14:40 <plntyk> hello, does somebody have any info on odroid-c1 / meson8b situation ? i filed two bugs/regressions in mainline kernel but its strange that those are *still* happening while there is continued development on that SoC - is it just some bug with different bootloader, kernel config and/or distro ?

14:41 <plntyk> bugzilla 205787 and 207481

16:18 <jbrunet> What bugzilla are you referring to ? This channel is about linux mainline kernel dev and we don't have a bugzilla.

16:26 <plntyk> https://bugzilla.kernel.org/show_bug.cgi?id=205787 that kernel.org bugzilla

16:28 <plntyk> https://bugzilla.kernel.org/show_bug.cgi?id=207481

17:17 <xdarklight> plntyk: RE https://bugzilla.kernel.org/show_bug.cgi?id=207481 - what do you have for CONFIG_MESON_MX_EFUSE in your .config?

17:20 <xdarklight> plntyk: RE https://bugzilla.kernel.org/show_bug.cgi?id=205787 - I don't know about this one (never seen it). in my .config I have: CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND=y

17:23 <xdarklight> I've never used this kernel bugtracker before, so any issue there is "under my radar"

17:24 <plntyk> xdarklight: ah ok CONFIG_MESON_MX_EFUSE is not set on the nonworking ones - a newer nightly with 5.4.31 does seem to set it but I havent tested that yet because I broke one SD card - as i did some kernel compilation on that

17:24 <plntyk> *didnt test 5.4.31 yet*

17:25 <xdarklight> plntyk: you need the eFuse driver since https://github.com/torvalds/linux/commit/6b14dd7267126931e9a95c68a442e8f2dabdc3c4

17:25 <xdarklight> (that's v5.4-rc1)

17:25 <plntyk> i have CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND=y on those with rcu stalls

17:25 <xdarklight> arch/arm/configs/multi_v7_defconfig also enables the eFuse driver, so please make sure you stay in sync with that

17:28 <xdarklight> plntyk: do you get those RCU stalls only when putting traffic through that USB network adapter? I use USB only from time to time so not sure about that. Either way: I'm working on fixing Ethernet on Odroid-C1 (so you get as close to Gbit/s speeds with the built-in Ethernet controller)

17:28 <plntyk> the rcu stalls are with usb ethernet - if i set performance the error is gone

17:30 <plntyk> however there might be some side effect on a defconfig so i will try with a newer nightly and new compile - now i managed to use a usb rootfs and boot kernel only from sd-card since its a "buggy" sd-card that isnt working in that c1 permanently due to "hw issues"/bugs as stated on hardkernel side

17:44 <xdarklight> OK, great

17:44 <xdarklight> if you have any link regarding these "hw issues" then please share it, I don't remember hearing about that so far

18:15 <plntyk> xdarklight: https://wiki.odroid.com/odroid-c1/troubleshooting/microsd_selection - a Transcend 32GB microsdxc seems to have an issue if used at rootfs - currently trying to reflash and test with new nightly but now armbian server seems down

18:15 <plntyk> and at least one new nightly buster image does not have a uImage in /boot

18:16 <plntyk> damn maybe time to go to bed or watch tv

19:28 <khilman> chewitt: I assume you've been testing the GX audio patches from jbrunet: will you be giving a tested-by on list? https://lore.kernel.org/linux-amlogic/20200421163935.775935-1-jbrunet@baylibre.com/

20:48 <hexdump0815> xdarklight: i think i saw those rcu stalls too on heavy usb use on my m8s and/or mxq with around 5.4 (did not test anything newer yet)

20:49 <hexdump0815> xdarklight: and also the fact that the performance governor fixed it sounds familiar

20:53 <hexdump0815> xdarklight: another strange thing is that my mxq only runs stable with "threadirqs" kernel option - i have to retest this one and send you the details if you are interested