Hello Chaoyi, On Tue, Jul 28, 2026 at 10:33:48AM +0800, Chaoyi Chen wrote: > >> Changes in v4: > >> - Link to v3: > >> https://lore.kernel.org/all/[email protected]/ > >> - Scan the entire typec_bus and attempt to register the hpd bridge, > >> so as not to miss devices that were already added during initialization. > > > > While that may help with the module being loaded late, it does not > > help with triggering the module load in the first place. > > > > The module does not have any MODULE_DEVICE_TABLE() helping with > > that. So basically any system relying on the automatic hpd bridge > > registration and using a modular kernel (e.g. the arm64 defconfig or > > the kernels usually provided by Linux distributions) requires > > manually loading that module to get any graphical output. > > > > Thank you for pointing this out. I will add it in v5.
Add what exactly? As this code is quite generic, you don't have
anything for a MODULE_DEVICE_TABLE() (i.e. no OF, no I2C, ...).
One solution I see to keep the notifier design would be to add
MODULE_SOFTDEP("pre: aux-hpd-typec-dp-bridge")
to some other module and piggy-back on it. Alternatively integrate
it directly to some other module, such as "drm_display_helper",
which already has DRM_DISPLAY_DP_HELPER and DRM_DISPLAY_DP_TUNNEL.
Adding TypeC DP AltMode as part of it seems sensible to me.
FWIW I suggest you test booting your patches work with the arm64
defconfig before your next submission ;)
Greetings,
-- Sebastian
signature.asc
Description: PGP signature
