15:03 <phh> I have a weird behaviour on 4.9-rc3, on rk3288. Setting gpio 222 (gpio7-6) to out, then reading gpio222/direction back gives "in". Setting it to "in" reports "out". the GPIO can be written only after I wrote "in" to it.

16:35 <mmind00> phh: in the TRM gpio7 A2-A6 are marked as reserved, so should not be accessed as gpios it seems

16:35 <mmind00> phh: scrap that ... the are only gpios

16:35 <mmind00> (i.e. no other iomux)

16:36 <phh> for a second I thought my chinese IDH were crazy magicians

16:36 <phh> (ok they are anyway)

16:37 <mmind00> phh: but interestingly, no currently in-kernel dts seems to use gpio7_a6 ... does this happen with this pin only?

16:39 <phh> I have also a problem with gpio7_a5, not exactly the same I think, and I think I had it with yet another one

16:39 <phh> but I think I see the problem only when using sysfs-exported gpio

16:40 <phh> right, considering /sys/kernel/debug/gpio I think I have this issue on most I/Os

16:40 <phh> gpio-17 ( |vcc_dvp ) in lo

16:40 <phh> gpio-148 ( |reset ) in hi

16:40 <phh> their name doesn't sound like it should be "in"

16:42 <phh> vcc_dvp is connected to a regulator-fixed, so should definitely be "out", reset-gpios is on a mmc-pwrseq-simple so should be out as well

16:47 <mmind00> phh: not necessarily ... if you look at http://lxr.free-electrons.com/source/drivers/gpio/gpiolib.c#L2398 and the open_drain and open_source calls, there might be other use-cases as well

16:48 <mmind00> phh: but I don't know if we fall into these cases, as I don't remember rockchip boards setting any flags in this regard