[EMAIL PROTECTED] wrote on 08/04/2006 08:31:37 PM: > Charles P Wright: > > Unionfs only gets hardlinks that were created on Unionfs right. > > If you use the persistent inode number feature, you can handle hardlinks > created by branch fs, can't you? Yes.
> > > What will happen after the cached dentry and inode are freed, by > > > shrink_dcache() or something? Is your i_stacked_inode freed too? Is it > > > still available? > > The lower-level inode can not be thrown out if there is a reference from a > > stacked inode, because the stacked inode takes a reference count on the > > lower-level inode. > > Yes it is. > But how about the 1-1 mapping ballance? For example, if a user does > 'find / -ls' or something, > - ecryptfs/wrapfs/unionfs inode cache may be discarded > - lower_inode may be survived, remains in cache (since unionfs branch is > accessible by 'find') > - lower_inode->i_stacked_inode will point illegal address > - next time when ecryptfs/wrapfs/unionfs access the inode, something bad > will happen Mike nulled out the lower_inode->i_stacked_inode when the ecryptfs inode was freed. > The posted code might be just a prototype to show the concept, so > the handlling the referrence counter of i_stacked_inode might be just > lacked. > But I am afraid if ecryptfs is going to handle iput/iget to > i_stacked_inode, it may create a 'never-freed-inode' since they refers > each other. The lower-level inode doesn't need a reference count, because it doesn't hold onto the upper-level inode. Charles _______________________________________________ unionfs mailing list unionfs@mail.fsl.cs.sunysb.edu http://www.fsl.cs.sunysb.edu/mailman/listinfo/unionfs
