drm/v3d: Introduce a new DRM driver for Broadcom V3D V3.x+

This driver will be used to support Mesa on the Broadcom 7268 and 7278
platforms.

V3D 3.3 introduces an MMU, which means we no longer need CMA or vc4's
complicated CL/shader validation scheme.  This massively changes the
GEM behavior, so I've forked off to a new driver.

v2: Mark SUBMIT_CL as needing DRM_AUTH.  coccinelle fixes from kbuild
    test robot. Drop personal git link from MAINTAINERS.  Don't
    double-map dma-buf imported BOs.  Add kerneldoc about needing MMU
    eviction.  Drop prime vmap/unmap stubs.  Delay mmap offset setup
    to mmap time.  Use drm_dev_init instead of _alloc.  Use
    ktime_get() for wait_bo timeouts.  Drop drm_can_sleep() usage,
    since we don't modeset.  Switch page tables back to WC (debug
    change to coherent had slipped in).  Switch
    drm_gem_object_unreference_unlocked() to
    drm_gem_object_put_unlocked().  Simplify overflow mem handling by
    not sharing overflow mem between jobs.
v3: no changes
v4: align submit_cl to 64 bits (review by airlied), check zero flags in
    other ioctls.

Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch> (v4)
Acked-by: Dave Airlie <airlied@linux.ie> (v3, requested submit_cl change)
Link: https://patchwork.freedesktop.org/patch/msgid/20180430181058.30181-3-eric@anholt.net

16 files changed, 2943 insertions, 0 deletions

diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
index 757825ac60df..1c73a455fdb1 100644
--- a/drivers/gpu/drm/Kconfig
+++ b/drivers/gpu/drm/Kconfig
@@ -267,6 +267,8 @@ source "drivers/gpu/drm/amd/amdkfd/Kconfig"
 
 source "drivers/gpu/drm/imx/Kconfig"
 
+source "drivers/gpu/drm/v3d/Kconfig"
+
 source "drivers/gpu/drm/vc4/Kconfig"
 
 source "drivers/gpu/drm/etnaviv/Kconfig"
diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile
index 9d66657ea117..7a401edd8761 100644
--- a/drivers/gpu/drm/Makefile
+++ b/drivers/gpu/drm/Makefile
@@ -61,6 +61,7 @@ obj-$(CONFIG_DRM_MGA)	+= mga/
 obj-$(CONFIG_DRM_I810)	+= i810/
 obj-$(CONFIG_DRM_I915)	+= i915/
 obj-$(CONFIG_DRM_MGAG200) += mgag200/
+obj-$(CONFIG_DRM_V3D)  += v3d/
 obj-$(CONFIG_DRM_VC4)  += vc4/
 obj-$(CONFIG_DRM_CIRRUS_QEMU) += cirrus/
 obj-$(CONFIG_DRM_SIS)   += sis/
diff --git a/drivers/gpu/drm/v3d/Kconfig b/drivers/gpu/drm/v3d/Kconfig
new file mode 100644
index 000000000000..a0c0259355bd
--- /dev/null
+++ b/drivers/gpu/drm/v3d/Kconfig
@@ -0,0 +1,9 @@
+config DRM_V3D
+	tristate "Broadcom V3D 3.x and newer"
+	depends on ARCH_BCM || ARCH_BCMSTB || COMPILE_TEST
+	depends on DRM
+	depends on COMMON_CLK
+	select DRM_SCHED
+	help
+	  Choose this option if you have a system that has a Broadcom
+	  V3D 3.x or newer GPU, such as BCM7268.
diff --git a/drivers/gpu/drm/v3d/Makefile b/drivers/gpu/drm/v3d/Makefile
new file mode 100644
index 000000000000..34446e1de64f
--- /dev/null
+++ b/drivers/gpu/drm/v3d/Makefile
@@ -0,0 +1,18 @@
+# Please keep these build lists sorted!
+
+# core driver code
+v3d-y := \
+	v3d_bo.o \
+	v3d_drv.o \
+	v3d_fence.o \
+	v3d_gem.o \
+	v3d_irq.o \
+	v3d_mmu.o \
+	v3d_trace_points.o \
+	v3d_sched.o
+
+v3d-$(CONFIG_DEBUG_FS) += v3d_debugfs.o
+
+obj-$(CONFIG_DRM_V3D)  += v3d.o
+
+CFLAGS_v3d_trace_points.o := -I$(src)
diff --git a/drivers/gpu/drm/v3d/v3d_bo.c b/drivers/gpu/drm/v3d/v3d_bo.c
new file mode 100644
index 000000000000..7b1e2a549a71
--- /dev/null
+++ b/drivers/gpu/drm/v3d/v3d_bo.c
@@ -0,0 +1,389 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* Copyright (C) 2015-2018 Broadcom */
+
+/**
+ * DOC: V3D GEM BO management support
+ *
+ * Compared to VC4 (V3D 2.x), V3D 3.3 introduces an MMU between the
+ * GPU and the bus, allowing us to use shmem objects for our storage
+ * instead of CMA.
+ *
+ * Physically contiguous objects may still be imported to V3D, but the
+ * driver doesn't allocate physically contiguous objects on its own.
+ * Display engines requiring physically contiguous allocations should
+ * look into Mesa's "renderonly" support (as used by the Mesa pl111
+ * driver) for an example of how to integrate with V3D.
+ *
+ * Long term, we should support evicting pages from the MMU when under
+ * memory pressure (thus the v3d_bo_get_pages() refcounting), but
+ * that's not a high priority since our systems tend to not have swap.
+ */
+
+#include <linux/dma-buf.h>
+#include <linux/pfn_t.h>
+
+#include "v3d_drv.h"
+#include "uapi/drm/v3d_drm.h"
+
+/* Pins the shmem pages, fills in the .pages and .sgt fields of the BO, and maps
+ * it for DMA.
+ */
+static int
+v3d_bo_get_pages(struct v3d_bo *bo)
+{
+	struct drm_gem_object *obj = &bo->base;
+	struct drm_device *dev = obj->dev;
+	int npages = obj->size >> PAGE_SHIFT;
+	int ret = 0;
+
+	mutex_lock(&bo->lock);
+	if (bo->pages_refcount++ != 0)
+		goto unlock;
+
+	if (!obj->import_attach) {
+		bo->pages = drm_gem_get_pages(obj);
+		if (IS_ERR(bo->pages)) {
+			ret = PTR_ERR(bo->pages);
+			goto unlock;
+		}
+
+		bo->sgt = drm_prime_pages_to_sg(bo->pages, npages);
+		if (IS_ERR(bo->sgt)) {
+			ret = PTR_ERR(bo->sgt);
+			goto put_pages;
+		}
+
+		/* Map the pages for use by the GPU. */
+		dma_map_sg(dev->dev, bo->sgt->sgl,
+			   bo->sgt->nents, DMA_BIDIRECTIONAL);
+	} else {
+		bo->pages = kcalloc(npages, sizeof(*bo->pages), GFP_KERNEL);
+		if (!bo->pages)
+			goto put_pages;
+
+		drm_prime_sg_to_page_addr_arrays(bo->sgt, bo->pages,
+						 NULL, npages);
+
+		/* Note that dma-bufs come in mapped. */
+	}
+
+	mutex_unlock(&bo->lock);
+
+	return 0;
+
+put_pages:
+	drm_gem_put_pages(obj, bo->pages, true, true);
+	bo->pages = NULL;
+unlock:
+	bo->pages_refcount--;
+	mutex_unlock(&bo->lock);
+	return ret;
+}
+
+static void
+v3d_bo_put_pages(struct v3d_bo *bo)
+{
+	struct drm_gem_object *obj = &bo->base;
+
+	mutex_lock(&bo->lock);
+	if (--bo->pages_refcount == 0) {
+		if (!obj->import_attach) {
+			dma_unmap_sg(obj->dev->dev, bo->sgt->sgl,
+				     bo->sgt->nents, DMA_BIDIRECTIONAL);
+			sg_free_table(bo->sgt);
+			kfree(bo->sgt);
+			drm_gem_put_pages(obj, bo->pages, true, true);
+		} else {
+			kfree(bo->pages);
+		}
+	}
+	mutex_unlock(&bo->lock);
+}
+
+static struct v3d_bo *v3d_bo_create_struct(struct drm_device *dev,
+					   size_t unaligned_size)
+{
+	struct v3d_dev *v3d = to_v3d_dev(dev);
+	struct drm_gem_object *obj;
+	struct v3d_bo *bo;
+	size_t size = roundup(unaligned_size, PAGE_SIZE);
+	int ret;
+
+	if (size == 0)
+		return ERR_PTR(-EINVAL);
+
+	bo = kzalloc(sizeof(*bo), GFP_KERNEL);
+	if (!bo)
+		return ERR_PTR(-ENOMEM);
+	obj = &bo->base;
+
+	INIT_LIST_HEAD(&bo->vmas);
+	INIT_LIST_HEAD(&bo->unref_head);
+	mutex_init(&bo->lock);
+
+	ret = drm_gem_object_init(dev, obj, size);
+	if (ret)
+		goto free_bo;
+
+	spin_lock(&v3d->mm_lock);
+	ret = drm_mm_insert_node_generic(&v3d->mm, &bo->node,
+					 obj->size >> PAGE_SHIFT,
+					 GMP_GRANULARITY >> PAGE_SHIFT, 0, 0);
+	spin_unlock(&v3d->mm_lock);
+	if (ret)
+		goto free_obj;
+
+	return bo;
+
+free_obj:
+	drm_gem_object_release(obj);
+free_bo:
+	kfree(bo);
+	return ERR_PTR(ret);
+}
+
+struct v3d_bo *v3d_bo_create(struct drm_device *dev, struct drm_file *file_priv,
+			     size_t unaligned_size)
+{
+	struct v3d_dev *v3d = to_v3d_dev(dev);
+	struct drm_gem_object *obj;
+	struct v3d_bo *bo;
+	int ret;
+
+	bo = v3d_bo_create_struct(dev, unaligned_size);
+	if (IS_ERR(bo))
+		return bo;
+	obj = &bo->base;
+
+	bo->resv = &bo->_resv;
+	reservation_object_init(bo->resv);
+
+	ret = v3d_bo_get_pages(bo);
+	if (ret)
+		goto free_mm;
+
+	v3d_mmu_insert_ptes(bo);
+
+	mutex_lock(&v3d->bo_lock);
+	v3d->bo_stats.num_allocated++;
+	v3d->bo_stats.pages_allocated += obj->size >> PAGE_SHIFT;
+	mutex_unlock(&v3d->bo_lock);
+
+	return bo;
+
+free_mm:
+	spin_lock(&v3d->mm_lock);
+	drm_mm_remove_node(&bo->node);
+	spin_unlock(&v3d->mm_lock);
+
+	drm_gem_object_release(obj);
+	kfree(bo);
+	return ERR_PTR(ret);
+}
+
+/* Called DRM core on the last userspace/kernel unreference of the
+ * BO.
+ */
+void v3d_free_object(struct drm_gem_object *obj)
+{
+	struct v3d_dev *v3d = to_v3d_dev(obj->dev);
+	struct v3d_bo *bo = to_v3d_bo(obj);
+
+	mutex_lock(&v3d->bo_lock);
+	v3d->bo_stats.num_allocated--;
+	v3d->bo_stats.pages_allocated -= obj->size >> PAGE_SHIFT;
+	mutex_unlock(&v3d->bo_lock);
+
+	reservation_object_fini(&bo->_resv);
+
+	v3d_bo_put_pages(bo);
+
+	if (obj->import_attach)
+		drm_prime_gem_destroy(obj, bo->sgt);
+
+	v3d_mmu_remove_ptes(bo);
+	spin_lock(&v3d->mm_lock);
+	drm_mm_remove_node(&bo->node);
+	spin_unlock(&v3d->mm_lock);
+
+	mutex_destroy(&bo->lock);
+
+	drm_gem_object_release(obj);
+	kfree(bo);
+}
+
+struct reservation_object *v3d_prime_res_obj(struct drm_gem_object *obj)
+{
+	struct v3d_bo *bo = to_v3d_bo(obj);
+
+	return bo->resv;
+}
+
+static void
+v3d_set_mmap_vma_flags(struct vm_area_struct *vma)
+{
+	vma->vm_flags &= ~VM_PFNMAP;
+	vma->vm_flags |= VM_MIXEDMAP;
+	vma->vm_page_prot = pgprot_writecombine(vm_get_page_prot(vma->vm_flags));
+}
+
+int v3d_gem_fault(struct vm_fault *vmf)
+{
+	struct vm_area_struct *vma = vmf->vma;
+	struct drm_gem_object *obj = vma->vm_private_data;
+	struct v3d_bo *bo = to_v3d_bo(obj);
+	unsigned long pfn;
+	pgoff_t pgoff;
+	int ret;
+
+	/* We don't use vmf->pgoff since that has the fake offset: */
+	pgoff = (vmf->address - vma->vm_start) >> PAGE_SHIFT;
+	pfn = page_to_pfn(bo->pages[pgoff]);
+
+	ret = vm_insert_mixed(vma, vmf->address, __pfn_to_pfn_t(pfn, PFN_DEV));
+
+	switch (ret) {
+	case -EAGAIN:
+	case 0:
+	case -ERESTARTSYS:
+	case -EINTR:
+	case -EBUSY:
+		/*
+		 * EBUSY is ok: this just means that another thread
+		 * already did the job.
+		 */
+		return VM_FAULT_NOPAGE;
+	case -ENOMEM:
+		return VM_FAULT_OOM;
+	default:
+		return VM_FAULT_SIGBUS;
+	}
+}
+
+int v3d_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	int ret;
+
+	ret = drm_gem_mmap(filp, vma);
+	if (ret)
+		return ret;
+
+	v3d_set_mmap_vma_flags(vma);
+
+	return ret;
+}
+
+int v3d_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
+{
+	int ret;
+
+	ret = drm_gem_mmap_obj(obj, obj->size, vma);
+	if (ret < 0)
+		return ret;
+
+	v3d_set_mmap_vma_flags(vma);
+
+	return 0;
+}
+
+struct sg_table *
+v3d_prime_get_sg_table(struct drm_gem_object *obj)
+{
+	struct v3d_bo *bo = to_v3d_bo(obj);
+	int npages = obj->size >> PAGE_SHIFT;
+
+	return drm_prime_pages_to_sg(bo->pages, npages);
+}
+
+struct drm_gem_object *
+v3d_prime_import_sg_table(struct drm_device *dev,
+			  struct dma_buf_attachment *attach,
+			  struct sg_table *sgt)
+{
+	struct drm_gem_object *obj;
+	struct v3d_bo *bo;
+
+	bo = v3d_bo_create_struct(dev, attach->dmabuf->size);
+	if (IS_ERR(bo))
+		return ERR_CAST(bo);
+	obj = &bo->base;
+
+	bo->resv = attach->dmabuf->resv;
+
+	bo->sgt = sgt;
+	v3d_bo_get_pages(bo);
+
+	v3d_mmu_insert_ptes(bo);
+
+	return obj;
+}
+
+int v3d_create_bo_ioctl(struct drm_device *dev, void *data,
+			struct drm_file *file_priv)
+{
+	struct drm_v3d_create_bo *args = data;
+	struct v3d_bo *bo = NULL;
+	int ret;
+
+	if (args->flags != 0) {
+		DRM_INFO("unknown create_bo flags: %d\n", args->flags);
+		return -EINVAL;
+	}
+
+	bo = v3d_bo_create(dev, file_priv, PAGE_ALIGN(args->size));
+	if (IS_ERR(bo))
+		return PTR_ERR(bo);
+
+	args->offset = bo->node.start << PAGE_SHIFT;
+
+	ret = drm_gem_handle_create(file_priv, &bo->base, &args->handle);
+	drm_gem_object_put_unlocked(&bo->base);
+
+	return ret;
+}
+
+int v3d_mmap_bo_ioctl(struct drm_device *dev, void *data,
+		      struct drm_file *file_priv)
+{
+	struct drm_v3d_mmap_bo *args = data;
+	struct drm_gem_object *gem_obj;
+	int ret;
+
+	if (args->flags != 0) {
+		DRM_INFO("unknown mmap_bo flags: %d\n", args->flags);
+		return -EINVAL;
+	}
+
+	gem_obj = drm_gem_object_lookup(file_priv, args->handle);
+	if (!gem_obj) {
+		DRM_DEBUG("Failed to look up GEM BO %d\n", args->handle);
+		return -ENOENT;
+	}
+
+	ret = drm_gem_create_mmap_offset(gem_obj);
+	if (ret == 0)
+		args->offset = drm_vma_node_offset_addr(&gem_obj->vma_node);
+	drm_gem_object_put_unlocked(gem_obj);
+
+	return ret;
+}
+
+int v3d_get_bo_offset_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *file_priv)
+{
+	struct drm_v3d_get_bo_offset *args = data;
+	struct drm_gem_object *gem_obj;
+	struct v3d_bo *bo;
+
+	gem_obj = drm_gem_object_lookup(file_priv, args->handle);
+	if (!gem_obj) {
+		DRM_DEBUG("Failed to look up GEM BO %d\n", args->handle);
+		return -ENOENT;
+	}
+	bo = to_v3d_bo(gem_obj);
+
+	args->offset = bo->node.start << PAGE_SHIFT;
+
+	drm_gem_object_put_unlocked(gem_obj);
+	return 0;
+}
diff --git a/drivers/gpu/drm/v3d/v3d_debugfs.c b/drivers/gpu/drm/v3d/v3d_debugfs.c
new file mode 100644
index 000000000000..4db62c545748
--- /dev/null
+++ b/drivers/gpu/drm/v3d/v3d_debugfs.c
@@ -0,0 +1,191 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* Copyright (C) 2014-2018 Broadcom */
+
+#include <linux/circ_buf.h>
+#include <linux/ctype.h>
+#include <linux/debugfs.h>
+#include <linux/pm_runtime.h>
+#include <linux/seq_file.h>
+#include <drm/drmP.h>
+
+#include "v3d_drv.h"
+#include "v3d_regs.h"
+
+#define REGDEF(reg) { reg, #reg }
+struct v3d_reg_def {
+	u32 reg;
+	const char *name;
+};
+
+static const struct v3d_reg_def v3d_hub_reg_defs[] = {
+	REGDEF(V3D_HUB_AXICFG),
+	REGDEF(V3D_HUB_UIFCFG),
+	REGDEF(V3D_HUB_IDENT0),
+	REGDEF(V3D_HUB_IDENT1),
+	REGDEF(V3D_HUB_IDENT2),
+	REGDEF(V3D_HUB_IDENT3),
+	REGDEF(V3D_HUB_INT_STS),
+	REGDEF(V3D_HUB_INT_MSK_STS),
+};
+
+static const struct v3d_reg_def v3d_gca_reg_defs[] = {
+	REGDEF(V3D_GCA_SAFE_SHUTDOWN),
+	REGDEF(V3D_GCA_SAFE_SHUTDOWN_ACK),
+};
+
+static const struct v3d_reg_def v3d_core_reg_defs[] = {
+	REGDEF(V3D_CTL_IDENT0),
+	REGDEF(V3D_CTL_IDENT1),
+	REGDEF(V3D_CTL_IDENT2),
+	REGDEF(V3D_CTL_MISCCFG),
+	REGDEF(V3D_CTL_INT_STS),
+	REGDEF(V3D_CTL_INT_MSK_STS),
+	REGDEF(V3D_CLE_CT0CS),
+	REGDEF(V3D_CLE_CT0CA),
+	REGDEF(V3D_CLE_CT0EA),
+	REGDEF(V3D_CLE_CT1CS),
+	REGDEF(V3D_CLE_CT1CA),
+	REGDEF(V3D_CLE_CT1EA),
+
+	REGDEF(V3D_PTB_BPCA),
+	REGDEF(V3D_PTB_BPCS),
+
+	REGDEF(V3D_MMU_CTL),
+	REGDEF(V3D_MMU_VIO_ADDR),
+
+	REGDEF(V3D_GMP_STATUS),
+	REGDEF(V3D_GMP_CFG),
+	REGDEF(V3D_GMP_VIO_ADDR),
+};
+
+static int v3d_v3d_debugfs_regs(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *)m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct v3d_dev *v3d = to_v3d_dev(dev);
+	int i, core;
+
+	for (i = 0; i < ARRAY_SIZE(v3d_hub_reg_defs); i++) {
+		seq_printf(m, "%s (0x%04x): 0x%08x\n",
+			   v3d_hub_reg_defs[i].name, v3d_hub_reg_defs[i].reg,
+			   V3D_READ(v3d_hub_reg_defs[i].reg));
+	}
+
+	for (i = 0; i < ARRAY_SIZE(v3d_gca_reg_defs); i++) {
+		seq_printf(m, "%s (0x%04x): 0x%08x\n",
+			   v3d_gca_reg_defs[i].name, v3d_gca_reg_defs[i].reg,
+			   V3D_GCA_READ(v3d_gca_reg_defs[i].reg));
+	}
+
+	for (core = 0; core < v3d->cores; core++) {
+		for (i = 0; i < ARRAY_SIZE(v3d_core_reg_defs); i++) {
+			seq_printf(m, "core %d %s (0x%04x): 0x%08x\n",
+				   core,
+				   v3d_core_reg_defs[i].name,
+				   v3d_core_reg_defs[i].reg,
+				   V3D_CORE_READ(core,
+						 v3d_core_reg_defs[i].reg));
+		}
+	}
+
+	return 0;
+}
+
+static int v3d_v3d_debugfs_ident(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *)m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct v3d_dev *v3d = to_v3d_dev(dev);
+	u32 ident0, ident1, ident2, ident3, cores;
+	int ret, core;
+
+	ret = pm_runtime_get_sync(v3d->dev);
+	if (ret < 0)
+		return ret;
+
+	ident0 = V3D_READ(V3D_HUB_IDENT0);
+	ident1 = V3D_READ(V3D_HUB_IDENT1);
+	ident2 = V3D_READ(V3D_HUB_IDENT2);
+	ident3 = V3D_READ(V3D_HUB_IDENT3);
+	cores = V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_NCORES);
+
+	seq_printf(m, "Revision:   %d.%d.%d.%d\n",
+		   V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_TVER),
+		   V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_REV),
+		   V3D_GET_FIELD(ident3, V3D_HUB_IDENT3_IPREV),
+		   V3D_GET_FIELD(ident3, V3D_HUB_IDENT3_IPIDX));
+	seq_printf(m, "MMU:        %s\n",
+		   (ident2 & V3D_HUB_IDENT2_WITH_MMU) ? "yes" : "no");
+	seq_printf(m, "TFU:        %s\n",
+		   (ident1 & V3D_HUB_IDENT1_WITH_TFU) ? "yes" : "no");
+	seq_printf(m, "TSY:        %s\n",
+		   (ident1 & V3D_HUB_IDENT1_WITH_TSY) ? "yes" : "no");
+	seq_printf(m, "MSO:        %s\n",
+		   (ident1 & V3D_HUB_IDENT1_WITH_MSO) ? "yes" : "no");
+	seq_printf(m, "L3C:        %s (%dkb)\n",
+		   (ident1 & V3D_HUB_IDENT1_WITH_L3C) ? "yes" : "no",
+		   V3D_GET_FIELD(ident2, V3D_HUB_IDENT2_L3C_NKB));
+
+	for (core = 0; core < cores; core++) {
+		u32 misccfg;
+		u32 nslc, ntmu, qups;
+
+		ident0 = V3D_CORE_READ(core, V3D_CTL_IDENT0);
+		ident1 = V3D_CORE_READ(core, V3D_CTL_IDENT1);
+		ident2 = V3D_CORE_READ(core, V3D_CTL_IDENT2);
+		misccfg = V3D_CORE_READ(core, V3D_CTL_MISCCFG);
+
+		nslc = V3D_GET_FIELD(ident1, V3D_IDENT1_NSLC);
+		ntmu = V3D_GET_FIELD(ident1, V3D_IDENT1_NTMU);
+		qups = V3D_GET_FIELD(ident1, V3D_IDENT1_QUPS);
+
+		seq_printf(m, "Core %d:\n", core);
+		seq_printf(m, "  Revision:     %d.%d\n",
+			   V3D_GET_FIELD(ident0, V3D_IDENT0_VER),
+			   V3D_GET_FIELD(ident1, V3D_IDENT1_REV));
+		seq_printf(m, "  Slices:       %d\n", nslc);
+		seq_printf(m, "  TMUs:         %d\n", nslc * ntmu);
+		seq_printf(m, "  QPUs:         %d\n", nslc * qups);
+		seq_printf(m, "  Semaphores:   %d\n",
+			   V3D_GET_FIELD(ident1, V3D_IDENT1_NSEM));
+		seq_printf(m, "  BCG int:      %d\n",
+			   (ident2 & V3D_IDENT2_BCG_INT) != 0);
+		seq_printf(m, "  Override TMU: %d\n",
+			   (misccfg & V3D_MISCCFG_OVRTMUOUT) != 0);
+	}
+
+	pm_runtime_mark_last_busy(v3d->dev);
+	pm_runtime_put_autosuspend(v3d->dev);
+
+	return 0;
+}
+
+static int v3d_debugfs_bo_stats(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *)m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct v3d_dev *v3d = to_v3d_dev(dev);
+
+	mutex_lock(&v3d->bo_lock);
+	seq_printf(m, "allocated bos:          %d\n",
+		   v3d->bo_stats.num_allocated);
+	seq_printf(m, "allocated bo size (kb): %ld\n",
+		   (long)v3d->bo_stats.pages_allocated << (PAGE_SHIFT - 10));
+	mutex_unlock(&v3d->bo_lock);
+
+	return 0;
+}
+
+static const struct drm_info_list v3d_debugfs_list[] = {
+	{"v3d_ident", v3d_v3d_debugfs_ident, 0},
+	{"v3d_regs", v3d_v3d_debugfs_regs, 0},
+	{"bo_stats", v3d_debugfs_bo_stats, 0},
+};
+
+int
+v3d_debugfs_init(struct drm_minor *minor)
+{
+	return drm_debugfs_create_files(v3d_debugfs_list,
+					ARRAY_SIZE(v3d_debugfs_list),
+					minor->debugfs_root, minor);
+}
diff --git a/drivers/gpu/drm/v3d/v3d_drv.c b/drivers/gpu/drm/v3d/v3d_drv.c
new file mode 100644
index 000000000000..38e8041b5f0c
--- /dev/null
+++ b/drivers/gpu/drm/v3d/v3d_drv.c
@@ -0,0 +1,371 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* Copyright (C) 2014-2018 Broadcom */
+
+/**
+ * DOC: Broadcom V3D Graphics Driver
+ *
+ * This driver supports the Broadcom V3D 3.3 and 4.1 OpenGL ES GPUs.
+ * For V3D 2.x support, see the VC4 driver.
+ *
+ * Currently only single-core rendering using the binner and renderer
+ * is supported.  The TFU (texture formatting unit) and V3D 4.x's CSD
+ * (compute shader dispatch) are not yet supported.
+ */
+
+#include <linux/clk.h>
+#include <linux/device.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/of_platform.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+#include <drm/drm_fb_cma_helper.h>
+#include <drm/drm_fb_helper.h>
+
+#include "uapi/drm/v3d_drm.h"
+#include "v3d_drv.h"
+#include "v3d_regs.h"
+
+#define DRIVER_NAME "v3d"
+#define DRIVER_DESC "Broadcom V3D graphics"
+#define DRIVER_DATE "20180419"
+#define DRIVER_MAJOR 1
+#define DRIVER_MINOR 0
+#define DRIVER_PATCHLEVEL 0
+
+#ifdef CONFIG_PM
+static int v3d_runtime_suspend(struct device *dev)
+{
+	struct drm_device *drm = dev_get_drvdata(dev);
+	struct v3d_dev *v3d = to_v3d_dev(drm);
+
+	v3d_irq_disable(v3d);
+
+	clk_disable_unprepare(v3d->clk);
+
+	return 0;
+}
+
+static int v3d_runtime_resume(struct device *dev)
+{
+	struct drm_device *drm = dev_get_drvdata(dev);
+	struct v3d_dev *v3d = to_v3d_dev(drm);
+	int ret;
+
+	ret = clk_prepare_enable(v3d->clk);
+	if (ret != 0)
+		return ret;
+
+	/* XXX: VPM base */
+
+	v3d_mmu_set_page_table(v3d);
+	v3d_irq_enable(v3d);
+
+	return 0;
+}
+#endif
+
+static const struct dev_pm_ops v3d_v3d_pm_ops = {
+	SET_RUNTIME_PM_OPS(v3d_runtime_suspend, v3d_runtime_resume, NULL)
+};
+
+static int v3d_get_param_ioctl(struct drm_device *dev, void *data,
+			       struct drm_file *file_priv)
+{
+	struct v3d_dev *v3d = to_v3d_dev(dev);
+	struct drm_v3d_get_param *args = data;
+	int ret;
+	static const u32 reg_map[] = {
+		[DRM_V3D_PARAM_V3D_UIFCFG] = V3D_HUB_UIFCFG,
+		[DRM_V3D_PARAM_V3D_HUB_IDENT1] = V3D_HUB_IDENT1,
+		[DRM_V3D_PARAM_V3D_HUB_IDENT2] = V3D_HUB_IDENT2,
+		[DRM_V3D_PARAM_V3D_HUB_IDENT3] = V3D_HUB_IDENT3,
+		[DRM_V3D_PARAM_V3D_CORE0_IDENT0] = V3D_CTL_IDENT0,
+		[DRM_V3D_PARAM_V3D_CORE0_IDENT1] = V3D_CTL_IDENT1,
+		[DRM_V3D_PARAM_V3D_CORE0_IDENT2] = V3D_CTL_IDENT2,
+	};
+
+	if (args->pad != 0)
+		return -EINVAL;
+
+	/* Note that DRM_V3D_PARAM_V3D_CORE0_IDENT0 is 0, so we need
+	 * to explicitly allow it in the "the register in our
+	 * parameter map" check.
+	 */
+	if (args->param < ARRAY_SIZE(reg_map) &&
+	    (reg_map[args->param] ||
+	     args->param == DRM_V3D_PARAM_V3D_CORE0_IDENT0)) {
+		u32 offset = reg_map[args->param];
+
+		if (args->value != 0)
+			return -EINVAL;
+
+		ret = pm_runtime_get_sync(v3d->dev);
+		if (args->param >= DRM_V3D_PARAM_V3D_CORE0_IDENT0 &&
+		    args->param <= DRM_V3D_PARAM_V3D_CORE0_IDENT2) {
+			args->value = V3D_CORE_READ(0, offset);
+		} else {
+			args->value = V3D_READ(offset);
+		}
+		pm_runtime_mark_last_busy(v3d->dev);
+		pm_runtime_put_autosuspend(v3d->dev);
+		return 0;
+	}
+
+	/* Any params that aren't just register reads would go here. */
+
+	DRM_DEBUG("Unknown parameter %d\n", args->param);
+	return -EINVAL;
+}
+
+static int
+v3d_open(struct drm_device *dev, struct drm_file *file)
+{
+	struct v3d_dev *v3d = to_v3d_dev(dev);
+	struct v3d_file_priv *v3d_priv;
+	int i;
+
+	v3d_priv = kzalloc(sizeof(*v3d_priv), GFP_KERNEL);
+	if (!v3d_priv)
+		return -ENOMEM;
+
+	v3d_priv->v3d = v3d;
+
+	for (i = 0; i < V3D_MAX_QUEUES; i++) {
+		drm_sched_entity_init(&v3d->queue[i].sched,
+				      &v3d_priv->sched_entity[i],
+				      &v3d->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL],
+				      32, NULL);
+	}
+
+	file->driver_priv = v3d_priv;
+
+	return 0;
+}
+
+static void
+v3d_postclose(struct drm_device *dev, struct drm_file *file)
+{
+	struct v3d_dev *v3d = to_v3d_dev(dev);
+	struct v3d_file_priv *v3d_priv = file->driver_priv;
+	enum v3d_queue q;
+
+	for (q = 0; q < V3D_MAX_QUEUES; q++) {
+		drm_sched_entity_fini(&v3d->queue[q].sched,
+				      &v3d_priv->sched_entity[q]);
+	}
+
+	kfree(v3d_priv);
+}
+
+static const struct file_operations v3d_drm_fops = {
+	.owner = THIS_MODULE,
+	.open = drm_open,
+	.release = drm_release,
+	.unlocked_ioctl = drm_ioctl,
+	.mmap = v3d_mmap,
+	.poll = drm_poll,
+	.read = drm_read,
+	.compat_ioctl = drm_compat_ioctl,
+	.llseek = noop_llseek,
+};
+
+/* DRM_AUTH is required on SUBMIT_CL for now, while we don't have GMP
+ * protection between clients.  Note that render nodes would be be
+ * able to submit CLs that could access BOs from clients authenticated
+ * with the master node.
+ */
+static const struct drm_ioctl_desc v3d_drm_ioctls[] = {
+	DRM_IOCTL_DEF_DRV(V3D_SUBMIT_CL, v3d_submit_cl_ioctl, DRM_RENDER_ALLOW | DRM_AUTH),
+	DRM_IOCTL_DEF_DRV(V3D_WAIT_BO, v3d_wait_bo_ioctl, DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(V3D_CREATE_BO, v3d_create_bo_ioctl, DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(V3D_MMAP_BO, v3d_mmap_bo_ioctl, DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(V3D_GET_PARAM, v3d_get_param_ioctl, DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(V3D_GET_BO_OFFSET, v3d_get_bo_offset_ioctl, DRM_RENDER_ALLOW),
+};
+
+static const struct vm_operations_struct v3d_vm_ops = {
+	.fault = v3d_gem_fault,
+	.open = drm_gem_vm_open,
+	.close = drm_gem_vm_close,
+};
+
+static struct drm_driver v3d_drm_driver = {
+	.driver_features = (DRIVER_GEM |
+			    DRIVER_RENDER |
+			    DRIVER_PRIME |
+			    DRIVER_SYNCOBJ),
+
+	.open = v3d_open,
+	.postclose = v3d_postclose,
+
+#if defined(CONFIG_DEBUG_FS)
+	.debugfs_init = v3d_debugfs_init,
+#endif
+
+	.gem_free_object_unlocked = v3d_free_object,
+	.gem_vm_ops = &v3d_vm_ops,
+
+	.prime_handle_to_fd = drm_gem_prime_handle_to_fd,
+	.prime_fd_to_handle = drm_gem_prime_fd_to_handle,
+	.gem_prime_import = drm_gem_prime_import,
+	.gem_prime_export = drm_gem_prime_export,
+	.gem_prime_res_obj = v3d_prime_res_obj,
+	.gem_prime_get_sg_table	= v3d_prime_get_sg_table,
+	.gem_prime_import_sg_table = v3d_prime_import_sg_table,
+	.gem_prime_mmap = v3d_prime_mmap,
+
+	.ioctls = v3d_drm_ioctls,
+	.num_ioctls = ARRAY_SIZE(v3d_drm_ioctls),
+	.fops = &v3d_drm_fops,
+
+	.name = DRIVER_NAME,
+	.desc = DRIVER_DESC,
+	.date = DRIVER_DATE,
+	.major = DRIVER_MAJOR,
+	.minor = DRIVER_MINOR,
+	.patchlevel = DRIVER_PATCHLEVEL,
+};
+
+static const struct of_device_id v3d_of_match[] = {
+	{ .compatible = "brcm,7268-v3d" },
+	{ .compatible = "brcm,7278-v3d" },
+	{},
+};
+MODULE_DEVICE_TABLE(of, v3d_of_match);
+
+static int
+map_regs(struct v3d_dev *v3d, void __iomem **regs, const char *name)
+{
+	struct resource *res =
+		platform_get_resource_byname(v3d->pdev, IORESOURCE_MEM, name);
+
+	*regs = devm_ioremap_resource(v3d->dev, res);
+	return PTR_ERR_OR_ZERO(*regs);
+}
+
+static int v3d_platform_drm_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct drm_device *drm;
+	struct v3d_dev *v3d;
+	int ret;
+	u32 ident1;
+
+	dev->coherent_dma_mask = DMA_BIT_MASK(36);
+
+	v3d = kzalloc(sizeof(*v3d), GFP_KERNEL);
+	if (!v3d)
+		return -ENOMEM;
+	v3d->dev = dev;
+	v3d->pdev = pdev;
+	drm = &v3d->drm;
+
+	ret = map_regs(v3d, &v3d->bridge_regs, "bridge");
+	if (ret)
+		goto dev_free;
+
+	ret = map_regs(v3d, &v3d->hub_regs, "hub");
+	if (ret)
+		goto dev_free;
+
+	ret = map_regs(v3d, &v3d->core_regs[0], "core0");
+	if (ret)
+		goto dev_free;
+
+	ident1 = V3D_READ(V3D_HUB_IDENT1);
+	v3d->ver = (V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_TVER) * 10 +
+		    V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_REV));
+	v3d->cores = V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_NCORES);
+	WARN_ON(v3d->cores > 1); /* multicore not yet implemented */
+
+	if (v3d->ver < 41) {
+		ret = map_regs(v3d, &v3d->gca_regs, "gca");
+		if (ret)
+			goto dev_free;
+	}
+
+	v3d->mmu_scratch = dma_alloc_wc(dev, 4096, &v3d->mmu_scratch_paddr,
+					GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO);
+	if (!v3d->mmu_scratch) {
+		dev_err(dev, "Failed to allocate MMU scratch page\n");
+		ret = -ENOMEM;
+		goto dev_free;
+	}
+
+	pm_runtime_use_autosuspend(dev);
+	pm_runtime_set_autosuspend_delay(dev, 50);
+	pm_runtime_enable(dev);
+
+	ret = drm_dev_init(&v3d->drm, &v3d_drm_driver, dev);
+	if (ret)
+		goto dma_free;
+
+	platform_set_drvdata(pdev, drm);
+	drm->dev_private = v3d;
+
+	ret = v3d_gem_init(drm);
+	if (ret)
+		goto dev_destroy;
+
+	v3d_irq_init(v3d);
+
+	ret = drm_dev_register(drm, 0);
+	if (ret)
+		goto gem_destroy;
+
+	return 0;
+
+gem_destroy:
+	v3d_gem_destroy(drm);
+dev_destroy:
+	drm_dev_put(drm);
+dma_free:
+	dma_free_wc(dev, 4096, v3d->mmu_scratch, v3d->mmu_scratch_paddr);
+dev_free:
+	kfree(v3d);
+	return ret;
+}
+
+static int v3d_platform_drm_remove(struct platform_device *pdev)
+{
+	struct drm_device *drm = platform_get_drvdata(pdev);
+	struct v3d_dev *v3d = to_v3d_dev(drm);
+
+	drm_dev_unregister(drm);
+
+	v3d_gem_destroy(drm);
+
+	drm_dev_put(drm);
+
+	dma_free_wc(v3d->dev, 4096, v3d->mmu_scratch, v3d->mmu_scratch_paddr);
+
+	return 0;
+}
+
+static struct platform_driver v3d_platform_driver = {
+	.probe		= v3d_platform_drm_probe,
+	.remove		= v3d_platform_drm_remove,
+	.driver		= {
+		.name	= "v3d",
+		.of_match_table = v3d_of_match,
+	},
+};
+
+static int __init v3d_drm_register(void)
+{
+	return platform_driver_register(&v3d_platform_driver);
+}
+
+static void __exit v3d_drm_unregister(void)
+{
+	platform_driver_unregister(&v3d_platform_driver);
+}
+
+module_init(v3d_drm_register);
+module_exit(v3d_drm_unregister);
+
+MODULE_ALIAS("platform:v3d-drm");
+MODULE_DESCRIPTION("Broadcom V3D DRM Driver");
+MODULE_AUTHOR("Eric Anholt <eric@anholt.net>");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/gpu/drm/v3d/v3d_drv.h b/drivers/gpu/drm/v3d/v3d_drv.h
new file mode 100644
index 000000000000..a043ac3aae98
--- /dev/null
+++ b/drivers/gpu/drm/v3d/v3d_drv.h
@@ -0,0 +1,294 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* Copyright (C) 2015-2018 Broadcom */
+
+#include <linux/reservation.h>
+#include <drm/drmP.h>
+#include <drm/drm_encoder.h>
+#include <drm/drm_gem.h>
+#include <drm/gpu_scheduler.h>
+
+#define GMP_GRANULARITY (128 * 1024)
+
+/* Enum for each of the V3D queues.  We maintain various queue
+ * tracking as an array because at some point we'll want to support
+ * the TFU (texture formatting unit) as another queue.
+ */
+enum v3d_queue {
+	V3D_BIN,
+	V3D_RENDER,
+};
+
+#define V3D_MAX_QUEUES (V3D_RENDER + 1)
+
+struct v3d_queue_state {
+	struct drm_gpu_scheduler sched;
+
+	u64 fence_context;
+	u64 emit_seq