In my hobby project, I started always passing an allocator argument to every function or object which requires allocation (inspired by Zig) and I love it so far. Often I can just pass a bump pointer allocator or a stack-based allocator and do not care about deallocation of individual objects. I also wrote a simple unit testing framework to test out-of-memory conditions because it's easy to do when you're in control of allocators. Basically I inject an allocator which calculates how many allocations are done when a unit test is run, and then unit tests are later rerun again by injecting OOM at every known allocation point. A lot of bugs and crashes happen when an OOM is encountered because such paths are rarely tested. The idea of my pet project is a very resilient HTTP server with request-scoped allocations and recovery from OOM without crashes.
Adding that when many linux distributions face OOM, a killer daemon steps in and might kill your service even if you were handling the situation properly.
Interestingly (confusingly), Linux's OOM killer is invoked for a different notion of OOM than a null return from malloc / bad_alloc exception. On a 64-bit machine, the latter will pretty much only ever happen if you set a vsize ulimit or you pass an absurd size into malloc. The OOM killer is the only response when you actually run out of memory.
If you want to avoid your program triggering the OOM killer all on its own, you need to set up a vsize such that you'll get an application level error before actually exhausting memory. Even that isn't completely foolproof (obviously anyone with a shell can allocate a large amount of RAM), but in practice -- if your program is the only significant thing on the system -- you can get it to be very reliable this way.
Add in some cgroup settings and you should be able to keep your program from being OOM killed at all, though that step is a bit more complex.
The idea is that the server must have a known allocation budget, similar to Java's max heap size. There's a tree of allocators, i.e. a temporarily created arena allocator needs initial memory for its arena, so it can grab it from the root allocator. And the root allocator ultimately must be fixed-size and deterministic. Sure if there are other processes in the system allocating without concern for other apps, then the OOM killer can kill the server. But if there's no such process, I think it should be pretty stable.
Oh wow that is a really interesting test solution. That would be an interesting thing to add to all zig tests (I know they already have the testing allocator and good valgrind support but I don't think that tests/simulates oom).
I love things like these that use existing tests and expand the to just test further thing in already covered flows. We have done similar things at my work where we test expansion of data models against old models to check that we cover upgrade scenarios.
I've been using this helper: https://github.com/judofyr/zini/blob/ea91f645b7dc061adcedc91.... It starts by making the first allocation fail, then the second, then the third, and so on. As long as it returns OutOfMemory (without leaking memory) then everything is fine.
A very old trick for running Lua in your PlayStation 2 game (where the PS2 is a machine with 32MB of RAM and no memory paging) is to hook Lua’s realloc function into the venerable Doug Lea’s Malloc (https://gee.cs.oswego.edu/dl/html/malloc.html) set up to run in arena mode (ONLY_MSPACES? It’s been a decade or two…). That way Lua can fragment the arena all it wants without making a mess of the rest of the tiny address space.
> (where the PS2 is a machine with 32MB of RAM and no memory paging)
The PS2 hardware does have full support for memory paging (at least on the main cpu core). PS2 Linux makes full use of it.
But the default TLB configuration from the BIOS is just a single static 31MB page (the other 1MB is reserved for the BIOS) and the SDK doesn't provide any tooling for dynamic pages.
And this is MIPS, so it's a software managed TLB with 48 entries. I wouldn't be surprised if some games did have dynamic paging, but they would need to provide their own TLB exception handler.
I recently needed to write a memory allocator and being lazy asked ChatGPT for help. Interestingly it came up with implementation eerily similar to what is described in that document.
Nonetheless everything worked like a charm from the start.
Writing a memory allocator is a standard homework assignment in any CS program. The solutions are mostly similar, and ChatGPT has probably learned hundreds of examples.
There is also the double-sided arena allocator which uses one contiguous buffer but grows in both directions (front-to-back and back-to-front). When allocating memory from it you need to indicate whether you want memory from the front or back. The allocator is out-of-memory when both font and back meet.
The double-sided approach is useful for various purposes. For instance you can allocate short-lived data from the front and long-lived data from the back. It also makes better use of free space: with two separate arena allocators one could be out-of-memory but the other might have free space. With the double-sided approach all memory is fair game.
Fantastic article! I have a project with a similar arena allocator, so I'll definitely be taking some of these tricks. One thing my allocator does do is organize arenas into a linked list so that you can grow your size dynamically.
However I really like the article's point that you're always going to be living within _some_ memory budget, so you might as well allocate everything up front into a giant arena, and then divide the giant arena up into smaller arenas.
Also I've heard that you can save an instruction when checking if your allocator is full by subtracting from the top, and checking the zero flag. It seems to complicate alignment logic. Does that ever end up mattering?
> However I really like the article's point that you're always going to be living within _some_ memory budget, so you might as well allocate everything up front into a giant arena, and then divide the giant arena up into smaller arenas.
That depends. If you’re running on e.g. a video game console where you’re the sole user of a block of pretty much all memory, go ahead. On a system with other things running, you generally don’t want to assume you can just take some amount of memory, even if it’s “just the free memory”, or even “I probably won’t use it so it will be overcommitted”. Changing system conditions and other system pressure are outside of your control and your reservation may prevent the system from doing its job effectively and prioritizing your application appropriately.
Yeah, profiling is your friend. I forget if it's called a sharded slab or a buddy allocator, but the one where you have different preallocated buffers chunked at different sizes. Any time you allocate you are given the smallest chunk that will hold what you asked for. Profiling gives you optimal size boundaries as well as the number of each. Add a safety margin and off you go. Super fast allocation and guaranteed no fragmentation. In a c++ codebase overloading std::new to do this is probably the easiest way to get your allocation performance back and avoid fragmentation.
> If you’re running on e.g. a video game console where you’re the sole user of a block of pretty much all memory
Games consoles haven't been that for a long time. PS5 and XSS are full blown multi-user multi-application systems. PS4 and Xbox One were multi user systems with reserved blocks for the OS, but still very close to a modern OS.
> so you might as well allocate everything up front into a giant arena, and then divide the giant arena up into smaller arenas
However if you do this note how the article hints at this strategy needing a bit more code on Windows: Windows doesn't do overcommit by default. If you do one big malloc Windows will grow the page file to ensure it can page that much memory in if you start writing to it. That's fine if you allocate a couple megabytes, but if your area is gigabytes in size you want to call VirtualAlloc with MEM_RESERVE to get a big contiguous memory area, then call VirtualAlloc with MEM_COMMIT as needed on chunks you actually want to use.
> While you could make a big, global char[] array to back your arena, it’s technically not permitted (strict aliasing).
Aren't char pointers/arrays allowed to alias everything?
I used that technique in my programming language and its allocator. It's freestanding so I couldn't use malloc. I had to get memory from somewhere so I just statically allocated a big array of bytes. It worked perfectly but I do disable strict aliasing as a matter of course since in systems programming there's aliasing everywhere.
char can alias everything, i.e. you can deference a char pointer with impunity, even if the actual dynamic type[1] of an object is a different type. The reverse is not true: if the dynamic type of an object is char, just by using the alias rules, you can't deference it as an object of a different type.
In C++ you can just use placement new to change the dynamic type of (part of ) a char array (but beware of pointer provenance). In C is more complex: I don't claim to understand this fully, but my understanding is you can't change the type of a named object, but you should be able to change the type of anonymous memory (for example, what is allocated with malloc) by simply writing into it.
In practice at least GCC considers the full alias rules unimplementable and my understanding is that it uses a conservative model where every store can change the type of an object, and uses purely structural equivalence.
[1] of course most C implementations don't actually track dynamic types at runtime.
> The reverse is not true: if the dynamic type of an object is char, just by using the alias rules, you can't deference it as an object of a different type.
A limitation like that simply makes no sense to me. Everything is a valid char array but I can't place structs on top of one? Oh well, nothing I can do about it. I'll just keep strict aliasing disabled. If we're writing C, it's because we want to do stuff like that without the compiler getting clever about it.
> you should be able to change the type of anonymous memory (for example, what is allocated with malloc) by simply writing into it
Well, in my case, I'm the one writing the malloc and the buffer is the anonymous memory. I remember months ago I scoured the GCC documentation for some kind of builtin that would allow me to mark the memory as such but there was nothing. I did add some malloc attributes to my allocation function just like TFA suggested but apparently its main purpose is to optimize based on aliasing nonsense which I disabled anyway.
If you overlay a struct in your (char) buffer and dereference a pointer to it you would be accessing something with a different type than its declared type(char* as struct something *), it’s strict aliasing violation
To do stay in the rules you could set up a void* to suitable region in a linkerscript
Some other good cases for arenas are rendering of a frame in a video game and handling of a http request. The memory is contained within that context and short lived.
That's the lifetime of the objects in the arena, but wouldn't you recycle the arena itself across frames?
For requests it might make sense to have low and high water marks so that additional arenas are created during request peaks and destroyed after if you want to limit long term memory usage of your application.
Typically you would have different arenas for different lifetimes ('per frame', 'per level', 'per game session' - or maybe even more specialized, like the duration of an IO operation), and 'reset' the arenas at the end of their respective lifetimes (which is basically just setting the current 'top offset' to zero). This sort-of expects that object destruction is a no-op (e.g. no destructors need to be called).
The general idea being that you don't need to track granular 'per-object lifetimes', but only a handful of 'arena lifetimes', and all objects share the lifetime of the arena they've been allocated in.
Of course it's also possible to manually call a destructor on an object in the arena without recycling its memory, but I heavily prefer using plain POD structs without owning pointers and which can be safely 'abandondend' at any time without leaking memory.
I never heard the term "arena allocation" before. I always thought it was called "bump pointer allocation" since all you do is add to (bump) a pointer. One useful trick when designing a bump allocator is to allocate word size bytes (8 on 64-bit) extra to store object headers in. For example, if you store objects' sizes in the header you can iterate all allocated objects and you can often also reallocate objects in the middle of the arena without having to free every object.
Readit News
If you want to avoid your program triggering the OOM killer all on its own, you need to set up a vsize such that you'll get an application level error before actually exhausting memory. Even that isn't completely foolproof (obviously anyone with a shell can allocate a large amount of RAM), but in practice -- if your program is the only significant thing on the system -- you can get it to be very reliable this way.
Add in some cgroup settings and you should be able to keep your program from being OOM killed at all, though that step is a bit more complex.
I love things like these that use existing tests and expand the to just test further thing in already covered flows. We have done similar things at my work where we test expansion of data models against old models to check that we cover upgrade scenarios.
- https://github.com/ziglang/zig/blob/1606717b5fed83ee64ba1a91...
- https://www.ryanliptak.com/blog/zig-intro-to-check-all-alloc...
Deleted Comment
The PS2 hardware does have full support for memory paging (at least on the main cpu core). PS2 Linux makes full use of it.
But the default TLB configuration from the BIOS is just a single static 31MB page (the other 1MB is reserved for the BIOS) and the SDK doesn't provide any tooling for dynamic pages.
And this is MIPS, so it's a software managed TLB with 48 entries. I wouldn't be surprised if some games did have dynamic paging, but they would need to provide their own TLB exception handler.
The double-sided approach is useful for various purposes. For instance you can allocate short-lived data from the front and long-lived data from the back. It also makes better use of free space: with two separate arena allocators one could be out-of-memory but the other might have free space. With the double-sided approach all memory is fair game.
Also I've heard that you can save an instruction when checking if your allocator is full by subtracting from the top, and checking the zero flag. It seems to complicate alignment logic. Does that ever end up mattering?
That depends. If you’re running on e.g. a video game console where you’re the sole user of a block of pretty much all memory, go ahead. On a system with other things running, you generally don’t want to assume you can just take some amount of memory, even if it’s “just the free memory”, or even “I probably won’t use it so it will be overcommitted”. Changing system conditions and other system pressure are outside of your control and your reservation may prevent the system from doing its job effectively and prioritizing your application appropriately.
Games consoles haven't been that for a long time. PS5 and XSS are full blown multi-user multi-application systems. PS4 and Xbox One were multi user systems with reserved blocks for the OS, but still very close to a modern OS.
Deleted Comment
Anyways, you can find a full article about up vs down at https://fitzgeraldnick.com/2019/11/01/always-bump-downwards....
However if you do this note how the article hints at this strategy needing a bit more code on Windows: Windows doesn't do overcommit by default. If you do one big malloc Windows will grow the page file to ensure it can page that much memory in if you start writing to it. That's fine if you allocate a couple megabytes, but if your area is gigabytes in size you want to call VirtualAlloc with MEM_RESERVE to get a big contiguous memory area, then call VirtualAlloc with MEM_COMMIT as needed on chunks you actually want to use.
> While you could make a big, global char[] array to back your arena, it’s technically not permitted (strict aliasing).
Aren't char pointers/arrays allowed to alias everything?
I used that technique in my programming language and its allocator. It's freestanding so I couldn't use malloc. I had to get memory from somewhere so I just statically allocated a big array of bytes. It worked perfectly but I do disable strict aliasing as a matter of course since in systems programming there's aliasing everywhere.
In C++ you can just use placement new to change the dynamic type of (part of ) a char array (but beware of pointer provenance). In C is more complex: I don't claim to understand this fully, but my understanding is you can't change the type of a named object, but you should be able to change the type of anonymous memory (for example, what is allocated with malloc) by simply writing into it.
In practice at least GCC considers the full alias rules unimplementable and my understanding is that it uses a conservative model where every store can change the type of an object, and uses purely structural equivalence.
[1] of course most C implementations don't actually track dynamic types at runtime.
A limitation like that simply makes no sense to me. Everything is a valid char array but I can't place structs on top of one? Oh well, nothing I can do about it. I'll just keep strict aliasing disabled. If we're writing C, it's because we want to do stuff like that without the compiler getting clever about it.
> you should be able to change the type of anonymous memory (for example, what is allocated with malloc) by simply writing into it
Well, in my case, I'm the one writing the malloc and the buffer is the anonymous memory. I remember months ago I scoured the GCC documentation for some kind of builtin that would allow me to mark the memory as such but there was nothing. I did add some malloc attributes to my allocation function just like TFA suggested but apparently its main purpose is to optimize based on aliasing nonsense which I disabled anyway.
To do stay in the rules you could set up a void* to suitable region in a linkerscript
https://lwn.net/Articles/316126/
https://lkml.org/lkml/2003/2/26/158
Deleted Comment
Some other good cases for arenas are rendering of a frame in a video game and handling of a http request. The memory is contained within that context and short lived.
For requests it might make sense to have low and high water marks so that additional arenas are created during request peaks and destroyed after if you want to limit long term memory usage of your application.
The general idea being that you don't need to track granular 'per-object lifetimes', but only a handful of 'arena lifetimes', and all objects share the lifetime of the arena they've been allocated in.
Of course it's also possible to manually call a destructor on an object in the arena without recycling its memory, but I heavily prefer using plain POD structs without owning pointers and which can be safely 'abandondend' at any time without leaking memory.
Deleted Comment