> Date: Sun, 31 Mar 2013 09:38:25 -0400 > From: "Cliff McDiarmid" <[email protected]> > To: "BLFS Support List" <[email protected]> > Subject: Re: [blfs-support] Autofs problem on LFS7.2 > > > Cliff McDiarmid wrote: > > > > > Got it! After starting autofs with 'automount -v -f -d . . > > > > Cliff, it would be nice to write up a short summary of what was wrong > > and the ultimate fix. > > Right, I'm going to attempt to summarise a very long thread which addressed > an automount problem(once again I must thank akhiezer for his help and > patience in solving it). > > Autofs-5.0.7 was built according to the compile/install instructions for > autofs in the blfs book, but I did a 'client-only' install of OpenLDAP-2.4.33 > and at one point compiled it against MIT Kerberos V5-1.10.3, but this was > later removed. > Either way I don't believe this had any bearing on the outcome. > > Autofs- 5.0.7 always started from boot up but failed create any dir. > specified in the '/etc/automaster' file or to mount it when a usbstick was > inserted. > > I believe(correct me if I'm wrong akhiezer)that the solution boiled down to > two changes that were made. > > 1) I reverted a change made by the > 'autofs-5.0.6-fix-libtirpc-name-clash.patch' to the file 'lib/rpc_subs.c' in > the source distrib, and commented-out the code: > > --- > #ifdef WITH_LIBTIRPC > #undef auth_destroy > #define auth_destroy(auth) \ > do { \ > int refs; \ > if ((refs = auth_put((auth))) == 0) \ > ((*((auth)->ah_ops->ah_destroy))(auth));\ > } while (0) > #endif > --- > thus: > === > /* > #ifdef WITH_LIBTIRPC > #undef auth_destroy > #define auth_destroy(auth) \ > do { \ > int refs; \ > if ((refs = auth_put((auth))) == 0) \ > ((*((auth)->ah_ops->ah_destroy))(auth));\ > } while (0) > #endif > */ > === > > I then recompiled Autofs. > > 2) I adjusted the '/etc/auto.master' file to read: > > --- > /testautomount file:/etc/auto.testautomount > --- > > rather than: > > --- > /testautomount /etc/auto.testautomount > --- > > This was to be more explicit about the type of map that was being used. A > minor change, but one that worked. > > > *Please note that the 'auto.testautomount' dir. was a test point and this > would normally be 'auto.misc'.* > > > Cliff >
I'd expect that 'reason 2)' is a bit of a red-herring: it's likely a 'chmod 644 /etc/auto.testautomount' change that would make the difference; the 'file:' stuff is really only to override any exec-permissions that are still on '/etc/auto.testautomount' . Here, on autofs-5.0.5, having any exec-perms on '/etc/auto.testautomount', and no 'file:' or 'file,sun:' &c labels in auto.master, causes automount to erroneously try to interpret the map-file as being a program. Then, if I do use 'file:' or 'file,sun:' in auto.master, and whether '/etc/auto.testautomount' has any exec-permissions set or not, then automount interprets the map-file as being an ordinary file and not as a program or any other type. As for 'reason 1)': per earlier note today, one would really want to do the recompile with the original libtirpc code in its original form (i.e. _NOT_ commented-out), and using the './configure ...... --without-... --without-hesiod --without-libtirpc --without-... ....', and with '/etc/auto.testautomount' having permissions 644; and see if all works ok. If it works ok, then one would go 'round the loop again with the sole, single change of omitting the '--without-libtirpc' from the './configure ...' line, and see if things now _don't_ work ok. If it does break, then go 'round the loop again but this time make the sole, single change of commenting-out the libtirpc block of code, and see if things now work ok. Without that or similar methodology, in particular (in this case) changing just one thing at a time, then one can't really put good reliable info into book/wiki, or in report to upstream, or even into one's own notes. rgds, akh -- -- http://linuxfromscratch.org/mailman/listinfo/blfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
