00:21 <steev> bamse: i see people doing reviewed by at the top of mails, under the signed off - would it be better for me to put my T-b there instead of at the bottom under the patch?

00:21 <steev> in thinking about it, it does seem like it would be faster, and less time wasted

02:22 <bamse> steev: it doesn't matter, as long as you reply inline somewhere

03:03 <leezu_> robclark: on lazor with Chromium 97 and Wayland, the Conformance Test Runner gives Passed: 10200/27092 Timeout: 691/27092 in 15049.81 seconds

03:03 <steev> https://www.phoronix.com/news/Linux-5.20-SoCs-8cx-Gen3-Arm no pressure bamse :D

03:04 <_> Is that SoC actually powerful?

03:04 <steev> it's okay?

03:06 <_> :\

03:06 <_> Does the X13s have SODIMM RAM slots?

03:06 <steev> no

03:07 <_> Do any qualcomm SoCs even support ECC UDIMMs?

03:07 <steev> probably not

03:09 <bamse> _: i compile my kernels on another machine, but other than that it's powerful enough

03:09 <bamse> _: and you get 20+ hours battery life and 5g connectivity

03:09 <bamse> well, right now we're at ~10 hours, but lenovo claims up to 28 hours...

03:13 <steev> 28 hours of video playback, and screen at ~50%

03:14 <bamse> yeah, obviously the real test is how long you can hold an irssi session alive in alacritty over lte/5g...

03:30 <davidebeatrici[m]> Hello!

03:32 <davidebeatrici[m]> I have a PiPO W12 and I'm willing to test Linux on it. Is there anybody else who owns the device?

03:42 <steev> not i, but it seems like it would be similar to the galaxybook 2

03:47 <davidebeatrici[m]> You mean the Galaxy Book S?

03:50 <gwolf> steev: Huh? what should make me happy? (your line at 00:10:21) :-] )

03:51 <gwolf> FWIW, I usually get ~10hr of effective battery use with the C630. And that is enough for me to be quite happy!

03:51 <gwolf> (still haven't upgraded from 5.16)

04:01 <steev> ah, 5.19 should be better battery usage than that

04:06 <steev> davidebeatrici[m]: no, i mean the galaxybook2 which is pretty much identical to that pipo w12

04:06 <davidebeatrici[m]> Isn't the Galaxy Book 2 x86?

04:06 <steev> no

04:07 <davidebeatrici[m]> Ah, sorry, it's the Pro that is x86. Samsung naming as usual.

04:07 <steev> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/qcom/sdm850-samsung-w737.dts

04:08 <steev> yeah... samsung naming is... almost as good as lenovo ;)

04:08 <davidebeatrici[m]> Yeah...

04:08 <steev> lenovo has the "yoga c630" and the "yoga chromebook c630"

04:09 <steev> but that dts i linked above should be quite similar to the device you have

04:11 <davidebeatrici[m]> Very similar, except for the fact that the PiPO W12 costed less than half the price of the Galaxy Book2.

04:11 <davidebeatrici[m]> With 256 GB of eMMC and 8 GB of RAM.

04:12 <steev> what i'm saying is, you can use that dts as a base, not that the hardware is "close enough"

04:12 <steev> or more as a guide

04:14 <steev> i have the galaxybook2, the yoga c630, the flex 5g, and the thinkpad x13s, and of them, the gb2 is likely the closest to get you to booted

04:15 <davidebeatrici[m]> I assume I should boot with that DTS and eventually change it to fix stuff that is not working, right?

04:25 <steev> essentially, yeah

09:35 <jhovold> steev: wasn't loading the dsp fw before as I mentioned earlier. Tried doing so now and ran into the remoteproc regression that mani_s apparently had also fixed yesterday. Seems to work now.

14:30 <steev> jhovold: yep :D

19:00 <clover[m]> wow i think "LENOVO - System Hardware Update - 7/25/2022" windows update just solved my BIOS issue

19:02 <steev> that's good!

19:02 <clover[m]> yep, no longer seeing the TCO logo

19:03 <steev> there's another update from a few days ago that updates MDM thermal

19:03 <steev> whatever that is

19:03 <clover[m]> yeah got that one

19:03 <steev> i'm guessing that one doesn't let it get as hot

19:04 <clover[m]> now that you mention it i haven't noticed it getting as hot lately

19:06 <AlexMarty[m]> Should I get another ssd and install windows again to get these firmware updates?

19:07 <steev> the bios one is definitely downloadable from their site

19:07 <steev> i don't know about the others

19:07 <clover[m]> i think Leo was able to do some updates from linux

19:07 <steev> or if the others even matter

19:08 <AlexMarty[m]> Ok I’ll try and update soon

19:08 <clover[m]> temperature controls probably help the overall longevity of the chip, no?

19:08 <clover[m]> i'd say that's pretty important

19:08 <steev> if they're windows specific, they won't help us in linux

19:08 <clover[m]> i see

19:09 <steev> unless they write something to an eeprom or some such

19:09 <AlexMarty[m]> I guess I’ll find out.

19:10 <steev> i really need to go through all my stuff when 5.19 releases

19:10 <steev> currently the thinkpad stuff is in a different directory and such

19:10 <steev> i want to move all the c630/flex/thinkpad stuff together... i also wanna try an install on the galaxybook2

19:17 <steev> the 5.20/next stuff is gonna be incompatible with what we've currently got

19:18 <clover[m]> i'll probably need to change the dtb name for my ISO, right?

19:18 <steev> yeah, currently we're using sc8280xp-thinkpad.dtb, upstream it'll properly be sc8280xp-lenovo-thinkpad-x13s.dtb

19:22 <steev> but need to see what made it in, and what didn't, once the patches roll in to torvalds after release

19:32 <leezu_> Unfortunately 5.19 does not resolve all hanging issues on the lazor / sc7180. The screen still freezes up completely sometimes. Connecting to the serial console is still possible and one can see some dmesg messages (eg related to the charging state), but the console is also very slow and it takes a minute for characters typed to actually be acknowledged / show up. Have you

19:32 <leezu_> seen similar issues or have hypotheses for what's wrong? I've observed this on two separate devices, so it's very unlikely to be a hardware issue

19:41 <leezu> Supposedly "Alt + Volume-Up is treated as the Magic SysRq key", but when I pressed "Alt + Volume-Up + R" (as part of the REISUB SysRq sequence), the system immediately restarted. I see that "Alt + Volume-Up + R" on Chromebook also refers to "Warm AP reset". Is that supposed to restart the system?

19:48 <steev> no ideas here, i hate the chromebook keyboard with a passion

19:56 <amstan> ap means the cpu, so yes

19:57 <amstan> that's unfortunate that they clash

19:57 <amstan> if you feel adventurous i could give you pointers to what pieces of code in the ec to disable and you could flash yourself a modified copy

20:20 <robclark> leezu: hmm, I'm still on v5.19-rc5 (plus drm things for v5.20 and `arm64: dts: qcom: Remove duplicate sc7180-trogdor include on lazor/homestar`).. but haven't had any such issues.. pastebin dmesg and I'll look at it in a bit (just heading out to store).. dianders and sboyd are probably also running upstream kernel on sc7180 chromebook things regularly

20:21 <robclark> (I'm running on lazor, fwiw)

20:42 <leezu> robclark: The weird thing is that there is no dmesg output when the problem occurs. At least that was the case on 5.17.15 (where I kept an intel laptop connected for many hours via SuzyQ to collect dmesg upon the hang). I've reconnected the SuzyQ now and will post here once the issue happens again. If connecting the SuzyQ after the problem happened, while I can type some

20:42 <leezu> letters on the keyboard and they'll show up in the console after a minute, the actual terminal "login:" prompt never shows up. So debugging this issue requires being connected on SuzyQ in a root shell prior to the occurance..

20:49 <leezu> amstan: I think my kernel isn't configured correctly for "Alt + Volume-Up is treated as the Magic SysRq key" either. I'm able to trigger it via "echo t > /proc/sysrq-trigger" though

21:23 <leezu> But I'm unable to send SysRq via serial console, despite CONFIG_MAGIC_SYSRQ=y CONFIG_MAGIC_SYSRQ_SERIAL=y. My expectation was that triggering BREAK via Ctrl+T Ctrl+B from pySerial miniterm and then typing a SysRq command key such as `l` for "Shows a stack backtrace for all active CPUs" would work..

21:43 <robclark> leezu: weird, so this has been happening a while? Check `free`, maybe something is leaking memory and you are hitting swap hard or something like that? If not that, build kernel w/ lockstat and use that to see if you are hitting some lock contention issue

21:43 <robclark> https://docs.kernel.org/locking/lockstat.html

21:54 <leezu> Hm, I hit the issue again and now even `free` isn't executed when triggered via SuzyQ. Memory leak is possible. Actually the system runs without swap, so I'd expect the kernel to kill processes before freezing up, but maybe something isn't working as expected

21:56 <leezu> I'll try to repro again while running a `watch free` over serial

22:36 <leezu> Ok, so at the time of hang we have Memory (in MB) total 4065 used 2976 free 134 shared 782 buff/cache 955 available 74

22:37 <leezu> Most of the memory should be used by chromium tabs. I don't know why the OOM handler does not kill any processes prior to the hang

22:39 <leezu> My understanding is that some of the memory is used by the GPU / DPU? I tried disabling memory overcommitting and noticed that starting a desktop session then would fail with "msm_dpu ae01000.mdp: [drm:get_pages] *ERROR* could not get pages: -28" so I wonder if that indicates any bug?

23:26 <steev> 5.19 is out

23:26 <HdkR> Party time

23:34 <robclark> leezu: hmm, I'd have expected oom to kill some things (although maybe not the things you'd want it to).. tbh I always use zram swap (which is similar to the CrOS setup) although not sure offhand why oom wouldn't kick in without swap (but I do recommend zram swap regardless)

23:35 <steev> clover[m]: i'm gonna rip the band-aid off for the 5.19 release, and moved the dts to the same name that 5.20 will have

23:37 <steev> it'll be easier to pull in things that might be needed to 5.19 point releases for those who don't like to live in -next

23:41 <davidebeatrici[m]> I will attempt to run OpenMandriva on the PiPO W12. For reference: https://github.com/OpenMandrivaSoftware/os-image-builder

23:45 <leezu> robclark: zram sounds pretty useful. Do you know how much zram cros would allocate on a system with 4gb ram? For now I have `apt install earlyoom` which should kill processes whenever system memory availability goes below 10%

23:48 <steev> i usually go 100% myself with zram

23:50 <robclark> I think we are currently going w/ 150% of ram on CrOS but want to raise it to 2x (although most of the benefit there comes with addition of MGLRU.. at least AFAIU)

23:53 <robclark> (well, combination of MGLRU + 2x ram zram swap plus msm.enable_eviction=1 is the current ideal setup.. enable_eviction still defaults to off upstream but plan to switch the upstream default after we get a bit more hammering on that on CrOS)

23:57 <steev> what is mglru?

23:58 <robclark> https://lwn.net/Articles/880393/