On Tue, Feb 01, 2011 at 09:47:15AM +0100, Corinna Vinschen wrote: >On Jan 31 20:44, Christian Franke wrote: >> If used on raw devices like /dev/sda fsync() always fails with >> EBADRQC (54) because FlushFileBuffers() always fails with >> ERROR_INVALID_FUNCTION (1). >> >> The attached patch fixes this by simply ignoring this error in the >> fhandler_base implementation. This should not affect any real flush >> errors which likely would return other error codes. >> >> An alternative approach would be to ignore the error only in a new >> fhandler_raw_dev/floppy::fsync(). IMO not worth the effort is this >> case. > >I agree. I applied the patch.
Shouldn't this patch be a little more robust and check that you're getting ERROR_INVALID_FUNCTION only for raw devices? cgf
