gpu: nvgpu: Use vmalloc only when size >4K

When allocation size is 4k or below, we should use kmalloc. vmalloc
should be used only for larged allocations.

Introduce nvgpu_alloc, which checks the size, and decides the API
to use.

Change-Id: I593110467cd319851b27e57d1bfe8d228d3f2909
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/743974

diff --git a/drivers/gpu/nvgpu/gk20a/channel_gk20a.c b/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
index f311a91a8e0c5c2545b62a45fa3caa2c65daf8ae..b4bf27fa06be6e81149669b6768aba40791798e3 100644 (file)
--- a/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
@@ -2063,7 +2063,7 @@ static int gk20a_ioctl_channel_submit_gpfifo(
         * synchronization; we might still wait and do an increment */
        size = args->num_entries * sizeof(struct nvgpu_gpfifo);
        if (size) {
-               gpfifo = vmalloc(size);
+               gpfifo = nvgpu_alloc(size, false);
                if (!gpfifo)
                        return -ENOMEM;
 
@@ -2098,7 +2098,7 @@ static int gk20a_ioctl_channel_submit_gpfifo(
        gk20a_fence_put(fence_out);
 
 clean_up:
-       vfree(gpfifo);
+       nvgpu_free(gpfifo);
        return ret;
 }
 

diff --git a/drivers/gpu/nvgpu/gk20a/mm_gk20a.c b/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
index 2ce390a493da0de7e4e54c7d6b5b7e3b9205f7ac..5a512edabd797c17fdcfc7b4cf3922934ad203cf 100644 (file)
--- a/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
@@ -745,8 +745,8 @@ int gk20a_vm_get_buffers(struct vm_gk20a *vm,
 
        mutex_lock(&vm->update_gmmu_lock);
 
-       buffer_list = kzalloc(sizeof(*buffer_list) *
-                             vm->num_user_mapped_buffers, GFP_KERNEL);
+       buffer_list = nvgpu_alloc(sizeof(*buffer_list) *
+                             vm->num_user_mapped_buffers, true);
        if (!buffer_list) {
                mutex_unlock(&vm->update_gmmu_lock);
                return -ENOMEM;
@@ -795,7 +795,7 @@ void gk20a_vm_put_buffers(struct vm_gk20a *vm,
 
        mutex_unlock(&vm->update_gmmu_lock);
 
-       kfree(mapped_buffers);
+       nvgpu_free(mapped_buffers);
 }
 
 static void gk20a_vm_unmap_user(struct vm_gk20a *vm, u64 offset)

diff --git a/drivers/gpu/nvgpu/gk20a/mm_gk20a.h b/drivers/gpu/nvgpu/gk20a/mm_gk20a.h
index 82003cd0b57c6cb5c398d7c6803043745fb3c743..7a627354ebd8d618931a8207a17c362052b354f1 100644 (file)
--- a/drivers/gpu/nvgpu/gk20a/mm_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/mm_gk20a.h
@@ -22,6 +22,7 @@
 #include <linux/dma-attrs.h>
 #include <linux/iommu.h>
 #include <linux/tegra-soc.h>
+#include <linux/vmalloc.h>
 #include <asm/dma-iommu.h>
 #include <asm/cacheflush.h>
 #include "gk20a_allocator.h"
@@ -628,4 +629,31 @@ void gk20a_mm_init_pdb(struct gk20a *g, void *inst_ptr, u64 pdb_addr);
 extern const struct gk20a_mmu_level gk20a_mm_levels_64k[];
 extern const struct gk20a_mmu_level gk20a_mm_levels_128k[];
 
+static inline void *nvgpu_alloc(size_t size, bool clear)
+{
+       void *p;
+
+       if (size > PAGE_SIZE) {
+               if (clear)
+                       p = vzalloc(size);
+               else
+                       p = vmalloc(size);
+       } else {
+               if (clear)
+                       p = kzalloc(size, GFP_KERNEL);
+               else
+                       p = kmalloc(size, GFP_KERNEL);
+       }
+
+       return p;
+}
+
+static inline void nvgpu_free(void *p)
+{
+       if (virt_addr_valid(p))
+               kfree(p);
+       else
+               vfree(p);
+}
+
 #endif /* MM_GK20A_H */