00:52 <steev> shawnguo: i tested the 3 uaf qce patches but they don't help with the crash when the crypto manager tests are enabled :(

00:56 <shawnguo> yeah, that bug is quite tricky, I do not understand how it happens exactly.

02:05 <steev> it's unfortunate because i'd probably get a LOT of pushback trying to disable that option in debian

03:17 <bamse> steev: what's the qce bug?

03:17 <steev> bamse: with # CONFIG_CRYPTO_MANAGER_DISABLE_TESTS is not set

03:18 <steev> you get a kernel panic when it probes the crypto stuff

03:18 <steev> CONFIG_CRYPTO_MANAGER_DISABLE_TESTS=y

03:18 <steev> is required to boot, basically

03:20 <steev> nearly every distro runs with CRYPTO_MANAGER_DISABLE_TESTS is not set

03:23 <bamse> steev: so if i just disable DISABLE_TESTS my device will crash on boot?

03:23 <steev> yes

03:24 <steev> i also really hate that naming because talking about it makes it seem like a double negative

03:25 <bamse> apparently didn't have QCE enabled...testing again

03:25 <steev> iirc shawn and thara are somewhat looking into it (i don't wanna put words in their mouths or commit them to being wworking on it)

03:26 <bamse> i don't see that there's anything c630 specific there, so i'm giving it a spin on the db845c...

03:26 <steev> tho

03:26 <steev> https://bpa.st/KYNA

03:26 <steev> maybe if we had the devices it wouldn't ?

03:27 <bamse> i did get [ 6.526400] WARNING: CPU: 3 PID: 275 at crypto/testmgr.c:5681 alg_test.part.0+0x12c/0x430

03:27 <bamse> but it didn't crash

03:27 <steev> oh they're incompatible

03:28 <steev> we get crashes here, shawn and i

03:28 <steev> on c630

03:28 <bamse> a panic? or just a reboot

03:28 <steev> panic

03:28 <bamse> do you have a pastbin of the log?

03:29 <steev> i do not because it requires a reboot

03:29 <steev> i can rebuild and get a picture :D

03:30 <steev> oh i still have it, lemme install and reboot

03:33 <steev> https://usercontent.irccloud-cdn.com/file/5gPkzO6K/IMG_5060.JPG

03:36 <steev> maybe disabling just skcipher would work?

03:36 <bamse> ohh, how lovely

03:36 <steev> i don't think i've actually tried that

03:37 <bamse> i think somehow you end up trying to play qce with out clocks..or perhaps interconnect votes

03:37 <bamse> send me a patch to status = "disable" the crypto node in the c630.dts for now

03:37 <steev> i can do that

03:38 <bamse> with a "FIXME: qce_start triggers an SError" comment

03:38 <bamse> thanks

03:38 <bamse> then we need to figure out why it does that on the c630...but not on db845c or the mtp

03:38 <bamse> but let's make sure it boots first

03:39 <steev> hm

03:44 <steev> since we just inherit crypto from the dtsi, i can just /delete-node/ yeah?

03:44 <bamse> nah, do &crypto { /* FIXME: foobar */ status = "disable"; }

03:45 <steev> okay that's what i thought, but was hoping for the former :D

03:45 <bamse> it's more explicit than the delete-node

03:49 <bamse> it might be good, as you suggested...to disable skcipher and see that it still crashes

03:50 <bamse> but feel free to do that when you have time

03:56 <steev> we can either have all, or we can only have one of the 3

03:58 <bamse> right, but looking at qce_start()...it's just 3 functions that probably gets inlines...and one of those are skcipher

03:58 <bamse> most likely the whole thing is dead...but in theory it could be just the skcipher function

03:58 <bamse> but looking more at it makes me feel that's quite unlikely

04:05 <steev> i can test, it's not that difficult :)

04:09 <steev> yeah, happens with hash too

04:16 <steev> well i get the blue screen with my patch applied and # CONFIG_CRYPTO_MANAGER_DISABLE_TESTS is not set... so that's... good? lol

04:23 <steev> robclark: with https://patchwork.kernel.org/project/linux-arm-msm/patch/20211104222840.781314-1-robdclark@gmail.com/ applied, dmesg is starting to get "a few" (depends on how long the system runs i suppose) devfreq 5000000.gpu: Couldn't update frequency transition information. - is that intended?

04:52 <robclark> nope

20:36 <steev> i think, this weekend, i'm gonna try to go over the audio stuff and see if i can figure out wtf all is going on there

20:36 <steev> i want audio working again before my trip