FYI, since iget() is going to be disappearing in 2.6.25, and akpm has
disabled ext4 in the -mm tree as a result, I'm folding the following
patch into the defrag-free-space-fragmentation.patch in the ext4 tree.
We also have an issue where the read-only bind patches from Dave Hansen
are making changes which conflict with the ext4 patch tree, which is
making akpm very grumpy. I'll try to take a look at this later in the
weekend...
- Ted
diff --git a/fs/ext4/defrag.c b/fs/ext4/defrag.c
index d9a14e4..6370fb8 100644
--- a/fs/ext4/defrag.c
+++ b/fs/ext4/defrag.c
@@ -287,9 +287,13 @@ static int ext4_ext_extents_info(struct ext4_extents_info *ext_info,
int entries = 0;
int err = 0;
- inode = iget(sb, ext_info->ino);
+ inode = iget_locked(sb, ext_info->ino);
if (!inode)
return -EACCES;
+ if (inode->i_state & I_NEW) {
+ sb->s_op->read_inode(inode);
+ unlock_new_inode(inode);
+ }
down_write(&EXT4_I(inode)->i_data_sem);
@@ -588,9 +592,13 @@ static int ext4_ext_defrag_victim(struct file *target_filp,
ext.len = 0;
/* Get the inode of the victim file */
- victim_inode = iget(sb, ex_info->ino);
+ victim_inode = iget_locked(sb, ex_info->ino);
if (!victim_inode)
return -EACCES;
+ if (victim_inode->i_state & I_NEW) {
+ sb->s_op->read_inode(victim_inode);
+ unlock_new_inode(victim_inode);
+ }
/* Setup file for the victim file */
victim_dent.d_inode = victim_inode;
On Sat, 2008-02-02 at 11:40 -0500, Theodore Ts'o wrote:
> FYI, since iget() is going to be disappearing in 2.6.25, and akpm has
> disabled ext4 in the -mm tree as a result, I'm folding the following
> patch into the defrag-free-space-fragmentation.patch in the ext4 tree.
>
>
I noticed that you have merged the changes in the patch queue...
diff --git a/fs/ext4/defrag.c b/fs/ext4/defrag.c
> index d9a14e4..6370fb8 100644
> --- a/fs/ext4/defrag.c
> +++ b/fs/ext4/defrag.c
> @@ -287,9 +287,13 @@ static int ext4_ext_extents_info(struct ext4_extents_info *ext_info,
> int entries = 0;
> int err = 0;
>
> - inode = iget(sb, ext_info->ino);
> + inode = iget_locked(sb, ext_info->ino);
> if (!inode)
> return -EACCES;
> + if (inode->i_state & I_NEW) {
> + sb->s_op->read_inode(inode);
> + unlock_new_inode(inode);
> + }
>
> down_write(&EXT4_I(inode)->i_data_sem);
>
read_inode() also sunset, as removed in
http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.24/2.6.24-mm1/broken-out/iget-remove-iget-and-the-read_inode-super-op-as.patch
so above changes won't compile in mm tree.
I have fixed it in patch queue, using ext4_iget() to replace iget() and read-inode() as suggested in iget-stop-ext4-from-using-iget-and-read_inode-try.patch
http://repo.or.cz/w/ext4-patch-queue.git?a=blob;f=ext4-online-defrag-iget-read-inode-fix.patch;h=94660ca7371680fcd02bb5804016b4cae4f20846;hb=85589744f9d53579a12aa463ccd42fb450aec786
> We also have an issue where the read-only bind patches from Dave Hansen
> are making changes which conflict with the ext4 patch tree, which is
> making akpm very grumpy. I'll try to take a look at this later in the
> weekend...
>
> - Ted
>
> @@ -588,9 +592,13 @@ static int ext4_ext_defrag_victim(struct file *target_filp,
> ext.len = 0;
>
> /* Get the inode of the victim file */
> - victim_inode = iget(sb, ex_info->ino);
> + victim_inode = iget_locked(sb, ex_info->ino);
> if (!victim_inode)
> return -EACCES;
> + if (victim_inode->i_state & I_NEW) {
> + sb->s_op->read_inode(victim_inode);
> + unlock_new_inode(victim_inode);
> + }
>
> /* Setup file for the victim file */
> victim_dent.d_inode = victim_inode;
On Sat, 2008-02-02 at 11:40 -0500, Theodore Ts'o wrote:
> We also have an issue where the read-only bind patches from Dave Hansen
> are making changes which conflict with the ext4 patch tree, which is
> making akpm very grumpy. I'll try to take a look at this later in the
> weekend...
>
> - Ted
BTW, Ted, could you clarify about this one? I am able to build the patch
queue on akpm's 2.6.24-mm1 tree with Dave Hansen's read-only bind
changes. Maybe I missed something.
Mingming
On Tue, Feb 05, 2008 at 04:55:52PM -0800, Mingming Cao wrote:
> On Sat, 2008-02-02 at 11:40 -0500, Theodore Ts'o wrote:
> > We also have an issue where the read-only bind patches from Dave Hansen
> > are making changes which conflict with the ext4 patch tree, which is
> > making akpm very grumpy. I'll try to take a look at this later in the
> > weekend...
> >
> > - Ted
>
> BTW, Ted, could you clarify about this one? I am able to build the patch
> queue on akpm's 2.6.24-mm1 tree with Dave Hansen's read-only bind
> changes. Maybe I missed something.
That's because I've already removed the BKL patches that conflict.
- Ted