I can't understand why address instability is a problem: if a Mutex is moved, then it can't be locked (because you need to hold a borrow while locked, which impedes moving), so using addresses is perfectly fine and there is absolutely no need to use IDs.
Also the fact that it doesn't detect locking the same mutex twice makes no sense: a static order obviously detects that and when locking multiple mutexes at the same level all you need to do is check for equal consecutive addresses after sorting, which is trivial.
Overall it seems like the authors are weirdly both quite competent and very incompetent. This is typical of LLMs, but it doesn't seem ZlLM-made.
The Level<> abstraction is a really neat way to have your cake and eat it too: you only need a consistent arbitrary order to avoid deadlocks, but the order can have performance consequences when some locks are more coarse than others.
But the example seems backwards to me: unless every callsite that locks any item always locks the big global lock first (probably not true, because if you serialize all item access on a global lock then a per-item lock serves no purpose...), aren't you begging for priority inversions by acquiring the big global lock before you acquire the item lock?
My only gripe is missing the obvious opportunity for Ferengi memes ("rules of acquisition") :D :D
There’s no global lock. There’s a linear MutexKey<N> that a lock of Level >= N has to be acquired with. Aquiring it consumes MutexKey<N> and hands you back MutexKey<Level+1> where Level is the N of the level you’re locking.
There’s no priority inversion possible because locks can only ever be held in decreasing orders of priority - you can’t acquire a low priority lock and then a high priority lock since your remaining MutexKey won’t have the right level.
...and now any other thread that needs big_lock() spins waiting for T2 to release it, but T2 is spinning waiting for T1 to release the (presumably less critical) small lock.
If small_lock is never ever acquired without acquiring big_lock first, small_lock serves no purpose and should be deleted from the program.
Usually a global lock is a lock that is taken outside all others and is taken for large parts of the runtime (or even, everywhere the thread isn't waiting on a condition variable, file descriptor and the like).
Mutex::new(AppConfig::default()) might very well be a small, leaf mutex.
I feel like Fuschia’s DAG approach can still be made compile time lock free by either disallowing holding locks from different branches or requiring an ordering when that does happen to prevent cycles (ie you can’t acquire them independently, you have to acquire all independent branches as a single group.
I really agree with jandrewrogers' point about the insularity of the database domain. While working on a custom C++ engine to handle 10M vectors in minimal RAM, I’ve noticed that many 'mainstream' concurrency patterns simply don't scale when cache-locality is your primary bottleneck.
In the DB world, we often trade complex locking for deterministic ordering or latch-free structures, but translating those to general-purpose app code (like what this Rust crate tries to do) is where the friction happens. It’s great to see more 'DB-style' rigour (like total ordering for locks) making its way into library design.
>This is a deliberate design decision. lock_tree uses a DAG, which lets you declare that branches A and B are independent — neither needs to come before the other. Sounds great, but it has a subtle problem: if thread 1 acquires A then B, and thread 2 acquires B then A, and both orderings are valid in the DAG, you have a deadlock that the compiler happily approved.
Would it be possible to build one at compile time? Static levels seem like they won't let you share code without level-collaboration, so that might be kinda important for larger-scale use.
I don't know enough about Rust's type system to know if that's possible though. Feels like it's pushing into "maybe" territory, like maybe not with just linear types but what about proc macros?
I can definitely see why it's easier to build this way though, and for some contexts that limitation seems entirely fine. Neat library, and nice post :)
I appreciate that this appears to be an incremental improvement on Fuschia's tree_lock, with the sharp edges sanded off. Good work! I hope I won't have to use it :p
I'm intrigued! I was fighting deadlocks in some Java code this week, and I'm working on a Rust project to maybe replace some of that.
One thing I didn't see in the post or the repo: does this work with async code?
I couldn't find the "search" button on Codeberg, and tests/integration.rs didn't have any async.
For embedded, I have had my eye on https://github.com/embassy-rs/embassy (which has an async runtime for embedded) and would love a nice locking crate to go with it.
IIUC, this crate has similar restrictions to the std Mutex. So it depends on what you mean by "work with async code."
First, lock acquisition seems to be a blocking method. And I don't see a `try_lock` method, so the naive pattern of spinning on `try_lock` and yielding on failure won't work. It'll still work in an async function, you'll just block the executor if the lock is contested and be sad.
Second, the key and guard types are not Send, otherwise it would be possible to send a key of a lower level to a thread that has already acquired a lock of a higher level, allowing deadlocks. (Or to pass a mutex guard of a higher level to a thread that has a key of a lower level.)
Therefore, holding a lock or a key across an await point makes your Future not Send.
Technically, this is fine. Nothing about Rust async in general requires that your Futures are Send. But in practice, most of the popular async runtimes require this. So if you want to use this with Tokio, for example, then you have to design your system to not hold locks or keys across await points.
This first restriction seems like it could be improved with the addition of an `AsyncLockable` trait. But the second restriction seems to me to be fundamental to the design.
Also to note, regarding “future not send,” that, in tokio codebases where the general expectation is that futures will be Send, enabling the clippy lint “future_not_send” is extremely helpful in avoiding these kinds of issues and also in keeping the error localized to the offending function, rather than it being miles away somewhere it happens to be getting indirectly spawned or whatever: https://rust-lang.github.io/rust-clippy/stable/index.html?se...
I agree with the author: it's a shame that TVars aren't catching on in more languages. They are a great idea from the database world, that we could use in the rest of computing, too.
The entire programming (or even computing) ecosystem suffers from this issue where very useful ideas don't always propagate across domains even though they just make a whole lot of sense. I'm not sure if it's because they truly wouldn't work out in practice, or if it's just a discovery/communication thing.
One thing that I think do affect things, is that language design discussions tend to be concentrated into their own communities based on the programming language itself, rather than one "programming language discussions" place where everyone can easier cross-pollinate ideas across languages. Luckily, there are some individuals who move between communities without effort, which does lead to a bit of ideas making it across, but it feels like we're missing out on so much evolution and ideas from various languages across the ecosystem.
The cross-fertilization of ideas across computer science domains is more limited than I think people assume. Databases are just one area that contains a lot of good ideas that never seem to leak into other parts of the software world.
Supercomputing is another domain that has deep insights into scalable systems that is famously so insular that ideas rarely cross over into mainstream scalable systems. My detour through supercomputing probably added as much to my database design knowledge as anything I actually did in databases.
> Models can be pulled along other axes, however, such as whether memory locations must be tagged in order to be used in a transaction or not, etc. Haskell requires this tagging (via TVars) so that side-effects are evident in the type system as with any other kind of monad. We quickly settled on unbounded transactions.
Snip
> In hindsight, this was a critical decision that had far-reaching implications. And to be honest, I now frequently doubt that it was the right call. We had our hearts in the right places, and the entire industry was trekking down the same path at the same time (with the notable exception of Haskell)
So basically not that TM isn’t workable, but unbounded TM is likely a fool’s errand but Haskell’s is bounded TM that requires explicit annotation of memory that will participate in atomicity.
Hrm. I'm not immediately impressed by the "Level<>" construct. That feels like a lot of new cognitive burden. It's also not at all obvious to me that multiple levels of mutex is a common pattern? I'm not sure I've ever encountered a situation where locking Account also and always requires locking Config? Heaven help you if you have 3 or more levels.
I dunno. I appreciate the opposition to "just be careful". But this feels to me like it's inducing bad design patterns. So it feels like it's wandering down the wrong path.
That's pretty awesome. Dead locks are extremely tough to debug. There are even cases where I saw behavior in code that might have been a dead lock. I never found out though.
So tired of this sort of comment. LLMs are trained using (primarily, generally) online material. It sounds like online humans, in aggregate, plus or minus a bit of policy on the part of the model builders.
Email the mods about it rather than replying, subject “Accusation of AI in FP comment” or whatever. It’s a guidelines violation to make the accusation in a comment rather than to them by email, and they have tools to deal with it!
Nobody is making an accusation of an AI comment - people are pointing out that the article is at least partially AI generate, which does not go against any HN guidelines, and neither does complaining about those comments.
That's exactly the problem. It sounds like one aggregate person. It's quite unpleasant to read the same turns of phrase again and again and again, especially when it means that the author copped out of writing it themselves.
In fairness I think in this case they mostly did write it themselves.
Except nobody writes like the aggregate, hence why it's so jarring.
The closest actually human style to LLM writing is obnoxious marketing speak. So that also sucks.
So many people who are not great writers lean on LLMs to write, but aren't good enough to see how bad it is. They should be criticised for this. Either use them and be good enough to make it read as human, or just don't use them. No free lunch.
I can't understand why address instability is a problem: if a Mutex is moved, then it can't be locked (because you need to hold a borrow while locked, which impedes moving), so using addresses is perfectly fine and there is absolutely no need to use IDs.
Also the fact that it doesn't detect locking the same mutex twice makes no sense: a static order obviously detects that and when locking multiple mutexes at the same level all you need to do is check for equal consecutive addresses after sorting, which is trivial.
Overall it seems like the authors are weirdly both quite competent and very incompetent. This is typical of LLMs, but it doesn't seem ZlLM-made.
What about mutexes living in shared memory, and each process having a different address mapping?
All bets go out the window with adversarial multi-process shared memory mutexes. The other process may not even be running the same locking code.
Don't address introduce ambiguous locking order across attempts?
While not obviously problematic, that seems weird enough you would need to validate that it is explicitly safe.
The Level<> abstraction is a really neat way to have your cake and eat it too: you only need a consistent arbitrary order to avoid deadlocks, but the order can have performance consequences when some locks are more coarse than others.
But the example seems backwards to me: unless every callsite that locks any item always locks the big global lock first (probably not true, because if you serialize all item access on a global lock then a per-item lock serves no purpose...), aren't you begging for priority inversions by acquiring the big global lock before you acquire the item lock?
My only gripe is missing the obvious opportunity for Ferengi memes ("rules of acquisition") :D :D
There’s no global lock. There’s a linear MutexKey<N> that a lock of Level >= N has to be acquired with. Aquiring it consumes MutexKey<N> and hands you back MutexKey<Level+1> where Level is the N of the level you’re locking.
There’s no priority inversion possible because locks can only ever be held in decreasing orders of priority - you can’t acquire a low priority lock and then a high priority lock since your remaining MutexKey won’t have the right level.
In the example it seems pretty clear to me that:
...is meant to be acquiring a mutex protecting some global config object, yes? That's what I'm calling a "global lock".> There’s no priority inversion possible because locks can only ever be held in decreasing orders of priority
...and now any other thread that needs big_lock() spins waiting for T2 to release it, but T2 is spinning waiting for T1 to release the (presumably less critical) small lock.If small_lock is never ever acquired without acquiring big_lock first, small_lock serves no purpose and should be deleted from the program.
Usually a global lock is a lock that is taken outside all others and is taken for large parts of the runtime (or even, everywhere the thread isn't waiting on a condition variable, file descriptor and the like).
Mutex::new(AppConfig::default()) might very well be a small, leaf mutex.
I feel like Fuschia’s DAG approach can still be made compile time lock free by either disallowing holding locks from different branches or requiring an ordering when that does happen to prevent cycles (ie you can’t acquire them independently, you have to acquire all independent branches as a single group.
I really agree with jandrewrogers' point about the insularity of the database domain. While working on a custom C++ engine to handle 10M vectors in minimal RAM, I’ve noticed that many 'mainstream' concurrency patterns simply don't scale when cache-locality is your primary bottleneck.
In the DB world, we often trade complex locking for deterministic ordering or latch-free structures, but translating those to general-purpose app code (like what this Rust crate tries to do) is where the friction happens. It’s great to see more 'DB-style' rigour (like total ordering for locks) making its way into library design.
>Why a Total Order, Not a DAG?
>This is a deliberate design decision. lock_tree uses a DAG, which lets you declare that branches A and B are independent — neither needs to come before the other. Sounds great, but it has a subtle problem: if thread 1 acquires A then B, and thread 2 acquires B then A, and both orderings are valid in the DAG, you have a deadlock that the compiler happily approved.
Would it be possible to build one at compile time? Static levels seem like they won't let you share code without level-collaboration, so that might be kinda important for larger-scale use.
I don't know enough about Rust's type system to know if that's possible though. Feels like it's pushing into "maybe" territory, like maybe not with just linear types but what about proc macros?
I can definitely see why it's easier to build this way though, and for some contexts that limitation seems entirely fine. Neat library, and nice post :)
I appreciate that this appears to be an incremental improvement on Fuschia's tree_lock, with the sharp edges sanded off. Good work! I hope I won't have to use it :p
I'm intrigued! I was fighting deadlocks in some Java code this week, and I'm working on a Rust project to maybe replace some of that.
One thing I didn't see in the post or the repo: does this work with async code?
I couldn't find the "search" button on Codeberg, and tests/integration.rs didn't have any async.
For embedded, I have had my eye on https://github.com/embassy-rs/embassy (which has an async runtime for embedded) and would love a nice locking crate to go with it.
IIUC, this crate has similar restrictions to the std Mutex. So it depends on what you mean by "work with async code."
First, lock acquisition seems to be a blocking method. And I don't see a `try_lock` method, so the naive pattern of spinning on `try_lock` and yielding on failure won't work. It'll still work in an async function, you'll just block the executor if the lock is contested and be sad.
Second, the key and guard types are not Send, otherwise it would be possible to send a key of a lower level to a thread that has already acquired a lock of a higher level, allowing deadlocks. (Or to pass a mutex guard of a higher level to a thread that has a key of a lower level.)
Therefore, holding a lock or a key across an await point makes your Future not Send.
Technically, this is fine. Nothing about Rust async in general requires that your Futures are Send. But in practice, most of the popular async runtimes require this. So if you want to use this with Tokio, for example, then you have to design your system to not hold locks or keys across await points.
This first restriction seems like it could be improved with the addition of an `AsyncLockable` trait. But the second restriction seems to me to be fundamental to the design.
Just wanted to add to your great summary a link to tokio’s docs on which kind of mutex to use, which seem applicable to the mutex in TFA as well: https://docs.rs/tokio/latest/tokio/sync/struct.Mutex.html#wh...
Also to note, regarding “future not send,” that, in tokio codebases where the general expectation is that futures will be Send, enabling the clippy lint “future_not_send” is extremely helpful in avoiding these kinds of issues and also in keeping the error localized to the offending function, rather than it being miles away somewhere it happens to be getting indirectly spawned or whatever: https://rust-lang.github.io/rust-clippy/stable/index.html?se...
I agree with the author: it's a shame that TVars aren't catching on in more languages. They are a great idea from the database world, that we could use in the rest of computing, too.
The entire programming (or even computing) ecosystem suffers from this issue where very useful ideas don't always propagate across domains even though they just make a whole lot of sense. I'm not sure if it's because they truly wouldn't work out in practice, or if it's just a discovery/communication thing.
One thing that I think do affect things, is that language design discussions tend to be concentrated into their own communities based on the programming language itself, rather than one "programming language discussions" place where everyone can easier cross-pollinate ideas across languages. Luckily, there are some individuals who move between communities without effort, which does lead to a bit of ideas making it across, but it feels like we're missing out on so much evolution and ideas from various languages across the ecosystem.
> Luckily, there are some individuals who move between communities without effort, [...]
Oh, many of these travelers spend a lot of effort!
It's discovery and communication. Public education for adults is way under-appreciated in many many scopes.
The cross-fertilization of ideas across computer science domains is more limited than I think people assume. Databases are just one area that contains a lot of good ideas that never seem to leak into other parts of the software world.
Supercomputing is another domain that has deep insights into scalable systems that is famously so insular that ideas rarely cross over into mainstream scalable systems. My detour through supercomputing probably added as much to my database design knowledge as anything I actually did in databases.
The canonical industrial explanation “why not” is probably this 2010 piece from Joe Duffy @ Microsoft:
http://joeduffyblog.com/2010/01/03/a-brief-retrospective-on-...
I don’t think we read the same thing.
> Models can be pulled along other axes, however, such as whether memory locations must be tagged in order to be used in a transaction or not, etc. Haskell requires this tagging (via TVars) so that side-effects are evident in the type system as with any other kind of monad. We quickly settled on unbounded transactions.
Snip
> In hindsight, this was a critical decision that had far-reaching implications. And to be honest, I now frequently doubt that it was the right call. We had our hearts in the right places, and the entire industry was trekking down the same path at the same time (with the notable exception of Haskell)
So basically not that TM isn’t workable, but unbounded TM is likely a fool’s errand but Haskell’s is bounded TM that requires explicit annotation of memory that will participate in atomicity.
It is a big reason why I picked Scala3/Zio over Rust for my most recent project.
Well, what means to support, truly, TVars?
Is easy, or hard?
Demand a new paradigm at large, or is only a inconvenience in the few places is used?
Because if the answer is "turns the language into Haskell" then is a big NOPE!
Hrm. I'm not immediately impressed by the "Level<>" construct. That feels like a lot of new cognitive burden. It's also not at all obvious to me that multiple levels of mutex is a common pattern? I'm not sure I've ever encountered a situation where locking Account also and always requires locking Config? Heaven help you if you have 3 or more levels.
I dunno. I appreciate the opposition to "just be careful". But this feels to me like it's inducing bad design patterns. So it feels like it's wandering down the wrong path.
What is the "graph" view on the right side?
That's pretty awesome. Dead locks are extremely tough to debug. There are even cases where I saw behavior in code that might have been a dead lock. I never found out though.
I'd read this, but I can't stomach this ChatGPT voice. It's absolutely grating.
> Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something.
tbh I'm not getting GPT-voice from this
I'm not either. If this was GPT-voice, I'd be happy. It's concise, technical, with good emphasis but no drama or AI tropes.
It's there in places ("The honest answer is...") but I think most of this is human written. They probably started with an AI draft I'd guess.
So tired of this sort of comment. LLMs are trained using (primarily, generally) online material. It sounds like online humans, in aggregate, plus or minus a bit of policy on the part of the model builders.
> So tired of this sort of comment.
Email the mods about it rather than replying, subject “Accusation of AI in FP comment” or whatever. It’s a guidelines violation to make the accusation in a comment rather than to them by email, and they have tools to deal with it!
Nobody is making an accusation of an AI comment - people are pointing out that the article is at least partially AI generate, which does not go against any HN guidelines, and neither does complaining about those comments.
They write like the worst possible person. It's terrible and obnoxious, there is no reason to put up with it.
> It sounds like online humans, in aggregate
That's exactly the problem. It sounds like one aggregate person. It's quite unpleasant to read the same turns of phrase again and again and again, especially when it means that the author copped out of writing it themselves.
In fairness I think in this case they mostly did write it themselves.
Except nobody writes like the aggregate, hence why it's so jarring.
The closest actually human style to LLM writing is obnoxious marketing speak. So that also sucks.
So many people who are not great writers lean on LLMs to write, but aren't good enough to see how bad it is. They should be criticised for this. Either use them and be good enough to make it read as human, or just don't use them. No free lunch.