SIGSEGV不会终止该过程

Question

I have a multithreaded application that runs 30-something threads. I know there is a bug where sometimes two threads attempt to sort one list simultaneously and this usually results in one of the threads accessing invalid memory. Thus, a SIGSEGV is generated for that thread.

Now, from what I understand about signals, the thread should call SIG_DFL for SIGSEGV , which is abnormal termination of the process and a coredump. However, I saw the process being still alive (in a kind of limbo state) where the execution halted, but the process was still alive. When I tried to kill it, the SIGTERM was actually propagated to my custom signal handler (which attempts to nicely shut down all the threads), but there it hang up, because none of the threads were actually executing anymore. I finally managed to kill it using SIGQUIT and the corefile was generated after that.

So my question is what is meant by "abnormal termination"? How can a process not be removed from the memory if the SIG_DFL is called for SIGSEGV ? What could possible be going on that caused such behaviour? My Linux is Red Hat Enterprise Linux Server release 5.11 (Tikanga)

EDIT: I know (more or less) how to debug it and I even know what the bug is. My question is more or less: What exactly RedHat does when SIG_DFL is called with SIGSEGV? The problem here is that the process was not responding, but was not dead either - therefore the automatic restart procedure did not kick in and we had some unpleasant downtime.

Answer 1

The situation was not as straightforward as I thought. The original problem was a deadlock between two of the threads. When I issued SIGTERM, then actually my custom signal handler caused a segfault in the (now un-)deadlocked threads.