14:50:34 pmooney: About RDMA. Is it possible to pass NIC to bhyve Linux VM with native drivers and setup and use there RDMA? 14:51:25 In theory, if you did PCI-pass-thru for the NIC 14:52:35 I don't personally know of anyone who has tried 14:52:46 And PPT support on AMD has not been wired up, FYI 14:54:11 I have tried PPT passthru for a NIC. 14:54:38 I tried Intel 800 but I couldn't TEST it because the Linux driver for the 800 series REALLY enforces Intel's pick "use our brand of SFP+". 14:55:01 I've passed through igb/I210 ports successfully, but it's been a while so I don't wanna say it'll work 100%. 14:55:42 If you pass a port (say one out of a 4-port card, like I tried with I210) you need to be nore careful in your ppt entries. 14:56:51 I do not recommend per-port passthru on chelsio cards, given how they structure their resource exposition 14:57:19 Good. Thanks! 14:59:45 i thought there was a tunable w/ intel's drivers that allowed you to disable the 'intel sfp' check... but maybe that didn't make it in that driver.. 15:01:31 jbk ==> Linux guest; no idea how to do it. 15:01:45 And from what I can tell the 800/ice driver is a binary download. :vomit: 15:02:13 @pmooney important safety tip about Chelsio,. thanks. Good thing they have a native driver however. 15:04:43 kebe: Not for RDMA, however 15:09:28 [illumos-gate] 16604 Add IPsec policy bypass for netcat -- Dan McDonald 15:10:04 kebe: not in drivers/net/ethernet/intel/ice ? 15:10:26 Ooof, I didn't think to check there, pardon my Linux naivete. Next time I have cycles I should check that. 15:10:43 (Will have to make sure the MNX HQ has it still wired up too, but I think they do.) 15:10:58 pmooney: yeah, not for RDMA on illumos, understood. 15:11:04 i have a local opengrok instance on my home system that includes linux and freebsd as well as illumos 15:11:59 (speaking of which, i probably need to update it since it's been a while) 15:56:19 kebe: Please notify if you have success :) 17:11:24 meanwhile in fbsd land: src: Use gnu17 as the default C standard for userland instead of gnu99 ; sys: Use gnu17 as the default C standard for the kernel 17:38:06 what is it with so called free software nazis? does it really hurt them if I want to watch netflix and chromium or firefox just won't work? 17:40:07 and they'll provide all the framework you need to install steam but make sure it won't fit with the package manager well 17:41:19 yeah, afaik google doesn't make chrome builds for illumos so i have no clue if I would have trouble using such if it were available 17:44:24 the fsf folks are the worst in this grouping, you can't even run guix unless you painstakingly shop for hardware that requires no closed firmware blobs of any sort 17:44:28 is there even a chromium port? I remember some effort at porting in the opensolaris days, but thought it never merged or kept up with the last 15 years worth of changes 17:45:09 alanc, I dunno actually, I'm using X over vnc so no need for a browser 17:45:39 the fsf position on microcode/firmware blobs is just stupid - it's okay to run the binary blob embedded in the hardware but not a bug-fixed version downloaded from the OS 17:46:23 vnc only takes one monitor on my win11 host so I always have a browser with cut/paste to my illumos desktop 17:46:52 alanc, if fsf were actually serious they would not run on AMD or Intel CPU's at ALL 17:47:23 like I'm running a copy of minix on this intel CPU's ME right now and i can't even access it 17:48:25 from the extreme purist POV only RISC-V with a totally open GPU would be acceptable 17:51:54 the steam and chrome parts were very directed at linux in particular nut i see license bigotry in other projects/spaces too 17:52:10 s/nut/but 17:53:39 once upon a time, OpenSPARC was fully open, but that's 15 years dead now 17:55:39 I'm a fan of RV and a member of RV International and support the idea but i don't have a budget for any hardware and it won't stop me from using CPUs that run minix against my will 17:59:32 and I can totally see wantng to use non-GPL when you are making an embedded device and don't want to host a FOSS repo 18:00:09 but almost all consumer routers are linux based and more or less comply with providing the needed repo 18:02:12 someone should pay me real money to build them a consumer mass market router based on bsd or illumos 18:04:13 I have an GPL-bigotry-causing-more-problems story from almost 30 years ago... some of you have heard it. I should tell it again sometime... 18:04:42 @vetal I'm not going to be able to test it out for a bit, alas. 18:05:25 danmcd, this a pro or anti GPL bigotry in this case? 18:05:52 both can just be taken to the point of silly 18:05:58 sun nic driver porting kit?;) 18:06:11 GPL bigot caused more problems because he didn't think it through. He got it straight in the end, but the process took far longer than it should have. 18:06:21 @tsoome NOT the story I was thinking but that IS a good one too. 18:06:28 My story predates my arrival at Sun. 18:09:33 nightly was ok with gnu17 in Makefile.uts. but it does imply gcc14 for primary. anyhow, that gnu17 built kernel does boot ok. os-tests: FAIL: 1 (secflags_dts) 18:11:36 I needed a linux machine and picked one without systemd at random, not devuan because they are debian exiles and license bigots. tried void for this, they provide 32 bit userspace (multilib) and all the hooks for steam (only reason to support 32 bit on linux for the vast majority of users) but do not have an installable steam package 18:12:01 I don't actually want steam on that machine anyway 18:12:17 but just little passive agressive things get annoying 18:15:18 some problems on *bsd and illumos are pretty difficult to debug without seeing the thing actually work on linux where it was developed 18:16:23 working on the vde networking stuff for simh, using opensimh-master because they stopped doing releases years ago 18:17:20 it so far doesn't work on simh on linux for me either so maybe I'm doing it wrong, the docs aren't great 18:21:00 Hello, just a little curiosity, I was looking through the OpenIndiana github and came across its website Astro source https://github.com/OpenIndiana/website and one of the first lines in the README.md is "Minimal styling (make it your own!)". Does this mean that the site is freely available for outside modification and use? I didn't find a license 18:21:00 in the repo, so I'm led to believe it's not freely available 18:21:01 when I get it sorted, I'll make sure it works on both Illumos and dragonfly bsd before I send upstream 18:23:07 this looks like a fun toy otherwise, I can run a vax sim with openvms on one host, a pdp-11 sim with rsts/e on another, and a pdp-10/decsystem 20 with tops-20 on a third and connect them with decnet 18:24:48 building a lab of ancient dec sims networked together may not be for everyone but I'm having fun 18:26:06 if only i could find somewhere original emacs written in teco it would be full on bliss 18:26:26 wait, bliss is another DEC language 18:28:20 https://en.wikipedia.org/wiki/BLISS 18:31:28 AIUI most of VMS was written in BLISS (like how most of S/360 & beyond was written in PL/I or derivatives). 18:32:06 (One of my grad-school friends had worked at DEC and told me as much.) 18:32:13 nifty 18:32:41 I found source tapes for rsts/e but haven't found same for VMS 18:33:24 I'm going to build a primos sim too, there's source for primos 19 out there but not 22 which i thnk was the final one 18:33:36 i think it's largely in PL/M 18:35:45 there is a dump of su-tac, a live major tops-20 site before it was shutdown though 18:37:56 I mess with PLATO some too fwiw, that's based on a CDC Cyber machine 18:39:55 I need to do some paying work, I kinda really want a vaxstation 4000 to show off to my grandson :) 18:42:01 I would play with old IBM stuff too but they still sell VM/CMS afaik and certainly don't have classic OSes available on the www 18:46:17 danmcd: guy from DEC I worked with at Project Athena had a few BLISS war stories. one of the language quirks was that identifiers were always memory addresses and you had to always dereference them with the "." operator to fetch the value. so "i = .i+1". He wrote some code. It didn't work. He tried to debug it but the debugger couldn't find the function he wrote. what had happened: only one call site. He left out a '.' on its parameter, 18:46:17 which meant that the parameter was a compile-time constant (think address of i instead of value of i), and the compiler computed the value at compile time and made the function body go away... 18:46:51 Wow wow wow! 18:47:23 rolf 18:47:30 rofl even 18:47:35 If (and it's a big IF) I wanted any vintage IBM kit, it'd be an AS/400 from 1990, one of the small ones. Would flash me back to my two summers in Rochester. 18:48:14 (Honestly, that stuff still weirds me out, plus I don't know COBOL or RPG, plus I barely grokked the TWO PL/I derivatives AS/400 used for low-level things.) 18:48:39 the vaxsation is small, I could never afford to run a huge old power hungry box and don't have the space 18:48:43 Speaking of PL/I, the one bit of emulated retrocomputing I've set up is a multics (also under a simh fork). Only they called it PL/1 ... 18:49:40 * m1ari wishes he'd had more space, time and knowledge when younger. I had a vaxstation rescued from a skip at Uni but never knew enough to do anything with it and it was too big to hang onto with all the other things I had at the time. 18:49:41 sommerfeld, I played with multics a bit too, out of curiousity. primos, vms, rsts/e and rsx-11 I all actually did paid work with 18:49:51 Heh heh... yeah. IIRC Corbato & friends published first papers on using higher-level languages for an OS, even if S/360 came first. 18:50:38 dangergrrl: There were AS/400 machines that were comparable to the Microvax, I think. In slightly-future-from-then terms, something the size of a Sun Fire 480. 18:51:03 had a canadian friend who got a sweet deal on a cray y/mp but he declared it as $1 value and customs canada took a hammer to the boards to make sure it was only worth $1 18:51:36 danmcd, without publicly available OS images I can't see too much use 18:51:45 Agreed. 18:52:10 I have a full set of license keys for *everything* for VMS 6.2 I found on a russian site with a keygen 18:53:05 AS/400 and its descendants had a "Machine Interface" it presented to higher-level OS primitives, where you used one PL/I derivative (don't know if I can even mention its name?!?), and the implementation of the MI itself, where you used a different one. 18:53:30 with the apropriate adapter I could probably retrofit a vaxstation with an SSD :) 18:53:46 Later in the 1990s the actual hardware swapped out from custom to POWER. These days AS/400-descendants and RS/6000-descendants are the same HW. 18:54:27 I have not so fond memories of 1990s AIX, worked at IBM twice 18:57:41 I used some box running VM/CMS in high school, we had some one day a week program where we went to IBM for that day every week 18:57:42 my cousin would refer to AIX as 'Ain't unIX' :P 18:58:10 jbk, I couldn't do any admin type stuff on AIX without smit 18:58:22 the commands just don't make sense in a unix way 18:58:39 my experience with it was with the SP/2 cluster Sprint PCS had for their CDR (call detail record) archive 18:59:03 but even that, most of the memories were of swapping out 10-12 disks/day on the attached storage 18:59:12 because of oil contamination 19:00:01 (that and the time an electrical contractor mistook the EPO for a light switch in the DC where it was running, and it took like 6 of us to go through, flip all the switched on everything so it could be properly sequenced on power on) 19:01:32 Sprint PCS makes sense, I interviewed for some bellcore job with AIX and worked on the NSFnet backbone project for IBM/Advantis 19:02:07 it was the only AIX there at the time 19:02:47 aside from the S/390 'IT'S NOT A MAINFRAME!!!!!' that came in later 19:03:28 err I should say only IBM 19:04:18 hard to believe that in 1994 the nsfnet backbone went from half-t3 to full-t3 links because IBM got the routers up to that speed (only 54mbps) 19:04:18 (basically we were told to never refer to it as a mainframe because it would make some VPs and the CIO look bad for various reasons) 19:05:41 jbk: oof. parallels in my experience: the RA81 issue (DEC winchester drive, ~400MB, maybe 5 or 6RU high). bad materials used for sealant -> crumbs on the disk platters..); Gary (our Dec Field Service guy) had to replace about 100 of them when the fixed version came out. And the time an electrician bumped the EPO button with a ladder and it took a large committee of electricians to figure out how to turn the power back on again. 19:06:31 the same electrician also was caught standing on one of the cabinets of disk (or I guess you could call the DASD since it's IBM) 19:07:19 my recollection is that the oil or whatever could end up forming a wave on the surface of the platter and if the head ran into it... dead drive 19:10:55 oh, there's people! :) 19:12:44 so... I bought an ft232 board and am ready mostly to try: https://www.vinnie.work/blog/2020-11-06-baremetal-rpi4-setup/ 19:14:04 mine doesn't have headers on the side pins so i'll have to figure that out but the board was only $6 and I already used it to unbrick a linksys wrt1900acs so if i fail at JTAG it was worth it anyway 19:15:17 I have a laptop running OI, I presume I can make a omnios vm with virtualbox and use usb pass through 19:16:44 I'm wanting to try to help with illumos on arm and I think i have what I need for hardware debugging but I've not looked at the initialization code for illumos so I'm not sure where to start 19:17:19 trying to come up with a better bug report than the screen on the pi flashes twice or so and 5 seconds later the bootloader restarts 19:22:26 I have to move equipment around to do this so I want to be prepared. I should start with omni on arm in qemu and get a look at how that works when it's working? 20:51:40 smb/cifs is slightly tricky with //hostname/path\with\backslashes 20:52:39 use \\with\\backslashes ? 20:55:20 in the shell, in fstab no need. none of the examples I found show share names with \, just //hostname/share 20:56:09 again, i got it to work on linux and probably can get it to work on BSD and illumos based on that 20:57:04 had no clue what i was doing wrong, the path seperator starts out as / like normal posix then switches to \ because parsed by the windows fileserver 21:00:00 my ft232 board didn't come with headers for the side pins (needed for jtag) and I didn't know what pitch to order (2.5mm?) so i may have to order one or jerry rig it 21:01:03 It should be able to connect to the UART pins on the big header though, and that's where the serial console is for the omnios ARM images. 21:01:36 andyf: which ARM board do you recommend for trying OmniOS ARM images? 21:01:53 andyf, it's not booting that far for me, 2 or so screen flashes and back to the pi bootloader after a few secs 21:02:09 otis - a raspberry pi 4 is what I use. 21:02:23 dangergrrl - screen? As in via HDMI or something? 21:02:32 thanks. 21:02:36 if a serial console would help I can wire it up though 21:02:43 We definitely don't have drivers for that - everything at present is serial console 21:02:48 yes, the hdmi screen 21:02:50 aha 21:02:52 so 21:03:05 i should get some diags on the serial console 21:03:09 I can do 21:04:39 i figured talking to the gpu would be a pain, i've been shopping for risc-v boards (that i can't afford just yet) and most have closed GPUs which defeats my intent 21:07:56 but if the gpu doesn't work at all this will be easier, i can bring the pi in here where i don't have a spare hdmi monitor nearby 21:08:41 This is my setup https://paste.omnios.org/?906e29d5a1a662be#DqY7qfV6UkhZPx6u1EnhAGFL2W6UbopjVeVGcQNSfupG 21:10:21 is that an ft232 just with female to female wires/connectors 21:10:32 i have that 21:12:42 Yes - that one was designed to go straight onto the header, but one with individual dupont cables should work fine. 21:14:54 okay, i have the stuff then, I'll bring the pi in here later and just put it back when I'm done. it's normally hooked to the TV running retropie 21:18:08 several things just got less troublesome when i figured out the slash oddness with cifs 21:18:51 scp works but I end up with copies of the same file a bunch of places 22:08:49 https://www.digikey.com/en/maker/tutorials/2023/how-to-connect-to-a-headless-raspberry-pi-using-serial 22:10:48 andyf, looking at that, I'm not powering the pi so won't hook up 5v but the rx/tx lines, do i use 3.3 or 5v? 22:16:22 nm, google knew 22:16:48 3v3? 22:17:24 https://www.mathworks.com/help/matlab/supportpkg/use-the-serial-interface-on-raspberry-pi-hardware.html#:~:text=Raspberry%20Pi%20hardware%20uses%20%2B3.3,pin%20on%20the%20serial%20device. that says 3 22:18:31 sorry, that was me being lazy :) i have all the docs i need and the stuff 22:21:32 I'm signing off for the night. Good luck! 22:24:08 thanks, tomorrow I should at least have seen boot message spillage