nv-tegra-mirror/linux-nv-oot
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nv-oot.git synced 2025-12-22 09:11:26 +03:00
Code Issues Packages Projects Releases Wiki Activity

tegra: nvmap: Fix nvmap_query_heap for multi numa node platform

Browse Source 
In current implementation, when NvRmMemQueryHeapParams is
called and multiple numa nodes are online:
1. For iovmm carveout, numa_id is set to garbage value,
   and we are calling compute_memory_stat with it.
2. For gpu carveout, we are returning values for
   numa_id 0.
3. For other carveouts, we are returning params for the
   first matching entry in nvmap_dev->heaps[i].

Correct this behavior as follows:
Regardless of carveout type, return params for numa_id 0
when NvRmMemQueryHeapParams is called and multiple numa
nodes are online.

In long-term, we need to disable NvRmMemQueryHeapParams
when multiple numa nodes are online. Clients should use
NvRmMemQueryHeapParamsNuma instead.

Jira TMM-5970

Change-Id: Id49289e51eda187b1d676e5192583f320835c2f4
Signed-off-by: N V S Abhishek <nabhishek@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3290730
Reviewed-by: Ketan Patil <ketanp@nvidia.com>
Reviewed-by: Pritesh Raithatha <praithatha@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
This commit is contained in:
N V S Abhishek
2025-01-24 17:04:51 +00:00
committed by Jon Hunter
parent 73f74915ee
commit a3e0e896a0
1 changed files with 22 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
drivers/video/tegra/nvmap/nvmap_heap.c
View File

@@ -1,6 +1,6 @@
// SPDX-License-Identifier: GPL-2.0-only
/*
* SPDX-FileCopyrightText: Copyright (c) 2011-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2011-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
*
* GPU heap allocator.
*/
@@ -222,7 +222,26 @@ int nvmap_query_heap(struct nvmap_query_heap_params *op, bool is_numa_aware)
struct nvmap_heap *heap;
unsigned int type;
int i;
int numa_id;
/*
* Default value of numa_id will be 0. It can be overwritten by user's
* input only if NvRmMemQueryHeapParamsNuma is called (making
* is_numa_aware true).
* If NvRmMemQueryHeapParamsNuma is called:
* 1. When single numa node present
* a. Return params for only op->numa_id = 0.
* b. Return error if op->numa_id is not 0.
* 2. When multiple numa nodes present
* a. Return params for op->numa_id
* If NvRmMemQueryHeapParams is called:
* 1. When single numa node present
* a. Return params using system_heap_total_mem &
* system_heap_free_mem for iovmm carveout.
* b. Return params for numa id 0 for any other
* carveout heaps.
* 2. When multiple numa nodes present
* a. Return params for numa id 0.
*/
int numa_id = 0;
unsigned long free_mem = 0;
int ret = 0;
@@ -255,9 +274,6 @@ int nvmap_query_heap(struct nvmap_query_heap_params *op, bool is_numa_aware)
* hugetlbfs, so we need to return the HugePages_Total, HugePages_Free values
*/
if (type & NVMAP_HEAP_CARVEOUT_GPU) {
if (!is_numa_aware)
numa_id = 0;
ret = compute_hugetlbfs_stat(&op->total, &op->free, numa_id);
if (ret)
goto exit;
@@ -267,10 +283,8 @@ int nvmap_query_heap(struct nvmap_query_heap_params *op, bool is_numa_aware)
} else if (type & NVMAP_HEAP_CARVEOUT_MASK) {
for (i = 0; i < nvmap_dev->nr_carveouts; i++) {
if ((type & nvmap_get_heap_bit(nvmap_dev->heaps[i])) &&
(is_numa_aware ?
(numa_id == nvmap_get_heap_nid(nvmap_get_heap_ptr(
nvmap_dev->heaps[i]))) :
true)) {
nvmap_dev->heaps[i])))) {
heap = nvmap_get_heap_ptr(nvmap_dev->heaps[i]);
op->total = nvmap_query_heap_size(heap);
op->free = nvmap_get_heap_free_size(heap);