drm/nv40/vpe: add support for PMPEG
authorBen Skeggs <bskeggs@redhat.com>
Mon, 4 Apr 2011 06:08:24 +0000 (16:08 +1000)
committerBen Skeggs <bskeggs@redhat.com>
Mon, 16 May 2011 00:48:56 +0000 (10:48 +1000)
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
drivers/gpu/drm/nouveau/Makefile
drivers/gpu/drm/nouveau/nouveau_drv.h
drivers/gpu/drm/nouveau/nouveau_state.c
drivers/gpu/drm/nouveau/nv40_fifo.c
drivers/gpu/drm/nouveau/nv40_mpeg.c [new file with mode: 0644]

index 2408242..11ff675 100644 (file)
@@ -21,6 +21,7 @@ nouveau-y := nouveau_drv.o nouveau_state.o nouveau_channel.o nouveau_mem.o \
              nv40_grctx.o nv50_grctx.o nvc0_grctx.o \
              nv84_crypt.o \
              nva3_copy.o nvc0_copy.o \
+             nv40_mpeg.o \
              nv04_instmem.o nv50_instmem.o nvc0_instmem.o \
              nv50_evo.o nv50_crtc.o nv50_dac.o nv50_sor.o \
              nv50_cursor.o nv50_display.o \
index 7dd5217..e1e29ba 100644 (file)
@@ -153,6 +153,7 @@ enum nouveau_flags {
 #define NVOBJ_ENGINE_CRYPT     2
 #define NVOBJ_ENGINE_COPY0     3
 #define NVOBJ_ENGINE_COPY1     4
+#define NVOBJ_ENGINE_MPEG      5
 #define NVOBJ_ENGINE_DISPLAY   15
 #define NVOBJ_ENGINE_NR                16
 
@@ -1154,6 +1155,9 @@ extern int  nva3_copy_create(struct drm_device *dev);
 /* nvc0_copy.c */
 extern int  nvc0_copy_create(struct drm_device *dev, int engine);
 
+/* nv40_mpeg.c */
+extern int  nv40_mpeg_create(struct drm_device *dev);
+
 /* nv04_instmem.c */
 extern int  nv04_instmem_init(struct drm_device *);
 extern void nv04_instmem_takedown(struct drm_device *);
index 769df65..f6c2535 100644 (file)
@@ -617,6 +617,9 @@ nouveau_card_init(struct drm_device *dev)
                break;
        }
 
+       if (dev_priv->card_type == NV_40)
+               nv40_mpeg_create(dev);
+
        if (!nouveau_noaccel) {
                for (e = 0; e < NVOBJ_ENGINE_NR; e++) {
                        if (dev_priv->eng[e]) {
index 49b9a35..68cb2d9 100644 (file)
@@ -115,6 +115,7 @@ nv40_fifo_do_load_context(struct drm_device *dev, int chid)
        nv_wr32(dev, 0x32e8, nv_ri32(dev, fc + 68));
        nv_wr32(dev, 0x2088, nv_ri32(dev, fc + 76));
        nv_wr32(dev, 0x3300, nv_ri32(dev, fc + 80));
+       nv_wr32(dev, 0x330c, nv_ri32(dev, fc + 84));
 
        nv_wr32(dev, NV03_PFIFO_CACHE1_GET, 0);
        nv_wr32(dev, NV03_PFIFO_CACHE1_PUT, 0);
@@ -186,6 +187,7 @@ nv40_fifo_unload_context(struct drm_device *dev)
        tmp |= (nv_rd32(dev, NV04_PFIFO_CACHE1_PUT) << 16);
        nv_wi32(dev, fc + 72, tmp);
 #endif
+       nv_wi32(dev, fc + 84, nv_rd32(dev, 0x330c));
 
        nv40_fifo_do_load_context(dev, pfifo->channels - 1);
        nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1,
diff --git a/drivers/gpu/drm/nouveau/nv40_mpeg.c b/drivers/gpu/drm/nouveau/nv40_mpeg.c
new file mode 100644 (file)
index 0000000..6d2af29
--- /dev/null
@@ -0,0 +1,311 @@
+/*
+ * Copyright 2011 Red Hat Inc.
+ *
+ * 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 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 COPYRIGHT HOLDER(S) OR AUTHOR(S) 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 Skeggs
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_ramht.h"
+
+struct nv40_mpeg_engine {
+       struct nouveau_exec_engine base;
+};
+
+static int
+nv40_mpeg_context_new(struct nouveau_channel *chan, int engine)
+{
+       struct drm_device *dev = chan->dev;
+       struct drm_nouveau_private *dev_priv = dev->dev_private;
+       struct nouveau_gpuobj *ctx = NULL;
+       unsigned long flags;
+       int ret;
+
+       NV_DEBUG(dev, "ch%d\n", chan->id);
+
+       ret = nouveau_gpuobj_new(dev, NULL, 264 * 4, 16, NVOBJ_FLAG_ZERO_ALLOC |
+                                NVOBJ_FLAG_ZERO_FREE, &ctx);
+       if (ret)
+               return ret;
+
+       nv_wo32(ctx, 0x78, 0x02001ec1);
+
+       spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
+       nv_mask(dev, 0x002500, 0x00000001, 0x00000000);
+       if ((nv_rd32(dev, 0x003204) & 0x1f) == chan->id)
+               nv_wr32(dev, 0x00330c, ctx->pinst >> 4);
+       nv_wo32(chan->ramfc, 0x54, ctx->pinst >> 4);
+       nv_mask(dev, 0x002500, 0x00000001, 0x00000001);
+       spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+
+       chan->engctx[engine] = ctx;
+       return 0;
+}
+
+static void
+nv40_mpeg_context_del(struct nouveau_channel *chan, int engine)
+{
+       struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
+       struct nouveau_gpuobj *ctx = chan->engctx[engine];
+       struct drm_device *dev = chan->dev;
+       unsigned long flags;
+       u32 inst = 0x80000000 | (ctx->pinst >> 4);
+
+       spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
+       nv_mask(dev, 0x00b32c, 0x00000001, 0x00000000);
+       if (nv_rd32(dev, 0x00b318) == inst)
+               nv_mask(dev, 0x00b318, 0x80000000, 0x00000000);
+       nv_mask(dev, 0x00b32c, 0x00000001, 0x00000001);
+       spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+
+       nouveau_gpuobj_ref(NULL, &ctx);
+       chan->engctx[engine] = NULL;
+}
+
+static int
+nv40_mpeg_object_new(struct nouveau_channel *chan, int engine,
+                     u32 handle, u16 class)
+{
+       struct drm_device *dev = chan->dev;
+       struct nouveau_gpuobj *obj = NULL;
+       int ret;
+
+       ret = nouveau_gpuobj_new(dev, chan, 20, 16, NVOBJ_FLAG_ZERO_ALLOC |
+                                NVOBJ_FLAG_ZERO_FREE, &obj);
+       if (ret)
+               return ret;
+       obj->engine = 2;
+       obj->class  = class;
+
+       nv_wo32(obj, 0x00, class);
+
+       ret = nouveau_ramht_insert(chan, handle, obj);
+       nouveau_gpuobj_ref(NULL, &obj);
+       return ret;
+}
+
+static int
+nv40_mpeg_init(struct drm_device *dev, int engine)
+{
+       struct drm_nouveau_private *dev_priv = dev->dev_private;
+       struct nv40_mpeg_engine *pmpeg = nv_engine(dev, engine);
+       int i;
+
+       /* VPE init */
+       nv_mask(dev, 0x000200, 0x00000002, 0x00000000);
+       nv_mask(dev, 0x000200, 0x00000002, 0x00000002);
+       nv_wr32(dev, 0x00b0e0, 0x00000020); /* nvidia: rd 0x01, wr 0x20 */
+       nv_wr32(dev, 0x00b0e8, 0x00000020); /* nvidia: rd 0x01, wr 0x20 */
+
+       for (i = 0; i < dev_priv->engine.fb.num_tiles; i++)
+               pmpeg->base.set_tile_region(dev, i);
+
+       /* PMPEG init */
+       nv_wr32(dev, 0x00b32c, 0x00000000);
+       nv_wr32(dev, 0x00b314, 0x00000100);
+       nv_wr32(dev, 0x00b220, 0x00000044);
+       nv_wr32(dev, 0x00b300, 0x02001ec1);
+       nv_mask(dev, 0x00b32c, 0x00000001, 0x00000001);
+
+       nv_wr32(dev, 0x00b100, 0xffffffff);
+       nv_wr32(dev, 0x00b140, 0xffffffff);
+
+       if (!nv_wait(dev, 0x00b200, 0x00000001, 0x00000000)) {
+               NV_ERROR(dev, "PMPEG init: 0x%08x\n", nv_rd32(dev, 0x00b200));
+               return -EBUSY;
+       }
+
+       return 0;
+}
+
+static int
+nv40_mpeg_fini(struct drm_device *dev, int engine)
+{
+       /*XXX: context save? */
+       nv_mask(dev, 0x00b32c, 0x00000001, 0x00000000);
+       nv_wr32(dev, 0x00b140, 0x00000000);
+       return 0;
+}
+
+static int
+nv40_mpeg_mthd_dma(struct nouveau_channel *chan, u32 class, u32 mthd, u32 data)
+{
+       struct drm_device *dev = chan->dev;
+       u32 inst = data << 4;
+       u32 dma0 = nv_ri32(dev, inst + 0);
+       u32 dma1 = nv_ri32(dev, inst + 4);
+       u32 dma2 = nv_ri32(dev, inst + 8);
+       u32 base = (dma2 & 0xfffff000) | (dma0 >> 20);
+       u32 size = dma1 + 1;
+
+       /* only allow linear DMA objects */
+       if (!(dma0 & 0x00002000))
+               return -EINVAL;
+
+       if (mthd == 0x0190) {
+               /* DMA_CMD */
+               nv_mask(dev, 0x00b300, 0x00030000, (dma0 & 0x00030000));
+               nv_wr32(dev, 0x00b334, base);
+               nv_wr32(dev, 0x00b324, size);
+       } else
+       if (mthd == 0x01a0) {
+               /* DMA_DATA */
+               nv_mask(dev, 0x00b300, 0x000c0000, (dma0 & 0x00030000) << 2);
+               nv_wr32(dev, 0x00b360, base);
+               nv_wr32(dev, 0x00b364, size);
+       } else {
+               /* DMA_IMAGE, VRAM only */
+               if (dma0 & 0x000c0000)
+                       return -EINVAL;
+
+               nv_wr32(dev, 0x00b370, base);
+               nv_wr32(dev, 0x00b374, size);
+       }
+
+       return 0;
+}
+
+static int
+nv40_mpeg_isr_chid(struct drm_device *dev, u32 inst)
+{
+       struct drm_nouveau_private *dev_priv = dev->dev_private;
+       struct nouveau_gpuobj *ctx;
+       unsigned long flags;
+       int i;
+
+       spin_lock_irqsave(&dev_priv->channels.lock, flags);
+       for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+               if (!dev_priv->channels.ptr[i])
+                       continue;
+
+               ctx = dev_priv->channels.ptr[i]->engctx[NVOBJ_ENGINE_MPEG];
+               if (ctx && ctx->pinst == inst)
+                       break;
+       }
+       spin_unlock_irqrestore(&dev_priv->channels.lock, flags);
+       return i;
+}
+
+static void
+nv40_vpe_set_tile_region(struct drm_device *dev, int i)
+{
+       struct drm_nouveau_private *dev_priv = dev->dev_private;
+       struct nouveau_tile_reg *tile = &dev_priv->tile.reg[i];
+
+       nv_wr32(dev, 0x00b008 + (i * 0x10), tile->pitch);
+       nv_wr32(dev, 0x00b004 + (i * 0x10), tile->limit);
+       nv_wr32(dev, 0x00b000 + (i * 0x10), tile->addr);
+}
+
+static void
+nv40_mpeg_isr(struct drm_device *dev)
+{
+       u32 inst = (nv_rd32(dev, 0x00b318) & 0x000fffff) << 4;
+       u32 chid = nv40_mpeg_isr_chid(dev, inst);
+       u32 stat = nv_rd32(dev, 0x00b100);
+       u32 type = nv_rd32(dev, 0x00b230);
+       u32 mthd = nv_rd32(dev, 0x00b234);
+       u32 data = nv_rd32(dev, 0x00b238);
+       u32 show = stat;
+
+       if (stat & 0x01000000) {
+               /* happens on initial binding of the object */
+               if (type == 0x00000020 && mthd == 0x0000) {
+                       nv_mask(dev, 0x00b308, 0x00000000, 0x00000000);
+                       show &= ~0x01000000;
+               }
+
+               if (type == 0x00000010) {
+                       if (!nouveau_gpuobj_mthd_call2(dev, chid, 0x3174, mthd, data))
+                               show &= ~0x01000000;
+               }
+       }
+
+       nv_wr32(dev, 0x00b100, stat);
+       nv_wr32(dev, 0x00b230, 0x00000001);
+
+       if (show && nouveau_ratelimit()) {
+               NV_INFO(dev, "PMPEG: Ch %d [0x%08x] 0x%08x 0x%08x 0x%08x 0x%08x\n",
+                       chid, inst, stat, type, mthd, data);
+       }
+}
+
+static void
+nv40_vpe_isr(struct drm_device *dev)
+{
+       if (nv_rd32(dev, 0x00b100))
+               nv40_mpeg_isr(dev);
+
+       if (nv_rd32(dev, 0x00b800)) {
+               u32 stat = nv_rd32(dev, 0x00b800);
+               NV_INFO(dev, "PMSRCH: 0x%08x\n", stat);
+               nv_wr32(dev, 0xb800, stat);
+       }
+}
+
+static void
+nv40_mpeg_destroy(struct drm_device *dev, int engine)
+{
+       struct nv40_mpeg_engine *pmpeg = nv_engine(dev, engine);
+
+       nouveau_irq_unregister(dev, 0);
+
+       NVOBJ_ENGINE_DEL(dev, MPEG);
+       kfree(pmpeg);
+}
+
+int
+nv40_mpeg_create(struct drm_device *dev)
+{
+       struct nv40_mpeg_engine *pmpeg;
+
+       pmpeg = kzalloc(sizeof(*pmpeg), GFP_KERNEL);
+       if (!pmpeg)
+               return -ENOMEM;
+
+       pmpeg->base.destroy = nv40_mpeg_destroy;
+       pmpeg->base.init = nv40_mpeg_init;
+       pmpeg->base.fini = nv40_mpeg_fini;
+       pmpeg->base.context_new = nv40_mpeg_context_new;
+       pmpeg->base.context_del = nv40_mpeg_context_del;
+       pmpeg->base.object_new = nv40_mpeg_object_new;
+
+       /* ISR vector, PMC_ENABLE bit,  and TILE regs are shared between
+        * all VPE engines, for this driver's purposes the PMPEG engine
+        * will be treated as the "master" and handle the global VPE
+        * bits too
+        */
+       pmpeg->base.set_tile_region = nv40_vpe_set_tile_region;
+       nouveau_irq_register(dev, 0, nv40_vpe_isr);
+
+       NVOBJ_ENGINE_ADD(dev, MPEG, &pmpeg->base);
+       NVOBJ_CLASS(dev, 0x3174, MPEG);
+       NVOBJ_MTHD (dev, 0x3174, 0x0190, nv40_mpeg_mthd_dma);
+       NVOBJ_MTHD (dev, 0x3174, 0x01a0, nv40_mpeg_mthd_dma);
+       NVOBJ_MTHD (dev, 0x3174, 0x01b0, nv40_mpeg_mthd_dma);
+
+#if 0
+       NVOBJ_ENGINE_ADD(dev, ME, &pme->base);
+       NVOBJ_CLASS(dev, 0x4075, ME);
+#endif
+       return 0;
+
+}