gpu: nvgpu: Use device instead of platform_device

Use struct device instead of struct platform_device wherever possible. This allows adding other bus types later. Change-Id: I1657287a68d85a542cdbdd8a00d1902c3d6e00ed Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com> Reviewed-on: http://git-master/r/1120466
author: Terje Bergstrom <tbergstrom@nvidia.com> 2016-03-29 19:02:34 -0400
committer: Terje Bergstrom <tbergstrom@nvidia.com> 2016-04-08 12:42:41 -0400
commit: e8bac374c0ed24f05bf389e1e8b5aca47f61bd3a (patch)
tree: 36b6e111f8706c0560ee552bc9d6e15b87fc9621 /drivers
parent: 2382a8433fddaee3baecff6ae941944850787ab7 (diff)
37 files changed, 691 insertions, 736 deletions
diff --git a/drivers/gpu/nvgpu/gk20a/cde_gk20a.c b/drivers/gpu/nvgpu/gk20a/cde_gk20a.c
index a2f7e7a4..7818f046 100644
--- a/drivers/gpu/nvgpu/gk20a/cde_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/cde_gk20a.c
@@ -1,7 +1,7 @@
 /*
 * Color decompression engine support
 *
- * Copyright (c) 2014-2015, NVIDIA Corporation.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA Corporation.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -227,14 +227,14 @@ static int gk20a_init_cde_buf(struct gk20a_cde_ctx *cde_ctx,
        /* check that the file can hold the buf */
        if (buf->data_byte_offset != 0 &&
            buf->data_byte_offset + buf->num_bytes > img->size) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: invalid data section. buffer idx = %d",
+                gk20a_warn(cde_ctx->dev, "cde: invalid data section. buffer idx = %d",
                           cde_ctx->num_bufs);
                return -EINVAL;
        }
        /* check that we have enough buf elems available */
        if (cde_ctx->num_bufs >= MAX_CDE_BUFS) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: invalid data section. buffer idx = %d",
+                gk20a_warn(cde_ctx->dev, "cde: invalid data section. buffer idx = %d",
                           cde_ctx->num_bufs);
                return -ENOMEM;
        }
@@ -243,7 +243,7 @@ static int gk20a_init_cde_buf(struct gk20a_cde_ctx *cde_ctx,
        mem = cde_ctx->mem + cde_ctx->num_bufs;
        err = gk20a_gmmu_alloc_map(cde_ctx->vm, buf->num_bytes, mem);
        if (err) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: could not allocate device memory. buffer idx = %d",
+                gk20a_warn(cde_ctx->dev, "cde: could not allocate device memory. buffer idx = %d",
                           cde_ctx->num_bufs);
                return -ENOMEM;
        }
@@ -281,7 +281,7 @@ static int gk20a_replace_data(struct gk20a_cde_ctx *cde_ctx, void *target,
                current_value = (u64)(current_value >> 32) |
                        (u64)(current_value << 32);
        } else {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: unknown type. type=%d",
+                gk20a_warn(cde_ctx->dev, "cde: unknown type. type=%d",
                           type);
                return -EINVAL;
        }
@@ -315,7 +315,7 @@ static int gk20a_init_cde_replace(struct gk20a_cde_ctx *cde_ctx,
        if (replace->target_buf >= cde_ctx->num_bufs ||
            replace->source_buf >= cde_ctx->num_bufs) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: invalid buffer. target_buf=%u, source_buf=%u, num_bufs=%d",
+                gk20a_warn(cde_ctx->dev, "cde: invalid buffer. target_buf=%u, source_buf=%u, num_bufs=%d",
                           replace->target_buf, replace->source_buf,
                           cde_ctx->num_bufs);
                return -EINVAL;
@@ -327,7 +327,7 @@ static int gk20a_init_cde_replace(struct gk20a_cde_ctx *cde_ctx,
        if (source_mem->size < (replace->source_byte_offset + 3) ||
            target_mem->size < (replace->target_byte_offset + 3)) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: invalid buffer offsets. target_buf_offs=%lld, source_buf_offs=%lld, source_buf_size=%zu, dest_buf_size=%zu",
+                gk20a_warn(cde_ctx->dev, "cde: invalid buffer offsets. target_buf_offs=%lld, source_buf_offs=%lld, source_buf_size=%zu, dest_buf_size=%zu",
                           replace->target_byte_offset,
                           replace->source_byte_offset,
                         source_mem->size,
@@ -344,7 +344,7 @@ static int gk20a_init_cde_replace(struct gk20a_cde_ctx *cde_ctx,
                                 replace->shift, replace->mask,
                                 vaddr);
        if (err) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: replace failed. err=%d, target_buf=%u, target_buf_offs=%lld, source_buf=%u, source_buf_offs=%lld",
+                gk20a_warn(cde_ctx->dev, "cde: replace failed. err=%d, target_buf=%u, target_buf_offs=%lld, source_buf=%u, source_buf_offs=%lld",
                           err, replace->target_buf,
                           replace->target_byte_offset,
                           replace->source_buf,
@@ -431,7 +431,7 @@ static int gk20a_cde_patch_params(struct gk20a_cde_ctx *cde_ctx)
                                         param->shift, param->mask, new_data);
                if (err) {
-                        gk20a_warn(&cde_ctx->pdev->dev, "cde: patch failed. err=%d, idx=%d, id=%d, target_buf=%u, target_buf_offs=%lld, patch_value=%llu",
+                        gk20a_warn(cde_ctx->dev, "cde: patch failed. err=%d, idx=%d, id=%d, target_buf=%u, target_buf_offs=%lld, patch_value=%llu",
                                   err, i, param->id, param->target_buf,
                                   param->target_byte_offset, new_data);
                        return err;
@@ -448,7 +448,7 @@ static int gk20a_init_cde_param(struct gk20a_cde_ctx *cde_ctx,
        struct mem_desc *target_mem;
        if (param->target_buf >= cde_ctx->num_bufs) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: invalid buffer parameter. param idx = %d, target_buf=%u, num_bufs=%u",
+                gk20a_warn(cde_ctx->dev, "cde: invalid buffer parameter. param idx = %d, target_buf=%u, num_bufs=%u",
                           cde_ctx->num_params, param->target_buf,
                           cde_ctx->num_bufs);
                return -EINVAL;
@@ -456,7 +456,7 @@ static int gk20a_init_cde_param(struct gk20a_cde_ctx *cde_ctx,
        target_mem = cde_ctx->mem + param->target_buf;
        if (target_mem->size< (param->target_byte_offset + 3)) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: invalid buffer parameter. param idx = %d, target_buf_offs=%lld, target_buf_size=%zu",
+                gk20a_warn(cde_ctx->dev, "cde: invalid buffer parameter. param idx = %d, target_buf_offs=%lld, target_buf_size=%zu",
                           cde_ctx->num_params, param->target_byte_offset,
                           target_mem->size);
                return -EINVAL;
@@ -464,14 +464,14 @@ static int gk20a_init_cde_param(struct gk20a_cde_ctx *cde_ctx,
        /* does this parameter fit into our parameter structure */
        if (cde_ctx->num_params >= MAX_CDE_PARAMS) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: no room for new parameters param idx = %d",
+                gk20a_warn(cde_ctx->dev, "cde: no room for new parameters param idx = %d",
                           cde_ctx->num_params);
                return -ENOMEM;
        }
        /* is the given id valid? */
        if (param->id >= NUM_RESERVED_PARAMS + MAX_CDE_USER_PARAMS) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: parameter id is not valid. param idx = %d, id=%u, max=%u",
+                gk20a_warn(cde_ctx->dev, "cde: parameter id is not valid. param idx = %d, id=%u, max=%u",
                           param->id, cde_ctx->num_params,
                           NUM_RESERVED_PARAMS + MAX_CDE_USER_PARAMS);
                return -EINVAL;
@@ -498,7 +498,7 @@ static int gk20a_init_cde_required_class(struct gk20a_cde_ctx *cde_ctx,
        err = gk20a_alloc_obj_ctx(cde_ctx->ch, &alloc_obj_ctx);
        if (err) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: failed to allocate ctx. err=%d",
+                gk20a_warn(cde_ctx->dev, "cde: failed to allocate ctx. err=%d",
                           err);
                return err;
        }
@@ -524,7 +524,7 @@ static int gk20a_init_cde_command(struct gk20a_cde_ctx *cde_ctx,
                gpfifo = &cde_ctx->convert_cmd;
                num_entries = &cde_ctx->convert_cmd_num_entries;
        } else {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: unknown command. op=%u",
+                gk20a_warn(cde_ctx->dev, "cde: unknown command. op=%u",
                           op);
                return -EINVAL;
        }
@@ -533,7 +533,7 @@ static int gk20a_init_cde_command(struct gk20a_cde_ctx *cde_ctx,
        *gpfifo = kzalloc(sizeof(struct nvgpu_gpfifo) * num_elems,
                          GFP_KERNEL);
        if (!*gpfifo) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: could not allocate memory for gpfifo entries");
+                gk20a_warn(cde_ctx->dev, "cde: could not allocate memory for gpfifo entries");
                return -ENOMEM;
        }
@@ -543,7 +543,7 @@ static int gk20a_init_cde_command(struct gk20a_cde_ctx *cde_ctx,
                /* validate the current entry */
                if (cmd_elem->target_buf >= cde_ctx->num_bufs) {
-                        gk20a_warn(&cde_ctx->pdev->dev, "cde: target buffer is not available (target=%u, num_bufs=%u)",
+                        gk20a_warn(cde_ctx->dev, "cde: target buffer is not available (target=%u, num_bufs=%u)",
                                   cmd_elem->target_buf, cde_ctx->num_bufs);
                        return -EINVAL;
                }
@@ -551,7 +551,7 @@ static int gk20a_init_cde_command(struct gk20a_cde_ctx *cde_ctx,
                target_mem = cde_ctx->mem + cmd_elem->target_buf;
                if (target_mem->size<
                    cmd_elem->target_byte_offset + cmd_elem->num_bytes) {
-                        gk20a_warn(&cde_ctx->pdev->dev, "cde: target buffer cannot hold all entries (target_size=%zu, target_byte_offset=%lld, num_bytes=%llu)",
+                        gk20a_warn(cde_ctx->dev, "cde: target buffer cannot hold all entries (target_size=%zu, target_byte_offset=%lld, num_bytes=%llu)",
                                   target_mem->size,
                                   cmd_elem->target_byte_offset,
                                   cmd_elem->num_bytes);
@@ -585,7 +585,7 @@ static int gk20a_cde_pack_cmdbufs(struct gk20a_cde_ctx *cde_ctx)
        /* allocate buffer that has space for both */
        combined_cmd = kzalloc(total_bytes, GFP_KERNEL);
        if (!combined_cmd) {
-                gk20a_warn(&cde_ctx->pdev->dev,
+                gk20a_warn(cde_ctx->dev,
                                "cde: could not allocate memory for gpfifo entries");
                return -ENOMEM;
        }
@@ -618,7 +618,7 @@ static int gk20a_init_cde_img(struct gk20a_cde_ctx *cde_ctx,
        min_size += 2 * sizeof(u32);
        if (img->size < min_size) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: invalid image header");
+                gk20a_warn(cde_ctx->dev, "cde: invalid image header");
                return -EINVAL;
        }
@@ -627,7 +627,7 @@ static int gk20a_init_cde_img(struct gk20a_cde_ctx *cde_ctx,
        min_size += num_of_elems * sizeof(*elem);
        if (img->size < min_size) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: bad image");
+                gk20a_warn(cde_ctx->dev, "cde: bad image");
                return -EINVAL;
        }
@@ -664,7 +664,7 @@ static int gk20a_init_cde_img(struct gk20a_cde_ctx *cde_ctx,
                                MAX_CDE_ARRAY_ENTRIES*sizeof(u32));
                        break;
                default:
-                        gk20a_warn(&cde_ctx->pdev->dev, "cde: unknown header element");
+                        gk20a_warn(cde_ctx->dev, "cde: unknown header element");
                        err = -EINVAL;
                }
@@ -675,13 +675,13 @@ static int gk20a_init_cde_img(struct gk20a_cde_ctx *cde_ctx,
        }
        if (!cde_ctx->init_convert_cmd || !cde_ctx->init_cmd_num_entries) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: convert command not defined");
+                gk20a_warn(cde_ctx->dev, "cde: convert command not defined");
                err = -EINVAL;
                goto deinit_image;
        }
        if (!cde_ctx->convert_cmd || !cde_ctx->convert_cmd_num_entries) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: convert command not defined");
+                gk20a_warn(cde_ctx->dev, "cde: convert command not defined");
                err = -EINVAL;
                goto deinit_image;
        }
@@ -714,12 +714,12 @@ static int gk20a_cde_execute_buffer(struct gk20a_cde_ctx *cde_ctx,
                gpfifo = cde_ctx->convert_cmd;
                num_entries = cde_ctx->convert_cmd_num_entries;
        } else {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: unknown buffer");
+                gk20a_warn(cde_ctx->dev, "cde: unknown buffer");
                return -EINVAL;
        }
        if (gpfifo == NULL || num_entries == 0) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: buffer not available");
+                gk20a_warn(cde_ctx->dev, "cde: buffer not available");
                return -ENOSYS;
        }
@@ -757,7 +757,7 @@ __releases(&cde_app->mutex)
        struct gk20a_cde_ctx *cde_ctx = container_of(delay_work,
                        struct gk20a_cde_ctx, ctx_deleter_work);
        struct gk20a_cde_app *cde_app = &cde_ctx->g->cde_app;
-        struct platform_device *pdev = cde_ctx->pdev;
+        struct device *dev = cde_ctx->dev;
        int err;
        /* someone has just taken it? engine deletion started? */
@@ -767,11 +767,11 @@ __releases(&cde_app->mutex)
        gk20a_dbg(gpu_dbg_fn | gpu_dbg_cde_ctx,
                        "cde: attempting to delete temporary %p", cde_ctx);
-        err = gk20a_busy(pdev);
+        err = gk20a_busy(dev);
        if (err) {
                /* this context would find new use anyway later, so not freeing
                 * here does not leak anything */
-                gk20a_warn(&pdev->dev, "cde: cannot set gk20a on, postponing"
+                gk20a_warn(dev, "cde: cannot set gk20a on, postponing"
                                " temp ctx deletion");
                return;
        }
@@ -795,7 +795,7 @@ __releases(&cde_app->mutex)
 out:
        mutex_unlock(&cde_app->mutex);
-        gk20a_idle(pdev);
+        gk20a_idle(dev);
 }
 static struct gk20a_cde_ctx *gk20a_cde_do_get_context(struct gk20a *g)
@@ -839,7 +839,7 @@ __must_hold(&cde_app->mutex)
        cde_ctx = gk20a_cde_allocate_context(g);
        if (IS_ERR(cde_ctx)) {
-                gk20a_warn(&g->dev->dev, "cde: cannot allocate context: %ld",
+                gk20a_warn(g->dev, "cde: cannot allocate context: %ld",
                                PTR_ERR(cde_ctx));
                return cde_ctx;
        }
@@ -888,7 +888,7 @@ static struct gk20a_cde_ctx *gk20a_cde_allocate_context(struct gk20a *g)
                return ERR_PTR(-ENOMEM);
        cde_ctx->g = g;
-        cde_ctx->pdev = g->dev;
+        cde_ctx->dev = g->dev;
        ret = gk20a_cde_load(cde_ctx);
        if (ret) {
@@ -951,7 +951,7 @@ __releases(&cde_app->mutex)
        /* First, map the buffer to local va */
        /* ensure that the compbits buffer has drvdata */
-        err = gk20a_dmabuf_alloc_drvdata(compbits_scatter_buf, &g->dev->dev);
+        err = gk20a_dmabuf_alloc_drvdata(compbits_scatter_buf, g->dev);
        if (err)
                goto exit_unlock;
@@ -1007,7 +1007,7 @@ __releases(&cde_app->mutex)
                surface = dma_buf_vmap(compbits_scatter_buf);
                if (IS_ERR(surface)) {
-                        gk20a_warn(&g->dev->dev,
+                        gk20a_warn(g->dev,
                                   "dma_buf_vmap failed");
                        err = -EINVAL;
                        goto exit_unlock;
@@ -1017,9 +1017,9 @@ __releases(&cde_app->mutex)
                gk20a_dbg(gpu_dbg_cde, "surface=0x%p scatterBuffer=0x%p",
                          surface, scatter_buffer);
-                sgt = gk20a_mm_pin(&g->dev->dev, compbits_scatter_buf);
+                sgt = gk20a_mm_pin(g->dev, compbits_scatter_buf);
                if (IS_ERR(sgt)) {
-                        gk20a_warn(&g->dev->dev,
+                        gk20a_warn(g->dev,
                                   "mm_pin failed");
                        err = -EINVAL;
                        goto exit_unlock;
@@ -1029,7 +1029,7 @@ __releases(&cde_app->mutex)
                                        scatterbuffer_size);
                        WARN_ON(err);
-                        gk20a_mm_unpin(&g->dev->dev, compbits_scatter_buf,
+                        gk20a_mm_unpin(g->dev, compbits_scatter_buf,
                                       sgt);
                        if (err)
                                goto exit_unlock;
@@ -1041,7 +1041,7 @@ __releases(&cde_app->mutex)
        }
        /* store source buffer compression tags */
-        gk20a_get_comptags(&g->dev->dev, compbits_scatter_buf, &comptags);
+        gk20a_get_comptags(g->dev, compbits_scatter_buf, &comptags);
        cde_ctx->surf_param_offset = comptags.offset;
        cde_ctx->surf_param_lines = comptags.lines;
@@ -1067,7 +1067,7 @@ __releases(&cde_app->mutex)
                int id = param->id - NUM_RESERVED_PARAMS;
                if (id < 0 || id >= MAX_CDE_USER_PARAMS) {
-                        gk20a_warn(&cde_ctx->pdev->dev, "cde: unknown user parameter");
+                        gk20a_warn(cde_ctx->dev, "cde: unknown user parameter");
                        err = -EINVAL;
                        goto exit_unlock;
                }
@@ -1077,7 +1077,7 @@ __releases(&cde_app->mutex)
        /* patch data */
        err = gk20a_cde_patch_params(cde_ctx);
        if (err) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: failed to patch parameters");
+                gk20a_warn(cde_ctx->dev, "cde: failed to patch parameters");
                goto exit_unlock;
        }
@@ -1140,19 +1140,19 @@ __releases(&cde_app->mutex)
        if (ch->has_timedout) {
                if (cde_ctx->is_temporary) {
-                        gk20a_warn(&cde_ctx->pdev->dev,
+                        gk20a_warn(cde_ctx->dev,
                                        "cde: channel had timed out"
                                        " (temporary channel)");
                        /* going to be deleted anyway */
                } else {
-                        gk20a_warn(&cde_ctx->pdev->dev,
+                        gk20a_warn(cde_ctx->dev,
                                        "cde: channel had timed out"
                                        ", reloading");
                        /* mark it to be deleted, replace with a new one */
                        mutex_lock(&cde_app->mutex);
                        cde_ctx->is_temporary = true;
                        if (gk20a_cde_create_context(g)) {
-                                gk20a_err(&cde_ctx->pdev->dev,
+                                gk20a_err(cde_ctx->dev,
                                                "cde: can't replace context");
                        }
                        mutex_unlock(&cde_app->mutex);
@@ -1181,14 +1181,14 @@ static int gk20a_cde_load(struct gk20a_cde_ctx *cde_ctx)
        img = gk20a_request_firmware(g, "gpu2cde.bin");
        if (!img) {
-                dev_err(&cde_ctx->pdev->dev, "cde: could not fetch the firmware");
+                dev_err(cde_ctx->dev, "cde: could not fetch the firmware");
                return -ENOSYS;
        }
        ch = gk20a_open_new_channel_with_cb(g, gk20a_cde_finished_ctx_cb,
                        cde_ctx);
        if (!ch) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: gk20a channel not available");
+                gk20a_warn(cde_ctx->dev, "cde: gk20a channel not available");
                err = -ENOMEM;
                goto err_get_gk20a_channel;
        }
@@ -1198,7 +1198,7 @@ static int gk20a_cde_load(struct gk20a_cde_ctx *cde_ctx)
        ch->vm = &g->mm.cde.vm;
        err = channel_gk20a_commit_va(ch);
        if (err) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: could not bind vm");
+                gk20a_warn(cde_ctx->dev, "cde: could not bind vm");
                goto err_commit_va;
        }
@@ -1206,7 +1206,7 @@ static int gk20a_cde_load(struct gk20a_cde_ctx *cde_ctx)
        err = gk20a_alloc_channel_gpfifo(ch,
                &(struct nvgpu_alloc_gpfifo_args){1024, 0});
        if (err) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: unable to allocate gpfifo");
+                gk20a_warn(cde_ctx->dev, "cde: unable to allocate gpfifo");
                goto err_alloc_gpfifo;
        }
@@ -1218,7 +1218,7 @@ static int gk20a_cde_load(struct gk20a_cde_ctx *cde_ctx)
                               false);
        if (!vaddr) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: cannot map compression bit backing store");
+                gk20a_warn(cde_ctx->dev, "cde: cannot map compression bit backing store");
                err = -ENOMEM;
                goto err_map_backingstore;
        }
@@ -1231,7 +1231,7 @@ static int gk20a_cde_load(struct gk20a_cde_ctx *cde_ctx)
        /* initialise the firmware */
        err = gk20a_init_cde_img(cde_ctx, img);
        if (err) {
-                gk20a_warn(&cde_ctx->pdev->dev, "cde: image initialisation failed");
+                gk20a_warn(cde_ctx->dev, "cde: image initialisation failed");
                goto err_init_cde_img;
        }
@@ -1248,7 +1248,7 @@ err_alloc_gpfifo:
 err_commit_va:
 err_get_gk20a_channel:
        release_firmware(img);
-        dev_err(&cde_ctx->pdev->dev, "cde: couldn't initialise buffer converter: %d",
+        dev_err(cde_ctx->dev, "cde: couldn't initialise buffer converter: %d",
                err);
        return err;
 }
@@ -1386,17 +1386,17 @@ static int gk20a_buffer_convert_gpu_to_cde_v1(
                g->ops.cde.get_program_numbers(g, block_height_log2,
                                               &hprog, &vprog);
        else {
-                gk20a_warn(&g->dev->dev, "cde: chip not supported");
+                gk20a_warn(g->dev, "cde: chip not supported");
                return -ENOSYS;
        }
        if (hprog < 0 || vprog < 0) {
-                gk20a_warn(&g->dev->dev, "cde: could not determine programs");
+                gk20a_warn(g->dev, "cde: could not determine programs");
                return -ENOSYS;
        }
        if (xtiles > 8192 / 8 || ytiles > 8192 / 8)
-                gk20a_warn(&g->dev->dev, "cde: surface is exceptionally large (xtiles=%d, ytiles=%d)",
+                gk20a_warn(g->dev, "cde: surface is exceptionally large (xtiles=%d, ytiles=%d)",
                           xtiles, ytiles);
        gk20a_dbg(gpu_dbg_cde, "w=%d, h=%d, bh_log2=%d, compbits_hoffset=0x%llx, compbits_voffset=0x%llx, scatterbuffer_offset=0x%llx",
@@ -1645,9 +1645,9 @@ static const struct file_operations gk20a_cde_reload_fops = {
        .write          = gk20a_cde_reload_write,
 };
-void gk20a_cde_debugfs_init(struct platform_device *dev)
+void gk20a_cde_debugfs_init(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct gk20a *g = get_gk20a(dev);
        debugfs_create_u32("cde_parameter", S_IWUSR | S_IRUGO,
diff --git a/drivers/gpu/nvgpu/gk20a/cde_gk20a.h b/drivers/gpu/nvgpu/gk20a/cde_gk20a.h
index fc5736ad..16d6b4ef 100644
--- a/drivers/gpu/nvgpu/gk20a/cde_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/cde_gk20a.h
@@ -1,7 +1,7 @@
 /*
 * GK20A color decompression engine support
 *
- * Copyright (c) 2014-2015, NVIDIA Corporation.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA Corporation.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -215,7 +215,7 @@ struct gk20a_cde_param {
 struct gk20a_cde_ctx {
        struct gk20a *g;
-        struct platform_device *pdev;
+        struct device *dev;
        /* channel related data */
        struct channel_gk20a *ch;
@@ -288,7 +288,7 @@ int gk20a_cde_convert(struct gk20a *g,
                struct nvgpu_fence *fence,
                u32 __flags, struct gk20a_cde_param *params,
                int num_params, struct gk20a_fence **fence_out);
-void gk20a_cde_debugfs_init(struct platform_device *dev);
+void gk20a_cde_debugfs_init(struct device *dev);
 int gk20a_prepare_compressible_read(
                struct gk20a *g, u32 buffer_fd, u32 request, u64 offset,
diff --git a/drivers/gpu/nvgpu/gk20a/channel_gk20a.c b/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
index 58175141..61211239 100644
--- a/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
@@ -74,7 +74,9 @@ static void gk20a_channel_clean_up_jobs(struct work_struct *work);
 static struct channel_gk20a *allocate_channel(struct fifo_gk20a *f)
 {
        struct channel_gk20a *ch = NULL;
-        struct gk20a_platform *platform = gk20a_get_platform(f->g->dev);
+        struct gk20a_platform *platform;
+        platform = gk20a_get_platform(f->g->dev);
        mutex_lock(&f->free_chs_mutex);
        if (!list_empty(&f->free_chs)) {
@@ -156,7 +158,7 @@ int gk20a_channel_get_timescale_from_timeslice(struct gk20a *g,
                int timeslice_period,
                int *__timeslice_timeout, int *__timeslice_scale)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        int value = scale_ptimer(timeslice_period,
                        ptimer_scalingfactor10x(platform->ptimer_src_freq));
        int shift = 0;
@@ -1068,7 +1070,7 @@ int gk20a_channel_release(struct inode *inode, struct file *filp)
        if (!ch)
                return 0;
-        trace_gk20a_channel_release(dev_name(&g->dev->dev));
+        trace_gk20a_channel_release(dev_name(g->dev));
        err = gk20a_busy(g->dev);
        if (err) {
@@ -1189,7 +1191,9 @@ static int __gk20a_channel_open(struct gk20a *g, struct file *filp)
        int err;
        struct channel_gk20a *ch;
-        trace_gk20a_channel_open(dev_name(&g->dev->dev));
+        gk20a_dbg_fn("");
+        trace_gk20a_channel_open(dev_name(g->dev));
        err = gk20a_busy(g->dev);
        if (err) {
@@ -1235,7 +1239,7 @@ int gk20a_channel_open_ioctl(struct gk20a *g,
        fd = err;
        name = kasprintf(GFP_KERNEL, "nvhost-%s-fd%d",
-                        dev_name(&g->dev->dev), fd);
+                        dev_name(g->dev), fd);
        if (!name) {
                err = -ENOMEM;
                goto clean_up;
@@ -1562,7 +1566,7 @@ static void trace_write_pushbuffer(struct channel_gk20a *c,
                 */
                for (i = 0; i < words; i += 128U) {
                        trace_gk20a_push_cmdbuf(
-                                c->g->dev->name,
+                                dev_name(c->g->dev),
                                0,
                                min(words - i, 128U),
                                offset + i * sizeof(u32),
@@ -2051,7 +2055,7 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
                return err;
        }
-        trace_gk20a_channel_submit_gpfifo(c->g->dev->name,
+        trace_gk20a_channel_submit_gpfifo(dev_name(c->g->dev),
                                          c->hw_chid,
                                          num_entries,
                                          flags,
@@ -2069,11 +2073,11 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
                if (locked_path)
                        mutex_unlock(&c->ioctl_lock);
-                trace_gk20a_gpfifo_submit_wait_for_space(c->g->dev->name);
+                trace_gk20a_gpfifo_submit_wait_for_space(dev_name(c->g->dev));
                err = wait_event_interruptible(c->submit_wq,
                        get_gp_free_count(c) >= num_entries + extra_entries ||
                        c->has_timedout);
-                trace_gk20a_gpfifo_submit_wait_for_space_done(c->g->dev->name);
+                trace_gk20a_gpfifo_submit_wait_for_space_done(dev_name(c->g->dev));
                if (locked_path)
                        mutex_lock(&c->ioctl_lock);
@@ -2156,7 +2160,7 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
                gpfifo_mem[c->gpfifo.put].entry0 = u64_lo32(wait_cmd->gva);
                gpfifo_mem[c->gpfifo.put].entry1 = u64_hi32(wait_cmd->gva) |
                        pbdma_gp_entry1_length_f(wait_cmd->size);
-                trace_gk20a_push_cmdbuf(c->g->dev->name,
+                trace_gk20a_push_cmdbuf(dev_name(c->g->dev),
                        0, wait_cmd->size, 0, wait_cmd->ptr);
                c->gpfifo.put = (c->gpfifo.put + 1) &
@@ -2244,7 +2248,7 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
                gpfifo_mem[c->gpfifo.put].entry0 = u64_lo32(incr_cmd->gva);
                gpfifo_mem[c->gpfifo.put].entry1 = u64_hi32(incr_cmd->gva) |
                        pbdma_gp_entry1_length_f(incr_cmd->size);
-                trace_gk20a_push_cmdbuf(c->g->dev->name,
+                trace_gk20a_push_cmdbuf(dev_name(c->g->dev),
                        0, incr_cmd->size, 0, incr_cmd->ptr);
                c->gpfifo.put = (c->gpfifo.put + 1) &
@@ -2273,7 +2277,7 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
        mutex_unlock(&c->submit_lock);
-        trace_gk20a_channel_submitted_gpfifo(c->g->dev->name,
+        trace_gk20a_channel_submitted_gpfifo(dev_name(c->g->dev),
                                             c->hw_chid,
                                             num_entries,
                                             flags,
@@ -2357,7 +2361,7 @@ static int gk20a_channel_wait_semaphore(struct channel_gk20a *ch,
                                        ulong id, u32 offset,
                                        u32 payload, long timeout)
 {
-        struct platform_device *pdev = ch->g->dev;
+        struct device *dev = ch->g->dev;
        struct dma_buf *dmabuf;
        void *data;
        u32 *semaphore;
@@ -2370,14 +2374,13 @@ static int gk20a_channel_wait_semaphore(struct channel_gk20a *ch,
        dmabuf = dma_buf_get(id);
        if (IS_ERR(dmabuf)) {
-                gk20a_err(&pdev->dev, "invalid notifier nvmap handle 0x%lx",
+                gk20a_err(dev, "invalid notifier nvmap handle 0x%lx", id);
-                           id);
                return -EINVAL;
        }
        data = dma_buf_kmap(dmabuf, offset >> PAGE_SHIFT);
        if (!data) {
-                gk20a_err(&pdev->dev, "failed to map notifier memory");
+                gk20a_err(dev, "failed to map notifier memory");
                ret = -EINVAL;
                goto cleanup_put;
        }
@@ -2917,7 +2920,7 @@ long gk20a_channel_ioctl(struct file *filp,
        unsigned int cmd, unsigned long arg)
 {
        struct channel_gk20a *ch = filp->private_data;
-        struct platform_device *dev = ch->g->dev;
+        struct device *dev = ch->g->dev;
        u8 buf[NVGPU_IOCTL_CHANNEL_MAX_ARG_SIZE];
        int err = 0;
@@ -2956,7 +2959,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_ALLOC_OBJ_CTX:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -2968,7 +2971,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_FREE_OBJ_CTX:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -2980,7 +2983,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_ALLOC_GPFIFO:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -2996,7 +2999,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_WAIT:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3016,7 +3019,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_ZCULL_BIND:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3028,7 +3031,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_SET_ERROR_NOTIFIER:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3041,7 +3044,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_CYCLE_STATS:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3080,7 +3083,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_SET_PRIORITY:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3092,7 +3095,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_ENABLE:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3106,7 +3109,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_DISABLE:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3120,7 +3123,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_PREEMPT:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3131,7 +3134,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_FORCE_RESET:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3147,7 +3150,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_CYCLE_STATS_SNAPSHOT:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3164,7 +3167,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_SET_RUNLIST_INTERLEAVE:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3176,7 +3179,7 @@ long gk20a_channel_ioctl(struct file *filp,
        case NVGPU_IOCTL_CHANNEL_SET_TIMESLICE:
                err = gk20a_busy(dev);
                if (err) {
-                        dev_err(&dev->dev,
+                        dev_err(dev,
                                "%s: failed to host gk20a for ioctl cmd: 0x%x",
                                __func__, cmd);
                        break;
@@ -3186,7 +3189,7 @@ long gk20a_channel_ioctl(struct file *filp,
                gk20a_idle(dev);
                break;
        default:
-                dev_dbg(&dev->dev, "unrecognized ioctl cmd: 0x%x", cmd);
+                dev_dbg(dev, "unrecognized ioctl cmd: 0x%x", cmd);
                err = -ENOTTY;
                break;
        }
diff --git a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
index cdf36454..30bb6efe 100644
--- a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
@@ -330,13 +330,13 @@ gk20a_channel_syncpt_create(struct channel_gk20a *c)
        sp->host1x_pdev = c->g->host1x_dev;
        snprintf(syncpt_name, sizeof(syncpt_name),
-                "%s_%d", dev_name(&c->g->dev->dev), c->hw_chid);
+                "%s_%d", dev_name(c->g->dev), c->hw_chid);
        sp->id = nvhost_get_syncpt_host_managed(sp->host1x_pdev,
                                                c->hw_chid, syncpt_name);
        if (!sp->id) {
                kfree(sp);
-                gk20a_err(&c->g->dev->dev, "failed to get free syncpt");
+                gk20a_err(c->g->dev, "failed to get free syncpt");
                return NULL;
        }
@@ -387,7 +387,7 @@ static void gk20a_channel_semaphore_launcher(
                        fence, fence->name);
        err = sync_fence_wait(fence, -1);
        if (err < 0)
-                dev_err(&g->dev->dev, "error waiting pre-fence: %d\n", err);
+                dev_err(g->dev, "error waiting pre-fence: %d\n", err);
        gk20a_dbg_info(
                  "wait completed (%d) for fence %p '%s', triggering gpu work",
diff --git a/drivers/gpu/nvgpu/gk20a/clk_gk20a.c b/drivers/gpu/nvgpu/gk20a/clk_gk20a.c
index e10df6ac..26fd1eaf 100644
--- a/drivers/gpu/nvgpu/gk20a/clk_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/clk_gk20a.c
@@ -1,9 +1,7 @@
 /*
- * drivers/video/tegra/host/gk20a/clk_gk20a.c
- *
 * GK20A Clocks
 *
- * Copyright (c) 2011-2014, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2011-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -831,7 +829,7 @@ DEFINE_SIMPLE_ATTRIBUTE(monitor_fops, monitor_get, NULL, "%llu\n");
 static int clk_gk20a_debugfs_init(struct gk20a *g)
 {
        struct dentry *d;
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        d = debugfs_create_file(
                "rate", S_IRUGO|S_IWUSR, platform->debugfs, g, &rate_fops);
diff --git a/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c b/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
index b2ae224f..87f0683f 100644
--- a/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
@@ -144,7 +144,7 @@ static int gk20a_ctrl_alloc_as(
                struct gk20a *g,
                struct nvgpu_alloc_as_args *args)
 {
-        struct platform_device *dev = g->dev;
+        struct device *dev = g->dev;
        struct gk20a_as_share *as_share;
        int err;
        int fd;
@@ -157,7 +157,7 @@ static int gk20a_ctrl_alloc_as(
        fd = err;
        name = kasprintf(GFP_KERNEL, "nvhost-%s-fd%d",
-                        dev_name(&dev->dev), fd);
+                        dev_name(dev), fd);
        file = anon_inode_getfile(name, g->as.cdev.ops, NULL, O_RDWR);
        kfree(name);
@@ -187,7 +187,7 @@ clean_up:
 static int gk20a_ctrl_open_tsg(struct gk20a *g,
                               struct nvgpu_gpu_open_tsg_args *args)
 {
-        struct platform_device *dev = g->dev;
+        struct device *dev = g->dev;
        int err;
        int fd;
        struct file *file;
@@ -199,7 +199,7 @@ static int gk20a_ctrl_open_tsg(struct gk20a *g,
        fd = err;
        name = kasprintf(GFP_KERNEL, "nvgpu-%s-tsg%d",
-                         dev_name(&dev->dev), fd);
+                         dev_name(dev), fd);
        file = anon_inode_getfile(name, g->tsg.cdev.ops, NULL, O_RDWR);
        kfree(name);
@@ -621,7 +621,7 @@ static int nvgpu_gpu_get_cpu_time_correlation_info(
 long gk20a_ctrl_dev_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 {
-        struct platform_device *dev = filp->private_data;
+        struct device *dev = filp->private_data;
        struct gk20a *g = get_gk20a(dev);
        struct nvgpu_gpu_zcull_get_ctx_size_args *get_ctx_size_args;
        struct nvgpu_gpu_zcull_get_info_args *get_info_args;
diff --git a/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.c b/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.c
index d087d89e..321cebb2 100644
--- a/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.c
@@ -65,7 +65,6 @@ static int gk20a_dbg_gpu_do_dev_open(struct inode *inode,
        struct dbg_session_gk20a *dbg_session;
        struct gk20a *g;
-        struct platform_device *pdev;
        struct device *dev;
        int err;
@@ -76,8 +75,7 @@ static int gk20a_dbg_gpu_do_dev_open(struct inode *inode,
        else
                g = container_of(inode->i_cdev,
                                 struct gk20a, prof.cdev);
-        pdev = g->dev;
+        dev = g->dev;
-        dev  = &pdev->dev;
        gk20a_dbg(gpu_dbg_fn | gpu_dbg_gpu_dbg, "dbg session: %s", dev_name(dev));
@@ -86,7 +84,6 @@ static int gk20a_dbg_gpu_do_dev_open(struct inode *inode,
                return err;
        filp->private_data = dbg_session;
-        dbg_session->pdev  = pdev;
        dbg_session->dev   = dev;
        dbg_session->g     = g;
        dbg_session->is_profiler = is_profiler;
@@ -95,7 +92,7 @@ static int gk20a_dbg_gpu_do_dev_open(struct inode *inode,
        /* For vgpu, all power-gating features are currently disabled
         * in the server. Set is_pg_disable to true to reflect this
         * on the client side. */
-        if (gk20a_gpu_is_virtual(pdev))
+        if (gk20a_gpu_is_virtual(dev))
                dbg_session->is_pg_disabled = true;
        INIT_LIST_HEAD(&dbg_session->dbg_s_list_node);
@@ -486,7 +483,7 @@ static int nvgpu_dbg_gpu_ioctl_timeout(struct dbg_session_gk20a *dbg_s,
                         struct nvgpu_dbg_gpu_timeout_args *args)
 {
        int err;
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
        gk20a_dbg_fn("powergate mode = %d", args->enable);
@@ -501,7 +498,7 @@ static void nvgpu_dbg_gpu_ioctl_get_timeout(struct dbg_session_gk20a *dbg_s,
                         struct nvgpu_dbg_gpu_timeout_args *args)
 {
        int status;
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
        mutex_lock(&g->dbg_sessions_lock);
        status = g->timeouts_enabled;
@@ -532,7 +529,7 @@ long gk20a_dbg_gpu_dev_ioctl(struct file *filp, unsigned int cmd,
                             unsigned long arg)
 {
        struct dbg_session_gk20a *dbg_s = filp->private_data;
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
        u8 buf[NVGPU_DBG_GPU_IOCTL_MAX_ARG_SIZE];
        int err = 0;
@@ -672,7 +669,7 @@ static int nvgpu_ioctl_channel_reg_ops(struct dbg_session_gk20a *dbg_s,
        bool is_pg_disabled = false;
        struct device *dev = dbg_s->dev;
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
        struct nvgpu_dbg_gpu_reg_op *ops;
        u64 ops_size = sizeof(ops[0]) * args->num_ops;
@@ -689,7 +686,7 @@ static int nvgpu_ioctl_channel_reg_ops(struct dbg_session_gk20a *dbg_s,
        }
        /* be sure that ctx info is in place */
-        if (!gk20a_gpu_is_virtual(dbg_s->pdev) &&
+        if (!gk20a_gpu_is_virtual(dbg_s->dev) &&
                !gr_context_info_available(dbg_s, &g->gr)) {
                gk20a_err(dev, "gr context data not available\n");
                return -ENODEV;
@@ -757,7 +754,7 @@ static int dbg_set_powergate(struct dbg_session_gk20a *dbg_s,
                                __u32  powermode)
 {
        int err = 0;
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
         /* This function must be called with g->dbg_sessions_lock held */
@@ -786,7 +783,7 @@ static int dbg_set_powergate(struct dbg_session_gk20a *dbg_s,
                        if (err)
                                return err;
-                        err = gk20a_busy(dbg_s->pdev);
+                        err = gk20a_busy(dbg_s->dev);
                        if (err)
                                return -EPERM;
@@ -838,7 +835,7 @@ static int dbg_set_powergate(struct dbg_session_gk20a *dbg_s,
                                gk20a_pmu_enable_elpg(g);
                        gk20a_dbg(gpu_dbg_gpu_dbg | gpu_dbg_fn, "module idle");
-                        gk20a_idle(dbg_s->pdev);
+                        gk20a_idle(dbg_s->dev);
                        gk20a_idle(g->dev);
                }
@@ -862,7 +859,7 @@ static int nvgpu_ioctl_powergate_gk20a(struct dbg_session_gk20a *dbg_s,
                                struct nvgpu_dbg_gpu_powergate_args *args)
 {
        int err;
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
        gk20a_dbg_fn("%s  powergate mode = %d",
                      dev_name(dbg_s->dev), args->mode);
@@ -876,7 +873,7 @@ static int nvgpu_dbg_gpu_ioctl_smpc_ctxsw_mode(struct dbg_session_gk20a *dbg_s,
                               struct nvgpu_dbg_gpu_smpc_ctxsw_mode_args *args)
 {
        int err;
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
        struct channel_gk20a *ch_gk20a;
        gk20a_dbg_fn("%s smpc ctxsw mode = %d",
@@ -912,7 +909,7 @@ static int nvgpu_dbg_gpu_ioctl_hwpm_ctxsw_mode(struct dbg_session_gk20a *dbg_s,
                               struct nvgpu_dbg_gpu_hwpm_ctxsw_mode_args *args)
 {
        int err;
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
        struct channel_gk20a *ch_gk20a;
        gk20a_dbg_fn("%s pm ctxsw mode = %d",
@@ -948,7 +945,7 @@ static int nvgpu_dbg_gpu_ioctl_suspend_resume_sm(
                struct dbg_session_gk20a *dbg_s,
                struct nvgpu_dbg_gpu_suspend_resume_all_sms_args *args)
 {
-        struct gk20a *g = get_gk20a(dbg_s->pdev);
+        struct gk20a *g = get_gk20a(dbg_s->dev);
        struct channel_gk20a *ch = dbg_s->ch;
        bool ch_is_curr_ctx;
        int err = 0, action = args->mode;
diff --git a/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.h b/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.h
index a2e075a6..0ff6cfb3 100644
--- a/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.h
@@ -1,7 +1,7 @@
 /*
 * Tegra GK20A GPU Debugger Driver
 *
- * Copyright (c) 2013-2015, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2013-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -66,7 +66,6 @@ struct dbg_session_gk20a {
        /* gpu module vagaries */
        struct device             *dev;
-        struct platform_device    *pdev;
        struct gk20a              *g;
        /* bound channel, if any */
diff --git a/drivers/gpu/nvgpu/gk20a/debug_gk20a.c b/drivers/gpu/nvgpu/gk20a/debug_gk20a.c
index 67434e7f..369fbe29 100644
--- a/drivers/gpu/nvgpu/gk20a/debug_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/debug_gk20a.c
@@ -1,7 +1,7 @@
 /*
 * drivers/video/tegra/host/t20/debug_gk20a.c
 *
- * Copyright (C) 2011-2015 NVIDIA Corporation.  All rights reserved.
+ * Copyright (C) 2011-2016 NVIDIA Corporation.  All rights reserved.
 *
 * This software is licensed under the terms of the GNU General Public
 * License version 2, as published by the Free Software Foundation, and
@@ -32,7 +32,7 @@
 #include "hw_pbdma_gk20a.h"
 unsigned int gk20a_debug_trace_cmdbuf;
-static struct platform_device *gk20a_device;
+static struct device *gk20a_device;
 struct ch_state {
        int pid;
@@ -120,7 +120,7 @@ static void gk20a_debug_show_channel(struct gk20a *g,
        syncpointb = gk20a_mem_rd32(inst_ptr, ram_fc_syncpointb_w());
        gk20a_debug_output(o, "%d-%s, pid %d, refs: %d: ", hw_chid,
-                        g->dev->name,
+                        dev_name(g->dev),
                        ch_state->pid,
                        ch_state->refs);
        gk20a_debug_output(o, "%s in use %s %s\n",
@@ -181,7 +181,7 @@ void gk20a_debug_show_dump(struct gk20a *g, struct gk20a_debug_output *o)
                u32 status = gk20a_readl(g, fifo_pbdma_status_r(i));
                u32 chan_status = fifo_pbdma_status_chan_status_v(status);
-                gk20a_debug_output(o, "%s pbdma %d: ", g->dev->name, i);
+                gk20a_debug_output(o, "%s pbdma %d: ", dev_name(g->dev), i);
                gk20a_debug_output(o,
                                "id: %d (%s), next_id: %d (%s) status: %s\n",
                                fifo_pbdma_status_id_v(status),
@@ -206,7 +206,7 @@ void gk20a_debug_show_dump(struct gk20a *g, struct gk20a_debug_output *o)
                u32 status = gk20a_readl(g, fifo_engine_status_r(i));
                u32 ctx_status = fifo_engine_status_ctx_status_v(status);
-                gk20a_debug_output(o, "%s eng %d: ", g->dev->name, i);
+                gk20a_debug_output(o, "%s eng %d: ", dev_name(g->dev), i);
                gk20a_debug_output(o,
                                "id: %d (%s), next_id: %d (%s), ctx: %s ",
                                fifo_engine_status_id_v(status),
@@ -270,67 +270,67 @@ done:
        gk20a_idle(g->dev);
 }
-static int gk20a_gr_dump_regs(struct platform_device *pdev,
+static int gk20a_gr_dump_regs(struct device *dev,
                struct gk20a_debug_output *o)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(pdev);
+        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a *g = platform->g;
        int err;
-        err = gk20a_busy(g->dev);
+        err = gk20a_busy(dev);
        if (err) {
-                gk20a_err(&pdev->dev, "failed to power on gpu: %d\n", err);
+                gk20a_err(dev, "failed to power on gpu: %d\n", err);
                return -EINVAL;
        }
        gr_gk20a_elpg_protected_call(g, g->ops.gr.dump_gr_regs(g, o));
-        gk20a_idle(g->dev);
+        gk20a_idle(dev);
        return 0;
 }
-int gk20a_gr_debug_dump(struct platform_device *pdev)
+int gk20a_gr_debug_dump(struct device *dev)
 {
        struct gk20a_debug_output o = {
                .fn = gk20a_debug_write_printk
        };
-        gk20a_gr_dump_regs(pdev, &o);
+        gk20a_gr_dump_regs(dev, &o);
        return 0;
 }
 static int gk20a_gr_debug_show(struct seq_file *s, void *unused)
 {
-        struct platform_device *pdev = s->private;
+        struct device *dev = s->private;
        struct gk20a_debug_output o = {
                .fn = gk20a_debug_write_to_seqfile,
                .ctx = s,
        };
-        gk20a_gr_dump_regs(pdev, &o);
+        gk20a_gr_dump_regs(dev, &o);
        return 0;
 }
-void gk20a_debug_dump(struct platform_device *pdev)
+void gk20a_debug_dump(struct device *dev)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(pdev);
+        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a *g = platform->g;
        struct gk20a_debug_output o = {
                .fn = gk20a_debug_write_printk
        };
        if (platform->dump_platform_dependencies)
-                platform->dump_platform_dependencies(pdev);
+                platform->dump_platform_dependencies(dev);
        /* HAL only initialized after 1st power-on */
        if (g->ops.debug.show_dump)
                g->ops.debug.show_dump(g, &o);
 }
-void gk20a_debug_dump_device(struct platform_device *pdev)
+void gk20a_debug_dump_device(void *data)
 {
        struct gk20a_debug_output o = {
                .fn = gk20a_debug_write_printk
@@ -341,15 +341,7 @@ void gk20a_debug_dump_device(struct platform_device *pdev)
        if (!tegra_platform_is_silicon())
                return;
-        /* Dump the first device if no info is provided */
+        g = gk20a_from_dev(gk20a_device);
-        if (!pdev) {
-                if (!gk20a_device)
-                        return;
-                pdev = gk20a_device;
-        }
-        g = gk20a_get_platform(pdev)->g;
        /* HAL only initialized after 1st power-on */
        if (g->ops.debug.show_dump)
                g->ops.debug.show_dump(g, &o);
@@ -358,14 +350,14 @@ EXPORT_SYMBOL(gk20a_debug_dump_device);
 static int gk20a_debug_show(struct seq_file *s, void *unused)
 {
-        struct platform_device *pdev = s->private;
+        struct device *dev = s->private;
        struct gk20a_debug_output o = {
                .fn = gk20a_debug_write_to_seqfile,
                .ctx = s,
        };
        struct gk20a *g;
-        g = gk20a_get_platform(pdev)->g;
+        g = gk20a_get_platform(dev)->g;
        /* HAL only initialized after 1st power-on */
        if (g->ops.debug.show_dump)
                g->ops.debug.show_dump(g, &o);
@@ -401,24 +393,24 @@ void gk20a_init_debug_ops(struct gpu_ops *gops)
        gops->debug.show_dump = gk20a_debug_show_dump;
 }
-void gk20a_debug_init(struct platform_device *pdev)
+void gk20a_debug_init(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        /* Store the first device */
        if (!gk20a_device)
-                gk20a_device = pdev;
+                gk20a_device = dev;
-        platform->debugfs = debugfs_create_dir(pdev->name, NULL);
+        platform->debugfs = debugfs_create_dir(dev_name(dev), NULL);
        if (platform->debugfs) {
                platform->debugfs_alias =
-                        debugfs_create_symlink("gpu.0", NULL, pdev->name);
+                        debugfs_create_symlink("gpu.0", NULL, dev_name(dev));
        }
        debugfs_create_file("status", S_IRUGO, platform->debugfs,
-                        pdev, &gk20a_debug_fops);
+                        dev, &gk20a_debug_fops);
        debugfs_create_file("gr_status", S_IRUGO, platform->debugfs,
-                        pdev, &gk20a_gr_debug_fops);
+                        dev, &gk20a_gr_debug_fops);
        debugfs_create_u32("trace_cmdbuf", S_IRUGO|S_IWUSR, platform->debugfs,
                        &gk20a_debug_trace_cmdbuf);
diff --git a/drivers/gpu/nvgpu/gk20a/debug_gk20a.h b/drivers/gpu/nvgpu/gk20a/debug_gk20a.h
index 620688cb..1044f881 100644
--- a/drivers/gpu/nvgpu/gk20a/debug_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/debug_gk20a.h
@@ -1,7 +1,7 @@
 /*
 * GK20A Debug functionality
 *
- * Copyright (C) 2011-2015 NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (C) 2011-2016 NVIDIA CORPORATION.  All rights reserved.
 *
 * This software is licensed under the terms of the GNU General Public
 * License version 2, as published by the Free Software Foundation, and
@@ -32,10 +32,10 @@ struct gk20a_debug_output {
 void gk20a_debug_output(struct gk20a_debug_output *o,
                                        const char *fmt, ...);
-void gk20a_debug_dump(struct platform_device *pdev);
+void gk20a_debug_dump(struct device *pdev);
 void gk20a_debug_show_dump(struct gk20a *g, struct gk20a_debug_output *o);
-int gk20a_gr_debug_dump(struct platform_device *pdev);
+int gk20a_gr_debug_dump(struct device *pdev);
-void gk20a_debug_init(struct platform_device *pdev);
+void gk20a_debug_init(struct device *dev);
 void gk20a_init_debug_ops(struct gpu_ops *gops);
 #endif
diff --git a/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c b/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
index bac36403..aef0c9b3 100644
--- a/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
@@ -545,7 +545,7 @@ DEFINE_SIMPLE_ATTRIBUTE(gk20a_fecs_trace_debugfs_write_fops,
 static void gk20a_fecs_trace_debugfs_init(struct gk20a *g)
 {
-        struct gk20a_platform *plat = platform_get_drvdata(g->dev);
+        struct gk20a_platform *plat = dev_get_drvdata(g->dev);
        debugfs_create_file("ctxsw_trace_read", 0600, plat->debugfs, g,
                &gk20a_fecs_trace_debugfs_read_fops);
@@ -557,7 +557,7 @@ static void gk20a_fecs_trace_debugfs_init(struct gk20a *g)
 static void gk20a_fecs_trace_debugfs_cleanup(struct gk20a *g)
 {
-        struct gk20a_platform *plat = platform_get_drvdata(g->dev);
+        struct gk20a_platform *plat = dev_get_drvdata(g->dev);
        debugfs_remove_recursive(plat->debugfs);
 }
diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
index 27a3992d..4422bca4 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
@@ -348,7 +348,7 @@ int gk20a_init_fifo_reset_enable_hw(struct gk20a *g)
        u32 mask;
        u32 timeout;
        int i;
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        gk20a_dbg_fn("");
        /* enable pmc pfifo */
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.c b/drivers/gpu/nvgpu/gk20a/gk20a.c
index f2e6a683..f3bf0517 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.c
@@ -44,10 +44,6 @@
 #include <linux/sched.h>
-#ifdef CONFIG_TEGRA_GK20A
-#include <linux/nvhost.h>
-#endif
 #include "gk20a.h"
 #include "debug_gk20a.h"
 #include "ctrl_gk20a.h"
@@ -79,7 +75,6 @@
 #define CLASS_NAME "nvidia-gpu"
 /* TODO: Change to e.g. "nvidia-gpu%s" once we have symlinks in place. */
-#define INTERFACE_NAME "nvhost%s-gpu"
 #define GK20A_NUM_CDEVS 7
@@ -92,12 +87,11 @@ u32 gk20a_dbg_ftrace;
 #define GK20A_WAIT_FOR_IDLE_MS  2000
-static int gk20a_pm_finalize_poweron(struct device *dev);
 static int gk20a_pm_prepare_poweroff(struct device *dev);
-static inline void set_gk20a(struct platform_device *dev, struct gk20a *gk20a)
+static inline void set_gk20a(struct platform_device *pdev, struct gk20a *gk20a)
 {
-        gk20a_get_platform(dev)->g = gk20a;
+        gk20a_get_platform(&pdev->dev)->g = gk20a;
 }
 static const struct file_operations gk20a_channel_ops = {
@@ -292,38 +286,38 @@ static void __iomem *gk20a_ioremap_resource(struct platform_device *dev, int i,
 }
 /* TBD: strip from released */
-static int gk20a_init_sim_support(struct platform_device *dev)
+static int gk20a_init_sim_support(struct platform_device *pdev)
 {
        int err = 0;
+        struct device *dev = &pdev->dev;
        struct gk20a *g = get_gk20a(dev);
-        struct device *d = &dev->dev;
        u64 phys;
        g->sim.g = g;
-        g->sim.regs = gk20a_ioremap_resource(dev, GK20A_SIM_IORESOURCE_MEM,
+        g->sim.regs = gk20a_ioremap_resource(pdev, GK20A_SIM_IORESOURCE_MEM,
                                             &g->sim.reg_mem);
        if (IS_ERR(g->sim.regs)) {
-                dev_err(d, "failed to remap gk20a sim regs\n");
+                dev_err(dev, "failed to remap gk20a sim regs\n");
                err = PTR_ERR(g->sim.regs);
                goto fail;
        }
        /* allocate sim event/msg buffers */
-        err = alloc_and_kmap_iopage(d, &g->sim.send_bfr.kvaddr,
+        err = alloc_and_kmap_iopage(dev, &g->sim.send_bfr.kvaddr,
                                    &g->sim.send_bfr.phys,
                                    &g->sim.send_bfr.page);
-        err = err || alloc_and_kmap_iopage(d, &g->sim.recv_bfr.kvaddr,
+        err = err || alloc_and_kmap_iopage(dev, &g->sim.recv_bfr.kvaddr,
                                           &g->sim.recv_bfr.phys,
                                           &g->sim.recv_bfr.page);
-        err = err || alloc_and_kmap_iopage(d, &g->sim.msg_bfr.kvaddr,
+        err = err || alloc_and_kmap_iopage(dev, &g->sim.msg_bfr.kvaddr,
                                           &g->sim.msg_bfr.phys,
                                           &g->sim.msg_bfr.page);
        if (!(g->sim.send_bfr.kvaddr && g->sim.recv_bfr.kvaddr &&
              g->sim.msg_bfr.kvaddr)) {
-                dev_err(d, "couldn't allocate all sim buffers\n");
+                dev_err(dev, "couldn't allocate all sim buffers\n");
                goto fail;
        }
@@ -566,25 +560,25 @@ void gk20a_pbus_isr(struct gk20a *g)
                gk20a_err(dev_from_gk20a(g), "pmc_enable : 0x%x",
                        gk20a_readl(g, mc_enable_r()));
                gk20a_err(dev_from_gk20a(g), "NV_PBUS_INTR_0 : 0x%x", val);
-                gk20a_err(&g->dev->dev,
+                gk20a_err(g->dev,
                        "NV_PTIMER_PRI_TIMEOUT_SAVE_0: 0x%x\n",
                        gk20a_readl(g, timer_pri_timeout_save_0_r()));
-                gk20a_err(&g->dev->dev,
+                gk20a_err(g->dev,
                        "NV_PTIMER_PRI_TIMEOUT_SAVE_1: 0x%x\n",
                        gk20a_readl(g, timer_pri_timeout_save_1_r()));
                err_code = gk20a_readl(g, timer_pri_timeout_fecs_errcode_r());
-                gk20a_err(&g->dev->dev,
+                gk20a_err(g->dev,
                        "NV_PTIMER_PRI_TIMEOUT_FECS_ERRCODE: 0x%x\n",
                        err_code);
                if (err_code == 0xbadf13)
-                        gk20a_err(&g->dev->dev,
+                        gk20a_err(g->dev,
                        "NV_PGRAPH_PRI_GPC0_GPCCS_FS_GPC: 0x%x\n",
                        gk20a_readl(g, gr_gpc0_fs_gpc_r()));
        }
        if (val)
-                gk20a_err(&g->dev->dev,
+                gk20a_err(g->dev,
                        "Unhandled pending pbus interrupt\n");
        gk20a_writel(g, bus_intr_0_r(), val);
@@ -602,7 +596,7 @@ static irqreturn_t gk20a_intr_thread_nonstall(int irq, void *dev_id)
        return g->ops.mc.isr_thread_nonstall(g);
 }
-static void gk20a_remove_support(struct platform_device *dev)
+static void gk20a_remove_support(struct device *dev)
 {
        struct gk20a *g = get_gk20a(dev);
@@ -642,7 +636,7 @@ static void gk20a_remove_support(struct platform_device *dev)
 static int gk20a_init_support(struct platform_device *dev)
 {
        int err = 0;
-        struct gk20a *g = get_gk20a(dev);
+        struct gk20a *g = get_gk20a(&dev->dev);
 #ifdef CONFIG_TEGRA_COMMON
        tegra_register_idle_unidle(gk20a_do_idle, gk20a_do_unidle);
@@ -689,14 +683,13 @@ static int gk20a_init_support(struct platform_device *dev)
        return 0;
 fail:
-        gk20a_remove_support(dev);
+        gk20a_remove_support(&dev->dev);
        return err;
 }
 static int gk20a_pm_prepare_poweroff(struct device *dev)
 {
-        struct platform_device *pdev = to_platform_device(dev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(pdev);
        int ret = 0;
        gk20a_dbg_fn("");
@@ -706,7 +699,7 @@ static int gk20a_pm_prepare_poweroff(struct device *dev)
        if (!g->power_on)
                goto done;
-        gk20a_scale_suspend(pdev);
+        gk20a_scale_suspend(dev);
        /* cancel any pending cde work */
        gk20a_cde_suspend(g);
@@ -767,11 +760,10 @@ static int gk20a_detect_chip(struct gk20a *g)
        return gpu_init_hal(g);
 }
-static int gk20a_pm_finalize_poweron(struct device *dev)
+int gk20a_pm_finalize_poweron(struct device *dev)
 {
-        struct platform_device *pdev = to_platform_device(dev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a_platform *platform = gk20a_get_platform(dev);
-        struct gk20a_platform *platform = gk20a_get_platform(pdev);
        int err, nice_value;
        gk20a_dbg_fn("");
@@ -904,7 +896,7 @@ static int gk20a_pm_finalize_poweron(struct device *dev)
        gk20a_channel_resume(g);
        set_user_nice(current, nice_value);
-        gk20a_scale_resume(pdev);
+        gk20a_scale_resume(dev);
        trace_gk20a_finalize_poweron_done(dev_name(dev));
@@ -955,13 +947,13 @@ static struct of_device_id tegra_gk20a_of_match[] = {
 };
 static int gk20a_create_device(
-        struct platform_device *pdev, int devno, const char *cdev_name,
+        struct device *dev, int devno,
+        const char *interface_name, const char *cdev_name,
        struct cdev *cdev, struct device **out,
        const struct file_operations *ops)
 {
-        struct device *dev;
+        struct device *subdev;
        int err;
-        struct gk20a *g = get_gk20a(pdev);
        gk20a_dbg_fn("");
@@ -970,127 +962,112 @@ static int gk20a_create_device(
        err = cdev_add(cdev, devno, 1);
        if (err) {
-                dev_err(&pdev->dev,
+                dev_err(dev, "failed to add %s cdev\n", cdev_name);
-                        "failed to add %s cdev\n", cdev_name);
                return err;
        }
-        dev = device_create(g->class, NULL, devno, NULL,
+        subdev = device_create(&nvgpu_class, NULL, devno, NULL,
-                (pdev->id <= 0) ? INTERFACE_NAME : INTERFACE_NAME ".%d",
+                interface_name, cdev_name);
-                cdev_name, pdev->id);
-        if (IS_ERR(dev)) {
+        if (IS_ERR(subdev)) {
                err = PTR_ERR(dev);
                cdev_del(cdev);
-                dev_err(&pdev->dev,
+                dev_err(dev, "failed to create %s device for %s\n",
-                        "failed to create %s device for %s\n",
+                        cdev_name, dev_name(dev));
-                        cdev_name, pdev->name);
                return err;
        }
-        *out = dev;
+        *out = subdev;
        return 0;
 }
-void gk20a_user_deinit(struct platform_device *dev)
+void gk20a_user_deinit(struct device *dev)
 {
-        struct gk20a *g = get_gk20a(dev);
+        struct gk20a *g = gk20a_from_dev(dev);
        if (g->channel.node) {
-                device_destroy(g->class, g->channel.cdev.dev);
+                device_destroy(&nvgpu_class, g->channel.cdev.dev);
                cdev_del(&g->channel.cdev);
        }
        if (g->as.node) {
-                device_destroy(g->class, g->as.cdev.dev);
+                device_destroy(&nvgpu_class, g->as.cdev.dev);
                cdev_del(&g->as.cdev);
        }
        if (g->ctrl.node) {
-                device_destroy(g->class, g->ctrl.cdev.dev);
+                device_destroy(&nvgpu_class, g->ctrl.cdev.dev);
                cdev_del(&g->ctrl.cdev);
        }
        if (g->dbg.node) {
-                device_destroy(g->class, g->dbg.cdev.dev);
+                device_destroy(&nvgpu_class, g->dbg.cdev.dev);
                cdev_del(&g->dbg.cdev);
        }
        if (g->prof.node) {
-                device_destroy(g->class, g->prof.cdev.dev);
+                device_destroy(&nvgpu_class, g->prof.cdev.dev);
                cdev_del(&g->prof.cdev);
        }
        if (g->tsg.node) {
-                device_destroy(g->class, g->tsg.cdev.dev);
+                device_destroy(&nvgpu_class, g->tsg.cdev.dev);
                cdev_del(&g->tsg.cdev);
        }
        if (g->ctxsw.node) {
-                device_destroy(g->class, g->ctxsw.cdev.dev);
+                device_destroy(&nvgpu_class, g->ctxsw.cdev.dev);
                cdev_del(&g->ctxsw.cdev);
        }
        if (g->cdev_region)
                unregister_chrdev_region(g->cdev_region, GK20A_NUM_CDEVS);
-        if (g->class)
-                class_destroy(g->class);
 }
-int gk20a_user_init(struct platform_device *dev)
+int gk20a_user_init(struct device *dev, const char *interface_name)
 {
        int err;
        dev_t devno;
-        struct gk20a *g = get_gk20a(dev);
+        struct gk20a *g = gk20a_from_dev(dev);
-        g->class = class_create(THIS_MODULE, CLASS_NAME);
-        if (IS_ERR(g->class)) {
-                err = PTR_ERR(g->class);
-                g->class = NULL;
-                dev_err(&dev->dev,
-                        "failed to create " CLASS_NAME " class\n");
-                goto fail;
-        }
-        err = alloc_chrdev_region(&devno, 0, GK20A_NUM_CDEVS, CLASS_NAME);
+        err = alloc_chrdev_region(&devno, 0, GK20A_NUM_CDEVS, dev_name(dev));
        if (err) {
-                dev_err(&dev->dev, "failed to allocate devno\n");
+                dev_err(dev, "failed to allocate devno\n");
                goto fail;
        }
        g->cdev_region = devno;
-        err = gk20a_create_device(dev, devno++, "",
+        err = gk20a_create_device(dev, devno++, interface_name, "",
                                  &g->channel.cdev, &g->channel.node,
                                  &gk20a_channel_ops);
        if (err)
                goto fail;
-        err = gk20a_create_device(dev, devno++, "-as",
+        err = gk20a_create_device(dev, devno++, interface_name, "-as",
                                  &g->as.cdev, &g->as.node,
                                  &gk20a_as_ops);
        if (err)
                goto fail;
-        err = gk20a_create_device(dev, devno++, "-ctrl",
+        err = gk20a_create_device(dev, devno++, interface_name, "-ctrl",
                                  &g->ctrl.cdev, &g->ctrl.node,
                                  &gk20a_ctrl_ops);
        if (err)
                goto fail;
-        err = gk20a_create_device(dev, devno++, "-dbg",
+        err = gk20a_create_device(dev, devno++, interface_name, "-dbg",
                                  &g->dbg.cdev, &g->dbg.node,
                                  &gk20a_dbg_ops);
        if (err)
                goto fail;
-        err = gk20a_create_device(dev, devno++, "-prof",
+        err = gk20a_create_device(dev, devno++, interface_name, "-prof",
                                  &g->prof.cdev, &g->prof.node,
                                  &gk20a_prof_ops);
        if (err)
                goto fail;
-        err = gk20a_create_device(dev, devno++, "-tsg",
+        err = gk20a_create_device(dev, devno++, interface_name, "-tsg",
                                  &g->tsg.cdev, &g->tsg.node,
                                  &gk20a_tsg_ops);
        if (err)
@@ -1190,7 +1167,7 @@ static void gk20a_pm_shutdown(struct platform_device *pdev)
 #endif
        /* Be ready for rail-gate after this point */
-        if (gk20a_gpu_is_virtual(pdev))
+        if (gk20a_gpu_is_virtual(&pdev->dev))
                vgpu_pm_prepare_poweroff(&pdev->dev);
        else
                gk20a_pm_prepare_poweroff(&pdev->dev);
@@ -1205,12 +1182,12 @@ static const struct dev_pm_ops gk20a_pm_ops = {
 };
 #endif
-static int _gk20a_pm_railgate(struct platform_device *pdev)
+static int _gk20a_pm_railgate(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        int ret = 0;
        if (platform->railgate)
-                ret = platform->railgate(pdev);
+                ret = platform->railgate(dev);
        return ret;
 }
@@ -1223,14 +1200,14 @@ static int gk20a_pm_railgate(struct generic_pm_domain *domain)
        return _gk20a_pm_railgate(g->dev);
 }
-static int _gk20a_pm_unrailgate(struct platform_device *pdev)
+static int _gk20a_pm_unrailgate(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        int ret = 0;
        if (platform->unrailgate) {
                mutex_lock(&platform->railgate_lock);
-                ret = platform->unrailgate(pdev);
+                ret = platform->unrailgate(dev);
                mutex_unlock(&platform->railgate_lock);
        }
@@ -1242,7 +1219,7 @@ static int gk20a_pm_unrailgate(struct generic_pm_domain *domain)
        struct gk20a_domain_data *gk20a_domain = container_of(domain,
                                     struct gk20a_domain_data, gpd);
        struct gk20a *g = gk20a_domain->gk20a;
-        trace_gk20a_pm_unrailgate(dev_name(&g->dev->dev));
+        trace_gk20a_pm_unrailgate(dev_name(g->dev));
        return _gk20a_pm_unrailgate(g->dev);
 }
@@ -1273,11 +1250,14 @@ static int gk20a_pm_resume(struct device *dev)
 }
 #ifdef CONFIG_PM_GENERIC_DOMAINS_OF
-static int gk20a_pm_initialise_domain(struct platform_device *pdev)
+static int gk20a_pm_initialise_domain(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct dev_power_governor *pm_domain_gov = NULL;
-        struct generic_pm_domain *domain = dev_to_genpd(&pdev->dev);
+        struct generic_pm_domain *domain = dev_to_genpd(dev);
+        if (IS_ERR(domain))
+                return 0;
 #ifdef CONFIG_PM
        if (!platform->can_railgate)
@@ -1288,14 +1268,14 @@ static int gk20a_pm_initialise_domain(struct platform_device *pdev)
        if (platform->railgate_delay)
                pm_genpd_set_poweroff_delay(domain, platform->railgate_delay);
-        device_set_wakeup_capable(&pdev->dev, 0);
+        device_set_wakeup_capable(dev, 0);
        return 0;
 }
 #else
-static int gk20a_pm_initialise_domain(struct platform_device *pdev)
+static int gk20a_pm_initialise_domain(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct dev_power_governor *pm_domain_gov = NULL;
        struct generic_pm_domain *domain = NULL;
        int ret = 0;
@@ -1326,8 +1306,8 @@ static int gk20a_pm_initialise_domain(struct platform_device *pdev)
        domain->dev_ops.suspend = gk20a_pm_suspend;
        domain->dev_ops.resume = gk20a_pm_resume;
-        device_set_wakeup_capable(&pdev->dev, 0);
+        device_set_wakeup_capable(dev, 0);
-        ret = pm_genpd_add_device(domain, &pdev->dev);
+        ret = pm_genpd_add_device(domain, dev);
        if (platform->railgate_delay)
                pm_genpd_set_poweroff_delay(domain, platform->railgate_delay);
@@ -1336,23 +1316,23 @@ static int gk20a_pm_initialise_domain(struct platform_device *pdev)
 }
 #endif
-static int gk20a_pm_init(struct platform_device *dev)
+static int gk20a_pm_init(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        int err = 0;
        gk20a_dbg_fn("");
        /* Initialise pm runtime */
        if (platform->clockgate_delay) {
-                pm_runtime_set_autosuspend_delay(&dev->dev,
+                pm_runtime_set_autosuspend_delay(dev,
                                                 platform->clockgate_delay);
-                pm_runtime_use_autosuspend(&dev->dev);
+                pm_runtime_use_autosuspend(dev);
        }
-        pm_runtime_enable(&dev->dev);
+        pm_runtime_enable(dev);
-        if (!pm_runtime_enabled(&dev->dev))
+        if (!pm_runtime_enabled(dev))
-                gk20a_pm_enable_clk(&dev->dev);
+                gk20a_pm_enable_clk(dev);
        /* Enable runtime railgating if possible. If not,
         * turn on the rail now. */
@@ -1374,7 +1354,7 @@ static int gk20a_secure_page_alloc(struct platform_device *pdev)
        int err = 0;
        if (platform->secure_page_alloc) {
-                err = platform->secure_page_alloc(pdev);
+                err = platform->secure_page_alloc(&pdev->dev);
                if (!err)
                        platform->secure_alloc_ready = true;
        }
@@ -1418,7 +1398,7 @@ static int gk20a_probe(struct platform_device *dev)
        platform_set_drvdata(dev, platform);
-        if (gk20a_gpu_is_virtual(dev))
+        if (gk20a_gpu_is_virtual(&dev->dev))
                return vgpu_probe(dev);
        gk20a = kzalloc(sizeof(struct gk20a), GFP_KERNEL);
@@ -1437,7 +1417,7 @@ static int gk20a_probe(struct platform_device *dev)
 #endif
        set_gk20a(dev, gk20a);
-        gk20a->dev = dev;
+        gk20a->dev = &dev->dev;
        gk20a->irq_stall = platform_get_irq(dev, 0);
        gk20a->irq_nonstall = platform_get_irq(dev, 1);
@@ -1468,7 +1448,7 @@ static int gk20a_probe(struct platform_device *dev)
        disable_irq(gk20a->irq_stall);
        disable_irq(gk20a->irq_nonstall);
-        err = gk20a_user_init(dev);
+        err = gk20a_user_init(&dev->dev, INTERFACE_NAME);
        if (err)
                return err;
@@ -1485,10 +1465,10 @@ static int gk20a_probe(struct platform_device *dev)
                platform->reset_control = NULL;
 #endif
-        gk20a_debug_init(dev);
+        gk20a_debug_init(&dev->dev);
        /* Initialize the platform interface. */
-        err = platform->probe(dev);
+        err = platform->probe(&dev->dev);
        if (err) {
                dev_err(&dev->dev, "platform probe failed");
                return err;
@@ -1499,7 +1479,7 @@ static int gk20a_probe(struct platform_device *dev)
                dev_err(&dev->dev,
                        "failed to allocate secure buffer %d\n", err);
-        err = gk20a_pm_init(dev);
+        err = gk20a_pm_init(&dev->dev);
        if (err) {
                dev_err(&dev->dev, "pm init failed");
                return err;
@@ -1509,7 +1489,7 @@ static int gk20a_probe(struct platform_device *dev)
        /* Initialise scaling */
        if (IS_ENABLED(CONFIG_GK20A_DEVFREQ))
-                gk20a_scale_init(dev);
+                gk20a_scale_init(&dev->dev);
        /* Set DMA parameters to allow larger sgt lists */
        dev->dev.dma_parms = &gk20a->dma_parms;
@@ -1547,14 +1527,14 @@ static int gk20a_probe(struct platform_device *dev)
        gk20a->pmu.aelpg_param[4] = APCTRL_CYCLES_PER_SAMPLE_MAX_DEFAULT;
        if (platform->late_probe) {
-                err = platform->late_probe(dev);
+                err = platform->late_probe(&dev->dev);
                if (err) {
                        dev_err(&dev->dev, "late probe failed");
                        return err;
                }
        }
-        gk20a_create_sysfs(dev);
+        gk20a_create_sysfs(&dev->dev);
 #ifdef CONFIG_DEBUG_FS
        spin_lock_init(&gk20a->debugfs_lock);
@@ -1612,10 +1592,10 @@ static int gk20a_probe(struct platform_device *dev)
                                        &gk20a->runlist_interleave);
        gr_gk20a_debugfs_init(gk20a);
-        gk20a_pmu_debugfs_init(dev);
+        gk20a_pmu_debugfs_init(&dev->dev);
-        gk20a_cde_debugfs_init(dev);
+        gk20a_cde_debugfs_init(&dev->dev);
        gk20a_alloc_debugfs_init(dev);
-        gk20a_mm_debugfs_init(dev);
+        gk20a_mm_debugfs_init(&dev->dev);
 #endif
        gk20a_init_gr(gk20a);
@@ -1623,8 +1603,9 @@ static int gk20a_probe(struct platform_device *dev)
        return 0;
 }
-static int __exit gk20a_remove(struct platform_device *dev)
+static int __exit gk20a_remove(struct platform_device *pdev)
 {
+        struct device *dev = &pdev->dev;
        struct gk20a *g = get_gk20a(dev);
        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a_domain_data *gk20a_gpd;
@@ -1632,7 +1613,7 @@ static int __exit gk20a_remove(struct platform_device *dev)
        gk20a_dbg_fn("");
        if (gk20a_gpu_is_virtual(dev))
-                return vgpu_remove(dev);
+                return vgpu_remove(pdev);
        if (platform->has_cde)
                gk20a_cde_destroy(g);
@@ -1650,7 +1631,7 @@ static int __exit gk20a_remove(struct platform_device *dev)
        debugfs_remove_recursive(platform->debugfs);
        debugfs_remove_recursive(platform->debugfs_alias);
-        gk20a_remove_sysfs(&dev->dev);
+        gk20a_remove_sysfs(dev);
        if (platform->secure_buffer.destroy)
                platform->secure_buffer.destroy(dev,
@@ -1660,15 +1641,15 @@ static int __exit gk20a_remove(struct platform_device *dev)
        gk20a_gpd->gk20a = NULL;
        kfree(gk20a_gpd);
-        if (pm_runtime_enabled(&dev->dev))
+        if (pm_runtime_enabled(dev))
-                pm_runtime_disable(&dev->dev);
+                pm_runtime_disable(dev);
        else
-                gk20a_pm_disable_clk(&dev->dev);
+                gk20a_pm_disable_clk(dev);
        if (platform->remove)
                platform->remove(dev);
-        set_gk20a(dev, NULL);
+        set_gk20a(pdev, NULL);
        kfree(g);
        gk20a_dbg_fn("removed");
@@ -1751,11 +1732,20 @@ static int gk20a_domain_init(struct of_device_id *matches)
 #endif
+struct class nvgpu_class = {
+        .owner = THIS_MODULE,
+        .name = CLASS_NAME,
+};
 static int __init gk20a_init(void)
 {
        int ret;
+        ret = class_register(&nvgpu_class);
+        if (ret)
+                return ret;
        ret = gk20a_domain_init(tegra_gpu_domain_match);
        if (ret)
                return ret;
@@ -1766,50 +1756,51 @@ static int __init gk20a_init(void)
 static void __exit gk20a_exit(void)
 {
        platform_driver_unregister(&gk20a_driver);
+        class_unregister(&nvgpu_class);
 }
-void gk20a_busy_noresume(struct platform_device *pdev)
+void gk20a_busy_noresume(struct device *dev)
 {
-        pm_runtime_get_noresume(&pdev->dev);
+        pm_runtime_get_noresume(dev);
 }
-int gk20a_busy(struct platform_device *pdev)
+int gk20a_busy(struct device *dev)
 {
        int ret = 0;
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
 #ifdef CONFIG_PM
-        struct gk20a_platform *platform = gk20a_get_platform(pdev);
+        struct gk20a_platform *platform = gk20a_get_platform(dev);
 #endif
        down_read(&g->busy_lock);
 #ifdef CONFIG_PM
        if (platform->busy) {
-                ret = platform->busy(pdev);
+                ret = platform->busy(dev);
                if (ret < 0) {
-                        dev_err(&pdev->dev, "%s: failed to poweron platform dependency\n",
+                        dev_err(dev, "%s: failed to poweron platform dependency\n",
                                __func__);
                        goto fail;
                }
        }
-        ret = pm_runtime_get_sync(&pdev->dev);
+        ret = pm_runtime_get_sync(dev);
        if (ret < 0) {
-                pm_runtime_put_noidle(&pdev->dev);
+                pm_runtime_put_noidle(dev);
                if (platform->idle)
-                        platform->idle(pdev);
+                        platform->idle(dev);
                goto fail;
        }
 #else
        if (!g->power_on) {
                ret = gk20a_gpu_is_virtual(pdev) ?
-                        vgpu_pm_finalize_poweron(&pdev->dev)
+                        vgpu_pm_finalize_poweron(dev)
-                        : gk20a_pm_finalize_poweron(&pdev->dev);
+                        : gk20a_pm_finalize_poweron(dev);
                if (ret)
                        goto fail;
        }
 #endif
-        gk20a_scale_notify_busy(pdev);
+        gk20a_scale_notify_busy(dev);
 fail:
        up_read(&g->busy_lock);
@@ -1817,19 +1808,19 @@ fail:
        return ret < 0 ? ret : 0;
 }
-void gk20a_idle(struct platform_device *pdev)
+void gk20a_idle(struct device *dev)
 {
 #ifdef CONFIG_PM
-        struct gk20a_platform *platform = gk20a_get_platform(pdev);
+        struct gk20a_platform *platform = gk20a_get_platform(dev);
-        if (atomic_read(&pdev->dev.power.usage_count) == 1)
+        if (atomic_read(&dev->power.usage_count) == 1)
-                gk20a_scale_notify_idle(pdev);
+                gk20a_scale_notify_idle(dev);
-        pm_runtime_mark_last_busy(&pdev->dev);
+        pm_runtime_mark_last_busy(dev);
-        pm_runtime_put_sync_autosuspend(&pdev->dev);
+        pm_runtime_put_sync_autosuspend(dev);
        if (platform->idle)
-                platform->idle(pdev);
+                platform->idle(dev);
 #else
-        gk20a_scale_notify_idle(pdev);
+        gk20a_scale_notify_idle(dev);
 #endif
 }
@@ -1882,10 +1873,10 @@ void gk20a_reset(struct gk20a *g, u32 units)
 * In success, we hold these locks and return
 * In failure, we release these locks and return
 */
-int __gk20a_do_idle(struct platform_device *pdev, bool force_reset)
+int __gk20a_do_idle(struct device *dev, bool force_reset)
 {
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a_platform *platform = dev_get_drvdata(&pdev->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        unsigned long timeout = jiffies +
                msecs_to_jiffies(GK20A_WAIT_FOR_IDLE_MS);
        int ref_cnt;
@@ -1898,7 +1889,7 @@ int __gk20a_do_idle(struct platform_device *pdev, bool force_reset)
        mutex_lock(&platform->railgate_lock);
        /* check if it is already railgated ? */
-        if (platform->is_railgated(pdev))
+        if (platform->is_railgated(dev))
                return 0;
        /*
@@ -1906,17 +1897,17 @@ int __gk20a_do_idle(struct platform_device *pdev, bool force_reset)
         * re-acquire railgate_lock
         */
        mutex_unlock(&platform->railgate_lock);
-        pm_runtime_get_sync(&pdev->dev);
+        pm_runtime_get_sync(dev);
        mutex_lock(&platform->railgate_lock);
        /* check and wait until GPU is idle (with a timeout) */
        do {
                msleep(1);
-                ref_cnt = atomic_read(&pdev->dev.power.usage_count);
+                ref_cnt = atomic_read(&dev->power.usage_count);
        } while (ref_cnt != 1 && time_before(jiffies, timeout));
        if (ref_cnt != 1) {
-                gk20a_err(&pdev->dev, "failed to idle - refcount %d != 1\n",
+                gk20a_err(dev, "failed to idle - refcount %d != 1\n",
                        ref_cnt);
                goto fail_drop_usage_count;
        }
@@ -1931,7 +1922,7 @@ int __gk20a_do_idle(struct platform_device *pdev, bool force_reset)
                 * if GPU is now idle, we will have only one ref count,
                 * drop this ref which will rail gate the GPU
                 */
-                pm_runtime_put_sync(&pdev->dev);
+                pm_runtime_put_sync(dev);
                /* add sufficient delay to allow GPU to rail gate */
                msleep(platform->railgate_delay);
@@ -1941,13 +1932,13 @@ int __gk20a_do_idle(struct platform_device *pdev, bool force_reset)
                /* check in loop if GPU is railgated or not */
                do {
                        msleep(1);
-                        is_railgated = platform->is_railgated(pdev);
+                        is_railgated = platform->is_railgated(dev);
                } while (!is_railgated && time_before(jiffies, timeout));
                if (is_railgated) {
                        return 0;
                } else {
-                        gk20a_err(&pdev->dev, "failed to idle in timeout\n");
+                        gk20a_err(dev, "failed to idle in timeout\n");
                        goto fail_timeout;
                }
        } else {
@@ -1964,12 +1955,12 @@ int __gk20a_do_idle(struct platform_device *pdev, bool force_reset)
                 */
                /* Save the GPU state */
-                gk20a_pm_prepare_poweroff(&pdev->dev);
+                gk20a_pm_prepare_poweroff(dev);
-                gk20a_pm_disable_clk(&pdev->dev);
+                gk20a_pm_disable_clk(dev);
                /* railgate GPU */
-                platform->railgate(pdev);
+                platform->railgate(dev);
                udelay(10);
@@ -1978,7 +1969,7 @@ int __gk20a_do_idle(struct platform_device *pdev, bool force_reset)
        }
 fail_drop_usage_count:
-        pm_runtime_put_noidle(&pdev->dev);
+        pm_runtime_put_noidle(dev);
 fail_timeout:
        mutex_unlock(&platform->railgate_lock);
        up_write(&g->busy_lock);
@@ -1997,7 +1988,7 @@ int gk20a_do_idle(void)
                        of_find_matching_node(NULL, tegra_gk20a_of_match);
        struct platform_device *pdev = of_find_device_by_node(node);
-        int ret =  __gk20a_do_idle(pdev, true);
+        int ret =  __gk20a_do_idle(&pdev->dev, true);
        of_node_put(node);
@@ -2007,25 +1998,25 @@ int gk20a_do_idle(void)
 /**
 * __gk20a_do_unidle() - unblock all the tasks blocked by __gk20a_do_idle()
 */
-int __gk20a_do_unidle(struct platform_device *pdev)
+int __gk20a_do_unidle(struct device *dev)
 {
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a_platform *platform = dev_get_drvdata(&pdev->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        if (g->forced_reset) {
                /*
                 * If we did a forced-reset/railgate
                 * then unrailgate the GPU here first
                 */
-                platform->unrailgate(pdev);
+                platform->unrailgate(dev);
-                gk20a_pm_enable_clk(&pdev->dev);
+                gk20a_pm_enable_clk(dev);
                /* restore the GPU state */
-                gk20a_pm_finalize_poweron(&pdev->dev);
+                gk20a_pm_finalize_poweron(dev);
                /* balance GPU usage counter */
-                pm_runtime_put_sync(&pdev->dev);
+                pm_runtime_put_sync(dev);
                g->forced_reset = false;
        }
@@ -2046,7 +2037,7 @@ int gk20a_do_unidle(void)
                        of_find_matching_node(NULL, tegra_gk20a_of_match);
        struct platform_device *pdev = of_find_device_by_node(node);
-        int ret = __gk20a_do_unidle(pdev);
+        int ret = __gk20a_do_unidle(&pdev->dev);
        of_node_put(node);
@@ -2057,7 +2048,7 @@ int gk20a_do_unidle(void)
 int gk20a_init_gpu_characteristics(struct gk20a *g)
 {
        struct nvgpu_gpu_characteristics *gpu = &g->gpu_characteristics;
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        gpu->L2_cache_size = g->ops.ltc.determine_L2_size_bytes(g);
        gpu->on_board_video_memory_size = 0; /* integrated GPU */
@@ -2164,7 +2155,7 @@ do_request_firmware(struct device *dev, const char *prefix, const char *fw_name)
 const struct firmware *
 gk20a_request_firmware(struct gk20a *g, const char *fw_name)
 {
-        struct device *dev = &g->dev->dev;
+        struct device *dev = g->dev;
        const struct firmware *fw;
        /* current->fs is NULL when calling from SYS_EXIT.
@@ -2177,8 +2168,10 @@ gk20a_request_firmware(struct gk20a *g, const char *fw_name)
 #ifdef CONFIG_TEGRA_GK20A
        /* TO BE REMOVED - Support loading from legacy SOC specific path. */
-        if (!fw)
+        if (!fw) {
-                fw = nvhost_client_request_firmware(g->dev, fw_name);
+                struct gk20a_platform *platform = gk20a_get_platform(dev);
+                fw = do_request_firmware(dev, platform->soc_name, fw_name);
+        }
 #endif
        if (!fw) {
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.h b/drivers/gpu/nvgpu/gk20a/gk20a.h
index 273eeaf4..9d8dc5f7 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.h
@@ -18,7 +18,6 @@
 #ifndef GK20A_H
 #define GK20A_H
 struct gk20a;
 struct fifo_gk20a;
 struct channel_gk20a;
@@ -206,7 +205,7 @@ struct gpu_ops {
                        bool *post_event, struct channel_gk20a *fault_ch);
                int (*handle_tex_exception)(struct gk20a *g, u32 gpc, u32 tpc,
                                                bool *post_event);
-                void (*create_gr_sysfs)(struct platform_device *dev);
+                void (*create_gr_sysfs)(struct device *dev);
                u32 (*get_lrf_tex_ltc_dram_override)(struct gk20a *g);
        } gr;
        const char *name;
@@ -517,7 +516,7 @@ struct gpu_ops {
 };
 struct gk20a {
-        struct platform_device *dev;
+        struct device *dev;
        struct platform_device *host1x_dev;
        struct resource *reg_mem;
@@ -602,7 +601,7 @@ struct gk20a {
         */
        u64 separate_fixed_allocs;
-        void (*remove_support)(struct platform_device *);
+        void (*remove_support)(struct device *);
        u64 pg_ingating_time_us;
        u64 pg_ungating_time_us;
@@ -648,7 +647,6 @@ struct gk20a {
        int client_refcount; /* open channels and ctrl nodes */
        dev_t cdev_region;
-        struct class *class;
        struct gpu_ops ops;
@@ -685,7 +683,7 @@ static inline unsigned long gk20a_get_gr_idle_timeout(struct gk20a *g)
                g->gr_idle_timeout_default : MAX_SCHEDULE_TIMEOUT;
 }
-static inline struct gk20a *get_gk20a(struct platform_device *dev)
+static inline struct gk20a *get_gk20a(struct device *dev)
 {
        return gk20a_get_platform(dev)->g;
 }
@@ -880,7 +878,11 @@ static inline u32 gk20a_bar1_readl(struct gk20a *g, u32 b)
 /* convenience */
 static inline struct device *dev_from_gk20a(struct gk20a *g)
 {
-        return &g->dev->dev;
+        return g->dev;
+}
+static inline struct gk20a *gk20a_from_dev(struct device *dev)
+{
+        return ((struct gk20a_platform *)dev_get_drvdata(dev))->g;
 }
 static inline struct gk20a *gk20a_from_as(struct gk20a_as *as)
 {
@@ -927,14 +929,14 @@ enum {
        KEPLER_CHANNEL_GPFIFO_C   = 0xA26F,
 };
-static inline bool gk20a_gpu_is_virtual(struct platform_device *dev)
+static inline bool gk20a_gpu_is_virtual(struct device *dev)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        return platform->virtual_dev;
 }
-static inline int support_gk20a_pmu(struct platform_device *dev)
+static inline int support_gk20a_pmu(struct device *dev)
 {
        if (IS_ENABLED(CONFIG_GK20A_PMU)) {
                /* gPMU is not supported for vgpu */
@@ -944,23 +946,23 @@ static inline int support_gk20a_pmu(struct platform_device *dev)
        return 0;
 }
-void gk20a_create_sysfs(struct platform_device *dev);
+void gk20a_create_sysfs(struct device *dev);
 void gk20a_remove_sysfs(struct device *dev);
 #define GK20A_BAR0_IORESOURCE_MEM 0
 #define GK20A_BAR1_IORESOURCE_MEM 1
 #define GK20A_SIM_IORESOURCE_MEM 2
-void gk20a_busy_noresume(struct platform_device *pdev);
+void gk20a_busy_noresume(struct device *dev);
-int __must_check gk20a_busy(struct platform_device *pdev);
+int __must_check gk20a_busy(struct device *dev);
-void gk20a_idle(struct platform_device *pdev);
+void gk20a_idle(struct device *dev);
 void gk20a_disable(struct gk20a *g, u32 units);
 void gk20a_enable(struct gk20a *g, u32 units);
 void gk20a_reset(struct gk20a *g, u32 units);
 int gk20a_do_idle(void);
 int gk20a_do_unidle(void);
-int __gk20a_do_idle(struct platform_device *pdev, bool force_reset);
+int __gk20a_do_idle(struct device *dev, bool force_reset);
-int __gk20a_do_unidle(struct platform_device *pdev);
+int __gk20a_do_unidle(struct device *dev);
 const struct firmware *
 gk20a_request_firmware(struct gk20a *g, const char *fw_name);
@@ -981,10 +983,10 @@ int gk20a_init_gpu_characteristics(struct gk20a *g);
 void gk20a_pbus_isr(struct gk20a *g);
-int gk20a_user_init(struct platform_device *dev);
+int gk20a_user_init(struct device *dev, const char *interface_name);
-void gk20a_user_deinit(struct platform_device *dev);
+void gk20a_user_deinit(struct device *dev);
-extern void gk20a_debug_dump_device(struct platform_device *pdev);
+void gk20a_debug_dump_device(void *dev);
 static inline u32 ptimer_scalingfactor10x(u32 ptimer_src_freq)
 {
@@ -999,4 +1001,8 @@ static inline u32 scale_ptimer(u32 timeout , u32 scale10x)
 }
 u64 gk20a_read_ptimer(struct gk20a *g);
+extern struct class nvgpu_class;
+#define INTERFACE_NAME "nvhost%s-gpu"
 #endif /* GK20A_H */
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a_scale.c b/drivers/gpu/nvgpu/gk20a/gk20a_scale.c
index 65772d3b..e5c93db5 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a_scale.c
+++ b/drivers/gpu/nvgpu/gk20a/gk20a_scale.c
@@ -1,7 +1,7 @@
 /*
 * gk20a clock scaling profile
 *
- * Copyright (c) 2013-2015, NVIDIA Corporation. All rights reserved.
+ * Copyright (c) 2013-2016, NVIDIA Corporation. All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -47,8 +47,8 @@ static int gk20a_scale_qos_notify(struct notifier_block *nb,
        struct gk20a_scale_profile *profile =
                container_of(nb, struct gk20a_scale_profile,
                             qos_notify_block);
-        struct gk20a_platform *platform = platform_get_drvdata(profile->pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(profile->dev);
-        struct gk20a *g = get_gk20a(profile->pdev);
+        struct gk20a *g = get_gk20a(profile->dev);
        unsigned long freq;
        if (!platform->postscale)
@@ -56,7 +56,7 @@ static int gk20a_scale_qos_notify(struct notifier_block *nb,
        /* get the frequency requirement. if devfreq is enabled, check if it
         * has higher demand than qos */
-        freq = platform->clk_round_rate(profile->pdev,
+        freq = platform->clk_round_rate(profile->dev,
                                pm_qos_request(platform->qos_id));
        if (g->devfreq)
                freq = max(g->devfreq->previous_freq, freq);
@@ -64,7 +64,7 @@ static int gk20a_scale_qos_notify(struct notifier_block *nb,
        /* Update gpu load because we may scale the emc target
         * if the gpu load changed. */
        gk20a_pmu_load_update(g);
-        platform->postscale(profile->pdev, freq);
+        platform->postscale(profile->dev, freq);
        return NOTIFY_OK;
 }
@@ -77,12 +77,12 @@ static int gk20a_scale_qos_notify(struct notifier_block *nb,
 static int gk20a_scale_make_freq_table(struct gk20a_scale_profile *profile)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(profile->pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(profile->dev);
        int num_freqs, err;
        unsigned long *freqs;
        /* get gpu frequency table */
-        err = platform->get_clk_freqs(profile->pdev, &freqs,
+        err = platform->get_clk_freqs(profile->dev, &freqs,
                                        &num_freqs);
        if (err)
                return -ENOSYS;
@@ -103,22 +103,21 @@ static int gk20a_scale_target(struct device *dev, unsigned long *freq,
                              u32 flags)
 {
        struct gk20a_platform *platform = dev_get_drvdata(dev);
-        struct platform_device *pdev = to_platform_device(dev);
        unsigned long rounded_rate =
-                platform->clk_round_rate(pdev, *freq);
+                platform->clk_round_rate(dev, *freq);
-        if (platform->clk_get_rate(pdev) == rounded_rate)
+        if (platform->clk_get_rate(dev) == rounded_rate)
                *freq = rounded_rate;
        else {
-                platform->clk_set_rate(pdev, rounded_rate);
+                platform->clk_set_rate(dev, rounded_rate);
-                *freq = platform->clk_get_rate(pdev);
+                *freq = platform->clk_get_rate(dev);
        }
        /* postscale will only scale emc (dram clock) if evaluating
         * gk20a_tegra_get_emc_rate() produces a new or different emc
         * target because the load or_and gpufreq has changed */
        if (platform->postscale)
-                platform->postscale(pdev, rounded_rate);
+                platform->postscale(dev, rounded_rate);
        return 0;
 }
@@ -130,9 +129,9 @@ static int gk20a_scale_target(struct device *dev, unsigned long *freq,
 * based on the time it was asked last time.
 */
-static void update_load_estimate_gpmu(struct platform_device *pdev)
+static void update_load_estimate_gpmu(struct device *dev)
 {
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
        struct gk20a_scale_profile *profile = g->scale_profile;
        unsigned long dt;
        u32 busy_time;
@@ -148,14 +147,14 @@ static void update_load_estimate_gpmu(struct platform_device *pdev)
 }
 /*
- * gk20a_scale_suspend(pdev)
+ * gk20a_scale_suspend(dev)
 *
 * This function informs devfreq of suspend
 */
-void gk20a_scale_suspend(struct platform_device *pdev)
+void gk20a_scale_suspend(struct device *dev)
 {
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
        struct devfreq *devfreq = g->devfreq;
        if (!devfreq)
@@ -165,14 +164,14 @@ void gk20a_scale_suspend(struct platform_device *pdev)
 }
 /*
- * gk20a_scale_resume(pdev)
+ * gk20a_scale_resume(dev)
 *
 * This functions informs devfreq of resume
 */
-void gk20a_scale_resume(struct platform_device *pdev)
+void gk20a_scale_resume(struct device *dev)
 {
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
        struct devfreq *devfreq = g->devfreq;
        if (!devfreq)
@@ -182,15 +181,15 @@ void gk20a_scale_resume(struct platform_device *pdev)
 }
 /*
- * gk20a_scale_notify(pdev, busy)
+ * gk20a_scale_notify(dev, busy)
 *
 * Calling this function informs that the device is idling (..or busy). This
 * data is used to estimate the current load
 */
-static void gk20a_scale_notify(struct platform_device *pdev, bool busy)
+static void gk20a_scale_notify(struct device *dev, bool busy)
 {
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
        struct gk20a_scale_profile *profile = g->scale_profile;
        struct devfreq *devfreq = g->devfreq;
@@ -204,15 +203,15 @@ static void gk20a_scale_notify(struct platform_device *pdev, bool busy)
        mutex_unlock(&devfreq->lock);
 }
-void gk20a_scale_notify_idle(struct platform_device *pdev)
+void gk20a_scale_notify_idle(struct device *dev)
 {
-        gk20a_scale_notify(pdev, false);
+        gk20a_scale_notify(dev, false);
 }
-void gk20a_scale_notify_busy(struct platform_device *pdev)
+void gk20a_scale_notify_busy(struct device *dev)
 {
-        gk20a_scale_notify(pdev, true);
+        gk20a_scale_notify(dev, true);
 }
 /*
@@ -224,24 +223,23 @@ void gk20a_scale_notify_busy(struct platform_device *pdev)
 static int gk20a_scale_get_dev_status(struct device *dev,
                                      struct devfreq_dev_status *stat)
 {
-        struct gk20a *g = get_gk20a(to_platform_device(dev));
+        struct gk20a *g = get_gk20a(dev);
        struct gk20a_scale_profile *profile = g->scale_profile;
-        struct platform_device *pdev = to_platform_device(dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
        /* update the software shadow */
        gk20a_pmu_load_update(g);
        /* inform edp about new constraint */
        if (platform->prescale)
-                platform->prescale(pdev);
+                platform->prescale(dev);
        /* Make sure there are correct values for the current frequency */
        profile->dev_stat.current_frequency =
-                                platform->clk_get_rate(profile->pdev);
+                                platform->clk_get_rate(profile->dev);
        /* Update load estimate */
-        update_load_estimate_gpmu(to_platform_device(dev));
+        update_load_estimate_gpmu(dev);
        /* Copy the contents of the current device status */
        *stat = profile->dev_stat;
@@ -254,12 +252,12 @@ static int gk20a_scale_get_dev_status(struct device *dev,
 }
 /*
- * gk20a_scale_init(pdev)
+ * gk20a_scale_init(dev)
 */
-void gk20a_scale_init(struct platform_device *pdev)
+void gk20a_scale_init(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct gk20a *g = platform->g;
        struct gk20a_scale_profile *profile;
        int err;
@@ -269,7 +267,7 @@ void gk20a_scale_init(struct platform_device *pdev)
        profile = kzalloc(sizeof(*profile), GFP_KERNEL);
-        profile->pdev = pdev;
+        profile->dev = dev;
        profile->dev_stat.busy = false;
        /* Create frequency table */
@@ -290,7 +288,7 @@ void gk20a_scale_init(struct platform_device *pdev)
                profile->devfreq_profile.get_dev_status =
                        gk20a_scale_get_dev_status;
-                devfreq = devfreq_add_device(&pdev->dev,
+                devfreq = devfreq_add_device(dev,
                                        &profile->devfreq_profile,
                                        platform->devfreq_governor, NULL);
@@ -316,9 +314,9 @@ err_get_freqs:
        kfree(profile);
 }
-void gk20a_scale_exit(struct platform_device *pdev)
+void gk20a_scale_exit(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct gk20a *g = platform->g;
        int err;
@@ -344,9 +342,9 @@ void gk20a_scale_exit(struct platform_device *pdev)
 * Initialize hardware portion of the device
 */
-void gk20a_scale_hw_init(struct platform_device *pdev)
+void gk20a_scale_hw_init(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct gk20a_scale_profile *profile = platform->g->scale_profile;
        /* make sure that scaling has bee initialised */
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a_scale.h b/drivers/gpu/nvgpu/gk20a/gk20a_scale.h
index 889a3973..5c8618eb 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a_scale.h
+++ b/drivers/gpu/nvgpu/gk20a/gk20a_scale.h
@@ -1,7 +1,7 @@
 /*
 * gk20a clock scaling profile
 *
- * Copyright (c) 2013-2015, NVIDIA Corporation. All rights reserved.
+ * Copyright (c) 2013-2016, NVIDIA Corporation. All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -21,11 +21,10 @@
 #include <linux/devfreq.h>
-struct platform_device;
 struct clk;
 struct gk20a_scale_profile {
-        struct platform_device          *pdev;
+        struct device                   *dev;
        ktime_t                         last_event_time;
        struct devfreq_dev_profile      devfreq_profile;
        struct devfreq_dev_status       dev_stat;
@@ -34,25 +33,25 @@ struct gk20a_scale_profile {
 };
 /* Initialization and de-initialization for module */
-void gk20a_scale_init(struct platform_device *);
+void gk20a_scale_init(struct device *);
-void gk20a_scale_exit(struct platform_device *);
+void gk20a_scale_exit(struct device *);
-void gk20a_scale_hw_init(struct platform_device *pdev);
+void gk20a_scale_hw_init(struct device *dev);
 #if defined(CONFIG_GK20A_DEVFREQ)
 /*
 * call when performing submit to notify scaling mechanism that the module is
 * in use
 */
-void gk20a_scale_notify_busy(struct platform_device *);
+void gk20a_scale_notify_busy(struct device *);
-void gk20a_scale_notify_idle(struct platform_device *);
+void gk20a_scale_notify_idle(struct device *);
-void gk20a_scale_suspend(struct platform_device *);
+void gk20a_scale_suspend(struct device *);
-void gk20a_scale_resume(struct platform_device *);
+void gk20a_scale_resume(struct device *);
 #else
-static inline void gk20a_scale_notify_busy(struct platform_device *pdev) {}
+static inline void gk20a_scale_notify_busy(struct device *dev) {}
-static inline void gk20a_scale_notify_idle(struct platform_device *pdev) {}
+static inline void gk20a_scale_notify_idle(struct device *dev) {}
-static inline void gk20a_scale_suspend(struct platform_device *pdev) {}
+static inline void gk20a_scale_suspend(struct device *dev) {}
-static inline void gk20a_scale_resume(struct platform_device *pdev) {}
+static inline void gk20a_scale_resume(struct device *dev) {}
 #endif
 #endif
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c b/drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c
index d433c9bb..9d25cfdd 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c
+++ b/drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c
@@ -18,7 +18,7 @@
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
-#include <linux/platform_device.h>
+#include <linux/device.h>
 #include <linux/pm_runtime.h>
 #include <linux/kernel.h>
 #include <linux/fb.h>
@@ -34,11 +34,10 @@
 #define ROOTRW (S_IRWXU|S_IRGRP|S_IROTH)
-static ssize_t elcg_enable_store(struct device *device,
+static ssize_t elcg_enable_store(struct device *dev,
        struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        int err;
@@ -60,28 +59,26 @@ static ssize_t elcg_enable_store(struct device *device,
        }
        gk20a_idle(g->dev);
-        dev_info(device, "ELCG is %s.\n", g->elcg_enabled ? "enabled" :
+        dev_info(dev, "ELCG is %s.\n", g->elcg_enabled ? "enabled" :
                        "disabled");
        return count;
 }
-static ssize_t elcg_enable_read(struct device *device,
+static ssize_t elcg_enable_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d\n", g->elcg_enabled ? 1 : 0);
 }
 static DEVICE_ATTR(elcg_enable, ROOTRW, elcg_enable_read, elcg_enable_store);
-static ssize_t blcg_enable_store(struct device *device,
+static ssize_t blcg_enable_store(struct device *dev,
        struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        int err;
@@ -118,17 +115,16 @@ static ssize_t blcg_enable_store(struct device *device,
                        g->blcg_enabled);
        gk20a_idle(g->dev);
-        dev_info(device, "BLCG is %s.\n", g->blcg_enabled ? "enabled" :
+        dev_info(dev, "BLCG is %s.\n", g->blcg_enabled ? "enabled" :
                        "disabled");
        return count;
 }
-static ssize_t blcg_enable_read(struct device *device,
+static ssize_t blcg_enable_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d\n", g->blcg_enabled ? 1 : 0);
 }
@@ -136,11 +132,10 @@ static ssize_t blcg_enable_read(struct device *device,
 static DEVICE_ATTR(blcg_enable, ROOTRW, blcg_enable_read, blcg_enable_store);
-static ssize_t slcg_enable_store(struct device *device,
+static ssize_t slcg_enable_store(struct device *dev,
        struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        int err;
@@ -185,17 +180,16 @@ static ssize_t slcg_enable_store(struct device *device,
                g->ops.clock_gating.slcg_xbar_load_gating_prod(g, g->slcg_enabled);
        gk20a_idle(g->dev);
-        dev_info(device, "SLCG is %s.\n", g->slcg_enabled ? "enabled" :
+        dev_info(dev, "SLCG is %s.\n", g->slcg_enabled ? "enabled" :
                        "disabled");
        return count;
 }
-static ssize_t slcg_enable_read(struct device *device,
+static ssize_t slcg_enable_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d\n", g->slcg_enabled ? 1 : 0);
 }
@@ -203,8 +197,8 @@ static ssize_t slcg_enable_read(struct device *device,
 static DEVICE_ATTR(slcg_enable, ROOTRW, slcg_enable_read, slcg_enable_store);
 static ssize_t ptimer_scale_factor_show(struct device *dev,
-                                                struct device_attribute *attr,
+                                        struct device_attribute *attr,
-                                                char *buf)
+                                        char *buf)
 {
        struct gk20a_platform *platform = dev_get_drvdata(dev);
        u32 src_freq_hz = platform->ptimer_src_freq;
@@ -240,8 +234,7 @@ static ssize_t railgate_enable_store(struct device *dev,
 {
        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct generic_pm_domain *genpd = dev_to_genpd(dev);
-        struct platform_device *ndev = to_platform_device(dev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long railgate_enable = 0;
        int err;
@@ -255,7 +248,7 @@ static ssize_t railgate_enable_store(struct device *dev,
                /* release extra ref count:if power domains not enabled */
                if ((platform->railgate) && \
                                !IS_ENABLED(CONFIG_PM_GENERIC_DOMAINS))
-                        err = platform->railgate(ndev);
+                        err = platform->railgate(dev);
                mutex_unlock(&platform->railgate_lock);
        } else if (railgate_enable == 0 && platform->can_railgate) {
                mutex_lock(&platform->railgate_lock);
@@ -265,7 +258,7 @@ static ssize_t railgate_enable_store(struct device *dev,
                /* take extra ref count - incase of power domains not enabled */
                if ((platform->unrailgate) && \
                                !IS_ENABLED(CONFIG_PM_GENERIC_DOMAINS))
-                        err = platform->unrailgate(ndev);
+                        err = platform->unrailgate(dev);
                mutex_unlock(&platform->railgate_lock);
        }
        dev_info(dev, "railgate is %s.\n", platform->can_railgate ?
@@ -279,11 +272,10 @@ static ssize_t railgate_enable_store(struct device *dev,
        return count;
 }
-static ssize_t railgate_enable_read(struct device *device,
+static ssize_t railgate_enable_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
-        struct gk20a_platform *platform = dev_get_drvdata(&ndev->dev);
        return sprintf(buf, "%d\n", platform->can_railgate ? 1 : 0);
 }
@@ -297,9 +289,8 @@ static ssize_t railgate_delay_store(struct device *dev,
                                    const char *buf, size_t count)
 {
        struct gk20a_platform *platform = dev_get_drvdata(dev);
-        struct platform_device *ndev = to_platform_device(dev);
        int railgate_delay = 0, ret = 0;
-        struct gk20a *g = get_gk20a(ndev);
+        struct gk20a *g = get_gk20a(dev);
        int err;
        if (!platform->can_railgate) {
@@ -373,8 +364,7 @@ static DEVICE_ATTR(clockgate_delay, ROOTRW, clockgate_delay_show,
 static ssize_t counters_show(struct device *dev,
                             struct device_attribute *attr, char *buf)
 {
-        struct platform_device *pdev = to_platform_device(dev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(pdev);
        u32 busy_cycles, total_cycles;
        ssize_t res;
@@ -390,8 +380,7 @@ static ssize_t counters_show_reset(struct device *dev,
                                   struct device_attribute *attr, char *buf)
 {
        ssize_t res = counters_show(dev, attr, buf);
-        struct platform_device *pdev = to_platform_device(dev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(pdev);
        gk20a_pmu_reset_load_counters(g);
@@ -403,8 +392,7 @@ static ssize_t gk20a_load_show(struct device *dev,
                                     struct device_attribute *attr,
                                     char *buf)
 {
-        struct platform_device *pdev = to_platform_device(dev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(pdev);
        u32 busy_time;
        ssize_t res;
        int err;
@@ -427,11 +415,10 @@ static ssize_t gk20a_load_show(struct device *dev,
 }
 static DEVICE_ATTR(load, S_IRUGO, gk20a_load_show, NULL);
-static ssize_t elpg_enable_store(struct device *device,
+static ssize_t elpg_enable_store(struct device *dev,
        struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        int err;
@@ -455,28 +442,26 @@ static ssize_t elpg_enable_store(struct device *device,
        }
        gk20a_idle(g->dev);
-        dev_info(device, "ELPG is %s.\n", g->elpg_enabled ? "enabled" :
+        dev_info(dev, "ELPG is %s.\n", g->elpg_enabled ? "enabled" :
                        "disabled");
        return count;
 }
-static ssize_t elpg_enable_read(struct device *device,
+static ssize_t elpg_enable_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d\n", g->elpg_enabled ? 1 : 0);
 }
 static DEVICE_ATTR(elpg_enable, ROOTRW, elpg_enable_read, elpg_enable_store);
-static ssize_t aelpg_param_store(struct device *device,
+static ssize_t aelpg_param_store(struct device *dev,
        struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        int status = 0;
        union pmu_ap_cmd ap_cmd;
        int *paramlist = (int *)g->pmu.aelpg_param;
@@ -514,11 +499,10 @@ static ssize_t aelpg_param_store(struct device *device,
        return count;
 }
-static ssize_t aelpg_param_read(struct device *device,
+static ssize_t aelpg_param_read(struct device *dev,
                struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d %d %d %d %d\n", g->pmu.aelpg_param[0],
                g->pmu.aelpg_param[1], g->pmu.aelpg_param[2],
@@ -528,11 +512,10 @@ static ssize_t aelpg_param_read(struct device *device,
 static DEVICE_ATTR(aelpg_param, ROOTRW,
                aelpg_param_read, aelpg_param_store);
-static ssize_t aelpg_enable_store(struct device *device,
+static ssize_t aelpg_enable_store(struct device *dev,
                struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        int status = 0;
        union pmu_ap_cmd ap_cmd;
@@ -558,21 +541,20 @@ static ssize_t aelpg_enable_store(struct device *device,
                        status = gk20a_pmu_ap_send_command(g, &ap_cmd, false);
                }
        } else {
-                dev_info(device, "PMU is not ready, AELPG request failed\n");
+                dev_info(dev, "PMU is not ready, AELPG request failed\n");
        }
        gk20a_idle(g->dev);
-        dev_info(device, "AELPG is %s.\n", g->aelpg_enabled ? "enabled" :
+        dev_info(dev, "AELPG is %s.\n", g->aelpg_enabled ? "enabled" :
                        "disabled");
        return count;
 }
-static ssize_t aelpg_enable_read(struct device *device,
+static ssize_t aelpg_enable_read(struct device *dev,
                struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d\n", g->aelpg_enabled ? 1 : 0);
 }
@@ -581,19 +563,17 @@ static DEVICE_ATTR(aelpg_enable, ROOTRW,
                aelpg_enable_read, aelpg_enable_store);
-static ssize_t allow_all_enable_read(struct device *device,
+static ssize_t allow_all_enable_read(struct device *dev,
                struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d\n", g->allow_all ? 1 : 0);
 }
-static ssize_t allow_all_enable_store(struct device *device,
+static ssize_t allow_all_enable_store(struct device *dev,
                struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        int err;
@@ -610,11 +590,10 @@ static ssize_t allow_all_enable_store(struct device *device,
 static DEVICE_ATTR(allow_all, ROOTRW,
                allow_all_enable_read, allow_all_enable_store);
-static ssize_t emc3d_ratio_store(struct device *device,
+static ssize_t emc3d_ratio_store(struct device *dev,
        struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        if (kstrtoul(buf, 10, &val) < 0)
@@ -625,22 +604,20 @@ static ssize_t emc3d_ratio_store(struct device *device,
        return count;
 }
-static ssize_t emc3d_ratio_read(struct device *device,
+static ssize_t emc3d_ratio_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d\n", g->emc3d_ratio);
 }
 static DEVICE_ATTR(emc3d_ratio, ROOTRW, emc3d_ratio_read, emc3d_ratio_store);
-static ssize_t fmax_at_vmin_safe_read(struct device *device,
+static ssize_t fmax_at_vmin_safe_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long gpu_fmax_at_vmin_hz = 0;
        gpu_fmax_at_vmin_hz = tegra_dvfs_get_fmax_at_vmin_safe_t(
@@ -652,11 +629,10 @@ static ssize_t fmax_at_vmin_safe_read(struct device *device,
 static DEVICE_ATTR(fmax_at_vmin_safe, S_IRUGO, fmax_at_vmin_safe_read, NULL);
 #ifdef CONFIG_PM
-static ssize_t force_idle_store(struct device *device,
+static ssize_t force_idle_store(struct device *dev,
        struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        int err = 0;
@@ -667,10 +643,10 @@ static ssize_t force_idle_store(struct device *device,
                if (g->forced_idle)
                        return count; /* do nothing */
                else {
-                        err = __gk20a_do_idle(ndev, false);
+                        err = __gk20a_do_idle(dev, false);
                        if (!err) {
                                g->forced_idle = 1;
-                                dev_info(device, "gpu is idle : %d\n",
+                                dev_info(dev, "gpu is idle : %d\n",
                                        g->forced_idle);
                        }
                }
@@ -678,10 +654,10 @@ static ssize_t force_idle_store(struct device *device,
                if (!g->forced_idle)
                        return count; /* do nothing */
                else {
-                        err = __gk20a_do_unidle(ndev);
+                        err = __gk20a_do_unidle(dev);
                        if (!err) {
                                g->forced_idle = 0;
-                                dev_info(device, "gpu is idle : %d\n",
+                                dev_info(dev, "gpu is idle : %d\n",
                                        g->forced_idle);
                        }
                }
@@ -690,11 +666,10 @@ static ssize_t force_idle_store(struct device *device,
        return count;
 }
-static ssize_t force_idle_read(struct device *device,
+static ssize_t force_idle_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        return sprintf(buf, "%d\n", g->forced_idle ? 1 : 0);
 }
@@ -702,11 +677,10 @@ static ssize_t force_idle_read(struct device *device,
 static DEVICE_ATTR(force_idle, ROOTRW, force_idle_read, force_idle_store);
 #endif
-static ssize_t tpc_fs_mask_store(struct device *device,
+static ssize_t tpc_fs_mask_store(struct device *dev,
        struct device_attribute *attr, const char *buf, size_t count)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        unsigned long val = 0;
        if (kstrtoul(buf, 10, &val) < 0)
@@ -730,11 +704,10 @@ static ssize_t tpc_fs_mask_store(struct device *device,
        return count;
 }
-static ssize_t tpc_fs_mask_read(struct device *device,
+static ssize_t tpc_fs_mask_read(struct device *dev,
        struct device_attribute *attr, char *buf)
 {
-        struct platform_device *ndev = to_platform_device(device);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(ndev);
        struct gr_gk20a *gr = &g->gr;
        u32 gpc_index;
        u32 tpc_fs_mask = 0;
@@ -760,7 +733,7 @@ static DEVICE_ATTR(tpc_fs_mask, ROOTRW, tpc_fs_mask_read, tpc_fs_mask_store);
 void gk20a_remove_sysfs(struct device *dev)
 {
-        struct gk20a *g = get_gk20a(to_platform_device(dev));
+        struct gk20a *g = get_gk20a(dev);
        device_remove_file(dev, &dev_attr_elcg_enable);
        device_remove_file(dev, &dev_attr_blcg_enable);
@@ -797,50 +770,50 @@ void gk20a_remove_sysfs(struct device *dev)
        }
 }
-void gk20a_create_sysfs(struct platform_device *dev)
+void gk20a_create_sysfs(struct device *dev)
 {
-        struct gk20a *g = get_gk20a(dev);
+        struct gk20a *g = gk20a_from_dev(dev);
        int error = 0;
-        error |= device_create_file(&dev->dev, &dev_attr_elcg_enable);
+        error |= device_create_file(dev, &dev_attr_elcg_enable);
-        error |= device_create_file(&dev->dev, &dev_attr_blcg_enable);
+        error |= device_create_file(dev, &dev_attr_blcg_enable);
-        error |= device_create_file(&dev->dev, &dev_attr_slcg_enable);
+        error |= device_create_file(dev, &dev_attr_slcg_enable);
-        error |= device_create_file(&dev->dev, &dev_attr_ptimer_scale_factor);
+        error |= device_create_file(dev, &dev_attr_ptimer_scale_factor);
-        error |= device_create_file(&dev->dev, &dev_attr_elpg_enable);
+        error |= device_create_file(dev, &dev_attr_elpg_enable);
-        error |= device_create_file(&dev->dev, &dev_attr_emc3d_ratio);
+        error |= device_create_file(dev, &dev_attr_emc3d_ratio);
-        error |= device_create_file(&dev->dev, &dev_attr_fmax_at_vmin_safe);
+        error |= device_create_file(dev, &dev_attr_fmax_at_vmin_safe);
-        error |= device_create_file(&dev->dev, &dev_attr_counters);
+        error |= device_create_file(dev, &dev_attr_counters);
-        error |= device_create_file(&dev->dev, &dev_attr_counters_reset);
+        error |= device_create_file(dev, &dev_attr_counters_reset);
-        error |= device_create_file(&dev->dev, &dev_attr_load);
+        error |= device_create_file(dev, &dev_attr_load);
-        error |= device_create_file(&dev->dev, &dev_attr_railgate_delay);
+        error |= device_create_file(dev, &dev_attr_railgate_delay);
-        error |= device_create_file(&dev->dev, &dev_attr_is_railgated);
+        error |= device_create_file(dev, &dev_attr_is_railgated);
-        error |= device_create_file(&dev->dev, &dev_attr_clockgate_delay);
+        error |= device_create_file(dev, &dev_attr_clockgate_delay);
 #ifdef CONFIG_PM
-        error |= device_create_file(&dev->dev, &dev_attr_force_idle);
+        error |= device_create_file(dev, &dev_attr_force_idle);
 #if defined(CONFIG_PM_GENERIC_DOMAINS)
-        error |= device_create_file(&dev->dev, &dev_attr_railgate_enable);
+        error |= device_create_file(dev, &dev_attr_railgate_enable);
 #endif
 #endif
-        error |= device_create_file(&dev->dev, &dev_attr_aelpg_param);
+        error |= device_create_file(dev, &dev_attr_aelpg_param);
-        error |= device_create_file(&dev->dev, &dev_attr_aelpg_enable);
+        error |= device_create_file(dev, &dev_attr_aelpg_enable);
-        error |= device_create_file(&dev->dev, &dev_attr_allow_all);
+        error |= device_create_file(dev, &dev_attr_allow_all);
-        error |= device_create_file(&dev->dev, &dev_attr_tpc_fs_mask);
+        error |= device_create_file(dev, &dev_attr_tpc_fs_mask);
-        if (g->host1x_dev && (dev->dev.parent != &g->host1x_dev->dev)) {
+        if (g->host1x_dev && (dev->parent != &g->host1x_dev->dev)) {
                error |= sysfs_create_link(&g->host1x_dev->dev.kobj,
-                                           &dev->dev.kobj,
+                                           &dev->kobj,
-                                           dev_name(&dev->dev));
+                                           dev_name(dev));
-                if (strcmp(dev_name(&dev->dev), "gpu.0")) {
+                if (strcmp(dev_name(dev), "gpu.0")) {
-                        struct kobject *kobj = &dev->dev.kobj;
+                        struct kobject *kobj = &dev->kobj;
                        struct device *parent = container_of((kobj->parent),
                                                struct device, kobj);
                        error |= sysfs_create_link(&parent->kobj,
-                                           &dev->dev.kobj, "gpu.0");
+                                           &dev->kobj, "gpu.0");
                }
        }
        if (error)
-                dev_err(&dev->dev, "Failed to create sysfs attributes!\n");
+                dev_err(dev, "Failed to create sysfs attributes!\n");
 }
diff --git a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
index cc3a6813..b0b5571f 100644
--- a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
@@ -2439,10 +2439,10 @@ static int gk20a_gr_alloc_ctx_buffer(struct gk20a *g,
 static int gr_gk20a_alloc_global_ctx_buffers(struct gk20a *g)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        struct gr_gk20a *gr = &g->gr;
        int i, attr_buffer_size, err;
-        struct platform_device *pdev = g->dev;
+        struct device *dev = g->dev;
        u32 cb_buffer_size = gr->bundle_cb_default_size *
                gr_scc_bundle_cb_size_div_256b_byte_granularity_v();
@@ -2462,7 +2462,7 @@ static int gr_gk20a_alloc_global_ctx_buffers(struct gk20a *g)
                goto clean_up;
        if (platform->secure_alloc)
-                platform->secure_alloc(pdev,
+                platform->secure_alloc(dev,
                                       &gr->global_ctx_buffer[CIRCULAR_VPR],
                                       cb_buffer_size);
@@ -2474,7 +2474,7 @@ static int gr_gk20a_alloc_global_ctx_buffers(struct gk20a *g)
                goto clean_up;
        if (platform->secure_alloc)
-                platform->secure_alloc(pdev,
+                platform->secure_alloc(dev,
                                       &gr->global_ctx_buffer[PAGEPOOL_VPR],
                                       pagepool_buffer_size);
@@ -2486,12 +2486,12 @@ static int gr_gk20a_alloc_global_ctx_buffers(struct gk20a *g)
                goto clean_up;
        if (platform->secure_alloc)
-                platform->secure_alloc(pdev,
+                platform->secure_alloc(dev,
                                       &gr->global_ctx_buffer[ATTRIBUTE_VPR],
                                       attr_buffer_size);
        if (platform->secure_buffer.destroy)
-                platform->secure_buffer.destroy(pdev, &platform->secure_buffer);
+                platform->secure_buffer.destroy(dev, &platform->secure_buffer);
        gk20a_dbg_info("golden_image_size : %d",
                   gr->ctx_vars.golden_image_size);
@@ -5016,7 +5016,7 @@ static int gr_gk20a_handle_sw_method(struct gk20a *g, u32 addr,
 {
        gk20a_dbg_fn("");
-        trace_gr_gk20a_handle_sw_method(g->dev->name);
+        trace_gr_gk20a_handle_sw_method(dev_name(g->dev));
        if (class_num == KEPLER_COMPUTE_A) {
                switch (offset << 2) {
@@ -8113,7 +8113,7 @@ static int gr_gk20a_dump_gr_status_regs(struct gk20a *g,
 #ifdef CONFIG_DEBUG_FS
 int gr_gk20a_debugfs_init(struct gk20a *g)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        g->debugfs_gr_default_attrib_cb_size =
                debugfs_create_u32("gr_default_attrib_cb_size",
diff --git a/drivers/gpu/nvgpu/gk20a/hal.c b/drivers/gpu/nvgpu/gk20a/hal.c
index 10607af0..7755bf7d 100644
--- a/drivers/gpu/nvgpu/gk20a/hal.c
+++ b/drivers/gpu/nvgpu/gk20a/hal.c
@@ -1,7 +1,7 @@
 /*
 * NVIDIA GPU HAL interface.
 *
- * Copyright (c) 2014, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -41,7 +41,7 @@ int gpu_init_hal(struct gk20a *g)
                break;
 #endif
        default:
-                gk20a_err(&g->dev->dev, "no support for %x", ver);
+                gk20a_err(g->dev, "no support for %x", ver);
                return -ENODEV;
        }
diff --git a/drivers/gpu/nvgpu/gk20a/ltc_gk20a.c b/drivers/gpu/nvgpu/gk20a/ltc_gk20a.c
index 0d9a98b4..963f6bb7 100644
--- a/drivers/gpu/nvgpu/gk20a/ltc_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/ltc_gk20a.c
@@ -111,7 +111,7 @@ static int gk20a_ltc_cbc_ctrl(struct gk20a *g, enum gk20a_cbc_op op,
        gk20a_dbg_fn("");
-        trace_gk20a_ltc_cbc_ctrl_start(g->dev->name, op, min, max);
+        trace_gk20a_ltc_cbc_ctrl_start(dev_name(g->dev), op, min, max);
        if (gr->compbit_store.mem.size == 0)
                return 0;
@@ -163,7 +163,7 @@ static int gk20a_ltc_cbc_ctrl(struct gk20a *g, enum gk20a_cbc_op op,
                }
        }
 out:
-        trace_gk20a_ltc_cbc_ctrl_done(g->dev->name);
+        trace_gk20a_ltc_cbc_ctrl_done(dev_name(g->dev));
        mutex_unlock(&g->mm.l2_op_lock);
        return err;
 }
@@ -204,7 +204,7 @@ static void gk20a_mm_g_elpg_flush_locked(struct gk20a *g)
        gk20a_dbg_fn("");
-        trace_gk20a_mm_g_elpg_flush_locked(g->dev->name);
+        trace_gk20a_mm_g_elpg_flush_locked(dev_name(g->dev));
        /* Make sure all previous writes are committed to the L2. There's no
           guarantee that writes are to DRAM. This will be a sysmembar internal
@@ -227,7 +227,7 @@ static void gk20a_mm_g_elpg_flush_locked(struct gk20a *g)
                gk20a_warn(dev_from_gk20a(g),
                            "g_elpg_flush too many retries");
-        trace_gk20a_mm_g_elpg_flush_locked_done(g->dev->name);
+        trace_gk20a_mm_g_elpg_flush_locked_done(dev_name(g->dev));
 }
diff --git a/drivers/gpu/nvgpu/gk20a/mc_gk20a.c b/drivers/gpu/nvgpu/gk20a/mc_gk20a.c
index 0a773d10..ea8b8024 100644
--- a/drivers/gpu/nvgpu/gk20a/mc_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/mc_gk20a.c
@@ -1,7 +1,7 @@
 /*
- * GK20A memory interface
+ * GK20A Master Control
 *
- * Copyright (c) 2014-2015, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -24,7 +24,7 @@ irqreturn_t mc_gk20a_isr_stall(struct gk20a *g)
 {
        u32 mc_intr_0;
-        trace_mc_gk20a_intr_stall(g->dev->name);
+        trace_mc_gk20a_intr_stall(dev_name(g->dev));
        if (!g->power_on)
                return IRQ_NONE;
@@ -42,7 +42,7 @@ irqreturn_t mc_gk20a_isr_stall(struct gk20a *g)
        atomic_inc(&g->hw_irq_stall_count);
-        trace_mc_gk20a_intr_stall_done(g->dev->name);
+        trace_mc_gk20a_intr_stall_done(dev_name(g->dev));
        return IRQ_WAKE_THREAD;
 }
@@ -77,7 +77,7 @@ irqreturn_t mc_gk20a_intr_thread_stall(struct gk20a *g)
        gk20a_dbg(gpu_dbg_intr, "interrupt thread launched");
-        trace_mc_gk20a_intr_thread_stall(g->dev->name);
+        trace_mc_gk20a_intr_thread_stall(dev_name(g->dev));
        mc_intr_0 = gk20a_readl(g, mc_intr_0_r());
        hw_irq_count = atomic_read(&g->hw_irq_stall_count);
@@ -111,7 +111,7 @@ irqreturn_t mc_gk20a_intr_thread_stall(struct gk20a *g)
        wake_up_all(&g->sw_irq_stall_last_handled_wq);
-        trace_mc_gk20a_intr_thread_stall_done(g->dev->name);
+        trace_mc_gk20a_intr_thread_stall_done(dev_name(g->dev));
        return IRQ_HANDLED;
 }
diff --git a/drivers/gpu/nvgpu/gk20a/mm_gk20a.c b/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
index 7a02d68e..ac4625e0 100644
--- a/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
@@ -3515,7 +3515,7 @@ int gk20a_mm_fb_flush(struct gk20a *g)
        gk20a_busy_noresume(g->dev);
        if (!g->power_on) {
-                pm_runtime_put_noidle(&g->dev->dev);
+                pm_runtime_put_noidle(g->dev);
                return 0;
        }
@@ -3525,7 +3525,7 @@ int gk20a_mm_fb_flush(struct gk20a *g)
           guarantee that writes are to DRAM. This will be a sysmembar internal
           to the L2. */
-        trace_gk20a_mm_fb_flush(g->dev->name);
+        trace_gk20a_mm_fb_flush(dev_name(g->dev));
        gk20a_writel(g, flush_fb_flush_r(),
                flush_fb_flush_pending_busy_f());
@@ -3552,11 +3552,11 @@ int gk20a_mm_fb_flush(struct gk20a *g)
                ret = -EBUSY;
        }
-        trace_gk20a_mm_fb_flush_done(g->dev->name);
+        trace_gk20a_mm_fb_flush_done(dev_name(g->dev));
        mutex_unlock(&mm->l2_op_lock);
-        pm_runtime_put_noidle(&g->dev->dev);
+        pm_runtime_put_noidle(g->dev);
        return ret;
 }
@@ -3566,7 +3566,7 @@ static void gk20a_mm_l2_invalidate_locked(struct gk20a *g)
        u32 data;
        s32 retry = 200;
-        trace_gk20a_mm_l2_invalidate(g->dev->name);
+        trace_gk20a_mm_l2_invalidate(dev_name(g->dev));
        /* Invalidate any clean lines from the L2 so subsequent reads go to
           DRAM. Dirty lines are not affected by this operation. */
@@ -3592,7 +3592,7 @@ static void gk20a_mm_l2_invalidate_locked(struct gk20a *g)
                gk20a_warn(dev_from_gk20a(g),
                        "l2_system_invalidate too many retries");
-        trace_gk20a_mm_l2_invalidate_done(g->dev->name);
+        trace_gk20a_mm_l2_invalidate_done(dev_name(g->dev));
 }
 void gk20a_mm_l2_invalidate(struct gk20a *g)
@@ -3604,7 +3604,7 @@ void gk20a_mm_l2_invalidate(struct gk20a *g)
                gk20a_mm_l2_invalidate_locked(g);
                mutex_unlock(&mm->l2_op_lock);
        }
-        pm_runtime_put_noidle(&g->dev->dev);
+        pm_runtime_put_noidle(g->dev);
 }
 void gk20a_mm_l2_flush(struct gk20a *g, bool invalidate)
@@ -3621,7 +3621,7 @@ void gk20a_mm_l2_flush(struct gk20a *g, bool invalidate)
        mutex_lock(&mm->l2_op_lock);
-        trace_gk20a_mm_l2_flush(g->dev->name);
+        trace_gk20a_mm_l2_flush(dev_name(g->dev));
        /* Flush all dirty lines from the L2 to DRAM. Lines are left in the L2
           as clean, so subsequent reads might hit in the L2. */
@@ -3646,7 +3646,7 @@ void gk20a_mm_l2_flush(struct gk20a *g, bool invalidate)
                gk20a_warn(dev_from_gk20a(g),
                        "l2_flush_dirty too many retries");
-        trace_gk20a_mm_l2_flush_done(g->dev->name);
+        trace_gk20a_mm_l2_flush_done(dev_name(g->dev));
        if (invalidate)
                gk20a_mm_l2_invalidate_locked(g);
@@ -3654,7 +3654,7 @@ void gk20a_mm_l2_flush(struct gk20a *g, bool invalidate)
        mutex_unlock(&mm->l2_op_lock);
 hw_was_off:
-        pm_runtime_put_noidle(&g->dev->dev);
+        pm_runtime_put_noidle(g->dev);
 }
 void gk20a_mm_cbc_clean(struct gk20a *g)
@@ -3696,7 +3696,7 @@ void gk20a_mm_cbc_clean(struct gk20a *g)
        mutex_unlock(&mm->l2_op_lock);
 hw_was_off:
-        pm_runtime_put_noidle(&g->dev->dev);
+        pm_runtime_put_noidle(g->dev);
 }
 int gk20a_vm_find_buffer(struct vm_gk20a *vm, u64 gpu_va,
@@ -3746,7 +3746,7 @@ void gk20a_mm_tlb_invalidate(struct vm_gk20a *vm)
        mutex_lock(&tlb_lock);
-        trace_gk20a_mm_tlb_invalidate(g->dev->name);
+        trace_gk20a_mm_tlb_invalidate(dev_name(g->dev));
        do {
                data = gk20a_readl(g, fb_mmu_ctrl_r());
@@ -3783,7 +3783,7 @@ void gk20a_mm_tlb_invalidate(struct vm_gk20a *vm)
                gk20a_warn(dev_from_gk20a(g),
                        "mmu invalidate too many retries");
-        trace_gk20a_mm_tlb_invalidate_done(g->dev->name);
+        trace_gk20a_mm_tlb_invalidate_done(dev_name(g->dev));
 out:
        mutex_unlock(&tlb_lock);
@@ -3868,11 +3868,11 @@ clean_up:
        return err;
 }
-void gk20a_mm_debugfs_init(struct platform_device *pdev)
+void gk20a_mm_debugfs_init(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct dentry *gpu_root = platform->debugfs;
-        struct gk20a *g = gk20a_get_platform(pdev)->g;
+        struct gk20a *g = gk20a_get_platform(dev)->g;
        debugfs_create_x64("separate_fixed_allocs", 0664, gpu_root,
                           &g->separate_fixed_allocs);
diff --git a/drivers/gpu/nvgpu/gk20a/mm_gk20a.h b/drivers/gpu/nvgpu/gk20a/mm_gk20a.h
index 833d896d..52d6c4e5 100644
--- a/drivers/gpu/nvgpu/gk20a/mm_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/mm_gk20a.h
@@ -288,7 +288,7 @@ struct channel_gk20a;
 int gk20a_init_mm_support(struct gk20a *g);
 int gk20a_init_mm_setup_sw(struct gk20a *g);
 int gk20a_init_mm_setup_hw(struct gk20a *g);
-void gk20a_mm_debugfs_init(struct platform_device *pdev);
+void gk20a_mm_debugfs_init(struct device *dev);
 int gk20a_mm_fb_flush(struct gk20a *g);
 void gk20a_mm_l2_flush(struct gk20a *g, bool invalidate);
diff --git a/drivers/gpu/nvgpu/gk20a/platform_gk20a.h b/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
index 6bffed9e..b2f3b870 100644
--- a/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
@@ -1,9 +1,7 @@
 /*
- * drivers/video/tegra/host/gk20a/soc/platform_gk20a.h
- *
 * GK20A Platform (SoC) Interface
 *
- * Copyright (c) 2014-2015, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -18,7 +16,7 @@
 #ifndef _GK20A_PLATFORM_H_
 #define _GK20A_PLATFORM_H_
-#include <linux/platform_device.h>
+#include <linux/device.h>
 #include <linux/pm_domain.h>
 #include <linux/dma-attrs.h>
@@ -28,7 +26,7 @@ struct gr_ctx_buffer_desc;
 struct gk20a_scale_profile;
 struct secure_page_buffer {
-        void (*destroy)(struct platform_device *, struct secure_page_buffer *);
+        void (*destroy)(struct device *, struct secure_page_buffer *);
        size_t size;
        u64 iova;
        struct dma_attrs attrs;
@@ -114,28 +112,28 @@ struct gk20a_platform {
         * After this function is finished, the driver will initialise
         * pm runtime and genpd based on the platform configuration.
         */
-        int (*probe)(struct platform_device *dev);
+        int (*probe)(struct device *dev);
        /* Second stage initialisation - called once all power management
         * initialisations are done.
         */
-        int (*late_probe)(struct platform_device *dev);
+        int (*late_probe)(struct device *dev);
        /* Remove device after power management has been done
         */
-        int (*remove)(struct platform_device *dev);
+        int (*remove)(struct device *dev);
        /* Poweron platform dependencies */
-        int (*busy)(struct platform_device *dev);
+        int (*busy)(struct device *dev);
        /* Powerdown platform dependencies */
-        void (*idle)(struct platform_device *dev);
+        void (*idle)(struct device *dev);
        /* This function is called to allocate secure memory (memory that the
         * CPU cannot see). The function should fill the context buffer
         * descriptor (especially fields destroy, sgt, size).
         */
-        int (*secure_alloc)(struct platform_device *dev,
+        int (*secure_alloc)(struct device *dev,
                            struct gr_ctx_buffer_desc *desc,
                            size_t size);
@@ -143,7 +141,7 @@ struct gk20a_platform {
         * This is also helpful to trigger secure memory resizing
         * while GPU is off
         */
-        int (*secure_page_alloc)(struct platform_device *dev);
+        int (*secure_page_alloc)(struct device *dev);
        struct secure_page_buffer secure_buffer;
        bool secure_alloc_ready;
@@ -151,33 +149,33 @@ struct gk20a_platform {
        int (*suspend)(struct device *);
        /* Called to turn off the device */
-        int (*railgate)(struct platform_device *dev);
+        int (*railgate)(struct device *dev);
        /* Called to turn on the device */
-        int (*unrailgate)(struct platform_device *dev);
+        int (*unrailgate)(struct device *dev);
        struct mutex railgate_lock;
        /* Called to check state of device */
-        bool (*is_railgated)(struct platform_device *dev);
+        bool (*is_railgated)(struct device *dev);
        /* get supported frequency list */
-        int (*get_clk_freqs)(struct platform_device *pdev,
+        int (*get_clk_freqs)(struct device *pdev,
                                unsigned long **freqs, int *num_freqs);
        /* clk related supported functions */
-        unsigned long (*clk_get_rate)(struct platform_device *pdev);
+        unsigned long (*clk_get_rate)(struct device *dev);
-        long (*clk_round_rate)(struct platform_device *pdev,
+        long (*clk_round_rate)(struct device *dev,
                                unsigned long rate);
-        int (*clk_set_rate)(struct platform_device *pdev,
+        int (*clk_set_rate)(struct device *dev,
                                unsigned long rate);
        /* Postscale callback is called after frequency change */
-        void (*postscale)(struct platform_device *pdev,
+        void (*postscale)(struct device *dev,
                          unsigned long freq);
        /* Pre callback is called before frequency change */
-        void (*prescale)(struct platform_device *pdev);
+        void (*prescale)(struct device *dev);
        /* Devfreq governor name. If scaling is enabled, we request
         * this governor to be used in scaling */
@@ -193,11 +191,11 @@ struct gk20a_platform {
         * hw units which may interact with the gpu without direct supervision
         * of the CPU.
         */
-        void (*dump_platform_dependencies)(struct platform_device *dev);
+        void (*dump_platform_dependencies)(struct device *dev);
        /* Callbacks to assert/deassert GPU reset */
-        int (*reset_assert)(struct platform_device *pdev);
+        int (*reset_assert)(struct device *dev);
-        int (*reset_deassert)(struct platform_device *pdev);
+        int (*reset_deassert)(struct device *dev);
        struct clk *clk_reset;
        struct dvfs_rail *gpu_rail;
@@ -210,12 +208,15 @@ struct gk20a_platform {
        u32 ptimer_src_freq;
        bool has_cde;
+        /* soc name for finding firmware files */
+        const char *soc_name;
 };
 static inline struct gk20a_platform *gk20a_get_platform(
-                struct platform_device *dev)
+                struct device *dev)
 {
-        return (struct gk20a_platform *)platform_get_drvdata(dev);
+        return (struct gk20a_platform *)dev_get_drvdata(dev);
 }
 extern struct gk20a_platform gk20a_generic_platform;
@@ -227,14 +228,14 @@ extern struct gk20a_platform vgpu_tegra_platform;
 #endif
 #endif
-static inline bool gk20a_platform_has_syncpoints(struct platform_device *dev)
+static inline bool gk20a_platform_has_syncpoints(struct device *dev)
 {
-        struct gk20a_platform *p = gk20a_get_platform(dev);
+        struct gk20a_platform *p = dev_get_drvdata(dev);
        return p->has_syncpoints;
 }
-int gk20a_tegra_busy(struct platform_device *dev);
+int gk20a_tegra_busy(struct device *dev);
-void gk20a_tegra_idle(struct platform_device *dev);
+void gk20a_tegra_idle(struct device *dev);
-void gk20a_tegra_debug_dump(struct platform_device *pdev);
+void gk20a_tegra_debug_dump(struct device *pdev);
 #endif
diff --git a/drivers/gpu/nvgpu/gk20a/platform_gk20a_generic.c b/drivers/gpu/nvgpu/gk20a/platform_gk20a_generic.c
index 74e59527..33e2fb37 100644
--- a/drivers/gpu/nvgpu/gk20a/platform_gk20a_generic.c
+++ b/drivers/gpu/nvgpu/gk20a/platform_gk20a_generic.c
@@ -31,9 +31,9 @@
 * the clock information to gk20a platform data.
 */
-static int gk20a_generic_get_clocks(struct platform_device *pdev)
+static int gk20a_generic_get_clocks(struct device *pdev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(pdev);
        platform->clk[0] = clk_get_sys("tegra_gk20a.0", "PLLG_ref");
        platform->clk[1] = clk_get_sys("tegra_gk20a.0", "pwr");
@@ -65,19 +65,19 @@ err_get_clock:
        return -ENODEV;
 }
-static int gk20a_generic_probe(struct platform_device *dev)
+static int gk20a_generic_probe(struct device *dev)
 {
        gk20a_generic_get_clocks(dev);
        return 0;
 }
-static int gk20a_generic_late_probe(struct platform_device *dev)
+static int gk20a_generic_late_probe(struct device *dev)
 {
        return 0;
 }
-static int gk20a_generic_remove(struct platform_device *dev)
+static int gk20a_generic_remove(struct device *dev)
 {
        return 0;
 }
diff --git a/drivers/gpu/nvgpu/gk20a/platform_gk20a_tegra.c b/drivers/gpu/nvgpu/gk20a/platform_gk20a_tegra.c
index 15d6609d..468c9257 100644
--- a/drivers/gpu/nvgpu/gk20a/platform_gk20a_tegra.c
+++ b/drivers/gpu/nvgpu/gk20a/platform_gk20a_tegra.c
@@ -1,6 +1,4 @@
 /*
- * drivers/video/tegra/host/gk20a/platform_gk20a_tegra.c
- *
 * GK20A Tegra Platform Interface
 *
 * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
@@ -66,7 +64,7 @@ static inline void pmc_write(u32 val, unsigned long reg)
 #define MHZ_TO_HZ(x) ((x) * 1000000)
 #define HZ_TO_MHZ(x) ((x) / 1000000)
-static void gk20a_tegra_secure_page_destroy(struct platform_device *pdev,
+static void gk20a_tegra_secure_page_destroy(struct device *dev,
                                       struct secure_page_buffer *secure_buffer)
 {
        dma_free_attrs(&tegra_vpr_dev, secure_buffer->size,
@@ -74,9 +72,9 @@ static void gk20a_tegra_secure_page_destroy(struct platform_device *pdev,
                        secure_buffer->iova, &secure_buffer->attrs);
 }
-int gk20a_tegra_secure_page_alloc(struct platform_device *pdev)
+int gk20a_tegra_secure_page_alloc(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct secure_page_buffer *secure_buffer = &platform->secure_buffer;
        DEFINE_DMA_ATTRS(attrs);
        dma_addr_t iova;
@@ -113,12 +111,11 @@ static void gk20a_tegra_secure_destroy(struct gk20a *g,
        }
 }
-int gk20a_tegra_secure_alloc(struct platform_device *pdev,
+int gk20a_tegra_secure_alloc(struct device *dev,
                             struct gr_ctx_buffer_desc *desc,
                             size_t size)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
-        struct device *dev = &pdev->dev;
        DEFINE_DMA_ATTRS(attrs);
        dma_addr_t iova;
        struct sg_table *sgt;
@@ -198,13 +195,13 @@ static unsigned long gk20a_tegra_get_emc_rate(struct gk20a *g,
 * This function sets emc frequency based on current gpu frequency
 */
-static void gk20a_tegra_postscale(struct platform_device *pdev,
+static void gk20a_tegra_postscale(struct device *dev,
                                  unsigned long freq)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct gk20a_scale_profile *profile = platform->g->scale_profile;
        struct gk20a_emc_params *emc_params = profile->private_data;
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
        struct clk *emc_clk = platform->clk[2];
        enum tegra_chipid chip_id = tegra_get_chip_id();
        unsigned long emc_target;
@@ -257,9 +254,9 @@ static void gk20a_tegra_postscale(struct platform_device *pdev,
 * This function informs EDP about changed constraints.
 */
-static void gk20a_tegra_prescale(struct platform_device *pdev)
+static void gk20a_tegra_prescale(struct device *dev)
 {
-        struct gk20a *g = get_gk20a(pdev);
+        struct gk20a *g = get_gk20a(dev);
        u32 avg = 0;
        gk20a_pmu_load_norm(g, &avg);
@@ -271,7 +268,7 @@ static void gk20a_tegra_prescale(struct platform_device *pdev)
 *
 */
-static void gk20a_tegra_calibrate_emc(struct platform_device *pdev,
+static void gk20a_tegra_calibrate_emc(struct device *dev,
                               struct gk20a_emc_params *emc_params)
 {
        enum tegra_chipid cid = tegra_get_chipid();
@@ -308,9 +305,9 @@ static void gk20a_tegra_calibrate_emc(struct platform_device *pdev,
 * Check status of gk20a power rail
 */
-static bool gk20a_tegra_is_railgated(struct platform_device *pdev)
+static bool gk20a_tegra_is_railgated(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        bool ret = false;
        if (!tegra_platform_is_linsim())
@@ -325,9 +322,9 @@ static bool gk20a_tegra_is_railgated(struct platform_device *pdev)
 * Gate (disable) gk20a power rail
 */
-static int gk20a_tegra_railgate(struct platform_device *pdev)
+static int gk20a_tegra_railgate(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        int ret = 0;
        if (tegra_platform_is_linsim() ||
@@ -344,7 +341,7 @@ static int gk20a_tegra_railgate(struct platform_device *pdev)
        udelay(10);
-        platform->reset_assert(pdev);
+        platform->reset_assert(dev);
        udelay(10);
@@ -367,7 +364,7 @@ static int gk20a_tegra_railgate(struct platform_device *pdev)
        return 0;
 err_power_off:
-        gk20a_err(&pdev->dev, "Could not railgate GPU");
+        gk20a_err(dev, "Could not railgate GPU");
        return ret;
 }
@@ -377,9 +374,9 @@ err_power_off:
 * Gate (disable) gm20b power rail
 */
-static int gm20b_tegra_railgate(struct platform_device *pdev)
+static int gm20b_tegra_railgate(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        int ret = 0;
        if (tegra_platform_is_linsim() ||
@@ -396,7 +393,7 @@ static int gm20b_tegra_railgate(struct platform_device *pdev)
        udelay(10);
-        platform->reset_assert(pdev);
+        platform->reset_assert(dev);
        udelay(10);
@@ -422,7 +419,7 @@ static int gm20b_tegra_railgate(struct platform_device *pdev)
        return 0;
 err_power_off:
-        gk20a_err(&pdev->dev, "Could not railgate GPU");
+        gk20a_err(dev, "Could not railgate GPU");
        return ret;
 }
@@ -432,9 +429,9 @@ err_power_off:
 * Ungate (enable) gk20a power rail
 */
-static int gk20a_tegra_unrailgate(struct platform_device *pdev)
+static int gk20a_tegra_unrailgate(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        int ret = 0;
        bool first = false;
@@ -457,19 +454,19 @@ static int gk20a_tegra_unrailgate(struct platform_device *pdev)
        if (!first) {
                ret = clk_enable(platform->clk[0]);
                if (ret) {
-                        gk20a_err(&pdev->dev, "could not turn on gpu pll");
+                        gk20a_err(dev, "could not turn on gpu pll");
                        goto err_clk_on;
                }
                ret = clk_enable(platform->clk[1]);
                if (ret) {
-                        gk20a_err(&pdev->dev, "could not turn on pwr clock");
+                        gk20a_err(dev, "could not turn on pwr clock");
                        goto err_clk_on;
                }
        }
        udelay(10);
-        platform->reset_assert(pdev);
+        platform->reset_assert(dev);
        udelay(10);
@@ -478,7 +475,7 @@ static int gk20a_tegra_unrailgate(struct platform_device *pdev)
        udelay(10);
-        platform->reset_deassert(pdev);
+        platform->reset_deassert(dev);
        /* Flush MC after boot/railgate/SC7 */
        tegra_mc_flush(MC_CLIENT_GPU);
@@ -503,9 +500,9 @@ err_clk_on:
 * Ungate (enable) gm20b power rail
 */
-static int gm20b_tegra_unrailgate(struct platform_device *pdev)
+static int gm20b_tegra_unrailgate(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        int ret = 0;
        bool first = false;
@@ -530,25 +527,25 @@ static int gm20b_tegra_unrailgate(struct platform_device *pdev)
        if (!first) {
                ret = clk_enable(platform->clk_reset);
                if (ret) {
-                        gk20a_err(&pdev->dev, "could not turn on gpu_gate");
+                        gk20a_err(dev, "could not turn on gpu_gate");
                        goto err_clk_on;
                }
                ret = clk_enable(platform->clk[0]);
                if (ret) {
-                        gk20a_err(&pdev->dev, "could not turn on gpu pll");
+                        gk20a_err(dev, "could not turn on gpu pll");
                        goto err_clk_on;
                }
                ret = clk_enable(platform->clk[1]);
                if (ret) {
-                        gk20a_err(&pdev->dev, "could not turn on pwr clock");
+                        gk20a_err(dev, "could not turn on pwr clock");
                        goto err_clk_on;
                }
        }
        udelay(10);
-        platform->reset_assert(pdev);
+        platform->reset_assert(dev);
        udelay(10);
@@ -558,7 +555,7 @@ static int gm20b_tegra_unrailgate(struct platform_device *pdev)
        udelay(10);
        clk_disable(platform->clk_reset);
-        platform->reset_deassert(pdev);
+        platform->reset_deassert(dev);
        clk_enable(platform->clk_reset);
        /* Flush MC after boot/railgate/SC7 */
@@ -594,16 +591,14 @@ static struct {
 * the clock information to gk20a platform data.
 */
-static int gk20a_tegra_get_clocks(struct platform_device *pdev)
+static int gk20a_tegra_get_clocks(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        char devname[16];
        int i;
        int ret = 0;
-        snprintf(devname, sizeof(devname),
+        snprintf(devname, sizeof(devname), "tegra_%s", dev_name(dev));
-                 (pdev->id <= 0) ? "tegra_%s" : "tegra_%s.%d\n",
-                 pdev->name, pdev->id);
        platform->num_clks = 0;
        for (i = 0; i < ARRAY_SIZE(tegra_gk20a_clocks); i++) {
@@ -630,7 +625,7 @@ err_get_clock:
        return ret;
 }
-static int gk20a_tegra_reset_assert(struct platform_device *dev)
+static int gk20a_tegra_reset_assert(struct device *dev)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
@@ -642,7 +637,7 @@ static int gk20a_tegra_reset_assert(struct platform_device *dev)
        return 0;
 }
-static int gk20a_tegra_reset_deassert(struct platform_device *dev)
+static int gk20a_tegra_reset_deassert(struct device *dev)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
@@ -654,14 +649,14 @@ static int gk20a_tegra_reset_deassert(struct platform_device *dev)
        return 0;
 }
-static int gm20b_tegra_reset_assert(struct platform_device *dev)
+static int gm20b_tegra_reset_assert(struct device *dev)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
        if (!platform->clk_reset) {
-                platform->clk_reset = clk_get(&dev->dev, "gpu_gate");
+                platform->clk_reset = clk_get(dev, "gpu_gate");
                if (IS_ERR(platform->clk_reset)) {
-                        gk20a_err(&dev->dev, "fail to get gpu reset clk\n");
+                        gk20a_err(dev, "fail to get gpu reset clk\n");
                        return PTR_ERR(platform->clk_reset);
                }
        }
@@ -671,9 +666,9 @@ static int gm20b_tegra_reset_assert(struct platform_device *dev)
        return 0;
 }
-static void gk20a_tegra_scale_init(struct platform_device *pdev)
+static void gk20a_tegra_scale_init(struct device *dev)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(pdev);
+        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a_scale_profile *profile = platform->g->scale_profile;
        struct gk20a_emc_params *emc_params;
@@ -685,28 +680,28 @@ static void gk20a_tegra_scale_init(struct platform_device *pdev)
                return;
        emc_params->freq_last_set = -1;
-        gk20a_tegra_calibrate_emc(pdev, emc_params);
+        gk20a_tegra_calibrate_emc(dev, emc_params);
        profile->private_data = emc_params;
 }
-static void gk20a_tegra_scale_exit(struct platform_device *pdev)
+static void gk20a_tegra_scale_exit(struct device *dev)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct gk20a_scale_profile *profile = platform->g->scale_profile;
        if (profile)
                kfree(profile->private_data);
 }
-void gk20a_tegra_debug_dump(struct platform_device *pdev)
+void gk20a_tegra_debug_dump(struct device *dev)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(pdev);
+        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a *g = platform->g;
        nvhost_debug_dump_device(g->host1x_dev);
 }
-int gk20a_tegra_busy(struct platform_device *dev)
+int gk20a_tegra_busy(struct device *dev)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a *g = platform->g;
@@ -716,7 +711,7 @@ int gk20a_tegra_busy(struct platform_device *dev)
        return 0;
 }
-void gk20a_tegra_idle(struct platform_device *dev)
+void gk20a_tegra_idle(struct device *dev)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a *g = platform->g;
@@ -725,10 +720,10 @@ void gk20a_tegra_idle(struct platform_device *dev)
                nvhost_module_idle_ext(g->host1x_dev);
 }
-static int gk20a_tegra_probe(struct platform_device *dev)
+static int gk20a_tegra_probe(struct device *dev)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
-        struct device_node *np = dev->dev.of_node;
+        struct device_node *np = dev->of_node;
        const __be32 *host1x_ptr;
        struct platform_device *host1x_pdev = NULL;
@@ -739,13 +734,13 @@ static int gk20a_tegra_probe(struct platform_device *dev)
                host1x_pdev = of_find_device_by_node(host1x_node);
                if (!host1x_pdev) {
-                        dev_warn(&dev->dev, "host1x device not available");
+                        dev_warn(dev, "host1x device not available");
                        return -EPROBE_DEFER;
                }
        } else {
-                host1x_pdev = to_platform_device(dev->dev.parent);
+                host1x_pdev = to_platform_device(dev->parent);
-                dev_warn(&dev->dev, "host1x reference not found. assuming host1x to be parent");
+                dev_warn(dev, "host1x reference not found. assuming host1x to be parent");
        }
        platform->g->host1x_dev = host1x_pdev;
@@ -761,7 +756,7 @@ static int gk20a_tegra_probe(struct platform_device *dev)
                np = of_find_node_by_path("/gpu-dvfs-rework");
                if (!(np && of_device_is_available(np))) {
                        platform->devfreq_governor = "";
-                        dev_warn(&dev->dev, "board does not support scaling");
+                        dev_warn(dev, "board does not support scaling");
                }
        }
@@ -770,10 +765,10 @@ static int gk20a_tegra_probe(struct platform_device *dev)
        return 0;
 }
-static int gk20a_tegra_late_probe(struct platform_device *dev)
+static int gk20a_tegra_late_probe(struct device *dev)
 {
        /* Make gk20a power domain a subdomain of host1x */
-        nvhost_register_client_domain(dev_to_genpd(&dev->dev));
+        nvhost_register_client_domain(dev_to_genpd(dev));
        /* Initialise tegra specific scaling quirks */
        gk20a_tegra_scale_init(dev);
@@ -781,15 +776,15 @@ static int gk20a_tegra_late_probe(struct platform_device *dev)
        return 0;
 }
-static int gk20a_tegra_remove(struct platform_device *dev)
+static int gk20a_tegra_remove(struct device *dev)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        if (platform->g->host1x_dev)
                nvhost_unregister_dump_device(platform->g->host1x_dev);
        /* remove gk20a power subdomain from host1x */
-        nvhost_unregister_client_domain(dev_to_genpd(&dev->dev));
+        nvhost_unregister_client_domain(dev_to_genpd(dev));
        /* deinitialise tegra specific scaling quirks */
        gk20a_tegra_scale_exit(dev);
@@ -804,7 +799,7 @@ static int gk20a_tegra_suspend(struct device *dev)
 }
 #ifdef CONFIG_TEGRA_CLK_FRAMEWORK
-static unsigned long gk20a_get_clk_rate(struct platform_device *dev)
+static unsigned long gk20a_get_clk_rate(struct device *dev)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a *g = platform->g;
@@ -813,8 +808,7 @@ static unsigned long gk20a_get_clk_rate(struct platform_device *dev)
 }
-static long gk20a_round_clk_rate(struct platform_device *dev,
+static long gk20a_round_clk_rate(struct device *dev, unsigned long rate)
-                                                unsigned long rate)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a *g = platform->g;
@@ -822,7 +816,7 @@ static long gk20a_round_clk_rate(struct platform_device *dev,
        return gk20a_clk_round_rate(g, rate);
 }
-static int gk20a_set_clk_rate(struct platform_device *dev, unsigned long rate)
+static int gk20a_set_clk_rate(struct device *dev, unsigned long rate)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
        struct gk20a *g = platform->g;
@@ -830,7 +824,7 @@ static int gk20a_set_clk_rate(struct platform_device *dev, unsigned long rate)
        return gk20a_clk_set_rate(g, rate);
 }
-static int gk20a_clk_get_freqs(struct platform_device *dev,
+static int gk20a_clk_get_freqs(struct device *dev,
                                unsigned long **freqs, int *num_freqs)
 {
        struct gk20a_platform *platform = gk20a_get_platform(dev);
@@ -900,6 +894,8 @@ struct gk20a_platform gk20a_tegra_platform = {
        .secure_alloc = gk20a_tegra_secure_alloc,
        .secure_page_alloc = gk20a_tegra_secure_page_alloc,
        .dump_platform_dependencies = gk20a_tegra_debug_dump,
+        .soc_name = "tegra12x",
 };
 struct gk20a_platform gm20b_tegra_platform = {
@@ -958,4 +954,6 @@ struct gk20a_platform gm20b_tegra_platform = {
        .dump_platform_dependencies = gk20a_tegra_debug_dump,
        .has_cde = true,
+        .soc_name = "tegra21x",
 };
diff --git a/drivers/gpu/nvgpu/gk20a/platform_vgpu_tegra.c b/drivers/gpu/nvgpu/gk20a/platform_vgpu_tegra.c
index ed65756a..bb8d05b8 100644
--- a/drivers/gpu/nvgpu/gk20a/platform_vgpu_tegra.c
+++ b/drivers/gpu/nvgpu/gk20a/platform_vgpu_tegra.c
@@ -1,7 +1,7 @@
 /*
 * Tegra Virtualized GPU Platform Interface
 *
- * Copyright (c) 2014, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -19,10 +19,10 @@
 #include "hal_gk20a.h"
 #include "platform_gk20a.h"
-static int gk20a_tegra_probe(struct platform_device *dev)
+static int gk20a_tegra_probe(struct device *dev)
 {
-        struct gk20a_platform *platform = gk20a_get_platform(dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
-        struct device_node *np = dev->dev.of_node;
+        struct device_node *np = dev->of_node;
        const __be32 *host1x_ptr;
        struct platform_device *host1x_pdev = NULL;
@@ -33,13 +33,13 @@ static int gk20a_tegra_probe(struct platform_device *dev)
                host1x_pdev = of_find_device_by_node(host1x_node);
                if (!host1x_pdev) {
-                        dev_warn(&dev->dev, "host1x device not available");
+                        dev_warn(dev, "host1x device not available");
                        return -EPROBE_DEFER;
                }
        } else {
-                host1x_pdev = to_platform_device(dev->dev.parent);
+                host1x_pdev = to_platform_device(dev->parent);
-                dev_warn(&dev->dev, "host1x reference not found. assuming host1x to be parent");
+                dev_warn(dev, "host1x reference not found. assuming host1x to be parent");
        }
        platform->g->host1x_dev = host1x_pdev;
diff --git a/drivers/gpu/nvgpu/gk20a/pmu_gk20a.c b/drivers/gpu/nvgpu/gk20a/pmu_gk20a.c
index 4edfe90c..3b154f30 100644
--- a/drivers/gpu/nvgpu/gk20a/pmu_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/pmu_gk20a.c
@@ -1814,7 +1814,7 @@ int pmu_reset(struct pmu_gk20a *pmu)
 int pmu_bootstrap(struct pmu_gk20a *pmu)
 {
        struct gk20a *g = gk20a_from_pmu(pmu);
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        struct mm_gk20a *mm = &g->mm;
        struct pmu_ucode_desc *desc = pmu->desc;
        u64 addr_code, addr_data, addr_load;
@@ -2955,7 +2955,7 @@ static u8 get_perfmon_id(struct pmu_gk20a *pmu)
                break;
 #endif
        default:
-                gk20a_err(&g->dev->dev, "no support for %x", ver);
+                gk20a_err(g->dev, "no support for %x", ver);
                BUG();
        }
@@ -4667,10 +4667,10 @@ static const struct file_operations security_fops = {
        .release        = single_release,
 };
-int gk20a_pmu_debugfs_init(struct platform_device *dev)
+int gk20a_pmu_debugfs_init(struct device *dev)
 {
        struct dentry *d;
-        struct gk20a_platform *platform = platform_get_drvdata(dev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct gk20a *g = get_gk20a(dev);
        d = debugfs_create_file(
diff --git a/drivers/gpu/nvgpu/gk20a/pmu_gk20a.h b/drivers/gpu/nvgpu/gk20a/pmu_gk20a.h
index c533ba8d..4edb3283 100644
--- a/drivers/gpu/nvgpu/gk20a/pmu_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/pmu_gk20a.h
@@ -1394,7 +1394,7 @@ int pmu_mutex_release(struct pmu_gk20a *pmu, u32 id, u32 *token);
 int gk20a_pmu_destroy(struct gk20a *g);
 int gk20a_pmu_load_norm(struct gk20a *g, u32 *load);
 int gk20a_pmu_load_update(struct gk20a *g);
-int gk20a_pmu_debugfs_init(struct platform_device *dev);
+int gk20a_pmu_debugfs_init(struct device *dev);
 void gk20a_pmu_reset_load_counters(struct gk20a *g);
 void gk20a_pmu_get_load_counters(struct gk20a *g, u32 *busy_cycles,
                u32 *total_cycles);
diff --git a/drivers/gpu/nvgpu/gk20a/regops_gk20a.c b/drivers/gpu/nvgpu/gk20a/regops_gk20a.c
index a9c7fdfd..fda480cf 100644
--- a/drivers/gpu/nvgpu/gk20a/regops_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/regops_gk20a.c
@@ -400,7 +400,7 @@ int exec_regops_gk20a(struct dbg_session_gk20a *dbg_s,
         * regops implementation, so we return -ENOSYS. This will allow
         * compute apps to run with vgpu. Tools will not work in this
         * configuration and are not required to work at this time. */
-        if (gk20a_gpu_is_virtual(dbg_s->pdev))
+        if (gk20a_gpu_is_virtual(dbg_s->dev))
                return -ENOSYS;
        ok = validate_reg_ops(dbg_s,
diff --git a/drivers/gpu/nvgpu/gk20a/tsg_gk20a.c b/drivers/gpu/nvgpu/gk20a/tsg_gk20a.c
index 51e6407c..7b98c80f 100644
--- a/drivers/gpu/nvgpu/gk20a/tsg_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/tsg_gk20a.c
@@ -470,7 +470,7 @@ long gk20a_tsg_dev_ioctl(struct file *filp, unsigned int cmd,
                {
                err = gk20a_busy(g->dev);
                if (err) {
-                        gk20a_err(&g->dev->dev,
+                        gk20a_err(g->dev,
                           "failed to host gk20a for ioctl cmd: 0x%x", cmd);
                        return err;
                }
@@ -483,7 +483,7 @@ long gk20a_tsg_dev_ioctl(struct file *filp, unsigned int cmd,
                {
                err = gk20a_busy(g->dev);
                if (err) {
-                        gk20a_err(&g->dev->dev,
+                        gk20a_err(g->dev,
                           "failed to host gk20a for ioctl cmd: 0x%x", cmd);
                        return err;
                }
@@ -496,7 +496,7 @@ long gk20a_tsg_dev_ioctl(struct file *filp, unsigned int cmd,
                {
                err = gk20a_busy(g->dev);
                if (err) {
-                        gk20a_err(&g->dev->dev,
+                        gk20a_err(g->dev,
                           "failed to host gk20a for ioctl cmd: 0x%x", cmd);
                        return err;
                }
diff --git a/drivers/gpu/nvgpu/gm20b/acr_gm20b.c b/drivers/gpu/nvgpu/gm20b/acr_gm20b.c
index 8a0be106..79f90f0b 100644
--- a/drivers/gpu/nvgpu/gm20b/acr_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/acr_gm20b.c
@@ -1257,7 +1257,7 @@ static int gm20b_init_pmu_setup_hw1(struct gk20a *g,
        struct pmu_gk20a *pmu = &g->pmu;
        int err;
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        gk20a_dbg_fn("");
diff --git a/drivers/gpu/nvgpu/gm20b/clk_gm20b.c b/drivers/gpu/nvgpu/gm20b/clk_gm20b.c
index 16f929a8..54406008 100644
--- a/drivers/gpu/nvgpu/gm20b/clk_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/clk_gm20b.c
@@ -1,7 +1,7 @@
 /*
 * GM20B Clocks
 *
- * Copyright (c) 2014, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -1651,7 +1651,7 @@ static const struct file_operations pll_param_fops = {
 static int clk_gm20b_debugfs_init(struct gk20a *g)
 {
        struct dentry *d;
-        struct gk20a_platform *platform = platform_get_drvdata(g->dev);
+        struct gk20a_platform *platform = dev_get_drvdata(g->dev);
        d = debugfs_create_file(
                "rate", S_IRUGO|S_IWUSR, platform->debugfs, g, &rate_fops);
diff --git a/drivers/gpu/nvgpu/gm20b/ltc_gm20b.c b/drivers/gpu/nvgpu/gm20b/ltc_gm20b.c
index ffc36903..e4e27764 100644
--- a/drivers/gpu/nvgpu/gm20b/ltc_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/ltc_gm20b.c
@@ -112,7 +112,7 @@ int gm20b_ltc_cbc_ctrl(struct gk20a *g, enum gk20a_cbc_op op,
        gk20a_dbg_fn("");
-        trace_gk20a_ltc_cbc_ctrl_start(g->dev->name, op, min, max);
+        trace_gk20a_ltc_cbc_ctrl_start(dev_name(g->dev), op, min, max);
        if (gr->compbit_store.mem.size == 0)
                return 0;
@@ -162,7 +162,7 @@ int gm20b_ltc_cbc_ctrl(struct gk20a *g, enum gk20a_cbc_op op,
                }
        }
 out:
-        trace_gk20a_ltc_cbc_ctrl_done(g->dev->name);
+        trace_gk20a_ltc_cbc_ctrl_done(dev_name(g->dev));
        mutex_unlock(&g->mm.l2_op_lock);
        return err;
 }
@@ -230,7 +230,7 @@ void gm20b_ltc_g_elpg_flush_locked(struct gk20a *g)
        gk20a_dbg_fn("");
-        trace_gk20a_mm_g_elpg_flush_locked(g->dev->name);
+        trace_gk20a_mm_g_elpg_flush_locked(dev_name(g->dev));
        for (i = 0; i < g->ltc_count; i++)
                done[i] = 0;
@@ -264,7 +264,7 @@ void gm20b_ltc_g_elpg_flush_locked(struct gk20a *g)
                gk20a_warn(dev_from_gk20a(g),
                            "g_elpg_flush too many retries");
-        trace_gk20a_mm_g_elpg_flush_locked_done(g->dev->name);
+        trace_gk20a_mm_g_elpg_flush_locked_done(dev_name(g->dev));
 }
 u32 gm20b_ltc_cbc_fix_config(struct gk20a *g, int base)
diff --git a/drivers/gpu/nvgpu/gm20b/mm_gm20b.c b/drivers/gpu/nvgpu/gm20b/mm_gm20b.c
index 621ee687..7bc19972 100644
--- a/drivers/gpu/nvgpu/gm20b/mm_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/mm_gm20b.c
@@ -1,7 +1,7 @@
 /*
 * GM20B MMU
 *
- * Copyright (c) 2014-2015, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -52,7 +52,7 @@ int gm20b_mm_mmu_vpr_info_fetch(struct gk20a *g)
        gk20a_busy_noresume(g->dev);
 #ifdef CONFIG_PM
-        if (!pm_runtime_active(&g->dev->dev))
+        if (!pm_runtime_active(g->dev))
                goto fail;
 #endif
@@ -67,7 +67,7 @@ int gm20b_mm_mmu_vpr_info_fetch(struct gk20a *g)
        ret = gm20b_mm_mmu_vpr_info_fetch_wait(g, VPR_INFO_FETCH_WAIT);
 fail:
-        pm_runtime_put(&g->dev->dev);
+        pm_runtime_put(g->dev);
        return ret;
 }
diff --git a/drivers/gpu/nvgpu/platform_tegra.h b/drivers/gpu/nvgpu/platform_tegra.h
index 6fa13831..815ac657 100644
--- a/drivers/gpu/nvgpu/platform_tegra.h
+++ b/drivers/gpu/nvgpu/platform_tegra.h
@@ -1,9 +1,7 @@
 /*
- * drivers/video/tegra/host/gk20a/soc/platform_gk20a.h
- *
 * GK20A Platform (SoC) Interface
 *
- * Copyright (c) 2014, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
@@ -23,9 +21,9 @@
 struct platform_device;
 struct gr_ctx_buffer_desc;
-int gk20a_tegra_secure_alloc(struct platform_device *pdev,
+int gk20a_tegra_secure_alloc(struct device *dev,
                             struct gr_ctx_buffer_desc *desc,
                             size_t size);
-int gk20a_tegra_secure_page_alloc(struct platform_device *pdev);
+int gk20a_tegra_secure_page_alloc(struct device *dev);
 #endif
diff --git a/drivers/gpu/nvgpu/vgpu/vgpu.c b/drivers/gpu/nvgpu/vgpu/vgpu.c
index ab063c9b..b67f4d95 100644
--- a/drivers/gpu/nvgpu/vgpu/vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/vgpu.c
@@ -140,7 +140,7 @@ static int vgpu_intr_thread(void *dev_id)
        return 0;
 }
-static void vgpu_remove_support(struct platform_device *dev)
+static void vgpu_remove_support(struct device *dev)
 {
        struct gk20a *g = get_gk20a(dev);
        struct gk20a_platform *platform = gk20a_get_platform(dev);
@@ -174,10 +174,10 @@ static void vgpu_remove_support(struct platform_device *dev)
        }
 }
-static int vgpu_init_support(struct platform_device *dev)
+static int vgpu_init_support(struct platform_device *pdev)
 {
-        struct resource *r = platform_get_resource(dev, IORESOURCE_MEM, 0);
+        struct resource *r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
-        struct gk20a *g = get_gk20a(dev);
+        struct gk20a *g = get_gk20a(&pdev->dev);
        int err = 0;
        if (!r) {
@@ -186,7 +186,7 @@ static int vgpu_init_support(struct platform_device *dev)
                goto fail;
        }
-        g->bar1 = devm_ioremap_resource(&dev->dev, r);
+        g->bar1 = devm_ioremap_resource(&pdev->dev, r);
        if (IS_ERR(g->bar1)) {
                dev_err(dev_from_gk20a(g), "failed to remap gk20a bar1\n");
                err = PTR_ERR(g->bar1);
@@ -200,14 +200,13 @@ static int vgpu_init_support(struct platform_device *dev)
        return 0;
 fail:
-        vgpu_remove_support(dev);
+        vgpu_remove_support(&pdev->dev);
        return err;
 }
 int vgpu_pm_prepare_poweroff(struct device *dev)
 {
-        struct platform_device *pdev = to_platform_device(dev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(pdev);
        int ret = 0;
        gk20a_dbg_fn("");
@@ -284,7 +283,7 @@ static int vgpu_init_hal(struct gk20a *g)
                break;
 #endif
        default:
-                gk20a_err(&g->dev->dev, "no support for %x", ver);
+                gk20a_err(g->dev, "no support for %x", ver);
                err = -ENODEV;
                break;
        }
@@ -294,8 +293,7 @@ static int vgpu_init_hal(struct gk20a *g)
 int vgpu_pm_finalize_poweron(struct device *dev)
 {
-        struct platform_device *pdev = to_platform_device(dev);
+        struct gk20a *g = get_gk20a(dev);
-        struct gk20a *g = get_gk20a(pdev);
        int err;
        gk20a_dbg_fn("");
@@ -342,9 +340,9 @@ done:
        return err;
 }
-static int vgpu_pm_initialise_domain(struct platform_device *pdev)
+static int vgpu_pm_initialise_domain(struct device *dev)
 {
-        struct gk20a_platform *platform = platform_get_drvdata(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(dev);
        struct dev_power_governor *pm_domain_gov = NULL;
        struct gk20a_domain_data *vgpu_pd_data;
        struct generic_pm_domain *domain;
@@ -369,17 +367,17 @@ static int vgpu_pm_initialise_domain(struct platform_device *pdev)
        domain->dev_ops.save_state = vgpu_pm_prepare_poweroff;
        domain->dev_ops.restore_state = vgpu_pm_finalize_poweron;
-        device_set_wakeup_capable(&pdev->dev, 0);
+        device_set_wakeup_capable(dev, 0);
-        return pm_genpd_add_device(domain, &pdev->dev);
+        return pm_genpd_add_device(domain, dev);
 }
-static int vgpu_pm_init(struct platform_device *dev)
+static int vgpu_pm_init(struct device *dev)
 {
        int err = 0;
        gk20a_dbg_fn("");
-        pm_runtime_enable(&dev->dev);
+        pm_runtime_enable(dev);
        /* genpd will take care of runtime power management if it is enabled */
        if (IS_ENABLED(CONFIG_PM_GENERIC_DOMAINS))
@@ -388,14 +386,15 @@ static int vgpu_pm_init(struct platform_device *dev)
        return err;
 }
-int vgpu_probe(struct platform_device *dev)
+int vgpu_probe(struct platform_device *pdev)
 {
        struct gk20a *gk20a;
        int err;
+        struct device *dev = &pdev->dev;
        struct gk20a_platform *platform = gk20a_get_platform(dev);
        if (!platform) {
-                dev_err(&dev->dev, "no platform data\n");
+                dev_err(dev, "no platform data\n");
                return -ENODATA;
        }
@@ -403,18 +402,18 @@ int vgpu_probe(struct platform_device *dev)
        gk20a = kzalloc(sizeof(struct gk20a), GFP_KERNEL);
        if (!gk20a) {
-                dev_err(&dev->dev, "couldn't allocate gk20a support");
+                dev_err(dev, "couldn't allocate gk20a support");
                return -ENOMEM;
        }
        platform->g = gk20a;
        gk20a->dev = dev;
-        err = gk20a_user_init(dev);
+        err = gk20a_user_init(dev, INTERFACE_NAME);
        if (err)
                return err;
-        vgpu_init_support(dev);
+        vgpu_init_support(pdev);
        vgpu_dbg_init();
        init_rwsem(&gk20a->busy_lock);
@@ -424,33 +423,33 @@ int vgpu_probe(struct platform_device *dev)
        /* Initialize the platform interface. */
        err = platform->probe(dev);
        if (err) {
-                dev_err(&dev->dev, "platform probe failed");
+                dev_err(dev, "platform probe failed");
                return err;
        }
        err = vgpu_pm_init(dev);
        if (err) {
-                dev_err(&dev->dev, "pm init failed");
+                dev_err(dev, "pm init failed");
                return err;
        }
        if (platform->late_probe) {
                err = platform->late_probe(dev);
                if (err) {
-                        dev_err(&dev->dev, "late probe failed");
+                        dev_err(dev, "late probe failed");
                        return err;
                }
        }
-        err = vgpu_comm_init(dev);
+        err = vgpu_comm_init(pdev);
        if (err) {
-                dev_err(&dev->dev, "failed to init comm interface\n");
+                dev_err(dev, "failed to init comm interface\n");
                return -ENOSYS;
        }
        platform->virt_handle = vgpu_connect();
        if (!platform->virt_handle) {
-                dev_err(&dev->dev, "failed to connect to server node\n");
+                dev_err(dev, "failed to connect to server node\n");
                vgpu_comm_deinit();
                return -ENOSYS;
        }
@@ -462,8 +461,8 @@ int vgpu_probe(struct platform_device *dev)
        gk20a_debug_init(dev);
        /* Set DMA parameters to allow larger sgt lists */
-        dev->dev.dma_parms = &gk20a->dma_parms;
+        dev->dma_parms = &gk20a->dma_parms;
-        dma_set_max_seg_size(&dev->dev, UINT_MAX);
+        dma_set_max_seg_size(dev, UINT_MAX);
        gk20a->gr_idle_timeout_default =
                        CONFIG_GK20A_DEFAULT_TIMEOUT;
@@ -475,8 +474,9 @@ int vgpu_probe(struct platform_device *dev)
        return 0;
 }
-int vgpu_remove(struct platform_device *dev)
+int vgpu_remove(struct platform_device *pdev)
 {
+        struct device *dev = &pdev->dev;
        struct gk20a *g = get_gk20a(dev);
        struct gk20a_domain_data *vgpu_gpd;
        gk20a_dbg_fn("");
author	Terje Bergstrom <tbergstrom@nvidia.com>	2016-03-29 19:02:34 -0400
committer	Terje Bergstrom <tbergstrom@nvidia.com>	2016-04-08 12:42:41 -0400
commit	e8bac374c0ed24f05bf389e1e8b5aca47f61bd3a (patch)
tree	36b6e111f8706c0560ee552bc9d6e15b87fc9621 /drivers
parent	2382a8433fddaee3baecff6ae941944850787ab7 (diff)