On Wed, Dec 13, 2023 at 03:45:41PM -0700, Andreas Dilger wrote: > It should be possible for userspace and the kernel to increase the size of > struct statx independently, and not have any issues. If userspace requests > a field via STATX_* flags that the kernel doesn't understand, then it will > be masked out by the kernel, and any extended fields in the struct will not > be referenced. Likewise, if the kernel understands more fields than what > userspace requests, it shouldn't spend time to fill in those fields, since > userspace will ignores them anyway, so it is just wasted cycles.
It's not the kernel <-> userspace boundary that's the problem, it's userspace <-> userspace if you're changing the size of struct statx and passing it across a dynamic lib boundary.
