thinking...
This commit is contained in:
parent
7750904cb1
commit
2cd51e474e
GPG Key ID:
6F42FE1A481818C2
@ -10,12 +10,12 @@
|
||||
* - 8MB Flash (Macronix MX25L6433F SPI NOR 8MB // dual bootloader, ART, etc.)
|
||||
* - UART
|
||||
* - USB
|
||||
* - wifi (QCA9984, QCA4019)
|
||||
*
|
||||
* What is NOT working?
|
||||
*
|
||||
* - 128MB nand (Micron MT29F1G08ABADA NAND 128MB // 2x kernel, 2x rootfs, 1x rootfs_data)
|
||||
* - switches (QCA8075, QCA8335)
|
||||
* - wifi (QCA9984, * wifi I did get working, I just have to copy the DTS code + board.bin binary back)
|
||||
*
|
||||
* WARNING: ENSURE BOARD.BIN AND CALDATA IS NOT COMMITTED TO GIT REPOSITORY (INTELLECTUAL PROPERTY CONCERNS)
|
||||
*
|
||||
@ -394,3 +394,91 @@
|
||||
|
||||
|
||||
/* CHANGE 3: Add wifi from existing/old DTS config attempts, and copy board*.bin (CALDATA) files */
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
/*
|
||||
* CRITICAL FIX: Explicitly define gmac0 and gmac1 inside &edma
|
||||
* because the base dtsi in your kernel tree lacks the global labels.
|
||||
*
|
||||
* human note: the above AI explanation seems hallucinatory and wrong.
|
||||
*
|
||||
*/
|
||||
/* disabled temporarily
|
||||
&gmac {
|
||||
status = "okay";
|
||||
|
||||
// GMAC0: SGMII -> QCA8337 (Switch 0)
|
||||
gmac0: ethernet@98000 {
|
||||
status = "okay";
|
||||
phy-mode = "sgmii";
|
||||
qcom,id = <1>;
|
||||
nvmem-cells = <&macaddr_art_0>;
|
||||
nvmem-cell-names = "mac-address";
|
||||
fixed-link { speed = <1000>; full-duplex; pause; };
|
||||
};
|
||||
|
||||
// GMAC1: RGMII -> QCA8075 (Switch 1)
|
||||
gmac1: ethernet@99000 {
|
||||
status = "okay";
|
||||
phy-mode = "rgmii";
|
||||
qcom,id = <2>;
|
||||
tx-internal-delay-ps = <1000>;
|
||||
rx-internal-delay-ps = <1000>;
|
||||
fixed-link { speed = <1000>; full-duplex; };
|
||||
};
|
||||
};
|
||||
/* disabled temporarily
|
||||
|
||||
/*
|
||||
|
||||
|
||||
*THINKING HARD*
|
||||
|
||||
8 LAN + 2 WAN (DSA, two switches QCA8337, QCA8075)
|
||||
This is definitely wrong, the manufacturer specifies gmac0, gmac1, gmac2 none of which are nested under eachother.
|
||||
Unsure as to their role but these are eth0-2
|
||||
QCA8075 being rgmii and QCA8337 being sgmii is confirmed, I know that.
|
||||
Manufacturer shows one of the GMAC with minimal config, only a mac of all 0's and a vlan_tag.
|
||||
The two switches are possibly GMAC0 and GMAC2 due to the presence of more configuration on two of the three gmac in manufacturer config.
|
||||
This leaves me the question, WTF is gmac1 for?
|
||||
Who cares. We need to replicate the manufacturer. They have gmac0, gmac1, gmac2 inside of an edma node. I don't know why.
|
||||
Do we just define gmac0, gmac1, gmac2 outside at top level, or do they need to be nested under edma node?
|
||||
Also one switch is "switch" and the other is "mdio". They both have 4 phy's in the manufacturer DTS. 4+4 = 8 - is that our 8 LAN? Where are our WAN? Does that mean our WAN are the gmac0 and gmac2? What is gmac1 then?
|
||||
Either gmac0 and gmac2 are switches, or they are WAN ports, or something.
|
||||
Although other routers have wan/wan6 in openwrt showing on the same eth device (in the AC1750 case, eth0.2)
|
||||
If wan1 and wan 2 are one device, this leads me to believe the other device, is the SWITCH?
|
||||
So in my case, wan1 and wan2 (its dual wan) would be on one eth?
|
||||
Or eth0 wan 1 and eth2 wan2 ? <--- this...?
|
||||
Then what eth1 ? <--- soooo....? (remember minimal config, vlan tag differs from the others)
|
||||
THE SWITCHES? BOTH SWITCHES ON ETH1? ETH0 = WAN1 and ETH2 = WAN2? <--- it must be,... does it sound right?
|
||||
|
||||
|
||||
** My best guess **
|
||||
qcom-ipq4019.dtsi has 5 ports per switch, daisy chained(?) - as per the manufacturer config they have 4 per switch. qcom-ipq4019.dtsi ain't us. We must emulate manufacturer here, switch config highly custom.
|
||||
|
||||
Wans being physical ports and not on switches is what I would do, and then have LANs on switches. Explains why manufacturer has 4 ports per switch in their config (8 LAN ports).
|
||||
Making gmac0/gmac2 aka eth0/eth2 the wan ports, leaving gmac1 aka eth1 for the two switches (QCA8337, QCA8075).
|
||||
|
||||
- eth0/gmac0 - wan 1 (physical, explains added config)
|
||||
- eth1/gmac1 - QCA8337, QCA8075 (virtual-ish? explains less config? patches the two switches together, enables patching switches to wan 1 or wan 2)
|
||||
- eth2/gmac2 - wan 2 (physical, explains added config)
|
||||
|
||||
I think this is it. It kinda makes sense. Even the config in the DTS, eth1 being minimal vs eth0 and eth2 being more config - speed, duplex, poll required, etc.
|
||||
|
||||
Open WRT then takes information coming/going to/from wan1/lan1-8/wan2 aka eth0/eth1/eth2 aka gmac0/gmac1/gmac2 respectively after applying firewall rules, then patches them together allowing the data transfer between them.
|
||||
|
||||
- This indicates a high security model, which would fit for a SMB router.
|
||||
- It also provides physical/visual separation of our concerns (wan1, wan2, and lan) for security purposes.
|
||||
- By default these are disconnected and physically separated.
|
||||
|
||||
And it just makes kinda sense no?
|
||||
|
||||
|
||||
*/
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user