The patch titled
fs: restore nobh
has been added to the -mm tree. Its filename is
fs-restore-nobh.patch
*** Remember to use Documentation/SubmitChecklist when testing your code ***
See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find
out what to do about this
------------------------------------------------------
Subject: fs: restore nobh
From: Nick Piggin <npiggin@xxxxxxx>
Implement nobh in new aops. This is a bit tricky. FWIW, nobh_truncate is
now implemented in a way that does not create blocks in sparse regions,
which is a silly thing for it to have been doing (isn't it?)
ext2 survives fsx and fsstress. jfs is converted as well... ext3
should be easy to do (but not done yet).
Cc: Badari Pulavarty <pbadari@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
fs/buffer.c | 227 ++++++++++++++++++++++------------
fs/ext2/inode.c | 20 ++
fs/jfs/inode.c | 7 -
include/linux/buffer_head.h | 10 +
4 files changed, 176 insertions(+), 88 deletions(-)
diff -puN fs/buffer.c~fs-restore-nobh fs/buffer.c
--- a/fs/buffer.c~fs-restore-nobh
+++ a/fs/buffer.c
@@ -2367,7 +2367,7 @@ out_unlock:
}
/*
- * nobh_prepare_write()'s prereads are special: the buffer_heads are freed
+ * nobh_write_begin()'s prereads are special: the buffer_heads are freed
* immediately, while under the page lock. So it needs a special end_io
* handler which does not touch the bh after unlocking it.
*/
@@ -2377,16 +2377,45 @@ static void end_buffer_read_nobh(struct
}
/*
+ * Attach the singly-linked list of buffers created by nobh_write_begin, to
+ * the page (converting it to circular linked list and taking care of page
+ * dirty races).
+ */
+static void attach_nobh_buffers(struct page *page, struct buffer_head *head)
+{
+ struct buffer_head *bh;
+
+ BUG_ON(!PageLocked(page));
+
+ spin_lock(&page->mapping->private_lock);
+ bh = head;
+ do {
+ if (PageDirty(page))
+ set_buffer_dirty(bh);
+ if (!bh->b_this_page)
+ bh->b_this_page = head;
+ bh = bh->b_this_page;
+ } while (bh != head);
+ attach_page_buffers(page, head);
+ spin_unlock(&page->mapping->private_lock);
+}
+
+/*
* On entry, the page is fully not uptodate.
* On exit the page is fully uptodate in the areas outside (from,to)
*/
-int nobh_prepare_write(struct page *page, unsigned from, unsigned to,
+int nobh_write_begin(struct file *file, struct address_space *mapping,
+ loff_t pos, unsigned len, unsigned flags,
+ struct page **pagep, void **fsdata,
get_block_t *get_block)
{
- struct inode *inode = page->mapping->host;
+ struct inode *inode = mapping->host;
const unsigned blkbits = inode->i_blkbits;
const unsigned blocksize = 1 << blkbits;
struct buffer_head *head, *bh;
+ struct page *page;
+ pgoff_t index;
+ unsigned from, to;
unsigned block_in_page;
unsigned block_start, block_end;
sector_t block_in_file;
@@ -2395,8 +2424,22 @@ int nobh_prepare_write(struct page *page
int ret = 0;
int is_mapped_to_disk = 1;
- if (page_has_buffers(page))
- return block_prepare_write(page, from, to, get_block);
+ index = pos >> PAGE_CACHE_SHIFT;
+ from = pos & (PAGE_CACHE_SIZE - 1);
+ to = from + len;
+
+ page = __grab_cache_page(mapping, index);
+ if (!page)
+ return -ENOMEM;
+ *pagep = page;
+ *fsdata = NULL;
+
+ if (page_has_buffers(page)) {
+ unlock_page(page);
+ page_cache_release(page);
+ *pagep = NULL;
+ return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata, get_block);
+ }
if (PageMappedToDisk(page))
return 0;
@@ -2411,8 +2454,10 @@ int nobh_prepare_write(struct page *page
* than the circular one we're used to.
*/
head = alloc_page_buffers(page, blocksize, 0);
- if (!head)
- return -ENOMEM;
+ if (!head) {
+ ret = -ENOMEM;
+ goto out_release;
+ }
block_in_file = (sector_t)page->index << (PAGE_CACHE_SHIFT - blkbits);
@@ -2481,15 +2526,12 @@ int nobh_prepare_write(struct page *page
if (is_mapped_to_disk)
SetPageMappedToDisk(page);
- do {
- bh = head;
- head = head->b_this_page;
- free_buffer_head(bh);
- } while (head);
+ *fsdata = head; /* to be released by nobh_write_end */
return 0;
failed:
+ BUG_ON(!ret);
/*
* Error recovery is a bit difficult. We need to zero out blocks that
* were newly allocated, and dirty them to ensure they get written out.
@@ -2497,64 +2539,56 @@ failed:
* the handling of potential IO errors during writeout would be hard
* (could try doing synchronous writeout, but what if that fails too?)
*/
- spin_lock(&page->mapping->private_lock);
- bh = head;
- block_start = 0;
- do {
- if (PageUptodate(page))
- set_buffer_uptodate(bh);
- if (PageDirty(page))
- set_buffer_dirty(bh);
+ attach_nobh_buffers(page, head);
+ page_zero_new_buffers(page, from, to);
- block_end = block_start+blocksize;
- if (block_end <= from)
- goto next;
- if (block_start >= to)
- goto next;
+out_release:
+ unlock_page(page);
+ page_cache_release(page);
+ *pagep = NULL;
- if (buffer_new(bh)) {
- clear_buffer_new(bh);
- if (!buffer_uptodate(bh)) {
- zero_user_page(page, block_start, bh->b_size, KM_USER0);
- set_buffer_uptodate(bh);
- }
- mark_buffer_dirty(bh);
- }
-next:
- block_start = block_end;
- if (!bh->b_this_page)
- bh->b_this_page = head;
- bh = bh->b_this_page;
- } while (bh != head);
- attach_page_buffers(page, head);
- spin_unlock(&page->mapping->private_lock);
+ if (pos + len > inode->i_size)
+ vmtruncate(inode, inode->i_size);
return ret;
}
-EXPORT_SYMBOL(nobh_prepare_write);
+EXPORT_SYMBOL(nobh_write_begin);
-/*
- * Make sure any changes to nobh_commit_write() are reflected in
- * nobh_truncate_page(), since it doesn't call commit_write().
- */
-int nobh_commit_write(struct file *file, struct page *page,
- unsigned from, unsigned to)
+int nobh_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;
- loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
+ struct buffer_head *head = NULL;
+ struct buffer_head *bh;
- if (page_has_buffers(page))
- return generic_commit_write(file, page, from, to);
+ if (!PageMappedToDisk(page)) {
+ if (unlikely(copied < len) && !page_has_buffers(page))
+ attach_nobh_buffers(page, head);
+ if (page_has_buffers(page))
+ return generic_write_end(file, mapping, pos, len, copied, page, fsdata);
+ }
SetPageUptodate(page);
set_page_dirty(page);
- if (pos > inode->i_size) {
- i_size_write(inode, pos);
+ if (pos+copied > inode->i_size) {
+ i_size_write(inode, pos+copied);
mark_inode_dirty(inode);
}
- return 0;
+
+ unlock_page(page);
+ page_cache_release(page);
+
+ head = fsdata;
+ while (head) {
+ bh = head;
+ head = head->b_this_page;
+ free_buffer_head(bh);
+ }
+
+ return copied;
}
-EXPORT_SYMBOL(nobh_commit_write);
+EXPORT_SYMBOL(nobh_write_end);
/*
* nobh_writepage() - based on block_full_write_page() except
@@ -2607,44 +2641,79 @@ out:
}
EXPORT_SYMBOL(nobh_writepage);
-/*
- * This function assumes that ->prepare_write() uses nobh_prepare_write().
- */
-int nobh_truncate_page(struct address_space *mapping, loff_t from)
+int nobh_truncate_page(struct address_space *mapping,
+ loff_t from, get_block_t *get_block)
{
- struct inode *inode = mapping->host;
- unsigned blocksize = 1 << inode->i_blkbits;
pgoff_t index = from >> PAGE_CACHE_SHIFT;
unsigned offset = from & (PAGE_CACHE_SIZE-1);
- unsigned to;
+ unsigned blocksize;
+ sector_t iblock;
+ unsigned length, pos;
+ struct inode *inode = mapping->host;
struct page *page;
- const struct address_space_operations *a_ops = mapping->a_ops;
- int ret = 0;
+ struct buffer_head map_bh;
+ int err;
- if ((offset & (blocksize - 1)) == 0)
- goto out;
+ blocksize = 1 << inode->i_blkbits;
+ length = offset & (blocksize - 1);
+
+ /* Block boundary? Nothing to do */
+ if (!length)
+ return 0;
+
+ length = blocksize - length;
+ iblock = (sector_t)index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
- ret = -ENOMEM;
page = grab_cache_page(mapping, index);
+ err = -ENOMEM;
if (!page)
goto out;
- to = (offset + blocksize) & ~(blocksize - 1);
- ret = a_ops->prepare_write(NULL, page, offset, to);
- if (ret == 0) {
- zero_user_page(page, offset, PAGE_CACHE_SIZE - offset,
- KM_USER0);
- /*
- * It would be more correct to call aops->commit_write()
- * here, but this is more efficient.
- */
- SetPageUptodate(page);
- set_page_dirty(page);
+ if (page_has_buffers(page)) {
+has_buffers:
+ unlock_page(page);
+ page_cache_release(page);
+ return block_truncate_page(mapping, from, get_block);
+ }
+
+ /* Find the buffer that contains "offset" */
+ pos = blocksize;
+ while (offset >= pos) {
+ iblock++;
+ pos += blocksize;
}
+
+ err = get_block(inode, iblock, &map_bh, 0);
+ if (err)
+ goto unlock;
+ /* unmapped? It's a hole - nothing to do */
+ if (!buffer_mapped(&map_bh))
+ goto unlock;
+
+ /* Ok, it's mapped. Make sure it's up-to-date */
+ if (!PageUptodate(page)) {
+ err = mapping->a_ops->readpage(NULL, page);
+ if (err) {
+ page_cache_release(page);
+ goto out;
+ }
+ lock_page(page);
+ if (!PageUptodate(page)) {
+ err = -EIO;
+ goto unlock;
+ }
+ if (page_has_buffers(page))
+ goto has_buffers;
+ }
+ zero_user_page(page, offset, length, KM_USER0);
+ set_page_dirty(page);
+ err = 0;
+
+unlock:
unlock_page(page);
page_cache_release(page);
out:
- return ret;
+ return err;
}
EXPORT_SYMBOL(nobh_truncate_page);
diff -puN fs/ext2/inode.c~fs-restore-nobh fs/ext2/inode.c
--- a/fs/ext2/inode.c~fs-restore-nobh
+++ a/fs/ext2/inode.c
@@ -659,6 +659,20 @@ ext2_write_begin(struct file *file, stru
return __ext2_write_begin(file, mapping, pos, len, flags, pagep,fsdata);
}
+static int
+ext2_nobh_write_begin(struct file *file, struct address_space *mapping,
+ loff_t pos, unsigned len, unsigned flags,
+ struct page **pagep, void **fsdata)
+{
+ /*
+ * Dir-in-pagecache still uses ext2_write_begin. Would have to rework
+ * directory handling code to pass around offsets rather than struct
+ * pages in order to make this work easily.
+ */
+ return nobh_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
+ ext2_get_block);
+}
+
static int ext2_nobh_writepage(struct page *page,
struct writeback_control *wbc)
{
@@ -710,7 +724,8 @@ const struct address_space_operations ex
.readpages = ext2_readpages,
.writepage = ext2_nobh_writepage,
.sync_page = block_sync_page,
- /* XXX: todo */
+ .write_begin = ext2_nobh_write_begin,
+ .write_end = nobh_write_end,
.bmap = ext2_bmap,
.direct_IO = ext2_direct_IO,
.writepages = ext2_writepages,
@@ -927,7 +942,8 @@ void ext2_truncate (struct inode * inode
if (mapping_is_xip(inode->i_mapping))
xip_truncate_page(inode->i_mapping, inode->i_size);
else if (test_opt(inode->i_sb, NOBH))
- nobh_truncate_page(inode->i_mapping, inode->i_size);
+ nobh_truncate_page(inode->i_mapping,
+ inode->i_size, ext2_get_block);
else
block_truncate_page(inode->i_mapping,
inode->i_size, ext2_get_block);
diff -puN fs/jfs/inode.c~fs-restore-nobh fs/jfs/inode.c
--- a/fs/jfs/inode.c~fs-restore-nobh
+++ a/fs/jfs/inode.c
@@ -279,8 +279,7 @@ static int jfs_write_begin(struct file *
loff_t pos, unsigned len, unsigned flags,
struct page **pagep, void **fsdata)
{
- *pagep = NULL;
- return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
+ return nobh_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
jfs_get_block);
}
@@ -306,7 +305,7 @@ const struct address_space_operations jf
.writepages = jfs_writepages,
.sync_page = block_sync_page,
.write_begin = jfs_write_begin,
- .write_end = generic_write_end,
+ .write_end = nobh_write_end,
.bmap = jfs_bmap,
.direct_IO = jfs_direct_IO,
};
@@ -359,7 +358,7 @@ void jfs_truncate(struct inode *ip)
{
jfs_info("jfs_truncate: size = 0x%lx", (ulong) ip->i_size);
- block_truncate_page(ip->i_mapping, ip->i_size, jfs_get_block);
+ nobh_truncate_page(ip->i_mapping, ip->i_size, jfs_get_block);
IWRITE_LOCK(ip, RDWRLOCK_NORMAL);
jfs_truncate_nolock(ip, ip->i_size);
diff -puN include/linux/buffer_head.h~fs-restore-nobh include/linux/buffer_head.h
--- a/include/linux/buffer_head.h~fs-restore-nobh
+++ a/include/linux/buffer_head.h
@@ -226,9 +226,13 @@ sector_t generic_block_bmap(struct addre
int generic_commit_write(struct file *, struct page *, unsigned, unsigned);
int block_truncate_page(struct address_space *, loff_t, get_block_t *);
int file_fsync(struct file *, struct dentry *, int);
-int nobh_prepare_write(struct page*, unsigned, unsigned, get_block_t*);
-int nobh_commit_write(struct file *, struct page *, unsigned, unsigned);
-int nobh_truncate_page(struct address_space *, loff_t);
+int nobh_write_begin(struct file *, struct address_space *,
+ loff_t, unsigned, unsigned,
+ struct page **, void **, get_block_t*);
+int nobh_write_end(struct file *, struct address_space *,
+ loff_t, unsigned, unsigned,
+ struct page *, void *);
+int nobh_truncate_page(struct address_space *, loff_t, get_block_t *);
int nobh_writepage(struct page *page, get_block_t *get_block,
struct writeback_control *wbc);
_
Patches currently in -mm which might be from npiggin@xxxxxxx are
origin.patch
remove-zero_page.patch
mm-use-lockless-radix-tree-probe.patch
mm-improve-find_lock_page.patch
mm-clarify-__add_to_swap_cache-locking.patch
radix-tree-use-indirect-bit.patch
calculation-of-pgoff-in-do_linear_fault-uses-mixed.patch
fs-fix-nobh-error-handling.patch
fix-the-max-path-calculation-in-radix-treec-update.patch
mm-revert-kernel_ds-buffered-write-optimisation.patch
revert-81b0c8713385ce1b1b9058e916edcf9561ad76d6.patch
revert-6527c2bdf1f833cc18e8f42bd97973d583e4aa83.patch
mm-clean-up-buffered-write-code.patch
mm-debug-write-deadlocks.patch
mm-trim-more-holes.patch
mm-buffered-write-cleanup.patch
mm-write-iovec-cleanup.patch
mm-fix-pagecache-write-deadlocks.patch
mm-buffered-write-iterator.patch
fs-fix-data-loss-on-error.patch
fs-introduce-write_begin-write_end-and-perform_write-aops.patch
introduce-write_begin-write_end-aops-important-fix.patch
introduce-write_begin-write_end-aops-fix2.patch
mm-restore-kernel_ds-optimisations.patch
implement-simple-fs-aops.patch
implement-simple-fs-aops-fix.patch
block_dev-convert-to-new-aops.patch
ext2-convert-to-new-aops.patch
ext2-convert-to-new-aops-fix.patch
ext2-convert-to-new-aops-fix2.patch
ext3-convert-to-new-aops.patch
ext3-convert-to-new-aops-fix.patch
ext3-convert-to-new-aops-fix-fix.patch
ext4-convert-to-new-aops.patch
ext4-convert-to-new-aops-fix.patch
ext4-convert-to-new-aops-fix-fix.patch
xfs-convert-to-new-aops.patch
gfs2-convert-to-new-aops-fix.patch
fs-new-cont-helpers.patch
fat-convert-to-new-aops.patch
hfs-convert-to-new-aops.patch
hfsplus-convert-to-new-aops.patch
hpfs-convert-to-new-aops.patch
bfs-convert-to-new-aops.patch
qnx4-convert-to-new-aops.patch
reiserfs-use-generic-write.patch
reiserfs-convert-to-new-aops.patch
reiserfs-convert-to-new-aops-fix.patch
reiserfs-convert-to-new-aops-fix2.patch
reiserfs-use-generic_cont_expand_simple.patch
with-reiserfs-no-longer-using-the-weird-generic_cont_expand-remove-it-completely.patch
nfs-convert-to-new-aops.patch
smb-convert-to-new-aops.patch
fuse-convert-to-new-aops.patch
hostfs-convert-to-new-aops.patch
hostfs-convert-to-new-aops-fix.patch
hostfs-convert-to-new-aops-fix-fix.patch
jffs2-convert-to-new-aops.patch
ufs-convert-to-new-aops.patch
ufs-convert-to-new-aops-fix.patch
ufs-convert-to-new-aops-fix2.patch
udf-convert-to-new-aops.patch
udf-convert-to-new-aops-fix.patch
sysv-convert-to-new-aops.patch
sysv-convert-to-new-aops-fix.patch
sysv-convert-to-new-aops-fix2.patch
minix-convert-to-new-aops.patch
minix-convert-to-new-aops-fix.patch
minix-convert-to-new-aops-fix2.patch
jfs-convert-to-new-aops.patch
fs-adfs-convert-to-new-aops.patch
fs-affs-convert-to-new-aops.patch
affs-convert-to-new-aops-fix.patch
affs-convert-to-new-aops-fix-fix.patch
ocfs2-convert-to-new-aops.patch
fs-restore-nobh.patch
fs-remove-some-aop_truncated_page.patch
make-swappiness-safer-to-use.patch
mm-document-tree_lock-zonelock-lockorder.patch
fs-reiserfs-cleanups.patch
atomic_opstxt-has-incorrect-misleading-and-insufficient-information.patch
fs-introduce-write_begin-write_end-and-perform_write-aops-revoke.patch
fs-introduce-write_begin-write_end-and-perform_write-aops-revoke-fix.patch
bitops-introduce-lock-ops.patch
alpha-fix-bitops.patch
alpha-lock-bitops.patch
alpha-lock-bitops-fix.patch
ia64-lock-bitops.patch
mips-fix-bitops.patch
mips-lock-bitops.patch
powerpc-lock-bitops.patch
powerpc-lock-bitops-fix.patch
bit_spin_lock-use-lock-bitops.patch
reiser4-fix-for-new-aops-patches.patch
-
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
