<Grommish>
I don't know what to do with a .rs file yet
<neggles>
prints some nice ascii art gradients
<neggles>
oh i mean i can push it as v0.1.1 :P
<Grommish>
Since I don't have rustc on deviice
<Grommish>
I need to build a bin to cross
<Grommish>
neggles: Go for it, it'll be a few hours to rebuild rust before I process the rest of it
<Grommish>
I'll just update the commit head when the time comes
<Grommish>
but, yeah, libc didn't have O_LARGEFILE or NOTSUPsomething-or-other for mips64 in the llvm libc
<neggles>
it is done
<neggles>
whether it runs on anything other than x86_64 i have no idea
<neggles>
but it's hardly complicated
<Grommish>
Well, I only have mips64 setup, so that's my test bed :D
<neggles>
maybe i should make it print how long it took to run :P
<neggles>
or see if there's some way i can detect hardfloat vs softfloat
<Grommish>
haha You got whatever time you want to devote to it.. I just will pull the commit head when the time comes
<Grommish>
If you could get it to print out CPU features would be helpful for arm
<neggles>
it appears there is no reliable way to tell :P
<Grommish>
because that is going to suck
<Grommish>
neggles: Well.. The issue I was seeing is that any float instructions SIGILL'd
<Grommish>
neggles: because apparently rust tuples have to explicitly define soft-float, regardless of the compile-time flags, or it doesn't include any of them in the toolchain
<neggles>
Grommish: fair enough! I have shrunk it from dumping a 256x256 char output down to 80x25 just for the sake of sanity, and it prints the OS/ARCH values first
valku has quit [Remote host closed the connection]
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
GNUmoon has joined #openwrt-devel
Luke-Jr has quit [Ping timeout: 480 seconds]
goliath has joined #openwrt-devel
Luke-Jr has joined #openwrt-devel
nitroshift has quit [Remote host closed the connection]
mrkiko has joined #openwrt-devel
<mrkiko>
Hhi all.
srslypascal is now known as Guest432
srslypascal has joined #openwrt-devel
<stintel>
> Sorry for the delay but a review of the software for GPL needed to be conducted. A link to the files which are identified as GPL can be found at this download.
<stintel>
requested in November
goliath has quit [Quit: SIGSEGV]
Guest432 has quit [Ping timeout: 480 seconds]
<neggles>
stintel: hah, i'm going through that with watchguard atm
<stintel>
this is with wg
<neggles>
oh interesting
bluew has quit [Ping timeout: 480 seconds]
<stintel>
only kernel sources, no u-boot
<neggles>
i'm after the files for the firebox T20 though, LS1023A
<neggles>
yeah they gave me literally a vanilla/unmodified kernel tree
<stintel>
wtf :P
<neggles>
so i shot back a "yeah, no, that's not how that works"
<stintel>
I was thinking they might have done the same but I diffed with vanilla 3.12.19 and it's not
<neggles>
with citations :P
<neggles>
to the support guy's credit, he asked me where to find Sophos's GPL ISO for their firewall OS so he could pass it to their engineers as a "what this should look like"
<neggles>
guess we'll see how that turns out
<stintel>
it's lacking all the .gitignore files for one
<stintel>
nice way to mess with people wanting to make a diff
<neggles>
indeed
<neggles>
we're considering changing away from sophos to someone else at work too & it's my call as to whether we do / who we go with, so i threw in a "how vendors handle these requests can have a significant impact on our relationship"
<neggles>
can't hurt
Tapper has quit [Ping timeout: 480 seconds]
<neggles>
stintel: i take it you hit the same stupid u-boot password wall?
<neggles>
I tried tracking what it does to validate the password in ghidra but it's an absolute friggin mess
<stintel>
I've not looked at u-boot really, M300 boots fine
<neggles>
T20 boots from eMMC or i'd just swap out the OS image
<neggles>
it doesn't actually verify the signature from the looks of things, just has a password, and while i assume there's a way to get a root shell on it I don't know how
gladiac has quit [Quit: k thx bye]
gladiac has joined #openwrt-devel
<stintel>
wait, this *is* the vanilla kernel just with some files revmoed
<rsalvaterra>
Grommish: pong
<Grommish>
rsalvaterra: I actually don't remember what I was going to ask you now, so it must not have been important :D
<rsalvaterra>
:P
<rsalvaterra>
I swear, ramips is starting to really test my patience.
eduardo010174 has joined #openwrt-devel
<\x>
is ramips the last mips, or will there any be upcoming mips soc
<stintel>
> As you did not include any of the changes made to the Linux kernel, nor did you include Das U-Boot, you are not complying with the GPL. Please try again.
<stintel>
fucking wankers
<rsalvaterra>
stintel: Frustrating… :(
<stintel>
took them 3 months to recompress a vanilla kernel with some files missing
<eduardo010174>
Good morning ahhahah. From Portugal.
<rsalvaterra>
eduardo010174: Likewise. :)
<rsalvaterra>
Not so good morning, though.
<eduardo010174>
Whait, why?
<Grommish>
rsalvaterra: Well, hopefully it gets better
* rsalvaterra
is pissed off at the diff between our local mt7621.dtsi and upstream mt7621.dtsi in staging.
<Namidairo>
dare I ask
<Grommish>
Well, if you could make it work with the rtl8367rb.... :D
<rsalvaterra>
It's making the port to 5.15 *really fun*.
<stintel>
what's 5.15?
* stintel
hides
<rsalvaterra>
And I even haven't gotten yet to the drivers…
<rsalvaterra>
… the old ramips eth drivers.
<Namidairo>
i thought they just replaced those
<Grommish>
rsalvaterra: When I tested 5.15 with the octeon tree, it was only like 3 or 4 commits I had to remove hehe What you get for having real, supported devices!
<rsalvaterra>
Does paraka (Sergio Paracuellos) usually hang around here?
<neggles>
stintel: that's what i thought, out of curiosity what's the sha256sum of the files they gave you?
<Grommish>
Wait, you found someone worse than Marvell?
<stintel>
I wonder if my legal insurance would cover something like this so I can just lawyer up :P
<rsalvaterra>
Grommish: And a *lot* of effort has been spent upstreaming ramips patches during the last months.
<neggles>
i bought this T20 from a pawn shop's ebay store
<neggles>
for like $100
<Namidairo>
I saw a pallet of used 11n aironets earlier today...
<neggles>
it had never been registered.
shibboleth has joined #openwrt-devel
<neggles>
came with a 3 year license and support contract.
<Grommish>
stintel: Honestly, if it's work or business related, I wonder if you could attempt compensation after x number of "reasonable" days
<stintel>
mine were registered, had to send them a photo of the serial number
<neggles>
it was also absolute cake to get the "warranty void if removed" sticker off in one piece
<neggles>
isopropyl alcohol and a bit of thin glossy plastic from some packaging
<neggles>
never mind that those stickers are utterly unenforceable
<Habbie>
that sign can't stop me because i can't read
<neggles>
hahaha
<stintel>
:P
<neggles>
i'd just like to work out what the secret magic command to get a root shell is
<stintel>
I didn't bother, just pushed the screwdriver through the label
<Namidairo>
"please"
<neggles>
though i note there's a sha256crypt password in the firmware image for `root`
<neggles>
which appears to be the same for every single box...
<neggles>
i sure hope you can't log in remotely with that...
<Grommish>
Kali could tell you :D
<Grommish>
neggles: No one would be that irresponsible.. Tsk
<neggles>
I guess I should throw the hash into hashcat and see if the usual 4-8 character bruteforce works
<Namidairo>
or... google it.
<neggles>
alas, google only finds me info about the older gen devices / software versions too old for this unit to run
<neggles>
the T20 is very new
<neggles>
LS1023A, the T40 is a 1043A and the T...60? 80? is a 1046A
<neggles>
T80 yeah - M290/390/590/690 are 1046A/2084A/2120A/2160A
<eduardo010174>
rsalvaterra: What is link for you staging tree? If it's have ...
<neggles>
managed to wrangle myself a juniper NFX250 for peanuts today too, i wonder how it will feel about running OpenWrt as a VNF :P
<rsalvaterra>
eduardo010174: I actually don't have a staging tree yet. Thanks for reminding me, though. I should poke jow about it… :) I do have Git{Hub,Lab} repos.
<neggles>
they want US$220 to ship that to australia, classic
<stintel>
next thing I'm buying has to have 10GbE copper
<mrkiko>
someone able to help me out with issue #4721 and can edit my posts? I am using "ghi" command-ine tool which uses the github API, but I can't find a way to edit specific comments trough it
<stintel>
ideally with PoE++ in ;)
<Namidairo>
that's a bold case colour
<neggles>
stintel: sounds like you want a supermicro X10SDV board :P no poe-in though
<stintel>
nope. supermicro is on my vendor blacklist. broken af
<neggles>
Namidairo: it's plastic and i, an imbecile, dropped mine within an hour of getting it and cracked a corner slightly :(
<Namidairo>
oh I thought it was alu
<stintel>
do that with an m300 and you crack your floor :P
<stintel>
or your toes
<neggles>
(T20 and T40 are the same, just dual-core / quad-core and the T40 boots from an SSD instead of eMMC)
<neggles>
heh, true
<neggles>
the m300 has the dual? quad? powerpc in it yeah?
<stintel>
4c/8t
<neggles>
they made/make hyperthreaded qoriqs?
danitool has joined #openwrt-devel
<neggles>
wow
<neggles>
there is a 12c24t @ 1.8GHz
<neggles>
that's absurd
<rsalvaterra>
neggles: Not too surprising. It's POWER, IBM's been doing SMT for years, and much better than Intel.
<neggles>
rsalvaterra: well yeah, but freescale :P
<neggles>
this chip has to have ended up in some device or another, but what...
<Grommish>
neggles: rust is finally done compiling, but not yet done installing :D
<neggles>
stintel: the mars rovers use a pair of single-core chips though, the RAD750
<neggles>
which is pretty mcuh a radiation-hardened G3
<neggles>
Grommish: sounds like you need more cores :P
<neggles>
got a tree i can build?
<stintel>
pfff I have 40 and considering a replacement because this thing is slooooow
<Grommish>
neggles: I'm only running a i7-3610QM :D
<neggles>
oh dear lord.
<Grommish>
I do run with 32Gb RAM and a SSD build drive though
<neggles>
my buildbox's dual E5-2673v3s do a decent job
<stintel>
oh
<stintel>
E5-2673 v4 here
<neggles>
the v4s are worse :P
<neggles>
lower tCASE so they run louder and throttle sooner :(
<stintel>
well it was the box I could buy when the previous one said no
<stintel>
same day pickup etc
<neggles>
yeah, i think it works out to about the same overall performance, just less cores
<neggles>
24c48t is nice
<Grommish>
neggles: Yeah, it takes forever evey time I make a change to the Makefile, it rebuilds lol
eduardo010174 has quit [Remote host closed the connection]
<stintel>
unfortunately 2.3-2.4GHz is not nice
<neggles>
stintel: 2.4GHz base, I get 2.9GHz all-core
<stintel>
2.9 all-core? wtf
aleksander has joined #openwrt-devel
<stintel>
anyway, considering an EPYC 74F3 based workstation
<neggles>
did a bunch of experimenting with power profile settings in the Precision R7910 BIOS and cpupower settings
<neggles>
mmmm we do love the F series chips
<neggles>
upping the RAM to 128GiB and moving to ZFS made a bigger difference than winding it up from 2.5 to 2.9 though
<neggles>
whole build runs out of ARC
<stintel>
I'm on 256GB RAM already
<neggles>
the other day i built a raspberry pi kernel and it took so little time i was convinced it couldn't possibly have finished
<neggles>
though on paper this new work laptop should be about the same
<stintel>
ah yes, had the same when I tried to build a kernel for my old sparc
<neggles>
this i7-11850H outpaces my desktop's i7-8086K... by, a lot
<neggles>
but the price gap to get anything significantly better than the 2x2673v3-ish performance tier is just huge
<stintel>
waiting for the ThinkPad Z series
<neggles>
work laptop is a P15 G2 :D
<neggles>
it is very nice
<stintel>
yeah that's not a laptop
<neggles>
you're right, it's a big heatsink with a keyboard and an LCD glued to it
<neggles>
it's actually ~250g heavier than the P52 it replaced...
<stintel>
2.867kg wtf
<neggles>
yeah, it's about 50% heatsink by volume
<neggles>
(well volume of the lower/non-screen part)
<stintel>
if I can't walk around it while holding it in one hand and typing with the other, I don't want it
<neggles>
fair, i have other devices for that :P
<stintel>
s/it//
<neggles>
this is only a laptop for the once or twice a month i go out to a client site
<neggles>
or datacenter
<neggles>
rest of the time it's hooked up to a TB3 eGPU pretending to be a desktop
<rsalvaterra>
stintel: Speaking of laptops, seems like Qualcomm is preparing some nice laptop SoCs for Microsoft…
<stintel>
but it's qca, so it's a no
<neggles>
never mind that the eGPU is slower than the internal GPU in the P15
<neggles>
yeah
<neggles>
QCA's exclusivity deal with microsoft ends this year
<neggles>
mediatek windows arm laptop when?
<rsalvaterra>
As long as the SoC does *not* contain a modem, though. :P
<neggles>
I'd take a kompanio 1380 (the chromebook 4xA78+4xA55)
<rsalvaterra>
(Qualcomm modems have access to the entire system memory.)
<mattytap_>
woke up and there is a conversation on watchguards
<neggles>
doesn't that technically apply to every GPU/DSP/other random closed-off IP block?
<rsalvaterra>
neggles: Not in the case of MediaTek, for example. The modem is isolated.
<neggles>
"IOMMUs are for squares" - Qualcomm, probably
<neggles>
rsalvaterra: interesting
<neggles>
mattytap_: mmmyes, stintel and i are both attempting to get something vaguely resembling a compliant GPL source bundle from them
<rsalvaterra>
It's interesting, but it sucks. MediaTek SoCs are great, but open support for it is terrible. Good luck finding a MediaTek phone with LineageOS support, for example.
<neggles>
perhaps with two people pestering them for the same thing at once they'll do it? find out after this message from our sponsors
<stintel>
mattytap_: how about you also open a case with them for the same ;)
<neggles>
none of us is as annoying as all of us >:D
<neggles>
rsalvaterra: h-uh, i thought mediatek's stuff was better supported than most? they seem to put a lot more effort into mainlining things than many SoC vendors
<neggles>
iirc this lenovo chromebook duet i have here (MT818...3?) has nearly everything at least partially working in mainline now
<rsalvaterra>
neggles: That's true for laptop SoCs. For phone SoCs, not so much.
<neggles>
phones are always an absolute crapshoot though
<neggles>
so much dumb crap with OEMs and bootloaders
<mattytap_>
I will, but dont hold out much hope - they will be the last of the closed source code.
<neggles>
I have a device with a 3 year support contract
<neggles>
I will keep this ticket open / open tickets for 3 years if i must
<mattytap_>
anyway - judging by their crappy device trees - hm
<neggles>
it seemed to catch their attention when i pointed out that Sophos post comprehensive and actually-kind-of-documented source bundles for public download
<neggles>
"the competition is happy to do it, so why aren't you?"
<neggles>
heh, the DT for the T20/T40 is just the LS1043A-RDB tree
<neggles>
with maybe half a dozen tweaks
<mattytap_>
For me, the WatchGuard monitoring is the best I've seen which is partly why I used WG, its their USP
<neggles>
I haven't actually used this little box for its intended purpose very much, but it does seem pretty nice
<neggles>
palo alto are sending me a PA-440 to try and twist my arm
<neggles>
fortinet are sending something too, i think a 60F?
<mattytap_>
Im running at home on a T30, Im on my third though as they have a habit of failing and the SD card is held down with tape
<stintel>
heh
<mattytap_>
looking forward to moving to M200
<neggles>
not really interested in the fortigates though, they use "custom SoCs" which appear to just be mildly-custom octeons/armadas
<stintel>
that's the nice part about M200/M300, normal SD card slot and Flex ATX PSU
<stintel>
and replaceable RAM too
<mattytap_>
I could move today if it were not for the alturistic intent to fix musl
<stintel>
how's that progressing :)
<stintel>
I'm not subscribed to the musl ML so in case you wrote there I have missed it
<mattytap_>
on ebay UK the M200 are plentyful, go for around 60EUR now
<neggles>
WG are not so common down under
<stintel>
they really are a steal at the price
<mattytap_>
there is ONE M200 for 150EUR, otherwise you have to import from States
<mattytap_>
sorry ONE M300
<neggles>
one M200 in aus at the moment, AU$200 so about US$150
<neggles>
not worth
<mattytap_>
Ive even bid on M270, M500 (which I used when I was working)
<neggles>
M270s are lame
<mattytap_>
I ignore XTM
<neggles>
world's got enough Atom C3558 boxes in it
<mattytap_>
agreed and bad that WG have goe that route
<neggles>
the new stuff is NXP ARM
<mattytap_>
I think stintel has it right with his M300
<neggles>
M290/390/490/590
<mattytap_>
didnt know that, - thats good
<neggles>
various assortments of cortex-A72
<neggles>
the T20 is an LS1023A (2xA53), T40 is LS1043A (4xA53), T80 is LS1046A (4xA72)
<neggles>
everyone else is doing [some kind of X86] + [an octeon]
<neggles>
or straight x86
<mattytap_>
I'm really love NXP, I find their documentation great and its layered, I meant there is summary stuff, architecture discussions and timelines, tech manuals
<mattytap_>
all freely available, some via a free login
<neggles>
yup, it is quite good
<neggles>
will be keeping an eye out for the T80 and the Mx90s
<neggles>
only major annoyance is the u-boot password prompt, which is looking for a 24-char hex string that seems to be partially dependent on the serial number
<mattytap_>
Ive seen three T80's too expensive for me though.
<mattytap_>
Ah yes, I still have not cracked how to break into my XTM25
<neggles>
iirc the XTMs all have the same password
<neggles>
F5BA25AB44724fb5A6DD37554809CE34
<mattytap_>
I use it as a tunnel endpoint
<neggles>
this T20 was a stroke of luck, unregistered, 3yrs of license/support pre-attached to it, build date july 2021, AU$1200 worth
<neggles>
paid a tenth of that
<mattytap_>
I tred the one floting around for the XTM21
<neggles>
ah, no dice?
<mattytap_>
I like a PoE port though
<neggles>
i should just binary patch the T20 u-boot
<Namidairo>
but that's only 850 us dollarydoos
<mattytap_>
no, the XTM1 is based on Redland I think and the Uboot passwords a defo no the same, others on forums also state
<mattytap_>
you know, that will be my next, next project but I know zero about uboot
<Namidairo>
well at least you can take a look at the check when you're poking around with the image in your disassembler
shibboleth has quit [Quit: shibboleth]
<neggles>
the main() func sets a variable that i'm 99% sure is "password valid" to 0, then runs through a bunch of really annoying indirect function calls full of stack variable shenanigans
<mattytap_>
g trashed a t30 by overwriting the NOR and s25 flash I know how to apply iy
<neggles>
Namidairo: yeah, they made a very silly decision
<neggles>
it initializes the global var to 0, then it checks if it's 0 before it calls the password prompt code
<neggles>
if it's not, it skips straight to the prompt
<neggles>
er, to the u-boot prompt
<Namidairo>
got a c dev to do it
<Namidairo>
clearly
<neggles>
there's a goto() target between the initialization and the check
<mattytap_>
on the M200 and T30 I run with a third SD partition with no uImage, that way it boots to uboot prompt without having remove SD card
<neggles>
if I had a T40 that would work, T40 uses a SATA M.2 for boot
<neggles>
T20 uses an eMMC
<neggles>
though every single component needed for the M.2 is on the PCB *except* the actual M.2 socket
<neggles>
all the passives and whatnot are there... only bothered to mark one part as "do not place"
<neggles>
do you know if there's a way to get a root shell on fireware?
<neggles>
i have been spoiled by sophos, where you ssh in, press 5, press 3, congrats you're root
<mattytap_>
no, the ONLY way in Ive ever found for WG is if you can remove the UImage (by removing the SD card in T30/M200 case)
<neggles>
ew
<mattytap_>
and Ive googled the heck out of it too
<neggles>
they did conveniently leave the EMI resistors for the eMMC right next to it in a neat little line
<mattytap_>
BUT, once in you can work with the uboot environment and save it
<neggles>
I tried manually overwriting the u-boot environment with an spi flash clip but it seemed to ignore it
<mattytap_>
I prefer to leave it OEM and rename the uImage and dtb files
<neggles>
hmmm, i can desolder the resistor connecting the eMMC clock, then it won't boot
<neggles>
should drop to shell
<neggles>
the difficulty of accessing eMMC chips has become the bane of my existence lately; half a dozen different devices I can't get into
<mattytap_>
neggles: and Namidairo: thanks for chat, gotta run
<neggles>
openwrt kernel has fpu emulation disabled due to softfloat userland?
<Grommish>
But, we have software float in the kernel
<neggles>
evidently not (at least as of 2018)
<neggles>
lemme see
<Grommish>
It's under CPU Selection? Same screen as the SEGV numbers
<Grommish>
It's been a while.. its under target options too
<neggles>
there's soft-float, and there's soft-float with in-kernel float emulation
<neggles>
Grommish: heh. no CONFIG_MIPS_FP_SUPPORT
<neggles>
Select y to include support for floating point in the kernel including initialization of FPU hardware, FP context save & restore and emulation of an FPU where necessary. Without this support any userland program attempting to use floating point instructions will receive a SIGILL.
<jow>
Grommish: kernel side floating emulation is disabled in openwrt
<jow>
Grommish: if you want the package to work with vanilla images you need to do softfloat
<Grommish>
I do call -msoft-float
<neggles>
why do we not enable this? octeon II and III have FPUs at least...
<Grommish>
Doesn't seem to matter
<neggles>
Grommish: doesn't do anything if the kernel doesn't have float emulation
<neggles>
apparently
<Grommish>
neggles: Right, I just was lettng jow know I've tried that hehe
<Grommish>
neggles: Let me open kernel_menuconfig and see if I can find it.. we are going source_only so it won't really matter
<neggles>
kernel type menu
<neggles>
"floating point support"
<Grommish>
In menuconfig or kernel_menuconfig?
<neggles>
kernel_menuconfig
<neggles>
"if you know that your userland will not attempt to use floating point instructions then you can say N here to shrink the kernel a little"
<Grommish>
Oof, but I think I have that on.. I'm checking
<aparcar>
hauke: mind bumping the last missing 5.4 Kernel?
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<ynezz>
aparcar: are there any conflicts? Otherwise it should be pretty automagic with maintainer-tools/update_kernel.sh script
<ynezz>
aparcar: it's like running `maintainer-tools/update_kernel.sh -u -v 4.14 4.14.266` and then commit the results
<ynezz>
it gets tricky once some of the patches doesn't apply and you get some rejects
<aparcar>
ynezz: oh sorry it's about bumping the last target to 5.10
<rsalvaterra>
aparcar: Will bcm63xx bump itself to 5.10 if we ask it nicely? :)
<aparcar>
I'll give it an evil look
<rsalvaterra>
In other news, I'm compiling a 5.15-based ramips (mt7621) image.
<rsalvaterra>
I'm not run-testing it, however. I'm pretty sure the device tree needs some love and the eyes of someone who knows it better than me (i.e., probably everyone else :P).
f00b4r0 has quit [Ping timeout: 480 seconds]
f00b4r0 has joined #openwrt-devel
aleksander has quit [Quit: Leaving]
<russell-->
rsalvaterra: what's the worst thing that can happen?
minimal has joined #openwrt-devel
<rsalvaterra>
russell--: Without a serial console?
<rsalvaterra>
… I was actually convinced it didn't have holes. That's a bit better, yes.
<aparcar>
nbd: do you get openwrt-21.02 to compile on macOS? I'm seeing some --pie error on toolchain/gcc/initial
<aparcar>
I'm not on the latest commit to avoid confusion, but was it recently fixed there?
<nbd>
i'm seeing the same
<nbd>
gcc8 doesn't have darwin-aarch64 host support
<aparcar>
dang it
* aparcar
needs to invest in a linux build server
<aparcar>
nbd: do you know a workaround?
<Slimey>
:)
<stintel>
use Linux
<nbd>
aparcar: running ubuntu on the mac via multipass
* stintel
hides
<rsalvaterra>
stintel: Amen!
<nbd>
aparcar: i usually build master natively and whenever i need to build 21.02 or some other tree, i use multipass+ubuntu
<rsalvaterra>
And I don't get why Ubuntu gets so much flak. It's basically a polished Debian Sid (although snap can go forth and multiply, as far as I'm concerned).
dedeckeh has joined #openwrt-devel
minimal has quit [Read error: Network is unreachable]
minimalist has joined #openwrt-devel
<aparcar>
nbd: ohh multipass, curious. I'll have a look thank you
Tapper has quit [Ping timeout: 480 seconds]
minimalist has quit []
minimal has joined #openwrt-devel
T-Bone has joined #openwrt-devel
minimalist has joined #openwrt-devel
minimal has quit [Read error: Connection reset by peer]
goliath has quit [Quit: SIGSEGV]
<rsalvaterra>
Got ~100 kB difference from 5.10 to 5.15 with the same-ish config. Not great, not terrible.
f00b4r0 has quit [Ping timeout: 480 seconds]
minimalist has quit []
<aparcar>
rsalvaterra: how do you test kernel bumps? I'd like to glue that into CI
<rsalvaterra>
aparcar: Well, after the manual process (patch rebasing), I build an image and yolorun it. :P
<rsalvaterra>
As far as CI integration is concerned, you probably only care about the building part, no?
<rsalvaterra>
And since the the kernel bump is rather self-contained, you probably could get away with just building the kernel and nothing else.
* stintel
summons dangole
<aparcar>
rsalvaterra: yea that's the plan, I'd build a kernel for every target
<aparcar>
and sure enough I'm working on some hardware CI testing with labgrid :)
<mangix>
aparcar: no idea. I see exec format error, which indicates trying to run a binary for the wrong architecture
<aparcar>
mangix: yea Felix already helped me out, it requires real linux to work. sorry for not unpinging
<mangix>
rsalvaterra: why modifiy DTS?
<rsalvaterra>
mangix: It's required due to the upstream changes. Because we carry our own copy of the .dtsi, which differs greatly from the upstream one.
<rsalvaterra>
stintel: Is it possible to learn that power? :P
<f00b4r0>
what I have here is _slightly_ newer. If memory serves me right it's a pair of 1.2GHz 2xMadison rx2600. With nondescript RAM (a few dozen GB iirc)
<Grommish>
rsalvaterra: I have no idea what either of those mean, sir
<rsalvaterra>
Oh, sorry. I mean, if that's with (userspace) software float emulation, or with kernel-side emulation.
<Grommish>
Required both on at least mips64
<Grommish>
Or it would SIGILL on the Float instructions
<Grommish>
I have no idea about any other arch yet
<rsalvaterra>
Both? That's… not right. o_O
<Grommish>
I had to have the kernel CONFIG_MIPS_FP_somehing or other
<Grommish>
In addition ot the CONFIG_FLOAT
<rsalvaterra>
Nice process monitor there. Is that btop?
<Grommish>
and the MIPS_FP_xxx was in the kernel
<Grommish>
bottom
<Grommish>
rust app
<Grommish>
I needed something to test with eheh
<Grommish>
I have ripgrep too
<rsalvaterra>
Ah, the definition of overkill, then. :P
<Grommish>
Well, suricata refused to work andI didn't know if it was Me, the build system, the toolchain, or the project tiself.. After all, I'm the definition of outlier for any issues
<Grommish>
Or a mix :D
<Grommish>
Now, I've not slept but I'll be damned if I don't at least look at Suricata
<rsalvaterra>
Of course, I was joking. When nothing works, we need a simple "hello world". ;)
<Grommish>
I have one of those too, and neggles wrote it.. But, it broked
<rsalvaterra>
I wager your kernel config is custom and you had to manually update it.
<stintel>
nope
<rsalvaterra>
Puzzling. I just checked the mt7622 subtarget kconfig and there are none of the related symbols there. It should be inheriting them from the generic kconfig. :/
<mattytap>
stintel: finally found an affordable WatchGuard M300, being delivered for the weekend
<rsalvaterra>
You noticed because the kernel failed to build due to a missing ksym, right?
<stintel>
I noticed because my build didn't build because it was stuck on that question
<stintel>
what's weirder is that it continued
<stintel>
which shouldn't happen since some commit some months ago
<rsalvaterra>
Right, it should have just aborted. Doubly odd.
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<rsalvaterra>
I'm doing an mt7622 build here to see what happens.
danitool has joined #openwrt-devel
<stintel>
and it just asked again
<Slimey>
you lack discipline
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
eduardo010174 has quit [Remote host closed the connection]
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
svlobanov has joined #openwrt-devel
<rsalvaterra>
stintel: No issues building here (mt7622 defconfig) at all. :/
<paper_>
hi, is there a documentation for ALTERNATIVES? is the number priority? Is lower number prefered? Do I need to specify ALTERNATIVES in both conflicting packages?
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
svlobanov has quit [Remote host closed the connection]
<PaulFertser>
minimal: in other words, it should be specified in the documentation for the specific hw rng, and quality is "bits of entropy per 1024 bits of input", I see, thank you!
minimal has quit []
srslypascal has quit [Remote host closed the connection]
<neggles>
Grommish: it was 3am yes 😝
<neggles>
rsalvaterra: it appears that rustc only supports kernel side emulation, at least on mips64
<rsalvaterra>
Wait, I haven't done that, I just cat'ted /dev/hwrng. ;)
<zx2c4>
`dd if=/dev/hwrng of=blah count=1 bs=512` does this work and give you a file of `bs` length, for trying various different `bs` values? and when you open up that file with `hexdump -C`, does it look pretty random?
<rsalvaterra>
Actually, I was just going to do a dieharder test.
<zx2c4>
good idea
<zx2c4>
whats the output of `cat /sys/devices/virtual/misc/hw_random/rng_available` though?
<zx2c4>
and `cat /sys/devices/virtual/misc/hw_random/rng_current`
<zx2c4>
should make sure ath9k is selected
<neggles>
it’s not exactly a good move to use the hwrng directly, in most cases anyway - that’s what urandom is for, as long as the hwrng is crediting entropy to the pool
<rsalvaterra>
I haven't specifically done 'echo ath9k > /sys/devices/virtual/misc/hw_random/rng_current', so the kernel is probably using it by default.
rua has quit [Ping timeout: 480 seconds]
<neggles>
arguably “does urandom still pass a dieharder test with the hwrng enabled” is a better question
<zx2c4>
neggles: we're talking about debugging the driver i wrote today
<neggles>
yeah okay fair
<zx2c4>
nobody is talking about using /dev/hwrng directly for real usage
<zx2c4>
wrt your second comment, /dev/urandom goes through chacha20, so it will always pass the dieharder test regardless of the input
<zx2c4>
rsalvaterra: excellent thank you. so it looks like its enabled
<zx2c4>
before/after you do the dieharder tests, can you try `dd if=/dev/hwrng of=blah count=1 bs=512` for various values of bs=?
<zx2c4>
i want to know if it starts failing at some point
<neggles>
not necessarily, it *should* but with sufficiently bad entropy sources it may not
<zx2c4>
neggles: if you pass an all zero key to chacha20 and spit out a stream, it will pass the diehard tests
<zx2c4>
rsalvaterra: oh awesome. you can add that on top of my patch!
<rsalvaterra>
zx2c4: Way ahead of you. ;)
<zx2c4>
haha
kenny has joined #openwrt-devel
<neggles>
zx2c4: yes, but it’s *possible* to feed it a stream of bits that will make it not pass :P (though I guess it’s more that dieharder isn’t the be-all-end-all of RNG tests)
<rsalvaterra>
I actually have an AR5008 card here to test, but I've been too lazy. :P
<neggles>
buuuut I’ve been reading a lot of LKML discussion on entropy generation lately, heh, sorry 😅
<zx2c4>
neggles: if you can compute with a chacha key such that the output stream passes or fails certain tests, that's probably an indication that chacha is broken in some important way
<zx2c4>
s/with//
bluew has joined #openwrt-devel
<neggles>
i said possible, not likely…
<zx2c4>
if the computation takes 2^128, is it still "possible" in your mind?
<zx2c4>
i figure once you're there, the whole thing is meaningless
<neggles>
zx2c4: I am currently being very pedantic and slightly tongue-in-cheek yes :P
<zx2c4>
rsalvaterra: do you mean to say you're testing it on your mom's computer which has something supported by the current driver, or that it has something that requires your new patch on top?
<neggles>
AFAIK the various wifi card RNGs are considered mostly trustworthy anyway
<rsalvaterra>
zx2c4: Exactly. It has an AR9462 Wi-Fi adapter.
<zx2c4>
rsalvaterra: that was a "1" or "2" question. when you say "yes" i dont know which one you mean ;-)
<rsalvaterra>
zx2c4: Sorry, lazy evaluation. xD
<zx2c4>
(i dont know anything about ath9k to know which cards are supported by the driver i apparently wrote today)
<zx2c4>
rsalvaterra: well, anyway, i put a copy of the email in your inbox, in case you want to reply to it with `Tested-by: Rui ...` and some info about your experiments
<zx2c4>
(check spam if it doesnt show up in a few min)
<rsalvaterra>
zx2c4: The hwrng (ADC entropy collection) feature of ath9k hardware has only been thoroughly tested in AR9300+ adapters, so it's only available on those. My patch adds support for all ath9k hardware, all the way back to the first AR5008 devices.
<zx2c4>
ohhh okay
<zx2c4>
and your mom's computer has a new one. gotcha
<rsalvaterra>
So, let's see if I still know how to invoke dieharder correctly…
<rsalvaterra>
I think it's the generator 201.
<zx2c4>
dieharder -a -f /dev/urandom -g 201
<zx2c4>
that seems to work on my machine
<zx2c4>
i guess try with /dev/hwrng
<rsalvaterra>
Yep, that's what I'm using. Needs root to access /dev/hwrng, though.
<rsalvaterra>
And it's probably going to take a while..
<zx2c4>
rsalvaterra: anyway what im most interested in is not how much ath9k sucks to be an rng but rather whether the kernel programming there is correct. then after the patch lands we can talk about just all together removing the feature if it truly does not work
<zx2c4>
Habbie: rsalvaterra wtf just happened here. rsalvaterra almost quoted verbatim the dilbert "thats the problem with randomness...", but then professed to only have seen the xkcd. meanwhile Habbie used the number 4, which is the number in the xkcd, but professed to only have seen the dilbert...
<rsalvaterra>
I have no idea! :)
<zx2c4>
hehe
<rsalvaterra>
Oh, wow…! I remember reading that Dilbert strip at the time. Damn, I'm old. :P
<mangix>
ath9k still gets worked on years after the hardware gets obsolete. Cool.
<Habbie>
zx2c4, oh no, i had seen both, but was mostly trying to emulate the dilbert - clearly with the number i subconsciously took from the xkcd :D
<Habbie>
zx2c4, i cannot explain rsalvaterra though ;)
dedeckeh has quit [Remote host closed the connection]
<rsalvaterra>
I'm me and even I can't explain myself.
<Habbie>
oh yes, i did not say "I cannot explain that one thing rsalvaterra said" ;)
rua has quit [Ping timeout: 480 seconds]
<rsalvaterra>
zx2c4: This email you sent, shall I just reply to it normally with my Tested-by?
<zx2c4>
rsalvaterra: yes, reply-to-all. it's the original email & message id as the one from lkml
<mangix>
Wonder of mediately wifi supports rng stuff
<mangix>
Mediatek. Love autocorrect.
<zx2c4>
rsalvaterra: include some notes on what you tried and then Tested-by i guess
<mangix>
Much love to wireguard btw. I'm currently using it to access my NAS and access IRC
<zx2c4>
gladyoudigit
rua has joined #openwrt-devel
<rsalvaterra>
Speaking of which…
<rsalvaterra>
… no clue about my roaming issue? :)
* mangix
has no idea
<rsalvaterra>
mangix: Sorry, that was for zx2c4, not you. ;)
<rsalvaterra>
Apart from that minor snag I've hit, WireGuard has truly been a life saver. Especially for using banking apps while travelling (without triggering safety measures). Phone's been at home all the time. :P
<Habbie>
<3 wireguard
philipp64 has joined #openwrt-devel
<Borromini>
very happy here as well. Way easier than OpenVPN to set up as well <3
rua has quit [Ping timeout: 480 seconds]
Borromini has quit [Quit: leaving]
rua has joined #openwrt-devel
Thagabe has quit [Quit: Connection closed for inactivity]