linux/fs/gfs2/aops.c

1183 lines
30 KiB
C
Raw Normal View History

/*
* Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
* Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
*
* This copyrighted material is made available to anyone wishing to use,
* modify, copy, or redistribute it subject to the terms and conditions
* of the GNU General Public License version 2.
*/
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/completion.h>
#include <linux/buffer_head.h>
#include <linux/pagemap.h>
#include <linux/pagevec.h>
#include <linux/mpage.h>
#include <linux/fs.h>
#include <linux/writeback.h>
#include <linux/swap.h>
#include <linux/gfs2_ondisk.h>
#include <linux/backing-dev.h>
#include "gfs2.h"
#include "incore.h"
#include "bmap.h"
#include "glock.h"
#include "inode.h"
#include "log.h"
#include "meta_io.h"
#include "quota.h"
#include "trans.h"
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
#include "rgrp.h"
#include "super.h"
#include "util.h"
#include "glops.h"
void gfs2_page_add_databufs(struct gfs2_inode *ip, struct page *page,
unsigned int from, unsigned int to)
{
struct buffer_head *head = page_buffers(page);
unsigned int bsize = head->b_size;
struct buffer_head *bh;
unsigned int start, end;
for (bh = head, start = 0; bh != head || !start;
bh = bh->b_this_page, start = end) {
end = start + bsize;
if (end <= from || start >= to)
continue;
if (gfs2_is_jdata(ip))
set_buffer_uptodate(bh);
gfs2_trans_add_bh(ip->i_gl, bh, 0);
}
}
/**
* gfs2_get_block_noalloc - Fills in a buffer head with details about a block
* @inode: The inode
* @lblock: The block number to look up
* @bh_result: The buffer head to return the result in
* @create: Non-zero if we may add block to the file
*
* Returns: errno
*/
static int gfs2_get_block_noalloc(struct inode *inode, sector_t lblock,
struct buffer_head *bh_result, int create)
{
int error;
error = gfs2_block_map(inode, lblock, bh_result, 0);
if (error)
return error;
if (!buffer_mapped(bh_result))
return -EIO;
return 0;
}
static int gfs2_get_block_direct(struct inode *inode, sector_t lblock,
struct buffer_head *bh_result, int create)
{
return gfs2_block_map(inode, lblock, bh_result, 0);
}
/**
* gfs2_writepage_common - Common bits of writepage
* @page: The page to be written
* @wbc: The writeback control
*
* Returns: 1 if writepage is ok, otherwise an error code or zero if no error.
*/
static int gfs2_writepage_common(struct page *page,
struct writeback_control *wbc)
{
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
struct inode *inode = page->mapping->host;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_sbd *sdp = GFS2_SB(inode);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
loff_t i_size = i_size_read(inode);
pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
unsigned offset;
if (gfs2_assert_withdraw(sdp, gfs2_glock_is_held_excl(ip->i_gl)))
goto out;
if (current->journal_info)
goto redirty;
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
/* Is the page fully outside i_size? (truncate in progress) */
offset = i_size & (PAGE_CACHE_SIZE-1);
if (page->index > end_index || (page->index == end_index && !offset)) {
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
page->mapping->a_ops->invalidatepage(page, 0);
goto out;
}
return 1;
redirty:
redirty_page_for_writepage(wbc, page);
out:
unlock_page(page);
return 0;
}
/**
* gfs2_writeback_writepage - Write page for writeback mappings
* @page: The page
* @wbc: The writeback control
*
*/
static int gfs2_writeback_writepage(struct page *page,
struct writeback_control *wbc)
{
int ret;
ret = gfs2_writepage_common(page, wbc);
if (ret <= 0)
return ret;
return nobh_writepage(page, gfs2_get_block_noalloc, wbc);
}
/**
* gfs2_ordered_writepage - Write page for ordered data files
* @page: The page to write
* @wbc: The writeback control
*
*/
static int gfs2_ordered_writepage(struct page *page,
struct writeback_control *wbc)
{
struct inode *inode = page->mapping->host;
struct gfs2_inode *ip = GFS2_I(inode);
int ret;
ret = gfs2_writepage_common(page, wbc);
if (ret <= 0)
return ret;
if (!page_has_buffers(page)) {
create_empty_buffers(page, inode->i_sb->s_blocksize,
(1 << BH_Dirty)|(1 << BH_Uptodate));
}
gfs2_page_add_databufs(ip, page, 0, inode->i_sb->s_blocksize-1);
return block_write_full_page(page, gfs2_get_block_noalloc, wbc);
}
/**
* __gfs2_jdata_writepage - The core of jdata writepage
* @page: The page to write
* @wbc: The writeback control
*
* This is shared between writepage and writepages and implements the
* core of the writepage operation. If a transaction is required then
* PageChecked will have been set and the transaction will have
* already been started before this is called.
*/
static int __gfs2_jdata_writepage(struct page *page, struct writeback_control *wbc)
{
struct inode *inode = page->mapping->host;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_sbd *sdp = GFS2_SB(inode);
if (PageChecked(page)) {
ClearPageChecked(page);
if (!page_has_buffers(page)) {
create_empty_buffers(page, inode->i_sb->s_blocksize,
(1 << BH_Dirty)|(1 << BH_Uptodate));
}
gfs2_page_add_databufs(ip, page, 0, sdp->sd_vfs->s_blocksize-1);
}
return block_write_full_page(page, gfs2_get_block_noalloc, wbc);
}
/**
* gfs2_jdata_writepage - Write complete page
* @page: Page to write
*
* Returns: errno
*
*/
static int gfs2_jdata_writepage(struct page *page, struct writeback_control *wbc)
{
struct inode *inode = page->mapping->host;
struct gfs2_sbd *sdp = GFS2_SB(inode);
int ret;
int done_trans = 0;
if (PageChecked(page)) {
if (wbc->sync_mode != WB_SYNC_ALL)
goto out_ignore;
ret = gfs2_trans_begin(sdp, RES_DINODE + 1, 0);
if (ret)
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
goto out_ignore;
done_trans = 1;
}
ret = gfs2_writepage_common(page, wbc);
if (ret > 0)
ret = __gfs2_jdata_writepage(page, wbc);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
if (done_trans)
gfs2_trans_end(sdp);
return ret;
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
out_ignore:
redirty_page_for_writepage(wbc, page);
unlock_page(page);
return 0;
}
/**
* gfs2_writeback_writepages - Write a bunch of dirty pages back to disk
* @mapping: The mapping to write
* @wbc: Write-back control
*
* For the data=writeback case we can already ignore buffer heads
* and write whole extents at once. This is a big reduction in the
* number of I/O requests we send and the bmap calls we make in this case.
*/
static int gfs2_writeback_writepages(struct address_space *mapping,
struct writeback_control *wbc)
{
return mpage_writepages(mapping, wbc, gfs2_get_block_noalloc);
}
/**
* gfs2_write_jdata_pagevec - Write back a pagevec's worth of pages
* @mapping: The mapping
* @wbc: The writeback control
* @writepage: The writepage function to call for each page
* @pvec: The vector of pages
* @nr_pages: The number of pages to write
*
* Returns: non-zero if loop should terminate, zero otherwise
*/
static int gfs2_write_jdata_pagevec(struct address_space *mapping,
struct writeback_control *wbc,
struct pagevec *pvec,
int nr_pages, pgoff_t end)
{
struct inode *inode = mapping->host;
struct gfs2_sbd *sdp = GFS2_SB(inode);
loff_t i_size = i_size_read(inode);
pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
unsigned offset = i_size & (PAGE_CACHE_SIZE-1);
unsigned nrblocks = nr_pages * (PAGE_CACHE_SIZE/inode->i_sb->s_blocksize);
int i;
int ret;
ret = gfs2_trans_begin(sdp, nrblocks, nrblocks);
if (ret < 0)
return ret;
for(i = 0; i < nr_pages; i++) {
struct page *page = pvec->pages[i];
lock_page(page);
if (unlikely(page->mapping != mapping)) {
unlock_page(page);
continue;
}
if (!wbc->range_cyclic && page->index > end) {
ret = 1;
unlock_page(page);
continue;
}
if (wbc->sync_mode != WB_SYNC_NONE)
wait_on_page_writeback(page);
if (PageWriteback(page) ||
!clear_page_dirty_for_io(page)) {
unlock_page(page);
continue;
}
/* Is the page fully outside i_size? (truncate in progress) */
if (page->index > end_index || (page->index == end_index && !offset)) {
page->mapping->a_ops->invalidatepage(page, 0);
unlock_page(page);
continue;
}
ret = __gfs2_jdata_writepage(page, wbc);
if (ret || (--(wbc->nr_to_write) <= 0))
ret = 1;
}
gfs2_trans_end(sdp);
return ret;
}
/**
* gfs2_write_cache_jdata - Like write_cache_pages but different
* @mapping: The mapping to write
* @wbc: The writeback control
* @writepage: The writepage function to call
* @data: The data to pass to writepage
*
* The reason that we use our own function here is that we need to
* start transactions before we grab page locks. This allows us
* to get the ordering right.
*/
static int gfs2_write_cache_jdata(struct address_space *mapping,
struct writeback_control *wbc)
{
int ret = 0;
int done = 0;
struct pagevec pvec;
int nr_pages;
pgoff_t index;
pgoff_t end;
int scanned = 0;
int range_whole = 0;
pagevec_init(&pvec, 0);
if (wbc->range_cyclic) {
index = mapping->writeback_index; /* Start from prev offset */
end = -1;
} else {
index = wbc->range_start >> PAGE_CACHE_SHIFT;
end = wbc->range_end >> PAGE_CACHE_SHIFT;
if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
range_whole = 1;
scanned = 1;
}
retry:
while (!done && (index <= end) &&
(nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
PAGECACHE_TAG_DIRTY,
min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
scanned = 1;
ret = gfs2_write_jdata_pagevec(mapping, wbc, &pvec, nr_pages, end);
if (ret)
done = 1;
if (ret > 0)
ret = 0;
pagevec_release(&pvec);
cond_resched();
}
if (!scanned && !done) {
/*
* We hit the last page and there is more work to be done: wrap
* back to the start of the file
*/
scanned = 1;
index = 0;
goto retry;
}
if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
mapping->writeback_index = index;
return ret;
}
/**
* gfs2_jdata_writepages - Write a bunch of dirty pages back to disk
* @mapping: The mapping to write
* @wbc: The writeback control
*
*/
static int gfs2_jdata_writepages(struct address_space *mapping,
struct writeback_control *wbc)
{
struct gfs2_inode *ip = GFS2_I(mapping->host);
struct gfs2_sbd *sdp = GFS2_SB(mapping->host);
int ret;
ret = gfs2_write_cache_jdata(mapping, wbc);
if (ret == 0 && wbc->sync_mode == WB_SYNC_ALL) {
gfs2_log_flush(sdp, ip->i_gl);
ret = gfs2_write_cache_jdata(mapping, wbc);
}
return ret;
}
/**
* stuffed_readpage - Fill in a Linux page with stuffed file data
* @ip: the inode
* @page: the page
*
* Returns: errno
*/
static int stuffed_readpage(struct gfs2_inode *ip, struct page *page)
{
struct buffer_head *dibh;
u64 dsize = i_size_read(&ip->i_inode);
void *kaddr;
int error;
/*
* Due to the order of unstuffing files and ->fault(), we can be
* asked for a zero page in the case of a stuffed file being extended,
* so we need to supply one here. It doesn't happen often.
*/
if (unlikely(page->index)) {
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-05 06:28:29 +00:00
zero_user(page, 0, PAGE_CACHE_SIZE);
SetPageUptodate(page);
return 0;
}
error = gfs2_meta_inode_buffer(ip, &dibh);
if (error)
return error;
kaddr = kmap_atomic(page, KM_USER0);
if (dsize > (dibh->b_size - sizeof(struct gfs2_dinode)))
dsize = (dibh->b_size - sizeof(struct gfs2_dinode));
memcpy(kaddr, dibh->b_data + sizeof(struct gfs2_dinode), dsize);
memset(kaddr + dsize, 0, PAGE_CACHE_SIZE - dsize);
kunmap_atomic(kaddr, KM_USER0);
flush_dcache_page(page);
brelse(dibh);
SetPageUptodate(page);
return 0;
}
/**
* __gfs2_readpage - readpage
* @file: The file to read a page for
* @page: The page to read
*
* This is the core of gfs2's readpage. Its used by the internal file
* reading code as in that case we already hold the glock. Also its
* called by gfs2_readpage() once the required lock has been granted.
*
*/
static int __gfs2_readpage(void *file, struct page *page)
{
struct gfs2_inode *ip = GFS2_I(page->mapping->host);
struct gfs2_sbd *sdp = GFS2_SB(page->mapping->host);
int error;
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
if (gfs2_is_stuffed(ip)) {
error = stuffed_readpage(ip, page);
unlock_page(page);
} else {
error = mpage_readpage(page, gfs2_block_map);
}
if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
return -EIO;
return error;
}
/**
* gfs2_readpage - read a page of a file
* @file: The file to read
* @page: The page of the file
*
* This deals with the locking required. We have to unlock and
* relock the page in order to get the locking in the right
* order.
*/
static int gfs2_readpage(struct file *file, struct page *page)
{
struct address_space *mapping = page->mapping;
struct gfs2_inode *ip = GFS2_I(mapping->host);
struct gfs2_holder gh;
int error;
unlock_page(page);
gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
error = gfs2_glock_nq(&gh);
if (unlikely(error))
goto out;
error = AOP_TRUNCATED_PAGE;
lock_page(page);
if (page->mapping == mapping && !PageUptodate(page))
error = __gfs2_readpage(file, page);
else
unlock_page(page);
gfs2_glock_dq(&gh);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
out:
gfs2_holder_uninit(&gh);
if (error && error != AOP_TRUNCATED_PAGE)
lock_page(page);
return error;
}
/**
* gfs2_internal_read - read an internal file
* @ip: The gfs2 inode
* @ra_state: The readahead state (or NULL for no readahead)
* @buf: The buffer to fill
* @pos: The file position
* @size: The amount to read
*
*/
int gfs2_internal_read(struct gfs2_inode *ip, struct file_ra_state *ra_state,
char *buf, loff_t *pos, unsigned size)
{
struct address_space *mapping = ip->i_inode.i_mapping;
unsigned long index = *pos / PAGE_CACHE_SIZE;
unsigned offset = *pos & (PAGE_CACHE_SIZE - 1);
unsigned copied = 0;
unsigned amt;
struct page *page;
void *p;
do {
amt = size - copied;
if (offset + size > PAGE_CACHE_SIZE)
amt = PAGE_CACHE_SIZE - offset;
page = read_cache_page(mapping, index, __gfs2_readpage, NULL);
if (IS_ERR(page))
return PTR_ERR(page);
p = kmap_atomic(page, KM_USER0);
memcpy(buf + copied, p + offset, amt);
kunmap_atomic(p, KM_USER0);
mark_page_accessed(page);
page_cache_release(page);
copied += amt;
index++;
offset = 0;
} while(copied < size);
(*pos) += size;
return size;
}
/**
* gfs2_readpages - Read a bunch of pages at once
*
* Some notes:
* 1. This is only for readahead, so we can simply ignore any things
* which are slightly inconvenient (such as locking conflicts between
* the page lock and the glock) and return having done no I/O. Its
* obviously not something we'd want to do on too regular a basis.
* Any I/O we ignore at this time will be done via readpage later.
* 2. We don't handle stuffed files here we let readpage do the honours.
* 3. mpage_readpages() does most of the heavy lifting in the common case.
* 4. gfs2_block_map() is relied upon to set BH_Boundary in the right places.
*/
static int gfs2_readpages(struct file *file, struct address_space *mapping,
struct list_head *pages, unsigned nr_pages)
{
struct inode *inode = mapping->host;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_sbd *sdp = GFS2_SB(inode);
struct gfs2_holder gh;
int ret;
gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
ret = gfs2_glock_nq(&gh);
if (unlikely(ret))
goto out_uninit;
if (!gfs2_is_stuffed(ip))
ret = mpage_readpages(mapping, pages, nr_pages, gfs2_block_map);
gfs2_glock_dq(&gh);
out_uninit:
gfs2_holder_uninit(&gh);
if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
ret = -EIO;
return ret;
}
/**
* gfs2_write_begin - Begin to write to a file
* @file: The file to write to
* @mapping: The mapping in which to write
* @pos: The file offset at which to start writing
* @len: Length of the write
* @flags: Various flags
* @pagep: Pointer to return the page
* @fsdata: Pointer to return fs data (unused by GFS2)
*
* Returns: errno
*/
static int gfs2_write_begin(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
struct page **pagep, void **fsdata)
{
struct gfs2_inode *ip = GFS2_I(mapping->host);
struct gfs2_sbd *sdp = GFS2_SB(mapping->host);
struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
unsigned int data_blocks = 0, ind_blocks = 0, rblocks;
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
int alloc_required;
int error = 0;
struct gfs2_alloc *al = NULL;
pgoff_t index = pos >> PAGE_CACHE_SHIFT;
unsigned from = pos & (PAGE_CACHE_SIZE - 1);
struct page *page;
gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &ip->i_gh);
error = gfs2_glock_nq(&ip->i_gh);
if (unlikely(error))
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
goto out_uninit;
if (&ip->i_inode == sdp->sd_rindex) {
error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE,
GL_NOCACHE, &m_ip->i_gh);
if (unlikely(error)) {
gfs2_glock_dq(&ip->i_gh);
goto out_uninit;
}
}
alloc_required = gfs2_write_alloc_required(ip, pos, len);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
if (alloc_required || gfs2_is_jdata(ip))
gfs2_write_calc_reserv(ip, len, &data_blocks, &ind_blocks);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
if (alloc_required) {
al = gfs2_alloc_get(ip);
if (!al) {
error = -ENOMEM;
goto out_unlock;
}
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
error = gfs2_quota_lock_check(ip);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
if (error)
goto out_alloc_put;
al->al_requested = data_blocks + ind_blocks;
error = gfs2_inplace_reserve(ip);
if (error)
goto out_qunlock;
}
rblocks = RES_DINODE + ind_blocks;
if (gfs2_is_jdata(ip))
rblocks += data_blocks ? data_blocks : 1;
if (ind_blocks || data_blocks)
rblocks += RES_STATFS + RES_QUOTA;
if (&ip->i_inode == sdp->sd_rindex)
rblocks += 2 * RES_STATFS;
if (alloc_required)
rblocks += gfs2_rg_blocks(al);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
error = gfs2_trans_begin(sdp, rblocks,
PAGE_CACHE_SIZE/sdp->sd_sb.sb_bsize);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
if (error)
goto out_trans_fail;
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
error = -ENOMEM;
flags |= AOP_FLAG_NOFS;
fs: symlink write_begin allocation context fix With the write_begin/write_end aops, page_symlink was broken because it could no longer pass a GFP_NOFS type mask into the point where the allocations happened. They are done in write_begin, which would always assume that the filesystem can be entered from reclaim. This bug could cause filesystem deadlocks. The funny thing with having a gfp_t mask there is that it doesn't really allow the caller to arbitrarily tinker with the context in which it can be called. It couldn't ever be GFP_ATOMIC, for example, because it needs to take the page lock. The only thing any callers care about is __GFP_FS anyway, so turn that into a single flag. Add a new flag for write_begin, AOP_FLAG_NOFS. Filesystems can now act on this flag in their write_begin function. Change __grab_cache_page to accept a nofs argument as well, to honour that flag (while we're there, change the name to grab_cache_page_write_begin which is more instructive and does away with random leading underscores). This is really a more flexible way to go in the end anyway -- if a filesystem happens to want any extra allocations aside from the pagecache ones in ints write_begin function, it may now use GFP_KERNEL (rather than GFP_NOFS) for common case allocations (eg. ocfs2_alloc_write_ctxt, for a random example). [kosaki.motohiro@jp.fujitsu.com: fix ubifs] [kosaki.motohiro@jp.fujitsu.com: fix fuse] Signed-off-by: Nick Piggin <npiggin@suse.de> Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: <stable@kernel.org> [2.6.28.x] Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> [ Cleaned up the calling convention: just pass in the AOP flags untouched to the grab_cache_page_write_begin() function. That just simplifies everybody, and may even allow future expansion of the logic. - Linus ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-04 20:00:53 +00:00
page = grab_cache_page_write_begin(mapping, index, flags);
*pagep = page;
if (unlikely(!page))
goto out_endtrans;
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
if (gfs2_is_stuffed(ip)) {
error = 0;
if (pos + len > sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)) {
error = gfs2_unstuff_dinode(ip, page);
if (error == 0)
goto prepare_write;
} else if (!PageUptodate(page)) {
error = stuffed_readpage(ip, page);
}
goto out;
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
}
prepare_write:
error = __block_write_begin(page, from, len, gfs2_block_map);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
out:
if (error == 0)
return 0;
unlock_page(page);
page_cache_release(page);
gfs2_trans_end(sdp);
if (pos + len > ip->i_inode.i_size)
gfs2_trim_blocks(&ip->i_inode);
goto out_trans_fail;
out_endtrans:
gfs2_trans_end(sdp);
out_trans_fail:
if (alloc_required) {
gfs2_inplace_release(ip);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
out_qunlock:
gfs2_quota_unlock(ip);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
out_alloc_put:
gfs2_alloc_put(ip);
}
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
out_unlock:
if (&ip->i_inode == sdp->sd_rindex) {
gfs2_glock_dq(&m_ip->i_gh);
gfs2_holder_uninit(&m_ip->i_gh);
}
gfs2_glock_dq(&ip->i_gh);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
out_uninit:
gfs2_holder_uninit(&ip->i_gh);
return error;
}
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
/**
* adjust_fs_space - Adjusts the free space available due to gfs2_grow
* @inode: the rindex inode
*/
static void adjust_fs_space(struct inode *inode)
{
struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
struct buffer_head *m_bh, *l_bh;
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
u64 fs_total, new_free;
/* Total up the file system space, according to the latest rindex. */
fs_total = gfs2_ri_total(sdp);
if (gfs2_meta_inode_buffer(m_ip, &m_bh) != 0)
return;
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
spin_lock(&sdp->sd_statfs_spin);
gfs2_statfs_change_in(m_sc, m_bh->b_data +
sizeof(struct gfs2_dinode));
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
if (fs_total > (m_sc->sc_total + l_sc->sc_total))
new_free = fs_total - (m_sc->sc_total + l_sc->sc_total);
else
new_free = 0;
spin_unlock(&sdp->sd_statfs_spin);
fs_warn(sdp, "File system extended by %llu blocks.\n",
(unsigned long long)new_free);
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
gfs2_statfs_change(sdp, new_free, new_free, 0);
if (gfs2_meta_inode_buffer(l_ip, &l_bh) != 0)
goto out;
update_statfs(sdp, m_bh, l_bh);
brelse(l_bh);
out:
brelse(m_bh);
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
}
/**
* gfs2_stuffed_write_end - Write end for stuffed files
* @inode: The inode
* @dibh: The buffer_head containing the on-disk inode
* @pos: The file position
* @len: The length of the write
* @copied: How much was actually copied by the VFS
* @page: The page
*
* This copies the data from the page into the inode block after
* the inode data structure itself.
*
* Returns: errno
*/
static int gfs2_stuffed_write_end(struct inode *inode, struct buffer_head *dibh,
loff_t pos, unsigned len, unsigned copied,
struct page *page)
{
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_sbd *sdp = GFS2_SB(inode);
struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
u64 to = pos + copied;
void *kaddr;
unsigned char *buf = dibh->b_data + sizeof(struct gfs2_dinode);
struct gfs2_dinode *di = (struct gfs2_dinode *)dibh->b_data;
BUG_ON((pos + len) > (dibh->b_size - sizeof(struct gfs2_dinode)));
kaddr = kmap_atomic(page, KM_USER0);
memcpy(buf + pos, kaddr + pos, copied);
memset(kaddr + pos + copied, 0, len - copied);
flush_dcache_page(page);
kunmap_atomic(kaddr, KM_USER0);
if (!PageUptodate(page))
SetPageUptodate(page);
unlock_page(page);
page_cache_release(page);
if (copied) {
if (inode->i_size < to)
i_size_write(inode, to);
gfs2_dinode_out(ip, di);
mark_inode_dirty(inode);
}
if (inode == sdp->sd_rindex) {
adjust_fs_space(inode);
ip->i_gh.gh_flags |= GL_NOCACHE;
}
brelse(dibh);
gfs2_trans_end(sdp);
if (inode == sdp->sd_rindex) {
gfs2_glock_dq(&m_ip->i_gh);
gfs2_holder_uninit(&m_ip->i_gh);
}
gfs2_glock_dq(&ip->i_gh);
gfs2_holder_uninit(&ip->i_gh);
return copied;
}
/**
* gfs2_write_end
* @file: The file to write to
* @mapping: The address space to write to
* @pos: The file position
* @len: The length of the data
* @copied:
* @page: The page that has been written
* @fsdata: The fsdata (unused in GFS2)
*
* The main write_end function for GFS2. We have a separate one for
* stuffed files as they are slightly different, otherwise we just
* put our locking around the VFS provided functions.
*
* Returns: errno
*/
static int gfs2_write_end(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned copied,
struct page *page, void *fsdata)
{
struct inode *inode = page->mapping->host;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_sbd *sdp = GFS2_SB(inode);
struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
struct buffer_head *dibh;
struct gfs2_alloc *al = ip->i_alloc;
unsigned int from = pos & (PAGE_CACHE_SIZE - 1);
unsigned int to = from + len;
int ret;
BUG_ON(gfs2_glock_is_locked_by_me(ip->i_gl) == NULL);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
ret = gfs2_meta_inode_buffer(ip, &dibh);
if (unlikely(ret)) {
unlock_page(page);
page_cache_release(page);
goto failed;
}
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
gfs2_trans_add_bh(ip->i_gl, dibh, 1);
if (gfs2_is_stuffed(ip))
return gfs2_stuffed_write_end(inode, dibh, pos, len, copied, page);
if (!gfs2_is_writeback(ip))
gfs2_page_add_databufs(ip, page, from, to);
ret = generic_write_end(file, mapping, pos, len, copied, page, fsdata);
if (ret > 0) {
gfs2_dinode_out(ip, dibh->b_data);
mark_inode_dirty(inode);
}
if (inode == sdp->sd_rindex) {
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
adjust_fs_space(inode);
ip->i_gh.gh_flags |= GL_NOCACHE;
}
[GFS2] kernel changes to support new gfs2_grow command This is another revision of my gfs2 kernel patch that allows gfs2_grow to function properly. Steve Whitehouse expressed some concerns about the previous patch and I restructured it based on his comments. The previous patch was doing the statfs_change at file close time, under its own transaction. The current patch does the statfs_change inside the gfs2_commit_write function, which keeps it under the umbrella of the inode transaction. I can't call ri_update to re-read the rindex file during the transaction because the transaction may have outstanding unwritten buffers attached to the rgrps that would be otherwise blown away. So instead, I created a new function, gfs2_ri_total, that will re-read the rindex file just to total the file system space for the sake of the statfs_change. The ri_update will happen later, when gfs2 realizes the version number has changed, as it happened before my patch. Since the statfs_change is happening at write_commit time and there may be multiple writes to the rindex file for one grow operation. So one consequence of this restructuring is that instead of getting one kernel message to indicate the change, you may see several. For example, before when you did a gfs2_grow, you'd get a single message like: GFS2: File system extended by 247876 blocks (968MB) Now you get something like: GFS2: File system extended by 207896 blocks (812MB) GFS2: File system extended by 39980 blocks (156MB) This version has also been successfully run against the hours-long "gfs2_fsck_hellfire" test that does several gfs2_grow and gfs2_fsck while interjecting file system damage. It does this repeatedly under a variety Resource Group conditions. Signed-off-By: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-09 14:37:57 +00:00
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
brelse(dibh);
failed:
gfs2_trans_end(sdp);
if (al) {
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
gfs2_inplace_release(ip);
gfs2_quota_unlock(ip);
gfs2_alloc_put(ip);
}
if (inode == sdp->sd_rindex) {
gfs2_glock_dq(&m_ip->i_gh);
gfs2_holder_uninit(&m_ip->i_gh);
}
gfs2_glock_dq(&ip->i_gh);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 11:50:51 +00:00
gfs2_holder_uninit(&ip->i_gh);
return ret;
}
/**
* gfs2_set_page_dirty - Page dirtying function
* @page: The page to dirty
*
* Returns: 1 if it dirtyed the page, or 0 otherwise
*/
static int gfs2_set_page_dirty(struct page *page)
{
SetPageChecked(page);
return __set_page_dirty_buffers(page);
}
/**
* gfs2_bmap - Block map function
* @mapping: Address space info
* @lblock: The block to map
*
* Returns: The disk address for the block or 0 on hole or error
*/
static sector_t gfs2_bmap(struct address_space *mapping, sector_t lblock)
{
struct gfs2_inode *ip = GFS2_I(mapping->host);
struct gfs2_holder i_gh;
sector_t dblock = 0;
int error;
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, &i_gh);
if (error)
return 0;
if (!gfs2_is_stuffed(ip))
dblock = generic_block_bmap(mapping, lblock, gfs2_block_map);
gfs2_glock_dq_uninit(&i_gh);
return dblock;
}
static void gfs2_discard(struct gfs2_sbd *sdp, struct buffer_head *bh)
{
struct gfs2_bufdata *bd;
lock_buffer(bh);
gfs2_log_lock(sdp);
clear_buffer_dirty(bh);
bd = bh->b_private;
if (bd) {
if (!list_empty(&bd->bd_le.le_list) && !buffer_pinned(bh))
list_del_init(&bd->bd_le.le_list);
else
gfs2_remove_from_journal(bh, current->journal_info, 0);
}
bh->b_bdev = NULL;
clear_buffer_mapped(bh);
clear_buffer_req(bh);
clear_buffer_new(bh);
gfs2_log_unlock(sdp);
unlock_buffer(bh);
}
static void gfs2_invalidatepage(struct page *page, unsigned long offset)
{
struct gfs2_sbd *sdp = GFS2_SB(page->mapping->host);
struct buffer_head *bh, *head;
unsigned long pos = 0;
BUG_ON(!PageLocked(page));
if (offset == 0)
ClearPageChecked(page);
if (!page_has_buffers(page))
goto out;
bh = head = page_buffers(page);
do {
if (offset <= pos)
gfs2_discard(sdp, bh);
pos += bh->b_size;
bh = bh->b_this_page;
} while (bh != head);
out:
if (offset == 0)
try_to_release_page(page, 0);
}
/**
* gfs2_ok_for_dio - check that dio is valid on this file
* @ip: The inode
* @rw: READ or WRITE
* @offset: The offset at which we are reading or writing
*
* Returns: 0 (to ignore the i/o request and thus fall back to buffered i/o)
* 1 (to accept the i/o request)
*/
static int gfs2_ok_for_dio(struct gfs2_inode *ip, int rw, loff_t offset)
{
/*
* Should we return an error here? I can't see that O_DIRECT for
* a stuffed file makes any sense. For now we'll silently fall
* back to buffered I/O
*/
if (gfs2_is_stuffed(ip))
return 0;
if (offset >= i_size_read(&ip->i_inode))
return 0;
return 1;
}
static ssize_t gfs2_direct_IO(int rw, struct kiocb *iocb,
const struct iovec *iov, loff_t offset,
unsigned long nr_segs)
{
struct file *file = iocb->ki_filp;
struct inode *inode = file->f_mapping->host;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_holder gh;
int rv;
/*
* Deferred lock, even if its a write, since we do no allocation
* on this path. All we need change is atime, and this lock mode
* ensures that other nodes have flushed their buffered read caches
* (i.e. their page cache entries for this inode). We do not,
* unfortunately have the option of only flushing a range like
* the VFS does.
*/
gfs2_holder_init(ip->i_gl, LM_ST_DEFERRED, 0, &gh);
rv = gfs2_glock_nq(&gh);
if (rv)
return rv;
rv = gfs2_ok_for_dio(ip, rw, offset);
if (rv != 1)
goto out; /* dio not valid, fall back to buffered i/o */
rv = __blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
offset, nr_segs, gfs2_get_block_direct,
NULL, NULL, 0);
out:
gfs2_glock_dq_m(1, &gh);
gfs2_holder_uninit(&gh);
return rv;
}
/**
* gfs2_releasepage - free the metadata associated with a page
* @page: the page that's being released
* @gfp_mask: passed from Linux VFS, ignored by us
*
* Call try_to_free_buffers() if the buffers in this page can be
* released.
*
* Returns: 0
*/
int gfs2_releasepage(struct page *page, gfp_t gfp_mask)
{
struct address_space *mapping = page->mapping;
struct gfs2_sbd *sdp = gfs2_mapping2sbd(mapping);
struct buffer_head *bh, *head;
struct gfs2_bufdata *bd;
if (!page_has_buffers(page))
return 0;
gfs2_log_lock(sdp);
spin_lock(&sdp->sd_ail_lock);
head = bh = page_buffers(page);
do {
if (atomic_read(&bh->b_count))
goto cannot_release;
bd = bh->b_private;
if (bd && bd->bd_ail)
goto cannot_release;
if (buffer_pinned(bh) || buffer_dirty(bh))
goto not_possible;
bh = bh->b_this_page;
} while(bh != head);
spin_unlock(&sdp->sd_ail_lock);
gfs2_log_unlock(sdp);
head = bh = page_buffers(page);
do {
gfs2_log_lock(sdp);
bd = bh->b_private;
if (bd) {
gfs2_assert_warn(sdp, bd->bd_bh == bh);
gfs2_assert_warn(sdp, list_empty(&bd->bd_list_tr));
if (!list_empty(&bd->bd_le.le_list)) {
if (!buffer_pinned(bh))
list_del_init(&bd->bd_le.le_list);
else
bd = NULL;
}
if (bd)
bd->bd_bh = NULL;
bh->b_private = NULL;
}
gfs2_log_unlock(sdp);
if (bd)
kmem_cache_free(gfs2_bufdata_cachep, bd);
bh = bh->b_this_page;
} while (bh != head);
return try_to_free_buffers(page);
not_possible: /* Should never happen */
WARN_ON(buffer_dirty(bh));
WARN_ON(buffer_pinned(bh));
cannot_release:
spin_unlock(&sdp->sd_ail_lock);
gfs2_log_unlock(sdp);
return 0;
}
static const struct address_space_operations gfs2_writeback_aops = {
.writepage = gfs2_writeback_writepage,
.writepages = gfs2_writeback_writepages,
.readpage = gfs2_readpage,
.readpages = gfs2_readpages,
.write_begin = gfs2_write_begin,
.write_end = gfs2_write_end,
.bmap = gfs2_bmap,
.invalidatepage = gfs2_invalidatepage,
.releasepage = gfs2_releasepage,
.direct_IO = gfs2_direct_IO,
.migratepage = buffer_migrate_page,
GFS2: Pagecache usage optimization on GFS2 I introduced "is_partially_uptodate" aops for GFS2. A page can have multiple buffers and even if a page is not uptodate, some buffers can be uptodate on pagesize != blocksize environment. This aops checks that all buffers which correspond to a part of a file that we want to read are uptodate. If so, we do not have to issue actual read IO to HDD even if a page is not uptodate because the portion we want to read are uptodate. "block_is_partially_uptodate" function is already used by ext2/3/4. With the following patch random read/write mixed workloads or random read after random write workloads can be optimized and we can get performance improvement. I did a performance test using the sysbench. #sysbench --num-threads=16 --max-requests=200000 --test=fileio --file-num=1 --file-block-size=8K --file-total-size=2G --file-test-mode=rndrw --file-fsync-freq=0 --file-rw-ratio=1 run -2.6.29-rc6 Test execution summary: total time: 202.6389s total number of events: 200000 total time taken by event execution: 2580.0480 per-request statistics: min: 0.0000s avg: 0.0129s max: 49.5852s approx. 95 percentile: 0.0462s -2.6.29-rc6-patched Test execution summary: total time: 177.8639s total number of events: 200000 total time taken by event execution: 2419.0199 per-request statistics: min: 0.0000s avg: 0.0121s max: 52.4306s approx. 95 percentile: 0.0444s arch: ia64 pagesize: 16k blocksize: 4k Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2009-03-03 02:45:20 +00:00
.is_partially_uptodate = block_is_partially_uptodate,
.error_remove_page = generic_error_remove_page,
};
static const struct address_space_operations gfs2_ordered_aops = {
.writepage = gfs2_ordered_writepage,
.readpage = gfs2_readpage,
.readpages = gfs2_readpages,
.write_begin = gfs2_write_begin,
.write_end = gfs2_write_end,
.set_page_dirty = gfs2_set_page_dirty,
.bmap = gfs2_bmap,
.invalidatepage = gfs2_invalidatepage,
.releasepage = gfs2_releasepage,
.direct_IO = gfs2_direct_IO,
.migratepage = buffer_migrate_page,
GFS2: Pagecache usage optimization on GFS2 I introduced "is_partially_uptodate" aops for GFS2. A page can have multiple buffers and even if a page is not uptodate, some buffers can be uptodate on pagesize != blocksize environment. This aops checks that all buffers which correspond to a part of a file that we want to read are uptodate. If so, we do not have to issue actual read IO to HDD even if a page is not uptodate because the portion we want to read are uptodate. "block_is_partially_uptodate" function is already used by ext2/3/4. With the following patch random read/write mixed workloads or random read after random write workloads can be optimized and we can get performance improvement. I did a performance test using the sysbench. #sysbench --num-threads=16 --max-requests=200000 --test=fileio --file-num=1 --file-block-size=8K --file-total-size=2G --file-test-mode=rndrw --file-fsync-freq=0 --file-rw-ratio=1 run -2.6.29-rc6 Test execution summary: total time: 202.6389s total number of events: 200000 total time taken by event execution: 2580.0480 per-request statistics: min: 0.0000s avg: 0.0129s max: 49.5852s approx. 95 percentile: 0.0462s -2.6.29-rc6-patched Test execution summary: total time: 177.8639s total number of events: 200000 total time taken by event execution: 2419.0199 per-request statistics: min: 0.0000s avg: 0.0121s max: 52.4306s approx. 95 percentile: 0.0444s arch: ia64 pagesize: 16k blocksize: 4k Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2009-03-03 02:45:20 +00:00
.is_partially_uptodate = block_is_partially_uptodate,
.error_remove_page = generic_error_remove_page,
};
static const struct address_space_operations gfs2_jdata_aops = {
.writepage = gfs2_jdata_writepage,
.writepages = gfs2_jdata_writepages,
.readpage = gfs2_readpage,
.readpages = gfs2_readpages,
.write_begin = gfs2_write_begin,
.write_end = gfs2_write_end,
.set_page_dirty = gfs2_set_page_dirty,
.bmap = gfs2_bmap,
.invalidatepage = gfs2_invalidatepage,
.releasepage = gfs2_releasepage,
GFS2: Pagecache usage optimization on GFS2 I introduced "is_partially_uptodate" aops for GFS2. A page can have multiple buffers and even if a page is not uptodate, some buffers can be uptodate on pagesize != blocksize environment. This aops checks that all buffers which correspond to a part of a file that we want to read are uptodate. If so, we do not have to issue actual read IO to HDD even if a page is not uptodate because the portion we want to read are uptodate. "block_is_partially_uptodate" function is already used by ext2/3/4. With the following patch random read/write mixed workloads or random read after random write workloads can be optimized and we can get performance improvement. I did a performance test using the sysbench. #sysbench --num-threads=16 --max-requests=200000 --test=fileio --file-num=1 --file-block-size=8K --file-total-size=2G --file-test-mode=rndrw --file-fsync-freq=0 --file-rw-ratio=1 run -2.6.29-rc6 Test execution summary: total time: 202.6389s total number of events: 200000 total time taken by event execution: 2580.0480 per-request statistics: min: 0.0000s avg: 0.0129s max: 49.5852s approx. 95 percentile: 0.0462s -2.6.29-rc6-patched Test execution summary: total time: 177.8639s total number of events: 200000 total time taken by event execution: 2419.0199 per-request statistics: min: 0.0000s avg: 0.0121s max: 52.4306s approx. 95 percentile: 0.0444s arch: ia64 pagesize: 16k blocksize: 4k Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2009-03-03 02:45:20 +00:00
.is_partially_uptodate = block_is_partially_uptodate,
.error_remove_page = generic_error_remove_page,
};
void gfs2_set_aops(struct inode *inode)
{
struct gfs2_inode *ip = GFS2_I(inode);
if (gfs2_is_writeback(ip))
inode->i_mapping->a_ops = &gfs2_writeback_aops;
else if (gfs2_is_ordered(ip))
inode->i_mapping->a_ops = &gfs2_ordered_aops;
else if (gfs2_is_jdata(ip))
inode->i_mapping->a_ops = &gfs2_jdata_aops;
else
BUG();
}