SIGSEGV This! SIGBUS That!

看文档看到的一则说明：

We saw that some virtual memory sections aren't mapped to physical memory. While some of them were simply paged out, others were never allocated by the process. When a process runs, its virtual memory table is small. As it allocates more memory pages, the table grows. However, if the process tries to access a virtual memory address of a section it hasn't allocated yet, the operating system has no where to bring this page from. The designers of the Unix system decided that this situation indicates a program bug, and thus instead of making an automatic allocation of a memory page in such a case, they chose to send a signal to the process. This signal is a SEGV signal (or SIGSEGV), and its default signal handler prints out a "Segmentation violation - core dumped" message, and dumps the memory image of the process into a file named 'core' in the process's current directory.

Another way to cause a 'segmentation violation' is trying to access an illegal location of virtual memory. Because many invalid pointer problems occur with very low pointer values, the operating system does not allow a process to allocate a memory page for the virtual memory section beginning with the virtual address '0'. This is what causes programs to receive a SEGV signal when trying to dereference a NULL pointer (NULL on _most_ machine architectures is defined as '0').

What about a BUS (or SIGBUS) signal? this signal is sent to a program that tries to access a non-aligned pointer. For instance, on many machine architectures, access to 'long' (4 byte) numbers must be done using a memory address that divides by 4. Trying to access such an entity using an address that does not abide by this rule will cause the CPU to emit a trap. The operating system's kernel catches this trap, and then sends a BUS signal to the program. The default signal handler for this signal emits a "Bus error - core dumped" message, and dumps the memory contents to a 'core' file, much like the handler for the SEGV signal does