Community > ReactOS Newsletter Archive > ReactOS Newsletter: Newsletter 45
ReactOS Newsletter - Newsletter 45 (#45)
|
by Z98 on 2008-08-11
topLinuxWorld Report
For the past week, Art Yerkes has been representing ReactOS at the Linux World Convention in California. He was unfortunately the only ReactOS developer or community member that could make it, but things seem to have turned out well. In the process of preparing for the convention, Art stumbled over an issue with drive letters that I will cover later. This forced him to rely on running ReactOS in QEMU, which actually turned out nicely. Many of the people expressing interest in ReactOS were Linux users and were specifically looking to run in virtual machines. Some of the applications Art demoed were Abiword, Firefox using static html pages due to lack of an internet connection at the booth, WinRAR, 7zip, the LMarbles game, and xemacs.
Some of the people Art met was a report from Ars Technica and another from the Alternageek online tech show. Their video has been posted at this location: http://alternageek.com/linuxworldexpo/art-yerkes-reactos-linuxworld-2008/ so be sure to check it out. He also had a chat with Hans Peter Anvin who expressed interested in our winldr code, likely for SYSLINUX.
topCommon Cache
Work on the Cc has been touched on before, where I explained NoCc and Aleksey's intentions with it. Art Yerkes has since picked up on the work and basically finished the extracting of Cc from the memory manager. A major problem originally was a shared data structure that polluted both the memory manager and the common cache. Both accessed and worked on it, but the data structure shouldn't have existed in the first place. This extricates the Cc from the Mm and allowed Art to begin writing a cache that only depended on interfaces exported by the Mm. Now, the Cc uses Mm section objects in order to build its cache.
The cache Art is currently working on remains fairly simplistic and follows the clock scheme. For people not familiar with memory management algorithms, the clock algorithm is basically where the entries of the cache are in a list and are walked by a 'hand' whenever the need arises to find a free spot. The entries themselves are marked with a flag that determines whether the hand will evict that entry for use. Every time the clock encounters a marked entry, it will unmark it so that the next time it is encountered the hand gets to free it. This is a simplistic explanation, but should get the idea across. Many variations exist, some with multiple hands and others with varying levels of priority for each entry.
The one Art is implementing is just a one handed clock cache and in this instance the hand also cannot unmark any of the entries. Not surprisingly, this means the system needs to wait for the application to unmark its entry or else the entire system crashes. The hand in this instance only traverses the unmarked entries when looking for a free spot for new entries. The entire cache is described by a bitmap and the unmarked entries are zeroes in the bitmap, which is how the hand knows which entries to go to next.
This work is nowhere complete as Art has to work out the kinks in the system. He is currently examining the various components around the Cc, including the section code in the memory manager as well as the filesystem drivers themselves. It will be a while before all this is functioning properly, but Art has already committed some code into a new branch so everyone can follow his progress and help.
topInstallation Quirks
Simplistically, ReactOS currently does these things wrong when trying to install: doesn't verify the volume that files are being copied from is actually the install CD, report the wrong volume for the CD when another volume is present on a disk, and report a volume even in a partition not marked in use. This was the drive letters issue mentioned above, as ROS currently does not fail right away if the CD is not at the drive letter expected. At this point, we want it to fail, as it means something has gone wrong with the drive detection. While trying to install ReactOS on a machine for demonstration at the convention, Art ran into these problems since he had an ext2 partition for testing purposes. These problems can actually manifest in other situations, but obviously install is the first place you would encounter it if your system is configured in such a manner. Fixing the second issue would at least partially eliminate the first issue, while the third issue should be fairly easy to deal with.
top
, Website Coordinator: Klemens Friedl 
