2.6.29 merge window, part 2

Did you know...? LWN.net is a subscriber-supported publication; we rely on subscribers to keep the entire operation going. Please help out by buying a subscription and keeping LWN on the net.

Linus Torvalds released 2.6.29-rc1 and closed the 2.6.29 merge window on January 10. A little over 2000 changesets were merged after the writing of last week's merge window summary; this article completes the summary for this development cycle.

Before getting into the details, though, it is worth pointing out that the 2.6.29-rc1 kernel has a couple of unusual traps for developers and testers. If you are playing with this kernel, you should be aware of the following:

• The Btrfs merge brought with it the entire development history for that project. One interesting result is that, if one uses git to check out a tree within that development history, the result will be a tree containing only Btrfs. In particular, this can happen in the middle of a bisection process, yielding a tree which cannot be built or tested - almost certainly not the desired result. The solution is easy, though; simply run:

      git bisect good
  

  and continue with the bisection process as usual.

• There is a portion of the kernel history which contains a badly broken version of reiserfs. Again, only developers running kernels from arbitrary points in the history will be affected by this problem; if you run reiserfs, though, read the summary and take care.

So what else was merged for 2.6.29? User-visible changes include:

• At the top of the list, of course, is the merge of the Btrfs filesystem. It cannot be repeated too many times, though, that Btrfs is still a development filesystem. Things are changing quickly, and it still will panic the system if you run out of space. Now is a good time for people to play with Btrfs - especially those who are willing to report bugs or submit enhancements. But it is not, yet, time to entrust your Valuable Intellectual Property to this filesystem.

• Also merged was the squashfs compressed, read-only filesystem. Squashfs has been packaged by distributors for years; its merger into the mainline was certainly overdue.

• There is now kernel support for WiMAX networking. The current code supports Intel's Wireless Wimax Connection 2400m devices, but others are expected for the future. See this documentation file for a bit of information on the WiMAX stack.

• There are new drivers for Atmel AVR32-based Hammerhead boards, Linear Technology LTC4245 Multiple Supply Hot Swap Controller I2C interfaces, Oxford OXU210HP USB host/OTG/device controllers, MIPS CI13412 USB controllers, Freescale IMX USB peripheral controllers, TI TWL4030 USB transceivers, Dell-specific laptop backlight and rfkill devices, ALIX.2 and ALIX.3 series LED controllers, PIKA FPGA watchdog devices, GE Fanuc watchdog timers, and NXP PCF50633 multifunction chips (as seen in OpenMoko devices).

• The Blackfin architecture has gained symmetric multiprocessing support. Also added is support for the BF51x family of processors.

• The memory controller has been extended to control swap usage as well. Previously, it would be possible for a memory-controlled group to exhaust swap space.

• The new "xenfs" virtual filesystem allows for information sharing and control between Xen domains, the hypervisor, and the host system.

• It is now possible to create and run ext4 filesystems without a journal. One loses the benefits of journaling, obviously, but there is a notable increase in performance.

• The filesystem freeze feature, allowing a suitably-privileged user to suspend changes to a filesystem (for backup purposes, perhaps) has been merged.

Changes visible to kernel developers include:

• The exclusive I/O memory allocation functions have been merged.

• The exports for a number of SUNRPC functions have been changed to GPL-only.

• The internal MTD (memory technology device) API has seen significant changes aimed at supporting larger devices (those requiring 64-bit sizes).

• An infrastructure for asynchronous function calls has been merged. This code is still a work in progress, though, and, for 2.6.29, it will not be activated in the absence of the fastboot command-line parameter.

And that completes the set of major changes added for 2.6.29 - with one possible exception. Linus has indicated that he would be willing to slip in an updated version of the spinning mutex code (as described in this Btrfs article) if it passes review in the near future.

Index entries for this article
Kernel	Releases/2.6.29

---

(Log in to post comments)