02:44 <wens> lol

02:49 <wens> IIRC proprietary tools are still needed to make the u-boot image bootable on meson?

05:49 <afaerber> wens, for S905 I have the meson-tools but for the others they haven't been adapted yet

06:08 <wens> afaerber: reverse engineered I see

06:08 <wens> afaerber: thanks

07:43 <narmstrong> The fip structure and headers have been completely changed for s905x and later, this needing new tools, our only option now it use the tools delivered by Amlogic

07:56 <afaerber> narmstrong, I was still planning to have that as command line option in my tool or at least an ifdef to share code

07:57 <afaerber> but mainly it needs input files and time for the RE effort

07:58 <narmstrong> anyway, it's good to have GXBB support, but this SoC is now legacy and S905X/S905D/S912 deserves some love aswell !

09:38 <Darkmatter66> Hi narmstrong, I've built gstreamer1.0-plugins-good from the topic/v4l2-m2m branch but the resulting libgstvideo4linux2.so doesn't have v4l2video0dec

09:38 <Darkmatter66> narmstrong, only v4l2deviceprovider, v4l2radio, v4l2sink and v4l2src

09:39 <narmstrong> Darkmatter66: from which poky version ?

09:40 <Darkmatter66> narmstrong, sumo, does this branch target a different version ?

09:41 <narmstrong> nop, sumo works

09:42 <Darkmatter66> narmstrong, well when I built from the sumo branch of meta-meson I used to have the v4l2video0dec but the topic branch is messing something

09:42 <Darkmatter66> narmstrong, maybe one of the patches you add in the bbappend is causing this

09:43 <narmstrong> Darkmatter66: should not

09:47 <Darkmatter66> narmstrong, does it work for you on the topic branch ? i cleaned the recipe and rebuilt multiple times and the issue remains

09:48 <narmstrong> Darkmatter66: it worked last time I tried, I need to retry

09:49 <Darkmatter66> narmstrong, sure, take your time

10:21 <mjourdan> AntonioND: that's really cool, nice work

11:59 <brads> How about the M4 in the S905 on C2? illegal instruction errors seem to be the theme on random kernel/ userspace processes when A53 is overclocked above 1.7Gh even with 1 only core enabled. Cooling is not the problem or even overvolt / undervolt nor seem memory freq. Can the M4 be hinted to run a little faster to get the instructions in the right place at the right time? Can its

11:59 <brads> microcode be ammended on boot through the BL blobs?

12:28 <xdarklight> I think it's a Cortex-M3 in GXBB (S905), GXL (S905X, S905D, ...) and GXM (S912)

12:29 <xdarklight> but I don't know anything about it (so sorry for the noise)

12:51 <AntonioND> well, that was something else I was thinking about. when the TF is in a nice shape, usptreamed and working nicely with uboot and the kernel, it would be nice to take a look at bl30.bin and see what's going on in there

12:51 <AntonioND> but I'm sure it's going to be way more obscure than the TF

13:15 <mjourdan> AntonioND: is bl31 acting as secure monitor on gxbb ?

13:16 <AntonioND> yes

13:17 <AntonioND> https://www.cnx-software.com/wp-content/uploads/2016/10/ARM-Trusted-Firmware-Architecture.jpg

13:17 <AntonioND> they are using a weirdly modified bootflow

13:17 <AntonioND> bl30 is the scp binary

13:17 <AntonioND> bl301 is non standard, but I think that it is actually open

13:18 <AntonioND> bl31 is the runtime part of the firmware

13:19 <AntonioND> I'm actually not sure of how many services are needed from the TF

13:19 <AntonioND> PSCI for sure

13:20 <AntonioND> but, for example, uboot uses some system monitor calls to get the serial number to generate the mac address using it

13:20 <AntonioND> this is obviously not critical

13:20 <AntonioND> but nice to have

13:21 <mjourdan> Thanks :)

13:22 <AntonioND> np

19:57 <AntonioND> can turn on and off cpus from linux! :P

19:57 <AntonioND> echo 0 > /sys/devices/system/cpu/cpu3/online

19:57 <AntonioND> etc

19:57 <AntonioND> lscpu lists them nicely

19:57 <vagrantc> yeah :)

20:00 <AntonioND> meh, cpu0 doesn't want to be nice

20:01 <AntonioND> the psci interfaces in the firmware aren't exactly intuitive

20:11 <AntonioND> ok, so I can turn on/off at will, except for cpu0, which can only be turned off, and if I try to turn it on it fails and then I get an endless loop of "can't change cpu frequency" messages lol

20:11 <AntonioND> it will eventually work

20:27 <xdarklight> not sure about ARMv8 but with older 32-bit cores you can't turn off CPU 0 because some interrupts are tied to it

20:28 <AntonioND> i'm not sure if linux understands that, though

20:28 <AntonioND> it hasn't blocked me from trying

20:28 <AntonioND> and it has actually succeeded

20:29 <AntonioND> so cpu0 isn't critical

20:29 <AntonioND> the problem is that the psci implementation I have in assembly code is a really old one that I have to reverse engineer + migrate to the new psci interfaces

20:29 <AntonioND> and this is a bit crazy

20:29 <AntonioND> the arguments of the hooks are different, etc

20:35 <narmstrong> You can turn off/on the cpu0 on amlogic implementation, at least the latest one...

20:45 <AntonioND> I can actually do it, but I think that when I try to power it on again it fails, but it doesn't fail to setup the gic

20:45 <AntonioND> so that's why I get error messages probably

20:50 <AntonioND> at least I don't have to worry about cluster on/off states

21:10 <AntonioND> lol

21:10 <AntonioND> okay

21:10 <AntonioND> with the original TF that command turns off the full board

21:10 <AntonioND> that makes me feel a lot better about my problems :P

21:11 <AntonioND> echo 0 > /sys/devices/system/cpu/cpu0/online

21:11 <AntonioND> this, I mean

21:12 <AntonioND> the other cpus work fine

21:13 <AntonioND> but cpu 0 turns off the full system

21:16 <AntonioND> i mean, even if the other cpus are on

21:17 <AntonioND> hmmm I guess they couldn't get it to work either. there is a workaround, I could just handle cpu0 differently and keep it in a wfi loop instead of trying to actually turn it off

21:24 <AntonioND> i think that their scp code may have a bug and that's why this happens... interesting

22:13 <AntonioND> ok, so I've implemented a hacky way to have cpu off in CPU0 for consistency

22:13 <AntonioND> when entering off state I enter a wfe loop

22:13 <AntonioND> and I wait for a flag to be set to 1

22:16 <AntonioND> but well, linux really doesn't like this. it shows messages that it tries to change the cpu freq and it fails, so well, I guess this workaround is really not great