Search logs:

channel logs for 2004 - 2010 are archived at http://tunes.org/~nef/logs/old/ ·· can't be searched

#osdev2 = #osdev @ Libera from 23may2021 to present

#osdev @ OPN/FreeNode from 3apr2001 to 23may2021

all other channels are on OPN/FreeNode from 2004 to present


http://bespin.org/~qz/search/?view=1&c=osdev2&y=21&m=10&d=12

Tuesday, 12 October 2021

00:00:00 <clever> nixos also has a qemu-kvm shell script, that will test if /dev/kvm exists, and maybe include -enable-kvm
00:00:00 <clever> so it cant fail when kvm is missing
00:01:00 <klange> You could just do -accel kvm:tcg (or something like that) and qemu will do that itself
00:10:00 <geist> oh huh yes the kvm is. i thought it was something that used libvirt or some sort of more complex wrapper
00:14:00 <clever> geist: at a closer glance, `perf kvm` doesnt run kvm, but rather, will monitor an already existing kvm guest?
00:28:00 <geist> ah i was thinking 'perf' just ran another app and then watched it's perf counters
00:28:00 <clever> yeah, that is one of its many modes
01:18:00 <Qubasa> clever, geist: Yeah 'perf kvm' is a bit different and watches all running vms and reads the perf counters from them.
01:19:00 <Qubasa> its really cool
01:26:00 <klange> Hm, should I further delay my release schedule and waste time implementing VT-x support...
01:26:00 <vdamewood> klange: As guest or host?
01:27:00 <klange> Host.
01:27:00 <vdamewood> Naw. Version 1.1 feature.
01:27:00 <klange> 2.1
01:27:00 <vdamewood> Okay, 2.1.
01:27:00 <klange> We're at 1.99.7, the release on the roadmap is 2.0 ;)
01:38:00 <klange> Right... ToaruOS 2.0... So this is the roadmap ticket I've had up for the last few months: https://github.com/klange/toaruos/issues/205
01:40:00 <klange> Main release blockers in my mind are the socket improvements and related demos, I keep avoiding AHCI, and I should probably rebuild at least the old Python 3.6 port since 3.7+ are proving annoying with pthread synchronization stuff.
01:42:00 <klange> SMP is stable and I've been running all of my VMs with at least 2 cores. ATAPI driver at least works across QEMU, VBox, and VMware Workstation again and I've set those up to auto-mount the CD. I've got audio support in all three of those now. All three support an e1000-series NIC, and those are all working nicely...
01:43:00 <klange> USB is being pushed back to 2.1. Maybe AHCI should be, too. I should focus on just the socket stuff.
01:43:00 <klange> I really want to have 2.0 out so I can finally say, yes, ToaruOS is 64-bit, and someone can update the Wikipedia article...
01:44:00 <klange> And 1.x was missing all of this network support anyway, so what does it really matter if 2.0's initial release can't do TCP servers... 1.x could barely manage a connection...
01:44:00 <Griwes> it seems that the next thing I need to implement is some sort of a binary tree... I think I'm reaching the point where I have enough kernel-specific data structures that I will need to start actually having unit tests for all the things...
01:52:00 <vdamewood> Ohh... I just thought of a good name for my OS... OkeanOS.
01:59:00 <vdamewood> Well, damn. The domain is taken.
02:02:00 <kazinsal> yeah getting domains these days is a pain
02:02:00 <kazinsal> I ended up with a .net instead of a .com
02:02:00 <eryjus> the-okeanos.com
02:06:00 <vdamewood> I think I'd rather go with a different TLD than add a dash to the domain.
02:11:00 <Oli> okean.net seems free.
02:12:00 <vdamewood> Well, nuts. Okeanos.* is taken for nearly all * except shop.
02:12:00 <vdamewood> Oli: Alas, the OS suffix is part of the word.
02:12:00 <Oli> I now acknowledge so.
02:22:00 <vdamewood> Well, nuts. .io domins are expensive.
02:22:00 <vdamewood> domains*
02:23:00 <Oli> Does okeanos.github.io satisfy to?
02:26:00 <vdamewood> At that point, I'd rather just do okeanos.vdamewood.com
03:08:00 <catern> earlier today someone mentioned as an aside that having the same memory mapped twice has a performance penalty on x86/x86_64 (or maybe just Intel, since they said "Intel")
03:09:00 <catern> this sounds vaguely familiar but I don't remember the cause, can anyone remind me?
03:19:00 <moon-child> vdamewood: okeanos.theoperatingsystem.com
03:19:00 <vdamewood> taken
03:26:00 <moon-child> catern: no idea, but random thought: pseudo 'ssb' on tlb miss, figure out later whether you have permission to store to the given v address or not; so entries are keyed by vaddr. If you make multiple store to the same cache line through different virtual addresses, you have to consolidate them prior to writeback. Maybe?
03:49:00 <klange> chase has returned!
03:51:00 <kazinsal> it's the second coming of the admin
03:51:00 <klange> he says he changed the captcha question for the forum, so hopefully we get a bit less spam now
03:51:00 <klange> I just purged 7 pages of things from the queue...
03:52:00 <moon-child> what was it before, vga textmode addr?
03:52:00 <moon-child> what is it now?
03:52:00 <klange> I don't even remember anymore.
03:54:00 <klange> oh no my thinkpad crashed; looks like it tried to allocate ~16MiB for the kernel heap, which was caught as suspicious and yielded a panic... a useless panic with no further details because only kernel page faults produce the useful ones...
03:55:00 <moon-child> caught as suspicious by what?
03:55:00 <klange> There's just a check >15MiB.
03:57:00 <klange> Previously it was useful for catching places where corrupt sizes got passed in, which was common during initial SMP debugging. This looks "legit" insofar as it appears to have actually been some code making a very large heap allocation, but that's not a good sign in itself.
03:58:00 <klange> Most likely culprit is the network stack, that is doing way more with the heap than it really should and also has a few instances where it believes incoming packets.
03:59:00 <klange> But with no traceback and no debugger for real hardware, no way to know :)
04:03:00 <moon-child> kernel debugger when?
13:24:00 <klange> Little bit of cheating, but looks like I can spin up my Surface with all of its cores pretty nicely: https://cdn.discordapp.com/attachments/783369041661394995/897473404205170728/image0.jpg
13:25:00 <klange> No way to interact with it, though. Supposedly the Type Cover keyboard is just USB HID on the software side (and I've heard it's USB HID on the hardware side... over I2C...)
13:26:00 <klange> And it should be a bog-standard xhci chipset... so maybe with some luck can have the Surface as a viable piece of test hardware.
13:26:00 <klange> The touchscreen is, uh, a very different matter.
17:12:00 <Ermine> Hello, I want to compile gdb with simulator for arm. Should I specify --target for configure script or something else?
18:56:00 <paulusASol> Ermine: According to a quick search for "cross gdb" with my preferred search engine, this (the --target part) seems to be the case.
20:26:00 <Ermine> paulusASol: thank you! Seems like I'm bad at searching