Hacker News new | past | comments | ask | show | jobs | submit login
Linux 3.0 release (lwn.net)
191 points by timf on July 22, 2011 | hide | past | favorite | 17 comments



Human-readable list of changes in 3.0.0, since 2.6.39: http://kernelnewbies.org/Linux_3.0

(For those who don't know, Kernelnewbies is a kernel developers community. They maintain by far the best changelogs.)


I disagree. The h-online.com articles, while not in bulletpoint style, are usually more detailed and better written, and currently the best way to dig into a new release.

Overview: http://www.h-online.com/open/features/What-s-new-in-Linux-3-...

Networking: http://www.h-online.com/open/features/Kernel-Log-Coming-in-3...

Filesystems: http://www.h-online.com/open/features/Kernel-Log-Coming-in-3...

Architectures and general infrastructure, including virtualization: http://www.h-online.com/open/features/Kernel-Log-Coming-in-3...

Drivers: http://www.h-online.com/open/features/Kernel-Log-Coming-in-3...



I wonder if they have the fixed the battery drain bug in this release.

When I installed Ubuntu 11.04 on my netbook, it had that bug. Read it on reddit that it has something to do with regression or something. So I had to downgrade to 10.10 to go back to previous version of the kernel.


Not exactly a regression. Current Linux started working around a BIOS bug that causes many people's systems to hang: namely, those BIOSes configure PCI Express devices to do Active State Power Management (ASPM), but tell the OS not to support ASPM, causing the system to go pear-shaped when the devices try to go into a low-power state. Linux now turns off ASPM on devices if the BIOS says not to use ASPM; that makes many people's systems more stable, at the cost of a bit more power usage.

So, complain to your BIOS vendor, or turn off that workaround in Linux and take the risk that your system will become less stable.


Im sure complaining to the BIOS vendor will work cough cough cough


How does Windows work around this?


"Windows" doesn't, the device vendors do.

I don't know anything about this particular bug, but it's no big deal: no doubt fixable for any given device with a quick patch ("If the DMI string is this and the BIOS says this, then it's lying and pretend it said this").

In the PC world, all the vendors have huge teams of engineers doing exactly this: applying hacks to get Windows running correctly on each new device. Then they package those hacks in per-device "driver" packages. This is the reason why, for example, every motherboard has "Windows drivers" for download where in Linux it's all part of the same kernel.

But this bites linux in the face of BIOS bugs: it has to make the assumption that the BIOS follows the relevant standards. It's not that you can't do the work to make this happen on any given device, it's that no one is paying the teams of engineers to actually do the work for every device.


Indeed, this is the crux of Linux compatibility problems: people will break their firmware to ensure it runs smoothly on Windows (e.g., sending ASPM information in an unconventional, broken way) while Linux is left holding the bag. All Linux can do is assume that something works according to spec until proven otherwise, and then in cases like this, it's hard to "undo" it because some things will function correctly and other things won't.

In other words, vendors only test on Windows and only care about Windows compatibility. Compatibility with other systems is almost always accidental. This puts MS at an advantage and everyone else at a keen disadvantage.


More precisely, it's that the team of engineers which is generally willing to do the work for every device, for free, is prevented from doing so by not having access to the internals of the product.


I'm not sure I buy that. If they had real access to the internals, they'd be fixing the BIOS bugs instead of hacking around it in the driver suite.

I think the real problem runs deeper: the culture of "firmware" in the PC world is fundamentally broken. PC BIOS exists more to protect the business models of the important players than it does to abstract the hardware in a meaningful way. And frankly it's hurting those players (like the big one that starts with an "I") badly in the embedded space, where 12-second POST times, duplicated drivers, and thousands of lines of workaround code aren't considered acceptable.


We don't know right now. Someone is cheating.




A little late to reply, but thankyou!


This article mentions that the kernel now has support for Microsoft Kinect. How so? It seems odd to support that in the core kernel...


It's support for access to the video and IR sensors.

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6...




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: