this presentation is on the recentdevelopments with the device driversthat are Kanoa is working on some of youmay have already heard some of thisinformation but what I did is I try toinclude anything that happens since thelast warps back to Europe so let's getgoing and see if we can create somecontroversy there's a lot of confusiongoing on around your systems BIOS whatit does and what it what uses it andwhat doesn't use it and whether or notit's going away well it's not going awaythe only thing that's changing is theinterface to it so whatever you call itUEFI BIOS or whatever it'll always bethere because it's what runs when youturn on the power switch so to avoid allthis confusion and to talk about allthis stuff that the vendor supplies withyour computer I'm just going to call itvendor supplied firmware so it's what weused to call the BIOS but itincorporates everything like the BIOSthe ACPI some startup codemaybe it's UEFI maybe it's not maybe ithas a CSM maybe it doesn't the point isfor most things it just doesn't matterthe ACPI PSD for example doesn't use ordepend on any of the old-style bio stuffit does depend on a CPI however which isin there so anyway just wanted to bringthat up because I'll be mentioning thatand you'll be hearing about that inother in other places let's move on tothe a CPI driver project it's currentversion is 3.23 11 most of the recentchanges have been to add features or fixproblems that are elsewhere in thesystem you know even though the PSD ispart of the a CPI project driving the aCPI is only a small part of what the PSDdoes it configures the system it fixesup the PCI spacebased on you know what thevendor-supplied firmware did or didn'tdo it patches the kernelit adds system enhancements it does abunch of things and we use the PSD to dothese things because it is strategicallylocated in the system and it loads veryearly in the boot process it'sessentially a kernel plug-in and itoperates at roughly the same levels asthe kernel so it can do pretty muchwhatever the kernel does so some of therecent changes have been to fix thekernel ipi timeout issue some of you mayhave already known about this it wasreported originally as the fullscreenhanging problem even though it reallyhad absolutely nothing to do withfullscreen sessions it actually wasa lock-up that occurred when the kernelwas doing inter-process communicationsso anyway that was one of the thingsthat got fixed the the things thatdetermined it was really fast CPUs inother words greater than 3 gigahertz andhaving more than one of them the otherthing that got fixed was some GPE stormssome of the vendors supplied firmwarewould get into loops generating constantinterrupts to the system which wouldbasically hang the computer so I putworkarounds for that in as well andthere have been other enhancements tohandle other problems with variousfirmware supplied by vendors and ofcourse we always get the obligatory aCPI CA updates from Intel there havealso been some recent changes in thedaemon particularly involving powermanagement I changed the algorithm andhow thethe power manager determines whether theCPU is busy or not and it's now a lotmore accurate and it also handles ifother applications try and turn the CPUson and off via that via the kernel APIso it detects those and deal with dealswith it accordingly Multi Mac has seensome work done on it the current releaseis from November 2006 2018 and itincludes all the drivers version 1.1point 4 and the RL mm RL is version zeropoint one point three obviously the themulti Mac release is a bundle of a bunchof different drivers and each driver hastheir own version so when you reportthings make sure you report whichversion of the driver using not the dateof the bundle one of the things that didget enhanced was memory management stuffmainly to handle a buffer buffer stuffthat transfers between the driver andthe kernel some of these changes werealso necessary to accommodate some ofthe newer drivers that got added thereis also some better intermodules I am[Music]acquainted things that the freebsddrivers do is they link various modulestogether as they load and this was alsoneeded for some of the more newerdrivers that got added to the pack and Ifixed a packet transmit deadlock issuethere was under some very rareconditions on multi CPU systems wherethe driver route would just stop becausewell there was a deadlock situation whentransmitting packets that got fixed andthere were improvements the interrupthandlers and the interrupt handler isone of the major difference betweenthe way FreeBSD does things in the wayos/2 does things and just in caseanybody didn't already know all themultimatic drivers are essentiallyrecompiled freebsd drivers they're notas exactly ported they're simplycompiled to run on os/2 and whateverminor changes are needed to make themrun on os/2 that's that is done and oneof those minor changes is the interrupthandler any questions on multi Mac yesin the back yeah there is one from KeithI'm going over here for the headset soyou can understand him hi Keith hereyou said multi CPU do you meanmulti-core or multi CPU either waythere's more than one CPU running Iconsider it multi CPU okay it doesn'tmatter whether they're all in the samechip or if they're multiple chipsdoesn't matter at all okay clarifythanks you're welcome okay moving on toUSB the current version is 1201 you'llbe seeing this really soon it has beenreally thoroughly tested and it is theversion that's going to be in ARCA oh sfive point zero point four do out anytime now I understand version 12 has all32 bit drivers except for audio and MSDif those two are either planned or inprogressto be converted to 32 bit drivers theversion 12 stack is significantlydifferent than the old 16 bits tax theyhave faster code executionmainly because you get a lot more workdone with a 32-bit instruction thanmultiple 16-bit instructions largeportions of the code has been completelyrewritten or just gotten rid of becauseit's been replaced by common code in mydrive 32 library it has better memorymanagement than the 16-bit drivers didand it has better faster inter drivercommunications doesn't require anythunking basically direct into drivercalls all of that means that there's alot less overhead the bottom line isthat although the data transfer isn'tany different between you know acrossthe USB bus because we all know the USBbus is what it is these drivers needless CPU time and less basically justless computer resources to execute soyou'll see less less CPU load whenrunning them here's a picture of how theUSB stuff kind of fits together at thebottom we have all of the hostcontroller drivers the OAH CDU hcd hcdand x HDD drivers these are whatactually talked to the hardware and allof those moving up the ladder talk tothe USB D which is the USB driver thatis the major interface to all of theother class drivers on the left-handside you'll see the human interfacedrivers the mouse keyboard in the hidand you'll notice the dotted box aroundthe the lower section the the hostcontroller drivers in USB D all of thoseare a tightly coupled matched set youcannot interchange any of those driverswith any other version also the hid hasa little dotted box around it those areall a tightly coupled match that youcannot swap out any of those with otherversions either and in addition thosetwo dotted boxes all have to be the 32bitversion 12 drivers you cannot swap outany of those with other versions thenyou've got the class drivers up in theupper right-hand side the printer driverthe comm driver and the USB calls driverthose are all 32-bit drivers now andthey all interface directly to the USB Ddriver through the the fast 32-bitdirect call interface the two kind oforphans are in pink MSD and audio theytalked to us BD through this little 16bit compatibility wort attacked on theside of USB D just so that they arestill able to be used it's not quite asfast as the other stuff but it stilldoes work so just kind of wanted to giveyou an idea of how all the USB stufffits together now let's talk aboutsomething that arab that's oneverybody's mind upcoming versions thecurrently release test version 2 thetest team is version 12.04 USB 3 it iscurrently working and being tested itcurrently works on every single system ihave ever tried it on however i haveheard that there are some issues withcompatibility with certain intelchipsets leave it to Intel to beincompatible that's kind of usually theway it is until Intel Harbor alwaystakes the most amount of workarounds andcomplex code to get them working theother other thing that is still notimplemented is isochronous support andthe reason that's still not done isbecause I've been trying to figure outwhat's wrong with the Intel hardware ohand one other thing sometimes errorrecovery doesn't quite work properly soI'm still working on that but other thanthat the USB 3 stuff is built it existsand it works as soon as I figure out theIntel problem it'll probably go intofinal beta test and thenyou should be able to see it okaydisplay drivers let's talk aboutpanorama for a second the currentversion is 1.13 it has had a few changesthere was a compatibility problem with avery rare thing with certain firmwaresfrom banders caused hangs in the driverI've managed to track that down and fixit that took some time there have beensome enhancement to the screen objectthis doesn't affect the drivernecessarily but it's some you know it'sit's when you open the screen object youcan see you can change some of thecustom resolution settings and some ofthe other things that's been enhanced alittle bit and one of the major things Idid was I enhanced the protect modeinterface and this gave it a lot morecompatibility with aval more variety ofsystems the the protect mode interfacethat was provided by IBM was kind ofcomplex and didn't always work in allsituations so I kind of modified that alittle bit and it gave us a better abetter system that's more compatible andthe other thing I did is I enhanced theedid capability or compatibility so thatit'll recognize some of the newermonitors it used to only detect up toversion 1.3 and now we have 1.4 so itwill detect though so it can get theresolution of those monitors as well oneof the other things involving displaydrivers is the Grad sis driver a lot oftimes this gets forgotten because ittends to just work however there havebeen a few enhancements to that as wellthe gradua is used in all the Graddosages and all of the Grad driverspanorama snap gen grab all of them soit's kind ofcommon to them all the current versionis 2.0 1 and version 2.0 2 is inprogressversion 2.0 1 fixed a trap if the systemdidn't even have any MTR RS and by theway that's one of the major things thatthe gratzes driver is responsible for issetting the MTR ARS in the system forvideo performance it turned out thereare some systems out there that don'thave any MTR artha it was discoveredthat there was a defect in there thatwould cause a trap if that didn't haveany the other thing that is currentlybeing worked on is apparently some ofthe newer firmware in some of thesesystems is already setting rightcombining for the display memory so thevideo driver doesn't have to do itanymore the newer version which is notreleased yet will detect if this hasalready happened and then not try and doit again interesting interesting thingspeople are doing all these vendors aredoing with their firmware these days anyquestions about display drivers - yep weare now seeing more and more high resdisplays and the older drivers onlysupported roughly 2 K by 1.5 K or all if2 monitors were combined you could getto full HD screen side-by-side willkatara pass this and support 4k displaysor what you have to reduce it somehowwell panorama is a visa based driver itdoesn't really have any limits orrequirements other than what is imposedby the VESA BIOS so it will be able todo whatever the VESA BIOS does that's inthe particular hardware I know thatdoesn't really answer a question verywell but that's really the real answerpanorama does not limit you to anyparticular size of screen and it doesnot limit you to any number of monitorsit's whatever the BIOS can do that'sthat's what pammer is for panorama isfor if you need multi head displays andbigger bigger displays than what theBIOS can handle if there is a bios ofvideo bios anymorethen you'll need actually a hardwaredrive or something like snap that makessense okay moving on disc drivers theAHCI drivercurrent version is 2.0 six and whatchanged in that since the last warpstock was the ioctl interface where thesmart utilities got fixed thatapparently wasn't working I removed somestartup restrictions primarily becausesome firmware didn't set up things quiteright even though the driver would stillwork it was detecting certain problemsin the PCI configuration so I justdecided to ignore those things go aheadand load and run anyway it the majorthing that got added in this most recentversion is a feature to ignore unusabledisks and what this feature does is itreads the first sector on the disk andit tries to determine if it's an MBRstyle disk or if it's been wiped andthose are theonly two conditions which cause it to bea useable disc it must have an MBR diskstructure or it must be wiped otherwiseit'll just ignore the disk and what thisis for is to prevent accidental damageto non MBR disks like GPT disks andstuff like that and also I cleaned upsome of the debug output messages thatyou couldn't do you might have seen insome of the test log outputs that'sgonna be cleaner now at this version anyquestions on AHCI so very good all thesestuff I've been working on the test blogprogram is obviously a very importantthing it provides the developer with acomprehensive log of your system it'sactually pretty invasive it goes andqueries all kinds of stuff and prettymuch gives the developer a fullinventory of everything in your systemhow it's working what's working what'snot workingthere have been enhancements andupgrades to it so make sure you alwaysuse the latest version we always saythat but it's never hurts to remindeverybody but one thing that I stillseem people seem to forget is thatrunning it clears all the debug bufferswhich means you only can run it once perboot because if you run it once just tokind of see if it works and then run ita second time to actually produce thelog well you lost all the debug data andit really is not very useful and theother thing that you always want to makesure you can do if possible is run thetest log program when the problemactually exists if you can't I meanobviously if you get a trap or somethingyou can't but you know if like thenetwork has stopped well run it when thenetwork is stopped that way it cancapture the state of the system whenwhen the problem actually existsanyway so that about covers the driversstuff that I've been working on but howmany people would like to see moredriver development at work nobody huhone person in the back two people threepeople there we go four or five yeahlots of people board driver developmentwould be good because there's a lot ofprojects that we want to work on andwant to get done and a lot of work to bedone however here's how my time getsused up handling tickets and otherproblems which is not just dealing withtickets but problems get reported andneed to be fixed or things come updifferent Hardware comes up and itdoesn't work and that needs to be fixedor enhancements and things that's sortof the big time consumer however thereare parts of tickets that do take up alot of time when people don't read thebest practices page the other thing thattakes up my time is you know preparingwarp stock presentations reading emailsmeetings you know other things likegetting the next version of our Co OSout and stuff like that and then we havecustomer support not core not customersfor a corporate customer support andthis thing is important becausecorporate customers maintain always torelevance and use in the world and italso provides income for our Ginolaand then after all of that stuff is donewhatever times leftover gets intowriting X HCI or disk drivers or youknow Wireless drivers and all of thatnow do the math here real quick addthose up sometimes which is partly whyyou don't see a whole lot of reallyrapid progress on some things what youcan do to help this what people can doto help this a little bit is if you haveproblems read the docsand the read means and the wiki's and doas much of the work as you can beforeopening a ticketmake sure your problems are reproducibleread the best practices page you know ifthere's not a whole lot of developersdeveloping for os/2 these days so if wehave to spend a lot of time readingtickets and in searching for Oh where'sthe test log file or or where's where isthe trap screen or where is this sowhere's that and then I have to ask forit and then wait back for a reply thatjust wastes our time so try and do asmuch of the work as you can to help thedevelopers out as much as possible andit really will save a lot of time anywaythat's all I have for you guys today Iwill I can take some questions if youguys like ah the fate of wireless LAN isstill on the list to be done we figuredpeople would want USB 3 first and thenthere's this thing called nvm EE whichis probably going to be pretty importantso yeah it's on the listif anybody one has them some goodcloning technology and maybe that wouldworkwhat's that somebody he did for thevoting system so we can vote what getsdone first Wi-Fi usually free and via mewhatever or cloning clone you yeah wellwe try and determine what comes nextbased on what's really important likefor example if we if if we're startingto see systems being developed withfirmwarethat just won't boot anymore then wekind of need to address that first andthat may involve PSD development orsomething like that or in the case ofthe UEFI stuff it might involve thatbecause we figure something likewireless support doesn't even come intoplay if your system won't boot and thenwhat if we start suddenly have systemsthat have disks that don't have a HDIanymore they've all got nvme well if youcan't boot your system yet again thenWi-Fi doesn't matter either and we gotto deal with the disk stuff so and thenwe've got you know corporate customersand things that actually pay a bunch ofmoney for licenses and things and whattheir priorities are matters is as wellwhat gets worked on first so you know wetry and we try and prioritize things thebest as possible assuming thateverything can't get done all at once ifwe had a hundred developers yeaheverything could get done all at oncebut we don't alright guys thanks a lotwish I could have been there but I guessthis is the next best thing all right[Applause]