nvmeadm connect -t tcp -n foo:bar:baz -a 10.91.230.250 -s 4420 would be dope

dumb question: With how swap works when the memory page is either backed by disk or anonymous (in swap), how does the memory allocation mechanism work for VMs? Is there an option to hard allocate memory?

Or was that written in when KVM was originally ported?

copec in native app case pagein/pageout will handle data to/from disk. No idea about KVM.

[illumos-gate] 15178 libmp: unused parameter -- Toomas Soome <tsoome⊙mc>

[illumos-gate] 15179 libmd: unused function 'ROTATE_LEFT' -- Toomas Soome <tsoome⊙mc>

tsoome, did you typo your commit messages, or did you really fix two libraries named libmd and libmp?

not a typo

That is an interesting coincidence C:

not really. I just picked from completed clang fix list libraries with least patches:)

Also, unix has oft favoured two letter names for things.

The original UI was literal Teletypes.

Not known for easy keypresses. A Selectric might have given us "copy" instead of 'cp'.

beta.birdsite.live/users/illumosorg is this one of ours?

I think birdsite is just a twitter gateway

i see

.oO 194 errors generated.

oh

does the illumos x86/32 ABI require a stack alignment of 16 bytes?

props for a ninety second deadman though. I just deadlocked a linux box for five minutes over the same thing that gives us a 90 sec kernel panic.

is there any guide on debugging user space memory allocations? i.e. i'm trying to find where did all the memory go in 32bit process

libumem?

:)

pmap will give robust view

08CD6000 255848K rw--- [ heap ]

running now with UMEM_DEBUG=default UMEM_LOGGING=transactions

you can try to force a core dump and use mdb on it

not sure how ::findleaks will work on a live process

i did use gcore yeah, findleaks says it's only leaking 64K :)

e

The other thing that can help then yuripv is to use ::umastat to get a sense of where the memory is going and then separately you can also use the bufctl information to see the paths that are allocating / freeing memory.

At least, that's the easy starting point from just a core file / running process.

If it's reproducible, then there's other things that you can consider evaluating.

thank you

Ok, I just tried again to boot omniosce on oracle cloud free tier this is as far as I got pasteboard.co/afrrEp3IqWff.png

yuripv you can set breakpoint on _exit and get the memory analysis done from there - so you wont need core/running process..

tsoome I'm on the boot loader what do I need to type to have a clear view on where the boot process is stuck?

you mean, the kernel is loaded and is starting?

you would need to esc out of menu, type clear and then enter boot

tsoome according to the picture I don't know if the kernel is loaded

Oh I can enter the menu before boot in omnios to use the boot console

It seems is stalling at the ominos installer

it looks like that console can't scroll :)

sorry, i do not have picture:)

tsoome here it is pasteboard.co/afrrEp3IqWff.png

yuripv yes, you need to manually scroll down but is stuck after that.

neirac that is looking like loader is hung in middle of drawing brand, and that is something unexpected really.

That screen is rather surprising, is it from serial console?

tsoome no is not from serial I provisioned a free instance on oracle cloud and check the console they have in there could be vnc

I'm going to try smartos to see if it changes

They have free arm64 instances that could be useful for the guys working on the illumos port