00:32:51 What does ada1 mean? 00:40:02 after running freebsd-update, my system is booting from /boot/kernel/INS@n0RyGu instead of /boot/kernel/kernel. what is a safe way to correct this situation? 00:47:56 mdekstrand: that's uniquely bizarre 00:48:17 no wonder I can't find any helpful search results 00:48:23 anything special in loader.conf? 00:48:25 nope 00:48:41 it looks like the INS@ file results from an interrupted 'install -S' 00:49:10 but i'm not sure what's telling it to look at that instead of 'kernel', so I'm hesitant to just delete the file & reboot 00:49:24 how is loader finding it, though? Is this forthloader or lualoader? 00:50:44 kenv script.lang should answer that 00:53:11 lualoader basically only uses literal kernel as the name 00:53:42 there's no way we'd even try a kernel there unless kernel= was explicitly set in loader.conf 00:56:09 lua 00:56:36 at least that's what script.lang reports 00:57:22 the only file that seems to reference this INS@ file is kernel/linker.hints 00:57:58 what does "ada1" mean? 00:58:21 !man ada 00:58:22 ada(4) - ATA Direct Access device driver https://www.freebsd.org/cgi/man.cgi?query=ada 00:58:29 device 1 of an ada driver 00:58:40 starting at device 0 01:02:21 does anyone know why they removed twa from -current? the commit message only said that it was supposed to happen before release. 01:03:35 link to commit 01:04:05 and sX is a msdos partition? does it use the same numeration as linux? 01:04:19 ? 01:04:27 dksnd: s is usually 'slice' 01:04:33 effectively partition 01:05:33 mdekstrand: does /boot/kernel/kernel load and boot if you do it manually? 01:05:55 so is it 1..4 for each partition except the extended one where it is 5+? 01:05:56 mdekstrand: if so, I'd nuke the INS one and kldxref /boot/kernel/kernel, which should fix it 01:06:41 @kevans have not tried (this machine is headless... I can dig around and plugin in a keyboard and mouse and try) 01:06:55 or just delete the file and roll the dice and dig it out if it breaks 01:07:25 mdekstrand: if kernel file is there and a reasonable size I'd go that route 01:07:41 also does freebsd use avx? 01:07:47 kldxref -d and see if it complains about the kernel first if you want, to be sure 01:10:33 kldxref -d seems ok, so here goes nothing 01:15:13 grub has suddently stopped booting freebsd via kfreebsd 01:15:42 @kevans delted, kldxrefed, and it booted and now reports /boot/kernel/kernel. 01:15:44 thanks! 01:16:01 I thought the issue could be with "mountfrom" but it looks correct 01:25:18 grub finds the partition at (hd1,msdos11) so it should correspond to ada1s11, but it doesn't seem to work 01:29:42 mdekstrand: awesome, happy booting 01:40:08 also is it normal not to have /boot/loader? 02:00:55 That would be ... abnormal 02:02:26 On stable/13 that is linked to /boot/loader_lua 02:09:46 on x86 02:19:06 Right; should have mentioned the platform of amd64 in my response 02:45:39 does freensd have /home? 02:47:33 I have loader_lua but no loader 02:48:27 are the freebsd partitions bootable? 02:57:48 On recent fresh installs (in last ~2 years) I found /home to be a symbolic link to /usr/home 03:05:59 anyone know what is the replacement for -DNO_CLEAN for buildworld/buildkernel ? 03:08:07 WITHOUT_CLEAN=yes 03:08:15 nice thanks 04:14:43 koobs: https://reviews.freebsd.org/D33114 04:14:44 Title: ⚙ D33114 twa: Remove 08:20:52 There is dead link to "FreeBSD Journal" -- http:// freebsdjournal .cc (redirects to freebsd .com, available for sale) -- in "Community" menu on freebsd.org web site. 08:28:30 is a freebsd partition bootable? 09:02:08 Hi 09:02:16 how to delete a route permanently ? 09:02:33 i have a vps with a nginx jail, with rdr redirection through pf 09:03:02 i have a route on the vps which redirect all packets to the public ip, through lo interface 09:03:20 problem is then that acme-tiny can't validate my challenges through internet 09:03:29 because the packet is routed locally, and not over internet 09:03:51 i need to tell to freebsd to remove that route, permanently 09:03:52 "194.163.181.239 link#1 UHS lo0" 10:06:01 I should set up a jail for something or other 10:31:58 is a freebsd partition bootable? 10:32:33 it depends on setup 10:32:42 and which partition.... 10:52:08 dksnd, on what hardware platform? x86 standard BIOS, x86 EFI, ARM, SPARC, what? 10:52:33 x86 bios 11:27:10 dksnd: https://docs.freebsd.org/en/books/handbook/boot/ 11:27:13 Title: Chapter 13. The FreeBSD Booting Process | FreeBSD Documentation Portal 11:27:45 see stage one and stage two 11:29:06 the commands you would often use these days are a bit different though 11:29:55 but for a ZFS BIOS system you have a small partition of type freebsd-boot that serves strictly to host zfsboot/gptzfsboot 11:30:10 so how the boot1 is maintained? is there a hook that installs it during the system update? 11:30:15 and then there is pmbr 11:30:47 dksnd: FreeBSD writes /boot/pmbr if you do not use the boot manager 11:30:55 to the first sector of the disk 11:31:04 that is if your disk is GPT 11:31:13 pmbr is protective MBR 11:31:31 I do not remember how it works for non-GPT disks, sorry 11:33:03 and with UEFI there is no such thing as a ‘bootable partition’, there is ESP 11:33:35 it is a mbr disk 11:33:37 speaking of partitions you need to keep in mind that in FreeBSD on non-GPT disks you deal with slices first 11:34:13 the DOS primary/logical partitions are slices to FreeBSD 11:34:29 adXsY 11:35:05 while with GPT disks you have partitions, adXpY 11:35:23 let's see if chainloading works 11:35:28 the FreeBSD slice is usually bootable as you install the bootcode into that 11:36:01 I think it was also possible to just bsdlabel the raw disk without using a partition table? but never tried 12:35:05 nope it does not boot 12:35:32 the first sector has invalid signature 12:36:14 chainloading /boot/boot0 prints a menu that has the other disk, pxe, and only prints hashes 12:36:35 chainloading /boot/boot1 just prints "boot error" 12:36:47 loading the kernel simply reboots 12:44:56 someone have issue with packet loss in a jail (not vnet) ? 13:23:01 dksnd: if the boot manager does not show a FreeBSD option it means there is no bootable slice/partition present 13:29:52 what can stop the partition from being bootable? how can I tell it WAS bootable in the first place? 13:30:42 is there a way to boot it anyway? 14:13:19 does freebsd partition know where it is located? 14:13:54 if the partition structure is changed, should freebsd be reinstalled? 14:19:58 you should probably boot a USB stick or a CD with FreeBSD on board and try to figure out your geometry 14:25:36 bsd partition is located at disk 1 partition 11 (which should be ada1s11) but kernel does not want boot from there and just reboots the machine 14:27:41 are there some loader.conf parameters to make the kernel print something and do not reboot? 14:39:43 boot0 is the 510 byte hand-written assembly an x86 BIOS is supposed to jump to and execute, as described on https://docs.freebsd.org/en/books/arch-handbook/boot/ 14:39:44 Title: Chapter 1. Bootstrapping and Kernel Initialization | FreeBSD Documentation Portal 14:40:15 https://cgit.freebsd.org/src/tree/stand/i386/boot0/boot0.S is the actual code 14:40:16 Title: boot0.S « boot0 « i386 « stand - src - FreeBSD source tree 14:40:30 it's bonkers complex for being hand-rolled assembly 14:42:13 I don't need boot0, I need the boot record of the ufs partition 14:42:35 it's first 512 bytes are not formatted as a boot record 14:43:02 that's stored in the vfs.root.mountfrom kernel environment variable, accessible at runtime with kenv(8) 14:43:34 in the loader, you can see it with 'show' 14:43:46 in the freebsd loader, i mean 14:43:50 I don't use freebsd loader 14:43:55 then i can't help you. 14:45:47 gpart(8) can also install bootcode if you're in freebsd and know where you want to install it, but it'll overwrite anything that's there if you tell it to 14:45:58 see the 'bootcode' subcommand 14:46:02 if the vfs.root.mountfrom is incorrect, would the kernel print something? 14:46:59 yes, it'd ask you whether you want to load /rescue/sh 14:47:11 that's the rescue shell, obviously 14:47:31 you might wanna read the BOOTSTRAPPING subheader in gpart(8) too 14:47:36 but how does it know where / is? 14:48:35 well if you've gotten past kernel initialization then you'll have to have some kind of filesystem to load the kernel from 14:49:08 freebsds loader understands the filesystems (both ufs and zfs) enough to not need an ram disk like other unix-likes 14:49:59 the kernel and loader configuration are both stored on the filesystem that you're booting from 14:50:27 so any third-party loader would have to understand ufs and zfs equally well 14:50:51 it's done this way to avoid complexities when maintaining kernels by having to copy files back and forth onto the ramdisk 14:51:23 the installkernel target for make in /usr/src autommatically makes a copy of the working kernel if the one you're building is different 14:52:21 also, are you talking about hardware partitions or bsd labels (software partitioning)? 14:52:37 or some other form of software partitioning? 14:53:22 i'm thinking it's the last one, since hardware doesn't support up to 11 partitions and you probably aren't using bsd labels if you're just installing freebsd? 14:54:10 partition 11 is the 7th logical disk in the extended partition 14:54:14 but you're also chainloading which implies that your boot manager loads boot0 14:54:44 Oh, EBR.. It's been too long for me to remember that. 14:54:53 I used grub that booted the kernel directly but it is broken right now 14:54:54 it _is_ supported 14:55:10 what's broken? 14:55:14 now I'm trying to bootstrap it 14:55:26 booting via grub 14:55:44 booting freebsd via grub or booting grub? 14:56:00 booting freebsd via grub, yes 14:56:08 well, what changed? 14:56:29 dksnd: If it's helpful, https://wiki.freebsd.org/MasonLoringBliss/BootingFreeBSDfromUEFIGRUB 14:56:30 Title: MasonLoringBliss/BootingFreeBSDfromUEFIGRUB - FreeBSD Wiki 14:56:59 mason: this is an x86 BIOS setup, as far as I've understood 14:57:02 I have no idea it used to work on april 21st 14:57:23 debdrup: I think the GRUB parts will be pretty similar. 14:57:35 mason: maybe, I don't know nothing about no grub. 14:57:51 That TLC reference didn't quite work out as well as I wanted. :/ 14:57:54 there could be grub updates... but it should not break the setup 14:57:54 debdrup: GRUB config doesn't really vary whether it's UEFI or legacy - there's a little stub grub config in the ESP, but you can ignore that for this purpose. 14:58:08 heh 14:58:48 dksnd: sorry, I'm in over my head here. I used Linux for a week in 2000 before getting frustrated with the lack of documentation, and haven't touched anything but FreeBSD (except in anger) since. 14:58:50 dksnd: Anyway, even with legacy, this is something that you'd invoke from the level of GRUB that gives you your boot menus. 14:59:25 dksnd: the easiest way might be to figure out what changed, though 14:59:56 dksnd: This shows a legacy example. Much the same: https://wiki.archlinux.org/title/GRUB_Legacy#chainloader_and_configfile 14:59:57 Title: GRUB Legacy - ArchWiki 15:00:35 In this case GRUB's using the legacy specifier (hdwhatever) to point to the partition with the bootloader, and then "chainloader +1" to leap into that new root. 15:00:56 freebsd partition does not have a valid boot secrot 15:01:00 That said, I don't use much legacy any more, so this would take some experimenting. 15:01:11 dksnd: Well. Reinstall one from rescue media maybe. 15:01:15 gpart bootcode whatever 15:01:53 /sbin/gpart bootcode -b /boot/pmbr your-correct-FreeBSD-boot-partition-or-disk 15:01:59 If I chainload directly into /boot/boot1 or boot0, they just give errors 15:02:07 pmbr is for efi 15:02:36 Well. Whatever the right one is. That was the first bootcode EXAMPLE from the gpart manpage. 15:03:29 Maybe something from the following MBR section then. 15:03:43 the /boot/boot2 also doesn't look chainloadable, as it starts with 512 zero bytes 15:04:33 I could also try /boot/boot 15:05:08 I find the fact that no one knows how their OS boot quite disconcerning 15:05:11 dksnd: I guess. But putting that in place with gpart bootcode is probably better. Maybe functionally equivalent I guess. 15:05:31 dksnd: Well. I mean, of everyone here, you're the one not booting. Just saying. 15:06:16 dksnd: Also, verify your assumptions. pmbr is for GPT, but it isn't for UEFI, it's just for GPT. 15:07:42 For instance, legacy booting here: gpart bootcode -b /boot/pmbr -p /boot/gptzfsboot -i 1 ada0 16:00:58 Assuming that there's a single FreeBSD partition is probably the first error. 16:01:02 Welp, seems they left. 16:03:25 There are lots of variables and there's no one true way to do any of this, especially with the multibooting. 16:25:20 Anyone seen this before? "bhyve: vm_setup_pptdev_msix: No space left on device" Got it while trying to pass 4 pcie Intel ice network adapters through to a bhyve VM. Passthrough of 3 NICs works fine, but 4 or more fails. 17:13:01 What kind of hardware do I need to use the `ow` device on a PC? Would it easier just to allocate an old r-pi to interface with ow? 17:16:49 The gpiobus(4) man page mentions arm/mips/ppc drivers, but nothing for, say, amd64.. 17:18:28 could anybody tell me if there was an answer to my morning question ? 17:18:34 It would be nice if I could hijack a pin on my parallel or RS232 port. 17:18:40 and kindly paste me the answer ? 17:19:02 eoli3n_: I wasn't here this morning. Might help if you asked it again. 17:20:26 why freebsd populate a route for the main if IP to route locally ? linux seems to not do this. Is that a route push by my vps provider or something that freebsd set up automatically ? 17:20:27 taht route 17:20:52 194.163.181.239 194.163.128.1 UGHS lo0 17:20:57 no 17:20:58 Is this an IP your machine learns via DHCP? 17:20:59 wrong line 17:21:03 ghoti nop 17:21:18 194.163.181.239 link#1 UGHS lo0 17:21:24 static ip 17:21:31 that route breaks things 17:21:39 i want to delete it, but i don't know how to persist deletion 17:21:50 static_route in rc.conf allow only adding route 17:21:57 not "change" nor "delete" 17:23:13 eoli3n_: if the gateway is "link#N" then it is an IP assigned to the interface, not routed. Try `ifconfig lo- 194.163.181.239 -alias`. 17:23:14 'S' means 'manually added', iirc 17:23:18 lo0 17:23:43 i forged the line 17:23:51 let me reboot to give you the correct one 17:24:56 ghoti i didn't get what you mean 17:25:04 what does that ifconfig line do ? 17:25:39 this is the line : 194.163.181.239 link#1 UHS lo0 17:25:42 eoli3n_: it removes the IP from the interface. Best to confirm it really is there first, of course, Run `ifconfig lo0` and see if you see it. 17:26:17 why do you want that I remove the ip from the interface ? 17:26:43 ifconfig lo0, to see if I see what ? 17:26:47 sorry I don't follow you 17:26:48 You wanted it out of your routing table. All local IPs are implicitly in your routing table. That's how to remove it. 17:27:19 here local IP si explictly 17:27:24 s/si/is 17:27:45 lo0 doesn't have ip 194.163.181.239 17:27:47 Okay, you said this morning that this was an IP on a jail, did you not? 17:27:57 i don't follow you at all 17:28:08 i'm just asking for a way to delete the route 17:28:18 `route delete ` then. 17:28:19 why would you want to remove an public IP from lo1 ? 17:28:26 persistently 17:28:32 route delete ip, reboot, come again 17:28:44 What do you mean by "persistently"? What is re-adding it after you reboot? 17:28:54 eoli3n_: this is the line : 194.163.181.239 link#1 UHS lo0 17:29:02 after each reboot, the route come back 17:29:10 what sets up that route ? 17:29:10 Is it mentioned in /etc/rc.conf? Something in one of your /usr/local/etc/rc.d/ scripts? 17:29:17 lets check 17:30:12 in /etc/rc.conf i only set : defaultrouter=194.163.128.1 17:30:25 grep -ri route /usr/local/etc/rc.d returns nothing 17:30:52 it seems that this route comes by default on a freebsd install 17:30:58 but not on linux nor windows 17:31:02 What about a recursive grep for the IP in /etc and /usr/local/etc ? 17:31:27 what are you searching for ? 17:31:35 The answer to your question. 17:31:38 And mine 17:31:40 ok ok 17:31:43 13:28 < ghoti> What do you mean by "persistently"? What is re-adding it after you reboot? 17:31:48 13:29 < eoli3n_> what sets up that route ? 17:32:03 in etc i only foudn the line which set up the interface 17:32:13 in /usr/local/etc nothing 17:32:23 /etc/rc.conf:ifconfig_eth0='194.163.181.239 netmask 255.255.192.0' 17:32:24 Might as well check /var/log as well 17:32:36 and there's nothing like blocklisting software that adds stupid routes via lo0? :) 17:32:47 pf ? 17:32:55 Ah, well, there you go. It is the main IP address on your eth0 interface. 17:33:13 why lo0 though? 17:33:17 it is yes, that's what i'm trying to say since the beginning 17:33:23 why lo0 what ? 17:34:15 yuripv sorry, didn't get here again 17:34:17 Why would you send traffic that is local out an interface? 17:34:27 stop 17:34:29 let's restart 17:34:34 we are going nowhere here 17:34:35 . 17:34:38 So 17:34:40 It is literally the local machine, no? 17:34:49 i really meant stop 17:34:52 oh right, it's same here for my local ip :) 17:34:54 let me reexplain the whole thing 17:35:11 yuripv yes it is, for all freebsd install 17:35:20 and that's what i'm trying to understand 17:35:25 that route is illogical 17:35:33 and it came by default on all your installations too 17:35:41 so how to remove it persistently 17:36:37 the problem is that i'm on a VPS, and that route breaks my acme-tiny challenge verification from a nginx jail 17:36:52 because the route catch all packets to my eoli3n.eu.org public IP 17:36:53 Why would you want to remove it? Where else should a host send traffic destined for itself? 17:37:01 What kind of a VPS are you using? 17:37:02 and that public IP is locally routed by default 17:37:05 contabo 17:37:08 Does it virtualize a network stack? 17:37:13 yes, proxmox 17:37:49 "ghoti: Why would you want to remove it?" : because i want to reach my public IP from internet, to validate my ssl certificate 17:37:57 i don't want to route locally for a public ip 17:38:06 it breaks acme-tiny 17:38:54 my question is, why does that route exist anyway ? 17:39:02 no, that's not really my question 17:39:08 Reaching the public Internet requires a default route. Traffic to yourself is generated from an IP that matches the subnet of the closest interface, the route should be immaterial. 17:39:13 my question is more "why linux/windows doesn't add this route ?" 17:39:33 i have a default route 17:39:50 what means immaterial ? 17:39:53 not english native here 17:39:56 unimportant. 17:40:04 but it is created by default 17:40:12 let me paste all my routing table 17:40:22 Because traffic from your machine to your machine does not need to leave your machine. 17:40:27 https://0x0.st/o3RU.txt 17:40:38 yes, that answer why freebsd creates it 17:40:43 but it creates problem too 17:40:55 Yes, that looks perfectly normal. 17:41:05 the route seems logical, but not on a vps, when using a public ip directly on the main interface 17:41:07 Are you saying the outside world can't reach your machine? 17:41:12 i'm not 17:41:16 This is really an X-Y problem? 17:41:27 did you understand my main problem ? 17:41:38 do you know what is acme-tiny and how it works ? 17:42:01 i can explain from the begenning 17:42:04 beginning 17:42:05 Apparently not. How does running in a VPS affect anything? Your outside interface is bridged with the host's. Your localhost is emulated. 17:42:29 because, the public ip is configured directly on the interface, and my dns solve my domain name by this IP 17:42:53 so when the ssl certifacation process try to reach eoli3n.eu.org from the jail, passing through internet, the route catch it, and run it locally 17:43:02 so, this is breaking 17:43:15 because I need to prove that i own eoli3n.eu.org 17:43:17 Sure, you are renewing your lets-encrypt certs. But the problem does not make sense to me. There is no reason for it to do that as far as I can see. 17:43:18 from internet 17:43:24 Oh you do? 17:43:38 i don't get your question 17:43:48 i autoamted ssl renewing from the jail directly 17:43:56 So, is it nginx that is answering the door when someone rings at port 443? 17:44:02 port 80 17:44:17 You are using certs on port 80? 17:44:34 Or are you just using this for domain verification? 17:44:35 try this at home : https://0x0.st/o3RU.txt 17:44:44 this is how acme-tiny works 17:45:00 check this : https://github.com/diafygi/acme-tiny#step-3-make-your-website-host-challenge-files 17:45:01 Title: GitHub - diafygi/acme-tiny: A tiny script to issue and renew TLS certs from Let's Encrypt 17:45:15 my challenge is reachable from outside 17:45:20 but not from the VPS 17:45:22 because of the route 17:45:40 which is sad, because the only purpose of this, is to be reachable FOR the vps FROM internet 17:46:30 I haven't run acme-tiny in a few years, but I never had problems like what you are describing. I was running from jails. 17:46:34 wrong paste : 17:46:38 try THIS at home : curl http://eoli3n.eu.org/.well-known/acme-challenge/GGsMeFaRa16F1tSGBuQY9C7G9q2-5v_yvYtKisYRVSs 17:47:12 if you use a hosting server which have an interface configured with a public ip, and that route, you can't use acme-tiny 17:47:16 or i'm missing something 17:47:19 but here i am 17:48:06 So.. You say you want the IP accessible from the outside, but your problem is that you can't reach it from the inside? 17:48:33 I still don't see the problem. nginx binds to the port on the IP. How traffic gets there should not matter. 17:48:33 i want my jail -> gateway -> public ip -> rdr -> nginx 17:48:55 ghoti : if i remove the route, acme-tiny rework 17:49:13 if i change the route to pass throught my default gateway : acme-tiny rework 17:49:32 i explained 3 times why it is breaking, i don't know precisly how acme-tiny work 17:49:37 but the route IS the problem 17:49:44 or acme-tiny is 17:49:50 but i mean that both can't work together 17:50:41 So, are you saying that acme-tiny can't reach the IP, or gets something unexpected when it connects? 17:51:13 I think changing acme-tiny, if that's required, is probably safer than breaking your network. 17:52:14 removign the route is not breaking the network 17:52:19 I haven't experienced the problem you describe, and from what I know of the two parts here, it should not occur. So I'm clearly missing a part of your situation. 17:52:24 that route only exist on freebsd 17:52:31 maybe other bsd distro too 17:52:45 ghoti i tried my best to explain 17:52:50 i need a pause :) 17:52:57 thanks for you attention, anyway 17:53:00 and your help 17:53:19 did you read acme-tiny readme ? 17:53:21 the link i gave 17:53:28 if you don't get that, you will not get my problem 17:53:42 that one -> check this : https://github.com/diafygi/acme-tiny#step-3-make-your-website-host-challenge-files 17:53:43 Title: GitHub - diafygi/acme-tiny: A tiny script to issue and renew TLS certs from Let's Encrypt 17:56:26 fwiw, the route exists in macos as well, but that is expected. I have a vps with haproxy answering public queries, and it has no problem reaching itself. But my setup is a little different from yours. 17:56:55 The link you gave was not a readme, but I have read it before. I will refresh though. 17:57:31 Ok, the step-3 link was fine. 17:57:32 it is a readme .. why do you say it is not ? 17:57:36 ah :) 17:57:56 I clicked the acme-challenge URL without reading it closely enough. :) 17:57:59 xse, i copy/paste the backlog here if you have the energy 17:58:49 https://0x0.st/o37Q.txt 17:58:52 xse ^ 17:59:56 I have haproxy in one jail and the web server in another jail. The jails have private IPs, and I do not use NAT. Inbound access to the haproxy jail is via a reverse-nat firewall rule, and outbound queries from both are through tinyproxy, so I always know what is happening. 18:00:58 Why do I need the chat log? 18:02:51 ghoti not for you 18:02:55 for xse 18:03:19 ah 18:08:33 I have not touched jails for quite some time but since deleting it seems to fix it for you i'd either: 18:08:39 1) script smth that deletes it every time the jail is started 18:09:42 or 2) skip the whole .well-known thing and use dns challenge via rfc-smth-smth 18:11:57 s/-smth-smth/2136 18:12:35 NerdyMcNerdface: if you're looking to virtualize lots of NICs, you want to get a NIC and a motherboard that supports PCI SR-IOV as that lets you assign virtual functions via iovctl(8). 18:15:31 Alternatively, if that's not an option, you're limited to using vNICs via netgraph as described on https://klarasystems.com/articles/using-netgraph-for-freebsds-bhyve-networking/ if you want proper network virtualization, vale (via netmap) if you need fast interconnects, or a simple software bridge(4) with if_vtnet or the like. 18:15:32 Title: Using Netgraph for FreeBSD's Bhyve Networking - Klara Inc. 18:46:24 here is the fix https://github.com/diafygi/acme-tiny/blob/master/acme_tiny.py#L187 18:46:25 Title: acme-tiny/acme_tiny.py at master · diafygi/acme-tiny · GitHub 19:14:28 https://lists.freebsd.org/archives/freebsd-current/2022-October/002732.html 19:14:29 Title: morse(6) sound 19:15:06 I wonder if you could use… eject to make sounds with your cdrom drive 19:15:26 remember cdrom drives? and cdroms?? 19:24:55 meena: i doubt it 19:57:32 debdrup: thanks! Will look into iovctl and see if passthrough of virtual functions works better, and then vale/netmap and netgraph if that fails. 19:58:49 In my use case I have just 1 VM using each physical NIC, so PCI passthrough of the whole physical interface to just 1 VM is fine. But when one thing is broken I guess I'll have to iterate to the next. ;) 20:24:51 SR-IOV is by far the best option as it just gives you a bunch of what appears to the OS as NICs, that you only have to ensure that the host doesn't try to do anything with. 20:25:11 Netmap is for +50Gbps interconnects, mind you. It isn't really worth it if you're not using those speeds. 20:25:52 So, short of special-purpose hardware where SR-IOV is validated (which it isn't for a lot of gear, even today), netgraph is always the best option. 20:26:27 As several people in here can testify to, just because a vendor claims that SR-IOV is supported, it doesn't mean it is. 20:28:41 the VMs are going to be allocated a mix of 2x25gbps and 4x25gbps, so performance is definitely a plus 20:29:36 Netgraph should be able to do 25Gbps 20:30:41 I should probably add that it depends on the CPU speed, as does vale/netmap. 20:31:10 vale is made to go _exceptionally_ fast, though - back when I used it in production at a campus network, it was doing +70Gbps 20:32:03 nice, I think both netgraph and vale will be performant enough 20:33:09 I've got a stack of mellanox/nvidia connectx-4 and intel E810 cards for testing + a couple of dual socket xeon e5 2680 v4 servers 20:33:59 maybe I'll beat your 70 Gbps record ;) 21:15:18 koobs: You a dev? 21:15:37 jb1277976: depends on definition 21:15:50 /dev/koobs 21:15:57 jb1277976: and depends on what youre trying to ascertain about me :) 21:16:05 i am not a device driver no 21:16:07 :] 21:16:15