On Sat, Oct 07, 2017 at 08:54:06AM +0300, Amir Goldstein wrote: > > If fallocate(2) is called with the FALLOC_FL_KEEP_SIZE flag, either > > According to patch without keep size flag? Oops, thanks for catching that. I'll fix the commit description. > BTW looking at the patch, does truncate racing with fallocate that > would increase size beyond truncated size works correctly? Yes, because we the inode is locked via i_mutex (well, now an write lock on i_rwsem) during both truncate and fallocate operations. - Ted