The init script for openafs searches only an ext2 partition mounted on /usr/vice/cache. Actually, openafs requires that the /vicepX partitions reside on non-journaled filesystems, but, anyway, if someone wants the cache to be on ext2 partitions, it's mandatory to check whether it is on a non-jouraled partition, not if it's an ext2 mount point. Usually, the cache as well as the etc directories on /usr/vice are symlinks. I propose a modification for the init script. Reproducible: Always Steps to Reproduce: 1. Use a symlink for the afs cache directory Actual Results: No matter if it points to an ext2 partition, the check will always fail. Expected Results: Check whether the cache directory is on an ext2 partiton, checking also symlinks.
Created attachment 37469 [details, diff] The init scripts checks the afs cache over the filesystem, not only as a partition With this patch, OpenAFS checks whether /usr/vice/cache is mounted or points as a symlink to an ext2 partition. There's another issue we should take care of: we must check if the cache is on a non-journaled partition, not if it's simply an ext2 filesystem.
Hey, openafs runs on journalled filesystems, sure. I think you mess it up a bit. Currently, openafs on linux can run the server part (i.e. the stuff forked under bosserver) on *any* partition. The fileserver process is on linux to be namei based by default (inode based version isn't available at all). So, I run 1.3.73 on xfs partititon. With the afsd kernel process, called cache manager, yes it can run safely only on ext2 or ext3.
The check for cache *must* take care of symlinks. I proposed a patch. The cache can reside on a journaled fs, but anyway it's suggested to use ext2, as you can see from the original init script. I'm not arguing about using ext3, but just allowing symlinks.
OpenAFS needs a developer to take up maintenance.
New ebuilds for openafs 1.2.13 (stable) and 1.3.85 (experimental) are available for testing. According to openafs-ml, 1.3.85 is currently undergoing testing so it can become 1.4rc. The new init-scripts don't bother with the check where the cache partition is mounted anymore, it's left as the administrator's responsibility (choose symlink, ext2/ext3, dedicated/non-dedicated partition as you like)... It'd be great if the new proposed ebuilds could be tested
The check for an ext2 partition as a cache has been obsoleted. Hence the problem will not appear anymore.
