drm/nouveau: Add DRM driver for NVIDIA GPUs

This adds a drm/kms staging non-API stable driver for GPUs from NVIDIA.

This driver is a KMS-based driver and requires a compatible nouveau
userspace libdrm and nouveau X.org driver.

This driver requires firmware files not available in this kernel tree,
interested parties can find them via the nouveau project git archive.

This driver is reverse engineered, and is in no way supported by nVidia.

Support for nearly the complete range of nvidia hw from nv04->g80 (nv50)
is available, and the kms driver should support driving nearly all
output types (displayport is under development still) along with supporting
suspend/resume.

This work is all from the upstream nouveau project found at
nouveau.freedesktop.org.

The original authors list from nouveau git tree is:
Anssi Hannula <anssi.hannula@iki.fi>
Ben Skeggs <bskeggs@redhat.com>
Francisco Jerez <currojerez@riseup.net>
Maarten Maathuis <madman2003@gmail.com>
Marcin Kościelnicki <koriakin@0x04.net>
Matthew Garrett <mjg@redhat.com>
Matt Parnell <mparnell@gmail.com>
Patrice Mandin <patmandin@gmail.com>
Pekka Paalanen <pq@iki.fi>
Xavier Chantry <shiningxc@gmail.com>
along with project founder Stephane Marchesin <marchesin@icps.u-strasbg.fr>

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

1 files changed, 321 insertions, 0 deletions

diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
new file mode 100644
index 00000000000..4c7f1e403e8
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
@@ -0,0 +1,321 @@
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include <linux/pagemap.h>
+
+#define NV_CTXDMA_PAGE_SHIFT 12
+#define NV_CTXDMA_PAGE_SIZE  (1 << NV_CTXDMA_PAGE_SHIFT)
+#define NV_CTXDMA_PAGE_MASK  (NV_CTXDMA_PAGE_SIZE - 1)
+
+struct nouveau_sgdma_be {
+	struct ttm_backend backend;
+	struct drm_device *dev;
+
+	dma_addr_t *pages;
+	unsigned nr_pages;
+
+	unsigned pte_start;
+	bool bound;
+};
+
+static int
+nouveau_sgdma_populate(struct ttm_backend *be, unsigned long num_pages,
+		       struct page **pages, struct page *dummy_read_page)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+	struct drm_device *dev = nvbe->dev;
+
+	NV_DEBUG(nvbe->dev, "num_pages = %ld\n", num_pages);
+
+	if (nvbe->pages)
+		return -EINVAL;
+
+	nvbe->pages = kmalloc(sizeof(dma_addr_t) * num_pages, GFP_KERNEL);
+	if (!nvbe->pages)
+		return -ENOMEM;
+
+	nvbe->nr_pages = 0;
+	while (num_pages--) {
+		nvbe->pages[nvbe->nr_pages] =
+			pci_map_page(dev->pdev, pages[nvbe->nr_pages], 0,
+				     PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
+		if (pci_dma_mapping_error(dev->pdev,
+					  nvbe->pages[nvbe->nr_pages])) {
+			be->func->clear(be);
+			return -EFAULT;
+		}
+
+		nvbe->nr_pages++;
+	}
+
+	return 0;
+}
+
+static void
+nouveau_sgdma_clear(struct ttm_backend *be)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+	struct drm_device *dev = nvbe->dev;
+
+	NV_DEBUG(nvbe->dev, "\n");
+
+	if (nvbe && nvbe->pages) {
+		if (nvbe->bound)
+			be->func->unbind(be);
+
+		while (nvbe->nr_pages--) {
+			pci_unmap_page(dev->pdev, nvbe->pages[nvbe->nr_pages],
+				       PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
+		}
+		kfree(nvbe->pages);
+		nvbe->pages = NULL;
+		nvbe->nr_pages = 0;
+	}
+}
+
+static inline unsigned
+nouveau_sgdma_pte(struct drm_device *dev, uint64_t offset)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	unsigned pte = (offset >> NV_CTXDMA_PAGE_SHIFT);
+
+	if (dev_priv->card_type < NV_50)
+		return pte + 2;
+
+	return pte << 1;
+}
+
+static int
+nouveau_sgdma_bind(struct ttm_backend *be, struct ttm_mem_reg *mem)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+	struct drm_device *dev = nvbe->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = dev_priv->gart_info.sg_ctxdma;
+	unsigned i, j, pte;
+
+	NV_DEBUG(dev, "pg=0x%lx\n", mem->mm_node->start);
+
+	dev_priv->engine.instmem.prepare_access(nvbe->dev, true);
+	pte = nouveau_sgdma_pte(nvbe->dev, mem->mm_node->start << PAGE_SHIFT);
+	nvbe->pte_start = pte;
+	for (i = 0; i < nvbe->nr_pages; i++) {
+		dma_addr_t dma_offset = nvbe->pages[i];
+		uint32_t offset_l = lower_32_bits(dma_offset);
+		uint32_t offset_h = upper_32_bits(dma_offset);
+
+		for (j = 0; j < PAGE_SIZE / NV_CTXDMA_PAGE_SIZE; j++) {
+			if (dev_priv->card_type < NV_50)
+				nv_wo32(dev, gpuobj, pte++, offset_l | 3);
+			else {
+				nv_wo32(dev, gpuobj, pte++, offset_l | 0x21);
+				nv_wo32(dev, gpuobj, pte++, offset_h & 0xff);
+			}
+
+			dma_offset += NV_CTXDMA_PAGE_SIZE;
+		}
+	}
+	dev_priv->engine.instmem.finish_access(nvbe->dev);
+
+	if (dev_priv->card_type == NV_50) {
+		nv_wr32(dev, 0x100c80, 0x00050001);
+		if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+			NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+			NV_ERROR(dev, "0x100c80 = 0x%08x\n",
+						nv_rd32(dev, 0x100c80));
+			return -EBUSY;
+		}
+
+		nv_wr32(dev, 0x100c80, 0x00000001);
+		if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+			NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+			NV_ERROR(dev, "0x100c80 = 0x%08x\n",
+						nv_rd32(dev, 0x100c80));
+			return -EBUSY;
+		}
+	}
+
+	nvbe->bound = true;
+	return 0;
+}
+
+static int
+nouveau_sgdma_unbind(struct ttm_backend *be)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+	struct drm_device *dev = nvbe->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = dev_priv->gart_info.sg_ctxdma;
+	unsigned i, j, pte;
+
+	NV_DEBUG(dev, "\n");
+
+	if (!nvbe->bound)
+		return 0;
+
+	dev_priv->engine.instmem.prepare_access(nvbe->dev, true);
+	pte = nvbe->pte_start;
+	for (i = 0; i < nvbe->nr_pages; i++) {
+		dma_addr_t dma_offset = dev_priv->gart_info.sg_dummy_bus;
+
+		for (j = 0; j < PAGE_SIZE / NV_CTXDMA_PAGE_SIZE; j++) {
+			if (dev_priv->card_type < NV_50)
+				nv_wo32(dev, gpuobj, pte++, dma_offset | 3);
+			else {
+				nv_wo32(dev, gpuobj, pte++, dma_offset | 0x21);
+				nv_wo32(dev, gpuobj, pte++, 0x00000000);
+			}
+
+			dma_offset += NV_CTXDMA_PAGE_SIZE;
+		}
+	}
+	dev_priv->engine.instmem.finish_access(nvbe->dev);
+
+	nvbe->bound = false;
+	return 0;
+}
+
+static void
+nouveau_sgdma_destroy(struct ttm_backend *be)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+
+	if (be) {
+		NV_DEBUG(nvbe->dev, "\n");
+
+		if (nvbe) {
+			if (nvbe->pages)
+				be->func->clear(be);
+			kfree(nvbe);
+		}
+	}
+}
+
+static struct ttm_backend_func nouveau_sgdma_backend = {
+	.populate		= nouveau_sgdma_populate,
+	.clear			= nouveau_sgdma_clear,
+	.bind			= nouveau_sgdma_bind,
+	.unbind			= nouveau_sgdma_unbind,
+	.destroy		= nouveau_sgdma_destroy
+};
+
+struct ttm_backend *
+nouveau_sgdma_init_ttm(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_sgdma_be *nvbe;
+
+	if (!dev_priv->gart_info.sg_ctxdma)
+		return NULL;
+
+	nvbe = kzalloc(sizeof(*nvbe), GFP_KERNEL);
+	if (!nvbe)
+		return NULL;
+
+	nvbe->dev = dev;
+
+	nvbe->backend.func	= &nouveau_sgdma_backend;
+
+	return &nvbe->backend;
+}
+
+int
+nouveau_sgdma_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = NULL;
+	uint32_t aper_size, obj_size;
+	int i, ret;
+
+	if (dev_priv->card_type < NV_50) {
+		aper_size = (64 * 1024 * 1024);
+		obj_size  = (aper_size >> NV_CTXDMA_PAGE_SHIFT) * 4;
+		obj_size += 8; /* ctxdma header */
+	} else {
+		/* 1 entire VM page table */
+		aper_size = (512 * 1024 * 1024);
+		obj_size  = (aper_size >> NV_CTXDMA_PAGE_SHIFT) * 8;
+	}
+
+	ret = nouveau_gpuobj_new(dev, NULL, obj_size, 16,
+				      NVOBJ_FLAG_ALLOW_NO_REFS |
+				      NVOBJ_FLAG_ZERO_ALLOC |
+				      NVOBJ_FLAG_ZERO_FREE, &gpuobj);
+	if (ret) {
+		NV_ERROR(dev, "Error creating sgdma object: %d\n", ret);
+		return ret;
+	}
+
+	dev_priv->gart_info.sg_dummy_page =
+		alloc_page(GFP_KERNEL|__GFP_DMA32);
+	set_bit(PG_locked, &dev_priv->gart_info.sg_dummy_page->flags);
+	dev_priv->gart_info.sg_dummy_bus =
+		pci_map_page(dev->pdev, dev_priv->gart_info.sg_dummy_page, 0,
+			     PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	if (dev_priv->card_type < NV_50) {
+		/* Maybe use NV_DMA_TARGET_AGP for PCIE? NVIDIA do this, and
+		 * confirmed to work on c51.  Perhaps means NV_DMA_TARGET_PCIE
+		 * on those cards? */
+		nv_wo32(dev, gpuobj, 0, NV_CLASS_DMA_IN_MEMORY |
+				       (1 << 12) /* PT present */ |
+				       (0 << 13) /* PT *not* linear */ |
+				       (NV_DMA_ACCESS_RW  << 14) |
+				       (NV_DMA_TARGET_PCI << 16));
+		nv_wo32(dev, gpuobj, 1, aper_size - 1);
+		for (i = 2; i < 2 + (aper_size >> 12); i++) {
+			nv_wo32(dev, gpuobj, i,
+				    dev_priv->gart_info.sg_dummy_bus | 3);
+		}
+	} else {
+		for (i = 0; i < obj_size; i += 8) {
+			nv_wo32(dev, gpuobj, (i+0)/4,
+				    dev_priv->gart_info.sg_dummy_bus | 0x21);
+			nv_wo32(dev, gpuobj, (i+4)/4, 0);
+		}
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	dev_priv->gart_info.type      = NOUVEAU_GART_SGDMA;
+	dev_priv->gart_info.aper_base = 0;
+	dev_priv->gart_info.aper_size = aper_size;
+	dev_priv->gart_info.sg_ctxdma = gpuobj;
+	return 0;
+}
+
+void
+nouveau_sgdma_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->gart_info.sg_dummy_page) {
+		pci_unmap_page(dev->pdev, dev_priv->gart_info.sg_dummy_bus,
+			       NV_CTXDMA_PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
+		unlock_page(dev_priv->gart_info.sg_dummy_page);
+		__free_page(dev_priv->gart_info.sg_dummy_page);
+		dev_priv->gart_info.sg_dummy_page = NULL;
+		dev_priv->gart_info.sg_dummy_bus = 0;
+	}
+
+	nouveau_gpuobj_del(dev, &dev_priv->gart_info.sg_ctxdma);
+}
+
+int
+nouveau_sgdma_get_page(struct drm_device *dev, uint32_t offset, uint32_t *page)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = dev_priv->gart_info.sg_ctxdma;
+	struct nouveau_instmem_engine *instmem = &dev_priv->engine.instmem;
+	int pte;
+
+	pte = (offset >> NV_CTXDMA_PAGE_SHIFT);
+	if (dev_priv->card_type < NV_50) {
+		instmem->prepare_access(dev, false);
+		*page = nv_ro32(dev, gpuobj, (pte + 2)) & ~NV_CTXDMA_PAGE_MASK;
+		instmem->finish_access(dev);
+		return 0;
+	}
+
+	NV_ERROR(dev, "Unimplemented on NV50\n");
+	return -EINVAL;
+}