Re: [PATCH] ramfs: pretend dirent sizes

[Jan Blunck <j.blunck@tu-harburg.de>]

Chris Wedgwood wrote:
> 
> So the size you want to reflect is n*<stack-depth> i take it?  Where
> in this case n is 20?
> 
> So you can seek to m*<stack-depth>+<offset> to access an offset into
> something at depth m?
> 

Yes.

>>The i_size of a directory isn't covered by the POSIX standard. IMO,
>>it should be possible to seek in the range of i_size and a following
>>readdir() on the directory should succeed.
> 
> With what defined semantics?  What if an entry is added in between
> seek and readdir?
> 

You have the same problem with regular files. This is a user and not a 
kernel problem.

> 
> Why?  It seems perfectly reasonable that we can return 0 in such
> cases.  Zero seems to make more sense as 'magical/unknown' than say
> any other arbitrary value.
> 

I disagree. Where is the information value of i_size if we always could 
return 0? IMO it should be at least an upper bound for the "number" of 
informations that could actually be read (in terms of a seek offset) 
like it is in the case of regular files. Better, if it is a strict upper 
bound so that you can seek to every value smaller than i_size. For this 
purpose the i_size of directories doesn't need to reflect any unit.

Jan