USB: xhci: Allocate and address USB devices

xHCI needs to get a "Slot ID" from the host controller and allocate other
data structures for every USB device.  Make usb_alloc_dev() and
usb_release_dev() allocate and free these device structures.  After
setting up the xHC device structures, usb_alloc_dev() must wait for the
hardware to respond to an Enable Slot command.  usb_alloc_dev() fires off
a Disable Slot command and does not wait for it to complete.

When the USB core wants to choose an address for the device, the xHCI
driver must issue a Set Address command and wait for an event for that
command.

Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

drivers/usb/host/xhci-dbg.c

@@ -410,3 +410,82 @@ void xhci_dbg_cmd_ptrs(struct xhci_hcd *xhci)
 	val = xhci_readl(xhci, &xhci->op_regs->cmd_ring[1]);
 	xhci_dbg(xhci, "// xHC command ring deq ptr high bits = 0x%x\n", val);
 }
+
+void xhci_dbg_ctx(struct xhci_hcd *xhci, struct xhci_device_control *ctx, dma_addr_t dma, unsigned int last_ep)
+{
+	int i, j;
+	int last_ep_ctx = 31;
+	/* Fields are 32 bits wide, DMA addresses are in bytes */
+	int field_size = 32 / 8;
+
+	xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - drop flags\n",
+			(unsigned int) &ctx->drop_flags,
+			dma, ctx->drop_flags);
+	dma += field_size;
+	xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - add flags\n",
+			(unsigned int) &ctx->add_flags,
+			dma, ctx->add_flags);
+	dma += field_size;
+	for (i = 0; i > 6; ++i) {
+		xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - rsvd[%d]\n",
+				(unsigned int) &ctx->rsvd[i],
+				dma, ctx->rsvd[i], i);
+		dma += field_size;
+	}
+
+	xhci_dbg(xhci, "Slot Context:\n");
+	xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - dev_info\n",
+			(unsigned int) &ctx->slot.dev_info,
+			dma, ctx->slot.dev_info);
+	dma += field_size;
+	xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - dev_info2\n",
+			(unsigned int) &ctx->slot.dev_info2,
+			dma, ctx->slot.dev_info2);
+	dma += field_size;
+	xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - tt_info\n",
+			(unsigned int) &ctx->slot.tt_info,
+			dma, ctx->slot.tt_info);
+	dma += field_size;
+	xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - dev_state\n",
+			(unsigned int) &ctx->slot.dev_state,
+			dma, ctx->slot.dev_state);
+	dma += field_size;
+	for (i = 0; i > 4; ++i) {
+		xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - rsvd[%d]\n",
+				(unsigned int) &ctx->slot.reserved[i],
+				dma, ctx->slot.reserved[i], i);
+		dma += field_size;
+	}
+
+	if (last_ep < 31)
+		last_ep_ctx = last_ep + 1;
+	for (i = 0; i < last_ep_ctx; ++i) {
+		xhci_dbg(xhci, "Endpoint %02d Context:\n", i);
+		xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - ep_info\n",
+				(unsigned int) &ctx->ep[i].ep_info,
+				dma, ctx->ep[i].ep_info);
+		dma += field_size;
+		xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - ep_info2\n",
+				(unsigned int) &ctx->ep[i].ep_info2,
+				dma, ctx->ep[i].ep_info2);
+		dma += field_size;
+		xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - deq[0]\n",
+				(unsigned int) &ctx->ep[i].deq[0],
+				dma, ctx->ep[i].deq[0]);
+		dma += field_size;
+		xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - deq[1]\n",
+				(unsigned int) &ctx->ep[i].deq[1],
+				dma, ctx->ep[i].deq[1]);
+		dma += field_size;
+		xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - tx_info\n",
+				(unsigned int) &ctx->ep[i].tx_info,
+				dma, ctx->ep[i].tx_info);
+		dma += field_size;
+		for (j = 0; j < 3; ++j) {
+			xhci_dbg(xhci, "@%08x (virt) @%08x (dma) %#08x - rsvd[%d]\n",
+					(unsigned int) &ctx->ep[i].reserved[j],
+					dma, ctx->ep[i].reserved[j], j);
+			dma += field_size;
+		}
+	}
+}

drivers/usb/host/xhci-hcd.c

@@ -318,6 +318,16 @@ void event_ring_work(unsigned long arg)
 	xhci_debug_segment(xhci, xhci->cmd_ring->deq_seg);
 	xhci_dbg_ring_ptrs(xhci, xhci->cmd_ring);
 	xhci_dbg_cmd_ptrs(xhci);
+	for (i = 0; i < MAX_HC_SLOTS; ++i) {
+		if (xhci->devs[i]) {
+			for (j = 0; j < 31; ++j) {
+				if (xhci->devs[i]->ep_rings[j]) {
+					xhci_dbg(xhci, "Dev %d endpoint ring %d:\n", i, j);
+					xhci_debug_segment(xhci, xhci->devs[i]->ep_rings[j]->deq_seg);
+				}
+			}
+		}
+	}
 
 	if (xhci->noops_submitted != NUM_TEST_NOOPS)
 		if (setup_one_noop(xhci))
@@ -499,6 +509,197 @@ void xhci_shutdown(struct usb_hcd *hcd)
 
 /*-------------------------------------------------------------------------*/
 
+/*
+ * At this point, the struct usb_device is about to go away, the device has
+ * disconnected, and all traffic has been stopped and the endpoints have been
+ * disabled.  Free any HC data structures associated with that device.
+ */
+void xhci_free_dev(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	unsigned long flags;
+
+	if (udev->slot_id == 0)
+		return;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	if (queue_slot_control(xhci, TRB_DISABLE_SLOT, udev->slot_id)) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		xhci_dbg(xhci, "FIXME: allocate a command ring segment\n");
+		return;
+	}
+	ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	/*
+	 * Event command completion handler will free any data structures
+	 * associated with the slot
+	 */
+}
+
+/*
+ * Returns 0 if the xHC ran out of device slots, the Enable Slot command
+ * timed out, or allocating memory failed.  Returns 1 on success.
+ */
+int xhci_alloc_dev(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	unsigned long flags;
+	int timeleft;
+	int ret;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	ret = queue_slot_control(xhci, TRB_ENABLE_SLOT, 0);
+	if (ret) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		xhci_dbg(xhci, "FIXME: allocate a command ring segment\n");
+		return 0;
+	}
+	ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* XXX: how much time for xHC slot assignment? */
+	timeleft = wait_for_completion_interruptible_timeout(&xhci->addr_dev,
+			USB_CTRL_SET_TIMEOUT);
+	if (timeleft <= 0) {
+		xhci_warn(xhci, "%s while waiting for a slot\n",
+				timeleft == 0 ? "Timeout" : "Signal");
+		/* FIXME cancel the enable slot request */
+		return 0;
+	}
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	if (!xhci->slot_id) {
+		xhci_err(xhci, "Error while assigning device slot ID\n");
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return 0;
+	}
+	if (!xhci_alloc_virt_device(xhci, xhci->slot_id, udev, GFP_KERNEL)) {
+		/* Disable slot, if we can do it without mem alloc */
+		xhci_warn(xhci, "Could not allocate xHCI USB device data structures\n");
+		if (!queue_slot_control(xhci, TRB_DISABLE_SLOT, udev->slot_id))
+			ring_cmd_db(xhci);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return 0;
+	}
+	udev->slot_id = xhci->slot_id;
+	/* Is this a LS or FS device under a HS hub? */
+	/* Hub or peripherial? */
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return 1;
+}
+
+/*
+ * Issue an Address Device command (which will issue a SetAddress request to
+ * the device).
+ * We should be protected by the usb_address0_mutex in khubd's hub_port_init, so
+ * we should only issue and wait on one address command at the same time.
+ *
+ * We add one to the device address issued by the hardware because the USB core
+ * uses address 1 for the root hubs (even though they're not really devices).
+ */
+int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	unsigned long flags;
+	int timeleft;
+	struct xhci_virt_device *virt_dev;
+	int ret = 0;
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	u32 temp;
+
+	if (!udev->slot_id) {
+		xhci_dbg(xhci, "Bad Slot ID %d\n", udev->slot_id);
+		return -EINVAL;
+	}
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	virt_dev = xhci->devs[udev->slot_id];
+
+	/* If this is a Set Address to an unconfigured device, setup ep 0 */
+	if (!udev->config)
+		xhci_setup_addressable_virt_dev(xhci, udev);
+	/* Otherwise, assume the core has the device configured how it wants */
+
+	ret = queue_address_device(xhci, virt_dev->in_ctx_dma, udev->slot_id);
+	if (ret) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		xhci_dbg(xhci, "FIXME: allocate a command ring segment\n");
+		return ret;
+	}
+	ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* ctrl tx can take up to 5 sec; XXX: need more time for xHC? */
+	timeleft = wait_for_completion_interruptible_timeout(&xhci->addr_dev,
+			USB_CTRL_SET_TIMEOUT);
+	/* FIXME: From section 4.3.4: "Software shall be responsible for timing
+	 * the SetAddress() "recovery interval" required by USB and aborting the
+	 * command on a timeout.
+	 */
+	if (timeleft <= 0) {
+		xhci_warn(xhci, "%s while waiting for a slot\n",
+				timeleft == 0 ? "Timeout" : "Signal");
+		/* FIXME cancel the address device command */
+		return -ETIME;
+	}
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	switch (virt_dev->cmd_status) {
+	case COMP_CTX_STATE:
+	case COMP_EBADSLT:
+		xhci_err(xhci, "Setup ERROR: address device command for slot %d.\n",
+				udev->slot_id);
+		ret = -EINVAL;
+		break;
+	case COMP_TX_ERR:
+		dev_warn(&udev->dev, "Device not responding to set address.\n");
+		ret = -EPROTO;
+		break;
+	case COMP_SUCCESS:
+		xhci_dbg(xhci, "Successful Address Device command\n");
+		break;
+	default:
+		xhci_err(xhci, "ERROR: unexpected command completion "
+				"code 0x%x.\n", virt_dev->cmd_status);
+		ret = -EINVAL;
+		break;
+	}
+	if (ret) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return ret;
+	}
+	temp = xhci_readl(xhci, &xhci->op_regs->dcbaa_ptr[0]);
+	xhci_dbg(xhci, "Op regs DCBAA ptr[0] = %#08x\n", temp);
+	temp = xhci_readl(xhci, &xhci->op_regs->dcbaa_ptr[1]);
+	xhci_dbg(xhci, "Op regs DCBAA ptr[1] = %#08x\n", temp);
+	xhci_dbg(xhci, "Slot ID %d dcbaa entry[0] @%08x = %#08x\n",
+			udev->slot_id,
+			(unsigned int) &xhci->dcbaa->dev_context_ptrs[2*udev->slot_id],
+			xhci->dcbaa->dev_context_ptrs[2*udev->slot_id]);
+	xhci_dbg(xhci, "Slot ID %d dcbaa entry[1] @%08x = %#08x\n",
+			udev->slot_id,
+			(unsigned int) &xhci->dcbaa->dev_context_ptrs[2*udev->slot_id+1],
+			xhci->dcbaa->dev_context_ptrs[2*udev->slot_id+1]);
+	xhci_dbg(xhci, "Output Context DMA address = %#08x\n",
+			virt_dev->out_ctx_dma);
+	xhci_dbg(xhci, "Slot ID %d Input Context:\n", udev->slot_id);
+	xhci_dbg_ctx(xhci, virt_dev->in_ctx, virt_dev->in_ctx_dma, 2);
+	xhci_dbg(xhci, "Slot ID %d Output Context:\n", udev->slot_id);
+	xhci_dbg_ctx(xhci, virt_dev->out_ctx, virt_dev->out_ctx_dma, 2);
+	/*
+	 * USB core uses address 1 for the roothubs, so we add one to the
+	 * address given back to us by the HC.
+	 */
+	udev->devnum = (virt_dev->out_ctx->slot.dev_state & DEV_ADDR_MASK) + 1;
+	/* FIXME: Zero the input context control for later use? */
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	xhci_dbg(xhci, "Device address = %d\n", udev->devnum);
+	/* XXX Meh, not sure if anyone else but choose_address uses this. */
+	set_bit(udev->devnum, udev->bus->devmap.devicemap);
+
+	return 0;
+}
+
 int xhci_get_frame(struct usb_hcd *hcd)
 {
 	struct xhci_hcd *xhci = hcd_to_xhci(hcd);

drivers/usb/host/xhci-mem.c

@@ -188,12 +188,187 @@ fail:
 	return 0;
 }
 
+void xhci_free_virt_device(struct xhci_hcd *xhci, int slot_id)
+{
+	struct xhci_virt_device *dev;
+	int i;
+
+	/* Slot ID 0 is reserved */
+	if (slot_id == 0 || !xhci->devs[slot_id])
+		return;
+
+	dev = xhci->devs[slot_id];
+	xhci->dcbaa->dev_context_ptrs[2*slot_id] = 0;
+	xhci->dcbaa->dev_context_ptrs[2*slot_id + 1] = 0;
+	if (!dev)
+		return;
+
+	for (i = 0; i < 31; ++i)
+		if (dev->ep_rings[i])
+			xhci_ring_free(xhci, dev->ep_rings[i]);
+
+	if (dev->in_ctx)
+		dma_pool_free(xhci->device_pool,
+				dev->in_ctx, dev->in_ctx_dma);
+	if (dev->out_ctx)
+		dma_pool_free(xhci->device_pool,
+				dev->out_ctx, dev->out_ctx_dma);
+	kfree(xhci->devs[slot_id]);
+	xhci->devs[slot_id] = 0;
+}
+
+int xhci_alloc_virt_device(struct xhci_hcd *xhci, int slot_id,
+		struct usb_device *udev, gfp_t flags)
+{
+	dma_addr_t	dma;
+	struct xhci_virt_device *dev;
+
+	/* Slot ID 0 is reserved */
+	if (slot_id == 0 || xhci->devs[slot_id]) {
+		xhci_warn(xhci, "Bad Slot ID %d\n", slot_id);
+		return 0;
+	}
+
+	xhci->devs[slot_id] = kzalloc(sizeof(*xhci->devs[slot_id]), flags);
+	if (!xhci->devs[slot_id])
+		return 0;
+	dev = xhci->devs[slot_id];
+
+	/* Allocate the (output) device context that will be used in the HC */
+	dev->out_ctx = dma_pool_alloc(xhci->device_pool, flags, &dma);
+	if (!dev->out_ctx)
+		goto fail;
+	dev->out_ctx_dma = dma;
+	xhci_dbg(xhci, "Slot %d output ctx = 0x%x (dma)\n", slot_id, dma);
+	memset(dev->out_ctx, 0, sizeof(*dev->out_ctx));
+
+	/* Allocate the (input) device context for address device command */
+	dev->in_ctx = dma_pool_alloc(xhci->device_pool, flags, &dma);
+	if (!dev->in_ctx)
+		goto fail;
+	dev->in_ctx_dma = dma;
+	xhci_dbg(xhci, "Slot %d input ctx = 0x%x (dma)\n", slot_id, dma);
+	memset(dev->in_ctx, 0, sizeof(*dev->in_ctx));
+
+	/* Allocate endpoint 0 ring */
+	dev->ep_rings[0] = xhci_ring_alloc(xhci, 1, true, flags);
+	if (!dev->ep_rings[0])
+		goto fail;
+
+	/*
+	 * Point to output device context in dcbaa; skip the output control
+	 * context, which is eight 32 bit fields (or 32 bytes long)
+	 */
+	xhci->dcbaa->dev_context_ptrs[2*slot_id] =
+		(u32) dev->out_ctx_dma + (32);
+	xhci_dbg(xhci, "Set slot id %d dcbaa entry 0x%x to 0x%x\n",
+			slot_id,
+			(unsigned int) &xhci->dcbaa->dev_context_ptrs[2*slot_id],
+			dev->out_ctx_dma);
+	xhci->dcbaa->dev_context_ptrs[2*slot_id + 1] = 0;
+
+	return 1;
+fail:
+	xhci_free_virt_device(xhci, slot_id);
+	return 0;
+}
+
+/* Setup an xHCI virtual device for a Set Address command */
+int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *udev)
+{
+	struct xhci_virt_device *dev;
+	struct xhci_ep_ctx	*ep0_ctx;
+	struct usb_device	*top_dev;
+
+	dev = xhci->devs[udev->slot_id];
+	/* Slot ID 0 is reserved */
+	if (udev->slot_id == 0 || !dev) {
+		xhci_warn(xhci, "Slot ID %d is not assigned to this device\n",
+				udev->slot_id);
+		return -EINVAL;
+	}
+	ep0_ctx = &dev->in_ctx->ep[0];
+
+	/* 2) New slot context and endpoint 0 context are valid*/
+	dev->in_ctx->add_flags = SLOT_FLAG | EP0_FLAG;
+
+	/* 3) Only the control endpoint is valid - one endpoint context */
+	dev->in_ctx->slot.dev_info |= LAST_CTX(1);
+
+	switch (udev->speed) {
+	case USB_SPEED_SUPER:
+		dev->in_ctx->slot.dev_info |= (u32) udev->route;
+		dev->in_ctx->slot.dev_info |= (u32) SLOT_SPEED_SS;
+		break;
+	case USB_SPEED_HIGH:
+		dev->in_ctx->slot.dev_info |= (u32) SLOT_SPEED_HS;
+		break;
+	case USB_SPEED_FULL:
+		dev->in_ctx->slot.dev_info |= (u32) SLOT_SPEED_FS;
+		break;
+	case USB_SPEED_LOW:
+		dev->in_ctx->slot.dev_info |= (u32) SLOT_SPEED_LS;
+		break;
+	case USB_SPEED_VARIABLE:
+		xhci_dbg(xhci, "FIXME xHCI doesn't support wireless speeds\n");
+		return -EINVAL;
+		break;
+	default:
+		/* Speed was set earlier, this shouldn't happen. */
+		BUG();
+	}
+	/* Find the root hub port this device is under */
+	for (top_dev = udev; top_dev->parent && top_dev->parent->parent;
+			top_dev = top_dev->parent)
+		/* Found device below root hub */;
+	dev->in_ctx->slot.dev_info2 |= (u32) ROOT_HUB_PORT(top_dev->portnum);
+	xhci_dbg(xhci, "Set root hub portnum to %d\n", top_dev->portnum);
+
+	/* Is this a LS/FS device under a HS hub? */
+	/*
+	 * FIXME: I don't think this is right, where does the TT info for the
+	 * roothub or parent hub come from?
+	 */
+	if ((udev->speed == USB_SPEED_LOW || udev->speed == USB_SPEED_FULL) &&
+			udev->tt) {
+		dev->in_ctx->slot.tt_info = udev->tt->hub->slot_id;
+		dev->in_ctx->slot.tt_info |= udev->ttport << 8;
+	}
+	xhci_dbg(xhci, "udev->tt = 0x%x\n", (unsigned int) udev->tt);
+	xhci_dbg(xhci, "udev->ttport = 0x%x\n", udev->ttport);
+
+	/* Step 4 - ring already allocated */
+	/* Step 5 */
+	ep0_ctx->ep_info2 = EP_TYPE(CTRL_EP);
+	/*
+	 * See section 4.3 bullet 6:
+	 * The default Max Packet size for ep0 is "8 bytes for a USB2
+	 * LS/FS/HS device or 512 bytes for a USB3 SS device"
+	 * XXX: Not sure about wireless USB devices.
+	 */
+	if (udev->speed == USB_SPEED_SUPER)
+		ep0_ctx->ep_info2 |= MAX_PACKET(512);
+	else
+		ep0_ctx->ep_info2 |= MAX_PACKET(8);
+	/* EP 0 can handle "burst" sizes of 1, so Max Burst Size field is 0 */
+	ep0_ctx->ep_info2 |= MAX_BURST(0);
+	ep0_ctx->ep_info2 |= ERROR_COUNT(3);
+
+	ep0_ctx->deq[0] =
+		dev->ep_rings[0]->first_seg->dma;
+	ep0_ctx->deq[0] |= dev->ep_rings[0]->cycle_state;
+	ep0_ctx->deq[1] = 0;
+
+	/* Steps 7 and 8 were done in xhci_alloc_virt_device() */
+
+	return 0;
+}
+
 void xhci_mem_cleanup(struct xhci_hcd *xhci)
 {
 	struct pci_dev	*pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
 	int size;
-
-	/* XXX: Free all the segments in the various rings */
+	int i;
 
 	/* Free the Event Ring Segment Table and the actual Event Ring */
 	xhci_writel(xhci, 0, &xhci->ir_set->erst_size);
@@ -218,16 +393,27 @@ void xhci_mem_cleanup(struct xhci_hcd *xhci)
 		xhci_ring_free(xhci, xhci->cmd_ring);
 	xhci->cmd_ring = NULL;
 	xhci_dbg(xhci, "Freed command ring\n");
+
+	for (i = 1; i < MAX_HC_SLOTS; ++i)
+		xhci_free_virt_device(xhci, i);
+
 	if (xhci->segment_pool)
 		dma_pool_destroy(xhci->segment_pool);
 	xhci->segment_pool = NULL;
 	xhci_dbg(xhci, "Freed segment pool\n");
+
+	if (xhci->device_pool)
+		dma_pool_destroy(xhci->device_pool);
+	xhci->device_pool = NULL;
+	xhci_dbg(xhci, "Freed device context pool\n");
+
 	xhci_writel(xhci, 0, &xhci->op_regs->dcbaa_ptr[1]);
 	xhci_writel(xhci, 0, &xhci->op_regs->dcbaa_ptr[0]);
 	if (xhci->dcbaa)
 		pci_free_consistent(pdev, sizeof(*xhci->dcbaa),
 				xhci->dcbaa, xhci->dcbaa->dma);
 	xhci->dcbaa = NULL;
+
 	xhci->page_size = 0;
 	xhci->page_shift = 0;
 }
@@ -280,8 +466,8 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 		goto fail;
 	memset(xhci->dcbaa, 0, sizeof *(xhci->dcbaa));
 	xhci->dcbaa->dma = dma;
-	xhci_dbg(xhci, "// Setting device context base array address to 0x%x\n",
-			xhci->dcbaa->dma);
+	xhci_dbg(xhci, "// Device context base array address = 0x%x (DMA), 0x%x (virt)\n",
+			xhci->dcbaa->dma, (unsigned int) xhci->dcbaa);
 	xhci_writel(xhci, (u32) 0, &xhci->op_regs->dcbaa_ptr[1]);
 	xhci_writel(xhci, dma, &xhci->op_regs->dcbaa_ptr[0]);
 
@@ -293,7 +479,12 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	 */
 	xhci->segment_pool = dma_pool_create("xHCI ring segments", dev,
 			SEGMENT_SIZE, 64, xhci->page_size);
-	if (!xhci->segment_pool)
+	/* See Table 46 and Note on Figure 55 */
+	/* FIXME support 64-byte contexts */
+	xhci->device_pool = dma_pool_create("xHCI input/output contexts", dev,
+			sizeof(struct xhci_device_control),
+			64, xhci->page_size);
+	if (!xhci->segment_pool || !xhci->device_pool)
 		goto fail;
 
 	/* Set up the command ring to have one segments for now. */
@@ -385,6 +576,9 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	 * something other than the default (~1ms minimum between interrupts).
 	 * See section 5.5.1.2.
 	 */
+	init_completion(&xhci->addr_dev);
+	for (i = 0; i < MAX_HC_SLOTS; ++i)
+		xhci->devs[i] = 0;
 
 	return 0;
 fail:

drivers/usb/host/xhci-pci.c

@@ -108,6 +108,13 @@ static const struct hc_driver xhci_pci_hc_driver = {
 	.stop =			xhci_stop,
 	.shutdown =		xhci_shutdown,
 
+	/*
+	 * managing i/o requests and associated device resources
+	 */
+	.alloc_dev =		xhci_alloc_dev,
+	.free_dev =		xhci_free_dev,
+	.address_device =	xhci_address_device,
+
 	/*
 	 * scheduling support
 	 */

drivers/usb/host/xhci-ring.c

@@ -252,13 +252,10 @@ void ring_cmd_db(struct xhci_hcd *xhci)
 static void handle_cmd_completion(struct xhci_hcd *xhci,
 		struct xhci_event_cmd *event)
 {
+	int slot_id = TRB_TO_SLOT_ID(event->flags);
 	u64 cmd_dma;
 	dma_addr_t cmd_dequeue_dma;
 
-	/* Check completion code */
-	if (GET_COMP_CODE(event->status) != COMP_SUCCESS)
-		xhci_dbg(xhci, "WARN: unsuccessful no-op command\n");
-
 	cmd_dma = (((u64) event->cmd_trb[1]) << 32) + event->cmd_trb[0];
 	cmd_dequeue_dma = trb_virt_to_dma(xhci->cmd_ring->deq_seg,
 			xhci->cmd_ring->dequeue);
@@ -273,6 +270,21 @@ static void handle_cmd_completion(struct xhci_hcd *xhci,
 		return;
 	}
 	switch (xhci->cmd_ring->dequeue->generic.field[3] & TRB_TYPE_BITMASK) {
+	case TRB_TYPE(TRB_ENABLE_SLOT):
+		if (GET_COMP_CODE(event->status) == COMP_SUCCESS)
+			xhci->slot_id = slot_id;
+		else
+			xhci->slot_id = 0;
+		complete(&xhci->addr_dev);
+		break;
+	case TRB_TYPE(TRB_DISABLE_SLOT):
+		if (xhci->devs[slot_id])
+			xhci_free_virt_device(xhci, slot_id);
+		break;
+	case TRB_TYPE(TRB_ADDR_DEV):
+		xhci->devs[slot_id]->cmd_status = GET_COMP_CODE(event->status);
+		complete(&xhci->addr_dev);
+		break;
 	case TRB_TYPE(TRB_CMD_NOOP):
 		++xhci->noops_handled;
 		break;
@@ -400,3 +412,17 @@ void *setup_one_noop(struct xhci_hcd *xhci)
 	xhci->noops_submitted++;
 	return ring_cmd_db;
 }
+
+/* Queue a slot enable or disable request on the command ring */
+int queue_slot_control(struct xhci_hcd *xhci, u32 trb_type, u32 slot_id)
+{
+	return queue_command(xhci, 0, 0, 0,
+			TRB_TYPE(trb_type) | SLOT_ID_FOR_TRB(slot_id));
+}
+
+/* Queue an address device command TRB */
+int queue_address_device(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr, u32 slot_id)
+{
+	return queue_command(xhci, in_ctx_ptr, 0, 0,
+			TRB_TYPE(TRB_ADDR_DEV) | SLOT_ID_FOR_TRB(slot_id));
+}

drivers/usb/host/xhci.h

@@ -285,12 +285,21 @@ struct xhci_op_regs {
  * 4 - super speed
  * 5-15 reserved
  */
-#define DEV_SPEED_MASK		(0xf<<10)
+#define DEV_SPEED_MASK		(0xf << 10)
+#define	XDEV_FS			(0x1 << 10)
+#define	XDEV_LS			(0x2 << 10)
+#define	XDEV_HS			(0x3 << 10)
+#define	XDEV_SS			(0x4 << 10)
 #define DEV_UNDEFSPEED(p)	(((p) & DEV_SPEED_MASK) == (0x0<<10))
-#define DEV_FULLSPEED(p)	(((p) & DEV_SPEED_MASK) == (0x1<<10))
-#define DEV_LOWSPEED(p)		(((p) & DEV_SPEED_MASK) == (0x2<<10))
-#define DEV_HIGHSPEED(p)	(((p) & DEV_SPEED_MASK) == (0x3<<10))
-#define DEV_SUPERSPEED(p)	(((p) & DEV_SPEED_MASK) == (0x4<<10))
+#define DEV_FULLSPEED(p)	(((p) & DEV_SPEED_MASK) == XDEV_FS)
+#define DEV_LOWSPEED(p)		(((p) & DEV_SPEED_MASK) == XDEV_LS)
+#define DEV_HIGHSPEED(p)	(((p) & DEV_SPEED_MASK) == XDEV_HS)
+#define DEV_SUPERSPEED(p)	(((p) & DEV_SPEED_MASK) == XDEV_SS)
+/* Bits 20:23 in the Slot Context are the speed for the device */
+#define	SLOT_SPEED_FS		(XDEV_FS << 10)
+#define	SLOT_SPEED_LS		(XDEV_LS << 10)
+#define	SLOT_SPEED_HS		(XDEV_HS << 10)
+#define	SLOT_SPEED_SS		(XDEV_SS << 10)
 /* Port Indicator Control */
 #define PORT_LED_OFF	(0 << 14)
 #define PORT_LED_AMBER	(1 << 14)
@@ -471,14 +480,19 @@ struct xhci_slot_ctx {
 /* Set if the device is a hub - bit 26 */
 #define DEV_HUB		(0x1 << 26)
 /* Index of the last valid endpoint context in this device context - 27:31 */
-#define LAST_EP_MASK	(0x1f << 27)
-#define LAST_EP(p)	((p) << 27)
+#define LAST_CTX_MASK	(0x1f << 27)
+#define LAST_CTX(p)	((p) << 27)
+#define LAST_CTX_TO_EP_NUM(p)	(((p) >> 27) - 1)
+/* Plus one for the slot context flag */
+#define EPI_TO_FLAG(p)	(1 << ((p) + 1))
+#define SLOT_FLAG	(1 << 0)
+#define EP0_FLAG	(1 << 1)
 
 /* dev_info2 bitmasks */
 /* Max Exit Latency (ms) - worst case time to wake up all links in dev path */
 #define MAX_EXIT	(0xffff)
 /* Root hub port number that is needed to access the USB device */
-#define ROOT_HUB_PORT	(0xff << 16)
+#define ROOT_HUB_PORT(p)	(((p) & 0xff) << 16)
 
 /* tt_info bitmasks */
 /*
@@ -495,7 +509,7 @@ struct xhci_slot_ctx {
 
 /* dev_state bitmasks */
 /* USB device address - assigned by the HC */
-#define DEV_ADDR	(0xff)
+#define DEV_ADDR_MASK	(0xff)
 /* bits 8:26 reserved */
 /* Slot state */
 #define SLOT_STATE	(0x1f << 27)
@@ -507,12 +521,13 @@ struct xhci_slot_ctx {
  * @ep_info2:	information on endpoint type, max packet size, max burst size,
  * 		error count, and whether the HC will force an event for all
  * 		transactions.
- * @ep_ring:	64-bit ring address.  If the endpoint only defines one flow,
- * 		this points to the endpoint transfer ring.  Otherwise, it points
- * 		to a flow context array, which has a ring pointer for each flow.
- * @intr_target:
- * 		64-bit address of the Interrupter Target that will receive
- * 		events from this endpoint.
+ * @deq:	64-bit ring dequeue pointer address.  If the endpoint only
+ * 		defines one stream, this points to the endpoint transfer ring.
+ * 		Otherwise, it points to a stream context array, which has a
+ * 		ring pointer for each flow.
+ * @tx_info:
+ * 		Average TRB lengths for the endpoint ring and
+ * 		max payload within an Endpoint Service Interval Time (ESIT).
  *
  * Endpoint Context - section 6.2.1.2.  This assumes the HC uses 32-byte context
  * structures.  If the HC uses 64-byte contexts, there is an additional 32 bytes
@@ -521,12 +536,10 @@ struct xhci_slot_ctx {
 struct xhci_ep_ctx {
 	u32	ep_info;
 	u32	ep_info2;
-	/* 64-bit endpoint ring address */
-	u32	ep_ring[2];
-	/* 64-bit address of the interrupter target */
-	u32	intr_target[2];
+	u32	deq[2];
+	u32	tx_info;
 	/* offset 0x14 - 0x1f reserved for HC internal use */
-	u32	reserved[2];
+	u32	reserved[3];
 } __attribute__ ((packed));
 
 /* ep_info bitmasks */
@@ -589,6 +602,28 @@ struct xhci_device_control {
 #define	ADD_EP(x)	(0x1 << x)
 
 
+struct xhci_virt_device {
+	/*
+	 * Commands to the hardware are passed an "input context" that
+	 * tells the hardware what to change in its data structures.
+	 * The hardware will return changes in an "output context" that
+	 * software must allocate for the hardware.  We need to keep
+	 * track of input and output contexts separately because
+	 * these commands might fail and we don't trust the hardware.
+	 */
+	struct xhci_device_control	*out_ctx;
+	dma_addr_t			out_ctx_dma;
+	/* Used for addressing devices and configuration changes */
+	struct xhci_device_control	*in_ctx;
+	dma_addr_t			in_ctx_dma;
+	/* FIXME when stream support is added */
+	struct xhci_ring		*ep_rings[31];
+	dma_addr_t			ep_dma[31];
+	/* Status of the last command issued for this device */
+	u32				cmd_status;
+};
+
+
 /**
  * struct xhci_device_context_array
  * @dev_context_ptr	array of 64-bit DMA addresses for device contexts
@@ -711,6 +746,11 @@ struct xhci_event_cmd {
 	u32 flags;
 } __attribute__ ((packed));
 
+/* flags bitmasks */
+/* bits 16:23 are the virtual function ID */
+/* bits 24:31 are the slot ID */
+#define TRB_TO_SLOT_ID(p)	(((p) & (0xff<<24)) >> 24)
+#define SLOT_ID_FOR_TRB(p)	(((p) & 0xff) << 24)
 
 /* Port Status Change Event TRB fields */
 /* Port ID - bits 31:24 */
@@ -931,6 +971,11 @@ struct xhci_hcd {
 	struct xhci_ring	*cmd_ring;
 	struct xhci_ring	*event_ring;
 	struct xhci_erst	erst;
+	/* slot enabling and address device helpers */
+	struct completion	addr_dev;
+	int slot_id;
+	/* Internal mirror of the HW's dcbaa */
+	struct xhci_virt_device	*devs[MAX_HC_SLOTS];
 
 	/* DMA pools */
 	struct dma_pool	*device_pool;
@@ -1002,10 +1047,14 @@ void xhci_debug_ring(struct xhci_hcd *xhci, struct xhci_ring *ring);
 void xhci_dbg_erst(struct xhci_hcd *xhci, struct xhci_erst *erst);
 void xhci_dbg_cmd_ptrs(struct xhci_hcd *xhci);
 void xhci_dbg_ring_ptrs(struct xhci_hcd *xhci, struct xhci_ring *ring);
+void xhci_dbg_ctx(struct xhci_hcd *xhci, struct xhci_device_control *ctx, dma_addr_t dma, unsigned int last_ep);
 
 /* xHCI memory managment */
 void xhci_mem_cleanup(struct xhci_hcd *xhci);
 int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags);
+void xhci_free_virt_device(struct xhci_hcd *xhci, int slot_id);
+int xhci_alloc_virt_device(struct xhci_hcd *xhci, int slot_id, struct usb_device *udev, gfp_t flags);
+int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *udev);
 
 #ifdef CONFIG_PCI
 /* xHCI PCI glue */
@@ -1022,6 +1071,9 @@ void xhci_stop(struct usb_hcd *hcd);
 void xhci_shutdown(struct usb_hcd *hcd);
 int xhci_get_frame(struct usb_hcd *hcd);
 irqreturn_t xhci_irq(struct usb_hcd *hcd);
+int xhci_alloc_dev(struct usb_hcd *hcd, struct usb_device *udev);
+void xhci_free_dev(struct usb_hcd *hcd, struct usb_device *udev);
+int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev);
 
 /* xHCI ring, segment, TRB, and TD functions */
 dma_addr_t trb_virt_to_dma(struct xhci_segment *seg, union xhci_trb *trb);
@@ -1029,6 +1081,8 @@ void ring_cmd_db(struct xhci_hcd *xhci);
 void *setup_one_noop(struct xhci_hcd *xhci);
 void handle_event(struct xhci_hcd *xhci);
 void set_hc_event_deq(struct xhci_hcd *xhci);
+int queue_slot_control(struct xhci_hcd *xhci, u32 trb_type, u32 slot_id);
+int queue_address_device(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr, u32 slot_id);
 
 /* xHCI roothub code */
 int xhci_hub_control(struct usb_hcd *hcd, u16 typeReq, u16 wValue, u16 wIndex,