hckrnws

Not being able to chown() caused us grief developing Frame Maker back in the 80s. The responsible way to handle "save" was to write the document into a new file mydoc.new, then rename mydoc.cur to mydoc.backup and then rename mydoc.new to mydoc.cur, so that failure never left you in the lurch. The only problem was that there was no way to create mydoc.new to have the same owner as mydoc.cur and customers complained that we'd keep changing the owner of their files. If only the semantics of the unix filesystem supported file generation numbers, like on Tops20 or VaxVMS, where the default for writing to a file isn't "yeah, sure, write over top of the old data, and let's hope nothing fails along the way" this would not have been a problem.

> caused us grief developing Frame Maker back in the 80s

To be fair, Frame Maker caused the rest of us a whole lot of grief back then, too. :)

The license manager daemon, lmgrd (?) would crash regularly enough that we just patched the dependency out of our binaries. Sorry about that!

ive always felt that file systems are by far the weakest point in the entire computing industry as we know it.

something like zfs should have been bog standard, yet its touted as an 'enterprise-grade' filesystem. why is common sense restricted to 'elite' status?

ofcourse i want transparent compression, dedup, copy on write, free snapshots, logical partitions, dynamic resizing, per-user/partition capabilities & qos. i want it now, here, by default, on everything! (just to clarify, ive ever used zfs.)

its so strange when in the compute space you have docker & cgroups, software defined networking, and on the harddrve space i'm dragging boxes in gparted like its the victorian era.

why can't we just... have cool storage stuff? out the box?

All of those things come with tradeoffs.

Compression tradesoff compute vs i/o, if your system has weak compute, it's a bad deal. Most modern systems should do well with compression.

Dedupe needs indexing to find duplicates and makes writes complex (at least for realtime dedupe). I think online dedupe has pretty limited application, but offline dedupe is interesting.

Copy on write again makes writes complex, and tends to fragmentation of files that are modified. Free snapshots are only free when copy on write is the norm (otherwise, you have to copy on write while a snapshot is open, as on FreeBSD UFS). Copy on write offers a lot, but some applications would suffer.

Dynamic resizing (upwards) is pretty common now. Resize down less so. Zfs downsizing is available, but at least when I tried it, the filesystem became unbootable, so maybe not super useful IMHO.

Logical partitions, per user stuff, qos adds complexity probably not needed for everyone.

> Compression tradesoff compute vs i/o, if your system has weak compute, it's a bad deal. Most modern systems should do well with compression.

Older systems with worse compute also had worse i/o. There are cases where fast compression slows things down, but they're rare enough to make compression the better default.

I certainly don't want my compiler to get slower, because we now compress the files that are gone in a few minutes anyway. Compression is useful for archiving files, but for anything that you currently work with, it's useless and only wastes compute.

If you're limited by your SSD, one core running lz4 (or zstd) will double your write speed for object files. If you're not writing hundreds of megabytes per second, then you'll barely notice the overhead at that phase while it makes later phases that load the data back snappier.

If everything fits in ram then compression could be postponed.

And for that area in between, where your files don't fit in ram but compressed they would fit in ram, compression can give you a big speed boost.

That's true. I think less memory is often accompanied with less compute. For example "one core running" is 50% of my available compute.

Because the vast majority of personal computer users have no need for the complexity of zfs. That doesn't come for free, and if something goes wrong the average user is going to have no hope of solving it.

FAT, ext4, FFS, are all pretty simple and bulletproof and do everything the typical user needs.

Servers in enterprise settings have higher demands but they can afford an administrator who knows how to manage them and handle problems. In theory.

FAT bulletproof? The newest versions have a few improvements but this is a line of filesystems for disposable sneakernet data.

Maybe bulletproof is a bit strong but I mean, it was fine on DOS/Windows for decades. I never lost data due to filesystem corruption on those computers. Media failures, yes frequently in the days of floppy disks.

Running scandisk and chkdsk was also a thing for decades. Luckily the lost sectors were often in unimportant files. But definitely a gamble.

The bulk of the safety came from the redundancy of copying the file across machines, not filesystem protections.

I had a HD fail on me while using Windows 98 as main OS, yet thanks to ext, I think it was ext2 at the time this happened, I still managed to repurpose it for Linux, for several months.

It was ok from possible data failures point of view, I didn't had much data other than the distro and the stuff I also needed to compile under Linux.

Somehow it managed to still work with the disk, with the sectors that were not damaged.

Because it was extremely difficult to create something like zfs? And it was proprietary and patent-encumbered, and the permissively licensed versions were buggy until about 5 minutes ago?

That's like saying the Romans should have just used computers.

I would guess that many early systems just didn't have the storage space for a lot of multiple versions of files. Was VMS saving diffs or full copies of files?

Once storage space was plentiful, the pattern of "overwrite the existing file" was already well established.

Typical TOPS-20 and VMS hardware of the time would have less than a gigabyte of spinning disk space, to be shared among many dozens of users. Full copies of files were saved, and there were strict per-user disk allotments. Creating Generation 2 of a file would mark the Generation 1 version as deleted. When you ran out of allotment during execution, the OS would pause your program and give you the chance to issue an Expunge command to really recycle all (or a subset) of the deleted files, and then you'd just Continue the paused process. Similar to desktop "Trash" folders where deleted things go, and that you may have to Empty once in a while.

If you could chown files to an arbitrary other user, you could use that to evade disk quotas.

The protocol for changing ownership should be two step.

1. The file is put into an "offered" state, e.g. "offered to bob". Only the owner or superuser can make this state change.

2. Bob can take an "offered to bob" file and change ownership to bob.

Files can always be in an offered state; i.e. have an offered user which is normaly equal to their owner. So when ownership is taken, the two match again.

Why would you use ownership data for disk quotas? That's crazy. It should be based on file location. Anything in the user's home directory comes out of their quota. I assume that's how it does work surely?

What if Alice leave her home directory permissions open? Mallory can then move a large file there and have it assigned to Alice's quota.

I suppose it could work, but Mallory would then risk Alice blowing that file away (perhaps innocently, not even realizing it was Mallory's file).

If I leave my front door unlocked and you put a large cow in my house, it's taking up space in my house.

Not if the file is in the shared /tmp/

What's the deal with disk quotas? Saw that in the OT as well. Why would you measure folder size seperately for each and every user? Would it not be a lot easier to just use the disk space of a folder regardless of whomever the file belongs to?

Because, in the early days of Unix systems actually being used as multiple, simultaneous, user systems, you might have one group of users collaborating on a project, and they would have a shared directory (via the 'group' owner) where they would store shared items. Each user would create various files, and each file's space consumption was charged to that user, but the shared directory might contain multiple files each owned by different users (but all owned by the shared 'group' identifier, so the group could access them).

For a group shared directory, assigning the disk space usage of files therein to one single user (ignoring the aspect of "which single user do you pick") is unfair to that user (his/her allowed maximum disk space is consumed) while everyone else is not charged for their actual usage.

This all came about to try to enforce rules to prevent one (or a few) rogue users from using up all disk space on the system for themselves, leaving no one else with any disk space available for their own usage.

It's not folder size that you measure, but a user's usage: how many blocks are occupied by files belonging to a certain user, no matter where they are.

That's what quotas are: per-user storage limits.

If Bob has a large file which is sitting in Alice's home directory, that counts toward's Bob's quota, not Alice's. If Bob could sneakily change the ownership to Alice, while leaving the permissions open so he could access the file, then the file counts toward Alice's quota.

One reason why Unix quotas are generally not maintained and imposed by path is that it's a lot easier to update quotas as things are created, deleted, modified, and so on if the only thing that matters for who gets charged is some attribute of the inode, which you always have available. This was especially the case in the 1980s (when UCB added disk quotas), because that was before kernels tracked name to inode associations in RAM the way they generally do today. (But even today things like hardlinks raise questions.)

(I'm the author of the linked-to article.)

> Forbidden

> You don't have permission to access /~cks/space/blog/unix/ChownRestrictionEarlyHistory on this server.

I laughed out loud.

https://web.archive.org/web/20251018101005/https://utcc.utor...

It was one of those restrictions that seemed unjustified to me but I figured someone smarter than I had seen a reason.

Yeah.. I'm sitting here wondering how many years would it take to remove equally stupid error that says 'private key permissions too open' from ssh-add and friends.

Would save me a wrapper script on my flashdrive that does hacks like loading it from stdin or moving it to temp file.

It's just a nice security measure.

It would need at least a little bit of thought with suid binaries.

Suid binaries were a bad idea and should be removed anyways.

They're elegant at first glance, it's just the attack surface of "exec this other program" turns out to be enormous. eg. You wouldn't necessarily think that you'd have to clean out all environment variables, but what if your program is linked to some library that tries to parse some obscure environment variable and that parser has a bug. And the list goes on.

The alternative is sending messages to daemons, but as it turns out, the attack surface of those is pretty large too, albeit not as large as setuid.

The whole "do some work on my behalf with elevated privs" is not exactly a solved problem in Unix.

Imagine if you wanted to enter a bank safe, but your key doesn't fit the lock. If you were able to change the lock, you would bypass the lock mechanism, rendering it useless

But imagine if you were the bank-safe owner. Shouldn't you be able to change the lock?

That would be what root is.

I think a more appropriate question would be, if the key fits, couldn't you change the lock?

Maybe, that would give you 3 abilities.

1 Lock yourself out if you please? Not terrible

2 Provide access to others, which makes sense since you already have access to the file, you could theoretically share it through other channels, you naturally cannot prevent this.

3. Lock others out. This one is less of a security risk and more of a nuisance risk.

I think the unix model is simple, maybe selinux offers more sophistication. That said the unix chown behaviour could have gone either way in terms of security, but in terms of design it makes sense as is.

> That would be what root is.

In this analogy, I think the analogue of the owner of the bank safe is the owner of the file. Unless you're envisioning the bank safe as representing all the files, rather than just one ...?

No, the bank owns and administers each safe, they lease it to you. They would be root.

UNIX ownership isn't necessarily legal ownership, files are not real property.

Wait. You can use chown as non root?

Yes, in SysIII and SysV. Per the article.

It was possible to chown/chgrp as non root in Solaris up to some version that I forget.

This reminds me of an interesting security incident that occurred on the undergrad.math.uwaterloo.ca unix cluster while I was an undergrad, circa 1996.

When I'd started, the cluster had three SunOS servers, named cayley, descartes, and napier; undergrad math students had their home directory allocated on a local disk on one of these three machines, which each cross-mounted the others' via NFS. At this time, however, the Math Faculty Computing Facility had just received a fancy new dedicated NFS file server from (IIRC) NetApp, and all our home directories had been moved there instead, presumably freeing up desperately-needed CPU cycles on the three compute servers so we could run the Modula-3 and μC++ compilers.

One evening I was in one of the XTerm labs in the Math and Computer centre working on a CS assignment (the only alternative being to do from my dorm room via 2400 buad dialup). As was tradition, I had left the assignment until the night before it was due to start work on. Indeed, it seems that we all must have, because after getting part way through I needed to access some input data files that were shared from the home directory of the course account—something like ~csXYZ/assignments/N/input—only to find I could not read them.

These files were of course owned by the csXYZ course account and should have been either world-readable or readable by the corresponding csXYZ group to which all students registered that term belonged. Unfortunately something had gone wrong, and although the files were rw-r-----, they belonged to the wrong group, so that I and the other students in the class were not able to access them.

It now being after 6pm there was no hope of tracking down one of the course professors or the tutor to rectify this before morning (and it's quite likley the assignment submission deadline was 9am).

Fortunately, I was a naive and ignorant undergrad student, and not knowing what should and should not have been possible I began to think about how I might obtain access to the needed files.

I knew about suid and sgid binaries, and knew that on these modern SunOS 4 machines you could also have suid and sgid script, so I created a script to cat the needed files, then changed its group to match that to which the files belonged, then tried to chmod g+s the script—but of course this (correctly) failed with a message informing that I could not make my file sgid if I didn't belong to the group in question. I then took a different tack: I chgrped the script back to a gropu I did belong to, ran chmod g+s, then chgrped the script back to the group that owned the files I wanted to read.

I now know that this should have resulted in the script losing its setgid bit, but at the time I was unaware of the expected behaviour—and it seemed that the computer was as ignorant as I was because it duly changed the group as requested without resetting the setgid bit, and I was able to run the script, obtain the files I needed, and finish the assignment.

I then headed over to the CS Club office to discuss what had happened, because I was somewhat surprised this had worked and I wanted to understand why, and I knew that despite the lateness of the hour the office would certainly be open and very likely contain someone more expert than I who would be able to explain.

The office was indeed open but no explanation was forthcoming; instead, I was admonished not to discuss this security hole with anyone until I had reported it, in person, to the system administrators.

Thus it was that bright and early the next morning I found myself in Bill Ince's office with a printout of the terminal history containing a demonstration of the exploit in hand. I informed him I had a security issue to report, and handed him the printout.

He scanned the paper for a moment or two, and then replied simply "ahh, you found it".

It seems I was not the first to report the issue, and he explained that it was due to a bug in the new NetApp file server. He then turned monitor of the terminal on his desk around to show me a long list of filenames scrolling by, and (in hindsight rather unwisely) informed me that it was displaying a list of files that were vulnerable to being WRITTEN to due to the same hole.

He duly swore me to secrecy until the issue could be resolved by NetApp (which it was a few days later), thanked me, and sent me on my way.