Fresh Install Samsung MZAL NVMe drops to D3cold and fails to wake (Ryzen 5 5600U)
Hello,
I am attempting a fresh installation of CachyOS using the latest live ISO, but the boot sequence consistently fails due to my main NVMe drive dropping off the bus.
Hardware:
CPU: AMD Ryzen 5 5600U
RAM: 16GB
Motherboard: HP 8861 (KBC Version 41.35.00)
Storage: SAMSUNG MZAL8256HDJD-00BL2 (Firmware Version: 1L1QKXD7, NVMe Version: 2.0)
Boot Method: Ventoy loading CachyOS LTS ISO (Kernel 6.10.4)
During the live boot initialization, the system hangs and throws an endless loop of I/O timeout errors:
nvme0n1: Read(0x2) @ LBA... Host Aborted Command
nvme 0000:04:00.0: Unable to change power state from D3cold to D0, device inaccessible
Because the drive physically drops offline, all dependent targets (like mounting /boot and creating swap on /dev/zram0) fail, resulting in a system hard lock before I can reach the live environment login.
Troubleshooting Steps Already Taken:
Graphics Isolation: I initially suspected an AMD integrated graphics hang. I passed nomodeset 3 in the boot parameters, which correctly threw an amdgpu failed with error -22, confirming the GPU was bypassed and isolating the issue entirely to the storage controller.
Kernel Flags Added: To combat the aggressive APST/ASPM power management forcing the drive into deep sleep, I modified the boot parameters with the following flags:
nvmecore.defaultpsmaxlatencyus=0 pcieaspm=off pcieportpm=off iommu=pt nvme.noacpi=1
Hardware Reset: Between attempts, I performed a 30-second physical power drain to clear the motherboard capacitors and ensure the PCIe port was reset from the frozen D3cold state.
Despite these modifications, the OEM firmware on this Samsung drive still ignores the kernel overrides and drops the connection. I cannot provide exported journalctl or dmesg text files since I cannot reach a stable shell, but I have attached photos of the kernel logs.
Has anyone encountered a specific patch or module flag for this Samsung OEM firmware, or is this particular drive fundamentally incompatible with the current kernel's ASPM routing?
https://redd.it/1vcibtw
@r_linux
Post #42890
66