Hi Simon, On 2026-09-10 Simon Glass wrote: > I wonder how many other drivers probe in of_to_plat() ? It would be > interested to get the scale of the problem, if you can.
I grepped next for gpio_request_by_name(), gpio_request_list_by_name() and dm_gpio_request() called from an of_to_plat() method: 50 driver files do it. By subsystem: video 22 (panels and backlights), net 6, mmc 5, spi 4, usb 3 (ehci-vf among them, which is where I started), pci 2, i2c 2, power 2 (the regulators in this series), w1, tpm, sound and misc one each. Most of them claim reset or enable lines that they then drive from probe(), so they would convert the same way the regulators do here: parse in of_to_plat(), request in probe(). I can only test two of them: fec_mxc and fsl_esdhc_imx run on the Colibri VF50 I have, and I will convert those in a follow-up series once the helper is settled. I would rather not touch the other drivers, the panels in particular, without hardware to check that moving the claim out of of_to_plat() does not change what they power up and when. I can post the list so their owners can pick them up. I will answer the per-patch comments in their threads and send v2 once those are settled. Regards, Mehmet
