> > Hmmm. I'm wondering if it would be better to return -EOPNOTSUPP here > > so that the application knows it can't do non-blocking write AIO to > > this file. > > I wondered the same too. In the end, I decided to go with silently ignoring > the flag (for now) since raw block device accesses do the same (the NOWAIT > support is not complete and IOs may wait on free tags). I have an idea for > fixing simply the out-of-order issuing that may result from using nowait. I find this aspect interesting for further clarification. > I will send a patch for that later and can then remove this. Should another software improvement be integrated into the initial proposal? Regards, Markus
