Hi Krzysztof, I have a BugHopper now and have attached the serial boot logs and the DTB procedure you requested.
This capture is on Ubuntu kernel 6.8.0-1089-qcom, package 6.8.0-1089.94. It includes firmware, UEFI, GRUB, Linux and the login prompt. The board boots from eMMC, but the NVMe still does not enumerate. The relevant lines are: [ 5.245430] qcom-qmp-pcie-phy 1c14000.phy: supply refgen not found, using dummy regulator [ 6.494836] qcom-pcie 1c10000.pci: Phy link never came up One distinction from the original report: this later boot no longer has my local REFGEN backport active. This is a capture of the missing-drive/link-training problem on 1089, not the original 1084 boot hang or a 1089 test with the backport applied. The command line still includes pcie_aspm.policy=performance. For the earlier DTB change, I compiled the attached refgen.dts with dtc -@ and used the attached patch_combined.py to apply it with fdtoverlay to only the arduino,monza trees in the boot and kernel combined DTBs. It added regulator@891c000 and refgen-supply links for both PCIe PHYs, leaving the existing supply properties and other board trees unchanged. The boot copy was deployed to the dtb_a FAT partition and the kernel copy under /lib/firmware/6.8.0-1084-qcom/device-tree/qcom/. The attached DTB-PROCEDURE.txt gives the commands, validation and capture context. I also rebuilt both historical patched DTBs from the saved originals without installing them; both outputs match the saved patched files byte for byte. The archive contains that hash comparison, the original UART bytes, readable logs and the matching kernel journal. Could you advise whether the next test should restore the REFGEN backport on 1089, or whether there is another supported DTB/PCIe fix to test first? Thanks, Karl On Thu, 05 Nov 2026 18:23:06 -0000, Krzysztof Adamski <[email protected]> wrote: > Hi Karl, sorry but I didn't have enough time to read through the linked > discussions but could you please describe the actual procedure that you > used to create your DTB change? > > Also it will be hard to find a fix without logs. Would it be possible > for you to use bughopper device or some other compatible UART converter > to get the boot logs? > > ** Changed in: linux-qcom (Ubuntu) > Status: New => Incomplete > > ** Changed in: linux-qcom (Ubuntu) > Assignee: Luca Corna (lucacorna) => Krzysztof Adamski (kadamski) > > -- > You received this bug notification because you are subscribed to the bug > report. > https://bugs.launchpad.net/bugs/2172593 > > Title: > VENTUNO Q fails to boot with SK hynix BC511 NVMe installed > > Status in linux-qcom package in Ubuntu: > Incomplete > > Bug description: > My VENTUNO Q boots with the M.2 slot empty. With a 512 GB SK hynix > BC511 installed, the fan spins briefly but the board never becomes > reachable over Ethernet or USB ADB. Removing the SSD gets it booting > again. I’ve repeated this several times. > > Expected: the board boots from eMMC and detects the SSD as a data > drive. > > The setup is: > > - Arduino 65 W power supply. > - Ubuntu 24.04, updated with apt. > - Kernel `6.8.0-1084-qcom`, package `6.8.0-1084.89`. > - BC511 firmware `11001101`, reporting NVMe 1.3. > > This is the Arduino Ubuntu image with updates and the local device- > tree change described below. I haven’t tested the Qualcomm 6.18 > branch. > > ## The SSD works over USB on another machine > > Through a Realtek RTL9210 enclosure on a separate Linux host, SMART > passed with zero media errors and zero error-log entries. Nine spaced > 128 MiB direct reads passed, covering 1.125 GiB. GPT and ext4 checks > also passed, along with a small file write/read test. > > That wasn’t a full-drive test, and it doesn’t test the Ventuno’s PCIe > connection. > > ## What I’ve tried > > 1. **Fresh partitioning.** The SSD had old MBR boot code but no usable > partitions. I removed it and created one GPT/ext4 partition. The boot > problem remained. The OS is still on eMMC. > > 2. **ASPM workaround.** I added `pcie_aspm.policy=performance`, > regenerated GRUB, and confirmed the setting was active after booting > without the SSD. Reinstalling the SSD gave the same result. > > 3. **REFGEN device-tree backport.** The kernel had the REFGEN driver > enabled, but the running tree lacked the regulator node and PCIe > `refgen-supply` links. It logged: > > ```text > qcom-qmp-pcie-phy 1c14000.phy: supply refgen not found, using dummy > regulator > ``` > > Using [PR #948](https://github.com/qualcomm-linux/kernel/pull/948) > and the REFGEN node included in [PR > #1013](https://github.com/qualcomm-linux/kernel/pull/1013), I added > the regulator at `0x0891c000` and connected both PCIe PHYs to it. The > existing `vdda-qref-supply` and `vdda-refgen-supply` entries were > already present. > > The board rebooted without the SSD. The driver bound, the NVMe PHY > used the real REFGEN regulator, and the warning disappeared. With the > SSD installed, it still failed to become reachable. > > ## Where I’m stuck > > The failed boots haven’t left a saved journal or pstore dump. I don’t > know whether the failure happens in firmware or early Linux boot. I > haven’t captured a serial trace yet. > > Is there a supported Ubuntu kernel, device-tree, or firmware fix > beyond the ASPM workaround? Does this REFGEN backport need another > dependency on the 6.8 kernel? > > Related Arduino discussion: [Issues with Ventuno Q (m.2, tmp, App > Lab)](https://forum.arduino.cc/t/issues-with-ventuno-q-m-2-tmp-app- > lab/1460599). > > Earlier report, with the same test results: > https://github.com/qualcomm-linux/kernel/issues/1360 > > To manage notifications about this bug go to: > https://bugs.launchpad.net/ubuntu/+source/linux-qcom/+bug/2172593/+subscriptions ** Bug watch added: github.com/qualcomm-linux/kernel/issues #1360 https://github.com/qualcomm-linux/kernel/issues/1360 ** Attachment added: "ventunoq-bug2172593-bootlogs-dtb-procedure.tar.gz" https://bugs.launchpad.net/bugs/2172593/+attachment/6017495/+files/ventunoq-bug2172593-bootlogs-dtb-procedure.tar.gz -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2172593 Title: VENTUNO Q fails to boot with SK hynix BC511 NVMe installed To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux-qcom/+bug/2172593/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
