"W. Wilson Ho" wrote: > Andreas Dilger wrote: > > > > > Lastly, if a (single) user process creates files f1, f2, f3, ... > > > in that order, does the file system guarantee the order of > > > creation? That is, is it possible that after a journal playback, > > > that f2 is gone but f1 and f3 exist? > > > > I'm pretty sure this IS a guarantee of a journaling filesystem, otherwise > > if f3 is a file inside a directory f2, you have an inconsistent filesystem > > (i.e. missing blocks, unlinked file). Granted, it is possible after a > > crash that both f2 and f3 are gone, or all are gone, but that is not your > > question. > > Thanks for your reply! Now, if I do the following: > > 1) create/write a large file data.tmp > 2) sync > 3) mv data.tmp data > 4) touch data.stamp > > If data.stamp survives a system crash, does it mean the content > of data will be correct? Yes. > Basically, I'd like to have some simple > way (like stat'ing a file) to check if my large data file is > corrupted. > Thanks, vs
