Firmware v2.49.0
bmcd 2.43.0 and BMC-UI 4.6.0; tpi 1.10.0 is unchanged.
Firmware v2.49.0, 14 November 2026 — the first image to carry bmcd 2.43.0, BMC-UI 4.6.0. Every item below is the repository's own changelog entry, taken apart; the whole entry is a click away under each one.
Answers https://github.com/excavador-turing/BMC-Firmware/issues/76: the console speed you set for a node now applies, and tools on the BMC can share a node's console with the web page.
A console speed for each node, in Power Control (BMC-UI 4.6.0)
Section titled “A console speed for each node, in Power Control (BMC-UI 4.6.0)”Each node's speed shows beside its name and Edit offers the usual rates, from 9600 up to 1500000. The default stays 115200. The Console page now tells you where the local console devices are.
Shared console devices /dev/ttyNode1 to /dev/ttyNode4 (bmcd 2.43.0)
Section titled “Shared console devices /dev/ttyNode1 to /dev/ttyNode4 (bmcd 2.43.0)”On the BMC, open /dev/ttyNodeN with microcom or picocom instead of /dev/ttyS*.
The whole entry
It shows what the web console shows, and what you type reaches the node, so the terminal and the web console no longer split a line between them. Nobody has to be reading: output nobody takes is dropped and never holds up the node. The speed of these devices is the node's setting; a speed chosen in the terminal program has no effect on them. /dev/ttyS* stays locked by the daemon, so a program opened on it still reports that it cannot lock the port.
A shared console device per node for tools on the BMC: /dev/ttyNode1 to
Section titled “A shared console device per node for tools on the BMC: /dev/ttyNode1 to”/dev/ttyNode4
bmcd 2.43.0
microcom or picocom on /dev/ttyS1 read the same UART as the daemon, so every character went to one reader or the other and a line came out split between the terminal and the web console.
The whole entry
The daemon now makes a pseudo-terminal for each node when it starts, with a stable symlink to it, and puts everything the node sends on it as well as in the web console. What is typed in a terminal on the device reaches the node, as typing in the web console does, so both show the same thing. Open /dev/ttyNodeN instead of /dev/ttyS*; the daemon's lock on the ttyS* ports stays, and means just that. A terminal program's own speed setting has no effect on it: the speed is the node's uart_baud setting. Nobody has to be reading: output that nobody takes is dropped and never holds up the node's other readers. One line per node in the log says where the device is.
A console speed for each node
Section titled “A console speed for each node”BMC-UI 4.6.0
Power Control now shows each node's serial speed beside its name and module name, and Edit offers it as a select of the common rates: 9600, 19200, 38400, 57600, 115200 (the default, shown as "115200 baud (default)" while nothing is stored), 230400, 460800, 921600, 1000000 and 1500000.
The whole entry
It saves through the same node_info call as the names, as uart_baud, and only when the speed was changed, so a node with none stored stays unset until somebody chooses one. A speed stored by hand that is not in the list is added to that node's select rather than replaced by the nearest listed one. The Fleet's Power Control tab is the same component and gets the field with it. In the demo, Save is refused and the toast says it is a demo; before, that call came back as a 200 and the form closed as if it had saved. bmcd has stored uart_baud for a long time but ignored it; a bmcd that applies it as the port's real speed is what makes the field take effect. Reported in BMC-Firmware#76.
The Console page names the local console devices
Section titled “The Console page names the local console devices”BMC-UI 4.6.0
Under "The line-oriented alternative", a note says that on the BMC itself /dev/ttyNodeN is the node's shared console (N is the node number), shows exactly what the page shows, and that a tool such as microcom or picocom can open it, with microcom -s 115200 /dev/ttyNode1 as the example.
The whole entry
It also says the speed is set in the node's settings, not by the tool. The devices come with the same bmcd change as the speed.
A gate for both
Section titled “A gate for both”BMC-UI 4.6.0
npm run baud-field-test drives the demo build at a desktop and a phone width: the default shows for every node, the select offers the ten rates in order, a chosen speed reaches the form for that node only, the demo's Save refuses and stays in Edit, and the Console note is there.
-
The console speed set for a node was stored and never used (bmcd 2.43.0). Every port opened at 115200 whatever the setting said, so a node at 1500000 showed garbage until a terminal program changed the speed.
The whole entry
The port now opens at the stored speed and follows a change at once, without restarting the daemon or disconnecting the web console. A speed that is not a standard rate (300 to 3000000) is refused and nothing in that request is saved.
-
tpi aboutshowedboardfollowed by garbage on a board whose information EEPROM was never programmed (bmcd 2.43.0). The board name, and the hardware revision (v31.31.63on an erased EEPROM), are now blank, as they are when the EEPROM cannot be read. -
The serial speed of a node was stored and never used.
uart_baudin a node's settings was kept and shown but every port was opened at 115200 whatever it said. bmcd 2.43.0The whole entry
The port now opens at the stored speed (115200 when there is none) and follows a change as soon as it is made, without restarting the daemon or disconnecting the web console. A value that is not a standard rate (300 to 3000000, which includes 1500000) is now refused with a
400that names the node and the rates, and nothing in that request is changed.GET /api/bmc/nodesshows it asuart_baud,nullfor the default. -
tpi aboutshowedboard ����on a board whose information EEPROM was never programmed. The serial number was already cleaned of padding; the model name was not, nor the hardware revision, which read asv31.31.63on an erased EEPROM. bmcd 2.43.0The whole entry
Both are now empty, as they are when the EEPROM cannot be read, and text that is noise in the middle is not shown either.
Every release of the firmware · the roadmap · follow by feed