On 2024/1/15 15:58, Ethan Zhao wrote:
-static int qi_check_fault(struct intel_iommu *iommu, int index, int
wait_index)
+static int qi_check_fault(struct intel_iommu *iommu, int index, int
wait_index,
+ pci_dev *target_pdev)
{
u32 fault;
int head, tail;
+ u64 iqe_err, ice_sid;
struct q_inval *qi = iommu->qi;
int shift = qi_shift(iommu);
if (qi->desc_status[wait_index] == QI_ABORT)
return -EAGAIN;
+ /*
+ * If the ATS invalidation target device is gone this moment
(surprise
+ * removed, died, no response) don't try this request again. this
+ * request will not get valid result anymore. but the request was
+ * already submitted to hardware and we predict to get a ITE in
+ * followed batch of request, if so, it will get handled then.
+ */
We can't leave the ITE triggered by this request for the next one, which
has no context about why this happened. Perhaps move below code down to
the segment that handles ITEs.
Another concern is about qi_dump_fault(), which pr_err's the fault
message as long as the register is set. Some faults are predictable,
such as cache invalidation for surprise-removed devices. Unconditionally
reporting errors with pr_err() may lead the user to believe that a more
serious hardware error has occurred. Probably we can refine this part of
the code as well.
Others look sane to me.
+ if (target_pdev && !pci_device_is_present(target_pdev))
+ return -EINVAL;
+
fault = readl(iommu->reg + DMAR_FSTS_REG);
if (fault & (DMA_FSTS_IQE | DMA_FSTS_ITE | DMA_FSTS_ICE))
qi_dump_fault(iommu, fault);
@@ -1315,6 +1327,13 @@ static int qi_check_fault(struct intel_iommu
*iommu, int index, int wait_index)
tail = readl(iommu->reg + DMAR_IQT_REG);
tail = ((tail >> shift) - 1 + QI_LENGTH) % QI_LENGTH;
+ /*
+ * SID field is valid only when the ITE field is Set in
FSTS_REG
+ * see Intel VT-d spec r4.1, section 11.4.9.9
+ */
+ iqe_err = dmar_readq(iommu->reg + DMAR_IQER_REG);
+ ice_sid = DMAR_IQER_REG_ITESID(iqe_err);
+
writel(DMA_FSTS_ITE, iommu->reg + DMAR_FSTS_REG);
pr_info("Invalidation Time-out Error (ITE) cleared\n");
@@ -1324,6 +1343,16 @@ static int qi_check_fault(struct intel_iommu
*iommu, int index, int wait_index)
head = (head - 2 + QI_LENGTH) % QI_LENGTH;
} while (head != tail);
+ /*
+ * If got ITE, we need to check if the sid of ITE is the
same as
+ * current ATS invalidation target device, if yes, don't
try this
+ * request anymore, the target device has a response
time beyound
+ * expected. 0 value of ice_sid means old device, no
ice_sid value.
+ */
+ if (target_pdev && ice_sid && ice_sid ==
+ pci_dev_id(pci_physfn(target_pdev))
+ return -ETIMEDOUT;
+
if (qi->desc_status[wait_index] == QI_ABORT)
return -EAGAIN;
}
Best regards,
baolu
