On 27/10/2023 21:22, Rob Herring wrote: > On Fri, Oct 27, 2023 at 01:25:56PM +0200, Krzysztof Kozlowski wrote: >> On 27/10/2023 11:09, Wolfram Sang wrote: >>> On Fri, Oct 27, 2023 at 04:31:03PM +1300, Chris Packham wrote: >>>> Add bus-reset-gpios and bus-reset-duration-us properties to the >>>> marvell,mv64xxx-i2c binding. These can be used to describe hardware >>>> where a common reset GPIO is connected to all downstream devices on and >>>> I2C bus. This reset will be asserted then released before the downstream >>>> devices on the bus are probed. >>>> >>>> Signed-off-by: Chris Packham <chris.packham@xxxxxxxxxxxxxxxxxxx> >>>> Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx> >>> >>> Krzysztof, are you fine with this change? >> >> Actually no. NAK. >> >> Not because of the naming, but because the new name triggered some new >> paths in my brain which brought the point - this is old problem of power >> sequencing of children. >> >> I believe this must be solved in more generic way. First - generic for >> all I2C devices. Second - generic also matching other buses/subsystems, >> which have similar problem. We did it for USB (onboard USB), MMC >> (unloved MMC power sequence) and now we are doing it for PCIe and few >> others (Cc: Abel) > > Unlike the others I2C doesn't expect to access the bus/device before > devices probe, right? > >> https://lpc.events/event/17/contributions/1507/ > > Oh, good! > >> Current solution is heavily limited. What about regulators? What about >> buses having 2 reset lines (still the same bus)? What about sequence? > > A more complicated case should be handled by the device's driver. If the > GPIO reset was not shared we'd be handling it there too. I think what's > needed is to solve the shared aspect. That's already done with reset > subsys, so I think making 'reset-gpios' handled by it too is the way > forward. That would handle the QCA WiFi/BT case I think. > > I'm not sure waiting for that or something else to happen is worth > holding up this simple case. It's not the only case of a common reset > for a bus (MDIO). I argue also that this bus-reset-gpios is not a property of this I2C controller. IIUC, the I2C controller does not have a line to reset all children. It's the children who have reset lines and it happens it is shared. Just like my WSA884x case: https://lore.kernel.org/alsa-devel/84f9f1c4-0627-4986-8160-b4ab99469b81@xxxxxxxxxx/ Best regards, Krzysztof
