gpu: nvgpu: ModeE perfbuffer feature development

perfbuf mapping fixes:
- Allocate VM specifically for perfbuf use: using the PMU's
  results in mmu faults for larger buffers where 64k pages are used.
- Make 4GB boundary check work for large address spaces
- remove requirement to have allow_all flag set
- track perfbuf ownership and clean up appropriately

Bug 1880196
JIRA EVLR-1074

Change-Id: Ieee4eb17b64acf9b6ede37bf8e6a91892cda4a7e
Signed-off-by: Peter Daifuku <pdaifuku@nvidia.com>
Reviewed-on: http://git-master/r/1460809
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 6 insertions, 0 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.h b/drivers/gpu/nvgpu/gk20a/gk20a.h
index 76251000..2f25e406 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.h
@@ -1027,6 +1027,12 @@ struct gk20a {
         struct nvgpu_dbg_gpu_reg_op *dbg_regops_tmp_buf;
         u32 dbg_regops_tmp_buf_ops;
 
+        /* For perfbuf mapping */
+        struct {
+                struct dbg_session_gk20a *owner;
+                u64 offset;
+        } perfbuf;
+
         /* For profiler reservations */
         struct nvgpu_list_node profiler_objects;
         bool global_profiler_reservation_held;