On 1 Apr 2016, at 16:08, Eric Blake <ebl...@redhat.com> wrote: > But yes, I'm favoring a) as well, for the simplicity factor. There's > still the issue that if we document a behavior, a new client talking to > an older server can't reliably tell if the behavior will be guaranteed.
Existing clients should not be sending FUA on anything other than NBD_CMD_WRITE *and* relying on the behaviour, as the behaviour is not documented (hence this discussion). Therefore it shouldn't break anything. I also think it won't break anything in practice as qemu doesn't use FUA on write and the kernel doesn't use FUA at all; I realise that is not an exhaustive list. -- Alex Bligh
signature.asc
Description: Message signed with OpenPGP using GPGMail
