I can't figure how to get changes to ~/.login.conf working

I'm trying to set times.allow for my own user, and I've been trying to set times.deny to a short interval of a few minutes to test whether the changes are taking

I obviously read the manpage, I ran cap_mkdb

mtll: you cannot set times.login in ~/.login.conf, this is supposed to be set by administrator in /etc/login.conf

times.allow rather

I see

is there some documentation on what can and can not be set in .login.conf

because the manpage is very vague about it

as far as i can tell, no, but there probably should be. if you don't find any documenation on this feel free to open a PR and cc ivy⊙fo

" typically those which do not involve authentication, resource limits and accounting."

[tj]: this is quite vague though

[tj]: yeah, like I said, pretty vague

uhuh

i agree it seems "obvious" that times.allow shouldn't be allowed, but there is a thing recently where users want to lock themselves out of their own accounts

so I guess I gotta make some login class, add my user to it, and my changes to /etc/login.conf

like, "i spend too much time on freebsd, only allow me to log in during these particular times"

ivy: yeah that's exactly what I want to do. and I mean it doesn't really matter where I put it, it's my personal desktop, and I'm the only user and administrator anyway

that is probably the sentence to clarify, is ~/.login.conf applied to anything other than environment variables and does it apply to all

otherwise the man page change is going to be quite intense

i think we can just say that nothing listed in the 'authentication' table may be added by the user themselves

we'll see what makes sense, at least i think a PR is justified here

just trying to stop myself from going 4 hours past my bedtime writing code when I get hyperfocused, and end up losing a bunch of sleep, and writing a bunch of shitty code that I'll have to rewrite anyway :P

well login.conf won't help you there either and that is in the man page

"Further enforcement over the life of a session requires a separate daemon to monitor transitions from an allowed period to a non-allowed one"

well times.allow seems like it'll work, along with a cronjob to lock my screen, right?

I don't think your screen locker is going via login

thats a guess based on a feeling

hrm. I did this before on linux with pam_time, although that doesn't seem to exist for openpam

it probably should do, but a lot of things don't...

(i noticed this while implementing pam_userjail)

worst case, I could always just cronjob a reboot with a 5 minute warning to save any files and such. be a bit annoying, but better sleep is worth it

this is what I liked about pam_time; pretty much anything that touches authentication uses pam, so this wasn't an issue

I guess I could port pam_time to openpam myself, it can't be that complicated. More effort than I'd like to invest right now though

Just don’t start right before bedtime

:D

well, the good news is xdm does interact with login, because I was just able to set a short test window with times.deny

let's see if xlock does as well, more doubtful about that

it does not

well xdm is creating a login, but xlock isn't

can you create a new terminal in a deny time?

yes

you mean cause of the login shell

I guess that doesn't actually do any authentication

eh fine I'll just set up a cronjob with a reboot

I guess an alternative way to do all this would be to add two cronjobs as root with pw lock/unlock <user>

I used to have a pomodoro timer which locked the screen after each interval

but I could just log back in

I used to have gum that helped me quit smoking

but I could just smoke a cigarette

probably fine just as a reminder yeah

it was "sleep 1800; i3lock "

hah

just trying to figure out the best way to warn me(via stumpwm) from the cronjob now

oh right, someone already made a little script for running stumpwm commands from the shell; I can just use that

echo 'go to sleep, neo...'

huh, stumpish doesn't work, because stumpwm doesn't show up in pgrep

shows up in htop just fine, under my user

maybe because it's spawned by xdm, which is run as root?

but the xterms started from stumpwm do show up

ps wwaux?

mtl 39183 0.0 0.1 1436272 91916 - Ss 11:25 0:02.43 /home/mtl/stumpwm/stumpwm

or pgrep -f

gotta go though, I'll have to figure this out later...

pgrep has stupid footshooting option -v, which is not verbose

i have at least killed everything in my machine with pkill -v

:p

at least once

pgrep -Slf is good way, but i like ps more to search stuff, eg this is for finding swapped out procs: ps wwauxd | awk 'NR==1||$8~/W/{print$0}' | less -InS

:p

i was not aware freebsd even supported swapping out entire process anymore

partially out i would guess

why does it matter?

are you sure W doesn't mean "Marks an idle interrupt thread", per ps(1)?

root@ixia:/usr/local/etc/bird.d # ps wwauxd|awk '$8 ~ /W/ {print}'

root 2 0.0 0.0 0 16 - WL 10:30 0:01.62 - [clock]

root 11 0.0 0.0 0 256 - WL 10:30 0:03.21 - [intr]

these are certainly not swapped out...

actually

there are multiple w's there

stupid eh

where's swappers tho?

top -bao swap 1000000 ?

i'm not sure i follow what you mean... you are matching on $8, which is 'state', and there is only one documented meaning for W

if by "where's swappers" you mean how do you find processes which have been swapped to disk, as i said, i do not believe freebsd supports this

i may be wrong here but i don't think swapping processes to disk has been a thing since like, 3BSD

it's basically pointless as soon as you have proper VM paging, which we got with VAX

top -bwao swap 1000000 maybe

there are two w's

another is down in manpage

can you please show the parts of the manpage you are talking about? i only see one

W The process is swapped out.

i only saw this

well top shows something?

hm, that text does not occur in my ps(1). what freebsd version are you on?

it's changed now eh?

damn

in current it's all gone

and there i can't check what top would show

ah, cgit.freebsd.org/src/commit/?id=62e…a0f07e448da27cb2cc8165e749e7fdfcd7e

but it exists there

so only top?!

surely there's still ps option

except i don't see it

no, there is no ps option because freebsd doesn't support swapping out processes

this commit was just removing the vestigates of obsolete code

hmm

what does it swap out then?

parts of proc memory so sure

that's what top shows

maybe definition issue

but it definitely swaps out some memory the proc is using

hence that one is swapped out, even if partially

what's the other way?

by definition whole idea is to move memory out of ram

i think someone swapped me out

i have no idea how different oses do it

right after birth?

:p

unsure what complete swapout means

kernel still knows it has it running

and what benefits are

i don't even know fbsd design that much

ketas: historically, "swapping" means moving the entire address space of a process to disk. "paging" means moving individual pages of a process to disk. freebsd does paging, not swapping

ketas, no, like, to flash

there is no ps flag for "this process has 1 or more pages paged out", as far as i know, since this is not really interesting

ah that

riiight

but who calls it like that

everyone, really?

no one says "this process is swapped out" nowadays, or if they do, they're probably wrong

but the location is still referred a swap, hence...

yes, and "swapping" is used to refer to pages being moved to/from disk, but that's different from the terminology used to refer to the actual state of a process

statistically it does not really matter if you have 1 or 100 pages paged out to swap and counting those will likely require too much work.

what top does then?

is this about virt and rss

sure, it mentiones it being approx

it probably fudges the maths

rss is ram use iirc

virt and rss is nothing to do with swap

virt = virtual memory allocated to the process, rss = actual process size in memory (sort of, it's complicated)

neither of those tell you if any pages have been paged to disk

why is it virtual tho?

yes; the only reason I connect swap to that is that those two are traditionally considered memory stats and swapped pages are also

well, "traditionally" if you're me, idk if it's a real tradition

to say it's imaginary memory?

ketas, correct

may or may not exist as real ram?

ketas: it's virtual because BSD is a virtual memory operating system

imagine you mmap() a 4TB file into memory, your virtual memory size instantly jumps to 4TB

this doesn't mean you're using 4TB of memory, it just means 4TB worth of virtual address space is mapped to your process

yeah that's not ram space of memory

memory overcommit at work

huh?

Koston: no, nothing to do with overcommit

hence virtual

ivy, well if you *do* have well over 4TB, then not

I was assuming you don't, because most people don't.

ketas: right, it's virtual in the sense it doesn't refer to physical memory pages. it may do, but it may not

it's address space in use, not memory in use

what's opposite of virtual memory os?

Koston: mmap() does not rely on overcommit. overcommit is when you attempt to map pages that don't exist. having a page in virtual memory doesn't mean it's mapper

s/mapper/mapped

i think this was before even my sperm existed

excuse you.

i don't know history that much

ketas, the opposet of a VM OS might be an OS where the entire address space is in memory, though I am too eepy to do the very basic groundwork to search about it

ketas, smaller embedded systems running on hardware without an MMU. it just means an address refers directly to a location in physical RAM.

oh that

aha

yes they have them still

without an MMU or not using the MMU if there is one

typically these are systems that don't have an OS, just some software running directly on the hardware

the nature of OSness can be debated there

let's not

i vague recall wrt54g

is the stand libc you're using an OS?

some early Unix systems ran without virtual memory - for example, Unix on PDP-11. that's why PDP-11 had to swap out the entire process, it couldn't page individual pages

iirc that has no mmu?

or someone lied

but since VAX gave us virtual memory, Unix doesn't work like this

it has no pages?

and page was like individual memory allocation unit?

should i attach a tape machine with a drum roll sound to the access bus for the swap disk to fill the lab with drum roll noise when someone tries to continue running a programme they had swapped out

a page is the smallest amount of memory you can address via the MMU

since it would be infeasible to address every byte individually, memory is divided into pages

ketas, you should pick up a copy of Modern Operating Systems, or maybe Design and Implementation of the FreeBSD Operating System

for how much money, from which bookstores

MelMalik: free from anna's archive...

it's too difficult i'm afraid

and if I don't want to commit piracy?

i would also need to read on hw first maybe

then pay some $ to support the freebsd developers who put a huge amount of effort into writing it

former is much easier to read and much easier than you'd expect imho

would buying the book not do that? or is there not a way to buy the book

i would maybe choose lecture videos first

okay, changing tack: what are their ISBNs

the lecture video series by McKusick is wicked expensive

a book without anything else is just hell

MelMalik: oh, maybe i misunderstood the question, i thought you were suggesting the books were hard to find

Modern Operating Systems is definitely good read, worth every penny

ketas, a book is hell?

ivy, hard to find for me, maybe.

but for someone who is actually looking, probably not

i mean reading a thick book, without doing any practical tests of something by the side

is super difficult somehow

i don't actually own FreeBSD D&I but i think i bought my copy of 4.4BSD D&I on Amazon

i haven't been in uni but there's a reason why they have mix of all this probably :p

my main complaint about anything ever is that the Tcl documentation in troff edition does not have cross references because it's troff and not mdoc

the cross references are just stated naked, i.e. as the manpage names

with (n)

funnily it depends on written material too, eg out manpages are really good

whoever made it all over decadea

s

FreeBSD documentation quality has suffered over the years, unfortunately

but it's still heaps ahead of Linux

why?

writing documentation is a lot of work and interest in operating systems has waned

interest in what

oses exist?

and it like doesn't need to be one person...

meh docs are hard

i have only wrote one manpage, for my own stuff

write doxygen comments and call it a day :3

hey Im trying to bring up freebsd powerpc/power on a ibm system p5 520 power5+ but it's only detecting 128MB which isn't even enough to load the kernel

*CANUCKED*-

flyback: my friend sent one of those to the e-waste bin a week ago :)

yeah I might have to do the same

although I just made progress

disableing msi and iommu it didn't kernel panic or opps as much

it couldn't find rootfs but that might be the fault of gentoo not having the ide driver loaded

im playing with freebsd also to see which one has more progress

yep looks like that is the fault of gentoo not having the ide driver compiled in

actually I could run both freebsd and linux if I have a licence for this power5 for their virtualization engine

or I *cough* just so happen to find a licence code lying around

flyback: power5 is like 20 years old, isn't it?

i think fbsd only supports power7 and higher from the power isa

the main ppc page says power7, the wiki says power5. hmm.