Re: Duplication of dirent names in JFFS2 summary

[David Woodhouse <dwmw2@infradead.org>]

On Fri, 2006-05-19 at 19:26 +0400, Artem B. Bityutskiy wrote:
> 1. During scan nobody can access JFFS2, so "giving incorrect values to
> userspace until we've finished the scan" is wrong.

I shouldn't have said 'scan'. I should have said 'until the GC thread
has finished checking nodes'. 

> 2. Before accessing *any* inode, read_inode() is called. So what I'm
> saying is to calculate *correct* nlink at the read_inode() time.

You can't do that. You don't have correct nlink until you've checked all
the dirents in _all_ the directories on the file system.

> I'm repeating for the 3rd time my statement: names (or their hashes) are
> only needed to calculate correct nlinks. We don't have to calculate
> correct nlinks at the *scan* time, we may try to do this at the
> *read_inode()* time. See what I mean? 

I know what you mean, and it's _still_ not correct.

-- 
dwmw2