Re: [PATCH v8 7/9] vfio-pci/zdev: Add a device feature for error information

[Farhan Ali <alifm@linux.ibm.com>]

On 1/27/2026 2:53 AM, Niklas Schnelle wrote:
> On Thu, 2026-01-22 at 11:44 -0800, Farhan Ali wrote:
>> For zPCI devices, we have platform specific error information. The platform
>> firmware provides this error information to the operating system in an
>> architecture specific mechanism. To enable recovery from userspace for
>> these devices, we want to expose this error information to userspace. Add a
>> new device feature to expose this information.
>>
>> Signed-off-by: Farhan Ali <alifm@linux.ibm.com>
>> ---
>>   drivers/vfio/pci/vfio_pci_core.c |  2 ++
>>   drivers/vfio/pci/vfio_pci_priv.h |  9 ++++++++
>>   drivers/vfio/pci/vfio_pci_zdev.c | 35 ++++++++++++++++++++++++++++++++
>>   include/uapi/linux/vfio.h        | 16 +++++++++++++++
>>   4 files changed, 62 insertions(+)
>>
> --- snip ---
>>   
>> +int vfio_pci_zdev_feature_err(struct vfio_device *device, u32 flags,
>> +			      void __user *arg, size_t argsz)
>> +{
>> +	struct vfio_device_feature_zpci_err err;
>> +	struct vfio_pci_core_device *vdev;
>> +	struct zpci_dev *zdev;
>> +	int head = 0;
>> +	int ret;
>> +
>> +	vdev = container_of(device, struct vfio_pci_core_device, vdev);
>> +	zdev = to_zpci(vdev->pdev);
>> +	if (!zdev)
>> +		return -ENODEV;
>> +
>> +	ret = vfio_check_feature(flags, argsz, VFIO_DEVICE_FEATURE_GET,
>> +				 sizeof(err));
>> +	if (ret != 1)
>> +		return ret;
>> +
>> +	mutex_lock(&zdev->pending_errs_lock);
>> +	if (zdev->pending_errs.count) {
>> +		head = zdev->pending_errs.head % ZPCI_ERR_PENDING_MAX;
>> +		err.pec = zdev->pending_errs.err[head].pec;
> In the previous patch you saved the entire struct zpci_ccdf_err now you
> only copy out and expose the PCI event code, though? If you do want to
> only expose that the commit message should state this and the reason
> for this restriction. Additionally I think the struct
> vfio_device_feature_zpci_err should include a mechanism (version +
> size?) to allow upgrading it to the full error information in the
> future.

I think having explicit version variable in the struct should be 
sufficient (the __pad can be replaced with version). I don't think we 
need explicit size variable? I have looked at some of the capability 
structures in vfio_zdev.h, as examples and so we could use a similar 
approach here when we need to extend the vfio_device_feature_zpci_err?

Though I don't see any other vfio device feature structure being 
explicitly versioned. I am open to any guidance/suggestions on the best 
practices on how to we version VFIO device feature structs.


>
> Then again why not just expose the entire CCDF? It's an architected
> data structure without and if you add it at the end of struct
> vfio_device_feature_zpci_err and add a size you should even be able to
> handle if it ever needs to grow. Of course you'd have to create a copy
> of the struct to use the the uAPI types so I'd probably also add a
> BUILD_BUG_ON() check on matching size. Or am I missing a reason to keep
> just the PEC?

I wanted to keep the information exposed to userspace minimal. The CCDF 
exposes far more information and may not be needed by userspace/VM. 
Today the PEC is sufficient for user space(QEMU) to take bubble up to a 
VM. I also wanted to avoid having a copy of the struct in 2 places.

Thanks

Farhan

>> +		zdev->pending_errs.head++;
>> +		zdev->pending_errs.count--;
>> +		err.pending_errors = zdev->pending_errs.count;
>> +	}
>> +	mutex_unlock(&zdev->pending_errs_lock);
>> +
>> +	if (copy_to_user(arg, &err, sizeof(err)))
>> +		return -EFAULT;
>> +
>> +	return 0;
>> +}
>> +