add support for hardware contexts on recent intel hardware

based on the code in linux 3.8.13
ok kettenis@

7 files changed, 628 insertions, 10 deletions

diff --git a/sys/dev/pci/drm/files.drm b/sys/dev/pci/drm/files.drm
index e8852a8fbd9..dc4783b9d16 100644
--- a/sys/dev/pci/drm/files.drm
+++ b/sys/dev/pci/drm/files.drm
@@ -1,5 +1,5 @@
 # $NetBSD: files.drm,v 1.2 2007/03/28 11:29:37 jmcneill Exp $
-# $OpenBSD: files.drm,v 1.25 2013/03/31 13:30:25 kettenis Exp $
+# $OpenBSD: files.drm,v 1.26 2013/08/07 00:04:27 jsg Exp $
 
 # direct rendering modules
 define drmbase {}
@@ -26,6 +26,7 @@ attach	inteldrm at drmdev
 file	dev/pci/drm/i915/i915_dma.c		inteldrm
 file	dev/pci/drm/i915/i915_drv.c		inteldrm
 file	dev/pci/drm/i915/i915_gem.c		inteldrm
+file	dev/pci/drm/i915/i915_gem_context.c	inteldrm
 file	dev/pci/drm/i915/i915_gem_evict.c	inteldrm
 file	dev/pci/drm/i915/i915_gem_execbuffer.c	inteldrm
 file	dev/pci/drm/i915/i915_gem_gtt.c		inteldrm
diff --git a/sys/dev/pci/drm/i915/i915_dma.c b/sys/dev/pci/drm/i915/i915_dma.c
index d596c0764e8..8b13310db12 100644
--- a/sys/dev/pci/drm/i915/i915_dma.c
+++ b/sys/dev/pci/drm/i915/i915_dma.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: i915_dma.c,v 1.8 2013/07/05 07:20:27 jsg Exp $	*/
+/*	$OpenBSD: i915_dma.c,v 1.9 2013/08/07 00:04:27 jsg Exp $	*/
 /* i915_dma.c -- DMA support for the I915 -*- linux-c -*-
  */
 /*
@@ -400,6 +400,8 @@ i915_driver_open(struct drm_device *dev, struct drm_file *file)
 	mtx_init(&file_priv->mm.lock, IPL_NONE);
 	INIT_LIST_HEAD(&file_priv->mm.request_list);
 
+	SPLAY_INIT(&file_priv->ctx_tree);
+
 	return 0;
 }
 
@@ -408,6 +410,7 @@ i915_driver_close(struct drm_device *dev, struct drm_file *file)
 {
 	struct drm_i915_file_private *file_priv = file->driver_priv;
 
+	i915_gem_context_close(dev, file);
 	i915_gem_release(dev, file);
 	free(file_priv, M_DRM);
 }
diff --git a/sys/dev/pci/drm/i915/i915_drv.c b/sys/dev/pci/drm/i915/i915_drv.c
index f3d23578ee9..a77f55403fc 100644
--- a/sys/dev/pci/drm/i915/i915_drv.c
+++ b/sys/dev/pci/drm/i915/i915_drv.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: i915_drv.c,v 1.35 2013/07/05 07:20:27 jsg Exp $ */
+/* $OpenBSD: i915_drv.c,v 1.36 2013/08/07 00:04:28 jsg Exp $ */
 /*
  * Copyright (c) 2008-2009 Owain G. Ainsworth <oga@openbsd.org>
  *
@@ -1390,6 +1390,12 @@ inteldrm_doioctl(struct drm_device *dev, u_long cmd, caddr_t data,
 		case DRM_IOCTL_I915_GEM_GET_CACHING:
 			return (i915_gem_get_caching_ioctl(dev, data,
 			    file_priv));
+		case DRM_IOCTL_I915_GEM_CONTEXT_CREATE:
+			return (i915_gem_context_create_ioctl(dev, data,
+			    file_priv));
+		case DRM_IOCTL_I915_GEM_CONTEXT_DESTROY:
+			return (i915_gem_context_destroy_ioctl(dev, data,
+			    file_priv));
 		default:
 			break;
 		}
@@ -2042,8 +2048,8 @@ i915_reset(struct drm_device *dev)
 		for_each_ring(ring, dev_priv, i)
 			ring->init(ring);
 
-#ifdef notyet
 		i915_gem_context_init(dev);
+#ifdef notyet
 		i915_gem_init_ppgtt(dev);
 #endif
 
diff --git a/sys/dev/pci/drm/i915/i915_drv.h b/sys/dev/pci/drm/i915/i915_drv.h
index a47ae71d308..7012b11cb1f 100644
--- a/sys/dev/pci/drm/i915/i915_drv.h
+++ b/sys/dev/pci/drm/i915/i915_drv.h
@@ -1,4 +1,4 @@
-/* $OpenBSD: i915_drv.h,v 1.24 2013/07/05 07:20:27 jsg Exp $ */
+/* $OpenBSD: i915_drv.h,v 1.25 2013/08/07 00:04:28 jsg Exp $ */
 /* i915_drv.h -- Private header for the I915 driver -*- linux-c -*-
  */
 /*
@@ -203,6 +203,22 @@ struct intel_device_info {
 	u8 has_llc:1;
 };
 
+/* This must match up with the value previously used for execbuf2.rsvd1. */
+#define DEFAULT_CONTEXT_ID 0
+struct i915_hw_context {
+	int id;
+	bool is_initialized;
+	struct drm_i915_file_private *file_priv;
+	struct intel_ring_buffer *ring;
+	struct drm_i915_gem_object *obj;
+};
+
+struct i915_ctx_handle {
+	SPLAY_ENTRY(i915_ctx_handle)	 entry;
+	struct i915_hw_context		*ctx;
+	uint32_t			 handle;
+};
+
 enum no_fbc_reason {
 	FBC_NO_OUTPUT, /* no outputs enabled to compress */
 	FBC_STOLEN_TOO_SMALL, /* not enough space to hold compressed buffers */
@@ -798,6 +814,9 @@ struct inteldrm_softc {
 	struct drm_property *broadcast_rgb_property;
 	struct drm_property *force_audio_property;
 
+	bool hw_contexts_disabled;
+	uint32_t hw_context_size;
+
 	u32 fdi_rx_config;
 
 	struct i915_suspend_saved_registers regfile;
@@ -973,6 +992,8 @@ struct drm_i915_file_private {
 		struct mutex lock;
 		struct list_head request_list;
 	} mm;
+	SPLAY_HEAD(i915_ctx_tree, i915_ctx_handle) ctx_tree;
+	uint32_t ctx_id;
 };
 
 /**
@@ -1180,6 +1201,17 @@ int	i915_reset(struct drm_device *);
 void	inteldrm_timeout(void *);
 bool	i915_semaphore_is_enabled(struct drm_device *);
 
+/* i915_gem_context.c */
+void i915_gem_context_init(struct drm_device *dev);
+void i915_gem_context_fini(struct drm_device *dev);
+void i915_gem_context_close(struct drm_device *dev, struct drm_file *file);
+int i915_switch_context(struct intel_ring_buffer *ring,
+			struct drm_file *file, int to_id);
+int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *file);
+int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
+				   struct drm_file *file);
+
 /* i915_gem_evict.c */
 int i915_gem_evict_everything(struct drm_device *);
 int i915_gem_evict_something(struct inteldrm_softc *, size_t);
diff --git a/sys/dev/pci/drm/i915/i915_gem.c b/sys/dev/pci/drm/i915/i915_gem.c
index 1302b764036..b5fccc52d4f 100644
--- a/sys/dev/pci/drm/i915/i915_gem.c
+++ b/sys/dev/pci/drm/i915/i915_gem.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: i915_gem.c,v 1.29 2013/07/29 18:32:15 kettenis Exp $	*/
+/*	$OpenBSD: i915_gem.c,v 1.30 2013/08/07 00:04:28 jsg Exp $	*/
 /*
  * Copyright (c) 2008-2009 Owain G. Ainsworth <oga@openbsd.org>
  *
@@ -1801,11 +1801,9 @@ i915_gpu_idle(struct drm_device *dev)
 
 	/* Flush everything onto the inactive list. */
 	for_each_ring(ring, dev_priv, i) {
-#ifdef notyet
 		ret = i915_switch_context(ring, NULL, DEFAULT_CONTEXT_ID);
 		if (ret)
 			return ret;
-#endif
 
 		ret = intel_ring_idle(ring);
 		if (ret)
@@ -3173,8 +3171,8 @@ i915_gem_init_hw(struct drm_device *dev)
 	 * XXX: There was some w/a described somewhere suggesting loading
 	 * contexts before PPGTT.
 	 */
-#ifdef notyet
 	i915_gem_context_init(dev);
+#ifdef notyet
 	i915_gem_init_ppgtt(dev);
 #endif
 
diff --git a/sys/dev/pci/drm/i915/i915_gem_context.c b/sys/dev/pci/drm/i915/i915_gem_context.c
new file mode 100644
index 00000000000..f4de9be0932
--- /dev/null
+++ b/sys/dev/pci/drm/i915/i915_gem_context.c
@@ -0,0 +1,573 @@
+/*	$OpenBSD: i915_gem_context.c,v 1.1 2013/08/07 00:04:28 jsg Exp $	*/
+/*
+ * Copyright © 2011-2012 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Ben Widawsky <ben@bwidawsk.net>
+ *
+ */
+
+/*
+ * This file implements HW context support. On gen5+ a HW context consists of an
+ * opaque GPU object which is referenced at times of context saves and restores.
+ * With RC6 enabled, the context is also referenced as the GPU enters and exists
+ * from RC6 (GPU has it's own internal power context, except on gen5). Though
+ * something like a context does exist for the media ring, the code only
+ * supports contexts for the render ring.
+ *
+ * In software, there is a distinction between contexts created by the user,
+ * and the default HW context. The default HW context is used by GPU clients
+ * that do not request setup of their own hardware context. The default
+ * context's state is never restored to help prevent programming errors. This
+ * would happen if a client ran and piggy-backed off another clients GPU state.
+ * The default context only exists to give the GPU some offset to load as the
+ * current to invoke a save of the context we actually care about. In fact, the
+ * code could likely be constructed, albeit in a more complicated fashion, to
+ * never use the default context, though that limits the driver's ability to
+ * swap out, and/or destroy other contexts.
+ *
+ * All other contexts are created as a request by the GPU client. These contexts
+ * store GPU state, and thus allow GPU clients to not re-emit state (and
+ * potentially query certain state) at any time. The kernel driver makes
+ * certain that the appropriate commands are inserted.
+ *
+ * The context life cycle is semi-complicated in that context BOs may live
+ * longer than the context itself because of the way the hardware, and object
+ * tracking works. Below is a very crude representation of the state machine
+ * describing the context life.
+ *                                         refcount     pincount     active
+ * S0: initial state                          0            0           0
+ * S1: context created                        1            0           0
+ * S2: context is currently running           2            1           X
+ * S3: GPU referenced, but not current        2            0           1
+ * S4: context is current, but destroyed      1            1           0
+ * S5: like S3, but destroyed                 1            0           1
+ *
+ * The most common (but not all) transitions:
+ * S0->S1: client creates a context
+ * S1->S2: client submits execbuf with context
+ * S2->S3: other clients submits execbuf with context
+ * S3->S1: context object was retired
+ * S3->S2: clients submits another execbuf
+ * S2->S4: context destroy called with current context
+ * S3->S5->S0: destroy path
+ * S4->S5->S0: destroy path on current context
+ *
+ * There are two confusing terms used above:
+ *  The "current context" means the context which is currently running on the
+ *  GPU. The GPU has loaded it's state already and has stored away the gtt
+ *  offset of the BO. The GPU is not actively referencing the data at this
+ *  offset, but it will on the next context switch. The only way to avoid this
+ *  is to do a GPU reset.
+ *
+ *  An "active context' is one which was previously the "current context" and is
+ *  on the active list waiting for the next context switch to occur. Until this
+ *  happens, the object must remain at the same gtt offset. It is therefore
+ *  possible to destroy a context, but it is still active.
+ *
+ */
+
+#include <dev/pci/drm/drmP.h>
+#include <dev/pci/drm/i915_drm.h>
+#include "i915_drv.h"
+
+/* This is a HW constraint. The value below is the largest known requirement
+ * I've seen in a spec to date, and that was a workaround for a non-shipping
+ * part. It should be safe to decrease this, but it's more future proof as is.
+ */
+#define CONTEXT_ALIGN (64<<10)
+
+static struct i915_hw_context *
+i915_gem_context_get(struct drm_i915_file_private *file_priv, u32 id);
+static int do_switch(struct i915_hw_context *to);
+
+int i915_ctx_handle_cmp(struct i915_ctx_handle *, struct i915_ctx_handle *);
+SPLAY_PROTOTYPE(i915_ctx_tree, i915_ctx_handle, entry, i915_ctx_handle_cmp);
+
+static int get_context_size(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	int ret;
+	u32 reg;
+
+	switch (INTEL_INFO(dev)->gen) {
+	case 6:
+		reg = I915_READ(CXT_SIZE);
+		ret = GEN6_CXT_TOTAL_SIZE(reg) * 64;
+		break;
+	case 7:
+		reg = I915_READ(GEN7_CXT_SIZE);
+		if (IS_HASWELL(dev))
+			ret = HSW_CXT_TOTAL_SIZE(reg) * 64;
+		else
+			ret = GEN7_CXT_TOTAL_SIZE(reg) * 64;
+		break;
+	default:
+		BUG();
+	}
+
+	return ret;
+}
+
+static void do_destroy(struct i915_hw_context *ctx)
+{
+	struct drm_device *dev = ctx->obj->base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	if (ctx->file_priv) {
+		struct i915_ctx_handle *han, find;
+		find.handle = ctx->id;
+		han = SPLAY_FIND(i915_ctx_tree, &ctx->file_priv->ctx_tree, &find);
+		if (han != NULL)
+			SPLAY_REMOVE(i915_ctx_tree, &ctx->file_priv->ctx_tree, han);
+	} else
+		BUG_ON(ctx != dev_priv->ring[RCS].default_context);
+
+	drm_gem_object_unreference(&ctx->obj->base);
+	free(ctx, M_DRM);
+}
+
+static struct i915_hw_context *
+create_hw_context(struct drm_device *dev,
+		  struct drm_i915_file_private *file_priv)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct i915_hw_context *ctx;
+	struct i915_ctx_handle *han;
+	int ret;
+
+	ctx = malloc(sizeof(*ctx), M_DRM, M_WAITOK | M_ZERO);
+	if (ctx == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	ctx->obj = i915_gem_alloc_object(dev, dev_priv->hw_context_size);
+	if (ctx->obj == NULL) {
+		free(ctx, M_DRM);
+		DRM_DEBUG_DRIVER("Context object allocated failed\n");
+		return ERR_PTR(-ENOMEM);
+	}
+
+	if (INTEL_INFO(dev)->gen >= 7) {
+		ret = i915_gem_object_set_cache_level(ctx->obj,
+						      I915_CACHE_LLC_MLC);
+		if (ret)
+			goto err_out;
+	}
+
+	/* The ring associated with the context object is handled by the normal
+	 * object tracking code. We give an initial ring value simple to pass an
+	 * assertion in the context switch code.
+	 */
+	ctx->ring = &dev_priv->ring[RCS];
+
+	/* Default context will never have a file_priv */
+	if (file_priv == NULL)
+		return ctx;
+
+	ctx->file_priv = file_priv;
+
+	if ((han = drm_calloc(1, sizeof(*han))) == NULL) {
+		ret = -ENOMEM;
+		DRM_DEBUG_DRIVER("idr allocation failed\n");
+		goto err_out;
+	}
+	han->ctx = ctx;
+
+again:
+	han->handle = ++file_priv->ctx_id;
+
+	if (han->handle <= DEFAULT_CONTEXT_ID + 1 || SPLAY_INSERT(i915_ctx_tree,
+	    &file_priv->ctx_tree, han))
+		goto again;
+
+	ctx->id = han->handle;
+
+	return ctx;
+
+err_out:
+	do_destroy(ctx);
+	return ERR_PTR(ret);
+}
+
+static inline bool is_default_context(struct i915_hw_context *ctx)
+{
+	return (ctx == ctx->ring->default_context);
+}
+
+/**
+ * The default context needs to exist per ring that uses contexts. It stores the
+ * context state of the GPU for applications that don't utilize HW contexts, as
+ * well as an idle case.
+ */
+static int create_default_context(struct drm_i915_private *dev_priv)
+{
+	struct i915_hw_context *ctx;
+	struct drm_device *dev = (struct drm_device *)dev_priv->drmdev;
+	int ret;
+
+	rw_assert_wrlock(&dev->dev_lock);
+
+	ctx = create_hw_context(dev, NULL);
+	if (IS_ERR(ctx))
+		return (long)(ctx);
+
+	/* We may need to do things with the shrinker which require us to
+	 * immediately switch back to the default context. This can cause a
+	 * problem as pinning the default context also requires GTT space which
+	 * may not be available. To avoid this we always pin the
+	 * default context.
+	 */
+	dev_priv->ring[RCS].default_context = ctx;
+	ret = i915_gem_object_pin(ctx->obj, CONTEXT_ALIGN, false, false);
+	if (ret)
+		goto err_destroy;
+
+	ret = do_switch(ctx);
+	if (ret)
+		goto err_unpin;
+
+	DRM_DEBUG_DRIVER("Default HW context loaded\n");
+	return 0;
+
+err_unpin:
+	i915_gem_object_unpin(ctx->obj);
+err_destroy:
+	do_destroy(ctx);
+	return ret;
+}
+
+void i915_gem_context_init(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	uint32_t ctx_size;
+
+	if (!HAS_HW_CONTEXTS(dev)) {
+		dev_priv->hw_contexts_disabled = true;
+		return;
+	}
+
+	/* If called from reset, or thaw... we've been here already */
+	if (dev_priv->hw_contexts_disabled ||
+	    dev_priv->ring[RCS].default_context)
+		return;
+
+	ctx_size = get_context_size(dev);
+	dev_priv->hw_context_size = get_context_size(dev);
+	dev_priv->hw_context_size = roundup2(dev_priv->hw_context_size, 4096);
+
+	if (ctx_size <= 0 || ctx_size > (1<<20)) {
+		dev_priv->hw_contexts_disabled = true;
+		return;
+	}
+
+	if (create_default_context(dev_priv)) {
+		dev_priv->hw_contexts_disabled = true;
+		return;
+	}
+
+	DRM_DEBUG_DRIVER("HW context support initialized\n");
+}
+
+void i915_gem_context_fini(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->hw_contexts_disabled)
+		return;
+
+	/* The only known way to stop the gpu from accessing the hw context is
+	 * to reset it. Do this as the very last operation to avoid confusing
+	 * other code, leading to spurious errors. */
+	intel_gpu_reset(dev);
+
+	i915_gem_object_unpin(dev_priv->ring[RCS].default_context->obj);
+
+	do_destroy(dev_priv->ring[RCS].default_context);
+}
+
+static int context_idr_cleanup(int id, void *p, void *data)
+{
+	struct i915_hw_context *ctx = p;
+
+	BUG_ON(id == DEFAULT_CONTEXT_ID);
+
+	do_destroy(ctx);
+
+	return 0;
+}
+
+void i915_gem_context_close(struct drm_device *dev, struct drm_file *file)
+{
+	struct drm_i915_file_private *file_priv = file->driver_priv;
+	struct i915_ctx_handle *han, *nxt;
+
+	DRM_LOCK();
+	for (han = SPLAY_MIN(i915_ctx_tree, &file_priv->ctx_tree); han != NULL;
+	    han = nxt) {
+		nxt = SPLAY_NEXT(i915_ctx_tree, &file_priv->ctx_tree, han);
+		context_idr_cleanup(han->handle, han->ctx, NULL);
+		SPLAY_REMOVE(i915_ctx_tree, &file_priv->ctx_tree, han);
+		drm_free(han);
+	}
+	DRM_UNLOCK();
+}
+
+static struct i915_hw_context *
+i915_gem_context_get(struct drm_i915_file_private *file_priv, u32 id)
+{
+	struct i915_ctx_handle *han, search;
+	search.handle = id;
+	han = SPLAY_FIND(i915_ctx_tree, &file_priv->ctx_tree, &search);
+	if (han == NULL)
+		return NULL;
+
+	return (han->ctx);
+}
+
+static inline int
+mi_set_context(struct intel_ring_buffer *ring,
+	       struct i915_hw_context *new_context,
+	       u32 hw_flags)
+{
+	int ret;
+
+	/* w/a: If Flush TLB Invalidation Mode is enabled, driver must do a TLB
+	 * invalidation prior to MI_SET_CONTEXT. On GEN6 we don't set the value
+	 * explicitly, so we rely on the value at ring init, stored in
+	 * itlb_before_ctx_switch.
+	 */
+	if (IS_GEN6(ring->dev) && ring->itlb_before_ctx_switch) {
+		ret = ring->flush(ring, I915_GEM_GPU_DOMAINS, 0);
+		if (ret)
+			return ret;
+	}
+
+	ret = intel_ring_begin(ring, 6);
+	if (ret)
+		return ret;
+
+	if (IS_GEN7(ring->dev))
+		intel_ring_emit(ring, MI_ARB_ON_OFF | MI_ARB_DISABLE);
+	else
+		intel_ring_emit(ring, MI_NOOP);
+
+	intel_ring_emit(ring, MI_NOOP);
+	intel_ring_emit(ring, MI_SET_CONTEXT);
+	intel_ring_emit(ring, new_context->obj->gtt_offset |
+			MI_MM_SPACE_GTT |
+			MI_SAVE_EXT_STATE_EN |
+			MI_RESTORE_EXT_STATE_EN |
+			hw_flags);
+	/* w/a: MI_SET_CONTEXT must always be followed by MI_NOOP */
+	intel_ring_emit(ring, MI_NOOP);
+
+	if (IS_GEN7(ring->dev))
+		intel_ring_emit(ring, MI_ARB_ON_OFF | MI_ARB_ENABLE);
+	else
+		intel_ring_emit(ring, MI_NOOP);
+
+	intel_ring_advance(ring);
+
+	return ret;
+}
+
+static int do_switch(struct i915_hw_context *to)
+{
+	struct intel_ring_buffer *ring = to->ring;
+	struct drm_i915_gem_object *from_obj = ring->last_context_obj;
+	u32 hw_flags = 0;
+	int ret;
+
+	BUG_ON(from_obj != NULL && from_obj->pin_count == 0);
+
+	if (from_obj == to->obj)
+		return 0;
+
+	ret = i915_gem_object_pin(to->obj, CONTEXT_ALIGN, false, false);
+	if (ret)
+		return ret;
+
+	/* Clear this page out of any CPU caches for coherent swap-in/out. Note
+	 * that thanks to write = false in this call and us not setting any gpu
+	 * write domains when putting a context object onto the active list
+	 * (when switching away from it), this won't block.
+	 * XXX: We need a real interface to do this instead of trickery. */
+	ret = i915_gem_object_set_to_gtt_domain(to->obj, false);
+	if (ret) {
+		i915_gem_object_unpin(to->obj);
+		return ret;
+	}
+
+#ifdef notyet
+	if (!to->obj->has_global_gtt_mapping)
+		i915_gem_gtt_bind_object(to->obj, to->obj->cache_level);
+#else
+	i915_gem_gtt_rebind_object(to->obj, to->obj->cache_level);
+#endif
+
+	if (!to->is_initialized || is_default_context(to))
+		hw_flags |= MI_RESTORE_INHIBIT;
+	else if (/*WARN_ON_ONCE*/(from_obj == to->obj)) /* not yet expected */
+		hw_flags |= MI_FORCE_RESTORE;
+
+	ret = mi_set_context(ring, to, hw_flags);
+	if (ret) {
+		i915_gem_object_unpin(to->obj);
+		return ret;
+	}
+
+	/* The backing object for the context is done after switching to the
+	 * *next* context. Therefore we cannot retire the previous context until
+	 * the next context has already started running. In fact, the below code
+	 * is a bit suboptimal because the retiring can occur simply after the
+	 * MI_SET_CONTEXT instead of when the next seqno has completed.
+	 */
+	if (from_obj != NULL) {
+		from_obj->base.read_domains = I915_GEM_DOMAIN_INSTRUCTION;
+		i915_gem_object_move_to_active(from_obj, ring);
+		/* As long as MI_SET_CONTEXT is serializing, ie. it flushes the
+		 * whole damn pipeline, we don't need to explicitly mark the
+		 * object dirty. The only exception is that the context must be
+		 * correct in case the object gets swapped out. Ideally we'd be
+		 * able to defer doing this until we know the object would be
+		 * swapped, but there is no way to do that yet.
+		 */
+		from_obj->dirty = 1;
+		BUG_ON(from_obj->ring != ring);
+		i915_gem_object_unpin(from_obj);
+
+		drm_gem_object_unreference(&from_obj->base);
+	}
+
+	drm_gem_object_reference(&to->obj->base);
+	ring->last_context_obj = to->obj;
+	to->is_initialized = true;
+
+	return 0;
+}
+
+/**
+ * i915_switch_context() - perform a GPU context switch.
+ * @ring: ring for which we'll execute the context switch
+ * @file_priv: file_priv associated with the context, may be NULL
+ * @id: context id number
+ * @seqno: sequence number by which the new context will be switched to
+ * @flags:
+ *
+ * The context life cycle is simple. The context refcount is incremented and
+ * decremented by 1 and create and destroy. If the context is in use by the GPU,
+ * it will have a refoucnt > 1. This allows us to destroy the context abstract
+ * object while letting the normal object tracking destroy the backing BO.
+ */
+int i915_switch_context(struct intel_ring_buffer *ring,
+			struct drm_file *file,
+			int to_id)
+{
+	struct drm_i915_private *dev_priv = ring->dev->dev_private;
+	struct i915_hw_context *to;
+
+	if (dev_priv->hw_contexts_disabled)
+		return 0;
+
+	if (ring != &dev_priv->ring[RCS])
+		return 0;
+
+	if (to_id == DEFAULT_CONTEXT_ID) {
+		to = ring->default_context;
+	} else {
+		if (file == NULL)
+			return -EINVAL;
+
+		to = i915_gem_context_get(file->driver_priv, to_id);
+		if (to == NULL)
+			return -ENOENT;
+	}
+
+	return do_switch(to);
+}
+
+int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *file)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_i915_gem_context_create *args = data;
+	struct drm_i915_file_private *file_priv = file->driver_priv;
+	struct i915_hw_context *ctx;
+	int ret;
+
+	if (!drm_core_check_feature(dev, DRIVER_GEM))
+		return -ENODEV;
+
+	if (dev_priv->hw_contexts_disabled)
+		return -ENODEV;
+
+	ret = i915_mutex_lock_interruptible(dev);
+	if (ret)
+		return ret;
+
+	ctx = create_hw_context(dev, file_priv);
+	DRM_UNLOCK();
+	if (IS_ERR(ctx))
+		return (long)(ctx);
+
+	args->ctx_id = ctx->id;
+	DRM_DEBUG_DRIVER("HW context %d created\n", args->ctx_id);
+
+	return 0;
+}
+
+int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
+				   struct drm_file *file)
+{
+	struct drm_i915_gem_context_destroy *args = data;
+	struct drm_i915_file_private *file_priv = file->driver_priv;
+	struct i915_hw_context *ctx;
+	int ret;
+
+	if (!drm_core_check_feature(dev, DRIVER_GEM))
+		return -ENODEV;
+
+	ret = i915_mutex_lock_interruptible(dev);
+	if (ret)
+		return ret;
+
+	ctx = i915_gem_context_get(file_priv, args->ctx_id);
+	if (!ctx) {
+		DRM_UNLOCK();
+		return -ENOENT;
+	}
+
+	do_destroy(ctx);
+
+	DRM_UNLOCK();
+
+	DRM_DEBUG_DRIVER("HW context %d destroyed\n", args->ctx_id);
+	return 0;
+}
+
+int
+i915_ctx_handle_cmp(struct i915_ctx_handle *a, struct i915_ctx_handle *b)
+{
+        return a->handle < b->handle ? -1 : a->handle > b->handle;
+}
+
+SPLAY_GENERATE(i915_ctx_tree, i915_ctx_handle, entry, i915_ctx_handle_cmp);
diff --git a/sys/dev/pci/drm/i915/i915_gem_execbuffer.c b/sys/dev/pci/drm/i915/i915_gem_execbuffer.c
index a96d3cb9ec4..cfc81528992 100644
--- a/sys/dev/pci/drm/i915/i915_gem_execbuffer.c
+++ b/sys/dev/pci/drm/i915/i915_gem_execbuffer.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: i915_gem_execbuffer.c,v 1.7 2013/05/05 13:55:36 kettenis Exp $	*/
+/*	$OpenBSD: i915_gem_execbuffer.c,v 1.8 2013/08/07 00:04:28 jsg Exp $	*/
 /*
  * Copyright (c) 2008-2009 Owain G. Ainsworth <oga@openbsd.org>
  *
@@ -425,6 +425,7 @@ i915_gem_execbuffer2(struct drm_device *dev, void *data,
 	struct drm_obj				**object_list = NULL;
 	struct drm_obj				*batch_obj, *obj;
 	struct intel_ring_buffer		*ring;
+	uint32_t				 ctx_id = i915_execbuffer2_get_context_id(*args);
 	size_t					 oflow;
 	int					 ret, ret2, i;
 	int					 pinned = 0, pin_tries;
@@ -630,6 +631,10 @@ i915_gem_execbuffer2(struct drm_device *dev, void *data,
 	if (ret)
 		goto err;
 
+	ret = i915_switch_context(ring, file_priv, ctx_id);
+	if (ret)
+		goto err;
+
 	if (ring == &dev_priv->ring[RCS] &&
 	    mode != dev_priv->relative_constants_mode) {
 		ret = intel_ring_begin(ring, 4);