owrt-1907-builds has quit [Server closed connection]
owrt-1907-builds has joined #openwrt-devel
srslypascal has quit [Remote host closed the connection]
srslypascal has joined #openwrt-devel
srslypascal is now known as Guest1463
srslypascal has joined #openwrt-devel
Guest1463 has quit [Ping timeout: 480 seconds]
cmonroe has quit [Ping timeout: 480 seconds]
clayface has joined #openwrt-devel
<f00b4r0>
hauke: re #3271 I had suggested some changes in my review, these seems to have not been picked up by whoever grabbed that patch and pushed it to our tree. Not a big deal but some of said changes may help mainlining
clayface has quit [Ping timeout: 480 seconds]
clayface has joined #openwrt-devel
<stintel>
hmmm looks like I'm having performance issues on realtek
<stintel>
switched traffic of ~80Mbps results in very high sirq (seen up to 85%)
<hurricos>
stintel: which board?
<stintel>
gs108t-v3
<hurricos>
:nods: I'm working on installing some ZyXEL realtek switches in our building, so I'll keep a look out for any issues.
<stintel>
hope the flash is large enough to include perf
<hurricos>
You can always install to RAM temporarily, too.
<hurricos>
nasty as that might be. I recall seeing a way to force that with some env vars to opkg
<hurricos>
svanheule: I mean the notes you took before you published a nice page :^)
<hurricos>
PaulFertser: I will patch it in and recompile. Thanks.
<svanheule>
hurricos: errrr... not really :-/ I basically refactored that page as I went along discovering more things, having annotations (variable names, rebuilt structs) in the decompiled code
<svanheule>
hurricos: and probably some notes on paper so I didn't have to keep everything in my head, but I'm sure I wouldn't be able to make head or tails out of that if I saw them again
<hurricos>
No problem, I understand. Just trying to get a sense of where to start with this crash. PaulFertser: Funny enough, objdump pointed at a `mode.3` somewhere. I can't see what this should be, what's your diff to fix this?
<PaulFertser>
hurricos: the crash must be something different. mode.3 sounds like an inlined instance of a function probably.
swegener has quit [Quit: leaving]
swegener has joined #openwrt-devel
<hurricos>
I think I misread the error anyways. It's an invalid memory access *to* 04020cc, then looked for 4020cc in less and found a word of machine code '004020cc'.
<hurricos>
so once misread by looking at where it was accessing, whether or not it was trying to execute at that address, and then once because I read machine code as an address.
<hurricos>
I don't even know what's supposed to be in the ra register in mips.
<hurricos>
Oh, the return address.
<hurricos>
Oh yeah, so it crashes the first time it runs ULOG_ERR.
<Borromini>
hurricos: for me it compiles/runs fine, on recent master
<Borromini>
except for the weirdness with my GS1900-8HP v1 not doing any PoE anymore on OpenWrt altogether
<Borromini>
but realtek-poe doesn't segfault here or anything
<hurricos>
Borromini: Does it work on oem firmware?
<Borromini>
hurricos: yep. without a hitch
<Borromini>
first thought was hardware failure too, but PoE stopped working for me somewhere between a post-release 21.02 build and master
<Borromini>
lowering the PoE budget seemed to fix it very briefly but it broke down anyway somehow
<hurricos>
I wish I knew which framework was going to be adopted for PoE stuff. I notice ynezz posted robimarko's driver for a ... TPS? PoE injector.
<hurricos>
err. Injector. Hardware.
<Borromini>
rebooting into OEM powers up my PoE client consistently
<hurricos>
but that's in-kernel, it gives a good idea for how it might ought to work
<hurricos>
which I don't really yet understand beyond that it exports some basic stuff to sysfs
<Borromini>
my GS1900-10HP, however, it works flawlessly with 21.02 HEAD and realtek-poe
Ntemis has joined #openwrt-devel
<Ntemis>
hello
<Borromini>
hurricos: to me the hex values realtek-poe -d outputs is chinese though, i have no idea how to debug it
<Borromini>
hurricos: was reading back the C daemon discussion on Patchwork and bumped into this as well, haven't given this a shot yet: https://github.com/blocktrron/poemgr
<Borromini>
so far seems to support just one type of PoE hardware - Microsemi PD69104
<hurricos>
Borromini: Each line is a 12-byte frame.
<Borromini>
hurricos: ok thanks. gonna run that on my 8HP v1.
noltari has quit [Server closed connection]
noltari has joined #openwrt-devel
<hurricos>
Well, no, sorry, I know. You just menitoned it was realtek. svanheule clarified that they all use Broadcom stuff.
gladiac has quit [Quit: k thx bye]
<hurricos>
poemgr would need to have code written for it to work here. I also read, but may have misunderstood, that the intent with poemgr isn't to provide a ubus API, which is sort of a shame
<Borromini>
no worries :)
<Borromini>
i think the 8HP v1 uses an older Broadcom chip
<hurricos>
Borromini: Oh, hmm.
<Ntemis>
There is an issue with zyxel_nbg-419n-v2 ramips-rt305x
<Ntemis>
never worked on -21.02.x
<Ntemis>
i think is a kernel issue
<Ntemis>
works fine up to 19.07.9
<Borromini>
hurricos: STMicro PoE controller (?) seems to be identical between both - STM32F100C8
<hurricos>
You should (probably) bisect for the issue.
<Borromini>
on realtek-poe? or on the openwrt codebase itself you mean?
<Borromini>
realtek-poe hasn't changed in ages, AFAIK.
<hurricos>
OpenWrt codebase. You did mention it stopped working somewhere between versions.
<Borromini>
yeah. wife will kill me though.
<hurricos>
Not that that makes much sense, but maybe it stopped working because of driver changes that impacted serial communications
<hurricos>
Oh. Buy another, then.
<Borromini>
fair question
<Borromini>
hahaha :D
<Borromini>
the 8HP v1 is a testing device, luckily
<Borromini>
the 10HP is production. you can understand I am reluctant to upgrade at this point :^)
<Borromini>
i've had my share of flash dumping / debugging / timewasting / guinea piggery with the RTL930x platform lately :(
<hurricos>
Sorry to hear :\
<hurricos>
I am thankful they are dual-partition devices. I can just fw_setenv bootpart and reboot
<hurricos>
(if I need ZyXEL's functionality)
<Borromini>
yeah. i'm happy with that too.
<Borromini>
except the XGS1250-12 e.g. doesn't seem to be dual-partition.
<Borromini>
but yeah, the GS1900 series is :)
ephemer0l has quit [Server closed connection]
ephemer0l has joined #openwrt-devel
GNUmoon has joined #openwrt-devel
Borromini has quit [Quit: leaving]
<Grommish>
neggles: stintel: I just tried 5.10(.103) again, I've lost nearly 250Mb in 17 hours of uptime without an apparently source of what's using it :(
<stintel>
I've also hit the leak again after trying to reproduce it for 2 weeks and acking the switch to 5.10 because I was unable to
<stintel>
typical
romany has quit [Server closed connection]
romany has joined #openwrt-devel
<neggles>
guh
<neggles>
that's just weird
<neggles>
mine's been idle for a while, I managed to kernel panic the system it's in a few days ago (by running ripgrep???) and didn't reload
drikus_ has quit [Server closed connection]
drikus_ has joined #openwrt-devel
Sagi has quit [Server closed connection]
Sagi has joined #openwrt-devel
Ntemis has quit [Quit: Page closed]
* Slimey
chokes hurricos with a smiley printed in china
<stintel>
ツ
clayface has quit [Ping timeout: 480 seconds]
szy_mat has quit [Server closed connection]
szy_mat has joined #openwrt-devel
clayface has joined #openwrt-devel
cmonroe_ has joined #openwrt-devel
<stintel>
Grommish: which device were you seeing this on?
<hurricos>
OK, so you're not going to believe this
<hurricos>
The very common Powerdsine 9012G ...
<stintel>
Grommish: and do you happen to use sqm-scripts or qosify? or manually configure cake?
<hurricos>
is an m68k CPU connected to a tiny Freescale mcu driving a PD69104
<Grommish>
stintel: Nope.. It's on the Itus Shield. It's a fairly defconfig build
<hurricos>
m68k hasn't been OpenWrt supported since ever, and in fact I'm pretty sure these ... yeah, no mmu
<hurricos>
CONFIG_M5272 depends on ! CONFIG_MMU
<hurricos>
so if they run linux it ain't much. But this thing probably doesn't
<Slimey>
i have some of those midspans sitting in storage
<hurricos>
I have this devilish idea -- strap the serial or a USB-to-serial (there's a CP2102 on-board) from an OpenWrt device and implement poemgr stuff for it
<hurricos>
and if need be, reverse engineer the firmware running on the little guy to make it easier
<Grommish>
hurricos: Sometimes, you just gotta do what you love...
<hurricos>
OR, and this one's all the sweeter, pull the MCF5257 module, whip up something in kicad with an AR9331 or similar, and plant it in the old socket
<hurricos>
actually
<hurricos>
the pitch is exactly right for raspi headers. Let me trace the voltages
<hurricos>
it probably won't work
<hurricos>
because you know, the chances of the right pins being SPI for the many 595 expanders and VCC etc are like 0