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: Steven Rostedt: Re: [PATCH 1/19] MUTEX: Introduce simple mutex implementation</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 Steven Rostedt" href="/groupie.php?aid=3543" /><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/2005/12/15"> [15]</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/2005/12/15/315" onclick="this.href='/lkml/headers'+'/2005/12/15/315';">[headers]</a>聽 <a href="/lkml/bounce/2005/12/15/315">[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/2005/12/15/149">First message in thread</a></li><li><a href="/lkml/2005/12/15/201">Linus Torvalds</a><ul><li><a href="/lkml/2005/12/15/216">Erik Mouw</a><ul><li><a href="/lkml/2005/12/15/224">(Dick Streefland)</a></li><li><a href="/lkml/2005/12/16/61">Erik Mouw</a><ul><li><a href="/lkml/2005/12/17/22">Sander</a><ul><li><a href="/lkml/2005/12/17/34">Douglas McNaught</a></li><li><a href="/lkml/2005/12/19/34">Erik Mouw</a></li></ul></li></ul></li></ul></li><li><a href="/lkml/2005/12/15/269">Nikita Danilov</a></li><li><a href="/lkml/2005/12/15/274"> linux@horizon ...</a><ul><li><a href="/lkml/2005/12/15/286">Linus Torvalds</a><ul><li><a href="/lkml/2005/12/15/422"> linux@horizon ...</a></li></ul></li><li class="origin"><a href="">Steven Rostedt</a></li></ul></li><li><a href="/lkml/2005/12/15/304">Steven Rostedt</a></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 1/19] MUTEX: Introduce simple mutex implementation</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Steven Rostedt <></td></tr><tr><td class="lp">Date</td><td class="rp" itemprop="datePublished">Thu, 15 Dec 2005 16:18:21 -0500</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody">On Thu, 2005-12-15 at 14:09 -0500, linux@horizon.com wrote:<br />> ><br />> <br />> As I said, as long as the -rt patch was not in the main tree, taking<br />> advantage of the fact that 95% of the down() and up() callers just want<br />> a mutex was a sensible implementation tradeoff. For merging it into the<br />> tree, it's ugly, and people don't like that. The -rt folks have gotten<br />> used to their naming perversions and so don't feel as much repugnance.<br />> <br /><br />The naming in the -rt side is to try to keep things as much in parallel<br />to the mainline as possible. I don't think Ingo would have a problem<br />with up / down just being used for semaphores, and having true mutex<br />names for just that. In fact that would help us a lot. A lot of bugs<br />fixes that I send to Ingo, is finding places that use mutex when they<br />are really counting semaphores, and thus cant have PI.<br /><br />><br /><br />...<br /><br />> <br />> > So when you say "This isn't about speed, this is about bug-free code", <br />> > you're just making that up.<br />> ><br />> > It's doubly silly when your "safer" <br />> > implementation uses totally illogical names. THAT is what creates bugs.<br />> <br />> If you want to argue about names, go discuss gay marriage.<br /><br />Are you suggesting a "mutex union"?<br /><br />> <br />> I don't care what it's *called*. I care that we have stronger<br />> conditions that we can test for correctness.<br /><br />Well a name is helpful in understanding what's going on. Especially if<br />we want new up and coming kernel programmers to help out. Instead of<br />staying with what is there now.<br /><br />Also, while we're at it, lets fix that damn down_trylock (or<br />mutex_trylock) to return 1 on success, 0 on contention, just like the<br />spin_trylock does!!!<br /><br />Actually, that alone is a good argument to not keep the same names. We<br />can keep down_trylock as the same perverted self, and have mutex_trylock<br />do it right. Of course, special care is needed when doing this<br />conversion, but a wrong pick should show itself right away.<br /><br />> <br />> > So go away.<br />> > <br />> > Come back if you have pondered, and accepted reality, and perhaps have an <br />> > acceptable patch that introduces a separate data structure. <br />> <br />> Ha! I still say you're wrong, and I'm not going to fold over an obvious<br />> technical point just because of flaming.<br />> <br />> Are we having some communication problems? I find it hard to believe<br />> that you're actually this *stupid*, but we might not be talking about<br />> the same thing.<br />> <br />> I took your posting to say that<br />> <br />> a) Using the names "struct semaphore", "up()" and "down()" for a mutex<br />> is monumentally brain-dead. I'm not arguing, although I understand<br />> the pragmatic reasons for the original abuse of notation.<br />> <br />> b) There is no need for a mutex implementation, because a semaphore can<br />> do anything that a mutex can. Here, I absolutely disagree. There<br />> are things you can do with a mutex that you CANNOT do with a<br />> general semaphore, because a mutex has stronger invariants.<br />> <br />> A counting semaphore can do MOST of what a mutex does, and is<br />> demonstrably close enough for a lot of uses.<br />> <br />> > And no, we're not switching users over whole-sale. First you introduce the <br />> > new concept. Only THEN can you can switch over INDIVIDUAL LOCKS with <br />> > reasons for why it's worth it.<br />> <br />> Given that 95% of callers are using it as mutex, you're making this 20<br />> times more work than necessary. Convert 'em all and change the 5%<br />> that need the counting back.<br /><br />I disagree with doing that. Especially since I've argued that a mutex<br />is a semaphore, but a semaphore is not a mutex. So I rather go slowly<br />changing the semaphores that are acting as mutexes, (since those that<br />are not changed are not broken) than doing the change all mutexes to<br />semaphores, where a mutex can not always act like a semaphore, and then<br />go and break those 5%.<br /><br />In reality, this is what the RT patch did. All semaphores (up / down)<br />became mutexes, and then we manually found the counting semaphores and<br />started switching them to compat_semaphores (what semaphore is today).<br />I'm still sending in patches to fix these.<br /><br />-- Steve<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: 2005-12-15 22:21 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>