Slashdot Mirror
Some Linux Distros Found Vulnerable By Default
TuringTest writes "Security Focus carries an article about a security compromise found on several major distros due to bad default settings in the Linux kernel. 'It's a sad day when an ancient fork bomb attack can still take down most of the latest Linux distributions', says the writer. The attack was performed by spawning lots of processes from a normal user shell. Is interesting to note that Debian was not among the distros that fell to the attack. The writer also praises the OpenBSD policy of Secure by Default."
Kittens are vulnerable to forks by default as well - you can easily get at the kernel if you just - oh, hang on, a different kind of fork, you say?
Thank god I use Windows, I'm safe!
Sorry but the ability for a non-privileged user to run as many programs as the like is a feature, not a bug. Inability to turn that feature off would be a bug, but given that few modern Linux boxes are actually used as multi-user remote-login accounts, it's a completely unecessary overhead.
And if you are administrating a true multi-user old-style-Unix type server, you should know enough to stop people fork bombing you (i.e. quotas).
Athletic Scholarships to universities make as much sense as academic scholarships to sports teams.
So what would a good limit to the number of processes spawned be?
I mean what can say what is good for everyone?
Saying that if you think the fork bomb is good grep bombs are more fun and particularly good for silincing the mass of Quake 3 players in an undergraduate lab:
'grep foo
Oh hang on did i just discover a new exploit
Fork bombs only work if you can log into the system in question. This is a bit lower priority than your usual vulnerabilities which allow outside attacks.
Sorry, but this article seems pretty retarded to me. Windows is insecure because people can use IE bugs to install scumware that takes over your entire machine... Linux is insecure because ordinary users who are legitimately logged into your machine can fork off as many processes as they want? Huh?
Sure, maybe if you're running a server that allows remote logins, you want to restrict how many processes a user can run. But as a single-user system, I want to be able to run as many processes as I choose, not be restricted by the distribution author's ideas of what's good for me.
I really wonder what kind of Debian installation he runs. Just a couple of weeks ago I had to reboot my Debian box after some experimenting with an obfuscated fork bomb. Won't work again now that I set some ulimits, but they're not there by default.
:(){ :&:;};:
In case anyone is interested, here's the obfuscated fork bomb:
A forkbomb is just a relatively simplistic way to mount a resource exhaustion attack. I would be extremely wary of anyone who claims that their UNIX class operating system is immune to resource exhaustion from a local user. There's just too many resources that can be commandeered, and to lock them all down would leave you with a system that's so restricted as to be nearly useless as a general computing platform.
/. if they're reporting this as news.
It must be a slow day on
No, I understand the article. I just couldn't resist the jab. The fact is that GNU/Linux ought to be the best it can be in and of itself. That some distributions are screwing that up and making very poor defaults is not to be forgiven. Not at all. Especially when it isn't difficult to do better.
How to use coral cache: http://slashdot.org.nyud.net:8090/~oscartheduck
On the 3 distros listed as vulnerable, the default settings would stop any remote person from having a chance of getting a shell open on the box to perform the fork attack in the first place.
If a person has enough access to the machine to be able to "forkbomb" it, then there's plenty of other nasty things you could do to it.
Unprivileged user can take down entire system by unplugging machine from power socket.
All my servers have multiple users. Those users are system accounts to run different software, and I do not want any of them to be able to cause a problem to the entire server. Reasonable limits should be in place by default, and those of us who actually need higher limits for certain users, can raise those limits.
Even on a single user desktop machine, its nice to have limits so shitty software can't take down my entire machine. With limits I can just log in on another terminal and kill the offending program, without limits you get to reboot, and lose any work you were doing.
You were running bash then :p
:) .login or .bash_rc files.
I recognise that one... which is always good
just don't leave your box unlocked and have some "funny" person drop it in your
Looks like everyone out there on slashdot think this is not really a problem. Remember when it was discovered that you could get into a xp installation locally with a win 2000 boot cd? Oh, the howling that was heard.
Here is a issue that can be done remotely with only a user account.
Humor from a Genetically Molested Mind
I came up with the idea of a ping/fork DoS attack (mostly as a joke)
:)
In pseudocode:
while (true) {
ping(target)
fork()
}
I seriously thought of posting this to a few script kiddie sites, so the kiddies could crash themselves long before the pinging does any damage
--buddy
It's funny, isn't it, that on the same day we have a story about Linux distros being insecure by default, EXCEPT Debian, we have another story where Debian is being criticized for not releasing updates more often.
Maybe, and here's a thought, just maybe, it's wise to take a decent, stable distro and perfect it, instead of taking a distro and submerging it in a state of perpetual flux with constant updates.
Just a thought. I might be biased because it's a Debian-based distro that finally put a working Linux on my laptop. But you know what? Every now and then the bias is there for a reason...
Most linux systems are used as desktops, if you use them as a server you don't use the defaults. Now a user being able to crash his own system is nothing new. It ain't nice but as long as it is the user doing it then no problem. Now if this fork could be used to make apache explode and bring down the system THAT would be a boo boo.
Ideally yes the system should not do things that bring it coming crashing down but this is close to blaming a car for allowing me to plow into a wall. Not sure if I want a car/computer telling me what I can and cannot do.
As to how to set the limits on the number of forks. Maybe I got this completly wrong but could it be that this depends entirely on your hardware? Perhaps the latest IBM mainframe can handle a few more then an ancient 386? How the hell is the distro supposed to know what I got?
Security is other people doing stuff on my computer that I don't want and or know about. Me screwing stuff up is my business.
BSD is very solid, this is known. It is also known that BSD has been along long before linux and but has been sucking it exhaust fumes ever since it arrived. For every story about how much more secure BSD is there are a dozen stories about linux actually making a mark on the world. So good. Your BSD survived a forkbomb. But why exactly was the author running a linux desktop then if BSD is so much better?
Another non-story on /. Is the internet going to the way of tv?
MMO Quests are like orgasms:
You may solo them, I prefer them in a group.
A good vm should do enough accoutning to allow you to log back in and kill those.
So, try this in FreeBSD, and be amazed, now try it in any 2.4 or 2.6 linux kernel, and be disgusted.
while(1) { malloc(1); }
That won't work on modern systems, or systems with a lot of virtual memory available (lots of RAM or large swap).
A modern OS will not actually commit memory until it is actually used, and while malloc() involves some bookkeeping, most of the bookkeeping is very little. It's quite likely you'll actually run out of process RAM (2GB or 3GB, depending on settings on a 32 bit machine) space first before the system starts to strain. On Linux, the recent kernels will kill processes that start hogging RAM when free memory falls below the low-water mark. And each malloc() really allocates 8/16/32 bytes of RAM for even a 1 byte allocation.
Unless you use genkernel, there is NO default kerenel configuration, verions or anything else. No serious admin uses genkerenel as anything other than a starting point - PERIOD.
Choose your kernel version, patch set, etc. No defaults. I guess he has never actually installed gentoo himself. The author should get a clue about the distro's he's talking about before making clames about their security.
Spell check? Why bother. That is what grammer/spelling Nazi freaks who waiste band width posting "spell right" are for.
while [1] do
fubar &
done
Then I did chmod +x fubar, and typed "fubar &"
oops.
The system load started climbing and everyone else on the machine started bitching. I thought it would crash, and went over to the local admin and fessed up. Of course, we were all interested in what would happen. Nobody could get in to kill it, and the processes were spawning so fast that we couldn't catch it. It was taking forever just to log int. But the load leveled off, and it wasn't going to crash. The admin was going to reboot it, but then I said "wait a second!" I went back to my window that was open. Know what I typed?
rm -f fubar
I suppose you could make it more nasty by making the file name create a copy of itself and name it the process id, so that you wouldn't be able to rename it.
cp $0 .$$
./.$$ &
This will leave all the process files laying around, but you could code something to remove them. But this gets the point across.
My beliefs do not require that you agree with them.
Right, it's a feature, but the question isn't whether it should ever be allowed, but what the default setting should be. I think the article made a pretty good case that default should be no.
And if you are administrating a true multi-user old-style-Unix type server, you should know enough to stop people fork bombing you (i.e. quotas).
First, I think a lot of unix people would be shocked to find that's on by default as the writer was. Second, that basically means that anyone who successfully hacks into a user account takes the machine down. That applies for your desktop machine, not just "old-style" unix type servers. Third, you mention the relative scarcity of old style servers these days - they're still more common than a user who needs to run an INFINITE number of programs. Even capping somewhere in the thousands would work, keeping anyone from being hampered in their work.
Basically, this is a case of idea vs. reality. You want the IDEA that you can run as many programs as you want, though you'll never need to. So in REALITY, a sane cap never hurts you. However, a lack of a cap provides very REAL security problems, either from a user or from someone who manages to hack a user account. Again, you really don't want EVERY userland exploit to lead to a kernel takedown, do you?
If you had bothered to read the thread the article points to, the forkbomb vulnerability wasn't in the kernel per se, but in the /etc/security/limits file, which on most distros has a bunch of example lines commented out by default.
The kernel can't/shouldn't implement limits that are commented out.
Edit the file(s) to your taste and reboot.
No kernel patching necessary.
On my Win2k box, running ":(){ :|:& };:" at a Cygwin bash prompt DOES kill the system. I don't know enough about Windows admin (and I don't care enough to learn) what would prevent a forkbomb.
If God had meant for man to see the sunrise, He would have scheduled it later in the day.
Just how many regular users expect to run 20000 processes at once? (or even 200?) When that happens it's almost always caused by a bug (or malicious activity). Right now, I have 50 user processes running. I'm a power user, but I'd probably never get blocked by a limit of 1000 unless I was doing something really wierd -- and something that weird should come with instructions on modifying the kernel settings.
Yes, it should always remain possible to set up your system so that you can run massive numbers of processes and/or threads, but the default should be to keep numbers to a dull roar in favour of system stability. People whose needs are such that they actually and legitimately want to fork massive numbers of processes are also the kinds of people who wouldn't have a hard time figuring out how to change the kernel settings to allow it.
As such, the default should err on the side of security, but allow a knowledgable user to do whatever the heck he wants.
Thing is, though, that local resource-exhaustion exploits are difficult to set. You want to allow a user felatively free reign -- even to the point of stressing the system, but still allow enough of a reserve so that an admin cam login and shut down a user who'se gone overboard. You also want to set a limit that will be reasonable 5 years down the road when processors are 10 times as fast (and/or 20-way SMP is standard issue)
Something to note here in Linux's favour: Even though the forkbomb brought the system to it's knees it stayed up. Although it might have taken 1/2hour to do, an admin may have been actually able to login and kill the offending process.
Free Software: Like love, it grows best when given away.
Please explain how the ftpd binary (not suid) can be used to exploit the system in ways that the user otherwise could not.
Taking away the ftpd binary wouldn't stop the user from doing exactly the same thing by some other means. For example, they could simply download the source and compile a new one by themselves. Or use Perl. Or compile a binary somewhere else and download that.
I rarely criticize things I don't care about.
We aren't saying that default limits will be perfect for everyone. We are saying that its better to have to raise your limits IF YOU NEED TO, then to have your machine vulnerable to being completely taken down trivially, very possibly by remote users with no accounts, just from making your services work harder than you expected.
If you are running a server than needs hundreds of apache processes running, then you know that and can raise it. Someone who is new to linux won't need that, and won't know how to setup limits for themselves. So you make the machine secure by default, and allowed advanced users with advanced needs to tweak things as they need.
The best thing I can think of to illustrate the point to you is your apache example. By default apache won't let you have more than 150 users connected. This is a sane default to protect from resource exhaustion. If you need more than that, you can set it yourself. People have some protection by default, but advanced users can customize the settings for their needs.
I cannot believe in 2005 I am arguing with someone who thinks secure by default is a bad idea because it might invonvenience you.
In the past I've been awakened in the dead of night by my IDS'es detecting a fork bomb and it's always been self inflicted - some dumbass^H^H^H^H^H^H^Hmisinformed programmer not understanding why it's important to check the return status of a fork() or set alarms to kill off hung sockets.
I found some earlier kernels ignored RLIMIT_CPU, RLIMIT_RSS, and RLIMIT_NPROC, and setting the CPU and RSS limits in Apache was ineffective. This was in the Red Hat 9 / 2.4.20 kernel days. I have not researched this in a year or so. If all this stuff works now, let me know so I can insert a "ulimit -Hu 10" in future startup scripts as a "courtesy" to inattentive programmers.
Give a man a fish and you have fed him for today. Teach a man to fish, and he'll say "WHERE'S MY FISH, YOU IDIOT?"
Many Linux users found to be insecure whenever the faults in their OS are pointed out. ;)
+5:offtopic,but anti-American
This level of fanboyness is unbelievable. Well actually I should not be surprised, blindness to linux's faults is endemic here on Slashdot.
The author is not "Running around screaming...", he is simply very surprised that a local user can exhaust a system so easily. Maybe every single admin should think of all n possible security problems every single time they take a box live, but people are human.
Which one is worse: limits in place by default so that an admin needs to know how to raise them when necessary and the forkbomb would not work, or no limits in place and having to know to set them or else the box can be brought to its knees? Secure by default or free for all.
I suppose you think every default install should also have telnetd enabled by default because any admin with half a brain should know how to turn that off? Point is admins are fallible, the default should be the lower total risk/cost option. I think which one that is is clear here.