<f00b4r0>
robimarko: very nice reply too. No fuss and to the point ;)
Tapper has quit [Ping timeout: 480 seconds]
<russell-->
engenius ews2910p (realtek switch) arrived today, saved a copy of flash, tftpboot'd, sysupgraded. mounted the vendor jffs2 partitons, found a bunch of pem's and ssh/authorized_keys, scary!
SherlockDomes2 has quit [Quit: Nettalk6 - www.ntalk.de]
goliath has quit [Quit: SIGSEGV]
goliath has joined #openwrt-devel
Tapper has joined #openwrt-devel
isak has quit [Read error: Connection reset by peer]
isak has joined #openwrt-devel
<\x>
hap ax2 is announced?
<\x>
think is it ipq60xx?
<robimarko>
Or IPQ807x
<robimarko>
But they will release it next week
<robimarko>
No idea why they even released the video
Tapper has quit [Ping timeout: 480 seconds]
<\x>
1GB ram
<\x>
so not like ac2 which is starved
<\x>
secureboot enabled likely? haha
<robimarko>
Yeah, they figured out that getting it to work with less is more work
<robimarko>
Would be surprised if secure boot was not enabled
<robimarko>
Though, they havent been doing that on the IPQ40xx devices
<robimarko>
Probably as its not easy to sign the RouterBoot
<\x>
is your 807x work ready for prime time?
<robimarko>
Depends on your definition
<robimarko>
It works fine I would say.
<robimarko>
Ethernet driver is the biggest issue, its really badly optimized as there is no offloads
<robimarko>
Not even checksum, let alone TSO
<robimarko>
As they only focused on using NSS offloading which implements all of those in FW
<robimarko>
So, I need to backport all of the upstreamed stuff
<robimarko>
As its way cleaner than the current 5.15 tree, ath11k needs some cleanup and the target should be ready for a PR
torv has quit [Remote host closed the connection]
<f00b4r0>
robimarko: birger apparently deleted his GH account. Looks like ragequit ;P
<robimarko>
I saw that he closed the PR
<f00b4r0>
his comments now appear as "ghost"
torv has joined #openwrt-devel
<robimarko>
I really dont get him
<robimarko>
He is mad because people actually reviewed his PR
<robimarko>
And not just applied it
<f00b4r0>
yeah he seems to have trouble coping with reality.
<robimarko>
Still, I would say a bit of overreaction, especially since nobody was rude or just ignored him
<f00b4r0>
I'm not surprised. His comments were out of place and clearly "edgy". Seems he doesn't tolerate contradiction.
<robimarko>
That would explain why he refuses to even try and upstream simple parts like SPI controller support
<f00b4r0>
*nod*
<\x>
is 807x cpu cores enough for gigabit SQM
<hitech95>
Is there a status page about the work done/missing for the rtl838x switches? the realtek page in the wiki is basically empty.
<robimarko>
\x: Well, if the ethernet driver was better then yeah
<robimarko>
Also, depends on whether you use IPQ8070/1A or others
<robimarko>
As these two are 1.4GHz while others are 2.2GHz
<robimarko>
Ethernet driver really needs improvements, I looked at it and only managed to add NAPI GRO
<robimarko>
As there are no clues what the bits for checksum and TSO are
MaxSoniX has quit [Quit: Konversation terminated!]
<russell-->
so, what's the current recommendation for poe on realtek?
<mrnuke_>
russell--: I'm biased because I wrote half of it, so I'll say realtek-poe :p
<hauke>
mrnuke_: is someone working on getting one POE interface into the mainline kernel?
<robimarko>
hauke: None that I know of
<robimarko>
I can tell you that there is some interest and scoping regarding what would it take
cbeznea has quit [Quit: Leaving.]
<hauke>
robimarko: ok
<hauke>
I am just asking, currently I am not looking into PoE
<svanheule>
hauke: I know there's a PSE chip from TI in the kernel, which is a pure hwmon driver IIRC
<mrnuke_>
svanheule: with a patch pending to fix the byte order...
<svanheule>
mrnuke_: right, your patch ;)
<svanheule>
but that only runs in some automatic/unmanaged mode, right?
<mrnuke_>
svanheule: Let's not get into who did what. We're all in this together. :p
<svanheule>
:P
<mrnuke_>
You can set the TI 23816(something something) to "auto" mode, and it will start providing power
<mrnuke_>
There's a toggle in the hwmon driver to enable power, but doesn't work if the chip is not in "auto" mode
<mrnuke_>
svanheule: I just recieved today a switch with two of these chips. It will be ineteresting to see how it does
<svanheule>
mrnuke_: more patches coming up? :^)
<mrnuke_>
Probably, but next on my list is EAP610-Outdoor!
<svanheule>
ath11k!
<mrnuke_>
Yes, I think so!
<svanheule>
there's a reason I never started looking at those, after the EAP2xx devices...
<svanheule>
also because money was flowing into switches maybe
<mrnuke_>
yeah, that realtek target is more popular than I ever thought it would be
<svanheule>
it's cheap for managed switches, right? And then you find out *why* they are cheap
<svanheule>
although I am running a Cisco SG220-26P with vendor FW at work, doing its job quite nicely
<svanheule>
heard people say OpenWrt is the only reason DGS-1210 devices are useful though :P
<f00b4r0>
<svanheule> it's cheap for managed switches, right? *And then you find out *why* they are cheap*
<f00b4r0>
yeah that, exactly ;P
<robimarko>
I wrote the driver for tps23861
<robimarko>
And the BE fix was merged AFAIK
<robimarko>
Driver only support auto mode
<svanheule>
robimarko: did the output-enable-via-hwmon cause much of a fuss?
<robimarko>
svanheule: none at all
<robimarko>
As it uses an existing standard property
<mrnuke_>
robimarko: Hmm, I thought something looked familiar with that driver
<robimarko>
And its in auto-mode anyway
<robimarko>
Driver is there just to read basic info out and be able to manually disable ports
<mrnuke_>
I swear by my TPS's sense resistors that the chip did not come up in auto mode on my TL-SG2008P
<robimarko>
Driver does not touch that
<robimarko>
You can bootstrap the IC in manual or some semi-auto mode AFAIK
<mrnuke_>
I know it doesn't. My workaround was to modprobe -r && i2cset && modprobe :p
bluew has joined #openwrt-devel
<hurricos>
RE: birger, some people just get really angry for a moment and do self-destructive things.
<robimarko>
mrnuke_: Just figured out that you sent the BE fix
<hurricos>
mrnuke_: I'm going to go order one of those I2C switches, unless svanheule has recommendations for one less than $175 (GS1920-24HP)?
<mrnuke_>
hurricos: the more time it takes you to get one, the less I have to hurry on my end. No rush. Take your time, do this right :p
<mrnuke_>
robimarko: knowing that you wrote the driver, I can no longer poke fun at the original author :p
<robimarko>
Feel free to continue doing so
<hurricos>
I lowball offered the seller, telling them my intent, asking them to give a dev a deal, and adding one too many winky faces. ;);)
<robimarko>
Was one of my first "drivers" that I managed to upstream
<robimarko>
Also, had to learn quite a lot about PSE, but still managed to mess up stuff
<svanheule>
hurricos: :P I think I got mine for <€100; got the SG220-26P-s for €70 a piece; a steal, really
<hurricos>
svanheule: finally, we americans are getting worse deals on hardware.
<f00b4r0>
lies ;P
<mrnuke_>
robimarko: to be honest, C is one of the few languages that allows someone to mess up byte order. Normally you can't convert from a byte stream to an int -- C allows it
<robimarko>
mrnuke_: Well, I kind of got used to little-endian as its so prevalent
<robimarko>
I am sure there is tons of drivers in the kernel that forgot this as well
<f00b4r0>
my first too hacking archs were parisc and ppc. Biased me the other way around ;)
<hurricos>
svanheule: never mind, it's definitely realtek.
* f00b4r0
has a couple HPE V1910 and no idea what the hardware actually is
<svanheule>
f00b4r0: HPE-1910 *gigabit* switch are some Marvell thing IIRC
<f00b4r0>
svanheule: i see, thanks
<svanheule>
the 100Mb ones are realtek
<hurricos>
svanheule: probably Prestera
<hurricos>
linkstreet + prestera is an extremely common combo. the linkstreet CPUs I see are MMUless though
<robimarko>
Which Prestera gen?
<f00b4r0>
i have the gige variants. Nothing to complain about so far
<hurricos>
Rumbling through my brain, I pulled one apart and showed a kid the internals at our repair cafe
<f00b4r0>
(except that they no longer receive updates, of course)
<hurricos>
robimarko: not the one you want though
<hurricos>
98DXBCW1
<hurricos>
Predates when they started adding the management CPU to the switch ASIC
<robimarko>
Oh yeah
<hurricos>
Not an OpenWrt target ;(
<robimarko>
For me only AC3X and AC5X are interesting
<hurricos>
robimarko: I saw those and wanted to ask. Haven't seen any hints of a switchdev drive yet though.
<hurricos>
At least they're not Prestera :^)
<robimarko>
There is an upstream driver for AC3X
<robimarko>
AC5X will eventually get covered, ps: you can use the "DENT" version of driver alredy on it
<f00b4r0>
stintel: btw, do we have a todo list for the m300? I might play with mine a little more before I put it to actual use ;)
<hurricos>
I never bothered pursuing DENT ultimately, I care about the practical :P
<hurricos>
If OpenWrt can't build a toolchain for it ... ahem
<robimarko>
I dont use DENT, only use the Prestera driver
<robimarko>
As Marvell added quite more features there
<hurricos>
I'm also very unfamiliar with switchdev. Thankfully I don't work in the embedded world, yet :^)
<hurricos>
robimarko: I was wrong. I pulled an older switch apart than the HPE V1910.
<hurricos>
robimarko: I do not see separate RAM on the V1910 motherboard, but it uses 5 Marvell QS?GMII to 1000BASE-T PHYs, all connected to ... probably still a Prestera.
<hurricos>
the 6th QSGMII PHY provides both copper and SFP
<hurricos>
but yeah, no hope for drivers for the ASICs
<russell-->
is there any documentation for realtek-poe?
<mrnuke_>
russell--: '#ubus -v list poe' is probably the least bad documentation about it :D
<mrnuke_>
There really isn't too much
<russell-->
what is "option priority"?
goliath has quit [Quit: SIGSEGV]
robimarko has quit [Quit: Leaving]
bluew has quit [Quit: Leaving]
<mrnuke_>
Doesn't ring a bell. What's the context of this "option priority" ?
bluew has joined #openwrt-devel
<mrnuke_>
russell--: You probably mean "priority" in the config file. As the power budget is reached, the controller will start shuttin down ports. The order is based on priority. Lowest priority ports get shut down first
<russell-->
mrnuke_: thanks
* russell--
apparently doing something wrong, no power to lan8
Rayyan_ has quit [Remote host closed the connection]
Rayyan has joined #openwrt-devel
<mrnuke_>
russell--: `logread -te tek-poe` show anything scary?
<mrnuke_>
russell--: I presume `ubus call poe info` will show "unknown" for all ports
<russell-->
I only have one port configured, but it does show status unknown and firmware v0.0
<mrnuke_>
As I suspected. It's not even applying the config.
<russell-->
i just added configuration for all the ports
<mrnuke_>
Does -- ubus call poe sendframe '{"frame" : "09 00 01" }' -- do anything (That's the reset command, BTW)
<russell-->
doesn't seem to
<russell-->
this is current HEAD
<mrnuke_>
There isn't much I can do remotely if the MCU isn't responding. On the other hand, if this is a comms issue, you might be able to see it with strace
<mrnuke_>
if you run realtek-poe under 'strace', does it receive _any_ data over ttyS1 ?
* russell--
builds strace into firmware
<mrnuke_>
russell--: I'll be gone for a while. If there's no data, the other thing you can try is boot vendor FW, see if that works, then boot OpenWRT, see if that works.
<russell-->
not seeing anything come back from ttyS1 from strace
cmonroe has quit [Ping timeout: 480 seconds]
<russell-->
using the pin 16 trick, got u-boot prompt, bootm'd the linux2 partition which has the vendor firmware, poe was enabled, rebooted into openwrt without powering off, port remains on