ptribble: I actually run a SunRay platform at home for the family. I also got one of the sunray compatible "laptops", so we can sit in the garden and use wifi with it.

i still have some sunray2's at home, not using them. one of them is even dualhead (two DVI outputs)

i've used them at my ex-gf company (where there were sunrays+country kit on desks, with one server. sunray software ran on linux)

I have on every TV a sunray, the kids have a SunRay2+, a SunRay3, another SunRay3 is in the living room, a SunRay2+ sits in the server-rack, we have the laptop and I have the SunRay3i with the embedded display in my office

the server runs in a KVM on OI (will move it to bhyve in the next weeks) and the kids use it for homework or during covid for remote learning

i'm moving to a new house next week, i might also get sunrays working again.

it was especially useful when my kids were younger and couldn't remember strong passwords. the smartcard makes it nice and easy

yeah, i only have my 20+-years-old smartcard, now yellowish...

now when we were packing our stuff, i found also my sun certification cards (solaris 6, 8, 10, 11)

cool :)

otis: yellowfish?

err

otis: yellowwish?

oh

right

the white cards turned

gotcha :)

exactly.

I got new glasses and I still need to get used to them...seeing is hard currently

you should be able to use any smartcard with sunray, except that you would need to write small code to read out unique id from it:)

I once did it for estonian ID card, wasnt too hard, just few lines of .... forth :)

tsoome: I bought a bunch of smartcards on amazon, but they didn't worked

yes, the system needs to recognize them

default is support for specific ones

tsoome: so the code needs to be on the sunray server?

yes.

interesting

and the script is ascii file, so it is possible to update

I actually love that about the SRSS, mostly shell - easy to customize

I need to search around, *may* have the sample stored, but I doubt a bit.....

tsoome: would be cool!

but it was not that hard to find:)

because the sun cards are starting to break (physically)

yes, that is the problem with EID as well -- it was popular to build access solutions based on EID cards - because everyone supposed to have them here (by law), but the contacts started to wear off and then you had to replace the card...

then they did create mobile ID (based on sim) and then smart ID (entirely software) solutions

that's the good thing about germany being so behind on digitalization - we will notice that way later ;)

:D

well, you can do some things with ~1.2 million people which you can not do with 100+ million people

true, but I guess it's more about fearing the invest here ;)

Hi, I bough a beelink s12 (a minipc with intel n100 in it) and I'm trying to run illumos on it. openindiana 20240426 boots, omni r151050 resets once the screen gets reset past loader, omni bloody 20240520 hangs at the same place. Any pointers on what to try to get omni to boot? I tried asking in #omnios without much success

Do you know where it is hanging? A verbose boot (one of the options in loader) will show at least the last driver that was loaded before the hang.

For r50, that resets, try booting with KMDB loaded (another loader option) and it should stop in the kernel debugger instead of resetting, then we can find out where it is.

It might be that OI has a patch that OmniOS doesn't have yet, so another option is to try the next bloody version that will be published on Friday/Saturday - OmniOS does a weekly upstream merge on a Thursday.

I don't know where it hangs. Even if I turn on verbose, I don't get anything. I get a couple of lines from loader (loading /unix/... , the last one is "Booting ..."), then the screen resets (r50 dies at this point), on bloody the screen resets and I get a single block cursor in top left corner and thats it

I'll try r50 with kmdb later today and we'll see

although I do have a vague feeling I tried all the options and it didn't change anything, but maybe I'm misremembering

It's possible, if it's aborting early enough. Another tip is to enable the earlier boot debug with "-v -B prom_debug=true,kbm_debug=true" from the loader's ok prompt.

If OI is working but OmniOS is not, it's most likely something that was fixed in illumos-gate since the last bloody media were made.

but that's only 9 days, interesting.

I was under the impression that the version of oi that comes up is older than latest bloody?

I'll try the early boot debug and report back if it gives me anything

I'll just test some options and post them here

thank you, appreciate that

the install isos of OI are usually older while the package updates are usually a bit faster

This should work - press escape in the loader screen to drop to the "ok" prompt, and then:

boot -kvd -B prom_debug=true,kbm_debug=true

The -k and -d will drop you into the kernel debugger, if you get that far, the others should increase verbosity a bit for the early boot code

i worked at a company where we made "RayMote W*Admin" where we could also personalize certain types of smart cards

andyf: nope, no luck. it still does the same thing

That's disappointing. No additional output or clues?

is this BIOS or UEFI boot? can you test other?

what the heck, I got it to boot... by disabling a serial port in bios?

let me guess - UEFI?

yup

it can be this specific system does not like us using IO ports for serial ports

physical hw?

yes

firmware update may improve this

I don't really care about serial port so I'm fine with keeping it off

I also have this update waiting for reviews: code.illumos.org/c/illumos-gate/+/3422

→ CODE REVIEW 3422: 16475 loader.efi: replace comconsole with efiserialio (NEW) | illumos.org/issues/16475

it would be interesting if you can check it (I can help by providing test binary)

if you could give me something I can dd to a usbstick then I'd be glad to help

altho, it is curious it is affecting kernel startup.... we may need to review those bits

so the hanging on bloody was probably it just waiting for input on the serial port? which didn't have anything attached to it

[illumos-gate] 16583 Want ability to use libtermcap to filter over libcurses(3xcurses) -- Andy Fiddaman <illumos⊙fn>

was it appearing hung or was it resetting?

bloody appeared hung, omni-r151050 was resetting

hm.

if you can test, enable serial again and drop out of boot menu, then enter: show console - what is the value there? also show os_console

otherwise, there is 54-115-50-84.sta.estpak.ee/loader.efi, you can copy it to the efi system partition on stick to /efi/boot/bootx64.efi and on ok prompt enter command: console

also if i disable the serial port, the loader looks different, the fonts are bigger and so on, more text-y

different resolution - omnios sets 800x600 as default

this is done to keep the boot loader screen layout

on omni-r151050 show console says "text,ttya", show os_console gives me "variable 'os_console' not found"

ok, so the kernel will use "text" as console, but *early kernel* probably does output on both text and ttya.

you can enter: set console=text and then boot

with loader.efi you sent me it works ootb

do you want me to collect anything else?

did you get output from command: console

with your loader or the original one?

bothe are mine:P

both*

you know what I meant :)

but yes, mine

you can mail the output to tsoome⊙mc

if you do not mind:)

the thing is, this new version is using UEFI Serial IO Protocol to communicate with port(s), so it can silently just drop the output when there is no terminal connacted; older version is using inb/outb on port address

but also early kernel is using inb/outb; so it may be that we are sending out "too much data" from old loader + kernel....

is pasting a pastebin link here also ok?

yes

that will do just fine

Not Found (#404)

what the heck

try paste.ec ?

better

ou

that is interesting.

and i guess not in a good way?

so, there appeares to be no ioports

the new version was unable to recognize the port and names it tty0 (as fallback name), and therefore console is only set to text (note the missing ttya)

so, neither loader nor kernel is attempting to output on serial port and this is the reason you have good boot

the rare case where two wrongs actually make a right?

do you have os running now? does it have anything in /dev/term and in /dev/cua ?

sort of yes.

UEFI is kind of pain for naming serial ports -- the API does not provide name for serial ports, so we attempt to check for "known" protocols. in this case, we fail to recognize

oh, there is one test you can help with, on loader ok prompt, enter lsefi -- it will print out long list of data, if you can find this path you see for tty0, that section would be interesting -- it may help me to identify the io protocol used there

there's stuff once the system is boot paste.ec/paste/TUoNuKW0#zNWpXwkK7C4gzgGIB6-Sai0MPPRTzIpQS45AIILeqI+

and here's the thing from lsefi paste.ec/paste/5LQcm7sX#ohBSl2V3HEGHn2LtpoHOh2rvhUg5ZFGc0-3e5ljPGAf

so, the asy driver does find it.

you do not have serial cable and device to connect with (usb-serial dongle or like) ?

sadly not

ok, np

I have to think a bit, this new code does check for ACPI IO, but you have ACPIEX, so we need to identify it

sure, I'll hang around in case you need me to run anything. It might take me some time to respond, but eventually I will

this is code.illumos.org/c/illumos-gate/+/3…c/boot/efi/libefi/efiserialio.c#291 this function basically

→ CODE REVIEW 3422: 16475 loader.efi: replace comconsole with efiserialio (NEW) | illumos.org/issues/16475

it's been a while since i had to read c

was that handle (second link) the only handle with this AcpiEx ?

unless I missed it then I'd say yes. It would be much easier if I could grep through the output

yes, I have thought to add search for pager, but still not done:)

hrm, seems I missed one more ocurrence, there's also this paste.ec/paste/aEZ08XFi#bPRuCvghBB1p+brCmDHduHV+xSaArRGLR8WFwHGY2eW

ok, good, that was expected one.

sorry for missing it before

efi-show -g global -v ConOut -- does it only list acpiaddr device?

np, it is long list to scroll through

one entry, "global NV,BS,RS ConOut = PciRoot(0x0)/Pci(0x2,0x0)/AcpiAdr(0x80012305)"

yes, thats video console

sorry, one more: efi-show -g global -v ConOutDev

one entry "global BS,RS ConOutDev = PciRoot(0x0)/Pci(0x2,0x0)/AcpiAdr(0x80012305)"

good, ok.

ConOutDev is list of possible devices. So your system does not provide console over serial (or it is not configured to do so)

an added note for fenix issue 16584: twitter.com/MengTangmu/status/1795920863206011111

BUG 16584: posix_fadvise(3C) should do something (New)

↳ illumos.org/issues/16584

issue '1795920863206011111' not found...

lol, if your issue ids ever get that high (while still being sequentially numbered), I'd worry

more directly useful: openzfs/zfs #13958

I updated #16584 (fenix) with those links.

BUG 16584: posix_fadvise(3C) should do something (New)

↳ illumos.org/issues/16584

right, the tweet just explained why you should look at that (an unfinished implementation for FreeBSD/OpenZFS based on the finished work for OpenZFS on Linux)

Thanks for the pointer..

alanc: obviously you should encode category information into the issue IDs, ala ASN.1

(finding ways to design a worse issue tracker is a fun hobby!)

is the asn.1 encoding done in perl?