Re: lockdep and threaded IRQs (was: ...)

On Mon, 2009-03-02 at 14:40 -0800, David Brownell wrote:

On Monday 02 March 2009, Peter Zijlstra wrote:

quoted

I state that every !IRQF_DISABLED usage is a bug, either
due to broken hardware or broken drivers.

That's a novel position.  You do realize that removing that
capability breaks drivers?

Then we fix them.

But if that's what is keeping you from fixing the lockdep bug,
why haven't you submitted patches to remove IRQF_DISABLED from
the kernel, and update all the drivers relying on IRQs being
enabled when their handlers run?

I did so today. Just didn't realize things actually relied on it since
lockdep turned them off and my system has been working fine.

Your driver needs threaded interrupts, Thomas is working on that now,
and I saw a conversion of your driver to use that.

IDE PIO can hopefully also be converted to threaded interrupts.

After that I'll post patches to remove IRQF_DISABLED and provide a
another flag to quick-'fix' other iffy drivers.

Once such drivers are found we can work on proper fixes for them.