Re: [Libguestfs] [PATCH nbdkit] sh: Remove assert and replace with smarter file descriptor duplication. (was: Re: [nbdkit PATCH v2 14/17] sh: Use pipe2 with CLOEXEC when possible)

On 8/27/19 6:47 AM, Richard W.M. Jones wrote: Hmm, so we've closed stdin/out because the client connection is no longer around, but still give the shell script one more callback. Yeah, the failure is definitely explainable, and worth fixing. It's also possible to open /dev/null in a loop until getting one larger than STDERR_FILENO (so you'd have up to four open() calls), then do pipe() with the assurance of no collisions, then close the scratch fds. But that's probably more syscall overhead and not worth doing, so your attempt at doing the fd shuffling correctly is better. Not quite. I don't think POSIX guarantees that fd[0] < fd[1] when pipe() succeeds. But in the situation we're hitting, fd 0 and 1 are closed when we start pipe()ing, so our first pipe will claim 0 and 1 in either order, then the first check will either be a no-op (because fd[0] is 0 and doesn't need moving) or else fd[0] will be 1 and needs to overwrite stdin. Overwriting is fine; but if fd[0] is 0, we HAVE to clear FD_CLOEXEC, otherwise stdin will be closed by exec(). But it's close; so I'll use it as the starting point and push the corrected version soon. -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org