gpu: nvgpu: User-space managed address space support

Implement NVGPU_GPU_IOCTL_ALLOC_AS_FLAGS_USERSPACE_MANAGED, which
enables creating userspace-managed GPU address spaces.

When an address space is marked as userspace-managed, the following
changes are in effect:

- Only fixed-address mappings are allowed.
- VA space allocation for fixed-address mappings is not required,
  except to mark space as sparse.
- Maps and unmaps are always immediate. In particular, the mapping
  ref increments at kickoffs and decrements at job completion are
  skipped.

Bug 1614735
Bug 1623949
Bug 1660392

Change-Id: I834fe19b3f65e9b02c268952383eddee0e465759
Signed-off-by: Sami Kiminki <skiminki@nvidia.com>
Reviewed-on: http://git-master/r/738558
Reviewed-on: http://git-master/r/833253
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>

1 files changed, 18 insertions, 1 deletions

diff --git a/include/uapi/linux/nvgpu.h b/include/uapi/linux/nvgpu.h
index 6a8e44c5..d2b5ceb8 100644
--- a/include/uapi/linux/nvgpu.h
+++ b/include/uapi/linux/nvgpu.h
@@ -106,6 +106,8 @@ struct nvgpu_gpu_zbc_query_table_args {
 #define NVGPU_GPU_FLAGS_SUPPORT_CYCLE_STATS             (1 << 4)
 /* NVGPU_IOCTL_CHANNEL_CYCLE_STATS_SNAPSHOT is available */
 #define NVGPU_GPU_FLAGS_SUPPORT_CYCLE_STATS_SNAPSHOT    (1 << 6)
+/* User-space managed address spaces support */
+#define NVGPU_GPU_FLAGS_SUPPORT_USERSPACE_MANAGED_AS    (1 << 7)
 
 struct nvgpu_gpu_characteristics {
         __u32 arch;
@@ -239,7 +241,22 @@ struct nvgpu_gpu_mark_compressible_write_args {
 struct nvgpu_alloc_as_args {
         __u32 big_page_size;
         __s32 as_fd;
-        __u64 reserved;                 /* must be zero */
+
+/*
+ * The GPU address space will be managed by the userspace. This has
+ * the following changes in functionality:
+ *   1. All non-fixed-offset user mappings are rejected (i.e.,
+ *      fixed-offset only)
+ *   2. Address space does not need to be allocated for fixed-offset
+ *      mappings, except to mark sparse address space areas.
+ *   3. Maps and unmaps are immediate. In particular, mapping ref
+ *      increments at kickoffs and decrements at job completion are
+ *      bypassed.
+ */
+#define NVGPU_GPU_IOCTL_ALLOC_AS_FLAGS_USERSPACE_MANAGED (1 << 0)
+        __u32 flags;
+
+        __u32 reserved;                 /* must be zero */
 };
 
 struct nvgpu_gpu_open_tsg_args {