BMC/OpenBmc/openbmc.git, branch master

linux-nuvoton: Backport AST2700 SoC support

2026-11-01T22:54:31+00:00

Billy Tsai (6): dt-bindings: pinctrl: Add aspeed,ast2700-soc0-pinctrl dt-bindings: mfd: aspeed,ast2x00-scu: Describe AST2700 SCU0 pinctrl: aspeed: Add AST2700 SoC0 support dt-bindings: mfd: aspeed,ast2x00-scu: Support AST2700 SoC1 pinctrl dt-bindings: pinctrl: Add aspeed,ast2700-soc1-pinctrl pinctrl: aspeed: Add AST2700 SoC1 support Chin-Ting Kuo (13): spi: aspeed: Update clock selection strategy spi: aspeed: Improve timing calibration algorithm for AST2600 platform spi: aspeed: Force default address decoding range assignment for each CS spi: aspeed: Centralize address decoding region management spi: aspeed: Add per-platform adjust_window callback for decoding range spi: aspeed: Only map necessary address window region spi: dt-bindings: aspeed,ast2600-fmc: Add AST2700 SoC support spi: aspeed: Enable Quad SPI mode for page program spi: aspeed: Use phys_addr_t for bus addresses to support 64-bit platforms spi: aspeed: Add support for the AST2700 SPI controller dt-bindings: watchdog: aspeed,ast2400-wdt: Add support for AST2700 watchdog: aspeed: Support variable number of reset mask registers watchdog: aspeed: Add support for AST2700 platform Michael Pesa (1): irqchip/ast2700-intc: Disable all interrupt merge banks on probe Ryan Chen (13): clk: aspeed: Move the existing ASPEED clk drivers into aspeed subdirectory. MAINTAINERS: Add entry for ASPEED clock drivers. clk: aspeed: add AST2700 clock driver dt-bindings: interrupt-controller: Describe AST2700-A2 hardware instead of A0 irqchip/ast2700-intc: Add AST2700-A2 support irqchip/ast2700-intc: Add KUnit tests for route resolution irqchip/aspeed-intc: Remove AST2700-A0 support irqchip/ast2700-intc: Avoid allocating in the irq_domain activate() callback dt-bindings: arm: aspeed: Add AST2700 board compatible arm64: Kconfig: Add ASPEED SoC family Kconfig support arm64: dts: aspeed: Add initial AST27xx SoC device tree arm64: configs: Update defconfig for AST2700 platform support arm64: dts: aspeed: Fix duplicate pinctrl labels and address scheme bui duc phuc (1): watchdog: aspeed: Propagate errors from optional IRQ lookup Change-Id: Ie57f80fe94eb921d26f25aa6c7a6de19bcaf9616 Signed-off-by: Andrew Jeffery

linux-aspeed: Backport AST2700 SoC support

2026-11-01T22:54:31+00:00

Billy Tsai (6): dt-bindings: pinctrl: Add aspeed,ast2700-soc0-pinctrl dt-bindings: mfd: aspeed,ast2x00-scu: Describe AST2700 SCU0 pinctrl: aspeed: Add AST2700 SoC0 support dt-bindings: mfd: aspeed,ast2x00-scu: Support AST2700 SoC1 pinctrl dt-bindings: pinctrl: Add aspeed,ast2700-soc1-pinctrl pinctrl: aspeed: Add AST2700 SoC1 support Chin-Ting Kuo (13): spi: aspeed: Update clock selection strategy spi: aspeed: Improve timing calibration algorithm for AST2600 platform spi: aspeed: Force default address decoding range assignment for each CS spi: aspeed: Centralize address decoding region management spi: aspeed: Add per-platform adjust_window callback for decoding range spi: aspeed: Only map necessary address window region spi: dt-bindings: aspeed,ast2600-fmc: Add AST2700 SoC support spi: aspeed: Enable Quad SPI mode for page program spi: aspeed: Use phys_addr_t for bus addresses to support 64-bit platforms spi: aspeed: Add support for the AST2700 SPI controller dt-bindings: watchdog: aspeed,ast2400-wdt: Add support for AST2700 watchdog: aspeed: Support variable number of reset mask registers watchdog: aspeed: Add support for AST2700 platform Michael Pesa (1): irqchip/ast2700-intc: Disable all interrupt merge banks on probe Ryan Chen (13): clk: aspeed: Move the existing ASPEED clk drivers into aspeed subdirectory. MAINTAINERS: Add entry for ASPEED clock drivers. clk: aspeed: add AST2700 clock driver dt-bindings: interrupt-controller: Describe AST2700-A2 hardware instead of A0 irqchip/ast2700-intc: Add AST2700-A2 support irqchip/ast2700-intc: Add KUnit tests for route resolution irqchip/aspeed-intc: Remove AST2700-A0 support irqchip/ast2700-intc: Avoid allocating in the irq_domain activate() callback dt-bindings: arm: aspeed: Add AST2700 board compatible arm64: Kconfig: Add ASPEED SoC family Kconfig support arm64: dts: aspeed: Add initial AST27xx SoC device tree arm64: configs: Update defconfig for AST2700 platform support arm64: dts: aspeed: Fix duplicate pinctrl labels and address scheme bui duc phuc (1): watchdog: aspeed: Propagate errors from optional IRQ lookup Change-Id: Id57ca31e9e4412596d368fe61bc9814ab3fdd006 Signed-off-by: Andrew Jeffery

meta-google: google-usb-network: Install the static gadget network config

2026-10-31T16:20:10+00:00

The static gadgets are attached once systemd-networkd runs, and usb_network.sh then writes their network config to /run and has networkd reload it. On a busy boot, that reload is most of the ~8s between networkd starting and the gadget link coming up. The config of a static gadget is known at build time, so generate it with the rest of the gadget config and install it with the image, where networkd loads it when it starts. Add a --network-config option to usb_network.sh to skip writing and removing the runtime config, and use it for the static gadgets. The bare metal setups are unaffected: the gadgets they keep from being created leave the installed config unused, and the server3 ++-bmc-cn.network override that takes a host link down still sorts before the installed +-bmc-cn.network. Tested: Built the recipe for sunfish: the generated gusb0 .network and .link and the bloomusb .network are identical to what usb_network.sh writes at runtime on a sunfish tray. Booted astoria, izumi and imperia BMCs with the gusb0 config installed and the gadget started with --network-config 0: the gadget no longer triggers a networkd reload, and comes up 1-6s after networkd starts, from 2-8s. Change-Id: I3e94f2b72289a740fd6367a26ab7e93ce34b0020 Signed-off-by: William A. Kennington III

meta-google: google-usb-network: Let networkd bring up static gadgets

2026-10-31T16:20:10+00:00

The static gadgets are attached at the start of boot, well before systemd-networkd runs. The host or peer BMC on the other side sees carrier right away (as soon as it enumerates an EEM gadget, or once usb_network.sh sets an NCM or ECM gadget up) and sends its router solicitations into a link nothing answers on yet. It then has to wait for the unsolicited RAs once networkd starts, 30-45s later. Add a --link-up option to usb_network.sh so the caller can leave bringing the link up to systemd-networkd, which already does it from the +-bmc-.network the script writes before binding the UDC. Use it for the static gadgets, and order them after systemd-networkd.service so the gadget is only attached once networkd manages the link and answers the peer right away. This also keeps the server3 host links down until networkd applies their bare metal ActivationPolicy, rather than the script bringing them up first. The dynamic gadgets and usb_network.service keep bringing the link up themselves. Tested: Ran usb_network.sh with a mocked ip and gbmc-net-lib, and checked it only skips `ip link set up` with --link-up 0 and rejects values other than 0 and 1. Change-Id: I8a6e07fbc474bf5cd8758c8acc5b009313585977 Signed-off-by: William A. Kennington III

meta-google: gbmc-bridge: Restart a port's RA solicitation

2026-10-31T16:19:59+00:00

A gbmcbr port can lose and regain its carrier while the gbmc-br-port-rs@ run for its previous carrier is still going. That happens on a ghostlite tray while its head BMC boots: the head's USB host enumerates the tray gadget well before its networkd runs, the tray resets its USB link because the head stopped answering, and the second carrier up comes ~7s into the 10s solicitation. Starting the already active oneshot did nothing, so the tray never solicited for the link it ended up with. Restart the port's solicitation instead, so it always follows the latest carrier up. The gbmcbr instance keeps coalescing. Tested: Rebooted a head BMC with its ghostlite trays patched. On the tray whose link came back twice, the second carrier up restarted gbmc-br-port-rs@gusb0, and the solicitation ran for the link it ended up with. Unpatched, the second carrier up was dropped. Change-Id: Iee53fda91596678153724a51c935f26ff14234f3 Signed-off-by: William A. Kennington III

meta-google: gbmc-systemd-config: Keep the plain USB net names on bus 1

2026-10-31T16:18:07+00:00

The bus number suffix is only needed to keep interfaces on different USB host controllers apart. Most BMCs only have one, bus 1, and scripts and configs, such as the TPU tray link and NC-SI sideband interfaces, refer to its interfaces by their predictable names (e.g. enu1u1u3), which the suffix renamed. Only append the suffix for the other buses, names stay unique since a USB port path never contains a 'b'. Tested: udevadm test on a single controller BMC, ID_NET_NAME_PATH stays enu1u1u1u3 on bus 1, and the same rule matching bus 1 still gives enu1u1u1u3b1. Change-Id: I10d04c277458b7590d708c4481c69b55fa6ae6ac Signed-off-by: William A. Kennington III

meta-facebook: request maintainers access

2026-10-31T01:43:49+00:00

Add myself to the meta-facebook owners so I can review and approve changes to the Meta platform layers, such as meta-anacapa. Change-Id: I40544e04b008e46c655871d976cfb0435a93cd09 Signed-off-by: Pete O_o

meta-intel-openbmc: add mctp interface initialisation

2026-10-31T00:39:06+00:00

Added mctp interface initialisation service for Intel platforms. Change-Id: Id164a2f35885ffdf459abee35250ea154cce00ee Signed-off-by: Nidhin MS Signed-off-by: Mayank Metha

meta-intel-openbmc: Add mctpd config

2026-10-31T00:38:42+00:00

Add mctp daemon config required for Intel platforms. Change-Id: I4b70077603f128a40e17ba674d109072276d1b68 Signed-off-by: Mayank Metha

meta-facebook: drop gzip FIT kernel compression override

2026-10-30T20:21:15+00:00

The gzip override in rainiera7, ventura2a7 and yosemite5a7 was a workaround for fit-image.bbclass lacking lzma support. Now that 82d4dac8c2 ("meta-phosphor: fit-image.bbclass: add lzma compression") has added it, drop the override and fall back to the ast2700.inc default (lzma). This shrinks the compressed kernel in the FIT image by about 2.3 MiB. Change-Id: I070e660211ccda2ef28286bac53d03ed500e9813 Signed-off-by: Cosmo Chou