Hi Ian! Hi Mathias! Please see below why I think this might be important for you as well. > > I *think* I understand what you describe there. If I were more > > knowledgeable about the internals of Linux and autofs I could be of more > > help :( > > No matter, the point was only that this unexpected case needed to be > caught so I could send a mount request to the daemon. The VFS is quite a > complicated beast and I'm certainly no expert but have become familiar > with much of it, to some small extent, through working on autofs.
Must have bean painful ;) > > Looks fairly simple - shall I test it even with your warning in the > > other mail? > > Think so. It didn't apply cleanly to 2.6.20 so I changed it manually. I just tested with only rebuilding the autofs4.ko-module and my kernel now thinks it's tainted (wtf...), BUT it works! I can't believe it! A one-liner, or, actually, a one-bitter ;)! I'm currently rebuilding the whole kernel so that I have a new .deb that I can roll out and test in a broader environment, but the exact use case of a recursive automount now seems to work perfectly well. Congratulations and thank you! After testing, I'll push this forward to Ubuntu, but I don't have high hopes of it to be included in any update. It should definitly go into 2.6.24, I'll report back if it breaks anything else. If I understand this correctly, would this be the fix needed for the 'nested mounts with loop'-problem Mathias Koenig sees, as the problem seems to be quite a similar one as ours ... > Turns out that I had some other problems, not the least of which is that > I'm running a version of autofs that that has changes I've just started > testing and I have a kernel that requires a "nosharecache" mount option > (that wasn't in my autofs configuration) for autofs to work (don't ask > it's not worth it). So there were lots of fails to begin with, but in > reality the patched module seem to be working ok. I recently read about nosharecache & stuff on kerneltrap.org so I feel with you ;( > > > > > Ian > > > > I really appreciate what you're doing, Ian. This is tough stuff and to > > me it looks very scary. > > Ya, I guess it is but I've spent a long time working on autofs so I > should know what's going on (you'd think). The other thing to remember > is that, with version 5, we've only just now reached a position where > our feature set is comparable to other industry standard automounters > that have had many years of maturation. People tend to forget this and > can't understand why we continue to find bugs and things that don't > quite work, but we are making strong progress. How right you are. On older Linux versions, we used a custom-patched amd that required a working hesiod installation which we definitly want to get rid of ... > Ian -- Lukas _______________________________________________ autofs mailing list [email protected] http://linux.kernel.org/mailman/listinfo/autofs
