CINXE.COM

LKML: Ville Herva: Re: filesystem transactions API

<?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: Ville Herva: Re: filesystem transactions API</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 Ville Herva" href="/groupie.php?aid=384" /><!--Matomo--><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><!--End Matomo Code--></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/2005/4"> [Apr]</a> 聽 <a class="nb" href="/lkml/2005/4/27"> [27]</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/2005/4/27/94" onclick="this.href='/lkml/headers'+'/2005/4/27/94';">[headers]</a>聽 <a href="/lkml/bounce/2005/4/27/94">[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/2005/4/24/76">First message in thread</a></li><li><a href="/lkml/2005/4/26/181">Jamie Lokier</a><ul><li><a href="/lkml/2005/4/26/229">Diego Calleja</a><ul><li><a href="/lkml/2005/4/26/240">Jamie Lokier</a></li></ul></li><li><a href="/lkml/2005/4/27/46">Jan Hudec</a><ul><li class="origin"><a href="/lkml/2005/4/27/126">Ville Herva</a><ul><li><a href="/lkml/2005/4/27/126">Jamie Lokier</a><ul><li><a href="/lkml/2005/4/27/150">Bernd Eckenfels</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">Date</td><td class="rp" itemprop="datePublished">Wed, 27 Apr 2005 16:43:31 +0300</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Ville Herva &lt;&gt;</td></tr><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: filesystem transactions API</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody">On Wed, Apr 27, 2005 at 11:34:12AM +0200, you [Jan Hudec] wrote:<br />&gt; On Tue, Apr 26, 2005 at 16:24:34 +0100, Jamie Lokier wrote:<br />&gt; &gt; John Stoffel wrote:<br />&gt; &gt; &gt; &gt;&gt;&gt;&gt;&gt; "Jamie" == Jamie Lokier &lt;jamie&#64;shareable.org&gt; writes:<br />&gt; &gt; &gt; <br />&gt; &gt; &gt; Jamie&gt; No. A transaction means that _all_ processes will see the<br />&gt; &gt; &gt; Jamie&gt; whole transaction or not.<br />&gt; &gt; &gt; <br />&gt; &gt; &gt; This is really hard. How do you handle the case where process X<br />&gt; &gt; &gt; starts a transaction modifies files a, b &amp; c, but process Y has file b<br />&gt; &gt; &gt; open for writing, and never lets it go? Or the file gets unlinked? <br />&gt; &gt; <br />&gt; &gt; Then it starts to depend on what kind of transactions you want to<br />&gt; &gt; implement.<br />&gt; &gt; <br />&gt; &gt; You can say that a transaction isn't allowed when a process has one of<br />&gt; &gt; the files opened for writing. Or you can say a transaction is<br />&gt; &gt; equivalent to calling all of the I/O system calls at once. You can<br />&gt; &gt; also decide if you want the reads and directory lookups performed in<br />&gt; &gt; the transactions to become prerequisites for the transaction<br />&gt; &gt; completing (so it's aborted if another process writes to those file<br />&gt; &gt; regions or changes the directory structure in a way which breaks a<br />&gt; &gt; prerequisite), or if you want those to lock the things which are read<br />&gt; &gt; for the duration of the transaction, or even just ignore reads for<br />&gt; &gt; transaction purposes. Or, you can say that transactions are limited<br />&gt; &gt; to just directory structure, and not file contents (that's good enough<br />&gt; &gt; for package management), or you can say they're limited to just file<br />&gt; &gt; contents (that's good enough for databases and text file edits).<br />&gt; &gt; <br />&gt; &gt; Etc, etc, quite a lot of semantic choices.<br />&gt; <br />&gt; How do we specify which calls belong to a transaction? By some kind of<br />&gt; extra file handle?<br />&gt; <br />&gt; I'd think having global per-process transaction is not the best way.<br />&gt; So I think we should have some kind of transaction handle (probably in<br />&gt; the file handle space) and a way to say that a syscall is done within<br />&gt; a transaction. To avoid duplicating all syscalls, we could have<br />&gt; set_active_transaction() operation.<br /><br />That's more or less what NTFS does. See the example at<br /><a href="http://blogs.msdn.com/because_we_can/">http://blogs.msdn.com/because_we_can/</a><br /><br /><br /><br />-- v -- <br /><br />v&#64;iki.fi<br /><br />-<br />To unsubscribe from this list: send the line "unsubscribe linux-kernel" in<br />the body of a message to majordomo&#64;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: 2005-04-27 15:51 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>

Pages: 1 2 3 4 5 6 7 8 9 10