PSOD with error "Failed to ack TLB invalidate" while handling a Non-Maskable Interrupt

Products

• VMware vSphere ESXi

Issue/Introduction

Symptoms

• An ESXi 8.x host fails with a purple diagnostic screen (PSOD).
• The panic message is similar to below:

####-##-##T##:##:## cpu6:2097493)World: 3359: CR0 0x8001003d CR3 0x80006000 CR4 0x14016c
####-##-##T##:##:## cpu6:2097493)Panic: 630: Panic from another CPU (cpu 6, world 2097493): ip=0x42002858e93e randomOff=0x28400000:
PCPU 114 locked up. Failed to ack TLB invalidate (at least 1 locked up, PCPU(s): 114).
####-##-##T##:##:## cpu6:2097493)Panic: 823: Saved backtrace: pcpu 114 TLB NMI
####-##-##T##:##:## cpu6:2097493)pcpu 114 TLB NMI: 0x453bc391bdc8:[0x420028490ebe]Power_ArchPerformWait@vmkernel#nover+0x157 stack: 0x42005c801880, 0x800000000, 0x100000814, 0x42005c800000, 0x42005c800000
####-##-##T##:##:## cpu6:2097493)pcpu 114 TLB NMI: 0x453bc391bdd0:[0x420028490f8d]Power_ArchSetCState@vmkernel#nover+0xba stack: 0x800000000, 0x100000814, 0x42005c800000, 0x42005c800000, 0x0
####-##-##T##:##:## cpu6:2097493)pcpu 114 TLB NMI: 0x453bc391be20:[0x420028ada1ed]CpuSchedIdleLoopInt@vmkernel#nover+0x292 stack: 0x0, 0x7fffffffffffffff, 0x1, 0x7fffffffffffffff, 0xfffffffffffffff6
####-##-##T##:##:## cpu6:2097493)pcpu 114 TLB NMI: 0x453bc391be90:[0x420028adedc5]CpuSched_IdleLoop@vmkernel#nover+0x12 stack: 0x72, 0x4200284804e0, 0x0, 0x0, 0x0
####-##-##T##:##:## cpu6:2097493)pcpu 114 TLB NMI: 0x453bc391bea0:[0x42002855f3aa]Init_APIdle@vmkernel#nover+0x3f stack: 0x0, 0x0, 0x0, 0x0, 0x0
####-##-##T##:##:## cpu6:2097493)pcpu 114 TLB NMI: 0x453bc391beb0:[0x4200284804df]SMPAPIdle@vmkernel#nover+0x27c stack: 0x0, 0x0, 0x0, 0x0, 0x0
####-##-##T##:##:## cpu6:2097493)Panic: 783: Halting PCPU 6.

 

• VMkernel logs will have indications that there are no heartbeats received before the panic:

2026-06-29T21:17:36.548Z cpu40:5389621)WARNING: Heartbeat: 961: PCPU 114 didn't have a heartbeat for 5 seconds, timeout is 10, 1 IPIs sent; may be locked up.
2026-06-29T21:17:42.546Z cpu22:2102660)WARNING: Heartbeat: 961: PCPU 114 didn't have a heartbeat for 5 seconds, timeout is 10, 1 IPIs sent; may be locked up.

Environment

VMware vSphere ESXi

Cause

• This is typically caused when one CPU in sleep idle state fails to wake up to an interrupt request
• One of the reasons may be an outdated CPU microcode or BIOS firmware failing to manage voltage transitions correctly during idle periods

2026-06-29T22:57:44Z Al(177) vmkalert: 0:00:00:00.000 cpu0:1)ALERT: MicrocodeUpdate: 421: This machine's BIOS should be updated to pick up the latest microcode

Resolution

• Contact the hardware vendor to understand the root cause
• Update the system BIOS and CPU microcode to the latest version provided by the hardware vendor.