View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0000635 | AlmaLinux-9 | kernel | public | 2026-06-25 12:48 | 2026-06-25 12:58 |
| Reporter | titoviejo | Assigned To | |||
| Priority | normal | Severity | major | Reproducibility | always |
| Status | new | Resolution | open | ||
| Summary | 0000635: AMDI0015 (DesignWare I3C master) driver initialization failure (Timeout -110 / UBSAN shift-out-of-bounds) on AMD EPYC 5th Gen | ||||
| Description | IMPORTANT NOTE FOR TRIAGE: We are opening this ticket here under the explicit direction of Benny (AlmaLinux). Although the attached logs were captured from a Proxmox VE kernel environment, this is a critical upstream Linux kernel driver/hardware compatibility issue on the new AMD EPYC 5th Gen platform. It directly impacts any modern enterprise Linux ecosystem trying to leverage native I3C telemetry on this Gigabyte hardware. --- A critical issue has been identified during the initialization of the native MIPI I3C bus on an AMD EPYC 5th Generation platform hosted on a Gigabyte server (Model: XV23-ZU0-AAJ1-000). During the boot sequence, the Synopsys/DesignWare I3C master control driver (dw-i3c-master.c) fails entirely. This driver failure triggers two explicit kernel errors: 1. UBSAN Math Error (Underflow): A shift-out-of-bounds error occurs when the driver tries to parse an invalid or corrupt hardware initialization response as an unsigned 64-bit integer. 2. Driver Timeout (Error -110): The integrated AMD ACPI I3C devices (AMDI0015) do not respond to the OS probing requests within the time limit (ETIMEDOUT), causing the controller to fail and disable itself. Secondary Impact (Power Capping): Concurrently, the CPU performance is degraded; the processor persistently caps its power consumption at ~245W despite its nominal 300W TDP specification. This is highly suspected to be a security Power Capping mechanism triggered by the AMD System Management Unit (SMU) or BMC due to the lack of high-speed telemetry from the VRMs/PMICs over the failed I3C bus. Hardware & Firmware Context: * Server Model: Gigabyte XV23-ZU0-AAJ1-000 (Motherboard MZU3-MG0-000) * CPU: AMD EPYC 5th Generation * Firmware: BIOS version R22_F17 | ||||
| Steps To Reproduce | 1. Boot the server on a modern Linux kernel supporting the native DesignWare I3C subsystem (observed on kernel 7.0.0-3-pve). 2. Check 'dmesg' or system logs right after initialization. 3. Observe the UBSAN shift-out-of-bounds trace and subsequent driver probe timeouts (-110) for AMDI0015 devices. Workaround tested: Forcing the operation mode of the 4 AMDI0015 ACPI devices from I3C to I2C (Legacy Mode) inside the UEFI settings stops the kernel crash and timeout errors. The OS boots stably under the standard I2C subsystem, but this removes the high-speed telemetry capabilities required for optimal power management. | ||||
| Additional Information | [Mon May 18 11:28:24 2026] UBSAN: shift-out-of-bounds in drivers/i3c/master/dw-i3c-master.c:1646:21 [Mon May 18 11:28:24 2026] shift exponent 18446744073709486145 is too large for 64-bit type 'long unsigned int' [Mon May 18 11:28:24 2026] Hardware name: Giga Computing XV23-ZUO-AAJ1-000/MZU3-MG0-000, BIOS R22_F17 04/01/2026 Call Trace: <TASK> dump_stack_lvl+0x5f/0x90 ubsan_handle_shift_out_of_bounds.cold+0x5e/0x113 dw_i3c_common_probe+0x34c/0x360 [dw_i3c_master] dw_i3c_probe+0x30/0x50 [dw_i3c_master] platform_probe+0x47/0xc0 ... dw-i3c-master AMDI0015:00: probe with driver dw-i3c-master failed with error -110 dw-i3c-master AMDI0015:01: probe with driver dw-i3c-master failed with error -110 dw-i3c-master AMDI0015:03: probe with driver dw-i3c-master failed with error -110 </TASK> | ||||
| Tags | No tags attached. | ||||
| Attached Files | |||||
