15:26 karolherbst: mhhh.. "gr: SKED: 00002000 [REGISTER_COUNT]" mhhhh
16:02 pmoreau: I’m guessing you have a block using too many threads considering the amount of regs used?
16:04 karolherbst: I think that was just fallout though
16:04 karolherbst: or maybe the test didn't fail.. mhhh
16:56 lockna: Hey! I'm working on a falcon emulator with a friend. Actually he's the brain behind the whole thing. I just help where I can. So, for he gave me a archive with some nvidia drivers, I need the falcon firmware blob in these drivers, how can I get them, in the fastest way? I know, I could use a hexeditor and search for falcon instructions, but yeah, that takes some time
16:59 RSpliet: lockna: do you need to start off with NVIDIA falcon binaries? Might be easier to start off with nouveau binaries first, so you can step through execution and see the corresponding source code, and even create your toy programs
16:59 RSpliet: Unless ofc you've already done all that :-)
17:01 lockna: Nah, I need the nvidia ones, because of their AES keys. ^^
17:21 pmoreau: So that’s for a Maxwell or more recent then? AFAIK no one here has extracted them.
17:34 lockna: Yes, it is. Well, then I'll do it myself. Do you guys have any tips on how I could go about this? Would be nice.
18:14 pmoreau: I do not, sorry. Maybe imirkin has, since IIRC he has tried to look for them.
18:29 lockna: Is he in this channel, or can I contact him in another way?
18:33 karolherbst: lockna: the keys won't be in the driver package
18:33 karolherbst: they are on the hw directly
18:33 lockna: I know
18:34 karolherbst: ahh, I see
18:34 karolherbst: the painful part is that they are also compressed but I think imirkin had something.. let's see
18:35 karolherbst: lockna: https://github.com/envytools/firmware
18:36 karolherbst: I think...
18:36 karolherbst: there was something new
18:36 karolherbst: ahh right
18:36 karolherbst: scanner.go
18:36 karolherbst: but no idea if that extracts the video stuff as well
18:37 lockna: Thanks, I'm gonna have a look at it.
18:38 karolherbst: but the thing is, extracting the firmware is annoying and potentially requires the hw, the driver and some kernel hacks to dump it somehow...
18:40 lockna: Yeah, but I find this totally interesting ^^.
18:41 imirkin: lockna: https://github.com/envytools/firmware
18:41 imirkin: this is what we got.
18:41 imirkin: oh. karol already pointed you at it. nevermind.
18:41 karolherbst: :)
18:42 karolherbst: I guess the video accel firmware is stored in a similiar way
18:42 imirkin: lockna: also this: https://nouveau.freedesktop.org/wiki/NVC0_Firmware/
18:42 imirkin: this is how we extracted firmware back in the bad old days
18:42 karolherbst: imirkin: you put both scripts into the repository :p
18:42 imirkin: karolherbst: extract_firmware actually works
18:43 imirkin: scanner is a lot more speculative
18:43 karolherbst: sure
18:43 karolherbst: but doesn't work for maxwell stuff
18:43 imirkin: i don't think i ever located the video firmware with it
18:43 imirkin: i expect it's stored differently than the other things
18:43 karolherbst: that would be annoying
18:43 imirkin: but it could probably be recovered using similar techniques. dunno.
18:45 lockna: Yeah, thanks. I must say, since I just joined the project, I still have a lot to learn. ^^ Thank you guys
18:46 imirkin: the idea of scanner was to look at the elf object directly (since clearly there'd be references to the blob from within the code)
18:46 imirkin: extract_firwmare just looks for gzip signatures within the file
18:47 imirkin: and it was written by knowing what the firmware actually looked like
18:47 imirkin: and using that knowledge to locate it in the blob in full
18:47 imirkin: whereas scanner takes a zero-knowledge approach and just finds blobs.
18:49 imirkin: as for AES keys, the firmware itself doesn't have AES keys
18:49 imirkin: the VP2 firmware (G84 - G96) did have some sort of variance between groups of GPUs, ostensibly for some sort of AES thing which differed between groups of GPUs, but there were only a handful of variations
23:43 imirkin: skeggsb: hey, so i'm seeing a case where the nouveau ddx will get "stuck" drm events when the screens go into dpms
23:43 imirkin: any thoughts on that?