http://www.secuobs.com Observatoire de la securite Internet fr webmaster@secuobs.com 2008-04-17 13:03:14 - KernelTrap Kernel news : "It's been long promised, but there it is now," began Linux creatorLinus Torvalds, announcing the 2625 Linux kernel He continued, "specialthanks to Ingo who found and fixed a nasty-looking regression thatturned out to not be a regression at all, but an old bug that just hadnot been triggering as reliably before That said, that was just thelast particular regression fix I was holding things up for, and it'snot like there weren't a lot of other fixes too, they just didn't endup being the final things that triggered my particular worries" Linusadded:"The full changelog from 2624 is 75M, with a 12MB compressedpatch Tons and tons has changed, but if you've been following the-rc releases, you'll already know about the big things Thechanges from the last rc -rc9 are fairly small and mostly prettytrivial, and the shortlog is appended So it's mostly one-liners,with some updates to drivers net and usb and to networking thatare a bit larger although a number of the driver updates arethings like just new ID's etc"More information about the latest release can be found on theKernelNewbies Linux 2625 wiki pageread morehttp://www.secuobs.com/revue/news/19022.shtmlhttp://www.secuobs.com/revue/news/19022.shtml Secuobs.com : 2008-04-16 15:03:01 - KernelTrap Kernel news - "Finally found it the patch below solves the sparsemem crash andthe test system boots up fine now," announced Ingo Molnar Hedescribed the patch as fixing a "memory corruption and crash on 32-bitx86 systems If a PAE x86 kernel is booted on a 32-bit system withmore than 4GB of RAM, then we call memory_present with a start/endthat goes outside the scope of MAX_PHYSMEM_BITS" He included a sourcesnippet with the loop that could corrupt memory, "depending on whatthat memory is, we might crash, misbehave or just not notice the bug"Ingo went on to note that the bug was first introduced with sparsememsupport in the 2616 kernel:"I believe this was the reason why my many bisection attempts wereunsuccessful: the bug pattern was not stable and seemingly workingkernels had the memory corruption too It was pure luck thatv2624 'worked' and v2625-rc9 broke visibly"Linux creator Linus Torvalds replied, "good job I've pushed this out,and will let this simmer at least overnight to see if there are anybrown-paper-bag issues either with this or with some last changesfrom Andrew, but I'm happy, and I think I'll do the real 2625tomorrow"read morehttp://www.secuobs.com/revue/news/18822.shtmlhttp://www.secuobs.com/revue/news/18822.shtml Secuobs.com : 2008-04-15 21:02:45 - KernelTrap Kernel news - "We are working on a new I/O scheduler based on CFQ, aiming atimproved predictability and fairness of the service, while maintainingthe high throughput it already provides," began Fabio Checconi,announcing the BFQ I/O scheduler "The Budget Fair Queueing BFQscheduler turns the CFQ Round-Robin scheduling policy of time slicesinto a fair queuing scheduling of sector budgets," he continued, "moreprecisely, each task is assigned a budget measured in number ofsectors instead of amount of time, and budgets are scheduled using aslightly modified version of WF2Q+ The budget assigned to each taskvaries over time as a function of its behaviour However, one can setthe maximum value of the budget that BFQ can assign to any task"Fabio went on to explain:"The time-based allocation of the disk service in CFQ, whilehaving the desirable effect of implicitly charging eachapplication for the seek time it incurs, suffers from unfairnessproblems also towards processes making the best possible use ofthe disk bandwidth In fact, even if the same time slice isassigned to two processes, they may get a different throughputeach, as a function of the positions on the disk of theirrequests On the contrary, BFQ can provide strong guarantees onbandwidth distribution because the assigned budgets are measuredin number of sectors Moreover, due to its Round Robin policy, CFQis characterized by an ON worst-case delay jitter in requestcompletion time, where N is the number of tasks competing for thedisk On the contrary, given the accurate service distribution ofthe internal WF2Q+ scheduler, BFQ exhibits O1 delay"Jens Axboe reacted favorably, "Fabio, I've merged the scheduler forsome testing Overall the code looks great, you've done a good job"He noted that the scheduler should soon appear in the -mm tree, andthat it was worth considering merging the two I/O schedulers togetherread morehttp://www.secuobs.com/revue/news/18671.shtmlhttp://www.secuobs.com/revue/news/18671.shtml Secuobs.com : 2008-04-15 01:32:41 - KernelTrap Kernel news - A thread on the Linux Kernel mailing list discussed the process inplace for reporting, bisecting and fixing bugs In response to asuggestion that some of the issues could be solved by introducing newprocedures, Al Viro retorted, "we've got ourselves a developingbeaurocracy As in 'more and more ways of generating activity withoutdoing anything even remotely useful' Complete with tendency tooperate in the ways that make sense only to bureaucracy in questionand an ever-growing set of bylaws" Later in the thread, DavidMiller agreed and noted that ,"the resulting 'bureaucracy' or whateveryou want to call it is perceived to undercut the very thing that makesthe Linux kernel fun to work on It's still largely free form, loose,and flexible And that's a notable accomplishment considering how muchthings have changed That feeling is why I got involved in the firstplace, and I know it's what gets other new people in and addicted too"Andrew Morton tried to return the discussion to its original topic, "theproblem we're discussing here is the apparently-large number of bugswhich are in the kernel, the apparently-large number of new bugs whichwe're adding to the kernel, and our apparent tardiness in addressingthem" Al noted that some of the problem is that git is so efficientat merging code, "the patches going in during a merge especially fora tree that collects from secondaries are not visible enough Andit's too late at that point, since one has to do somethingmonumentally ugly to get Linus revert a large merge On the scale ofGreat IDE Mess in 25" Another suggestion was made to replacebugzilla with something better, to which Andrew replied, "swapping outbugzilla for something else wouldn't help We'd end up with lots ofpeople ignoring a good bug tracking system just like they wereignoring a bad one"read morehttp://www.secuobs.com/revue/news/18472.shtmlhttp://www.secuobs.com/revue/news/18472.shtml Secuobs.com : 2008-04-13 01:32:22 - KernelTrap Kernel news - Open Graphics Project founder Timothy Miller recently noted on theproject's mailing list that they are set to announce that their firsthardware, the OGD1, is ready for pre-order "The OGD1 design hasactually been finished for a couple of months now," he began,explaining that they've been setting up a way to process pre-ordersfor the first 100 boards The board will retail at $1,500, with a $100discount offered for the first 100 pre-orders "These are pre-orders,not orders, Timothy continued, "that means the lead time isunpredictable We don't have a stock We will purchase a stock basedon the number of pre-orders we get Also, this means that if we neverget a large enough number of pre-orders, we will be unable to fulfillthem; all pre-orders would be canceled, and no one would be chargedanything" He then explained that though the OGD1 could function as agraphics card, it is instead offered as a competitively priced FPGAdevelopment kit, "we need to make it clear what OGD1 is and why buyingone is an important step for Free Software," adding:"OGD1 is for hardware hackers This isn't just about graphics Ifall you wanted was a graphics card that worked with Free Software,we've had that for a long time with Matrox, for some time withIntel, and most recently and significantly with ATI Where ourgraphics pipeline will be competitive is in embedded systems Asfor long-term goals of this project, there are many differenttypes of peripherals for which we do not have good Free Softwaresupport; for instance, wifi But let's not get ahead of ourselveshere OGD1 is for hardware hackers It's for the community ofpeople who want to tinker with their own hardware ideas, studentswho want to learn, and professionals who need a prototypingplatform And of course full schematics and design details forOGD1 are offered under the GPL"read morehttp://www.secuobs.com/revue/news/18154.shtmlhttp://www.secuobs.com/revue/news/18154.shtml Secuobs.com : 2008-04-12 01:32:25 - KernelTrap Kernel news - "I really don't want to do this, and I was actually hoping to release2625 last weekend which is why -rc9 is a few days late - just mehoping to not do another -rc at all, but I've done an -rc9," LinusTorvalds said, announcing the 2625-rc9 kernel "The changes in -rc9are pretty small shortlog appended," he continued, "and 60% of themare m68k updates - mostly defconfigs And some doc updates Butthere's some network driver updates tg3 and wireless hostap standout, some late XFS patches and a mvsas driver update the mvsasdriver is new in 2625, so that's not going to regress anything ;The rest is mostly one-liners, with a few reverts going on" Linusthen explained why he was putting out another release candidate:"The reason for not doing a 2625 is that some people are makingnoises about slab/page-alloc setup issues, and I wanted somethingout this week, but didn't feel comfy doing a final release"That said, I think I'll have to do 2625 early next weekregardless, because we can't just keep holding things backforever At some point it will have to turn into a 2625x issue,and the developers with stuff pending for the next version need tobe able to start merging"read morehttp://www.secuobs.com/revue/news/18097.shtmlhttp://www.secuobs.com/revue/news/18097.shtml Secuobs.com : 2008-04-11 19:02:58 - KernelTrap Kernel news - "Twice a year I get to release the song et lyrics, and write a littlecommentary on something the project dealt with other than therelease Hope you guys enjoy," said OpenBSD creator Theo de Raadt,including a link to the latest OpenBSD song The OpenBSD projectmaintains a six month release cycle, with the upcoming 43 releaseofficially scheduled for May 1st, 2008 Each release includes a songrelevant to issues faced by the project during the past six monthsThe song for the upcoming 43 release is titled, "Home to Hypocrisy",with scathing references to some recent postings on the OpenBSD -miscmailing list by Free Software Foundation creator Richard Stallman Inhis commentary, Theo explained, "we release our software in ways thatare maximally free We remove all restrictions on use anddistribution, but leave a requirement to be known as the authors" Hecontinued, describing the recent confrontation on the OpenBSD -miscmailing list:"We have a development sub-tree called 'ports' Our 'ports' treebuilds software that is 'found on the net' into packages thatOpenBSD users can use more easily A scaffold of Makefiles andscripts automatically fetch these pieces of software, applypatches as required by OpenBSD, and then build them into nice neatlittle tarballs Richard felt that this 'ports tree' of oursmade OpenBSD non-free He came to our mailing lists and lecturedto us specifically, yet he said nothing to the many other vendorswho do the same; many of them donate to the FSF and perhaps thathas something to do with it Meanwhile, Richard has personallymade sure that all the official GNU software -- including Emacs --compiles and runs on Windows"That man is a false leader He is a hypocrite There may be somepeople who listen to him But we don't listen to people who do notfollow their own stupid rules"read morehttp://www.secuobs.com/revue/news/18037.shtmlhttp://www.secuobs.com/revue/news/18037.shtml Secuobs.com : 2008-04-10 23:02:33 - KernelTrap Kernel news - "I have a question for the PF/ALTQ masters out there," Matthew Dillonbegan on the DragonFlyBSD kernel mailing list, having recentlyswitched from using a Cisco router to a DragonFlySD server running PF"I am trying to configure PF in a manner similar to what Cisco'sfair-queue algorithm does Cisco's algorithm basically hashes TCP andUDP traffic based on the port/IP pairs, creating a bunch of lists ofbacklogged packets and then schedules the packets at the head of eachlist" He went on to explain that he was unsuccessfully trying toconfigure the same thing with PF, "neither CBQ nor HFSC seem to workwell I can separate certain types of traffic but the real problem iswhen there are multiple TCP connections that are essentiallyclassified the same, and one is hogging the outgoing bandwidth So thequestion is, is there a PF solution for that or do I need to write anew ALTQ mechanic to implement fair queueing"Not finding a solution, he followed with a series of patchesimplementing what he needed He explained the resulting logic noting,"unless something comes up I am going to commit this to DragonFly onFriday and call it done I would be pleased if other projects pickedup some or all of the work":"The queues are scanned from highest priority to lowest priority;if the packet bandwidth on the queue does not exceed the bandwidthparameter and a packet is available, a packet will be chosen frothat queue; if a packet is available but the queue has exceededthe specified bandwidth, the next lower priority queue is scannedand so forth; if NO lower priority queues either have packets orare all over the bandwidth limit, then a packet will be taken fromthe highest priority queue with a packet ready; packet rate canexceed the queue bandwidth specification but will not exceed theinterface bandwidth specification, of course, but under fullsaturation the average bandwidth for any given queue will belimited to the specified value"read morehttp://www.secuobs.com/revue/news/17859.shtmlhttp://www.secuobs.com/revue/news/17859.shtml Secuobs.com : 2008-04-09 20:12:01 - KernelTrap Kernel news - "The latest feature release GIT 155 is available at the usualplaces," began Git maintainer Junio Hamano, adding "we kept this cyclejust slightly over two months, as the previous 154 cycle waspainfully tooooo long"Git is a distributed version control system that was originallywritten by Linus Torvalds in April of 2005 It was written to be onlya temporary replacement for BitKeeper, which Linus had been using tomanage kernel source code since February of 2002 Junio Hamano tookover maintainership of Git in July of 2005, and the tool has longsince become quite popular outside of even Linux kernel developmentRegarding the latest stable release, Junio highlighted some of thechanges, including:"Comes with git-gui 0101; bunch of portability improvementpatches coming from an effort to port to Solaris has been applied;'git fetch' over the native git protocol used to make a connectionto find out the set of current remote refs and another to actuallydownload the pack data We now use only one connection for thesetasks; 'git commit' does not run lstat2 more than necessaryanymore; bash completion script in contrib are aware of morecommands and options; a catch-all 'colorui' configurationvariable can be used to enable coloring of all color-capablecommands, instead of individual ones such as 'colorstatus' and'colorbranch'; bash completion's prompt helper function can talkabout operation in-progress eg merge, rebase, etc; 'git help'can use different backends to show manual pages and this can beconfigured using 'manviewer' configuration; 'git gui' learned anauto-spell checking; 'git checkout' and 'git remote' are rewrittenin C; two conflict hunks that are separated by a very short spanof common lines are now coalesced into one larger hunk, to makethe result easier to read"read morehttp://www.secuobs.com/revue/news/17522.shtmlhttp://www.secuobs.com/revue/news/17522.shtml Secuobs.com : 2008-04-08 19:32:41 - KernelTrap Kernel news - "The following three patches are intended to start the redesign of thesuspend and hibernation framework for devices," began Rafael WysockiHe noted that the first patch introduces new callbacks for suspend andhibernation, while the other two patches implement the new suspend andhibernation callbacks for the platform and PCI bus types Indescribing the first patch in the series, he noted that previouscallbacks were being phased out, explaining:"The main purpose of doing this is to separate suspend aka S2RAMand standby callbacks from hibernation callbacks in such a waythat the new callbacks won't take arguments and the semantics ofeach of them will be clearly specified This has been requestedfor multiple times by many people, including Linus himself, andthe reason is that within the current scheme if ->resume iscalled, for example, it's difficult to say why it's been calledie is it a resume from RAM or from hibernation or asuspend/hibernation failure etcThe second purpose is to make the suspend/hibernation callbacksmore flexible so that device drivers can handle more than they canwithin the current scheme For example, some drivers may need toprevent new children of the device from being registered beforetheir ->suspend callbacks are executed or they may want to carryout some operations requiring the availability of some otherdevices, not directly bound via the parent-child relationship, inorder to prepare for the execution of ->suspend, etc"read morehttp://www.secuobs.com/revue/news/17199.shtmlhttp://www.secuobs.com/revue/news/17199.shtml Secuobs.com : 2008-04-08 00:02:17 - KernelTrap Kernel news - Jörn Engel posted the sixth version of patches introducing his newLogFS filesystem for flash devices to the Linux kernel He highlightedsome areas of the code that need some more work, and cc'd theappropriate people for further review Regarding LogFS itself, henoted that one of its big advantages compared to other solutions wasimproved mount time and reduced memory consumption compared to othersolutions, "LogFS has an on-medium tree, fairly similar to Ext2 instructure, so mount times are O1" He went on to add that flash isbecoming more and more common in standard PC hardware, explaining:"Flash behaves significantly different to hard disks In order touse flash, the current standard practice is to add an emulationlayer and an old-fashioned hard disk filesystem As can beexpected, this is eating up some of the benefits flash can offerover hard disks In principle it is possible to achieve betterperformance with a flash filesystem than with the current emulatedapproach In practice our current flash filesystems are not evennear that theoretical goal LogFS in its current state is alreadycloser"read morehttp://www.secuobs.com/revue/news/16962.shtmlhttp://www.secuobs.com/revue/news/16962.shtml Secuobs.com : 2008-04-04 18:02:25 - KernelTrap Kernel news - "I skipped the public announcements for versions 5 and 6, but here is7 :," noted Vegard Nossum, announcing the latest release of hiskmemcheck patch, currently applying against the 2625-rc8 kernelVegard noted he is now hoping to get the patch merged into themainline kernel during the upcoming 2626 merge window He describedthe patch:"kmemcheck is a patch to the linux kernel that detects use ofuninitialized memory It does this by trapping every read andwrite to memory that was allocated dynamically eg usingkmalloc If a memory address is read that has not previouslybeen written to, a message is printed to the kernel log"Among the changes compared to earlier releases, v7 has undergone a lotof cleanup, some preparation has begun for an x86_64 port, errorreporting stability has been improved, boot time and run time optionshave been added, and there have been several bug fixesread morehttp://www.secuobs.com/revue/news/16499.shtmlhttp://www.secuobs.com/revue/news/16499.shtml Secuobs.com : 2008-04-02 12:03:11 - KernelTrap Kernel news - "No cute April 1st shenanigans, just a regular -rc release thathappened to come up today because I was waiting for the input layeroops-fixes to be ready and tested," began Linus Torvalds, announcingthe 2625-rc8 kernel on April 1st He continued, "the bulk of thefixes are the usual random one-liners A lot of the one-linersare some sparse cleanups, which is probably unnecessary noise at thispoint, but when Al sends me a series I just tend to apply it becausehis patches tend to be rather careful and basically always correct"Linus added:"The big thing that is actually *noticeable* to most people isthat this should fix the two top regressions: we've had somesuspend-resume regressions due to the stupid ACPI _PTS orderingissues, and while the cleanups were left, the ordering changeswere reverted So that should fix issues for some people ofcourse, the people who had it fixed are unhappy, but regressionsare worse The other thing that bit a number of people and is nowfixed and that also probably often showed up as a suspend/resumeregression was some 'struct device' lifetime changes that brokethe input layer Thanks to people who debugged that one"read morehttp://www.secuobs.com/revue/news/15829.shtmlhttp://www.secuobs.com/revue/news/15829.shtml Secuobs.com : 2008-04-01 16:32:35 - KernelTrap Kernel news - "I and the other Kernelorg admins would like to announce downtime forALL kernelorg machines this includes all of the mirror machines, thepublic machines and the backend master The downtime is scheduled tostart on or around April 2nd, 2008 on or around 0001 UTC," began a GPGsigned message on the Linux kernel mailing list from John 'Warthog9'Hawley, one of the kernelorg admins Referencing a recent Slashdotdiscussion that compared Linux and FreeBSD performance, he continued:"After much deliberation, research and argument in #korg alongwith screaming matches between HPA and I over dinner we areupgrading the kernelorg machines from Fedora Core 5 to FreeBSD70 This decision does not come lightly to the Kernelorg admins,and we would like to point out several key things that helped usform our decision:"John concluded, "we feel that we can better serve our mirrors, ourusers and the community by making the switch, and we hope to have thetransition done very shortly"read morehttp://www.secuobs.com/revue/news/15566.shtmlhttp://www.secuobs.com/revue/news/15566.shtml Secuobs.com : 2008-03-31 18:02:48 - KernelTrap Kernel news - Following the recent announcement that UBIFS is nearly productionready, it was asked how UBIFS compares to LogFS LogFS author JörnEngel suggested, "both share similar design goals Biggest differenceis that ubifs works on top of ubi and depends on ubi support, whilelogfs works on plain mtd or block devices and does everythingitself Code size difference is huge Ubi weighs some 11kloc, ubifssome 30, logfs some 8" He continued:"Ubi scales linearly, as it does a large scan at init time It isstill reasonably fast, as it reads just a few bytes worth ofheader per block Logfs mounts in O1 but will currently becomemindbogglingly slow when the filesystem nears 100% full and writeare purely random Not that any other flash filesystem wouldperform well under these conditions - it is the known worst casescenario"Artem Bityutskiy replied, "I personally refuse to compare a finishedFS with handles all the crucial flash features to a non-finished FSIt just makes no sense LogFS was talked about back 2005 in LinuxKongress, but is not finished yet Let's talk about it when it isproduction ready"read morehttp://www.secuobs.com/revue/news/15325.shtmlhttp://www.secuobs.com/revue/news/15325.shtml Secuobs.com : 2008-03-28 15:02:58 - KernelTrap Kernel news - "Here is a new flash file system developed by Nokia engineers withhelp from the University of Szeged The new file-system is calledUBIFS, which stands for UBI file system UBI is the wear-leveling/bad-block handling/volume management layer which is already inmainline see drivers/mtd/ubi," began Artem Bityutskiy He explainedthat UBIFS is stable and "very close to being production ready",aiming to offer improved performance and scalability compared to JFFS2by implementing write-back caching, and storing a file-system indexrather than rebuilding it each time the media is mounted Thewrite-back cache implementation claims to offer around a 100 timeimprovement in write performance over JFFS2 Artem went on to note:"UBIFS works on top of UBI, not on top of bare flash devices Itdelegates crucial things like garbage-collection and baderaseblock handling to UBI One important thing to note is MLCNAND flashes which tend to have very small eraseblock lifetime -just few thousand erase-cycles some have even about 3000 orless This makes JFFS2 random wear-leveling algorithm to be notgood enough In opposite, UBI provides good wear-leveling based onsaved erase-counters"read morehttp://www.secuobs.com/revue/news/14407.shtmlhttp://www.secuobs.com/revue/news/14407.shtml Secuobs.com : 2008-03-27 16:32:59 - KernelTrap Kernel news - "Now that we are presumably approaching the next merge window, can Iask what use if any will you be making of the linux-next treeAlternatively, is there any information you want from it" StephenRothwell asked regarding the tree he started maintaining last monthfor tracking upcoming stable mergesAndrew Morton replied, "afacit it's already working The level ofmerge and build errors in the subsystem trees this time around is atiny fraction of what it was at the same stage in 2624-rcX" He wenton to note, "there are 60 or 80 "susbsytem" trees hosted in -mm atpresent," adding:"I need to find a way to a get matureish parts of those treesinto linux-next and to b base the rest of -mm off linux-next Ihaven't started thinking about that yet There seem to be sometrees which aren't yet in linux-next, some of them significant"read morehttp://www.secuobs.com/revue/news/14141.shtmlhttp://www.secuobs.com/revue/news/14141.shtml Secuobs.com : 2008-03-26 16:33:50 - KernelTrap Kernel news - "So this hopefully continues closing various regressions, and most ofthe changes are pretty small ie diffstat shows a lot of onelinersThe biggest patches are the trivial powerpc defconfig updates whichshow up pretty clearly in the dirstat, ie if it weren't for those, thearch/ updates would hardly show up at all," began Linus Torvalds,announcing the 2625-rc7 Linux kernel He noted that the ps2esdidriver was removed after being marked broken for years, and a newmetronomefbc driver was added for the E-Ink Metronome controllerLinus continued:"Apart from those, most of the changes really are fairly small andspread out The scheduler got some tweaking, the memstick drivergot some TLC, and cifs and reiserfs had some fixes The shortloghas more details, but it boils down to some reverts, some docbookfixes, some sparse annotation fixups, a number of trivial patches,and a healthy sprinkling of small fixups"In summary, Linus suggested, "give it a good testing, because we'rehopefully now well on our way towards that eventual real 2625release"read morehttp://www.secuobs.com/revue/news/13909.shtmlhttp://www.secuobs.com/revue/news/13909.shtml Secuobs.com : 2008-03-25 15:27:57 - KernelTrap Kernel news - Matthew Dillon posted on update on his evolving HAMMER filesystem,noting that it "passes all standard filesystem stress tests andbuildworld will run with a HAMMER /usr/obj" He also noted, "pruningand reblocking code is in and partially tested, but now needs morestringent testing; full historical access appears to be working butneeds testing" He added, "there are two big-ticket and severallittle-ticket items left HAMMER will officially go Alpha when thebig-ticket items are done, and beta when we get a few of thelittle-ticket items done" The two "big-ticket" items left to becompleted are UNDO crash recovery code, and handling for fullfilesystems Matt summarized:"I have no time frame for these items yet It will depend on howquickly HAMMER moves to Alpha and Beta status I will say,however, now that HAMMER's on-disk format has solidified, that Ihave a very precise understanding of the protocols that will beneeded to accomplish fully cache coherent remote access for bothreplicated and non-replicated remote mount style access And, asyou know, fully coherent filesystem access across machines isgoing to be the basis for DragonFly's clustering across saidmachines In summary, things are progressing very well"read morehttp://www.secuobs.com/revue/news/13664.shtmlhttp://www.secuobs.com/revue/news/13664.shtml Secuobs.com : 2008-03-24 12:44:48 - KernelTrap Kernel news - "Google Summer of Code 2008 is on Over the past three years, theprogram has brought together over 1500 students and 2000 mentors from90 countries worldwide, all for the love of code We look forward towelcoming more new contributors and projects this year," begins a pagelisting all the projects planning to participate in this year's GSoCAmong the numerous planned participtants there are many kernelprojects, including DragonFly BSD, FreeBSD, Git, GNU/Hurd, Linux, andNetBSDStudent applications for GSoC projects begin today, running throughthe end of the month Read on for many of the participationannouncements from the above projects For more information about theGSoC, the program's FAQ explains:"Google Summer of Code GSoC is a program that offers studentdevelopers stipends to write code for various open sourceprojects Google will be working with a several open source, freesoftware, and technology-related groups to identify and fundseveral projects over a three month period Historically, theprogram has brought together over 1,500 students with over 130open source projects to create millions of lines of code Theprogram, which kicked off in 2005, is now in its fourth year"read morehttp://www.secuobs.com/revue/news/13354.shtmlhttp://www.secuobs.com/revue/news/13354.shtml Secuobs.com : 2008-03-22 01:32:55 - KernelTrap Kernel news - "Allocations of larger pages are not reliable in Linux If largerpages have to be allocated then one faces various choices of allowinggraceful fallback or using vmalloc with a performance penalty due tothe use of a page table," began Christoph Lameter, describing thethird version of his virtual compound page support patchset Hecontinued, "a virtual compound allocation means that there will befirst of all an attempt to satisfy the request with physicallycontiguous memory If that is not possible then a virtually contiguousmemory will be created" Christopher proposed two advantages:"1 Current uses of vmalloc can be converted to allocate virtualcompounds instead In most cases physically contiguous memory canbe used which avoids the vmalloc performance penalty 2 Uses ofhigher order allocations stacks, buffers etc can be converted touse virtual compounds instead Physically contiguous memory willstill be used for those higher order allocs in general but thesystem can degrade to the use of vmalloc should memory becomeheavily fragmented"read morehttp://www.secuobs.com/revue/news/13118.shtmlhttp://www.secuobs.com/revue/news/13118.shtml Secuobs.com : 2008-03-18 03:54:14 - KernelTrap Kernel news - "I lost a day-and-a-half this week due to a disk that decided to getread errors due to an unfortunate power outage, and had to spend toomuch time regenerating my normal setup," began Linus Torvalds,announcing the 2625-rc6 kernel, "but I don't think I lost anyemails, and things seemed to have calmed down a bit, so here's tohoping that -rc6 is starting to look better" He then summarized thechanges:"The dirstat shows the usual pattern of most changes being indrivers and architecture updates, although this time it's a bitskewed by the parisc and powerpc updates hopefully closing theparisc compile regression among other things, which means thatarch is about half, and drivers are just under a third of thepatch it seems to be usually the other way around"read morehttp://www.secuobs.com/revue/news/11949.shtmlhttp://www.secuobs.com/revue/news/11949.shtml Secuobs.com : 2008-03-13 20:45:40 - KernelTrap Kernel news - "A change after 2624 broke ndiswrapper by accidentally removing itsaccess to GPL-only symbols," noted Pavel Roskin, offering a patch toaddress the issue Linux creator Linus Torvalds was unimpressed, "I'mnot seeing why ndiswrapper should be treated separately If it loadsnon-GPL modules, it shouldn't be able to use GPLONLY symbols" TheNDISwrapper project page explains, "many vendors do not releasespecifications of the hardware or provide a Linux driver for theirwireless network cards This project implements Windows kernel API andNDIS Network Driver Interface Specification API within Linux kernelA Windows driver for wireless network card is then linked to thisimplementation so that the driver runs natively, as though it is inWindows, without binary emulation" Due to this, Linus explained:"Ndiswrapper itself is *not* compatible with the GPL Trying toclaim that ndiswrapper somehow itself is GPL'd even though it thenloads modules that aren't is stupid and pointless Clearly it justre-exports those GPLONLY functions to code that is *not* GPL'd"read morehttp://www.secuobs.com/revue/news/10875.shtmlhttp://www.secuobs.com/revue/news/10875.shtml Secuobs.com : 2008-03-13 20:45:40 - KernelTrap Kernel news - "It's a few days late, but I was waiting for some updates for some ofthe most annoying regressions until releasing it, so the end result ishopefully more useful as a result," Linus Torvalds began, announcingthe 2625-rc4 kernel He offered a dirstat summary, noting, "thedirstat shows that as usual most of the changes are in drivers andarch ~51% and ~17% respectively, with about half the driver updatesbeing in network drivers" Linus continued:"In particular, the block layer changes should hopefully havesorted themselves out, and CD burning etc hopefully works forpeople again Same goes for the the scheduler regressions, and anumber of annoying boot-time problems It's really a fairamount of small changes spread all over, with most of the changesbeing quite small 604 commits, most of them small, with the BNX2Xnetwork driver and the new fsldma driver the only ones that gotsome bigger changes"read morehttp://www.secuobs.com/revue/news/10874.shtmlhttp://www.secuobs.com/revue/news/10874.shtml Secuobs.com : 2008-03-01 21:03:05 - KernelTrap Kernel news - "Ok, it's out there, ready for your enjoyment," Linus Torvalds said,announcing the 2625-rc3 kernel He summarized the changes:"As usual, most of the updates are in architecture and drivers,with the dirstat showing about 37% in arch and that's with renamedetection: there's some file movement in arch/xtensa that wouldbring it up to 43% if you looked at it as a traditional diff andalmost 50% in drivers Much of the include file stuff is alsoarchitecture-related updates The driver updates are mostly fairlyspread out, but some of it comes from a couple of new drivers: themvsas SCSI driver, a new adt7473 driver, and a couple of newwatchdog drivers"Linus continued, "if you ignore the architecture-specific stuff anddrivers, the rest is mostly in networking, some Documentation updates,and a few filesystem updates mainly efs and xfs Anyway, the upshotof it all Quite frankly, it's all over the place The changes in -rc3are bigger than -rc2, probably mostly because we had some more time-rc2 was a couple of days early because of the long weekend in theUS, but hopefully also because people have started to findregressions" Among the bug fixes, he highlighted, "we had a nastySLUB corruption issue in -rc2 that is fixed not that very many peopleprobably saw it, and we've hopfully fixed a number of regressions innetworking and suspend/resume"read morehttp://www.secuobs.com/revue/news/10166.shtmlhttp://www.secuobs.com/revue/news/10166.shtml Secuobs.com : 2008-03-01 21:03:05 - KernelTrap Kernel news - "Hello everyone We are happy to say that the 112 release is nowavailable" began Matthew Dillon, announcing the latest stable versionof DragonFly BSD The project's home page explains, "DragonFly is anoperating system and environment originally based on FreeBSDDragonFly branched from FreeBSD in 2003 in order to develop aradically different approach to concurrency, SMP, and most otherkernel subsystems" Regarding the latest release, Matt explained:"This release is primarily a maintenance update A lot of work hasbeen done all over the kernel and userland There are no newbig-ticket items though we have pushed the MP lock further intothe kernel"The 20 release is scheduled for mid-year Of the currentbig-ticket item work, the new HAMMER filesystem is almost to thealpha stage of development and is expected to be production readyby the mid-year 20 release"read morehttp://www.secuobs.com/revue/news/10165.shtmlhttp://www.secuobs.com/revue/news/10165.shtml Secuobs.com : 2008-02-25 11:07:07 - KernelTrap Kernel news - "To quote you a number of years ago: 'Linux is evolution, notintelligent design'," noted Greg KH, quoting Linux creator LinusTorvalds Linus expanded on the statement, "evolution often does oddand 'suboptimal' things exactly because it does incremental changesthat DO NOT BREAK at any point" He continued, "in other words,exactly *because* evolution requires 'bisectability' any non-viablepoint in between is a dead end by definition and does thingsincrementally, it doesn't do big flips" When alternative examples inevolution were pointed out, Linus suggested that the kernel was muchsimpler than a mammal and more similar to bacteria:"In bacteria and viruses, duplication of DNA/RNA is a big costof living in general, and as a result there is *much* less junkDNA So in an evolutionary sense, it's much closer to what thekernel should have with occasional duplication of code andinterfaces to allow new functionality, but rather aggressivepruning of the excess baggage In other words, all of thesechoices are a matter of 'balance' In some areas, excess code isnot a sufficient downside, and we keep even broken source codearound with no actual function, 'just because' or rather, becausethe cost of carrying it around is so small that nobody cares"read morehttp://www.secuobs.com/revue/news/8981.shtmlhttp://www.secuobs.com/revue/news/8981.shtml Secuobs.com : 2008-02-25 11:07:07 - KernelTrap Kernel news - "These patches add local caching for network filesystems such as NFS,"began David Howells describing an updated set of thirty-seven patchesto introduce FS-Cache When asked how the patches affect performance,he noted that this was dependent on the use case, highlighting issueswhen dealing with lots of metadata, "getting metadata from the localdisk fs is slower than pulling it across an unshared gigabit ethernetfrom a server that already has it in memory"David continued "these points don't mean that fscache is no use, justthat you have to consider carefully whether it's of use to *you* givenyour particular situation, and that depends on various factors,"adding, "note that currently FS-Caching is disabled for individual NFSfiles opened for writing as there's no way to handle the coherencyproblems thereby introduced" He concluded with a number of simpleperformance benchmarksread morehttp://www.secuobs.com/revue/news/8980.shtmlhttp://www.secuobs.com/revue/news/8980.shtml Secuobs.com : 2008-02-25 11:07:07 - KernelTrap Kernel news - Issues reported during the suspend-to-disk process lead Linux creatorLinus Torvalds to suggest, "please - just make the f*ckingsuspend-to-disk use other routines already 99% of all hardware needsto do exactly *nothing* on suspend-to-disk, and the ones that reallydo need things tend to need to not do a whole lot" He went on toexplain why sharing the code path for suspend-to-disk and freezing toRAM is wrong:"For example, the 'freeze' action for USB which is one of thehardest things to suspend should literally be something like justsetting the controller STOP bit, and waiting for it to havestopped The 'unfreeze' should be to just clear the stop bit,while the 'restart' should be just a controller reset to use thecurrent memory image NONE OF THIS HAS ABSOLUTELY ANYTHING TO DOWITH SUSPEND It never did I've told people so for years Maybeactually seeing the problems will make people realize"Linus also noted another advantage to having separate code paths forthe two actions, "the other issue is that I've long wanted to makesure that when people fix suspend-to-ram, they don't screw upsuspend-to-disk by mistake and vice versa" During the discussion,Rafael Wysocki noted that he would be fixing this up presently, "I'malready convinced, really :-"read morehttp://www.secuobs.com/revue/news/8979.shtmlhttp://www.secuobs.com/revue/news/8979.shtml Secuobs.com : 2008-02-17 01:47:30 - KernelTrap Kernel news - "Ok, this kernel is a winner," began Linux creator Linus Torvalds,playfully announcing the 2625-rc2 kernel which gained the name "FunkyWeasel is Jiggy wit it" He continued:"Just to show how _much_ of a winner it is, it's been awarded acoveted 'weasel' series name, which should tell you just how goodit's going to be It's a name revered in Linux kernel history, andas such this brings back the good old days where if you find abug, you're almost certainly simply mistaken, and you probablyjust did something wrong But hey, you can try to prove me wrongI dare you"Linus went on to describe some of the changes using 'git dirstat', "inparticular, it shows that almost exactly half of the updates are todrivers, with network drivers alone being a third of the whole patchAnd of the remaining half, about half was architecture updates,notably to SH" He then noted, "I'm optimistic that this release cyclewon't be anywhere near the pain of what 24 was, which is why I'm justgoing to go off for the long weekend and stay at the beach"read morehttp://www.secuobs.com/revue/news/8152.shtmlhttp://www.secuobs.com/revue/news/8152.shtml Secuobs.com : 2008-02-14 17:30:46 - KernelTrap Kernel news - "Andrew Morton was looking for someone to run a linux-next tree thatjust contained the subsystem git and quilt trees for 26x+1 and I ina moment of madness volunteered So, this is to announce the creatingof such a tree," began Stephen Rothwell, resulting in a lengthy threaddiscussing the current Linux kernel development process In a followup email announcing the first linux-next release, Stephen went on toexplain, "it has two branches - master and stable Stable is currentlyjust Linus' tree and will never rebase Master will rebase on analmost daily basis maybe slower at the start" He then detailed themaster branch:"The tree consists of subsystem git and quilt trees Currently,the quilt trees are integrated by importing them intoappropriately based git branches and then merging those branchesThis has the advantage that any conflict resolution will onlt haveto happen once at the merge point rather than, possibly, severaltimes during the series However, I am considering just applyingthe quilt trees on top of the current tree to get a result morelike Linus' tree - we will see The git trees are obviously justmerged"One of the goals of the new tree is to get subsystem maintainers moreinvolved in managing merge conflicts by quickly notifying all involvedwhen things break, and automatically dropping the offenders untilbuild problems are fixed Andrew plans to base his -mm kernel on thenew linux-next tree, providing a stabler test branch for code beforeit's merged into Linus' mainline kernel treeread morehttp://www.secuobs.com/revue/news/7540.shtmlhttp://www.secuobs.com/revue/news/7540.shtml Secuobs.com : 2008-02-12 14:17:21 - KernelTrap Kernel news - "HAMMER won't be ready for sure things take however long they take,but the hardest part is working and stable and I'm just down togarbage collection and crash recovery," noted Matthew Dillon,discussing the status of what is ultimately intended to be a highlyavailable clustering filesystem The upcoming DragonFlyBSD releasethis month was originally intended to be 20 with a beta qualityHAMMER, but the decision was recently made to call the release 112while HAMMER continues to stabilize Matt continued, "HAMMER is reallyshaping up now Here's what works now: all filesystem operations; allhistorical operations; all Pruning features" During the discussion,he was asked how he planned to support multi-master replication, inreply to which he began:"My current plan is to use a quorum algorithm similar to the one Iwrote for the backplane database years ago But there are reallytwo major and very complex pieces to the puzzle Not only do weneed a quorum algorithm, but we need a distributed cache coherencyalgorithm as well With those two pieces individual machines willbe able to proactively cache filesystem data and guaranteetransactional consistency across the cluster"read morehttp://www.secuobs.com/revue/news/6891.shtmlhttp://www.secuobs.com/revue/news/6891.shtml Secuobs.com : 2008-02-12 13:39:19 - KernelTrap Kernel news - "With a lot of help from Ingo Molnar and Pekka Enberg over the lastcouple of weeks, we've been able to produce a new version ofkmemcheck" announced Vegard Nossum, adding, "the current version ofthe patch boots on real hardware, but we've seen freezes on somemachines, so it's not perfect yet In other words, this patch isHIGHLY experimental, and run at your own risk, etc" He also offereda high level summary of the patch:"kmemcheck is a patch to the linux kernel that detects use ofuninitialized memory It does this by trapping every read andwrite to memory that was allocated dynamically eg usingkmalloc If a memory address is read that has not previouslybeen written to, a message is printed to the kernel log"Ingo Molnar credited the new patch with already finding 4 kernel bugs,and offered some more insights into how the patch works, and why it'suseful, "it should also be made clear that not only does kmemcheckconsume half of the RAM to do byte granular tracking of the other halfof RAM, it's also slow, very slow, because almost every kernel-spaceinstruction will generate a pagefault and then it will besingle-stepped and it takes a debug fault as well That's of coursetotally crazy, but that's also OK and it's what makes the feature sointeresting and powerful"read morehttp://www.secuobs.com/revue/news/6746.shtmlhttp://www.secuobs.com/revue/news/6746.shtml Secuobs.com : 2008-02-12 13:39:19 - KernelTrap Kernel news - "While this is probably one of the last days of the merge window,please still consider pulling the 'kgdb light' git tree," began IngoMolnar, explaining:"This is a slimmed-down and cleaned up version of KGDB that i'vecreated out of the original patches that we submitted two weeksago I went over the kgdb patches with Thomas and we cut outeverything that we did not like, and cleaned up the result KGDBis still just as functional as it was before i tested it on32-bit and 64-bit x86 - and any desired extra capability orcomplexity should be added as a delta improvement, not in thisinitial merge"Ingo noted that the previous merge request modified 41 files, whilethis new merge request modifies only 22 files Among the changes, hehighlighted, "removed _all_ critical path impact, even if KGDB isenabled and active; removed all the lowlevel serial drivers; added aredesigned and cleaned up version of the 'KGDB over polled consoles'approach; removed the longjump code; removed the module symbol hacks;removed the GTOD/clocksource hacks; removed the softlockup hacks;removed the toplevel Makefile changes; removed the might_sleepscheduler hack; and did lots of other cleanups and rewrites as well"Ingo summarized, "as a result, this kgdb series has _obviously_ zeroimpact on the kernel, because it just does not touch any dangerouscodepath From this point on KGDB can evolve in small, well-controlledbaby steps, as all other kernel code as well And the resulting kgdbis still very functional: it can still break into a kernel viaSysRq-G, can catch crashes, can single-step, etc It's already aquite usable first step"read morehttp://www.secuobs.com/revue/news/6745.shtmlhttp://www.secuobs.com/revue/news/6745.shtml Secuobs.com : 2008-02-12 13:39:19 - KernelTrap Kernel news - "Ok, it's a bloody large -rc as was 24-rc1, for that matter,probably because the 2624 release cycle dragged out, so people had alot of things pending," noted Linus Torvalds, announcing the2625-rc1 kernel He added, "the full diff is something like 11MB and14M lines of diffs, with the bulk of the stuff being in architectureupdates and drivers" Linus continued:"Just to have some fun, I did trivial statistics, and of the 14Mlines of diffs, about 38% - 530k lines - were in architecturefiles 400k+ lines of diffs in arch/, 100k+ lines of diffs ininclude/asm-*, and another big chunk is in drivers includingsound at about 44% - 610k lines - of changes The rest comes inmuch smaller, but still noticeable is networking 8% - 110klines, with filesystems at 4%, and documentation at about 2% Theremaining crumbles being spread out mostly over block layer,crypto, kernel core, and security layer updates ie SElinux andsmack"Linus highlighted a few of the changes, including, "the Intel graphicsdriver is starting to do suspend/resume natively ie even without Xsupport, which is a welcome sign of the times and may help somepeople; lots of cleanups from the x86 merge making more and more useof common files, but also the big page attribute stuff is in andcaused a fair amount of churn, and while most of the issues shouldhave been very obvious and all got fixed, this is definitely one ofthose things that we want a lot of very wide testing of to make surenothing regressed; fair number of changes to things like the legacyIDE drivers too, and a totally new driver for the very common PCIEversion of the Intel e1000 network card etc; and I've probably totallyforgotten about tons of other stuff I should have mentioned, but thepoint is that not only do we have lots of new core, we do have a fairamout of changes to basic stuff that can actually affect perfectlybog-standard hardware setups So give it all a good testing"read morehttp://www.secuobs.com/revue/news/6744.shtmlhttp://www.secuobs.com/revue/news/6744.shtml Secuobs.com : 2008-02-12 13:39:19 - KernelTrap Kernel news - Patches for a much publicized Linux kernel local root exploit werereleased today as 26242, 262316, and 262218 The latest bug,labeled as CVE-2008-0600, was introduced by the vmsplice system calland added into the 26 kernel in 2617 It is the third in a seriesof root exploits surrounding the same system call, the two earlierbugs being CVE-2008-0009 and CVE-2008-0010 Easily obtained exploitsexist for both the older CVE-2008-0010 which affected the 2623 and2624 kernels, and the latest CVE-2008-0600, allowing a localnon-root user to gain root permissionsread morehttp://www.secuobs.com/revue/news/6743.shtmlhttp://www.secuobs.com/revue/news/6743.shtml Secuobs.com : 2008-02-08 15:27:39 - KernelTrap Kernel news - "The latest feature release GIT 154 is available at the usualplaces," began Git maintainer Junio Hamano He continued, "it has beenan unusually long cycle 5 months since the last feature release 153was really a bit too long But I hope it was worth waiting for Thankseverybody for working hard to improve it" He noted that there were165 contributers resulting in 684 changed files, included 70,435insertions and 28,984 deletionsThe Git distributed version control system was originally written byLinus Torvalds in April of 2005 to temporarily replace BitKeeper,which he had been using to manage kernel source code since February of2002 Junio Hamano took over maintainership of Git a few months later,in July of 2005, and the tool has long since become quite popularoutside of even Linux kernel development Regarding the latest stablerelease, Junio highlighted some of the changes, including:"Comes with much improved gitk, with i18n; comes with git-gui092 with i18n; progress displays from many commands are a lotnicer to the eye; rename detection of diff family while detectingexact matches has been greatly optimized; 'git diff' sometimes didnot quote paths with funny characters properly; various Perforceimporter updates; 'git clean' has been rewritten in C; 'git push'learned --dry-run option to show what would happen if a push isrun; 'cvs' is recognized as a synonym for 'git cvsserver', so thatCVS users can be switched to git just by changing their loginshell; 'git add -i' UI has been colorized; 'git commit' has beenrewritten in C; 'git bisect' learned 'skip' action to markuntestable commits; 'git svn' wasted way too much disk to recordrevision mappings between svn and git, a new representation thatis much more compact for this information has been introduced tocorrect this; in addition there are quite a few internalclean-ups"read morehttp://www.secuobs.com/revue/news/6106.shtmlhttp://www.secuobs.com/revue/news/6106.shtml Secuobs.com : 2008-02-07 11:19:07 - KernelTrap Kernel news - "Work continues to progress well but I've hit a few snags," notedMatthew Dillon, referring to the ongoing development of his HAMMERfilesystem He began by highlighting a number of problems with thecurrent design, then adding, "everything else now works, and workswell, including and most especially the historical access features"He continued:"I've come to the conclusion that I am going to have to make afairly radical change to the on-disk structures to solve theseproblems On the plus side, these changes will greatly simplifythe filesystem topology and greatly reduce its complexity On thenegative side, recovering space will not be instantaneous and willbasically require data to be copied from one part of thefilesystem to another"Matt detailed his solution, which included getting rid of thepreviously described clusters, super-clusters, A-lists, andper-cluster B-Tree's, "instead have just one global B-Tree for theentire filesystem, able to access any record anywhere in thefilesystem", adding that the filesystem would be implemented "as onebig huge circular FIFO, pretty much laid down linearly on disk, with aB-Tree to locate and access data" He detailed the many improvements,noting that this also makes it possible to provide efficient real-timemirroring He concluded, "it will probably take a week or two torewire the topology and another week or two to debug it Despite themassive rewiring, the new model is much, MUCH simpler then the old,and all the B-Tree code is retained just extended to operate acrossthe entire filesystem instead of just within a single cluster"read morehttp://www.secuobs.com/revue/news/5499.shtmlhttp://www.secuobs.com/revue/news/5499.shtml Secuobs.com : 2008-02-06 22:53:08 - KernelTrap Kernel news - "I wasn't planning on releasing v012 yet, and it was supposed to havesome initial support for multiple devices But, I have made a numberof performance fixes and small bug fixes, and I wanted to get them outthere before the destabilizing work on multiple-devices took over,"explained Chris Mason regarding the 012 release of his new btrfsfilesytem Btrfs was first announced in June of 2007, as analpha-quality filesystem offering checksumming of all files andmetadata, extent based file storage, efficient packing of small files,dynamic inode allocation, writable snapshots, object level mirroringand striping, and fast offline filesystem checks, among otherfeatures The project's website explains, "Linux has a wealth offilesystems to choose from, but we are facing a number of challengeswith scaling to the large storage subsystems that are becoming commonin today's data centers Filesystems need to scale in their ability toaddress and manage large storage, and also in their ability to detect,repair and tolerate errors in the data stored on disk" Regarding thelatest release, Chris offered:"So, here's v012 It comes with a shiny new disk format sorry,but the gain is dramatically better random writes to existingfiles In testing here, the random write phase of tiobench wentfrom 1MB/s to 30MB/s The fix was to change the way backreferences for file extents were hashed"read morehttp://www.secuobs.com/revue/news/5446.shtmlhttp://www.secuobs.com/revue/news/5446.shtml Secuobs.com : 2008-02-06 01:53:38 - KernelTrap Kernel news - It was recently pointed out that most of the x86 architecture patcheshad been merged into the mainline 2625 kernel, except for the kgdbpatches Linus Torvalds replied, "I won't even consider pulling itunless it's offered as a separate tree, not mixed up with otherthings At that point I can give a look" He continued:"That said, I explained to Ingo why I'm not particularlyinterested in it I don't think that 'developer-centric' debuggingis really even remotely our problem, and that I'm personally a lotmore interested in infrastructure that helps normal users givebetter bug-reports And kgdb isn't even _remotely_ it"So I'd merge a patch that puts oops information or the wholeconsole printout in the Intel management stuff in a heartbeatThat code is likely much grottier than any kgdb thing will ever beIntel really screwed up the interface and made it some insane XMLthing, but it's also fundamentally more important - if it meansthat normal users can give oops reports after they happened in Xor, these days, probably more commonly during suspend/resume andthe machine just died"read morehttp://www.secuobs.com/revue/news/5160.shtmlhttp://www.secuobs.com/revue/news/5160.shtml Secuobs.com : 2008-02-04 23:18:55 - KernelTrap Kernel news - Joseph Myers announced the availability of GCC 423 saying, "GCC423 is a bug-fix release, containing fixes for regressions in GCC422 relative to previous GCC releases" He adds, "as always, a vastnumber of people contributed to this GCC release -- far too many tothank individually"GCC is the GNU Compiler Collection which includes C, C++, Objective-C,Fortran, Java, and Ada compilers Download GCC 423 from your nearestgccgnuorg mirrorread morehttp://www.secuobs.com/revue/news/4947.shtmlhttp://www.secuobs.com/revue/news/4947.shtml Secuobs.com : 2008-02-03 09:28:02 - KernelTrap Kernel news - "I re-ran some statistics the other day on our kernel developmentrate, and changed my formula after Andrew accused me of severelyundercounting the rate of change," noted Greg KH during a discussionabout the stability of the Linux kernel while undergoing significantchanges He continued, "turns out that as of 2624-rc8 for the 2624kernel release we did: lines added per day, 4945; lines removed perday, 2006; lines modified per day, 1702" Greg continued:"And note, that is real stuff, not renames or file moves at all,git handles not reporting that That's for the 99 days that ittook to do 2624-rc8 I need to re-run the scripts now that2624 is out It's fricken scarily amazing that things are stillworking at all Just something to make you all sleep well atnight :"read morehttp://www.secuobs.com/revue/news/4764.shtmlhttp://www.secuobs.com/revue/news/4764.shtml Secuobs.com : 2008-02-01 22:47:02 - KernelTrap Kernel news - Avi Kivity summarized the kvm patches bound for the 2625 kernel:"Changes include performance and scalability improvements,completion of the portability work though no new architecturesare supported with this submission, support for new hardwarefeatures, using general userspace memory for kvm which enablesswapping guest memory as well as sharing memory among guests, aswell as the usual cleanups and incremental fixes"The Kernel-based Virtual Machine project, kvm, was started inmid-2006, and has been part of the Linux kernel since the 2620release in February of 2007 The recent changes can be browsed withgitwebread morehttp://www.secuobs.com/revue/news/4666.shtmlhttp://www.secuobs.com/revue/news/4666.shtml Secuobs.com : 2008-02-01 22:47:02 - KernelTrap Kernel news - Ingo Molnar summarized his pull request for changes to the x86architecture bound for mainline inclusion in 2625 noting, "it's nota small merge, it consists of 908 commits from 96 individual arch/x86developers " He continued, "a number of core files are changed aswell: most notably percpu, debugging details, timers, the firewireremote debugging patch and the KGDB remote debugging stub inkernel/kgdbc" He went on to detail the extent of the testing thistree has received, "in the past few weeks tens of thousands of randomx86git bzImages were successfully built and booted on a number ofcommodity 32-bit and64-bit testsystems - and there has been a fair amount of test exposureon -mm as well" Regarding the remote kernel debugger, Ingo explained:"We tested KGDB to be merge-worthy within the x86 architecturethe only supported architecture for now and it's better to havekernel/kgdbc than arch/x86/kernel/kgdbc The code is reasonablyclean and the user-space exposure is small - the only realexposure is the decades-old remote GDB protocol We are happy tofix up any further cleanliness comments that people might have -but we really wanted to start somewhere and get this thing movingAs an added bonus: finally a kernel debugger that can be readwithout puking too much ;- anyone remember KDB"read morehttp://www.secuobs.com/revue/news/4665.shtmlhttp://www.secuobs.com/revue/news/4665.shtml Secuobs.com : 2008-01-31 20:47:18 - KernelTrap Kernel news - "As you probably know there is a trend in enterprise computing towardsnetworked storage This is illustrated by the emergence during thepast few years of standards like SRP SCSI RDMA Protocol, iSCSIInternet SCSI and iSER iSCSI Extensions for RDMA," began Bart VanAssche, proposing that SCST be merged into the mainline kernel Hesuggested that while similar to the STGT project which has been partof the mainline kernel since 2620, "SCST is superior to STGT withrespect to features, performance, maturity, stability, and number ofexisting target drivers Unfortunately the SCST kernel code livesoutside the kernel tree, which makes SCST harder to use than STGT"SCSI subsystem maintainer, James Bottomley, was not convinced,explaining:"The two target architectures perform essentially identicalfunctions, so there's only really room for one in the kernelRight at the moment, it's STGT Problems in STGT come from theuserkernel boundary which can be mitigated in a variety ofways The fact that the figures are pretty much comparable on nonIB networks shows this I really need a whole lot more evidencethan at worst a 20% performance difference on IB to pull oneimplementation out and replace it with another Particularly asthere's no real evidence that STGT can't be tweaked to recover the20% even on IB"read morehttp://www.secuobs.com/revue/news/4456.shtmlhttp://www.secuobs.com/revue/news/4456.shtml Secuobs.com : 2008-01-30 03:55:42 - KernelTrap Kernel news - Prefacing a series of 196 patches, Greg KH noted, "due to the lowlevel nature of these patches, and because they touch so manydifferent parts of the kernel, a number of the subsystem maintainershave asked me to get them in first to make merging other trees easier"Linus Torvalds agreed and quickly merged the patches into his treeGreg summarized the many changes:"They can be broken down into these major areas: Documentationupdates language translations and fixes, as well as kobject andkset documentation updates; major kset/kobject/ktype rework andfixes; struct bus_type has been reworked to now handle thelifetime rules properly, as the kobject is properly dynamic;struct driver has also been reworked, and now the lifetime issuesare resolved; the block subsystem has been converted to use structdevice now, and not 'raw' kobjects; nozomi driver is added; lotsof class_device conversions to use struct device instead"read morehttp://www.secuobs.com/revue/news/4108.shtmlhttp://www.secuobs.com/revue/news/4108.shtml Secuobs.com : 2008-01-26 16:18:36 - KernelTrap Kernel news - Ingo Molnar posted a merge request for the latest git scheduler treesummarizing, "it contains various enhancements to the scheduler - findthe full shortlog is below 96 commits from 19 authors - schedulerdevelopers have been busy again :-/" He added, "the schedulingbehavior of the kernel to normal users should not change over v2624,but there are a good number of new features and enhancements under thehood" Ingo went on to list a number of these new features, including:"Various instrumentation and debugging enhancements from Arjan vande Ven; Peter Zijlstra's RT time limit and RT throttling code forthe RT scheduling class; Paul E McKenney's preemptible RCU code;refcount based CPU-hotplug rework by Gautham R Shenoy; there'sserious interest in running RT tasks on enterprise-class hardware,so Steven Rostedt and Gregory Haskins wrote a large number ofenhancements to the RT scheduling class and load-balancer; PeterZijlstra's high-resolution scheduler tick code; and a goodnumber of other, smaller enhancements"read morehttp://www.secuobs.com/revue/news/3448.shtmlhttp://www.secuobs.com/revue/news/3448.shtml Secuobs.com : 2008-01-24 17:10:03 - KernelTrap Kernel news - "I'm happy to announce that I've implemented a Block I/O bandwidthcontroller," began Ryo Tsuruta, explaining that it was intended to beused in a cgroup or virtual machine environment, implemented as adevice-mapper driver He detailed a token-based implementation inwhich dm-band passes out to the various groups, "a group passes on I/Orequests that its job issues to the underlying layer so long as it hastokens left, while requests are blocked if there aren't any tokensleft in the group One token is consumed each time the group passes ona request Dm-band will refill groups with tokens once all of groupsthat have requests on a given physical device use up their tokens"Ryo explained:"Dm-band is an I/O bandwidth controller implemented as adevice-mapper driver Several jobs using the same physical devicehave to share the bandwidth of the device Dm-band gives bandwidthto each job according to its weight, which each job can set itsown value to At this time, a job is a group of processes with thesame pid or pgrp or uid There is also a plan to make it supportcgroup A job can also be a virtual machine such as KVM or Xen"read morehttp://www.secuobs.com/revue/news/3051.shtmlhttp://www.secuobs.com/revue/news/3051.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - "The text below is mostly for the benefit of newbies - it's morealong the lines of 'how to get from a bug report to the source ofthe bug', with more details than normal," began Al Viro, offering afull review of another Linux kernel oops in an effort to educate morepeople on how this is done Al's walk through included a patch to fixthe bug that caused the oops He noted:"This might be worth doing on a more or less regular basis,especially if more people join the fun; everyone has their ownset of tricks in this area and making it easier to gather mighthelp a lot of people It's not just about oops-tracing per se, ofcourse - Arjan's site gives a nice collection of those, so thatmakes an obvious starting point"read morehttp://www.secuobs.com/revue/news/2959.shtmlhttp://www.secuobs.com/revue/news/2959.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - Michael Meeuwisse started Project VGA in September of 2007 Theproject aims to develop a simple, low budget, open source, VGAcompatible video card available this year Michael is also a member ofthe Open Graphic's Project, but started Project VGA in order to getsomething affordable on the market as soon as possibleIn this interview, Michael explains his inspiration for the projectand talks about the first development cards that will be functional bythe end of the month He details the costs involved in building thecards, as well as when the cards will be available for purchase andwhat they will be capable of doingread morehttp://www.secuobs.com/revue/news/2958.shtmlhttp://www.secuobs.com/revue/news/2958.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - Chris Mason announced version 010 of his new Btrfs filesystem,listing the following new features, "explicit back references, onlineresizing including shrinking, in place conversion from Ext3 toBtrfs, data=ordered support, mount options to disable data COW andchecksumming, and barrier support for sata and IDE drives" He notedthat the disk format in v010 has changed, and is not compatible withthe v09 disk format Regarding back reference support, Chrisexplained, "the core of this release is explicit back references forall metadata blocks, data extents, and directory items These are acrucial building block for future features such as online fsck andmigration between devices The back references are verified duringdeletes, and the extent back references are checked by the existingoffline fsck tool" He then detailed the new Ext3 to Btrfs conversionutility:"The conversion program uses the copy on write nature of Btrfs topreserve the original Ext3 FS, sharing the data blocks betweenBtrfs and Ext3 metadata Btrfs metadata is created inside the freespace of the Ext3 filesystem, and it is possible to either makethe conversion permanent reclaiming the space used by Ext3 orroll back the conversion to the original Ext3 filesystem"read morehttp://www.secuobs.com/revue/news/2956.shtmlhttp://www.secuobs.com/revue/news/2956.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - "'Const' has *never* been about the thing not being modified Forgetall that claptrap C does not have such a notion," began LinusTorvalds, responding to a query about why kfree takes a constpointer He continued, "'const' is a pointer type issue, and is meantto make certain mis-uses more visible at compile time It has *no*other meaning, and anybody who thinks it has is just setting himselfup for problems" He offered two explanations, beginning with simple Csemantics, "from a very obvious and very *real* caller perspective,'free' really doesn't change the thing the pointer points to Itdoes something totally different: it makes the *pointer* itselfinvalid" He then added his second reason, "anything that *can* take aconst pointer should always do so Why Because we want the types tobe as tight as possible, and normal code should need as few casts aspossible" When it was pointed out that GCC 42 displays warnings whencasting a const pointer to a non-const, Linus replied:"Either don't use a broken compiler casting a const pointer to anon-const is definitely not a bug, or cast to 'unsigned long' ifit still complains, now the compiler is not just stupid, it'sbroken The whole point of memory management is that we know howpointers work, and understand that they have a *bit*representation, not just the C semantics"read morehttp://www.secuobs.com/revue/news/2955.shtmlhttp://www.secuobs.com/revue/news/2955.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - "Slow servers, Skipping audio, Jerky video --everyone knows thesymptoms of latency But to know what's really going on in the system,what's causing the latency, and how to fix it those are difficultquestions without good answers right now," began Arjan van de Van,announcing version 01 of LatencyTop, "a tool for developers tovisualize system latencies" He continued:"LatencyTOP is a Linux tool for software developers both kerneland userspace, aimed at identifying where system latency occurs,and what kind of operation/action is causing the latency tohappen By identifying this, developers can then change the codeto avoid the worst latency hiccups"There are many types and causes of latency, and LatencyTOP focuson type that causes audio skipping and desktop stuttersSpecifically, LatencyTOP focuses on the cases where theapplications want to run and execute useful code, but there's someresource that's not currently available and the kernel thenblocks the process This is done both on a system level and on aper process level, so that you can see what's happening to thesystem, and which process is suffering and/or causing the delays"read morehttp://www.secuobs.com/revue/news/2954.shtmlhttp://www.secuobs.com/revue/news/2954.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - "Yes, I know another tree, just what everyone wants," quippedJames Bottomley, announcing his new merge candidate -mc tree:"This one has a specific purpose: It's my tree tracking everyoneelse's git and quilt trees so I get early warning if there aregoing to be any merge issues However, it struck me it might beuseful to anyone wishing to track what's going upstream moreclosely"James noted that his new tree is available in git, and beingautomatically built each night "As you can see from the reverts andthe skips, we have trouble even now and that's after I fixed up mostof the failures in SCSI ACPI and the x86 trees clash hideously, so Ikicked out x86 Jens' block tree has two patches which clash withBart's ide quilt Greg actually has one patch in his tree that clasheswith one of mine" He also noted, "this tree is currently very storagecentric ie I haven't included net trees or quilts because I didn'tthink they'd be likely to clash with my SCSI trees However, if itcould be more generally useful, I could add other trees and quilts toit"read morehttp://www.secuobs.com/revue/news/2953.shtmlhttp://www.secuobs.com/revue/news/2953.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - The final 2624 Linux kernel is expected any day now, so the varioussubsystem maintainers have begun summarizing what changes are expectedto be merged into the mainline kernel during the 2625 merge windowIngo Molnar spoke to changes for the x86 architecture, "there are 763commits in x86git so far, from more than 90 contributors, so it wouldbe difficult to mention and credit every contribution in this mail"Along with a lengthy list of other changes, he included:"Continued, intense arch/x86 unification and cleanup work by lotsof people; FIFO ticket spinlocks for better spinlock scalability;'regset' generalizations - the most important step towards utracesupport ==next-gen ptrace; support for more than 255 CPUs up to4096 - in theory up to 65535; almost complete 64-bit paravirtguest support; KGDB support on x86, finally"read morehttp://www.secuobs.com/revue/news/2952.shtmlhttp://www.secuobs.com/revue/news/2952.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - "The following patches have been in the -mm tree for a while, and Iplan to push them to Linus when the 2625 merge window opens," beganTheodore Ts'o, offering the patches for review before they are mergedHe explained that the patches introduce some of the final changes tothe ext4 on-disk format, "ext4, shouldn't be deployed to productionsystems yet, although we do salute those who are willing to be guineapigs and play with this code" He continued:"With this patch series, it is expected that the ext4 formatshould be settling down We still have delayed allocation andonline defrag which aren't quite ready to merge, but thoseshouldn't affect the on-disk format I don't expect any otheron-disk format changes to show up after this point, but I've beenwrong before any such changes would have to have a Really GoodReason, though"read morehttp://www.secuobs.com/revue/news/2951.shtmlhttp://www.secuobs.com/revue/news/2951.shtml Secuobs.com : 2008-01-24 15:42:16 - KernelTrap Kernel news - "I'd like to send a small update on my progress on the PerformanceTracker project," noted Erik Cederstrand on the FreeBSD -currentmailing list He continued, "I now have a small setup of a server anda slave chugging along, currently collecting data I'm followingCURRENT and collecting results from super-smack and unixbench" Theproject performs regular benchmarks of the FreeBSD -current sourcetree using Unixbench and Super Smack, allowing you to chart theresults over time Erik highlighted an example of a visible change inperformance when the generic kernel moved from the 4BSD scheduler tothe ULE scheduler on October 19th, 2007Kris Kennaway responded favorably, then noted, "one suggestion I haveis that as more metrics are added it becomes important for an 'at aglance; overview of changes so we can monitor for performanceimprovements and regressions among many workloads" He went on tosuggest, "at some point the ability to annotate the data will becomeimportant eg 'We understand the cause of this, it was r1123 offooc, which was corrected in r1124 The developer responsible hasbeen shot"" Erik agreed with both recommendations, and noted that hewould continue to work in that directionread morehttp://www.secuobs.com/revue/news/2950.shtmlhttp://www.secuobs.com/revue/news/2950.shtml