Re: [PATCH v4 10/18] iommu/vt-d: Restore IOMMU state and reclaimed domain ids

[Samiullah Khawaja <skhawaja@google.com>]

On Thu, Aug 27, 2026 at 03:22:26PM +0800, Baolu Lu wrote:
>On 8/8/26 10:27, Samiullah Khawaja wrote:
>>During boot fetch the preserved state of IOMMU unit and if found then
>>restore the state.
>>
>>- Reuse the root_table that was preserved in the previous kernel.
>>- Reclaim the domain ids of the preserved domains for each preserved
>>   devices so these are not acquired by another domain.
>>
>>Signed-off-by: Samiullah Khawaja <skhawaja@google.com>
>>---
>>  drivers/iommu/intel/iommu.c      | 111 +++++++++++++++++++------------
>>  drivers/iommu/intel/iommu.h      |   7 ++
>>  drivers/iommu/intel/liveupdate.c |  69 +++++++++++++++++++
>>  3 files changed, 144 insertions(+), 43 deletions(-)
>>

[snip]

>>+
>>+static int _restore_used_domain_ids(struct iommu_device_ser *ser, void *arg)
>>+{
>>+	int id = ser->domain_iommu_ser.attachment_id;
>>+	struct iommu_hw_ser *iommu_hw_ser;
>>+	struct intel_iommu *iommu = arg;
>>+
>>+	if (WARN_ON(!ser->domain_iommu_ser.iommu_phys))
>>+		return 0;
>>+
>>+	iommu_hw_ser = phys_to_virt(ser->domain_iommu_ser.iommu_phys);
>>+	if (iommu_hw_ser->type != IOMMU_INTEL)
>>+		return 0;
>>+
>>+	/* Only allocate domain ID from associated IOMMU HW unit */
>>+	if (iommu_hw_ser->intel.phys_addr != iommu->reg_phys)
>>+		return 0;
>>+
>>+	/*
>>+	 * This can fail as multiple preserved devices can share the same domain
>>+	 * ID. Since this is done during DMAR init so these failures can be
>>+	 * ignored.
>>+	 */
>>+	ida_alloc_range(&iommu->domain_ida, id, id, GFP_ATOMIC);
>
>This mixes two different cases:
>
>- another preserved device has already reserved the same DID.
>- the DID is not reserved because ida_alloc_range() failed (for example,
>  memory allocation failure).
>
>Case #1 is expected. Case #2 must not be ignored, because it means
>restore failed. So this should probably be something like (not tested):
>
>	mutex_lock(&iommu->did_lock);
>	if (ida_find_first_range(&iommu->domain_ida, id, id) >= 0)
>		BUG_ON(ida_alloc_range(&iommu->domain_ida, id, id, GFP_KERNEL) < 0)
>	mutex_unlock(&iommu->did_lock);

This is a good point, I will update this. Actually, I was thinking of
adding a new struct in the preserved state that represents the
association between domain-iommu-device. All the drivers have this, so
it is better to have a representation of this. Let me evaluate that.
>
>?
>
>By the way, why GFP_ATOMIC here at all?

Will update as you suggested.
>
>>+	return 0;
>>+}
>>+
>>+/**
>>+ * intel_iommu_liveupdate_restore_root_table() - Restore root table and reclaim domain IDs
>>+ * @iommu: Target IOMMU
>>+ * @iommu_ser: Serialized IOMMU hardware state from previous kernel
>>+ *
>>+ * Restores the preserved root table and context tables for the IOMMU hardware
>>+ * instance across Live Update, and reclaims all domain IDs previously allocated
>>+ * to preserved devices so they are not reused.
>>+ */
>>+void intel_iommu_liveupdate_restore_root_table(struct intel_iommu *iommu,
>>+					       struct iommu_hw_ser *iommu_ser)
>>+{
>>+	if (!iommu_ser->intel.restored)
>>+		iommu_restore_pages(iommu_ser->intel.root_table);
>>+
>>+	iommu->root_entry = __va(iommu_ser->intel.root_table);
>>+
>>+	if (!iommu_ser->intel.restored)
>>+		restore_iommu_context(iommu);
>>+
>>+	iommu_ser->intel.restored = 1;
>
>This uses iommu_ser->intel.restored to prevent restoring the root and
>context tables multiple times. For this to work safely, iommu_ser-
>>intel.restored must be 0 the first time restore runs after kexec.
>
>The problem is: this field is inside struct iommu_hw_ser, and that
>struct is allocated in the old kernel. How to ensure that the previous
>kernel has zeroed this out? Maybe I’m overthinking this.

This is a valid point. However, the old kernel allocates this struct
with GFP_ZERO and it never touches the restored field prior to kexec.
So it should be zero in the new kernel.
>
>>+	BUG_ON(iommu_for_each_preserved_device(_restore_used_domain_ids, iommu));
>>+}
>>+
>>  /**
>>   * intel_iommu_preserve_device() - Intel IOMMU callback to preserve device state
>>   * @dev: Target device
>
>Thanks,
>baolu

Thanks,
Sami