| Message ID | 20200923065900.658666-1-jhp@endlessos.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/rockchip: skip probed failed device | expand |
Hi, Am Mittwoch, 23. September 2020, 08:59:00 CEST schrieb Jian-Hong Pan: > The cdn-dp sub driver probes the device failed on PINEBOOK Pro. > > kernel: cdn-dp fec00000.dp: [drm:cdn_dp_probe [rockchipdrm]] *ERROR* missing extcon or phy > kernel: cdn-dp: probe of fec00000.dp failed with error -22 > > Then, the device halts all of the DRM related device jobs. For example, > the operations: vop_component_ops, vop_component_ops and > rockchip_dp_component_ops cannot be bound to corresponding devices. So, > Xorg cannot find the correct DRM device. > > This patch skips the probing failed devices to fix this issue. > > Link: http://lists.infradead.org/pipermail/linux-rockchip/2020-September/022352.html > Signed-off-by: Jian-Hong Pan <jhp@endlessos.org> > --- > drivers/gpu/drm/rockchip/rockchip_drm_drv.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > index 0f3eb392fe39..de13588602b4 100644 > --- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > +++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > @@ -331,6 +331,12 @@ static struct component_match *rockchip_drm_match_add(struct device *dev) > > if (!d) > break; > + if (!d->driver) { > + DRM_DEV_ERROR(d, > + "%s did not probe successfully", > + drv->driver.name); > + continue; > + } How does this relate to drivers doing EPROBE_DEFER? Very often you have sub-drivers defering probe as they still need another resource, so excluding them in that case would not work? Heiko > > device_link_add(dev, d, DL_FLAG_STATELESS); > component_match_add(dev, &match, compare_dev, d); >
On 2020-09-23 07:59, Jian-Hong Pan wrote: > The cdn-dp sub driver probes the device failed on PINEBOOK Pro. > > kernel: cdn-dp fec00000.dp: [drm:cdn_dp_probe [rockchipdrm]] *ERROR* missing extcon or phy > kernel: cdn-dp: probe of fec00000.dp failed with error -22 Wouldn't it make more sense to simply not enable the DisplayPort node in the upstream DT, until the type-C phy work has been done to make it usable at all? AIUI the "official" Manjaro kernel is carrying a bunch of hacks to make type-C work via extcon, but they know that isn't an upstreamable solution. Robin. > Then, the device halts all of the DRM related device jobs. For example, > the operations: vop_component_ops, vop_component_ops and > rockchip_dp_component_ops cannot be bound to corresponding devices. So, > Xorg cannot find the correct DRM device. > > This patch skips the probing failed devices to fix this issue. > > Link: http://lists.infradead.org/pipermail/linux-rockchip/2020-September/022352.html > Signed-off-by: Jian-Hong Pan <jhp@endlessos.org> > --- > drivers/gpu/drm/rockchip/rockchip_drm_drv.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > index 0f3eb392fe39..de13588602b4 100644 > --- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > +++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > @@ -331,6 +331,12 @@ static struct component_match *rockchip_drm_match_add(struct device *dev) > > if (!d) > break; > + if (!d->driver) { > + DRM_DEV_ERROR(d, > + "%s did not probe successfully", > + drv->driver.name); > + continue; > + } > > device_link_add(dev, d, DL_FLAG_STATELESS); > component_match_add(dev, &match, compare_dev, d); >
Am Mittwoch, 23. September 2020, 13:05:26 CEST schrieb Robin Murphy: > On 2020-09-23 07:59, Jian-Hong Pan wrote: > > The cdn-dp sub driver probes the device failed on PINEBOOK Pro. > > > > kernel: cdn-dp fec00000.dp: [drm:cdn_dp_probe [rockchipdrm]] *ERROR* missing extcon or phy > > kernel: cdn-dp: probe of fec00000.dp failed with error -22 > > Wouldn't it make more sense to simply not enable the DisplayPort node in > the upstream DT, until the type-C phy work has been done to make it > usable at all? Or alternatively just disable the cdn-dp Rockchip driver in the kernel config, which results in it also not getting probed. > AIUI the "official" Manjaro kernel is carrying a bunch of > hacks to make type-C work via extcon, but they know that isn't an > upstreamable solution. > > Robin. > > > Then, the device halts all of the DRM related device jobs. For example, > > the operations: vop_component_ops, vop_component_ops and > > rockchip_dp_component_ops cannot be bound to corresponding devices. So, > > Xorg cannot find the correct DRM device. > > > > This patch skips the probing failed devices to fix this issue. > > > > Link: http://lists.infradead.org/pipermail/linux-rockchip/2020-September/022352.html > > Signed-off-by: Jian-Hong Pan <jhp@endlessos.org> > > --- > > drivers/gpu/drm/rockchip/rockchip_drm_drv.c | 6 ++++++ > > 1 file changed, 6 insertions(+) > > > > diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > > index 0f3eb392fe39..de13588602b4 100644 > > --- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > > +++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > > @@ -331,6 +331,12 @@ static struct component_match *rockchip_drm_match_add(struct device *dev) > > > > if (!d) > > break; > > + if (!d->driver) { > > + DRM_DEV_ERROR(d, > > + "%s did not probe successfully", > > + drv->driver.name); > > + continue; > > + } > > > > device_link_add(dev, d, DL_FLAG_STATELESS); > > component_match_add(dev, &match, compare_dev, d); > > >
Heiko Stübner <heiko@sntech.de> 於 2020年9月23日 週三 下午7:16寫道： > > Am Mittwoch, 23. September 2020, 13:05:26 CEST schrieb Robin Murphy: > > On 2020-09-23 07:59, Jian-Hong Pan wrote: > > > The cdn-dp sub driver probes the device failed on PINEBOOK Pro. > > > > > > kernel: cdn-dp fec00000.dp: [drm:cdn_dp_probe [rockchipdrm]] *ERROR* missing extcon or phy > > > kernel: cdn-dp: probe of fec00000.dp failed with error -22 > > > > Wouldn't it make more sense to simply not enable the DisplayPort node in > > the upstream DT, until the type-C phy work has been done to make it > > usable at all? > > Or alternatively just disable the cdn-dp Rockchip driver in the kernel config, > which results in it also not getting probed. This may be the simplest way. However, considering generic distro kernels have a policy to enable all drivers, disabling the DisplayPort node in the upstream DT, until the type-C phy work has been done may be a better solution for now. I can prepare a patch for this. Jian-Hong Pan > > AIUI the "official" Manjaro kernel is carrying a bunch of > > hacks to make type-C work via extcon, but they know that isn't an > > upstreamable solution. > > > > Robin. > > > > > Then, the device halts all of the DRM related device jobs. For example, > > > the operations: vop_component_ops, vop_component_ops and > > > rockchip_dp_component_ops cannot be bound to corresponding devices. So, > > > Xorg cannot find the correct DRM device. > > > > > > This patch skips the probing failed devices to fix this issue. > > > > > > Link: http://lists.infradead.org/pipermail/linux-rockchip/2020-September/022352.html > > > Signed-off-by: Jian-Hong Pan <jhp@endlessos.org> > > > --- > > > drivers/gpu/drm/rockchip/rockchip_drm_drv.c | 6 ++++++ > > > 1 file changed, 6 insertions(+) > > > > > > diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > > > index 0f3eb392fe39..de13588602b4 100644 > > > --- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > > > +++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c > > > @@ -331,6 +331,12 @@ static struct component_match *rockchip_drm_match_add(struct device *dev) > > > > > > if (!d) > > > break; > > > + if (!d->driver) { > > > + DRM_DEV_ERROR(d, > > > + "%s did not probe successfully", > > > + drv->driver.name); > > > + continue; > > > + } > > > > > > device_link_add(dev, d, DL_FLAG_STATELESS); > > > component_match_add(dev, &match, compare_dev, d);
diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c index 0f3eb392fe39..de13588602b4 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c @@ -331,6 +331,12 @@ static struct component_match *rockchip_drm_match_add(struct device *dev) if (!d) break; + if (!d->driver) { + DRM_DEV_ERROR(d, + "%s did not probe successfully", + drv->driver.name); + continue; + } device_link_add(dev, d, DL_FLAG_STATELESS); component_match_add(dev, &match, compare_dev, d);
The cdn-dp sub driver probes the device failed on PINEBOOK Pro. kernel: cdn-dp fec00000.dp: [drm:cdn_dp_probe [rockchipdrm]] *ERROR* missing extcon or phy kernel: cdn-dp: probe of fec00000.dp failed with error -22 Then, the device halts all of the DRM related device jobs. For example, the operations: vop_component_ops, vop_component_ops and rockchip_dp_component_ops cannot be bound to corresponding devices. So, Xorg cannot find the correct DRM device. This patch skips the probing failed devices to fix this issue. Link: http://lists.infradead.org/pipermail/linux-rockchip/2020-September/022352.html Signed-off-by: Jian-Hong Pan <jhp@endlessos.org> --- drivers/gpu/drm/rockchip/rockchip_drm_drv.c | 6 ++++++ 1 file changed, 6 insertions(+)