Date: Sun, 18 Aug 2019 16:24:49 +0000 From: bugzilla-noreply@freebsd.org To: bugs@FreeBSD.org Subject: [Bug 229616] bin/dd does not check output's close() return code Message-ID: <bug-229616-227-zuEoOOx5ND@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-229616-227@https.bugs.freebsd.org/bugzilla/> References: <bug-229616-227@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D229616 --- Comment #3 from Thomas Hurst <tom@hur.st> --- So, did a bit of research to see what everyone else is doing. NetBSD: * Checks close() on output * fsync() on output if it's stdout (where close() may simply decrement a refcount) OpenBSD: * Does not close() at all * Supports conv=3Dfsync Illumos: * Checks close() on output, and also fclose(stdout) * Supports conv=3Dfsync and fdatasync * Supports oflag=3Dsync and dsync for synchronous writes Coreutils: * Checks close() on both input and output * Checks ferror()/fclose() on stdout if arg parsing fails * Supports conv=3Dfsync and fdatasync * Supports oflag=3Dsync, dsync, and direct * EINTR-checking loops for close() and f*sync() So FreeBSD's bit of an outlier, being the only dd without any fsync support= at all, and calling close() explicitly but not checking it succeeds. I'm thinking we could do a hybrid of NetBSD and OpenBSD's approaches: 1. Check close() 2. Also check fsync() if output is stdout 3. Add conv=3Dfsync support Other options might be nice to have, but aren't needed to ensure data is on stable storage after dd is finished, which is the most important point. --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-229616-227-zuEoOOx5ND>