virtio: bugfixes

just a bunch of bugfixes all over the place.
 
 Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
 -----BEGIN PGP SIGNATURE-----
 
 iQFDBAABCAAtFiEEXQn9CHHI+FuUyooNKB8NuNKNVGkFAmTVQCMPHG1zdEByZWRo
 YXQuY29tAAoJECgfDbjSjVRpclIH/A1l9iFK6v0eQ7cfdRjONTsd0YzlAdqU9J1V
 yoygkdd3xZIXSwAUuRTEHq7AbTyHeHb2aTV6cPFSNzTckjAQV7v6Nk6vbxeO2uLM
 d620V6cOI2QJGcib0nEkk1QMvyCJ/gipYRDvXwakyZTsga/VztpOG+EdXbTcuDUk
 ud8lY90LgiVc/JozMT+bCUmbYsorfe/Xo52twLwb+irtHJ/w5h8sQbq0M7jd1B7k
 hM/KebGlg9vGHH/G4UdfQx/3Td1GnNhNBDLVZLvqmie9FkA3c0S4SQwoKLXMbHfu
 By0bnvmJaQFs64Qr2AvYGmyaUniu9qnpXDGNMBweASAzJA04tg0=
 =h3Pg
 -----END PGP SIGNATURE-----

Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost

Pull virtio fixes from Michael Tsirkin:
 "Just a bunch of bugfixes all over the place"

* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost: (26 commits)
  virtio-mem: check if the config changed before fake offlining memory
  virtio-mem: keep retrying on offline_and_remove_memory() errors in Sub Block Mode (SBM)
  virtio-mem: convert most offline_and_remove_memory() errors to -EBUSY
  virtio-mem: remove unsafe unplug in Big Block Mode (BBM)
  pds_vdpa: fix up debugfs feature bit printing
  pds_vdpa: alloc irq vectors on DRIVER_OK
  pds_vdpa: clean and reset vqs entries
  pds_vdpa: always allow offering VIRTIO_NET_F_MAC
  pds_vdpa: reset to vdpa specified mac
  virtio-net: Zero max_tx_vq field for VIRTIO_NET_CTRL_MQ_HASH_CONFIG case
  vdpa/mlx5: Fix crash on shutdown for when no ndev exists
  vdpa/mlx5: Delete control vq iotlb in destroy_mr only when necessary
  vdpa/mlx5: Fix mr->initialized semantics
  vdpa/mlx5: Correct default number of queues when MQ is on
  virtio-vdpa: Fix cpumask memory leak in virtio_vdpa_find_vqs()
  vduse: Use proper spinlock for IRQ injection
  vdpa: Enable strict validation for netlinks ops
  vdpa: Add max vqp attr to vdpa_nl_policy for nlattr length check
  vdpa: Add queue index attr to vdpa_nl_policy for nlattr length check
  vdpa: Add features attr to vdpa_nl_policy for nlattr length check
  ...

MAINTAINERS

@@ -22474,7 +22474,6 @@ L:	virtualization@lists.linux-foundation.org
 S:	Maintained
 F:	drivers/block/virtio_blk.c
 F:	drivers/scsi/virtio_scsi.c
-F:	drivers/vhost/scsi.c
 F:	include/uapi/linux/virtio_blk.h
 F:	include/uapi/linux/virtio_scsi.h
 
@@ -22573,6 +22572,16 @@ F:	include/linux/vhost_iotlb.h
 F:	include/uapi/linux/vhost.h
 F:	kernel/vhost_task.c
 
+VIRTIO HOST (VHOST-SCSI)
+M:	"Michael S. Tsirkin" <mst@redhat.com>
+M:	Jason Wang <jasowang@redhat.com>
+M:	Mike Christie <michael.christie@oracle.com>
+R:	Paolo Bonzini <pbonzini@redhat.com>
+R:	Stefan Hajnoczi <stefanha@redhat.com>
+L:	virtualization@lists.linux-foundation.org
+S:	Maintained
+F:	drivers/vhost/scsi.c
+
 VIRTIO I2C DRIVER
 M:	Conghui Chen <conghui.chen@intel.com>
 M:	Viresh Kumar <viresh.kumar@linaro.org>

drivers/net/virtio_net.c

@@ -2761,7 +2761,7 @@ static void virtnet_init_default_rss(struct virtnet_info *vi)
 		vi->ctrl->rss.indirection_table[i] = indir_val;
 	}
 
-	vi->ctrl->rss.max_tx_vq = vi->curr_queue_pairs;
+	vi->ctrl->rss.max_tx_vq = vi->has_rss ? vi->curr_queue_pairs : 0;
 	vi->ctrl->rss.hash_key_length = vi->rss_key_size;
 
 	netdev_rss_key_fill(vi->ctrl->rss.key, vi->rss_key_size);

drivers/vdpa/mlx5/core/mlx5_vdpa.h

@@ -31,6 +31,7 @@ struct mlx5_vdpa_mr {
 	struct list_head head;
 	unsigned long num_directs;
 	unsigned long num_klms;
+	/* state of dvq mr */
 	bool initialized;
 
 	/* serialize mkey creation and destruction */
@@ -121,6 +122,7 @@ int mlx5_vdpa_handle_set_map(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *io
 int mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *iotlb,
 			unsigned int asid);
 void mlx5_vdpa_destroy_mr(struct mlx5_vdpa_dev *mvdev);
+void mlx5_vdpa_destroy_mr_asid(struct mlx5_vdpa_dev *mvdev, unsigned int asid);
 
 #define mlx5_vdpa_warn(__dev, format, ...)                                                         \
 	dev_warn((__dev)->mdev->device, "%s:%d:(pid %d) warning: " format, __func__, __LINE__,     \

drivers/vdpa/mlx5/core/mr.c

@@ -489,60 +489,103 @@ static void destroy_user_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr *mr
 	}
 }
 
-void mlx5_vdpa_destroy_mr(struct mlx5_vdpa_dev *mvdev)
+static void _mlx5_vdpa_destroy_cvq_mr(struct mlx5_vdpa_dev *mvdev, unsigned int asid)
+{
+	if (mvdev->group2asid[MLX5_VDPA_CVQ_GROUP] != asid)
+		return;
+
+	prune_iotlb(mvdev);
+}
+
+static void _mlx5_vdpa_destroy_dvq_mr(struct mlx5_vdpa_dev *mvdev, unsigned int asid)
 {
 	struct mlx5_vdpa_mr *mr = &mvdev->mr;
 
-	mutex_lock(&mr->mkey_mtx);
+	if (mvdev->group2asid[MLX5_VDPA_DATAVQ_GROUP] != asid)
-	if (!mr->initialized)
+		return;
-		goto out;
+
+	if (!mr->initialized)
+		return;
 
-	prune_iotlb(mvdev);
 	if (mr->user_mr)
 		destroy_user_mr(mvdev, mr);
 	else
 		destroy_dma_mr(mvdev, mr);
 
 	mr->initialized = false;
-out:
+}
+
+void mlx5_vdpa_destroy_mr_asid(struct mlx5_vdpa_dev *mvdev, unsigned int asid)
+{
+	struct mlx5_vdpa_mr *mr = &mvdev->mr;
+
+	mutex_lock(&mr->mkey_mtx);
+
+	_mlx5_vdpa_destroy_dvq_mr(mvdev, asid);
+	_mlx5_vdpa_destroy_cvq_mr(mvdev, asid);
+
 	mutex_unlock(&mr->mkey_mtx);
 }
 
+void mlx5_vdpa_destroy_mr(struct mlx5_vdpa_dev *mvdev)
+{
+	mlx5_vdpa_destroy_mr_asid(mvdev, mvdev->group2asid[MLX5_VDPA_CVQ_GROUP]);
+	mlx5_vdpa_destroy_mr_asid(mvdev, mvdev->group2asid[MLX5_VDPA_DATAVQ_GROUP]);
+}
+
+static int _mlx5_vdpa_create_cvq_mr(struct mlx5_vdpa_dev *mvdev,
+				    struct vhost_iotlb *iotlb,
+				    unsigned int asid)
+{
+	if (mvdev->group2asid[MLX5_VDPA_CVQ_GROUP] != asid)
+		return 0;
+
+	return dup_iotlb(mvdev, iotlb);
+}
+
+static int _mlx5_vdpa_create_dvq_mr(struct mlx5_vdpa_dev *mvdev,
+				    struct vhost_iotlb *iotlb,
+				    unsigned int asid)
+{
+	struct mlx5_vdpa_mr *mr = &mvdev->mr;
+	int err;
+
+	if (mvdev->group2asid[MLX5_VDPA_DATAVQ_GROUP] != asid)
+		return 0;
+
+	if (mr->initialized)
+		return 0;
+
+	if (iotlb)
+		err = create_user_mr(mvdev, iotlb);
+	else
+		err = create_dma_mr(mvdev, mr);
+
+	if (err)
+		return err;
+
+	mr->initialized = true;
+
+	return 0;
+}
+
 static int _mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev,
 				struct vhost_iotlb *iotlb, unsigned int asid)
 {
-	struct mlx5_vdpa_mr *mr = &mvdev->mr;
 	int err;
 
-	if (mr->initialized)
+	err = _mlx5_vdpa_create_dvq_mr(mvdev, iotlb, asid);
-		return 0;
+	if (err)
+		return err;
 
-	if (mvdev->group2asid[MLX5_VDPA_DATAVQ_GROUP] == asid) {
+	err = _mlx5_vdpa_create_cvq_mr(mvdev, iotlb, asid);
-		if (iotlb)
+	if (err)
-			err = create_user_mr(mvdev, iotlb);
+		goto out_err;
-		else
-			err = create_dma_mr(mvdev, mr);
 
-		if (err)
-			return err;
-	}
-
-	if (mvdev->group2asid[MLX5_VDPA_CVQ_GROUP] == asid) {
-		err = dup_iotlb(mvdev, iotlb);
-		if (err)
-			goto out_err;
-	}
-
-	mr->initialized = true;
 	return 0;
 
 out_err:
-	if (mvdev->group2asid[MLX5_VDPA_DATAVQ_GROUP] == asid) {
+	_mlx5_vdpa_destroy_dvq_mr(mvdev, asid);
-		if (iotlb)
-			destroy_user_mr(mvdev, mr);
-		else
-			destroy_dma_mr(mvdev, mr);
-	}
 
 	return err;
 }

drivers/vdpa/mlx5/net/mlx5_vnet.c

@@ -2517,7 +2517,15 @@ static int mlx5_vdpa_set_driver_features(struct vdpa_device *vdev, u64 features)
 	else
 		ndev->rqt_size = 1;
 
-	ndev->cur_num_vqs = 2 * ndev->rqt_size;
+	/* Device must start with 1 queue pair, as per VIRTIO v1.2 spec, section
+	 * 5.1.6.5.5 "Device operation in multiqueue mode":
+	 *
+	 * Multiqueue is disabled by default.
+	 * The driver enables multiqueue by sending a command using class
+	 * VIRTIO_NET_CTRL_MQ. The command selects the mode of multiqueue
+	 * operation, as follows: ...
+	 */
+	ndev->cur_num_vqs = 2;
 
 	update_cvq_info(mvdev);
 	return err;
@@ -2636,7 +2644,7 @@ static int mlx5_vdpa_change_map(struct mlx5_vdpa_dev *mvdev,
 		goto err_mr;
 
 	teardown_driver(ndev);
-	mlx5_vdpa_destroy_mr(mvdev);
+	mlx5_vdpa_destroy_mr_asid(mvdev, asid);
 	err = mlx5_vdpa_create_mr(mvdev, iotlb, asid);
 	if (err)
 		goto err_mr;
@@ -2652,7 +2660,7 @@ static int mlx5_vdpa_change_map(struct mlx5_vdpa_dev *mvdev,
 	return 0;
 
 err_setup:
-	mlx5_vdpa_destroy_mr(mvdev);
+	mlx5_vdpa_destroy_mr_asid(mvdev, asid);
 err_mr:
 	return err;
 }
@@ -3548,17 +3556,6 @@ static void mlx5v_remove(struct auxiliary_device *adev)
 	kfree(mgtdev);
 }
 
-static void mlx5v_shutdown(struct auxiliary_device *auxdev)
-{
-	struct mlx5_vdpa_mgmtdev *mgtdev;
-	struct mlx5_vdpa_net *ndev;
-
-	mgtdev = auxiliary_get_drvdata(auxdev);
-	ndev = mgtdev->ndev;
-
-	free_irqs(ndev);
-}
-
 static const struct auxiliary_device_id mlx5v_id_table[] = {
 	{ .name = MLX5_ADEV_NAME ".vnet", },
 	{},
@@ -3570,7 +3567,6 @@ static struct auxiliary_driver mlx5v_driver = {
 	.name = "vnet",
 	.probe = mlx5v_probe,
 	.remove = mlx5v_remove,
-	.shutdown = mlx5v_shutdown,
 	.id_table = mlx5v_id_table,
 };
 

drivers/vdpa/pds/Makefile

@@ -5,6 +5,5 @@ obj-$(CONFIG_PDS_VDPA) := pds_vdpa.o
 
 pds_vdpa-y := aux_drv.o \
 	      cmds.o \
+	      debugfs.o \
 	      vdpa_dev.o
-
-pds_vdpa-$(CONFIG_DEBUG_FS) += debugfs.o

drivers/vdpa/pds/debugfs.c

@@ -176,6 +176,7 @@ static int identity_show(struct seq_file *seq, void *v)
 {
 	struct pds_vdpa_aux *vdpa_aux = seq->private;
 	struct vdpa_mgmt_dev *mgmt;
+	u64 hw_features;
 
 	seq_printf(seq, "aux_dev:            %s\n",
 		   dev_name(&vdpa_aux->padev->aux_dev.dev));
@@ -183,8 +184,9 @@ static int identity_show(struct seq_file *seq, void *v)
 	mgmt = &vdpa_aux->vdpa_mdev;
 	seq_printf(seq, "max_vqs:            %d\n", mgmt->max_supported_vqs);
 	seq_printf(seq, "config_attr_mask:   %#llx\n", mgmt->config_attr_mask);
-	seq_printf(seq, "supported_features: %#llx\n", mgmt->supported_features);
+	hw_features = le64_to_cpu(vdpa_aux->ident.hw_features);
-	print_feature_bits_all(seq, mgmt->supported_features);
+	seq_printf(seq, "hw_features:        %#llx\n", hw_features);
+	print_feature_bits_all(seq, hw_features);
 
 	return 0;
 }
@@ -200,7 +202,6 @@ static int config_show(struct seq_file *seq, void *v)
 {
 	struct pds_vdpa_device *pdsv = seq->private;
 	struct virtio_net_config vc;
-	u64 driver_features;
 	u8 status;
 
 	memcpy_fromio(&vc, pdsv->vdpa_aux->vd_mdev.device,
@@ -223,12 +224,8 @@ static int config_show(struct seq_file *seq, void *v)
 	status = vp_modern_get_status(&pdsv->vdpa_aux->vd_mdev);
 	seq_printf(seq, "dev_status:           %#x\n", status);
 	print_status_bits(seq, status);
-
+	seq_printf(seq, "negotiated_features:  %#llx\n", pdsv->negotiated_features);
-	seq_printf(seq, "req_features:         %#llx\n", pdsv->req_features);
+	print_feature_bits_all(seq, pdsv->negotiated_features);
-	print_feature_bits_all(seq, pdsv->req_features);
-	driver_features = vp_modern_get_driver_features(&pdsv->vdpa_aux->vd_mdev);
-	seq_printf(seq, "driver_features:      %#llx\n", driver_features);
-	print_feature_bits_all(seq, driver_features);
 	seq_printf(seq, "vdpa_index:           %d\n", pdsv->vdpa_index);
 	seq_printf(seq, "num_vqs:              %d\n", pdsv->num_vqs);
 

drivers/vdpa/pds/vdpa_dev.c

@@ -126,11 +126,9 @@ static void pds_vdpa_release_irq(struct pds_vdpa_device *pdsv, int qid)
 static void pds_vdpa_set_vq_ready(struct vdpa_device *vdpa_dev, u16 qid, bool ready)
 {
 	struct pds_vdpa_device *pdsv = vdpa_to_pdsv(vdpa_dev);
-	struct pci_dev *pdev = pdsv->vdpa_aux->padev->vf_pdev;
 	struct device *dev = &pdsv->vdpa_dev.dev;
 	u64 driver_features;
 	u16 invert_idx = 0;
-	int irq;
 	int err;
 
 	dev_dbg(dev, "%s: qid %d ready %d => %d\n",
@@ -143,19 +141,6 @@ static void pds_vdpa_set_vq_ready(struct vdpa_device *vdpa_dev, u16 qid, bool re
 		invert_idx = PDS_VDPA_PACKED_INVERT_IDX;
 
 	if (ready) {
-		irq = pci_irq_vector(pdev, qid);
-		snprintf(pdsv->vqs[qid].irq_name, sizeof(pdsv->vqs[qid].irq_name),
-			 "vdpa-%s-%d", dev_name(dev), qid);
-
-		err = request_irq(irq, pds_vdpa_isr, 0,
-				  pdsv->vqs[qid].irq_name, &pdsv->vqs[qid]);
-		if (err) {
-			dev_err(dev, "%s: no irq for qid %d: %pe\n",
-				__func__, qid, ERR_PTR(err));
-			return;
-		}
-		pdsv->vqs[qid].irq = irq;
-
 		/* Pass vq setup info to DSC using adminq to gather up and
 		 * send all info at once so FW can do its full set up in
 		 * one easy operation
@@ -164,7 +149,6 @@ static void pds_vdpa_set_vq_ready(struct vdpa_device *vdpa_dev, u16 qid, bool re
 		if (err) {
 			dev_err(dev, "Failed to init vq %d: %pe\n",
 				qid, ERR_PTR(err));
-			pds_vdpa_release_irq(pdsv, qid);
 			ready = false;
 		}
 	} else {
@@ -172,7 +156,6 @@ static void pds_vdpa_set_vq_ready(struct vdpa_device *vdpa_dev, u16 qid, bool re
 		if (err)
 			dev_err(dev, "%s: reset_vq failed qid %d: %pe\n",
 				__func__, qid, ERR_PTR(err));
-		pds_vdpa_release_irq(pdsv, qid);
 	}
 
 	pdsv->vqs[qid].ready = ready;
@@ -318,6 +301,7 @@ static int pds_vdpa_set_driver_features(struct vdpa_device *vdpa_dev, u64 featur
 	struct device *dev = &pdsv->vdpa_dev.dev;
 	u64 driver_features;
 	u64 nego_features;
+	u64 hw_features;
 	u64 missing;
 
 	if (!(features & BIT_ULL(VIRTIO_F_ACCESS_PLATFORM)) && features) {
@@ -325,21 +309,26 @@ static int pds_vdpa_set_driver_features(struct vdpa_device *vdpa_dev, u64 featur
 		return -EOPNOTSUPP;
 	}
 
-	pdsv->req_features = features;
-
 	/* Check for valid feature bits */
-	nego_features = features & le64_to_cpu(pdsv->vdpa_aux->ident.hw_features);
+	nego_features = features & pdsv->supported_features;
-	missing = pdsv->req_features & ~nego_features;
+	missing = features & ~nego_features;
 	if (missing) {
 		dev_err(dev, "Can't support all requested features in %#llx, missing %#llx features\n",
-			pdsv->req_features, missing);
+			features, missing);
 		return -EOPNOTSUPP;
 	}
 
+	pdsv->negotiated_features = nego_features;
+
 	driver_features = pds_vdpa_get_driver_features(vdpa_dev);
 	dev_dbg(dev, "%s: %#llx => %#llx\n",
 		__func__, driver_features, nego_features);
 
+	/* if we're faking the F_MAC, strip it before writing to device */
+	hw_features = le64_to_cpu(pdsv->vdpa_aux->ident.hw_features);
+	if (!(hw_features & BIT_ULL(VIRTIO_NET_F_MAC)))
+		nego_features &= ~BIT_ULL(VIRTIO_NET_F_MAC);
+
 	if (driver_features == nego_features)
 		return 0;
 
@@ -352,7 +341,7 @@ static u64 pds_vdpa_get_driver_features(struct vdpa_device *vdpa_dev)
 {
 	struct pds_vdpa_device *pdsv = vdpa_to_pdsv(vdpa_dev);
 
-	return vp_modern_get_driver_features(&pdsv->vdpa_aux->vd_mdev);
+	return pdsv->negotiated_features;
 }
 
 static void pds_vdpa_set_config_cb(struct vdpa_device *vdpa_dev,
@@ -389,6 +378,72 @@ static u8 pds_vdpa_get_status(struct vdpa_device *vdpa_dev)
 	return vp_modern_get_status(&pdsv->vdpa_aux->vd_mdev);
 }
 
+static int pds_vdpa_request_irqs(struct pds_vdpa_device *pdsv)
+{
+	struct pci_dev *pdev = pdsv->vdpa_aux->padev->vf_pdev;
+	struct pds_vdpa_aux *vdpa_aux = pdsv->vdpa_aux;
+	struct device *dev = &pdsv->vdpa_dev.dev;
+	int max_vq, nintrs, qid, err;
+
+	max_vq = vdpa_aux->vdpa_mdev.max_supported_vqs;
+
+	nintrs = pci_alloc_irq_vectors(pdev, max_vq, max_vq, PCI_IRQ_MSIX);
+	if (nintrs < 0) {
+		dev_err(dev, "Couldn't get %d msix vectors: %pe\n",
+			max_vq, ERR_PTR(nintrs));
+		return nintrs;
+	}
+
+	for (qid = 0; qid < pdsv->num_vqs; ++qid) {
+		int irq = pci_irq_vector(pdev, qid);
+
+		snprintf(pdsv->vqs[qid].irq_name, sizeof(pdsv->vqs[qid].irq_name),
+			 "vdpa-%s-%d", dev_name(dev), qid);
+
+		err = request_irq(irq, pds_vdpa_isr, 0,
+				  pdsv->vqs[qid].irq_name,
+				  &pdsv->vqs[qid]);
+		if (err) {
+			dev_err(dev, "%s: no irq for qid %d: %pe\n",
+				__func__, qid, ERR_PTR(err));
+			goto err_release;
+		}
+
+		pdsv->vqs[qid].irq = irq;
+	}
+
+	vdpa_aux->nintrs = nintrs;
+
+	return 0;
+
+err_release:
+	while (qid--)
+		pds_vdpa_release_irq(pdsv, qid);
+
+	pci_free_irq_vectors(pdev);
+
+	vdpa_aux->nintrs = 0;
+
+	return err;
+}
+
+static void pds_vdpa_release_irqs(struct pds_vdpa_device *pdsv)
+{
+	struct pci_dev *pdev = pdsv->vdpa_aux->padev->vf_pdev;
+	struct pds_vdpa_aux *vdpa_aux = pdsv->vdpa_aux;
+	int qid;
+
+	if (!vdpa_aux->nintrs)
+		return;
+
+	for (qid = 0; qid < pdsv->num_vqs; qid++)
+		pds_vdpa_release_irq(pdsv, qid);
+
+	pci_free_irq_vectors(pdev);
+
+	vdpa_aux->nintrs = 0;
+}
+
 static void pds_vdpa_set_status(struct vdpa_device *vdpa_dev, u8 status)
 {
 	struct pds_vdpa_device *pdsv = vdpa_to_pdsv(vdpa_dev);
@@ -399,6 +454,11 @@ static void pds_vdpa_set_status(struct vdpa_device *vdpa_dev, u8 status)
 	old_status = pds_vdpa_get_status(vdpa_dev);
 	dev_dbg(dev, "%s: old %#x new %#x\n", __func__, old_status, status);
 
+	if (status & ~old_status & VIRTIO_CONFIG_S_DRIVER_OK) {
+		if (pds_vdpa_request_irqs(pdsv))
+			status = old_status | VIRTIO_CONFIG_S_FAILED;
+	}
+
 	pds_vdpa_cmd_set_status(pdsv, status);
 
 	/* Note: still working with FW on the need for this reset cmd */
@@ -409,6 +469,8 @@ static void pds_vdpa_set_status(struct vdpa_device *vdpa_dev, u8 status)
 			pdsv->vqs[i].avail_idx = 0;
 			pdsv->vqs[i].used_idx = 0;
 		}
+
+		pds_vdpa_cmd_set_mac(pdsv, pdsv->mac);
 	}
 
 	if (status & ~old_status & VIRTIO_CONFIG_S_FEATURES_OK) {
@@ -418,6 +480,20 @@ static void pds_vdpa_set_status(struct vdpa_device *vdpa_dev, u8 status)
 							i, &pdsv->vqs[i].notify_pa);
 		}
 	}
+
+	if (old_status & ~status & VIRTIO_CONFIG_S_DRIVER_OK)
+		pds_vdpa_release_irqs(pdsv);
+}
+
+static void pds_vdpa_init_vqs_entry(struct pds_vdpa_device *pdsv, int qid,
+				    void __iomem *notify)
+{
+	memset(&pdsv->vqs[qid], 0, sizeof(pdsv->vqs[0]));
+	pdsv->vqs[qid].qid = qid;
+	pdsv->vqs[qid].pdsv = pdsv;
+	pdsv->vqs[qid].ready = false;
+	pdsv->vqs[qid].irq = VIRTIO_MSI_NO_VECTOR;
+	pdsv->vqs[qid].notify = notify;
 }
 
 static int pds_vdpa_reset(struct vdpa_device *vdpa_dev)
@@ -441,14 +517,17 @@ static int pds_vdpa_reset(struct vdpa_device *vdpa_dev)
 			if (err)
 				dev_err(dev, "%s: reset_vq failed qid %d: %pe\n",
 					__func__, i, ERR_PTR(err));
-			pds_vdpa_release_irq(pdsv, i);
-			memset(&pdsv->vqs[i], 0, sizeof(pdsv->vqs[0]));
-			pdsv->vqs[i].ready = false;
 		}
 	}
 
 	pds_vdpa_set_status(vdpa_dev, 0);
 
+	if (status & VIRTIO_CONFIG_S_DRIVER_OK) {
+		/* Reset the vq info */
+		for (i = 0; i < pdsv->num_vqs && !err; i++)
+			pds_vdpa_init_vqs_entry(pdsv, i, pdsv->vqs[i].notify);
+	}
+
 	return 0;
 }
 
@@ -532,7 +611,6 @@ static int pds_vdpa_dev_add(struct vdpa_mgmt_dev *mdev, const char *name,
 	struct device *dma_dev;
 	struct pci_dev *pdev;
 	struct device *dev;
-	u8 mac[ETH_ALEN];
 	int err;
 	int i;
 
@@ -563,7 +641,7 @@ static int pds_vdpa_dev_add(struct vdpa_mgmt_dev *mdev, const char *name,
 
 	if (add_config->mask & BIT_ULL(VDPA_ATTR_DEV_FEATURES)) {
 		u64 unsupp_features =
-			add_config->device_features & ~mgmt->supported_features;
+			add_config->device_features & ~pdsv->supported_features;
 
 		if (unsupp_features) {
 			dev_err(dev, "Unsupported features: %#llx\n", unsupp_features);
@@ -614,29 +692,30 @@ static int pds_vdpa_dev_add(struct vdpa_mgmt_dev *mdev, const char *name,
 	}
 
 	/* Set a mac, either from the user config if provided
-	 * or set a random mac if default is 00:..:00
+	 * or use the device's mac if not 00:..:00
+	 * or set a random mac
 	 */
 	if (add_config->mask & BIT_ULL(VDPA_ATTR_DEV_NET_CFG_MACADDR)) {
-		ether_addr_copy(mac, add_config->net.mac);
+		ether_addr_copy(pdsv->mac, add_config->net.mac);
-		pds_vdpa_cmd_set_mac(pdsv, mac);
 	} else {
 		struct virtio_net_config __iomem *vc;
 
 		vc = pdsv->vdpa_aux->vd_mdev.device;
-		memcpy_fromio(mac, vc->mac, sizeof(mac));
+		memcpy_fromio(pdsv->mac, vc->mac, sizeof(pdsv->mac));
-		if (is_zero_ether_addr(mac)) {
+		if (is_zero_ether_addr(pdsv->mac) &&
-			eth_random_addr(mac);
+		    (pdsv->supported_features & BIT_ULL(VIRTIO_NET_F_MAC))) {
-			dev_info(dev, "setting random mac %pM\n", mac);
+			eth_random_addr(pdsv->mac);
-			pds_vdpa_cmd_set_mac(pdsv, mac);
+			dev_info(dev, "setting random mac %pM\n", pdsv->mac);
 		}
 	}
+	pds_vdpa_cmd_set_mac(pdsv, pdsv->mac);
 
 	for (i = 0; i < pdsv->num_vqs; i++) {
-		pdsv->vqs[i].qid = i;
+		void __iomem *notify;
-		pdsv->vqs[i].pdsv = pdsv;
+
-		pdsv->vqs[i].irq = VIRTIO_MSI_NO_VECTOR;
+		notify = vp_modern_map_vq_notify(&pdsv->vdpa_aux->vd_mdev,
-		pdsv->vqs[i].notify = vp_modern_map_vq_notify(&pdsv->vdpa_aux->vd_mdev,
+						 i, &pdsv->vqs[i].notify_pa);
-							      i, &pdsv->vqs[i].notify_pa);
+		pds_vdpa_init_vqs_entry(pdsv, i, notify);
 	}
 
 	pdsv->vdpa_dev.mdev = &vdpa_aux->vdpa_mdev;
@@ -746,24 +825,19 @@ int pds_vdpa_get_mgmt_info(struct pds_vdpa_aux *vdpa_aux)
 
 	max_vqs = min_t(u16, dev_intrs, max_vqs);
 	mgmt->max_supported_vqs = min_t(u16, PDS_VDPA_MAX_QUEUES, max_vqs);
-	vdpa_aux->nintrs = mgmt->max_supported_vqs;
+	vdpa_aux->nintrs = 0;
 
 	mgmt->ops = &pds_vdpa_mgmt_dev_ops;
 	mgmt->id_table = pds_vdpa_id_table;
 	mgmt->device = dev;
 	mgmt->supported_features = le64_to_cpu(vdpa_aux->ident.hw_features);
+
+	/* advertise F_MAC even if the device doesn't */
+	mgmt->supported_features |= BIT_ULL(VIRTIO_NET_F_MAC);
+
 	mgmt->config_attr_mask = BIT_ULL(VDPA_ATTR_DEV_NET_CFG_MACADDR);
 	mgmt->config_attr_mask |= BIT_ULL(VDPA_ATTR_DEV_NET_CFG_MAX_VQP);
 	mgmt->config_attr_mask |= BIT_ULL(VDPA_ATTR_DEV_FEATURES);
 
-	err = pci_alloc_irq_vectors(pdev, vdpa_aux->nintrs, vdpa_aux->nintrs,
-				    PCI_IRQ_MSIX);
-	if (err < 0) {
-		dev_err(dev, "Couldn't get %d msix vectors: %pe\n",
-			vdpa_aux->nintrs, ERR_PTR(err));
-		return err;
-	}
-	vdpa_aux->nintrs = err;
-
 	return 0;
 }

drivers/vdpa/pds/vdpa_dev.h

@@ -35,10 +35,11 @@ struct pds_vdpa_device {
 	struct pds_vdpa_aux *vdpa_aux;
 
 	struct pds_vdpa_vq_info vqs[PDS_VDPA_MAX_QUEUES];
-	u64 supported_features;		/* specified device features */
+	u64 supported_features;		/* supported device features */
-	u64 req_features;		/* features requested by vdpa */
+	u64 negotiated_features;	/* negotiated features */
 	u8 vdpa_index;			/* rsvd for future subdevice use */
 	u8 num_vqs;			/* num vqs in use */
+	u8 mac[ETH_ALEN];		/* mac selected when the device was added */
 	struct vdpa_callback config_cb;
 	struct notifier_block nb;
 };

drivers/vdpa/vdpa.c

@@ -1247,44 +1247,41 @@ static const struct nla_policy vdpa_nl_policy[VDPA_ATTR_MAX + 1] = {
 	[VDPA_ATTR_MGMTDEV_DEV_NAME] = { .type = NLA_STRING },
 	[VDPA_ATTR_DEV_NAME] = { .type = NLA_STRING },
 	[VDPA_ATTR_DEV_NET_CFG_MACADDR] = NLA_POLICY_ETH_ADDR,
+	[VDPA_ATTR_DEV_NET_CFG_MAX_VQP] = { .type = NLA_U16 },
 	/* virtio spec 1.1 section 5.1.4.1 for valid MTU range */
 	[VDPA_ATTR_DEV_NET_CFG_MTU] = NLA_POLICY_MIN(NLA_U16, 68),
+	[VDPA_ATTR_DEV_QUEUE_INDEX] = { .type = NLA_U32 },
+	[VDPA_ATTR_DEV_FEATURES] = { .type = NLA_U64 },
 };
 
 static const struct genl_ops vdpa_nl_ops[] = {
 	{
 		.cmd = VDPA_CMD_MGMTDEV_GET,
-		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 		.doit = vdpa_nl_cmd_mgmtdev_get_doit,
 		.dumpit = vdpa_nl_cmd_mgmtdev_get_dumpit,
 	},
 	{
 		.cmd = VDPA_CMD_DEV_NEW,
-		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 		.doit = vdpa_nl_cmd_dev_add_set_doit,
 		.flags = GENL_ADMIN_PERM,
 	},
 	{
 		.cmd = VDPA_CMD_DEV_DEL,
-		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 		.doit = vdpa_nl_cmd_dev_del_set_doit,
 		.flags = GENL_ADMIN_PERM,
 	},
 	{
 		.cmd = VDPA_CMD_DEV_GET,
-		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 		.doit = vdpa_nl_cmd_dev_get_doit,
 		.dumpit = vdpa_nl_cmd_dev_get_dumpit,
 	},
 	{
 		.cmd = VDPA_CMD_DEV_CONFIG_GET,
-		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 		.doit = vdpa_nl_cmd_dev_config_get_doit,
 		.dumpit = vdpa_nl_cmd_dev_config_get_dumpit,
 	},
 	{
 		.cmd = VDPA_CMD_DEV_VSTATS_GET,
-		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 		.doit = vdpa_nl_cmd_dev_stats_get_doit,
 		.flags = GENL_ADMIN_PERM,
 	},

drivers/vdpa/vdpa_user/vduse_dev.c

@@ -935,10 +935,10 @@ static void vduse_dev_irq_inject(struct work_struct *work)
 {
 	struct vduse_dev *dev = container_of(work, struct vduse_dev, inject);
 
-	spin_lock_irq(&dev->irq_lock);
+	spin_lock_bh(&dev->irq_lock);
 	if (dev->config_cb.callback)
 		dev->config_cb.callback(dev->config_cb.private);
-	spin_unlock_irq(&dev->irq_lock);
+	spin_unlock_bh(&dev->irq_lock);
 }
 
 static void vduse_vq_irq_inject(struct work_struct *work)
@@ -946,10 +946,10 @@ static void vduse_vq_irq_inject(struct work_struct *work)
 	struct vduse_virtqueue *vq = container_of(work,
 					struct vduse_virtqueue, inject);
 
-	spin_lock_irq(&vq->irq_lock);
+	spin_lock_bh(&vq->irq_lock);
 	if (vq->ready && vq->cb.callback)
 		vq->cb.callback(vq->cb.private);
-	spin_unlock_irq(&vq->irq_lock);
+	spin_unlock_bh(&vq->irq_lock);
 }
 
 static bool vduse_vq_signal_irqfd(struct vduse_virtqueue *vq)

drivers/vhost/scsi.c

@@ -25,6 +25,8 @@
 #include <linux/fs.h>
 #include <linux/vmalloc.h>
 #include <linux/miscdevice.h>
+#include <linux/blk_types.h>
+#include <linux/bio.h>
 #include <asm/unaligned.h>
 #include <scsi/scsi_common.h>
 #include <scsi/scsi_proto.h>
@@ -75,6 +77,9 @@ struct vhost_scsi_cmd {
 	u32 tvc_prot_sgl_count;
 	/* Saved unpacked SCSI LUN for vhost_scsi_target_queue_cmd() */
 	u32 tvc_lun;
+	u32 copied_iov:1;
+	const void *saved_iter_addr;
+	struct iov_iter saved_iter;
 	/* Pointer to the SGL formatted memory from virtio-scsi */
 	struct scatterlist *tvc_sgl;
 	struct scatterlist *tvc_prot_sgl;
@@ -328,8 +333,13 @@ static void vhost_scsi_release_cmd_res(struct se_cmd *se_cmd)
 	int i;
 
 	if (tv_cmd->tvc_sgl_count) {
-		for (i = 0; i < tv_cmd->tvc_sgl_count; i++)
+		for (i = 0; i < tv_cmd->tvc_sgl_count; i++) {
-			put_page(sg_page(&tv_cmd->tvc_sgl[i]));
+			if (tv_cmd->copied_iov)
+				__free_page(sg_page(&tv_cmd->tvc_sgl[i]));
+			else
+				put_page(sg_page(&tv_cmd->tvc_sgl[i]));
+		}
+		kfree(tv_cmd->saved_iter_addr);
 	}
 	if (tv_cmd->tvc_prot_sgl_count) {
 		for (i = 0; i < tv_cmd->tvc_prot_sgl_count; i++)
@@ -504,6 +514,28 @@ static void vhost_scsi_evt_work(struct vhost_work *work)
 	mutex_unlock(&vq->mutex);
 }
 
+static int vhost_scsi_copy_sgl_to_iov(struct vhost_scsi_cmd *cmd)
+{
+	struct iov_iter *iter = &cmd->saved_iter;
+	struct scatterlist *sg = cmd->tvc_sgl;
+	struct page *page;
+	size_t len;
+	int i;
+
+	for (i = 0; i < cmd->tvc_sgl_count; i++) {
+		page = sg_page(&sg[i]);
+		len = sg[i].length;
+
+		if (copy_page_to_iter(page, 0, len, iter) != len) {
+			pr_err("Could not copy data while handling misaligned cmd. Error %zu\n",
+			       len);
+			return -1;
+		}
+	}
+
+	return 0;
+}
+
 /* Fill in status and signal that we are done processing this command
  *
  * This is scheduled in the vhost work queue so we are called with the owner
@@ -527,15 +559,20 @@ static void vhost_scsi_complete_cmd_work(struct vhost_work *work)
 
 		pr_debug("%s tv_cmd %p resid %u status %#02x\n", __func__,
 			cmd, se_cmd->residual_count, se_cmd->scsi_status);
-
 		memset(&v_rsp, 0, sizeof(v_rsp));
-		v_rsp.resid = cpu_to_vhost32(cmd->tvc_vq, se_cmd->residual_count);
+
-		/* TODO is status_qualifier field needed? */
+		if (cmd->saved_iter_addr && vhost_scsi_copy_sgl_to_iov(cmd)) {
-		v_rsp.status = se_cmd->scsi_status;
+			v_rsp.response = VIRTIO_SCSI_S_BAD_TARGET;
-		v_rsp.sense_len = cpu_to_vhost32(cmd->tvc_vq,
+		} else {
-						 se_cmd->scsi_sense_length);
+			v_rsp.resid = cpu_to_vhost32(cmd->tvc_vq,
-		memcpy(v_rsp.sense, cmd->tvc_sense_buf,
+						     se_cmd->residual_count);
-		       se_cmd->scsi_sense_length);
+			/* TODO is status_qualifier field needed? */
+			v_rsp.status = se_cmd->scsi_status;
+			v_rsp.sense_len = cpu_to_vhost32(cmd->tvc_vq,
+							 se_cmd->scsi_sense_length);
+			memcpy(v_rsp.sense, cmd->tvc_sense_buf,
+			       se_cmd->scsi_sense_length);
+		}
 
 		iov_iter_init(&iov_iter, ITER_DEST, cmd->tvc_resp_iov,
 			      cmd->tvc_in_iovs, sizeof(v_rsp));
@@ -613,12 +650,12 @@ static int
 vhost_scsi_map_to_sgl(struct vhost_scsi_cmd *cmd,
 		      struct iov_iter *iter,
 		      struct scatterlist *sgl,
-		      bool write)
+		      bool is_prot)
 {
 	struct page **pages = cmd->tvc_upages;
 	struct scatterlist *sg = sgl;
-	ssize_t bytes;
+	ssize_t bytes, mapped_bytes;
-	size_t offset;
+	size_t offset, mapped_offset;
 	unsigned int npages = 0;
 
 	bytes = iov_iter_get_pages2(iter, pages, LONG_MAX,
@@ -627,13 +664,53 @@ vhost_scsi_map_to_sgl(struct vhost_scsi_cmd *cmd,
 	if (bytes <= 0)
 		return bytes < 0 ? bytes : -EFAULT;
 
+	mapped_bytes = bytes;
+	mapped_offset = offset;
+
 	while (bytes) {
 		unsigned n = min_t(unsigned, PAGE_SIZE - offset, bytes);
+		/*
+		 * The block layer requires bios/requests to be a multiple of
+		 * 512 bytes, but Windows can send us vecs that are misaligned.
+		 * This can result in bios and later requests with misaligned
+		 * sizes if we have to break up a cmd/scatterlist into multiple
+		 * bios.
+		 *
+		 * We currently only break up a command into multiple bios if
+		 * we hit the vec/seg limit, so check if our sgl_count is
+		 * greater than the max and if a vec in the cmd has a
+		 * misaligned offset/size.
+		 */
+		if (!is_prot &&
+		    (offset & (SECTOR_SIZE - 1) || n & (SECTOR_SIZE - 1)) &&
+		    cmd->tvc_sgl_count > BIO_MAX_VECS) {
+			WARN_ONCE(true,
+				  "vhost-scsi detected misaligned IO. Performance may be degraded.");
+			goto revert_iter_get_pages;
+		}
+
 		sg_set_page(sg++, pages[npages++], n, offset);
 		bytes -= n;
 		offset = 0;
 	}
+
 	return npages;
+
+revert_iter_get_pages:
+	iov_iter_revert(iter, mapped_bytes);
+
+	npages = 0;
+	while (mapped_bytes) {
+		unsigned int n = min_t(unsigned int, PAGE_SIZE - mapped_offset,
+				       mapped_bytes);
+
+		put_page(pages[npages++]);
+
+		mapped_bytes -= n;
+		mapped_offset = 0;
+	}
+
+	return -EINVAL;
 }
 
 static int
@@ -657,25 +734,80 @@ vhost_scsi_calc_sgls(struct iov_iter *iter, size_t bytes, int max_sgls)
 }
 
 static int
-vhost_scsi_iov_to_sgl(struct vhost_scsi_cmd *cmd, bool write,
+vhost_scsi_copy_iov_to_sgl(struct vhost_scsi_cmd *cmd, struct iov_iter *iter,
-		      struct iov_iter *iter,
+			   struct scatterlist *sg, int sg_count)
-		      struct scatterlist *sg, int sg_count)
+{
+	size_t len = iov_iter_count(iter);
+	unsigned int nbytes = 0;
+	struct page *page;
+	int i;
+
+	if (cmd->tvc_data_direction == DMA_FROM_DEVICE) {
+		cmd->saved_iter_addr = dup_iter(&cmd->saved_iter, iter,
+						GFP_KERNEL);
+		if (!cmd->saved_iter_addr)
+			return -ENOMEM;
+	}
+
+	for (i = 0; i < sg_count; i++) {
+		page = alloc_page(GFP_KERNEL);
+		if (!page) {
+			i--;
+			goto err;
+		}
+
+		nbytes = min_t(unsigned int, PAGE_SIZE, len);
+		sg_set_page(&sg[i], page, nbytes, 0);
+
+		if (cmd->tvc_data_direction == DMA_TO_DEVICE &&
+		    copy_page_from_iter(page, 0, nbytes, iter) != nbytes)
+			goto err;
+
+		len -= nbytes;
+	}
+
+	cmd->copied_iov = 1;
+	return 0;
+
+err:
+	pr_err("Could not read %u bytes while handling misaligned cmd\n",
+	       nbytes);
+
+	for (; i >= 0; i--)
+		__free_page(sg_page(&sg[i]));
+	kfree(cmd->saved_iter_addr);
+	return -ENOMEM;
+}
+
+static int
+vhost_scsi_map_iov_to_sgl(struct vhost_scsi_cmd *cmd, struct iov_iter *iter,
+			  struct scatterlist *sg, int sg_count, bool is_prot)
 {
 	struct scatterlist *p = sg;
+	size_t revert_bytes;
 	int ret;
 
 	while (iov_iter_count(iter)) {
-		ret = vhost_scsi_map_to_sgl(cmd, iter, sg, write);
+		ret = vhost_scsi_map_to_sgl(cmd, iter, sg, is_prot);
 		if (ret < 0) {
+			revert_bytes = 0;
+
 			while (p < sg) {
-				struct page *page = sg_page(p++);
+				struct page *page = sg_page(p);
-				if (page)
+
+				if (page) {
 					put_page(page);
+					revert_bytes += p->length;
+				}
+				p++;
 			}
+
+			iov_iter_revert(iter, revert_bytes);
 			return ret;
 		}
 		sg += ret;
 	}
+
 	return 0;
 }
 
@@ -685,7 +817,6 @@ vhost_scsi_mapal(struct vhost_scsi_cmd *cmd,
 		 size_t data_bytes, struct iov_iter *data_iter)
 {
 	int sgl_count, ret;
-	bool write = (cmd->tvc_data_direction == DMA_FROM_DEVICE);
 
 	if (prot_bytes) {
 		sgl_count = vhost_scsi_calc_sgls(prot_iter, prot_bytes,
@@ -698,9 +829,9 @@ vhost_scsi_mapal(struct vhost_scsi_cmd *cmd,
 		pr_debug("%s prot_sg %p prot_sgl_count %u\n", __func__,
 			 cmd->tvc_prot_sgl, cmd->tvc_prot_sgl_count);
 
-		ret = vhost_scsi_iov_to_sgl(cmd, write, prot_iter,
+		ret = vhost_scsi_map_iov_to_sgl(cmd, prot_iter,
-					    cmd->tvc_prot_sgl,
+						cmd->tvc_prot_sgl,
-					    cmd->tvc_prot_sgl_count);
+						cmd->tvc_prot_sgl_count, true);
 		if (ret < 0) {
 			cmd->tvc_prot_sgl_count = 0;
 			return ret;
@@ -716,8 +847,14 @@ vhost_scsi_mapal(struct vhost_scsi_cmd *cmd,
 	pr_debug("%s data_sg %p data_sgl_count %u\n", __func__,
 		  cmd->tvc_sgl, cmd->tvc_sgl_count);
 
-	ret = vhost_scsi_iov_to_sgl(cmd, write, data_iter,
+	ret = vhost_scsi_map_iov_to_sgl(cmd, data_iter, cmd->tvc_sgl,
-				    cmd->tvc_sgl, cmd->tvc_sgl_count);
+					cmd->tvc_sgl_count, false);
+	if (ret == -EINVAL) {
+		sg_init_table(cmd->tvc_sgl, cmd->tvc_sgl_count);
+		ret = vhost_scsi_copy_iov_to_sgl(cmd, data_iter, cmd->tvc_sgl,
+						 cmd->tvc_sgl_count);
+	}
+
 	if (ret < 0) {
 		cmd->tvc_sgl_count = 0;
 		return ret;

drivers/virtio/virtio_mem.c

@@ -38,11 +38,6 @@ module_param(bbm_block_size, ulong, 0444);
 MODULE_PARM_DESC(bbm_block_size,
 		 "Big Block size in bytes. Default is 0 (auto-detection).");
 
-static bool bbm_safe_unplug = true;
-module_param(bbm_safe_unplug, bool, 0444);
-MODULE_PARM_DESC(bbm_safe_unplug,
-	     "Use a safe unplug mechanism in BBM, avoiding long/endless loops");
-
 /*
  * virtio-mem currently supports the following modes of operation:
  *
@@ -173,6 +168,13 @@ struct virtio_mem {
 			/* The number of subblocks per Linux memory block. */
 			uint32_t sbs_per_mb;
 
+			/*
+			 * Some of the Linux memory blocks tracked as "partially
+			 * plugged" are completely unplugged and can be offlined
+			 * and removed -- which previously failed.
+			 */
+			bool have_unplugged_mb;
+
 			/* Summary of all memory block states. */
 			unsigned long mb_count[VIRTIO_MEM_SBM_MB_COUNT];
 
@@ -746,11 +748,15 @@ static int virtio_mem_offline_and_remove_memory(struct virtio_mem *vm,
 		 * immediately instead of waiting.
 		 */
 		virtio_mem_retry(vm);
-	} else {
+		return 0;
-		dev_dbg(&vm->vdev->dev,
-			"offlining and removing memory failed: %d\n", rc);
 	}
-	return rc;
+	dev_dbg(&vm->vdev->dev, "offlining and removing memory failed: %d\n", rc);
+	/*
+	 * We don't really expect this to fail, because we fake-offlined all
+	 * memory already. But it could fail in corner cases.
+	 */
+	WARN_ON_ONCE(rc != -ENOMEM && rc != -EBUSY);
+	return rc == -ENOMEM ? -ENOMEM : -EBUSY;
 }
 
 /*
@@ -766,6 +772,34 @@ static int virtio_mem_sbm_offline_and_remove_mb(struct virtio_mem *vm,
 	return virtio_mem_offline_and_remove_memory(vm, addr, size);
 }
 
+/*
+ * Try (offlining and) removing memory from Linux in case all subblocks are
+ * unplugged. Can be called on online and offline memory blocks.
+ *
+ * May modify the state of memory blocks in virtio-mem.
+ */
+static int virtio_mem_sbm_try_remove_unplugged_mb(struct virtio_mem *vm,
+						  unsigned long mb_id)
+{
+	int rc;
+
+	/*
+	 * Once all subblocks of a memory block were unplugged, offline and
+	 * remove it.
+	 */
+	if (!virtio_mem_sbm_test_sb_unplugged(vm, mb_id, 0, vm->sbm.sbs_per_mb))
+		return 0;
+
+	/* offline_and_remove_memory() works for online and offline memory. */
+	mutex_unlock(&vm->hotplug_mutex);
+	rc = virtio_mem_sbm_offline_and_remove_mb(vm, mb_id);
+	mutex_lock(&vm->hotplug_mutex);
+	if (!rc)
+		virtio_mem_sbm_set_mb_state(vm, mb_id,
+					    VIRTIO_MEM_SBM_MB_UNUSED);
+	return rc;
+}
+
 /*
  * See virtio_mem_offline_and_remove_memory(): Try to offline and remove a
  * all Linux memory blocks covered by the big block.
@@ -1155,7 +1189,8 @@ static void virtio_mem_fake_online(unsigned long pfn, unsigned long nr_pages)
  * Try to allocate a range, marking pages fake-offline, effectively
  * fake-offlining them.
  */
-static int virtio_mem_fake_offline(unsigned long pfn, unsigned long nr_pages)
+static int virtio_mem_fake_offline(struct virtio_mem *vm, unsigned long pfn,
+				   unsigned long nr_pages)
 {
 	const bool is_movable = is_zone_movable_page(pfn_to_page(pfn));
 	int rc, retry_count;
@@ -1168,6 +1203,14 @@ static int virtio_mem_fake_offline(unsigned long pfn, unsigned long nr_pages)
 	 * some guarantees.
 	 */
 	for (retry_count = 0; retry_count < 5; retry_count++) {
+		/*
+		 * If the config changed, stop immediately and go back to the
+		 * main loop: avoid trying to keep unplugging if the device
+		 * might have decided to not remove any more memory.
+		 */
+		if (atomic_read(&vm->config_changed))
+			return -EAGAIN;
+
 		rc = alloc_contig_range(pfn, pfn + nr_pages, MIGRATE_MOVABLE,
 					GFP_KERNEL);
 		if (rc == -ENOMEM)
@@ -1917,7 +1960,7 @@ static int virtio_mem_sbm_unplug_sb_online(struct virtio_mem *vm,
 	start_pfn = PFN_DOWN(virtio_mem_mb_id_to_phys(mb_id) +
 			     sb_id * vm->sbm.sb_size);
 
-	rc = virtio_mem_fake_offline(start_pfn, nr_pages);
+	rc = virtio_mem_fake_offline(vm, start_pfn, nr_pages);
 	if (rc)
 		return rc;
 
@@ -1989,20 +2032,10 @@ static int virtio_mem_sbm_unplug_any_sb_online(struct virtio_mem *vm,
 	}
 
 unplugged:
-	/*
+	rc = virtio_mem_sbm_try_remove_unplugged_mb(vm, mb_id);
-	 * Once all subblocks of a memory block were unplugged, offline and
+	if (rc)
-	 * remove it. This will usually not fail, as no memory is in use
+		vm->sbm.have_unplugged_mb = 1;
-	 * anymore - however some other notifiers might NACK the request.
+	/* Ignore errors, this is not critical. We'll retry later. */
-	 */
-	if (virtio_mem_sbm_test_sb_unplugged(vm, mb_id, 0, vm->sbm.sbs_per_mb)) {
-		mutex_unlock(&vm->hotplug_mutex);
-		rc = virtio_mem_sbm_offline_and_remove_mb(vm, mb_id);
-		mutex_lock(&vm->hotplug_mutex);
-		if (!rc)
-			virtio_mem_sbm_set_mb_state(vm, mb_id,
-						    VIRTIO_MEM_SBM_MB_UNUSED);
-	}
-
 	return 0;
 }
 
@@ -2111,38 +2144,32 @@ static int virtio_mem_bbm_offline_remove_and_unplug_bb(struct virtio_mem *vm,
 			 VIRTIO_MEM_BBM_BB_ADDED))
 		return -EINVAL;
 
-	if (bbm_safe_unplug) {
+	/*
-		/*
+	 * Start by fake-offlining all memory. Once we marked the device
-		 * Start by fake-offlining all memory. Once we marked the device
+	 * block as fake-offline, all newly onlined memory will
-		 * block as fake-offline, all newly onlined memory will
+	 * automatically be kept fake-offline. Protect from concurrent
-		 * automatically be kept fake-offline. Protect from concurrent
+	 * onlining/offlining until we have a consistent state.
-		 * onlining/offlining until we have a consistent state.
+	 */
-		 */
+	mutex_lock(&vm->hotplug_mutex);
-		mutex_lock(&vm->hotplug_mutex);
+	virtio_mem_bbm_set_bb_state(vm, bb_id, VIRTIO_MEM_BBM_BB_FAKE_OFFLINE);
-		virtio_mem_bbm_set_bb_state(vm, bb_id,
-					    VIRTIO_MEM_BBM_BB_FAKE_OFFLINE);
 
-		for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
+	for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
-			page = pfn_to_online_page(pfn);
+		page = pfn_to_online_page(pfn);
-			if (!page)
+		if (!page)
-				continue;
+			continue;
 
-			rc = virtio_mem_fake_offline(pfn, PAGES_PER_SECTION);
+		rc = virtio_mem_fake_offline(vm, pfn, PAGES_PER_SECTION);
-			if (rc) {
+		if (rc) {
-				end_pfn = pfn;
+			end_pfn = pfn;
-				goto rollback_safe_unplug;
+			goto rollback;
-			}
 		}
-		mutex_unlock(&vm->hotplug_mutex);
 	}
+	mutex_unlock(&vm->hotplug_mutex);
 
 	rc = virtio_mem_bbm_offline_and_remove_bb(vm, bb_id);
 	if (rc) {
-		if (bbm_safe_unplug) {
+		mutex_lock(&vm->hotplug_mutex);
-			mutex_lock(&vm->hotplug_mutex);
+		goto rollback;
-			goto rollback_safe_unplug;
-		}
-		return rc;
 	}
 
 	rc = virtio_mem_bbm_unplug_bb(vm, bb_id);
@@ -2154,7 +2181,7 @@ static int virtio_mem_bbm_offline_remove_and_unplug_bb(struct virtio_mem *vm,
 					    VIRTIO_MEM_BBM_BB_UNUSED);
 	return rc;
 
-rollback_safe_unplug:
+rollback:
 	for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
 		page = pfn_to_online_page(pfn);
 		if (!page)
@@ -2260,12 +2287,13 @@ static int virtio_mem_unplug_request(struct virtio_mem *vm, uint64_t diff)
 
 /*
  * Try to unplug all blocks that couldn't be unplugged before, for example,
- * because the hypervisor was busy.
+ * because the hypervisor was busy. Further, offline and remove any memory
+ * blocks where we previously failed.
  */
-static int virtio_mem_unplug_pending_mb(struct virtio_mem *vm)
+static int virtio_mem_cleanup_pending_mb(struct virtio_mem *vm)
 {
 	unsigned long id;
-	int rc;
+	int rc = 0;
 
 	if (!vm->in_sbm) {
 		virtio_mem_bbm_for_each_bb(vm, id,
@@ -2287,6 +2315,27 @@ static int virtio_mem_unplug_pending_mb(struct virtio_mem *vm)
 					    VIRTIO_MEM_SBM_MB_UNUSED);
 	}
 
+	if (!vm->sbm.have_unplugged_mb)
+		return 0;
+
+	/*
+	 * Let's retry (offlining and) removing completely unplugged Linux
+	 * memory blocks.
+	 */
+	vm->sbm.have_unplugged_mb = false;
+
+	mutex_lock(&vm->hotplug_mutex);
+	virtio_mem_sbm_for_each_mb(vm, id, VIRTIO_MEM_SBM_MB_MOVABLE_PARTIAL)
+		rc |= virtio_mem_sbm_try_remove_unplugged_mb(vm, id);
+	virtio_mem_sbm_for_each_mb(vm, id, VIRTIO_MEM_SBM_MB_KERNEL_PARTIAL)
+		rc |= virtio_mem_sbm_try_remove_unplugged_mb(vm, id);
+	virtio_mem_sbm_for_each_mb(vm, id, VIRTIO_MEM_SBM_MB_OFFLINE_PARTIAL)
+		rc |= virtio_mem_sbm_try_remove_unplugged_mb(vm, id);
+	mutex_unlock(&vm->hotplug_mutex);
+
+	if (rc)
+		vm->sbm.have_unplugged_mb = true;
+	/* Ignore errors, this is not critical. We'll retry later. */
 	return 0;
 }
 
@@ -2368,9 +2417,9 @@ retry:
 		virtio_mem_refresh_config(vm);
 	}
 
-	/* Unplug any leftovers from previous runs */
+	/* Cleanup any leftovers from previous runs */
 	if (!rc)
-		rc = virtio_mem_unplug_pending_mb(vm);
+		rc = virtio_mem_cleanup_pending_mb(vm);
 
 	if (!rc && vm->requested_size != vm->plugged_size) {
 		if (vm->requested_size > vm->plugged_size) {
@@ -2382,6 +2431,13 @@ retry:
 		}
 	}
 
+	/*
+	 * Keep retrying to offline and remove completely unplugged Linux
+	 * memory blocks.
+	 */
+	if (!rc && vm->in_sbm && vm->sbm.have_unplugged_mb)
+		rc = -EBUSY;
+
 	switch (rc) {
 	case 0:
 		vm->retry_timer_ms = VIRTIO_MEM_RETRY_TIMER_MIN_MS;

drivers/virtio/virtio_mmio.c

@@ -607,9 +607,8 @@ static void virtio_mmio_release_dev(struct device *_d)
 	struct virtio_device *vdev =
 			container_of(_d, struct virtio_device, dev);
 	struct virtio_mmio_device *vm_dev = to_virtio_mmio_device(vdev);
-	struct platform_device *pdev = vm_dev->pdev;
 
-	devm_kfree(&pdev->dev, vm_dev);
+	kfree(vm_dev);
 }
 
 /* Platform device */
@@ -620,7 +619,7 @@ static int virtio_mmio_probe(struct platform_device *pdev)
 	unsigned long magic;
 	int rc;
 
-	vm_dev = devm_kzalloc(&pdev->dev, sizeof(*vm_dev), GFP_KERNEL);
+	vm_dev = kzalloc(sizeof(*vm_dev), GFP_KERNEL);
 	if (!vm_dev)
 		return -ENOMEM;
 

drivers/virtio/virtio_pci_common.c

@@ -557,8 +557,6 @@ static int virtio_pci_probe(struct pci_dev *pci_dev,
 
 	pci_set_master(pci_dev);
 
-	vp_dev->is_legacy = vp_dev->ldev.ioaddr ? true : false;
-
 	rc = register_virtio_device(&vp_dev->vdev);
 	reg_dev = vp_dev;
 	if (rc)

drivers/virtio/virtio_pci_legacy.c

@@ -223,6 +223,7 @@ int virtio_pci_legacy_probe(struct virtio_pci_device *vp_dev)
 	vp_dev->config_vector = vp_config_vector;
 	vp_dev->setup_vq = setup_vq;
 	vp_dev->del_vq = del_vq;
+	vp_dev->is_legacy = true;
 
 	return 0;
 }

drivers/virtio/virtio_vdpa.c

@@ -393,11 +393,13 @@ static int virtio_vdpa_find_vqs(struct virtio_device *vdev, unsigned int nvqs,
 	cb.callback = virtio_vdpa_config_cb;
 	cb.private = vd_dev;
 	ops->set_config_cb(vdpa, &cb);
+	kfree(masks);
 
 	return 0;
 
 err_setup_vq:
 	virtio_vdpa_del_vqs(vdev);
+	kfree(masks);
 	return err;
 }