07:18 <ldevulder> hi! I tried the last s905x dts(i) from Linux-4.19rc4 for u-boot and I found an issue with this commit: https://github.com/torvalds/linux/commit/6f95c8cd76c19e1fb9b87e8c176b8df975fb14a4#diff-21be33a3ffb6232113847fffb8387099

07:19 <ldevulder> without the 'reg = <0x0 0x0 0x0 0x3db>;' line u-boot stuck at boot...

07:39 <Amit_T> narmstorng: Hello, Good Morning. One query, is LEGACY_SCPI_CMD_SYS_PWR_STATE equivalent of SCPI_CMD_SET_SYS_PWR_STATE ?

09:11 <ldevulder> narmstrong, I'm not sure, but the USB issue on kvim2 could be related to a clock issue, I have this message when I increase log level with 'usb start':

09:11 <ldevulder> clk_set_defaults() clk_set_defaults(reset-controller@4404)

09:11 <ldevulder> clk_set_default_parents() clk_set_default_parents: could not read assigned-clock-parents for 00000000bdf45490

09:12 <narmstrong> for 4.19, the clock probing must be changed, the DT has chanaged

09:12 <narmstrong> for kivm2, it may be missing a clock only available on S912

09:16 <ldevulder> narmstrong, ok, I will check later for the clock probing/DT how to (try) to do this, and yes it seems for kvim2, now I need to understand how to add it! :D

09:17 <narmstrong> ldevulder: you will need to switch the clock register acess to regmap and change the probe like this :

09:17 <narmstrong> priv->map = syscon_node_to_regmap(dev_get_parent(dev)->node);

09:17 <narmstrong> return PTR_ERR(priv->map);

09:17 <narmstrong> if (IS_ERR(priv->map))

09:19 <narmstrong> ldevulder: looking at the DT, no new clocks are used... only a third PHY, but my code should handle 2 or 3 PHYs

09:21 <ldevulder> narmstrong, so this clock error should also be visible on s905x as well, I will check later on kvim or libretech-cc

09:22 <narmstrong> ldevulder: yes

09:23 <ldevulder> narmstrong, on the kvim2 thie issue is that no PHY0 are found for dwc3@c9000000, so it's like somthing is missing in the DT, I'll continue my investigation :)

09:24 <narmstrong> ldevulder: ok, keep me in touch !

09:24 <ldevulder> narmstrong, sure! thx

10:44 <mjourdan> ndufresne: there are a lot of debug registers but mostly undocumented :<

10:45 <mjourdan> ndufresne: quick gstreamer question, do you have a plugin that can render dmabuf buffers by importing them into egl images via EGL_LINUX_DMA_BUF_EXT ?

13:23 <ndufresne> mjourdan, yes, glupload does that, if you are on wayland, automatically, if you are on X11, you have to switch from GLX to EGL, because this is not supported by GLX, GST_GL_PLATFORM=egl iirc

13:24 <ndufresne> oh, and if you are on GBM, it also works

13:25 <ndufresne> mjourdan, though, with the Mali blob for 4XX series, you might needed something like https://bugzilla.gnome.org/show_bug.cgi?id=783521

13:26 <ndufresne> as we required R8 and RG88 texture importation support in the generic case that we implement, works on Mali 6XX and 8XX

13:27 <ndufresne> and it's the blob the limitation, not the GPU

13:30 <mjourdan> ndufresne: good to know, thanks

13:30 <mjourdan> I know in the case of the mali 450, importing a nv12 dmabuf buffer works great

13:32 <ndufresne> yes, there is only 2 importation format in the 4XX blob, NV12 and BGRA (called ARGB in DRM)

13:33 <ndufresne> mjourdan, the "direct" dmabuf upload code is exactly that, instead of breaking down NV12 into R8 (Y plane) and RG88 (UV plane), it will give both plane for a single EGLImage

13:33 <ndufresne> There is only two GPU have have such a support, the Mali 4XX, and the Adreno (with Freedreno driver)

13:34 <ndufresne> and Etnaviv too, but Mesa calls into another IP, it's not the GPU

13:34 <mjourdan> Interesting, so what you said about the 6XX and 8XX is that they no longer support importing something else than rgb ?

13:35 <ndufresne> they do, but it's stupid to use that, because internally they will split it up in R8 and RG88 and run shaders on it

13:35 <ndufresne> it's better to use our own shader, for which we have control over

13:35 <mjourdan> well sure, but isn't it easier for userspace to just import nv12 and not worry about conversions ?

13:35 <ndufresne> we also support more color formats in our shaders

13:36 <ndufresne> Well, easy is not important, the question is does it work on Intel and Nvidia

13:36 <ndufresne> the embedded stuff came afterward

13:36 <ndufresne> e.g. R8 and RG88 works just fine on Freedreno

13:37 <mjourdan> kk

13:38 <ndufresne> anyway, we are 1 bug away from merging the patches in 783521

13:41 <ndufresne> mjourdan, just a reminder, Endless workaround was to use the blitter to convert to something for which gstreamer and mali blob will agree ;-D

13:42 <mjourdan> Yeh they supported both V4L2_PIX_FMT_ABGR32 and V4L2_PIX_FMT_NV12M

13:42 <mjourdan> But since we don't have a mainline driver for ge2d (the blitter) yet, we're limited to NV12 for now :P .

18:13 <AntonioND> well, after quite some time of reverse engineering of the BL31 provided with the odroid C2, I've managed to write an open source BL31 get uboot to hang while looking for a kernel image in my sd card :P

18:13 <AntonioND> system_off and system_reset work

18:15 <AntonioND> and before trying to get psci_cpu_on working, I need to get linux to actually boot...

18:17 <AntonioND> is there any place with instructions on how to set up the SD card for this? I've just flashed an image that hardkernel provides and replaced the firmware and uboot, so I'm not sure if that will work