this post was submitted on 05 Oct 2023
586 points (99.0% liked)

Linux

47344 readers
1297 users here now

From Wikipedia, the free encyclopedia

Linux is a family of open source Unix-like operating systems based on the Linux kernel, an operating system kernel first released on September 17, 1991 by Linus Torvalds. Linux is typically packaged in a Linux distribution (or distro for short).

Distributions include the Linux kernel and supporting system software and libraries, many of which are provided by the GNU Project. Many Linux distributions use the word "Linux" in their name, but the Free Software Foundation uses the name GNU/Linux to emphasize the importance of GNU software, causing some controversy.

Rules

Related Communities

Community icon by Alpár-Etele Méder, licensed under CC BY 3.0

founded 5 years ago
MODERATORS
 
all 47 comments
sorted by: hot top controversial new old
[–] [email protected] 45 points 11 months ago (1 children)

I have no idea what is going on but this looks good. I agree with you guys. Upvoted.

[–] [email protected] 11 points 11 months ago

It's for more granular access permissions for files and folders.

[–] [email protected] 29 points 11 months ago (3 children)

Cool. Does ACL support also depend on the filesystem?

[–] [email protected] 31 points 11 months ago

Yeah, but I think all reasonably-modern Unixy filesystems on Linux will support ACLs. ext2/3/4, btrfs, xfs, zfs, jfs, etc.

[–] [email protected] 12 points 11 months ago (2 children)

Yes. Some filesystems straight up do not support ACL of any kind (eg: fat32)

[–] [email protected] 9 points 11 months ago (1 children)

Fat32 doesn't support regular file permissions either, right? I was under the impression that it was permissionless.

[–] [email protected] 5 points 11 months ago (1 children)

You are entirely correct, it has no permission system to speak of

[–] davidgro 4 points 11 months ago (1 children)

I'll speak of it anyway: There's a "Read-only" bit on every file/directory and The User (there's only one!) can change it for any of them at any time.

[–] [email protected] 24 points 11 months ago (1 children)

Technically, this is also possible by creating extra groups, but this kind of access control presumably exists because the old-school method can be a pain to administer. Choosing group names can also be an "interesting" secondary challenge.

i.e. Dude's not going to be best pleased if they ls -l and see the group on the file is xyzgroup-but-not-dude even if it is with good reason. (Shouldn't have deleted the database, dude.)

[–] [email protected] 15 points 11 months ago

I don't really think that that's a realistic goal for ACLs. I mean, getfacl showing the user specifically being excluded probably isn't any more-polite.

[–] cybersandwich 21 points 11 months ago

I understand it all less after reading that. I need more coffee.

[–] [email protected] 19 points 11 months ago (1 children)

Does anybody use ACL without regretting it?

[–] InternetCitizen2 5 points 11 months ago

Whoaa, are you ACLing sober? Slow down champ!!

[–] doktorseven 17 points 11 months ago (1 children)

Confusing. You set a mask for a user and somehow it propagated down to a group and then you change permission on that group suddenly it applies to the user? Either something is wrong here or ACL permissions make absolutely zero sense. It is 5 billion times easier to use normal permissions to set these things up.

[–] Deiskos 12 points 11 months ago* (last edited 11 months ago) (1 children)

I think it's like this: what used to be group in regular permissions output is a union of group and ACL mask in ACL output. Mask sets the upper limit of what ACL can do, so if mask is rw- then it's impossible to set a r-x ACL permission because allowing execution is not allowed.

This seems to be more geared towards enterprise environment where it could be complicated to cleanly define groups, since you can only give access to one you might run into a problem where dept. A needs access to that directory but also person G from dept. B and person K from dept. C.

[–] [email protected] 1 points 11 months ago

Yep, it's basically a way to define new groups per directory. But these groups are hidden from the normal group commands!

[–] [email protected] 17 points 11 months ago* (last edited 11 months ago)

Cool, I didn't know ACLs were a widely available thing but the infographic explains pretty well! Sounds really useful when granular controls are needed, but I could also imagine it being a huge pain in environments already built out and scripted around regular permissions. Still as always, options are good and an ounce of planning is worth a pound of troubleshooting.

I do low-key hate seeing a directory named "dir" and a group named "me" though. That's chaotic neutral shit at the very least.

[–] [email protected] 13 points 11 months ago (1 children)

ACL Access-Control Lists

Access Control List Access Control Lists?

[–] [email protected] 18 points 11 months ago (1 children)

I mean in this case it's to show what ACL means, not like RIP in peace

[–] Akareth 1 points 11 months ago

That's what parentheses are for.

[–] [email protected] 10 points 11 months ago (2 children)

That's all fine and dandy, but why does it say rxw on the left?

[–] Zangoose 11 points 11 months ago (2 children)

Permissions are listed as "user", "group", "other". I.e. the user who made the file, the group of the user who made the file (usually just their name as a group), and everyone else. In this case the rxw is for the user.

For chmod, you can also represent these as binary numbers: 111 would mean having all 3, 101 would mean having read and write, etc. These binary numbers then get turned back into regular numbers (7 in the first example, since it's 111) for chmod. Giving a file "chmod 777" means the user, group, and other all have full permissions on the file. "chmod 700" gives the creator full control, but no one else can view, modify, or execute the file.

[–] redbeardgecko 15 points 11 months ago (1 children)

It's supposed to be rwx, not rxw.

[–] Zangoose 5 points 11 months ago

Oh I completely missed that lol. Oh well, it's probably still a useful explanation for someone else reading this

[–] [email protected] 6 points 11 months ago (1 children)

I feel bad for you because you were just trying to help.

[–] Zangoose 3 points 11 months ago

🤷‍♂️ They're just internet points, lemmy doesn't notify about up/downvotes so I will only see it if people respond. Either way it's hopefully still useful to someone else looking at the post who isn't familiar with basic permissions or acl

[–] mvirts 1 points 11 months ago (1 children)
[–] [email protected] 1 points 11 months ago (1 children)
[–] mvirts 3 points 11 months ago

The guide has w and x switched on the left side

[–] orl0pl 7 points 11 months ago (1 children)
[–] objectionist 2 points 11 months ago

it’s a really awesome website

[–] [email protected] 4 points 11 months ago (1 children)

The only thing you need to know about file acls is not to use them. Similar thing can be said for Network ACLs to be honest.

[–] c0mbatbag3l 4 points 11 months ago (1 children)

I've been a network engineer for five years at three companies and not a one has used switch or router based ACL's. It's all in the FW appliance.

[–] [email protected] 0 points 11 months ago (1 children)

Network ACLs are my bane. Someone long ago decided we needed to "isolate" the network, so they put ACLs everywhere and so now 50% of my teams time is spend fucking with ACLs :/ It's awful.

[–] c0mbatbag3l 3 points 11 months ago (1 children)

Yeah don't get me wrong it's an excellent part of network security but if it's not defined primarily on one device it's a hassle.

[–] [email protected] -5 points 11 months ago (1 children)

Only if you assume IP Addresses act as authentication for what that host is. But since they don't, I see ACLs as a security blanket.
I can change the IP of a server I control and bypass any ACL easily. If I have control of my network as well, then no ACL you apply can stop any of my servers from hitting whatever server you have allowed any of my servers to hit. So why not just allow my entire network block?

[–] c0mbatbag3l 3 points 11 months ago (1 children)

I don't assume that, and that's why I only consider IP based ACL's as a "part of this balanced security solution" because while handy, modern attacks are smarter everyday and heuristics based NIP systems are essential.

In the military we called it the "swiss cheese model", in ORM you use as many layers of security as you can to prevent a mishap. Controlling what subnets can access certain others keeps Becky from accounts payable from getting access into accounts receivable's data and writing her own checks. Sure, a network admin/sysadmin could just change their IP, but Becky doesn't have that access. I usually define network access by the subnet, if we aren't comfortable with all devices in a LAN having access then it's a pretty locked down solution, in which case we most likely have higher level requirements like application/port number or port security .1X.

I'm assuming your servers all reside in the same subnet? If not, changing the IP without changing the VLAN and/or trunking it to the access layer switch you're attached to would only result in a loss of connection.

For your use case I'd just allow the whole LAN and define applications we are ok with having communications between the two subnets, and as always a well thought out DMZ goes a really long way.

[–] [email protected] -1 points 11 months ago (1 children)

Right but if you want to start doing application level blocking, then the proper tool for the job is a stateful firewall and even better, a RADIUS/Kerberos system that authenticates every connection between servers.

Basically I use ACLs to prevent spoofing attacks from originating out of my network, and also to lock down the management plane of my network devices to specific subnets. In all other cases a stateful firewall should be used exclusively.

In any other case ACLs provide the illusion of security and create a huge amount of operational friction especially in a dynamic environment.

[–] dylanTheDeveloper 1 points 11 months ago