On Friday, 8 November 2024 11:39:57 EST Detlev Casanova wrote: > On Thursday, 23 May 2024 23:09:26 EST Andy Yan wrote: > > Hi Detlev, > > > > At 2024-05-23 02:57:48, "Detlev Casanova" <detlev.casanova@xxxxxxxxxxxxx> > > wrote: > > >At the end of initialization, each VP clock needs to be reset before > > >they can be used. > > > > > >Failing to do so can put the VOP in an undefined state where the > > >generated HDMI signal is either lost or not matching the selected mode. > > > > Would you please provide a detailed description of your test case? > > The test case was to switch modes (using modetest) until the HDMI signal was > lost on the TV side. It was also possible to detect the issue by tracking > the HDMI TX Controller_VIDEO_MONITOR_STATUS[1-6] registers, especially at > address 0x890, where the register would take the value `0x0000018c`. > > After adding these resets, the issue cannot be reproduced. I can share a > script that reproduced this in the past (but this is an old patchset now, so > things could have changed) > > > >Signed-off-by: Detlev Casanova <detlev.casanova@xxxxxxxxxxxxx> > > >--- > > > > > > drivers/gpu/drm/rockchip/rockchip_drm_vop2.c | 30 ++++++++++++++++++++ > > > 1 file changed, 30 insertions(+) > > > > > >diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c > > >b/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c index > > >fdd768bbd487c..e81a67161d29a 100644 > > >--- a/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c > > >+++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c > > >@@ -17,6 +17,7 @@ > > > > > > #include <linux/platform_device.h> > > > #include <linux/pm_runtime.h> > > > #include <linux/regmap.h> > > > > > >+#include <linux/reset.h> > > > > > > #include <linux/swab.h> > > > > > > #include <drm/drm.h> > > > > > >@@ -157,6 +158,7 @@ struct vop2_win { > > > > > > struct vop2_video_port { > > > > > > struct drm_crtc crtc; > > > struct vop2 *vop2; > > > > > >+ struct reset_control *dclk_rst; > > > > > > struct clk *dclk; > > > unsigned int id; > > > const struct vop2_video_port_data *data; > > > > > >@@ -1915,6 +1917,26 @@ static int us_to_vertical_line(struct > > >drm_display_mode *mode, int us)> > > > > > > return us * mode->clock / mode->htotal / 1000; > > > > > > } > > > > > >+static int vop2_clk_reset(struct vop2_video_port *vp) > > >+{ > > >+ struct reset_control *rstc = vp->dclk_rst; > > >+ struct vop2 *vop2 = vp->vop2; > > >+ int ret; > > >+ > > >+ if (!rstc) > > >+ return 0; > > > > In fact, this check is not necessary here. The following reset control > > api > > will check for NULL pointer > > Agreed, I'll do a rebased v3 and remove the check. Actually, re-thinking about it, the check is done to avoid the udelay(10); if there is no resets configured, so I'd rather keep it that way. > > >+ > > >+ ret = reset_control_assert(rstc); > > >+ if (ret < 0) > > >+ drm_warn(vop2->drm, "failed to assert reset\n"); > > >+ udelay(10); > > >+ ret = reset_control_deassert(rstc); > > >+ if (ret < 0) > > >+ drm_warn(vop2->drm, "failed to deassert reset\n"); > > >+ > > >+ return ret; > > >+} > > >+ > > > > > > static void vop2_crtc_atomic_enable(struct drm_crtc *crtc, > > > > > > struct drm_atomic_state > > *state) > > > > { > > > > > >@@ -2055,6 +2077,8 @@ static void vop2_crtc_atomic_enable(struct drm_crtc > > >*crtc,> > > > > > > vop2_vp_write(vp, RK3568_VP_DSP_CTRL, dsp_ctrl); > > > > > >+ vop2_clk_reset(vp); > > >+ > > > > > > drm_crtc_vblank_on(crtc); > > > > > > vop2_unlock(vop2); > > > > > >@@ -2706,6 +2730,12 @@ static int vop2_create_crtcs(struct vop2 *vop2) > > > > > > vp->data = vp_data; > > > > > > snprintf(dclk_name, sizeof(dclk_name), "dclk_vp%d", vp- > > > >id); > > > > >+ vp->dclk_rst = devm_reset_control_get_optional(vop2- > > > >dev, dclk_name); > > > > >+ if (IS_ERR(vp->dclk_rst)) { > > >+ drm_err(vop2->drm, "failed to get %s reset\n", > > dclk_name); > > > >+ return PTR_ERR(vp->dclk_rst); > > >+ } > > >+ > > > > > > vp->dclk = devm_clk_get(vop2->dev, dclk_name); > > > if (IS_ERR(vp->dclk)) { > > > > > > drm_err(vop2->drm, "failed to get %s\n", > > dclk_name);