Re: [PATCH 1/5] drm/i915: Fix forcewake counts for gen8

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Fri, Feb 21, 2014 at 11:38:48AM -0800, Ben Widawsky wrote:
> On Fri, Feb 21, 2014 at 05:31:59PM +0200, mika.kuoppala@intel.com wrote:
> > From: Mika Kuoppala <mika.kuoppala@intel.com>
> > 
> > Sometimes generic driver code gets forcewake explicitly by
> > gen6_gt_force_wake_get(), which check forcewake_count before accessing
> > hardware. However the register access with gen8_write function access
> > low level hw accessors directly, ignoring the forcewake_count. This
> > leads to nested forcewake get from hardware, in ring init and possibly
> > elsewhere, causing forcewake ack clear errors and/or hangs.
> > 
> > Fix this by checking the forcewake count also in gen8_write
> > 
> > v2: Read side doesn't care about shadowed registers,
> >     Remove __needs_put funkiness from gen8_write. (Ville)
> >     Improved commit message.
> > 
> > References: https://bugs.freedesktop.org/show_bug.cgi?id=74007
> > Cc: Ben Widawsky <benjamin.widawsky@intel.com>
> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
> > Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
> 
> For those concerned with the performance implication of the extra if
> (if anyone at all cares, it's Chris) - I suppose we could also just add
> the lock to gen6_gt_force_wake_get/put.

Two things:
a) I don't understand what you mean here. uncore.lock already protects
the forcewake count in gen6_gt_force_wake_get/put. Also there's no way
to avoid the branch here since doing a .force_wake_put() w/o checking
the forcewake count is never ok.

b) The cost of branch should be a drop in the ocean compared to the
cost of the register reads/writes in .forcewake_get/put.

-- 
Ville Syrjälä
Intel OTC