CINXE.COM

<?xml version="1.0" encoding="UTF-8" standalone="yes"?> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>LKML: Thomas Gleixner: Re: [patch 00/21] hrtimer - High-resolution timer subsystem</title><link href="/css/message.css" rel="stylesheet" type="text/css" /><link href="/css/wrap.css" rel="alternate stylesheet" type="text/css" title="wrap" /><link href="/css/nowrap.css" rel="stylesheet" type="text/css" title="nowrap" /><link href="/favicon.ico" rel="shortcut icon" /><script src="/js/simple-calendar.js" type="text/javascript"></script><script src="/js/styleswitcher.js" type="text/javascript"></script><link rel="alternate" type="application/rss+xml" title="lkml.org : last 100 messages" href="/rss.php" /><link rel="alternate" type="application/rss+xml" title="lkml.org : last messages by Thomas Gleixner" href="/groupie.php?aid=7748" /><script> var _paq = window._paq = window._paq || []; /* tracker methods like "setCustomDimension" should be called before "trackPageView" */ _paq.push(["setDoNotTrack", true]); _paq.push(["disableCookies"]); _paq.push(['trackPageView']); _paq.push(['enableLinkTracking']); (function() { var u="//m.lkml.org/"; _paq.push(['setTrackerUrl', u+'matomo.php']); _paq.push(['setSiteId', '1']); var d=document, g=d.createElement('script'), s=d.getElementsByTagName('script')[0]; g.async=true; g.src=u+'matomo.js'; s.parentNode.insertBefore(g,s); })(); </script></head><body onload="es.jasper.simpleCalendar.init();" itemscope="itemscope" itemtype="http://schema.org/BlogPosting"><table border="0" cellpadding="0" cellspacing="0"><tr><td width="180" align="center"><a href="/"><img style="border:0;width:135px;height:32px" src="/images/toprowlk.gif" alt="lkml.org" /></a></td><td width="32">聽</td><td class="nb"><div><a class="nb" href="/lkml"> [lkml]</a> 聽 <a class="nb" href="/lkml/2005"> [2005]</a> 聽 <a class="nb" href="/lkml/2005/12"> [Dec]</a> 聽 <a class="nb" href="/lkml/2005/12/13"> [13]</a> 聽 <a class="nb" href="/lkml/last100"> [last100]</a> 聽 <a href="/rss.php"><img src="/images/rss-or.gif" border="0" alt="RSS Feed" /></a></div><div>Views: <a href="#" class="nowrap" onclick="setActiveStyleSheet('wrap');return false;">[wrap]</a><a href="#" class="wrap" onclick="setActiveStyleSheet('nowrap');return false;">[no wrap]</a> 聽 <a class="nb" href="/lkml/mheaders/2005/12/13/89" onclick="this.href='/lkml/headers'+'/2005/12/13/89';">[headers]</a>聽 <a href="/lkml/bounce/2005/12/13/89">[forward]</a>聽 </div></td><td width="32">聽</td></tr><tr><td valign="top"><div class="es-jasper-simpleCalendar" baseurl="/lkml/"></div><div class="threadlist">Messages in this thread</div><ul class="threadlist"><li class="root"><a href="/lkml/2005/12/5/313">First message in thread</a></li><li><a href="/lkml/2005/12/12/122">Thomas Gleixner</a><ul><li><a href="/lkml/2005/12/12/154">Thomas Gleixner</a></li><li><a href="/lkml/2005/12/12/278">George Anzinger</a><ul><li class="origin"><a href="">Thomas Gleixner</a></li><li><a href="/lkml/2005/12/14/406">Roman Zippel</a><ul><li><a href="/lkml/2005/12/14/426">George Anzinger</a><ul><li><a href="/lkml/2005/12/19/63">Roman Zippel</a></li></ul></li></ul></li></ul></li><li><a href="/lkml/2005/12/14/269">Roman Zippel</a><ul><li><a href="/lkml/2005/12/14/299">Thomas Gleixner</a><ul><li><a href="/lkml/2005/12/14/394">George Anzinger</a></li><li><a href="/lkml/2005/12/15/151">Steven Rostedt</a></li><li><a href="/lkml/2005/12/19/62">Roman Zippel</a><ul><li><a href="/lkml/2005/12/19/180">Thomas Gleixner</a></li></ul></li></ul></li></ul></li></ul></li></ul></td><td width="32" rowspan="2" class="c" valign="top"><img src="/images/icornerl.gif" width="32" height="32" alt="/" /></td><td class="c" rowspan="2" valign="top" style="padding-top: 1em"><table><tr><td><table><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: [patch 00/21] hrtimer - High-resolution timer subsystem</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Thomas Gleixner <></td></tr><tr><td class="lp">Date</td><td class="rp" itemprop="datePublished">Tue, 13 Dec 2005 10:18:08 +0100</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody">On Mon, 2005-12-12 at 17:25 -0800, George Anzinger wrote:<br />> >>The rationale for example talks about "a periodic timer with an absolute <br />> >>_initial_ expiration time", so I could also construct a valid example with <br />> >>this expectation. The more I read the spec the more I think the current <br />> >>behaviour is not correct, e.g. that ABS_TIME is only relevant for <br />> >>it_value.<br />> >>So I'm interested in specific interpretations of the spec which support <br />> >>the current behaviour.<br />> <br />> My $0.02 worth: It is clear (from the standard) that the initial time <br />> is to be ABS_TIME. It is also clear that the interval is to be added <br />> to that time. IMHO then, the result should have the same property, <br />> i.e. ABS_TIME. Sort of like adding an offset to a relative address. <br />> The result is still relative.<br /><br />So the only difference between a timer with ABSTIME set and one without<br />is the notion of the initial expiry value, aside the<br />clock_settime(CLOCK_REALTIME) speciality.<br /><br />ABSTIME:<br />firstexp = it_value<br />firstexp, firstexp + it_interval, ... firstexp + n * it_interval<br /><br />non ABSTIME:<br />firstexp = now + it_value<br />firstexp, firstexp + it_interval, ... firstexp + n * it_interval<br /><br />The only limitation of this is that the interval value can not be less<br />than the resolution of the clock in order to avoid the wrong accounting<br />of the overflow.<br /><br />> > Unfortunately you find just the spec all over the place. I fear we have<br />> > to find and agree on an interpretation ourself.<br />> > <br />> > I agree, that the restriction to the initial it_value is definitely<br />> > something you can read out of the spec. But it does not make a lot of<br />> > sense for me. Also the restriction to TIMER_ABSTIME is somehow strange<br />> > as it converts an CLOCK_REALTIME timer to a CLOCK_MONOTONIC timer. I<br />> > never understood the rationale behind that.<br />> <br />> I don't think it really does that. The TIMER_ABSTIME flag just says <br />> that the time requested is to be taken as "clock" time (which ever <br />> clock) AND that this is to be the expire time regardless of clock <br />> setting. We, in an attempt to simplify the lists, convert the expire <br />> time into some common time notation (in most cases we convert relative <br />> times to absolute times) but this introduces problems because the <br />> caller has _asked_ for a relative or absolute time and not the other. <br />> If the clock can not be set this is not a problem. If it can, well, <br />> we need to keep track of what the caller wanted, absolute or relative.<br />> <br />> It might help others to understand this if you were to remove the <br />> clock names from your queues and instead call them "absolute_real" and <br />> "up_time". Then it would be more clear, I think, that we are mapping <br />> user requests onto these queues based on the desired functionality <br />> without a predilection to put a timer on a given queue just because a <br />> particular clock was requested. At this point it becomes clear, for <br />> example, that a TIMER_ABSTIME request on the real clock is the _only_ <br />> request that should be mapped to the "absolute_real" list.<br /><br />In other words. If there is only CLOCK_REALTIME, then the implementation<br />has to keep track of absolute and relative timers.<br /><br />The existance of CLOCK_MONOTONIC and the fact that CLOCK_MONOTONIC is<br />using the same clock source as CLOCK_REALTIME allows us to optimize the<br />implementation by putting the relative timers on the monotonic list.<br /><br /> tglx<br /><br /><br />-<br />To unsubscribe from this list: send the line "unsubscribe linux-kernel" in<br />the body of a message to majordomo@vger.kernel.org<br />More majordomo info at <a href="http://vger.kernel.org/majordomo-info.html">http://vger.kernel.org/majordomo-info.html</a><br />Please read the FAQ at <a href="http://www.tux.org/lkml/">http://www.tux.org/lkml/</a><br /><br /></pre></td><td width="32" rowspan="2" class="c" valign="top"><img src="/images/icornerr.gif" width="32" height="32" alt="\" /></td></tr><tr><td align="right" valign="bottom"> 聽 </td></tr><tr><td align="right" valign="bottom">聽</td><td class="c" valign="bottom" style="padding-bottom: 0px"><img src="/images/bcornerl.gif" width="32" height="32" alt="\" /></td><td class="c">聽</td><td class="c" valign="bottom" style="padding-bottom: 0px"><img src="/images/bcornerr.gif" width="32" height="32" alt="/" /></td></tr><tr><td align="right" valign="top" colspan="2"> 聽 </td><td class="lm">Last update: 2005-12-13 10:14 聽聽 [from the cache]<br />漏2003-2020 <a href="http://blog.jasper.es/"><span itemprop="editor">Jasper Spaans</span></a>|hosted at <a href="https://www.digitalocean.com/?refcode=9a8e99d24cf9">Digital Ocean</a> and my Meterkast|<a href="http://blog.jasper.es/categories.html#lkml-ref">Read the blog</a></td><td>聽</td></tr></table><script language="javascript" src="/js/styleswitcher.js" type="text/javascript"></script></body></html>