scudo

It is not nearly as terrible as it was, but scudo, memalloc, and glibc's malloc are better.

Surely the scudo[1] memory allocator, which was implemented in Android 11 has helped here as well?

For example in Chimera Linux, musl uses the llvm scudo allocator instead of the default allocator packaged with musl libc.

The Maltese scudo is divided into twelve tarì, which are divided into twenty grani, which are divided into six piccioli.

Chimera uses the scudo allocator, which is Ajay also the default on Android and Fuscia.

Speed can be a problem with the stock malloc but you are free to include scudo, jemalloc, or any malloc you want.

Another option for dynamic bounds checking on heap allocations is Electric Fence or scudo.

However, I recall Scudo is woefully underperforming on 32-bit systems (it may not even work?).

If we just wouldn't use inline metadata, or verify it's integrity (smth like scudo but not broken), all of the security issues with the heap would just be gone.

* Chimera Linux uses BSD userland while Adelie uses GNU coreutils * Chimera has the dinit init system / service manager while Adelie uses OpenRC * Adelie uses gcc as far as I can tell while Chimera uses clang as the default system compiler * Chimera uses the scudo memory allocator rather than musl's default memory allocator.

I have reproduced parts of it that I found interesting (Heavily edited for telegraphic brevity): "Like many of America’s new generation of users, Ms Scudo's ddiction began in 2000 when, after a hip injury, a doctor prescribed her...

As if from nowhere, Ms Scudo had become a heroin addict." "last year 11m Americans used illicitly-acquired prescription painkillers, more than the number who used cocaine, ecstasy, methamphetamine and LSD combined." "Though Afghanistan accounts for 80% of global opium production, America gets most of its heroin from Mexico.

Or, are the likes of memory tagging, sanitizers, hardened allocators (Scudo in bionic / Arena/dl/Ros in ART) enough of a defense?

You can fully enable userspace heap MTE for the stock OS via the standard allocator implementation (Scudo) which is currently not particularly hardened.

Scudo is apparently faster than musl's default allocator.