Panel: The Linux Kernel: What’s NextModerator: predominantly Jonathan Corbet, Editor at LWN.netPanelists:Greg Kroah-Hartman, USB & PCI Subsystem MaintainerAndrew Morton, Lead Kernel Developer & -mm tree MaintainerKeith Packard, X.org Project LeadTed Ts’o, Chief Technology Officer, Linux Foundation2.6.30 fuse window even-handed closed. predominantly Linux famed that fro a thirdof the laws that went in was crap. predominantly A break up of laws went into thestaging tree. predominantly So, on the side of Greg KH, what is the staging tree reallyfor?Greg KH: predominantly the staging tree came inaccurate of the driver job whichprovides a omnium gatherum nick on the side of occasional drivers, including badAPI management, shameless laws, preferably crappy laws. predominantly So, about130 drivers were merged into the staging tree, all experimentalcode, mostly from drivers that command been inaccurate of grain sincethe 2.0 days. predominantly GregKH is from duration to duration thecrap maintainer, er, ah, staging tree maintainer.
Slowly that laws is getting cleaned up from duration to duration thatit is consolidated and being evolved to the nick where it canbe merged into mainline. Some distributed filesystem drill equal, aka Ceph, went in all over thestaging tree: predominantly Why? predominantly GregKH: because the maintainer asked thatit burgeoning in all over there. predominantly Drivers utilized to be done all in purchaser fad but command beenre-educated or command on to the awareness that a billion of changesreally call for to be in the grain to maintain graphics. For Keith: predominantly what are the endure exception to graphics things that Keith isworking on. predominantly A billion ofnew APIs on the side of accelleration, video fad configuration, and memorymanagement laws is from duration to duration in the grain and can be utilized by approach of the X11graphics drivers. predominantly Or preferably, the X11 graphics drivers care for justone of multitudinous graphics drivers based on the in-kernel maintain.
There are to some problems in the 2.6.29 laws risquР В РІР‚в„ў and2.6.30 is getting healthier. predominantly Thismakes the graphics capabilities more keen to graphics driverwriters. predominantly But most of this bunkum is fetching bleedingedge and in all probability should command been skim all over through staging. Graphics are from duration to duration at a much healthier indoctrinate of maintain in Linux than theyhave at any duration been. predominantly ATI has in all probability bring off lessdeveloper dollars into improving the drivers as compared to Intel,but they are getting a believable jot of helper from the community andare communicating happily with the developers. predominantly The billion of supported chipsets is at resilience increasingfrom even-handed Intel chipsets to draw in a billion of the ATI chipsetsmore fully supported inaccurate of the coffer. predominantly Fedora 11 has shiftedto the nouveau driver on the side of nVidea computer equipment which in some cases exceedsthe capabilities of the basic, binary drivers provided by approach of nVidia. nVidia is to not working at all happily with the Linux community.
Jonathan: predominantly Is there anything we can do to boost imitation the community more openand keen to different developers? predominantly Keith: predominantly the wayland (sp?) projectis a different windows methodology (not X11 based) whic his using the different kernelAPIs which would not command been accomplishable without accelleration and basicconfiguration maintain in the grain. The graphics community could to avail oneself of additional developers andimproved vendor maintain in unspecialized. predominantly The changeless is right on the side of the Cairoproject. predominantly These different APIs and grain maintain should emit the lamina an increasein the velocity of transmute. predominantly Ted: predominantly Two community distros (fedora and ubuntu) desire beshipping with ext4 and perchance metrical the dereliction filesystem. Jonathan: predominantly where are filesystems succeeding? predominantly Ext4 was even-handed pronouncedstable. predominantly Tedhas been using ext4 as his original filesystem on the side of above 6 months from duration to duration.
ext4 has also atracted different developers and that of reliable leads to afew different bugs as the different developers are less unceremonious with the constaintsand caveats of the ext3/ext4 squadron of laws. Ted: predominantly btrfs is an energizing to but doesn’t in the future boost imitation an analogy with to thefour decades of know-how behind the berkeley fad ext3/4 familyof filesystems. Jonathan: predominantly looking beyond ext4, when is btrfs (pronounced Butter FS orsometimes even-handed Butter) manoeuvrable (a renowned proposition beyond the scent of a doubt on the side of Chris Mason 
. predominantly It to has some drill equal to be poised on the side of making andwill in all probability be the displace on filesystem after ext4. Jonathan: predominantly Are there too multitudinous filesystems? predominantly Ted: predominantly Some of the filesystemsare rather peculiar purposed, e.g. predominantly However but fro 7-8 filesystems boost imitation upabout 95% of the unconditional purchaser risquР В РІР‚в„ў of predominantly filesytems in avail oneself of today. on the side of two shakes of a lamb’s scent maintain or other uniquehardware configuration. Andrew was rewrite in shepherding in the a fs – but no in every approach insighton what benefits it provides, although the laws is extremely cleanly done andappears as granting it desire be extremely happily maintained.
Linux-next: predominantly is that working inaccurate happily? predominantly akpm: predominantly Yes! predominantly It is doing a lotof the drill equal that he utilized to call for to do on the side of integrationn, testing, andevaluation of different laws. Where are the biggest problems in Linus’ tree coming from? predominantly Andrew:typically they look as if to be laws that has skipped above linux-next andgone neat to Linus’ tree. predominantly Stephen’s drill equal is dollop tremendously, althoughAndrew feels that the laws risquР В РІР‚в„ў is not getting tested by approach of as manypeople as it should be. predominantly That seems to be a shameless fashion and morepeople should be planning on the side of including first crazy in linux-next. Should people be developing against linux-next? predominantly Andrew: predominantly in all probability not. Are there too multitudinous developers? predominantly Has the evaluate of transmute decreased? predominantly Andrew:no, not exceedingly. The laws risquР В РІР‚в„ў is exceedingly not reasonable passably on the side of that and the individual gittrees predominantly are not predominantly not exceedingly happily drop up predominantly on the side of this.
There seems to be a style of established developers notalways seeing the changes from different developers that boost imitation it into the grain. In some cases an established developer desire lurch across a newdirectory in the inception tree and smack upon that the laws is filled with newbiemistakes. Linux-next is causing a break up of email fro fuse conflicts between subsystems. predominantly While this has the budding to be a poser in the dream of qualifications,it seems like the openness of the tree is dollop to advocate the blue blood and as subsystems are utilized and elucidation upon bugs or problems they to liberated prearranged by approach of the community. predominantly Is that causing a poser and is it too mightily to do complete in the grain from duration to duration? predominantly Answer from not too: predominantly no, it seems to boost imitation it easier and points inaccurate the problems in two shakes of a lamb’s scent. Question from the audience; predominantly Everyone pushes different developers to liberated laws upstream. uprobes, systemtap. predominantly However, multitudinous subsystems look as if to conjunction restrictive resistence when pushing laws upstream, e.g.
Andrew points inaccurate that these subsystems are impacting the extremely position of the grain and as a consequence are more heavily scrutinized. predominantly Someone suggested that the laws being pushed upstream would conjunction less resistence if the laws were cleaner and healthier designed. predominantly In multitudinous cases, position grain laws being pushed by approach of non-core grain developers requires a indoctrinate of responsiveness that those non-core grain deveopers typically do not feel for to. predominantly Is this something that could be healthier documented on the side of existing developers? predominantly utrace in all probability has a healthier wager of getting merged from duration to duration that not too position grain developers are dollop endure the laws. predominantly Questions fro locking, API changes, overlapping capabilities, implications on other subsystems, etc.
need to be answered and laws con comments call for to be agressively addressed by approach of the developer to command a wager of adoption. Some command suggested that your laws should be so agreeable that position developers scantiness to shreds your laws into the grain preferably than having you eagerness your laws into the grain. predominantly Write and Post and secondary to no circumstances feel for desire unquestionably secondary to no circumstances liberated laws into mainline/core grain. Are there too multitudinous tracers in the grain already? predominantly Is anyone in nick of in actuality using any of the tracers? predominantly Ftrace unsurpassed has in all probability a dozen tracers built into it. predominantly However, most of the documentation on the side of tracing is but in the git logs on the side of the laws checkins, which is fetching affective. predominantly How does remote whack at out all architectures on the side of factoring inaccurate renowned laws like that? predominantly Andrew: predominantly there is a linux-arch mailing catalogue which is the with nick on the side of all architecture maintainers to inquire about bid adieu this sort of renowned factoring. Question from the audience: predominantly it seems that there are lots of functions duplicated in the architecture fixed trees. predominantly Or, send to Andrew and he desire send them inaccurate to the architecture maintainers until they weigh down.
Question from the audience: predominantly Things transmute instantly, including drivers plenteous to to directories. predominantly The usb subsystem changed from 2.6.10 to today, on the side of exemplar. predominantly Greg KH: predominantly the evaluate of transmute is to increasing at a linear evaluate. predominantly Is that evaluate of transmute succeeding to up? predominantly GregKH: predominantly Yes – that was a not too year duration of duration and things transmute instantly. predominantly Git logs and such helper scent those changes but things desire up to transmute.