misc: bcm-vk: add VK messaging support

Add message support in order to be able to communicate
to VK card via message queues.

This info is used for debug purposes via collection of logs via direct
read of BAR space and by sysfs access (in a follow on commit).

Co-developed-by: Desmond Yan <desmond.yan@broadcom.com>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Desmond Yan <desmond.yan@broadcom.com>
Signed-off-by: Scott Branden <scott.branden@broadcom.com>
Link: https://lore.kernel.org/r/20210120175827.14820-10-scott.branden@broadcom.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

1 files changed, 1187 insertions, 0 deletions

diff --git a/drivers/misc/bcm-vk/bcm_vk_msg.c b/drivers/misc/bcm-vk/bcm_vk_msg.c
index 2d9a6b4e5f61..b05e20a72a8b 100644
--- a/drivers/misc/bcm-vk/bcm_vk_msg.c
+++ b/drivers/misc/bcm-vk/bcm_vk_msg.c
@@ -3,8 +3,200 @@
  * Copyright 2018-2020 Broadcom.
  */
 
+#include <linux/delay.h>
+#include <linux/fs.h>
+#include <linux/hash.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/poll.h>
+#include <linux/sizes.h>
+#include <linux/spinlock.h>
+#include <linux/timer.h>
+
 #include "bcm_vk.h"
 #include "bcm_vk_msg.h"
+#include "bcm_vk_sg.h"
+
+/* functions to manipulate the transport id in msg block */
+#define BCM_VK_MSG_Q_SHIFT	 4
+#define BCM_VK_MSG_Q_MASK	 0xF
+#define BCM_VK_MSG_ID_MASK	 0xFFF
+
+#define BCM_VK_DMA_DRAIN_MAX_MS	  2000
+
+/* number x q_size will be the max number of msg processed per loop */
+#define BCM_VK_MSG_PROC_MAX_LOOP 2
+
+/* module parameter */
+static bool hb_mon = true;
+module_param(hb_mon, bool, 0444);
+MODULE_PARM_DESC(hb_mon, "Monitoring heartbeat continuously.\n");
+static int batch_log = 1;
+module_param(batch_log, int, 0444);
+MODULE_PARM_DESC(batch_log, "Max num of logs per batch operation.\n");
+
+static bool hb_mon_is_on(void)
+{
+	return hb_mon;
+}
+
+static u32 get_q_num(const struct vk_msg_blk *msg)
+{
+	u32 q_num = msg->trans_id & BCM_VK_MSG_Q_MASK;
+
+	if (q_num >= VK_MSGQ_PER_CHAN_MAX)
+		q_num = VK_MSGQ_NUM_DEFAULT;
+	return q_num;
+}
+
+static void set_q_num(struct vk_msg_blk *msg, u32 q_num)
+{
+	msg->trans_id = (msg->trans_id & ~BCM_VK_MSG_Q_MASK) |
+		(q_num >= VK_MSGQ_PER_CHAN_MAX) ? VK_MSGQ_NUM_DEFAULT : q_num;
+}
+
+static u32 get_msg_id(const struct vk_msg_blk *msg)
+{
+	return ((msg->trans_id >> BCM_VK_MSG_Q_SHIFT) & BCM_VK_MSG_ID_MASK);
+}
+
+static void set_msg_id(struct vk_msg_blk *msg, u32 val)
+{
+	msg->trans_id = (val << BCM_VK_MSG_Q_SHIFT) | get_q_num(msg);
+}
+
+static u32 msgq_inc(const struct bcm_vk_sync_qinfo *qinfo, u32 idx, u32 inc)
+{
+	return ((idx + inc) & qinfo->q_mask);
+}
+
+static
+struct vk_msg_blk __iomem *msgq_blk_addr(const struct bcm_vk_sync_qinfo *qinfo,
+					 u32 idx)
+{
+	return qinfo->q_start + (VK_MSGQ_BLK_SIZE * idx);
+}
+
+static u32 msgq_occupied(const struct bcm_vk_msgq __iomem *msgq,
+			 const struct bcm_vk_sync_qinfo *qinfo)
+{
+	u32 wr_idx, rd_idx;
+
+	wr_idx = readl_relaxed(&msgq->wr_idx);
+	rd_idx = readl_relaxed(&msgq->rd_idx);
+
+	return ((wr_idx - rd_idx) & qinfo->q_mask);
+}
+
+static
+u32 msgq_avail_space(const struct bcm_vk_msgq __iomem *msgq,
+		     const struct bcm_vk_sync_qinfo *qinfo)
+{
+	return (qinfo->q_size - msgq_occupied(msgq, qinfo) - 1);
+}
+
+/* number of retries when enqueue message fails before returning EAGAIN */
+#define BCM_VK_H2VK_ENQ_RETRY 10
+#define BCM_VK_H2VK_ENQ_RETRY_DELAY_MS 50
+
+bool bcm_vk_drv_access_ok(struct bcm_vk *vk)
+{
+	return (!!atomic_read(&vk->msgq_inited));
+}
+
+void bcm_vk_set_host_alert(struct bcm_vk *vk, u32 bit_mask)
+{
+	struct bcm_vk_alert *alert = &vk->host_alert;
+	unsigned long flags;
+
+	/* use irqsave version as this maybe called inside timer interrupt */
+	spin_lock_irqsave(&vk->host_alert_lock, flags);
+	alert->notfs |= bit_mask;
+	spin_unlock_irqrestore(&vk->host_alert_lock, flags);
+
+	if (test_and_set_bit(BCM_VK_WQ_NOTF_PEND, vk->wq_offload) == 0)
+		queue_work(vk->wq_thread, &vk->wq_work);
+}
+
+/*
+ * Heartbeat related defines
+ * The heartbeat from host is a last resort.  If stuck condition happens
+ * on the card, firmware is supposed to detect it.  Therefore, the heartbeat
+ * values used will be more relaxed on the driver, which need to be bigger
+ * than the watchdog timeout on the card.  The watchdog timeout on the card
+ * is 20s, with a jitter of 2s => 22s.  We use a value of 27s here.
+ */
+#define BCM_VK_HB_TIMER_S 3
+#define BCM_VK_HB_TIMER_VALUE (BCM_VK_HB_TIMER_S * HZ)
+#define BCM_VK_HB_LOST_MAX (27 / BCM_VK_HB_TIMER_S)
+
+static void bcm_vk_hb_poll(struct timer_list *t)
+{
+	u32 uptime_s;
+	struct bcm_vk_hb_ctrl *hb = container_of(t, struct bcm_vk_hb_ctrl,
+						 timer);
+	struct bcm_vk *vk = container_of(hb, struct bcm_vk, hb_ctrl);
+
+	if (bcm_vk_drv_access_ok(vk) && hb_mon_is_on()) {
+		/* read uptime from register and compare */
+		uptime_s = vkread32(vk, BAR_0, BAR_OS_UPTIME);
+
+		if (uptime_s == hb->last_uptime)
+			hb->lost_cnt++;
+		else /* reset to avoid accumulation */
+			hb->lost_cnt = 0;
+
+		dev_dbg(&vk->pdev->dev, "Last uptime %d current %d, lost %d\n",
+			hb->last_uptime, uptime_s, hb->lost_cnt);
+
+		/*
+		 * if the interface goes down without any activity, a value
+		 * of 0xFFFFFFFF will be continuously read, and the detection
+		 * will be happened eventually.
+		 */
+		hb->last_uptime = uptime_s;
+	} else {
+		/* reset heart beat lost cnt */
+		hb->lost_cnt = 0;
+	}
+
+	/* next, check if heartbeat exceeds limit */
+	if (hb->lost_cnt > BCM_VK_HB_LOST_MAX) {
+		dev_err(&vk->pdev->dev, "Heartbeat Misses %d times, %d s!\n",
+			BCM_VK_HB_LOST_MAX,
+			BCM_VK_HB_LOST_MAX * BCM_VK_HB_TIMER_S);
+
+		bcm_vk_blk_drv_access(vk);
+		bcm_vk_set_host_alert(vk, ERR_LOG_HOST_HB_FAIL);
+	}
+	/* re-arm timer */
+	mod_timer(&hb->timer, jiffies + BCM_VK_HB_TIMER_VALUE);
+}
+
+void bcm_vk_hb_init(struct bcm_vk *vk)
+{
+	struct bcm_vk_hb_ctrl *hb = &vk->hb_ctrl;
+
+	timer_setup(&hb->timer, bcm_vk_hb_poll, 0);
+	mod_timer(&hb->timer, jiffies + BCM_VK_HB_TIMER_VALUE);
+}
+
+void bcm_vk_hb_deinit(struct bcm_vk *vk)
+{
+	struct bcm_vk_hb_ctrl *hb = &vk->hb_ctrl;
+
+	del_timer(&hb->timer);
+}
+
+static void bcm_vk_msgid_bitmap_clear(struct bcm_vk *vk,
+				      unsigned int start,
+				      unsigned int nbits)
+{
+	spin_lock(&vk->msg_id_lock);
+	bitmap_clear(vk->bmap, start, nbits);
+	spin_unlock(&vk->msg_id_lock);
+}
 
 /*
  * allocate a ctx per file struct
@@ -39,12 +231,47 @@ static struct bcm_vk_ctx *bcm_vk_get_ctx(struct bcm_vk *vk, const pid_t pid)
 	/* increase kref */
 	kref_get(&vk->kref);
 
+	/* clear counter */
+	atomic_set(&ctx->pend_cnt, 0);
+	atomic_set(&ctx->dma_cnt, 0);
+	init_waitqueue_head(&ctx->rd_wq);
+
 all_in_use_exit:
 	spin_unlock(&vk->ctx_lock);
 
 	return ctx;
 }
 
+static u16 bcm_vk_get_msg_id(struct bcm_vk *vk)
+{
+	u16 rc = VK_MSG_ID_OVERFLOW;
+	u16 test_bit_count = 0;
+
+	spin_lock(&vk->msg_id_lock);
+	while (test_bit_count < (VK_MSG_ID_BITMAP_SIZE - 1)) {
+		/*
+		 * first time come in this loop, msg_id will be 0
+		 * and the first one tested will be 1.  We skip
+		 * VK_SIMPLEX_MSG_ID (0) for one way host2vk
+		 * communication
+		 */
+		vk->msg_id++;
+		if (vk->msg_id == VK_MSG_ID_BITMAP_SIZE)
+			vk->msg_id = 1;
+
+		if (test_bit(vk->msg_id, vk->bmap)) {
+			test_bit_count++;
+			continue;
+		}
+		rc = vk->msg_id;
+		bitmap_set(vk->bmap, vk->msg_id, 1);
+		break;
+	}
+	spin_unlock(&vk->msg_id_lock);
+
+	return rc;
+}
+
 static int bcm_vk_free_ctx(struct bcm_vk *vk, struct bcm_vk_ctx *ctx)
 {
 	u32 idx;
@@ -82,6 +309,636 @@ static int bcm_vk_free_ctx(struct bcm_vk *vk, struct bcm_vk_ctx *ctx)
 	return count;
 }
 
+static void bcm_vk_free_wkent(struct device *dev, struct bcm_vk_wkent *entry)
+{
+	int proc_cnt;
+
+	bcm_vk_sg_free(dev, entry->dma, VK_DMA_MAX_ADDRS, &proc_cnt);
+	if (proc_cnt)
+		atomic_dec(&entry->ctx->dma_cnt);
+
+	kfree(entry->to_h_msg);
+	kfree(entry);
+}
+
+static void bcm_vk_drain_all_pend(struct device *dev,
+				  struct bcm_vk_msg_chan *chan,
+				  struct bcm_vk_ctx *ctx)
+{
+	u32 num;
+	struct bcm_vk_wkent *entry, *tmp;
+	struct bcm_vk *vk;
+	struct list_head del_q;
+
+	if (ctx)
+		vk = container_of(ctx->miscdev, struct bcm_vk, miscdev);
+
+	INIT_LIST_HEAD(&del_q);
+	spin_lock(&chan->pendq_lock);
+	for (num = 0; num < chan->q_nr; num++) {
+		list_for_each_entry_safe(entry, tmp, &chan->pendq[num], node) {
+			if ((!ctx) || (entry->ctx->idx == ctx->idx)) {
+				list_del(&entry->node);
+				list_add_tail(&entry->node, &del_q);
+			}
+		}
+	}
+	spin_unlock(&chan->pendq_lock);
+
+	/* batch clean up */
+	num = 0;
+	list_for_each_entry_safe(entry, tmp, &del_q, node) {
+		list_del(&entry->node);
+		num++;
+		if (ctx) {
+			struct vk_msg_blk *msg;
+			int bit_set;
+			bool responded;
+			u32 msg_id;
+
+			/* if it is specific ctx, log for any stuck */
+			msg = entry->to_v_msg;
+			msg_id = get_msg_id(msg);
+			bit_set = test_bit(msg_id, vk->bmap);
+			responded = entry->to_h_msg ? true : false;
+			if (num <= batch_log)
+				dev_info(dev,
+					 "Drained: fid %u size %u msg 0x%x(seq-%x) ctx 0x%x[fd-%d] args:[0x%x 0x%x] resp %s, bmap %d\n",
+					 msg->function_id, msg->size,
+					 msg_id, entry->seq_num,
+					 msg->context_id, entry->ctx->idx,
+					 msg->cmd, msg->arg,
+					 responded ? "T" : "F", bit_set);
+			if (responded)
+				atomic_dec(&ctx->pend_cnt);
+			else if (bit_set)
+				bcm_vk_msgid_bitmap_clear(vk, msg_id, 1);
+		}
+		bcm_vk_free_wkent(dev, entry);
+	}
+	if (num && ctx)
+		dev_info(dev, "Total drained items %d [fd-%d]\n",
+			 num, ctx->idx);
+}
+
+/*
+ * Function to sync up the messages queue info that is provided by BAR1
+ */
+int bcm_vk_sync_msgq(struct bcm_vk *vk, bool force_sync)
+{
+	struct bcm_vk_msgq __iomem *msgq;
+	struct device *dev = &vk->pdev->dev;
+	u32 msgq_off;
+	u32 num_q;
+	struct bcm_vk_msg_chan *chan_list[] = {&vk->to_v_msg_chan,
+					       &vk->to_h_msg_chan};
+	struct bcm_vk_msg_chan *chan;
+	int i, j;
+	int ret = 0;
+
+	/*
+	 * If the driver is loaded at startup where vk OS is not up yet,
+	 * the msgq-info may not be available until a later time.  In
+	 * this case, we skip and the sync function is supposed to be
+	 * called again.
+	 */
+	if (!bcm_vk_msgq_marker_valid(vk)) {
+		dev_info(dev, "BAR1 msgq marker not initialized.\n");
+		return -EAGAIN;
+	}
+
+	msgq_off = vkread32(vk, BAR_1, VK_BAR1_MSGQ_CTRL_OFF);
+
+	/* each side is always half the total  */
+	num_q = vkread32(vk, BAR_1, VK_BAR1_MSGQ_NR) / 2;
+	if (!num_q || (num_q > VK_MSGQ_PER_CHAN_MAX)) {
+		dev_err(dev,
+			"Advertised msgq %d error - max %d allowed\n",
+			num_q, VK_MSGQ_PER_CHAN_MAX);
+		return -EINVAL;
+	}
+
+	vk->to_v_msg_chan.q_nr = num_q;
+	vk->to_h_msg_chan.q_nr = num_q;
+
+	/* first msgq location */
+	msgq = vk->bar[BAR_1] + msgq_off;
+
+	/*
+	 * if this function is called when it is already inited,
+	 * something is wrong
+	 */
+	if (bcm_vk_drv_access_ok(vk) && !force_sync) {
+		dev_err(dev, "Msgq info already in sync\n");
+		return -EPERM;
+	}
+
+	for (i = 0; i < ARRAY_SIZE(chan_list); i++) {
+		chan = chan_list[i];
+		memset(chan->sync_qinfo, 0, sizeof(chan->sync_qinfo));
+
+		for (j = 0; j < num_q; j++) {
+			struct bcm_vk_sync_qinfo *qinfo;
+			u32 msgq_start;
+			u32 msgq_size;
+			u32 msgq_nxt;
+			u32 msgq_db_offset, q_db_offset;
+
+			chan->msgq[j] = msgq;
+			msgq_start = readl_relaxed(&msgq->start);
+			msgq_size = readl_relaxed(&msgq->size);
+			msgq_nxt = readl_relaxed(&msgq->nxt);
+			msgq_db_offset = readl_relaxed(&msgq->db_offset);
+			q_db_offset = (msgq_db_offset & ((1 << DB_SHIFT) - 1));
+			if (q_db_offset  == (~msgq_db_offset >> DB_SHIFT))
+				msgq_db_offset = q_db_offset;
+			else
+				/* fall back to default */
+				msgq_db_offset = VK_BAR0_Q_DB_BASE(j);
+
+			dev_info(dev,
+				 "MsgQ[%d] type %d num %d, @ 0x%x, db_offset 0x%x rd_idx %d wr_idx %d, size %d, nxt 0x%x\n",
+				 j,
+				 readw_relaxed(&msgq->type),
+				 readw_relaxed(&msgq->num),
+				 msgq_start,
+				 msgq_db_offset,
+				 readl_relaxed(&msgq->rd_idx),
+				 readl_relaxed(&msgq->wr_idx),
+				 msgq_size,
+				 msgq_nxt);
+
+			qinfo = &chan->sync_qinfo[j];
+			/* formulate and record static info */
+			qinfo->q_start = vk->bar[BAR_1] + msgq_start;
+			qinfo->q_size = msgq_size;
+			/* set low threshold as 50% or 1/2 */
+			qinfo->q_low = qinfo->q_size >> 1;
+			qinfo->q_mask = qinfo->q_size - 1;
+			qinfo->q_db_offset = msgq_db_offset;
+
+			msgq++;
+		}
+	}
+	atomic_set(&vk->msgq_inited, 1);
+
+	return ret;
+}
+
+static int bcm_vk_msg_chan_init(struct bcm_vk_msg_chan *chan)
+{
+	u32 i;
+
+	mutex_init(&chan->msgq_mutex);
+	spin_lock_init(&chan->pendq_lock);
+	for (i = 0; i < VK_MSGQ_MAX_NR; i++)
+		INIT_LIST_HEAD(&chan->pendq[i]);
+
+	return 0;
+}
+
+static void bcm_vk_append_pendq(struct bcm_vk_msg_chan *chan, u16 q_num,
+				struct bcm_vk_wkent *entry)
+{
+	struct bcm_vk_ctx *ctx;
+
+	spin_lock(&chan->pendq_lock);
+	list_add_tail(&entry->node, &chan->pendq[q_num]);
+	if (entry->to_h_msg) {
+		ctx = entry->ctx;
+		atomic_inc(&ctx->pend_cnt);
+		wake_up_interruptible(&ctx->rd_wq);
+	}
+	spin_unlock(&chan->pendq_lock);
+}
+
+static u32 bcm_vk_append_ib_sgl(struct bcm_vk *vk,
+				struct bcm_vk_wkent *entry,
+				struct _vk_data *data,
+				unsigned int num_planes)
+{
+	unsigned int i;
+	unsigned int item_cnt = 0;
+	struct device *dev = &vk->pdev->dev;
+	struct bcm_vk_msg_chan *chan = &vk->to_v_msg_chan;
+	struct vk_msg_blk *msg = &entry->to_v_msg[0];
+	struct bcm_vk_msgq __iomem *msgq;
+	struct bcm_vk_sync_qinfo *qinfo;
+	u32 ib_sgl_size = 0;
+	u8 *buf = (u8 *)&entry->to_v_msg[entry->to_v_blks];
+	u32 avail;
+	u32 q_num;
+
+	/* check if high watermark is hit, and if so, skip */
+	q_num = get_q_num(msg);
+	msgq = chan->msgq[q_num];
+	qinfo = &chan->sync_qinfo[q_num];
+	avail = msgq_avail_space(msgq, qinfo);
+	if (avail < qinfo->q_low) {
+		dev_dbg(dev, "Skip inserting inband SGL, [0x%x/0x%x]\n",
+			avail, qinfo->q_size);
+		return 0;
+	}
+
+	for (i = 0; i < num_planes; i++) {
+		if (data[i].address &&
+		    (ib_sgl_size + data[i].size) <= vk->ib_sgl_size) {
+			item_cnt++;
+			memcpy(buf, entry->dma[i].sglist, data[i].size);
+			ib_sgl_size += data[i].size;
+			buf += data[i].size;
+		}
+	}
+
+	dev_dbg(dev, "Num %u sgl items appended, size 0x%x, room 0x%x\n",
+		item_cnt, ib_sgl_size, vk->ib_sgl_size);
+
+	/* round up size */
+	ib_sgl_size = (ib_sgl_size + VK_MSGQ_BLK_SIZE - 1)
+		       >> VK_MSGQ_BLK_SZ_SHIFT;
+
+	return ib_sgl_size;
+}
+
+void bcm_to_v_q_doorbell(struct bcm_vk *vk, u32 q_num, u32 db_val)
+{
+	struct bcm_vk_msg_chan *chan = &vk->to_v_msg_chan;
+	struct bcm_vk_sync_qinfo *qinfo = &chan->sync_qinfo[q_num];
+
+	vkwrite32(vk, db_val, BAR_0, qinfo->q_db_offset);
+}
+
+static int bcm_to_v_msg_enqueue(struct bcm_vk *vk, struct bcm_vk_wkent *entry)
+{
+	static u32 seq_num;
+	struct bcm_vk_msg_chan *chan = &vk->to_v_msg_chan;
+	struct device *dev = &vk->pdev->dev;
+	struct vk_msg_blk *src = &entry->to_v_msg[0];
+
+	struct vk_msg_blk __iomem *dst;
+	struct bcm_vk_msgq __iomem *msgq;
+	struct bcm_vk_sync_qinfo *qinfo;
+	u32 q_num = get_q_num(src);
+	u32 wr_idx; /* local copy */
+	u32 i;
+	u32 avail;
+	u32 retry;
+
+	if (entry->to_v_blks != src->size + 1) {
+		dev_err(dev, "number of blks %d not matching %d MsgId[0x%x]: func %d ctx 0x%x\n",
+			entry->to_v_blks,
+			src->size + 1,
+			get_msg_id(src),
+			src->function_id,
+			src->context_id);
+		return -EMSGSIZE;
+	}
+
+	msgq = chan->msgq[q_num];
+	qinfo = &chan->sync_qinfo[q_num];
+
+	mutex_lock(&chan->msgq_mutex);
+
+	avail = msgq_avail_space(msgq, qinfo);
+
+	/* if not enough space, return EAGAIN and let app handles it */
+	retry = 0;
+	while ((avail < entry->to_v_blks) &&
+	       (retry++ < BCM_VK_H2VK_ENQ_RETRY)) {
+		mutex_unlock(&chan->msgq_mutex);
+
+		msleep(BCM_VK_H2VK_ENQ_RETRY_DELAY_MS);
+		mutex_lock(&chan->msgq_mutex);
+		avail = msgq_avail_space(msgq, qinfo);
+	}
+	if (retry > BCM_VK_H2VK_ENQ_RETRY) {
+		mutex_unlock(&chan->msgq_mutex);
+		return -EAGAIN;
+	}
+
+	/* at this point, mutex is taken and there is enough space */
+	entry->seq_num = seq_num++; /* update debug seq number */
+	wr_idx = readl_relaxed(&msgq->wr_idx);
+
+	if (wr_idx >= qinfo->q_size) {
+		dev_crit(dev, "Invalid wr_idx 0x%x => max 0x%x!",
+			 wr_idx, qinfo->q_size);
+		bcm_vk_blk_drv_access(vk);
+		bcm_vk_set_host_alert(vk, ERR_LOG_HOST_PCIE_DWN);
+		goto idx_err;
+	}
+
+	dst = msgq_blk_addr(qinfo, wr_idx);
+	for (i = 0; i < entry->to_v_blks; i++) {
+		memcpy_toio(dst, src, sizeof(*dst));
+
+		src++;
+		wr_idx = msgq_inc(qinfo, wr_idx, 1);
+		dst = msgq_blk_addr(qinfo, wr_idx);
+	}
+
+	/* flush the write pointer */
+	writel(wr_idx, &msgq->wr_idx);
+
+	/* log new info for debugging */
+	dev_dbg(dev,
+		"MsgQ[%d] [Rd Wr] = [%d %d] blks inserted %d - Q = [u-%d a-%d]/%d\n",
+		readl_relaxed(&msgq->num),
+		readl_relaxed(&msgq->rd_idx),
+		wr_idx,
+		entry->to_v_blks,
+		msgq_occupied(msgq, qinfo),
+		msgq_avail_space(msgq, qinfo),
+		readl_relaxed(&msgq->size));
+	/*
+	 * press door bell based on queue number. 1 is added to the wr_idx
+	 * to avoid the value of 0 appearing on the VK side to distinguish
+	 * from initial value.
+	 */
+	bcm_to_v_q_doorbell(vk, q_num, wr_idx + 1);
+idx_err:
+	mutex_unlock(&chan->msgq_mutex);
+	return 0;
+}
+
+int bcm_vk_send_shutdown_msg(struct bcm_vk *vk, u32 shut_type,
+			     const pid_t pid, const u32 q_num)
+{
+	int rc = 0;
+	struct bcm_vk_wkent *entry;
+	struct device *dev = &vk->pdev->dev;
+
+	/*
+	 * check if the marker is still good.  Sometimes, the PCIe interface may
+	 * have gone done, and if so and we ship down thing based on broken
+	 * values, kernel may panic.
+	 */
+	if (!bcm_vk_msgq_marker_valid(vk)) {
+		dev_info(dev, "PCIe comm chan - invalid marker (0x%x)!\n",
+			 vkread32(vk, BAR_1, VK_BAR1_MSGQ_DEF_RDY));
+		return -EINVAL;
+	}
+
+	entry = kzalloc(sizeof(*entry) +
+			sizeof(struct vk_msg_blk), GFP_KERNEL);
+	if (!entry)
+		return -ENOMEM;
+
+	/* fill up necessary data */
+	entry->to_v_msg[0].function_id = VK_FID_SHUTDOWN;
+	set_q_num(&entry->to_v_msg[0], q_num);
+	set_msg_id(&entry->to_v_msg[0], VK_SIMPLEX_MSG_ID);
+	entry->to_v_blks = 1; /* always 1 block */
+
+	entry->to_v_msg[0].cmd = shut_type;
+	entry->to_v_msg[0].arg = pid;
+
+	rc = bcm_to_v_msg_enqueue(vk, entry);
+	if (rc)
+		dev_err(dev,
+			"Sending shutdown message to q %d for pid %d fails.\n",
+			get_q_num(&entry->to_v_msg[0]), pid);
+
+	kfree(entry);
+
+	return rc;
+}
+
+static int bcm_vk_handle_last_sess(struct bcm_vk *vk, const pid_t pid,
+				   const u32 q_num)
+{
+	int rc = 0;
+	struct device *dev = &vk->pdev->dev;
+
+	/*
+	 * don't send down or do anything if message queue is not initialized
+	 */
+	if (!bcm_vk_drv_access_ok(vk))
+		return -EPERM;
+
+	dev_dbg(dev, "No more sessions, shut down pid %d\n", pid);
+
+	rc = bcm_vk_send_shutdown_msg(vk, VK_SHUTDOWN_PID, pid, q_num);
+
+	return rc;
+}
+
+static struct bcm_vk_wkent *bcm_vk_dequeue_pending(struct bcm_vk *vk,
+						   struct bcm_vk_msg_chan *chan,
+						   u16 q_num,
+						   u16 msg_id)
+{
+	bool found = false;
+	struct bcm_vk_wkent *entry;
+
+	spin_lock(&chan->pendq_lock);
+	list_for_each_entry(entry, &chan->pendq[q_num], node) {
+		if (get_msg_id(&entry->to_v_msg[0]) == msg_id) {
+			list_del(&entry->node);
+			found = true;
+			bcm_vk_msgid_bitmap_clear(vk, msg_id, 1);
+			break;
+		}
+	}
+	spin_unlock(&chan->pendq_lock);
+	return ((found) ? entry : NULL);
+}
+
+s32 bcm_to_h_msg_dequeue(struct bcm_vk *vk)
+{
+	struct device *dev = &vk->pdev->dev;
+	struct bcm_vk_msg_chan *chan = &vk->to_h_msg_chan;
+	struct vk_msg_blk *data;
+	struct vk_msg_blk __iomem *src;
+	struct vk_msg_blk *dst;
+	struct bcm_vk_msgq __iomem *msgq;
+	struct bcm_vk_sync_qinfo *qinfo;
+	struct bcm_vk_wkent *entry;
+	u32 rd_idx, wr_idx;
+	u32 q_num, msg_id, j;
+	u32 num_blks;
+	s32 total = 0;
+	int cnt = 0;
+	int msg_processed = 0;
+	int max_msg_to_process;
+	bool exit_loop;
+
+	/*
+	 * drain all the messages from the queues, and find its pending
+	 * entry in the to_v queue, based on msg_id & q_num, and move the
+	 * entry to the to_h pending queue, waiting for user space
+	 * program to extract
+	 */
+	mutex_lock(&chan->msgq_mutex);
+
+	for (q_num = 0; q_num < chan->q_nr; q_num++) {
+		msgq = chan->msgq[q_num];
+		qinfo = &chan->sync_qinfo[q_num];
+		max_msg_to_process = BCM_VK_MSG_PROC_MAX_LOOP * qinfo->q_size;
+
+		rd_idx = readl_relaxed(&msgq->rd_idx);
+		wr_idx = readl_relaxed(&msgq->wr_idx);
+		msg_processed = 0;
+		exit_loop = false;
+		while ((rd_idx != wr_idx) && !exit_loop) {
+			u8 src_size;
+
+			/*
+			 * Make a local copy and get pointer to src blk
+			 * The rd_idx is masked before getting the pointer to
+			 * avoid out of bound access in case the interface goes
+			 * down.  It will end up pointing to the last block in
+			 * the buffer, but subsequent src->size check would be
+			 * able to catch this.
+			 */
+			src = msgq_blk_addr(qinfo, rd_idx & qinfo->q_mask);
+			src_size = readb(&src->size);
+
+			if ((rd_idx >= qinfo->q_size) ||
+			    (src_size > (qinfo->q_size - 1))) {
+				dev_crit(dev,
+					 "Invalid rd_idx 0x%x or size 0x%x => max 0x%x!",
+					 rd_idx, src_size, qinfo->q_size);
+				bcm_vk_blk_drv_access(vk);
+				bcm_vk_set_host_alert(vk,
+						      ERR_LOG_HOST_PCIE_DWN);
+				goto idx_err;
+			}
+
+			num_blks = src_size + 1;
+			data = kzalloc(num_blks * VK_MSGQ_BLK_SIZE, GFP_KERNEL);
+			if (data) {
+				/* copy messages and linearize it */
+				dst = data;
+				for (j = 0; j < num_blks; j++) {
+					memcpy_fromio(dst, src, sizeof(*dst));
+
+					dst++;
+					rd_idx = msgq_inc(qinfo, rd_idx, 1);
+					src = msgq_blk_addr(qinfo, rd_idx);
+				}
+				total++;
+			} else {
+				/*
+				 * if we could not allocate memory in kernel,
+				 * that is fatal.
+				 */
+				dev_crit(dev, "Kernel mem allocation failure.\n");
+				return -ENOMEM;
+			}
+
+			/* flush rd pointer after a message is dequeued */
+			writel(rd_idx, &msgq->rd_idx);
+
+			/* log new info for debugging */
+			dev_dbg(dev,
+				"MsgQ[%d] [Rd Wr] = [%d %d] blks extracted %d - Q = [u-%d a-%d]/%d\n",
+				readl_relaxed(&msgq->num),
+				rd_idx,
+				wr_idx,
+				num_blks,
+				msgq_occupied(msgq, qinfo),
+				msgq_avail_space(msgq, qinfo),
+				readl_relaxed(&msgq->size));
+
+			/*
+			 * No need to search if it is an autonomous one-way
+			 * message from driver, as these messages do not bear
+			 * a to_v pending item. Currently, only the shutdown
+			 * message falls into this category.
+			 */
+			if (data->function_id == VK_FID_SHUTDOWN) {
+				kfree(data);
+				continue;
+			}
+
+			msg_id = get_msg_id(data);
+			/* lookup original message in to_v direction */
+			entry = bcm_vk_dequeue_pending(vk,
+						       &vk->to_v_msg_chan,
+						       q_num,
+						       msg_id);
+
+			/*
+			 * if there is message to does not have prior send,
+			 * this is the location to add here
+			 */
+			if (entry) {
+				entry->to_h_blks = num_blks;
+				entry->to_h_msg = data;
+				bcm_vk_append_pendq(&vk->to_h_msg_chan,
+						    q_num, entry);
+
+			} else {
+				if (cnt++ < batch_log)
+					dev_info(dev,
+						 "Could not find MsgId[0x%x] for resp func %d bmap %d\n",
+						 msg_id, data->function_id,
+						 test_bit(msg_id, vk->bmap));
+				kfree(data);
+			}
+			/* Fetch wr_idx to handle more back-to-back events */
+			wr_idx = readl(&msgq->wr_idx);
+
+			/*
+			 * cap the max so that even we try to handle more back-to-back events,
+			 * so that it won't hold CPU too long or in case rd/wr idexes are
+			 * corrupted which triggers infinite looping.
+			 */
+			if (++msg_processed >= max_msg_to_process) {
+				dev_warn(dev, "Q[%d] Per loop processing exceeds %d\n",
+					 q_num, max_msg_to_process);
+				exit_loop = true;
+			}
+		}
+	}
+idx_err:
+	mutex_unlock(&chan->msgq_mutex);
+	dev_dbg(dev, "total %d drained from queues\n", total);
+
+	return total;
+}
+
+/*
+ * init routine for all required data structures
+ */
+static int bcm_vk_data_init(struct bcm_vk *vk)
+{
+	int i;
+
+	spin_lock_init(&vk->ctx_lock);
+	for (i = 0; i < ARRAY_SIZE(vk->ctx); i++) {
+		vk->ctx[i].in_use = false;
+		vk->ctx[i].idx = i;	/* self identity */
+		vk->ctx[i].miscdev = NULL;
+	}
+	spin_lock_init(&vk->msg_id_lock);
+	spin_lock_init(&vk->host_alert_lock);
+	vk->msg_id = 0;
+
+	/* initialize hash table */
+	for (i = 0; i < VK_PID_HT_SZ; i++)
+		INIT_LIST_HEAD(&vk->pid_ht[i].head);
+
+	return 0;
+}
+
+irqreturn_t bcm_vk_msgq_irqhandler(int irq, void *dev_id)
+{
+	struct bcm_vk *vk = dev_id;
+
+	if (!bcm_vk_drv_access_ok(vk)) {
+		dev_err(&vk->pdev->dev,
+			"Interrupt %d received when msgq not inited\n", irq);
+		goto skip_schedule_work;
+	}
+
+	queue_work(vk->wq_thread, &vk->wq_work);
+
+skip_schedule_work:
+	return IRQ_HANDLED;
+}
+
 int bcm_vk_open(struct inode *inode, struct file *p_file)
 {
 	struct bcm_vk_ctx *ctx;
@@ -112,16 +969,346 @@ int bcm_vk_open(struct inode *inode, struct file *p_file)
 	return rc;
 }
 
+ssize_t bcm_vk_read(struct file *p_file,
+		    char __user *buf,
+		    size_t count,
+		    loff_t *f_pos)
+{
+	ssize_t rc = -ENOMSG;
+	struct bcm_vk_ctx *ctx = p_file->private_data;
+	struct bcm_vk *vk = container_of(ctx->miscdev, struct bcm_vk,
+					 miscdev);
+	struct device *dev = &vk->pdev->dev;
+	struct bcm_vk_msg_chan *chan = &vk->to_h_msg_chan;
+	struct bcm_vk_wkent *entry = NULL;
+	u32 q_num;
+	u32 rsp_length;
+	bool found = false;
+
+	if (!bcm_vk_drv_access_ok(vk))
+		return -EPERM;
+
+	dev_dbg(dev, "Buf count %zu\n", count);
+	found = false;
+
+	/*
+	 * search through the pendq on the to_h chan, and return only those
+	 * that belongs to the same context.  Search is always from the high to
+	 * the low priority queues
+	 */
+	spin_lock(&chan->pendq_lock);
+	for (q_num = 0; q_num < chan->q_nr; q_num++) {
+		list_for_each_entry(entry, &chan->pendq[q_num], node) {
+			if (entry->ctx->idx == ctx->idx) {
+				if (count >=
+				    (entry->to_h_blks * VK_MSGQ_BLK_SIZE)) {
+					list_del(&entry->node);
+					atomic_dec(&ctx->pend_cnt);
+					found = true;
+				} else {
+					/* buffer not big enough */
+					rc = -EMSGSIZE;
+				}
+				goto read_loop_exit;
+			}
+		}
+	}
+read_loop_exit:
+	spin_unlock(&chan->pendq_lock);
+
+	if (found) {
+		/* retrieve the passed down msg_id */
+		set_msg_id(&entry->to_h_msg[0], entry->usr_msg_id);
+		rsp_length = entry->to_h_blks * VK_MSGQ_BLK_SIZE;
+		if (copy_to_user(buf, entry->to_h_msg, rsp_length) == 0)
+			rc = rsp_length;
+
+		bcm_vk_free_wkent(dev, entry);
+	} else if (rc == -EMSGSIZE) {
+		struct vk_msg_blk tmp_msg = entry->to_h_msg[0];
+
+		/*
+		 * in this case, return just the first block, so
+		 * that app knows what size it is looking for.
+		 */
+		set_msg_id(&tmp_msg, entry->usr_msg_id);
+		tmp_msg.size = entry->to_h_blks - 1;
+		if (copy_to_user(buf, &tmp_msg, VK_MSGQ_BLK_SIZE) != 0) {
+			dev_err(dev, "Error return 1st block in -EMSGSIZE\n");
+			rc = -EFAULT;
+		}
+	}
+	return rc;
+}
+
+ssize_t bcm_vk_write(struct file *p_file,
+		     const char __user *buf,
+		     size_t count,
+		     loff_t *f_pos)
+{
+	ssize_t rc;
+	struct bcm_vk_ctx *ctx = p_file->private_data;
+	struct bcm_vk *vk = container_of(ctx->miscdev, struct bcm_vk,
+					 miscdev);
+	struct bcm_vk_msgq __iomem *msgq;
+	struct device *dev = &vk->pdev->dev;
+	struct bcm_vk_wkent *entry;
+	u32 sgl_extra_blks;
+	u32 q_num;
+	u32 msg_size;
+	u32 msgq_size;
+
+	if (!bcm_vk_drv_access_ok(vk))
+		return -EPERM;
+
+	dev_dbg(dev, "Msg count %zu\n", count);
+
+	/* first, do sanity check where count should be multiple of basic blk */
+	if (count & (VK_MSGQ_BLK_SIZE - 1)) {
+		dev_err(dev, "Failure with size %zu not multiple of %zu\n",
+			count, VK_MSGQ_BLK_SIZE);
+		rc = -EINVAL;
+		goto write_err;
+	}
+
+	/* allocate the work entry + buffer for size count and inband sgl */
+	entry = kzalloc(sizeof(*entry) + count + vk->ib_sgl_size,
+			GFP_KERNEL);
+	if (!entry) {
+		rc = -ENOMEM;
+		goto write_err;
+	}
+
+	/* now copy msg from user space, and then formulate the work entry */
+	if (copy_from_user(&entry->to_v_msg[0], buf, count)) {
+		rc = -EFAULT;
+		goto write_free_ent;
+	}
+
+	entry->to_v_blks = count >> VK_MSGQ_BLK_SZ_SHIFT;
+	entry->ctx = ctx;
+
+	/* do a check on the blk size which could not exceed queue space */
+	q_num = get_q_num(&entry->to_v_msg[0]);
+	msgq = vk->to_v_msg_chan.msgq[q_num];
+	msgq_size = readl_relaxed(&msgq->size);
+	if (entry->to_v_blks + (vk->ib_sgl_size >> VK_MSGQ_BLK_SZ_SHIFT)
+	    > (msgq_size - 1)) {
+		dev_err(dev, "Blk size %d exceed max queue size allowed %d\n",
+			entry->to_v_blks, msgq_size - 1);
+		rc = -EINVAL;
+		goto write_free_ent;
+	}
+
+	/* Use internal message id */
+	entry->usr_msg_id = get_msg_id(&entry->to_v_msg[0]);
+	rc = bcm_vk_get_msg_id(vk);
+	if (rc == VK_MSG_ID_OVERFLOW) {
+		dev_err(dev, "msg_id overflow\n");
+		rc = -EOVERFLOW;
+		goto write_free_ent;
+	}
+	set_msg_id(&entry->to_v_msg[0], rc);
+	ctx->q_num = q_num;
+
+	dev_dbg(dev,
+		"[Q-%d]Message ctx id %d, usr_msg_id 0x%x sent msg_id 0x%x\n",
+		ctx->q_num, ctx->idx, entry->usr_msg_id,
+		get_msg_id(&entry->to_v_msg[0]));
+
+	if (entry->to_v_msg[0].function_id == VK_FID_TRANS_BUF) {
+		/* Convert any pointers to sg list */
+		unsigned int num_planes;
+		int dir;
+		struct _vk_data *data;
+
+		num_planes = entry->to_v_msg[0].cmd & VK_CMD_PLANES_MASK;
+		if ((entry->to_v_msg[0].cmd & VK_CMD_MASK) == VK_CMD_DOWNLOAD)
+			dir = DMA_FROM_DEVICE;
+		else
+			dir = DMA_TO_DEVICE;
+
+		/* Calculate vk_data location */
+		/* Go to end of the message */
+		msg_size = entry->to_v_msg[0].size;
+		if (msg_size > entry->to_v_blks) {
+			rc = -EMSGSIZE;
+			goto write_free_msgid;
+		}
+
+		data = (struct _vk_data *)&entry->to_v_msg[msg_size + 1];
+
+		/* Now back up to the start of the pointers */
+		data -= num_planes;
+
+		/* Convert user addresses to DMA SG List */
+		rc = bcm_vk_sg_alloc(dev, entry->dma, dir, data, num_planes);
+		if (rc)
+			goto write_free_msgid;
+
+		atomic_inc(&ctx->dma_cnt);
+		/* try to embed inband sgl */
+		sgl_extra_blks = bcm_vk_append_ib_sgl(vk, entry, data,
+						      num_planes);
+		entry->to_v_blks += sgl_extra_blks;
+		entry->to_v_msg[0].size += sgl_extra_blks;
+	} else if (entry->to_v_msg[0].function_id == VK_FID_INIT &&
+		   entry->to_v_msg[0].context_id == VK_NEW_CTX) {
+		/*
+		 * Init happens in 2 stages, only the first stage contains the
+		 * pid that needs translating.
+		 */
+		pid_t org_pid, pid;
+
+		/*
+		 * translate the pid into the unique host space as user
+		 * may run sessions inside containers or process
+		 * namespaces.
+		 */
+#define VK_MSG_PID_MASK 0xffffff00
+#define VK_MSG_PID_SH   8
+		org_pid = (entry->to_v_msg[0].arg & VK_MSG_PID_MASK)
+			   >> VK_MSG_PID_SH;
+
+		pid = task_tgid_nr(current);
+		entry->to_v_msg[0].arg =
+			(entry->to_v_msg[0].arg & ~VK_MSG_PID_MASK) |
+			(pid << VK_MSG_PID_SH);
+		if (org_pid != pid)
+			dev_dbg(dev, "In PID 0x%x(%d), converted PID 0x%x(%d)\n",
+				org_pid, org_pid, pid, pid);
+	}
+
+	/*
+	 * store work entry to pending queue until a response is received.
+	 * This needs to be done before enqueuing the message
+	 */
+	bcm_vk_append_pendq(&vk->to_v_msg_chan, q_num, entry);
+
+	rc = bcm_to_v_msg_enqueue(vk, entry);
+	if (rc) {
+		dev_err(dev, "Fail to enqueue msg to to_v queue\n");
+
+		/* remove message from pending list */
+		entry = bcm_vk_dequeue_pending
+			       (vk,
+				&vk->to_v_msg_chan,
+				q_num,
+				get_msg_id(&entry->to_v_msg[0]));
+		goto write_free_ent;
+	}
+
+	return count;
+
+write_free_msgid:
+	bcm_vk_msgid_bitmap_clear(vk, get_msg_id(&entry->to_v_msg[0]), 1);
+write_free_ent:
+	kfree(entry);
+write_err:
+	return rc;
+}
+
+__poll_t bcm_vk_poll(struct file *p_file, struct poll_table_struct *wait)
+{
+	__poll_t ret = 0;
+	int cnt;
+	struct bcm_vk_ctx *ctx = p_file->private_data;
+	struct bcm_vk *vk = container_of(ctx->miscdev, struct bcm_vk, miscdev);
+	struct device *dev = &vk->pdev->dev;
+
+	poll_wait(p_file, &ctx->rd_wq, wait);
+
+	cnt = atomic_read(&ctx->pend_cnt);
+	if (cnt) {
+		ret = (__force __poll_t)(POLLIN | POLLRDNORM);
+		if (cnt < 0) {
+			dev_err(dev, "Error cnt %d, setting back to 0", cnt);
+			atomic_set(&ctx->pend_cnt, 0);
+		}
+	}
+
+	return ret;
+}
+
 int bcm_vk_release(struct inode *inode, struct file *p_file)
 {
 	int ret;
 	struct bcm_vk_ctx *ctx = p_file->private_data;
 	struct bcm_vk *vk = container_of(ctx->miscdev, struct bcm_vk, miscdev);
+	struct device *dev = &vk->pdev->dev;
+	pid_t pid = ctx->pid;
+	int dma_cnt;
+	unsigned long timeout, start_time;
+
+	/*
+	 * if there are outstanding DMA transactions, need to delay long enough
+	 * to ensure that the card side would have stopped touching the host buffer
+	 * and its SGL list.  A race condition could happen if the host app is killed
+	 * abruptly, eg kill -9, while some DMA transfer orders are still inflight.
+	 * Nothing could be done except for a delay as host side is running in a
+	 * completely async fashion.
+	 */
+	start_time = jiffies;
+	timeout = start_time + msecs_to_jiffies(BCM_VK_DMA_DRAIN_MAX_MS);
+	do {
+		if (time_after(jiffies, timeout)) {
+			dev_warn(dev, "%d dma still pending for [fd-%d] pid %d\n",
+				 dma_cnt, ctx->idx, pid);
+			break;
+		}
+		dma_cnt = atomic_read(&ctx->dma_cnt);
+		cpu_relax();
+		cond_resched();
+	} while (dma_cnt);
+	dev_dbg(dev, "Draining for [fd-%d] pid %d - delay %d ms\n",
+		ctx->idx, pid, jiffies_to_msecs(jiffies - start_time));
+
+	bcm_vk_drain_all_pend(&vk->pdev->dev, &vk->to_v_msg_chan, ctx);
+	bcm_vk_drain_all_pend(&vk->pdev->dev, &vk->to_h_msg_chan, ctx);
 
 	ret = bcm_vk_free_ctx(vk, ctx);
+	if (ret == 0)
+		ret = bcm_vk_handle_last_sess(vk, pid, ctx->q_num);
+	else
+		ret = 0;
 
 	kref_put(&vk->kref, bcm_vk_release_data);
 
 	return ret;
 }
 
+int bcm_vk_msg_init(struct bcm_vk *vk)
+{
+	struct device *dev = &vk->pdev->dev;
+	int ret;
+
+	if (bcm_vk_data_init(vk)) {
+		dev_err(dev, "Error initializing internal data structures\n");
+		return -EINVAL;
+	}
+
+	if (bcm_vk_msg_chan_init(&vk->to_v_msg_chan) ||
+	    bcm_vk_msg_chan_init(&vk->to_h_msg_chan)) {
+		dev_err(dev, "Error initializing communication channel\n");
+		return -EIO;
+	}
+
+	/* read msgq info if ready */
+	ret = bcm_vk_sync_msgq(vk, false);
+	if (ret && (ret != -EAGAIN)) {
+		dev_err(dev, "Error reading comm msg Q info\n");
+		return -EIO;
+	}
+
+	return 0;
+}
+
+void bcm_vk_msg_remove(struct bcm_vk *vk)
+{
+	bcm_vk_blk_drv_access(vk);
+
+	/* drain all pending items */
+	bcm_vk_drain_all_pend(&vk->pdev->dev, &vk->to_v_msg_chan, NULL);
+	bcm_vk_drain_all_pend(&vk->pdev->dev, &vk->to_h_msg_chan, NULL);
+}
+