Hi Am 29.07.21 um 21:18 schrieb Thomas Zimmermann: > Hi > > Am 28.07.21 um 22:11 schrieb Sam Ravnborg: >> Hi Dan, >> >>>> >>>> I think I got it - we need to set irq_enabled to true. >>>> The documentation says so: >>>> " >>>>            * @irq_enabled: >>>>            * >>>>            * Indicates that interrupt handling is enabled, >>>> specifically vblank >>>>            * handling. Drivers which don't use drm_irq_install() >>>> need to set this >>>>            * to true manually. >>>> " >>>> >>>> Can you try to add the following line: >>>> >>>> >>>> +static int atmel_hlcdc_dc_irq_install(struct drm_device *dev, >>>> unsigned int irq) >>>> +{ >>>> +       int ret; >>>> + >>>> +       if (irq == IRQ_NOTCONNECTED) >>>> +               return -ENOTCONN; >>>> + >>>> >>>>           dev->irq_enabled = true;                <= THIS LINE >>>> >>>> >>>> +       atmel_hlcdc_dc_irq_disable(dev); >>>> + >>>> +       ret = request_irq(irq, atmel_hlcdc_dc_irq_handler, 0, >>>> dev->driver->name, dev); >>>> +       if (ret) >>>> +               return ret; >>>> >>>> I hope this fixes it. >>> >>> It does!  With the irq_enabled line added everything is looking good. > > Are you sure, you're testing with the latest drm-misc-next or drm-tip? > Because using irq_enabled is deprecated and the flag was recently > replaced by commit 1e4cd78ed493 ("drm: Don't test for IRQ support in > VBLANK ioctls"). For reference, the cover letter lists base-commit: 2bda1ca4d4acb4892556fec3a7ea1f02afcd40bb which is a recent drm-tip created on July 25. Best regards Thomas > > Best regards > Thomas > >> >> Great, thanks for testing. >> >> Thomas - I assume you will do a re-spin and there is likely some fixes >> for the applied IRQ conversions too. >> >> Note - irq_enabled must be cleared if request_irq fails. I did not >> include this in the testing here. >> >>     Sam >> > -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer