12:41 <deathcamel57> How would I go about diagnosing a kernel panic in `/init`? I'm getting `Unhandled fault: imprecise external abort (0x1c06) at 0x00218700` ... `Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000007`. Starting the kernel with `init=/bin/sh` yields the same error (likely because `/init` is linked to busybox, and so is `/bin/sh`.

13:09 <ardb> deathcamel57: 'imprecise' means that the reported address is approximate

13:09 <ardb> and so the issue is not actually in /init but in the kernel

13:10 <ardb> most likely some driver poking into physical address space that is not connected to anything, or the device in question is turned off (i.e., not decoding the accesses)

13:11 <ardb> this particular issue tends to occur also with flaky PCIe blocks that do not deal with error conditions on the PCIe link correctly, and just forward it to the CPU as a fatal exception

13:11 <ardb> which platform is this?

13:12 <deathcamel57> This is a niche SoC, which I'm currently working on porting mainline to.

14:43 <robmur01> yeah, external abort usually means a driver's trying to access something which isn't there, or is powered off. Is it loading modules and probing drivers at that point?

14:43 <robmur01> I'd also try seeing if the "clk_ignore_unused" and "pd_ignore_unused" arguments make any difference

14:45 <deathcamel57> Are those device tree arguments for the nodes?

14:46 <deathcamel57> Nevermind, figured out it's a bootarg

14:46 <robmur01> although if it's happening while the PC is in userspace, it could also potentially be something like describing the RAM layout wrong

14:47 <deathcamel57> I appreciate it so much! I'm getting some other issues now. I'll look into them myself and if I get stuck, I'll swing back around.

14:47 <robmur01> beware vendor BSPs which describe too much memory but then boot with "mem=..." to constrain it to the real amount ;)

14:49 <deathcamel57> I'll keep that in mind. The BSP for this device is Linux 3.0, so I thought I'd get mainline working on it and maybe upstream.

14:53 <Xogium> hoooooly crap that is old

14:54 <deathcamel57> Yep

14:54 <deathcamel57> U-Boot is 2010.06, but I have mainline for that *mostly* working so far

14:55 <Xogium> oof

18:31 <ardb> robher: i suppose you might be as tired as i am of dealing with the payload people, but please chime in if i am not making sense