考川大,操作系統(tǒng)重點(diǎn)打印版(一卷再手,os不愁)

1、進(jìn)程：線程：調(diào)度隊列的種類：1.jobqueue–setofallprocessesinthesystem2.readyqueue–setofallprocessesresidinginmainmemyreadywaitingtoexecutegenerallystedasalinkedlist3.devicequeues–setofprocesseswaitingfaparticularIOdeviceeg.atapedriverad

2、isk進(jìn)程狀態(tài)：1.new:beingcreated2.running:instructionsarebeingexecuted3.waiting:waitingfsomeeventtooccur4.ready:waitingtobeassignedtoaprocess5.terminated:hasfinishedexecutionFCB的內(nèi)容：1.Processstate2.Programcounter3.CPUregisters4

3、.CPUschedulinginfmation5.Memymanagementinfmation6.Accountinginfmation7.IOstatusinfmation8.pagetablerelocationregisterlimitregister9.fileopentable調(diào)度器scheduler分類：longtermscheduler(jobscheduler)–swhichprocessesshouldbeloade

4、dfexecutionshttermscheduler(CPUscheduler)–swhichprocessshouldbeexecutednextallocatesCPU生產(chǎn)者消費(fèi)者問題（指利用了n1個單元）shareddata#defineBUFFER_SIZE10typedefstruct...itemitembuffer[BUFFER_SIZE]intin=0intout=0itemnextProducedlocalvaria

5、blewhile(1)produceaniteminnextProducedwhile(((in1)%BUFFER_SIZE)==out)buffer[in]=nextProducedin=(in1)%BUFFER_SIZEitemnextConsumedlocalvariablewhile(1)while(in==out)bufferisemptynextConsumed=buffer[out]out=(out1)%BUFFER_SI

6、ZEconsumetheiteminnextConsumed用戶線程和內(nèi)核線程Userthreads：supptedabovethekernelimplementedbyathreadlibraryattheuserlevelthelibrarysupptthreadcreationschedulingmanagementwithnosupptfromthekernel(onasinglethreadkernel)anyuserleve

7、lthreadperfmsablockingsystemcallblocksthewholeprocessKernelthreads：suppteddirectlybytheoperatingsystem:thekernelperfmsthreadcreationschedulingmanagementinkernelspaceslowertocreatemanagethanuserthreadsindependentblockamon

8、gthreadsgoodsupptfmultiprocesssManytoone：manyuserlevelthreadsmappedtosinglekernelthreadthreadmanagementisdoneinuserspace，usedonsystemsthatdonotsupptkernelthreadsdrawbacks：badblockingbad，utilizationofmultiprocesssonetoone

9、：eachuserlevelthreadmapstoakernelthread：meconcurrencythanmanytoone，supptmultiprocesssdrawback：overheadofcreatingkernelthreadsexamples：Windows9598NT2000，OS2manytomany：multiplexesmanyuserlevelthreadstoasmallerequalnumberof

10、kernelthreadsallowstheprogrammertocreateasufficientnumberofuserthreadsavoidbadblockingsupptmultiprocesssExamples：Solaris2，WindowsNT2000withtheThreadFiberpackage線程的終止：threadcancellation?targetthread:thethreadtobecancelled

11、?asynchronouscancellation:onethreadimmediatelyterminatesthetargetthread，maybecancelledinthedleofupdatingdatasharedwithotherthreads，maynotfreeasystemwideresource?deferredcancellation:thetargetthreadcanperiodicallycheckifi

12、tshouldterminate(atsocalledcancellationpointsinPthread)同步信號和異步信號：synchronoussignalsanillegalmemyaccessadivisionbyzerodeliveredtothesameprocessthatcausethesignalasynchronoussignalsauserkeystroke(CtrlC)atimerexpirationtypi

13、callysenttoanotherprocess線程池：Whythethreadpools1.avoidcreationterminationoverheadsothatitisfastertoservicearequest2.putboundonnumberofthreadsthuslimitCPUmemyusageconsiderationsnumberofCPUsamountofphysicalmemyexpectednumbe

14、rofconcurrentrequests進(jìn)程同步種類：1.blocking(synchronous)versusnonblocking(asynchronous)2.blockingsend:thesendingprocessisblockeduntilthemessageisreceivedbythereceivingprocessbythemailbox3.nonblockingsend:thesendingprocesssend

15、sthemessageresumesoperation4.blockingreceive:thereceiverblocksuntilamessageisavailable5.nonblockingreceive:thereceiverretrieveseitheravalidmessageanullresumesoperationCPU調(diào)度算法1.FCFS：等待時間不穩(wěn)定，響應(yīng)時間不穩(wěn)定。特點(diǎn)：easytoundersteasytoi

16、mplement(anFIFOqueue)largevarianceofwaitingresponsetime.convoyeffect.nonpreemptivebadftimesharingsystem2.SJF：分為preemptivetarget=truereturnrv進(jìn)程共享變量booleanlock=false對于第i個進(jìn)程dowhile(TestSet(lock))CriticalSectionlock=falseRem

17、ainderSectionwhile(1)分析：滿足了互斥，但不滿足有限等待條件。假設(shè)系統(tǒng)調(diào)度不幸使得某個號數(shù)的進(jìn)程每次都如下述情況運(yùn)行則這個號數(shù)的進(jìn)程會一直霸占進(jìn)入臨界區(qū)的機(jī)會，使得其他號數(shù)的進(jìn)程無限等待。lock=falseRemainderSectionwhile(1)dowhile(TestSet(lock))CriticalSection死鎖檢測算法：1.LetWkFinishbevectsoflengthmnrespecti

18、vely.Initialize:(a)Wk=Available(b)Fi=12…nifAllocationi?0thenFinish[i]=falseotherwiseFinish[i]=true.2.Findanindexisuchthatboth:(a)Finish[i]==false(b)Requesti?WkIfnosuchiexistsgotostep4.3.Wk=WkAllocationiFinish[i]=truegoto

19、step2.4.IfFinish[i]==falsefsomei1?i?nthenthesystemisindeadlockstateMeoverifFinish[i]==falsethenPiisdeadlocked.ThealgithmrequiresanderofO(mxn2)operationstodetectwhetherthesystemisindeadlockedstate地址綁定：addressesintheprogra

20、maregenerallysymbolicacompilerbindthesesymbolicaddressestorelocatableaddressesalinkageedit(loader)bindtheserelocatableaddressestoabsoluteaddresses1.Compiletime:Ifmemylocationknownapriiabsolutecodecanbegeneratedmustrecomp

21、ilecodeifstartinglocationchanges.2.Loadtime:Mustgeneraterelocatablecodeifmemylocationisnotknownatcompiletime.3.Executiontime:Bindingdelayeduntilruntimeiftheprocesscanbemovedduringitsexecutionfromonememysegmenttoanother.N

22、eedhardwaresupptfaddressmaps(e.g.baselimitregisters).邏輯地址于物理地址：theconceptofalogicaladdressspacethatisboundtoaseparatephysicaladdressspaceiscentraltopropermemymanagement1.logicaladdress–generatedbytheCPUalsoreferredtoasvi

23、rtualaddress2.physicaladdress–addressseenbythememyunitlogicalphysicaladdressesarethesameincompiletimeloadtimeaddressbindingschemes.logical(virtual)physicaladdressesdifferinexecutiontimeaddressbindingscheme內(nèi)存管理單元，MMU：MMUi

24、sahardwaredevicethatmapsvirtualtophysicaladdressInMMUschemethevalueintherelocationregisterisaddedtoeveryaddressgeneratedbyauserprocessatthetimeitissenttomemy.Theuserprogramdealswithlogicaladdressesitneverseestherealphysi

25、caladdresses.動態(tài)連接：?linkingpostponeduntilexecutiontimesmallpieceofcodestubusedtolocatetheappropriatememyresidentlibraryroutinestubreplacesitselfwiththeaddressoftheroutineexecutestheroutine?dynamiclinkingisparticularlyusef

26、ulf(shared)libraries：WindowsDLLs?bettermemyusageeasierlibraryupdates?operatingsystemneedtocheckifroutineisinprocesses’memyaddress哲學(xué)家就餐問題：五個哲學(xué)家五個筷子，如何才能共享？采用信號量semaphechopstick[5]，初始值均為1。對于第i哲學(xué)家：dowait(chopstick[i])wait(c

27、hopstick[(i1)%5])eatsignal(chopstick[i])signal(chopstick[(i1)%5])thinkwhile(1)其他解決方法：1.allowatmostfourphilosopherstobesittingsimultaneouslyatthetableaspareresource2.allowaphilosophertopickupchopsticksonlyifbothareavailab

28、lepickupthemsimultaneously3.useaasymmetricsolution:anoddphilosopherpicksupfirsttheleftchopstickthentherightchopstickwhereasanevenonepicksupfirsttherightthentheleft銀行家算法1.IfRequesti?Needigotostep2.Otherwiseraiseerrconditi

29、onsinceprocesshasexceededitsmaximumclaim.2.IfRequesti?Availablegotostep3.OtherwisePimustwaitsinceresourcesarenotavailable.3.PretendtoallocaterequestedresourcestoPibymodifyingthestateasfollows:Available=AvailableRequestiA