>> Rejecting won't help much, there are so many other parts that may be >> corrupt that you cannot validate on mount. > For start we want to stop the kernel from crashing on mount.
So you'd rather have it crash at some unpredictable time after mount? Okay, that's mostly snark, but there's a serious point lurking. >> The goal should be to gracefully handle corrupted data structures by >> returning errors instead of crashing the kernel. > mbouyer@ wants to panic always, after a successful mount. And, sometimes, I think that's the rightest choice. But I also would like to be able to get errors instead of panics. Want to pull something off a thumbdrive? I'd rather have an error just forcibly unmount the filesystem and flush everything using it: mount -o onerror=unmount /dev/sd1e /mnt But for the machine's own filesystems? Corruption should panic: mount -o onerror=panic /dev/wd2a /builds Of course, actually making that work, well, I don't have any suggestions for cat-bellers. Unless and until I have a significant amount of spare time, it's all just "it might be nice if". /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mo...@rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B