[14:26] <zzarr> alf_ how much work is it to make Mir talk to fbdev?
[14:32] <anpok_> zzarr: is that really the right problem to solve?
[14:33] <anpok_> i mean .. if you have a drm capable kernel driver.. you already have the right mechanism to efficiently post to screen, and to pass buffers from clients to server
[14:34] <anpok_> the only missing piece in your case seems to be a mali driver that works with drm
[14:34] <anpok_> or as a temporary solution kms-swrast
[14:34] <anpok_> oh of course your drm kernel driver could mis some features
[14:35] <anpok_> then we would have to add those.. what mir needs there can be easily enabled via various drm-helpers..
[14:36] <anpok_> zzarr: you could ask the lima driver developers..
[14:36]  * alan_g thinks it really is about time we cleaned up support for 3rd party "platform" modules
[14:37] <anpok_> zzarr: i did look at that situation with mali recently
[14:42] <anpok_> zzarr: the mali drivers user ump to manage memory... in theory ump is capable to grant other processes access to buffers. But I havent seen any way to get egl/glev2 render to such a handle, their fbdev-egl drivers do not over an entry point with a native ump handle to render to..
[14:43] <anpok_> the x11-egl drivers require an x11 connection, and also allocation happens underneath..
[14:43] <anpok_> so the only arm-mali solution that might work - is the android drivers..
[14:44] <anpok_> s/not over/not offer/
[14:45] <anpok_> so if you can build an android kernel for your board.. you should be able to get mir working with hybris and the arm drivers that you can download from arm.com..
[14:47] <anpok_> the only other options are: build and select kms-swrast as dri driver, or look into lima.. http://limadriver.org/
[14:50] <zzarr> okey anpok_
[14:51] <zzarr> what changes needs to be done to the kernel to run hybris?
[14:55] <zzarr> anpok_, or is there a tutorial how to rewrite the X driver to become a Mir one?
[14:57] <zzarr> anpok_, is there a Mir driver I can have a look at (so I can compare the mali X driver to it?)
[14:58] <zzarr> I'll read this later, I'll leave the chat open, have to go
[14:58] <anpok_> zzarr: hm not sure what hybris actually needs.. but the mali driver for android will require a mali kernel module.. I am not sure how to integrate that one. I once looked how it was done for an Allwinner SOC, and it didnt look easy..
[15:01] <zzarr> anpok_, okey, you don't think it's possible to "just install" the user space driver?
[15:01] <anpok_> as in I tried to migrate that one to a newer kernel version, and got lost in the integration code that needs to access clocks and interrupts
[15:01] <zzarr> ohh.... I see
[15:01] <anpok_> so the easier path here would be to get an android kernel for rockchip
[15:01] <anpok_> and port ubuntu-touch to it
[15:02] <anpok_> but I cannot help with that
[15:02] <zzarr> I thought it was a matter of kernel configuration
[15:02] <anpok_> well .. of configuring the right kernel
[15:03] <zzarr> the thing is, everything except the graphics and closing the lid/flipping are working
[15:05] <anpok_> someone did that for rk2918 http://forum.xda-developers.com/showthread.php?t=2318728
[15:05] <zzarr> I have seen that there are android specific configuration options in the kernel, but... I'll have a look
[15:09] <zzarr> If I manage to make an Android version of the kernel (or get one from some RK3288 based device) will the other drivers (/lib/modules... /lib/firmware...) stop working so I need a WiFi driver and so forth too?
[15:13] <anpok_> which kernel did you use?
[15:14] <zzarr> I used one for Kali linux for my device
[15:15] <zzarr> (it actually uses both KERN-A and KERN-B)
[15:15] <anpok_> so .. you would need one that has drivers/gpu/mali
[15:16] <zzarr> anpok_, the kernel(s) are basically the default one(s) but with console activated
[15:17] <zzarr> and as it have GPU acceleration when Chrome OS is booted it have some driver
[15:17] <zzarr> but it's drm
[15:21] <zzarr> anpok_, I hope that I'm not too tiresome, but as many devices are based on the RK3288 SoC I think it would be very nice to get Mir support for it :-)
[15:22] <anpok_> hm yeah it would be interesting to see how the mali driver that they probably use interact with the drm module
[15:23] <anpok_> chrome os has the advantage that they only have a single rendering process in the system, and the different browser rendering process communicate the command streams to that process..
[15:24] <zzarr> anpok_, yea, but I think that's only the user space driver
[15:24] <zzarr> and freon
[15:25] <zzarr> I'll download the code for the mali X driver
[15:25] <anpok_> oh the x11 user space parts are open source?
[15:27] <zzarr> I think they are, if I don't remember wrong
[15:28] <zzarr> here http://malideveloper.arm.com/resources/drivers/open-source-mali-gpus-ump-user-space-drivers-source-code-2/
[15:30] <zzarr> I have to go now, I'll check the chat later (I'll leave it open)
[15:30] <zzarr> thanks so much anpok_
[15:30] <zzarr> I'll figure it out some how ;-)
[15:30] <anpok_> ok
[16:12] <zzarr> anpok_, now I'm here again for a little while
[16:12] <zzarr> I'm sitting on a bus
[16:13] <zzarr> I'm remote controlling a computer at home with the chat
[16:20] <zzarr> anpok_, I got the ump driver, is that enough to write a Mir driver? (I don't see anything interesting yet)
[16:22] <zzarr> I'm soon to get off the bus
[16:22] <anpok_> enough to write a server i guess. but not enough to get a client to render with glesv2 as far as i understood it
[16:23] <zzarr> are you looking at the code?
[16:24] <anpok_> i was looking at the glesv2 sdk examples from arm ... looking for a way to initialize egl with an ump handle..
[16:24] <anpok_> but maybe i was missing something
[16:24] <zzarr> I have to close the connecton now
[16:24] <zzarr> okey
[16:24] <zzarr> bbl
[17:09] <zzarr> I'm back... for now, I was going to a postal office, but the guy who tried to deliver the package to my door wrote the wrong office on the paper he put in my mailbox
[17:09] <zzarr> so I have to go to another office....
[17:12] <zzarr> anpok_, the code I downloaded have a ump_arch.c, ump_frontend.c and ump_ref_drv.c are they useful?
[17:12] <zzarr> anpok_, there are other ump* files as well
[17:13] <zzarr> I have to go to the other postal office now
[17:14] <zzarr> bbl (again)