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: Adam Belay: Re: [PATCH] PCI: Add pci shutdown ability</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 Adam Belay" href="/groupie.php?aid=6696" /><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/2005/4"> [Apr]</a> 聽 <a class="nb" href="/lkml/2005/4/26"> [26]</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/2005/4/26/39" onclick="this.href='/lkml/headers'+'/2005/4/26/39';">[headers]</a>聽 <a href="/lkml/bounce/2005/4/26/39">[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/2005/4/25/154">First message in thread</a></li><li><a href="/lkml/2005/4/25/253">Adam Belay</a><ul><li><a href="/lkml/2005/4/26/14">Benjamin Herrenschmidt</a><ul><li class="origin"><a href="/lkml/2005/4/26/57">Adam Belay</a><ul><li><a href="/lkml/2005/4/26/57">Nigel Cunningham</a></li><li><a href="/lkml/2005/4/26/82">Pavel Machek</a></li></ul></li></ul></li><li><a href="/lkml/2005/4/26/92">Pavel Machek</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">Date</td><td class="rp" itemprop="datePublished">Tue, 26 Apr 2005 02:23:14 -0400</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Adam Belay <></td></tr><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: [PATCH] PCI: Add pci shutdown ability</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody">On Tue, Apr 26, 2005 at 02:32:29PM +1000, Benjamin Herrenschmidt wrote:<br /><br />-->snip<br /><br />> <br />> I don't like this notion of "stop" separated from power states anyway, I<br />> think it just doesn't work in practice.<br /><br />Yeah, after giving it some additional thought, I think there are better ways.<br /><br />> <br />> Ben.<br />> <br /><br />Ok, here's a new idea.<br /><br />For many devices "->suspend" and "->resume" with pm_message_t is exactly what<br />we need. However, as we support more advanced power management features, such<br />as runtime power management, or power containers, we need something a little<br />more specific. The exact power state must be specified among other issues.<br /><br />We might do something like this:<br /><br />Keep "->suspend" and "->resume" around unchanged. (so the states would<br />probably remain as PMSG_FREEZE and PMSG_SUSPEND). If the driver doesn't<br />support the more advanced PM methods just use these. They work well enough<br />for system sleep states etc.<br /><br />Alternatively drivers could support a more rich power management interface<br />via the following methods:<br /><br /><br />change_state - changes a device's power state<br /><br />change_state(struct device * dev, pm_state_t state, struct system_state * sys_state, int reason); <br />@dev - the device<br />@state - the target device-specific power state<br />@sys_state - a data structure containing information about the intended global system power state<br />@reason - why the state must be changed (ex. RUNTIME_PM, SYSTEM_SLEEP, SYSTEM_RESUME, etc.)<br /><br /><br />halt - acts somewhat like PMSG_FREEZE, stops device activity, doesn't change power state<br /><br />halt(struct device * dev, struct system_state * sys_state, int reason);<br />@dev - the device<br />@sys_state - a data structure containing information about the intended global system power state<br />@reason - why we are halting operation (ex. RUNTIME_CHANGES (like cpufreq), SYSTEM_SLEEP, SHUTDOWN, REBOOT) <br /><br /><br />contine - resumes from a "halt"<br /><br />continue(struct device * dev, struct system_state * sys_state, int reason);<br />@dev - the device<br />@sys_state - a data structure containing information about the intended global system power state<br />@reason - why we are resuming operation (ex. RUNTIME_CHANGES (like cpufreq), SYSTEM_RESUME) <br /><br /><br />When changing system state, we call "change_state" for every device with power<br />resources. Devices that do not directly consume power or have power states<br />will not implement "change_state" so we will call "halt" and "continue"<br />instead.<br /><br />When shutting down the system, halt has the option of turning off the device,<br />as it will see the SHUTDOWN reason. So it's a driver-knows-best approach<br />instead of assuming everything must be turned off, or everything must just be<br />stopped.<br /><br />So in theory, with cpufreq, we could stop userspace, ->halt every device<br />(drivers won't do anything if they know it's not necessary), change the<br />frequency, and then resume operation.<br /><br />We may want to create structures like pm_message_t for "change_state", "halt",<br />and "continue". Pavel, do you have any thoughts on this?<br /><br />This is just a rough idea... I look forward to any comments or suggestions.<br /><br />Thanks,<br />Adam<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: 2005-04-26 08:30 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>