msm: kgsl: Add facility to BUG_ON for a gpu fault

Few gpu faults need system level information like voltages,
interface clock etc. This information can't be extracted
through the snapshot dump. Add a facility to force panic
on gpu fault, which will help to extract additional
information from the ramdumps.

To trigger BUG_ON for a gpu fault:
echo 1  > /sys/class/kgsl/kgsl-3d0/snapshot/force_panic

Change-Id: I020b5518fcaedd4fecc572b580a989bf117f6ceb
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>

2 files changed, 37 insertions, 2 deletions

diff --git a/drivers/gpu/msm/kgsl_device.h b/drivers/gpu/msm/kgsl_device.h
index 0df6dd8628a5..25021063d438 100644
--- a/drivers/gpu/msm/kgsl_device.h
+++ b/drivers/gpu/msm/kgsl_device.h
@@ -261,6 +261,7 @@ struct kgsl_device {
 	struct kgsl_snapshot *snapshot;
 
 	u32 snapshot_faultcount;	/* Total number of faults since boot */
+	bool force_panic;		/* Force panic after snapshot dump */
 	struct kobject snapshot_kobj;
 
 	struct kobject ppd_kobj;
diff --git a/drivers/gpu/msm/kgsl_snapshot.c b/drivers/gpu/msm/kgsl_snapshot.c
index 2c9f17f9e7a4..a2e4a909062f 100644
--- a/drivers/gpu/msm/kgsl_snapshot.c
+++ b/drivers/gpu/msm/kgsl_snapshot.c
@@ -810,6 +810,29 @@ static ssize_t faultcount_store(struct kgsl_device *device, const char *buf,
 	return count;
 }
 
+/* Show the force_panic request status */
+static ssize_t force_panic_show(struct kgsl_device *device, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%d\n", device->force_panic);
+}
+
+/* Store the panic request value to force_panic */
+static ssize_t force_panic_store(struct kgsl_device *device, const char *buf,
+	size_t count)
+{
+	unsigned int val = 0;
+	int ret;
+
+	if (device && count > 0)
+		device->force_panic = 0;
+
+	ret = kgsl_sysfs_store(buf, &val);
+
+	if (!ret && device)
+		device->force_panic = (bool)val;
+
+	return (ssize_t) ret < 0 ? ret : count;
+}
 /* Show the timestamp of the last collected snapshot */
 static ssize_t timestamp_show(struct kgsl_device *device, char *buf)
 {
@@ -835,6 +858,7 @@ struct kgsl_snapshot_attribute attr_##_name = { \
 
 static SNAPSHOT_ATTR(timestamp, 0444, timestamp_show, NULL);
 static SNAPSHOT_ATTR(faultcount, 0644, faultcount_show, faultcount_store);
+static SNAPSHOT_ATTR(force_panic, 0644, force_panic_show, force_panic_store);
 
 static ssize_t snapshot_sysfs_show(struct kobject *kobj,
 	struct attribute *attr, char *buf)
@@ -914,6 +938,7 @@ int kgsl_device_snapshot_init(struct kgsl_device *device)
 
 	device->snapshot = NULL;
 	device->snapshot_faultcount = 0;
+	device->force_panic = 0;
 
 	ret = kobject_init_and_add(&device->snapshot_kobj, &ktype_snapshot,
 		&device->dev->kobj, "snapshot");
@@ -929,7 +954,11 @@ int kgsl_device_snapshot_init(struct kgsl_device *device)
 		goto done;
 
 	ret  = sysfs_create_file(&device->snapshot_kobj, &attr_faultcount.attr);
+	if (ret)
+		goto done;
 
+	ret  = sysfs_create_file(&device->snapshot_kobj,
+			&attr_force_panic.attr);
 done:
 	return ret;
 }
@@ -954,6 +983,7 @@ void kgsl_device_snapshot_close(struct kgsl_device *device)
 	device->snapshot_memory.ptr = NULL;
 	device->snapshot_memory.size = 0;
 	device->snapshot_faultcount = 0;
+	device->force_panic = 0;
 }
 EXPORT_SYMBOL(kgsl_device_snapshot_close);
 
@@ -1032,6 +1062,7 @@ void kgsl_snapshot_save_frozen_objs(struct work_struct *work)
 {
 	struct kgsl_snapshot *snapshot = container_of(work,
 				struct kgsl_snapshot, work);
+	struct kgsl_device *device = kgsl_get_device(KGSL_DEVICE_3D0);
 	struct kgsl_snapshot_object *obj, *tmp;
 	size_t size = 0;
 	void *ptr;
@@ -1073,12 +1104,15 @@ done:
 	snapshot->process = NULL;
 
 	if (snapshot->ib1base && !snapshot->ib1dumped)
-		pr_warn("kgsl: snapshot: Active IB1:%016llx not dumped\n",
+		KGSL_DRV_ERR(device,
+				"snapshot: Active IB1:%016llx not dumped\n",
 				snapshot->ib1base);
 	else if (snapshot->ib2base && !snapshot->ib2dumped)
-		pr_warn("kgsl: snapshot: Active IB2:%016llx not dumped\n",
+		KGSL_DRV_ERR(device,
+			       "snapshot: Active IB2:%016llx not dumped\n",
 				snapshot->ib2base);
 
 	complete_all(&snapshot->dump_gate);
+	BUG_ON(device->force_panic);
 	return;
 }