TheSeven has quit [Ping timeout: 250 seconds]
TheSeven has joined #linux-exynos
ansiwon has joined #linux-exynos
amitk has joined #linux-exynos
krzk has joined #linux-exynos
wwilly has joined #linux-exynos
libv_ has joined #linux-exynos
libv has quit [Ping timeout: 276 seconds]
libv has joined #linux-exynos
libv_ has quit [Ping timeout: 276 seconds]
libv_ has joined #linux-exynos
libv has quit [Ping timeout: 244 seconds]
libv_ has quit [Ping timeout: 244 seconds]
libv has joined #linux-exynos
Vasco_O is now known as Vasco
Vasco is now known as Vasco_O
libv has quit [Ping timeout: 260 seconds]
paulk-collins has joined #linux-exynos
libv has joined #linux-exynos
amitk has quit [Quit: leaving]
libv_ has joined #linux-exynos
libv has quit [Ping timeout: 240 seconds]
libv has joined #linux-exynos
libv_ has quit [Ping timeout: 244 seconds]
krzk has quit [Quit: Ex-Chat]
masta has quit [Quit: Leaving]
masta has joined #linux-exynos
pekka10 has joined #linux-exynos
libv_ has joined #linux-exynos
libv has quit [Ping timeout: 246 seconds]
libv has joined #linux-exynos
libv_ has quit [Ping timeout: 252 seconds]
libv_ has joined #linux-exynos
libv has quit [Ping timeout: 264 seconds]
libv_ has quit [Ping timeout: 264 seconds]
libv has joined #linux-exynos
libv has quit [Ping timeout: 272 seconds]
libv has joined #linux-exynos
LiquidAcid has joined #linux-exynos
cosm_ has joined #linux-exynos
<
LiquidAcid>
Wizzup, so how did the testing go?
<
Wizzup>
didn't get to it yet
<
Wizzup>
LiquidAcid: let me have a look now.
<
Wizzup>
Is your kernel based on 4.5? Seems like it?
<
LiquidAcid>
the kernel in the public repo is based on 4.5.5
<
Wizzup>
OK, I am currently on 4.5 vanilla, so it should be easy to port over my .config
<
Wizzup>
I'll try to figure out what the best way is to replace my normal libdrm with yours
<
LiquidAcid>
i just put it into $HOME/local/lib and use it from there
<
LiquidAcid>
overwriting any kind of system libs while development is usually a very bad idea :)
<
Wizzup>
I'm cloning the kernel now. Will do a oldconfig
<
Wizzup>
and ensure that g2d is enabled
<
Wizzup>
do you have a xorg.conf for armsoc?
<
Wizzup>
(since I'm on modesetting atm)
<
Wizzup>
(also - going to test on peach pi)
<
LiquidAcid>
i think i have one, but it's not online (and it also just selects 1280x1024 res instead of the native one)
Vasco_O is now known as Vasco
<
Wizzup>
I guess that the armsoc build needs the right libdrm headers?
<
LiquidAcid>
Wizzup, yes, like i said i put everything in $HOME/local
<
Wizzup>
OK, so I will build the kernel and libdrm, install libdrm to $HOME/local, and then build armsoc and ensure it checks $HOME/local
<
Wizzup>
It's good that I get it right so I can also hack on this in the future)
<
LiquidAcid>
sounds like a plan
<
Wizzup>
(working on it atm)
isaque has joined #linux-exynos
<
Wizzup>
almost done building the kernel.
<
Wizzup>
An encrypted rootfs isn't always the easiest when testing
<
LiquidAcid>
Wizzup, i also once thought about this, but i'm only encryting /home for now
<
Wizzup>
also a 16MB /boot was another mistake
<
Wizzup>
can only fit two uImages on there
<
Wizzup>
configured like this: ./configure --with-drmmode=exynos --prefix=/home/merlijn/local/ --with-sysroot=/home/merlijn/local/
<
Wizzup>
libdrm compiled with, configured with exynos and libkms, installed in ~/local/
<
Wizzup>
s/compiled with/compiled fine/
<
LiquidAcid>
Wizzup, that doesn't change the directories the compiler searches for includes
<
LiquidAcid>
Wizzup, you need to change you cflags/ldflags
<
Wizzup>
to have -I$HOME/local/include in there?
<
Wizzup>
OK going to reboot in your kernel
<
Wizzup>
kernel booted, now trying to make X use the right driver.
* Wizzup
wonders why this has to be so painful
<
LiquidAcid>
Wizzup, so you checked the libdrm tests already?
<
Wizzup>
No, I'll do that first.
<
Wizzup>
p.s. plugging in hdmi gives a kernel panic
<
Wizzup>
(just a side note)
<
javier__>
Wizzup: do you have a crash log of that kernel panic?
<
Wizzup>
javier__: it happens on his kernel, but on 4.5 vanilla
<
LiquidAcid>
Wizzup, hmm, that doesn't sound right
<
Wizzup>
but not on*
<
javier__>
Wizzup: ah, Ok
<
Wizzup>
LiquidAcid: I hope you don't mind me asking - I'm digging through the logs - how do I run libdrm tests
<
Wizzup>
javier__: hdmi has been working pretty good for me on vanilla :)
<
Wizzup>
javier__: even seperate screens works
<
LiquidAcid>
Wizzup, the exynos tests are all in tests/exynos
<
javier__>
Wizzup: yeah, same here and that's why I was curious about your panic
<
javier__>
missed that it wasn't on mainline, sorry about that
<
Wizzup>
LiquidAcid: do you want me to just run them, or do you have suggested params and outcomes/results
<
Wizzup>
javier__: np, thanks for being ever so helpful
<
LiquidAcid>
although i shouldn't have a lot of patches in that tree that affect hdmi
<
Wizzup>
odroid might
<
LiquidAcid>
Wizzup, you could run the perf test first i guess, since it doesn't need a display
<
Wizzup>
$ ./exynos_fimg2d_perf -i 10
<
Wizzup>
exynos/fimg2d: failed to get version.
<
Wizzup>
error: failed to init G2D
<
Wizzup>
nothing in dmesg
<
LiquidAcid>
Wizzup, have you enabled drm debug output?
<
Wizzup>
Likely not.
<
Wizzup>
Here's to hoping I can do it at runtime. ;-)
<
LiquidAcid>
yeah, should work
<
Wizzup>
there is a bit more. hang on.
<
LiquidAcid>
i don't see the get_ver2 ioctl there, so that's probably not from the perf test run
<
Wizzup>
This is the relevant part
<
Wizzup>
I hope/believe
<
LiquidAcid>
ret = -13, that's EACCESS
<
Wizzup>
let me retry
<
LiquidAcid>
you sure you have control over the device
<
LiquidAcid>
like full control (no xserver or so running?)
<
Wizzup>
no, X is running. my bad.
<
Wizzup>
let me kill X
<
Wizzup>
the last part of dmesg is me running the test with X not running
<
LiquidAcid>
hmm, ret = -19, ENODEV
<
LiquidAcid>
on which platform is this?
<
LiquidAcid>
hmm, and you sure that g2d is probed?
<
LiquidAcid>
because it doesn't look like, so either the driver wasn't enabled or there is no node for the block in the dts
<
Wizzup>
CONFIG_DRM_EXYNOS_G2D=y according to /proc/config.gz
<
Wizzup>
It may indeed not be in the dts. I do not know.
<
LiquidAcid>
either this, or it isn't probed some reason
<
LiquidAcid>
check the kernel log, it should at least have a line about it being probed
<
Wizzup>
I'm afraid the ring buffer doesn't show my boot log at this point
Vasco is now known as Vasco_O
<
LiquidAcid>
aren't you using some logging daemon?
<
Wizzup>
s5p-g2d seems to be regged though
<
javier__>
Wizzup: I quick grep shows that only exynos5250.dtsi has a G2D node for Exynos5
<
Wizzup>
Yeah, I am.
<
Wizzup>
javier__: I also tried to grep but saw nothing
<
LiquidAcid>
i guess it was never enabled for all exynos5 devices then
<
Wizzup>
I do have a s5p-g2d entry in /sys
<
Wizzup>
/sys/bus/platforms/drivers/s5p-g2d
<
LiquidAcid>
yeah, but that doesn't mean that the driver successfully probed the device
<
javier__>
Wizzup: that's the driver, not the device
<
Wizzup>
javier__: ACK
<
LiquidAcid>
i think it was only enabled globally for exynos4 some while ago (i think either marek or krzh)
<
Wizzup>
grep -ri g2d in arch/arm/boot/dts shows nothing for me
<
Wizzup>
wait - now it does.
<
LiquidAcid>
Wizzup, it should show at least the exynos4 dtsi
<
Wizzup>
javier__: so it seems to be enabled for snow
<
Wizzup>
I have a snow at home that I can test with, but I won't be home for another week ;-)
<
javier__>
Wizzup: yes, that's what I said
<
LiquidAcid>
Wizzup, or just copy the node over
<
LiquidAcid>
hmm, however i don't think that will work
<
LiquidAcid>
i think there were some sysmmu differences there
<
LiquidAcid>
it has two mmus on exynos5420
<
javier__>
LiquidAcid: I don't think he is using with IOMMU enabled though, since that prevents the Exynos Chromebooks to boot
<
LiquidAcid>
javier__, oh, if that's the case the driver won't work anyway, since it relies on iommu caps
<
javier__>
LiquidAcid: I see...
<
LiquidAcid>
javier__, is that another iommu issue, or still the fimd/iommu one?
<
javier__>
LiquidAcid: the fimd/iommu one
<
javier__>
Wizzup: btw, the Exynos5420 g2d node should be quite similar to the 5250, if not the same
<
javier__>
the register address is the same and there is even a CLK_G2D in the 5420 clk driver
<
LiquidAcid>
javier__, i think the only difference is the mmus
<
javier__>
I've to check if the interrupts
<
LiquidAcid>
javier__, also the reserved memory region patch by marek is in the tree, so that might fix the iommu/fimd issue
<
LiquidAcid>
Wizzup, arch/arm/boot/dts/exynos5800-peach-pi.dts
<
LiquidAcid>
is that the one you're using?
<
javier__>
LiquidAcid: interestingly the sysmmu are defined in mainline for G2D: sysmmu_g2d{r,w}
<
LiquidAcid>
then it should reserve the memory regions
<
LiquidAcid>
javier__, yeah, it's read and write
<
javier__>
so it may be a matter of just copying the 5250 node and set iommus = <&sysmmu_g2dr>, <&sysmmu_g2dw>
<
LiquidAcid>
i guess so
<
LiquidAcid>
javier__, have you checked the interrupts?
<
Wizzup>
Thanks for the help trying your code. I'm all set to try it again later on - I need to get a rest first. It's getting late here.
<
LiquidAcid>
i need to beat up some villians now ;)
<
LiquidAcid>
javier__, the idea would be to move the node down to exynos5.dtsi and just configure iommus in the corresponding dtsis
<
LiquidAcid>
javier__, i'm not sure though if the driver needs specific handling if read and write go through different mmus
<
LiquidAcid>
i guess one would need to check this
<
javier__>
LiquidAcid: sorry, I was looking at the Exynos manual
<
javier__>
yes, same IRQ: GIC SPI 91
<
javier__>
LiquidAcid: and yeah, dunno if the driver handles that. I don't see G2D enabled even in DTS of the vendor tree
<
javier__>
s/enabled/enabled in Exynos5420 DTS
isaque has quit [Ping timeout: 276 seconds]
paulk-collins has quit [Quit: Leaving]
isaque has joined #linux-exynos
cosm_ has quit [Quit: Leaving]
wwilly has quit [Quit: Leaving]
Vasco has joined #linux-exynos
sjoerd` has joined #linux-exynos
mturquette has quit [*.net *.split]
indy has quit [*.net *.split]
Vasco_O has quit [*.net *.split]
sjoerd has quit [*.net *.split]
indy_ has joined #linux-exynos
LiquidAcid has quit [Quit: Leaving]