Linux Returns To Parallel CPU Microcode Updates To Reduce Cloud Disruption
Following the Spectre mitigations coming to light last year, the late microcode update process for CPUs was serialized. However, this has led to complaints from cloud vendors and other customers with large core count servers where downtime/disruptions need to be minimal. So now the CPU microcode update process is being parallelized again.
Serial late microcode updates on large core count servers was slow with it being done one CPU at a time. In the change by an Intel engineer, he acknowledged "cloud customers have expressed discontent." So now it's being parallelized again while ensuring only one core is updated at a time, i.e. not the SMT sibling thread.
At least one of the cloud vendors seeking this serial/parallel CPU microcode change was Oracle. That change is now queued in tip.git as material waiting around until Linux 5.5 unless some urgency is determined to get it into 5.4 as a "fix" for the kernel.
Serial late microcode updates on large core count servers was slow with it being done one CPU at a time. In the change by an Intel engineer, he acknowledged "cloud customers have expressed discontent." So now it's being parallelized again while ensuring only one core is updated at a time, i.e. not the SMT sibling thread.
At least one of the cloud vendors seeking this serial/parallel CPU microcode change was Oracle. That change is now queued in tip.git as material waiting around until Linux 5.5 unless some urgency is determined to get it into 5.4 as a "fix" for the kernel.
11 Comments