Re: [PATCH 02/12] DRBD: activity_log

[Andi Kleen <andi@firstfloor.org>]

On Wed, Mar 25, 2009 at 11:27:22AM +0100, Philipp Reisner wrote:
> md_io_page gets allocated with GFP_KERNEL (no GFP_HIGHMEM either).
> 
> 
> [...]
> > > +
> > > +		spin_lock_irq(&mdev->al_lock);
> > > +		lc_changed(mdev->act_log, al_ext);
> > > +		spin_unlock_irq(&mdev->al_lock);
> > > +		wake_up(&mdev->al_wait);
> >
> > The wake_up outside the lock looks a little dangerous.
> >
> 
> Please share you thoughts, why this looks a little dangerous ?

I haven't double checked the whole path, but unlocked wake up
is often a good recipe to potentially lose wake ups.

> [...]
> > > +	mutex_lock(&mdev->md_io_mutex); /* protects md_io_buffer, al_tr_cycle,
> > > ... */
> >
> > Doing checksumming inside a lock looks nasty.
> >
> 
> Well, that is a mutex, not a spinlock. We need to hold that lock here,

Yes it's independent. If it takes a lot of CPU time you'll likely have
a bottle neck. It's normally a bad idea to do anything CPU intensive
under a lock covering more than your current limited object.

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only.