https://bugzilla.novell.com/show_bug.cgi?id=732910 https://bugzilla.novell.com/show_bug.cgi?id=732910#c41 --- Comment #41 from Neil Rickert <nrickert@ameritech.net> 2011-12-26 16:49:46 UTC --- Responding to Phil Berlin in comment 40: That those inodes are the same is a symptom of the underlying problem. Let me repeat what I see as the probable bug(s) here: 1: It seems that "dhclient" is inappropriately closing file descriptor 2. If you are examing the source code, this would be happening in the child process after the fork(), since the parent process continues to have fd=2 open to "/dev/null" (based on "lsof" output). When the bash interpreter opens "/sbin/dhclient-script" to run the commands, it is presumably getting file descriptor 2 for that, so that stderr is now a read-only file descriptor to "dhclient-script". This is compounded by what I see as a kernel bug. And that is, writing to "/dev/stderr" (which is really "/dev/fd/2") does actually write to the file, even though fd=2 was a read-only file descriptor. I would have expected that either the attempt to open "/dev/stderr" for writing would fail, or the attempt to actually write to "/dev/stderr" would fail in these circumstances. I don't know enough about how "/dev/fd/n" was supposed to be implemented to be certain that this is a bug, but it sure looks wrong to me. I'll add another comment with my suggestions on what the opensuse team should do with this bug. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.