gpu: nvgpu: fix arbiter teardown on PCI

The driver is not properly tearing down the arbiter on the PCI driver
unload. This change makes sure that the workqueues are drained before
tearing down the driver

bug 200277762
JIRA: EVLR-1023

Change-Id: If98fd00e27949ba1569dd26e2af02b75897231a7
Signed-off-by: David Nieto <dmartineznie@nvidia.com>
Reviewed-on: http://git-master/r/1320147
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>

1 files changed, 13 insertions, 5 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c b/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
index 5a4a2251..753623fa 100644
--- a/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
@@ -39,6 +39,7 @@
 
 struct gk20a_ctrl_priv {
         struct device *dev;
+        struct gk20a *g;
 #ifdef CONFIG_ARCH_TEGRA_18x_SOC
         struct nvgpu_clk_session *clk_session;
 #endif
@@ -58,35 +59,42 @@ int gk20a_ctrl_dev_open(struct inode *inode, struct file *filp)
         priv = kzalloc(sizeof(struct gk20a_ctrl_priv), GFP_KERNEL);
         if (!priv)
                 return -ENOMEM;
-
         filp->private_data = priv;
         priv->dev = g->dev;
+        /*
+         * We dont close the arbiter fd's after driver teardown to support
+         * GPU_LOST events, so we store g here, instead of dereferencing the
+         * dev structure on teardown
+         */
+        priv->g = g;
 
         if (!g->gr.sw_ready) {
                 err = gk20a_busy(g->dev);
                 if (err)
                         return err;
-
                 gk20a_idle(g->dev);
         }
 
 #ifdef CONFIG_ARCH_TEGRA_18x_SOC
         err = nvgpu_clk_arb_init_session(g, &priv->clk_session);
+        if (err)
+                return err;
 #endif
+
         return err;
 }
-
 int gk20a_ctrl_dev_release(struct inode *inode, struct file *filp)
 {
         struct gk20a_ctrl_priv *priv = filp->private_data;
+        struct gk20a *g = priv->g;
 
         gk20a_dbg_fn("");
 
 #ifdef CONFIG_ARCH_TEGRA_18x_SOC
         if (priv->clk_session)
-                nvgpu_clk_arb_release_session(gk20a_from_dev(priv->dev),
-                                priv->clk_session);
+                nvgpu_clk_arb_release_session(g, priv->clk_session);
 #endif
+
         kfree(priv);
 
         return 0;