http://man7.org/linux/man-pages/man2/mmap.2.html says of MAP_NORESERVE > When swap space is not reserved one might get SIGSEGV upon a write if > no physical memory is available. I think this sentence should be removed, or perhaps qualified with specific circumstances or past kernel versions. It likely descends from documentation for some other operating system with different behavior, but as best I can tell from reading the kernel sources and experimenting, it does not describe Linux. In Linux, the effect of MAP_NORESERVE is only to set VM_NORESERVE in the vma flags, exempting the mapping from the "heuristic overcommit" accounting; it doesn't affect the fault path. And I know of no out-of-memory situation in which the kernel generates SIGSEGV. The closest I can see is touching a MAP_SHARED region backed by a full tmpfs, which generates SIGBUS. For anonymous mappings the kernel invokes the OOM killer. The man page later hedges > BUGS > > On Linux there are no guarantees like those suggested above under > MAP_NORESERVE. Perhaps this is meant to negate the SIGSEGV remark, but it seems more meant for the previous sentence which actually contains the word "guarantee". "One might get SIGSEGV" doesn't sound like any sort of guarantee to me. -- To unsubscribe from this list: send the line "unsubscribe linux-man" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
