From: Paul L. <pl...@li...> - 2003-01-29 14:53:31
|
On Wed, 2003-01-29 at 01:58, Dan Kegel wrote: > Reading > http://marc.theaimsgroup.com/?l=3Dlinux-kernel&m=3D104382026530895&w=3D2 > made me wonder: how does one test the write error handling > behavior of the kernel? Can we perhaps add a configuration > option to the ramdisk driver that lets us inject error > conditions via an ioctl? Then LTP could build that as > a module out-of-tree, and load it into the kernel, like > the module loading tests presumably do. >=20 > OK, it's a bit complex :-) Yes it is. :) Actually it seems there might be a potentially useful mechanism for fault injection in dprobes/kprobes. I've thought so for a long time but havn't had time to pursue it. I think I did hear something about a project recently that is aiming to build a fault injection tool using kprobes though. I .... was going to tell you to google for it but you piqued my interest again. It's at fault-injection.sourceforge.net. I don't know anything more about it right now, or about how far along they are though. Thanks, Paul Larson |