Jeff Mahoney:
>  sysaufs_sbinfo is a pretty ugly hack. It has a refcount, a kset, some
>  attributes, and a way to track if it's been added. This is pretty much
>  what a kobject was designed to do.
> 
>  Rather than duplicating all the code inside aufs, we can leverage the
>  kobject code to do all of that for us.
> 
>  For the attribute functions, we don't need to search for anything anymore
>  because the kobject can be used to reach a properly protected and
>  refcounted sbinfo structure.

Is this case safe too?
- a user opens /sys/fs/aufs/sbi_*/mntpnt1
- anothere user unmount the same aufs
- the first user reads mntpnt1 which was opened previously

I am afraid your approach may corrupt the file structure because of the
lifetime of sysaufs_sbinfo structutre. Your patch makes it freed when
the aufs is unmounted.
This scenario is similar to your approach for sysaufs_br.


Junjiro Okajima

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone

Reply via email to