drm/i915: Replace lockless_dereference(bool) with READ_ONCE()
authorChris Wilson <chris@chris-wilson.co.uk>
Tue, 5 Jul 2016 07:54:36 +0000 (08:54 +0100)
committerChris Wilson <chris@chris-wilson.co.uk>
Tue, 5 Jul 2016 09:58:07 +0000 (10:58 +0100)
After Joonas complained about using READ_ONCE() on the only use of the
variable in the function, where the intent was to simply document that
the read was intentionally racy and unlocked, I switched the READ_ONCE()
over to lockless_dereference(). However, in linux-next that has a
stronger type-check to only allow pointers and is no longer
interchangeable with READ_ONCE(), see commit 331b6d8c7afc
("locking/barriers: Validate lockless_dereference() is used on a pointer
type")

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 67d97da34917 ("drm/i915: Only start retire worker when idle")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1467705276-707-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
drivers/gpu/drm/i915/i915_gem.c
drivers/gpu/drm/i915/i915_irq.c

index a7852dd..20123c0 100644 (file)
@@ -3290,7 +3290,7 @@ i915_gem_retire_work_handler(struct work_struct *work)
         * We do not need to do this test under locking as in the worst-case
         * we queue the retire worker once too often.
         */
-       if (lockless_dereference(dev_priv->gt.awake))
+       if (READ_ONCE(dev_priv->gt.awake))
                queue_delayed_work(dev_priv->wq,
                                   &dev_priv->gt.retire_work,
                                   round_jiffies_up_relative(HZ));
index f6de8dd..2f01b0b 100644 (file)
@@ -3095,7 +3095,7 @@ static void i915_hangcheck_elapsed(struct work_struct *work)
        if (!i915.enable_hangcheck)
                return;
 
-       if (!lockless_dereference(dev_priv->gt.awake))
+       if (!READ_ONCE(dev_priv->gt.awake))
                return;
 
        /* As enabling the GPU requires fairly extensive mmio access,