ai and ml
You can probably guess what he blames for the excessive workload.
Arm’s Lorenzo Stoakes is not abandoning Linux kernel development, although he is stepping back from some patch review work he previously did.
Stoakes is one of the lead software engineers at Arm and he’s a very busy guy. Besides writing the next book The Linux memory managerin his CV (PDF), he says:
“I have four maintenance positions: memory mapping, memory mapping (locking), memory mapping (madvise) – basically the entire user space API for memory management – and reverse mapping which connects physical memory to virtual memory.”
However, he plans to soon do a little less. Earlier this week, he sent an email to the Linux kernel mailing list to announce that he planned to drop some of his workload.
He explained:
“Given the current kernel throughput resulting from AI-generated runs, I unfortunately don’t have the bandwidth to be able to generally offer revisions except in the areas I manage. (…) Apologies for that, but given current trends, I’d like to do something to make my inbox a little more manageable.”
In fact, this came just a day after Paramount Penguin itself said it saw signs that the tide of AI fixes was “starting to turn.”
Stoakes said The register:
“I maintain my responsibilities and I don’t intend to give them up, I just want my inbox to be less horrible.
“And yes, AI is part of the reason; in general, it has gone from “once a week” to “daily” to “half of what I see in review.”
“But the AI wave really hits (memory management) across the board.”
He added:
I maintain my responsibilities and I don’t intend to give them up, I just want my inbox to be less horrible.
“I have a habit of burning out on weekends and evenings – you can see my involvement in a thread discussing this here – which gives you an idea of my perspective if you re-read that thread.”
At the Open Source Summit last month, the Reg The FOSS office joined the “Ask the Expert” session with Greg Kroah-Hartman, and he too expressed some worrying sentiments:
“The onslaught of LLM-generated patches… we don’t have enough people to review them all… We need more reviewers. We don’t have enough. Every maintainer is exhausted, and has been for ten years.”
Kroah-Hartman also said:
“I want to limit myself to two long-term support releases – four at most. Six is too many. We’re too busy. We don’t have the manpower. It’s crazy.”
Following Red Hat’s model of offering 10 years of support for products with a given kernel version:
“Debian does things right. Debian and Yocto run the world. I have argued for years that the RHEL and SLES model is broken. That’s why I no longer work for SUSE.” ®
Gn tech
