philipp64 has joined #openwrt-devel
<neggles> Grommish: it does, but the way it boots is
<neggles> unpleasant
<neggles> it's probably fairly similar to the ER-4/6 though
philipp64 has quit [Quit: philipp64]
Tapper has quit [Ping timeout: 480 seconds]
<Grommish> neggles: Hmm.. mine is FAT-based loading and I just can tftp the .bin and load it into RAM. Keeps it from screwing up a working image
<Grommish> of course, nothing is persistent that way either, so.. *shrug*
<Grommish> neggles: Yeah, sitting idle for an hour (1.6KiB thru the br-lan), RAM has gone from 46204 to70516. It' a slow creep, but there
Grommish has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
<neggles> hmm.
<neggles> oh he left
goliath has quit [Quit: SIGSEGV]
cmonroe has joined #openwrt-devel
Grommish has joined #openwrt-devel
shibboleth has joined #openwrt-devel
minimal has quit []
SamantazFox_ has joined #openwrt-devel
<shibboleth> so, the openvpn people ask if openwrt is open to testing openvpn-dco/openvpn 2.6 (openvpn at wireguard speeds)
<shibboleth> offer it as a sep package for those bulding openwrt themselves
shibboleth has quit [Quit: shibboleth]
isak has quit [Remote host closed the connection]
Tapper has joined #openwrt-devel
isak has joined #openwrt-devel
cmonroe has quit [Ping timeout: 480 seconds]
dangole has quit [Quit: Leaving]
hanetzer has quit [Quit: WeeChat 3.4]
hanetzer has joined #openwrt-devel
philipp64 has joined #openwrt-devel
srslypascal has quit [Ping timeout: 480 seconds]
ecloud has quit [Ping timeout: 480 seconds]
<Namidairo> now how do I split up this long command in commit message without angering people's eyeballs
ecloud has joined #openwrt-devel
valku has quit [Remote host closed the connection]
GNUmoon2 has quit [Ping timeout: 480 seconds]
rua has quit [Ping timeout: 480 seconds]
srslypascal has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
GNUmoon2 has joined #openwrt-devel
pmelange has joined #openwrt-devel
goliath has joined #openwrt-devel
<aparcar> nbd: if you have the time please look at my changes of package-ipkg.mk. I can't figure out why "all" packages are added to the resulting image rather than only packages-y. It looks like all kmods are added to the linux.install file, which seems wrong.
<champtar> Anyone for my opkg patch ? it fixes some packages CI flake: https://patchwork.ozlabs.org/project/openwrt/patch/20210907224245.4993-1-champetier.etienne@gmail.com/
<champtar> jow: as you are opkg maintainer any opinion ?
ldir has joined #openwrt-devel
<aparcar> ldir: morning
<rmilecki> rsalvaterra: broken bisect is usually caused by changes from 1 specific subsystem maintainer
<ldir> aparcar: morning!
<rmilecki> rsalvaterra: git skip usually goes only one commit either way
<aparcar> ldir: since you're working on the new dnsmasq, should I close my PR?
<rmilecki> rsalvaterra: it's better to use "git bisect visualize" and manually jump to another top-git-level Merge commit
<rmilecki> rsalvaterra: with git reset --hard SHA
<rmilecki> and test it
<jow> champtar: lgtm
<ldir> it's up to you - I doubt I'll push to master as I'm not around much - I just keep my tree following latest things including the commits post the 'test5' - I guess the driver is the nftables support - which I have as another commit - feel free to cherry pick or whatever :-)
<aparcar> ldir: I'll cherry pick things to get more comment
<aparcar> thanks
<ldir> I wish simon would do another test release 'cos the later patches don't apply 'cos they patch a debian related file that doesn't exist in the tarball.
<ldir> that's why patches 19&20 don't exist and why 21 is edited to remove reference to the debian changelog
<ldir> gotta go to work - have fun
ldir has quit [Quit: Textual IRC Client: www.textualapp.com]
<rsalvaterra> rmilecki: git bisect visualize? Never used that, nice to know. :) However, every single merge commit I hit gave me a broken boot. I've skipped so many times that I've done 30+ builds and git keeps saying there are roughly 12 steps to go, the same as the beginning.
<rmilecki> rsalvaterra: if you are testing different Merge commits and all are bad, sounds fishy
<rsalvaterra> I'm starting to think I could probably do a better job than git selecting the commits. :P
<rmilecki> maybe your .config got wrong/
<rsalvaterra> rmilecki: That would be very strange. A bad .config giving a completely broken (black screen) boot? I'd understand missing functionality (drivers), but I'm not even getting a single printk… :/
<rmilecki> rsalvaterra: i had that happening on Broadcom devices
<rmilecki> can't tell you exact .config option that caused that
<rmilecki> but i remember that some config options could preven kernel from starting to boot (no printing anything)
<rsalvaterra> I'd expect x86-64, as a platform, to be a bit more robust in that regard… :)
pmelange has left #openwrt-devel [#openwrt-devel]
<rsalvaterra> I'll keep digging, however… I'm not in a hurry to find the root cause. It baffles me, however, that (of all the applications used here at home) only a solitaire game stopped working in Linux 5.17-rc1. I wonder what's so special about it…?
<champtar> jow: so there is no confusion you will merge it ? or I need to hunt down someone else ?
<champtar> I need conntrack support for nft bridge, I've made a patch to just add what's needed to kmod-nft-bridge package, is it ok or should I split it in another package ? https://github.com/openwrt/openwrt/pull/4995
<stintel> nbd: was anything wrong with my airtime settings that could explain ping >1000ms and barely usable connection ?
f00b4r0 has quit [Quit: brb]
f00b4r0 has joined #openwrt-devel
ecloud has quit [Ping timeout: 480 seconds]
ecloud has joined #openwrt-devel
<nbd> stintel: i don't see anything wrong. does it only happen with multiple clients, or is one client enough?
<stintel> nbd: most likely a regression then. I've not touched this config since I've initially set it up around the time it was introduced ... I've not tested a single client
<stintel> I could probably bring up an extra SSID just for testing that
<stintel> as for the performance regression in mt76.git master I'll try to bisect
<stintel> should have ordered more than 2 APs :/
<stintel> and right now they seem to be sold oud everywhere
<stintel> actually, if I can figure out that NAND issue with the ZR-2662 I'll can use those for testing
<stintel> they're mt7915e like the EAP615-Wall
rua has quit [Ping timeout: 480 seconds]
<stintel> s/'ll//
pmelange has joined #openwrt-devel
rua has joined #openwrt-devel
<nbd> stintel: btw. which chip produced the performance regression with your phone?
<nbd> 7915?
<stintel> nbd: 7915e
<stintel> 01:00.0 Unclassified device [0002]: MEDIATEK Corp. Device 7916
<stintel> 02:00.0 Unclassified device [0002]: MEDIATEK Corp. MT7915E 802.11ax PCI Express Wireless Network Adapter
Tapper has joined #openwrt-devel
pmelange has left #openwrt-devel [#openwrt-devel]
GNUmoon2 has quit [Remote host closed the connection]
GNUmoon has joined #openwrt-devel
danitool has joined #openwrt-devel
Acinonyx_ has joined #openwrt-devel
Acinonyx has quit [Ping timeout: 480 seconds]
arjun has joined #openwrt-devel
<arjun> I have a QUALCOMM development board based on IPQ6018. It currently has chaos calmer installed. How do i integrate the drivers with the current release 21.02.1.
<stintel> you don't
<stintel> qualcomm is a disaster
<stintel> once we branch the next release, and bump kernel in master to 5.15, things might become easier
<stintel> btw, your ipq6018 board does not run chaos calmer, it runs QSDK. it's absolutely not OpenWrt
<arjun> didnt know that thanks for the info
<PaulFertser> At certain point of time Atheros tried contributing to OpenWrt properly...
<arjun> It does have feeds set to openwrt and i have installed a few packages from it. Not sure whats going on
<stintel> arjun: QSDK is based on OpenWrt Choas Calmer, which is ancient, EOL for many years. QCA uses their own toolchain, kernel, it's as far away from upstream as you can imagine
<neggles> iirc the reason it's marked as chaos calmer has something to do with a licensing change
<neggles> cause it seems like every single proprietary vendor SDK was forked at 15.05
<stintel> it doesn't really matter. it's old, it's crap, it should not be used
<neggles> the closer you look at QSDK 11 the more you discover it's actually 19.07ish
<neggles> and yes, crap
<neggles> arjun: what kernel is it running?
<stintel> I bet 4.4 :P
<stintel> but not even close to 4.4.299
<neggles> QSDK 12 runs 3.14, 3.18, 4.4, 4.9
<neggles> https://wiki.codeaurora.org/xwiki/bin/QSDK/ they have not yet updated the wiki to contain qsdk11 / ipq6000
<neggles> s/qsdk11/qsdk12
<neggles> hot damn, QSDK 12.0.r9 uses kernel 5.4
<neggles> what a time to be alive
<arjun> its running version 4.4
<neggles> aaaand stintel takes home the gold
<neggles> what patchver?
<neggles> 4.4.79?
<stintel> 4.4.60 ;)
<stintel> also 32bit arm ;)
<neggles> there are 32-bit and 64-bit profiles for the ipq6000 :P
<stintel> I don't believe 64 bit was supported
<stintel> to summarize: "Total Fucking Crap"
<neggles> this is 11.2 actually
<stintel> it might be there, doesn't mean it's officially supported
<neggles> well the ubiquiti ipq5000 images are 64-bit...
robimarko has joined #openwrt-devel
<robimarko> They actually do support 64 bit
<neggles> how do you *do that*
<robimarko> What?
<neggles> I mean it's not really a mystery but
<neggles> the whole ~appear in the middle of a conversation~ thing
<stintel> whitequark, matrix, ... ? :)
<robimarko> I am lurking on ht elog
<robimarko> *the log
<neggles> I figured it had to be that :P
<neggles> but yeah, the ipq807x target in mainline is aarch64 too
<robimarko> CPU cores in all of those are Cortex A53
<neggles> yup, not a lot of point in putting aarch64 cores in your SoCs if you run armv7l code on them
<robimarko> But QCA didnt actually support running them in 32 bit
<robimarko> Ahh, 64 bit
<robimarko> Until 11.0 or 11.1
<neggles> prior to QSDK 11 yea
<robimarko> But even now you gotta build the 32 bit version first
<neggles> ha
<robimarko> Copy some stuff over and then rebuild
<stintel> so basically everything out there is running 32 bit armv7 ;)
<robimarko> Yeah
<robimarko> Cause it takes extra effort
<stintel> and then people complain new mediatek is still a53 ;p
<robimarko> I was dissapointed in QCA as well
<robimarko> Cause despite IPQ807x being couple of years old now
<robimarko> They continued just using A53 on newer SoC-s
<neggles> yeah
<robimarko> IPQ60xx is the same CPU with some peripherals a bit updated
<neggles> I mean with how powerful the NSS processors are, that's *almost* justified
<robimarko> And everything toned down
<neggles> except... y'know... the software/drivers...
<neggles> NSS ain't much good if it doesn't work right
<robimarko> Well, they made it work in QSDK
<robimarko> Thats all they care
<neggles> they do seem to be putting some amount of effort in now, at least
<robimarko> And 99.99% of their customers
Tapper1 has joined #openwrt-devel
<neggles> they're trying more than they used to be, which is not a high bar
<neggles> but
<neggles> i'm more interested in the fun mediatek SoC that's in the pipeline
Tapper has quit [Ping timeout: 480 seconds]
<robimarko> Well, my experience with them is that they will kind of start upstreaming
<neggles> Filogic 830 aka mt7986
<robimarko> And as long as nobody has a comment or god forbid a request to change their shitty code
<robimarko> They will do it, but if it takes effort then they just give up
<robimarko> That MTK looks nice
<robimarko> I mean, they had to do quad core
<neggles> yeah
<neggles> and integrated 4x4 dual band AX :D
<neggles> with 2x2.5GbE no less
<robimarko> It was time
<robimarko> Though they kind of skimped on 2.5G only
<neggles> eh, 2.5G is where everyone's going, you can get actual 2.5G-T SFP modules now which is nice
<neggles> and in reality you're not pushing >2.5Gbps through the AP
<robimarko> Well, it wouldnt have killed them to make it USXGMII
<robimarko> Its not just about AP-s
<neggles> true
<robimarko> I just hope that they wont do some crazy preloader
<neggles> looks like it's got the 1G hookup for a 7531A as well which is interesting
<neggles> but I guess that makes sense for a router with 2.5G WAN + 2.5G LAN + 4x 1G LAN
<neggles> oh and there's a couple of lanes of PCIe in there so you could always bolt an aquantia 10G on
<neggles> a very silly but neat card.
arjun has quit [Quit: Page closed]
<robimarko> Gotta say that I am not fan on 10G copper
<robimarko> Its just a power hog
<robimarko> And quite expensive
<neggles> we've gotten down to <2W/PHY
<neggles> which is... fine
<robimarko> Compared to like 2-3 years ago thats great
<robimarko> Though it cant compare to fiber yet
<neggles> and compared to the first 20-port Arista 10G-T switch, which was iirc six watts a PHY
<stintel> problem with fiber is that it's much less accessible for home/soho
<neggles> maybe more?
<neggles> and the usual suspects (FS.com and MikroTik) have 1/2.5/5/10G-T SFP+ modules for quite cheap
<robimarko> I would argue that fiber is cheap
<robimarko> I mean, switches with SFP-s are nothing unusual
<robimarko> FS modules are cheap
<neggles> stintel: sorta... cheap/used enterprise 10G still works out cheaper than 2.5G most of the time (mikrotik CRS305/CRS312/etc + mellanox CX-3 cards)
<neggles> the problem 2.5G has is there aren't any cheap switches
<stintel> cable installation is kind of difficult with fiber...
<neggles> 'cause enterprise never had 2.5/5, so the years of cost optimization/amortization haven't happened
<robimarko> Oh yeah, 2.5G is weird currently
<robimarko> I have lots of SFP+ gear
<neggles> stintel: true, but I can take the $200-300 8-10x SFP+ switch, drop some $50-70 10G-T SFP modules into it
<robimarko> But nothing switch based that has 2.5G
<neggles> and still come out cheaper than buying a netgear 8x2.5G
<neggles> well, not quite, but close
<neggles> with a much better feature set, and... most people don't put cable in the walls
<neggles> even if they do, you've probably got maybe 2 or 3 things you actually care about >1G on - a NAS and an AP and a desktop PC, and even then
<stintel> and the other problem with fiber is no PoE ;)
<neggles> this is true. and that's why it's so frustrating
<neggles> USW-Enterprise-8-PoE is about the only reasonably priced 2.5G/SFP+ uplink/PoE+ game in town
<neggles> and, well, unifi.
dedeckeh has joined #openwrt-devel
<neggles> (with a marvell switch SoC that has mainline support, at least - gotta get my hands on one of those and get hacking)
<neggles> there *are* old-ish enterprise 10G-T/multigig switches that are cheap, but, uhhh, I don't want to run a switch that pulls 350W at idle thanks
<stintel> :P
<neggles> i think the worst offender I saw, apart from the *seven hundred watt Arista* which was the first real 10G-T switch, so gets a free pass, was a Cisco that wants 350W with no ports up and 600W with all ports up
<neggles> so at the end of the day, you're left with dropping copper modules into 10G SFP+ switches and a mess of PoE injector, or spending $$$$ :(
<neggles> curse you and your huawei :P
<stintel> I don't regret my Huawei S6702 :)
<stintel> s6720*
<stintel> I do regret the AP7060DN as it's QCA based
<neggles> it's a very nice switch indeed, but I can't get one for under AU$3000 here
srslypascal is now known as Guest538
srslypascal has joined #openwrt-devel
<neggles> and $3000 buys a lot of slightly less nice hardware :P
<stintel> neggles: I paid EUR2100 ex VAT early 2020, for the switch + a 1150W AC power supply
<neggles> so yeah about AU$3500
<neggles> I cannot even begin to justify that
<stintel> and 540 for an extra 1150W psu
<neggles> when i can get a switch with 48x1G PoE, 8xSFP+, and 4x QSFP+ for $500
<stintel> oops, the extra PSU was 830, not 540
<neggles> O_O
<neggles> a third of the price of the entire switch?
<stintel> so yeah, 3k EUR ex VAT for the switch with 2 PSUs is not cheap
dangole has joined #openwrt-devel
<stintel> neggles: apparently :)
<neggles> that's just criminal
Guest538 has quit [Ping timeout: 480 seconds]
<stintel> if I take the price of the extra PSU and subtract it from the switch+psu the switch alone is just under 1300 EUR :P
<stintel> so by that calculation the PSU is ~65% of the switch
<neggles> I mean, for the amount of money that cost you, I could have a 48xSFP28 + 6xQSFP28 switch running SONiC + something with 8-16 2.5GbE and an appropriate amount of uplink, and have enough money left over to cover the power bill for a couple years
<neggles> https://www.edge-core.com/productsInfo.php?cls=1&cls2=9&cls3=46&id=919 <-- this is the dream switch at the moment though, shame they're AU$6-8k
<neggles> to be clear i'm not questioning your decision or saying it's bad value at all (though the extra PSU is just... oof)
<stintel> that was actually a miscommunication
<stintel> but I don't mind the added redundancy
<neggles> ah
<stintel> and you don't buy something like that to replace after 2y
<stintel> should easily last >5y
<stintel> in that case it costs < 1% yearly revenue
<neggles> if you can find one of the uplink port modules for it without doubling its cost, sure :P
<stintel> so it was not a difficult call
<neggles> but by the time you actually care about >10G interfaces the cards will be cheap, probably
<neggles> that really just illustrates my complaint, though - you could spend <25% as much to get something that is just as fast, if not faster, but won't do copper (or won't do copper without chewing power like you wouldn't believe)
<neggles> and there's nothing in between - even the edgecore has this problem, 36x2.5G + 12x10G? where's the 32x1G, 8x2.5G, 8x2.5/5/10G? doesn't exist
<neggles> sort of putting my money where my mouth is, though - broadcom and WNC are surprisingly pleasant to talk to... anyway
srslypascal has quit [Remote host closed the connection]
srslypascal has joined #openwrt-devel
Gaspare has joined #openwrt-devel
Gaspare has quit [Read error: Connection reset by peer]
Tapper1 has quit [Ping timeout: 480 seconds]
<stintel> anyone familiar with usteer who can explain this: https://gist.github.com/stintel/dca2b0679db6359f5f88a82a5be87ebb ?
<stintel> why does it create the station entry on all 3 2.4 GHz interfaces ?
GNUmoon has quit [Remote host closed the connection]
GNUmoon has joined #openwrt-devel
<aparcar> stintel: can't you block upnp per default for wan but allow it within the network? Having some easy network printer stuff going on sounds nice
<stintel> aparcar: it has some knobs ... in my case only my xbox is allowed to use it
<aparcar> can't we add something to LuCI like "static IP" and "allow Upnp"?
<stintel> I need more context ;)
dangole has quit [Remote host closed the connection]
dangole has joined #openwrt-devel
<aparcar> it's regarding your email
<aparcar> Re: Pre-install MiniUPnPd on OpenWrt by default
<aparcar> I wonder if we can have it by default
<rsalvaterra> aparcar: Dear God, no.
rua has quit [Ping timeout: 480 seconds]
<PaulFertser> mdns working by default is probably OK, but UPnP-PMP certainly not.
<rsalvaterra> UPnP is, quite literally, a trojan horse.
<f00b4r0> can the bridge-vlan syntax work on non-DSA devices? I.e. could I use it to assign tagged eth0 VLANs to specific bridges?
<f00b4r0> rsalvaterra: preinstall != enable though ;)
<rsalvaterra> f00b4r0: Right, let's increase de default image size for nothing. :P
<f00b4r0> I wouldn't say "nothing" but you have a point :)
<jow> f00b4r0: yes
<jow> f00b4r0: its DSA-agnostic and will work on any system
rua has joined #openwrt-devel
<f00b4r0> jow: ah thanks. I'm trying to anticipate DSA on existing devices
<f00b4r0> trying to wrap my head around vlan bridges
<f00b4r0> jow: I assume it needs the "bridge" package on the system though?
<PaulFertser> Is it ok to show OpenWrt users this article as an intro to bridge VLAN concept: https://docs.nvidia.com/networking-ethernet-software/cumulus-linux-50/Layer-2/Ethernet-Bridging-VLANs/VLAN-aware-Bridge-Mode/ ?
<rsalvaterra> It's already bad enough having ISPs that provide UPnP-enabled CPEs… without having an option to *disable* it.
<rsalvaterra> I know because I'm one of the victims. :P
<PaulFertser> f00b4r0: as with all the other protocols supported by netifd internally it should be making its own netlink calls, without a dependency on cli tools.
<f00b4r0> PaulFertser: ok. But then the resulting configuration is not "visible" in cli?
<PaulFertser> f00b4r0: I'm afraid yes. "bridge" is preinstalled on realtek target but not on the others, I do not know why.
<PaulFertser> It's partially visible in brctl I guess, but not fully properly.
<PaulFertser> Please correct me if I'm wrong.
<f00b4r0> PaulFertser: I've read the article you listed. I find it fairly confusing and unhelpful in an OpenWRT context. Especially since it starts with "You cannot use the same port with multiple VLAN-aware bridges"
<stintel> OpenWrt* ;)
<PaulFertser> f00b4r0: is it not the case for OpenWrt too?
<PaulFertser> f00b4r0: I'm not sure how one would reason about such an arrangement.
<stintel> aparcar: "your email"?
<stintel> "Pre-install MiniUPnPd on OpenWrt by default" - I don't recall writing such email
<stintel> am I losing my mind?
<f00b4r0> I sure hope not. If I can't assign a port to multiple different bridges (as a tagged member), the feature seems rather useless. Or I'm missing something
<aparcar> stintel: I thought you responded to it
<stintel> aparcar: I have no idea what you are talking about
<aparcar> stintel: sorry my bad, I only read the first name
<aparcar> which was Stijn too
<stintel> Borromini I guess
<f00b4r0> PaulFertser: if I read the (very limited documentation) correctly, the "bridge-vlan" item is one per VID, so that's rather orthogonal to the article you pointed at
<PaulFertser> f00b4r0: with bridge vlan type of config you're assumed to have a single bridge for that case, and you can have different ports in it have same or different vlan ID.
<rsalvaterra> stintel: Yes, Borromini replied, not you. :)
<stintel> phew, a bit early for dementia to kick in :P
Tapper has joined #openwrt-devel
<f00b4r0> PaulFertser: ok, I need to try the thing because right now I really can't make sense of how it works
<aparcar> stintel: sorry for the inconvenience
<stintel> aparcar: no worries
<PaulFertser> f00b4r0: I might be wrong too, hence asking for opinions.
<f00b4r0> I don't have any DSA-enabled devices to play with so that's not helping, hence my earlier question
<PaulFertser> f00b4r0: in any case we need to have some references for people familiar with enterprise networking solutions.
<stintel> I would not be against including miniupnpd by default, but it would have to: find a maintainer, move to openwrt.git, changed to block all by default, and leave it up to the user to change it to allow all by default, clearly warning them that is a bad idea and insist on doing per-IP whitelist
* f00b4r0 runs miniupnpd enabled by default :)
valku has joined #openwrt-devel
<f00b4r0> I like a dangerous life
<stintel> which basically brings it to the same "pain" to configure as setting manual port forwards
<PaulFertser> f00b4r0: "config device" would correspond to a "VLAN-aware bridge", "config bridge-vlan" to a configuration for a specific VLAN for a specific bridge.
<stintel> and I've considered taking maintainership of miniupnpd until I went through the existing issues for it in the packages feed
<f00b4r0> PaulFertser: ok. I'm trying to understand how this plays with 8021q type devices. So far I never use implied vlans, I always define a 8021q device and assign it to a bridge as needed
<f00b4r0> I'm under the impression that bridge-vlan will create the related 8021q br.* interfaces
pmelange has joined #openwrt-devel
pmelange has left #openwrt-devel [#openwrt-devel]
<stintel> with bridge-vlan you'll only need one bridge interface
<f00b4r0> stintel: but what shows up in e.g. ifconfig? separate brN.* ? separate lanN.* ?
<PaulFertser> That cumulus article shows that one can create several "bridge domains" from a single switch, and then each bridge domain is configured individually as a "vlan-aware bridge".
<f00b4r0> stintel: thanks but that doesn't answer my question
<stintel> Forst: refresh
<stintel> f00b4r0: ^
<PaulFertser> stintel: no tagged ports?
<stintel> PaulFertser: many tagged ports
<PaulFertser> Ah, silly me
<f00b4r0> ok so it does create multiple brN.* devices
<stintel> f00b4r0: no, it doesn't
<f00b4r0> switch.30@switch:
<stintel> f00b4r0: check the tagged vlans configured on switch
<stintel> f00b4r0: there are more vlans than switch.VID interfaces
<stintel> these switch.VID interfaces are created by other sections in /etc/config/network
<stintel> bridge-vlan itself does not create them
<f00b4r0> darn. This is giving me a massive headache ;P
<jow> bridge-vlan does not create devices
<PaulFertser> So unless you create a switch.VID interface the bridge will just do the bridging for that VLAN but you won't be able to access that traffic from CPU?
<f00b4r0> also I see eth0 has an MTU of 1504? I suppose that's because VLAN tag?
<jow> PaulFertser: yes
<stintel> f00b4r0: no, dsa tag
<f00b4r0> hmm ok. So I guess you should never use eth0 directly in this context?
<stintel> you shouldn't
<f00b4r0> and that's supposed to be an improvement over the previous situation :-3
<stintel> it is
<stintel> explicit > implicit
<stintel> you can refresh the gist again, I've added /etc/config/network
<f00b4r0> thanks
Gaspare has joined #openwrt-devel
<stintel> yw
<stintel> took me a few devices to get used to it
<PaulFertser> stintel: what would you suggest as an introduction to the concept?
<stintel> PaulFertser: try it until it works :P
<stintel> but I think there's a forum post which handles most use-cases
<f00b4r0> in your current config, if you add config device / option type "8021q" / option ifname "switch.99" / option vid "99" / option name "vlan-guest" and then use 'option device "vlan-guest"' in config interface 'guest', would it work?
<PaulFertser> stintel: those config bridge-vlan sections look pretty similar to what you'd do with swconfig.
<stintel> f00b4r0: probably
<f00b4r0> The point is to name the vlan device
<jow> f00b4r0: yes it will work
<f00b4r0> thanks
<jow> f00b4r0: bridge-vlan just programs the filters on the bridge as switch-vlan would do on a hardware switch IC
<PaulFertser> So can you define several config device sections for the same switch IC?
<stintel> not sure if that will workin with dynamic VLAN though
<f00b4r0> ok so I'll try playing with bridge-vlan over a swconfig system. I assume that if I correctly setup switch_vlan sections and use 'eth0:t' in bridge-vlan sections, it should Just Work™?
<jow> "grabbing" the vlan tagged traffic off the bridge (or the cpu port in the swconfig case) is done by declaring vlan interfaces having the appropriate tag
<stintel> f00b4r0: I don't think this bridge-vlan stuff works with swconfig systems
<stintel> but I could be wrong
<f00b4r0> jow disagreed a few lines up :)
<stintel> be prepared to recover the device ;)
<stintel> so ideally experiment with something that has serial console
<f00b4r0> the reset button is close at hand :)
<f00b4r0> well I assume that failsafe will get me back from network config snafus :)
valku has quit [Quit: valku]
valku has joined #openwrt-devel
valku has quit []
valku has joined #openwrt-devel
<f00b4r0> stintel: how do you assign wlans to your VLAN bridge? I don't see that bit in the config
<jow> f00b4r0: /e/c/wireless, option network
<f00b4r0> ah right
<f00b4r0> silly me
<mrkiko> So, I am using a GL-MV1000 with eMMC and it gets formated with ext4. Someone may explain the how the choice between ext4 and f2fs works? Will OpenWRt prefer what of them? And your recommendations?
<f00b4r0> hmm "bridge" isn't the package that provides "bridge" :P
<jow> ip-bridge afair
<f00b4r0> thx
<jow> PaulFertser: in the sense that you can have multiple 8021q device declarations refering to the same cpu port netdev (e.g. eth0), yes
<jow> e.g. config device; option type 8021q; option vid 1; option ifname eth0; option name vlan1 ;; config device; option type 8021q; option vid 2; option ifname eth0; option name vlan2 etc.
minimal has joined #openwrt-devel
dwmw2_gone has joined #openwrt-devel
<f00b4r0> hmm. First attempt at using bridge-vlan is fail
<PaulFertser> jow: no, I meant e.g. having one config device of type bridge with ports lan0, lan1 and another with ports lan2, lan3, lan4.
<PaulFertser> f00b4r0: I think jow meant you can use the syntax to configure linux software bridges but not to control ports of a "swconfig bridge".
* stintel curses QCA and QSDK
<f00b4r0> PEBKAC
<f00b4r0> it does work
<PaulFertser> f00b4r0: to control bridge ports?
<f00b4r0> PaulFertser: how to configure a swconfig device "a la" bridge-vlan
<f00b4r0> using a single VLAN-aware bridge
<f00b4r0> interesting bridge -s gives all counters at 0
<f00b4r0> now the question I guess is do I like having a single bridge with every interface in it which means it's not immediately obvious which interface is actually bridge with which, or whether I prefer old style one bridge per vlan. Assuming an old-style config would also be applicable to DSA
<f00b4r0> oh. This does mean a single 8021q device regardless of how many physical interfaces should belong to that VLAN
<f00b4r0> ok I'm sold :)
<stintel> that didn't take so long ;)
<jow> f00b4r0: so far I always opted for the single bridge approach (I usually call mine "switch0")
<jow> PaulFertser: that will work too (having multiple bridges, each containing a distinct subset of the dsa ports)
<PaulFertser> jow: so same as in that Cumulus Linux article it seems.
<PaulFertser> (assuming you meant disjoint sets)
<f00b4r0> stintel: I saw the light ;^)
<stintel> > Debian 10’s default compiler is GCC 8. The GCC 8 documentation wasn’t packaged in time for Debian 10’s release, but it is available in backports
<stintel> le wtf
<f00b4r0> jow: yes I can see how this makes sense indeed. Besides it's the Future™ and I assume it'll make the setup compatible with Luci which is a desirable side effect.
<f00b4r0> stintel: classic debian move ;P
Tapper has quit [Ping timeout: 480 seconds]
<dangole> mrkiko: choice of the filesystem used for rootfs_data depends on the total size of the volume/partition hosting that filesystem. if it's too small for F2FS, ext4 is used.
<dangole> mrkiko: https://git.openwrt.org/?p=project/fstools.git;a=blob;f=libfstools/common.c;h=1b22d2cb19aefe1b928fad6045210b2656c1c16d;hb=HEAD#l94
<mrkiko> dangole: thanks!!
<mrkiko> dangole: 92.7 MB rootfs_data, 7.28 GB eMMC
Tapper has joined #openwrt-devel
<dangole> mrkiko: from 100 MB on it would use F2FS... i remember correctly that the partition layout comes from the vendor firmware?
<f00b4r0> is there a measurable performance downside to assigning a NIC to a vlan-aware bridge and using vlans to talk through that NIC instead of using the NIC directly?
<stintel> I would suggest you measure that ;)
<f00b4r0> heh :)
Borromini has joined #openwrt-devel
cmonroe_ has joined #openwrt-devel
* f00b4r0 notes that devstatus does not report tagged/untagged status for bridge-vlans
<mangix> stintel: why c++14 for Gcc < 11?
<stintel> mangix: because that's what the manpage says is default
<mangix> I ask as C++17 is supported by GCC8 even
<stintel> the point of the change is to build a GCC version with its default dialect so that when GCC 14 deprecates something used in GCC 8 or 11, it will still build the toolchain
<mangix> Hmm OK.
<mangix> prereq-build lists GCC6, which does not support c++17 for the std parameter. It supports c++1z
<mangix> Well, IIRC
<stintel> good point
<stintel> I'll just forget about it, the downstream response was "build on ubuntu 20.04"
<mangix> true I don't build on old stuff either
<mangix> Pin any case, it might make sense to bump prereq-build
<stintel> feel free to comment on the patch ;)
<stintel> I'm calling it a day
* f00b4r0 notices 'bridge' talks json.
<f00b4r0> I guess that'll have to do to complement devstatus
dangole_ has joined #openwrt-devel
swegener has joined #openwrt-devel
DLange has joined #openwrt-devel
dangole has quit [Ping timeout: 480 seconds]
owrt-1907-builds has quit [Ping timeout: 480 seconds]
nbd has joined #openwrt-devel
owrt-1907-builds has joined #openwrt-devel
<mrkiko> dangole_: don't know, didn't look much in the mv1000
<Slimey> heh big boi Found Spansion flash chip "S25FL512S" (65536 kB, SPI) on linux_spi.
Gaspare has quit [Quit: Gaspare]
Gaspare has joined #openwrt-devel
<dwmw2_gone> root@garden:~# sysupgrade /tmp/openwrt-ramips-mt7621-asus_rt-ac85p-squashfs-sysupgrade.bin
<dwmw2_gone> Tue Jan 25 20:09:23 GMT 2022 upgrade: The device is supported, but the config is incompatible to the new image (1.0->1.1). Please upgrade without keeping config (sysupgrade -n).
<dwmw2_gone> hm, but I upgraded it to OpenWrt master just *days* ago and it said that last time too (and I forced it like it said)
<dwmw2_gone> root@garden:/etc/config# uci -q set "system.@system[-1].compat_version=1.1"
<dwmw2_gone> better
robimarko has quit [Quit: Page closed]
ashkan has joined #openwrt-devel
minimal has quit []
shibboleth has joined #openwrt-devel
Tapper has quit [Read error: Connection reset by peer]
Tapper has joined #openwrt-devel
Borromini has left #openwrt-devel [#openwrt-devel]
dedeckeh has quit [Remote host closed the connection]
<hurricos> dwmw2_gone: OpenWrt master needs to have a base-files/etc/uci-defaults/05_fix-compat-version for your board.
<hurricos> Sounds like it deosn't
<hurricos> In your situation, that sounds intentional. It's just warning you that config is lost, it seems
<hurricos> seems like this commit is coming back for a haunting? https://github.com/openwrt/openwrt/commit/38f6d5d217ca0c42f7f42b08f835a8a9cee71ad7
<dwmw2_gone> hurricos: the config wasn't lost. I did a '-F' and it seemed to preserve everything anyway.
Gaspare has quit [Quit: Gaspare]
Slimey_ has joined #openwrt-devel
<hurricos> It's referring to DSA configuration
<dwmw2_gone> Right.
Slimey has quit [Ping timeout: 480 seconds]
<dwmw2_gone> I hadn't done anything special here and the config just worked anyway.
<hurricos> Looks like https://github.com/openwrt/openwrt/commit/38f6d5d217ca0c42f7f42b08f835a8a9cee71ad7 is incomplete. Perhaps intentionally. I don't know the DSA status of MT7621.
Slimey_ is now known as Slimey
<dwmw2_gone> It was the *next* upgrade which still complained as nothing had set the compat_version even after the upgrade.
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<hurricos> The conversation in that commit asserts the existence of an 05_fix-compat-version https://github.com/openwrt/openwrt/tree/master/package/base-files/files/etc/uci-defaults
<hurricos> The role you did manually ought to be done by an 05_fix-compat-version somewhere (but that hasn't been compiled in for you),.
goliath has quit [Quit: SIGSEGV]
shibboleth has quit [Quit: shibboleth]
dangole_ has quit [Remote host closed the connection]
danitool has joined #openwrt-devel
goliath has joined #openwrt-devel