<will[m]>
tmn505: what if i didn't want /my/path/sw_watcher/src but instead /my/path/other_watcher_src ? could I do like MAKE_PATH:=../other_watcher_src ?
<tmn505>
that should work.
<will[m]>
now if only USE_SOURCE_DIR behaved like that
<digitalcircuit>
Woo, I've recreated a crash when CPU governor is on "performance" with only using "yes >/dev/null", no Deja Dup complicated test case needed :D
<digitalcircuit>
I'll update my ipq806x test repo with details soon.
titanous has quit []
DragoonAethis has quit [Quit: hej-hej!]
DragoonAethis has joined #openwrt-devel
<will[m]>
is there any reason under the sun why a buildroot would output only packages, no firmware image, when running make / make world ?
<will[m]>
oh i see, somehow all the packages disappeared so i had to reupdate/install/make-download them
<digitalcircuit>
This triggered the crash on a month-old Linux 5.4 snapshot build; I'll retest on a snapshot from today using Linux 5.10 once it finishes building.
<digitalcircuit>
...at least, hopefully. I don't know if anyone has actually tried this script.
valku has quit [Quit: valku]
victhor has quit [Ping timeout: 480 seconds]
<digitalcircuit>
Welp, router on today's snapshot with 5.10 fails to mount the root partition. Time to figure out how to set the bootloader to boot the alternative partition
wulfy23 has quit [Remote host closed the connection]
<digitalcircuit>
If anyone else runs into this issue, follow the "ATSE" password unlock instructions from https://git.openwrt.org/project/ubus.git;git://git.openwrt.org/project/ubox.git?p=openwrt/openwrt.git;a=commit;h=459c8c9ef816156107e297964d088ddee2b4eef5 , then in the full u-boot environment, all you need to do is... "run boot_mmc_1" (instead of "run boot_mmc")
<digitalcircuit>
That was a pain. I'll try to figure out why current snapshots fail to boot on the NBG6817 when the router isn't in the bedroom of someone sleeping.
<digitalcircuit>
(To be clear, my current custom snapshot. I probably need to redo the configuration.)
<digitalcircuit>
("make menuconfig" from scratch instead of using the kernel 5.4 config I've kept around.)
<digitalcircuit>
I don't think I have a wiki account, but I'd imagine this might be useful for others in the future, too.
* digitalcircuit
notes debugging is always more fun when the router's located in the room of a family member that's sleeping. He'll return to this in the day time.
<slh>
digitalcircuit: keep in mind that we've had major toolchain updates in master over the last ~2 weeks, gcc was upgraded from 8 to 10, now to 11, musl from 1.1.x to 1.2.x. you do need to clean up your buildtree for that (tmp, build_dir, staging)
<digitalcircuit>
slh: Noted! I've been running "make distclean" before most builds (especially rebases, major branch switches) as I've not wanted to risk cross-build contamination. (I've been meaning to learn what exactly is needed at some point.)
<slh>
digitalcircuit: out of curiosity, have you ever found a way to toggle between primary and alternative installation without software access? by pressing some button or similar, something that doesn't require a fully booted system or serial console access?
<slh>
akin to Linksys' abort early boot three times to toggle the bootpartition?
<digitalcircuit>
I suspect my issue is trying to re-use a menuconfig based on OpenWRT 21.02 with the 5.4 kernel. It worked fine when building with kernel 5.4 on master branch, but I think since the switch to 5.10 as default, something went wrong. I'll probably just need to re-do menuconfig from scratch.
<digitalcircuit>
slh: I've not figured out any method yet. I noticed it happened once unexpectedly when I was flashing different versions, but I haven't yet looked into it. It's a good idea to try to figure out though.
<slh>
digitalcircuit: ./scripts/diffconfig.sh is a good start of getting your build config down to the parts that really matter (yes, there generally is more to drop, but it's a good start), you can then re-expand (the seed-) .config (-that you ended up with via diffconfig.sh) via make defconfig oldconfig
<slh>
digitalcircuit: I've tried to find a way, but haven't been successful yet (but I also don't have a serial console connected - and use the device as my main router/ in production, so I could never dig that much either)
<digitalcircuit>
slh: I *think* I've used diffconfig.sh, but I never tried trimming it down below that - I didn't know it was feasible. Noted on how to re-expand it, too, thank you!
<digitalcircuit>
slh: As to boot, noted. The router is my main router in production as well, so it's not exactly convenient, but I do have time and a serial console, so I'm happy to try to explore this in the future (probably after I run out of things for the CPU crash/reset tests). I had naively assumed that because the wiki didn't mention a button-press method, that it flat-out wasn't possible. Knowing it might exist just not explored is useful!
<digitalcircuit>
With the serial console and ATSE unlocking full uboot, knowing that there's two easy commands for booting 1st or 2nd partition makes it seem like there might be something to toggle it, unless they really just meant it as a developer aid. "bootcmd=boot_mmc", but maybe there's something which changes bootcmd to boot_mmc_1.
<rsalvaterra>
mangix: I will take a stab at converting the C6 device tree to DSA, but I'll likely ask you to review my changes, since I'm not yet comfortable doing device tree work. :/
<hrw>
ok, 35 is in beta - 3.10 is expected around f35 release
eck has quit [Quit: PIRCH98:WIN 95/98/WIN NT:1.0 (build 1.0.1.1190)]
eck has joined #openwrt-devel
<rsalvaterra>
rmilecki: Who doesn't love Harry Potter? :D
<stintel>
o/
<rsalvaterra>
stintel: Oh, come on… :P
<rsalvaterra>
I'm playing with Nick Terrel's kerne zstd update series…
<rsalvaterra>
*kernel
<rsalvaterra>
… and so far I got it to compile without warnings (tons of fall-through complains).
<rsalvaterra>
*complaints
<stintel>
I managed to reorder the M300 ethernet interfaces, renaming no joy however, so it's going to be eth0, eth1, eth2, sweth3, sweth4, sweth5, sweth6, sweth7
<stintel>
but the digit will match what is on the front panel
<rsalvaterra>
The size, however… 4814432 vs 4671392 bytes, zstd-compressed (hah!), for my personal .config.
<will[m]>
what's the biggest obvious difference between the sdk, imagebuilder, and buildroot other than how you obtain it?
<will[m]>
i've got a docker image here i'm pretty sure i created, and i'm pretty sure it's a raw buildroot, but i named it "sdk" and it's only generating packages not images
<will[m]>
so i'm wondering if there's a CONFIG_* or folder structure that makes the difference obvious
dedeckeh has joined #openwrt-devel
<will[m]>
ahh seems to be CONFIG_SDK
<will[m]>
and i did accidentally use the sdk image instead of my buildroot image, fun
<blocktrron>
mangix: I've tried your branch on my 4q - doesn't work
<blocktrron>
what irritates me is the fact the Generic PHY driver is used for the WAN port - it should use at803x
goliath has quit [Quit: SIGSEGV]
<Ansuel>
blocktrron: is this about qca8k ?
Borromini has quit [Quit: Lost terminal]
<rsalvaterra>
Ansuel: Being blocktrron, it's likely the AR9344 internal switch, if I'm not mistaken.
<rsalvaterra>
Ansuel: I'm looking at this now… https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/ath79/dts/qca9563_tplink_archer-x6-v2.dtsi;h=fc6206d891a92249eabad471934ccade3b1ff7f3;hb=HEAD
<rsalvaterra>
… but I'm afraid I have no idea on how to convert this safely to DSA. For starters, to which switch port is the SGMII connected? :/
<Ansuel>
rsalvaterra: give me a sec to convert the reg
ldir has quit [Quit: +++Divide by cucumber error+++]
schwicht has joined #openwrt-devel
<Ansuel>
rsalvaterra: sgmii with rsalvaterra edge set
<Ansuel>
nothing else
ldir has joined #openwrt-devel
<Ansuel>
failling edge set*
southey has joined #openwrt-devel
<rsalvaterra>
Ok, falling edge, noted, thanks. :)
<Ansuel>
i should reall start working on a script... or even a little script to conver the regs to usefull info...
<Ansuel>
shouldn't be that hard... do you think it would be useful ?
<rsalvaterra>
Ansuel: Oh, definitely! Won't exclude human reviewing, but it should make the whole process a whole lot less tedious.
<Ansuel>
also pay attention it seems no secondary cpu connected
<stintel>
how do you make sure of that :)
<stintel>
I'm still trying to figure that out for the M300
<rsalvaterra>
Ansuel: Yes, the C6 v2 is a one-armed router. Only one eth port. :)
<Ansuel>
they don't set any reg to pad6 regs.... i notice oem never trust initial switch config... so we can assume no pad6 = no secondary cpu... also no phy4 defined in the node
<Ansuel>
in that case probably the wan port is directly connected to the soc phy (i think.. would make sense... but we notice some oem using only one soc phy and using all the 5 availablòe port of the switch)
<rsalvaterra>
It's sad, however, as the QCA9563 SoC contains two eth interfaces…
<Ansuel>
so much wasted resource
foxtrot has quit [Ping timeout: 480 seconds]
<mangix>
A wild Ansuel appears
<rsalvaterra>
Well. the AR9344 also has an internal switch, but it's Fast Ethernet. No wonder it's paired with the AR8327N on the WDR3600. :)
<mangix>
blocktrron: yeah no idea. Maybe I messed something up when getting the patches to apply to 5.10
<mangix>
Hmm? I thought qca956x was all single eth interfaces
<Ansuel>
mangix: hi
<rsalvaterra>
mangix: Check the bottom… https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/ath79/dts/qca956x.dtsi;h=8627f7ee40b192dd2976cd28d03a7a7954e456f3;hb=HEAD
hanetzer has quit [resistance.oftc.net weber.oftc.net]
slingamn has quit [resistance.oftc.net weber.oftc.net]
sauce has quit [resistance.oftc.net weber.oftc.net]
felix has quit [resistance.oftc.net weber.oftc.net]
arifre has quit [resistance.oftc.net weber.oftc.net]
dedeckeh has quit [resistance.oftc.net weber.oftc.net]
hurricos has quit [resistance.oftc.net weber.oftc.net]
norris has quit [resistance.oftc.net weber.oftc.net]
chwba has quit [resistance.oftc.net weber.oftc.net]
awgh has quit [resistance.oftc.net weber.oftc.net]
agb has quit [resistance.oftc.net weber.oftc.net]
larsc has quit [resistance.oftc.net weber.oftc.net]
noahm has quit [resistance.oftc.net weber.oftc.net]
<mangix>
cool
<rsalvaterra>
Oh, wow…! Also has a built-in switch.
<mangix>
Wonder why TPLink decided to switch to single port for their Archer series
<rsalvaterra>
compatible = "qca,ar8229";
<rsalvaterra>
Fast Ethernet too, no?
<mangix>
I guess so
sauce has joined #openwrt-devel
norris has joined #openwrt-devel
slingamn has joined #openwrt-devel
chwba has joined #openwrt-devel
larsc has joined #openwrt-devel
awgh has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
felix has joined #openwrt-devel
noahm has joined #openwrt-devel
agb has joined #openwrt-devel
hanetzer has joined #openwrt-devel
hurricos has joined #openwrt-devel
arifre has joined #openwrt-devel
<mangix>
rsalvaterra: the g in gmii == gigabit
<rsalvaterra>
mangix: I know. Totally missed the g, thanks. :P
<mangix>
The gigabit part is just the connection to the CPU port
<mangix>
Huh. The more I look at it the more confusing it gets.
<hauke>
mangix: you can probably also connect a 1G PHY to the (R)GMII interfaces
<hauke>
an external 1G PHY
southey is now known as foxtrot
<mangix>
Right. What's confusing is the interaction between port 5 and 6.
<hauke>
it looks like you can conenct port 5 or 6 to the last PHY and use one of them in (R)GMII mode or both of them in (R)GMII mode.
goliath has joined #openwrt-devel
<blocktrron>
mangix: i have it working now
<blocktrron>
my memories lied to me, the at803x driver was never used, just threw a patch at it
<Ansuel>
anyway woring on a script to parse all the regs so conversion can be easier. (aka user don't need to ask everytime sgmii or rgmii) Where should i post it when it's ready?
<blocktrron>
Also, the 956x dts is kind of messy, the qca9563 does not have a esw, while the qca9561 does not have the sgmii if
<mangix>
On a device I have, phy 0 and 6 are swapped. The DTS binding is there and it works with swconfig but not with the DSA one. I don't see phy6 referenced either.
pmelange has joined #openwrt-devel
<mangix>
Ansuel: as a fit commit probably. Easier to pull.
<mangix>
*git
<mangix>
blocktrron: I noticed an issue with one of the patches on latest master. Conflicts with a bcm backport.
<rsalvaterra>
Ansuel, mangix, I've changed things. I'll push them out and brace for the insults. :)
<mangix>
Oh. So upstream only accepts GPIO currently?
<blocktrron>
I'm aware of the mtk GPIO support, but i assume this went through as it can be used as GPIO
<rsalvaterra>
mangix: The LEDs are connected to GPIO, unless (highly likely!) I'm not understanding correctly what I'm reading… https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/ath79/dts/qca9563_tplink_archer-c6-v2.dts;h=c40fe0f29238149801adc21e5c4ccae0a30b1953;hb=HEAD
<blocktrron>
just sharing my experiences, not sure where upstream is heading to this week
<mangix>
rsalvaterra: those are LEDs not connected to the switch
<rsalvaterra>
mangix: Oh, so I should be able to control them as usual, right?
<mangix>
I assume you have separate LEDs for all Ethernet ports
<mangix>
If you want to control them, you need to add entries in DTS.
<rsalvaterra>
mangix: LOL, no. It's a single LED. These is a *very* cheap device, and it shows.
<rsalvaterra>
*This
<mangix>
Never mind then.
<rsalvaterra>
mangix: I have a vague memory I had to set a mask in /etc/config/system to change the LED behaviour.
<rsalvaterra>
(I haven't powered up the C6 in a few months.)
<Ansuel>
in the current implementation switch leds are exported used cled framework but are controlled by switch regs not gpio
<rsalvaterra>
Ansuel: I'm not sure about the fixed-link properties. I added them in two files, but I'm not sure if that's how it should be done.
<mangix>
Redundant in the v2 DTS
<Ansuel>
any cpu port should have the fixed-link set to gigabit
<rsalvaterra>
Thought so. Will discard.
<Slimey>
how do i force 5.10 to build for ath79 target?
<mangix>
Select testing kernel version in make menuconfig
<Slimey>
ah thanks last time i edited config
<rsalvaterra>
Also… how do I know that the port numbers match the labels…? :P
<hauke>
rsalvaterra: did you change anything for the bcm2710 in the kernel 5.10.69 update?
<mangix>
rsalvaterra: plug and check ifconfig for traffic
<rsalvaterra>
hauke: Nope, nothing at all. :/
Tapper has quit [Ping timeout: 480 seconds]
<mangix>
that’s what I did for my E7350 and found it to have questionable assignment.
<mangix>
WAN/lan4-3-2-1
<rsalvaterra>
Ok, building an image…
<rsalvaterra>
Removed swconfig, enabled DSA QCA tagging and the qca8k driver in the kernel config…
<Ansuel>
i think the led output is a bit confusing if you want i can try to improve that
<Ansuel>
i can also elaborate more the data to report directly to use rgmii-id or the mode to use
<rsalvaterra>
Ansuel: The output does look a bit confusing, but the fact that you hacked this together in a couple of hours is impressive already.
<Ansuel>
the base code is there... the hard and tedious part of parting the various regs is already done... soo elaborate the data is not reallt that problematic... Will check what i can do
<Ansuel>
lets have some fun ahahah
<mangix>
Ansuel: the LED stuff is not important. The mii mode is
<mangix>
although...
<stintel>
rsalvaterra: no, this is an SD
<mangix>
it can be derived from pad0/6...nvm
<mangix>
Ansuel: what's this sel 1 and sel 2?
<Ansuel>
i hide the settings in later version. it's the actual delay applied to tx and rx... by default it's 1 for tx and 2 for rx. we don't need to set that as it'a applied by default
<mangix>
OK
<mangix>
I also don't see anything for qca,sgmii-rxclk-falling-edge. I assume that can't be derived.
<mangix>
oh wait...
<mangix>
ah I see. I meant qca,sgmii-disable-pll
<slh>
rsalvaterra: I have ~12 year old cheap sdhc cards that do fstrim just fine
<stintel>
ok so nothing new
<rsalvaterra>
slh: Interesting. I need to check my cards, then.
<Ansuel>
mangix: yes there is i actually check if pll is enabled
<Ansuel>
but i'm elaborating the data so i can make tha clear