On Mon, Mar 03, 2025 at 06:34:45PM +0100, Heinz Mauelshagen wrote: > It is, unless you ensure the pre-allocated and fully provisioned file to be > immutable. Which you can't. > > > The inode fiemap() operation would avoid issues relative to outdated > > bmap(). > > > > No, using that for I/O is even worse. > > > > Are you saying it imposes more overhead than the VFS APIs? No, but the reporting in fiemap is even less defined than the already overly lose bmap. fiemap for btrfs for example doesn't report actual block numbers.
