gpu: nvgpu: fix race between do_idle() and unrailgate()

While we are executing do_idle() API, it is possible that
unrailgate() gets invoked in midst of idling the GPU and
this can result in failure of do_idle()

To prevent simultaneous execution of these methods,
add a mutex railgate_lock and acquire it during
do_idle() and unrailgate() APIs

Also, keep this lock held if do_idle() is successful.
In success, lock will be released in do_unidle(),
otherwise release this lock before returning

Note that this lock should not be held in railgate() API
since we do not want it to be blocked during do_idle()

bug 1529160

Change-Id: I87114b5367eaa217376455a2699c0d21c451c889
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/434190
GVS: Gerrit_Virtual_Submit
Reviewed-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-by: Sachin Nikam <snikam@nvidia.com>

2 files changed, 19 insertions, 3 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.c b/drivers/gpu/nvgpu/gk20a/gk20a.c
index f3dbf31ed2b8..74138e36f5e7 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.c
@@ -1193,8 +1193,11 @@ static int gk20a_pm_unrailgate(struct generic_pm_domain *domain)
 	struct gk20a_platform *platform = platform_get_drvdata(g->dev);
 	int ret = 0;
 
-	if (platform->unrailgate)
+	if (platform->unrailgate) {
+		mutex_lock(&platform->railgate_lock);
 		ret = platform->unrailgate(platform->g->dev);
+		mutex_unlock(&platform->railgate_lock);
+	}
 
 	return ret;
 }
@@ -1259,6 +1262,8 @@ static int gk20a_pm_init(struct platform_device *dev)
 	struct gk20a_platform *platform = platform_get_drvdata(dev);
 	int err = 0;
 
+	mutex_init(&platform->railgate_lock);
+
 	/* Initialise pm runtime */
 	if (platform->clockgate_delay) {
 		pm_runtime_set_autosuspend_delay(&dev->dev,
@@ -1623,6 +1628,13 @@ int gk20a_do_idle(void)
 	/* acquire busy lock to block other busy() calls */
 	down_write(&g->busy_lock);
 
+	/* acquire railgate lock to prevent unrailgate in midst of do_idle() */
+	mutex_lock(&platform->railgate_lock);
+
+	/* check if it is already railgated ? */
+	if (platform->is_railgated(pdev))
+		return 0;
+
 	/* prevent suspend by incrementing usage counter */
 	pm_runtime_get_noresume(&pdev->dev);
 
@@ -1660,11 +1672,12 @@ int gk20a_do_idle(void)
 	}
 
 	/* GPU is not rail gated by now, return error */
-	up_write(&g->busy_lock);
-	return -EBUSY;
+	goto fail_timeout;
 
 fail:
 	pm_runtime_put_noidle(&pdev->dev);
+fail_timeout:
+	mutex_unlock(&platform->railgate_lock);
 	up_write(&g->busy_lock);
 	return -EBUSY;
 }
@@ -1678,8 +1691,10 @@ int gk20a_do_unidle(void)
 		bus_find_device_by_name(&platform_bus_type,
 		NULL, "gk20a.0"));
 	struct gk20a *g = get_gk20a(pdev);
+	struct gk20a_platform *platform = dev_get_drvdata(&pdev->dev);
 
 	/* release the lock and open up all other busy() calls */
+	mutex_unlock(&platform->railgate_lock);
 	up_write(&g->busy_lock);
 
 	return 0;
diff --git a/drivers/gpu/nvgpu/gk20a/platform_gk20a.h b/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
index 50358af6f11d..7ce7f95b620f 100644
--- a/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
@@ -103,6 +103,7 @@ struct gk20a_platform {
 
 	/* Called to turn on the device */
 	int (*unrailgate)(struct platform_device *dev);
+	struct mutex railgate_lock;
 
 	/* Called to check state of device */
 	bool (*is_railgated)(struct platform_device *dev);