On 2022/4/20 15:51, Maxime Ripard wrote: > On Wed, Apr 20, 2022 at 12:49:48AM +0000, Miaoqian Lin wrote: >> If the device is already in a runtime PM enabled state >> pm_runtime_get_sync() will return 1, so a test for negative >> value should be used to check for errors. >> >> Also, we need to call pm_runtime_put_noidle() when pm_runtime_get_sync() >> fails, so use pm_runtime_resume_and_get() instead. this function >> will handle this. >> >> Fixes: 4078f5757144 ("drm/vc4: Add DSI driver") >> Signed-off-by: Miaoqian Lin <linmq006@xxxxxxxxx> >> --- >> change in v2: >> - switch to pm_runtime_resume_and_get() to fix refcount leak. >> --- >> drivers/gpu/drm/vc4/vc4_dsi.c | 4 ++-- >> 1 file changed, 2 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/gpu/drm/vc4/vc4_dsi.c b/drivers/gpu/drm/vc4/vc4_dsi.c >> index 752f921735c6..9d7ffaf6bc70 100644 >> --- a/drivers/gpu/drm/vc4/vc4_dsi.c >> +++ b/drivers/gpu/drm/vc4/vc4_dsi.c >> @@ -846,8 +846,8 @@ static void vc4_dsi_encoder_enable(struct drm_encoder *encoder) >> unsigned long phy_clock; >> int ret; >> >> - ret = pm_runtime_get_sync(dev); >> - if (ret) { >> + ret = pm_runtime_resume_and_get(dev); >> + if (ret < 0) { > pm_runtime_resume_and_get will return 0 on success, so the previous check was correct previous check is for pm_runtime_get_sync() not for pm_runtime_resume_and_get (), I switch to pm_runtime_resume_and_get() to fix the refcount leak bug at the same time. Sure it's ok to use check if(ret) to check the retval, I just follow a more common way for usage of pm_runtime_resume_and_get() in the codebase—— check ret<0 Since pm_runtime_resume_and_get() return negative error code. > Maxime