<damex>
Grommish: we can split after (and that is if) we add more devices. need to get a hold of edgerouter 6p and find about bringing up that weird switch platform on er12/er12p
<rsalvaterra>
"QCN9074 uses 45MB of host ddr memory."
<rsalvaterra>
And here I was thinking ath10k was a memory hog and ath11k could only improve…
<rsalvaterra>
Oh, and 45 MB in *physically contiguous* chunks of 2 MB. Peachy.
<blogic>
rsalvaterra: yes
<blogic>
rsalvaterra: a 1GB ram unit will OOM at ~120 clients
<blogic>
rsalvaterra: there is also 16 mb reserved for uboot and 64 for the NSS 4 more for smem, another 8 for core dumps for the wmac ....
<blogic>
so a 512mb unit one it is up has ~360mb free ram ;) rest is hogged by the HW
<rsalvaterra>
blogic: How's MT79xx in comparison?
<blogic>
nice
<blogic>
has it running with 256 clients attached pumping traffic on a 512mb unit
<blogic>
also traffic was more stable
<blogic>
problem right now with mt7622 is that the hw is only just appearing on the market
<rsalvaterra>
So, so what are the QCA guys smoking? And where can I get it?
<blogic>
and the uboot is shitty, however dangole is working on a uboot replacement that has ubi support
<blogic>
so 2-3 months down the road mtk should be in a fit state, I am already running my network on mt7915
<blogic>
linksys e8450 is a nice unit
<blogic>
however needs to be imported from the states, not sold in EU
<rsalvaterra>
My Omnia with two QCA9880 hovers at about 120 MiB of RAM used. My Redmi RM2100 (MT7603 + MT7615) is using just under 60 MiB.
jas4711 has quit [Remote host closed the connection]
numero53 has joined #openwrt-devel
<numero53>
blogic: In Europe there is the Belkin RT3200 that should be the same exact router rebranded. I bought it from Amazon UK (I live in Italy)
<numero53>
Is your nand working correctly? I have problems with JFFS2 corrupting often... I tried the nbd's tree (which contains your work I think) and I have the same problem...
numero53 has quit [Quit: Connection closed]
numero53 has joined #openwrt-devel
Tapper has quit [Ping timeout: 240 seconds]
Tapper has joined #openwrt-devel
swalker has quit [Remote host closed the connection]
<karlp>
never going to get supported with that attitude ;)
<stintel>
hehe :P
adrianschmutzler has joined #openwrt-devel
<numero53>
ynezz: Surely there wasn't a sata port accessible on the board... Probably another uart or spi at most. There were 10 holes and some of them were powered with different voltages, but I didn't investigate further
lnslbrty has quit [Read error: Connection reset by peer]
lnslbrty has joined #openwrt-devel
csrf has quit [Remote host closed the connection]
<adrianschmutzler>
Hi, I have stupid question
<adrianschmutzler>
In file header for license or comments, it's frequently the case that the header is followed by an empty commit line, just "#"
<adrianschmutzler>
Is there a reason for that, or is that just the taste of the person who wrote it?
csrf has joined #openwrt-devel
SergioCabral has quit [Remote host closed the connection]
<adrianschmutzler>
I see this as a mere replacement, and do not think changing the date is really necessary there. In contrast, changing the copyright date is actually making this much more complicated, as we have seen when discussing your PR.
koniu has quit [Remote host closed the connection]
koniu has joined #openwrt-devel
feriman has joined #openwrt-devel
<Hauke>
damex: I have never used Cavium devices
<Hauke>
blogic: is Belkin RT3200 (mt7622 + mt7915) with 4x4 2.4GHz wifi (mt7622) + 4x4 5GHz (mt7915)?
<Hauke>
which switch is used=
<Hauke>
looks like it is this combination
victhor has joined #openwrt-devel
numero53 has joined #openwrt-devel
<numero53>
Hauke: Your specs are correct
<numero53>
The switch is the MT7531
<philipp64>
how much retesting is required when cherry-picking fixes into 19.07?
<philipp64>
there doesn't seem to be any automated tests and I don't have a testbed I can easily reflash and test on...
<adrianschmutzler>
depends on the nature of the change, obviously ...
<blogic>
I aked a US based friend to send me 10 additional unit after he sent 4 already
codehero has joined #openwrt-devel
<blogic>
the pita is that the wan port uses the mt7531 switch which only doe 1gbit
<blogic>
so even if the wireless can do beyond that you can only trunk 1gbit on fiber
<blogic>
Hauke: its mt7531 ...
<blogic>
been running the unit with wifi6 on 4 deployments over the last week and evryone is happy
<blogic>
yet the 2G is the built-in mt7622/7615 11n wmac
<blogic>
so no huge 2g rates
<blogic>
oh and the serial is a 1,5m pitch molex, so one needs a special molex2duponmt cable for usb serial
<blogic>
still a pretty f'in cool unit for that price
gromero_ has joined #openwrt-devel
codehero11 has joined #openwrt-devel
<blogic>
and we already booted it with 2021.4 uboot + a small amount of patches, so we can prolly drop the shitty mtk bmt-v2 nand wear leveling layer soonish and just do ubo on the spi-nand
<blogic>
*ubi
codehero11 has quit [Client Quit]
<blogic>
fun fact, the stock webui consumes raw owrt stylw sysupgrade images
<blogic>
*style
<blogic>
images buildable from nbd's staging tree with v5.10
gromero has quit [Ping timeout: 240 seconds]
<Borromini>
blogic: are you talking about the linksys e8450?
<blogic>
numero53: ordered a unit from the uk, import tax will be a pain, but really wanna see if it is also easy to flash
<blogic>
Borromini: yes
<Borromini>
:)
<blogic>
Borromini: first mtk wifi6 unit i found that is not atf/rsa locked
<blogic>
numero53: yeah, read that, it boils down to some crap in the hw eec, post chinese ne year one of the mtk devs promised to gpl the latest driver that they have that apparently fixes the problem
<blogic>
so close, my fingers are tingelining
<Hauke>
I ordered one unit on amazon uk, I think amazon will take care of taxes at least they chared me ~10£ for that
<ynezz>
+1 :p
<numero53>
blogic: thanks a lot for the great news! :)
<Hauke>
blogic: do you have the TP link unit with the same specs? does it use secure boot?
<blogic>
there is a unit it uses secure boot, we manged to hack into the stage 1 uboot but it has been obfuscated, the stage 2 uboot is rsa locked
<blogic>
i tried the usual webui injection crap but gave up
<blogic>
been talking to tplink about it and might find a vector to owrt'ify it
<Hauke>
ok, so the Belkin is the better device for now
<blogic>
I emphazied to them that this i not cool and newer units should not be shit-craped
<blogic>
Hauke: correct
<blogic>
there is also a adtran unit that will hit market shortly
<Hauke>
also to the EU or only US?
<blogic>
no idea
<blogic>
i just ordered one of the uk units
<Hauke>
looks promising when now multiple vendors have mt7622 + mt7915 devices in the market
<Hauke>
I was wondering why there were only so few mt7622 devices on the market before
<ynezz>
no official openwrt support :p
ivanich has quit [Quit: Konversation terminated!]
<blogic>
Hauke: so production capacity is locked down in .ch till end of the year
<blogic>
huawei bought all of it
<blogic>
there is a huge production traffic jam in .ch
<Hauke>
The belkin devices has Samba 3.6 in the source code delivery, if this is on the device and they would use secure boot I have the right to get the key ;-)
<blogic>
this is for cpe as well as silicon
<Hauke>
yes I know of the delays in the chip production
<Tapper>
I don't use luci-app-wireguard and luci-app-advanced-reboot on my r7800. advanced-reboot worked for me tho on the wrt3200acm.
<Tapper>
pkgadd: remember I am using a screen reader so if that one is about how it looks I will not get it.
<pkgadd>
yeah, I realize that there might be differences in that regard
<Tapper>
pkgadd: sorry I should of read more.
<Tapper>
It seems most of the ones I was on about have bin pointed out.
<Tapper>
The packages anyway.
nitdega has quit [Quit: ZNC - 1.6.0 - http://znc.in]
nitdega has joined #openwrt-devel
<mangix>
Pepe: I don't have control over such things
Acinonyx has quit [Ping timeout: 264 seconds]
Acinonyx has joined #openwrt-devel
<hurricos>
Got SFP+ ports working under OpenWrt on my CN6640-SNIC10E card :)
<hurricos>
there was a CSR poke that Cavium's ethernet driver does but which was noted to hang CN6880-based cards
<hurricos>
looks like they never checked CN6640
<hurricos>
since this code is shared with the cavium simple executive this could also be the bug that is preventing the CSE from working on these cards? not sure
<hurricos>
or rather liquidio from working on these cards*