> Subject: Re: [EXTERNAL] Re: [PATCH v2 0/2] vhost-vdpa: Add support for NO- > IOMMU mode > > On Mon, Oct 14, 2024 at 01: 18: 01PM +0000, Srujana Challa wrote: > > On Fri, > Sep 20, 2024 at 07: 35: 28PM +0530, Srujana Challa wrote: > > > This patchset > introduces support for an UNSAFE, no-IOMMU mode in the > > > vhost-vdpa > > On Mon, Oct 14, 2024 at 01:18:01PM +0000, Srujana Challa wrote: > > > On Fri, Sep 20, 2024 at 07:35:28PM +0530, Srujana Challa wrote: > > > > This patchset introduces support for an UNSAFE, no-IOMMU mode in > > > > the vhost-vdpa driver. When enabled, this mode provides no device > > > > isolation, no DMA translation, no host kernel protection, and > > > > cannot be used for device assignment to virtual machines. It > > > > requires RAWIO permissions and will taint the kernel. > > > > > > > > This mode requires enabling the > > > "enable_vhost_vdpa_unsafe_noiommu_mode" > > > > option on the vhost-vdpa driver and also negotiate the feature > > > > flag VHOST_BACKEND_F_NOIOMMU. This mode would be useful to get > > > > better performance on specifice low end machines and can be > > > > leveraged by embedded platforms where applications run in controlled > environment. > > > > > > ... and is completely broken and dangerous. > > Based on the discussions in this thread > > https://urldefense.proofpoint.com/v2/url?u=https-3A__www.spinics.net_l > > > ists_kvm_msg357569.html&d=DwIBAg&c=nKjWec2b6R0mOyPaz7xtfQ&r=Fj4O > oD5hcK > > > FpANhTWdwQzjT1Jpf7veC5263T47JVpnc&m=Kj2YVdoGecovW95oPf_fILveQer > 4EsJfWJ > > > XmzmACF_v1jROPwW343ZXF2nEc5JN7&s=Dw7EoKg_W8Ak7E0uGR4gT3vHBv > Em2uEP1Pvx0 > > 5dXVHI&e=, we have decided to proceed with this implementation. Could > > you please share any alternative ideas or suggestions you might have? > > Don't do this. It is inherently unsafe and dangerous and there is not valid > reason to implement it. > > Double-Nacked-by: Christoph Hellwig <hch@xxxxxx> When using the DPDK virtio user PMD, we’ve noticed a significant 70% performance improvement when IOMMU is disabled on specific low-end x86 machines. This performance improvement can be particularly advantageous for embedded platforms where applications operate in controlled environments. Therefore, we believe supporting the intel_iommu=off mode is beneficial. Thanks.
