· Content
· News
· Articles
· Mailinglists
· Knowledgebase
· Trouble Tickets
· Files
· Glossary
· Links
· Compatibility Lists
· Forums
Welcome to our website
To take full advantage of all features you need to login or register. Registration is completely free and takes only a few seconds.
Intel Hyper-threading Has Security Flaw?
Posted by: duke on: 05/13/2005 02:44 PM [ Print | 8 comment(s) ]
The Inquirer is reporting that Intel's Hyper-Threading may have a rather serious security flaw.
Colin Percival, a member of FreeBSD's security team, claimed the flaw allows local information disclosure, meaning that unprivileged users can steal RSA private keys being used on the same machine.You can read the entire post over here.
« Supermicro AMD Opteron box star of builder show · Intel Hyper-threading Has Security Flaw?
· Intel: next-gen dual-cores not NetBurst »
Comment
|
duraid SMP Qualified Posts: 387 Joined: 2002-03-31 |
#36327 Posted on: 05/14/2005 11:02 AM
I just read the paper, and had a quick chat with Colin himself. As will be clear if you read the paper (and not the inquirer's rubbish), this is not a problem specific to Intel hyperthreading. The situation occurs on any processor that allows multiple concurrent threads of execution that can "attack" a shared cache. Therefore, the same problem will be observed on: any multi-core processor that has a shared cache, regardless of whether or not it has any sort of hyperthreading. For example, IBM's POWER4 and POWER5 are vulnerable to the attack described, even if you disable SMT. Finally, the problem is easily (and in my opinion, most appropriately) fixed at the operating-system level, as you can again learn by reading Colin's paper (and, surprise surprise, not the inquirer, even though they seem to have updated their page.) In short: don't worry. There's nothing wrong with your hardware, and the software (your OS) is likely to be fixed very soon (if it hasn't already.) |
Comment
|
GNU_Raiz Registered User Posts: 140 Joined: 2003-02-26 |
#36328 Posted on: 05/14/2005 07:49 PM
Very informative, I noticed this when it first appeared on KernelTrap. This could be another reason why HyperTheading is not being pushed, as much as it used to be. Also I think its interesting that Netburst is not as big a player as it used to be. I just feel sorry for all those Xeon based servers who rely on HyperThreading as part of their job. You make it sound like its not a big deal, its a big deal if you have an application that actually uses, or needs HyperThreading! For many system admins, that was probably one of the reasons they bought into the Intel mantra, now it comes back and bites them in the backside. How many Xeon servers run the *BSD's, which it seems will be the first one to have this problem fixed? What about the other OS'es? I see it two ways, this could hurry up the death of HyperThreading since Intel is already moving away from Netburst, or it will push Intel to put heat on the affected OS'es nip this in the bud. Just wait for it, yea here it comes, when the mainsteam media gets a wiff of this look for Intel to go into damage control. Gnu_Raiz Use it, promote it, live it! Open Software! Open Source Libre Office! |
Comment
|
AssKoala Anti-Zealot @ GATech Posts: 3309 Joined: 2002-01-02 |
#36329 Posted on: 05/14/2005 08:34 PM
Ignoring GNU_Raiz's generally ignorant statements, I'm with duraid on this one. There's not a damned thing to worry about. Me Webpage | If you always think like an expert, you'll always be a beginner. | "A handful of knowledgeable people is more effective than an army of fools" -Writing Secure Code, 2nd Ed. |
Comment
|
Mr. Hahn Unregistered |
#36330 Posted on: 05/15/2005 10:33 AM
Never thought I'd be sorta happy that I don't have to patch something because that something was never supported properly on my os for me to want to enable it in the first place. |
Comment
|
audioaficionado SMP aficionado too Posts: 3536 Joined: 2003-08-23 |
#36331 Posted on: 05/15/2005 11:02 AM
[Alfred E. Newman voice] What? Me worry... Iwill DH800, 2GB (2x1024) OCZ4002048PFDC-K 2.5-3-3-8-1t Dual Prestionia Xeons 2.4 M0 @1.50v 220Mhz 3.3GHz Location: 42.3 N 122.9 W Medford, Oregon, USA |
Comment
|
Hydra Performance Junkie Posts: 174 Joined: 2001-03-09 |
#36332 Posted on: 05/17/2005 07:27 AM
This security flaw with Hyper-Threading seems to be only related to Intel processors with this feature: http://www.daemonology.net/hyperthreading-considered-harmful/
Overclocking is a blast! Do I smell something burning?  |
Comment
|
duraid SMP Qualified Posts: 387 Joined: 2002-03-31 |
#36333 Posted on: 05/17/2005 02:54 PM
Hydra, you're a retard mate. Not only have you just proven that you can't understand the very article you just linked to (if you could, you'd know that the hack isn't limited to hyperthreading or other SMT processors, much less Intel SMT processors)) but apparently you can't even read up this very ******* web page. Did you not see what I said? Could you not understand it? I am now almost certain that the POWER4 and POWER5 are both vulnerable to Colin's attack, as they provide the means for the spy process to _accurately_ determine the speed with which it is executing (which, ultimately, is how the information is transmitted to it.) (I had never needed to use POWER4 or POWER5 performance counters by hand before, but they are there and they work just like the RDTSC instruction on pentium/athlon/opteron/whatever.) For $1,000 I will go ahead and write the trojan/spy pair for POWER4 (a CPU which is NOT multithreaded, hydra) and if it turns out that they don't work, you keep your money.  Anyone else on this forum with a power4/power5 linux box who wants to give me an account can be the judge, and it would actually be quite a fun competition, because they wouldn't have to trust me. All they'd have to do is run the trojan program I give them and point it to some file on their hard disk. I could then tell them the contents of this file just by running the spy, even though they could verify using strace that the trojan didn't create any files, open any network sockets, or do anything at all like that.
So, do I have any takers? Hydra, are you feeling lucky?  |
Comment
|
Hydra Performance Junkie Posts: 174 Joined: 2001-03-09 |
#36334 Posted on: 05/18/2005 04:47 PM
duraid: You're just so rude that you really don't deserve a reply. If every Itanic fanboy is like you, I feel really sorry for the platform. People like you dosen't do it any good, and if you're representative it's no wonder why it's struggling so much to survive :rolleyes: Overclocking is a blast! Do I smell something burning? |