Following a GPU hang, we break out of the request loop in order to
unlock the struct_mutex for use by the GPU reset. However, if we retire
all the requests at that moment, we cannot identify the guilty request
after performing the reset.
v2: Not automatically retiring requests forces us to recheck for
available ringspace.
Fixes: f4457ae71fd6 ("drm/i915: Prevent leaking of -EIO from i915_wait_request()")
Testcase: igt/gem_reset_stats/ban-*
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Tested-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463137042-9669-4-git-send-email-chris@chris-wilson.co.uk
        if (ret)
                return ret;
 
-       __i915_gem_request_retire__upto(req);
+       /* If the GPU hung, we want to keep the requests to find the guilty. */
+       if (req->reset_counter == i915_reset_counter(&dev_priv->gpu_error))
+               __i915_gem_request_retire__upto(req);
+
        return 0;
 }
 
        else if (obj->last_write_req == req)
                i915_gem_object_retire__write(obj);
 
-       __i915_gem_request_retire__upto(req);
+       if (req->reset_counter == i915_reset_counter(&req->i915->gpu_error))
+               __i915_gem_request_retire__upto(req);
 }
 
 /* A nonblocking variant of the above wait. This is a highly dangerous routine
 
                        return ret;
 
                intel_ring_update_space(ringbuf);
+               if (unlikely(ringbuf->space < wait_bytes))
+                       return -EAGAIN;
        }
 
        if (unlikely(need_wrap)) {