rmalloc

Crates.iormalloc
lib.rsrmalloc
version1.1.0
sourcesrc
created_at2020-09-28 02:29:31.891069
updated_at2020-10-27 08:41:45.224447
descriptionrandomized malloc
homepage
repository
max_upload_size
id293566
size247,428
iximeow (iximeow)

documentation

README

rmalloc

what's the point of an address space this large if we don't use chunks of it randomly

  • a wise software engineer

rmalloc is a malloc (and calloc, and realloc, and, of course, free) implementation that improves upon traditional malloc implementations by taking advantage of the size of modern processor address spaces to provide cryptographically enhanced security in allocations.

usage

rmalloc is most easily used with a simple LD_PRELOAD - first, get the repo, then run the program you want to secure like normal, but with LD_PRELOAD=path/to/librmalloc.so in front of it. for example:

LD_PRELOAD=./target/release/librmalloc.so cargo build

in the rmalloc repo should complete without error.

for the security-minded user, safety-checks enables off-by-default checks to confirm that memory is not double-allocated. these checks can cause instability in many applications and are disabled by default. to build rmalloc with safety checks enabled, cargo build --release --features safety-checks. IMPORTANT: if you intend to use rmalloc with safety checks enabled, read the following section!

"help, my program reports that it crashed with Segmentation fault!!!!!"

it probably caught the segfault rmalloc uses to probe if a page can be used for a new allocation, and thought the fault was due to its own behavior. vim, bash, and collect2 both do this, to name a few. some applications do not chain signal handlers on the assumption they have exclusive interest in signals or signal handling, so naively overwriting the SIGSEGV handler will irreparably break rmalloc.

theory

at its core, rmalloc uses the Mersenne Twister algorithm to randomly select addresses for allocations. it will then probe to see if the chosen address has been allocated, and if not, will then allocate it with a high-performance mmap system call. because the mmap function is implemented inside the Linux kernel, it is secure to RCE exploits and supply chain (software update) attacks. because it is in the krenel, it is also fast.

thread safety

rmalloc is thread safe.

no_std

rmalloc is no_std. it is appropriate for embedded usage to replace glibc or other malloc.

the name rmalloc

this is a joke crate. if you'd like the name for pretty much any more serious purpose feel free to email me.

changelog

1.1.0

  • made safety checks optional to improve rmalloc compatibility

1.0.1

  • first release of a new secure and randomized mallocator
Commit count: 0

cargo fmt