User-triggerable NULL pointer dereference

https://jira.mongodb.org/browse/PYTHON-532

by jibberzmcgee • 13 years ago

124 78 13 years ago

jira.mongodb.org

Doublon2• 13 years ago

I like the style

It perfectly captures the mindstate of someone who's been debugging for 16 hours straight only to find out it's someone else's fault.

EDIT: That's not to say it's professional. It is amusing to read though.

Doublon2• 13 years ago

Exactly. And we all know how it feels. WTF MIKE?

D_hemming• 13 years ago

I've written similar bug reports at 4am.

atesti• 13 years ago

http://www.fredberinger.com/images/codereview6.jpg

(This picture is funny but very accurate: If you can review code without swear words, your soul just isn't into it)

akanet• 13 years ago

(I'm sure there's going to be a plethora of commenters in the near future roundly criticizing the author for his tone)

I strongly agree with the author on the rampant disconnect between Mongo's actual quality and its marketed quality.

chris_wot3• 13 years ago

Allow me to be the first. The tone is appalling.

coldtea10• 13 years ago

And since when do hackers care about "the tone"?

adlpz2• 13 years ago

I will join the others saying as me: as someone who has been in the situation of debugging something for hours only to find out the cause is someone's else incompetence, I totally find his tone acceptable.

At the same time, MongoDB is a free, OSS solution. While it's true it's marketed way above reality, you both don't have to pay for it (beyond support) and can contribute to it if something doesn't work as you wish it did.

lucian1900• 13 years ago

Sadly, it is also terrible. When something is that bad and marketed so strongly, it being open source and free is less of a saving grace.

exDM693• 13 years ago

> I will join the others saying as me: as someone who has been in the situation of debugging something for hours only to find out the cause is someone's else incompetence, I totally find his tone acceptable.

This tone is not acceptable in a formal bug report. It's fine to pour out your frustration in this tone to your coworkers over a pint of beer, but it is not fine to go and yell it at someones face in a formal environment.

Everyone has wasted hours and hours in a frustrating debugging spree, we all know that feeling. Get over it and be a professional, report the bug, fix it and shut up.

adlpz2• 13 years ago

Well, you are right. I would have said I found his tone understandable, not really acceptable.

antirez• 13 years ago

I guess the tone may be disturbing, but I hope that the MongoDB team will do the right thing and thank the user for the detailed bug report, fix the issue, and don't focus too much on the noisy part of the message. After all he tried to use the system, found an issue after debugging for a long time, and contributed the finding back instead of sharing it only with his friends to crash random servers on the internet.

coldtea10• 13 years ago

>I guess the tone may be disturbing, but I hope that the MongoDB team will do the right thing and thank the user for the detailed bug report, fix the issue, and don't focus too much on the noisy part of the message.

Actually, just fixing the bugs after public outrage will no do.

They should focus A LOT on the noisy part of the message.

And feel shame. And then do something about it.

I'm glad for the noisy part myself.

exDM693• 13 years ago

> They should focus A LOT on the noisy part of the message. > And feel shame. And then do something about it.

No the MongoDB team should not feel ashamed, the author of this bug report should. This kind of aggressive writing is unprofessional, rude and childish.

People make mistakes, even good engineers do. They should not be yelled at like this even if they screw up badly. Writing software is a team effort, and the users of open source software should be a part of that team and take the collective responsibility of finding, reporting and fixing bugs in an effective and civilized manner.

The tone of this bug report does not help fixing this bug faster or better, but it does make the reporter look like an ass.

pfortuny• 13 years ago

A little history shows anybody can make any mistake.

You should read about the pseudorandom number generators used ubiquitously in the 1970s and 1980s, for example. Or the Patriot missile bug. The second one cost lives and really, it was kind of obvious.

So: no, impoliteness is useless, unhelpful and a waate of time. Humor is difficult (just pouring a list of swear words and admiration marks is not funny).

We all have bad days and Mike is not an exception.

Hence: thanks for the bug report, keep your (meaning the OP's) shit to yourself.

nknighthb3• 13 years ago

All the noisy part accomplished is to convince me MongoDB's critics are so far off the deep end they should be ignored. I'll be considering MongoDB for my next project.

latch• 13 years ago

The l33sp3@k as well as the [attempt] at humor at 10gen's expense is fine.

Anyone who thinks that publicly attacking an individual at this level is in any way acceptable should work hard at being more empathic. No long hours of working justify this.

antihero2• 13 years ago

"Anyone who thinks that publicly attacking an individual at this level is in any way acceptable should work hard at being more empathic"

YAWN. Or people could just get thicker skin so we can actually have fun? This is just banter.

drcongo• 13 years ago

"Banter" is a vile word for a disgusting act.

untothebreach3• 13 years ago

Ridiculing others is fun? I know that was true in high school, but I thought we were adults now.

SamReidHughes• 13 years ago

Everybody, quick! Let's rain down condemnation on somebody!

aktau• 13 years ago

I personally don't fault the author too much for his tone, because I understand his situation (been there, done that). Something like this is on the verge (and probably over) of being unacceptable. In inherited a project with MongoDB as well and even though I haven't run in to many problems except for MongoDB 2.0.x (debian wheezy) removing the journal file somewhere during a shutdown but only removing the lock file later, which can and does lead to a race condition and a database that refuses to start up unless removing the lockfile manually)... I sincerely hold my heart I don't run into things like this. Luckily Mongo is only used for some kind of persistent object storage for a PHP webapp. (the db is all of 10MB...).

MichaelAza6• 13 years ago

Before I even get to the tone of this bug report, I'd like to ask the guy who made it - are you aware you just put many many users in risk here? Besides making their services unavailable, this might lead to further exploits. There is a way to disclose this sort of stuff. This is not it.

Now, about the tone - be a human first, programmer second. Even though it seems 10gen screwed up big time, everyone makes mistakes. You could inform them of what seems to be a major issue privately and politely, not ridicule them in public. I wouldn't want it done to me and I'm sure neither would you.

As said by Chazal (old timey Jewish scholars): "Proper behavior precedes the Torah ... and whoever humiliates another, it is as though he killed him"

EDIT: I'd just like to add that among the companies affected by this disclosure and which are now totally open to attack are Craiglist, Firebase, MTV, SourceForge, Codecademy and others. Here's the full list - http://www.mongodb.org/about/production-deployments/

tokenizerrr• 13 years ago

He found a bug, and posted to the bug tracker. Would you have preferred he just shared this with his buddies to crash servers, or attempt to sell it?

MichaelAza6• 13 years ago

When you find a bug which amounts to a security vulnerability, you should inform them privately and then make it public after a reasonable amount of time has passed and preferably after it's been fixed. By making it public right from the get-go, anyone who sees this bug report can now attack any system that uses MongoDB. Do you think that's OK?

ctz• 13 years ago

Disagree. Users ultimately are put at risk by shipping broken code. Everything after that doesn't increase risk, but it might realise that risk.

bcoates• 13 years ago

The tone argument is a distraction, but thank you for pointing out the real problem here: this is an irresponsible way to disclose an issue that hurts people whose only mistake was using MongoDB in production.

It's not hard to find out how to report security issues to 10gen: http://docs.mongodb.org/manual/administration/vulnerability-... and the terms they ask for look reasonable to me.

brudgers2• 13 years ago

"be a human first"

Dehumanization creates more harm than ten thousand crashing web apps. Packaging the evil as the wisdom of religious scholarship changes nothing.

Refined words and moralizing demeanor but gilly suit vileness and violence.

MichaelAza6• 13 years ago

You know what, you're right. I shouldn't have used "be a human" there. Haven't thought of it that way.

Even so, I still stand by the message even if the wording was wrong - one should be a human first in the sense that one should be empathetic and strive to be good. And yes, I still stand by the quote from Chazal which, even as an atheist, I appreciate as a part of my cultural heritage. All it's saying is that being a good person is more important than anything (even studying the Torah) and that one should not humiliate others. I think those are important and valid messages.

PommeDeTerre2• 13 years ago

There's no place for political correctness when it comes to developing critical software systems.

Databases are such a system. They're expected to attain and maintain an extremely high level of quality and reliability. Those who build them should expect to do everything in their power to avoid flaws, and these people should be more than willing to accept harsh reprisals when issues are found.

When the security and safety of data is at risk, some developer's hurt feelings should be the last thing anyone is concerned about. There are far more pressing problems at hand in such a situation.

PommeDeTerre2• 13 years ago

People building software systems, especially ones that are public-facing or otherwise widely accessible, have a responsibility to not use components that are questionable.

In my opinion, MongoDB is a questionable component.

A few years ago, when it was really hyped beyond belief, I think that anyone with at least some database experience, and who could look at it objectively, was inherently suspicious of MongoDB.

There were, and clearly still are, just too many things that are disconcerting about it. These range from how it rejects much of the proven knowledge and experience gained over the past four decades, right down to the community who embraced it and the people who hyped it the most.

Thankfully, many of us saw this, and we stood against the use of MongoDB and other unsuitable NoSQL "databases" for the projects we work on, or within the organizations we may work for.

While it is unfortunate if others suffer as a result of this bug, I can't help but feel that they should have known better than to use MongoDB in the first place. The unease and skepticism should've been there. Given the numerous other proven and reliable DBs, including free and open source ones, there's really no reason to have used MongoDB, in my mind.

lttlrck• 13 years ago

Step 0. Assume free open source project with publicly accessible bug tracking system with non-zero open issues is bug-free.

alcuadrado• 13 years ago

They are profiting out of mongodb, they are getting people to invest lots of time, trust, and money in their product.

The author has a strong point that quality of the development process for mongodb must not be good if a bug like this one gets into production, when it could be avoided using automated tools. This won't be such a fault for a random open source project, but if you are making a business out of it, and you are encouraging others to base use it as a key part of their business it do is a great problem.

OldSchool• 13 years ago

Agreed. This is one of the unintended results of giving out free stuff in general; no good deed goes unpunished. There are enough bugs in commercial APIs and products that hitting one in something you got for free should be no surprise at all.

I always describe programming as incredibly easy, except for the part when things don't work as they're supposed to work. Very often you eventually find this is outside of your control. You have to be more calm and patient than the bug.

He's not going to win any friends or help acting like a pompous ass who's never made a mistake in his life. Maybe he hasn't because this is his first real project.

viraptor2• 13 years ago

Could someone explain what's actually so critical here? The way I understand it is that if you can inject custom json contents into the mongodb data, you can cause the pymongo library to crash. This should not be allowed to begin with though - why would you allow anyone to store custom object without any checks? I would compare it to a null reference on a badly formatted SQL query - sure, it's a bug, but why did you allow the user to submit unescaped string to begin with.

Or am I missing something important?

tetha• 13 years ago

As other comments pointed out, it's not that critical per se. It crashes, restarts and then that's it.

Beyond that, it's an undefined pointer dereference - who knows what this could be use for in certain combinations and systems. Use a "not so critical" bug in that subsystem, a "not so critical" mistake over there, another "somewhat severe" error over there and you got a root shell going. It's simply disconcerning and annoying if you consider that static checkers could have caught it.

viraptor2• 13 years ago

It's a NULL pointer dereference, not undefined pointer from what I can see. Unless someone was able to mmap that memory, it should simply cause an instant crash. PyDict_GetItemString() is guaranteed to return NULL for missing fields.

acc01• 13 years ago

Is this a new kind of viral adverts for Coverity?

kmasters2• 13 years ago

Before getting religiously high and mighty it might be instructive to run a security scan on the whole of the Internet and see how many SQL injection vulnerabilities exist on the websites of major US companies and small time startups.

The fact that MongoDB has in itself a vulnerability to unchecked input is not great. But consider that if you are dealing with client side browser or server side software, the entire stack is rife with security vulnerabilities because the components themselves right down to TCP/IP are inherently insecure.

Be careful out there, and write nicer bug reports. Use the process. If you were on the other end of that bug report, you would feel differently.

kyllo• 13 years ago

Doing that will almost certainly get you thrown behind bars.

kmasters2• 13 years ago

lol I didnt say to actually DO THAT. It was to make a point that we all adopt things that bite us, many things.

"Nobody expects the Inquisition"

User-triggerable NULL pointer dereference

Related Stories