Digg

Kernel 2.6.30 has arrived, and along with Tux as the old new mascot, is accompanied by an assortment of modifications to the new data system Ext 4. This was preceded by a lengthy discussion relating to the lagging write cycle of the data system which lead to a loss of data in some applications. While Ted Ts’o saw this as a problem on the part of the application developers, Linus Torvald tended towards the opinion that this issue lay with the kernel and was therefore not the developers’ responsibility. In this respect, the new kernel reduced performance in favor of data security. At the same time kernel developers have focused more on Ext 3 to avoid issues with the read and write processes.

The boot time of the kernel has been accelerated and improved: Alain Knaff has added support to the kernel for compression algorithms LZMA and BZIP2 so that the kernel images and initramFS can be more quickly depacketized, up to 33%. Likewise, Arjan van de Ven accelerated the drives over the SCSI and Libata subsystems so that the kernel now works on a parallel basis instead of sequentially. The fastboot option now functions by default. The new kernel also allows for a more superior interaction between suspend and interrupts.

At the graphic level, the new kernel supports the GPUs from the R6xx/R7xx series from ATI. The ATI Mobility Radeon RV350 now operates over a working suspend/resume function. Intel’s IGD chip supports the kernel just as well as VGA Hotplugging for the 945er chip set family. For the wireless realm, the new kernel already supports the recommended standard IEEE 802.11w which is to ensure the exchange of data between WLANs. As this data passes non-protected over the ethernet, potential areas susceptible to Hacker attacks are created. In addition, there are a few additional supported WLAN cards, such as one from Atheros (AR9170), from Atmel (AT76c50x), as well as PRISM54 chips which can be found in Nokia’s N800 and N810 devices. Other WLAN units, among which Ralink, now function better with the energy savers and the construction of mesh networks.
The new data system NILFS2 comes from NTT and functions on a log-structured basis. It writes updates from data and meta-data in a continuous stream or log onto the entire hard drive which administers as a continuous list of blocks. NILFS2 works with modifictions and extensions sequentially so that potential crashes avoid damage to the data system. At the same time it allows snapshots of past system writing processes to trigger so the Admin can gain access. In addition, POHMELFS has come into play, offering up a network based data system. It reads highly performing data from various sources and commands the ability to simultaneously write onto many nodes within the network. Presently, the data system is still in the staging realm. Likewise, additional network block device storage can be found there in the form of DST (distributed network storage) which allows local and remote nodes to be brought into use and possesses both a linear mode and mirror mode.

Many distributions offer the packets with the vanilla kernel which can be installed next to the standard kernels. Depending on whether a distribution is based on the original kernel or not, the possibility that certain functions or devices will cease to function correctly after the update so it is not advisable to chuck out the old kernels just yet.

Digg