@@ -1312,6 +1312,8 @@ static int mtk_jpeg_probe(struct platform_device *pdev)
jpeg->dev = &pdev->dev;
jpeg->variant = of_device_get_match_data(jpeg->dev);
+ platform_set_drvdata(pdev, jpeg);
+
ret = devm_of_platform_populate(&pdev->dev);
if (ret) {
v4l2_err(&jpeg->v4l2_dev, "Master of platform populate failed.");
@@ -1383,8 +1385,6 @@ static int mtk_jpeg_probe(struct platform_device *pdev)
jpeg->variant->dev_name, jpeg->vdev->num,
VIDEO_MAJOR, jpeg->vdev->minor);
- platform_set_drvdata(pdev, jpeg);
-
pm_runtime_enable(&pdev->dev);
return 0;
In the multi-core JPEG encoder/decoder setup, the driver for the individual cores references the parent device's platform driver data. However, in the parent driver, this is only set at the end of the probe function, way later than devm_of_platform_populate(), which triggers the probe of the cores. This causes a kernel splat in the sub-device probe function. Move platform_set_drvdata() to before devm_of_platform_populate() to fix this. Fixes: 934e8bccac95 ("mtk-jpegenc: support jpegenc multi-hardware") Signed-off-by: Chen-Yu Tsai <wenst@chromium.org> --- Not sure what caused this to surface just now, given that the driver changes were merged in v6.2-rc1, and the corresponding dts entries in v6.3-rc1. This fixes boot failures on MT8195. Please apply and send to Linus ASAP. drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)