<robimarko>
I assume that load adress is too small, this happened to me previously
<robimarko>
Or it clashed with something else
<robimarko>
And then you are stuck with nothing after U-boot starts the kernel
<mangix>
yeah I tried multiple load addresses, no dice.
<mangix>
alright non fit boots
<mangix>
I think I'm done with this
<mangix>
I got it working...
<mangix>
now to figure out why...
AtomiclyCursed has quit [Quit: ZNC 1.8.2 - https://znc.in]
AtomiclyCursed has joined #openwrt-devel
gladiac has joined #openwrt-devel
SlimeyX_ has joined #openwrt-devel
winternull_ has joined #openwrt-devel
Slimey_ has joined #openwrt-devel
Vaughn_ has joined #openwrt-devel
_0x4a6f_ has joined #openwrt-devel
aparcar_ has joined #openwrt-devel
schwicht_ has joined #openwrt-devel
dwfreed_ has joined #openwrt-devel
jbowen_ has joined #openwrt-devel
mangix_ has joined #openwrt-devel
hgl_ has joined #openwrt-devel
winternull has quit [synthon.oftc.net larich.oftc.net]
SlimeyX has quit [synthon.oftc.net larich.oftc.net]
schwicht has quit [synthon.oftc.net larich.oftc.net]
matoro has quit [synthon.oftc.net larich.oftc.net]
mangix has quit [synthon.oftc.net larich.oftc.net]
Slimey has quit [synthon.oftc.net larich.oftc.net]
swalker has quit [synthon.oftc.net larich.oftc.net]
Vaughn_ is now known as Vaughn
SlimeyX_ is now known as SlimeyX
Slimey_ is now known as Slimey
hurricos has joined #openwrt-devel
mangix_ has quit []
mangix has joined #openwrt-devel
eigma has joined #openwrt-devel
aparcar_ is now known as aparcar
_0x4a6f_ is now known as _0x4a6f
matoro has joined #openwrt-devel
SlimeyX_ has joined #openwrt-devel
robimarko_ has joined #openwrt-devel
SlimeyX has quit [Ping timeout: 480 seconds]
SlimeyX_ is now known as SlimeyX
robimarko has quit [Ping timeout: 480 seconds]
srslypascal is now known as Guest1036
srslypascal has joined #openwrt-devel
Guest1036 has quit [Ping timeout: 480 seconds]
<mangix>
alright I sort of figured out why it's not working
<mangix>
when Data Size: 15670586 Bytes = 14.9 MiB , it does not boot
<mangix>
when Data Size: 6595575 Bytes = 6.3 MiB, it boots
<mangix>
I'm guessing 8MB limitation
<robimarko_>
Its gotta be a memory mapping issue then
<mangix>
keep in mind this is initramfs. I'm not flashing images yet.
<robimarko_>
I understand, but it really doesnt matter to U-boot
<mangix>
hmm? how so?
<robimarko_>
There is no difference if you are starting a kernel or kernel that has initramfs included
<robimarko_>
Unless you are passing the initramfs as a separate image in FIT
srslypascal has quit [Ping timeout: 480 seconds]
srslypascal has joined #openwrt-devel
StifflersMagic has joined #openwrt-devel
schwicht_ has quit [Read error: Connection reset by peer]
schwicht has joined #openwrt-devel
schwicht_ has joined #openwrt-devel
schwicht has quit [Ping timeout: 480 seconds]
hitech95 has joined #openwrt-devel
<hitech95>
Can I say that I'm dumb enough to don't understand if the my7621 have 3 or 2 MAC imterfaces in the CPU side?
<f00b4r0>
2
<hitech95>
F00b4r0, yeah but one of that is the rgmii2?
<f00b4r0>
yes
<hitech95>
So how they manage to get wan+4lan without using an external phy?
<hitech95>
They use vlan even on the official BSP?
<f00b4r0>
the internal switch can mux one of the phy to the second cpu port
<f00b4r0>
#10238 is precisely taking advantage of that
<hitech95>
Oh ok, so you loose rgmii2 but you gain the fifth port
<f00b4r0>
have to run sorry; bbl
<hitech95>
Np
<hitech95>
So if I want two ETH ports:
<hitech95>
or I have an independent eth port and i have to run a Linux bridge if I want all part of the LAN or I have to use vlans and use 2 of the first
<hitech95>
And I suppose that the mux works only for port 4 of the switch, and that port cannot be part of the switch itself by disabling one of the others ports....
<hitech95>
2 ports of the switch.
minimal has joined #openwrt-devel
<mangix>
hitech95: port0 and 4 can be mixed.
<hitech95>
So, port 0 can be both gmac0 or 1? And port4 can also be gmac0 or 1?
<mangix>
the gb-pc1 has two ports. That PR also made it work.
<mangix>
Yes
<hitech95>
I'm reading the programming guide for the gbit switch but I cannot find this information
<mangix>
No idea about that.
<hitech95>
So anyway the best solution would be to use port 0 and port 4. So in the worst case you always have 2 links to the CPU :D
<f00b4r0>
hitech95: you have the programming guide for the switch? nice :)
<hitech95>
Yea, it's available online ... I'm trying to make my Own LTE CPE based on a mt7621 SOM.
<jow>
ynezz: note that we define namelen as strlen() and we allocate namelen + 1 (strlen() + 1) bytes of storage, then copy up to strlen() bytes in a strlen() + 1 buffer
<jow>
so the gcc message seems to be a false positive
<ynezz>
jow: ok, I'll take a deeper look
<ynezz>
maybe we should stop using gcc? :P
<jow>
I guess gcc really want to be sure that the destination is strncpy() length + 1
<jow>
it does not understand implicit \0
<ynezz>
but that implicit \0 is from the calloca
<jow>
yes
<jow>
so one can't effectively use strncpy(..., s, strlen(s))
<jow>
it'll always yield a warning
<jow>
even if care is taken to arrange for a trailing \0 by sizing the destination appropriately
<ynezz>
I probably need to lookup first why it's visible only with -O2 flag, why its not reported with -Os
<ynezz>
does that size optimization optimize out this checks as well? :p
<jow>
I guess an easy fix would be using memcpy instead
<jow>
same semantics, no string specific checking
<stintel>
f00b4r0: copy-pasting what I wrote in Matri
<stintel>
Not really using the unmatched as it's too slow
<stintel>
At some point I might throw in an extra NIC and play with OpenWrt and test routing performance
<stintel>
But I have way too many things to play with already 😂
<stintel>
Meson, m200, bsap3040 to name a few
<stintel>
Recently distracted with woodpecker CI too
<ynezz>
jow: if that's your preference, fine with me
<jow>
ynezz: could you test it? simply replace the original strncpy() with memcpy()
<jow>
without adding - 1
<jow>
ah wait, gcc 10 with -O2
<jow>
thought it requires some fancy bleeding edge variant
<ynezz>
yeah, thats the point, check the referenced issue
<f00b4r0>
out of curiosity why are you using namelen - 1 when you already allocate +1 ?
<ynezz>
jow: BTW that latest commit of yours in rpcd doesn't compile "rpcd-2022-08-24-82904bd4/ucode.c:932:36: error: 'uc_parse_config_t' has no member named 'module_search_path'"
<jow>
f00b4r0: to appease gcc
<jow>
it unconditioanlly warns on strncpy(..., s, strlen(s))
<jow>
because one looses the \0 of s in this case
<jow>
which might or might not be desired
<jow>
in our case it is desired/expected, but gcc is too stupid to know
<jow>
hence ynezz' attempt to silence it with -1
<f00b4r0>
ah. I assume from what you wrote above that it's new behaviour? I have some code using this type of construct and no warning in gcc8, so I guess I should brace for impact then ;)
<f00b4r0>
stintel: got it thanks :)
gladiac has quit [Quit: k thx bye]
jbowen_ has quit []
<ynezz>
f00b4r0: are you using -O2/-O3 flag?
<jow>
ynezz: s/strncpy/memcpy/ fixes it too
<ynezz>
f00b4r0: it's not visible with -Os, so in order to reproduce it you need `-Wall -Werror -O2`
jbowen has joined #openwrt-devel
<ynezz>
jow: thanks, are you willing to fix with that change yourself or should I update the PR?
<jow>
ynezz: just pushed a fix
<ynezz>
jow: thanks!
<hitech95>
I'm reading the mt7630.txt the point about port 5 modes/configurations. And i cannot figure out point 2.
<hitech95>
Mac 0-3 and phy0-3 are the switch one.
<hitech95>
Gmac0 is connected to mac6, and mac5 is connected to gmac1.
<hitech95>
Port5 i suppose is the port4 of the datasheet.
<hitech95>
This is the part im not understanding: "2. Port 5 is muxed to PHY of port 0/4: Port 0/4 interfaces with 2nd GMAC."
goliath has quit [Quit: SIGSEGV]
Gaspare has joined #openwrt-devel
Gaspare has quit []
Gaspare has joined #openwrt-devel
guidosarducci has quit []
guidosarducci has joined #openwrt-devel
<jow>
hmm, do we have an example of a cmake package with host install?
<tmn505>
fwtool
gladiac has joined #openwrt-devel
<jow>
thanks
<hitech95>
fb0b40r0, i probably got how it works. H
<domon>
But not sf? sf command event no enabled in uboot :/
borek has joined #openwrt-devel
Misanthropos has quit [Ping timeout: 480 seconds]
borek1 has joined #openwrt-devel
borek has quit [Ping timeout: 480 seconds]
borek1 is now known as borek
winternull_ has quit [Ping timeout: 480 seconds]
jlsalvador has quit [Quit: jlsalvador]
<f00b4r0>
mrnuke: see, more PCB real estate used :^)
SlimeyX has quit [Read error: Connection reset by peer]
<nbd>
f00b4r0: hi
<f00b4r0>
nbd: hi
<nbd>
f00b4r0: i've made some good progress on unetd
Misanthropos has joined #openwrt-devel
<nbd>
it's in my staging tree now
<nbd>
you were interested in it, right?
<f00b4r0>
sure!
<f00b4r0>
is it ready for testing now?
<nbd>
yes. it can now synchronize signed network json data across nodes
<nbd>
and there's an admin script written in ucode
<nbd>
which makes setup really simple
<f00b4r0>
sounds good, I'll give it a spin later this week then
<nbd>
here's a simple example script that i made which sets up two openwrt routers with unetd to connect to each other: https://termbin.com/nlxe
<nbd>
it also sets up a vxlan tunnel between them
<f00b4r0>
got it. It sure looks simple
SlimeyX has joined #openwrt-devel
<nbd>
it uses ssh to connect to the routers and prepares the uci network interface
<nbd>
generates fresh keys
<nbd>
and puts the public keys in the json
<nbd>
auth key is also autogenerated if it's not present
<f00b4r0>
ok. I assume it needs ssh-key set on target routers beforehand, or will it prompt for passwd otherwise?
<nbd>
it will prompt for passwd if there is no key
<f00b4r0>
excellent
<nbd>
it runs a normal ssh command
<nbd>
and i also added a mode where you can configure a gateway for a host
<nbd>
this ensure that the host will only set up a wg connection to its gateway
<nbd>
and all others will route to it using the gateway
<f00b4r0>
that's really cool. It's gonna make my use case a heck of a lot simpler ;-)
<nbd>
you just need to add a firewall entry to allow routing from unet to unet
<nbd>
on the gateway host
<f00b4r0>
ok
<nbd>
if you have a node with a public ip, you can leave out the endpoint address for all other hosts
<f00b4r0>
ok. All but one node have public IPs in my case
winternull_ has joined #openwrt-devel
<nbd>
and you can even run unet-cli on a machine that does not run unetd
<nbd>
it can send data to unetd instances anyway
<f00b4r0>
i see you've been really thorough about this :)
<f00b4r0>
i'll definitely take it for a test drive later (got a couple more busy days before I can start playing again)
<nbd>
cool, let me know how it goes
<f00b4r0>
sure thing!
<f00b4r0>
btw since you're here, pardon my attention hijack but you wouldn't happen to have any idea about the mt7621 dsa issues (bridge-vlan roaming being broken and DSCP markings issue) that are plaguing current 22.03? I hit a deadend poking around tbh.
<nbd>
only took a superficial look so far, didn't have time to go into detail yet
<nbd>
didn't spot anything so far
<nbd>
i hope that i will have more time for it next week
<nbd>
please keep reminding me if i forget
<nbd>
too much wildly different stuff on my plate at the moment
<f00b4r0>
ok. Let me know if I can provide more details. Last check was me digging into the ethernet driver, with no luck. Really frustrating that everything seems to work correctly in 5.15 but no amount of backport solves it in 5.10
<f00b4r0>
sure; will do!
<nbd>
btw. regarding the sign/upload step at the end of my unetd script:
<nbd>
for the node that doesn't have a public ip, you can put in connect=<ip of one node> in the add-ssh-host step
<nbd>
and then you only need to upload the data to one of your nodes
<nbd>
even at the bootstrap stage
<nbd>
it will contact the others and spread it around
<nbd>
and once the network is up, updates will be distributed over the wireguard link
<f00b4r0>
that's brilliant
<nbd>
i also made sure that the data transfer outside of the link (for bootstrap) is encrypted and only accessible to members of the network
minimal has quit [Quit: Leaving]
<nbd>
it only contains public keys, but i did it for defense in depth and all that...
<ynezz>
looks almost like self-hosted tailscale
<ynezz>
cool
<nbd>
i did look at tailscale as well
<nbd>
i like their approach as well
seer has quit [Quit: quit]
<nbd>
it's just for a different use case
seer has joined #openwrt-devel
winternull_ has quit [Quit: Leaving]
Gaspare has quit [Quit: Gaspare]
<mangix>
f00b4r0: that's on mt7622 only, right?
Gaspare has joined #openwrt-devel
Borromini has joined #openwrt-devel
borek has quit [Ping timeout: 480 seconds]
<f00b4r0>
Mangix: what is?
cbeznea has quit [Quit: Leaving.]
cbeznea has joined #openwrt-devel
tlj has joined #openwrt-devel
<mangix>
f00b4r0: mt7530 issues
<mangix>
actually...doesn't mt7622 use mt7531?
cbeznea has quit [Quit: Leaving.]
Borromini has quit [Read error: Connection reset by peer]
Borromini has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
f00b4r0 has quit [Ping timeout: 480 seconds]
schwicht has joined #openwrt-devel
schwicht_ has quit [Read error: Connection reset by peer]
f00b4r0 has joined #openwrt-devel
<f00b4r0>
Mangix: the mt7530 issues are mt7621
<mangix>
what are they?
<mangix>
IIRC you were mentioning issues with backporting stuff to 5.10
<f00b4r0>
Mangix: my connexion is flimsy right now (on *3G* backup :P), see the mailing list. There's a roaming issue when there's a bridge-vlan involved (the stale fdb entry on the ethernet port blocks roaming to cpu port - aka wireless), and there's also a dscp issue with dropbear (there's a GH issue)
<mangix>
f00b4r0: DSCP issue is only on mt7622
<f00b4r0>
the roaming problem is fixed in 5.15, however backporting the "fix" triggers a different throughput problem where all switch ports throughput max at the slowest speed (while still reporting higher speed)
<f00b4r0>
ok
<f00b4r0>
i haven't experienced the dscp issue so I'm not sure
<mangix>
I have. It happens on my RT3200 but not my RT1800
<f00b4r0>
ok
<mangix>
mt7622 vs mt7621
<hauke>
Mangix: does the dscp issue happen after some time or after some data was transfered?
<mangix>
hauke: nope. Immediate.
<mangix>
example: I ssh into root@openwrt. After that I cannot input anything
<hauke>
for me ssh to my RT3200 over wifi still works
<mangix>
thiis is over WiFi. Over ethernet works
<mangix>
I think the issue has to do with Wireless Ethernet Dispatch
<hauke>
I am also going over wifi
<mangix>
I don't think that's supported by mt761`
<hauke>
on openwrt 22.03
<mangix>
*mt7621
<jow>
Mangix: it has been reported that the issue also happens when mt76 wifi is involved somewhere along the network path
<jow>
Mangix: so not sure if it's wireless ethernet dispatch related
<f00b4r0>
i have to bail. Hopefully regular internets will work again tomorrow. ttyl
<mangix>
So I'm currently running a WAX202 as a router. No issues. If I switch back to my RT3200, the issue will show up.