Re: [PATCH v7 4/5] iommu/of: Require reg for iommu-addresses
From: Bryan O'Donoghue
Date: Thu Nov 12 2026 - 09:55:47 EST
On 12/11/2026 14:04, Vikash Garodia wrote:
But this change is not about requiring a reg. It's about not reserving when reg and IOVA don't match.
On 11/11/2026 5:32 PM, Bryan O'Donoghue wrote:
On 06/11/2026 17:14, Vikash Garodia wrote:
@@ -257,6 +252,9 @@ void of_iommu_get_resv_regions(struct device *dev, struct list_head *list)
}
type = iommu_resv_region_get_type(dev, &phys, iova, length);
+ if (type != IOMMU_RESV_DIRECT)
+ continue;
+
This change is not about requiring a reg though, which is why it should be a standalone patch with its own justifications.
This is about reg, which determines the direct mapping, hence the code is very much applicable to this patch. Please refer this code [1]
[1] https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/ tree/drivers/iommu/of_iommu.c#n223
So reading a patch entitled "Require reg for iommu-addresses" doesn't actually transmit the information of what this second change does.
Which is why that change should have its own patch.
Require reg != don't reserve reg + non-matching IOVA.
A reasonable reviewer might ask "why is a mismatch dropped instead of reserved" - which is the question I ask myself as I read and IMO isn't answered - which is why I've asked for a patch which reasons that out.
It may be that conjoining iommu-addresses/reg and !reserving mismatch are in fact the same thing but I don't think that jumps out of the code as you read it and I doubt adding to this commit log would make it much clearer because it seems like two changes in one patch.
It would be something like "require a reg and don't reserve on mismatch"
The "and" part is the part I'm flagging as likely deserving its own patch which explains what/why but "require reg" ain't it.
---
bod