Merge branch 'perf-fixes-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel...
[firefly-linux-kernel-4.4.55.git] / drivers / gpu / drm / nouveau / nv40_graph.c
index d3e0a2a6acf8e982d5077ce587b8c4982973e77d..2b332bb55acf75b4cce5bafb9dc5110538e47428 100644 (file)
  *
  */
 
-#include <linux/firmware.h>
-
 #include "drmP.h"
 #include "drm.h"
 #include "nouveau_drv.h"
-
-MODULE_FIRMWARE("nouveau/nv40.ctxprog");
-MODULE_FIRMWARE("nouveau/nv40.ctxvals");
-MODULE_FIRMWARE("nouveau/nv41.ctxprog");
-MODULE_FIRMWARE("nouveau/nv41.ctxvals");
-MODULE_FIRMWARE("nouveau/nv42.ctxprog");
-MODULE_FIRMWARE("nouveau/nv42.ctxvals");
-MODULE_FIRMWARE("nouveau/nv43.ctxprog");
-MODULE_FIRMWARE("nouveau/nv43.ctxvals");
-MODULE_FIRMWARE("nouveau/nv44.ctxprog");
-MODULE_FIRMWARE("nouveau/nv44.ctxvals");
-MODULE_FIRMWARE("nouveau/nv46.ctxprog");
-MODULE_FIRMWARE("nouveau/nv46.ctxvals");
-MODULE_FIRMWARE("nouveau/nv47.ctxprog");
-MODULE_FIRMWARE("nouveau/nv47.ctxvals");
-MODULE_FIRMWARE("nouveau/nv49.ctxprog");
-MODULE_FIRMWARE("nouveau/nv49.ctxvals");
-MODULE_FIRMWARE("nouveau/nv4a.ctxprog");
-MODULE_FIRMWARE("nouveau/nv4a.ctxvals");
-MODULE_FIRMWARE("nouveau/nv4b.ctxprog");
-MODULE_FIRMWARE("nouveau/nv4b.ctxvals");
-MODULE_FIRMWARE("nouveau/nv4c.ctxprog");
-MODULE_FIRMWARE("nouveau/nv4c.ctxvals");
-MODULE_FIRMWARE("nouveau/nv4e.ctxprog");
-MODULE_FIRMWARE("nouveau/nv4e.ctxvals");
+#include "nouveau_grctx.h"
 
 struct nouveau_channel *
 nv40_graph_channel(struct drm_device *dev)
@@ -83,27 +57,30 @@ nv40_graph_create_context(struct nouveau_channel *chan)
 {
        struct drm_device *dev = chan->dev;
        struct drm_nouveau_private *dev_priv = dev->dev_private;
-       struct nouveau_gpuobj *ctx;
+       struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
        int ret;
 
-       /* Allocate a 175KiB block of PRAMIN to store the context.  This
-        * is massive overkill for a lot of chipsets, but it should be safe
-        * until we're able to implement this properly (will happen at more
-        * or less the same time we're able to write our own context programs.
-        */
-       ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, 175*1024, 16,
-                                         NVOBJ_FLAG_ZERO_ALLOC,
-                                         &chan->ramin_grctx);
+       ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, pgraph->grctx_size,
+                                    16, NVOBJ_FLAG_ZERO_ALLOC,
+                                    &chan->ramin_grctx);
        if (ret)
                return ret;
-       ctx = chan->ramin_grctx->gpuobj;
 
        /* Initialise default context values */
        dev_priv->engine.instmem.prepare_access(dev, true);
-       nv40_grctx_vals_load(dev, ctx);
-       nv_wo32(dev, ctx, 0, ctx->im_pramin->start);
-       dev_priv->engine.instmem.finish_access(dev);
+       if (!pgraph->ctxprog) {
+               struct nouveau_grctx ctx = {};
 
+               ctx.dev = chan->dev;
+               ctx.mode = NOUVEAU_GRCTX_VALS;
+               ctx.data = chan->ramin_grctx->gpuobj;
+               nv40_grctx_init(&ctx);
+       } else {
+               nouveau_grctx_vals_load(dev, chan->ramin_grctx->gpuobj);
+       }
+       nv_wo32(dev, chan->ramin_grctx->gpuobj, 0,
+                    chan->ramin_grctx->gpuobj->im_pramin->start);
+       dev_priv->engine.instmem.finish_access(dev);
        return 0;
 }
 
@@ -204,135 +181,6 @@ nv40_graph_unload_context(struct drm_device *dev)
        return ret;
 }
 
-struct nouveau_ctxprog {
-       uint32_t signature;
-       uint8_t  version;
-       uint16_t length;
-       uint32_t data[];
-} __attribute__ ((packed));
-
-struct nouveau_ctxvals {
-       uint32_t signature;
-       uint8_t  version;
-       uint32_t length;
-       struct {
-               uint32_t offset;
-               uint32_t value;
-       } data[];
-} __attribute__ ((packed));
-
-int
-nv40_grctx_init(struct drm_device *dev)
-{
-       struct drm_nouveau_private *dev_priv = dev->dev_private;
-       struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
-       const int chipset = dev_priv->chipset;
-       const struct firmware *fw;
-       const struct nouveau_ctxprog *cp;
-       const struct nouveau_ctxvals *cv;
-       char name[32];
-       int ret, i;
-
-       pgraph->accel_blocked = true;
-
-       if (!pgraph->ctxprog) {
-               sprintf(name, "nouveau/nv%02x.ctxprog", chipset);
-               ret = request_firmware(&fw, name, &dev->pdev->dev);
-               if (ret) {
-                       NV_ERROR(dev, "No ctxprog for NV%02x\n", chipset);
-                       return ret;
-               }
-
-               pgraph->ctxprog = kmalloc(fw->size, GFP_KERNEL);
-               if (!pgraph->ctxprog) {
-                       NV_ERROR(dev, "OOM copying ctxprog\n");
-                       release_firmware(fw);
-                       return -ENOMEM;
-               }
-               memcpy(pgraph->ctxprog, fw->data, fw->size);
-
-               cp = pgraph->ctxprog;
-               if (cp->signature != 0x5043564e || cp->version != 0 ||
-                   cp->length != ((fw->size - 7) / 4)) {
-                       NV_ERROR(dev, "ctxprog invalid\n");
-                       release_firmware(fw);
-                       nv40_grctx_fini(dev);
-                       return -EINVAL;
-               }
-               release_firmware(fw);
-       }
-
-       if (!pgraph->ctxvals) {
-               sprintf(name, "nouveau/nv%02x.ctxvals", chipset);
-               ret = request_firmware(&fw, name, &dev->pdev->dev);
-               if (ret) {
-                       NV_ERROR(dev, "No ctxvals for NV%02x\n", chipset);
-                       nv40_grctx_fini(dev);
-                       return ret;
-               }
-
-               pgraph->ctxvals = kmalloc(fw->size, GFP_KERNEL);
-               if (!pgraph->ctxprog) {
-                       NV_ERROR(dev, "OOM copying ctxprog\n");
-                       release_firmware(fw);
-                       nv40_grctx_fini(dev);
-                       return -ENOMEM;
-               }
-               memcpy(pgraph->ctxvals, fw->data, fw->size);
-
-               cv = (void *)pgraph->ctxvals;
-               if (cv->signature != 0x5643564e || cv->version != 0 ||
-                   cv->length != ((fw->size - 9) / 8)) {
-                       NV_ERROR(dev, "ctxvals invalid\n");
-                       release_firmware(fw);
-                       nv40_grctx_fini(dev);
-                       return -EINVAL;
-               }
-               release_firmware(fw);
-       }
-
-       cp = pgraph->ctxprog;
-
-       nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0);
-       for (i = 0; i < cp->length; i++)
-               nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA, cp->data[i]);
-
-       pgraph->accel_blocked = false;
-       return 0;
-}
-
-void
-nv40_grctx_fini(struct drm_device *dev)
-{
-       struct drm_nouveau_private *dev_priv = dev->dev_private;
-       struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
-
-       if (pgraph->ctxprog) {
-               kfree(pgraph->ctxprog);
-               pgraph->ctxprog = NULL;
-       }
-
-       if (pgraph->ctxvals) {
-               kfree(pgraph->ctxprog);
-               pgraph->ctxvals = NULL;
-       }
-}
-
-void
-nv40_grctx_vals_load(struct drm_device *dev, struct nouveau_gpuobj *ctx)
-{
-       struct drm_nouveau_private *dev_priv = dev->dev_private;
-       struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
-       struct nouveau_ctxvals *cv = pgraph->ctxvals;
-       int i;
-
-       if (!cv)
-               return;
-
-       for (i = 0; i < cv->length; i++)
-               nv_wo32(dev, ctx, cv->data[i].offset, cv->data[i].value);
-}
-
 /*
  * G70         0x47
  * G71         0x49
@@ -355,7 +203,26 @@ nv40_graph_init(struct drm_device *dev)
        nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) |
                         NV_PMC_ENABLE_PGRAPH);
 
-       nv40_grctx_init(dev);
+       if (nouveau_ctxfw) {
+               nouveau_grctx_prog_load(dev);
+               dev_priv->engine.graph.grctx_size = 175 * 1024;
+       }
+
+       if (!dev_priv->engine.graph.ctxprog) {
+               struct nouveau_grctx ctx = {};
+               uint32_t cp[256];
+
+               ctx.dev = dev;
+               ctx.mode = NOUVEAU_GRCTX_PROG;
+               ctx.data = cp;
+               ctx.ctxprog_max = 256;
+               nv40_grctx_init(&ctx);
+               dev_priv->engine.graph.grctx_size = ctx.ctxvals_pos * 4;
+
+               nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0);
+               for (i = 0; i < ctx.ctxprog_len; i++)
+                       nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA, cp[i]);
+       }
 
        /* No context present currently */
        nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, 0x00000000);
@@ -535,6 +402,7 @@ nv40_graph_init(struct drm_device *dev)
 
 void nv40_graph_takedown(struct drm_device *dev)
 {
+       nouveau_grctx_fini(dev);
 }
 
 struct nouveau_pgraph_object_class nv40_graph_grclass[] = {