> Subject: RE: [EXTERNAL] Re: [PATCH v2 0/2] vhost-vdpa: Add support for NO- > IOMMU mode > > > Subject: Re: [EXTERNAL] Re: [PATCH v2 0/2] vhost-vdpa: Add support for > > NO- IOMMU mode > > > > On Wed, Nov 06, 2024 at 12: 38: 02PM +0000, Srujana Challa wrote: > It > > is going in circles, let me give the summary, > Issue: We need to > > address the lack of no-IOMMU support in the vhost vDPA driver for > > better performance. > Measured > > On Wed, Nov 06, 2024 at 12:38:02PM +0000, Srujana Challa wrote: > > > It is going in circles, let me give the summary, > > > Issue: We need to address the lack of no-IOMMU support in the vhost > > > vDPA > > driver for better performance. > > > Measured Performance: On the machine "13th Gen Intel(R) Core(TM) > > > i9-13900K, 32 Cores", we observed > > > > Looks ike you are going in circles indeed. Lack of performance is > > never a reason to disable the basic memoy safety for userspace drivers. > If security is a priority, the IOMMU should stay enabled. In that case, this > patch wouldn’t cause any issues, right? Another important aspect of this patch is that it enables vhost-vdpa functionality on machines that do not support IOMMU, such as the Intel® Core™ i5-2320 CPU @ 3.00GHz. I believe, this enhancement is essential as it broadens the compatibility of our system, allowing users with embedded or less advanced hardware to benefit from vhost-vdpa features. Thanks. > > > > The (also quite bad) reason why vfio-nummu was added was to support > > hardware entirely with an iommu. > > > > There is absolutely no reason to add krnel code for new methods of > > unsafer userspace I/O without an IOMMU ever.
