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: Linus Torvalds: Re: [PATCH] i386: handle iret faults better</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 Linus Torvalds" href="/groupie.php?aid=1" /><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/2005/4/29"> [29]</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/2005/4/29/151" onclick="this.href='/lkml/headers'+'/2005/4/29/151';">[headers]</a>聽 <a href="/lkml/bounce/2005/4/29/151">[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/2005/4/23/129">First message in thread</a></li><li><a href="/lkml/2005/4/26/3">Linus Torvalds</a><ul><li><a href="/lkml/2005/4/28/280">Roland McGrath</a><ul><li><a href="/lkml/2005/4/29/122">Andi Kleen</a></li><li class="origin"><a href="/lkml/2005/5/2/124">Linus Torvalds</a><ul><li><a href="/lkml/2005/5/2/124">Andi Kleen</a></li></ul></li></ul></li><li><a href="/lkml/2005/4/28/281">Roland McGrath</a><ul><li><a href="/lkml/2005/4/29/121">Andi Kleen</a></li></ul></li></ul></li></ul><div class="threadlist">Patch in this message</div><ul class="threadlist"><li><a href="/lkml/diff/2005/4/29/151/1">Get diff 1</a></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">Fri, 29 Apr 2005 09:44:28 -0700 (PDT)</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Linus Torvalds <></td></tr><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: [PATCH] i386: handle iret faults better</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody"><br /><br />On Thu, 28 Apr 2005, Roland McGrath wrote:<br />><br />> I was never very happy with the special-case check for iret_exc either.<br />> But for the first crack, I went for the fix that didn't touch other<br />> infrastructure code.<br />> <br />> The fault.c changes here are really not necessary for the bug fix at all,<br />> it will never be used there. But to make it a clean infrastructure<br />> upgrade, I made every caller of fixup_exception consistently pass in the<br />> complete info it uses for signals in the user-mode case.<br /><br />I really ended up deciding that we can fix it with a simple one-liner <br />instead, which actually simplifies and cleans up the code, instead of <br />adding new special cases.<br /><br />I just committed the appended, which actually removes one line more than <br />it adds.<br /><br /> Linus<br />-----<br />commit a879cbbb34cbecfa9707fbb6e5a00c503ac1ecb9<br />tree fdf247f8dedea8f04d0989aeab6922ed073eee11<br />parent c06fec5022ebe014af876da2df4a0eee836e97c8<br />author Linus Torvalds <torvalds@ppc970.osdl.org> Fri, 29 Apr 2005 09:38:44 -0700<br />committer Linus Torvalds <torvalds@ppc970.osdl.org> Fri, 29 Apr 2005 09:38:44 -0700<br /><br /> x86: make traps on 'iret' be debuggable in user space<br /><br /> This makes a trap on the 'iret' that returns us to user space<br /> cause a nice clean SIGSEGV, instead of just a hard (and silent)<br /> exit.<br /><br /> That way a debugger can actually try to see what happened, and<br /> we also properly notify everybody who might be interested about<br /> us being gone.<br /><br /> This loses the error code, but tells the debugger what happened<br /> with ILL_BADSTK in the siginfo.<br /><br />--- k/arch/i386/kernel/entry.S (mode:100644)<br />+++ l/arch/i386/kernel/entry.S (mode:100644)<br />@@ -260,11 +260,9 @@ restore_nocheck:<br /> .section .fixup,"ax"<br /> iret_exc:<br /> sti<br />- movl $__USER_DS, %edx<br />- movl %edx, %ds<br />- movl %edx, %es<br />- movl $11,%eax<br />- call do_exit<br />+ pushl $0 # no error code<br />+ pushl $do_iret_error<br />+ jmp error_code<br /> .previous<br /> .section __ex_table,"a"<br /> .align 4<br />--- k/arch/i386/kernel/traps.c (mode:100644)<br />+++ l/arch/i386/kernel/traps.c (mode:100644)<br />@@ -451,6 +451,7 @@ DO_ERROR(10, SIGSEGV, "invalid TSS", inv<br /> DO_ERROR(11, SIGBUS, "segment not present", segment_not_present)<br /> DO_ERROR(12, SIGBUS, "stack segment", stack_segment)<br /> DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0)<br />+DO_ERROR_INFO(32, SIGSEGV, "iret exception", iret_error, ILL_BADSTK, 0)<br /> <br /> fastcall void do_general_protection(struct pt_regs * regs, long error_code)<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 /></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: 2009-11-18 23:46 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>