* Tony Lindgren <tony@xxxxxxxxxxx> [190530 05:47]: > * Tony Lindgren <tony@xxxxxxxxxxx> [190529 08:11]: > > * Tomi Valkeinen <tomi.valkeinen@xxxxxx> [190529 07:06]: > > > On 28/05/2019 13:18, Tony Lindgren wrote: > > > > > > > > My board is x15 rev A3, attached to AM5 EVM. I've also attached my kernel > > > > > config. > > > > > > > > Strange that this is not affecting other x15? I think timer12 would > > > > be blocked on HS devices though? > > > > > > Seems that the kernel config affects. omap2plus_defconfig boots ok. > > > > OK, this line in your oops: > > > > Unable to handle kernel paging request at virtual address 5a5a5a5a > > > > Probably means we hit some slab poison with DEBUG_SLAB set. > > Looks like your config boots fine with DEBUG_SLAB disabled > > for me. > > > > As this only happens for timer12, I wonder if we're again > > hitting some uncompress issue with corrupted dtb. Changing > > u-boot ftdaddr higher up might possibly make it go away. > > Or else there's a bug elsewhere :) > > Oh but CM_WKUPAON_TIMER12_CLKCTRL has no CLKSEL option unlike > CM_WKUPAON_TIMER1_CLKCTRL. Below is one part of the fix, but > it seems like we're missing handling somewhere as trying to > get a non-existing clock should just produce -ENODEV type error. > > And the clksel should be just handled with assigned-clocks > in general, but I think we still need it there until we > have drivers/clocksource/ timer drivers updated to boot > using early_platform_device. OK found it, we have the clkctrl clock potentially return uninitialized data. I posted two fixes for the issue: [PATCH] clk: ti: clkctrl: Fix returning uninitialized data [PATCH] ARM: dts: Drop bogus CLKSEL for timer12 on dra7 Regards, Tony
