Greg Freemyer
What is dangerous about double mounting?
I don't think this was answered.
The issue comes in with Enterprise Storage Area Networks. They often have redundant paths to the same filesystem. Thus in a SAN environment it is easy to mistakenly mount the same filesystem two different places simultaneously.
If all reads/writes were atomic, that would be fine. But in the real world there are buffers and caches. The data in them often conflicts with what is on the actual storage system. Having 2 independent sets of buffers/caches quickly leads to chaos on a busy system.
I guess in a typical local system the kernel is aware of every second mount and would use the same buffers and caches at least I would believe that would normally hold true. It would not make sense to not be aware of a second mount, in that case every --bind mount would start causing trouble as well, or wouldn't it?. So it seems we have a special case here where the kernel can't be aware. And now XFS is using filesystem descriptions/identifiers (whatever those may be) to identify the filesystem (which would typically be stored on some device) but it can't distinguish identical or near-identical copies because the UUID obviously does not describe the entire thing like a checksum would. And even if it did, it would still be wrong. But in actual fact the condition arises easily where two filesystems with identical UUID can be on a different device, which means the logic for not mounting twice does not hold true, which means the FS UUID logic is just flawed and WRONG. And it also means you can't just say a person is lazy or a turd or "has a problem with looking up information" (as per mr. Anton) as if the bad design of some feature suddenly warrants every individual have some obligation to the Unix world to design around that flaw. I don't know if it makes more sense in the context of that SAN, perhaps there are also conditions where network shares cannot be identified by device because the devices on different systems may be different, but the filesystem itself may not? I don't know. Apparently we are talking about local systems here still. So I think it is just not right. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org