Messages in this thread | | | Date | Sat, 20 Sep 2008 09:54:54 +0900 | Subject | Re: [PATCH] remove fullflush and nofullflush in IOMMU generic option | From | FUJITA Tomonori <> |
| |
On Sat, 20 Sep 2008 00:39:51 +0200 Joerg Roedel <joerg.roedel@amd.com> wrote:
> On Sat, Sep 20, 2008 at 07:18:11AM +0900, FUJITA Tomonori wrote: > > I don't know it's acceptable if removing the exported interface even > > if the maintainer of it wants to remove it and it's totally > > meaningless. > > > > > > > reverting is the wrong way. For AMD IOMMU I want to use the > > > iommu=fullflush way because I want to reuse a parameter thats already > > > there. Thats why I am against your reverting patch. > > > So now I stop repeating my points again and again. EOD. > > > > I understood you want to use iommu=fullflush but you can't touch the > > generic code without any discussion. > > > > And even if everyone is happy about the change, it's much better to > > start from scratch rather than try to fix the things done in the wrong > > way. > > Just sleep a night about the discussions for this issue and count then > how often you changed your opinions and points. It all started with a > objection from you against 'nofullflush' and ended in that you want to > explain me the develpment process. And then think again if this > unimportant minor change was worth all that ridiculous flaming.
Your patch made two changes to the interfaces exported to users, which we can't change in the future. And they are the changes to the interface that all IOMMUs can use. Any changes to the interfaces exported to users are always very important for me because we can't change or remove them later. I think that we should make such changes after other developers agree that making the changes is a good idea.
But seems that you have a different opinion about changes to the interfaces exported to users. Then I can see why we can't agree about these changes.
If Ingo doesn't marge my patch to revert your changes to the interfaces exported for users, then I'll fix only 'nofullflush' option, which is clearly a bad change, as you admitted. I'll move 'nofullflush' back to GART as before. If you think that you can remove the option, please send your patch to remove it because I don't think that any changes to the existing exported interface is a good idea and I just try to fix the problem that your patch introduced. Removing 'nofullflush' is a different topic.
| |