Re: [libvirt] [PATCH 09/12] storage: Cleanup failures in virStorageBackendCreateRaw

On Wed, Oct 14, 2015 at 09:35:24 -0400, John Ferlan wrote: > On 10/14/2015 08:16 AM, Peter Krempa wrote: >> On Tue, Oct 13, 2015 at 16:12:46 -0400, John Ferlan wrote: ... >> Erm, you've patched virFileOpenAs a few patches before this to do the >> right thing when creating the file so I don't see a reason to make sure >> again. >> > > I know that, look closer at the code after the call to virFileOpenAs. We > know that if virOpenFileAs succeeds then we've created the file (if it > didn't exist prior to calling virOpenFileAs); however, if the following > fails: > > > if ((ret = createRawFile(fd, vol, inputvol, reflink_copy)) < 0) > /* createRawFile already reported the exact error. */ > ret = -1; > > we would have left the function creating the file, but not deleting it > because some subsequent action failed. I agree that virStorageBackendCreateRaw needs to delete the file in some cases. I just don't think that checking if the file exists prior to the call to virOpenFileAs is necessary or makes sense in any way.