On Mon, Aug 03, 2026 at 07:02:43PM -0500, Mike Christie wrote: > On 7/23/26 5:54 AM, Jia Jia wrote: > > The owner must keep the old mappings valid until the ioctl returns. > > Remapping or dropping them from another userspace thread before then is > > outside the lifetime assumption of this transition barrier. > > Ok I get it now. For some reason I thought that was valid. > > For the patch we don't want to flush if vhost_dev_ioctl fails > and there's no reason to flush for operations like > VHOST_GET_FORK_FROM_OWNER or VHOST_SET_OWNER. > > Could vhost_dev_ioctl return when a flush is needed?
I agree but also this is consistent with all other vhost devices. So I'm taking v2 of this, and we can iterate to optimize all devices with patches on top. -- MST
