Il 13/08/2014 12:19, Ming Lei ha scritto: >> > The problem is that your patches to do touch too much code and subtly >> > break too much stuff. The one I wrote does have a little breakage > Could you give a hint about which stuff are broken? Last time, you mention > virtio-scsi need to keep AIOCB live after returning, I have fixed it in V1.
They are dataplane-specific, while there's no reason not to have the same benefits elsewhere. They are file-specific, while there's no reason not to have the same benefits for e.g. iSCSI (though iSCSI now uses coroutines instead of bdrv_aio_*). They touch AioContext for no reason, and introduce a bunch of layering violations everywhere. They are simply the wrong API. >> > because I don't understand bs->growable 100% and I didn't really put >> > much effort into it (my deadline being basically "be done as soon as the >> > shower is free"), and it is ugly as hell, _but_ it should be compatible >> > with the way the block layer works. > I will take a careful look to your patch later. > > If coroutine is still there, I think it still can slow down performance. No, it's not there. Please try the patch. Paolo