[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <5c4f5f0d-024a-46e5-9506-0b589f6cce5d@amd.com>
Date: Sun, 6 Jul 2025 16:44:27 -0400
From: Mario Limonciello <mario.limonciello@....com>
To: Samuel Zhang <guoqing.zhang@....com>, alexander.deucher@....com,
christian.koenig@....com, rafael@...nel.org, len.brown@...el.com,
pavel@...nel.org, gregkh@...uxfoundation.org, dakr@...nel.org,
airlied@...il.com, simona@...ll.ch, ray.huang@....com,
matthew.auld@...el.com, matthew.brost@...el.com,
maarten.lankhorst@...ux.intel.com, mripard@...nel.org, tzimmermann@...e.de
Cc: lijo.lazar@....com, victor.zhao@....com, haijun.chang@....com,
Qing.Ma@....com, linux-pm@...r.kernel.org, linux-kernel@...r.kernel.org,
amd-gfx@...ts.freedesktop.org, dri-devel@...ts.freedesktop.org
Subject: Re: [PATCH v2 1/5] drm/ttm: add ttm_device_prepare_hibernation() api
On 7/4/2025 6:12 AM, Samuel Zhang wrote:
> This new api is used for hibernation to move GTT BOs to shmem after
> VRAM eviction. shmem will be flushed to swap disk later to reduce
> the system memory usage for hibernation.
>
> Signed-off-by: Samuel Zhang <guoqing.zhang@....com>
> ---
> drivers/gpu/drm/ttm/ttm_device.c | 29 +++++++++++++++++++++++++++++
> include/drm/ttm/ttm_device.h | 1 +
> 2 files changed, 30 insertions(+)
>
> diff --git a/drivers/gpu/drm/ttm/ttm_device.c b/drivers/gpu/drm/ttm/ttm_device.c
> index 02e797fd1891..19ab35ffeead 100644
> --- a/drivers/gpu/drm/ttm/ttm_device.c
> +++ b/drivers/gpu/drm/ttm/ttm_device.c
> @@ -123,6 +123,35 @@ static int ttm_global_init(void)
> return ret;
> }
>
> +/**
> + * move GTT BOs to shmem for hibernation.
> + *
> + * returns 0 on success, negative on failure.
> + */
> +int ttm_device_prepare_hibernation(void)
> +{
> + struct ttm_operation_ctx ctx = {
> + .interruptible = false,
> + .no_wait_gpu = false,
> + .force_alloc = true
> + };
> + struct ttm_global *glob = &ttm_glob;
> + struct ttm_device *bdev;
> + int ret = 0;
> +
> + mutex_lock(&ttm_global_mutex);
> + list_for_each_entry(bdev, &glob->device_list, device_list) {
> + do {
> + ret = ttm_device_swapout(bdev, &ctx, GFP_KERNEL);
> + } while (ret > 0);
> + if (ret < 0)
> + break;
> + }
> + mutex_unlock(&ttm_global_mutex);
> + return ret;
I'd personally rather see scoped guard here so you can return
immediately and the guard will clean up but up to Christian what he thinks.
int ret;
scoped_guard(mutex, &ttm_global_mutex) {
list_for_each_entry(bdev, &glob->device_list, device_list) {
do {
ret = ttm_device_swapout(bdev, &ctx, GFP_KERNEL);
} while (ret > 0);
if (ret)
return ret;
}
return 0;
> +}
> +EXPORT_SYMBOL(ttm_device_prepare_hibernation);
> +
> /*
> * A buffer object shrink method that tries to swap out the first
> * buffer object on the global::swap_lru list.
> diff --git a/include/drm/ttm/ttm_device.h b/include/drm/ttm/ttm_device.h
> index 39b8636b1845..b45498b398dd 100644
> --- a/include/drm/ttm/ttm_device.h
> +++ b/include/drm/ttm/ttm_device.h
> @@ -272,6 +272,7 @@ struct ttm_device {
> int ttm_global_swapout(struct ttm_operation_ctx *ctx, gfp_t gfp_flags);
> int ttm_device_swapout(struct ttm_device *bdev, struct ttm_operation_ctx *ctx,
> gfp_t gfp_flags);
> +int ttm_device_prepare_hibernation(void);
>
> static inline struct ttm_resource_manager *
> ttm_manager_type(struct ttm_device *bdev, int mem_type)
Powered by blists - more mailing lists