2007-06-12 09:07:21 -04:00
|
|
|
/*
|
|
|
|
* Copyright (C) 2007 Oracle. All rights reserved.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public
|
|
|
|
* License v2 as published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU General Public
|
|
|
|
* License along with this program; if not, write to the
|
|
|
|
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
|
|
|
|
* Boston, MA 021110-1307, USA.
|
|
|
|
*/
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/pagemap.h>
|
|
|
|
#include <linux/highmem.h>
|
|
|
|
#include <linux/time.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/string.h>
|
|
|
|
#include <linux/smp_lock.h>
|
|
|
|
#include <linux/backing-dev.h>
|
|
|
|
#include <linux/mpage.h>
|
|
|
|
#include <linux/swap.h>
|
|
|
|
#include <linux/writeback.h>
|
|
|
|
#include <linux/statfs.h>
|
|
|
|
#include <linux/compat.h>
|
2007-07-25 12:31:35 -04:00
|
|
|
#include <linux/version.h>
|
2007-06-12 06:35:45 -04:00
|
|
|
#include "ctree.h"
|
|
|
|
#include "disk-io.h"
|
|
|
|
#include "transaction.h"
|
|
|
|
#include "btrfs_inode.h"
|
|
|
|
#include "ioctl.h"
|
|
|
|
#include "print-tree.h"
|
2008-09-05 16:13:11 -04:00
|
|
|
#include "tree-log.h"
|
|
|
|
#include "locking.h"
|
Btrfs: Add workaround for AppArmor changing remove_suid()
In openSUSE 10.3, AppArmor modifies remove_suid to take a struct path
rather than just a dentry. This patch tests that the kernel is openSUSE
10.3 or newer and adjusts the call accordingly.
Debian/Ubuntu with AppArmor applied will also need a similar patch.
Maintainers of btrfs under those distributions should build on this
patch or, alternatively, alter their package descriptions to add
-DREMOVE_SUID_PATH to the compiler command line.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
- --- /dev/null 1970-01-01 00:00:00.000000000 +0000
+++ b/compat.h 2008-02-06 16:46:13.000000000 -0500
@@ -0,0 +1,15 @@
+#ifndef _COMPAT_H_
+#define _COMPAT_H_
+
+
+/*
+ * Even if AppArmor isn't enabled, it still has different prototypes.
+ * Add more distro/version pairs here to declare which has AppArmor applied.
+ */
+#if defined(CONFIG_SUSE_KERNEL)
+# if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
+# define REMOVE_SUID_PATH 1
+# endif
+#endif
+
+#endif /* _COMPAT_H_ */
- --- a/file.c 2008-02-06 11:37:39.000000000 -0500
+++ b/file.c 2008-02-06 16:46:23.000000000 -0500
@@ -37,6 +37,7 @@
#include "ordered-data.h"
#include "ioctl.h"
#include "print-tree.h"
+#include "compat.h"
static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes,
@@ -790,7 +791,11 @@ static ssize_t btrfs_file_write(struct f
goto out_nolock;
if (count == 0)
goto out_nolock;
+#ifdef REMOVE_SUID_PATH
+ err = remove_suid(&file->f_path);
+#else
err = remove_suid(fdentry(file));
+#endif
if (err)
goto out_nolock;
file_update_time(file);
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-05-02 15:03:58 -04:00
|
|
|
#include "compat.h"
|
2007-06-12 06:35:45 -04:00
|
|
|
|
|
|
|
|
2008-09-29 15:18:18 -04:00
|
|
|
/* simple helper to fault in pages and copy. This should go away
|
|
|
|
* and be replaced with calls into generic code.
|
|
|
|
*/
|
2008-09-05 16:09:51 -04:00
|
|
|
static int noinline btrfs_copy_from_user(loff_t pos, int num_pages,
|
|
|
|
int write_bytes,
|
|
|
|
struct page **prepared_pages,
|
|
|
|
const char __user * buf)
|
2007-06-12 06:35:45 -04:00
|
|
|
{
|
|
|
|
long page_fault = 0;
|
|
|
|
int i;
|
|
|
|
int offset = pos & (PAGE_CACHE_SIZE - 1);
|
|
|
|
|
|
|
|
for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) {
|
|
|
|
size_t count = min_t(size_t,
|
|
|
|
PAGE_CACHE_SIZE - offset, write_bytes);
|
|
|
|
struct page *page = prepared_pages[i];
|
|
|
|
fault_in_pages_readable(buf, count);
|
|
|
|
|
|
|
|
/* Copy data from userspace to the current page */
|
|
|
|
kmap(page);
|
|
|
|
page_fault = __copy_from_user(page_address(page) + offset,
|
|
|
|
buf, count);
|
|
|
|
/* Flush processor's dcache for this page */
|
|
|
|
flush_dcache_page(page);
|
|
|
|
kunmap(page);
|
|
|
|
buf += count;
|
|
|
|
write_bytes -= count;
|
|
|
|
|
|
|
|
if (page_fault)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return page_fault ? -EFAULT : 0;
|
|
|
|
}
|
|
|
|
|
2008-09-29 15:18:18 -04:00
|
|
|
/*
|
|
|
|
* unlocks pages after btrfs_file_write is done with them
|
|
|
|
*/
|
2008-09-05 16:09:51 -04:00
|
|
|
static void noinline btrfs_drop_pages(struct page **pages, size_t num_pages)
|
2007-06-12 06:35:45 -04:00
|
|
|
{
|
|
|
|
size_t i;
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
if (!pages[i])
|
|
|
|
break;
|
2008-09-29 15:18:18 -04:00
|
|
|
/* page checked is some magic around finding pages that
|
|
|
|
* have been modified without going through btrfs_set_page_dirty
|
|
|
|
* clear it here
|
|
|
|
*/
|
2008-07-21 10:29:44 -04:00
|
|
|
ClearPageChecked(pages[i]);
|
2007-06-12 06:35:45 -04:00
|
|
|
unlock_page(pages[i]);
|
|
|
|
mark_page_accessed(pages[i]);
|
|
|
|
page_cache_release(pages[i]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-09-29 15:18:18 -04:00
|
|
|
/*
|
|
|
|
* after copy_from_user, pages need to be dirtied and we need to make
|
|
|
|
* sure holes are created between the current EOF and the start of
|
|
|
|
* any next extents (if required).
|
|
|
|
*
|
|
|
|
* this also makes the decision about creating an inline extent vs
|
|
|
|
* doing real data extents, marking pages dirty and delalloc as required.
|
|
|
|
*/
|
2008-01-03 10:01:48 -05:00
|
|
|
static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
|
2007-06-12 06:35:45 -04:00
|
|
|
struct btrfs_root *root,
|
|
|
|
struct file *file,
|
|
|
|
struct page **pages,
|
|
|
|
size_t num_pages,
|
|
|
|
loff_t pos,
|
|
|
|
size_t write_bytes)
|
|
|
|
{
|
|
|
|
int err = 0;
|
2007-08-27 16:49:44 -04:00
|
|
|
int i;
|
2007-12-18 16:15:09 -05:00
|
|
|
struct inode *inode = fdentry(file)->d_inode;
|
2008-01-24 16:13:08 -05:00
|
|
|
struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
|
2007-10-15 16:15:53 -04:00
|
|
|
u64 hint_byte;
|
|
|
|
u64 num_bytes;
|
2007-08-27 16:49:44 -04:00
|
|
|
u64 start_pos;
|
|
|
|
u64 end_of_last_block;
|
|
|
|
u64 end_pos = pos + write_bytes;
|
|
|
|
loff_t isize = i_size_read(inode);
|
2007-06-12 06:35:45 -04:00
|
|
|
|
2007-10-15 16:14:19 -04:00
|
|
|
start_pos = pos & ~((u64)root->sectorsize - 1);
|
2007-10-15 16:15:53 -04:00
|
|
|
num_bytes = (write_bytes + pos - start_pos +
|
|
|
|
root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
|
2007-06-12 06:35:45 -04:00
|
|
|
|
2007-10-15 16:15:53 -04:00
|
|
|
end_of_last_block = start_pos + num_bytes - 1;
|
|
|
|
|
2008-01-24 16:13:08 -05:00
|
|
|
lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS);
|
2008-07-31 10:48:37 -04:00
|
|
|
trans = btrfs_join_transaction(root, 1);
|
2007-08-27 16:49:44 -04:00
|
|
|
if (!trans) {
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
btrfs_set_trans_block_group(trans, inode);
|
2007-10-15 16:15:53 -04:00
|
|
|
hint_byte = 0;
|
2007-08-27 16:49:44 -04:00
|
|
|
|
|
|
|
if ((end_of_last_block & 4095) == 0) {
|
2007-09-10 20:02:22 -04:00
|
|
|
printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block);
|
2007-08-27 16:49:44 -04:00
|
|
|
}
|
2008-01-24 16:13:08 -05:00
|
|
|
set_extent_uptodate(io_tree, start_pos, end_of_last_block, GFP_NOFS);
|
2007-08-27 16:49:44 -04:00
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
/* check for reserved extents on each page, we don't want
|
|
|
|
* to reset the delalloc bit on things that already have
|
|
|
|
* extents reserved.
|
2007-08-27 16:49:44 -04:00
|
|
|
*/
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block);
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
struct page *p = pages[i];
|
|
|
|
SetPageUptodate(p);
|
|
|
|
ClearPageChecked(p);
|
|
|
|
set_page_dirty(p);
|
2007-08-27 16:49:44 -04:00
|
|
|
}
|
|
|
|
if (end_pos > isize) {
|
|
|
|
i_size_write(inode, end_pos);
|
|
|
|
btrfs_update_inode(trans, root, inode);
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
2008-07-28 15:32:51 -04:00
|
|
|
err = btrfs_end_transaction(trans, root);
|
2007-08-27 16:49:44 -04:00
|
|
|
out_unlock:
|
2008-01-24 16:13:08 -05:00
|
|
|
unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS);
|
2007-06-12 06:35:45 -04:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2008-09-29 15:18:18 -04:00
|
|
|
/*
|
|
|
|
* this drops all the extents in the cache that intersect the range
|
|
|
|
* [start, end]. Existing extents are split as required.
|
|
|
|
*/
|
2008-09-26 10:05:38 -04:00
|
|
|
int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
|
|
|
|
int skip_pinned)
|
2007-08-27 16:49:44 -04:00
|
|
|
{
|
|
|
|
struct extent_map *em;
|
2008-04-17 11:29:12 -04:00
|
|
|
struct extent_map *split = NULL;
|
|
|
|
struct extent_map *split2 = NULL;
|
2007-08-27 16:49:44 -04:00
|
|
|
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
|
2008-02-15 10:40:50 -05:00
|
|
|
u64 len = end - start + 1;
|
2008-04-17 11:29:12 -04:00
|
|
|
int ret;
|
|
|
|
int testend = 1;
|
2008-09-26 10:05:38 -04:00
|
|
|
unsigned long flags;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
int compressed = 0;
|
2007-08-27 16:49:44 -04:00
|
|
|
|
2008-07-17 12:53:50 -04:00
|
|
|
WARN_ON(end < start);
|
2008-04-17 11:29:12 -04:00
|
|
|
if (end == (u64)-1) {
|
2008-02-15 10:40:50 -05:00
|
|
|
len = (u64)-1;
|
2008-04-17 11:29:12 -04:00
|
|
|
testend = 0;
|
|
|
|
}
|
2007-08-27 16:49:44 -04:00
|
|
|
while(1) {
|
2008-04-17 11:29:12 -04:00
|
|
|
if (!split)
|
|
|
|
split = alloc_extent_map(GFP_NOFS);
|
|
|
|
if (!split2)
|
|
|
|
split2 = alloc_extent_map(GFP_NOFS);
|
|
|
|
|
2008-01-24 16:13:08 -05:00
|
|
|
spin_lock(&em_tree->lock);
|
2008-02-15 10:40:50 -05:00
|
|
|
em = lookup_extent_mapping(em_tree, start, len);
|
2008-01-24 16:13:08 -05:00
|
|
|
if (!em) {
|
|
|
|
spin_unlock(&em_tree->lock);
|
2007-08-27 16:49:44 -04:00
|
|
|
break;
|
2008-01-24 16:13:08 -05:00
|
|
|
}
|
2008-09-26 10:05:38 -04:00
|
|
|
flags = em->flags;
|
|
|
|
if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) {
|
|
|
|
spin_unlock(&em_tree->lock);
|
|
|
|
if (em->start <= start &&
|
|
|
|
(!testend || em->start + em->len >= start + len)) {
|
|
|
|
free_extent_map(em);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (start < em->start) {
|
|
|
|
len = em->start - start;
|
|
|
|
} else {
|
|
|
|
len = start + len - (em->start + em->len);
|
|
|
|
start = em->start + em->len;
|
|
|
|
}
|
|
|
|
free_extent_map(em);
|
|
|
|
continue;
|
|
|
|
}
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
|
2008-07-31 15:42:54 -04:00
|
|
|
clear_bit(EXTENT_FLAG_PINNED, &em->flags);
|
2007-08-27 16:49:44 -04:00
|
|
|
remove_extent_mapping(em_tree, em);
|
2008-04-17 11:29:12 -04:00
|
|
|
|
|
|
|
if (em->block_start < EXTENT_MAP_LAST_BYTE &&
|
|
|
|
em->start < start) {
|
|
|
|
split->start = em->start;
|
|
|
|
split->len = start - em->start;
|
|
|
|
split->block_start = em->block_start;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
|
|
|
|
if (compressed)
|
|
|
|
split->block_len = em->block_len;
|
|
|
|
else
|
|
|
|
split->block_len = split->len;
|
|
|
|
|
2008-04-17 11:29:12 -04:00
|
|
|
split->bdev = em->bdev;
|
2008-09-26 10:05:38 -04:00
|
|
|
split->flags = flags;
|
2008-04-17 11:29:12 -04:00
|
|
|
ret = add_extent_mapping(em_tree, split);
|
|
|
|
BUG_ON(ret);
|
|
|
|
free_extent_map(split);
|
|
|
|
split = split2;
|
|
|
|
split2 = NULL;
|
|
|
|
}
|
|
|
|
if (em->block_start < EXTENT_MAP_LAST_BYTE &&
|
|
|
|
testend && em->start + em->len > start + len) {
|
|
|
|
u64 diff = start + len - em->start;
|
|
|
|
|
|
|
|
split->start = start + len;
|
|
|
|
split->len = em->start + em->len - (start + len);
|
|
|
|
split->bdev = em->bdev;
|
2008-09-26 10:05:38 -04:00
|
|
|
split->flags = flags;
|
2008-04-17 11:29:12 -04:00
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
if (compressed) {
|
|
|
|
split->block_len = em->block_len;
|
|
|
|
split->block_start = em->block_start;
|
|
|
|
} else {
|
|
|
|
split->block_len = split->len;
|
|
|
|
split->block_start = em->block_start + diff;
|
|
|
|
}
|
2008-04-17 11:29:12 -04:00
|
|
|
|
|
|
|
ret = add_extent_mapping(em_tree, split);
|
|
|
|
BUG_ON(ret);
|
|
|
|
free_extent_map(split);
|
|
|
|
split = NULL;
|
|
|
|
}
|
2008-01-24 16:13:08 -05:00
|
|
|
spin_unlock(&em_tree->lock);
|
|
|
|
|
2007-08-27 16:49:44 -04:00
|
|
|
/* once for us */
|
|
|
|
free_extent_map(em);
|
|
|
|
/* once for the tree*/
|
|
|
|
free_extent_map(em);
|
|
|
|
}
|
2008-04-17 11:29:12 -04:00
|
|
|
if (split)
|
|
|
|
free_extent_map(split);
|
|
|
|
if (split2)
|
|
|
|
free_extent_map(split2);
|
2007-08-27 16:49:44 -04:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-01-22 16:47:59 -05:00
|
|
|
int btrfs_check_file(struct btrfs_root *root, struct inode *inode)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
#if 0
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_key found_key;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_file_extent_item *extent;
|
|
|
|
u64 last_offset = 0;
|
|
|
|
int nritems;
|
|
|
|
int slot;
|
|
|
|
int found_type;
|
|
|
|
int ret;
|
|
|
|
int err = 0;
|
|
|
|
u64 extent_end = 0;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
ret = btrfs_lookup_file_extent(NULL, root, path, inode->i_ino,
|
|
|
|
last_offset, 0);
|
|
|
|
while(1) {
|
|
|
|
nritems = btrfs_header_nritems(path->nodes[0]);
|
|
|
|
if (path->slots[0] >= nritems) {
|
|
|
|
ret = btrfs_next_leaf(root, path);
|
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
nritems = btrfs_header_nritems(path->nodes[0]);
|
|
|
|
}
|
|
|
|
slot = path->slots[0];
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
btrfs_item_key_to_cpu(leaf, &found_key, slot);
|
|
|
|
if (found_key.objectid != inode->i_ino)
|
|
|
|
break;
|
|
|
|
if (found_key.type != BTRFS_EXTENT_DATA_KEY)
|
|
|
|
goto out;
|
|
|
|
|
2008-02-08 13:49:28 -05:00
|
|
|
if (found_key.offset < last_offset) {
|
2008-01-22 16:47:59 -05:00
|
|
|
WARN_ON(1);
|
|
|
|
btrfs_print_leaf(root, leaf);
|
|
|
|
printk("inode %lu found offset %Lu expected %Lu\n",
|
|
|
|
inode->i_ino, found_key.offset, last_offset);
|
|
|
|
err = 1;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
extent = btrfs_item_ptr(leaf, slot,
|
|
|
|
struct btrfs_file_extent_item);
|
|
|
|
found_type = btrfs_file_extent_type(leaf, extent);
|
|
|
|
if (found_type == BTRFS_FILE_EXTENT_REG) {
|
|
|
|
extent_end = found_key.offset +
|
|
|
|
btrfs_file_extent_num_bytes(leaf, extent);
|
|
|
|
} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
|
|
|
|
struct btrfs_item *item;
|
|
|
|
item = btrfs_item_nr(leaf, slot);
|
|
|
|
extent_end = found_key.offset +
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
btrfs_file_extent_inline_len(leaf, extent);
|
2008-01-22 16:47:59 -05:00
|
|
|
extent_end = (extent_end + root->sectorsize - 1) &
|
|
|
|
~((u64)root->sectorsize -1 );
|
|
|
|
}
|
|
|
|
last_offset = extent_end;
|
|
|
|
path->slots[0]++;
|
|
|
|
}
|
2008-02-08 13:49:28 -05:00
|
|
|
if (0 && last_offset < inode->i_size) {
|
2008-01-22 16:47:59 -05:00
|
|
|
WARN_ON(1);
|
|
|
|
btrfs_print_leaf(root, leaf);
|
|
|
|
printk("inode %lu found offset %Lu size %Lu\n", inode->i_ino,
|
|
|
|
last_offset, inode->i_size);
|
|
|
|
err = 1;
|
|
|
|
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return err;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
/*
|
|
|
|
* this is very complex, but the basic idea is to drop all extents
|
|
|
|
* in the range start - end. hint_block is filled in with a block number
|
|
|
|
* that would be a good hint to the block allocator for this file.
|
|
|
|
*
|
|
|
|
* If an extent intersects the range but is not entirely inside the range
|
|
|
|
* it is either truncated or split. Anything entirely inside the range
|
|
|
|
* is deleted from the tree.
|
2008-09-29 15:18:18 -04:00
|
|
|
*
|
|
|
|
* inline_limit is used to tell this code which offsets in the file to keep
|
|
|
|
* if they contain inline extents.
|
2007-06-12 06:35:45 -04:00
|
|
|
*/
|
2008-09-05 16:09:51 -04:00
|
|
|
int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans,
|
2007-06-12 06:35:45 -04:00
|
|
|
struct btrfs_root *root, struct inode *inode,
|
2007-11-30 10:09:33 -05:00
|
|
|
u64 start, u64 end, u64 inline_limit, u64 *hint_byte)
|
2007-06-12 06:35:45 -04:00
|
|
|
{
|
2007-11-30 10:09:33 -05:00
|
|
|
u64 extent_end = 0;
|
2008-10-30 14:19:50 -04:00
|
|
|
u64 locked_end = end;
|
2007-11-30 10:09:33 -05:00
|
|
|
u64 search_start = start;
|
2008-09-23 13:14:14 -04:00
|
|
|
u64 leaf_start;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
u64 ram_bytes = 0;
|
|
|
|
u8 compression = 0;
|
|
|
|
u8 encryption = 0;
|
|
|
|
u16 other_encoding = 0;
|
2008-09-23 13:14:14 -04:00
|
|
|
u64 root_gen;
|
|
|
|
u64 root_owner;
|
2007-10-15 16:14:19 -04:00
|
|
|
struct extent_buffer *leaf;
|
2007-06-12 06:35:45 -04:00
|
|
|
struct btrfs_file_extent_item *extent;
|
|
|
|
struct btrfs_path *path;
|
2007-11-30 10:09:33 -05:00
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_file_extent_item old;
|
|
|
|
int keep;
|
|
|
|
int slot;
|
2007-06-12 06:35:45 -04:00
|
|
|
int bookend;
|
|
|
|
int found_type;
|
|
|
|
int found_extent;
|
|
|
|
int found_inline;
|
2007-06-28 15:57:36 -04:00
|
|
|
int recow;
|
2007-11-30 10:09:33 -05:00
|
|
|
int ret;
|
2007-06-12 06:35:45 -04:00
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
inline_limit = 0;
|
2008-09-26 10:05:38 -04:00
|
|
|
btrfs_drop_extent_cache(inode, start, end - 1, 0);
|
2007-08-27 16:49:44 -04:00
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
while(1) {
|
2007-06-28 15:57:36 -04:00
|
|
|
recow = 0;
|
2007-06-12 06:35:45 -04:00
|
|
|
btrfs_release_path(root, path);
|
|
|
|
ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
|
|
|
|
search_start, -1);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
if (ret > 0) {
|
|
|
|
if (path->slots[0] == 0) {
|
|
|
|
ret = 0;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
path->slots[0]--;
|
|
|
|
}
|
2007-06-18 09:57:58 -04:00
|
|
|
next_slot:
|
2007-06-12 06:35:45 -04:00
|
|
|
keep = 0;
|
|
|
|
bookend = 0;
|
|
|
|
found_extent = 0;
|
|
|
|
found_inline = 0;
|
2008-09-23 13:14:14 -04:00
|
|
|
leaf_start = 0;
|
|
|
|
root_gen = 0;
|
|
|
|
root_owner = 0;
|
2007-06-12 06:35:45 -04:00
|
|
|
extent = NULL;
|
2007-10-15 16:14:19 -04:00
|
|
|
leaf = path->nodes[0];
|
2007-06-12 06:35:45 -04:00
|
|
|
slot = path->slots[0];
|
2007-06-18 09:57:58 -04:00
|
|
|
ret = 0;
|
2007-10-15 16:14:19 -04:00
|
|
|
btrfs_item_key_to_cpu(leaf, &key, slot);
|
2008-02-05 15:40:36 -05:00
|
|
|
if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY &&
|
|
|
|
key.offset >= end) {
|
2007-06-12 06:35:45 -04:00
|
|
|
goto out;
|
|
|
|
}
|
2008-02-05 15:40:36 -05:00
|
|
|
if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
|
|
|
|
key.objectid != inode->i_ino) {
|
2007-06-12 06:35:45 -04:00
|
|
|
goto out;
|
|
|
|
}
|
2007-06-28 15:57:36 -04:00
|
|
|
if (recow) {
|
|
|
|
search_start = key.offset;
|
|
|
|
continue;
|
|
|
|
}
|
2007-06-18 09:57:58 -04:00
|
|
|
if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
|
|
|
|
extent = btrfs_item_ptr(leaf, slot,
|
|
|
|
struct btrfs_file_extent_item);
|
2007-10-15 16:14:19 -04:00
|
|
|
found_type = btrfs_file_extent_type(leaf, extent);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
compression = btrfs_file_extent_compression(leaf,
|
|
|
|
extent);
|
|
|
|
encryption = btrfs_file_extent_encryption(leaf,
|
|
|
|
extent);
|
|
|
|
other_encoding = btrfs_file_extent_other_encoding(leaf,
|
|
|
|
extent);
|
2007-06-18 09:57:58 -04:00
|
|
|
if (found_type == BTRFS_FILE_EXTENT_REG) {
|
2007-11-07 21:08:16 -05:00
|
|
|
extent_end =
|
|
|
|
btrfs_file_extent_disk_bytenr(leaf,
|
|
|
|
extent);
|
|
|
|
if (extent_end)
|
|
|
|
*hint_byte = extent_end;
|
|
|
|
|
2007-06-18 09:57:58 -04:00
|
|
|
extent_end = key.offset +
|
2007-10-15 16:15:53 -04:00
|
|
|
btrfs_file_extent_num_bytes(leaf, extent);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
ram_bytes = btrfs_file_extent_ram_bytes(leaf,
|
|
|
|
extent);
|
2007-06-18 09:57:58 -04:00
|
|
|
found_extent = 1;
|
|
|
|
} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
|
|
|
|
found_inline = 1;
|
|
|
|
extent_end = key.offset +
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
btrfs_file_extent_inline_len(leaf, extent);
|
2007-06-18 09:57:58 -04:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
extent_end = search_start;
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* we found nothing we can drop */
|
2007-06-18 09:57:58 -04:00
|
|
|
if ((!found_extent && !found_inline) ||
|
|
|
|
search_start >= extent_end) {
|
|
|
|
int nextret;
|
|
|
|
u32 nritems;
|
2007-10-15 16:14:19 -04:00
|
|
|
nritems = btrfs_header_nritems(leaf);
|
2007-06-18 09:57:58 -04:00
|
|
|
if (slot >= nritems - 1) {
|
|
|
|
nextret = btrfs_next_leaf(root, path);
|
|
|
|
if (nextret)
|
|
|
|
goto out;
|
2007-06-28 15:57:36 -04:00
|
|
|
recow = 1;
|
2007-06-18 09:57:58 -04:00
|
|
|
} else {
|
|
|
|
path->slots[0]++;
|
|
|
|
}
|
|
|
|
goto next_slot;
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
if (end <= extent_end && start >= key.offset && found_inline)
|
2007-11-01 11:28:41 -04:00
|
|
|
*hint_byte = EXTENT_MAP_INLINE;
|
2008-09-23 13:14:14 -04:00
|
|
|
|
|
|
|
if (found_extent) {
|
|
|
|
read_extent_buffer(leaf, &old, (unsigned long)extent,
|
|
|
|
sizeof(old));
|
|
|
|
root_gen = btrfs_header_generation(leaf);
|
|
|
|
root_owner = btrfs_header_owner(leaf);
|
|
|
|
leaf_start = leaf->start;
|
2007-11-01 11:28:41 -04:00
|
|
|
}
|
2008-09-23 13:14:14 -04:00
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
if (end < extent_end && end >= key.offset) {
|
2007-11-01 11:28:41 -04:00
|
|
|
bookend = 1;
|
2008-01-30 14:39:54 -05:00
|
|
|
if (found_inline && start <= key.offset)
|
2007-11-01 11:28:41 -04:00
|
|
|
keep = 1;
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
2008-10-30 14:19:50 -04:00
|
|
|
|
|
|
|
if (bookend && found_extent && locked_end < extent_end) {
|
|
|
|
ret = try_lock_extent(&BTRFS_I(inode)->io_tree,
|
|
|
|
locked_end, extent_end - 1, GFP_NOFS);
|
|
|
|
if (!ret) {
|
|
|
|
btrfs_release_path(root, path);
|
|
|
|
lock_extent(&BTRFS_I(inode)->io_tree,
|
|
|
|
locked_end, extent_end - 1, GFP_NOFS);
|
|
|
|
locked_end = extent_end;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
locked_end = extent_end;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (found_inline) {
|
|
|
|
u64 mask = root->sectorsize - 1;
|
|
|
|
search_start = (extent_end + mask) & ~mask;
|
|
|
|
} else
|
|
|
|
search_start = extent_end;
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
/* truncate existing extent */
|
|
|
|
if (start > key.offset) {
|
|
|
|
u64 new_num;
|
|
|
|
u64 old_num;
|
|
|
|
keep = 1;
|
2007-10-15 16:14:19 -04:00
|
|
|
WARN_ON(start & (root->sectorsize - 1));
|
2007-06-12 06:35:45 -04:00
|
|
|
if (found_extent) {
|
2007-10-15 16:15:53 -04:00
|
|
|
new_num = start - key.offset;
|
|
|
|
old_num = btrfs_file_extent_num_bytes(leaf,
|
|
|
|
extent);
|
|
|
|
*hint_byte =
|
|
|
|
btrfs_file_extent_disk_bytenr(leaf,
|
|
|
|
extent);
|
|
|
|
if (btrfs_file_extent_disk_bytenr(leaf,
|
|
|
|
extent)) {
|
2008-10-09 11:46:29 -04:00
|
|
|
inode_sub_bytes(inode, old_num -
|
|
|
|
new_num);
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
2007-10-15 16:15:53 -04:00
|
|
|
btrfs_set_file_extent_num_bytes(leaf, extent,
|
|
|
|
new_num);
|
2007-10-15 16:14:19 -04:00
|
|
|
btrfs_mark_buffer_dirty(leaf);
|
2007-11-30 10:09:33 -05:00
|
|
|
} else if (key.offset < inline_limit &&
|
|
|
|
(end > extent_end) &&
|
|
|
|
(inline_limit < extent_end)) {
|
2007-10-15 16:18:25 -04:00
|
|
|
u32 new_size;
|
|
|
|
new_size = btrfs_file_extent_calc_inline_size(
|
2007-11-30 10:09:33 -05:00
|
|
|
inline_limit - key.offset);
|
2008-10-09 11:46:29 -04:00
|
|
|
inode_sub_bytes(inode, extent_end -
|
|
|
|
inline_limit);
|
2007-10-15 16:18:25 -04:00
|
|
|
btrfs_truncate_item(trans, root, path,
|
2007-11-01 11:28:41 -04:00
|
|
|
new_size, 1);
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
/* delete the entire extent */
|
|
|
|
if (!keep) {
|
2008-10-09 11:46:29 -04:00
|
|
|
if (found_inline)
|
|
|
|
inode_sub_bytes(inode, extent_end -
|
|
|
|
key.offset);
|
2007-06-12 06:35:45 -04:00
|
|
|
ret = btrfs_del_item(trans, root, path);
|
2007-06-22 14:16:25 -04:00
|
|
|
/* TODO update progress marker and return */
|
2007-06-12 06:35:45 -04:00
|
|
|
BUG_ON(ret);
|
|
|
|
extent = NULL;
|
2008-09-23 13:14:14 -04:00
|
|
|
btrfs_release_path(root, path);
|
|
|
|
/* the extent will be freed later */
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
2008-01-30 14:39:54 -05:00
|
|
|
if (bookend && found_inline && start <= key.offset) {
|
2007-11-01 11:28:41 -04:00
|
|
|
u32 new_size;
|
|
|
|
new_size = btrfs_file_extent_calc_inline_size(
|
2008-01-30 14:39:54 -05:00
|
|
|
extent_end - end);
|
2008-10-09 11:46:29 -04:00
|
|
|
inode_sub_bytes(inode, end - key.offset);
|
2008-09-23 13:14:14 -04:00
|
|
|
ret = btrfs_truncate_item(trans, root, path,
|
|
|
|
new_size, 0);
|
|
|
|
BUG_ON(ret);
|
2007-11-01 11:28:41 -04:00
|
|
|
}
|
2007-06-12 06:35:45 -04:00
|
|
|
/* create bookend, splitting the extent in two */
|
|
|
|
if (bookend && found_extent) {
|
2008-09-23 13:14:14 -04:00
|
|
|
u64 disk_bytenr;
|
2007-06-12 06:35:45 -04:00
|
|
|
struct btrfs_key ins;
|
|
|
|
ins.objectid = inode->i_ino;
|
|
|
|
ins.offset = end;
|
|
|
|
btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY);
|
|
|
|
btrfs_release_path(root, path);
|
|
|
|
ret = btrfs_insert_empty_item(trans, root, path, &ins,
|
|
|
|
sizeof(*extent));
|
2008-09-23 13:14:14 -04:00
|
|
|
BUG_ON(ret);
|
2007-06-18 09:57:58 -04:00
|
|
|
|
2007-10-15 16:14:19 -04:00
|
|
|
leaf = path->nodes[0];
|
|
|
|
extent = btrfs_item_ptr(leaf, path->slots[0],
|
|
|
|
struct btrfs_file_extent_item);
|
|
|
|
write_extent_buffer(leaf, &old,
|
|
|
|
(unsigned long)extent, sizeof(old));
|
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
btrfs_set_file_extent_compression(leaf, extent,
|
|
|
|
compression);
|
|
|
|
btrfs_set_file_extent_encryption(leaf, extent,
|
|
|
|
encryption);
|
|
|
|
btrfs_set_file_extent_other_encoding(leaf, extent,
|
|
|
|
other_encoding);
|
2007-10-15 16:14:19 -04:00
|
|
|
btrfs_set_file_extent_offset(leaf, extent,
|
2007-10-15 16:15:53 -04:00
|
|
|
le64_to_cpu(old.offset) + end - key.offset);
|
|
|
|
WARN_ON(le64_to_cpu(old.num_bytes) <
|
|
|
|
(extent_end - end));
|
|
|
|
btrfs_set_file_extent_num_bytes(leaf, extent,
|
|
|
|
extent_end - end);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 14:49:59 -04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* set the ram bytes to the size of the full extent
|
|
|
|
* before splitting. This is a worst case flag,
|
|
|
|
* but its the best we can do because we don't know
|
|
|
|
* how splitting affects compression
|
|
|
|
*/
|
|
|
|
btrfs_set_file_extent_ram_bytes(leaf, extent,
|
|
|
|
ram_bytes);
|
2007-10-15 16:14:19 -04:00
|
|
|
btrfs_set_file_extent_type(leaf, extent,
|
2007-06-12 06:35:45 -04:00
|
|
|
BTRFS_FILE_EXTENT_REG);
|
2007-10-15 16:15:53 -04:00
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
btrfs_mark_buffer_dirty(path->nodes[0]);
|
2008-09-23 13:14:14 -04:00
|
|
|
|
|
|
|
disk_bytenr = le64_to_cpu(old.disk_bytenr);
|
|
|
|
if (disk_bytenr != 0) {
|
|
|
|
ret = btrfs_inc_extent_ref(trans, root,
|
|
|
|
disk_bytenr,
|
|
|
|
le64_to_cpu(old.disk_num_bytes),
|
|
|
|
leaf->start,
|
|
|
|
root->root_key.objectid,
|
2008-10-09 11:46:24 -04:00
|
|
|
trans->transid, ins.objectid);
|
2008-09-23 13:14:14 -04:00
|
|
|
BUG_ON(ret);
|
|
|
|
}
|
|
|
|
btrfs_release_path(root, path);
|
|
|
|
if (disk_bytenr != 0) {
|
2008-10-09 11:46:29 -04:00
|
|
|
inode_add_bytes(inode, extent_end - end);
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
2008-09-23 13:14:14 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
if (found_extent && !keep) {
|
|
|
|
u64 disk_bytenr = le64_to_cpu(old.disk_bytenr);
|
|
|
|
|
|
|
|
if (disk_bytenr != 0) {
|
2008-10-09 11:46:29 -04:00
|
|
|
inode_sub_bytes(inode,
|
|
|
|
le64_to_cpu(old.num_bytes));
|
2008-09-23 13:14:14 -04:00
|
|
|
ret = btrfs_free_extent(trans, root,
|
|
|
|
disk_bytenr,
|
|
|
|
le64_to_cpu(old.disk_num_bytes),
|
|
|
|
leaf_start, root_owner,
|
2008-10-09 11:46:24 -04:00
|
|
|
root_gen, key.objectid, 0);
|
2008-09-23 13:14:14 -04:00
|
|
|
BUG_ON(ret);
|
|
|
|
*hint_byte = disk_bytenr;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (search_start >= end) {
|
2007-06-12 06:35:45 -04:00
|
|
|
ret = 0;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
btrfs_free_path(path);
|
2008-10-30 14:19:50 -04:00
|
|
|
if (locked_end > end) {
|
|
|
|
unlock_extent(&BTRFS_I(inode)->io_tree, end, locked_end - 1,
|
|
|
|
GFP_NOFS);
|
|
|
|
}
|
2008-02-08 13:49:28 -05:00
|
|
|
btrfs_check_file(root, inode);
|
2007-06-12 06:35:45 -04:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2008-09-29 15:18:18 -04:00
|
|
|
* this gets pages into the page cache and locks them down, it also properly
|
|
|
|
* waits for data=ordered extents to finish before allowing the pages to be
|
|
|
|
* modified.
|
2007-06-12 06:35:45 -04:00
|
|
|
*/
|
2008-09-05 16:09:51 -04:00
|
|
|
static int noinline prepare_pages(struct btrfs_root *root, struct file *file,
|
2008-01-03 10:01:48 -05:00
|
|
|
struct page **pages, size_t num_pages,
|
|
|
|
loff_t pos, unsigned long first_index,
|
|
|
|
unsigned long last_index, size_t write_bytes)
|
2007-06-12 06:35:45 -04:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
unsigned long index = pos >> PAGE_CACHE_SHIFT;
|
2007-12-18 16:15:09 -05:00
|
|
|
struct inode *inode = fdentry(file)->d_inode;
|
2007-06-12 06:35:45 -04:00
|
|
|
int err = 0;
|
2007-06-18 09:57:58 -04:00
|
|
|
u64 start_pos;
|
2008-07-17 12:53:50 -04:00
|
|
|
u64 last_pos;
|
2007-06-18 09:57:58 -04:00
|
|
|
|
2007-10-15 16:14:19 -04:00
|
|
|
start_pos = pos & ~((u64)root->sectorsize - 1);
|
2008-07-17 12:53:50 -04:00
|
|
|
last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT;
|
2007-06-12 06:35:45 -04:00
|
|
|
|
2008-10-30 14:19:41 -04:00
|
|
|
if (start_pos > inode->i_size) {
|
|
|
|
err = btrfs_cont_expand(inode, start_pos);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
memset(pages, 0, num_pages * sizeof(struct page *));
|
2008-07-17 12:53:50 -04:00
|
|
|
again:
|
2007-06-12 06:35:45 -04:00
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
pages[i] = grab_cache_page(inode->i_mapping, index + i);
|
|
|
|
if (!pages[i]) {
|
|
|
|
err = -ENOMEM;
|
2007-08-27 16:49:44 -04:00
|
|
|
BUG_ON(1);
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
2007-06-28 15:57:36 -04:00
|
|
|
wait_on_page_writeback(pages[i]);
|
2007-06-12 06:35:45 -04:00
|
|
|
}
|
2008-02-19 11:29:24 -05:00
|
|
|
if (start_pos < inode->i_size) {
|
2008-07-17 12:53:50 -04:00
|
|
|
struct btrfs_ordered_extent *ordered;
|
2008-02-19 12:55:05 -05:00
|
|
|
lock_extent(&BTRFS_I(inode)->io_tree,
|
|
|
|
start_pos, last_pos - 1, GFP_NOFS);
|
2008-07-17 12:53:50 -04:00
|
|
|
ordered = btrfs_lookup_first_ordered_extent(inode, last_pos -1);
|
|
|
|
if (ordered &&
|
|
|
|
ordered->file_offset + ordered->len > start_pos &&
|
|
|
|
ordered->file_offset < last_pos) {
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
unlock_extent(&BTRFS_I(inode)->io_tree,
|
|
|
|
start_pos, last_pos - 1, GFP_NOFS);
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
unlock_page(pages[i]);
|
|
|
|
page_cache_release(pages[i]);
|
|
|
|
}
|
|
|
|
btrfs_wait_ordered_range(inode, start_pos,
|
|
|
|
last_pos - start_pos);
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
if (ordered)
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
|
2008-02-19 11:29:24 -05:00
|
|
|
clear_extent_bits(&BTRFS_I(inode)->io_tree, start_pos,
|
|
|
|
last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC,
|
|
|
|
GFP_NOFS);
|
2008-02-19 12:55:05 -05:00
|
|
|
unlock_extent(&BTRFS_I(inode)->io_tree,
|
|
|
|
start_pos, last_pos - 1, GFP_NOFS);
|
2008-02-19 11:29:24 -05:00
|
|
|
}
|
2008-07-17 12:53:50 -04:00
|
|
|
for (i = 0; i < num_pages; i++) {
|
2008-08-01 11:27:23 -04:00
|
|
|
clear_page_dirty_for_io(pages[i]);
|
2008-07-17 12:53:50 -04:00
|
|
|
set_page_extent_mapped(pages[i]);
|
|
|
|
WARN_ON(!PageLocked(pages[i]));
|
|
|
|
}
|
2007-06-12 06:35:45 -04:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
|
|
|
|
size_t count, loff_t *ppos)
|
|
|
|
{
|
|
|
|
loff_t pos;
|
2007-10-29 14:36:41 -04:00
|
|
|
loff_t start_pos;
|
|
|
|
ssize_t num_written = 0;
|
|
|
|
ssize_t err = 0;
|
2007-06-12 06:35:45 -04:00
|
|
|
int ret = 0;
|
2007-12-18 16:15:09 -05:00
|
|
|
struct inode *inode = fdentry(file)->d_inode;
|
2007-06-12 06:35:45 -04:00
|
|
|
struct btrfs_root *root = BTRFS_I(inode)->root;
|
2007-06-18 09:57:58 -04:00
|
|
|
struct page **pages = NULL;
|
|
|
|
int nrptrs;
|
2007-06-12 06:35:45 -04:00
|
|
|
struct page *pinned[2];
|
|
|
|
unsigned long first_index;
|
|
|
|
unsigned long last_index;
|
2008-10-03 12:30:02 -04:00
|
|
|
int will_write;
|
|
|
|
|
|
|
|
will_write = ((file->f_flags & O_SYNC) || IS_SYNC(inode) ||
|
|
|
|
(file->f_flags & O_DIRECT));
|
2007-06-18 09:57:58 -04:00
|
|
|
|
|
|
|
nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE,
|
|
|
|
PAGE_CACHE_SIZE / (sizeof(struct page *)));
|
2007-06-12 06:35:45 -04:00
|
|
|
pinned[0] = NULL;
|
|
|
|
pinned[1] = NULL;
|
2007-10-29 14:36:41 -04:00
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
pos = *ppos;
|
2007-10-29 14:36:41 -04:00
|
|
|
start_pos = pos;
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
|
|
|
|
current->backing_dev_info = inode->i_mapping->backing_dev_info;
|
|
|
|
err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
|
|
|
|
if (err)
|
2007-12-21 16:27:21 -05:00
|
|
|
goto out_nolock;
|
2007-06-12 06:35:45 -04:00
|
|
|
if (count == 0)
|
2007-12-21 16:27:21 -05:00
|
|
|
goto out_nolock;
|
2008-09-24 11:48:04 -04:00
|
|
|
|
2008-07-30 16:54:26 -04:00
|
|
|
err = file_remove_suid(file);
|
2007-06-12 06:35:45 -04:00
|
|
|
if (err)
|
2007-12-21 16:27:21 -05:00
|
|
|
goto out_nolock;
|
2007-06-12 06:35:45 -04:00
|
|
|
file_update_time(file);
|
|
|
|
|
2007-06-18 09:57:58 -04:00
|
|
|
pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
|
2007-06-12 06:35:45 -04:00
|
|
|
|
|
|
|
mutex_lock(&inode->i_mutex);
|
|
|
|
first_index = pos >> PAGE_CACHE_SHIFT;
|
|
|
|
last_index = (pos + count) >> PAGE_CACHE_SHIFT;
|
|
|
|
|
2008-04-22 09:24:20 -04:00
|
|
|
/*
|
|
|
|
* if this is a nodatasum mount, force summing off for the inode
|
|
|
|
* all the time. That way a later mount with summing on won't
|
|
|
|
* get confused
|
|
|
|
*/
|
|
|
|
if (btrfs_test_opt(root, NODATASUM))
|
|
|
|
btrfs_set_flag(inode, NODATASUM);
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
/*
|
|
|
|
* there are lots of better ways to do this, but this code
|
|
|
|
* makes sure the first and last page in the file range are
|
|
|
|
* up to date and ready for cow
|
|
|
|
*/
|
|
|
|
if ((pos & (PAGE_CACHE_SIZE - 1))) {
|
|
|
|
pinned[0] = grab_cache_page(inode->i_mapping, first_index);
|
|
|
|
if (!PageUptodate(pinned[0])) {
|
2007-06-15 13:50:00 -04:00
|
|
|
ret = btrfs_readpage(NULL, pinned[0]);
|
2007-06-12 06:35:45 -04:00
|
|
|
BUG_ON(ret);
|
|
|
|
wait_on_page_locked(pinned[0]);
|
|
|
|
} else {
|
|
|
|
unlock_page(pinned[0]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if ((pos + count) & (PAGE_CACHE_SIZE - 1)) {
|
|
|
|
pinned[1] = grab_cache_page(inode->i_mapping, last_index);
|
|
|
|
if (!PageUptodate(pinned[1])) {
|
2007-06-15 13:50:00 -04:00
|
|
|
ret = btrfs_readpage(NULL, pinned[1]);
|
2007-06-12 06:35:45 -04:00
|
|
|
BUG_ON(ret);
|
|
|
|
wait_on_page_locked(pinned[1]);
|
|
|
|
} else {
|
|
|
|
unlock_page(pinned[1]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
while(count > 0) {
|
|
|
|
size_t offset = pos & (PAGE_CACHE_SIZE - 1);
|
2007-06-22 14:16:24 -04:00
|
|
|
size_t write_bytes = min(count, nrptrs *
|
|
|
|
(size_t)PAGE_CACHE_SIZE -
|
2007-06-18 09:57:58 -04:00
|
|
|
offset);
|
2007-06-12 06:35:45 -04:00
|
|
|
size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
|
|
|
|
PAGE_CACHE_SHIFT;
|
|
|
|
|
2007-06-18 09:57:58 -04:00
|
|
|
WARN_ON(num_pages > nrptrs);
|
2007-06-12 06:35:45 -04:00
|
|
|
memset(pages, 0, sizeof(pages));
|
2007-12-21 16:27:21 -05:00
|
|
|
|
|
|
|
ret = btrfs_check_free_space(root, write_bytes, 0);
|
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
ret = prepare_pages(root, file, pages, num_pages,
|
|
|
|
pos, first_index, last_index,
|
2007-06-18 09:57:58 -04:00
|
|
|
write_bytes);
|
2007-06-22 14:16:25 -04:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
2007-06-12 06:35:45 -04:00
|
|
|
|
|
|
|
ret = btrfs_copy_from_user(pos, num_pages,
|
|
|
|
write_bytes, pages, buf);
|
2007-06-22 14:16:25 -04:00
|
|
|
if (ret) {
|
|
|
|
btrfs_drop_pages(pages, num_pages);
|
|
|
|
goto out;
|
|
|
|
}
|
2007-06-12 06:35:45 -04:00
|
|
|
|
|
|
|
ret = dirty_and_release_pages(NULL, root, file, pages,
|
|
|
|
num_pages, pos, write_bytes);
|
|
|
|
btrfs_drop_pages(pages, num_pages);
|
2007-06-22 14:16:25 -04:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
2007-06-12 06:35:45 -04:00
|
|
|
|
2008-10-03 12:30:02 -04:00
|
|
|
if (will_write) {
|
|
|
|
btrfs_fdatawrite_range(inode->i_mapping, pos,
|
|
|
|
pos + write_bytes - 1,
|
|
|
|
WB_SYNC_NONE);
|
|
|
|
} else {
|
|
|
|
balance_dirty_pages_ratelimited_nr(inode->i_mapping,
|
|
|
|
num_pages);
|
|
|
|
if (num_pages <
|
|
|
|
(root->leafsize >> PAGE_CACHE_SHIFT) + 1)
|
|
|
|
btrfs_btree_balance_dirty(root, 1);
|
|
|
|
btrfs_throttle(root);
|
|
|
|
}
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
buf += write_bytes;
|
|
|
|
count -= write_bytes;
|
|
|
|
pos += write_bytes;
|
|
|
|
num_written += write_bytes;
|
|
|
|
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
out:
|
2007-12-21 16:27:21 -05:00
|
|
|
mutex_unlock(&inode->i_mutex);
|
2008-01-03 13:46:11 -05:00
|
|
|
|
2007-12-21 16:27:21 -05:00
|
|
|
out_nolock:
|
2007-06-18 09:57:58 -04:00
|
|
|
kfree(pages);
|
2007-06-12 06:35:45 -04:00
|
|
|
if (pinned[0])
|
|
|
|
page_cache_release(pinned[0]);
|
|
|
|
if (pinned[1])
|
|
|
|
page_cache_release(pinned[1]);
|
|
|
|
*ppos = pos;
|
2007-10-29 14:36:41 -04:00
|
|
|
|
2008-10-03 12:30:02 -04:00
|
|
|
if (num_written > 0 && will_write) {
|
2008-09-05 16:13:11 -04:00
|
|
|
struct btrfs_trans_handle *trans;
|
|
|
|
|
2008-10-03 12:30:02 -04:00
|
|
|
err = btrfs_wait_ordered_range(inode, start_pos, num_written);
|
|
|
|
if (err)
|
2007-10-29 14:36:41 -04:00
|
|
|
num_written = err;
|
2008-09-05 16:13:11 -04:00
|
|
|
|
2008-10-03 12:30:02 -04:00
|
|
|
if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
|
|
|
|
trans = btrfs_start_transaction(root, 1);
|
|
|
|
ret = btrfs_log_dentry_safe(trans, root,
|
|
|
|
file->f_dentry);
|
|
|
|
if (ret == 0) {
|
|
|
|
btrfs_sync_log(trans, root);
|
|
|
|
btrfs_end_transaction(trans, root);
|
|
|
|
} else {
|
|
|
|
btrfs_commit_transaction(trans, root);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (file->f_flags & O_DIRECT) {
|
|
|
|
invalidate_mapping_pages(inode->i_mapping,
|
|
|
|
start_pos >> PAGE_CACHE_SHIFT,
|
|
|
|
(start_pos + num_written - 1) >> PAGE_CACHE_SHIFT);
|
2008-09-05 16:13:11 -04:00
|
|
|
}
|
2007-10-29 14:36:41 -04:00
|
|
|
}
|
2007-06-12 06:35:45 -04:00
|
|
|
current->backing_dev_info = NULL;
|
|
|
|
return num_written ? num_written : err;
|
|
|
|
}
|
|
|
|
|
2008-06-10 10:07:39 -04:00
|
|
|
int btrfs_release_file(struct inode * inode, struct file * filp)
|
2008-05-27 10:55:43 -04:00
|
|
|
{
|
2008-06-10 10:07:39 -04:00
|
|
|
if (filp->private_data)
|
|
|
|
btrfs_ioctl_trans_end(filp);
|
2008-05-27 10:55:43 -04:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-09-29 15:18:18 -04:00
|
|
|
/*
|
|
|
|
* fsync call for both files and directories. This logs the inode into
|
|
|
|
* the tree log instead of forcing full commits whenever possible.
|
|
|
|
*
|
|
|
|
* It needs to call filemap_fdatawait so that all ordered extent updates are
|
|
|
|
* in the metadata btree are up to date for copying to the log.
|
|
|
|
*
|
|
|
|
* It drops the inode mutex before doing the tree log commit. This is an
|
|
|
|
* important optimization for directories because holding the mutex prevents
|
|
|
|
* new operations on the dir while we write to disk.
|
|
|
|
*/
|
2008-09-05 16:13:11 -04:00
|
|
|
int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
|
2007-06-12 06:35:45 -04:00
|
|
|
{
|
|
|
|
struct inode *inode = dentry->d_inode;
|
|
|
|
struct btrfs_root *root = BTRFS_I(inode)->root;
|
2007-08-10 16:22:09 -04:00
|
|
|
int ret = 0;
|
2007-06-12 06:35:45 -04:00
|
|
|
struct btrfs_trans_handle *trans;
|
|
|
|
|
|
|
|
/*
|
2007-08-10 16:22:09 -04:00
|
|
|
* check the transaction that last modified this inode
|
|
|
|
* and see if its already been committed
|
2007-06-12 06:35:45 -04:00
|
|
|
*/
|
2007-08-10 16:22:09 -04:00
|
|
|
if (!BTRFS_I(inode)->last_trans)
|
|
|
|
goto out;
|
2008-06-25 16:01:30 -04:00
|
|
|
|
2007-08-10 16:22:09 -04:00
|
|
|
mutex_lock(&root->fs_info->trans_mutex);
|
|
|
|
if (BTRFS_I(inode)->last_trans <=
|
|
|
|
root->fs_info->last_trans_committed) {
|
|
|
|
BTRFS_I(inode)->last_trans = 0;
|
|
|
|
mutex_unlock(&root->fs_info->trans_mutex);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
mutex_unlock(&root->fs_info->trans_mutex);
|
|
|
|
|
2008-09-11 15:53:12 -04:00
|
|
|
root->fs_info->tree_log_batch++;
|
2008-09-05 16:13:11 -04:00
|
|
|
filemap_fdatawait(inode->i_mapping);
|
2008-09-11 15:53:12 -04:00
|
|
|
root->fs_info->tree_log_batch++;
|
2008-09-05 16:13:11 -04:00
|
|
|
|
2007-08-10 16:22:09 -04:00
|
|
|
/*
|
2007-08-27 16:49:44 -04:00
|
|
|
* ok we haven't committed the transaction yet, lets do a commit
|
|
|
|
*/
|
2008-06-10 10:07:39 -04:00
|
|
|
if (file->private_data)
|
|
|
|
btrfs_ioctl_trans_end(file);
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
trans = btrfs_start_transaction(root, 1);
|
|
|
|
if (!trans) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2008-09-05 16:13:11 -04:00
|
|
|
|
|
|
|
ret = btrfs_log_dentry_safe(trans, root, file->f_dentry);
|
2008-09-11 15:53:12 -04:00
|
|
|
if (ret < 0) {
|
2008-09-05 16:13:11 -04:00
|
|
|
goto out;
|
2008-09-11 15:53:12 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* we've logged all the items and now have a consistent
|
|
|
|
* version of the file in the log. It is possible that
|
|
|
|
* someone will come in and modify the file, but that's
|
|
|
|
* fine because the log is consistent on disk, and we
|
|
|
|
* have references to all of the file's extents
|
|
|
|
*
|
|
|
|
* It is possible that someone will come in and log the
|
|
|
|
* file again, but that will end up using the synchronization
|
|
|
|
* inside btrfs_sync_log to keep things safe.
|
|
|
|
*/
|
|
|
|
mutex_unlock(&file->f_dentry->d_inode->i_mutex);
|
|
|
|
|
2008-09-05 16:13:11 -04:00
|
|
|
if (ret > 0) {
|
|
|
|
ret = btrfs_commit_transaction(trans, root);
|
|
|
|
} else {
|
|
|
|
btrfs_sync_log(trans, root);
|
|
|
|
ret = btrfs_end_transaction(trans, root);
|
|
|
|
}
|
2008-09-11 15:53:12 -04:00
|
|
|
mutex_lock(&file->f_dentry->d_inode->i_mutex);
|
2007-06-12 06:35:45 -04:00
|
|
|
out:
|
|
|
|
return ret > 0 ? EIO : ret;
|
|
|
|
}
|
|
|
|
|
2007-06-15 13:50:00 -04:00
|
|
|
static struct vm_operations_struct btrfs_file_vm_ops = {
|
2007-07-25 12:31:35 -04:00
|
|
|
.fault = filemap_fault,
|
2007-06-15 13:50:00 -04:00
|
|
|
.page_mkwrite = btrfs_page_mkwrite,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma)
|
|
|
|
{
|
|
|
|
vma->vm_ops = &btrfs_file_vm_ops;
|
|
|
|
file_accessed(filp);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2007-06-12 06:35:45 -04:00
|
|
|
struct file_operations btrfs_file_operations = {
|
|
|
|
.llseek = generic_file_llseek,
|
|
|
|
.read = do_sync_read,
|
2007-06-15 13:50:00 -04:00
|
|
|
.aio_read = generic_file_aio_read,
|
2007-12-14 12:56:58 -05:00
|
|
|
.splice_read = generic_file_splice_read,
|
2007-06-12 06:35:45 -04:00
|
|
|
.write = btrfs_file_write,
|
2007-06-15 13:50:00 -04:00
|
|
|
.mmap = btrfs_file_mmap,
|
2007-06-12 06:35:45 -04:00
|
|
|
.open = generic_file_open,
|
2008-05-27 10:55:43 -04:00
|
|
|
.release = btrfs_release_file,
|
2007-06-12 06:35:45 -04:00
|
|
|
.fsync = btrfs_sync_file,
|
2007-09-14 10:22:47 -04:00
|
|
|
.unlocked_ioctl = btrfs_ioctl,
|
2007-06-12 06:35:45 -04:00
|
|
|
#ifdef CONFIG_COMPAT
|
2007-09-14 10:22:47 -04:00
|
|
|
.compat_ioctl = btrfs_ioctl,
|
2007-06-12 06:35:45 -04:00
|
|
|
#endif
|
|
|
|
};
|