nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-24 10:34:43 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: split nvgpu power on sequence into 2 stages

Browse Source 
1) nvgpu poweron sequence split into two stages:
    - nvgpu_early_init() - Initializes the sub units
      which are required to be initialized before the grgmr init.
      For creating dev node, grmgr init and its dependency unit
      needs to move to early stage of GPU power on.
      After successful nvgpu_early_init() sequence,
      NvGpu can indetify the number of MIG instance required
      for each physical GPU.
    - nvgpu_finalize_poweron() - Initializes the sub units which
      can be initialized at the later stage of GPU power on sequence.

    - grmgr init depends on the following HAL sub units,
      * device - To get the device caps.
      * priv_ring - To get the gpc count and other
        MIG config programming.
      * fb - MIG config programming.
      * ltc - MIG config programming.
      * bios, bus, ecc and clk - dependent module of
        priv_ring/fb/ltc.

2) g->ops.xve.reset_gpu() should be called before GPU sub unit
   initialization. Hence, added g->ops.xve.reset_gpu() HAL in the
   early stage of dGPU power on sequence.

3) Increased xve_reset timeout from 100ms to 200ms.

4) Added nvgpu_assert() for gpc_count, gpc_mask and
   max_veid_count_per_tsg for identify the GPU boot
   device probe failure during nvgpu_init_gr_manager().

JIRA NVGPU-6633

Change-Id: I5d43bf711198e6b3f8eebcec3027ba17c15fc692
Signed-off-by: Lakshmanan M <lm@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2521894
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: Debarshi Dutta <ddutta@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Vaibhav Kachore <vkachore@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
This commit is contained in:
Lakshmanan M
2021-04-28 21:53:41 +05:30
committed by mobile promotions
parent a18be36a15
commit c041ad5b4b
4 changed files with 129 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
drivers/gpu/nvgpu/include/nvgpu/nvgpu_init.h
View File

@@ -89,6 +89,24 @@ struct nvgpu_ref;
* the GPU dev node in the early stage of GPU power on sequence.
* Each sub-unit is responsible for HW initialization.
*
* nvgpu poweron sequence split into two stages:
* - nvgpu_early_init() - Initializes the sub units
* which are required to be initialized before the grgmr init.
* For creating dev node, grmgr init and its dependency unit
* needs to move to early stage of GPU power on.
* After successful nvgpu_early_init() sequence,
* NvGpu can indetify the number of MIG instance required
* for each physical GPU.
* - nvgpu_finalize_poweron() - Initializes the sub units which
* can be initialized at the later stage of GPU power on sequence.
*
* grmgr init depends on the following HAL sub units,
* device - To get the device caps.
* priv_ring - To get the gpc count and other MIG config programming.
* fb - MIG config programming.
* ltc - MIG config programming.
* bios, bus, ecc and clk - dependent module of priv_ring/fb/ltc.
*
* @return 0 in case of success, < 0 in case of failure.
*/
int nvgpu_early_poweron(struct gk20a *g);