| Torvalds comments on Core 2 Duo bugs |
Jul. 02, 2007
"Bugs happen," concluded Linus Torvalds after a discussion of Core 2 Duo errata on the Linux Kernel Mailing List (LKML) last week. The pragmatic iconoclast defended Intel for handling bugs publicly, in contrast to embedded chip makers that tend to hide chip bugs behind workarounds in their OS ports.
The discussion of Core 2 Duo bugs was prompted by an OpenBSD mailing list post last week by Theo de Raadt. De Raadt, who founded the high-security oriented OpenBSD distribution, had expressed frustration over what he termed "unfixable" Core 2 Duo bugs. He characterized the Core 2 Duo architecture as "buggy as hell," while adding that AMD has "serious errata lists [that] are growing rapidly too."
For his part, Torvalds defended Intel for aggressively publishing bug reports, to the point of publishing bugs that can only be duplicated by ignoring Intel's own documentation. "[Intel] did learn something from the F00F and FDIV bugs, I think. They're very good about making [bugs] public these days," he wrote.
Torvalds goes on to contrast Intel's pro-active, public bug reporting to that of "boutique" and "embedded" chip makers. Smaller chip houses sometimes opt to "just ship the buggy crud" and "never tell anybody" about bugs, Torvalds suggests, if they can manage to work around bugs in their firmware and OS ports.
One of Torvalds's posts reads in part:A lot of [embedded CPUs] are quite buggy. The way the embedded people define it, though, any bugs becomes just "specifications".
For example, the ARM cache control has always been pretty damn buggy. What is the solution? Document it as a bug, and tell people not to use it.
Or grep for "errata" in arch/arm in the Linux kernel. Trust me, they exist.
But yeah, you can make a microcontroller CPU that is based on some decade-old core that is fairly bug-free. Not that it probably really is, but over the years the bugs have all become "behavior" rather than "bugs".
[Embedded chip vendors] tend to fix the bugs that are user-visible, and then not fix the bugs that can be worked around on an OS level.
Also, boutique vendors tend to not talk about [bugs], because it's all internal to their own stuff. Of course, if they don't catch it in time, they'll have to release OS upgrades, but if they find an errata early, they can just work around it and need never tell anybody, exactly like the random embedded ones.
It's really simple: when you count your CPU's in thousands rather than millions, you generally don't want to do a whole new mask set that costs you months and a few megabucks. It's much cheaper to just ship the buggy crud.
Yeah, x86 errata get more attention. But those things are pretty damn well tested. Better than most. And since the OS is outside the control of the vendors, they get fixed too. In another post, Torvalds said Linux was not affected, as far as he knew, by changes to the Core 2 Duo architecture's TLB (translation look-aside buffer). While De Raadt complained that Intel had "defined 'new ways to handle page tables,'" Torvalds explained, "We had mis-used the TLB earlier, and fixing that software bug we rewrote the page table handling to be more robust, which means that the spec update from Intel didn't affect us at all, afaik."
Torvalds did say Intel should have better-documented its TLB behaviour in the past. Better documentation could have warned users that higher-level page table structures could be cached, in the future. "If you depended on it not happening (and a lot of people did), it's painful," he admitted, adding, "But it really does make the architecture definition better and clearer."
The complete thread can be found here.
Related Stories:
(Click here for further information)
|
|
|
FUEL Database on MontaVista Linux
Whether building a mobile handset, a car navigation system, a package tracking device, or a home entertainment console, developers need capable software systems, including an operating system, development tools, and supporting libraries, to gain maximum benefit from their hardware platform and to meet aggressive time-to-market goals.
Breaking New Ground: The Evolution of Linux Clustering
With a platform comprising a complete Linux distribution, enhanced for clustering, and tailored for HPC, Penguin Computing¿s Scyld Software provides the building blocks for organizations from enterprises to workgroups to deploy, manage, and maintain Linux clusters, regardless of their size.
Data Monitoring with NightStar LX
Unlike ordinary debuggers, NightStar LX doesn¿t leave you stranded in the dark. It¿s more than just a debugger, it¿s a whole suite of integrated diagnostic tools designed for time-critical Linux applications to reduce test time, increase productivity and lower costs. You can debug, monitor, analyze and tune with minimal intrusion, so you see real execution behavior. And that¿s positively illuminating.
Virtualizing Service Provider Networks with Vyatta
This paper highlights Vyatta's unique ability to virtualize networking functions using Vyatta's secure routing software in service provider environments.
High Availability Messaging Solution Using AXIGEN, Heartbeat and DRBD
This white paper discusses a high-availability messaging solution relying on the AXIGEN Mail Server, Heartbeat and DRBD. Solution architecture and implementation, as well as benefits of using AXIGEN for this setup are all presented in detail.
Understanding the Financial Benefits of Open Source
Will open source pay off? Open source is becoming standard within enterprises, often because of cost savings. Find out how much of a financial impact it can have on your organization. Get this methodology and calculator now, compliments of JBoss.
Embedded Hardware and OS Technology Empower PC-Based Platforms
The modern embedded computer is the jack of all trades appearing in many forms.
Data Management for Real-Time Distributed Systems
This paper provides an overview of the network-centric computing model, data distribution services, and distributed data management. It then describes how the SkyBoard integration and synchronization service, coupled with an implementation of the OMG¿s Data Distribution Service (DDS) standard, can be used to create an efficient data distribution, storage, and retrieval system.
7 Advantages of D2D Backup
For decades, tape has been the backup medium of choice. But, now, disk-to-disk (D2D) backup is gaining in favor. Learn why you should make the move in this whitepaper.
|
|
|
|
|
news feed