MIRRORS/linux-yocto
1
0
Fork 0
mirror of git://git.yoctoproject.org/linux-yocto.git synced 2025-10-22 23:13:01 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity

amd/amdkfd: correct mem limit calculation for small APUs

Browse Source 
[ Upstream commit 53503556273a5ead8b75534085e2dcb46e96f883 ]

Current mem limit check leaks some GTT memory (reserved_for_pt
reserved_for_ras + adev->vram_pin_size) for small APUs.

Since carveout VRAM is tunable on APUs, there are three case
regarding the carveout VRAM size relative to GTT:

1. 0 < carveout < gtt
   apu_prefer_gtt = true, is_app_apu = false

2. carveout > gtt / 2
   apu_prefer_gtt = false, is_app_apu = false

3. 0 = carveout
   apu_prefer_gtt = true, is_app_apu = true

It doesn't make sense to check below limitation in case 1
(default case, small carveout) because the values in the below
expression are mixed with carveout and gtt.

adev->kfd.vram_used[xcp_id] + vram_needed >
    vram_size - reserved_for_pt - reserved_for_ras -
    atomic64_read(&adev->vram_pin_size)

gtt: kfd.vram_used, vram_needed, vram_size
carveout: reserved_for_pt, reserved_for_ras, adev->vram_pin_size

In case 1, vram allocation will go to gtt domain, skip vram check
since ttm_mem_limit check already cover this allocation.

Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit fa7c99f04f6dd299388e9282812b14e95558ac8e)
Signed-off-by: Sasha Levin <sashal@kernel.org>
This commit is contained in:
Yifan Zhang 2025-08-20 16:10:51 +08:00 committed by Greg Kroah-Hartman
parent a01d1325e0
commit 0bcc5ea4bb
1 changed files with 32 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
View File

@ -213,19 +213,35 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
spin_lock(&kfd_mem_limit.mem_limit_lock);
if (kfd_mem_limit.system_mem_used + system_mem_needed >
kfd_mem_limit.max_system_mem_limit)
kfd_mem_limit.max_system_mem_limit) {
pr_debug("Set no_system_mem_limit=1 if using shared memory\n");
if (!no_system_mem_limit) {
ret = -ENOMEM;
goto release;
}
}
if ((kfd_mem_limit.system_mem_used + system_mem_needed >
kfd_mem_limit.max_system_mem_limit && !no_system_mem_limit) ||
(kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
kfd_mem_limit.max_ttm_mem_limit) ||
(adev && xcp_id >= 0 && adev->kfd.vram_used[xcp_id] + vram_needed >
vram_size - reserved_for_pt - reserved_for_ras - atomic64_read(&adev->vram_pin_size))) {
if (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
kfd_mem_limit.max_ttm_mem_limit) {
ret = -ENOMEM;
goto release;
}
/*if is_app_apu is false and apu_prefer_gtt is true, it is an APU with
* carve out < gtt. In that case, VRAM allocation will go to gtt domain, skip
* VRAM check since ttm_mem_limit check already cover this allocation
*/
if (adev && xcp_id >= 0 && (!adev->apu_prefer_gtt || adev->gmc.is_app_apu)) {
uint64_t vram_available =
vram_size - reserved_for_pt - reserved_for_ras -
atomic64_read(&adev->vram_pin_size);
if (adev->kfd.vram_used[xcp_id] + vram_needed > vram_available) {
ret = -ENOMEM;
goto release;
}
}
/* Update memory accounting by decreasing available system
* memory, TTM memory and GPU memory as computed above
*/
@ -1626,11 +1642,15 @@ size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
uint64_t vram_available, system_mem_available, ttm_mem_available;
spin_lock(&kfd_mem_limit.mem_limit_lock);
vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
- adev->kfd.vram_used_aligned[xcp_id]
- atomic64_read(&adev->vram_pin_size)
- reserved_for_pt
- reserved_for_ras;
if (adev->apu_prefer_gtt && !adev->gmc.is_app_apu)
vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
- adev->kfd.vram_used_aligned[xcp_id];
else
vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
- adev->kfd.vram_used_aligned[xcp_id]
- atomic64_read(&adev->vram_pin_size)
- reserved_for_pt
- reserved_for_ras;
if (adev->apu_prefer_gtt) {
system_mem_available = no_system_mem_limit ?