On 2010-08-06, Lino Sanfilippo <[email protected]> wrote:
> This patch implements writing via memory mapping for dazukofs:
>
> In writepage() we look for the corresponding lower page and create
> it if it does not exist.  Then we copy the content of the upper page
> to the lower page and mark it dirty.  This causes marked pages to be
> written to disk as soon as the pages are synced (i.e due to an
> msync() or sync()).
>
> Note that this patch also removes some of the functions that do also
> cause writepage() to be called, like generic_file_aio_[read|write]
> and splice_read() since mmap has not been tested in conjunction with
> these functions - also writebegin() should not be called any more
> now.
>
> With this the only address space operations that we should need are
> readpage() and writepage() and both are only needed for the memory
> mapping feature and nothing else.

I am not comfortable accepting a patch full of "#if 0" and "BUG()"
lines. I have the feeling you aren't confident that these major
changes are safe.

If you are certain that supporting mmap-write is safe, please send a
new patch against 3.1.4-rc1 that is not full of debug stuff.

John Ogness

-- 
Dazuko Maintainer

_______________________________________________
Dazuko-devel mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/dazuko-devel

Reply via email to