USB: xhci: Always align output device contexts to 64 bytes.

Make sure the xHCI output device context is 64-byte aligned.  Previous
code was using the same structure for both the output device context and
the input control context.  Since the structure had 32 bytes of flags
before the device context, the output device context wouldn't be 64-byte
aligned.  Define a new structure to use for the output device context and
clean up the debugging for these two structures.

The copy of the device context in the input control context does *not*
need to be 64-byte aligned.

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

4 files changed, 89 insertions, 53 deletions

diff --git a/drivers/usb/host/xhci-dbg.c b/drivers/usb/host/xhci-dbg.c
index 6d62e4abe3c6..d77f8de11256 100644
--- a/drivers/usb/host/xhci-dbg.c
+++ b/drivers/usb/host/xhci-dbg.c
@@ -393,78 +393,103 @@ void xhci_dbg_cmd_ptrs(struct xhci_hcd *xhci)
                         upper_32_bits(val));
 }
 
-void xhci_dbg_ctx(struct xhci_hcd *xhci, struct xhci_device_control *ctx, dma_addr_t dma, unsigned int last_ep)
+dma_addr_t xhci_dbg_slot_ctx(struct xhci_hcd *xhci, struct xhci_slot_ctx *slot, dma_addr_t dma)
 {
-        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, "@%p (virt) @%08llx (dma) %#08x - drop flags\n",
-                        &ctx->drop_flags, (unsigned long long)dma,
-                        ctx->drop_flags);
-        dma += field_size;
-        xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - add flags\n",
-                        &ctx->add_flags, (unsigned long long)dma,
-                        ctx->add_flags);
-        dma += field_size;
-        for (i = 0; i < 6; ++i) {
-                xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - rsvd[%d]\n",
-                                &ctx->rsvd[i], (unsigned long long)dma,
-                                ctx->rsvd[i], i);
-                dma += field_size;
-        }
+        int i;
 
         xhci_dbg(xhci, "Slot Context:\n");
         xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - dev_info\n",
-                        &ctx->slot.dev_info,
-                        (unsigned long long)dma, ctx->slot.dev_info);
+                        &slot->dev_info,
+                        (unsigned long long)dma, slot->dev_info);
         dma += field_size;
         xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - dev_info2\n",
-                        &ctx->slot.dev_info2,
-                        (unsigned long long)dma, ctx->slot.dev_info2);
+                        &slot->dev_info2,
+                        (unsigned long long)dma, slot->dev_info2);
         dma += field_size;
         xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - tt_info\n",
-                        &ctx->slot.tt_info,
-                        (unsigned long long)dma, ctx->slot.tt_info);
+                        &slot->tt_info,
+                        (unsigned long long)dma, slot->tt_info);
         dma += field_size;
         xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - dev_state\n",
-                        &ctx->slot.dev_state,
-                        (unsigned long long)dma, ctx->slot.dev_state);
+                        &slot->dev_state,
+                        (unsigned long long)dma, slot->dev_state);
         dma += field_size;
         for (i = 0; i < 4; ++i) {
                 xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - rsvd[%d]\n",
-                                &ctx->slot.reserved[i], (unsigned long long)dma,
-                                ctx->slot.reserved[i], i);
+                                &slot->reserved[i], (unsigned long long)dma,
+                                slot->reserved[i], i);
                 dma += field_size;
         }
 
+        return dma;
+}
+
+dma_addr_t xhci_dbg_ep_ctx(struct xhci_hcd *xhci, struct xhci_ep_ctx *ep, 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;
+
         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, "@%p (virt) @%08llx (dma) %#08x - ep_info\n",
-                                &ctx->ep[i].ep_info,
-                                (unsigned long long)dma, ctx->ep[i].ep_info);
+                                &ep[i].ep_info,
+                                (unsigned long long)dma, ep[i].ep_info);
                 dma += field_size;
                 xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - ep_info2\n",
-                                &ctx->ep[i].ep_info2,
-                                (unsigned long long)dma, ctx->ep[i].ep_info2);
+                                &ep[i].ep_info2,
+                                (unsigned long long)dma, ep[i].ep_info2);
                 dma += field_size;
                 xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08llx - deq\n",
-                                &ctx->ep[i].deq,
-                                (unsigned long long)dma, ctx->ep[i].deq);
+                                &ep[i].deq,
+                                (unsigned long long)dma, ep[i].deq);
                 dma += 2*field_size;
                 xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - tx_info\n",
-                                &ctx->ep[i].tx_info,
-                                (unsigned long long)dma, ctx->ep[i].tx_info);
+                                &ep[i].tx_info,
+                                (unsigned long long)dma, ep[i].tx_info);
                 dma += field_size;
                 for (j = 0; j < 3; ++j) {
                         xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - rsvd[%d]\n",
-                                        &ctx->ep[i].reserved[j],
+                                        &ep[i].reserved[j],
                                         (unsigned long long)dma,
-                                        ctx->ep[i].reserved[j], j);
+                                        ep[i].reserved[j], j);
                         dma += field_size;
                 }
         }
+        return dma;
+}
+
+void xhci_dbg_ctx(struct xhci_hcd *xhci, struct xhci_device_control *ctx, dma_addr_t dma, unsigned int last_ep)
+{
+        int i;
+        /* Fields are 32 bits wide, DMA addresses are in bytes */
+        int field_size = 32 / 8;
+
+        xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - drop flags\n",
+                        &ctx->drop_flags, (unsigned long long)dma,
+                        ctx->drop_flags);
+        dma += field_size;
+        xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - add flags\n",
+                        &ctx->add_flags, (unsigned long long)dma,
+                        ctx->add_flags);
+        dma += field_size;
+        for (i = 0; i < 6; ++i) {
+                xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - rsvd[%d]\n",
+                                &ctx->rsvd[i], (unsigned long long)dma,
+                                ctx->rsvd[i], i);
+                dma += field_size;
+        }
+        dma = xhci_dbg_slot_ctx(xhci, &ctx->slot, dma);
+        dma = xhci_dbg_ep_ctx(xhci, ctx->ep, dma, last_ep);
+}
+
+void xhci_dbg_device_ctx(struct xhci_hcd *xhci, struct xhci_device_ctx *ctx, dma_addr_t dma, unsigned int last_ep)
+{
+        dma = xhci_dbg_slot_ctx(xhci, &ctx->slot, dma);
+        dma = xhci_dbg_ep_ctx(xhci, ctx->ep, dma, last_ep);
 }
diff --git a/drivers/usb/host/xhci-hcd.c b/drivers/usb/host/xhci-hcd.c
index 008326d5bc52..921dd173d793 100644
--- a/drivers/usb/host/xhci-hcd.c
+++ b/drivers/usb/host/xhci-hcd.c
@@ -1013,7 +1013,7 @@ int xhci_check_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
         }
 
         xhci_dbg(xhci, "Output context after successful config ep cmd:\n");
-        xhci_dbg_ctx(xhci, virt_dev->out_ctx, virt_dev->out_ctx_dma,
+        xhci_dbg_device_ctx(xhci, virt_dev->out_ctx, virt_dev->out_ctx_dma,
                         LAST_CTX_TO_EP_NUM(virt_dev->in_ctx->slot.dev_info));
 
         xhci_zero_in_ctx(virt_dev);
@@ -1265,7 +1265,7 @@ int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev)
         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);
+        xhci_dbg_device_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.
@@ -1274,9 +1274,6 @@ int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev)
         /* Zero the input context control for later use */
         virt_dev->in_ctx->add_flags = 0;
         virt_dev->in_ctx->drop_flags = 0;
-        /* Mirror flags in the output context for future ep enable/disable */
-        virt_dev->out_ctx->add_flags = SLOT_FLAG | EP0_FLAG;
-        virt_dev->out_ctx->drop_flags = 0;
 
         xhci_dbg(xhci, "Device address = %d\n", udev->devnum);
         /* XXX Meh, not sure if anyone else but choose_address uses this. */
diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
index 71121d99235d..8d6bdf2f8015 100644
--- a/drivers/usb/host/xhci-mem.c
+++ b/drivers/usb/host/xhci-mem.c
@@ -235,7 +235,10 @@ int xhci_alloc_virt_device(struct xhci_hcd *xhci, int slot_id,
                 return 0;
         dev = xhci->devs[slot_id];
 
-        /* Allocate the (output) device context that will be used in the HC */
+        /* Allocate the (output) device context that will be used in the HC.
+         * The structure is 32 bytes smaller than the input context, but that's
+         * fine.
+         */
         dev->out_ctx = dma_pool_alloc(xhci->device_pool, flags, &dma);
         if (!dev->out_ctx)
                 goto fail;
@@ -260,16 +263,12 @@ int xhci_alloc_virt_device(struct xhci_hcd *xhci, int slot_id,
 
         init_completion(&dev->cmd_completion);
 
-        /*
-         * 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[slot_id] =
-                (u32) dev->out_ctx_dma + (32);
+        /* Point to output device context in dcbaa. */
+        xhci->dcbaa->dev_context_ptrs[slot_id] = dev->out_ctx_dma;
         xhci_dbg(xhci, "Set slot id %d dcbaa entry %p to 0x%llx\n",
                         slot_id,
                         &xhci->dcbaa->dev_context_ptrs[slot_id],
-                        (unsigned long long)dev->out_ctx_dma);
+                        (unsigned long long) xhci->dcbaa->dev_context_ptrs[slot_id]);
 
         return 1;
 fail:
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
index 5a09b9a26e0d..d4d3c7777fb8 100644
--- a/drivers/usb/host/xhci.h
+++ b/drivers/usb/host/xhci.h
@@ -584,15 +584,29 @@ struct xhci_ep_ctx {
 
 /**
  * struct xhci_device_control
- * Input/Output context; see section 6.2.5.
+ * Input context; see section 6.2.5.
  *
  * @drop_context:       set the bit of the endpoint context you want to disable
  * @add_context:        set the bit of the endpoint context you want to enable
  */
 struct xhci_device_control {
+        /* Input control context */
         u32     drop_flags;
         u32     add_flags;
         u32     rsvd[6];
+        /* Copy of device context */
+        struct xhci_slot_ctx    slot;
+        struct xhci_ep_ctx      ep[31];
+};
+
+/**
+ * struct xhci_device_ctx
+ * Device context; see section 6.2.1.
+ *
+ * @slot:               slot context for the device.
+ * @ep:                 array of endpoint contexts for the device.
+ */
+struct xhci_device_ctx {
         struct xhci_slot_ctx    slot;
         struct xhci_ep_ctx      ep[31];
 };
@@ -612,7 +626,7 @@ struct xhci_virt_device {
          * 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;
+        struct xhci_device_ctx          *out_ctx;
         dma_addr_t                      out_ctx_dma;
         /* Used for addressing devices and configuration changes */
         struct xhci_device_control      *in_ctx;
@@ -1126,6 +1140,7 @@ 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);
+void xhci_dbg_device_ctx(struct xhci_hcd *xhci, struct xhci_device_ctx *ctx, dma_addr_t dma, unsigned int last_ep);
 
 /* xHCI memory managment */
 void xhci_mem_cleanup(struct xhci_hcd *xhci);