15:00:20 #startmeeting Fedora ARM Working Group Meeting 15:00:20 Meeting started Tue Aug 22 15:00:20 2023 UTC. 15:00:20 This meeting is logged and archived in a public location. 15:00:20 The chair is coremodule. Information about MeetBot at https://fedoraproject.org/wiki/Zodbot#Meeting_Functions. 15:00:20 Useful Commands: #action #agreed #halp #info #idea #link #topic. 15:00:20 The meeting name has been set to 'fedora_arm_working_group_meeting' 15:00:21 #chair pwhalen pbrobinson jlinton jforbes coremodule 15:00:21 Current chairs: coremodule jforbes jlinton pbrobinson pwhalen 15:00:21 #topic roll call 15:00:29 morning 15:00:38 morning jforbes 15:01:50 * jlinton waves 15:02:08 hi jlinton 15:02:30 I don't have much outside the usual review, does anyone have anything special for today? 15:03:33 Not too much of interest. I did ask pjones about the grub MR for fixing the x13s. The maintainer was on PTO until yesterday 15:03:39 No, the grub2/x13s fedora PR remains open. I've bumped a couple people, not sure when it will happen. 15:04:30 But pjones mentioned he was working on a rebase to the upstream rc which might clobber this MR (but also hopefully fix the problem?) 15:05:02 #info grub/x13s issue is still prevalent, jforbes reached out about the status, the maintainer was on PTO until yesterday. It's safe to expect a fix in the near future. 15:05:07 Yes, he mentioned a rebase to me as well. Depends on whether he keeps the downstream startimage replacement code. 15:06:43 well, let's call this meeting for today, I don't have anything else and we've documented the movement on the issue 15:07:21 Only other tidbit, I picked up an m2 mini last week, got it running asahi headless. I think the next goal is to see if we can get stock fedora 39 running in some capacity on these boxes. It is decent hardware (builds kernels faster than koji) 15:08:05 wow! how much ram jforbes ? 15:08:10 We obviously won't be carrying the way out of tree patches (like graphics driver) 15:08:24 I have a near stock version running, although I have both the 16k and mainline on it. 15:08:51 Only 8G. but for $599 US, not bad. 16G version had a bigger drive and faster CPU, but was $1299 15:09:30 yeah, not bad at all. 15:09:33 jlinton: I am planning to turn on 16K as as standard build for Fedora once we have just a bit more upstreamed. 15:10:13 coremodule: 'fedpkg local' for the f38 kernel on it takes about 61 mins, compared to 116 minutes in koji 15:10:38 that's wild. do you plan to use it as a day-to-day machine? :) 15:10:56 Well thats some kind of config issue on the koji instances which are altra's now, right? 15:10:57 No, I plan to run it headless. The m2 mini doesn't even have working hdmi just yet 15:11:13 oh right, right, you said that. 15:11:42 jlinton: probably just the resources available. This is 10 cpu cores, not sure how many the koji instances are given for aarch64 15:12:48 still fast enought that it makes sense to use for bisect and such when I need aarch64 builds. Faster than anything else I have here 15:14:08 8 cores for koji 15:14:48 Yah, thats been the limit forever. Sorta sad when the machine has 100+ cores. 15:14:52 Also no idea what the disks are, kernel build is I/O heavy in a lot of the packaging places 15:16:35 I mean its gotta be some kind of NVMe at this point. I did a bunch of kernel nvme builds with some optane drives last year, and despite those things being a good order of magnitude faster for random 4k, there wasn't a huge difference in build perf. The kernel does to good a job caching the IO. 15:17:37 Anyway, I know asahi is a tangent objective, with their own SIG, but it does make a good build system for people doing aarch64 on a lot of smaller hardware. And I do plan to have things working as best we can with stock Fedora over time. 15:18:59 Yes, its the only perf/price competitive machine in that price range. Its been a big hole between SBCs/ and full blown servers. 15:19:17 If only it was native SBSA/SBBR/etc.. 15:20:08 Well, it does seem that there might be some market if another vendor chooses to really participate with more standard hardware 15:24:20 isn't the windows dev kit pretty comparable? (not that I think anyone has done the work to support linux on it...) 15:24:28 also not sbsa/sbbr/etc 15:24:48 pjones: I thought the windows dev kit was the same SOC as the X13s 15:24:56 oh, right, yeah 15:24:59 Perf wise? Not really. But uh... the some of the work required is grub2.. lol 15:26:08 ironically no, the way its firmware is implemented it doesn't have the same problem you're seeing, so most of the work is just fdt goo 15:28:01 ? I was under the impression all the QC machines were now failing to +X data allocations. Is that machine not doing that? 15:29:04 The hyper-v guest of course works with the current grub, but that's a different FW stack. 15:30:49 I just like the idea of being able to tell someone how to install on an X13s that doesn't start with "pull the nvme out of the laptop and put it in a USB cradle" 15:30:52 it's a completely different firmware implementation 15:31:53 I have the USB boot working on the x13s, which i'm assuming is part of the requirements for doing an install, but at the moment its a case of disable this driver, or remove this firmware from the volume to keep it from reprogramming the device in the middle of the boot process and killing it. 15:33:32 I will flip back to it in the near future and come up with something we can merge to fix it.. 15:35:33 would be nice if we could come up with something for F39 15:35:58 I think the 6.5 kernel itself is in decent shape 15:36:14 It boots.... 15:36:24 * jlinton chuckles 15:36:37 sometimes. 15:37:36 well, you have to manually add some things to the initramfs. 15:38:24 I don't have that at the moment, but I might have =Y'ed one of the drivers. 15:38:52 I ran out of time a couple weeks back before PTO, and haven't gotten back to it. 15:39:45 Its not pretty when it boots though, it turns the screen on/off a couple times, and leaves it off for a good couple seconds during the rootfs pivot. 15:41:43 "keep two more regulators than documented enabled, total: 15:41:45 regulator-edp-3p3, regulator-edp-bl, ldo3 and ldo6) or the screen 15:41:47 would blank.Y 15:43:23 I think those are the regulators that are would disable the screen in its entirely without the kernel command line parms? 15:44:21 vreg_l6b and vreg_edp_3p3 keep it from the full blank after 30 seconds 15:44:33 the others added are supposed to help with the other blanking 15:49:58 this the x13s? 15:50:15 yes. 15:52:24 pbrobinson: There were some linux-firmware, notes I have WRT to it, one of which was some of the GPU firmware was missing. I will poke you the info later today if your interested. 15:52:43 jlinton: yep, pls do 15:54:15 a660_gmu.bin ? 15:54:28 https://lore.kernel.org/lkml/20230811160505.174574-1-robdclark@gmail.com/ 15:55:11 hey guys, I'm not sure if there's another meeting at the top of the hour, so I'm going to close this one. feel free to keep going once it's closed. 15:55:13 #endmeeting