Alexander Graf <ag...@suse.de> writes: > On 26.04.2013, at 18:11, Christian Borntraeger wrote: > >> On 26/04/13 18:05, Alexander Graf wrote: >>> >>> On 26.04.2013, at 18:01, Christian Borntraeger wrote: >>> >>>> On 26/04/13 17:48, Alexander Graf wrote: >>>> >>>>>> So I suggest to go with this patch and implement later on what we >>>>>> agree upon? >>>>> >>>>> I don't see how having "first device we find" is any better than a >>>>> rushed interface we need to agree on right before 1.5 hard freeze. >>>> >>>> Exactly, find first device isnt better ;-) >>>> See, the current code chooses the first subchannel that matches. This >>>> boils down to "a random disk" as soon as you have more than one. >>>> >>>> With this patch the user can at least specify the devno of the disk.It >>>> even works out of the box with libvirt. >>>> >>>> Let's just release 1.5 with the very simple one and then go for something >>>> awesome in the next version. >>>> >>>> Exactly - and having a list is more in the awesome area. Beiing able to >>>> specify the first disk and pass that in a register to the bios is of >>>> course not a full-features interface, but it works and can be changed. >>> >>> My main concern is backwards compatibility. If we introduce a command line >>> interface now, we have to support it forever. I'd rather only support one >>> interface, rather than 2 out of which one is only legacy for 1.5 >>> compatibility. >> >> The cool thing is, that we dont introduce a command line interface in this >> patch. >> We just use the existing bootindex. > > Get an ack from Anthony on the bootindex thing and I'm fine with > pulling that in.
So I don't see a problem with the no-fallback behavior (nor with it changing down the road), but the bootindex change would cause a regression on x86. Regards, Anthony Liguori > > > Alex