serial&nbsp;port&nbsp;<br />some&nbsp;Ethernet&nbsp;cards&nbsp;<br />most&nbsp;harddisk/SCSI&nbsp;drivers&nbsp;<br />most&nbsp;CD-ROM&nbsp;drivers&nbsp;<br />Especially&nbsp;for&nbsp;the&nbsp;serial&nbsp;port,&nbsp;it&nbsp;would&nbsp;be&nbsp;very&nbsp;nice&nbsp;to&nbsp;have&nbsp;a&nbsp;way&nbsp;to&nbsp;<br />switch&nbsp;to&nbsp;low&nbsp;dispatch&nbsp;latency&nbsp;interrupt&nbsp;handling,&nbsp;such&nbsp;that&nbsp;an&nbsp;<br />application&nbsp;can&nbsp;react&nbsp;as&nbsp;fast&nbsp;as&nbsp;possible&nbsp;on&nbsp;any&nbsp;incoming&nbsp;serial&nbsp;byte.&nbsp;<br />Apart&nbsp;from&nbsp;the&nbsp;&quot;fast&quot;&nbsp;interrupt&nbsp;handler&nbsp;type,&nbsp;another&nbsp;important&nbsp;cause&nbsp;<br />of&nbsp;latency&nbsp;in&nbsp;the&nbsp;serial&nbsp;port&nbsp;are&nbsp;the&nbsp;16550&nbsp;FIFO&nbsp;UARTs.&nbsp;At&nbsp;the&nbsp;moment,&nbsp;<br />the&nbsp;UARTs&nbsp;are&nbsp;configured&nbsp;to&nbsp;trigger&nbsp;an&nbsp;interrupt&nbsp;(except&nbsp;for&nbsp;for&nbsp;bit&nbsp;<br />rates&nbsp;&lt;&nbsp;2400&nbsp;bit/s)&nbsp;only&nbsp;if&nbsp;the&nbsp;FIFO&nbsp;is&nbsp;filled&nbsp;with&nbsp;at&nbsp;least&nbsp;eight&nbsp;<br />bytes&nbsp;or&nbsp;if&nbsp;an&nbsp;UART&nbsp;timeout&nbsp;has&nbsp;occured&nbsp;(this&nbsp;happens&nbsp;after&nbsp;four&nbsp;<br />character&nbsp;transmission&nbsp;times).&nbsp;The&nbsp;16550&nbsp;chip&nbsp;can&nbsp;be&nbsp;configured&nbsp;to&nbsp;<br />trigger&nbsp;limits&nbsp;1,&nbsp;4,&nbsp;8,&nbsp;and&nbsp;14,&nbsp;but&nbsp;Linux&nbsp;currently&nbsp;provides&nbsp;no&nbsp;<br />ioctl()&nbsp;for&nbsp;this.&nbsp;For&nbsp;details&nbsp;about&nbsp;the&nbsp;FIFO&nbsp;UART,&nbsp;please&nbsp;consult&nbsp;the&nbsp;<br />National&nbsp;Semiconductor&nbsp;PC16550D&nbsp;data&nbsp;sheet.&nbsp;<br />For&nbsp;real-time&nbsp;applications,&nbsp;it&nbsp;would&nbsp;be&nbsp;very&nbsp;desirable&nbsp;to&nbsp;change&nbsp;the&nbsp;<br />FIFO&nbsp;trigger&nbsp;limit&nbsp;to&nbsp;one&nbsp;byte&nbsp;so&nbsp;that&nbsp;incoming&nbsp;serial&nbsp;bytes&nbsp;are&nbsp;<br />immediately&nbsp;delivered&nbsp;to&nbsp;the&nbsp;waiting&nbsp;process.&nbsp;Applications&nbsp;that&nbsp;depend&nbsp;<br />heavily&nbsp;on&nbsp;minimum&nbsp;serial&nbsp;port&nbsp;interrupt&nbsp;dispatch&nbsp;latency&nbsp;are&nbsp;for&nbsp;<br />example&nbsp;MIDI&nbsp;music&nbsp;systems,&nbsp;DCF77&nbsp;reference&nbsp;time&nbsp;long-wave&nbsp;radio&nbsp;signal&nbsp;<br />receivers&nbsp;in&nbsp;Europe,&nbsp;packet&nbsp;radio&nbsp;network&nbsp;interfaces,&nbsp;and&nbsp;ISO&nbsp;7816&nbsp;<br />smartcard&nbsp;interfaces.&nbsp;<br />Some&nbsp;existing&nbsp;work&nbsp;in&nbsp;this&nbsp;area:&nbsp;<br />Stuart&nbsp;Cheshire&nbsp;&lt;<a href="mailto:cheshire@DSG.Stanford.EDU>">cheshire@DSG.Stanford.EDU></a>&nbsp;has&nbsp;implemented&nbsp;a&nbsp;&quot;smart&quot;&nbsp;<br />interrupt&nbsp;handler&nbsp;that&nbsp;can&nbsp;switch&nbsp;between&nbsp;the&nbsp;classic&nbsp;&quot;fast&quot;&nbsp;and&nbsp;<br />&quot;slow&quot;&nbsp;alternatives&nbsp;(see&nbsp;&lt;URL:<a href="http://mosquitonet.stanford.edu/">http://mosquitonet.stanford.edu/</a>&nbsp;<br />smartirq.html&gt;&nbsp;for&nbsp;details).&nbsp;This&nbsp;way,&nbsp;you&nbsp;can&nbsp;get&nbsp;the&nbsp;real-time&nbsp;<br />advantage&nbsp;of&nbsp;&quot;slow&quot;&nbsp;handling&nbsp;for&nbsp;the&nbsp;serial&nbsp;port&nbsp;without&nbsp;introducing&nbsp;a&nbsp;<br />permanent&nbsp;system&nbsp;overhead&nbsp;even&nbsp;when&nbsp;no&nbsp;real-time&nbsp;application&nbsp;is&nbsp;<br />active.&nbsp;<br />Nick&nbsp;Simicich&nbsp;&lt;<a href="mailto:njs@scifi.emi.net>">njs@scifi.emi.net></a>&nbsp;and&nbsp;Rik&nbsp;Faith&nbsp;&lt;<a href="mailto:faith@cs.unc.edu>">faith@cs.unc.edu></a>&nbsp;<br />have&nbsp;written&nbsp;a&nbsp;program&nbsp;called&nbsp;&quot;cytune&quot;&nbsp;which&nbsp;can&nbsp;change&nbsp;the&nbsp;FIFO&nbsp;<br />thresholds&nbsp;for&nbsp;individual&nbsp;ports&nbsp;in&nbsp;the&nbsp;Cyclades&nbsp;async&nbsp;mux&nbsp;driver.&nbsp;<br />Problem&nbsp;2:&nbsp;Timer&nbsp;resolution&nbsp;<br />On&nbsp;most&nbsp;Linux&nbsp;architectures,&nbsp;a&nbsp;timer&nbsp;interrupt&nbsp;occurs&nbsp;every&nbsp;10&nbsp;ms&nbsp;(HZ&nbsp;<br />=&nbsp;100).&nbsp;One&nbsp;exception&nbsp;is&nbsp;the&nbsp;Alpha&nbsp;architecture,&nbsp;where&nbsp;the&nbsp;timer&nbsp;<br />interrupt&nbsp;comes&nbsp;every&nbsp;1&nbsp;ms&nbsp;(HZ&nbsp;=&nbsp;1024).&nbsp;The&nbsp;macro&nbsp;HZ,&nbsp;which&nbsp;specifies&nbsp;<br />the&nbsp;timer&nbsp;frequency,&nbsp;is&nbsp;defined&nbsp;in&nbsp;&lt;asm/param.h&gt;.&nbsp;At&nbsp;the&nbsp;moment,&nbsp;the&nbsp;<br />kernel&nbsp;timer&nbsp;mechanism,&nbsp;on&nbsp;which&nbsp;the&nbsp;itimer&nbsp;and&nbsp;nanosleep&nbsp;<br />implementation&nbsp;is&nbsp;based,&nbsp;checks&nbsp;after&nbsp;each&nbsp;timer&nbsp;interrupt&nbsp;whether&nbsp;a&nbsp;<br />software&nbsp;timer&nbsp;has&nbsp;expired,&nbsp;makes&nbsp;the&nbsp;corresponding&nbsp;process&nbsp;runnable&nbsp;<br />again,&nbsp;and&nbsp;calls&nbsp;the&nbsp;scheduler.&nbsp;This&nbsp;means&nbsp;that&nbsp;a&nbsp;highest&nbsp;priority&nbsp;<br />real-time&nbsp;process&nbsp;can&nbsp;have&nbsp;to&nbsp;wait&nbsp;up&nbsp;to&nbsp;2&nbsp;*&nbsp;(1&nbsp;s&nbsp;/&nbsp;HZ)&nbsp;=&nbsp;20&nbsp;ms&nbsp;longer&nbsp;<br />than&nbsp;requested.&nbsp;<br />An&nbsp;excellent&nbsp;solution&nbsp;would&nbsp;be&nbsp;to&nbsp;implement&nbsp;an&nbsp;interrupt-on-demand&nbsp;<br />timer&nbsp;facility.&nbsp;The&nbsp;8253&nbsp;timer/counter&nbsp;chip&nbsp;in&nbsp;the&nbsp;PC&nbsp;would&nbsp;then&nbsp;have&nbsp;<br />to&nbsp;be&nbsp;programmed&nbsp;such&nbsp;that&nbsp;it&nbsp;delivers&nbsp;an&nbsp;interrupt&nbsp;with&nbsp;microsecond&nbsp;<br />precision&nbsp;exactly&nbsp;at&nbsp;the&nbsp;time&nbsp;when&nbsp;a&nbsp;software&nbsp;timer&nbsp;expires.&nbsp;This&nbsp;<br />would&nbsp;increase&nbsp;the&nbsp;software&nbsp;timer&nbsp;precision&nbsp;by&nbsp;a&nbsp;factor&nbsp;of&nbsp;10&nbsp;000!&nbsp;As&nbsp;<br />the&nbsp;8253&nbsp;timer&nbsp;would&nbsp;be&nbsp;programmed&nbsp;in&nbsp;a&nbsp;single&nbsp;shot&nbsp;mode,&nbsp;a&nbsp;lost&nbsp;<br />interrupt&nbsp;might&nbsp;cause&nbsp;an&nbsp;accidental&nbsp;system&nbsp;halt.&nbsp;If&nbsp;this&nbsp;is&nbsp;problem,&nbsp;<br />the&nbsp;CMOS&nbsp;battery&nbsp;clock,&nbsp;which&nbsp;can&nbsp;also&nbsp;be&nbsp;used&nbsp;to&nbsp;implement&nbsp;periodic&nbsp;<br />interrupts,&nbsp;should&nbsp;be&nbsp;utilized&nbsp;to&nbsp;trigger&nbsp;periodic&nbsp;calls&nbsp;to&nbsp;the&nbsp;<br />scheduler&nbsp;in&nbsp;order&nbsp;to&nbsp;ensure&nbsp;process&nbsp;preemption&nbsp;and&nbsp;kernel&nbsp;clock&nbsp;<br />update.&nbsp;The&nbsp;time&nbsp;stamp&nbsp;counter&nbsp;(TSC)&nbsp;available&nbsp;in&nbsp;all&nbsp;Pentium&nbsp;<br />processors&nbsp;is&nbsp;a&nbsp;64-bit&nbsp;counter&nbsp;clocked&nbsp;at&nbsp;CPU&nbsp;frequency.&nbsp;It&nbsp;can&nbsp;be&nbsp;<br />used&nbsp;on&nbsp;Pentium&nbsp;systems&nbsp;in&nbsp;order&nbsp;to&nbsp;get&nbsp;extremely&nbsp;precise&nbsp;timing&nbsp;<br />information&nbsp;with&nbsp;theoretically&nbsp;close&nbsp;to&nbsp;nanosecond&nbsp;resolution&nbsp;and&nbsp;<br />without&nbsp;overflow&nbsp;problems.&nbsp;<br />Literature&nbsp;<br />----------&nbsp;<br />For&nbsp;those&nbsp;of&nbsp;you&nbsp;who&nbsp;have&nbsp;become&nbsp;interested&nbsp;in&nbsp;POSIX.1b,&nbsp;there&nbsp;exists&nbsp;<br />a&nbsp;good&nbsp;book:&nbsp;<br />Bill&nbsp;O.&nbsp;Gallmeister,&nbsp;POSIX.4&nbsp;--&nbsp;Programming&nbsp;for&nbsp;the&nbsp;Real&nbsp;World,&nbsp;<br />O'Reilly&nbsp;&amp;&nbsp;Associates,&nbsp;1995,&nbsp;ISBN&nbsp;1-56592-074-0.&nbsp;<br />This&nbsp;book&nbsp;is&nbsp;not&nbsp;only&nbsp;a&nbsp;good&nbsp;introduction&nbsp;into&nbsp;POSIX.1b&nbsp;(which&nbsp;was&nbsp;<br />originally&nbsp;called&nbsp;POSIX.4),&nbsp;it&nbsp;is&nbsp;also&nbsp;an&nbsp;easy&nbsp;reading&nbsp;nice&nbsp;way&nbsp;into&nbsp;<br />the&nbsp;world&nbsp;of&nbsp;real-time&nbsp;operating&nbsp;systems&nbsp;for&nbsp;those&nbsp;developers&nbsp;who&nbsp;have&nbsp;<br />so&nbsp;far&nbsp;been&nbsp;very&nbsp;UNIX&nbsp;and&nbsp;time-sharing&nbsp;oriented.&nbsp;<br />You&nbsp;can&nbsp;order&nbsp;the&nbsp;POSIX.1b&nbsp;standard&nbsp;(officially&nbsp;called&nbsp;IEEE&nbsp;Std&nbsp;<br />1003.1b-1993;&nbsp;this&nbsp;book&nbsp;includes&nbsp;also&nbsp;all&nbsp;text&nbsp;of&nbsp;POSIX.1&nbsp;and&nbsp;costs&nbsp;<br />114&nbsp;USD)&nbsp;as&nbsp;well&nbsp;as&nbsp;the&nbsp;other&nbsp;POSIX&nbsp;standards&nbsp;directly&nbsp;from&nbsp;IEEE:&nbsp;<br />phone:&nbsp;+1&nbsp;908&nbsp;981&nbsp;1393&nbsp;(TZ:&nbsp;eastern&nbsp;standard&nbsp;time)&nbsp;<br />1&nbsp;800&nbsp;678&nbsp;4333&nbsp;(from&nbsp;US+Canada&nbsp;only)&nbsp;<br />fax:&nbsp;+1&nbsp;908&nbsp;981&nbsp;9667&nbsp;<br />e-mail:&nbsp;<a href="mailto:stds.info@ieee.org">stds.info@ieee.org</a>&nbsp;<br />Information&nbsp;about&nbsp;POSIX&nbsp;and&nbsp;other&nbsp;IEEE&nbsp;standards&nbsp;is&nbsp;also&nbsp;available&nbsp;on&nbsp;<br />&lt;URL:<a href="http://stdsbbs.ieee.org/>">http://stdsbbs.ieee.org/></a>&nbsp;and&nbsp;&lt;URL:<a href="http://www.knosof.co.uk/">http://www.knosof.co.uk/</a>&nbsp;<br />posix.html&gt;,&nbsp;however&nbsp;unfortunately&nbsp;the&nbsp;full&nbsp;standard&nbsp;documents&nbsp;are&nbsp;<br />only&nbsp;available&nbsp;as&nbsp;books&nbsp;or&nbsp;on&nbsp;CD-ROM,&nbsp;not&nbsp;on&nbsp;the&nbsp;Internet.&nbsp;Having&nbsp;<br />access&nbsp;to&nbsp;the&nbsp;POSIX&nbsp;specs&nbsp;is&nbsp;certainly&nbsp;a&nbsp;good&nbsp;idea&nbsp;for&nbsp;any&nbsp;Linux&nbsp;<br />kernel&nbsp;hacker.&nbsp;<br />Here&nbsp;is&nbsp;a&nbsp;brief&nbsp;list&nbsp;of&nbsp;some&nbsp;of&nbsp;the&nbsp;POSIX&nbsp;standards:&nbsp;<br />POSIX.1&nbsp;Basic&nbsp;OS&nbsp;interface&nbsp;(C&nbsp;language)&nbsp;<br />POSIX.1a&nbsp;Misc.&nbsp;extensions&nbsp;(symlinks,&nbsp;etc.)&nbsp;<br />POSIX.1b&nbsp;Real-time&nbsp;and&nbsp;I/O&nbsp;extensions&nbsp;(was:&nbsp;POSIX.4)&nbsp;<br />POSIX.1c&nbsp;Threads&nbsp;(was:&nbsp;POSIX.4a)&nbsp;<br />POSIX.1d&nbsp;More&nbsp;real-time&nbsp;extensions&nbsp;(was:&nbsp;POSIX.4b)&nbsp;<br />POSIX.1e&nbsp;Security&nbsp;extensions,&nbsp;ACLs&nbsp;(was:&nbsp;POSIX.6)&nbsp;<br />POSIX.1f&nbsp;Transparent&nbsp;network&nbsp;file&nbsp;access&nbsp;(was:&nbsp;POSIX.8)&nbsp;<br />POSIX.1g&nbsp;Protocol&nbsp;independent&nbsp;communication,&nbsp;sockets&nbsp;(was:&nbsp;POSIX.12)&nbsp;<br />POSIX.1i&nbsp;Technical&nbsp;corrections&nbsp;to&nbsp;POSIX.1b&nbsp;<br />POSIX.2&nbsp;Shell&nbsp;and&nbsp;common&nbsp;utility&nbsp;programs&nbsp;(date,&nbsp;ln,&nbsp;...)&nbsp;<br />POSIX.3&nbsp;Test&nbsp;methods&nbsp;<br />POSIX.5&nbsp;ADA&nbsp;binding&nbsp;to&nbsp;POSIX.1&nbsp;<br />POSIX.7&nbsp;System&nbsp;administration&nbsp;<br />POSIX.9&nbsp;FORTRAN-77&nbsp;binding&nbsp;to&nbsp;POSIX.1&nbsp;<br />POSIX.15&nbsp;Supercomputing&nbsp;extensions&nbsp;(checkpoint/recovery,&nbsp;etc.)&nbsp;<br />and&nbsp;a&nbsp;few&nbsp;others&nbsp;which&nbsp;are&nbsp;still&nbsp;in&nbsp;early&nbsp;draft&nbsp;stage.&nbsp;If&nbsp;you&nbsp;want&nbsp;to&nbsp;<br />follow&nbsp;progress&nbsp;on&nbsp;POSIX&nbsp;standardization,&nbsp;you&nbsp;should&nbsp;read&nbsp;the&nbsp;<br />announcements&nbsp;in&nbsp;the&nbsp;moderated&nbsp;USENET&nbsp;group&nbsp;comp.std.unix.&nbsp;The&nbsp;current&nbsp;<br />status&nbsp;of&nbsp;POSIX&nbsp;drafts&nbsp;is&nbsp;summarized&nbsp;in&nbsp;&lt;URL:<a href="http://stdsbbs.ieee.org/">http://stdsbbs.ieee.org/</a>&nbsp;<br />groups/pasc/standing/sd11.html&gt;.&nbsp;<br />ISO&nbsp;has&nbsp;also&nbsp;published&nbsp;POSIX.1&nbsp;as&nbsp;ISO/IEC&nbsp;9945-1:1990.&nbsp;ISO&nbsp;and&nbsp;IEEE&nbsp;<br />will&nbsp;soon&nbsp;publish&nbsp;the&nbsp;new&nbsp;revision&nbsp;of&nbsp;this&nbsp;standard:&nbsp;ISO/IEC&nbsp;<br />9945-1:1996.&nbsp;This&nbsp;will&nbsp;be&nbsp;the&nbsp;new&nbsp;1996&nbsp;revision&nbsp;of&nbsp;POSIX.1,&nbsp;which&nbsp;will&nbsp;<br />contain&nbsp;in&nbsp;one&nbsp;single&nbsp;standard&nbsp;POSIX.1(1990),&nbsp;POSIX.1b(1993),&nbsp;<br />POSIX.1c(1995),&nbsp;and&nbsp;POSIX.1i(1995)&nbsp;(perhaps&nbsp;also&nbsp;POSIX.1a(1996)&nbsp;if&nbsp;it&nbsp;<br />gets&nbsp;ready&nbsp;in&nbsp;time).&nbsp;If&nbsp;you&nbsp;want&nbsp;to&nbsp;order&nbsp;the&nbsp;POSIX&nbsp;standard&nbsp;but&nbsp;are&nbsp;<br />not&nbsp;in&nbsp;a&nbsp;hurry,&nbsp;may&nbsp;be&nbsp;it&nbsp;is&nbsp;a&nbsp;good&nbsp;idea&nbsp;to&nbsp;wait&nbsp;a&nbsp;few&nbsp;months&nbsp;until&nbsp;<br />ISO/IEC&nbsp;9945-1:1996&nbsp;is&nbsp;available.&nbsp;All&nbsp;differences&nbsp;between&nbsp;<br />POSIX.1(1990)&nbsp;and&nbsp;POSIX.1(1996)&nbsp;will&nbsp;be&nbsp;marked&nbsp;by&nbsp;bars&nbsp;at&nbsp;the&nbsp;page&nbsp;<br />margins.&nbsp;<br />This&nbsp;text&nbsp;just&nbsp;summarizes&nbsp;POSIX.1b&nbsp;and&nbsp;related&nbsp;work&nbsp;on&nbsp;Linux.&nbsp;Many&nbsp;<br />people&nbsp;interested&nbsp;in&nbsp;POSIX.1b&nbsp;support&nbsp;seem&nbsp;also&nbsp;to&nbsp;be&nbsp;interested&nbsp;in&nbsp;<br />POSIX.1c&nbsp;support&nbsp;(threads).&nbsp;Some&nbsp;information&nbsp;about&nbsp;POSIX.1c&nbsp;support&nbsp;is&nbsp;<br />on&nbsp;&lt;URL:<a href="http://www.mit.edu:8001/people/proven/pthreads.html>">http://www.mit.edu:8001/people/proven/pthreads.html></a>&nbsp;and&nbsp;<br />&lt;URL:<a href="http://www.aa.net/~mtp/>.">http://www.aa.net/~mtp/>.</a>&nbsp;There&nbsp;is&nbsp;now&nbsp;a&nbsp;production&nbsp;release&nbsp;<br />package&nbsp;called&nbsp;&quot;PCthreads&nbsp;(tm)&nbsp;POSIX&nbsp;threads&nbsp;for&nbsp;Linux&quot;&nbsp;available&nbsp;from&nbsp;<br />&lt;URL:<a href="ftp://sunsite.unc.edu/pub/Linux/devel/lang/c/>.">ftp://sunsite.unc.edu/pub/Linux/devel/lang/c/>.</a>&nbsp;It&nbsp;is&nbsp;maintained&nbsp;<br />by&nbsp;Michael&nbsp;T.&nbsp;Peterson&nbsp;&lt;<a href="mailto:mtp@big.aa.net>.">mtp@big.aa.net>.</a>&nbsp;A&nbsp;POSIX.1c&nbsp;package&nbsp;that&nbsp;<br />provides&nbsp;real&nbsp;kernel-threads&nbsp;using&nbsp;the&nbsp;clone()&nbsp;system&nbsp;call&nbsp;is&nbsp;<br />available&nbsp;from&nbsp;&lt;URL:<a href="http://pauillac.inria.fr/~xleroy/linuxthreads/>">http://pauillac.inria.fr/~xleroy/linuxthreads/></a>&nbsp;<br />and&nbsp;has&nbsp;been&nbsp;developed&nbsp;by&nbsp;Xavier&nbsp;Leroy&nbsp;&lt;<a href="mailto:Xavier.Leroy@inria.fr>.">Xavier.Leroy@inria.fr>.</a>&nbsp;<br />Markus&nbsp;<br />--&nbsp;<br />Markus&nbsp;Kuhn,&nbsp;Computer&nbsp;Science&nbsp;student&nbsp;--&nbsp;University&nbsp;of&nbsp;Erlangen,&nbsp;<br />Internet&nbsp;Mail:&nbsp;&lt;<a href="mailto:mskuhn@cip.informatik.uni-erlangen.de>">mskuhn@cip.informatik.uni-erlangen.de></a>&nbsp;-&nbsp;Germany&nbsp;<br />WWW&nbsp;Home:&nbsp;&lt;<a href="http://wwwcip.informatik.uni-erlangen.de/user/mskuhn>">http://wwwcip.informatik.uni-erlangen.de/user/mskuhn></a>&nbsp;<br />&nbsp;<br />--&nbsp;<br />&nbsp;<br />※&nbsp;來源:·BBS&nbsp;水木清華站&nbsp;smth.org·[FROM:&nbsp;166.111.161.11]&nbsp;<br /><a href="00000008.htm">上一篇</a><a href="javascript:history.go(-1)">返回上一頁</a><a href="index.htm">回到目錄</a><a href="#top">回到頁首</a><a href="00000010.htm">下一篇</a></h1></center><center><h1>BBS 水木清華站∶精華區(qū)</h1></center></body></html>