Re: zram: lockdep spew for zram->init_lock

[Minchan Kim]

Hello Sasha,

On Thu, Feb 27, 2014 at 09:48:37AM -0500, Sasha Levin wrote:
> Hi all,
> 
> I've stumbled on the following spew while fuzzing with trinity
> inside a KVM tools guest running latest -next. It looks like a false
> positive (we only set size for uninitialized devices, so we can't
> deadlock on them being in-use) but I'd really like someone to
> confirm it before I write it down as such.

Thanks for the info!
As you said, it's false positive and introduced by [1] in recent
linux-next but I don't feel we need to annotate to prevent lockdep
spew.

Just enough to move zram_meta_alloc out of lock as old.
Sergey claimed that "let's avoid unnecessary allocation/free of zram_meta
on currently used device" but I think it's not worth to add annotate lock
so that lockdep would be void.

[1] zram: delete zram_init_device()