On Tuesday 06 January 2009 20:49, OGAWA Hirofumi wrote:

quoted text
> Hi,

>

> This series is mainly for logmap inode. Current one just use

> new_inode(), so blockget() doesn't work.

>

> So, this uses tux_new_inode() to allocate logmap inode. With it, inode

> will have tux_blk_apos, and blockget() will work correctly.

>

> blockget() uses ->write_begin() with block aligmented range. So,

> block_write_begin() almost no-op except buffer_head setup.

>

> And blockread() of logmap never be used, we can get it via blockdev

> (volume map later), I think it is more clean/efficient way than reading

> logs via logmap. Because if we read the logs via logmap, it would have

> data alias issue.

>

> I think this way is enough for now, and later we will revisit this

> with buffer_head change.

>

> And other patches is warning fix and cleanup.

>

> 	static-http://userweb.kernel.org/~hirofumi/tux3/

>

> Please review and pull.

Ah, nice, so the log gets very close to just being a file.  With

this, we can start adding logging code without segfaults.  :-)
I wonder if we need to have an #ifdef LOG or something like that so we

can add logging functionality that might destabilize things, without

breaking the default build, which people may be trying out now?
The iattr passed here seem to be uninitialized:
	sbi->logmap = tux_new_inode(sbi->rootdir, &iattr, 0);
Why not pass &(struct iattr){ }?
Regards,
Daniel
_______________________________________________

Tux3 mailing list

Tux3@tux3.org

http://mailman.tux3.org/cgi-bin/mailman/listinfo/tux3