Agatti/RB1 Boot regression between mainline and qcom-next

[lool]

Hi

I can't pinpoint this to a specific driver, but qcom-next is currently not booting on Agatti/RB1, while Linux mainline does, so this is an upcoming regression.

1) mainline

qcom-deb-images + mainline kernel from 2026-01-26 – works
a) dmesg after booting successfully locally:

dmesg-mainline-normal.txt

b) dmesg after changing kernel cmdline to add earlycon console=ttyMSM0,115200 and booting successfully locally:

dmesg-mainline-earlycon.txt

2) qcom-next

qcom-deb-images + qcom-next kernel from 2026-01-26
a) no output from kernel by default (systemd-boot starts and loads the kernel); boot output stops with:

scanning bus for devices...
starting USB...
No USB controllers found
       scanning usb for storage devices... 0 Storage Device(s) found
Hit <SPACE> key to stop autoboot in  1s
7[r[999;999H[6n8MMC: no card present
Failed to load EFI variables
FAT sector size mismatch (fs=4096, dev=512)
FAT sector size mismatch (fs=4096, dev=512)
FAT sector size mismatch (fs=4096, dev=512)
FAT sector size mismatch (fs=4096, dev=512)

b) after changing kernel cmdline to add earlycon console=ttyMSM0,115200, boot log captured from terminal stops at:

[    1.292941] NET: Registered PF_PACKET protocol family
[    1.298142] 9pnet: Installing 9P2000 support
[    1.302556] Key type dns_resolver registered
[    1.307451] NET: Registered PF_VSOCK protocol family
[    1.321525] registered taskstats version 1
[    1.325879] Loading compiled-in X.509 certificates
[    1.338275] Demotion targets for Node 0: null
[    1.384747] Freeing initrd memory: 14176K

hardcopy-dmesg-qcom-next-earlycon.txt

[shashim-quic]

@idlethread can someone check the Agatti boot issue.