Re: [PATCH v18 2/4] vfio-pci/zdev: Add a device feature for error information

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

On 6/3/2026 3:37 PM, Alex Williamson wrote:
> On Wed,  3 Jun 2026 11:24:13 -0700
> Farhan Ali <alifm@linux.ibm.com> wrote:
>> diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h
>> index 5de618a3a5ee..3eead86a00ab 100644
>> --- a/include/uapi/linux/vfio.h
>> +++ b/include/uapi/linux/vfio.h
>> @@ -1534,6 +1534,36 @@ struct vfio_device_feature_dma_buf {
>>    */
>>   #define VFIO_DEVICE_FEATURE_MIG_PRECOPY_INFOv2  12
>>   
>> +/**
>> + * VFIO_DEVICE_FEATURE_ZPCI_ERROR feature provides PCI error information to
>> + * userspace for vfio-pci devices on s390. On s390, PCI error recovery
>> + * involves platform firmware and notification to operating systems is done
>> + * by architecture specific mechanism. Exposing this information to
>> + * userspace allows it to take appropriate actions to handle an
>> + * error on the device. The ioctl returns -ENOMSG if there are no pending
>> + * PCI errors.
>> + */
>> +
>> +struct vfio_device_feature_zpci_err {
>> +	__u32 feature_flags;		/* Indicate future features */
>> +	__u32 reserved1;
>> +	__u32 fh;			/* function handle */
>> +	__u32 fid;			/* function id */
>> +	__u32 ett		:  4;	/* expected table type */
>> +	__u32 mvn		: 12;	/* MSI vector number */
>> +	__u32 dmaas		:  8;	/* DMA address space */
>> +	__u32 reserved2		:  6;
>> +	__u32 q			:  1;	/* event qualifier */
>> +	__u32 rw		:  1;	/* read/write */
> There's a 4-byte hole here.
>
>> +	__u64 faddr;			/* failing address */
>> +	__u32 reserved3;
>> +	__u16 reserved4;
>> +	__u16 pec;			/* PCI event code */
>> +	__u8 reserved5[28];		/* Allow for future expansion */
>> +};
>> +
>> +#define VFIO_DEVICE_FEATURE_ZPCI_ERROR 13
> Based on the discussion with Omar and FMB reporting, I don't suppose
> this could also be copied to a user buffer as opaque data with size
> reported elsewhere, could it?  Thanks,
>
> Alex

We currently don't expose any PCI error data to userspace and so 
userspace has no way of knowing the size of the buffer. Unless we 
provide the size, i don't think we can copy the user buffer as opaque data.

Thanks

Farhan