[v3,1/6] dmaengine: idxd: Add enable/disable device IOPF feature
Commit Message
The iommu subsystem requires IOMMU_DEV_FEAT_IOPF must be enabled before
and disabled after IOMMU_DEV_FEAT_SVA, if device's I/O page faults rely
on the IOMMU. Add explicit IOMMU_DEV_FEAT_IOPF enabling/disabling in this
driver.
At present, missing IOPF enabling/disabling doesn't cause any real issue,
because the IOMMU driver places the IOPF enabling/disabling in the path
of SVA feature handling. But this may change.
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Fenghua Yu <fenghua.yu@intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
drivers/dma/idxd/init.c | 31 +++++++++++++++++++++++++------
1 file changed, 25 insertions(+), 6 deletions(-)
Comments
On 3/24/23 8:02 PM, Lu Baolu wrote:
> The iommu subsystem requires IOMMU_DEV_FEAT_IOPF must be enabled before
> and disabled after IOMMU_DEV_FEAT_SVA, if device's I/O page faults rely
> on the IOMMU. Add explicit IOMMU_DEV_FEAT_IOPF enabling/disabling in this
> driver.
>
> At present, missing IOPF enabling/disabling doesn't cause any real issue,
> because the IOMMU driver places the IOPF enabling/disabling in the path
> of SVA feature handling. But this may change.
>
> Reviewed-by: Dave Jiang<dave.jiang@intel.com>
> Reviewed-by: Fenghua Yu<fenghua.yu@intel.com>
> Reviewed-by: Kevin Tian<kevin.tian@intel.com>
> Signed-off-by: Lu Baolu<baolu.lu@linux.intel.com>
Hi Dave and Fenghua,
The following iommu patches depends on this one. Can I route it to
Linus through the iommu tree?
Best regards,
baolu
On 4/2/23 10:49 PM, Baolu Lu wrote:
> On 3/24/23 8:02 PM, Lu Baolu wrote:
>> The iommu subsystem requires IOMMU_DEV_FEAT_IOPF must be enabled before
>> and disabled after IOMMU_DEV_FEAT_SVA, if device's I/O page faults rely
>> on the IOMMU. Add explicit IOMMU_DEV_FEAT_IOPF enabling/disabling in this
>> driver.
>>
>> At present, missing IOPF enabling/disabling doesn't cause any real issue,
>> because the IOMMU driver places the IOPF enabling/disabling in the path
>> of SVA feature handling. But this may change.
>>
>> Reviewed-by: Dave Jiang<dave.jiang@intel.com>
>> Reviewed-by: Fenghua Yu<fenghua.yu@intel.com>
>> Reviewed-by: Kevin Tian<kevin.tian@intel.com>
>> Signed-off-by: Lu Baolu<baolu.lu@linux.intel.com>
>
> Hi Dave and Fenghua,
>
> The following iommu patches depends on this one. Can I route it to
> Linus through the iommu tree?
Hi Baolu, you'll need an ack from Vinod, who is the dmaengine subsystem
maintainer. I have no objections.
>
> Best regards,
> baolu
On 4/4/23 12:57 AM, Dave Jiang wrote:
> On 4/2/23 10:49 PM, Baolu Lu wrote:
>> On 3/24/23 8:02 PM, Lu Baolu wrote:
>>> The iommu subsystem requires IOMMU_DEV_FEAT_IOPF must be enabled before
>>> and disabled after IOMMU_DEV_FEAT_SVA, if device's I/O page faults rely
>>> on the IOMMU. Add explicit IOMMU_DEV_FEAT_IOPF enabling/disabling in
>>> this
>>> driver.
>>>
>>> At present, missing IOPF enabling/disabling doesn't cause any real
>>> issue,
>>> because the IOMMU driver places the IOPF enabling/disabling in the path
>>> of SVA feature handling. But this may change.
>>>
>>> Reviewed-by: Dave Jiang<dave.jiang@intel.com>
>>> Reviewed-by: Fenghua Yu<fenghua.yu@intel.com>
>>> Reviewed-by: Kevin Tian<kevin.tian@intel.com>
>>> Signed-off-by: Lu Baolu<baolu.lu@linux.intel.com>
>>
>> Hi Dave and Fenghua,
>>
>> The following iommu patches depends on this one. Can I route it to
>> Linus through the iommu tree?
>
> Hi Baolu, you'll need an ack from Vinod, who is the dmaengine subsystem
> maintainer. I have no objections.
Hi Vinod,
Does this work work for you?
Best regards,
baolu
@@ -516,6 +516,27 @@ static void idxd_disable_system_pasid(struct idxd_device *idxd)
idxd->sva = NULL;
}
+static int idxd_enable_sva(struct pci_dev *pdev)
+{
+ int ret;
+
+ ret = iommu_dev_enable_feature(&pdev->dev, IOMMU_DEV_FEAT_IOPF);
+ if (ret)
+ return ret;
+
+ ret = iommu_dev_enable_feature(&pdev->dev, IOMMU_DEV_FEAT_SVA);
+ if (ret)
+ iommu_dev_disable_feature(&pdev->dev, IOMMU_DEV_FEAT_IOPF);
+
+ return ret;
+}
+
+static void idxd_disable_sva(struct pci_dev *pdev)
+{
+ iommu_dev_disable_feature(&pdev->dev, IOMMU_DEV_FEAT_SVA);
+ iommu_dev_disable_feature(&pdev->dev, IOMMU_DEV_FEAT_IOPF);
+}
+
static int idxd_probe(struct idxd_device *idxd)
{
struct pci_dev *pdev = idxd->pdev;
@@ -530,7 +551,7 @@ static int idxd_probe(struct idxd_device *idxd)
dev_dbg(dev, "IDXD reset complete\n");
if (IS_ENABLED(CONFIG_INTEL_IDXD_SVM) && sva) {
- if (iommu_dev_enable_feature(dev, IOMMU_DEV_FEAT_SVA)) {
+ if (idxd_enable_sva(pdev)) {
dev_warn(dev, "Unable to turn on user SVA feature.\n");
} else {
set_bit(IDXD_FLAG_USER_PASID_ENABLED, &idxd->flags);
@@ -578,21 +599,19 @@ static int idxd_probe(struct idxd_device *idxd)
if (device_pasid_enabled(idxd))
idxd_disable_system_pasid(idxd);
if (device_user_pasid_enabled(idxd))
- iommu_dev_disable_feature(dev, IOMMU_DEV_FEAT_SVA);
+ idxd_disable_sva(pdev);
return rc;
}
static void idxd_cleanup(struct idxd_device *idxd)
{
- struct device *dev = &idxd->pdev->dev;
-
perfmon_pmu_remove(idxd);
idxd_cleanup_interrupts(idxd);
idxd_cleanup_internals(idxd);
if (device_pasid_enabled(idxd))
idxd_disable_system_pasid(idxd);
if (device_user_pasid_enabled(idxd))
- iommu_dev_disable_feature(dev, IOMMU_DEV_FEAT_SVA);
+ idxd_disable_sva(idxd->pdev);
}
static int idxd_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
@@ -710,7 +729,7 @@ static void idxd_remove(struct pci_dev *pdev)
pci_free_irq_vectors(pdev);
pci_iounmap(pdev, idxd->reg_base);
if (device_user_pasid_enabled(idxd))
- iommu_dev_disable_feature(&pdev->dev, IOMMU_DEV_FEAT_SVA);
+ idxd_disable_sva(pdev);
pci_disable_device(pdev);
destroy_workqueue(idxd->wq);
perfmon_pmu_remove(idxd);