Hi, I'm the lead developer for the Puppy Linux project. The last release of Puppy used both unionfs and aufs, depending on the conditions. Normally unionfs was used, but booting from a Flash drive I have a system of periodically flushing ram to a save-file on the Flash drive. Aufs allowed true flushing, that is it can actually move files from the rw top layer down to a lower layer (branch), rather than just copying down. That was using the 2.6.18.1 kernel and unionfs 1.x.
I'm now developing Puppy with the 2.6.21.5 kernel, and I applied the unionfs2 patch. Note, I'm particularly interested in the improved branch management. For the last week or so I've been using this same kernel with aufs, and all is fine, I've been surfing with SeaMonkey. Everything is the same now, just switched to unionfs, and everything looks fine, but after using SeaMonkey for about a half-hour, it crashed with this in /var/log/messages: BUG: unable to handle kernel NULL pointer dereference at virtual address 00000010 15 minutes later, again the same crash. I've been trying to narrow it down. I was only visiting the same web pages as usual, always in the past SeaMonkey has been rock solid. The only different factor here is unionfs. I'm using the latest squashfs, with the lzma patch. The kernel otherwise has only a few minor patches, so is basically plain-vanilla. I know this is inadequate information to fix the bug, or even positively identify unionfs as the culprit! But I do know that I was using aufs before that, exactly same setup and it ran fine. Note, I would like the 2.6.21 patch to be supported/upgraded, as major distros like Slackware 12 are and will be coming out using it, including Puppy. I would also like to request being able to compile the module separately without having to recompile the kernel. Regards, Barry Kauler _______________________________________________ unionfs mailing list: http://unionfs.filesystems.org/ [email protected] http://www.fsl.cs.sunysb.edu/mailman/listinfo/unionfs
