Why GNU su does not support the `wheel' group (2002)(ftp.gnu.org) |
Why GNU su does not support the `wheel' group (2002)(ftp.gnu.org) |
Back in the day, it was about multiple OS users on one big machine, maintained by a university or a corporation.
Now I'm the only human user of my several machines. I have more than one interactive user account on some of them. I put these accounts to the wheel group, to avoid ever using a root password. (Void Linux has it pre-configured in /etc/sudoers.)
Basically https://xkcd.com/1200/
It's not wheel's fault, and so substituting root group for wheel was nothing more than a special snowflake move by someone who thought the sysadmins were holding back and the "enemy".
Why have multiple different unix users if all have the same power of the root user? They aren't isolated from each other, they aren't less powerful than root. So why not just use the already existing single root account?
Second, it can still be useful for bookkeeping.
Though I suppose, with further thought, it’s not significantly worse than having them in sudoers, in that particular respect.
Back in the day, university campus networks popularized a "cluster" approach of logging in to specific machines whereby hosts followed the pattern: $(uname)[0-9]{2}. You would soon setup authorized_keys and a script that would check the load of all machines by iterating ssh. The same home directory would be shared to all machines, regardless of OS, so portability of shells scripts and POSIX/C code was necessary.
actually if you think about it, lots of machines you OWN nowadays don't give you root access.
For example you can't get access at the actual filesystem on your iphone without a jailbreak.
https://en.wikipedia.org/wiki/Wheel_(computing)
The term wheel was first applied to computer user privilege levels after the introduction of the TENEX operating system, later distributed under the name TOPS-20 in the 1960s and early 1970s. The term was derived from the slang phrase big wheel, referring to a person with great power or influence.
In the 1980s, the term was imported into Unix culture due to the migration of operating system developers and users from TENEX/TOPS-20 to Unix.
By convention, "wheel" is a special Unix user group that determines who can use "su" and "sudo". Most "su" and "sudo" implementations allow the sysadmin to make their use exclusive to the trusted users inside the "wheel" group. In most systems, it's the default setting of "su", and optional for "sudo" (given as an example in /etc/sudoers).
> if someone has the root password, can't they change what groups they're a member of?
No. If "su" is configured to be "wheel"-exclusive, you can't log in as root even if you have the password, because you cannot use "su" - unless you have direct access to the system console that allows you to type "username: root", which is almost never the case on servers that disable remote root login.
Users who aren't in the wheel group aren't supposed to be able to become root, even if they have the password.
Isn't it just supposed to switch user? And what are the implications of not-checking whatever it was supposed to check?
Someone who steals the root password (say, by looking over the sysadmin's shoulder) would be able to become root.
And I also don't get: if someone has the root password, can't they change what groups they're a member of?
No, because they can't log in as root and (on non-broken systems) can't become root.
Or is the wheel group not really about being able to sudo?
It removes the section at the bottom of the man page, and has this addition to the source code comments:
+#ifdef USE_PAM
+
+ Actually, with PAM, su has nothing to do with whether or not a
+ wheel group is enforced by su. RMS tries to restrict your access
+ to a su which implements the wheel group, but PAM considers that
+ to be fascist, and gives the user/sysadmin the opportunity to
+ enforce a wheel group by proper editing of /etc/pam.conf
+
+#endif- it makes it smaller, less code which can go wrong
- su isn't limited to "set user root" but wheel tends to be
- it avoids having to handle many kind of subtle problems with group based permission handling in linux
It's just not a bad idea to have a very minimalist program like su and then delegate all more complicated "acting as user" permission handling to other programs like sudo or doas.
Through tbh. the more I do learn the more I come to believe that uid/gid based permission handling is fundamentally flawed (but also good enough inside of a single application OCI(docker) image). The facts that Linux had to add a (very limited) capability system or that enterprise permission handling often goes through stuff like pollkit adding additional handling then just "gid/uid match" is I think very telling.
And what was allowed by the social norms of the tiny 1980's *nix computing world, or what you can get away with when you're as famous as Mr. Stallman...those may not translate well to other contexts.
It's easy to get someone in trouble in either case. If the user foobar starts doing crazy stuff as root, then foobar is in trouble. If you get the root password and start doing crazy stuff, then your username is associated with the troublemaking. (Assume that the logs go to some machine where you don't have access to remove them.) RMS's mechanism shifts the responsibility for your actions onto you, so that someone who knows the root password is more likely to leak it to you.
The best of both worlds is to get the root password, then find a hapless coworker who left their screen unlocked while out to lunch. su with the root password there, cause your chaos, everyone blames lunch guy. (Do look for outside systems; people and cameras can see you using someone's computer. Was always funny to me how many people have tried something like this, only to be nailed by the security cameras.)
You either have the password, or you don't have it. But not something in-between.
Outside of any ideology, in a scenario where you use su to become root, it's a very odd choice to link the wheel group to su; because if you know the password to the "root" user, and you have physical or remote access to the computer, you can likely just login as root.
And if you can't, then it means you actually needed sudo su, not su.
Those who actually need to be root, usually use sudo instead of su.
In the other cases, if you just need to switch user, then no point at all to refer to wheel
and let sudo verify that the user belongs to the group of allowed sudoers.
No need for the password to the root account.
If you do have sudo you can be very restrictive on who can run what.
Have you seen the number of flags every command has? ls has almost the entire alphabet taken.
“9.5 Case Sensitivity There are 66 useful settings that affect case sensitivity, plus 10 settings that are nearly useless, with the remaining 116 settings being either redundant or useless.”
https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/Case-Sensitivit...
I think the convention itself is good - I prefer long flags in scripts for improved readability.
But GNU is anything but minimal - compare GNU ls[1] with BSD ls[2], and try to recall the last time you needed --dereference-command-line-symlink-to-dir.
[1]: https://linux.die.net/man/1/ls
[2]: http://man.openbsd.org/lsSo true. I remember early on when GNU was started, people in the project where saying and developing with something like this in mind (paraphrasing):
"Make sure it works and meet users needs, even if it is too heavy for current systems, the hardware will improve as time goes on"
And that came true, for example, emacs is a lite ballerina compared to current IDEs.
I guess there are ways that sudo/doas could be adapted to implement passwd, chfn, chsh and friends, but the approach appears to have been chosen in the '70, and codified by POSIX.
How do you think these should be implemented?
It was a different world, and having some basic speed bumps like not allowing random user accounts to su to root was useful at the time.
I used to think Stallman was an ideologue from a different era. When I started dealing with software projects measured in years and millions his thoughts made much more sense to me. When you're selling me a system that has a 6/7 digit implementation cost for it to stand any chance of meeting my goals, withholding the source code only serves to annoy me.
I do understand the point of view when I think back. Today, Unix-like systems are everywhere. Learning it and working with it is a given. Back then, having access to a unix system was not a given. It was very expensive for hardware and software. The idea that one would be so close to the system and could be denied enough access by an overzealous BOFH was too much to take.
It just goes to show that circumstances change, and things can get weird if we don't change with it.
One of my favorites is extensive rights management. Especially on CMS. More ofte than not it is part of a buying decision, but used Stallman style soon after.
The observation that these credentials leak is correct. Or that you grow permissions over time for no other reason than doing work. A wheel group would today quickly attract users, too.
So let it be. The latest iteration is "basically let everyone, but audited and short term only". I find that very close to Stallmanns idea, for very different reasons.
And that person may not be the systemadmin
Similarly, Stallman's coding expertise can sometimes overshadow any potential shortcomings in the non-IT subjects.
Should this be that far-fetched though? That employees might not be simple thralls of the capitalist, whose agency extends only as far as his master permits?
It reminds me of something I'd read that one of the reasons modern capitalism is so borked is because the founding fathers weren't conceiving of things like "Amazon" existing, where one entity employs a staggeringly large number of employees. Or that a small number of companies would employ such a large percentage of workers.
Their worldview was that where most people were "self-employed" - and if they weren't, employers were small and had a few or tens of employees at most. Or it was a matter of master and apprentices where both groups were investing heavily in each other in a trade and in the running of a shop.
So, while yes our current system finds it a matter of course that employees are utterly subject to the whims of their employer and the legal and economic system fully supports them in this, does it have to be that way?
(I know you can go be a contractor, but good luck with health insurance and etc etc etc all the other things that being yoked to an employer brings that I wish were just public taxpayer-funded services).
I'm not quite sure I buy that argument. They lived in the time of the East India Company, which owned something like 50% of the world's trade at the time and ruled several nations.
`su` predates `sudo` by a decade doesn’t offer the fine-grained control `sudo` has. With `su` if you have the root password, you can do anything you want as root. With `sudo` admins can configure what commands users are allowed to run as root and could specifically block `sudo bash` from running.
Wheel users can do anything because (default) sudoers contains
%wheel ALL=(ALL) ALL
I strongly prefer doas wherever it's available.
My point that it does not have to be coded in the kernel as a dedicated mechanism to circumvent protection. Use any IPC channel to send a message to another process which already runs under root and accept those messages.
There was no room in that environment for a running service to elevate privilege, so it was implemented as a kernel system call.
This poverty meant that efficiency was required, and setuid was the most efficient mechanism.
It was a reasonable and efficient mechanism for its time, and it has successfully scaled to the realm of modern supercomputers, and remains efficient on the lowliest of embedded systems.
Maybe there was a more secure option bearing in mind of all the places that UNIX was forced to go, but I cannot think of one.
Lots of attempts to standardize procedures, etc., etc. - but East India agents far from home often had enormous latitude, and there were plenty of disasters and atrocities. (Not that either the British Government proper, or other European powers, were notably better. But they could certainly be worse - just look at the Spanish Conquest of the Americas, or the Belgian Congo.)
> ...the fastest way to transport messages was to have fresh horses pre-positioned ...
When there were enough short-but-important messages to be passed along a given route, they did have a far-faster-than-a-horse technology available - https://en.wikipedia.org/wiki/Optical_telegraph#India
Sudoing into root from lowuser account is in some scenarios potentially more dangerous than just using both accounts separately, as the user who uses root regularly/very often gets accustomed to the fact his commands are powerful and can screw his system, so mistakes almost never happen. While sudoing all the time creates a false sense of security and the user is more likely to run harmful command with sudo.
> To prevent remote root exploits, the Amazon Linux AMI does not allow remote root login via SSH[...] By default, the only account that can log in remotely via SSH is ec2-user. The ec2-user has sudo privileges.
Can someone please explain how this makes any sense for better security. It seems to be just a security theater.
Join us now and share the software,
You'lll be free, hackers, you'll be freee
Also by his uncanny tendency to be proven correct in matters concerning software freedom. His coding expertise are tertiary at most. Honestly, I never see people praise RMS's coding expertise, where are you even getting that idea from? I don't think you get it why people like RMS.
He is good at one domain, and then by cognitive bias people think he is right on everything.
It’s not true at all, and I think you have to take a bit of distance with glorifying IT personalities.
Like Bill Gates, Elon Musk, Stallman, and many others (especially in the VC world) it’s important to take them with a grain of salt, and not accept them as perfect nice guys because they have money (Musk) or influence (Stallman).
Otherwise they can spread dangerous ideas that normally should 100% be challenged, but that are not, due to blind acceptance.
You really don't get it at all. You're out of touch. People think that Stallman is right about one thing only, software freedom, and think he's out of touch with virtually everything else.
rms famously refused to secure his account @gnu.ai.mit.edu, and so the machine basically became an open shell server for every hacker in the world, ca. 1992.
Lots of hijinks ensued, and it was usually not possible to do anything useful in that account, since it was usually broken or pwned in remarkable ways. But they were wild, fun times.
[1] https://en.wikipedia.org/wiki/John_Gilmore_(activist)#Activi...
I used to do dumb stuff by telnetting to mail servers, back in the day. I think I really annoyed a few coworkers with that; one said he almost called CERT.
I think it was one of the earliest Cybersecurity realizations I had about how insecure the Internet really was, based as it was on blind trust among hosts that were supposed to have legitimate admins in control. I found that it was so easy to telnet to a mail server and feed it whatever you wanted, this must be the tip of the iceberg. And it was!
It's good, for a second I thought you were supporting his views on non-IT topics.
The problem is that the political speeches are part of the person, and the scope way way beyond software.
They are really interleaved with (supposed) IT topics, like if IT was a bait.
Once I went to one of his conference, and I had "learnt" more about "sex" and "facism" than software engineering or freedom.
How big of a change is it? If you are, say, a Linux terminal native can you just pick up and run?
There's some learning curve for features which exist for valid reasons, especially around communicating between domains. For instance, copy-and-paste between qubes requires extra steps. Plugging in a USB keyboard or mouse doesn't just work - you have to authorize it first (just click the OK button using a PS/2 mouse, or laptop's touchpad). You have to learn how to move files between qubes. USB drives, cameras, and microphones aren't globally available to all applications - you have to attach them to a qube first. You can install software using apt-get inside a qube, but it won't persist across reboots - you have to update the OS template.
I want those extra steps and complications - they are features, not a bugs! The first few days you'll be looking things up in the FAQ. After that it's pretty easy.
There are a few sore points that don't go away. You don't get GPU acceleration in your web browser, so rendering is slower. Gaming is not an option. Your application qubes live behind a firewall qube, so things that require network broadcast like Chromecast won't work. Those are fine for me but not for everyone.
> You can install software using apt-get inside a qube, but it won't persist across reboots - you have to update the OS template.
> I want those extra steps and complications
Is it wrong of me to say that enabling persistence, with snapshots, on a qube should be a single toggle?
Doing so has proved hard and slow so far, and Spectrum hasn't had a usable release for the masses yet.
As a qubes user, I think this is interesting but it definitely does not sound more usable.
Of course, I’m not being targeted by the state so my threat model is much lower.
The idea your data on a PC connected to Internet can be really secured from the most powerful actors is very naive.
Am I wrong about that?
Upd: yes, also there is no GPU acceleration in the VMs.
Yes, if you only have a USB keyboard, it will work. Manual creation of a USB VM then is recommended for security: https://www.qubes-os.org/doc/usb-qubes/
> Is it wrong of me to say that enabling persistence, with snapshots, on a qube should be a single toggle?
Of course you are right. TemplateVMs provide /root partition to AppVMs and software should be installed normally to the former. At every AppVM reboot, their /root is reset to the one from TemplateVM. Ordinary, persistent VMs are also possible. Details: https://www.qubes-os.org/doc/getting-started/
> Never lead to escapes
Escape is the highest form of security failure. I'm talking about data access and exfiltration.
Do you store all your important data on a VM with no internet access? Even Qubes users don't, it's hard to work with. Then it's Firefox/ kernel bug away from being accessed remotely.
XSAs are publicly known vulnerabilities discovered by someone who wanted to make it public and later were published by the Xen developers. There very probably are publicly unknown vulnerabilities, both in HW and Xen, discovered/created by people who want to profit from exploiting them. There are whole teams focused on this kind of work, payed by states and criminal-enablers like NSO.
> What is restrictive about Qubes?
No GPU acceleration for video in a VM, legacy OS on dom0. Xen development in support of modern CPUs has fallen behind, didn't even boot on modern Zen X570 platform last time I tried, dysfunctional nested virtualization, using KVM from Linux does not work, can't run Android Studio with phone emulator.
Did you just call Snowden a "twitter persona"? You're not serious. Not sure if I should reply further after that.
> Do you store all your important data on a VM with no internet access? Even Qubes users don't
Yes, I do. And I'm a Qubes user. There are many more users like me on their forums. This is much more convenient than you think: you can easily and securely copy/paste passwords wherever needed.
What about that audit?
If you keep all your important data off internet, that's good for you, and Qubes helps.
The problem is not isolation or lack of it. The problem is that app require complex set of permissions for both users files and other apps.
App might want to send notification to notification daemon. But app should not be able to pretend to be another app, whether by name or icon. And good luck trying to stop malicious app from just making same/similar enough icon and spelling Firefox with some fancy UTF characters to go around it.
And that's pretty simple case! And already very hard on kernel/OS level to solve. Now look at files.
You might want to allow graphical editor to open any graphical file, regardless of location.
You might want to allow that same editor to only edit some of them.
But for browser, you might want to allow saving new files, but not editing/rewriting existing ones, because it is not an editor, and should have no business editing the files.
Or, allow browser tab browsing certain URL (say, web image editor) to modify the files, but not the image sharing webpage that only needs to read the file.
Now we not only have insanely granular permissions per app, the different actions from "app" (web browser is basically container for multiple applications at that point) also need different permissions.
It has nothing to do with "unix bad", or "unix wrong", to actually separate the applications without hardships on the user (like fucking with permissions every time one app needs to touch files of another app) is just very very hard
More likely, you want to temporarily give them permission to specific files you indicate. A graphical editor doesn’t have reason to read any file that the user didn’t explicitly picked for editing/viewing.
That’s how Mac OS works nowadays (possibly except for the ‘temporarily’; I don’t know the details): applications can only open files that the user selected in the system file open dialog. That runs in a separate process and opens up an app’s sandbox to allow access to the file the user selected.
> The FileChooser portal allows sandboxed applications to ask the user for access to files outside the sandbox. The portal backend will present the user with a file chooser dialog.
> The selected files will be made accessible to the application via the document portal, and the returned URI will point into the document portal fuse filesystem in /run/user/$UID/doc/.
[1]: https://docs.flatpak.org/en/latest/portal-api-reference.html...
In practice I don't have the time to debug every shitty little app armor integration for weeks. I lost days to libvirt-manager because its app armor support was enforced and not even half assed. Some configuration paths would automatically get whitelisted in its auto generated app armor profiles, others would just get you a file not found until you whitelisted them manually. The process responsible for generating these profiles would also silently kill itself if it encountered a path that was on its internal ban list, have fun debugging that when you do things like using an alternative bios rom, which by default are all stored in a blocked path.
Apparmor feels like security through obscurity, unless you already know that you are dealing with app armor fuckery there is no chance in hell that you will be able to run your application and not being able to run anything is the holy grail of security.
I used libvirt with apparmor and was pretty satisfied with it
because programs often to need to have part of the capabilities of the user which started them, just a very well controlled subset of them, something which the UNIX model can't properly represent (through you can hack it on-top of it)
There is also the problem of having by default "owner-user owner-group other" as permission sets for files and executable. This works if others is "other humans" (assuming it does work, security issues on shared systems based on that where not uncommon). But this works much less if you want to protect users from rogue programs because then "other" tends to be far to permissive.
Picture: local DB client, remote DB server. Server can stream a file to the client for the client to write to disk. “On the same machine, as a different user” is just the trivial case of “over the network.”
Next problem will be sharing data between programs that legitimately need to do so; if I had an _emacs user that owned my source code, how do I make it non-painful for the _gcc user to read the source and write the resulting executables (which would end up in a directory owned by _emacs)? What about git, various preprocessors/generators, formatters, linters?
You'd have to step out of the traditional UNIX authn/authz model to effectively implement that. It's what various security-focussed OS's have been doing for a while anyway; e.g. OpenBSD implements unveil, which "hides" entire branches of the VFS tree. For example, if git has no business reading or writing files outside of the currently operated on repository, it can restrict itself very early in the process life - before proceeding to perform any of the "tricky" operations that are the common sources of security bugs.
I think one problem with the UNIX design is that UIDs/GIDs are a flat namespace, and commonly only 32-bits in size (even on 64-bit systems), when what is really needed to meet contemporary requirements is a hierarchy, either with an unlimited number of levels, or at least generous limits. Allow a user to create sub-uids (such as one per an application) and even sub-sub-uids (a web browser might create a sub-sub-uid for each website the user visits).
I think the Windows design of variable-length SIDs is in principle superior to the POSIX approach.
(Although, not necessarily in practice - it isn’t uncommon for Windows to make design decisions which in theory are superior to those of UNIX, but the practical implementation of them is full of warts, backward compatibility hacks, arbitrary limitations, and undocumented black boxes, which end up canceling out a lot of the theoretical advantage.)
From what I understand, Linux user namespaces require you to reserve a UID range for each namespace to be mapped to its parent. Since you only have 32-bits to play with, you are forced to map multiple UIDs in the child namespace to the same UID in the parent, while many security decisions are based on the root user namespace UID only. So this is actually a lot more limiting and inflexible than Windows-style variable length UIDs would be.
The problem with this approach is that, of course, it really doesn’t scale easily. Eventually you need multi tenant, and eventually we ended just pushing everything into the database, using row level security and tenant IDs. It worked great but felt more fragile (eg, you can disable RLS)
I’m not an OS expert by any means, but I think ultimately the problem is that we’re using one operating system model for two orthogonal use cases.
I feel like need a well-defined client model - “one user with multiple apps” and a well-defined server model - “one app with multiple users”. But it’s not clear to me how the OS can help with the latter, since it’s going to be domain specific. Maybe Postgres’ model is the right answer after all.
I tried to use Apparmor and SELinux, but how policies work is beyond me. Snap's sandboxing seems to be the closest thing to user-friendly sandboxing, but it's still not that user-friendly.
I mean, the correct approach would be to have groups even for specific network protocols because capabilities are not enough to sandbox a binary correctly, and the network group is pretty much pointless.
And then there's icmp, which brings us to the ping binary which on lazy distributions still has an SUID flag set, as well as glibc which still allows LD_PRELOAD by default because it is intended functionality from the perspective of its developers.
Most of these privilege escalation exploits can be mitigated, if users and groups and capabilities are managed correctly.
In practice I probably would recommend to use the systemd seccomp sandboxes because most of these quirks have been abstracted away there and are configurable in the service files - like file/folder access, user/group randomization, chrooting, capabilities etc.
(Also, implied is that any untrusted logic lives in the spawned program, while the "client" program is simple and auditable. As I said: like a database client vs a database server. Or how about: like a client that wants to print something, vs. a print server embedding untrusted printer drivers!)
But users can be in multiple groups. You can have files with groups like "graphics, audio" etc. and give access to the application users by adding that user to the relevant groups.
>IPC other than file based
This isn't UNIX model though, is it?
Though I agree with you. Given the current state of programs, file permissions aren't enough for isolation.
Is that really a limit or just a thing for convenience?
I don't think besides 0 in namespace being the actual user in the actual system as a good convenience, that there is any "need" for pids per root-pid, and even if that happened it would save "root-pids".
And I find it unlikely as of now that a system would reach the 16-bit limits of running more that 65000 applications on a single system without hitting some other limit like /proc/sys/kernel/pid_max or /proc/sys/fs/file-max first.
What happens with filesystems though? I would assume the filesystem is using the root user namespace. Which means if you have two different UIDs and they map to the same UID in the root namespace, they get collapsed into one for file ownership/etc. That seems a rather major limitation.
> And I find it unlikely as of now that a system would reach the 16-bit limits of running more that 65000 applications on a single system
With 32-bit identifiers, if you make each level 16-bit, you only have room for two levels. What if you have need for a third?
Also, you have to design a mapping from however many levels you need to the 32-bit flat namespace. A mapping which works well for one use case might turn out to be a problematic limitation in another. With variable-length UIDs there is no mapping to bother with.
Yes, this thing is gonna make IPv4 NAT look like a nice thing in comparison.
Yes, it will probably mean horrible kludge mapping of isolated-applications to UIDs, but until you get to 2^15 ~ 2^16 count of isolated-applications it should work fine.
Yes, this will be on a per-system basis, the resulting filesystem will be only useable by your system, and no other system.
What I'm saying is that in theory the "filesystem" and the "UIDs are 32 bit" parts are mostly there. They're there from the multi-user-big-box days not being used (except by Android/Linux).
> With 32-bit identifiers, if you make each level 16-bit, you only have room for two levels. What if you have need for a third?
The main reason why 65536 UIDs and GIDs are often submapped to every user is because POSIX systems often have a hardcoded assumption that user nobody is UID 65534, GID 65534, and if you want to run nested POSIX systems under POSIX systems without too many changes, reserving that many UIDs and GIDs are required.
There's no good place for that universal "nobody" user anyways, and if you're rethinking how the UIDs and GID mechanisms relate to security, definitively no place for a universal nobody, so you might as well map only the required ammount of UIDs/GIDs per isolated-application.
That then leads to leaving unmapped UIDs unmapped on both host and isolated-applications.
Unless you're reaching the 2^15 ~ 2^16 count of isolated-applications then it should work fine.
Another option would be doing what Android (and supposedly flatpak) does: you should not be able to simply run whatever you want if you're a isolated-application. If as a isolated-application you need to run another isolated-application you need to invoke "the platform" via `am` or `flatpak-spawn` and use it to spawn another isolated-application.
> What happens with filesystems though? I would assume the filesystem is using the root user namespace. Which means if you have two different UIDs and they map to the same UID in the root namespace, they get collapsed into one for file ownership/etc. That seems a rather major limitation.
As far as I know most of what can be considered normal Linux filesystems (ext4, btrfs and I think xfs) support said 32-bit UIDs so you would not need to change filesystem code (and I believe changing and bugfixing filesystem code is always a scary proposition) to use a 32-bit mapping.
Nothing prevents you from using only UID/GIDs; there are other security mechanisms that could be used:
* present every isolated-applications with different overlay filesystems visible. So you can have several things read/write to the same places, but every one has it's own view of what is being read/written.
* displaying a entirely different file system for every isolated-application (bindfs as an example)
* every isolated-application has it's SELinux context (labels) or other forms of ACL applied to those files.
But I find this birthday attack scenario dubious, why would you "need" this UID overlap if the isolated-application don't overlap outside of both namespaces?
If they aren't the same isolated-application it's the wrong thing to do and a security risk.
If they do map to the same isolated-application with the same set of data then trusting everything will be fine is a reasonable assumption. It isn't getting any more data or more privileges from being at different UIDs or GIDs in different contexts.
https://linux.die.net/man/7/unix
SCM_RIGHTS
Send or receive a set of open file descriptors from another process. The data
portion contains an integer array of the file descriptors. The passed file
descriptors behave as though they have been created with dup(2).
Other nifty thing with UNIX sockets is that you can just... read which user sent the message and as it is kernel adding that metadata you're 100% sure it came from that user. That's for example how you can set postgresql so say a certain user in the system can log as themselves without having to have a password.
In addition to sending data, processes may send file descriptors across a Unix domain socket connection using the sendmsg() and recvmsg() system calls. This allows the sending processes to grant the receiving process access to a file descriptor for which the receiving process otherwise does not have access.[2][3] This can be used to implement a rudimentary form of capability-based security.