RE: [PATCH v6 09/10] hisi_acc_vfio_pci: Add support for VFIO live migration

[Shameerali Kolothum Thodi <shameerali.kolothum.thodi@huawei.com>]

> -----Original Message-----
> From: Jason Gunthorpe [mailto:jgg@nvidia.com]
> Sent: 28 February 2022 14:58
> To: Shameerali Kolothum Thodi <shameerali.kolothum.thodi@huawei.com>
> Cc: kvm@vger.kernel.org; linux-kernel@vger.kernel.org;
> linux-crypto@vger.kernel.org; alex.williamson@redhat.com;
> cohuck@redhat.com; mgurtovoy@nvidia.com; yishaih@nvidia.com; Linuxarm
> <linuxarm@huawei.com>; liulongfang <liulongfang@huawei.com>; Zengtao (B)
> <prime.zeng@hisilicon.com>; Jonathan Cameron
> <jonathan.cameron@huawei.com>; Wangzhou (B) <wangzhou1@hisilicon.com>
> Subject: Re: [PATCH v6 09/10] hisi_acc_vfio_pci: Add support for VFIO live
> migration
> 
> On Mon, Feb 28, 2022 at 09:01:20AM +0000, Shameer Kolothum wrote:
> 
> > +static int hisi_acc_vf_stop_copy(struct hisi_acc_vf_core_device
> *hisi_acc_vdev,
> > +				 struct hisi_acc_vf_migration_file *migf)
> > +{
> > +	struct acc_vf_data *vf_data = &migf->vf_data;
> 
> This now needs to hold the migf->lock
> 
> > +
> > +	if ((cur == VFIO_DEVICE_STATE_STOP || cur ==
> VFIO_DEVICE_STATE_PRE_COPY) &&
> > +	    new == VFIO_DEVICE_STATE_RUNNING) {
> > +		hisi_acc_vf_start_device(hisi_acc_vdev);
> 
> This should be two stanzas STOP->RUNNING should do start_device
> 
> And PRE_COPY->RUNNING should do disable_fds, and presumably nothing
> else - the device was never stopped.
> 

Ok. I will take care of all the above.

> > +	} else if (cmd == VFIO_DEVICE_MIG_PRECOPY) {
> > +		struct vfio_device_mig_precopy precopy;
> > +		enum vfio_device_mig_state curr_state;
> > +		unsigned long minsz;
> > +		int ret;
> > +
> > +		minsz = offsetofend(struct vfio_device_mig_precopy, dirty_bytes);
> > +
> > +		if (copy_from_user(&precopy, (void __user *)arg, minsz))
> > +			return -EFAULT;
> > +		if (precopy.argsz < minsz)
> > +			return -EINVAL;
> > +
> > +		ret = hisi_acc_vfio_pci_get_device_state(core_vdev, &curr_state);
> > +		if (!ret && curr_state == VFIO_DEVICE_STATE_PRE_COPY) {
> > +			precopy.initial_bytes = QM_MATCH_SIZE;
> > +			precopy.dirty_bytes = QM_MATCH_SIZE;
> 
> dirty_bytes should be 0
> 
> initial_bytes should be calculated based on the current file
> descriptor offset.
> 
> The use of curr_state should be eliminated
> 
> This ioctl should be on the saving file_operations, not here
> 
> + * This ioctl is used on the migration data FD in the precopy phase of the
> + * migration data transfer. It returns an estimate of the current data sizes
> 
> I see there is a bug in the qemu version:
> 
> @@ -215,12 +218,13 @@ static void vfio_save_precopy_pending(QEMUFile
> *f, void *>
>                                        uint64_t *res_postcopy_only)
>  {
>      VFIODevice *vbasedev = opaque;
> +    VFIOMigration *migration = vbasedev->migration;
>      struct vfio_device_mig_precopy precopy = {
>          .argsz = sizeof(precopy),
>      };
>      int ret;
> 
> -    ret = ioctl(vbasedev->fd, VFIO_DEVICE_MIG_PRECOPY, &precopy);
> +    ret = ioctl(migration->data_fd, VFIO_DEVICE_MIG_PRECOPY, &precopy);
>      if (ret) {
>          return;
>      }
> 
> I'll update my github.

Ok. Thanks for that.

And for the VFIO_DEVICE_MIG_PRECOPY ioctl, this is what I have now,

+static long hisi_acc_vf_save_unl_ioctl(struct file *filp,
+                                      unsigned int cmd, unsigned long arg)
+{
+       struct hisi_acc_vf_migration_file *migf = filp->private_data;
+       loff_t *pos = &filp->f_pos;
+       struct vfio_device_mig_precopy precopy;
+       unsigned long minsz;
+
+       if (cmd != VFIO_DEVICE_MIG_PRECOPY)
+               return -EINVAL;
+
+       minsz = offsetofend(struct vfio_device_mig_precopy, dirty_bytes);
+
+       if (copy_from_user(&precopy, (void __user *)arg, minsz))
+               return -EFAULT;
+       if (precopy.argsz < minsz)
+               return -EINVAL;
+
+       mutex_lock(&migf->lock);
+       if (*pos > migf->total_length) {
+               mutex_unlock(&migf->lock);
+               return -EINVAL;
+       }
+
+       precopy.dirty_bytes = 0;
+       precopy.initial_bytes = migf->total_length - *pos;
+       mutex_unlock(&migf->lock);
+       return copy_to_user((void __user *)arg, &precopy, minsz) ? -EFAULT : 0;
+}
+

I had a quick run with the above Qemu changes, and looks ok. Please let me know.

Thanks,
Shameer