On Wed, Dec 18, 2013 at 09:07:59PM -0500, Theodore Ts'o wrote:

On Wed, Dec 18, 2013 at 07:27:49AM -0700, Matthew Wilcox wrote:

quoted

I think there is a callback in xip_file_write(), and it's get_xip_mem().
From what you're saying, it sounds like it's just not doing enough.

The problem is that git_xip_mem() is called before we write to the
memory, right?

We need to convert the uninit extents to be marked as initialized in
*after* the write has been sent to the storage medium.

Now that I've spent the best part of a day looking at the ext4 code, I
still don't think there's a problem here.  With the way the XIP code is
currently written (calling ext4_get_block with create=1), we won't get an
uninitialised extent in the caller.  Instead, we'll get one that's been
zeroed (the zeroing is part of patch 3/3 and done only for xip files).

As I understand it, when ext4 uses direct I/O, it can pass
ext4_get_block_write() as the get_block method, which uses the
magic EXT4_GET_BLOCKS_IO_CREATE_EXT flag to permit the allocation of
uninitialised extents.  But the regular ext4_get_block cannot create
uninitialised extents (it can return them in the case of create=0,
and we handle that correctly as a hole).

Moreover, I don't see, eg, block_read_full_page() handling uninitialised
extents specially, so I'm pretty sure the regular ext4_get_block()
can't create uninitialised extents.

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."