Hacker News new | past | comments | ask | show | jobs | submit login
Fallout: Reading Kernel Writes from User Space (arxiv.org)
98 points by lainon 15 days ago | hide | past | web | favorite | 16 comments

'Remarkably, the patent explicitly states that: "if there is a hit at operation 302 [partial match using page offsets] and the physical address of the load or the store operations is not valid, the physical address check at operation 310 [full physical address match] may be considered as a hit"'

Wait, Intel thought the insecure optimization which lead to this vulnerability was such a great idea that they actually patented it? Wow.

The load and store buffers are an incredibly performance sensitive part of the processor. To be maximally conservative you'd fully resolve the physical address of the load and store (including the access check) before allowing the load to bypass a non-aliasing store, or before forwarding store data to a load. But you could be waiting awhile (for example, many processors have multiple levels of TLBs), during which time you've held up a load for no good reason.

The thing to understand about the quoted portion (here is the patent: https://patents.google.com/patent/US7603527B2/en?oq=7%2c603%...), is that you don't know if "the physical address of the load or the store operations is not valid" until long before you get to operation 302. The quoted paragraph is about how you get the correct result even if you optimistically forward the store data to the load.

The gist of the optimization is as follows. Operation 302 checks if the page offset of a load matches a store. Address translation won't effect the page offset (just the page number). So if there is a match, there is a good chance that the operations alias. Now, say you do a store followed by a load. At 302, you may not have the actual physical address of the load and/or store. But if there is a match at 302 (in the page offset), there is a good chance you're just loading from a location you recently stored to. Optimistically forwarding the data from the store buffer to the load allows the load to continue to make progress. Otherwise, you'd have to wait for both physical addresses to be resolved at Operation 310 before the load could continue.

As an aside, I think de-tuning this at the microcode or architectural level is probably a fool's errand. What you need is an architectural mode that basically says "this code needs to be protected from information leakage due to timing attacks." Then you can turn off speculation or whatever in such code.

Having a mode switch isn't enough; potentially any part of the program could contain a spectre gadget, so you'd need to run the whole program in that mode.

A perhaps better approach would be to ensure that speculation aborts actually do clean up the entire microarchitectural state - ensuring that the cache state (and other persistent state) isn't affected by aborted predicted execution.

Specifically, the question is whether the current process (user or kernel) has mapped anywhere in its address space data the attacker[0] should not be able to access.

0: ie, there exists a entity (other than the end user) that should not have access.

Also, good luck working the aforementioned mode switch into every darn programming language, and finding all the places in existing code bases where it has to be applied, and then doing the work.

It also means that this was laid out in writing for security researchers: but research came there none, before the last couple of years. As I've said before, the speculative-execution CPU bugs really seem to be the financial crisis or replication crisis of computing: https://news.ycombinator.com/item?id=16105385

Does anyone know of a list/summary of all the different ways OoO can be exploited to create vulnerabilities? It seems there's a new attack coming out every month, it'd be nice to see a list of attacks, affected configurations, and workarounds if they exist.

for those who want to know about who's impacted :

"Intel. We notified Intel about our findings, including a preliminary writeup and proof-of-concept code, on January 31st, 2019. Intel had acknowledged the issue and requested an embargo on the results in this paper, ending May 14th, 2019. Intel has further classified this issue as Microarchitectural Store Buffer Data Sampling (MSBDS), assigning it CVE-2018-12126 and a CVSS ranking of Medium. Finally, Intel had indicated that we are the first academic group to report this issue and that a similar issue was found internally as well.

AMD. We also notified AMD’s security response team regarding our findings, including our writeup. AMD had investigated this issue of their architectures and indicated that AMD CPUs are not vulnerable to the attacks described in this paper.

ARM. We have also notified ARM’s security response team regarding our findings. ARM had investigated this issue and found that ARM CPUs are not vulnerable to the attacks described in this paper. IBM. Finally, we also notified IBM security about the finding reported in this work. IBM had responded that none of their CPUs is affected, including System-V and PowerPC."

And here's a description (from the paper) of the mechanisms :

"The Mechanism Behind Fallout. Fallout exploits an optimization that we call Write Transient Forwarding (WTF), which incorrectly passes values from memory writes to subsequent memory reads. In a nutshell, when the program writes a value to memory, the processor needs to first translate the virtual address of the destination to a physical address and then acquire exclusive access to the location. Rather than stalling the store instruction and subsequent computation, the processor records the value and the address in the store buffer, and continues executing the program. The store buffer then resolves the address, acquires the access to the memory location and stores the data.

When a value is in the store buffer, care should be taken that subsequent loads from the same address do not read stale values from memory. To solve this, the processor matches the addresses of all load instructions against addresses in the store buffer. In the case of a match, the processor forwards the matching value from the store buffer to the load instruction. To increase efficiency, the processor uses partial address matches to rule out the need for store-to-load forwarding. WTF kicks in when a load instruction partially matches a preceding store and the processor determines that the load is bound to fail. In such cases, instead of cleaning up the state of the processor, it marks the load as faulty, and incorrectly forwards the value of the partially matched store.

Exploiting the WTF optimization. Fallout exploits this behavior to leak, through a microarchitectural channel, the value that WTF incorrectly forwards. The attacker deliberately performs a faulty load, causing the CPU to transiently forward an incorrect value from the store buffer. We subsequently leak the value using a Flush+Reload [58] side channel. As the store buffer is a shared resource used by all software running on a CPU core, the incorrectly-forwarded value might not even belong to the attacker’s process."

> ... we call Write Transient Forwarding (WTF)...

WTF starts to be rather overloaded acronym. Reminds me for example of "Wobbly Transformation Format" [0]...

[0]: https://en.wikipedia.org/wiki/UTF-8#WTF-8

"Fallout" also seems like a particularly bad choice of name to sell this vulnerability by, since the game series is so well known.

(I would also prefer descriptive names rather than this kind of marketing, but these are the times we live in.)

The game series is named after an even better known word. This is a bit like complaining Super Mario Odyssey shouldn't have been named after the Honda Odyssey and that Assassin's Creed: Odyssey needlessly muddied the waters further.

The instantly recognizable connotation of "fallout" is "nuclear fallout". The game series, which is post-apocalyptically themed, almost certainly derives its name from that.

Of course it does. But in the context of computing equipment, the computer game series was the first thing that came to my mind. Nuclear fallout is in a different domain.

> for those who want to know about who's impacted : ...

tldr: Intel. Intel is impacted. Again.

Shared resources, especially across privilege boundaries, almost always create undesirable side channels that leak information.

Hardware designers need to come up with better isolation mechanisms.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact