linux-kernel - [RPC]: iommu/dma: custom iommu_setup_dma

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <77a82fb5-873f-f57c-a6fc-9cd879273d57@synaptics.com>
Date:   Tue, 7 Mar 2023 18:52:44 +0800
From:   Hsia-Jun Li <Randy.Li@...aptics.com>
To:     iommu@...ts.linux.dev
Cc:     linux-kernel@...r.kernel.org, robin.murphy@....com,
        joro@...tes.org, will@...nel.org
Subject: [RPC]: iommu/dma: custom iommu_setup_dma_ops for secure iommu

Hello All

If I want to implement a secure IOMMU. Is there any way other than hook 
the iommu_setup_dma_ops() I still want to use the of(device tree) helper?

The secure IOMMU is a little different from the existing secure 
attribute page table. The physical address and iova address not 
available for the Linux environment(REE), even they are available, they 
can't be configured to the secure device directly. So we don't need the 
Linux kernel to allocate the iova address for us, means I need to 
replace the current iommu_dma_map_sg() implementation.

Even sometimes it didn't do the iommu work, just allocate a secure 
reference page id for a continuous memory. I think the iommu framework 
is the most proper framework to implement this feature, it just fit the 
steps of importing a buffer for the secure world.

Please give me some advice on this idea.

-- 
Hsia-Jun(Randy) Li