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

Pull virtio updates from Michael Tsirkin:
 "virtio,vhost,vdpa: features, fixes, and cleanups:

   - reduction in interrupt rate in virtio

   - perf improvement for VDUSE

   - scalability for vhost-scsi

   - non power of 2 ring support for packed rings

   - better management for mlx5 vdpa

   - suspend for snet

   - VIRTIO_F_NOTIFICATION_DATA

   - shared backend with vdpa-sim-blk

   - user VA support in vdpa-sim

   - better struct packing for virtio

  and fixes, cleanups all over the place"

* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost: (52 commits)
  vhost_vdpa: fix unmap process in no-batch mode
  MAINTAINERS: make me a reviewer of VIRTIO CORE AND NET DRIVERS
  tools/virtio: fix build caused by virtio_ring changes
  virtio_ring: add a struct device forward declaration
  vdpa_sim_blk: support shared backend
  vdpa_sim: move buffer allocation in the devices
  vdpa/snet: use likely/unlikely macros in hot functions
  vdpa/snet: implement kick_vq_with_data callback
  virtio-vdpa: add VIRTIO_F_NOTIFICATION_DATA feature support
  virtio: add VIRTIO_F_NOTIFICATION_DATA feature support
  vdpa/snet: support the suspend vDPA callback
  vdpa/snet: support getting and setting VQ state
  MAINTAINERS: add vringh.h to Virtio Core and Net Drivers
  vringh: address kdoc warnings
  vdpa: address kdoc warnings
  virtio_ring: don't update event idx on get_buf
  vdpa_sim: add support for user VA
  vdpa_sim: replace the spinlock with a mutex to protect the state
  vdpa_sim: use kthread worker
  vdpa_sim: make devices agnostic for work management
  ...

MAINTAINERS

@@ -22212,6 +22212,7 @@ F:	include/uapi/linux/virtio_console.h
 VIRTIO CORE AND NET DRIVERS
 M:	"Michael S. Tsirkin" <mst@redhat.com>
 M:	Jason Wang <jasowang@redhat.com>
+R:	Xuan Zhuo <xuanzhuo@linux.alibaba.com>
 L:	virtualization@lists.linux-foundation.org
 S:	Maintained
 F:	Documentation/ABI/testing/sysfs-bus-vdpa
@@ -22225,6 +22226,7 @@ F:	drivers/vdpa/
 F:	drivers/virtio/
 F:	include/linux/vdpa.h
 F:	include/linux/virtio*.h
+F:	include/linux/vringh.h
 F:	include/uapi/linux/virtio_*.h
 F:	tools/virtio/
 

drivers/s390/virtio/virtio_ccw.c

@@ -391,7 +391,7 @@ static void virtio_ccw_drop_indicator(struct virtio_ccw_device *vcdev,
 	ccw_device_dma_free(vcdev->cdev, thinint_area, sizeof(*thinint_area));
 }
 
-static bool virtio_ccw_kvm_notify(struct virtqueue *vq)
+static inline bool virtio_ccw_do_kvm_notify(struct virtqueue *vq, u32 data)
 {
 	struct virtio_ccw_vq_info *info = vq->priv;
 	struct virtio_ccw_device *vcdev;
@@ -402,12 +402,22 @@ static bool virtio_ccw_kvm_notify(struct virtqueue *vq)
 	BUILD_BUG_ON(sizeof(struct subchannel_id) != sizeof(unsigned int));
 	info->cookie = kvm_hypercall3(KVM_S390_VIRTIO_CCW_NOTIFY,
 				      *((unsigned int *)&schid),
-				      vq->index, info->cookie);
+				      data, info->cookie);
 	if (info->cookie < 0)
 		return false;
 	return true;
 }
 
+static bool virtio_ccw_kvm_notify(struct virtqueue *vq)
+{
+	return virtio_ccw_do_kvm_notify(vq, vq->index);
+}
+
+static bool virtio_ccw_kvm_notify_with_data(struct virtqueue *vq)
+{
+	return virtio_ccw_do_kvm_notify(vq, vring_notification_data(vq));
+}
+
 static int virtio_ccw_read_vq_conf(struct virtio_ccw_device *vcdev,
 				   struct ccw1 *ccw, int index)
 {
@@ -495,6 +505,7 @@ static struct virtqueue *virtio_ccw_setup_vq(struct virtio_device *vdev,
 					     struct ccw1 *ccw)
 {
 	struct virtio_ccw_device *vcdev = to_vc_device(vdev);
+	bool (*notify)(struct virtqueue *vq);
 	int err;
 	struct virtqueue *vq = NULL;
 	struct virtio_ccw_vq_info *info;
@@ -502,6 +513,11 @@ static struct virtqueue *virtio_ccw_setup_vq(struct virtio_device *vdev,
 	unsigned long flags;
 	bool may_reduce;
 
+	if (__virtio_test_bit(vdev, VIRTIO_F_NOTIFICATION_DATA))
+		notify = virtio_ccw_kvm_notify_with_data;
+	else
+		notify = virtio_ccw_kvm_notify;
+
 	/* Allocate queue. */
 	info = kzalloc(sizeof(struct virtio_ccw_vq_info), GFP_KERNEL);
 	if (!info) {
@@ -524,7 +540,7 @@ static struct virtqueue *virtio_ccw_setup_vq(struct virtio_device *vdev,
 	may_reduce = vcdev->revision > 0;
 	vq = vring_create_virtqueue(i, info->num, KVM_VIRTIO_CCW_RING_ALIGN,
 				    vdev, true, may_reduce, ctx,
-				    virtio_ccw_kvm_notify, callback, name);
+				    notify, callback, name);
 
 	if (!vq) {
 		/* For now, we fail if we can't get the requested size. */

drivers/vdpa/solidrun/Makefile

 # SPDX-License-Identifier: GPL-2.0
 obj-$(CONFIG_SNET_VDPA) += snet_vdpa.o
 snet_vdpa-$(CONFIG_SNET_VDPA) += snet_main.o
+snet_vdpa-$(CONFIG_SNET_VDPA) += snet_ctrl.o
 ifdef CONFIG_HWMON
 snet_vdpa-$(CONFIG_SNET_VDPA) += snet_hwmon.o
 endif

drivers/vdpa/solidrun/snet_ctrl.c

+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * SolidRun DPU driver for control plane
+ *
+ * Copyright (C) 2022-2023 SolidRun
+ *
+ * Author: Alvaro Karsz <alvaro.karsz@solid-run.com>
+ *
+ */
+
+#include <linux/iopoll.h>
+
+#include "snet_vdpa.h"
+
+enum snet_ctrl_opcodes {
+	SNET_CTRL_OP_DESTROY = 1,
+	SNET_CTRL_OP_READ_VQ_STATE,
+	SNET_CTRL_OP_SUSPEND,
+};
+
+#define SNET_CTRL_TIMEOUT	        2000000
+
+#define SNET_CTRL_DATA_SIZE_MASK	0x0000FFFF
+#define SNET_CTRL_IN_PROCESS_MASK	0x00010000
+#define SNET_CTRL_CHUNK_RDY_MASK	0x00020000
+#define SNET_CTRL_ERROR_MASK		0x0FFC0000
+
+#define SNET_VAL_TO_ERR(val)		(-(((val) & SNET_CTRL_ERROR_MASK) >> 18))
+#define SNET_EMPTY_CTRL(val)		(((val) & SNET_CTRL_ERROR_MASK) || \
+						!((val) & SNET_CTRL_IN_PROCESS_MASK))
+#define SNET_DATA_READY(val)		((val) & (SNET_CTRL_ERROR_MASK | SNET_CTRL_CHUNK_RDY_MASK))
+
+/* Control register used to read data from the DPU */
+struct snet_ctrl_reg_ctrl {
+	/* Chunk size in 4B words */
+	u16 data_size;
+	/* We are in the middle of a command */
+	u16 in_process:1;
+	/* A data chunk is ready and can be consumed */
+	u16 chunk_ready:1;
+	/* Error code */
+	u16 error:10;
+	/* Saved for future usage */
+	u16 rsvd:4;
+};
+
+/* Opcode register */
+struct snet_ctrl_reg_op {
+	u16 opcode;
+	/* Only if VQ index is relevant for the command */
+	u16 vq_idx;
+};
+
+struct snet_ctrl_regs {
+	struct snet_ctrl_reg_op op;
+	struct snet_ctrl_reg_ctrl ctrl;
+	u32 rsvd;
+	u32 data[];
+};
+
+static struct snet_ctrl_regs __iomem *snet_get_ctrl(struct snet *snet)
+{
+	return snet->bar + snet->psnet->cfg.ctrl_off;
+}
+
+static int snet_wait_for_empty_ctrl(struct snet_ctrl_regs __iomem *regs)
+{
+	u32 val;
+
+	return readx_poll_timeout(ioread32, &regs->ctrl, val, SNET_EMPTY_CTRL(val), 10,
+				  SNET_CTRL_TIMEOUT);
+}
+
+static int snet_wait_for_empty_op(struct snet_ctrl_regs __iomem *regs)
+{
+	u32 val;
+
+	return readx_poll_timeout(ioread32, &regs->op, val, !val, 10, SNET_CTRL_TIMEOUT);
+}
+
+static int snet_wait_for_data(struct snet_ctrl_regs __iomem *regs)
+{
+	u32 val;
+
+	return readx_poll_timeout(ioread32, &regs->ctrl, val, SNET_DATA_READY(val), 10,
+				  SNET_CTRL_TIMEOUT);
+}
+
+static u32 snet_read32_word(struct snet_ctrl_regs __iomem *ctrl_regs, u16 word_idx)
+{
+	return ioread32(&ctrl_regs->data[word_idx]);
+}
+
+static u32 snet_read_ctrl(struct snet_ctrl_regs __iomem *ctrl_regs)
+{
+	return ioread32(&ctrl_regs->ctrl);
+}
+
+static void snet_write_ctrl(struct snet_ctrl_regs __iomem *ctrl_regs, u32 val)
+{
+	iowrite32(val, &ctrl_regs->ctrl);
+}
+
+static void snet_write_op(struct snet_ctrl_regs __iomem *ctrl_regs, u32 val)
+{
+	iowrite32(val, &ctrl_regs->op);
+}
+
+static int snet_wait_for_dpu_completion(struct snet_ctrl_regs __iomem *ctrl_regs)
+{
+	/* Wait until the DPU finishes completely.
+	 * It will clear the opcode register.
+	 */
+	return snet_wait_for_empty_op(ctrl_regs);
+}
+
+/* Reading ctrl from the DPU:
+ * buf_size must be 4B aligned
+ *
+ * Steps:
+ *
+ * (1) Verify that the DPU is not in the middle of another operation by
+ *     reading the in_process and error bits in the control register.
+ * (2) Write the request opcode and the VQ idx in the opcode register
+ *     and write the buffer size in the control register.
+ * (3) Start readind chunks of data, chunk_ready bit indicates that a
+ *     data chunk is available, we signal that we read the data by clearing the bit.
+ * (4) Detect that the transfer is completed when the in_process bit
+ *     in the control register is cleared or when the an error appears.
+ */
+static int snet_ctrl_read_from_dpu(struct snet *snet, u16 opcode, u16 vq_idx, void *buffer,
+				   u32 buf_size)
+{
+	struct pci_dev *pdev = snet->pdev;
+	struct snet_ctrl_regs __iomem *regs = snet_get_ctrl(snet);
+	u32 *bfr_ptr = (u32 *)buffer;
+	u32 val;
+	u16 buf_words;
+	int ret;
+	u16 words, i, tot_words = 0;
+
+	/* Supported for config 2+ */
+	if (!SNET_CFG_VER(snet, 2))
+		return -EOPNOTSUPP;
+
+	if (!IS_ALIGNED(buf_size, 4))
+		return -EINVAL;
+
+	mutex_lock(&snet->ctrl_lock);
+
+	buf_words = buf_size / 4;
+
+	/* Make sure control register is empty */
+	ret = snet_wait_for_empty_ctrl(regs);
+	if (ret) {
+		SNET_WARN(pdev, "Timeout waiting for previous control data to be consumed\n");
+		goto exit;
+	}
+
+	/* We need to write the buffer size in the control register, and the opcode + vq index in
+	 * the opcode register.
+	 * We use a spinlock to serialize the writes.
+	 */
+	spin_lock(&snet->ctrl_spinlock);
+
+	snet_write_ctrl(regs, buf_words);
+	snet_write_op(regs, opcode | (vq_idx << 16));
+
+	spin_unlock(&snet->ctrl_spinlock);
+
+	while (buf_words != tot_words) {
+		ret = snet_wait_for_data(regs);
+		if (ret) {
+			SNET_WARN(pdev, "Timeout waiting for control data\n");
+			goto exit;
+		}
+
+		val = snet_read_ctrl(regs);
+
+		/* Error? */
+		if (val & SNET_CTRL_ERROR_MASK) {
+			ret = SNET_VAL_TO_ERR(val);
+			SNET_WARN(pdev, "Error while reading control data from DPU, err %d\n", ret);
+			goto exit;
+		}
+
+		words = min_t(u16, val & SNET_CTRL_DATA_SIZE_MASK, buf_words - tot_words);
+
+		for (i = 0; i < words; i++) {
+			*bfr_ptr = snet_read32_word(regs, i);
+			bfr_ptr++;
+		}
+
+		tot_words += words;
+
+		/* Is the job completed? */
+		if (!(val & SNET_CTRL_IN_PROCESS_MASK))
+			break;
+
+		/* Clear the chunk ready bit and continue */
+		val &= ~SNET_CTRL_CHUNK_RDY_MASK;
+		snet_write_ctrl(regs, val);
+	}
+
+	ret = snet_wait_for_dpu_completion(regs);
+	if (ret)
+		SNET_WARN(pdev, "Timeout waiting for the DPU to complete a control command\n");
+
+exit:
+	mutex_unlock(&snet->ctrl_lock);
+	return ret;
+}
+
+/* Send a control message to the DPU using the old mechanism
+ * used with config version 1.
+ */
+static int snet_send_ctrl_msg_old(struct snet *snet, u32 opcode)
+{
+	struct pci_dev *pdev = snet->pdev;
+	struct snet_ctrl_regs __iomem *regs = snet_get_ctrl(snet);
+	int ret;
+
+	mutex_lock(&snet->ctrl_lock);
+
+	/* Old mechanism uses just 1 register, the opcode register.
+	 * Make sure that the opcode register is empty, and that the DPU isn't
+	 * processing an old message.
+	 */
+	ret = snet_wait_for_empty_op(regs);
+	if (ret) {
+		SNET_WARN(pdev, "Timeout waiting for previous control message to be ACKed\n");
+		goto exit;
+	}
+
+	/* Write the message */
+	snet_write_op(regs, opcode);
+
+	/* DPU ACKs the message by clearing the opcode register */
+	ret = snet_wait_for_empty_op(regs);
+	if (ret)
+		SNET_WARN(pdev, "Timeout waiting for a control message to be ACKed\n");
+
+exit:
+	mutex_unlock(&snet->ctrl_lock);
+	return ret;
+}
+
+/* Send a control message to the DPU.
+ * A control message is a message without payload.
+ */
+static int snet_send_ctrl_msg(struct snet *snet, u16 opcode, u16 vq_idx)
+{
+	struct pci_dev *pdev = snet->pdev;
+	struct snet_ctrl_regs __iomem *regs = snet_get_ctrl(snet);
+	u32 val;
+	int ret;
+
+	/* If config version is not 2+, use the old mechanism */
+	if (!SNET_CFG_VER(snet, 2))
+		return snet_send_ctrl_msg_old(snet, opcode);
+
+	mutex_lock(&snet->ctrl_lock);
+
+	/* Make sure control register is empty */
+	ret = snet_wait_for_empty_ctrl(regs);
+	if (ret) {
+		SNET_WARN(pdev, "Timeout waiting for previous control data to be consumed\n");
+		goto exit;
+	}
+
+	/* We need to clear the control register and write the opcode + vq index in the opcode
+	 * register.
+	 * We use a spinlock to serialize the writes.
+	 */
+	spin_lock(&snet->ctrl_spinlock);
+
+	snet_write_ctrl(regs, 0);
+	snet_write_op(regs, opcode | (vq_idx << 16));
+
+	spin_unlock(&snet->ctrl_spinlock);
+
+	/* The DPU ACKs control messages by setting the chunk ready bit
+	 * without data.
+	 */
+	ret = snet_wait_for_data(regs);
+	if (ret) {
+		SNET_WARN(pdev, "Timeout waiting for control message to be ACKed\n");
+		goto exit;
+	}
+
+	/* Check for errors */
+	val = snet_read_ctrl(regs);
+	ret = SNET_VAL_TO_ERR(val);
+
+	/* Clear the chunk ready bit */
+	val &= ~SNET_CTRL_CHUNK_RDY_MASK;
+	snet_write_ctrl(regs, val);
+
+	ret = snet_wait_for_dpu_completion(regs);
+	if (ret)
+		SNET_WARN(pdev, "Timeout waiting for DPU to complete a control command, err %d\n",
+			  ret);
+
+exit:
+	mutex_unlock(&snet->ctrl_lock);
+	return ret;
+}
+
+void snet_ctrl_clear(struct snet *snet)
+{
+	struct snet_ctrl_regs __iomem *regs = snet_get_ctrl(snet);
+
+	snet_write_op(regs, 0);
+}
+
+int snet_destroy_dev(struct snet *snet)
+{
+	return snet_send_ctrl_msg(snet, SNET_CTRL_OP_DESTROY, 0);
+}
+
+int snet_read_vq_state(struct snet *snet, u16 idx, struct vdpa_vq_state *state)
+{
+	return snet_ctrl_read_from_dpu(snet, SNET_CTRL_OP_READ_VQ_STATE, idx, state,
+				       sizeof(*state));
+}
+
+int snet_suspend_dev(struct snet *snet)
+{
+	return snet_send_ctrl_msg(snet, SNET_CTRL_OP_SUSPEND, 0);
+}

drivers/vdpa/solidrun/snet_hwmon.c

@@ -2,7 +2,7 @@
 /*
  * SolidRun DPU driver for control plane
  *
- * Copyright (C) 2022 SolidRun
+ * Copyright (C) 2022-2023 SolidRun
  *
  * Author: Alvaro Karsz <alvaro.karsz@solid-run.com>
  *

drivers/vdpa/solidrun/snet_main.c

@@ -2,7 +2,7 @@
 /*
  * SolidRun DPU driver for control plane
  *
- * Copyright (C) 2022 SolidRun
+ * Copyright (C) 2022-2023 SolidRun
  *
  * Author: Alvaro Karsz <alvaro.karsz@solid-run.com>
  *
@@ -16,14 +16,12 @@
 /* SNET signature */
 #define SNET_SIGNATURE          0xD0D06363
 /* Max. config version that we can work with */
-#define SNET_CFG_VERSION        0x1
+#define SNET_CFG_VERSION        0x2
 /* Queue align */
 #define SNET_QUEUE_ALIGNMENT    PAGE_SIZE
 /* Kick value to notify that new data is available */
 #define SNET_KICK_VAL           0x1
 #define SNET_CONFIG_OFF         0x0
-/* ACK timeout for a message */
-#define SNET_ACK_TIMEOUT	2000000
 /* How long we are willing to wait for a SNET device */
 #define SNET_DETECT_TIMEOUT	5000000
 /* How long should we wait for the DPU to read our config */
@@ -32,61 +30,16 @@
 #define SNET_GENERAL_CFG_LEN	36
 #define SNET_GENERAL_CFG_VQ_LEN	40
 
-enum snet_msg {
-	SNET_MSG_DESTROY = 1,
-};
-
 static struct snet *vdpa_to_snet(struct vdpa_device *vdpa)
 {
 	return container_of(vdpa, struct snet, vdpa);
 }
 
-static int snet_wait_for_msg_ack(struct snet *snet)
-{
-	struct pci_dev *pdev = snet->pdev;
-	int ret;
-	u32 val;
-
-	/* The DPU will clear the messages offset once messages
-	 * are processed.
-	 */
-	ret = readx_poll_timeout(ioread32, snet->bar + snet->psnet->cfg.msg_off,
-				 val, !val, 10, SNET_ACK_TIMEOUT);
-	if (ret)
-		SNET_WARN(pdev, "Timeout waiting for message ACK\n");
-
-	return ret;
-}
-
-/* Sends a message to the DPU.
- * If blocking is set, the function will return once the
- * message was processed by the DPU (or timeout).
- */
-static int snet_send_msg(struct snet *snet, u32 msg, bool blocking)
-{
-	int ret = 0;
-
-	/* Make sure the DPU acked last message before issuing a new one */
-	ret = snet_wait_for_msg_ack(snet);
-	if (ret)
-		return ret;
-
-	/* Write the message */
-	snet_write32(snet, snet->psnet->cfg.msg_off, msg);
-
-	if (blocking)
-		ret = snet_wait_for_msg_ack(snet);
-	else /* If non-blocking, flush the write by issuing a read */
-		snet_read32(snet, snet->psnet->cfg.msg_off);
-
-	return ret;
-}
-
 static irqreturn_t snet_cfg_irq_hndlr(int irq, void *data)
 {
 	struct snet *snet = data;
 	/* Call callback if any */
-	if (snet->cb.callback)
+	if (likely(snet->cb.callback))
 		return snet->cb.callback(snet->cb.private);
 
 	return IRQ_HANDLED;
@@ -96,7 +49,7 @@ static irqreturn_t snet_vq_irq_hndlr(int irq, void *data)
 {
 	struct snet_vq *vq = data;
 	/* Call callback if any */
-	if (vq->cb.callback)
+	if (likely(vq->cb.callback))
 		return vq->cb.callback(vq->cb.private);
 
 	return IRQ_HANDLED;
@@ -153,12 +106,24 @@ static void snet_kick_vq(struct vdpa_device *vdev, u16 idx)
 {
 	struct snet *snet = vdpa_to_snet(vdev);
 	/* not ready - ignore */
-	if (!snet->vqs[idx]->ready)
+	if (unlikely(!snet->vqs[idx]->ready))
 		return;
 
 	iowrite32(SNET_KICK_VAL, snet->vqs[idx]->kick_ptr);
 }
 
+static void snet_kick_vq_with_data(struct vdpa_device *vdev, u32 data)
+{
+	struct snet *snet = vdpa_to_snet(vdev);
+	u16 idx = data & 0xFFFF;
+
+	/* not ready - ignore */
+	if (unlikely(!snet->vqs[idx]->ready))
+		return;
+
+	iowrite32((data & 0xFFFF0000) | SNET_KICK_VAL, snet->vqs[idx]->kick_ptr);
+}
+
 static void snet_set_vq_cb(struct vdpa_device *vdev, u16 idx, struct vdpa_callback *cb)
 {
 	struct snet *snet = vdpa_to_snet(vdev);
@@ -181,33 +146,48 @@ static bool snet_get_vq_ready(struct vdpa_device *vdev, u16 idx)
 	return snet->vqs[idx]->ready;
 }
 
-static int snet_set_vq_state(struct vdpa_device *vdev, u16 idx, const struct vdpa_vq_state *state)
+static bool snet_vq_state_is_initial(struct snet *snet, const struct vdpa_vq_state *state)
 {
-	struct snet *snet = vdpa_to_snet(vdev);
-	/* Setting the VQ state is not supported.
-	 * If the asked state is the same as the initial one
-	 * we can ignore it.
-	 */
 	if (SNET_HAS_FEATURE(snet, VIRTIO_F_RING_PACKED)) {
 		const struct vdpa_vq_state_packed *p = &state->packed;
 
 		if (p->last_avail_counter == 1 && p->last_used_counter == 1 &&
 		    p->last_avail_idx == 0 && p->last_used_idx == 0)
-			return 0;
+			return true;
 	} else {
 		const struct vdpa_vq_state_split *s = &state->split;
 
 		if (s->avail_index == 0)
-			return 0;
+			return true;
 	}
 
+	return false;
+}
+
+static int snet_set_vq_state(struct vdpa_device *vdev, u16 idx, const struct vdpa_vq_state *state)
+{
+	struct snet *snet = vdpa_to_snet(vdev);
+
+	/* We can set any state for config version 2+ */
+	if (SNET_CFG_VER(snet, 2)) {
+		memcpy(&snet->vqs[idx]->vq_state, state, sizeof(*state));
+		return 0;
+	}
+
+	/* Older config - we can't set the VQ state.
+	 * Return 0 only if this is the initial state we use in the DPU.
+	 */
+	if (snet_vq_state_is_initial(snet, state))
+		return 0;
+
 	return -EOPNOTSUPP;
 }
 
 static int snet_get_vq_state(struct vdpa_device *vdev, u16 idx, struct vdpa_vq_state *state)
 {
-	/* Not supported */
-	return -EOPNOTSUPP;
+	struct snet *snet = vdpa_to_snet(vdev);
+
+	return snet_read_vq_state(snet, idx, state);
 }
 
 static int snet_get_vq_irq(struct vdpa_device *vdev, u16 idx)
@@ -232,9 +212,9 @@ static int snet_reset_dev(struct snet *snet)
 	if (!snet->status)
 		return 0;
 
-	/* If DPU started, send a destroy message */
+	/* If DPU started, destroy it */
 	if (snet->status & VIRTIO_CONFIG_S_DRIVER_OK)
-		ret = snet_send_msg(snet, SNET_MSG_DESTROY, true);
+		ret = snet_destroy_dev(snet);
 
 	/* Clear VQs */
 	for (i = 0; i < snet->cfg->vq_num; i++) {
@@ -258,7 +238,7 @@ static int snet_reset_dev(struct snet *snet)
 	snet->dpu_ready = false;
 
 	if (ret)
-		SNET_WARN(pdev, "Incomplete reset to SNET[%u] device\n", snet->sid);
+		SNET_WARN(pdev, "Incomplete reset to SNET[%u] device, err: %d\n", snet->sid, ret);
 	else
 		SNET_DBG(pdev, "Reset SNET[%u] device\n", snet->sid);
 
@@ -356,7 +336,7 @@ static int snet_write_conf(struct snet *snet)
 	 * |             DESC AREA                |
 	 * |            DEVICE AREA               |
 	 * |            DRIVER AREA               |
-	 * |             RESERVED                 |
+	 * |    VQ STATE (CFG 2+)     |   RSVD    |
 	 *
 	 * Magic number should be written last, this is the DPU indication that the data is ready
 	 */
@@ -391,12 +371,15 @@ static int snet_write_conf(struct snet *snet)
 		off += 8;
 		snet_write64(snet, off, snet->vqs[i]->driver_area);
 		off += 8;
+		/* Write VQ state if config version is 2+ */
+		if (SNET_CFG_VER(snet, 2))
+			snet_write32(snet, off, *(u32 *)&snet->vqs[i]->vq_state);
+		off += 4;
+
 		/* Ignore reserved */
-		off += 8;
+		off += 4;
 	}
 
-	/* Clear snet messages address for this device */
-	snet_write32(snet, snet->psnet->cfg.msg_off, 0);
 	/* Write magic number - data is ready */
 	snet_write32(snet, snet->psnet->cfg.host_cfg_off, SNET_SIGNATURE);
 
@@ -512,10 +495,25 @@ static void snet_set_config(struct vdpa_device *vdev, unsigned int offset,
 		iowrite8(*buf_ptr++, cfg_ptr + i);
 }
 
+static int snet_suspend(struct vdpa_device *vdev)
+{
+	struct snet *snet = vdpa_to_snet(vdev);
+	int ret;
+
+	ret = snet_suspend_dev(snet);
+	if (ret)
+		SNET_ERR(snet->pdev, "SNET[%u] suspend failed, err: %d\n", snet->sid, ret);
+	else
+		SNET_DBG(snet->pdev, "Suspend SNET[%u] device\n", snet->sid);
+
+	return ret;
+}
+
 static const struct vdpa_config_ops snet_config_ops = {
 	.set_vq_address         = snet_set_vq_address,
 	.set_vq_num             = snet_set_vq_num,
 	.kick_vq                = snet_kick_vq,
+	.kick_vq_with_data	= snet_kick_vq_with_data,
 	.set_vq_cb              = snet_set_vq_cb,
 	.set_vq_ready           = snet_set_vq_ready,
 	.get_vq_ready           = snet_get_vq_ready,
@@ -537,6 +535,7 @@ static const struct vdpa_config_ops snet_config_ops = {
 	.set_status             = snet_set_status,
 	.get_config             = snet_get_config,
 	.set_config             = snet_set_config,
+	.suspend		= snet_suspend,
 };
 
 static int psnet_open_pf_bar(struct pci_dev *pdev, struct psnet *psnet)
@@ -697,7 +696,7 @@ static int psnet_read_cfg(struct pci_dev *pdev, struct psnet *psnet)
 	off += 4;
 	cfg->hwmon_off = psnet_read32(psnet, off);
 	off += 4;
-	cfg->msg_off = psnet_read32(psnet, off);
+	cfg->ctrl_off = psnet_read32(psnet, off);
 	off += 4;
 	cfg->flags = psnet_read32(psnet, off);
 	off += 4;
@@ -997,6 +996,10 @@ static int snet_vdpa_probe_vf(struct pci_dev *pdev)
 		goto free_irqs;
 	}
 
+	/* Init control mutex and spinlock */
+	mutex_init(&snet->ctrl_lock);
+	spin_lock_init(&snet->ctrl_spinlock);
+
 	/* Save pci device pointer */
 	snet->pdev = pdev;
 	snet->psnet = psnet;
@@ -1013,6 +1016,9 @@ static int snet_vdpa_probe_vf(struct pci_dev *pdev)
 	/* Create a VirtIO config pointer */
 	snet->cfg->virtio_cfg = snet->bar + snet->psnet->cfg.virtio_cfg_off;
 
+	/* Clear control registers */
+	snet_ctrl_clear(snet);
+
 	pci_set_master(pdev);
 	pci_set_drvdata(pdev, snet);
 

drivers/vdpa/solidrun/snet_vdpa.h

@@ -2,7 +2,7 @@
 /*
  * SolidRun DPU driver for control plane
  *
- * Copyright (C) 2022 SolidRun
+ * Copyright (C) 2022-2023 SolidRun
  *
  * Author: Alvaro Karsz <alvaro.karsz@solid-run.com>
  *
@@ -20,10 +20,15 @@
 #define SNET_INFO(pdev, fmt, ...) dev_info(&(pdev)->dev, "%s"fmt, "snet_vdpa: ", ##__VA_ARGS__)
 #define SNET_DBG(pdev, fmt, ...) dev_dbg(&(pdev)->dev, "%s"fmt, "snet_vdpa: ", ##__VA_ARGS__)
 #define SNET_HAS_FEATURE(s, f) ((s)->negotiated_features & BIT_ULL(f))
+/* Check if negotiated config version is at least @ver */
+#define SNET_CFG_VER(snet, ver) ((snet)->psnet->negotiated_cfg_ver >= (ver))
+
 /* VQ struct */
 struct snet_vq {
 	/* VQ callback */
 	struct vdpa_callback cb;
+	/* VQ state received from bus */
+	struct vdpa_vq_state vq_state;
 	/* desc base address */
 	u64 desc_area;
 	/* device base address */
@@ -51,6 +56,10 @@ struct snet {
 	struct vdpa_device vdpa;
 	/* Config callback */
 	struct vdpa_callback cb;
+	/* To lock the control mechanism */
+	struct mutex ctrl_lock;
+	/* Spinlock to protect critical parts in the control mechanism */
+	spinlock_t ctrl_spinlock;
 	/* array of virqueues */
 	struct snet_vq **vqs;
 	/* Used features */
@@ -117,8 +126,8 @@ struct snet_cfg {
 	u32 kick_off;
 	/* Offset in PCI BAR for HW monitoring */
 	u32 hwmon_off;
-	/* Offset in PCI BAR for SNET messages */
-	u32 msg_off;
+	/* Offset in PCI BAR for Control mechanism */
+	u32 ctrl_off;
 	/* Config general flags - enum snet_cfg_flags */
 	u32 flags;
 	/* Reserved for future usage */
@@ -191,4 +200,9 @@ static inline void snet_write64(struct snet *snet, u32 off, u64 val)
 void psnet_create_hwmon(struct pci_dev *pdev);
 #endif
 
+void snet_ctrl_clear(struct snet *snet);
+int snet_destroy_dev(struct snet *snet);
+int snet_read_vq_state(struct snet *snet, u16 idx, struct vdpa_vq_state *state);
+int snet_suspend_dev(struct snet *snet);
+
 #endif //_SNET_VDPA_H_

drivers/vdpa/vdpa_sim/vdpa_sim.h

@@ -39,33 +39,34 @@ struct vdpasim_dev_attr {
 	u64 supported_features;
 	size_t alloc_size;
 	size_t config_size;
-	size_t buffer_size;
 	int nvqs;
 	u32 id;
 	u32 ngroups;
 	u32 nas;
 
-	work_func_t work_fn;
+	void (*work_fn)(struct vdpasim *vdpasim);
 	void (*get_config)(struct vdpasim *vdpasim, void *config);
 	void (*set_config)(struct vdpasim *vdpasim, const void *config);
 	int (*get_stats)(struct vdpasim *vdpasim, u16 idx,
 			 struct sk_buff *msg,
 			 struct netlink_ext_ack *extack);
+	void (*free)(struct vdpasim *vdpasim);
 };
 
 /* State of each vdpasim device */
 struct vdpasim {
 	struct vdpa_device vdpa;
 	struct vdpasim_virtqueue *vqs;
-	struct work_struct work;
+	struct kthread_worker *worker;
+	struct kthread_work work;
+	struct mm_struct *mm_bound;
 	struct vdpasim_dev_attr dev_attr;
-	/* spinlock to synchronize virtqueue state */
-	spinlock_t lock;
+	/* mutex to synchronize virtqueue state */
+	struct mutex mutex;
 	/* virtio config according to device type */
 	void *config;
 	struct vhost_iotlb *iommu;
 	bool *iommu_pt;
-	void *buffer;
 	u32 status;
 	u32 generation;
 	u64 features;
@@ -78,6 +79,7 @@ struct vdpasim {
 
 struct vdpasim *vdpasim_create(struct vdpasim_dev_attr *attr,
 			       const struct vdpa_dev_set_config *config);
+void vdpasim_schedule_work(struct vdpasim *vdpasim);
 
 /* TODO: cross-endian support */
 static inline bool vdpasim_is_little_endian(struct vdpasim *vdpasim)

drivers/vdpa/vdpa_sim/vdpa_sim_blk.c

@@ -11,7 +11,6 @@
 #include <linux/module.h>
 #include <linux/device.h>
 #include <linux/kernel.h>
-#include <linux/sched.h>
 #include <linux/blkdev.h>
 #include <linux/vringh.h>
 #include <linux/vdpa.h>
@@ -44,8 +43,39 @@
 #define VDPASIM_BLK_AS_NUM	1
 #define VDPASIM_BLK_GROUP_NUM	1
 
+struct vdpasim_blk {
+	struct vdpasim vdpasim;
+	void *buffer;
+	bool shared_backend;
+};
+
+static struct vdpasim_blk *sim_to_blk(struct vdpasim *vdpasim)
+{
+	return container_of(vdpasim, struct vdpasim_blk, vdpasim);
+}
+
 static char vdpasim_blk_id[VIRTIO_BLK_ID_BYTES] = "vdpa_blk_sim";
 
+static bool shared_backend;
+module_param(shared_backend, bool, 0444);
+MODULE_PARM_DESC(shared_backend, "Enable the shared backend between virtio-blk devices");
+
+static void *shared_buffer;
+/* mutex to synchronize shared_buffer access */
+static DEFINE_MUTEX(shared_buffer_mutex);
+
+static void vdpasim_blk_buffer_lock(struct vdpasim_blk *blk)
+{
+	if (blk->shared_backend)
+		mutex_lock(&shared_buffer_mutex);
+}
+
+static void vdpasim_blk_buffer_unlock(struct vdpasim_blk *blk)
+{
+	if (blk->shared_backend)
+		mutex_unlock(&shared_buffer_mutex);
+}
+
 static bool vdpasim_blk_check_range(struct vdpasim *vdpasim, u64 start_sector,
 				    u64 num_sectors, u64 max_sectors)
 {
@@ -79,6 +109,7 @@ static bool vdpasim_blk_check_range(struct vdpasim *vdpasim, u64 start_sector,
 static bool vdpasim_blk_handle_req(struct vdpasim *vdpasim,
 				   struct vdpasim_virtqueue *vq)
 {
+	struct vdpasim_blk *blk = sim_to_blk(vdpasim);
 	size_t pushed = 0, to_pull, to_push;
 	struct virtio_blk_outhdr hdr;
 	bool handled = false;
@@ -144,9 +175,10 @@ static bool vdpasim_blk_handle_req(struct vdpasim *vdpasim,
 			break;
 		}
 
+		vdpasim_blk_buffer_lock(blk);
 		bytes = vringh_iov_push_iotlb(&vq->vring, &vq->in_iov,
-					      vdpasim->buffer + offset,
-					      to_push);
+					      blk->buffer + offset, to_push);
+		vdpasim_blk_buffer_unlock(blk);
 		if (bytes < 0) {
 			dev_dbg(&vdpasim->vdpa.dev,
 				"vringh_iov_push_iotlb() error: %zd offset: 0x%llx len: 0x%zx\n",
@@ -166,9 +198,10 @@ static bool vdpasim_blk_handle_req(struct vdpasim *vdpasim,
 			break;
 		}
 
+		vdpasim_blk_buffer_lock(blk);
 		bytes = vringh_iov_pull_iotlb(&vq->vring, &vq->out_iov,
-					      vdpasim->buffer + offset,
-					      to_pull);
+					      blk->buffer + offset, to_pull);
+		vdpasim_blk_buffer_unlock(blk);
 		if (bytes < 0) {
 			dev_dbg(&vdpasim->vdpa.dev,
 				"vringh_iov_pull_iotlb() error: %zd offset: 0x%llx len: 0x%zx\n",
@@ -248,8 +281,10 @@ static bool vdpasim_blk_handle_req(struct vdpasim *vdpasim,
 		}
 
 		if (type == VIRTIO_BLK_T_WRITE_ZEROES) {
-			memset(vdpasim->buffer + offset, 0,
+			vdpasim_blk_buffer_lock(blk);
+			memset(blk->buffer + offset, 0,
 			       num_sectors << SECTOR_SHIFT);
+			vdpasim_blk_buffer_unlock(blk);
 		}
 
 		break;
@@ -286,13 +321,12 @@ static bool vdpasim_blk_handle_req(struct vdpasim *vdpasim,
 	return handled;
 }
 
-static void vdpasim_blk_work(struct work_struct *work)
+static void vdpasim_blk_work(struct vdpasim *vdpasim)
 {
-	struct vdpasim *vdpasim = container_of(work, struct vdpasim, work);
 	bool reschedule = false;
 	int i;
 
-	spin_lock(&vdpasim->lock);
+	mutex_lock(&vdpasim->mutex);
 
 	if (!(vdpasim->status & VIRTIO_CONFIG_S_DRIVER_OK))
 		goto out;
@@ -323,10 +357,10 @@ static void vdpasim_blk_work(struct work_struct *work)
 		}
 	}
 out:
-	spin_unlock(&vdpasim->lock);
+	mutex_unlock(&vdpasim->mutex);
 
 	if (reschedule)
-		schedule_work(&vdpasim->work);
+		vdpasim_schedule_work(vdpasim);
 }
 
 static void vdpasim_blk_get_config(struct vdpasim *vdpasim, void *config)
@@ -355,6 +389,14 @@ static void vdpasim_blk_get_config(struct vdpasim *vdpasim, void *config)
 
 }
 
+static void vdpasim_blk_free(struct vdpasim *vdpasim)
+{
+	struct vdpasim_blk *blk = sim_to_blk(vdpasim);
+
+	if (!blk->shared_backend)
+		kvfree(blk->buffer);
+}
+
 static void vdpasim_blk_mgmtdev_release(struct device *dev)
 {
 }
@@ -368,6 +410,7 @@ static int vdpasim_blk_dev_add(struct vdpa_mgmt_dev *mdev, const char *name,
 			       const struct vdpa_dev_set_config *config)
 {
 	struct vdpasim_dev_attr dev_attr = {};
+	struct vdpasim_blk *blk;
 	struct vdpasim *simdev;
 	int ret;
 
@@ -378,16 +421,30 @@ static int vdpasim_blk_dev_add(struct vdpa_mgmt_dev *mdev, const char *name,
 	dev_attr.nvqs = VDPASIM_BLK_VQ_NUM;
 	dev_attr.ngroups = VDPASIM_BLK_GROUP_NUM;
 	dev_attr.nas = VDPASIM_BLK_AS_NUM;
-	dev_attr.alloc_size = sizeof(struct vdpasim);
+	dev_attr.alloc_size = sizeof(struct vdpasim_blk);
 	dev_attr.config_size = sizeof(struct virtio_blk_config);
 	dev_attr.get_config = vdpasim_blk_get_config;
 	dev_attr.work_fn = vdpasim_blk_work;
-	dev_attr.buffer_size = VDPASIM_BLK_CAPACITY << SECTOR_SHIFT;
+	dev_attr.free = vdpasim_blk_free;
 
 	simdev = vdpasim_create(&dev_attr, config);
 	if (IS_ERR(simdev))
 		return PTR_ERR(simdev);
 
+	blk = sim_to_blk(simdev);
+	blk->shared_backend = shared_backend;
+
+	if (blk->shared_backend) {
+		blk->buffer = shared_buffer;
+	} else {
+		blk->buffer = kvmalloc(VDPASIM_BLK_CAPACITY << SECTOR_SHIFT,
+				       GFP_KERNEL);
+		if (!blk->buffer) {
+			ret = -ENOMEM;
+			goto put_dev;
+		}
+	}
+
 	ret = _vdpa_register_device(&simdev->vdpa, VDPASIM_BLK_VQ_NUM);
 	if (ret)
 		goto put_dev;
@@ -437,6 +494,15 @@ static int __init vdpasim_blk_init(void)
 	if (ret)
 		goto parent_err;
 
+	if (shared_backend) {
+		shared_buffer = kvmalloc(VDPASIM_BLK_CAPACITY << SECTOR_SHIFT,
+					 GFP_KERNEL);
+		if (!shared_buffer) {
+			ret = -ENOMEM;
+			goto parent_err;
+		}
+	}
+
 	return 0;
 
 parent_err:
@@ -446,6 +512,7 @@ static int __init vdpasim_blk_init(void)
 
 static void __exit vdpasim_blk_exit(void)
 {
+	kvfree(shared_buffer);
 	vdpa_mgmtdev_unregister(&mgmt_dev);
 	device_unregister(&vdpasim_blk_mgmtdev);
 }

drivers/vdpa/vdpa_sim/vdpa_sim_net.c

@@ -11,7 +11,6 @@
 #include <linux/module.h>
 #include <linux/device.h>
 #include <linux/kernel.h>
-#include <linux/sched.h>
 #include <linux/etherdevice.h>
 #include <linux/vringh.h>
 #include <linux/vdpa.h>
@@ -59,6 +58,7 @@ struct vdpasim_net{
 	struct vdpasim_dataq_stats tx_stats;
 	struct vdpasim_dataq_stats rx_stats;
 	struct vdpasim_cq_stats cq_stats;
+	void *buffer;
 };
 
 static struct vdpasim_net *sim_to_net(struct vdpasim *vdpasim)
@@ -88,14 +88,15 @@ static bool receive_filter(struct vdpasim *vdpasim, size_t len)
 	size_t hdr_len = modern ? sizeof(struct virtio_net_hdr_v1) :
 				  sizeof(struct virtio_net_hdr);
 	struct virtio_net_config *vio_config = vdpasim->config;
+	struct vdpasim_net *net = sim_to_net(vdpasim);
 
 	if (len < ETH_ALEN + hdr_len)
 		return false;
 
-	if (is_broadcast_ether_addr(vdpasim->buffer + hdr_len) ||
-	    is_multicast_ether_addr(vdpasim->buffer + hdr_len))
+	if (is_broadcast_ether_addr(net->buffer + hdr_len) ||
+	    is_multicast_ether_addr(net->buffer + hdr_len))
 		return true;
-	if (!strncmp(vdpasim->buffer + hdr_len, vio_config->mac, ETH_ALEN))
+	if (!strncmp(net->buffer + hdr_len, vio_config->mac, ETH_ALEN))
 		return true;
 
 	return false;
@@ -192,9 +193,8 @@ static void vdpasim_handle_cvq(struct vdpasim *vdpasim)
 	u64_stats_update_end(&net->cq_stats.syncp);
 }
 
-static void vdpasim_net_work(struct work_struct *work)
+static void vdpasim_net_work(struct vdpasim *vdpasim)
 {
-	struct vdpasim *vdpasim = container_of(work, struct vdpasim, work);
 	struct vdpasim_virtqueue *txq = &vdpasim->vqs[1];
 	struct vdpasim_virtqueue *rxq = &vdpasim->vqs[0];
 	struct vdpasim_net *net = sim_to_net(vdpasim);
@@ -203,7 +203,7 @@ static void vdpasim_net_work(struct work_struct *work)
 	u64 rx_drops = 0, rx_overruns = 0, rx_errors = 0, tx_errors = 0;
 	int err;
 
-	spin_lock(&vdpasim->lock);
+	mutex_lock(&vdpasim->mutex);
 
 	if (!vdpasim->running)
 		goto out;
@@ -227,8 +227,7 @@ static void vdpasim_net_work(struct work_struct *work)
 
 		++tx_pkts;
 		read = vringh_iov_pull_iotlb(&txq->vring, &txq->out_iov,
-					     vdpasim->buffer,
-					     PAGE_SIZE);
+					     net->buffer, PAGE_SIZE);
 
 		tx_bytes += read;
 
@@ -247,7 +246,7 @@ static void vdpasim_net_work(struct work_struct *work)
 		}
 
 		write = vringh_iov_push_iotlb(&rxq->vring, &rxq->in_iov,
-					      vdpasim->buffer, read);
+					      net->buffer, read);
 		if (write <= 0) {
 			++rx_errors;
 			break;
@@ -260,13 +259,13 @@ static void vdpasim_net_work(struct work_struct *work)
 		vdpasim_net_complete(rxq, write);
 
 		if (tx_pkts > 4) {
-			schedule_work(&vdpasim->work);
+			vdpasim_schedule_work(vdpasim);
 			goto out;
 		}
 	}
 
 out:
-	spin_unlock(&vdpasim->lock);
+	mutex_unlock(&vdpasim->mutex);
 
 	u64_stats_update_begin(&net->tx_stats.syncp);
 	net->tx_stats.pkts += tx_pkts;
@@ -429,6 +428,13 @@ static void vdpasim_net_setup_config(struct vdpasim *vdpasim,
 		vio_config->mtu = cpu_to_vdpasim16(vdpasim, 1500);
 }
 
+static void vdpasim_net_free(struct vdpasim *vdpasim)
+{
+	struct vdpasim_net *net = sim_to_net(vdpasim);
+
+	kvfree(net->buffer);
+}
+
 static void vdpasim_net_mgmtdev_release(struct device *dev)
 {
 }
@@ -458,7 +464,7 @@ static int vdpasim_net_dev_add(struct vdpa_mgmt_dev *mdev, const char *name,
 	dev_attr.get_config = vdpasim_net_get_config;
 	dev_attr.work_fn = vdpasim_net_work;
 	dev_attr.get_stats = vdpasim_net_get_stats;
-	dev_attr.buffer_size = PAGE_SIZE;
+	dev_attr.free = vdpasim_net_free;
 
 	simdev = vdpasim_create(&dev_attr, config);
 	if (IS_ERR(simdev))
@@ -472,6 +478,12 @@ static int vdpasim_net_dev_add(struct vdpa_mgmt_dev *mdev, const char *name,
 	u64_stats_init(&net->rx_stats.syncp);
 	u64_stats_init(&net->cq_stats.syncp);
 
+	net->buffer = kvmalloc(PAGE_SIZE, GFP_KERNEL);
+	if (!net->buffer) {
+		ret = -ENOMEM;
+		goto reg_err;
+	}
+
 	/*
 	 * Initialization must be completed before this call, since it can
 	 * connect the device to the vDPA bus, so requests can arrive after

drivers/vhost/scsi.c

@@ -229,7 +229,10 @@ struct vhost_scsi_ctx {
 	struct iov_iter out_iter;
 };
 
-/* Global spinlock to protect vhost_scsi TPG list for vhost IOCTL access */
+/*
+ * Global mutex to protect vhost_scsi TPG list for vhost IOCTLs and LIO
+ * configfs management operations.
+ */
 static DEFINE_MUTEX(vhost_scsi_mutex);
 static LIST_HEAD(vhost_scsi_list);
 
@@ -1501,7 +1504,7 @@ static int vhost_scsi_setup_vq_cmds(struct vhost_virtqueue *vq, int max_cmds)
  * vhost_scsi_tpg with an active struct vhost_scsi_nexus
  *
  *  The lock nesting rule is:
- *    vhost_scsi_mutex -> vs->dev.mutex -> tpg->tv_tpg_mutex -> vq->mutex
+ *    vs->dev.mutex -> vhost_scsi_mutex -> tpg->tv_tpg_mutex -> vq->mutex
  */
 static int
 vhost_scsi_set_endpoint(struct vhost_scsi *vs,
@@ -1515,7 +1518,6 @@ vhost_scsi_set_endpoint(struct vhost_scsi *vs,
 	int index, ret, i, len;
 	bool match = false;
 
-	mutex_lock(&vhost_scsi_mutex);
 	mutex_lock(&vs->dev.mutex);
 
 	/* Verify that ring has been setup correctly. */
@@ -1536,6 +1538,7 @@ vhost_scsi_set_endpoint(struct vhost_scsi *vs,
 	if (vs->vs_tpg)
 		memcpy(vs_tpg, vs->vs_tpg, len);
 
+	mutex_lock(&vhost_scsi_mutex);
 	list_for_each_entry(tpg, &vhost_scsi_list, tv_tpg_list) {
 		mutex_lock(&tpg->tv_tpg_mutex);
 		if (!tpg->tpg_nexus) {
@@ -1551,6 +1554,7 @@ vhost_scsi_set_endpoint(struct vhost_scsi *vs,
 		if (!strcmp(tv_tport->tport_name, t->vhost_wwpn)) {
 			if (vs->vs_tpg && vs->vs_tpg[tpg->tport_tpgt]) {
 				mutex_unlock(&tpg->tv_tpg_mutex);
+				mutex_unlock(&vhost_scsi_mutex);
 				ret = -EEXIST;
 				goto undepend;
 			}
@@ -1565,6 +1569,7 @@ vhost_scsi_set_endpoint(struct vhost_scsi *vs,
 			if (ret) {
 				pr_warn("target_depend_item() failed: %d\n", ret);
 				mutex_unlock(&tpg->tv_tpg_mutex);
+				mutex_unlock(&vhost_scsi_mutex);
 				goto undepend;
 			}
 			tpg->tv_tpg_vhost_count++;
@@ -1574,6 +1579,7 @@ vhost_scsi_set_endpoint(struct vhost_scsi *vs,
 		}
 		mutex_unlock(&tpg->tv_tpg_mutex);
 	}
+	mutex_unlock(&vhost_scsi_mutex);
 
 	if (match) {
 		memcpy(vs->vs_vhost_wwpn, t->vhost_wwpn,
@@ -1629,7 +1635,6 @@ vhost_scsi_set_endpoint(struct vhost_scsi *vs,
 	kfree(vs_tpg);
 out:
 	mutex_unlock(&vs->dev.mutex);
-	mutex_unlock(&vhost_scsi_mutex);
 	return ret;
 }
 
@@ -1645,7 +1650,6 @@ vhost_scsi_clear_endpoint(struct vhost_scsi *vs,
 	int index, ret, i;
 	u8 target;
 
-	mutex_lock(&vhost_scsi_mutex);
 	mutex_lock(&vs->dev.mutex);
 	/* Verify that ring has been setup correctly. */
 	for (index = 0; index < vs->dev.nvqs; ++index) {
@@ -1666,11 +1670,10 @@ vhost_scsi_clear_endpoint(struct vhost_scsi *vs,
 		if (!tpg)
 			continue;
 
-		mutex_lock(&tpg->tv_tpg_mutex);
 		tv_tport = tpg->tport;
 		if (!tv_tport) {
 			ret = -ENODEV;
-			goto err_tpg;
+			goto err_dev;
 		}
 
 		if (strcmp(tv_tport->tport_name, t->vhost_wwpn)) {
@@ -1679,35 +1682,51 @@ vhost_scsi_clear_endpoint(struct vhost_scsi *vs,
 				tv_tport->tport_name, tpg->tport_tpgt,
 				t->vhost_wwpn, t->vhost_tpgt);
 			ret = -EINVAL;
-			goto err_tpg;
+			goto err_dev;
 		}
+		match = true;
+	}
+	if (!match)
+		goto free_vs_tpg;
+
+	/* Prevent new cmds from starting and accessing the tpgs/sessions */
+	for (i = 0; i < vs->dev.nvqs; i++) {
+		vq = &vs->vqs[i].vq;
+		mutex_lock(&vq->mutex);
+		vhost_vq_set_backend(vq, NULL);
+		mutex_unlock(&vq->mutex);
+	}
+	/* Make sure cmds are not running before tearing them down. */
+	vhost_scsi_flush(vs);
+
+	for (i = 0; i < vs->dev.nvqs; i++) {
+		vq = &vs->vqs[i].vq;
+		vhost_scsi_destroy_vq_cmds(vq);
+	}
+
+	/*
+	 * We can now release our hold on the tpg and sessions and userspace
+	 * can free them after this point.
+	 */
+	for (i = 0; i < VHOST_SCSI_MAX_TARGET; i++) {
+		target = i;
+		tpg = vs->vs_tpg[target];
+		if (!tpg)
+			continue;
+
+		mutex_lock(&tpg->tv_tpg_mutex);
+
 		tpg->tv_tpg_vhost_count--;
 		tpg->vhost_scsi = NULL;
 		vs->vs_tpg[target] = NULL;
-		match = true;
+
 		mutex_unlock(&tpg->tv_tpg_mutex);
-		/*
-		 * Release se_tpg->tpg_group.cg_item configfs dependency now
-		 * to allow vhost-scsi WWPN se_tpg->tpg_group shutdown to occur.
-		 */
+
 		se_tpg = &tpg->se_tpg;
 		target_undepend_item(&se_tpg->tpg_group.cg_item);
 	}
-	if (match) {
-		for (i = 0; i < vs->dev.nvqs; i++) {
-			vq = &vs->vqs[i].vq;
-			mutex_lock(&vq->mutex);
-			vhost_vq_set_backend(vq, NULL);
-			mutex_unlock(&vq->mutex);
-		}
-		/* Make sure cmds are not running before tearing them down. */
-		vhost_scsi_flush(vs);
 
-		for (i = 0; i < vs->dev.nvqs; i++) {
-			vq = &vs->vqs[i].vq;
-			vhost_scsi_destroy_vq_cmds(vq);
-		}
-	}
+free_vs_tpg:
 	/*
 	 * Act as synchronize_rcu to make sure access to
 	 * old vs->vs_tpg is finished.
@@ -1717,14 +1736,10 @@ vhost_scsi_clear_endpoint(struct vhost_scsi *vs,
 	vs->vs_tpg = NULL;
 	WARN_ON(vs->vs_events_nr);
 	mutex_unlock(&vs->dev.mutex);
-	mutex_unlock(&vhost_scsi_mutex);
 	return 0;
 
-err_tpg:
-	mutex_unlock(&tpg->tv_tpg_mutex);
 err_dev:
 	mutex_unlock(&vs->dev.mutex);
-	mutex_unlock(&vhost_scsi_mutex);
 	return ret;
 }
 
@@ -1965,8 +1980,6 @@ vhost_scsi_do_plug(struct vhost_scsi_tpg *tpg,
 	if (!vs)
 		return;
 
-	mutex_lock(&vs->dev.mutex);
-
 	if (plug)
 		reason = VIRTIO_SCSI_EVT_RESET_RESCAN;
 	else
@@ -1974,11 +1987,18 @@ vhost_scsi_do_plug(struct vhost_scsi_tpg *tpg,
 
 	vq = &vs->vqs[VHOST_SCSI_VQ_EVT].vq;
 	mutex_lock(&vq->mutex);
+	/*
+	 * We can't queue events if the backend has been cleared, because
+	 * we could end up queueing an event after the flush.
+	 */
+	if (!vhost_vq_get_backend(vq))
+		goto unlock;
+
 	if (vhost_has_feature(vq, VIRTIO_SCSI_F_HOTPLUG))
 		vhost_scsi_send_evt(vs, tpg, lun,
 				   VIRTIO_SCSI_T_TRANSPORT_RESET, reason);
+unlock:
 	mutex_unlock(&vq->mutex);
-	mutex_unlock(&vs->dev.mutex);
 }
 
 static void vhost_scsi_hotplug(struct vhost_scsi_tpg *tpg, struct se_lun *lun)
@@ -1997,15 +2017,10 @@ static int vhost_scsi_port_link(struct se_portal_group *se_tpg,
 	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
 				struct vhost_scsi_tpg, se_tpg);
 
-	mutex_lock(&vhost_scsi_mutex);
-
 	mutex_lock(&tpg->tv_tpg_mutex);
 	tpg->tv_tpg_port_count++;
-	mutex_unlock(&tpg->tv_tpg_mutex);
-
 	vhost_scsi_hotplug(tpg, lun);
-
-	mutex_unlock(&vhost_scsi_mutex);
+	mutex_unlock(&tpg->tv_tpg_mutex);
 
 	return 0;
 }
@@ -2016,15 +2031,10 @@ static void vhost_scsi_port_unlink(struct se_portal_group *se_tpg,
 	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
 				struct vhost_scsi_tpg, se_tpg);
 
-	mutex_lock(&vhost_scsi_mutex);
-
 	mutex_lock(&tpg->tv_tpg_mutex);
 	tpg->tv_tpg_port_count--;
-	mutex_unlock(&tpg->tv_tpg_mutex);
-
 	vhost_scsi_hotunplug(tpg, lun);
-
-	mutex_unlock(&vhost_scsi_mutex);
+	mutex_unlock(&tpg->tv_tpg_mutex);
 }
 
 static ssize_t vhost_scsi_tpg_attrib_fabric_prot_type_store(

drivers/vhost/vdpa.c

@@ -219,6 +219,28 @@ static int vhost_vdpa_reset(struct vhost_vdpa *v)
 	return vdpa_reset(vdpa);
 }
 
+static long vhost_vdpa_bind_mm(struct vhost_vdpa *v)
+{
+	struct vdpa_device *vdpa = v->vdpa;
+	const struct vdpa_config_ops *ops = vdpa->config;
+
+	if (!vdpa->use_va || !ops->bind_mm)
+		return 0;
+
+	return ops->bind_mm(vdpa, v->vdev.mm);
+}
+
+static void vhost_vdpa_unbind_mm(struct vhost_vdpa *v)
+{
+	struct vdpa_device *vdpa = v->vdpa;
+	const struct vdpa_config_ops *ops = vdpa->config;
+
+	if (!vdpa->use_va || !ops->unbind_mm)
+		return;
+
+	ops->unbind_mm(vdpa);
+}
+
 static long vhost_vdpa_get_device_id(struct vhost_vdpa *v, u8 __user *argp)
 {
 	struct vdpa_device *vdpa = v->vdpa;
@@ -599,9 +621,11 @@ static long vhost_vdpa_vring_ioctl(struct vhost_vdpa *v, unsigned int cmd,
 		if (vq->call_ctx.ctx) {
 			cb.callback = vhost_vdpa_virtqueue_cb;
 			cb.private = vq;
+			cb.trigger = vq->call_ctx.ctx;
 		} else {
 			cb.callback = NULL;
 			cb.private = NULL;
+			cb.trigger = NULL;
 		}
 		ops->set_vq_cb(vdpa, idx, &cb);
 		vhost_vdpa_setup_vq_irq(v, idx);
@@ -716,6 +740,17 @@ static long vhost_vdpa_unlocked_ioctl(struct file *filep,
 		break;
 	}
 
+	if (r)
+		goto out;
+
+	switch (cmd) {
+	case VHOST_SET_OWNER:
+		r = vhost_vdpa_bind_mm(v);
+		if (r)
+			vhost_dev_reset_owner(d, NULL);
+		break;
+	}
+out:
 	mutex_unlock(&d->mutex);
 	return r;
 }
@@ -851,11 +886,7 @@ static void vhost_vdpa_unmap(struct vhost_vdpa *v,
 		if (!v->in_batch)
 			ops->set_map(vdpa, asid, iotlb);
 	}
-	/* If we are in the middle of batch processing, delay the free
-	 * of AS until BATCH_END.
-	 */
-	if (!v->in_batch && !iotlb->nmaps)
-		vhost_vdpa_remove_as(v, asid);
+
 }
 
 static int vhost_vdpa_va_map(struct vhost_vdpa *v,
@@ -1112,8 +1143,6 @@ static int vhost_vdpa_process_iotlb_msg(struct vhost_dev *dev, u32 asid,
 		if (v->in_batch && ops->set_map)
 			ops->set_map(vdpa, asid, iotlb);
 		v->in_batch = false;
-		if (!iotlb->nmaps)
-			vhost_vdpa_remove_as(v, asid);
 		break;
 	default:
 		r = -EINVAL;
@@ -1287,6 +1316,7 @@ static int vhost_vdpa_release(struct inode *inode, struct file *filep)
 	vhost_vdpa_clean_irq(v);
 	vhost_vdpa_reset(v);
 	vhost_dev_stop(&v->vdev);
+	vhost_vdpa_unbind_mm(v);
 	vhost_vdpa_config_put(v);
 	vhost_vdpa_cleanup(v);
 	mutex_unlock(&d->mutex);

drivers/vhost/vhost.c

@@ -434,8 +434,7 @@ static size_t vhost_get_avail_size(struct vhost_virtqueue *vq,
 	size_t event __maybe_unused =
 	       vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
 
-	return sizeof(*vq->avail) +
-	       sizeof(*vq->avail->ring) * num + event;
+	return size_add(struct_size(vq->avail, ring, num), event);
 }
 
 static size_t vhost_get_used_size(struct vhost_virtqueue *vq,
@@ -444,8 +443,7 @@ static size_t vhost_get_used_size(struct vhost_virtqueue *vq,
 	size_t event __maybe_unused =
 	       vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0;
 
-	return sizeof(*vq->used) +
-	       sizeof(*vq->used->ring) * num + event;
+	return size_add(struct_size(vq->used, ring, num), event);
 }
 
 static size_t vhost_get_desc_size(struct vhost_virtqueue *vq,

drivers/vhost/vringh.c

@@ -636,9 +636,9 @@ static inline int xfer_to_user(const struct vringh *vrh,
  * @features: the feature bits for this ring.
  * @num: the number of elements.
  * @weak_barriers: true if we only need memory barriers, not I/O.
- * @desc: the userpace descriptor pointer.
- * @avail: the userpace avail pointer.
- * @used: the userpace used pointer.
+ * @desc: the userspace descriptor pointer.
+ * @avail: the userspace avail pointer.
+ * @used: the userspace used pointer.
  *
  * Returns an error if num is invalid: you should check pointers
  * yourself!
@@ -911,9 +911,9 @@ static inline int kern_xfer(const struct vringh *vrh, void *dst,
  * @features: the feature bits for this ring.
  * @num: the number of elements.
  * @weak_barriers: true if we only need memory barriers, not I/O.
- * @desc: the userpace descriptor pointer.
- * @avail: the userpace avail pointer.
- * @used: the userpace used pointer.
+ * @desc: the userspace descriptor pointer.
+ * @avail: the userspace avail pointer.
+ * @used: the userspace used pointer.
  *
  * Returns an error if num is invalid.
  */
@@ -1094,10 +1094,17 @@ EXPORT_SYMBOL(vringh_need_notify_kern);
 
 #if IS_REACHABLE(CONFIG_VHOST_IOTLB)
 
+struct iotlb_vec {
+	union {
+		struct iovec *iovec;
+		struct bio_vec *bvec;
+	} iov;
+	size_t count;
+};
+
 static int iotlb_translate(const struct vringh *vrh,
 			   u64 addr, u64 len, u64 *translated,
-			   struct bio_vec iov[],
-			   int iov_size, u32 perm)
+			   struct iotlb_vec *ivec, u32 perm)
 {
 	struct vhost_iotlb_map *map;
 	struct vhost_iotlb *iotlb = vrh->iotlb;
@@ -1107,9 +1114,11 @@ static int iotlb_translate(const struct vringh *vrh,
 	spin_lock(vrh->iotlb_lock);
 
 	while (len > s) {
-		u64 size, pa, pfn;
+		uintptr_t io_addr;
+		size_t io_len;
+		u64 size;
 
-		if (unlikely(ret >= iov_size)) {
+		if (unlikely(ret >= ivec->count)) {
 			ret = -ENOBUFS;
 			break;
 		}
@@ -1124,10 +1133,22 @@ static int iotlb_translate(const struct vringh *vrh,
 		}
 
 		size = map->size - addr + map->start;
-		pa = map->addr + addr - map->start;
-		pfn = pa >> PAGE_SHIFT;
-		bvec_set_page(&iov[ret], pfn_to_page(pfn), min(len - s, size),
-			      pa & (PAGE_SIZE - 1));
+		io_len = min(len - s, size);
+		io_addr = map->addr - map->start + addr;
+
+		if (vrh->use_va) {
+			struct iovec *iovec = ivec->iov.iovec;
+
+			iovec[ret].iov_len = io_len;
+			iovec[ret].iov_base = (void __user *)io_addr;
+		} else {
+			u64 pfn = io_addr >> PAGE_SHIFT;
+			struct bio_vec *bvec = ivec->iov.bvec;
+
+			bvec_set_page(&bvec[ret], pfn_to_page(pfn), io_len,
+				      io_addr & (PAGE_SIZE - 1));
+		}
+
 		s += size;
 		addr += size;
 		++ret;
@@ -1141,26 +1162,41 @@ static int iotlb_translate(const struct vringh *vrh,
 	return ret;
 }
 
+#define IOTLB_IOV_STRIDE 16
+
 static inline int copy_from_iotlb(const struct vringh *vrh, void *dst,
 				  void *src, size_t len)
 {
+	struct iotlb_vec ivec;
+	union {
+		struct iovec iovec[IOTLB_IOV_STRIDE];
+		struct bio_vec bvec[IOTLB_IOV_STRIDE];
+	} iov;
 	u64 total_translated = 0;
 
+	ivec.iov.iovec = iov.iovec;
+	ivec.count = IOTLB_IOV_STRIDE;
+
 	while (total_translated < len) {
-		struct bio_vec iov[16];
 		struct iov_iter iter;
 		u64 translated;
 		int ret;
 
 		ret = iotlb_translate(vrh, (u64)(uintptr_t)src,
 				      len - total_translated, &translated,
-				      iov, ARRAY_SIZE(iov), VHOST_MAP_RO);
+				      &ivec, VHOST_MAP_RO);
 		if (ret == -ENOBUFS)
-			ret = ARRAY_SIZE(iov);
+			ret = IOTLB_IOV_STRIDE;
 		else if (ret < 0)
 			return ret;
 
-		iov_iter_bvec(&iter, ITER_SOURCE, iov, ret, translated);
+		if (vrh->use_va) {
+			iov_iter_init(&iter, ITER_SOURCE, ivec.iov.iovec, ret,
+				      translated);
+		} else {
+			iov_iter_bvec(&iter, ITER_SOURCE, ivec.iov.bvec, ret,
+				      translated);
+		}
 
 		ret = copy_from_iter(dst, translated, &iter);
 		if (ret < 0)
@@ -1177,23 +1213,36 @@ static inline int copy_from_iotlb(const struct vringh *vrh, void *dst,
 static inline int copy_to_iotlb(const struct vringh *vrh, void *dst,
 				void *src, size_t len)
 {
+	struct iotlb_vec ivec;
+	union {
+		struct iovec iovec[IOTLB_IOV_STRIDE];
+		struct bio_vec bvec[IOTLB_IOV_STRIDE];
+	} iov;
 	u64 total_translated = 0;
 
+	ivec.iov.iovec = iov.iovec;
+	ivec.count = IOTLB_IOV_STRIDE;
+
 	while (total_translated < len) {
-		struct bio_vec iov[16];
 		struct iov_iter iter;
 		u64 translated;
 		int ret;
 
 		ret = iotlb_translate(vrh, (u64)(uintptr_t)dst,
 				      len - total_translated, &translated,
-				      iov, ARRAY_SIZE(iov), VHOST_MAP_WO);
+				      &ivec, VHOST_MAP_WO);
 		if (ret == -ENOBUFS)
-			ret = ARRAY_SIZE(iov);
+			ret = IOTLB_IOV_STRIDE;
 		else if (ret < 0)
 			return ret;
 
-		iov_iter_bvec(&iter, ITER_DEST, iov, ret, translated);
+		if (vrh->use_va) {
+			iov_iter_init(&iter, ITER_DEST, ivec.iov.iovec, ret,
+				      translated);
+		} else {
+			iov_iter_bvec(&iter, ITER_DEST, ivec.iov.bvec, ret,
+				      translated);
+		}
 
 		ret = copy_to_iter(src, translated, &iter);
 		if (ret < 0)
@@ -1210,20 +1259,36 @@ static inline int copy_to_iotlb(const struct vringh *vrh, void *dst,
 static inline int getu16_iotlb(const struct vringh *vrh,
 			       u16 *val, const __virtio16 *p)
 {
-	struct bio_vec iov;
-	void *kaddr, *from;
+	struct iotlb_vec ivec;
+	union {
+		struct iovec iovec[1];
+		struct bio_vec bvec[1];
+	} iov;
+	__virtio16 tmp;
 	int ret;
 
+	ivec.iov.iovec = iov.iovec;
+	ivec.count = 1;
+
 	/* Atomic read is needed for getu16 */
-	ret = iotlb_translate(vrh, (u64)(uintptr_t)p, sizeof(*p), NULL,
-			      &iov, 1, VHOST_MAP_RO);
+	ret = iotlb_translate(vrh, (u64)(uintptr_t)p, sizeof(*p),
+			      NULL, &ivec, VHOST_MAP_RO);
 	if (ret < 0)
 		return ret;
 
-	kaddr = kmap_atomic(iov.bv_page);
-	from = kaddr + iov.bv_offset;
-	*val = vringh16_to_cpu(vrh, READ_ONCE(*(__virtio16 *)from));
-	kunmap_atomic(kaddr);
+	if (vrh->use_va) {
+		ret = __get_user(tmp, (__virtio16 __user *)ivec.iov.iovec[0].iov_base);
+		if (ret)
+			return ret;
+	} else {
+		void *kaddr = kmap_local_page(ivec.iov.bvec[0].bv_page);
+		void *from = kaddr + ivec.iov.bvec[0].bv_offset;
+
+		tmp = READ_ONCE(*(__virtio16 *)from);
+		kunmap_local(kaddr);
+	}
+
+	*val = vringh16_to_cpu(vrh, tmp);
 
 	return 0;
 }
@@ -1231,20 +1296,36 @@ static inline int getu16_iotlb(const struct vringh *vrh,
 static inline int putu16_iotlb(const struct vringh *vrh,
 			       __virtio16 *p, u16 val)
 {
-	struct bio_vec iov;
-	void *kaddr, *to;
+	struct iotlb_vec ivec;
+	union {
+		struct iovec iovec;
+		struct bio_vec bvec;
+	} iov;
+	__virtio16 tmp;
 	int ret;
 
+	ivec.iov.iovec = &iov.iovec;
+	ivec.count = 1;
+
 	/* Atomic write is needed for putu16 */
-	ret = iotlb_translate(vrh, (u64)(uintptr_t)p, sizeof(*p), NULL,
-			      &iov, 1, VHOST_MAP_WO);
+	ret = iotlb_translate(vrh, (u64)(uintptr_t)p, sizeof(*p),
+			      NULL, &ivec, VHOST_MAP_RO);
 	if (ret < 0)
 		return ret;
 
-	kaddr = kmap_atomic(iov.bv_page);
-	to = kaddr + iov.bv_offset;
-	WRITE_ONCE(*(__virtio16 *)to, cpu_to_vringh16(vrh, val));
-	kunmap_atomic(kaddr);
+	tmp = cpu_to_vringh16(vrh, val);
+
+	if (vrh->use_va) {
+		ret = __put_user(tmp, (__virtio16 __user *)ivec.iov.iovec[0].iov_base);
+		if (ret)
+			return ret;
+	} else {
+		void *kaddr = kmap_local_page(ivec.iov.bvec[0].bv_page);
+		void *to = kaddr + ivec.iov.bvec[0].bv_offset;
+
+		WRITE_ONCE(*(__virtio16 *)to, tmp);
+		kunmap_local(kaddr);
+	}
 
 	return 0;
 }
@@ -1306,9 +1387,9 @@ static inline int putused_iotlb(const struct vringh *vrh,
  * @features: the feature bits for this ring.
  * @num: the number of elements.
  * @weak_barriers: true if we only need memory barriers, not I/O.
- * @desc: the userpace descriptor pointer.
- * @avail: the userpace avail pointer.
- * @used: the userpace used pointer.
+ * @desc: the userspace descriptor pointer.
+ * @avail: the userspace avail pointer.
+ * @used: the userspace used pointer.
  *
  * Returns an error if num is invalid.
  */
@@ -1318,11 +1399,39 @@ int vringh_init_iotlb(struct vringh *vrh, u64 features,
 		      struct vring_avail *avail,
 		      struct vring_used *used)
 {
+	vrh->use_va = false;
+
 	return vringh_init_kern(vrh, features, num, weak_barriers,
 				desc, avail, used);
 }
 EXPORT_SYMBOL(vringh_init_iotlb);
 
+/**
+ * vringh_init_iotlb_va - initialize a vringh for a ring with IOTLB containing
+ *                        user VA.
+ * @vrh: the vringh to initialize.
+ * @features: the feature bits for this ring.
+ * @num: the number of elements.
+ * @weak_barriers: true if we only need memory barriers, not I/O.
+ * @desc: the userspace descriptor pointer.
+ * @avail: the userspace avail pointer.
+ * @used: the userspace used pointer.
+ *
+ * Returns an error if num is invalid.
+ */
+int vringh_init_iotlb_va(struct vringh *vrh, u64 features,
+			 unsigned int num, bool weak_barriers,
+			 struct vring_desc *desc,
+			 struct vring_avail *avail,
+			 struct vring_used *used)
+{
+	vrh->use_va = true;
+
+	return vringh_init_kern(vrh, features, num, weak_barriers,
+				desc, avail, used);
+}
+EXPORT_SYMBOL(vringh_init_iotlb_va);
+
 /**
  * vringh_set_iotlb - initialize a vringh for a ring with IOTLB.
  * @vrh: the vring

drivers/virtio/virtio_mmio.c

@@ -286,6 +286,16 @@ static bool vm_notify(struct virtqueue *vq)
 	return true;
 }
 
+static bool vm_notify_with_data(struct virtqueue *vq)
+{
+	struct virtio_mmio_device *vm_dev = to_virtio_mmio_device(vq->vdev);
+	u32 data = vring_notification_data(vq);
+
+	writel(data, vm_dev->base + VIRTIO_MMIO_QUEUE_NOTIFY);
+
+	return true;
+}
+
 /* Notify all virtqueues on an interrupt. */
 static irqreturn_t vm_interrupt(int irq, void *opaque)
 {
@@ -364,12 +374,18 @@ static struct virtqueue *vm_setup_vq(struct virtio_device *vdev, unsigned int in
 				  const char *name, bool ctx)
 {
 	struct virtio_mmio_device *vm_dev = to_virtio_mmio_device(vdev);
+	bool (*notify)(struct virtqueue *vq);
 	struct virtio_mmio_vq_info *info;
 	struct virtqueue *vq;
 	unsigned long flags;
 	unsigned int num;
 	int err;
 
+	if (__virtio_test_bit(vdev, VIRTIO_F_NOTIFICATION_DATA))
+		notify = vm_notify_with_data;
+	else
+		notify = vm_notify;
+
 	if (!name)
 		return NULL;
 
@@ -398,7 +414,7 @@ static struct virtqueue *vm_setup_vq(struct virtio_device *vdev, unsigned int in
 
 	/* Create the vring */
 	vq = vring_create_virtqueue(index, num, VIRTIO_MMIO_VRING_ALIGN, vdev,
-				 true, true, ctx, vm_notify, callback, name);
+				 true, true, ctx, notify, callback, name);
 	if (!vq) {
 		err = -ENOMEM;
 		goto error_new_virtqueue;

drivers/virtio/virtio_pci_modern.c

@@ -288,6 +288,15 @@ static u16 vp_config_vector(struct virtio_pci_device *vp_dev, u16 vector)
 	return vp_modern_config_vector(&vp_dev->mdev, vector);
 }
 
+static bool vp_notify_with_data(struct virtqueue *vq)
+{
+	u32 data = vring_notification_data(vq);
+
+	iowrite32(data, (void __iomem *)vq->priv);
+
+	return true;
+}
+
 static struct virtqueue *setup_vq(struct virtio_pci_device *vp_dev,
 				  struct virtio_pci_vq_info *info,
 				  unsigned int index,
@@ -298,10 +307,16 @@ static struct virtqueue *setup_vq(struct virtio_pci_device *vp_dev,
 {
 
 	struct virtio_pci_modern_device *mdev = &vp_dev->mdev;
+	bool (*notify)(struct virtqueue *vq);
 	struct virtqueue *vq;
 	u16 num;
 	int err;
 
+	if (__virtio_test_bit(&vp_dev->vdev, VIRTIO_F_NOTIFICATION_DATA))
+		notify = vp_notify_with_data;
+	else
+		notify = vp_notify;
+
 	if (index >= vp_modern_get_num_queues(mdev))
 		return ERR_PTR(-EINVAL);
 
@@ -310,18 +325,13 @@ static struct virtqueue *setup_vq(struct virtio_pci_device *vp_dev,
 	if (!num || vp_modern_get_queue_enable(mdev, index))
 		return ERR_PTR(-ENOENT);
 
-	if (!is_power_of_2(num)) {
-		dev_warn(&vp_dev->pci_dev->dev, "bad queue size %u", num);
-		return ERR_PTR(-EINVAL);
-	}
-
 	info->msix_vector = msix_vec;
 
 	/* create the vring */
 	vq = vring_create_virtqueue(index, num,
 				    SMP_CACHE_BYTES, &vp_dev->vdev,
 				    true, true, ctx,
-				    vp_notify, callback, name);
+				    notify, callback, name);
 	if (!vq)
 		return ERR_PTR(-ENOMEM);
 

drivers/virtio/virtio_ring.c

@@ -231,10 +231,10 @@ static void vring_free(struct virtqueue *_vq);
  * Helpers.
  */
 
-#define to_vvq(_vq) container_of(_vq, struct vring_virtqueue, vq)
+#define to_vvq(_vq) container_of_const(_vq, struct vring_virtqueue, vq)
 
-static inline bool virtqueue_use_indirect(struct vring_virtqueue *vq,
-					  unsigned int total_sg)
+static bool virtqueue_use_indirect(const struct vring_virtqueue *vq,
+				   unsigned int total_sg)
 {
 	/*
 	 * If the host supports indirect descriptor tables, and we have multiple
@@ -269,7 +269,7 @@ static inline bool virtqueue_use_indirect(struct vring_virtqueue *vq,
  * unconditionally on data path.
  */
 
-static bool vring_use_dma_api(struct virtio_device *vdev)
+static bool vring_use_dma_api(const struct virtio_device *vdev)
 {
 	if (!virtio_has_dma_quirk(vdev))
 		return true;
@@ -289,7 +289,7 @@ static bool vring_use_dma_api(struct virtio_device *vdev)
 	return false;
 }
 
-size_t virtio_max_dma_size(struct virtio_device *vdev)
+size_t virtio_max_dma_size(const struct virtio_device *vdev)
 {
 	size_t max_segment_size = SIZE_MAX;
 
@@ -349,7 +349,7 @@ static void vring_free_queue(struct virtio_device *vdev, size_t size,
  * making all of the arch DMA ops work on the vring device itself
  * is a mess.
  */
-static inline struct device *vring_dma_dev(const struct vring_virtqueue *vq)
+static struct device *vring_dma_dev(const struct vring_virtqueue *vq)
 {
 	return vq->dma_dev;
 }
@@ -423,7 +423,7 @@ static void virtqueue_init(struct vring_virtqueue *vq, u32 num)
  */
 
 static void vring_unmap_one_split_indirect(const struct vring_virtqueue *vq,
-					   struct vring_desc *desc)
+					   const struct vring_desc *desc)
 {
 	u16 flags;
 
@@ -784,7 +784,7 @@ static void detach_buf_split(struct vring_virtqueue *vq, unsigned int head,
 	}
 }
 
-static inline bool more_used_split(const struct vring_virtqueue *vq)
+static bool more_used_split(const struct vring_virtqueue *vq)
 {
 	return vq->last_used_idx != virtio16_to_cpu(vq->vq.vdev,
 			vq->split.vring.used->idx);
@@ -854,6 +854,14 @@ static void virtqueue_disable_cb_split(struct virtqueue *_vq)
 
 	if (!(vq->split.avail_flags_shadow & VRING_AVAIL_F_NO_INTERRUPT)) {
 		vq->split.avail_flags_shadow |= VRING_AVAIL_F_NO_INTERRUPT;
+
+		/*
+		 * If device triggered an event already it won't trigger one again:
+		 * no need to disable.
+		 */
+		if (vq->event_triggered)
+			return;
+
 		if (vq->event)
 			/* TODO: this is a hack. Figure out a cleaner value to write. */
 			vring_used_event(&vq->split.vring) = 0x0;
@@ -1172,18 +1180,18 @@ static int virtqueue_resize_split(struct virtqueue *_vq, u32 num)
 /*
  * Packed ring specific functions - *_packed().
  */
-static inline bool packed_used_wrap_counter(u16 last_used_idx)
+static bool packed_used_wrap_counter(u16 last_used_idx)
 {
 	return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
 }
 
-static inline u16 packed_last_used(u16 last_used_idx)
+static u16 packed_last_used(u16 last_used_idx)
 {
 	return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
 }
 
 static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
-				     struct vring_desc_extra *extra)
+				     const struct vring_desc_extra *extra)
 {
 	u16 flags;
 
@@ -1206,7 +1214,7 @@ static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
 }
 
 static void vring_unmap_desc_packed(const struct vring_virtqueue *vq,
-				   struct vring_packed_desc *desc)
+				    const struct vring_packed_desc *desc)
 {
 	u16 flags;
 
@@ -1612,7 +1620,7 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
 	return avail == used && used == used_wrap_counter;
 }
 
-static inline bool more_used_packed(const struct vring_virtqueue *vq)
+static bool more_used_packed(const struct vring_virtqueue *vq)
 {
 	u16 last_used;
 	u16 last_used_idx;
@@ -1699,6 +1707,14 @@ static void virtqueue_disable_cb_packed(struct virtqueue *_vq)
 
 	if (vq->packed.event_flags_shadow != VRING_PACKED_EVENT_FLAG_DISABLE) {
 		vq->packed.event_flags_shadow = VRING_PACKED_EVENT_FLAG_DISABLE;
+
+		/*
+		 * If device triggered an event already it won't trigger one again:
+		 * no need to disable.
+		 */
+		if (vq->event_triggered)
+			return;
+
 		vq->packed.vring.driver->flags =
 			cpu_to_le16(vq->packed.event_flags_shadow);
 	}
@@ -2330,12 +2346,6 @@ void virtqueue_disable_cb(struct virtqueue *_vq)
 {
 	struct vring_virtqueue *vq = to_vvq(_vq);
 
-	/* If device triggered an event already it won't trigger one again:
-	 * no need to disable.
-	 */
-	if (vq->event_triggered)
-		return;
-
 	if (vq->packed_ring)
 		virtqueue_disable_cb_packed(_vq);
 	else
@@ -2752,6 +2762,23 @@ void vring_del_virtqueue(struct virtqueue *_vq)
 }
 EXPORT_SYMBOL_GPL(vring_del_virtqueue);
 
+u32 vring_notification_data(struct virtqueue *_vq)
+{
+	struct vring_virtqueue *vq = to_vvq(_vq);
+	u16 next;
+
+	if (vq->packed_ring)
+		next = (vq->packed.next_avail_idx &
+				~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))) |
+			vq->packed.avail_wrap_counter <<
+				VRING_PACKED_EVENT_F_WRAP_CTR;
+	else
+		next = vq->split.avail_idx_shadow;
+
+	return next << 16 | _vq->index;
+}
+EXPORT_SYMBOL_GPL(vring_notification_data);
+
 /* Manipulates transport-specific feature bits. */
 void vring_transport_features(struct virtio_device *vdev)
 {
@@ -2771,6 +2798,8 @@ void vring_transport_features(struct virtio_device *vdev)
 			break;
 		case VIRTIO_F_ORDER_PLATFORM:
 			break;
+		case VIRTIO_F_NOTIFICATION_DATA:
+			break;
 		default:
 			/* We don't understand this bit. */
 			__virtio_clear_bit(vdev, i);
@@ -2786,10 +2815,10 @@ EXPORT_SYMBOL_GPL(vring_transport_features);
  * Returns the size of the vring.  This is mainly used for boasting to
  * userspace.  Unlike other operations, this need not be serialized.
  */
-unsigned int virtqueue_get_vring_size(struct virtqueue *_vq)
+unsigned int virtqueue_get_vring_size(const struct virtqueue *_vq)
 {
 
-	struct vring_virtqueue *vq = to_vvq(_vq);
+	const struct vring_virtqueue *vq = to_vvq(_vq);
 
 	return vq->packed_ring ? vq->packed.vring.num : vq->split.vring.num;
 }
@@ -2819,9 +2848,9 @@ void __virtqueue_unbreak(struct virtqueue *_vq)
 }
 EXPORT_SYMBOL_GPL(__virtqueue_unbreak);
 
-bool virtqueue_is_broken(struct virtqueue *_vq)
+bool virtqueue_is_broken(const struct virtqueue *_vq)
 {
-	struct vring_virtqueue *vq = to_vvq(_vq);
+	const struct vring_virtqueue *vq = to_vvq(_vq);
 
 	return READ_ONCE(vq->broken);
 }
@@ -2868,9 +2897,9 @@ void __virtio_unbreak_device(struct virtio_device *dev)
 }
 EXPORT_SYMBOL_GPL(__virtio_unbreak_device);
 
-dma_addr_t virtqueue_get_desc_addr(struct virtqueue *_vq)
+dma_addr_t virtqueue_get_desc_addr(const struct virtqueue *_vq)
 {
-	struct vring_virtqueue *vq = to_vvq(_vq);
+	const struct vring_virtqueue *vq = to_vvq(_vq);
 
 	BUG_ON(!vq->we_own_ring);
 
@@ -2881,9 +2910,9 @@ dma_addr_t virtqueue_get_desc_addr(struct virtqueue *_vq)
 }
 EXPORT_SYMBOL_GPL(virtqueue_get_desc_addr);
 
-dma_addr_t virtqueue_get_avail_addr(struct virtqueue *_vq)
+dma_addr_t virtqueue_get_avail_addr(const struct virtqueue *_vq)
 {
-	struct vring_virtqueue *vq = to_vvq(_vq);
+	const struct vring_virtqueue *vq = to_vvq(_vq);
 
 	BUG_ON(!vq->we_own_ring);
 
@@ -2895,9 +2924,9 @@ dma_addr_t virtqueue_get_avail_addr(struct virtqueue *_vq)
 }
 EXPORT_SYMBOL_GPL(virtqueue_get_avail_addr);
 
-dma_addr_t virtqueue_get_used_addr(struct virtqueue *_vq)
+dma_addr_t virtqueue_get_used_addr(const struct virtqueue *_vq)
 {
-	struct vring_virtqueue *vq = to_vvq(_vq);
+	const struct vring_virtqueue *vq = to_vvq(_vq);
 
 	BUG_ON(!vq->we_own_ring);
 
@@ -2910,7 +2939,7 @@ dma_addr_t virtqueue_get_used_addr(struct virtqueue *_vq)
 EXPORT_SYMBOL_GPL(virtqueue_get_used_addr);
 
 /* Only available for split ring */
-const struct vring *virtqueue_get_vring(struct virtqueue *vq)
+const struct vring *virtqueue_get_vring(const struct virtqueue *vq)
 {
 	return &to_vvq(vq)->split.vring;
 }

drivers/virtio/virtio_vdpa.c

@@ -13,6 +13,7 @@
 #include <linux/kernel.h>
 #include <linux/slab.h>
 #include <linux/uuid.h>
+#include <linux/group_cpus.h>
 #include <linux/virtio.h>
 #include <linux/vdpa.h>
 #include <linux/virtio_config.h>
@@ -112,6 +113,17 @@ static bool virtio_vdpa_notify(struct virtqueue *vq)
 	return true;
 }
 
+static bool virtio_vdpa_notify_with_data(struct virtqueue *vq)
+{
+	struct vdpa_device *vdpa = vd_get_vdpa(vq->vdev);
+	const struct vdpa_config_ops *ops = vdpa->config;
+	u32 data = vring_notification_data(vq);
+
+	ops->kick_vq_with_data(vdpa, data);
+
+	return true;
+}
+
 static irqreturn_t virtio_vdpa_config_cb(void *private)
 {
 	struct virtio_vdpa_device *vd_dev = private;
@@ -138,6 +150,7 @@ virtio_vdpa_setup_vq(struct virtio_device *vdev, unsigned int index,
 	struct device *dma_dev;
 	const struct vdpa_config_ops *ops = vdpa->config;
 	struct virtio_vdpa_vq_info *info;
+	bool (*notify)(struct virtqueue *vq) = virtio_vdpa_notify;
 	struct vdpa_callback cb;
 	struct virtqueue *vq;
 	u64 desc_addr, driver_addr, device_addr;
@@ -154,6 +167,14 @@ virtio_vdpa_setup_vq(struct virtio_device *vdev, unsigned int index,
 	if (index >= vdpa->nvqs)
 		return ERR_PTR(-ENOENT);
 
+	/* We cannot accept VIRTIO_F_NOTIFICATION_DATA without kick_vq_with_data */
+	if (__virtio_test_bit(vdev, VIRTIO_F_NOTIFICATION_DATA)) {
+		if (ops->kick_vq_with_data)
+			notify = virtio_vdpa_notify_with_data;
+		else
+			__virtio_clear_bit(vdev, VIRTIO_F_NOTIFICATION_DATA);
+	}
+
 	/* Queue shouldn't already be set up. */
 	if (ops->get_vq_ready(vdpa, index))
 		return ERR_PTR(-ENOENT);
@@ -183,8 +204,7 @@ virtio_vdpa_setup_vq(struct virtio_device *vdev, unsigned int index,
 		dma_dev = vdpa_get_dma_dev(vdpa);
 	vq = vring_create_virtqueue_dma(index, max_num, align, vdev,
 					true, may_reduce_num, ctx,
-					virtio_vdpa_notify, callback,
-					name, dma_dev);
+					notify, callback, name, dma_dev);
 	if (!vq) {
 		err = -ENOMEM;
 		goto error_new_virtqueue;
@@ -195,6 +215,7 @@ virtio_vdpa_setup_vq(struct virtio_device *vdev, unsigned int index,
 	/* Setup virtqueue callback */
 	cb.callback = callback ? virtio_vdpa_virtqueue_cb : NULL;
 	cb.private = info;
+	cb.trigger = NULL;
 	ops->set_vq_cb(vdpa, index, &cb);
 	ops->set_vq_num(vdpa, index, virtqueue_get_vring_size(vq));
 
@@ -272,6 +293,66 @@ static void virtio_vdpa_del_vqs(struct virtio_device *vdev)
 		virtio_vdpa_del_vq(vq);
 }
 
+static void default_calc_sets(struct irq_affinity *affd, unsigned int affvecs)
+{
+	affd->nr_sets = 1;
+	affd->set_size[0] = affvecs;
+}
+
+static struct cpumask *
+create_affinity_masks(unsigned int nvecs, struct irq_affinity *affd)
+{
+	unsigned int affvecs = 0, curvec, usedvecs, i;
+	struct cpumask *masks = NULL;
+
+	if (nvecs > affd->pre_vectors + affd->post_vectors)
+		affvecs = nvecs - affd->pre_vectors - affd->post_vectors;
+
+	if (!affd->calc_sets)
+		affd->calc_sets = default_calc_sets;
+
+	affd->calc_sets(affd, affvecs);
+
+	if (!affvecs)
+		return NULL;
+
+	masks = kcalloc(nvecs, sizeof(*masks), GFP_KERNEL);
+	if (!masks)
+		return NULL;
+
+	/* Fill out vectors at the beginning that don't need affinity */
+	for (curvec = 0; curvec < affd->pre_vectors; curvec++)
+		cpumask_setall(&masks[curvec]);
+
+	for (i = 0, usedvecs = 0; i < affd->nr_sets; i++) {
+		unsigned int this_vecs = affd->set_size[i];
+		int j;
+		struct cpumask *result = group_cpus_evenly(this_vecs);
+
+		if (!result) {
+			kfree(masks);
+			return NULL;
+		}
+
+		for (j = 0; j < this_vecs; j++)
+			cpumask_copy(&masks[curvec + j], &result[j]);
+		kfree(result);
+
+		curvec += this_vecs;
+		usedvecs += this_vecs;
+	}
+
+	/* Fill out vectors at the end that don't need affinity */
+	if (usedvecs >= affvecs)
+		curvec = affd->pre_vectors + affvecs;
+	else
+		curvec = affd->pre_vectors + usedvecs;
+	for (; curvec < nvecs; curvec++)
+		cpumask_setall(&masks[curvec]);
+
+	return masks;
+}
+
 static int virtio_vdpa_find_vqs(struct virtio_device *vdev, unsigned int nvqs,
 				struct virtqueue *vqs[],
 				vq_callback_t *callbacks[],
@@ -282,9 +363,15 @@ static int virtio_vdpa_find_vqs(struct virtio_device *vdev, unsigned int nvqs,
 	struct virtio_vdpa_device *vd_dev = to_virtio_vdpa_device(vdev);
 	struct vdpa_device *vdpa = vd_get_vdpa(vdev);
 	const struct vdpa_config_ops *ops = vdpa->config;
+	struct irq_affinity default_affd = { 0 };
+	struct cpumask *masks;
 	struct vdpa_callback cb;
 	int i, err, queue_idx = 0;
 
+	masks = create_affinity_masks(nvqs, desc ? desc : &default_affd);
+	if (!masks)
+		return -ENOMEM;
+
 	for (i = 0; i < nvqs; ++i) {
 		if (!names[i]) {
 			vqs[i] = NULL;
@@ -298,6 +385,7 @@ static int virtio_vdpa_find_vqs(struct virtio_device *vdev, unsigned int nvqs,
 			err = PTR_ERR(vqs[i]);
 			goto err_setup_vq;
 		}
+		ops->set_vq_affinity(vdpa, i, &masks[i]);
 	}
 
 	cb.callback = virtio_vdpa_config_cb;
@@ -337,6 +425,32 @@ static const char *virtio_vdpa_bus_name(struct virtio_device *vdev)
 	return dev_name(&vdpa->dev);
 }
 
+static int virtio_vdpa_set_vq_affinity(struct virtqueue *vq,
+				       const struct cpumask *cpu_mask)
+{
+	struct virtio_vdpa_device *vd_dev = to_virtio_vdpa_device(vq->vdev);
+	struct vdpa_device *vdpa = vd_dev->vdpa;
+	const struct vdpa_config_ops *ops = vdpa->config;
+	unsigned int index = vq->index;
+
+	if (ops->set_vq_affinity)
+		return ops->set_vq_affinity(vdpa, index, cpu_mask);
+
+	return 0;
+}
+
+static const struct cpumask *
+virtio_vdpa_get_vq_affinity(struct virtio_device *vdev, int index)
+{
+	struct vdpa_device *vdpa = vd_get_vdpa(vdev);
+	const struct vdpa_config_ops *ops = vdpa->config;
+
+	if (ops->get_vq_affinity)
+		return ops->get_vq_affinity(vdpa, index);
+
+	return NULL;
+}
+
 static const struct virtio_config_ops virtio_vdpa_config_ops = {
 	.get		= virtio_vdpa_get,
 	.set		= virtio_vdpa_set,
@@ -349,6 +463,8 @@ static const struct virtio_config_ops virtio_vdpa_config_ops = {
 	.get_features	= virtio_vdpa_get_features,
 	.finalize_features = virtio_vdpa_finalize_features,
 	.bus_name	= virtio_vdpa_bus_name,
+	.set_vq_affinity = virtio_vdpa_set_vq_affinity,
+	.get_vq_affinity = virtio_vdpa_get_vq_affinity,
 };
 
 static void virtio_vdpa_release_dev(struct device *_d)

include/linux/vdpa.h

@@ -10,13 +10,19 @@
 #include <linux/if_ether.h>
 
 /**
- * struct vdpa_calllback - vDPA callback definition.
+ * struct vdpa_callback - vDPA callback definition.
  * @callback: interrupt callback function
  * @private: the data passed to the callback function
+ * @trigger: the eventfd for the callback (Optional).
+ *           When it is set, the vDPA driver must guarantee that
+ *           signaling it is functional equivalent to triggering
+ *           the callback. Then vDPA parent can signal it directly
+ *           instead of triggering the callback.
  */
 struct vdpa_callback {
 	irqreturn_t (*callback)(void *data);
 	void *private;
+	struct eventfd_ctx *trigger;
 };
 
 /**
@@ -114,7 +120,7 @@ struct vdpa_dev_set_config {
 };
 
 /**
- * Corresponding file area for device memory mapping
+ * struct vdpa_map_file - file area for device memory mapping
  * @file: vma->vm_file for the mapping
  * @offset: mapping offset in the vm_file
  */
@@ -143,6 +149,14 @@ struct vdpa_map_file {
  * @kick_vq:			Kick the virtqueue
  *				@vdev: vdpa device
  *				@idx: virtqueue index
+ * @kick_vq_with_data:		Kick the virtqueue and supply extra data
+ *				(only if VIRTIO_F_NOTIFICATION_DATA is negotiated)
+ *				@vdev: vdpa device
+ *				@data for split virtqueue:
+ *				16 bits vqn and 16 bits next available index.
+ *				@data for packed virtqueue:
+ *				16 bits vqn, 15 least significant bits of
+ *				next available index and 1 bit next_wrap.
  * @set_vq_cb:			Set the interrupt callback function for
  *				a virtqueue
  *				@vdev: vdpa device
@@ -165,10 +179,16 @@ struct vdpa_map_file {
  *				@vdev: vdpa device
  *				@idx: virtqueue index
  *				@state: pointer to returned state (last_avail_idx)
+ * @get_vendor_vq_stats:	Get the vendor statistics of a device.
+ *				@vdev: vdpa device
+ *				@idx: virtqueue index
+ *				@msg: socket buffer holding stats message
+ *				@extack: extack for reporting error messages
+ *				Returns integer: success (0) or error (< 0)
  * @get_vq_notification:	Get the notification area for a virtqueue (optional)
  *				@vdev: vdpa device
  *				@idx: virtqueue index
- *				Returns the notifcation area
+ *				Returns the notification area
  * @get_vq_irq:			Get the irq number of a virtqueue (optional,
  *				but must implemented if require vq irq offloading)
  *				@vdev: vdpa device
@@ -250,6 +270,15 @@ struct vdpa_map_file {
  *				@vdev: vdpa device
  *				Returns the iova range supported by
  *				the device.
+ * @set_vq_affinity:		Set the affinity of virtqueue (optional)
+ *				@vdev: vdpa device
+ *				@idx: virtqueue index
+ *				@cpu_mask: the affinity mask
+ *				Returns integer: success (0) or error (< 0)
+ * @get_vq_affinity:		Get the affinity of virtqueue (optional)
+ *				@vdev: vdpa device
+ *				@idx: virtqueue index
+ *				Returns the affinity mask
  * @set_group_asid:		Set address space identifier for a
  *				virtqueue group (optional)
  *				@vdev: vdpa device
@@ -290,6 +319,14 @@ struct vdpa_map_file {
  *				@vdev: vdpa device
  *				@idx: virtqueue index
  *				Returns pointer to structure device or error (NULL)
+ * @bind_mm:			Bind the device to a specific address space
+ *				so the vDPA framework can use VA when this
+ *				callback is implemented. (optional)
+ *				@vdev: vdpa device
+ *				@mm: address space to bind
+ * @unbind_mm:			Unbind the device from the address space
+ *				bound using the bind_mm callback. (optional)
+ *				@vdev: vdpa device
  * @free:			Free resources that belongs to vDPA (optional)
  *				@vdev: vdpa device
  */
@@ -300,6 +337,7 @@ struct vdpa_config_ops {
 			      u64 device_area);
 	void (*set_vq_num)(struct vdpa_device *vdev, u16 idx, u32 num);
 	void (*kick_vq)(struct vdpa_device *vdev, u16 idx);
+	void (*kick_vq_with_data)(struct vdpa_device *vdev, u32 data);
 	void (*set_vq_cb)(struct vdpa_device *vdev, u16 idx,
 			  struct vdpa_callback *cb);
 	void (*set_vq_ready)(struct vdpa_device *vdev, u16 idx, bool ready);
@@ -340,6 +378,10 @@ struct vdpa_config_ops {
 			   const void *buf, unsigned int len);
 	u32 (*get_generation)(struct vdpa_device *vdev);
 	struct vdpa_iova_range (*get_iova_range)(struct vdpa_device *vdev);
+	int (*set_vq_affinity)(struct vdpa_device *vdev, u16 idx,
+			       const struct cpumask *cpu_mask);
+	const struct cpumask *(*get_vq_affinity)(struct vdpa_device *vdev,
+						 u16 idx);
 
 	/* DMA ops */
 	int (*set_map)(struct vdpa_device *vdev, unsigned int asid,
@@ -351,6 +393,8 @@ struct vdpa_config_ops {
 	int (*set_group_asid)(struct vdpa_device *vdev, unsigned int group,
 			      unsigned int asid);
 	struct device *(*get_vq_dma_dev)(struct vdpa_device *vdev, u16 idx);
+	int (*bind_mm)(struct vdpa_device *vdev, struct mm_struct *mm);
+	void (*unbind_mm)(struct vdpa_device *vdev);
 
 	/* Free device resources */
 	void (*free)(struct vdpa_device *vdev);
@@ -506,6 +550,8 @@ struct vdpa_mgmtdev_ops {
  * @config_attr_mask: bit mask of attributes of type enum vdpa_attr that
  *		      management device support during dev_add callback
  * @list: list entry
+ * @supported_features: features supported by device
+ * @max_supported_vqs: maximum number of virtqueues supported by device
  */
 struct vdpa_mgmt_dev {
 	struct device *device;

include/linux/virtio.h

@@ -34,8 +34,8 @@ struct virtqueue {
 	unsigned int index;
 	unsigned int num_free;
 	unsigned int num_max;
-	void *priv;
 	bool reset;
+	void *priv;
 };
 
 int virtqueue_add_outbuf(struct virtqueue *vq,
@@ -84,14 +84,14 @@ bool virtqueue_enable_cb_delayed(struct virtqueue *vq);
 
 void *virtqueue_detach_unused_buf(struct virtqueue *vq);
 
-unsigned int virtqueue_get_vring_size(struct virtqueue *vq);
+unsigned int virtqueue_get_vring_size(const struct virtqueue *vq);
 
-bool virtqueue_is_broken(struct virtqueue *vq);
+bool virtqueue_is_broken(const struct virtqueue *vq);
 
-const struct vring *virtqueue_get_vring(struct virtqueue *vq);
-dma_addr_t virtqueue_get_desc_addr(struct virtqueue *vq);
-dma_addr_t virtqueue_get_avail_addr(struct virtqueue *vq);
-dma_addr_t virtqueue_get_used_addr(struct virtqueue *vq);
+const struct vring *virtqueue_get_vring(const struct virtqueue *vq);
+dma_addr_t virtqueue_get_desc_addr(const struct virtqueue *vq);
+dma_addr_t virtqueue_get_avail_addr(const struct virtqueue *vq);
+dma_addr_t virtqueue_get_used_addr(const struct virtqueue *vq);
 
 int virtqueue_resize(struct virtqueue *vq, u32 num,
 		     void (*recycle)(struct virtqueue *vq, void *buf));
@@ -147,7 +147,7 @@ int virtio_device_restore(struct virtio_device *dev);
 #endif
 void virtio_reset_device(struct virtio_device *dev);
 
-size_t virtio_max_dma_size(struct virtio_device *vdev);
+size_t virtio_max_dma_size(const struct virtio_device *vdev);
 
 #define virtio_device_for_each_vq(vdev, vq) \
 	list_for_each_entry(vq, &vdev->vqs, list)

include/linux/virtio_ring.h

@@ -58,6 +58,7 @@ do { \
 
 struct virtio_device;
 struct virtqueue;
+struct device;
 
 /*
  * Creates a virtqueue and allocates the descriptor ring.  If
@@ -117,4 +118,6 @@ void vring_del_virtqueue(struct virtqueue *vq);
 void vring_transport_features(struct virtio_device *vdev);
 
 irqreturn_t vring_interrupt(int irq, void *_vq);
+
+u32 vring_notification_data(struct virtqueue *_vq);
 #endif /* _LINUX_VIRTIO_RING_H */

include/linux/vringh.h

@@ -32,6 +32,9 @@ struct vringh {
 	/* Can we get away with weak barriers? */
 	bool weak_barriers;
 
+	/* Use user's VA */
+	bool use_va;
+
 	/* Last available index we saw (ie. where we're up to). */
 	u16 last_avail_idx;
 
@@ -54,6 +57,9 @@ struct vringh {
 	void (*notify)(struct vringh *);
 };
 
+struct virtio_device;
+typedef void vrh_callback_t(struct virtio_device *, struct vringh *);
+
 /**
  * struct vringh_config_ops - ops for creating a host vring from a virtio driver
  * @find_vrhs: find the host vrings and instantiate them
@@ -65,8 +71,6 @@ struct vringh {
  *	Returns 0 on success or error status
  * @del_vrhs: free the host vrings found by find_vrhs().
  */
-struct virtio_device;
-typedef void vrh_callback_t(struct virtio_device *, struct vringh *);
 struct vringh_config_ops {
 	int (*find_vrhs)(struct virtio_device *vdev, unsigned nhvrs,
 			 struct vringh *vrhs[], vrh_callback_t *callbacks[]);
@@ -81,6 +85,12 @@ struct vringh_range {
 
 /**
  * struct vringh_iov - iovec mangler.
+ * @iov: array of iovecs to operate on
+ * @consumed: number of bytes consumed within iov[i]
+ * @i: index of current iovec
+ * @used: number of iovecs present in @iov
+ * @max_num: maximum number of iovecs.
+ *           corresponds to allocated memory of @iov
  *
  * Mangles iovec in place, and restores it.
  * Remaining data is iov + i, of used - i elements.
@@ -93,6 +103,12 @@ struct vringh_iov {
 
 /**
  * struct vringh_kiov - kvec mangler.
+ * @iov: array of iovecs to operate on
+ * @consumed: number of bytes consumed within iov[i]
+ * @i: index of current iovec
+ * @used: number of iovecs present in @iov
+ * @max_num: maximum number of iovecs.
+ *           corresponds to allocated memory of @iov
  *
  * Mangles kvec in place, and restores it.
  * Remaining data is iov + i, of used - i elements.
@@ -284,6 +300,12 @@ int vringh_init_iotlb(struct vringh *vrh, u64 features,
 		      struct vring_avail *avail,
 		      struct vring_used *used);
 
+int vringh_init_iotlb_va(struct vringh *vrh, u64 features,
+			 unsigned int num, bool weak_barriers,
+			 struct vring_desc *desc,
+			 struct vring_avail *avail,
+			 struct vring_used *used);
+
 int vringh_getdesc_iotlb(struct vringh *vrh,
 			 struct vringh_kiov *riov,
 			 struct vringh_kiov *wiov,

include/uapi/linux/virtio_config.h

@@ -99,6 +99,12 @@
  */
 #define VIRTIO_F_SR_IOV			37
 
+/*
+ * This feature indicates that the driver passes extra data (besides
+ * identifying the virtqueue) in its device notifications.
+ */
+#define VIRTIO_F_NOTIFICATION_DATA	38
+
 /*
  * This feature indicates that the driver can reset a queue individually.
  */

lib/group_cpus.c

@@ -426,3 +426,4 @@ struct cpumask *group_cpus_evenly(unsigned int numgrps)
 	return masks;
 }
 #endif /* CONFIG_SMP */
+EXPORT_SYMBOL_GPL(group_cpus_evenly);

tools/include/linux/types.h

@@ -49,7 +49,12 @@ typedef __s8  s8;
 #endif
 
 #define __force
+/* This is defined in linux/compiler_types.h and is left for backward
+ * compatibility.
+ */
+#ifndef __user
 #define __user
+#endif
 #define __must_check
 #define __cold
 

tools/virtio/linux/compiler.h

@@ -2,6 +2,8 @@
 #ifndef LINUX_COMPILER_H
 #define LINUX_COMPILER_H
 
+#include "../../../include/linux/compiler_types.h"
+
 #define WRITE_ONCE(var, val) \
 	(*((volatile typeof(val) *)(&(var))) = (val))
 

tools/virtio/linux/kernel.h

@@ -10,6 +10,7 @@
 #include <stdarg.h>
 
 #include <linux/compiler.h>
+#include "../../../include/linux/container_of.h"
 #include <linux/log2.h>
 #include <linux/types.h>
 #include <linux/overflow.h>
@@ -107,10 +108,6 @@ static inline void free_page(unsigned long addr)
 	free((void *)addr);
 }
 
-#define container_of(ptr, type, member) ({			\
-	const typeof( ((type *)0)->member ) *__mptr = (ptr);	\
-	(type *)( (char *)__mptr - offsetof(type,member) );})
-
 # ifndef likely
 #  define likely(x)	(__builtin_expect(!!(x), 1))
 # endif

tools/virtio/linux/uaccess.h

@@ -6,15 +6,10 @@
 
 extern void *__user_addr_min, *__user_addr_max;
 
-static inline void __chk_user_ptr(const volatile void *p, size_t size)
-{
-	assert(p >= __user_addr_min && p + size <= __user_addr_max);
-}
-
 #define put_user(x, ptr)					\
 ({								\
 	typeof(ptr) __pu_ptr = (ptr);				\
-	__chk_user_ptr(__pu_ptr, sizeof(*__pu_ptr));		\
+	__chk_user_ptr(__pu_ptr);				\
 	WRITE_ONCE(*(__pu_ptr), x);				\
 	0;							\
 })
@@ -22,7 +17,7 @@ static inline void __chk_user_ptr(const volatile void *p, size_t size)
 #define get_user(x, ptr)					\
 ({								\
 	typeof(ptr) __pu_ptr = (ptr);				\
-	__chk_user_ptr(__pu_ptr, sizeof(*__pu_ptr));		\
+	__chk_user_ptr(__pu_ptr);				\
 	x = READ_ONCE(*(__pu_ptr));				\
 	0;							\
 })
@@ -37,7 +32,6 @@ static void volatile_memcpy(volatile char *to, const volatile char *from,
 static inline int copy_from_user(void *to, const void __user volatile *from,
 				 unsigned long n)
 {
-	__chk_user_ptr(from, n);
 	volatile_memcpy(to, from, n);
 	return 0;
 }
@@ -45,7 +39,6 @@ static inline int copy_from_user(void *to, const void __user volatile *from,
 static inline int copy_to_user(void __user volatile *to, const void *from,
 			       unsigned long n)
 {
-	__chk_user_ptr(to, n);
 	volatile_memcpy(to, from, n);
 	return 0;
 }

tools/virtio/virtio_test.c

@@ -134,7 +134,7 @@ static void vdev_info_init(struct vdev_info* dev, unsigned long long features)
 	dev->buf_size = 1024;
 	dev->buf = malloc(dev->buf_size);
 	assert(dev->buf);
-        dev->control = open("/dev/vhost-test", O_RDWR);
+	dev->control = open("/dev/vhost-test", O_RDWR);
 	assert(dev->control >= 0);
 	r = ioctl(dev->control, VHOST_SET_OWNER, NULL);
 	assert(r >= 0);
@@ -327,7 +327,7 @@ const struct option longopts[] = {
 	}
 };
 
-static void help(void)
+static void help(int status)
 {
 	fprintf(stderr, "Usage: virtio_test [--help]"
 		" [--no-indirect]"
@@ -337,6 +337,8 @@ static void help(void)
 		" [--batch=random/N]"
 		" [--reset=N]"
 		"\n");
+
+	exit(status);
 }
 
 int main(int argc, char **argv)
@@ -354,14 +356,12 @@ int main(int argc, char **argv)
 		case -1:
 			goto done;
 		case '?':
-			help();
-			exit(2);
+			help(2);
 		case 'e':
 			features &= ~(1ULL << VIRTIO_RING_F_EVENT_IDX);
 			break;
 		case 'h':
-			help();
-			goto done;
+			help(0);
 		case 'i':
 			features &= ~(1ULL << VIRTIO_RING_F_INDIRECT_DESC);
 			break;