Add ENC424J600 support #134

occheung · 2021-01-20T06:33:13Z

This PR adds support to Booster with ENC424J600 as the Ethernet device.
This is implemented by adding feature gates to conditionally compile W5500 and ENC424J600 codes. (i.e. phy_w5500 & phy_enc424j600)
With phy_enc424j600 enabled, Booster uses the factory preprogrammed MAC address of ENC424J600 and ignore the one stored in EEPROM.

Testing

Booster with ENC424J600 can send & receive MQTT messages properly, when using firmware built with phy_enc424j600 feature.
The expected MAC address is found in network trace and after invoking USB read mac command.
The firmware can be compiled with phy_w5500 feature enabled.

jordens

Thanks for this!
I think there is some room for structural improvements that will help with maintaining this code.
And we'll look into moving to raw sockets and smoltcp for the w5500 as well.

jordens · 2021-01-20T09:07:44Z

Cargo.toml

+git = "https://github.com/smoltcp-rs/smoltcp.git"
+rev = "2080152b"


I think this can be blank/crate-versioned since you patch it below.

83380a9 changes this reference to version 0.6.0.

Can't this be smoltcp 0.7.0 and the patch removed?

Should be possible after bumping the smoltcp version in enc424j600 crate.

enc424j600 should be able to use smoltcp 0.7.0 safely, let's bump that and keep smoltcp up-to-date.

src/settings/mac.rs

jordens · 2021-01-20T09:13:35Z

src/main.rs

+                    match enc424j600.init_dev(&mut delay) {
+                        Ok(_) => {}
+                        Err(_) => {
+                            panic!("ENC424J600 PHY initialization failed");
+                        }
+                    }
+
+                    enc424j600.read_from_mac(&mut eth_mac_addr).unwrap();
+
+                    // Init Rx/Tx buffers
+                    enc424j600.init_rxbuf().unwrap();
+                    enc424j600.init_txbuf().unwrap();
+
+                    let eth_iface = unsafe {
+                        let device = enc424j600::smoltcp_phy::SmoltcpDevice::new(enc424j600);
+
+                        // TODO: Restore IP config from EEPROM
+                        NET_STORE.ip_addrs[0] = {
+                            let ip = settings.ip().octets();
+                            let subnet = settings.subnet().octets();
+                            net::wire::IpCidr::new(
+                                net::wire::IpAddress::from(net::wire::Ipv4Address::from_bytes(&ip)),
+                                net::wire::IpAddress::from(net::wire::Ipv4Address::from_bytes(
+                                    &subnet,
+                                ))
+                                .to_prefix_len()
+                                .unwrap(),
+                            )
+                        };
+
+                        let routes = {
+                            // TODO: Restore gateway config from EEPROM
+                            let gateway =
+                                net::wire::Ipv4Address::from_bytes(&settings.gateway().octets());
+
+                            let mut routes =
+                                net::iface::Routes::new(&mut NET_STORE.routes_cache[..]);
+                            routes.add_default_ipv4_route(gateway).unwrap();
+                            routes
+                        };
+
+                        let neighbor_cache =
+                            net::iface::NeighborCache::new(&mut NET_STORE.neighbor_cache[..]);
+
+                        net::iface::EthernetInterfaceBuilder::new(device)
+                            .ethernet_addr(net::wire::EthernetAddress(eth_mac_addr))
+                            .neighbor_cache(neighbor_cache)
+                            .ip_addrs(&mut NET_STORE.ip_addrs[..])
+                            .routes(routes)
+                            .finalize()
+                    };
+
+                    let socket_set = unsafe {
+                        let mut sockets =
+                            net::socket::SocketSet::new(&mut NET_STORE.socket_storage[..]);
+
+                        let mut tcp_socket = {
+                            let tx_buffer =
+                                net::socket::TcpSocketBuffer::new(&mut DATA_STORE.tx_storage[..]);
+                            let rx_buffer =
+                                net::socket::TcpSocketBuffer::new(&mut DATA_STORE.rx_storage[..]);
+
+                            net::socket::TcpSocket::new(rx_buffer, tx_buffer)
+                        };
+                        tcp_socket.set_keep_alive(Some(net::time::Duration::from_millis(1000)));
+
+                        let _handle = sockets.add(tcp_socket);
+                        sockets
+                    };
+
+                    NetworkStack::new(eth_iface, socket_set)
+                }


Can this go into a different module together with the static's above (https://github.com/quartiq/booster/pull/134/files#diff-42cb6807ad74b3e201c5a7ca98b911c5fa08380e942be6e4ac5807f8377f87fcR167-R198)?

I think we have some code in quartiq/stabilizer#230 that looks like it might be similar.

@occheung Could be like this I think: https://github.com/quartiq/stabilizer/blob/573189bdd96b002d11dc152118d69599cb3dcf2d/src/hardware/configuration.rs#L435-L535

15f0964. Does it do what you want?

@jordens Stabilizer has moved all hardware init code to a single configuration.rs such that main.rs only needs to call hardware::setup() to return a set of ready-to-use hardware interfaces. Should Booster do the same, and if so, would it be better to do that, along with 15f0964 (splitting smoltcp init from main.rs), in a separate PR altogether?

jordens · 2021-01-20T09:14:29Z

src/smoltcp_nal.rs

@@ -0,0 +1,251 @@
+// Ripped off from minimq example


This should really be its own crate or in enc424j600 (like it is in w5500) I think. And the code provenience/copyright needs to be clearly explained.

We will add the NAL support to enc424j600.

jordens · 2021-01-21T09:32:49Z

src/settings/global_settings.rs

+    /// Override MAC address during initialization
+    pub fn set_mac(mut self, mac: [u8; 6]) -> Self {
+        // MAC address is designed to be unchanged in EEPROM
+        self.eui48.copy_from_slice(&mac);
+        self
+    }


This is broken I think. It'll guarantee that the checksum is always invalid.

The MAC isn't actually saved to EEPROM. Only BoosterMainBoardData is. The MAC is always read form the EUI-48 section, so there's no need to duplicate that data into EEPROM writable region.

Edit: Post-discussion with Robert revealed that this statement is incorrect - the MAC is actually used in salting the CRC32 calculation.

jordens · 2021-01-21T09:33:46Z

src/settings/global_settings.rs

+use w5500::MacAddress;
+
+#[cfg(feature = "phy_enc424j600")]
+use super::mac::MacAddress;


MacAddress should either be in the enc424j600 crate (like it is in the w5500 crate). Or it should be in smoltcp-nal, or in smoltcp. No need to copy and paste it here.

a3db62f, using 'EthernetAddress' from smoltcp as 'MacAddress'.

jordens · 2021-01-21T09:35:47Z

src/main.rs

+
+            #[cfg(feature = "phy_enc424j600")]
+            {
+                SerialTerminal::new(usb_device, usb_serial, settings.set_mac(eth_mac_addr))


I don't get the reason for the added code.
Setting the MAC doesn't have much to do with creating the serial terminal.

This is to relay MAC address from the one in enc424j600 to the serial terminal, we don't use the one on the eeprom.
Just for the read mac command.

Are you sure this doesn't factor into the checksum computation on settings save? Did you test that?

src/settings/mac.rs

jordens · 2021-01-21T09:43:57Z

Cargo.toml

+git = "https://github.com/smoltcp-rs/smoltcp.git"
+rev = "2080152b"


Can't this be smoltcp 0.7.0 and the patch removed?

ryan-summers · 2021-01-22T13:56:29Z

Cargo.toml

+
+# Coerce ENC424J600 to use the latest smoltcp to accomodate minimq
+[patch.crates-io]
+smoltcp = { git = "https://github.com/smoltcp-rs/smoltcp.git", rev = "2080152b" }


We should capture the revision in the Cargo.lock file - is there a specific reason why this commit is important? If so, we need to add documentation around it.

It will be bumped to 0.7.0 after the NAL is incorporated into the enc424j600 crate.

ryan-summers · 2021-01-22T13:57:22Z

src/enc424j600_config.rs

+    use enc424j600::EthController;
+    use smoltcp as net;
+
+    match enc424j600.init_dev(&mut delay) {


simplify to enc424j600.init_dev(&mut delay).expect("PHY initialization failed");

Fixed in e66d916.

ryan-summers · 2021-01-22T13:59:43Z

src/enc424j600_config.rs

+    let eth_iface = unsafe {
+        let device = enc424j600::smoltcp_phy::SmoltcpDevice::new(enc424j600);
+
+        // TODO: Restore IP config from EEPROM


Is this TODO still relevant?

Removed in e66d916.

ryan-summers · 2021-01-22T14:00:08Z

src/enc424j600_config.rs

+        };
+
+        let routes = {
+            // TODO: Restore gateway config from EEPROM


Is this TODO still relevant?

Removed in e66d916.

ryan-summers · 2021-01-22T14:01:26Z

src/enc424j600_config.rs

+    }
+
+    let mut eth_mac_addr: [u8; 6] = [0; 6];
+    enc424j600.read_from_mac(&mut eth_mac_addr).unwrap();


I'm honestly not a fan of using different MACs for the different PHYs. The MAC is derived from a EUI-48 identifier IC on the booster board, so in the case of the ENC424J600, we technically have two MAC addresses that we can use. Does the ENC424J600 allow us to override the MAC address with our own? That would simplify the system-level logic to always use the EUI48 identifier, as that is also used in identifying booster.

This diff actually becomes a lot smaller and keeps the code flow simpler if the ENC424J600 is updated to utilize boosters internal MAC address (while ignoring the ENC424J600's address).

The datasheet says this is possible:

This allows us to keep the change minimal, which increases maintainability in the future.

Ok. The MAC address overriding will be supported in enc424j600 crate.
We will use the EUI-48 ID from the eeprom.

@ryan-summers We would need to review some PRs on the ENC424J600 driver repo and do some cleanup first, before we can finalise this PR here. Thanks!

ryan-summers · 2021-01-22T14:06:48Z

src/settings/global_settings.rs

+    /// Override MAC address during initialization
+    pub fn set_mac(mut self, mac: [u8; 6]) -> Self {
+        // MAC address is designed to be unchanged in EEPROM
+        self.eui48.copy_from_slice(&mac);
+        self
+    }


The MAC isn't actually saved to EEPROM. Only BoosterMainBoardData is. The MAC is always read form the EUI-48 section, so there's no need to duplicate that data into EEPROM writable region.

Edit: Post-discussion with Robert revealed that this statement is incorrect - the MAC is actually used in salting the CRC32 calculation.

ryan-summers · 2021-01-22T14:09:05Z

src/settings/global_settings.rs

+        #[cfg(feature = "phy_enc424j600")]
+        {
+            MacAddress::from_bytes(&self.eui48)
+        }


The brackets are unnecessary here.

Fixed in 3869217.

ryan-summers · 2021-01-22T14:14:07Z

Cargo.toml

@@ -36,6 +36,20 @@ features = ["stm32f407", "rt", "usb_fs"]
 [dependencies.w5500]
 git = "https://github.com/quartiq/w5500"
 branch = "feature/tcp-nal"
+optional = true
+
+[dependencies.enc424j600]


Please update the .github/workflow/ci.yml file to build both PHYs.

Updated in 4ae6e82

HarryMakes · 2021-01-26T09:01:48Z

Cargo.toml

@@ -63,6 +77,13 @@ path = "tca9548"

 [features]
 unstable = []
+phy_enc424j600 = [ "enc424j600", "smoltcp" ]
+phy_w5500 = [ "w5500" ]
+default = [ "phy_w5500" ]


Looks like removing this line (i.e. not defining default features entirely) would avoid situations like running cargo build --features=phy_enc424j600 but forgetting to use --no-default-features. I would like to hear the opinion from others on this, thanks!

HarryMakes · 2021-03-19T02:39:45Z

Just an update: Since @occheung is on leave, I've been handling the addition of ENC424J600. I have an active PR on the ENC424J600 repo, so after that has been merged I'll probably open a new PR here on the Booster firmware repo with updated function calls to the driver/NAL. That new PR will have my own commits on top of all commits @occheung has pushed here.

HarryMakes · 2021-04-12T05:29:13Z

Replaced with #154. Closing.

occheung added 7 commits January 19, 2021 13:03

cargo: include enc424j600 dependencies

afdab5f

main: feature gate different eth device

39e0b55

nal: implement enc424j600 variant

e845f56

nal: fix unnecessary w5500 dependencies

befe70a

settings: override eeprom MAC

20efdcc

use cargo fmt

0d90470

cargo: add w5500 as default

6e13f4f

jordens reviewed Jan 20, 2021

View reviewed changes

occheung added 2 commits January 21, 2021 09:29

cargo: simplify smoltcp dependency

83380a9

settings/mac: remove redundant code

83109c3

jordens requested changes Jan 21, 2021

View reviewed changes

occheung added 2 commits January 22, 2021 11:43

mac: use smoltcp

a3db62f

eth_cfg: split off from main

15f0964

ryan-summers requested changes Jan 22, 2021

View reviewed changes

ryan-summers reviewed Jan 22, 2021

View reviewed changes

HarryMakes reviewed Jan 26, 2021

View reviewed changes

occheung added 3 commits January 28, 2021 17:44

nal: migrated to enc424j600

4ae6e82

phy_config: minor fix

e66d916

clean up

3869217

HarryMakes mentioned this pull request Apr 12, 2021

[outdated] Add support for ENC424J600 with bumped minimq #154

Merged

HarryMakes closed this Apr 12, 2021

jordens deleted the feature/enc424j600 branch October 18, 2022 10:26

		git = "https://github.com/smoltcp-rs/smoltcp.git"
		rev = "2080152b"

Add ENC424J600 support #134

Add ENC424J600 support #134

Conversation

occheung commented Jan 20, 2021

Testing

jordens left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

occheung Jan 22, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

HarryMakes Jan 21, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jordens Jan 20, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ryan-summers Jan 22, 2021 • edited Loading

Choose a reason for hiding this comment

jordens Jan 21, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ryan-summers Jan 22, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ryan-summers Jan 22, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

HarryMakes commented Mar 19, 2021

HarryMakes commented Apr 12, 2021

occheung Jan 22, 2021 •

edited

Loading

HarryMakes Jan 21, 2021 •

edited

Loading

jordens Jan 20, 2021 •

edited

Loading

ryan-summers Jan 22, 2021 •

edited

Loading

jordens Jan 21, 2021 •

edited

Loading

ryan-summers Jan 22, 2021 •

edited

Loading

ryan-summers Jan 22, 2021 •

edited

Loading