<div dir="ltr"><div dir="ltr"> </div> <div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Feb 25, 2020 at 3:38 PM Thomas Zimmermann <<a href="mailto:tzimmermann@suse.de">tzimmermann@suse.de</a>> wrote: </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hi Am 23.02.20 um 05:26 schrieb tang pengchuan: > > > On Sun, Feb 23, 2020 at 5:27 AM Sam Ravnborg <<a href="mailto:sam@ravnborg.org" target="_blank">sam@ravnborg.org</a> > <mailto:<a href="mailto:sam@ravnborg.org" target="_blank">sam@ravnborg.org</a>>> wrote: > > Hi Kevin/tang. > > Thanks for the quick and detailed feedback. > Your questions are addressed below. > > Sam > > > > > > +static int sprd_drm_bind(struct device *dev) > > > > +{ > > > > + struct drm_device *drm; > > > > + struct sprd_drm *sprd; > > > > + int err; > > > > + > > > > + drm = drm_dev_alloc(&sprd_drm_drv, dev); > > > > + if (IS_ERR(drm)) > > > > + return PTR_ERR(drm); > > > You should embed drm_device in struct sprd_drm. > > > See example code in drm/drm_drv.c > > > > > > This is what modern drm drivers do. > > > > > > I *think* you can drop the component framework if you do this. > > > > > I have read it(drm/drm_drv.c) carefully, if drop the component > framework, > > the whole our drm driver maybe need to redesign, so i still want > to keep > > current design. > OK, fine. > > > > > + sprd_drm_mode_config_init(drm); > > > > + > > > > + /* bind and init sub drivers */ > > > > + err = component_bind_all(drm->dev, drm); > > > > + if (err) { > > > > + DRM_ERROR("failed to bind all component.\n"); > > > > + goto err_dc_cleanup; > > > > + } > > > When you have a drm_device - which you do here. > > > Then please use drm_err() and friends for error messages. > > > Please verify all uses of DRM_XXX > > > > > modern drm drivers need drm_xxx to replace DRM_XXX? > Yes, use of DRM_XXX is deprecated - when you have a drm_device. > > > > > > > > + /* with irq_enabled = true, we can use the vblank > feature. */ > > > > + drm->irq_enabled = true; > > > I cannot see any irq being installed. This looks wrong. > > > > > Our display controller isr is been register on crtc > driver(sprd_dpu.c), not > > here. > > I think you just need to move this to next patch and then it is fine. > > Here is the advice given by Thomas Zimmermann, similar to the advice you > gave. > I have given thomas feedback about my questions, maybe thomas didn't see > my email, so there is no reply. I have been busy last week. Sorry for not getting back to you. > > But I've always been confused, because irq is initialized in drm driver > for other guys, why not for me? Do you have an example? </blockquote><div>Hi Thomas,</div><div><div>The the irq is initialized in the sub-device code, but the device state is set on kms driver.</div><div>E.g </div><div>Here is the device state set on kms driver: </div><div><pre><a class="gmail-l gmail-selected" name="206" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_drm_drv.c#206">206 </a><a style="cursor: pointer;" id="gmail-scope_id_c6d1be76_fold_icon"> </a>static int <a class="gmail-xf" name="mtk_drm_kms_init"></a><a href="http://10.0.1.79:8081/s?refs=mtk_drm_kms_init&project=sprdroidr_trunk" class="gmail-xf gmail-intelliWindow-symbol">mtk_drm_kms_init</a>(struct <a href="http://10.0.1.79:8081/s?defs=drm_device&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">drm_device</a> *<a class="gmail-xa" name="drm"></a><a href="http://10.0.1.79:8081/s?refs=drm&project=sprdroidr_trunk" class="gmail-xa gmail-intelliWindow-symbol">drm</a>) ...... </pre></div><div><pre><a class="gmail-l gmail-selected" name="298" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_drm_drv.c#298">298 </a> /* <a class="gmail-l gmail-selected" name="299" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_drm_drv.c#299">299 </a> * We don't use the drm_irq_install() helpers provided by the DRM <a class="gmail-hl gmail-selected" name="300" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_drm_drv.c#300">300 </a> * core, so we need to set this manually in order to allow the <a class="gmail-l gmail-selected" name="301" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_drm_drv.c#301">301 </a> * DRM_IOCTL_WAIT_VBLANK to operate correctly. <a class="gmail-l gmail-selected" name="302" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_drm_drv.c#302">302 </a> */ <a class="gmail-l gmail-selected" name="303" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_drm_drv.c#303">303 </a> <a href="http://10.0.1.79:8081/s?defs=drm&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">drm</a>-><a href="http://10.0.1.79:8081/s?defs=irq_enabled&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">irq_enabled</a> = true; Here is irq install on subdev: <a class="gmail-l gmail-selected" name="265" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_disp_rdma.c#265">265 </a><a style="cursor: pointer;" id="gmail-scope_id_b3948385_fold_icon"> </a>static int <a class="gmail-xf" name="mtk_disp_rdma_probe"></a><a href="http://10.0.1.79:8081/s?refs=mtk_disp_rdma_probe&project=sprdroidr_trunk" class="gmail-xf gmail-intelliWindow-symbol">mtk_disp_rdma_probe</a>(struct <a href="http://10.0.1.79:8081/s?defs=platform_device&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">platform_device</a> *<a class="gmail-xa" name="pdev"></a><a href="http://10.0.1.79:8081/s?refs=pdev&project=sprdroidr_trunk" class="gmail-xa gmail-intelliWindow-symbol">pdev</a>) ...... <a class="gmail-l gmail-target gmail-selected" name="298" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_disp_rdma.c#298">298 </a> <a href="http://10.0.1.79:8081/s?defs=ret&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">ret</a> = <a href="http://10.0.1.79:8081/s?defs=devm_request_irq&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">devm_request_irq</a>(<a href="http://10.0.1.79:8081/s?defs=dev&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">dev</a>, <a href="http://10.0.1.79:8081/s?defs=irq&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">irq</a>, <a class="gmail-d gmail-intelliWindow-symbol" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_disp_rdma.c#mtk_disp_rdma_irq_handler">mtk_disp_rdma_irq_handler</a>, <a class="gmail-l gmail-selected" name="299" href="http://10.0.1.79:8081/xref/sprdroidr_trunk/bsp/kernel/kernel5.4/drivers/gpu/drm/mediatek/mtk_disp_rdma.c#299">299 </a> <a href="http://10.0.1.79:8081/s?defs=IRQF_TRIGGER_NONE&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">IRQF_TRIGGER_NONE</a>, <a href="http://10.0.1.79:8081/s?defs=dev_name&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">dev_name</a>(<a href="http://10.0.1.79:8081/s?defs=dev&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">dev</a>), <a href="http://10.0.1.79:8081/s?defs=priv&project=sprdroidr_trunk" class="gmail-intelliWindow-symbol">priv</a>); I'm not sure if my understanding is wrong...</pre></div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> Best regards Thomas > Can you help to tell the reason in detail, looking forward to your answers. > > Thomas's suggestion： > ------------------------------------------------------------------------------------------- > > This line indicates the problem's design. The irq is initialized in the > sub-device code, but the device state is set here. Instead both should > be set in the same place. > >> + >> + /* reset all the states of crtc/plane/encoder/connector */ >> + drm_mode_config_reset(drm); >> + >> + /* init kms poll for handling hpd */ >> + drm_kms_helper_poll_init(drm); > > Most of this function's code should be moved into the sub-device bind > function. > > Here, maybe do: > > * allocate device structures > * call component_bind_all() > * on success, register device > > The sub-device function should then do > > * init modesetting, crtc, planes, etc. > * do drm_mode_config_reset() > * init vblanking > * init the irq > * do drm_kms_helper_poll_init() > > roughtly in that order. It makes sense to call drm_vblank_init() after > drm_mode_config_reset(), as vblanking uses some of the mode-config fields. > ------------------------------------------------------------------------------------------------------ > > > Sam > > > _______________________________________________ > dri-devel mailing list > <a href="mailto:dri-devel@lists.freedesktop.org" target="_blank">dri-devel@lists.freedesktop.org</a> > <a href="https://lists.freedesktop.org/mailman/listinfo/dri-devel" rel="noreferrer" target="_blank">https://lists.freedesktop.org/mailman/listinfo/dri-devel</a> > -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer </blockquote></div></div>