Re: [PATCH 0/7] RFC: coverage deduplication for KCOV

[Alexander Potapenko]

> > Below are the average stats from the runs.
> Is there test without trace collection? Is bitmap only enough?

If we bump the bitmap size to ~16K, it should be enough to keep all
the fuzzing results from a single run.
We haven't experimented with it much though, because syzkaller
currently processes coverage as an array of PCs, not a bitmap.
Changing this would require a major rework of syzkaller, given that
the positions in the bitmap may differ between different machines, so
we'll need to maintain a reverse mapping between bits and PCs in every
executor.

Such a mapping could be implemented on the kernel side on top of the
proposed patches, once someone proves a need for that.