USB: xhci: Ring allocation and initialization.

Allocate basic xHCI host controller data structures.  For every xHC, there
is a command ring, an event ring, and a doorbell array.

The doorbell array is used to notify the host controller that work has
been enqueued onto one of the rings.  The host controller driver enqueues
commands on the command ring.  The HW enqueues command completion events
on the event ring and interrupts the system (currently using PCI
interrupts, although the xHCI HW will use MSI interrupts eventually).

All rings and the doorbell array must be allocated by the xHCI host
controller driver.

Each ring is comprised of one or more segments, which consists of 16-byte
Transfer Request Blocks (TRBs) that can be chained to form a Transfer
Descriptor (TD) that represents a multiple-buffer request.  Segments are
linked into a ring using Link TRBs, which means they are dynamically
growable.

The producer of the ring enqueues a TD by writing one or more TRBs in the
ring and toggling the TRB cycle bit for each TRB.  The consumer knows it
can process the TRB when the cycle bit matches its internal consumer cycle
state for the ring.  The consumer cycle state is toggled an odd amount of
times in the ring.

An example ring (a ring must have a minimum of 16 TRBs on it, but that's
too big to draw in ASCII art):

              chain  cycle
               bit    bit
 ------------------------
| TD A TRB 1 |  1  |  1  |<-------------  <-- consumer dequeue ptr
 ------------------------               |     consumer cycle state = 1
| TD A TRB 2 |  1  |  1  |              |
 ------------------------               |
| TD A TRB 3 |  0  |  1  |  segment 1   |
 ------------------------               |
| TD B TRB 1 |  1  |  1  |              |
 ------------------------               |
| TD B TRB 2 |  0  |  1  |              |
 ------------------------               |
| Link TRB   |  0  |  1  |-----         |
 ------------------------     |         |
                              |         |
              chain  cycle    |         |
               bit    bit     |         |
 ------------------------     |         |
| TD C TRB 1 |  0  |  1  |<----         |
 ------------------------               |
| TD D TRB 1 |  1  |  1  |              |
 ------------------------               |
| TD D TRB 2 |  1  |  1  |   segment 2  |
 ------------------------               |
| TD D TRB 3 |  1  |  1  |              |
 ------------------------               |
| TD D TRB 4 |  1  |  1  |              |
 ------------------------               |
| Link TRB   |  1  |  1  |-----         |
 ------------------------     |         |
                              |         |
              chain  cycle    |         |
               bit    bit     |         |
 ------------------------     |         |
| TD D TRB 5 |  1  |  1  |<----         |
 ------------------------               |
| TD D TRB 6 |  0  |  1  |              |
 ------------------------               |
| TD E TRB 1 |  0  |  1  |   segment 3  |
 ------------------------               |
|            |  0  |  0  |              | <-- producer enqueue ptr
 ------------------------               |
|            |  0  |  0  |              |
 ------------------------               |
| Link TRB   |  0  |  0  |---------------
 ------------------------

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

4 files changed, 686 insertions, 1 deletions

diff --git a/drivers/usb/host/xhci-dbg.c b/drivers/usb/host/xhci-dbg.c
index a7798b46049..5724683cef1 100644
--- a/drivers/usb/host/xhci-dbg.c
+++ b/drivers/usb/host/xhci-dbg.c
@@ -56,6 +56,8 @@ void xhci_dbg_regs(struct xhci_hcd *xhci)
 	temp = xhci_readl(xhci, &xhci->cap_regs->db_off);
 	xhci_dbg(xhci, "// @%x = 0x%x DBOFF\n",
 			(unsigned int) &xhci->cap_regs->db_off, temp);
+	xhci_dbg(xhci, "// Doorbell array at 0x%x:\n",
+			(unsigned int) xhci->dba);
 }
 
 void xhci_print_cap_regs(struct xhci_hcd *xhci)
@@ -227,3 +229,82 @@ void xhci_print_registers(struct xhci_hcd *xhci)
 	xhci_print_cap_regs(xhci);
 	xhci_print_op_regs(xhci);
 }
+
+
+/**
+ * Debug a segment with an xHCI ring.
+ *
+ * @return The Link TRB of the segment, or NULL if there is no Link TRB
+ * (which is a bug, since all segments must have a Link TRB).
+ *
+ * Prints out all TRBs in the segment, even those after the Link TRB.
+ *
+ * XXX: should we print out TRBs that the HC owns?  As long as we don't
+ * write, that should be fine...  We shouldn't expect that the memory pointed to
+ * by the TRB is valid at all.  Do we care about ones the HC owns?  Probably,
+ * for HC debugging.
+ */
+void xhci_debug_segment(struct xhci_hcd *xhci, struct xhci_segment *seg)
+{
+	int i;
+	u32 addr = (u32) seg->dma;
+	union xhci_trb *trb = seg->trbs;
+
+	for (i = 0; i < TRBS_PER_SEGMENT; ++i) {
+		trb = &seg->trbs[i];
+		xhci_dbg(xhci, "@%08x %08x %08x %08x %08x\n", addr,
+				(unsigned int) trb->link.segment_ptr[0],
+				(unsigned int) trb->link.segment_ptr[1],
+				(unsigned int) trb->link.intr_target,
+				(unsigned int) trb->link.control);
+		addr += sizeof(*trb);
+	}
+}
+
+/**
+ * Debugging for an xHCI ring, which is a queue broken into multiple segments.
+ *
+ * Print out each segment in the ring.  Check that the DMA address in
+ * each link segment actually matches the segment's stored DMA address.
+ * Check that the link end bit is only set at the end of the ring.
+ * Check that the dequeue and enqueue pointers point to real data in this ring
+ * (not some other ring).
+ */
+void xhci_debug_ring(struct xhci_hcd *xhci, struct xhci_ring *ring)
+{
+	/* FIXME: Throw an error if any segment doesn't have a Link TRB */
+	struct xhci_segment *seg;
+	struct xhci_segment *first_seg = ring->first_seg;
+	xhci_debug_segment(xhci, first_seg);
+
+	for (seg = first_seg->next; seg != first_seg; seg = seg->next)
+		xhci_debug_segment(xhci, seg);
+}
+
+void xhci_dbg_erst(struct xhci_hcd *xhci, struct xhci_erst *erst)
+{
+	u32 addr = (u32) erst->erst_dma_addr;
+	int i;
+	struct xhci_erst_entry *entry;
+
+	for (i = 0; i < erst->num_entries; ++i) {
+		entry = &erst->entries[i];
+		xhci_dbg(xhci, "@%08x %08x %08x %08x %08x\n",
+				(unsigned int) addr,
+				(unsigned int) entry->seg_addr[0],
+				(unsigned int) entry->seg_addr[1],
+				(unsigned int) entry->seg_size,
+				(unsigned int) entry->rsvd);
+		addr += sizeof(*entry);
+	}
+}
+
+void xhci_dbg_cmd_ptrs(struct xhci_hcd *xhci)
+{
+	u32 val;
+
+	val = xhci_readl(xhci, &xhci->op_regs->cmd_ring[0]);
+	xhci_dbg(xhci, "// xHC command ring deq ptr low bits + flags = 0x%x\n", val);
+	val = xhci_readl(xhci, &xhci->op_regs->cmd_ring[1]);
+	xhci_dbg(xhci, "// xHC command ring deq ptr high bits = 0x%x\n", val);
+}
diff --git a/drivers/usb/host/xhci-hcd.c b/drivers/usb/host/xhci-hcd.c
index 64fcc22e9d5..011f4781066 100644
--- a/drivers/usb/host/xhci-hcd.c
+++ b/drivers/usb/host/xhci-hcd.c
@@ -266,6 +266,11 @@ int xhci_run(struct usb_hcd *hcd)
 			&xhci->ir_set->irq_pending);
 	xhci_print_ir_set(xhci, xhci->ir_set, 0);
 
+	xhci_dbg(xhci, "Command ring memory map follows:\n");
+	xhci_debug_ring(xhci, xhci->cmd_ring);
+	xhci_dbg(xhci, "ERST memory map follows:\n");
+	xhci_dbg_erst(xhci, &xhci->erst);
+
 	temp = xhci_readl(xhci, &xhci->op_regs->command);
 	temp |= (CMD_RUN);
 	xhci_dbg(xhci, "// Turn on HC, cmd = 0x%x.\n",
diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
index 0e383f9c380..7cf15ca854b 100644
--- a/drivers/usb/host/xhci-mem.c
+++ b/drivers/usb/host/xhci-mem.c
@@ -21,18 +21,215 @@
  */
 
 #include <linux/usb.h>
+#include <linux/pci.h>
 
 #include "xhci.h"
 
+/*
+ * Allocates a generic ring segment from the ring pool, sets the dma address,
+ * initializes the segment to zero, and sets the private next pointer to NULL.
+ *
+ * Section 4.11.1.1:
+ * "All components of all Command and Transfer TRBs shall be initialized to '0'"
+ */
+static struct xhci_segment *xhci_segment_alloc(struct xhci_hcd *xhci, gfp_t flags)
+{
+	struct xhci_segment *seg;
+	dma_addr_t	dma;
+
+	seg = kzalloc(sizeof *seg, flags);
+	if (!seg)
+		return 0;
+	xhci_dbg(xhci, "Allocating priv segment structure at 0x%x\n",
+			(unsigned int) seg);
+
+	seg->trbs = dma_pool_alloc(xhci->segment_pool, flags, &dma);
+	if (!seg->trbs) {
+		kfree(seg);
+		return 0;
+	}
+	xhci_dbg(xhci, "// Allocating segment at 0x%x (virtual) 0x%x (DMA)\n",
+			(unsigned int) seg->trbs, (u32) dma);
+
+	memset(seg->trbs, 0, SEGMENT_SIZE);
+	seg->dma = dma;
+	seg->next = NULL;
+
+	return seg;
+}
+
+static void xhci_segment_free(struct xhci_hcd *xhci, struct xhci_segment *seg)
+{
+	if (!seg)
+		return;
+	if (seg->trbs) {
+		xhci_dbg(xhci, "Freeing DMA segment at 0x%x"
+				" (virtual) 0x%x (DMA)\n",
+				(unsigned int) seg->trbs, (u32) seg->dma);
+		dma_pool_free(xhci->segment_pool, seg->trbs, seg->dma);
+		seg->trbs = NULL;
+	}
+	xhci_dbg(xhci, "Freeing priv segment structure at 0x%x\n",
+			(unsigned int) seg);
+	kfree(seg);
+}
+
+/*
+ * Make the prev segment point to the next segment.
+ *
+ * Change the last TRB in the prev segment to be a Link TRB which points to the
+ * DMA address of the next segment.  The caller needs to set any Link TRB
+ * related flags, such as End TRB, Toggle Cycle, and no snoop.
+ */
+static void xhci_link_segments(struct xhci_hcd *xhci, struct xhci_segment *prev,
+		struct xhci_segment *next, bool link_trbs)
+{
+	u32 val;
+
+	if (!prev || !next)
+		return;
+	prev->next = next;
+	if (link_trbs) {
+		prev->trbs[TRBS_PER_SEGMENT-1].link.segment_ptr[0] = next->dma;
+
+		/* Set the last TRB in the segment to have a TRB type ID of Link TRB */
+		val = prev->trbs[TRBS_PER_SEGMENT-1].link.control;
+		val &= ~TRB_TYPE_BITMASK;
+		val |= TRB_TYPE(TRB_LINK);
+		prev->trbs[TRBS_PER_SEGMENT-1].link.control = val;
+	}
+	xhci_dbg(xhci, "Linking segment 0x%x to segment 0x%x (DMA)\n",
+			prev->dma, next->dma);
+}
+
+/* XXX: Do we need the hcd structure in all these functions? */
+static void xhci_ring_free(struct xhci_hcd *xhci, struct xhci_ring *ring)
+{
+	struct xhci_segment *seg;
+	struct xhci_segment *first_seg;
+
+	if (!ring || !ring->first_seg)
+		return;
+	first_seg = ring->first_seg;
+	seg = first_seg->next;
+	xhci_dbg(xhci, "Freeing ring at 0x%x\n", (unsigned int) ring);
+	while (seg != first_seg) {
+		struct xhci_segment *next = seg->next;
+		xhci_segment_free(xhci, seg);
+		seg = next;
+	}
+	xhci_segment_free(xhci, first_seg);
+	ring->first_seg = NULL;
+	kfree(ring);
+}
+
+/**
+ * Create a new ring with zero or more segments.
+ *
+ * Link each segment together into a ring.
+ * Set the end flag and the cycle toggle bit on the last segment.
+ * See section 4.9.1 and figures 15 and 16.
+ */
+static struct xhci_ring *xhci_ring_alloc(struct xhci_hcd *xhci,
+		unsigned int num_segs, bool link_trbs, gfp_t flags)
+{
+	struct xhci_ring	*ring;
+	struct xhci_segment	*prev;
+
+	ring = kzalloc(sizeof *(ring), flags);
+	xhci_dbg(xhci, "Allocating ring at 0x%x\n", (unsigned int) ring);
+	if (!ring)
+		return 0;
+
+	if (num_segs == 0)
+		return ring;
+
+	ring->first_seg = xhci_segment_alloc(xhci, flags);
+	if (!ring->first_seg)
+		goto fail;
+	num_segs--;
+
+	prev = ring->first_seg;
+	while (num_segs > 0) {
+		struct xhci_segment	*next;
+
+		next = xhci_segment_alloc(xhci, flags);
+		if (!next)
+			goto fail;
+		xhci_link_segments(xhci, prev, next, link_trbs);
+
+		prev = next;
+		num_segs--;
+	}
+	xhci_link_segments(xhci, prev, ring->first_seg, link_trbs);
+
+	if (link_trbs) {
+		/* See section 4.9.2.1 and 6.4.4.1 */
+		prev->trbs[TRBS_PER_SEGMENT-1].link.control |= (LINK_TOGGLE);
+		xhci_dbg(xhci, "Wrote link toggle flag to"
+				" segment 0x%x (virtual), 0x%x (DMA)\n",
+				(unsigned int) prev, (u32) prev->dma);
+	}
+	/* The ring is empty, so the enqueue pointer == dequeue pointer */
+	ring->enqueue = ring->first_seg->trbs;
+	ring->dequeue = ring->enqueue;
+	/* The ring is initialized to 0. The producer must write 1 to the cycle
+	 * bit to handover ownership of the TRB, so PCS = 1.  The consumer must
+	 * compare CCS to the cycle bit to check ownership, so CCS = 1.
+	 */
+	ring->cycle_state = 1;
+
+	return ring;
+
+fail:
+	xhci_ring_free(xhci, ring);
+	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 */
+
+	/* Free the Event Ring Segment Table and the actual Event Ring */
+	xhci_writel(xhci, 0, &xhci->ir_set->erst_size);
+	xhci_writel(xhci, 0, &xhci->ir_set->erst_base[1]);
+	xhci_writel(xhci, 0, &xhci->ir_set->erst_base[0]);
+	xhci_writel(xhci, 0, &xhci->ir_set->erst_dequeue[1]);
+	xhci_writel(xhci, 0, &xhci->ir_set->erst_dequeue[0]);
+	size = sizeof(struct xhci_erst_entry)*(xhci->erst.num_entries);
+	if (xhci->erst.entries)
+		pci_free_consistent(pdev, size,
+				xhci->erst.entries, xhci->erst.erst_dma_addr);
+	xhci->erst.entries = NULL;
+	xhci_dbg(xhci, "Freed ERST\n");
+	if (xhci->event_ring)
+		xhci_ring_free(xhci, xhci->event_ring);
+	xhci->event_ring = NULL;
+	xhci_dbg(xhci, "Freed event ring\n");
+
+	xhci_writel(xhci, 0, &xhci->op_regs->cmd_ring[1]);
+	xhci_writel(xhci, 0, &xhci->op_regs->cmd_ring[0]);
+	if (xhci->cmd_ring)
+		xhci_ring_free(xhci, xhci->cmd_ring);
+	xhci->cmd_ring = NULL;
+	xhci_dbg(xhci, "Freed command ring\n");
+	if (xhci->segment_pool)
+		dma_pool_destroy(xhci->segment_pool);
+	xhci->segment_pool = NULL;
+	xhci_dbg(xhci, "Freed segment pool\n");
 	xhci->page_size = 0;
 	xhci->page_shift = 0;
 }
 
 int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 {
+	dma_addr_t	dma;
+	struct device	*dev = xhci_to_hcd(xhci)->self.controller;
 	unsigned int	val, val2;
+	struct xhci_segment	*seg;
 	u32 page_size;
 	int i;
 
@@ -65,7 +262,113 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 			(unsigned int) val);
 	xhci_writel(xhci, val, &xhci->op_regs->config_reg);
 
-	xhci->ir_set = &xhci->run_regs->ir_set[0];
+	/*
+	 * Initialize the ring segment pool.  The ring must be a contiguous
+	 * structure comprised of TRBs.  The TRBs must be 16 byte aligned,
+	 * however, the command ring segment needs 64-byte aligned segments,
+	 * so we pick the greater alignment need.
+	 */
+	xhci->segment_pool = dma_pool_create("xHCI ring segments", dev,
+			SEGMENT_SIZE, 64, xhci->page_size);
+	if (!xhci->segment_pool)
+		goto fail;
+
+	/* Set up the command ring to have one segments for now. */
+	xhci->cmd_ring = xhci_ring_alloc(xhci, 1, true, flags);
+	if (!xhci->cmd_ring)
+		goto fail;
+	xhci_dbg(xhci, "Allocated command ring at 0x%x\n", (unsigned int) xhci->cmd_ring);
+	xhci_dbg(xhci, "First segment DMA is 0x%x\n", (unsigned int) xhci->cmd_ring->first_seg->dma);
+
+	/* Set the address in the Command Ring Control register */
+	val = xhci_readl(xhci, &xhci->op_regs->cmd_ring[0]);
+	val = (val & ~CMD_RING_ADDR_MASK) |
+		(xhci->cmd_ring->first_seg->dma & CMD_RING_ADDR_MASK) |
+		xhci->cmd_ring->cycle_state;
+	xhci_dbg(xhci, "// Setting command ring address high bits to 0x0\n");
+	xhci_writel(xhci, (u32) 0, &xhci->op_regs->cmd_ring[1]);
+	xhci_dbg(xhci, "// Setting command ring address low bits to 0x%x\n", val);
+	xhci_writel(xhci, val, &xhci->op_regs->cmd_ring[0]);
+	xhci_dbg_cmd_ptrs(xhci);
+
+	val = xhci_readl(xhci, &xhci->cap_regs->db_off);
+	val &= DBOFF_MASK;
+	xhci_dbg(xhci, "// Doorbell array is located at offset 0x%x"
+			" from cap regs base addr\n", val);
+	xhci->dba = (void *) xhci->cap_regs + val;
+	xhci_dbg_regs(xhci);
+	xhci_print_run_regs(xhci);
+	/* Set ir_set to interrupt register set 0 */
+	xhci->ir_set = (void *) xhci->run_regs->ir_set;
+
+	/*
+	 * Event ring setup: Allocate a normal ring, but also setup
+	 * the event ring segment table (ERST).  Section 4.9.3.
+	 */
+	xhci_dbg(xhci, "// Allocating event ring\n");
+	xhci->event_ring = xhci_ring_alloc(xhci, ERST_NUM_SEGS, false, flags);
+	if (!xhci->event_ring)
+		goto fail;
+
+	xhci->erst.entries = pci_alloc_consistent(to_pci_dev(dev),
+			sizeof(struct xhci_erst_entry)*ERST_NUM_SEGS, &dma);
+	if (!xhci->erst.entries)
+		goto fail;
+	xhci_dbg(xhci, "// Allocated event ring segment table at 0x%x\n", dma);
+
+	memset(xhci->erst.entries, 0, sizeof(struct xhci_erst_entry)*ERST_NUM_SEGS);
+	xhci->erst.num_entries = ERST_NUM_SEGS;
+	xhci->erst.erst_dma_addr = dma;
+	xhci_dbg(xhci, "Set ERST to 0; private num segs = %i, virt addr = 0x%x, dma addr = 0x%x\n",
+			xhci->erst.num_entries,
+			(unsigned int) xhci->erst.entries,
+			xhci->erst.erst_dma_addr);
+
+	/* set ring base address and size for each segment table entry */
+	for (val = 0, seg = xhci->event_ring->first_seg; val < ERST_NUM_SEGS; val++) {
+		struct xhci_erst_entry *entry = &xhci->erst.entries[val];
+		entry->seg_addr[1] = 0;
+		entry->seg_addr[0] = seg->dma;
+		entry->seg_size = TRBS_PER_SEGMENT;
+		entry->rsvd = 0;
+		seg = seg->next;
+	}
+
+	/* set ERST count with the number of entries in the segment table */
+	val = xhci_readl(xhci, &xhci->ir_set->erst_size);
+	val &= ERST_SIZE_MASK;
+	val |= ERST_NUM_SEGS;
+	xhci_dbg(xhci, "// Write ERST size = %i to ir_set 0 (some bits preserved)\n",
+			val);
+	xhci_writel(xhci, val, &xhci->ir_set->erst_size);
+
+	xhci_dbg(xhci, "// Set ERST entries to point to event ring.\n");
+	/* set the segment table base address */
+	xhci_dbg(xhci, "// Set ERST base address for ir_set 0 = 0x%x\n",
+			xhci->erst.erst_dma_addr);
+	xhci_writel(xhci, 0, &xhci->ir_set->erst_base[1]);
+	val = xhci_readl(xhci, &xhci->ir_set->erst_base[0]);
+	val &= ERST_PTR_MASK;
+	val |= (xhci->erst.erst_dma_addr & ~ERST_PTR_MASK);
+	xhci_writel(xhci, val, &xhci->ir_set->erst_base[0]);
+
+	/* Set the event ring dequeue address */
+	xhci_dbg(xhci, "// Set ERST dequeue address for ir_set 0 = 0x%x%x\n",
+			xhci->erst.entries[0].seg_addr[1], xhci->erst.entries[0].seg_addr[0]);
+	val = xhci_readl(xhci, &xhci->run_regs->ir_set[0].erst_dequeue[0]);
+	val &= ERST_PTR_MASK;
+	val |= (xhci->erst.entries[0].seg_addr[0] & ~ERST_PTR_MASK);
+	xhci_writel(xhci, val, &xhci->run_regs->ir_set[0].erst_dequeue[0]);
+	xhci_writel(xhci, xhci->erst.entries[0].seg_addr[1],
+			&xhci->run_regs->ir_set[0].erst_dequeue[1]);
+	xhci_dbg(xhci, "Wrote ERST address to ir_set 0.\n");
+	xhci_print_ir_set(xhci, xhci->ir_set, 0);
+
+	/*
+	 * XXX: Might need to set the Interrupter Moderation Register to
+	 * something other than the default (~1ms minimum between interrupts).
+	 * See section 5.5.1.2.
+	 */
 
 	return 0;
 fail:
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
index 59fae2e5ea5..ed331310f1a 100644
--- a/drivers/usb/host/xhci.h
+++ b/drivers/usb/host/xhci.h
@@ -241,6 +241,18 @@ struct xhci_op_regs {
  */
 #define	DEV_NOTE_FWAKE		ENABLE_DEV_NOTE(1)
 
+/* CRCR - Command Ring Control Register - cmd_ring bitmasks */
+/* bit 0 is the command ring cycle state */
+/* stop ring operation after completion of the currently executing command */
+#define CMD_RING_PAUSE		(1 << 1)
+/* stop ring immediately - abort the currently executing command */
+#define CMD_RING_ABORT		(1 << 2)
+/* true: command ring is running */
+#define CMD_RING_RUNNING	(1 << 3)
+/* bits 4:5 reserved and should be preserved */
+/* Command Ring pointer - bit mask for the lower 32 bits. */
+#define CMD_RING_ADDR_MASK	(0xffffffc0)
+
 /* CONFIG - Configure Register - config_reg bitmasks */
 /* bits 0:7 - maximum number of device slots enabled (NumSlotsEn) */
 #define MAX_DEVS(p)	((p) & 0xff)
@@ -391,6 +403,7 @@ struct intr_reg {
  * a work queue (or delayed service routine)?
  */
 #define ERST_EHB		(1 << 3)
+#define ERST_PTR_MASK		(0xf)
 
 /**
  * struct xhci_run_regs
@@ -407,6 +420,275 @@ struct xhci_run_regs {
 	struct intr_reg	ir_set[128];
 } __attribute__ ((packed));
 
+/**
+ * struct doorbell_array
+ *
+ * Section 5.6
+ */
+struct xhci_doorbell_array {
+	u32	doorbell[256];
+} __attribute__ ((packed));
+
+#define	DB_TARGET_MASK		0xFFFFFF00
+#define	DB_STREAM_ID_MASK	0x0000FFFF
+#define	DB_TARGET_HOST		0x0
+#define	DB_STREAM_ID_HOST	0x0
+#define	DB_MASK			(0xff << 8)
+
+
+struct xhci_transfer_event {
+	/* 64-bit buffer address, or immediate data */
+	u32	buffer[2];
+	u32	transfer_len;
+	/* This field is interpreted differently based on the type of TRB */
+	u32	flags;
+} __attribute__ ((packed));
+
+/* Completion Code - only applicable for some types of TRBs */
+#define	COMP_CODE_MASK		(0xff << 24)
+#define GET_COMP_CODE(p)	(((p) & COMP_CODE_MASK) >> 24)
+#define COMP_SUCCESS	1
+/* Data Buffer Error */
+#define COMP_DB_ERR	2
+/* Babble Detected Error */
+#define COMP_BABBLE	3
+/* USB Transaction Error */
+#define COMP_TX_ERR	4
+/* TRB Error - some TRB field is invalid */
+#define COMP_TRB_ERR	5
+/* Stall Error - USB device is stalled */
+#define COMP_STALL	6
+/* Resource Error - HC doesn't have memory for that device configuration */
+#define COMP_ENOMEM	7
+/* Bandwidth Error - not enough room in schedule for this dev config */
+#define COMP_BW_ERR	8
+/* No Slots Available Error - HC ran out of device slots */
+#define COMP_ENOSLOTS	9
+/* Invalid Stream Type Error */
+#define COMP_STREAM_ERR	10
+/* Slot Not Enabled Error - doorbell rung for disabled device slot */
+#define COMP_EBADSLT	11
+/* Endpoint Not Enabled Error */
+#define COMP_EBADEP	12
+/* Short Packet */
+#define COMP_SHORT_TX	13
+/* Ring Underrun - doorbell rung for an empty isoc OUT ep ring */
+#define COMP_UNDERRUN	14
+/* Ring Overrun - isoc IN ep ring is empty when ep is scheduled to RX */
+#define COMP_OVERRUN	15
+/* Virtual Function Event Ring Full Error */
+#define COMP_VF_FULL	16
+/* Parameter Error - Context parameter is invalid */
+#define COMP_EINVAL	17
+/* Bandwidth Overrun Error - isoc ep exceeded its allocated bandwidth */
+#define COMP_BW_OVER	18
+/* Context State Error - illegal context state transition requested */
+#define COMP_CTX_STATE	19
+/* No Ping Response Error - HC didn't get PING_RESPONSE in time to TX */
+#define COMP_PING_ERR	20
+/* Event Ring is full */
+#define COMP_ER_FULL	21
+/* Missed Service Error - HC couldn't service an isoc ep within interval */
+#define COMP_MISSED_INT	23
+/* Successfully stopped command ring */
+#define COMP_CMD_STOP	24
+/* Successfully aborted current command and stopped command ring */
+#define COMP_CMD_ABORT	25
+/* Stopped - transfer was terminated by a stop endpoint command */
+#define COMP_STOP	26
+/* Same as COMP_EP_STOPPED, but the transfered length in the event is invalid */
+#define COMP_STOP_INVAL	27
+/* Control Abort Error - Debug Capability - control pipe aborted */
+#define COMP_DBG_ABORT	28
+/* TRB type 29 and 30 reserved */
+/* Isoc Buffer Overrun - an isoc IN ep sent more data than could fit in TD */
+#define COMP_BUFF_OVER	31
+/* Event Lost Error - xHC has an "internal event overrun condition" */
+#define COMP_ISSUES	32
+/* Undefined Error - reported when other error codes don't apply */
+#define COMP_UNKNOWN	33
+/* Invalid Stream ID Error */
+#define COMP_STRID_ERR	34
+/* Secondary Bandwidth Error - may be returned by a Configure Endpoint cmd */
+/* FIXME - check for this */
+#define COMP_2ND_BW_ERR	35
+/* Split Transaction Error */
+#define	COMP_SPLIT_ERR	36
+
+struct xhci_link_trb {
+	/* 64-bit segment pointer*/
+	u32 segment_ptr[2];
+	u32 intr_target;
+	u32 control;
+} __attribute__ ((packed));
+
+/* control bitfields */
+#define LINK_TOGGLE	(0x1<<1)
+
+
+union xhci_trb {
+	struct xhci_link_trb		link;
+	struct xhci_transfer_event	trans_event;
+};
+
+/* Normal TRB fields */
+/* transfer_len bitmasks - bits 0:16 */
+#define	TRB_LEN(p)		((p) & 0x1ffff)
+/* TD size - number of bytes remaining in the TD (including this TRB):
+ * bits 17 - 21.  Shift the number of bytes by 10. */
+#define TD_REMAINDER(p)		((((p) >> 10) & 0x1f) << 17)
+/* Interrupter Target - which MSI-X vector to target the completion event at */
+#define TRB_INTR_TARGET(p)	(((p) & 0x3ff) << 22)
+#define GET_INTR_TARGET(p)	(((p) >> 22) & 0x3ff)
+
+/* Cycle bit - indicates TRB ownership by HC or HCD */
+#define TRB_CYCLE		(1<<0)
+/*
+ * Force next event data TRB to be evaluated before task switch.
+ * Used to pass OS data back after a TD completes.
+ */
+#define TRB_ENT			(1<<1)
+/* Interrupt on short packet */
+#define TRB_ISP			(1<<2)
+/* Set PCIe no snoop attribute */
+#define TRB_NO_SNOOP		(1<<3)
+/* Chain multiple TRBs into a TD */
+#define TRB_CHAIN		(1<<4)
+/* Interrupt on completion */
+#define TRB_IOC			(1<<5)
+/* The buffer pointer contains immediate data */
+#define TRB_IDT			(1<<6)
+
+
+/* Control transfer TRB specific fields */
+#define TRB_DIR_IN		(1<<16)
+
+/* TRB bit mask */
+#define	TRB_TYPE_BITMASK	(0xfc00)
+#define TRB_TYPE(p)		((p) << 10)
+/* TRB type IDs */
+/* bulk, interrupt, isoc scatter/gather, and control data stage */
+#define TRB_NORMAL		1
+/* setup stage for control transfers */
+#define TRB_SETUP		2
+/* data stage for control transfers */
+#define TRB_DATA		3
+/* status stage for control transfers */
+#define TRB_STATUS		4
+/* isoc transfers */
+#define TRB_ISOC		5
+/* TRB for linking ring segments */
+#define TRB_LINK		6
+#define TRB_EVENT_DATA		7
+/* Transfer Ring No-op (not for the command ring) */
+#define TRB_TR_NOOP		8
+/* Command TRBs */
+/* Enable Slot Command */
+#define TRB_ENABLE_SLOT		9
+/* Disable Slot Command */
+#define TRB_DISABLE_SLOT	10
+/* Address Device Command */
+#define TRB_ADDR_DEV		11
+/* Configure Endpoint Command */
+#define TRB_CONFIG_EP		12
+/* Evaluate Context Command */
+#define TRB_EVAL_CONTEXT	13
+/* Reset Transfer Ring Command */
+#define TRB_RESET_RING		14
+/* Stop Transfer Ring Command */
+#define TRB_STOP_RING		15
+/* Set Transfer Ring Dequeue Pointer Command */
+#define TRB_SET_DEQ		16
+/* Reset Device Command */
+#define TRB_RESET_DEV		17
+/* Force Event Command (opt) */
+#define TRB_FORCE_EVENT		18
+/* Negotiate Bandwidth Command (opt) */
+#define TRB_NEG_BANDWIDTH	19
+/* Set Latency Tolerance Value Command (opt) */
+#define TRB_SET_LT		20
+/* Get port bandwidth Command */
+#define TRB_GET_BW		21
+/* Force Header Command - generate a transaction or link management packet */
+#define TRB_FORCE_HEADER	22
+/* No-op Command - not for transfer rings */
+#define TRB_CMD_NOOP		23
+/* TRB IDs 24-31 reserved */
+/* Event TRBS */
+/* Transfer Event */
+#define TRB_TRANSFER		32
+/* Command Completion Event */
+#define TRB_COMPLETION		33
+/* Port Status Change Event */
+#define TRB_PORT_STATUS		34
+/* Bandwidth Request Event (opt) */
+#define TRB_BANDWIDTH_EVENT	35
+/* Doorbell Event (opt) */
+#define TRB_DOORBELL		36
+/* Host Controller Event */
+#define TRB_HC_EVENT		37
+/* Device Notification Event - device sent function wake notification */
+#define TRB_DEV_NOTE		38
+/* MFINDEX Wrap Event - microframe counter wrapped */
+#define TRB_MFINDEX_WRAP	39
+/* TRB IDs 40-47 reserved, 48-63 is vendor-defined */
+
+/*
+ * TRBS_PER_SEGMENT must be a multiple of 4,
+ * since the command ring is 64-byte aligned.
+ * It must also be greater than 16.
+ */
+#define TRBS_PER_SEGMENT	64
+#define SEGMENT_SIZE		(TRBS_PER_SEGMENT*16)
+
+struct xhci_segment {
+	union xhci_trb		*trbs;
+	/* private to HCD */
+	struct xhci_segment	*next;
+	dma_addr_t		dma;
+} __attribute__ ((packed));
+
+struct xhci_ring {
+	struct xhci_segment	*first_seg;
+	union  xhci_trb		*enqueue;
+	union  xhci_trb		*dequeue;
+	/*
+	 * Write the cycle state into the TRB cycle field to give ownership of
+	 * the TRB to the host controller (if we are the producer), or to check
+	 * if we own the TRB (if we are the consumer).  See section 4.9.1.
+	 */
+	u32			cycle_state;
+};
+
+struct xhci_erst_entry {
+	/* 64-bit event ring segment address */
+	u32	seg_addr[2];
+	u32	seg_size;
+	/* Set to zero */
+	u32	rsvd;
+} __attribute__ ((packed));
+
+struct xhci_erst {
+	struct xhci_erst_entry	*entries;
+	unsigned int		num_entries;
+	/* xhci->event_ring keeps track of segment dma addresses */
+	dma_addr_t		erst_dma_addr;
+	/* Num entries the ERST can contain */
+	unsigned int		erst_size;
+};
+
+/*
+ * Each segment table entry is 4*32bits long.  1K seems like an ok size:
+ * (1K bytes * 8bytes/bit) / (4*32 bits) = 64 segment entries in the table,
+ * meaning 64 ring segments.
+ * Initial allocated size of the ERST, in number of entries */
+#define	ERST_NUM_SEGS	1
+/* Initial allocated size of the ERST, in number of entries */
+#define	ERST_SIZE	64
+/* Initial number of event segment rings allocated */
+#define	ERST_ENTRIES	1
+/* XXX: Make these module parameters */
+
 
 /* There is one ehci_hci structure per controller */
 struct xhci_hcd {
@@ -414,6 +696,7 @@ struct xhci_hcd {
 	struct xhci_cap_regs __iomem *cap_regs;
 	struct xhci_op_regs __iomem *op_regs;
 	struct xhci_run_regs __iomem *run_regs;
+	struct xhci_doorbell_array __iomem *dba;
 	/* Our HCD's current interrupter register set */
 	struct	intr_reg __iomem *ir_set;
 
@@ -441,6 +724,14 @@ struct xhci_hcd {
 	/* only one MSI vector for now, but might need more later */
 	int		msix_count;
 	struct msix_entry	*msix_entries;
+	/* data structures */
+	struct xhci_ring	*cmd_ring;
+	struct xhci_ring	*event_ring;
+	struct xhci_erst	erst;
+
+	/* DMA pools */
+	struct dma_pool	*device_pool;
+	struct dma_pool	*segment_pool;
 };
 
 /* convert between an HCD pointer and the corresponding EHCI_HCD */
@@ -488,6 +779,11 @@ static inline void xhci_writel(const struct xhci_hcd *xhci,
 /* xHCI debugging */
 void xhci_print_ir_set(struct xhci_hcd *xhci, struct intr_reg *ir_set, int set_num);
 void xhci_print_registers(struct xhci_hcd *xhci);
+void xhci_dbg_regs(struct xhci_hcd *xhci);
+void xhci_print_run_regs(struct xhci_hcd *xhci);
+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);
 
 /* xHCI memory managment */
 void xhci_mem_cleanup(struct xhci_hcd *xhci);