Repository navigation
Conversation
keks2293
force-pushed
the
initramfs-adsp
branch
from
October 11, 2026 12:16
bc2eef9 to
e618a1d
Compare
Disk image
Downloads as a zip holding |
keks2293
force-pushed
the
initramfs-adsp
branch
from
October 11, 2026 15:59
e618a1d to
72aad4f
Compare
Member
|
Wouldn't step 7 from the official Fedora WoA wiki be enough? Including a file like this instead of creating a whole module for them https://fedoraproject.org/wiki/Snapdragon_WoA_Laptop_Install |
Booting Armada from a USB-C stick (ABL "linux usb") hangs in dracut-initqueue: the port never sources VBUS, so neither the stick nor a keyboard ever enumerates. VBUS is gated behind the ADSP: without it pmic-glink never reports the charger PD service as up, ucsi_glink never reaches ucsi_register(), no connector shows up under /sys/class/typec, no usb role switch is registered and the port stays electrically invisible. On a full boot the ADSP comes up from the rootfs; only the initramfs lacks the pieces: - dracut's aarch64 module whitelist covers drivers/rpmsg and drivers/usb/* but not drivers/remoteproc; - dracut only pulls firmware named in a module's MODULE_FIRMWARE metadata, while the ADSP firmware name comes from the DT. None of this needs Armada-specific kernel code: qcom_q6v5_pas, pmic_glink and ucsi_glink are all upstream and the board DT already sets compatible = "qcom,sm8550-adsp-pas" plus firmware-name. Add a dracut module (91armada-adsp, explicit --add like the other armada modules) that installs drivers/remoteproc and the ADSP images listed in usr/share/armada/initramfs-adsp-firmware.list, and make 55-generate-initramfs.sh fail the build when either is missing. The kernel picks the right image itself at boot (firmware-name comes from the board DT, request_firmware() resolves it), so the list is globs over every board's ADSP set rather than one entry per board: one generic image then boots any board whose firmware the repo vendors. Raw cost is ~210 MB across boards; the initramfs is compressed, so the image grows by far less. Sm8250 boards reference qcom/sm8250/adsp.mbn, which the repo does not vendor, so they stay out until it is added. Also fix the initramfs checks that failed this change's first CI run with both ADSP images "missing" although they were bundled: under `set -o pipefail`, `lsinitrd | grep -q` exits at the first match, lsinitrd dies of SIGPIPE and the pipeline reports failure. Take one lsinitrd listing up front and match against it instead, and check firmware sources before dracut runs (dracut ignores module_install's exit code, so a broken list would otherwise surface only here).
… port The RP6 connector's mode-switch/orientation-switch provider is the usb0-sbu-mux node (compatible "gpio-sbu-mux"), but that driver was left as a module while QCOM_PMIC_GLINK, TYPEC_UCSI and UCSI_PMIC_GLINK are all built in. pmic_glink therefore probes and UCSI tries to register the typec port before udev has loaded gpio-sbu-mux, fwnode_typec_switch_get() returns -EPROBE_DEFER, and after the ~10s UCSI retry window the driver gives up for good: ucsi_glink.pmic_glink_ucsi pmic_glink.ucsi.0: PPM init failed, stop trying With no typec port there is no USB role switch, so the port never enters host mode / supplies VBUS and a root filesystem on USB is never found (dracut-initqueue waits forever). Promote TYPEC_MUX_GPIO_SBU to =y so the provider is present from boot, matching ROCKNIX's SM8550 config.
keks2293
force-pushed
the
initramfs-adsp
branch
from
October 12, 2026 05:53
21a8ed4 to
0d05baf
Compare
… dwc3 The SM8550 dwc3 reports GHWPARAMS0_MODE=HOST, so dwc3 forces host mode and never registers the usb_role_switch advertised by the usb-role-switch property inherited from upstream sm8550.dtsi. UCSI's ucsi_register_port() then gets -EPROBE_DEFER forever from fwnode_usb_role_switch_get() and gives up at 20s with "PPM init failed, stop trying": the typec port and VBUS stay unmanaged, and USB host boot hangs waiting for the boot drive. Drop the property with /delete-property/ in the two shared qcs8550 board files (qcs8550-ayn-common.dtsi and qcs8550-ayaneo-pocket-common.dtsi) instead of patching the upstream dtsi, so every qcs8550 board is covered, UCSI skips the role-switch lookup and registers the port normally (usb_role_switch_set_role() is NULL-safe).
keks2293
force-pushed
the
initramfs-adsp
branch
from
October 12, 2026 09:40
0d05baf to
c3aa050
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Booting Armada from a USB-C stick (ABL "linux usb") hangs in dracut-initqueue: the port never sources VBUS, so neither the stick nor a keyboard ever enumerates.
VBUS is gated behind the ADSP: without it pmic-glink never reports the charger PD service as up, ucsi_glink never reaches ucsi_register(), no connector shows up under /sys/class/typec, no usb role switch is registered and the port stays electrically invisible. On a full boot the ADSP comes up from the rootfs; only the initramfs lacks the pieces:
None of this needs Armada-specific kernel code: qcom_q6v5_pas, pmic_glink and ucsi_glink are all upstream and the board DT already sets compatible = "qcom,sm8550-adsp-pas" plus firmware-name.
Add a dracut module (91armada-adsp, explicit --add like the other armada modules) that installs drivers/remoteproc and the ADSP images listed in usr/share/armada/initramfs-adsp-firmware.list, and make 55-generate-initramfs.sh fail the build when either is missing.
No per-board selection is needed. The kernel picks the right image itself at boot: firmware-name comes from the board DT and request_firmware() resolves it. So the list is globs over every board's ADSP set (sm8550/sm8650/sm8750, plus the Ayaneo Pocket MDT+segment set) and one generic image then boots any board whose firmware the repo vendors. Raw cost is ~210 MB across boards; the initramfs is compressed, so the image grows by far less. Sm8250 boards reference qcom/sm8250/adsp.mbn, which the repo does not vendor, so they stay out until it is added.
The first CI run of this change failed with both ADSP images "missing" although they were bundled: under
set -o pipefail,lsinitrd | grep -qexits at the first match, lsinitrd dies of SIGPIPE and the pipeline reports failure. Fixed by taking one lsinitrd listing up front and matching against it, and by checking firmware sources before dracut runs (dracut ignores module_install's exit code, so a broken list would otherwise surface only in the post-check).Root cause verified on an RP6 with a diagnostic initramfs: xhci probes fine, but typec, usb_role and rpmsg devices are all absent and the kernel logs "sync_state() pending due to 6000000.remoteproc".