Bug 35154 – JavaScript JIT incompatible with SELinux (execmem)

• Component: JavaScriptCore
• Status: RESOLVED INVALID

[Jaroslav Reznik]

JavaScript JIT causes crash while SELinux in enforcing mode is involved. 

Dec 23 09:09:38 htpc kernel: plasma-netbook[4493]: segfault at bbadbeef ip
0621a5c0 sp bf8f4b20 error 6 in libQtWebKit.so.4.6.0[5fcb000+1090000]  
Dec 23 09:09:38 htpc kernel: Process 4493(plasma-netbook) has RLIMIT_CORE set
to 0
Dec 23 09:09:38 htpc kernel: Aborting core
Dec 23 09:09:45 htpc setroubleshoot: SELinux is preventing
/usr/bin/plasma-netbook "execmem" access on <Unknown>. For complete SELinux
messages. run sealert -l 94d4513e-7637-4087-9039-4340e3c4b452
Dec 23 09:09:48 htpc setroubleshoot: SELinux is preventing
/usr/bin/plasma-netbook "execmem" access on <Unknown>. For complete SELinux
messages. run sealert -l 94d4513e-7637-4087-9039-4340e3c4b452


See recent Fedora bugs for more information:
https://bugzilla.redhat.com/show_bug.cgi?id=549994
https://bugzilla.redhat.com/show_bug.cgi?id=527079
https://bugzilla.redhat.com/show_bug.cgi?id=516057

Currently we ship QtWebKit with JIT disabled as workaround. Can someone guess
performance impact? Or any possibility to find fix in trac?

I've found this https://bugs.webkit.org/show_bug.cgi?id=22033 bug ([GTK] CTI/Linux
r38064 crashes; JIT requires executable memory) that could be related to this 
issue.

Version: QtWebKit as found in Qt 4.6.0 (but it affects WebKitGtk too - 
JIT is disabled in Fedora package too).

Thanks.

[Rex Dieter]

Confirmed selinux/execmem issue remains with current qt-4.8.0-rc1/qtwebkit-2.2

[Alexey Proskuryakov]

Isn't this expected behavior for a platform that disallows making memory executable? Of course, that means taking a significant performance hit.

[Gavin Barraclough]

As Alexey says, sounds like this is behaving correctly for your system.