metapage_cache);
if (metapage_mempool == NULL) {
kmem_cache_destroy(metapage_cache);
return -ENOMEM;
}
return 0;
}
void metapage_exit(void)
{
mempool_destroy(metapage_mempool);
kmem_cache_destroy(metapage_cache);
}
static inline void drop_metapage(struct page *page, struct metapage *mp)
{
if (mp->count || mp->nohomeok || test_bit(META_dirty, &mp->flag) ||
test_bit(META_io, &mp->flag))
return;
remove_metapage(page, mp);
INCREMENT(mpStat.pagefree);
free_metapage(mp);
}
/*
* Metapage address space operations
*/
static sector_t metapage_get_blocks(struct inode *inode, sector_t lblock,
unsigned int *len)
{
int rc = 0;
int xflag;
s64 xaddr;
sector_t file_blocks = (inode->i_size + inode->i_blksize - 1) >>
inode->i_blkbits;
if (lblock >= file_blocks)
return 0;
if (lblock + *len > file_blocks)
*len = file_blocks - lblock;
if (inode->i_ino) {
rc = xtLookup(inode, (s64)lblock, *len, &xflag, &xaddr, len, 0);
if ((rc == 0) && *len)
lblock = (sector_t)xaddr;
else
lblock = 0;
} /* else no mapping */
return lblock;
}
static void last_read_complete(struct page *page)
{
if (!PageError(page))
SetPageUptodate(page);
unlock_page(page);
}
static int metapage_read_end_io(struct bio *bio, unsigned int bytes_done,
int err)
{
struct page *page = bio->bi_private;
if (bio->bi_size)
return 1;
if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
printk(KERN_ERR "metapage_read_end_io: I/O error\n");
SetPageError(page);
}
dec_io(page, last_read_complete);
bio_put(bio);
return 0;
}
static void remove_from_logsync(struct metapage *mp)
{
struct jfs_log *log = mp->log;
unsigned long flags;
/*
* This can race. Recheck that log hasn't been set to null, and after
* acquiring logsync lock, recheck lsn
*/
if (!log)
return;
LOGSYNC_LOCK(log, flags);
if (mp->lsn) {
mp->log = NULL;
mp->lsn = 0;
mp->clsn = 0;
log->count--;
list_del(&mp->synclist);
}
LOGSYNC_UNLOCK(log, flags);
}
static void last_write_complete(struct page *page)
{
struct metapage *mp;
unsigned int offset;
for (offset = 0; offset < PAGE_CACHE_SIZE; offset += PSIZE) {
mp = page_to_mp(page, offset);
if (mp && test_bit(META_io, &mp->flag)) {
if (mp->lsn)
remove_from_logsync(mp);
clear_bit(META_io, &mp->flag);
}
/*
* I'd like to call drop_metapage here, but I don't think it's
* safe unless I have the page locked
*/
}
end_page_writeback(page);
}
static int metapage_write_end_io(struct bio *bio, unsigned int bytes_done,
int err)
{
struct page *page = bio->bi_private;
BUG_ON(!PagePrivate(page));
if (bio->bi_size)
return 1;
if (! test_bit(BIO_UPTODATE, &bio->bi_flags)) {
printk(KERN_ERR "metapage_write_end_io: I/O error\n");
SetPageError(page);
}
dec_io(page, last_write_complete);
bio_put(bio);
return 0;
}
static int metapage_writepage(struct page *page, struct writeback_control *wbc)
{
struct bio *bio = NULL;
unsigned int block_offset; /* block offset of mp within page */
struct inode *inode = page->mapping->host;
unsigned int blocks_per_mp = JFS_SBI(inode->i_sb)->nbperpage;
unsigned int len;
unsigned int xlen;
struct metapage *mp;
int redirty = 0;
sector_t lblock;
sector_t pblock;
sector_t next_block = 0;
sector_t page_start;
unsigned long bio_bytes = 0;
unsigned long bio_offset = 0;
unsigned int offset;
page_start = (sector_t)page->index <<
(PAGE_CACHE_SHIFT - inode->i_blkbits);
BUG_ON(!PageLocked(page));
BUG_ON(PageWriteback(page));
for (offset = 0; offset < PAGE_CACHE_SIZE; offset += PSIZE) {
mp = page_to_mp(page, offset);
if (!mp || !test_bit(META_dirty, &mp->flag))
continue;
if (mp->nohomeok && !test_bit(META_forcewrite, &mp->flag)) {
redirty = 1;
/*
* Make sure this page isn't blocked indefinitely.
* If the journal isn't undergoing I/O, push it
*/
if (mp->log && !(mp->log->cflag & logGC_PAGEOUT))
jfs_flush_journal(mp->log, 0);
continue;
}
clear_bit(META_dirty, &mp->flag);
block_offset = offset >> inode->i_blkbits;
lblock = page_start + block_offset;
if (bio) {
if (xlen && lblock == next_block) {
/* Contiguous, in memory & on disk */
len = min(xlen, blocks_per_mp);
xlen -= len;
bio_bytes += len << inode->i_blkbits;
set_bit(META_io, &mp->flag);
continue;
}
/* Not contiguous */
if (bio_add_page(bio, page, bio_bytes, bio_offset) <
bio_bytes)
goto add_failed;
/*
* Increment counter before submitting i/o to keep
* count from hitting zero before we're through
*/
inc_io(page);
if (!bio->bi_size)
goto dump_bio;
submit_bio(WRITE, bio);
bio = NULL;
} else {
set_page_writeback(page);
inc_io(page);
}
xlen = (PAGE_CACHE_SIZE - offset) >> inode->i_blkbits;
pblock = metapage_get_blocks(inode, lblock, &xlen);
if (!pblock) {
/* Need better error handling */
printk(KERN_ERR "JFS: metapage_get_blocks failed\n");
dec_io(page, last_write_complete);
continue;
}
set_bit(META_io, &mp->flag);
len = min(xlen, (uint) JFS_SBI(inode->i_sb)->nbperpage);
bio = bio_alloc(GFP_NOFS, 1);
bio->bi_bdev = inode->i_sb->s_bdev;
bio->bi_sector = pblock << (inode->i_blkbits - 9);
bio->bi_end_io = metapage_write_end_io;
bio->bi_private = page;
/* Don't call bio_add_page yet, we may add to this vec */
bio_offset = offset;
bio_bytes = len << inode->i_blkbits;
xlen -= len;
next_block = lblock + len;
}
if (bio) {
if (bio_add_page(bio, page, bio_bytes, bio_offset) < bio_bytes)
goto add_failed;
if (!bio->bi_size)
goto dump_bio;
submit_bio(WRITE, bio);
}
if (redirty)
redirty_page_for_writepage(wbc, page);
unlock_page(page);
return 0;
add_failed:
/* We should never reach here, since we're only adding one vec */
printk(KERN_ERR "JFS: bio_add_page failed unexpectedly\n");
goto skip;
dump_bio:
dump_mem("bio", bio, sizeof(*bio));
skip:
bio_put(bio);
unlock_page(page);
dec_io(page, last_write_complete);
return -EIO;
}
static int metapage_readpage(struct file *fp, struct page *page)
{
struct inode *inode = page->mapping->host;
struct bio *bio = NULL;
unsigned int block_offset;
unsigned int blocks_per_page = PAGE_CACHE_SIZE >> inode->i_blkbits;
sector_t page_start; /* address of page in fs blocks */
sector_t pblock;
unsigned int xlen;
unsigned int len;
unsigned int offset;
BUG_ON(!PageLocked(page));
page_start = (sector_t)page->index <<
(PAGE_CACHE_SHIFT - inode->i_blkbits);
block_offset = 0;
while (block_offset < blocks_per_page) {
xlen = blocks_per_page - block_offset;
pblock = metapage_get_blocks(inode, page_start + block_offset,
&xlen);
if (pblock) {
if (!PagePrivate(page))
insert_metapage(page, NULL);
inc_io(page);
if (bio)
submit_bio(READ, bio);
bio = bio_alloc(GFP_NOFS, 1);
bio->bi_bdev = inode->i_sb->s_bdev;
bio->bi_sector = pblock << (inode->i_blkbits - 9);
bio->bi_end_io = metapage_read_end_io;
bio->bi_private = page;
len = xlen << inode->i_blkbits;
offset = block_offset << inode->i_blkbits;
if (bio_add_page(bio, page, len, offset) < len)
goto add_failed;
block_offset += xlen;
} else
block_offset++;
}
if (bio)
submit_bio(READ, bio);
else
unlock_page(page);
return 0;
add_failed:
printk(KERN_ERR "JFS: bio_add_page failed unexpectedly\n");
bio_put(bio);
dec_io(page, last_read_complete);
return -EIO;
}
static int metapage_releasepage(struct page *page, gfp_t gfp_mask)
{
struct metapage *mp;
int ret = 1;
unsigned int offset;
for (offset = 0; offset < PAGE_CACHE_SIZE; offset += PSIZE) {
mp = page_to_mp(page, offset);
if (!mp)
continue;
jfs_info("metapage_releasepage: mp = 0x%p", mp);
if (mp->count || mp->nohomeok ||
test_bit(META_dirty, &mp->flag)) {
jfs_info("count = %ld, nohomeok = %d", mp->count,
mp->nohomeok);
ret = 0;
continue;
}
if (mp->lsn)
remove_from_logsync(mp);
remove_metapage(page, mp);
INCREMENT(mpStat.pagefree);
free_metapage(mp);
}
return ret;
}
static void metapage_invalidatepage(struct page *page, unsigned long offset)
{
BUG_ON(offset);
BUG_ON(PageWriteback(page));
metapage_releasepage(page, 0);
}
const struct address_space_operations jfs_metapage_aops = {
.readpage = metapage_readpage,
.writepage = metapage_writepage,
.sync_page = block_sync_page,
.releasepage = metapage_releasepage,
.invalidatepage = metapage_invalidatepage,
.set_page_dirty = __set_page_dirty_nobuffers,
};
struct metapage *__get_metapage(struct inode *inode, unsigned long lblock,
unsigned int size, int absolute,
unsigned long new)
{
int l2BlocksPerPage;
int l2bsize;
struct address_space *mapping;
struct metapage *mp = NULL;
struct page *page;
unsigned long page_index;
unsigned long page_offset;
jfs_info("__get_metapage: ino = %ld, lblock = 0x%lx, abs=%d",
inode->i_ino, lblock, absolute);
l2bsize = inode->i_blkbits;
l2BlocksPerPage = PAGE_CACHE_SHIFT - l2bsize;
page_index = lblock >> l2BlocksPerPage;
page_offset = (lblock - (page_index << l2BlocksPerPage)) << l2bsize;
if ((page_offset + size) > PAGE_CACHE_SIZE) {
jfs_err("MetaData crosses page boundary!!");
jfs_err("lblock = %lx, size = %d", lblock, size);
dump_stack();
return NULL;
}
if (absolute)
mapping = JFS_SBI(inode->i_sb)->direct_inode->i_mapping;
else {
/*
* If an nfs client tries to read an inode that is larger
* than any existing inodes, we may try to read past the
* end of the inode map
*/
if ((lblock << inode->i_blkbits) >= inode->i_size)
return NULL;
mapping = inode->i_mapping;
}
if (new && (PSIZE == PAGE_CACHE_SIZE)) {
page = grab_cache_page(mapping, page_index);
if (!page) {
jfs_err("grab_cache_page failed!");
return NULL;
}
SetPageUptodate(page);
} else {
page = read_mapping_page(mapping, page_index, NULL);
if (IS_ERR(page) || !PageUptodate(page)) {
jfs_err("read_mapping_page failed!");
return NULL;
}
lock_page(page);
}
mp = page_to_mp(page, page_offset);
if (mp) {
if (mp->logical_size != size) {
jfs_error(inode->i_sb,
"__get_metapage: mp->logical_size != size");
jfs_err("logical_size = %d, size = %d",
mp->logical_size, size);
dump_stack();
goto unlock;
}
mp->count++;
lock_metapage(mp);
if (test_bit(META_discard, &mp->flag)) {
if (!new) {
jfs_error(inode->i_sb,
"__get_metapage: using a "
"discarded metapage");
discard_metapage(mp);
goto unlock;
}
clear_bit(META_discard, &mp->flag);
}
} else {
INCREMENT(mpStat.pagealloc);
mp = alloc_metapage(GFP_NOFS);
mp->page = page;
mp->flag = 0;
mp->xflag = COMMIT_PAGE;
mp->count = 1;
mp->nohomeok = 0;
mp->logical_size = size;
mp->data = page_address(page) + page_offset;
mp->index = lblock;
if (unlikely(insert_metapage(page, mp))) {
free_metapage(mp);
goto unlock;
}
lock_metapage(mp);
}
if (new) {
jfs_info("zeroing mp = 0x%p", mp);
memset(mp->data, 0, PSIZE);
}
unlock_page(page);
jfs_info("__get_metapage: returning = 0x%p data = 0x%p", mp, mp->data);
return mp;
unlock:
unlock_page(page);
return NULL;
}
void grab_metapage(struct metapage * mp)
{
jfs_info("grab_metapage: mp = 0x%p", mp);
page_cache_get(mp->page);
lock_page(mp->page);
mp->count++;
lock_metapage(mp);
unlock_page(mp->page);
}
void force_metapage(struct metapage *mp)
{
struct page *page = mp->page;
jfs_info("force_metapage: mp = 0x%p", mp);
set_bit(META_forcewrite, &mp->flag);
clear_bit(META_sync, &mp->flag);
page_cache_get(page);
lock_page(page);
set_page_dirty(page);
write_one_page(page, 1);
clear_bit(META_forcewrite, &mp->flag);
page_cache_release(page);
}
void hold_metapage(struct metapage *mp)
{
lock_page(mp->page);
}
void put_metapage(struct metapage *mp)
{
if (mp->count || mp->nohomeok) {
/* Someone else will release this */
unlock_page(mp->page);
return;
}
page_cache_get(mp->page);
mp->count++;
lock_metapage(mp);
unlock_page(mp->page);
release_metapage(mp);
}
void release_metapage(struct metapage * mp)
{
struct page *page = mp->page;
jfs_info("release_metapage: mp = 0x%p, flag = 0x%lx", mp, mp->flag);
BUG_ON(!page);
lock_page(page);
unlock_metapage(mp);
assert(mp->count);
if (--mp->count || mp->nohomeok) {
unlock_page(page);
page_cache_release(page);
return;
}
if (test_bit(META_dirty, &mp->flag)) {
set_page_dirty(page);
if (test_bit(META_sync, &mp->flag)) {
clear_bit(META_sync, &mp->flag);
write_one_page(page, 1);
lock_page(page); /* write_one_page unlocks the page */
}
} else if (mp->lsn) /* discard_metapage doesn't remove it */
remove_from_logsync(mp);
#if MPS_PER_PAGE == 1
/*
* If we know this is the only thing in the page, we can throw
* the page out of the page cache. If pages are larger, we
* don't want to do this.
*/
/* Retest mp->count since we may have released page lock */
if (test_bit(META_discard, &mp->flag) && !mp->count) {
clear_page_dirty(page);
ClearPageUptodate(page);
}
#else
/* Try to keep metapages from using up too much memory */
drop_metapage(page, mp);
#endif
unlock_page(page);
page_cache_release(page);
}
void __invalidate_metapages(struct inode *ip, s64 addr, int len)
{
sector_t lblock;
int l2BlocksPerPage = PAGE_CACHE_SHIFT - ip->i_blkbits;
int BlocksPerPage = 1 << l2BlocksPerPage;
/* All callers are interested in block device's mapping */
struct address_space *mapping =
JFS_SBI(ip->i_sb)->direct_inode->i_mapping;
struct metapage *mp;
struct page *page;
unsigned int offset;
/*
* Mark metapages to discard. They will eventually be
* released, but should not be written.
*/
for (lblock = addr & ~(BlocksPerPage - 1); lblock < addr + len;
lblock += BlocksPerPage) {
page = find_lock_page(mapping, lblock >> l2BlocksPerPage);
if (!page)
continue;
for (offset = 0; offset < PAGE_CACHE_SIZE; offset += PSIZE) {
mp = page_to_mp(page, offset);
if (!mp)
continue;
if (mp->index < addr)
continue;
if (mp->index >= addr + len)
break;
clear_bit(META_dirty, &mp->flag);
set_bit(META_discard, &mp->flag);
if (mp->lsn)
remove_from_logsync(mp);
}
unlock_page(page);
page_cache_release(page);
}
}
#ifdef CONFIG_JFS_STATISTICS
int jfs_mpstat_read(char *buffer, char **start, off_t offset, int length,
int *eof, void *data)
{
int len = 0;
off_t begin;
len += sprintf(buffer,
"JFS Metapage statistics\n"
"=======================\n"
"page allocations = %d\n"
"page frees = %d\n"
"lock waits = %d\n",
mpStat.pagealloc,
mpStat.pagefree,
mpStat.lockwait);
begin = offset;
*start = buffer + begin;
len -= begin;
if (len > length)
len = length;
else
*eof = 1;
if (len < 0)
len = 0;
return len;
}
#endif