On Wed, Sep 25, 2019, at 10:56 AM, Chris Mason wrote: > The data is verified while being decompressed, but that's a fairly large > fuzzing surface (all of zstd, zlib, and lzo). A lot of people will > correctly argue that we already have that fuzzing surface today, but I'd > rather not make a really easy way to stuff arbitrary bytes through the > kernel decompression code until all the projects involved sign off. Right. So maybe have this start of as a BTRFS ioctl and require privileges? I assume that's sufficient for what Omar wants. (Are there actually any other popular Linux filesystems that do transparent compression anyways?)
