Red Hat Bugzilla – Bug 1497846
autofs option parsing for maps with fstype=autofs no longer works
Last modified: 2018-04-10 14:18:00 EDT
Description of problem: Macro definitions which are specified as options to the mount entry are not being set, and are therefore not available for use in substitutions. Version-Release number of selected component (if applicable): autofs-5.0.7-69.el7.x86_64 How reproducible: see below Steps to Reproduce: samba server with share //server/share auto.master: /smb /etc/auto.cifs.top --negative-timeout=1 --timeout=20 auto.cifs.top: * -fstype=autofs,-Dhname=& file:/etc/auto.cifs.sub auto.cifs.sub: * -fstype=cifs,credentials=/root/cifs_mount_creds,noserverino,nounix ://$hname/& # ls /smb/server/share Actual results: ls: cannot access /smb/server/share: No such file or directory the 'hname' macro is not defined in the table, so the '$hname' is replaced by an empty string when used in 'auto.cifs.sub' Expected results: 'hname' macro gets defined, and is available Additional info: this has worked in previous autofs versions
(In reply to Frank Sorenson from comment #0) > Description of problem: > > Macro definitions which are specified as options to the mount entry are not > being set, and are therefore not available for use in substitutions. Indeed they aren't. It seems a "fix" I added actually introduced a regression. Ian
Created attachment 1333480 [details] Patch - revert fix argc off by one in mount_autofs.c
*** Bug 1498443 has been marked as a duplicate of this bug. ***
Hello, I understand that has been a commit that has seemingly resolved the issue. Customer in case 01940146 is interested in any potential test package that has result from this commit. Understanding that this seems to be in package 5.0.7-71 I do not observe this to be GA. Apologies if I am misunderstanding the current state. Thanks
(In reply to smazul from comment #7) > Hello, > > I understand that has been a commit that has seemingly resolved the issue. > Customer in case 01940146 is interested in any potential test package that > has result from this commit. Understanding that this seems to be in package > 5.0.7-71 I do not observe this to be GA. Apologies if I am misunderstanding > the current state. Right, the process is that changes are committed and built into a package during the release development phase, following development the release enters testing phase where QA verifies changed packages and finally it gets to a GA release. All of this can take quite a while. We can give the customer revision 71 (or the current latest revision) but that package can have other changes that have seen limited testing and isn't (strictly speaking) supported. Also the package may have further changes from the development phase or the testing phase so it may not (and likely won't) be the package that is released at GA. Locate the support documentation on providing an accelerated fix for the options and requirements for delivering an appropriate package to the customer. Ian
Now that I look at this patch it looks to me like this will affect option parsing which goes far beyond macro definitions. If I'm understanding correctly, effectively the bug is that the last 'option' in the chain gets ignored, so it could be any option in the list. I'm going to do a few more tests. If this is true, then an effective workaround may just be to add a garbage option at the end, which will be ignored.
BTW, I verified a simple 'ro' got ignored so it definitely is a more generic problem as the patch reads. For some reason adding a garbage option at the end does not work around the problem so there's something I'm missing there.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2018:0977