| Message ID | 20180302175757.28192-6-enric.balletbo@collabora.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Laurent, sorry, you're right, this patch should not be needed. the connector should be cleanup by drm_mode_config_cleanup->drm_connector_put. i did that in analogix_dp is to avoid a use-after-free issue not kmemleak, because the connector was allocated/freed in analogix_dp's bind/unbind. but i found a kmemleak issue(dma_mask not freed) in dw-hdmi when testing that, will send patch soon. On 03/03/2018 08:20 AM, JeffyChen wrote: > Hi Laurent, > > On 03/03/2018 05:49 AM, Laurent Pinchart wrote: >> Hi Enric, >> >> Thank you for the patch. >> >> On Friday, 2 March 2018 19:57:57 EET Enric Balletbo i Serra wrote: >>> From: Jeffy Chen <jeffy.chen@rock-chips.com> >>> >>> We inited connector in attach(), so need a detach() to cleanup. >> >> Do we ? The dw-hdmi driver already sets drm_connector_cleanup() as the >> connector .destroy() handler, and the .destroy() operation is called >> by the >> DRM core. None of the other bridge drivers call drm_connector_cleanup() >> directly. > > hmmm, checking the code, there are also lots of drivers do the > cleanup(drm_connector_cleanup or funcs->destroy): > drm# grep -r "connector.*funcs->destroy" . > ./rockchip/inno_hdmi.c: hdmi->connector.funcs->destroy(&hdmi->connector); > ./rockchip/cdn-dp-core.c: connector->funcs->destroy(connector); > ./bridge/analogix/analogix_dp_core.c: > dp->connector.funcs->destroy(&dp->connector); > ./msm/hdmi/hdmi.c: hdmi->connector->funcs->destroy(hdmi->connector); > ./msm/dsi/dsi.c: msm_dsi->connector->funcs->destroy(msm_dsi->connector); > ./msm/edp/edp.c: edp->connector->funcs->destroy(edp->connector); > ./zte/zx_hdmi.c: hdmi->connector.funcs->destroy(&hdmi->connector); > ./drm_connector.c: connector->funcs->destroy(connector); > ./drm_connector.c: connector->funcs->destroy(connector); > ./nouveau/dispnv04/disp.c: connector->funcs->destroy(connector); > ./nouveau/nv50_display.c: mstc->connector.funcs->destroy(&mstc->connector); > ./nouveau/nv50_display.c: connector->funcs->destroy(connector); > > > > when i debug analogix_dp bind/unbind, i found that we need to cleanup > the connector(reported by kmemleak). so i added it to > ./bridge/analogix/analogix_dp_core.c...after that i saw dw-hdmi missing > that too(by checking the code), so make this patch. > > but i didn't really tested it on devices using dw-hdmi, so i'm not very > sure the dw-hdmi(maybe also other bridges) is the same with analogix_dp. > > i can try to find a chromebook veyron to check it next week :) > > but even there's a leak, i'm still not very sure about: > should the caller of drm_connector_init cleanup it > or the caller of drm_bridge_attach should do it(for example > analogix_dp_bind/analogix_dp_unbind) > or should the DRM core take care of that? > >> >>> Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com> >>> Signed-off-by: Thierry Escande <thierry.escande@collabora.com> >>> Signed-off-by: Enric Balletbo i Serra <enric.balletbo@collabora.com> >>> ---
diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c index f9802399cc0d..5626922f95f9 100644 --- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c +++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c @@ -1985,6 +1985,13 @@ static int dw_hdmi_bridge_attach(struct drm_bridge *bridge) return 0; } +static void dw_hdmi_bridge_detach(struct drm_bridge *bridge) +{ + struct dw_hdmi *hdmi = bridge->driver_private; + + drm_connector_cleanup(&hdmi->connector); +} + static enum drm_mode_status dw_hdmi_bridge_mode_valid(struct drm_bridge *bridge, const struct drm_display_mode *mode) @@ -2041,6 +2048,7 @@ static void dw_hdmi_bridge_enable(struct drm_bridge *bridge) static const struct drm_bridge_funcs dw_hdmi_bridge_funcs = { .attach = dw_hdmi_bridge_attach, + .detach = dw_hdmi_bridge_detach, .enable = dw_hdmi_bridge_enable, .disable = dw_hdmi_bridge_disable, .mode_set = dw_hdmi_bridge_mode_set,