[illumos-gate] 13373 Import upstream improvements to bhyve VNC server -- Marko Kiiskila <marko⊙ao>

Does anyone know how to use `eeprom` to read the new bootnext list from the zfs label?

got it `eeprom -bp`

yep

tsoome: fwiw, isn't it worth to add SYMLINKS from kenv to eeprom also on freebsd?

symlinks?

LINKS, i mean

you mean, to have kenv aliased to eeprom?

yes

or wouldn't it make any sense?

i think, they are not that similar and probably would create more confusion

you can always just have local shell alias

Thanks tsoome, I knew you would know :D

I was working on code.illumos.org/c/illumos-gate/+/1167

alias xs="echo go to sleep!"

:)

`pmset -g sleepnow` ?

tsoome - I broke a box with `beadm activate -t xxx; beadm destroy xxx; init 6`

had to get it power cycled

hm?

went to loop?

the photograph they sent me was just a blank screen

it came back in the old BE fine after the power cycle

efi or bios?

efi

I tested it in a bhyve VM and it does the same

anyway, we should be removing the nvlist entry when we destroy a BE. Otherwise if another is created later with the same name, it gets the T flag

I also fixed `beadm rename`

yes, but, loader set_currdev() must survive this, not hung itself.

zfs_get_bootonce() will wipe the bootenv area, but then we call set_currdev() and that one probably will get stuck.

btw, you may have noticed, the other part of bootenv is illumos-nvstore - the nvlist to store non-volatile R/W variables.

yes, I saw it

and the generic store implementation

do you have an update for the eeprom man page in the wings?

no, not yet.

ok, I just might do it, just didn't want to duplicate

you can pick some bits from freebsd.org/cgi/man.cgi?query=zfsbo….0-current&arch=default&format=html

i did add eeprom as "reference", maybe something could be done better etc, so I wanted to avoid making it written in stone too early:D

That's fair enough, I just wanted to check the value when testing beadm destroy

ofc we can just mark it unstable on manual and thats it.

"The command line interface of eeprom is Uncommitted. The output of eeprom is Not-An-Interface and may change at any time."

:)

ah.

its there?

anyhow, need to go offline once more...

no, there's no stability stuff there at the moment

that's just the text from the prototype

I've got a question about boot modules, if anybody's around and/or has a few minutes.

I'm pxe booting with a module that I want as a file. It's there, but the path name is the full url, and not the name I specified

It's going to /system/boot/http:/hostname/file and not just /system/boot/file

You mean boot-time modules?

Yes

I'm trying to do the same thing we do for networking.json

but the name isn't what I expected it to be

The thing to look at is what are the actual module lines that end up passed through to the actual bootloader.

Before I spend hours digging into this, I was hoping maybe someone knew what I might ber doing wrong

This is what I've got in ipxe:

module metadata.platformequinix.com/metadata type=file name=metadata

And this is what we do for networking.json:

module ${next-server}/bootfs/1c697a6118e1/networking.json type=file name=networking.json

The only difference I can tell is that I'm not including a hash module for it, but it seems silly that that would be the difference.

In theory, name= is supposed to be what the name comes from. You're using the joyent ipxe?

No, it's not

it's equinix metal

OK. You're probably hitting the internal size limits in ipxe.

But you can see what name we actually got.

bootparams has this:

module-name-0=metadata.platformequinix.com/metadata

Well, that's what it got from the boot loader then.

Take a look at OS-3770

Which is in joyent/ipxe

but on triton I get this:

module-name-0=networking.json

Read the bug, please. I suspect that's the problem.

Ah, yeah, that looks like it might be what I'm running into.

I wonder if ipxe follows redirects...

maybe I could trick it into being shorter somehow.

Dunno, sorry.

But thanks for the pointer. I at least know what I'm dealing with now.

Chainload your own ipxe?

That's one option I'm thinking of

You've at least saved me many hours of digging and re-discovering OS-3770.

or maybe just using a variable for the URL will work

But either way, I can deal with the file in the current place until I can get that part fixed

v_a_b: the Japanese person, whose name I forgot,... has a riscv port already I believe

they sort of said "Here it is" and disappeared, which is a shame

[illumos-gate] 13378 unknown ld(1) guidance should be a warning -- Richard Lowe <richlowe⊙rn>

gitomat: for future reference, bot, that was enough latency to scare me. do better.

Not bad for government work, though.

richlowe: true, I vaguely remember something about not liking our build system so nothing got upstreamed

I think it was also a university student project? Or somehow attached to a university?

Memory hazy at best

It looked good (as did the arm one). Certainly a good starting point but all rewritten in GNU make to cross compile from Linux...

the cross work robert and I did almost certainly will work with minimal poking

Funnily enough, I was looking into that today for something else :)

Hello Illuminati! I think I've found a bug in /usr/bin/sh's wildcard expansion but not sure where to report it

bugs.illumos.org

not in omnios's tracker (as that's the distribution i'm using)?

sjorge Yes, I remember his mail. But I expect that some things would have to be changed and adapted to that new board.

DrHyde - I'd be interested in the details

Interesting, I can't replicate it immediately on omnios r151036

on my machine /usr/bin/sh is a symlink to i86/ksh93

yes, same here

which has md5 sum ffb847606e5e1752270c4afad4e8eb85

want me to create an account on there for you to have a look?

have you set any shell options, or is it all out of the box?

i don't think there's anything relevant in my .profile, and i don't think i've touched /etc/profile

sh -o says that 'glob' is on and 'globstar' is off

ok, globstar is on for me

but I changed it, and no difference

It probably won't make a difference, but what locale are you both in?

C

i'm in en_GB.UTF-8