Bug 199533
| Summary: | Regression in 4.14.34 panics Celeron-847 when reloading microcode | ||
|---|---|---|---|
| Product: | Platform Specific/Hardware | Reporter: | Kris Karas (bugs-a21) |
| Component: | x86-64 | Assignee: | platform_x86_64 (platform_x86_64) |
| Status: | NEW --- | ||
| Severity: | normal | ||
| Priority: | P1 | ||
| Hardware: | Intel | ||
| OS: | Linux | ||
| Kernel Version: | 4.14.34 | Subsystem: | |
| Regression: | No | Bisected commit-id: | |
I have a Celeron 847 platform (x86_64) that works fine on kernels <= 4.14.33 But a regression was introduced into 4.14.34 that causes a PANIC and reboot when system initialization executes this: echo 1 > /sys/devices/system/cpu/microcode/reload That line is in rc.local, and is used during boot to update to the latest microcode (from the intel-ucode repository). Bisecting to find the culprit, we have this: commit 8413a3a63d3717504f2db5d6c3b018cabf15d132 Author: Borislav Petkov <bp@suse.de> Date: Wed Mar 14 19:36:15 2018 +0100 x86/microcode: Fix CPU synchronization routine commit bb8c13d61a629276a162c1d2b1a20a815cbcfbb7 upstream. Reverting that patch (or simply not performing the microcode update) makes this Celeron-847 happy again. Here is some console output, thanks to netconsole: microcode: Timeout while waiting for CPUs rendezvous, remaining: 1 Kernel panic - not syncing: Timeout during microcode update! CPU: 0 PID: 10 Comm: migration/0 Not tainted 4.14.33+ #1 Hardware name: ECS NM70-TI/NM70-TI, BIOS 4.6.5 05/30/2013 Call Trace: dump_stack+0x46/0x68 panic+0xca/0x208 __reload_late+0xab/0xb0 multi_cpu_stop+0x4f/0xa0 ? cpu_stop_queue_work+0x70/0x70 cpu_stopper_thread+0x7f/0xf0 ? sort_range+0x20/0x20 smpboot_thread_fn+0xda/0x150 kthread+0x105/0x120 ? __kthread_parkme+0x70/0x70 ? __kthread_parkme+0x70/0x70 ret_from_fork+0x1f/0x30 Shutting down cpus with NMI ObBugzilla: Sorry for filing this in "platform", but I couldn't find a bugzilla "product" for CPU. I toyed with "device drivers" but didn't see anything relevant.