broly
81
not talking about openwrt.
regarding the possibility: i think it's supposed to be possible? isn't this how they define a tri-band router? to run three bands at once. in this case you should be able to run two 5 ghz on different channels plus a 2.4ghz.
if it's just a bunch of 5 ghz ssids on the same channel, it's no different than creating a bunch of vaps on a 5ghz radio.
these are the mac addr of my interfaces:
lan :01
wan: 05
wl0 : 02
wl1: 03
wl2: 02
it suggests to me that there could be 3 hardware-separate radios?
from iw list:
phy 0 (index 0): dual band. band 1 2 g, ch 1-14, HT20/Ht40. band 2: 5G, VHT80 short GI capable, 2 streams, CH 36-155, radar detection on ch 52-144
phy1 (index 1): 5g only, HT20/40 and VHT80/160 capable, 4 streams, CH 36-165, radar detection 52-144
phy2 (index 2): 5g only, HT 20/40, VHT80 short GI capable, 2 streams, CH 36-165, radar detection 52-144
these indices dont exactly line up with the radio designations in etc/config/wireless or the gui, since it looks like radio 1 / phy1 has VHT160/4 stream capability but my faastest wifi was on radio2 (?)
i quickly need to point out that i'm confused by some of this.
the pci designations in my wireless are:
radio 0: option path '1e140000.pcie/pci0000:00/0000:00:00.0/0000:01:00.0'
radio 1: option path '1e140000.pcie/pci0000:00/0000:00:01.0/0000:02:00.0'
radio 2: option path '1e140000.pcie/pci0000:00/0000:00:00.0/0000:01:00.0+1'
tgt79
83
the pci designations mean that radio2 is a subdevice/virtual function of radio0 and they are essentially the same device, radio 1 is a separate pcie card. which makes me wonder why radio2 was faster for you (or rather worked at all with dbdc not correctly set up). maybe the firmware for the radios isn't correctly extracted and uploaded?
broly
84
it is virtual at the pci level, as in they belong to the same interface, but but in another sense this interface has two 'radios' (capable of handling two bands at once on the same interface). it's actually one radio, but with the ability to have two bands at once. this is what 'DBDC' (dual band, dual channel).
the radio firmware handles the abstraction, not the operating system. so one physical interface does indeed create a virtual interface for the second 'radio' (again it's ust one radio) because of the above fact.
one of the 5ghz band interfaces is supposed to be the 'fastest'. one is rated at 1733, the other at 800, and the remaining 2.4ghz is rated at 600 or something like that.
welcome to the mind fack known as DBDC mode. where a single physical radio is capable of handling two bands/channels at the low-level, but abstracts this to the operating system (hence the virtual interface).
broly
85
looking at the OEM bootlog, it seems the configuration is as follows:
radio 1 (with two bands, 600 and 800): 2.4G and 5.0 G (ra0, rax0)
radio 2 (one band, 5ghz, 1733): 5.0G (rai0)
no wonder the dir853 guy is so eager for this router to get added, because it would allow the 853 to be added with ease.
the nuisance is the first radio with two bands and two channels.
tgt79
86
yes, this is kinda what i meant. radio 1 is supposed to be the fastest, radio 0 and 2 provide the slower 5g and 2g bands. I am currently confused about wether the mt7615 have their own flash with firmware or if the firmware is loaded from user-space by the driver.
wrt DBDC: The same abstraction happens on datacenter NICs, where you can configure a physical 10G/25G/100G... interface to be presented to the os as several virtual interfaces with varying data rates - whose sum can eaven be higher than the maximum rate of the physical interface in some implementations.
broly
87
gotcha, no problem.
with either the proprietary driver or the mt76 driver, it is loaded by the driver into the 'Factory' partition. this is why nothing is supposed to touch the Factory partition, since aside from the spots for the mac addresses (which the drivers know about), the rest of it is buffer/play space for either driver.
look at
to inspect the firmware files. very similar files are used for the proprietary driver, but i suspect they have minor differences to accommodate the different approaches taken by the respective teams.
i believe you when it comes to datacenter stuff. this device (the mt7615) is still a work in progress five years later. it is quite the marvel if you ask me, and i shouldn't be surprised it took the community approximately this long to even get to DBDC.
i've been working with this platform for approximately 3 years and change. in that time, i really did not address DBDC mode because the 853 was of little/no interest because getting it working well took some time.
i haven't looked in-depth at the AX line, but i'm curious if mediatek is using a similar approach on AX devices as well, which would create a large incentive for me to work with this platform to get a feel for how i need to add the additional interface.
i have a good idea of what i want to do already, and it seems it should be easily accommodated in DD-WRT (but you still have to compile it a bunch of times and test the additions), but i need to see if this would be a worthy investment for what lies on the horizon.
broly
88
seems the MT7915 in the DIR-X1860 is DBDC-ified too
UGH
as i previously wrote, the openwrt firmware is not quite stable on the dir3060.
after a power cycle, necessary for reboot, this is what comes up:
as in the oem bootlog, radio0 is dual band, radio1 is 5G. radio2 is tagging on radio0 but wont come up apparently without restarting the interfaces.
i also cannot get connected to radio1. the signal strength is weaker and previous connections were slower. i think however openwrt is inializing this radio, it does not have optimal parameters.
i really appreciate this help.
DBDC means dual-band dual-channel on one radio?
broly
90
that's correct, ghoffman.
DBDC is hard, bro. lol.
oh how nice, you already had the serial header on the board lol!
boy if you had to solder the 882/2640 board, let me tell you.... and ive had plenty of experience lol....
i was wondering how you were able to get a serial log so casually. and that's no disrespect intended to you. but removing silver solder and having to put in your own with the location of that 4 pin... it's not easy. i am surprised i was able to do it 3 years ago.
tgt79
91
I was wondering if there are maybe different firmwares for dbdc and openwrt is loading the wrong one. but then the second interface probably wouldn't come up at all.
How recent is the openwrt master wrt upstream linux? I saw some patches from march that supposedly improve dbdc performance (and maybe unintentionally decrease it for non-dbdc?). I still can't get a handle on why the non-dbdc 5G phy would be slower than the dbdc phy. that doesn't make any sense at all. Would be nice if someone else could test and verify ghoffman's results.
my board di not have serial header pins. i've been soldering for almost 60 years and this was a tough one. i thnk the solder in the holes is a high-melting-point one, and the holes seem a little smaller than the usual pins with that header spacing. i bypassed these challenges by soldering wires horizontally and bringing a header up off those wires.
broly
93
lol drats! the fcc spec page showed it with serial, but i guess that was added. lol
https://fcc.io/KA2/IR3060A1
brb!
broly
94
my friend ("SWIM" and all that) just got one of these, he was really excited reading this conversation
and look, he's already set up!!
subtract ~25 minutes from this post for all the time he spent trying to take good pictures on his QNX phone.
no silver solder in the holes made this fun.
Real Men Use Portasol.
oh, he also says he's aware some highlanders would critique the two left joints due to the excess.
his brief answer:
"who cares, everything is easy when you're using Sn63/Pb37." he said he could have ribboned the excess, but wanted to be sure it was a solid joint.
and i guess 'twas, 'twas.
if they were grading him, he would have taken the extra minute to do that.
clearly he is lacking no shortage of confidence after using silver solder for so long. what an interesting personality he has!
what kind of soldering iron??
broly
96
p2kc weller. butane solder hahaha.
broly
97
it seems that this router (Dir-2640) actually does have the mac address for lan/wan separated by 3.
he says they're: xx:xx:xx:xx:xx:c5 and xx:xx:xx:xx:xx:c9
but no wlan leds either on the proprietary driver (apparently), so it's not just the mt76.
this all has to probably do with the different gpios.
openwrt reboot is definitely problematic on the dir3060. seems to need 2 reboots.
another wireless update: with a manually edited /etc/config/wireless, and then power cycles x2, it looks like the wireless comes up correctly, with 1x 2.4g and 3x 5g (virtual) radios. the wireless interface gui and settings are messed up, but this looks good, from the wireless channel analysis tab:
note 4 interfaces, with radio0 being 2-band -
edit: it's really 3 interfaces; radio0 (5G) seems to be a pointer to radio2 (5G)
kar200
99
Luci does not "yet" handle wireless interfaces with DBDC. I do recall having to power off - back on the router sometimes on mine.
Also did you happen to enable software+hardware offload in (network-firewall)? My ISP decided to upgrade my connection free of charge a couple of days ago (now 500/200) and without them enabled I cannot reach the max (on cable). but after enabling them it seems to do the trick and speed test shows full speed. It might be worth enabling that as well.
I also monitor cpu usage using htop just in case one of the cores hit the limit (and I know that it is a cpu bottleneck). That's how I remembered to enable the flow offload.
kar200
100
I flashed the factory firmware to check again the mac address. here they are (PS: This is my actual device and what I posted earlier was a dump from a different device)
wlan2g = F6:8C:XX:C8:F0:66
wlan5g = F4:8C:XX:A8:F0:66
LAN = F4:8C:XX:A8:F0:65
WAN = F4:8C:XX:A8:F0:68
EDIT: Just realised that 2g has C8 and 5G is A8 on the 4th hex number
