[Unionfs] Re: 2.6.26-rc4: smack and unionfs deadlock

[Luiz Fernando N. Capitulino]

Em Thu, 05 Jun 2008 22:11:19 -0700
Casey Schaufler <casey at schaufler-ca.com> escreveu:

| Casey Schaufler wrote:
| > Luiz Fernando N. Capitulino wrote:
| >>  Hi Casey,
| >>
| >>  I've got another deadlock which only happens when CONFIG_SMACK is
| >> enabled, but this time it happens with unionfs and I don't know
| >> which of the two is causing the problem.
| >>
| >>   
| >
| > I need a little education. unionfs has a lock on its superblock,
| > which appears to be where we're getting hung up. What condition
| > is this lock for? It would seem that a getxattr would be pretty
| > safe without having the superblock locked, but no one is going
| > to accuse me of being excessively skilled at locking issues.
| > I'm going to see what goes wrong without that lock, and if removing
| > it takes care of the problem. There may be more than one dragon
| > in these waters.
| >
| > Thank you.
| >
| OK. The sb associated with unionfs has a magic number of 0,
| which implies you haven't set it. Is this intentional, or
| do you just figure no one cares, or should care?
| 
| I can relieve the hang by having the Smack code treat
| unionfs (actually the "0" file system) as one of the
| special cases. The "real" file shows the correct smack
| value, the unionfs shows "*", and access is correctly
| controlled in either case. I'm still looking to see what
| to do better, and I'd still like to understand the locking
| strategy better.

 Casey, I'm afraid that only Erez (unionfs maintainer)
can answer you here but it seems that he's away or
something.

 His last message on unionfs ML dates about one month
back.

 :(

-- 
Luiz Fernando N. Capitulino