gpu: nvgpu: Move kmem_caches to allocator

Instead of using a single static kmem_cache for each type of
data structure the allocators may want to allocate each
allocator now has its own instance of the kmem_cache. This is
done so that each GPU driver instance can accurately track how
much memory it is using.

In order to support this on older kernels a new NVGPU API has
been made,

  nvgpu_kmem_cache_create(struct gk20a *g, size_t size)

To handle the possibility that caches cannot be created with
the same name.

This patch also fixes numerous places where kfree() was wrongly
used to free kmem_cache allocs.

Bug 1799159
Bug 1823380

Change-Id: Id674f9a5445fde3f95db65ad6bf3ea990444603d
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: http://git-master/r/1283826
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
GVS: Gerrit_Virtual_Submit

drivers/gpu/nvgpu/include/nvgpu/kmem.h

@@ -0,0 +1,40 @@
+/*
+ * Copyright (c) 2017, 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,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef NVGPU_KMEM_H
+#define NVGPU_KMEM_H
+
+struct gk20a;
+
+/*
+ * In Linux there is support for the notion of a kmem_cache. It gives better
+ * memory usage characteristics for lots of allocations of the same size. Think
+ * structs that get allocated over and over. Normal kmalloc() type routines
+ * typically round to the next power-of-2 since that's easy.
+ *
+ * But if we know the size ahead of time the packing for the allocations can be
+ * much better. This is the benefit of a slab allocator. This type hides the
+ * underlying kmem_cache (or absense thereof).
+ */
+struct nvgpu_kmem_cache;
+
+struct nvgpu_kmem_cache *nvgpu_kmem_cache_create(struct gk20a *g, size_t size);
+void nvgpu_kmem_cache_destroy(struct nvgpu_kmem_cache *cache);
+
+void *nvgpu_kmem_cache_alloc(struct nvgpu_kmem_cache *cache);
+void nvgpu_kmem_cache_free(struct nvgpu_kmem_cache *cache, void *ptr);
+
+#endif