Re: [PATCH 5.10 392/530] iommu/vt-d: Preset Access/Dirty bits for IOVA over FL

From: Sasha Levin
Date: Sun May 16 2021 - 22:50:16 EST


On Mon, May 17, 2021 at 10:38:42AM +0800, Lu Baolu wrote:
Hi Wolfgang,

On 5/15/21 9:28 PM, Wolfgang Müller wrote:
Hi!

First of all, apologies if this is the wrong place to post a problem
report. I figured since I was going to reference a particular commit
anyway I might as well reply to the patch series that (seemed to have)
introduced the problem.

From: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>

[ Upstream commit a8ce9ebbecdfda3322bbcece6b3b25888217f8e3 ]

The Access/Dirty bits in the first level page table entry will be set
whenever a page table entry was used for address translation or write
permission was successfully translated. This is always true when using
the first-level page table for kernel IOVA. Instead of wasting hardware
cycles to update the certain bits, it's better to set them up at the
beginning.

This commit seems to trigger a kernel panic very early in boot for me in
5.10.37 (36 is fine):

It seems due to the back-ported patch:

- if (!sg) {
- sg_res = nr_pages;
- pteval = ((phys_addr_t)phys_pfn << VTD_PAGE_SHIFT) | attr;
+ if (domain->domain.type == IOMMU_DOMAIN_DMA) {
+ attr |= DMA_FL_PTE_ACCESS;
+ if (prot & DMA_PTE_WRITE)
+ attr |= DMA_FL_PTE_DIRTY;
+ }
}

+ pteval = ((phys_addr_t)phys_pfn << VTD_PAGE_SHIFT) | attr;

Greg, do you want me to rework this patch, or submit an incremental fix?

Could you send a reworked patch please?

--
Thanks,
Sasha