Grommish has quit [Read error: Connection reset by peer]
shibboleth has quit [Quit: shibboleth]
fblaese has quit [Quit: bye]
goliath has quit [Quit: SIGSEGV]
<neggles>
takimata: if it's been fixed in 21.02.2 there's your answer :P could have been a lot of things, but i'm not seeing any commits between .1 and .2 which affect overlayfs or x86_64 specifically, so it was probably a kernel bug fixed somewhere between 5.4.154 and 5.4.163
<takimata>
neggles: slh made a good observation on #openwrt. first of all it's the difference between f2fs and ext4 used for the overlay, something (probably some size threshold) triggered use of f2fs in 21.02.1 and ext4 in 21.02.2
<takimata>
I could confirm that 21.02.1 is indeed using f2fs, 21.02.2 is using ext4.
<takimata>
the overhead for f2fs seems to be _huge_ in this instance.
<neggles>
ahhhh
<neggles>
takimata: on the overhead, maybe, maybe not - it may just be a difference in how overlayfs reports used space between the two
<takimata>
possible.
<neggles>
takimata: what image did you use? there's two kinds of x86_64 image
<takimata>
squashfs-combined-efi on both versions.
<neggles>
interesting
<takimata>
(I use squashfs for its reset-ability.)
Grommish has joined #openwrt-devel
<Grommish>
Wow.. I just found one of my old HTC One S devices that not only w3orks, but is still running the custom ROM I created for it.. I wonder if I can repurpose it for anything
<mangix>
Grommish: clock
<neggles>
home assistant touchpanel
<Grommish>
It has Beats Audio before it went to shite
<Grommish>
Maybe for that as a media hub
<Grommish>
Dunno.. I'm surprised it still works at all
<neggles>
takimata: that's... kinda weird. both have the same exact config, only change made to the target between the trees is the kernel version bump, and the build scripts haven't changed,
<neggles>
weird
<mangix>
oh no nvm that's ancient
<Grommish>
Yah.. But, it's running Lollipop at least
<Grommish>
I doubt I'll do better anymore
<Grommish>
Ah, the days when "Predictive Text Input" was a feature
<mangix>
you could use is an a wifi adapter
<mangix>
*it as an
<takimata>
neggles: right? slh suspected something hit some set threshold to decide between f2fs and ext4, which 21.02.1 barely missed and 21.02.2 barely crossed.
<Grommish>
True..and it has a headphone jack which is nice
<mangix>
oh
<mangix>
also drivedroid
<Grommish>
Yep
<Grommish>
I love DriveDroid
<Grommish>
and I've already got root
<neggles>
takimata: i got the 21.02.2 kernel version wrong, it's .179
<neggles>
and the last commit applied to .179 before tagging
* takimata
uses an old smartphone as a clock/weather station (self-coded "big pixels" clock and regular API calls to openweathermap). phones are nice for that, good size and automatic brightness ajustment.
<neggles>
may have something to do with it
<mangix>
main annoyances I have with it are: no sparse file support and no UDF support
<mangix>
can't help the last part though
<neggles>
i wonder if building a forced-f2fs image would have the same difference or not
<neggles>
I didn't know about drivedroid and
<neggles>
welp guess it's time to buy a used [something i can root with a decent amount of storage]
<neggles>
preferably something smallish...
<Grommish>
neggles: Yah, it's very handy as long as you can get root
<neggles>
oh USB3 OTG support would be good to have as well hey
<Grommish>
OTG is built into most Android devices
<neggles>
USB 2.0 boot is painful by comparison
<neggles>
yeah but many are only USB HS in OTG mode
<Grommish>
Meh
<neggles>
hey google where's the pixel 6 mini
<takimata>
ooh, drivedroid is nifty. thanks for the pointer.
<neggles>
so this is just using the USB mass storage gadget and giving you a nice shiny UI to switch between / manage images isn't it
<takimata>
major flaw: doesn't have OpenWrt images in its repository :)
<neggles>
why would you want to liveboot openwrt? (I guess you could add a storage partition and use it to install openwrt)
<takimata>
I used OpenWrt on a USB stick to get it onto my x86 firewall thingy. first got OpenWrt booted, then dd-ed the usb stick wholesale to the DOM module inside. :)
<mangix>
honestly Ventoy has mostly replaced DriveDroid
<Grommish>
neggles: I've nly used ISOs but it's been a god-send when you actually need it.. Esp since you can just load the ISO on the fly
<takimata>
works a treat, no need to add storage and carry an image. the USB stick itself is the image. :)
<neggles>
i mean you can (and probably should) just download the image and use sysupgrade running against the disk you want to install on
<neggles>
or at least clean rootfs_data so firstboot runs again
<takimata>
yeah, firstboot-ed after that, ofc.
<neggles>
I wonder how hard it would be to emulate a displaylink chip with the usb gadget framework
<takimata>
but that was way easier than getting the DOM module out and figure out how to provide power.
<neggles>
do the other side of the udlfb driver (which is apparently very similar to regular fbdev)
<neggles>
fbdev is "dead" though
<neggles>
takimata: what system was this??
<takimata>
neggles: a "gateprotect gp-u 50", originally sold as a firewall appliance.
<neggles>
oh hey a cas-well box
<takimata>
oh, that's the OEM? interesting.
<neggles>
probably, it'll be them or ADLink or the third one i can't remember
<takimata>
I have another "gateprotect" here that's based on a lanner/lantec design.
<neggles>
ah yeah lanner and there's at least one more
<takimata>
nice little boxes that.
<neggles>
cas-well make All Of The Sophos And Checkpoint Devices
<neggles>
and a bunch of others
<takimata>
I believe my GP-U 50 is also sold by sophos, so you might be onto something there.
<neggles>
what CPU/etc?
<takimata>
or ... "was sold", it's EOL of course.
<takimata>
Atom E3805, four intel gigE, 2 gigs of ram, 32 gb SATA DOM.
<takimata>
not incredibly powerful of course, but fits my needs well and runs quiet (no fans).
<neggles>
ah yep that's a sophos UTM 100/110/120
<takimata>
yup.
<neggles>
i have one sitting on the floor here
<neggles>
except sophos used a 160GB HDD
<neggles>
I'd offer to give you it before I throw it into the ewaste pile but shipping would not be worth it :P
<takimata>
hang on to it for when I finished moving to Japan. :)
<neggles>
express airmail via your local postal service rather than DHL/FedEx/etc
<takimata>
neggles: depends on the destination country. they don't do economy air at all, but most countries can get the cheaper-and-better-anyway e-packets now.
<Grommish>
neggles: Ah.. Gotcha..
<neggles>
takimata: no no, it's a temporary thing, regular old airmail shipments are suspended to/from japan and several other countries
<takimata>
neggles: 2kg e-packet from japan to germany is something like 2000円, registered.
<neggles>
china's done the same atm
<takimata>
neggles: most people don't know about e-packets at all, and it's very well hidden on the japanpost website.
<neggles>
> All services from Japan to Australia are suspended except for airmail letters and surface mail.
<neggles>
going in the other direction it's EMS only atm, the other two options are disabled
<takimata>
yeah, "depends on the destination country" ... fwiw, EMS was the only option to europe until late 2021 too.
<neggles>
we had it available prior to The Pandemic and i'm sure it'll be back sooner or later, but, point being, I can't send this thing to japan for less than $50 and it is not worth $50 :P
<takimata>
true.
<takimata>
it's not worth considerably _less_ than that, IMHO, but $50 makes it a wash. :(
<neggles>
takimata: I couldn't sell it for $50 on ebay here so :P
<takimata>
that might be because noone except us lunatics know how to make use of it once the sophos/gateprotect license has run out.
<takimata>
(except for the pfsense/opnsense lunatics, of course)
<neggles>
i'd say it's more because the E3826 is catastrophically slow
<neggles>
though hey it's faster than what's in yours :P
<neggles>
oh that's right, this board has a CF slot on it
<neggles>
and a miniPCIe... i wonder if the USB is wired up
<takimata>
no worries, I have another one with a J1900 if this one doesn't hold up on a gigabit line. my current line is 100/40 so it's mostly bored anyway.
<neggles>
it definitely will not hold up on gigabit
<neggles>
J1900 won't either really, depending on how complex your setup is
<neggles>
the J1900 suffers from a severe lack of AES-NI, so VPN throughput is a disaster
<takimata>
I switched pretty much everything to WireGuard anyway.
<neggles>
WG still benefits from some AES-NI instructions
<neggles>
it's just not as night-and-day as ipsec :P
<takimata>
but I hear you. they might not be the greatest-ever. but they still run circles around every consumer router.
<takimata>
which is what I upgraded from.
<neggles>
the MT7622A probably outperforms a J1900 with hwnat accel... depends on your use case *shrug* none of the above will handle SQM/etc on a 1G link
<neggles>
fortunately you don't really need it on a 1G link
* takimata
nods.
<slh>
Grommish: a https://www.gsmarena.com/htc_legend-3141.php with CM 7.2.0 (gingerbread, 2.3.7) still makes a good dumb phone, with the wireless disabled up to 3 weeks battery runtime
<takimata>
maybe the CF slot is throwing me off, mine doesn't have it.
<takimata>
oh, wait, that is a RAM slot.
<neggles>
there's mPCIe, CF, DDR2 SODIMM
<takimata>
mine has none of that. everything except the sata drive is onboard.
<neggles>
ah ok
<neggles>
might be a nexcom then
<neggles>
on the customized/cut-down versions they often remove stuff, e.g. the XG 135 i have uses a board that's typically got a CF slot and a miniPCIe and a few things, the pads are there but not placed
<takimata>
I do remember I was looking for details on the mainboard, and I came up empty.
<takimata>
I would have to open it up again, sorry.
<takimata>
meh, I should have taken pictures.
<neggles>
not important :P
* takimata
nods, again.
<takimata>
although it would be nice to find the actual manufacturer and model, bios updates and such.
<takimata>
but next time I have it open I will remember your pointers.
<Grommish>
slh: Nice! This is running our 5.1.1 Lollipop ROM stiff
<neggles>
takimata: ok so this cannot be an E3826 b/c it has DDR2, I think what i have is a CAD-0205, which is an Atom N450
<neggles>
p a p e r w e i g h t
<takimata>
eek.
<takimata>
oh you're right ... I wasn't thinking straight ... DDR2 is a giveaway.
<neggles>
yeeeeep. this is an 0205
<neggles>
I knew there was a reason it was headed for the ewaste pile :P
<takimata>
bummer, really.
<neggles>
oh don't worry I have *plenty* of stupid firewall appliances
<neggles>
including one with a bypass adapter
<Grommish>
neggles: Ah, but not a Itus Shield ;p
<Grommish>
neggles: But, I wish you would pick one up haha
<neggles>
well, "bypass adapter", it's built into the board, there's some extra registers you can tweak to switch between bypass/dual-port/packet-sniffer mode
<neggles>
been meaning to write a kmod to expose that in sysfs
<neggles>
Grommish: never found one for sale around here :P
user95 has joined #openwrt-devel
<philipp64>
hi... i'm testing a patch in Asterisk and it's coredumping. What's the easiest way to decode a stacktrace?
<philipp64>
it looks like it's happening in an .so, so I can't just run "nm" on asterisk and use those addresses...
<Grommish>
philipp64: ./scripts/remote-gdb might help
minimal has quit [Quit: Leaving]
<Grommish>
as long as you have remotegdb on the device to act as the server end
<Grommish>
and -ggdb3 is a good tag to add to your TARGET_CFLAGS to force symbols and debugging info
<Grommish>
then its just ./scripts/remote-gdb <ip>:<port> <path/to/build_dir/nativebin>
<philipp64>
Grommish: you mean gdb-server on the target?
<Grommish>
philipp64: Yes, you'll have to put that on the device to connect, sorry :) I wasn't looking at the name
<Grommish>
but that'll let you use gdb as you normally would, which might help
<philipp64>
just wanted to make sure I was following...
<Grommish>
oh.. Hint: Set with gdb-server as One Once, or remember to 'monitor exit' before you quit gdb locally, or it'll just sit there (at least on mine) with no way to kill it unless you log in again
<Grommish>
err Only Once
<Grommish>
I ended up having to power cycle the device so many times before I forgot :D
<slh>
neggles: the j1900 should cope with routing/ NATing 1 GBit/s at wirespeed, the E3815 probably as well - in terms of SQM, I'd be hopeful for the j1900, not so sure about the E3815. just for reference, the dreadfully slow Atom n270 (1c2t) can do ~600 MBit/s without SQM
<slh>
based on my testing with Atom 330 (2c4t, 1.6 GHz), that one can host iperf3 at 850-930 MBit/s, depending on its mood - so it shouldn't be that far away from managing to route 1 GBit/s (without SQM, of course)
<neggles>
slh: depends on how long your firewall rule chain is :P
<slh>
neggles: sure, I was referring to 'normal' setups, close to vanilla OpenWrt (and not having to do PPPoE)
<neggles>
so what i'm hearing is "you should put openwrt on this N450 sophos"
<neggles>
s/this/that
<slh>
I was seriously considering to buy one of those (gateprotect gpo-150 v1, atom d525) for under a 10 EUR this week, but effectively let it pass in the end (it sold for 2.50 EUR, plus 4.99 shipping)
<slh>
just saying, in terms of performance they can still be useful - idle power consumption might be a money sink though
<neggles>
yeah, I am not short of intel firewalls
<neggles>
and it caps out at 2GB of RAM + no 64-bit support
<neggles>
time to let it die / stop taking up space in my house
<slh>
2 GB RAM wouldn't be an issue for me, Mem: 3929316 55512 2790080 3740 1083724 3788684 ;)
<neggles>
true :P oooof, yeah, uh, idle power draw is not great
<slh>
I can't get the atom n270 under ~25-28 watts idle, the atom 330 more around 34 watts - and j1900 ~6 watts; but in the end, I use an ivy-bridge c1037u running at 15 watts idle (but still looking for more ethernet ports and less idle power consumption, but there's not much to improve - so I'm picky ;)
<neggles>
slh: yeah, sitting at the "no boot device found" menu this thing is pulling 15w of real and 26w of apparent power
<neggles>
a sophos SG 230 with an i3-4150 in it idles at 21w
<slh>
yeah, haswell has both the power and can be low power (chosen correctly, down to 10-12 watts idle)
<neggles>
SG 230 has six i210-AT NICs and a 120gb intel DC SSD, plus an expansion slot - that was with VyOS on it and the governor set to performance, too
<neggles>
the 4x10G module you can drop into the expansion slot is absolutely cursed, though. It's got three 24-lane PLX chips on it
<neggles>
that card pulls more power idle than the whole rest of the thing.
<slh>
the sophos sg/ xg devices are sadly still a bit beyond of what I'm looking for in terms of used prices
<neggles>
oh? the SG 115/125/135 are usually <$100, i've seen them go for 50 (AUD)
<neggles>
they're Atom C2000 though so they need a little resistor soldered in to prevent them dying of AVR54
<neggles>
this 135 stopped POSTing because of AVR54, one 120 ohm resistor later and it's fine
<slh>
I've been actively hunting for the SG 105 or 115 rev 2, none seem to sell under 100 EUR in .de
<neggles>
boo :(
<slh>
that's why I settled on the c1037u, despite its 15 watts idle and only two e1000e ports
<neggles>
you may be able to find the 210 and 230 for cheaper since they're the big boi rackmounts
<neggles>
I was surprised the idle power was so low
<neggles>
with a haswell refresh i3 you get ECC memory support too which is nice
<slh>
around 150 EUR for the sg 210, around 180 EUR for the sg 230
<neggles>
ouch
<neggles>
the chip that they come with is a pentium, 2c2t, with no AES-NI (which is insane given that one of the main marketing points on these is SSL VPN...)
<philipp64>
-ggdb or -ggdb3?
<slh>
easily available, but kind of expensive
<neggles>
yeah, not worth it considering you'd want to swap the processor out anyway
<neggles>
welp. this thing's power draw has sealed its fate
<philipp64>
Grommish: or -g3 -ggdb ?
<Grommish>
philipp64: I've only use -ggdb3
<Grommish>
in TARGET_CFLAGS=-ggdb3
<slh>
on the other hand, two watchguard m300 sold for under 25 EUR each within the last month - but the idle power consumption (~23 watts), noise (3 small/ fast spinning fans) and the long term viability with OpenWrt kept me from pulling the trigger
<Grommish>
Or, I guessit should be: TARGET_CFLAGS+=-ggdb3
<Grommish>
Sorry
<neggles>
ooh, the bios chip is socketed, how convenient
<neggles>
8mbit, boo
<neggles>
slh: you can probably get the power consumption down just by tinkering with the fan speed control
<slh>
neggles: probably, but where to start, where to stop - at some poing, even ~100 bucks for a sophos end up being cheaper (after replacing PSU, fans, etc.)
<neggles>
slh: oh i wouldn't even replace them, just turn the speed down, but yeah, not a guarantee and you can often end up with an annoying growl
<takimata>
echo 80 > /sys/devices/platform/f71882fg.2576/pwm1_auto_point5_pwm ^_^ (that's on my Gateprotect GPO-150v2)
<takimata>
turned it virtually silent.
<neggles>
works great for intel systems, less so for freescale powerpc :P
<takimata>
the difficult part was which kmod to install to adress the io sensor chip.
<philipp64>
Grommish: I can't figure this one out. In Asterisk, which is threaded, why would alloca(24) return NULL? This makes no sense...
<Grommish>
philipp64: That I can't answer.. I don't know code. I just had reason recently to use remote-gdb. Best I know how to do is set args, run, bt, and x/$pc is about all I know how to do
<neggles>
philipp64: got a trace/dump? :P (also check dmesg and boot with loglevel=8)
<philipp64>
neggles: I'm wondering if it's not home hinkiness caused by over-optimizing inlines, so trying it as a regular function...
<neggles>
philipp64: maybe? without knowing where in the code it's failing i can't really comment :P
<nick[m]1234>
hey, it is no longer possible to flash m2/m5 nanostations, because sysupgrade is killed due to lack of ram ressources. I wanted to put it to tiny, but people said it still works in default. However, now it does not work anymore. https://github.com/openwrt/openwrt/pull/4879
kerneltoast has quit [Server closed connection]
kerneltoast has joined #openwrt-devel
<neggles>
philipp64: "You need to be logged in to view this paste." ewwwwww (i have an account though but)
<neggles>
the -O0 implies that it's something to do with inlining yea
goliath has joined #openwrt-devel
<neggles>
oh
<neggles>
philipp64: use ast_alloca() from astmm.h, not alloca()
<stintel>
hmmmm looks like usteer might be causing password popups in networkmanager on client
<neggles>
philipp64: sorry, from asterisk/astmm.h
<neggles>
stintel: repeatedly deauthenticating a client causes networkmanager to conclude that it has the wrong password? sounds like a WONTFIX to me :P
<aparcar[m]>
hauke: I think you added the "Netgear RAXE500" to the "supported devices" of 21.02.2, but I can't really find the reference, not even in snapshots. Was there a copy&paste error?
<aparcar[m]>
nick: is the problem only for sysupgrades or also for the initial firmware flash? I wonder if we could "kill" more services before performing the upgrade
<nick[m]1234>
aparcar: not sure, however I already killed uhttpd, dnsmasq and so on, and it did not help. :/
<nick[m]1234>
I need to tftpflash again, to go to a new version
<aparcar[m]>
rmilecki: hey you did some work on the RAXE500 right?
<aparcar[m]>
is it maybe "not ready" yet?
<aparcar[m]>
nick: the sysupgrade script stop quite a lot of services, maybe there is more we can stop
<aparcar[m]>
?
<rmilecki>
aparcar[m]: i started, yes
<rmilecki>
not ready yet
<rmilecki>
I need U-Boot for it & testing
<rmilecki>
that's why it's default n for now
rua has joined #openwrt-devel
<nick[m]1234>
aparcar: I can look again. I think we definitely get to some point where we can not disable more... ^^ why not just going directly to tiny again? I will have a look
<aparcar[m]>
nick: thank you
chder has quit [Server closed connection]
chder has joined #openwrt-devel
<aparcar[m]>
jow: what is your plan on documenting ucode? I'd write some docs for the libs (e.g. fs)
<jow>
planned to write more documentation throughout the summer
slingamn has quit [Server closed connection]
slingamn has joined #openwrt-devel
rua has quit [Remote host closed the connection]
<takimata>
so ... to bring this topic up again: x86 builds, more concrete the squashfs-combined-efi, seems to switch between f2fs and ext4 for the overlay, resulting in wildly varying free space (due to f2fs overhead?): https://pastebin.com/N3syNLrJ
<takimata>
collective puzzlement about the reason, guesses are some threshold is missed in 21.02.1 and hit in 21.02.2
rua has joined #openwrt-devel
<takimata>
no commit between 21.02.1 and 21.02.2 seems to point to any particular reason why this would happen.
<takimata>
not a problem, certainly not a breaking issue, but maybe some investigation is warranted?
<PaulFertser>
takimata: when available space for the overlay is more than F2FS_MINSIZE (100ULL * 1024ULL * 1024ULL) it uses F2FS, ext4 otherwise.
<takimata>
PaulFertser: that was our guess ... so this threshold is set juuuuuuuust about right for the x86 builds to switch, maybe it needs a bit finetuning?
<PaulFertser>
takimata: looks like it, yes.
<PaulFertser>
takimata: please create a github ticket so it's not forgotten. For the reference, the code in question is in fstools libfstools/common.c use_f2fs()
<takimata>
will do, thanks for looking into the reason.
srslypascal has quit [Remote host closed the connection]
goliath has quit [Quit: SIGSEGV]
srslypascal has joined #openwrt-devel
Spr0cket has quit [Server closed connection]
Spr0cket has joined #openwrt-devel
Lynx- has joined #openwrt-devel
<Lynx->
Do bash scripts sometime spawn child processes even when commands or functions are not explicitly run with '&'?
<aparcar[m]>
how do I add a proper header to a quilt patch?
<aparcar[m]>
why do we even use that while using Git in parallel?