It depends on Guest driver. If you look at virtio, the present guest drivers are implemented this way, I presume there cannot be a better way functionally. But if someone wants to reduce these notifications, they can do it in their Guest driver. There were some generic attempts like "TX Interrupt Moderation" and "polling mode" in virtio driver.
I have reduced notifications in my own implementation, but that’s specific to my Guest. Thanks, Venkatesh -----Original Message----- From: David Xu [mailto:[email protected]] Sent: Monday, June 23, 2014 8:43 AM To: Venkateswara Rao Nandigam Cc: kvm Subject: Re: vhost notification model 2014-06-22 22:57 GMT-04:00 Venkateswara Rao Nandigam <[email protected]>: > There are individual notifications for each packet/buffer on the both Rx and > Tx. So, it means there is no coalescing mechanism for Rx/Tx notification currently. Each request/response (Rx/Tx) will trigger a VM exit/entry. Right? Regards, Cong > > -----Original Message----- > From: [email protected] [mailto:[email protected]] On > Behalf Of David Xu > Sent: Sunday, June 22, 2014 11:55 PM > To: kvm > Subject: vhost notification model > > Hi All, > > When vhost receive packets, does it notify target VM for each request or a > bunch of requests? How about the tx path? When one VM send packets, does the > VM trap to privilege mode via VM exit and kick vhost for each packet or > buffer several packets then trap and kick once? > Thanks. > > Regards, > Cong > -- > To unsubscribe from this list: send the line "unsubscribe kvm" in the > body of a message to [email protected] More majordomo info at > http://vger.kernel.org/majordomo-info.html N�����r��y����b�X��ǧv�^�){.n�+����h����ܨ}���Ơz�&j:+v�������zZ+��+zf���h���~����i���z� �w���?�����&�)ߢf
