Hi Philippe,
On 10/9/25 10:19 PM, Philippe Mathieu-Daudé wrote:
Hi Alejandro,
On 10/10/25 03:33, Alejandro Jimenez wrote:
I know the commit log is not consistent so far, but going forward I
propose we adopt the shorter prefix "amd_iommu: " for commit
summaries. There is no ambiguity (only one arch has amd_iommu), so the
full path is not required (i.e. avoid 'hw/i386/amd_iommu: '). Shorter
boilerplate leaves more space for relevant details, and helps people
like me who struggle to comply with character limits :).
What about "hw/amd_iommu:" to keep 'hw' in subject?
Is there any tooling that relies on the hw prefix? Skipping the arch in
the prefix is confusing I think, since hw/amd_iommu is not a valid path
in the repository.
I was looking for precedent of any preferred format in the commit logs
under hw/i386/ and there is a lot of variance. But specifically for
IOMMU emulation code, my interpretation is that the short prefix style
is most commonly used e.g.
Common x86 IOMMU uses "x86-iommu: "
The VT-d changes are typically in the form:
"intel_iommu: XYZ", which Clément also pointed out recently in:
https://lore.kernel.org/qemu-devel/[email protected]/
virtio IOMMU uses "virtio-iommu: "
RISC-V IOMMU uses the full path: "hw/riscv/riscv-iommu: "
SPARC64 has a few commits with "sun4u_iommu: "
I don't believe the 'hw' component is required to avoid ambiguity, but
perhaps there is something else I am missing...
Thank you,
Alejandro
