[PATCH v3 1/6] iommu/core: split mapping to page sizes as supported by the hardware
From: Ohad Ben-Cohen <hidden>
Date: 2011-10-11 17:01:49
Also in:
kvm, linux-omap, lkml
Subsystem:
arm/qualcomm mailing list, iommu subsystem, kernel virtual machine (kvm), qualcomm iommu, the rest · Maintainers:
Joerg Roedel, Will Deacon, Paolo Bonzini, Rob Clark, Linus Torvalds
On Tue, Oct 11, 2011 at 12:38 PM, Roedel, Joerg [off-list ref] wrote:
You need to make sure that you don;t pass larger regions then requested to the low-level driver. Some logic like in the iommu_map function should do it.
You're right. But the solution IMHO should be getting rid of that size -> order -> size conversion which we do back and forth, as it's kinda pointless.
Please use PAGE_SIZE instead of 0x1000UL.
This one would go away too if we remove the size -> order -> size conversions. In fact, iommu_unmap() just becomes so much simpler when we do that. I'm attaching two patches: first one that removes the size->order->size conversions we have, and then the usual pgsize patch rebased to it. Please take a look, Thanks, Ohad.