We are seeing DMAR PTE read access not set errors when booting a
kernel with default passthrough, both with a test kernel and with
a 5.4.0 kernel. Previously we would see a number of identity mappings
being set related to the rmrrs, and now they aren't seen and we get
the dmar pte errors as devices touch those regions. From what I can tell
currently df4f3c603aeb ("iommu/vt-d: Remove static identity map code")
removed the bit of code in init_dmars that used to set up those
mappings:
- /*
- * For each rmrr
- * for each dev attached to rmrr
- * do
- * locate drhd for dev, alloc domain for dev
- * allocate free domain
- * allocate page table entries for rmrr
- * if context not allocated for bus
- * allocate and init context
- * set present in root table for this bus
- * init context with domain, translation etc
- * endfor
- * endfor
- */
- pr_info("Setting RMRR:\n");
- for_each_rmrr_units(rmrr) {
- /* some BIOS lists non-exist devices in DMAR table. */
- for_each_active_dev_scope(rmrr->devices, rmrr->devices_cnt,
- i, dev) {
- ret = iommu_prepare_rmrr_dev(rmrr, dev);
- if (ret)
- pr_err("Mapping reserved region failed\n");
- }
- }
si_domain_init now has code that sets identity maps for devices in rmrrs, but
only for certain devices.
With iommu=nopt, the system boots up without issue.
_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu
