09:50 <jannau> noisycoil[mds]: no, my initial hunch did not fix anything. havenm't tried anything else yet

10:46 <cy8aer> mps: is it really freezing? I have the effect on my m1pro: press suspend button, press it again, nothing happens (so like frozen). But: then close the lid and open it again: Everything is fine (min since 6.14...)

11:05 <cy8aer> It is somehow like the display is not switched on.

11:19 <mps> cy8aer: yes, it really freeze on j314s and j293

11:20 <mps> my suspend script have this line 'echo mem > /sys/power/state'

11:21 <mps> this started with 6.15 kernel, before it i did not had it ever

11:23 <mps> i thought at first it is related to dm-crypt but now i see this even when not using dm-crypt

11:24 <mps> yesterday I've rebuilt kernel changing some config options and now testing to see will it happen again

11:31 <cy8aer> Hm but then I need to ask: Does anyone have seen my behavior?

12:02 <jannau> noisycoil[mds]: sigh, I did not read. the issue is in drivers/video/fbdev/simplefb.c and not in simpledrm. That explains why there are no issues on fedora. Any idea why debian doesn't use simpledrm

12:07 <noisycoil[mds]> No I don't, I need to ask the kernel team

12:08 <noisycoil[mds]> Any idea why it fails on asahi/M2 only?

12:08 <jannau> the code is very similar there and I don't see an obvious issue there yet

12:09 <jannau> I expect it to fail on any device with 2 or more power-domains in the framebuffer node. Only m2 devices have that

12:13 <noisycoil[mds]> All M2s?

12:23 <jannau> not sure about the air and 13-inch pro. the 14"/16" mbp with the DT sent upstream, downstream DT might still have a always-on and skips the second power-domain

12:37 <jannau> yes, it's only the M2* desktop devices which use 2 power domains

12:38 <noisycoil[mds]> Would you rather I switch to simpledrm downstream?

12:41 <jannau> no, debian should switch unless there's a reason why it can't. I'll fix the issue in any case

12:43 <jannau> still eludes me why it's broken. looks like something corrupts par->genpd_links

12:48 <noisycoil[mds]> Ack. I'm afk today unfortunately

14:20 <jannau> I think I found the cause. completely unrelated to simplefb/simpledrm

14:32 <noisycoil[mds]> Is it?

14:33 <noisycoil[mds]> What was that?

14:49 <jannau> see #asahi-dev, not fully sure though because I'm just fixing an oob read and not store

14:51 <jannau> no, that's not the issue. it was just hiding it from kasan

14:59 <noisycoil[mds]> Saw it

15:21 <jannau> do you and/or the user care for "Reported-by:" tags

15:23 <noisycoil[mds]> Eh, not a lot, but the user has been doing a lot of testing for us and it could be nice to acknowledge him

15:27 <jannau> noisycoil[mds]: https://paste.centos.org/view/cd2896f0 is the fix. I'll amend Reported-by: before sending it out. I'll include the change in 6.16.4 (expected to be tagged later today)

15:28 <noisycoil[mds]> Daniel Huhardeaux is the name

15:28 <noisycoil[mds]> Thank you

15:29 <jannau> no idea why this triggers consistently with 6.16.3 but hasn't before

15:30 <noisycoil[mds]> Saw the fix, wonderful. And thanks a lot for going all the way and install Debian (or whatever you did) to fix this, it is truly appreciated

15:36 <jannau> relevant debian kernel PR for simpledrm https://salsa.debian.org/kernel-team/linux/-/merge_requests/1453

15:37 <noisycoil[mds]> Thanks, as I said I'm mostly afk today and could not check