Linus Torvalds recently released development kernel 2.5.24, listing the major changes:

"Big sound merges, NTFS merge, IrDA, IPv6, SCSI drivers, you name it. Rusty "it's two l's, dammit!" Russell sent in about a million of his trivial patches, hopefully they got somewhat correctly attributed and merged."

See the changelog for more details. Linus also mentioned that he was getting "ready to leave for the kernel summit and OLS" [earlier story]. The "invite only" kernel summit takes place next Monday and Tuesday, with the Ottawa Linux Symposium following it through Saturday. In addition to Linus' 2.5.24 announcement email, find Guillaume Boissiere's weekly kernel status report, this one pointing out what has and hasn't been accomplished since last years summit.

OpenSSH 3.3 was released today. This release includes improved support for privilege seperation (now enabled by default), and removal of the need for the sshd binary to be setuid root for protocol 2 hostbased authentication. (however the requirement was not removed for protocol 1 rhosts/rsa authentication) It can be downloaded from one of the many mirrors. The complete release announcement follows.

Darrin Jewell posted a lengthy review of the differences between NetBSD's FFS and Apple Darwin's UFS. He has worked to provide support for UFS in NetBSD.

He lists three main differences between the two filesystems. Most significantly, DIRBLKSIZ has been changed from 512 to 1024 "most likely for performance." The second difference he mentions is "that they shift the cylinder group cluster summary count array forward by 4 bytes." Finally, the third difference Darrin mentions is that UFS forces MAXSYMLINKLEN to 60, modifying "the maximum length of symlinks that can be stored directly in the inode."

The email goes on to explain how his patches manage to detect Apple's UFS, and what modifications they make to NetBSD. Darren says, "Since this is a substantial change to our file system code, I would like some guidance before committing it to our tree." His full email follows and makes for an interesting read.

Three months ago, in March of 2002, Bill Hayden announced that he was forking AtheOS [earlier story], porting it to run on top of the Linux kernel and providing wrappers for the BeOS API and Carbon API. The project is now officially named Cosmoe, the Compatible Open-Source Multi Operating-system.

Bill recently released version 0.5.2 of Cosmoe, saying:

"Cosmoe 0.5.2, the 'if only this had been the first release' release, is now available for download on www.cosmoe.com. If you've had trouble compiling Cosmoe in the past, check this release out."

This release solves numerous earlier compilation problems. Find a screenshot of Cosmoe running on a Linux 2.4.18 kernel here. Also of interest is this earlier OSNews interview with Bill. For the full release announcement, read on.

KernelTrap has spoken with guru Jordan Hubbard, one of the creators of FreeBSD and currently a manager of Apple's Darwin project. With just a high school education, Jordan has offered some impressive contributions to the world of computing.

In this interview, Jordan talks about his current involvement with Darwin, as well as his past efforts with FreeBSD and 386BSD. He also reflects on his recent decision to step down from the core FreeBSD team. Read on for the full interview.

A request for ideas on possible FreeBSD projects for a graduate student on the FreeBSD hackers mailing list led to an interesting email from Terry Lambert looking at several potential networking projects. He details several, including LRP (Lazy Receiver Processing), TCP Rate Halving (fast congestion recovery), SACK (Selective Acknowledgment), FACK (Forward Acknowledgment), ECN (Explicit Congestion Notification) and (the possibly patent blocked) VRRP (Virtual Router Redundancy Protocol).

A discussion on the lkml that began by pointing out a compile error in the 2.5 kernel tree evolved into an interesting discussion on Linux clusters. A brief exchange between Linus Torvalds and Eric Biederman started the discussion. During the exchange, Linus explained:

"I'm absolutely 100% conviced that you don't want to have a "single kernel" for a cluster, you want to run independent kernels with good communication infrastructure between them (ie global filesystem, and try to make the networking look uniform). Trying to have a single kernel for thousands of nodes is just crazy. Even if the system were ccNuma and _could_ do it in theory."

Larry McVoy [earlier interview] followed up with a reference to his cache coherent cluster slides, also forwarding a lengthy summary email from Paul McKenney. Altogether, it makes for an interesting read.

Craig Kulesa recently posted two VM patches for the 2.5 development Linux kernel, ports of Rik van Riel's [earlier interview] rmap VM [earlier story]. His intention is to keep Rik's rmap patches in sync with the 2.5 kernel tree.

The first patch is a complete port of Rik's work, along with some further cleanup. The second patch is a subset of the whole, about which Craig says, "It introduces the "minimal" rmap functionality of Rik van Riel's reverse mapping patches atop the current 2.5.23 classzone VM." His quick test of each reveals good results.

Rik seemed pleased with the effort, and indicated that after the Ottowa Linux Symposium [earlier story] he would also be working on rmap for the 2.5 kernel.

A lengthy thread on the stable FreeBSD mailing list discusses problems with the ATA code in the recently released FreeBSD 4.6 [earlier story]. One of the main problems with the ATA code appears to prevent certain fast ATA CD-ROM drives from working properly (AOpen 48x, 52x, and 56x drives were specifically reported with the problem). When hitting this bug, an error such as the following is displayed on the console and the kernel sometimes panics:

acd0: READ_BIG command timeout - resetting
ata1: resetting devices .. done

A workaround is available. The complete thread can be followed in the online stable-FreeBSD archive, starting here. A few relevant messages from the thread follow, including a discussion of the workaround.

Robert Love [earlier interview] recently submitted a patch to the O(1) scheduler, implementing scheduler hints. He explains the idea:

" Basically, scheduler hints are a way for a program to give a "hint" to the scheduler about its present behavior in the hopes of the scheduler subsequently making better scheduling decisions. After all, who knows better than the application what it is about to do?"

His patch currently supports three different hints: HINT_TIME (the task needs some more quanta, boost remaining timeslice), HINT_INTERACTIVE (the task is interactive, give it a small priority bonus to help), and HINT_BATCH (the task is a batch-processed task, give it a small priority penalty to be fair).

Though Robert considers this recent scheduler hints work "academic", a little testing revealed a 6% improvement in a five-thread test. He currently provides patches for Linux kernel versions 2.4.19-pre9 with the O(1) scheduler patch, and 2.5.20 (which already has the scheduler patch). Read on for his full post.

OpenBSD creator Theo de Raadt recently discussed the possibilty of removing 'pppd' from the OpenBSD source tree "because after a year+ of notices about the licenses in it not being completely correct, not not much has happened to get this fixed." The OpenBSD web site offers a review of licenses and a thorough explanation of applicable copyright policy.

Theo further explained the issue with pppd, "Most of the licenses in there are not acceptable. Some do not explicitly permit modification. Some do not explicitly permit any use (including sale)." He went on to request that OpenBSD users contact the pppd authors to help get the licensing issues cleaned up, saving pppd from being removed.

Much of the resulting thread follows.

Every now and again, the subject of Netcraft's OS detection system (part of their uptime graphing service) is raised on a Hurd mailing list. A recent thread discussed Netcraft's inability to detect the Hurd as a webhost, instead reporting it as Linux. Oystein Viggen explained:

"Unfortunately, the similarity between the Linux TCP stack and ours (which is also basically the Linux TCP stack) was so big that their TCP fingerprinting was not able to discern between the two. Obviously, they preferred to continue recognizing Linux with the small penalty that the Hurd is falsly seen as Linux to having both be "unknown OS"."

The full discussion follows.