On Sat, Dec 21, 2019 at 05:41:18AM -0800, Christoph Hellwig wrote: > On Fri, Dec 13, 2019 at 01:57:45PM -0600, Goldwyn Rodrigues wrote: > > From: Goldwyn Rodrigues <rgoldwyn@xxxxxxxx> > > > > Filesystems such as btrfs can perform direct I/O without holding the > > inode->i_rwsem in some of the cases like writing within i_size. > > So, remove the check for lockdep_assert_held() in iomap_dio_rw() > > As said last time: in the callers the assert is completely pointless, > as it is always very close to taking the lock. This was just intended > to deal with callers not adhering to the iomap_dio_rw calling > conventins, and moving the assert to the calllers doesn't help with > that at all. > > So if you think you need to remove it do just that and write a changelog > explaining the why much better. Uh... that's exactly what this patch does, and the commit message says that btrfs doesn't need to hold i_rwsem to guarantee concurrency correctness. Hm, but maybe the commit message is simply too subtle here? How about: "Filesystems do not necessarily need i_rwsem to ensure correct operation with multiple threads, so remove the i_rwsem assertion in iomap_dio_rw. For example, btrfs can perform directio within i_size without needing to hold i_rwsem." --D
