On 06/05/2015 06:15 AM, Adam Conrad wrote:
On Fri, Jun 05, 2015 at 05:23:35AM -0800, James Long wrote:Hi Andreas, My problem is actually with unshare(2), rather than unshare(1). Is there an equivalent patch for unshare(2)?That doesn't make much sense. Your bug report was about the command line utilite (unshare(1)), while unshare(2) is a C function. The commit pointed at does seem to address the issue that when running under systemd (which you are), the default mount sharing is "shared" instead of "private", which is mildly unintuitive. I'd assume the patch would fix your issue, but it would be nice to test that theory. ... Adam
I apologize for any confusion, but it seemed easier to illustrate the problem with unshare(1), whose new default behavior in jessie mimics exactly the problem I am having with unshare(2) in jessie.
More specifically, a call to unshare(CLONE_NEWNS) would previously (wheezy) let me mount resources for my process that were invisible to other processes, and the mounts would go away once my process exited.
Now (jessie) the same call results in mounts that are visible to all, and persist after the process exits.
Thanks, Jim -- James Long Information Systems Manager International Arctic Research Center University of Alaska Fairbanks jlong15 at alaska.edu (907) 474-2440 -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
