On 2019-11-11 21:35:35 [-0600], Rob Herring wrote:
> > 28d0e36bf9686 ("OF: Fixup resursive locking code paths")
> > d6d3c4e656513 ("OF: convert devtree lock from rw_lock to raw spinlock")
>
> So to summarize, we changed mainline to fix RT which then broke RT. :)
correct, but we were good until v4.17-rc1 :)
> > I've been looking into making devtree_lock a spinlock_t which would
> > avoid this patch. I haven't seen an issue during boot on arm64 even
> > with hotplug.
>
> Did you look into using RCU reader locks any more?
A little bit. The writers, which modify the node, would need to replace
the whole node. So this is where things got a little complicated.
Frank wasn't a big fan of it back then and he still wasn't a few weeks
back.
If you two agree to prefer RCU over this patch then I would look more
into adding RCU into the lookup path. The argument was that this isn't
time critical. I'm just trying to avoid to replace the locking for
nothing.
So, should I come up with a RCU patch?
> Rob
Sebastian
