incompatibility with code that stream `/proc/self/smaps` and doing small allocations per VMA

[inthewaves]

See GrapheneOS/platform_frameworks_base#361 for a failing test and sample parser code and GrapheneOS/os-issue-tracker#7916 (comment) for an explanation with an app that actually has code doing this.

Essentially, streaming /proc/self/smaps while doing allocations for each VMA will lead to a loop, as the stream will discover VMAs from hardened_malloc and make allocations for those

[thestinger]

I'm not sure what we can do about that. It's definitely a bug in the app.

#258 could help mitigate it depending on how it's integrated by avoiding making as many VMAs for both small and large allocations but it wouldn't stop new VMAs from being made.

[thestinger]

This isn't a bug in hardened_malloc and it's not going to drop the core security features to work around buggy apps. There's already #258 as an approach we can use to reduce VMA usage but it's not guaranteed to resolve this and has tradeoffs.