On Wed, Aug 13, 2008 at 09:40:34PM +0300, Avi Kivity wrote:
Gleb Natapov wrote:
but what bother me is that we
consciously drop user data that we can easily save. Why? Real HW tries
hard to save every bit of user data and we just decided to drop it. The
difference between cancel or complete a request may be corrupted or not
corrupted file system after a crash.
If a guest didn't wait for an I/O to complete, it shouldn't expect it to
be on disk.
If you'll save your swiss bank account details in file and suddenly
your guest will crash, you'll surely appreciate if qemu will save it
for you :)
Of course, we can have the IDE layer wait instead of cancelling, which
will get the request onto the disk.
That is the current approach, but the wait is done for all block IO not
just IDE.