elastic_dog has joined #maemo-leste
Guest95 has quit [Quit: Client closed]
panzeroceania_ has quit [Ping timeout: 268 seconds]
panzeroceania_ has joined #maemo-leste
nohit has joined #maemo-leste
joerg has quit [Ping timeout: 245 seconds]
joerg has joined #maemo-leste
_whitelogger has joined #maemo-leste
uvos has joined #maemo-leste
<tmlind> fyi, here's a little shell script for a few socs to use with pvrsrvinit: muru.com/linux/d4/pvrsgx-init
<tmlind> trying to do the init on nfsroot was just driving me nuts with different modules and separate library path for each soc..
<uvos> the fact that the blobs cant even deal with the core revision being different is nuts
<uvos> at runtime
<tmlind> if anybody wants to play with n9, looks like initializes with the ti335x blobs
<uvos> i wonder how there code looks
<tmlind> some of it could be fixed and detected in pvrsrvinit.c
<uvos> well i know accually its #ifdef CONDITION_60000000
<tmlind> detect the soc like my shell script does
pere has quit [Ping timeout: 265 seconds]
<freemangordon> uvos: did you manage to try sway on n900?
<freemangordon> or whatever it was
<freemangordon> tmlind: can I help somehow with fixing 33bc438d6d8883d77e37b369fe5144ee9b01fad8 ?
<freemangordon> wait, what? pvr_dri in blobs is with debug symbols?
<freemangordon> uvos: does it make sense to pull leste mesa master into 'your' mesa (the one with pvr driver)
<uvos> freemangordon: no not yet
<freemangordon> ok
<freemangordon> what about mesa?
<freemangordon> I am going to try to build and test it
<uvos> freemangordon: leste mesa is just upstream
<freemangordon> I know
<uvos> freemangordon: we just have the package because its newer
<uvos> freemangordon: so no i think i based the pvr mesa on the same version as the leste one
<freemangordon> does it make sense to merge it to your tree?
<uvos> freemangordon: not atm
<freemangordon> ok
<freemangordon> uvos: do you cross-compile it?
<freemangordon> or you build on d4
<uvos> no i just compiled it on d4
<freemangordon> oh
<uvos> was not that bad
<freemangordon> I will try to cross-compile
<uvos> 30 min or so
<freemangordon> yeah,but... still
<uvos> maybe less
<uvos> yeah corss-compile would be great
<freemangordon> will try
<freemangordon> uvos: mesa-pvr-meamo branch I guess?
<uvos> another reason for me to get leste to work on xt1602 - ts farily fast
<uvos> yes
<freemangordon> won.t happen
<freemangordon> will have to compile on the device
<freemangordon> and ofc battery is on 0 % :)
<uvos> why dont you do it on the n900 instead then? :P
<freemangordon> single 600MHz core
<freemangordon> will take ages to compile
<Wizzup> You can do it with qemu
<Wizzup> or with spinals methods
<freemangordon> I was looking for qemu debian armhf images ;)
<Wizzup> you would have to update systemd-nspawn with lxc or something
<freemangordon> trying to install armhf gl libs in x86 leste wanted to remove almost everything
<freemangordon> lemme try with a container
<Wizzup> maybe we need to provide a lxc file
<freemangordon> mhm
<freemangordon> I think for now I will just wait d4 battery to charge
<freemangordon> lets hope it is not broken
<freemangordon> hmm, I see 15%, SO IT SHOULD BE OK
<freemangordon> oops
<Wizzup> usually leaving it on a pc charger overnight
<uvos> frendly reminder that https://github.com/maemo-leste/leste-config/pull/25 makes leste shutdown before the battery is totaly empty avoiding this state on mapphones
<Wizzup> uvos: I think the phone still uses some when powered off no
<uvos> on d4 ist manageble
<uvos> 1 ish mA
<uvos> so takes quite some time for the battery to get to the sate where its hard to charge
<uvos> on bionic yeah
<uvos> shits bad
<uvos> hmm just booted 5.15 with new rootfs
<uvos> no wlan interface
<uvos> fun
<Wizzup> uvos: what if you manually probe the driver
<uvos> its loaded
<freemangordon> on n900?
<uvos> yeah
<freemangordon> you need to modprobe wl1251 and wl1251_spi
<freemangordon> also, wlan inteface gets renamed after every bood
<uvos> yeah they are there
<freemangordon> *boot
<uvos> i only have lo
<freemangordon> wlan0,1,2...
<freemangordon> check in dmesg what wlan0 was renamed to
<uvos> missing fw maybe
<freemangordon> this is debian not leste?
<uvos> pretty hard to check dmesg on device
<uvos> yeah
<freemangordon> ah, could be missing FW, yeah
<uvos> no serial :(
<freemangordon> yeah, its tricky, I know ;)
<uvos> anyhow ill have to pick this up later
<uvos> ttyl
<uvos> neat i cant even type wl1251 to reload the driver
<uvos> to look at dmesg
<uvos> beacuse no numbers
<freemangordon> burn leste image and move dsme from /etc/init.d
<uvos> but then i dont have new enough glibc
<uvos> anyhow i really need to go no ttyl for real
<uvos> *now
<freemangordon> ttyl
<freemangordon> hmm, I shall fix dsme to not spit errors when it cannot open a watchdog
uvos has quit [Quit: Konversation terminated!]
<freemangordon> oh, oops on d4 :(
<freemangordon> while it was upgrading
<Wizzup> this might be the 'dsme is restarting' issue
<freemangordon> no
<Wizzup> kernel oops?
<freemangordon> kernel oops in 8250_serial
<Wizzup> hm
<Wizzup> well is it new kernel?
<freemangordon> no, should be stock one
<freemangordon> Linux devuan-droid4 5.11.22 #1 SMP PREEMPT Thu Sep 30 10:40:46 UTC 2021 armv7l GNU/Linux
<freemangordon> though that is after the restart, no idea if kernel was upgraded
<Wizzup> it probably was
<freemangordon> hmm, glib was not upgraded with the fixed version?
<tmlind> freemangordon: let's debug that revert a bit later, trying to adding pvr_ocp_readl/writel to pvr-drv.c..
<freemangordon> ok
<freemangordon> hmm, it was upgraded
<freemangordon> ah, I guess I need hildon-mime-whatever
<freemangordon> wtf! meson (>= 0.52)
<freemangordon> parazyd: where shall I get that from! backports?
<freemangordon> uvos: hmm, I don;t see pvr driver being build
<freemangordon> do I need to tweak build scripts?
<Wizzup> we have a package to build it don't we?
<Wizzup> maybe that contains some instructions
<freemangordon> I had to tweak debian/rules
<freemangordon> still building so not sure what the result will be
<Wizzup> ok
<freemangordon> Wizzup: any clue why wlan interface get renamed on every reboot? on n900 with 5.15 that is
<freemangordon> *gets
<Wizzup> yes, it gets a random mac
<Wizzup> and our ifup thing is not called probably
<Wizzup> the one that assigns a static mac from a file
<freemangordon> ah
<freemangordon> hmm, I was under the impression this is fixed in the kernel
<freemangordon> like, it does FW load of a FW that was generated by reading CAL
<freemangordon> hmm, where is Pali
<Wizzup> I think the kernel folks didn't like it or something
<freemangordon> did they come up with a way to fix that?
<freemangordon> I don;t remember how did that discussion end
<Wizzup> I don't think so
<freemangordon> hmm, pvr_dri in 1.17 seems to have more stuff that chromiumos pvr_dri
<freemangordon> *than
<freemangordon> luckily it is compiled with debug symbols, so I can RE back the missing parts
<Wizzup> what parts are missing?
<freemangordon> like for example PVRDRIFlushBuffersForSwap calls (missing in chromeos) PVRDRICacheFlushSurfaces
<freemangordon> blobs support PVRDRI_API_GL it seems
<freemangordon> I think I've just hit zswap on d4
<freemangordon> and it looks dead :)
<freemangordon> we shall remove that
<Wizzup> hm
Guest62 has joined #maemo-leste
<Guest62> hi guys, can anyone tell me if sphone is working on Droid 4? (if yes, how do I get it to install? do I need to add a new package repository?)
<Wizzup> hi
<Wizzup> it works for me in the sense that I can smses and it vibrates on calls, but the audio switching does not work for me yet
<Guest62> I see; so how did you install it?
<Wizzup> I think you might need to enable the -devel repo(s)
<Wizzup> so you need additional repos in the apt sources
<Wizzup> add this/these:
<Wizzup> # cat /etc/apt/sources.list | grep devel
<Wizzup> deb https://maedevu.maemo.org/leste beowulf-devel main contrib non-free droid4
<Wizzup> deb-src https://maedevu.maemo.org/leste beowulf-devel main contrib non-free droid4
<Guest62> firing up the Droid rn
<Wizzup> then 'apt install sphone' should work
<Guest62> ha, how do I enter this "|" character on the Droid keyboard? :P
<Guest62> nevermind, used nano to edit the source.list file
<Guest62> and got this to install, thank you!
<Wizzup> great
<Wizzup> btw, got a nickname other than guest?
<Wizzup> and for |, it's mapped on p
Guest62 has left #maemo-leste [#maemo-leste]
pere has joined #maemo-leste
_inky has joined #maemo-leste
<freemangordon> hmm, mesa still builds
xmn has joined #maemo-leste
zhxt has quit [Ping timeout: 265 seconds]
<tmlind> freemangordon: maybe send your fix with a description etc to the linuxopenpvr mailing list? my changes will take some more time and testing..
<freemangordon> ok
Daanct12 has joined #maemo-leste
Danct12 has quit [Ping timeout: 265 seconds]
<freemangordon> hmm, glmark works with mesa I just build
<freemangordon> uvos said it doesn't, weird
<freemangordon> glmark2 Score: 36
<freemangordon> with TI provided blobs it is 37, but I guess this is within the margin of the error
<freemangordon> Xorg started with no issue, however, I hit the known bugs as soon as I tried xterm
<freemangordon> ok, seems we have x support :D
<freemangordon> uvos was right
<freemangordon> ok, everything runs fine, besides that es2gears_x11 are noit visible :D
<freemangordon> but, I am almost sure this is glamor bug I was fixing
<freemangordon> 234.848] (II) modeset(0): glamor X acceleration enabled on PowerVR SGX 530
<freemangordon> you know what! all this means we don;t need any shim or hacks :D
<freemangordon> worst thing is to patch 2-3 blobs to relax glibc requirement
<freemangordon> but nothing needs to be preloaded ;)
<sunshavi> good news. And a lot of work have been done
<Wizzup> freemangordon: right, and then glamor fixes or alt 2d path?
<freemangordon> I am waiting for uvos to report on the performance with wayland
<Wizzup> right
<freemangordon> hmm
<freemangordon> es2gears seem to run on swrast
<freemangordon> Wizzup: do you know how to tell mesae wich dri driver to use?
<freemangordon> *mesa
<freemangordon> looking at it, but cannot find anything useful
<Wizzup> yeah, I guess it's not gallium
<Wizzup> maybe MESA_LOADER_DRIVER_OVERRIDE ?
* freemangordon tries
<freemangordon> yeah, that's better :)
<freemangordon> I still see no gears, but fps is 57 instead of 19
<freemangordon> OMG I see gears!!!
<Wizzup> what did you change?
<freemangordon> initially I started xorg with 16bpp
<Wizzup> ah
<freemangordon> also, I added MESA_LOADER_DRIVER_OVERRIDE=pvr to xorg startup
<freemangordon> no idea which one helped
<freemangordon> will try
<Wizzup> next up, h-d? ;D
<freemangordon> no, glamor needs to be fixed first, IIUC
<freemangordon> yeah, depth 16 is not somehting that pvr likes it seems
<freemangordon> I can try to hack it though
<Wizzup> I don't think we use 16bpp even on the n900 now
<Wizzup> (otherwise the status applet was buggy)
<freemangordon> because of the buggy drivers
<Wizzup> yeah ok, I suppose I meant it's nice ot have, but we don't need it right now
<freemangordon> sure
<freemangordon> running glmark_es2 now, lets see
<Wizzup> in x11 mode I guess, not drm mode?
<freemangordon> could someone run glmark_es2 with old driver to have a base to compare to
<freemangordon> yes, in x11 mode
<freemangordon> hmm, x11 eats all the cpu, not good
<Wizzup> probably a bottleneck in the 2d part?
<freemangordon> yeah, maybe
<freemangordon> but, why 2d part?
xmn has quit [Quit: ZZZzzz…]
<Wizzup> freemangordon: isn't this also using glamor?
<Wizzup> I mean, for anything that is drawn on the screen, no/
<freemangordon> yes
<Wizzup> well that could be the bottleneck then?
<freemangordon> but, 3d only calls present, which should not take twice the time
<Wizzup> well what's the score? :)
<freemangordon> still running, but I expect something like 20
<freemangordon> or even less
<freemangordon> it does something with the CPU
<freemangordon> oh, wait
xmn has joined #maemo-leste
<freemangordon> I think I know what it is
<freemangordon> by default glamor doesn;t know about BGRA
<freemangordon> so i guess it does BGRA<->ARGB in software
<freemangordon> well, I have a patch for that s well :)
<Wizzup> heh
<freemangordon> oh, ok, I will have to get back to xorg
<freemangordon> abook will have to wait :)
<Wizzup> get back to what part in particular?
<freemangordon> start from where I stopped
<freemangordon> SubImage2D or whatever
<Wizzup> ah, I didn't know we were at this point before with 1.17
<Wizzup> I guess we were
<Wizzup> right
<freemangordon> maybe it will work with upstream mesa
<freemangordon> glmark2 Score: 25
<Wizzup> so what does not work right now
<freemangordon> xorg(pvr) hangs as soon as you start xterm
<Wizzup> there were the previous ones: https://pastebin.com/hXWt4HPP
<Wizzup> so maybe it is the swizzle as you mention
<freemangordon> mhm
<freemangordon> I will have to build xorg
<freemangordon> this will take a while :)
<freemangordon> xterm:
<freemangordon> [ 2195.355468] PVR_K:(Error): SGXOSTimer() detected SGX lockup (0x12 tasks)
<freemangordon> [ 2195.362487] PVR_K: HWRecoveryResetSGX: SGX Hardware Recovery triggered
<freemangordon> but, this is a known issue
<freemangordon> The fact we don;t need my shim is really great
<tmlind> cool :)
<freemangordon> yeah
<freemangordon> lets see what uvos will report about WL 2d rendering performace
<tmlind> what's the 2d test, glmark2 also?
<freemangordon> if it is acceptable, I am goint to resume fixing glamor
<freemangordon> tmlind: no, but I am not sure what it is
<freemangordon> video playback I guess
<tmlind> ok, there's some moving linaro windows around test in glamark2
<freemangordon> or scrolling long lists
<freemangordon> yeah, but it is still 3D
<tmlind> right, all 2d is using 3d with wl
<freemangordon> the point is - is n900 capable of video playback through 3D
<freemangordon> for example
<freemangordon> or - is it powerful enough so one can do kinetic scrolling of long list without (much) stutter
<freemangordon> tmlind: will omap2_defconfig kernel boot on droid4?
<freemangordon> your 5.15-rc2 tree that is
<freemangordon> oh, I need that reverted patch back I guess
<freemangordon> Wizzup: tmlind: d4 is ti443x, right?
<tmlind> freemangordon: yeah omap2plus_defconfig should boot on droid4 and n900
<freemangordon> ok
<tmlind> yeah ti443x
<freemangordon> thanks
<Wizzup> tmlind: what would be a good way to make a static minimal kernel for kexecboot purposes? I started with doing something based on the loaded list of modules on the d4, but that seems not to work too well, so now I am trying omap2plus_defconfig with all '=m' set to '=y' _+ make defconfig
<Wizzup> so this boots to a point, but then it just doesn't want to mount any rootfs
<Wizzup> even rootdelay doesn't make it panic after the delay
<tmlind> Wizzup: something uvos uses for the clownboot on bionic should do the trick
<tmlind> freemangordon: so which xorg driver did you end up tweaking, the armsoc driver?
<Wizzup> yeah, but he doesn't have the config/remember
<Wizzup> :)
<Wizzup> we'll see
<Wizzup> ok, I think I know what went wrong
<tmlind> ok
<Wizzup> yeah ok, the out of tree dtb was out of sync
<tmlind> ok
<tmlind> Wizzup, uvos: you might be able to recover the clownboot .config with zcat /proc/config.gz if it was built in
<Wizzup> yeah, I also suggested that a while aog
<Wizzup> I think I got it now though
<Wizzup> the dtb -mostly- worked fine, except that it didn't see the mmc sd card
<tmlind> ok
<Wizzup> but screen, modem, etc worked
<Wizzup> unlucky I guess :)
<Wizzup> going to pick a partition for kexecboot now
<tmlind> well the old dtbs are incomplete nowadays, you need a quite recent one for now
<Wizzup> yeah this was from 5.11
<Wizzup> and I am booting 5.14
<Wizzup> minor changes to the common dtsi
<Wizzup> enough to break it :D
<tmlind> yeah ok, we dropped all the old platform data for hwmod, so only dtb will have the interconnect data
<tmlind> you should see a warning about old dtb during boot
<freemangordon> tmlind: umm, what do you mean?
<freemangordon> I didn;t tweak anything, what I did is to get chromeos mesa that has pvr dri driver
<freemangordon> so, now it uses modesetting with glamor
<freemangordon> 'it' == xorg
<tmlind> oh ok nice :)
<tmlind> wow
<freemangordon> as uvos correctly guessed, with that mesa there is no need of my shim as X support will be there
<freemangordon> and yes, it is there
<tmlind> great
<Wizzup> what is a sensible way to see what partition is bpsw?
<tmlind> which kernel?
<Wizzup> just a second
<Wizzup> 2.6.35.7-g5fa4155
<Wizzup> 2.6.35.7-g5fa4155
<Wizzup> oops
<tmlind> cat /proc/cmdline might show it for android kernels
<Wizzup> shows: console=/dev/null rw mem=456M@0x80000000 mem=52M@0x9CC00000 vram=4M omapgpu.vram=0:4M init=/init ip=off mmcparts=mmcblk1:p7(pds),p15(boot),p16(recovery),p17(cdrom),p18(misc),p19(cid),p20(kpanic),p21(system),p22(cache),p23(preinstall),p24(userdata),p25(emstorage)
<tmlind> hmm maybe do strings on the cid partition?
<tmlind> hmm no cdt partition, not sure which one it is
<Wizzup> cid seems to be only 0xff
<Wizzup> It might just be the same as on bionic, so p14, but I don't really know how to verify that
<Wizzup> I just need to figure it out so that I can give the static zimage the right partition to load kexecboot from
<tmlind> if it's empty it's not in use, if it's not empty, don't overwrite
<Wizzup> clearly fastboot knows what bpsw is
<tmlind> do hd on it to see
<Wizzup> sorry, hd?
<Wizzup> hexudmp
<Wizzup> ok
<Wizzup> yeah I did that
<tmlind> could be p14, but only 2MB?
<tmlind> would be still enough for kexecboot, kexec and busybox
<Wizzup> going to flash to bspw and see what's in p14
<Wizzup> FAILED (remote: Battery Low)
<Wizzup> in another hour I guess :)
<Wizzup> yeh it is p14
<Wizzup> uvos: when you get back, I have trouble mounting /system as rw from android
<Wizzup> # mount -o remount,sw /system
<Wizzup> Usage: mount [-r] [-w] [-o options] [-t type] device directory
<Wizzup> maybe it needs none there or something?
pere has quit [Ping timeout: 265 seconds]
<Wizzup> it wasn't the sw typo
<Wizzup> it needs to be like this
<Wizzup> mount -o remount,rw /dev/block/system /system
<Wizzup> oh, looks like mmc1 is not visible, but mmc0 is [...]
<Wizzup> I guess I get to reflash android now
<Wizzup> just for the record the modem unlock survived android reflash, as expected
<Wizzup> ah, I think it might just be called mmcblk0p2 ....
<Wizzup> err mmcblk2**
<Wizzup> ah, no internal mmc on my 5.14.9 - will fix that later
<Wizzup> I think once that's solved for the droid3 then kexecboot will work
<freemangordon> hmm, h-d starts but nothing gets rendered
<freemangordon> I suspect I shall export MESA_LOADER_DRIVER_OVERRIDE=pvr systemwide
pere has joined #maemo-leste
<freemangordon> hmm, glmark2 does not work on d4
<freemangordon> glmark2-drm that is
<sicelo> i'm really impressed fmg :-)
<freemangordon> :)
<Wizzup> freemangordon: yeah that seems like a safe bet (system wide)
<freemangordon> where should I put that?
<Wizzup> don't know per se, but if you put it in /etc/profile it might work
<freemangordon> omg, hildon-desktop started :)
<freemangordon> on n900 that is
<sunshavi> awesome
<freemangordon> h-d scrolls with 58 fps
<sicelo> this is N900?
<freemangordon> mhm
<freemangordon> well, there are artifacts
<freemangordon> like the last 2-3 frames are 'recycled'
<freemangordon> also, it seems TFP does not work
<freemangordon> (texture-from-pixmap)
<freemangordon> could be a clutter bug though, but I doubt
<freemangordon> yeah, glamor is buggy as hell :(
<freemangordon> oh
<freemangordon> tmlind: any idea:
<freemangordon> cma: cma_alloc: reserved: alloc failed, req-size: 332 pages, ret: -12
<Wizzup> larger cma required maybe? sometimes there is a hardcoded limit in kernel
<freemangordon> yeah
<freemangordon> hmm, why h-d does not rotate?
<freemangordon> ok, xrand rotates it, but scrolling is with 10 fps
<Wizzup> freemangordon: n900 sensors are not set up for rotation
<Wizzup> if they would, current n900 would break
<freemangordon> I am sending signal with dbus-send
<freemangordon> run-standalone.sh dbus-send --system --type=signal /com/nokia/mce/signal com.nokia.mce.signal.sig_device_orientation_ind string:'portrait'
<freemangordon> hmm, I guess it needs mce running
<Wizzup> right
<Wizzup> rotated being slow could be many things I guess
<freemangordon> no, I mean - h-d does not rotate when I send signal with dbus-send
<freemangordon> if I rotate with xrandr, then scrolling is slow
<freemangordon> oh, maybe orientation is locked?
<freemangordon> what was the gconf key?
<Wizzup> sorry, do not know
<sicelo> /apps/osso/hildon-desktop/orientation_lock (at least here on my fremantle system)
<freemangordon> yeah, but it is true
<freemangordon> oh
<sicelo> :-)
<sicelo> should be false if you want orientation active
<freemangordon> orientation_lock = false
<freemangordon> sorry, I meant false
<freemangordon> anyway, I think I should stop for today
<sicelo> ah
belcher has quit [Ping timeout: 265 seconds]
belcher has joined #maemo-leste