On 5/4/2012 9:37 AM, Jan de Kruyf wrote: > Right, back from away. > > John: > You answered my question quite adequately. And I referred to > PREEMPT_RT when writing RT. Jus t my fingers are so thick. > > All: > I seem to miss the stacktrace leading up to the crash in the > reports. So, sadly, there is little help I can offer since you guys are > all so wrapped up in the idea that there is perhaps an underlying problem > in the kernel or in glibc.
I haven't been thinking there's a bug in glibc or the kernel. While the code is crashing in malloc() or (with electric fence) in vfprintf, this is almost certainly due to some other bug (likely memory corruption) that happened earlier in the linuxcnc code. After reviewing some code last night, I think the segfault I was seeing when running electric fence may simply be an earlier occurrence of the later malloc() segfault, and rather than implicating the printing routines it could be indicating partial corruption of the malloc structures (or some other memory) earlier than the nasty crash that happens later. As for the lost mutex, that seems to be due to the process crashing, rather than any fundamental problem with the code. A backwards summary of symptoms when things are broken: * The hal_data->mutex bit gets stuck * The rtapi_app process segfaults (in malloc()) * rtapi_print_msg called from rtapi_reset_pagefault_count triggers a segfault when running electric fence. This could indicate an issue with the message print routines, or be another side effect of already corrupted memory. ...that's about as far as I've gotten for things that I'm pretty certain should *NOT* be happening. Backing up further, I see the rtapi_print_msg calls causing writes to the stderr _IO_STRUCT, but this could be normal behavior. I was just guessing at memory to watch for corruption, and since prints to stderr seem to cause problems, watching the stderr _IO_STRUCT seemed like a good idea. -- Charles Steinkuehler char...@steinkuehler.net ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
