The subject should say something like "atomic unencrypted allocations.." as many other atomic allocations are fine. Which brings up that with the codebase in this patch we can't really support architectures that require both an atomic pool for uncached remapping for just some devices and unencrypted for others. We don't have such an archicture right now, and I hope we don't grow one, but we probably need a little safeguard with a BUILD_BUG_ON if both options are set. I can send an incremental patch for that if that is ok with you. _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
- [patch 0/7] unencrypted atomic DMA pools with dyn... David Rientjes via iommu
- [patch 5/7] dma-pool: add pool sizes to debu... David Rientjes via iommu
- Re: [patch 5/7] dma-pool: add pool sizes... Tom Lendacky
- Re: [patch 5/7] dma-pool: add pool s... Christoph Hellwig
- [patch 2/7] dma-pool: add additional coheren... David Rientjes via iommu
- [patch 7/7] dma-pool: scale the default DMA ... David Rientjes via iommu
- [patch 3/7] dma-pool: dynamically expanding ... David Rientjes via iommu
- [patch 6/7] x86/mm: unencrypted non-blocking... David Rientjes via iommu
- [patch 4/7] dma-direct: atomic allocations m... David Rientjes via iommu
- Re: [patch 4/7] dma-direct: atomic alloc... Christoph Hellwig
- [patch 1/7] dma-remap: separate DMA atomic p... David Rientjes via iommu
- Re: [patch 1/7] dma-remap: separate DMA ... Christoph Hellwig
- Re: [patch 0/7] unencrypted atomic DMA pools... Christoph Hellwig
- Re: [patch 0/7] unencrypted atomic DMA p... David Rientjes via iommu
