00:04imirkin: yeah
00:04imirkin: no worries
00:05imirkin:will try reverse psychology
00:05imirkin: i bet you couldn't even do it
00:06ccr: :D
00:07ccr: I bet I could .. if I wanted to! .. ..
00:07imirkin: =]
00:10ccr: to be honest, I am more scared for my Haswell-based desktop PC if the "let's get rid of classic Mesa drivers" Mesa-lts/old/something plan ever materializes. but I'm fairly sure that it would be somewhat insane to jump into the deep end and try to cobble together a Gallium-based Haswell driver without a deep understanding of .. everything involved. :P
00:30karolherbst: ccr: well, as long as there is somebody willing to maintain it, your are safe
00:31ccr: agreed
00:31karolherbst: but I don't think that writing a gallium driver is all that much work
00:31karolherbst: especially if you got the compiler already
09:14HdkR: Lyude: Help. What Displayport MST actually works with dual 4k60 displays. Every single one I try has problems D:
09:42skeggsb: HdkR: on nouveau? i wouldn't be surprised if there's just pain not the memory bandwidth
09:42skeggsb: plain*
09:42skeggsb: nfi what the boot mclk even is on recent boards, we don't even bother to try and read them
09:52HdkR: Well, testing on Nvidia + Windows atm. These MST hubs are so flakey that I want the best chance of them working :/
10:30captainchris: hi everybody
10:31captainchris: I have a old nvidia gt240. with nvidia-driver-340xx its works fine but with nouveau driver vdpauinfo says alll codecs no supported and when i watch movie the cpu usage up to 140% and my computer freeze
12:53SolarAquarion: what's going on with nv130 with Hardware Acceleration
13:00SolarAquarion: nv130 is i guess in "mesa"
13:00HdkR: It's Pascal so it's supported in mesa
13:02HdkR: But what sort of hardware acceleration? 3D? Video?
13:03SolarAquarion: HdkR: Video
13:06HdkR: ah, you're in for a bad time then
13:23captainchris: hi everybody
13:23captainchris: I have a old nvidia gt240. with nvidia-driver-340xx its works fine but with nouveau driver vdpauinfo says alll codecs no supported and when i watch movie the cpu usage up to 140% and my computer freeze
14:47SolarAquarion: skeggsb: there isn't much detail concerning video HW acceleration on nvidia pascal
17:38imirkin: SolarAquarion: https://nouveau.freedesktop.org/VideoAcceleration.html
18:12SolarAquarion: imirkin: what's holding up vp-6+
18:12SolarAquarion: VP6+
18:12imirkin: caring
18:13imirkin: the engine layout changed -- went from 3 engines to a single one
18:13imirkin: from what i've heard, the actual data format is similar/identical to the previous versions
18:13RSpliet: CPUs got beefy enough to do the decoding without much sweat. It's still useful, but less so with every gen
18:13imirkin: also you'd have to figure out how to extract the necessary firmware, as well as all the ancilary bits of driving an engine
18:14imirkin: and who knows if you run into "secure mode" problems
18:14SolarAquarion: well, i guess "Secure Mode" would need to be disabled in the bios
18:14imirkin: not that secure mode
18:14imirkin: the one that prevents us from loading arbitrary e.g. pmu firmware that can handle the full reclocking sequence
18:14SolarAquarion: oh, the nvidia bios secure mode
18:15RSpliet: falcon microcores, not (V)BIOS
18:15SolarAquarion: yes, falcon microcores, gotcha
18:15imirkin: no one's working on it though
18:16imirkin: mlankhorst did vp4/vp5, i did vp2 (and vp3, which is a lot like vp4)
18:16SolarAquarion: shouldn't the firmware be in "nvidia-utils", buti guess some of it is now in CUDA
18:17imirkin: could be
18:17RSpliet: Don't think NVIDIA has released any video decoding firmware in a redistributable form?
18:17imirkin: the person who cares about vp6+ would have to work that out
18:17imirkin: normally it'd be part of the blob kernel driver
18:17imirkin: vp4 and earlier have user-space "blobs" too, but vp5 does not, and i doubt vp6+ would employ that technique
18:18imirkin: (it's super-insecure)
18:18SolarAquarion: so they're now probably only in the kernel driver, ala vp5
18:18imirkin: that is most likely.
18:18RSpliet: imirkin: Is that to let userspace switch between codecs?
18:18imirkin: RSpliet: no, it's for entirely silly reasons
18:19imirkin: RSpliet: oh wait, you might be right
18:19RSpliet: the two aren't mutually exclusive :-D
18:19SolarAquarion: so i guess one idea is to try to do the extract_firmware, but try to extend it for vp6+
18:19imirkin: yeah, i think it's for the different codecs. but i have some recollection of there being 2 firmwares you have to execute in sequence for the same codec too
18:19imirkin: (on the same engine. i could be dreaming though.)
18:19imirkin: SolarAquarion: yeah, that's one of the 75 steps necessary
18:20imirkin: well, not necessarily extract_firmware
18:20imirkin: but some mechanism.
18:22imirkin: anyways, the nice thing is that with just a single engine, there's probably less effort that has to go into working out various layout things of the intermediate engines
18:22imirkin: but RE'ing one video decoding engine was one too many for me. i have no interest in a second one.
18:23imirkin: (and even that one i relied on mwk's work on a lot)
18:29SolarAquarion: the whole "space" of the kernel firmware is different
18:31SolarAquarion: imirkin: does this sound like the place with the binary that needs to be extracted nv-modeset-kernel.o_binary
18:32imirkin: usually it's nv-kernel.o
18:32imirkin: check the extract_firmware script
18:32imirkin: and/or the scanner one
18:32imirkin: https://github.com/envytools/firmware
18:34SolarAquarion: imirkin: i downloaded the last driver, and the only .o is the modeset one
18:34SolarAquarion: i guess the nv modeset is where all the stuff is now
18:34imirkin: yea dunno
18:34imirkin: good luck!
18:35SolarAquarion: so the extract_firmware stuff would probably need to be a bit rewritten
18:36SolarAquarion: afaik
18:36imirkin: probably
18:37SolarAquarion: imirkin: https://wiki.archlinux.org/index.php/NVIDIA#DRM_kernel_mode_setting
18:37SolarAquarion: to support kms, they moved stuff around
18:37imirkin: ok
18:41SolarAquarion: i'm not particularly a programmer, lmao, but i guess it first steps of how to do it with 6+
18:51SolarAquarion: imirkin: oh found it nv-kernel.o_binary
19:00SolarAquarion: imirkin: i'm getting somewhere thanks to binwalk, i see a ext2 filesystem image
21:12SolarAquarion: imirkin: all the piglet tests skipped, is that okay
21:12imirkin: i'll let you work that out yourself
21:23SolarAquarion: runnning gpu tests
21:52SolarAquarion: my computer crashed. lmao.
22:45RSpliet: You get used to that when doing nouveau stuff
22:51SolarAquarion: RSpliet: i'm trying to figure out the firmwares, i did a binwalk, but i don't know how to go further
22:51RSpliet: SolarAquarion: I know, I kept an eye on the channel. There's not much I can do to help sadly. Was pretty surprised you found an ext2 filesystem. Any idea what's in it?
22:52SolarAquarion: no, not yet
22:52SolarAquarion: 1A98DA8.ext2
22:52SolarAquarion: if i can figure out what's in that file i think it can go further in the RE department
22:53RSpliet: Extract and use loopback mount?
22:53RSpliet: The trick is presumably to find the end of it. I'm pretty surprised about this tbh
22:54imirkin: eh, used to be gzip-based things
22:54RSpliet: I knew the blob pretty much reimplements half the kernel (because it can't use GPL symbols), but would be a tad surprised if it also implements a whole filesystem.
22:54imirkin: could well be that they just stuck a fs in there
22:54imirkin: and register it in the vfs even maybe? who knows
22:54RSpliet: Kind of feels like the run up to christmas eve now. What could be inside? Could be firmware and debugging symbols, could be coal!
22:54airlied: or it's just coincindence
22:55airlied: not sure how good binwalk is at validating thing
22:55imirkin: also quite likely.
22:55RSpliet: airlied: right. Party pooper! :-D
23:15SolarAquarion: RSpliet: i extracted but i seem to have corrupted it
23:27SolarAquarion: 25660847 0x1878DAF Copyright string: "Copyright NVIDIA"
23:27SolarAquarion: there is encrypted data here
23:27imirkin: it'd be funnier if it were AMD
23:28SolarAquarion: 27131431 0x19DFE27 mcrypt 2.2 encrypted data, algorithm: SAFER-SK128, mode: CBC, keymode: SHA-1 hash
23:28SolarAquarion: bunch of encrypted things
23:29SolarAquarion: i don't know how to forward with the extracted files
23:34SolarAquarion: or any of the files i extracted
23:34SolarAquarion: things are here NVIDIA-Linux-x86_64-455.45.01
23:34SolarAquarion: i don't know what, but things are there
23:42ccr: * NVidia takes the mask off ... "hah, you never suspected it was me, AMD, all along!"
23:42ccr: * dramatic music plays *
23:50SolarAquarion: got a lot of things different via strings