* Jamal Hadi Salim [off-list ref] 2006-07-08 10:14

I dont know if i tested for the above specific setup. I have to look at
my testcases when i get the opportunity.
The problem is:
There is no easy way to detect such a deadlock. I think it reduces to
the same case as eth0->eth0, i.e:

It's very simple. Just use the same method and add a flag if a mirred
is currently in use as Herbert did to qdisc_run().

I have a dated  patch to mirred (may not apply cleanly) that i believe
will fix this specific one. Try to see if it also fixes this case you
have.

This is complete nonsense, fixing a case out of a generic
problem is useless.

Its the second pathological case tx deadlock essentially and mirred
could help - it is not the mirred lock really if the attached patch
fixes it, but i could be wrong.

__LINK_STATE_QDISC_RUNNING will prevent an eventual tx deadlock,
it will not prevent the mirred deadlock.

I think everything has been said about this, since I'm still not
getting the intend of many parts I'm not able to fix these, sorry.