AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Khatri, Sunil" <sukhatri@amd.com>
To: "Christian König" <ckoenig.leichtzumerken@gmail.com>,
	tursulin@ursulin.net, Alexander.Deucher@amd.com,
	Prike.Liang@amd.com, Yogesh.Mohanmarimuthu@amd.com,
	SRINIVASAN.SHANMUGAM@amd.com, Sunil.Khatri@amd.com,
	amd-gfx@lists.freedesktop.org
Subject: Re: [PATCH 06/11] drm/amdgpu: fix adding eviction fence
Date: Wed, 11 Mar 2026 17:56:44 +0530	[thread overview]
Message-ID: <533e3f43-d375-44a4-a449-10a8c19aee28@amd.com> (raw)
In-Reply-To: <20260310191327.2279-6-christian.koenig@amd.com>

Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>

On 11-03-2026 12:43 am, Christian König wrote:
> We can't add the eviction fence without validating the BO.
>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> ---
>   .../drm/amd/amdgpu/amdgpu_eviction_fence.c    | 19 ++++++++++++++++---
>   .../drm/amd/amdgpu/amdgpu_eviction_fence.h    |  4 ++--
>   drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c       |  9 ++++++---
>   3 files changed, 24 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c
> index ef4da6f2e2a3..6598823ec619 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c
> @@ -82,14 +82,27 @@ amdgpu_eviction_fence_suspend_worker(struct work_struct *work)
>   	mutex_unlock(&uq_mgr->userq_mutex);
>   }
>   
> -void amdgpu_evf_mgr_attach_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
> -				 struct amdgpu_bo *bo)
> +int amdgpu_evf_mgr_attach_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
> +				struct amdgpu_bo *bo)
>   {
>   	struct dma_fence *ev_fence = amdgpu_evf_mgr_get_fence(evf_mgr);
> +	struct ttm_operation_ctx ctx = { false, false };
>   	struct dma_resv *resv = bo->tbo.base.resv;
> +	int ret;
> +
> +	if (!dma_fence_is_signaled(ev_fence)) {
> +
> +		amdgpu_bo_placement_from_domain(bo, bo->allowed_domains);
> +		ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
> +		if (!ret)
> +			dma_resv_add_fence(resv, ev_fence,
> +					   DMA_RESV_USAGE_BOOKKEEP);
> +	} else {
> +		ret = 0;
> +	}
>   
> -	dma_resv_add_fence(resv, ev_fence, DMA_RESV_USAGE_BOOKKEEP);
>   	dma_fence_put(ev_fence);
> +	return ret;
>   }
>   
>   int amdgpu_evf_mgr_rearm(struct amdgpu_eviction_fence_mgr *evf_mgr,
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h
> index 132a13a5dc1c..2a750add4e7b 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h
> @@ -58,8 +58,8 @@ amdgpu_evf_mgr_get_fence(struct amdgpu_eviction_fence_mgr *evf_mgr)
>   	return ev_fence;
>   }
>   
> -void amdgpu_evf_mgr_attach_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
> -				 struct amdgpu_bo *bo);
> +int amdgpu_evf_mgr_attach_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
> +				struct amdgpu_bo *bo);
>   int amdgpu_evf_mgr_rearm(struct amdgpu_eviction_fence_mgr *evf_mgr,
>   			 struct drm_exec *exec);
>   void amdgpu_evf_mgr_detach_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> index e28abfd04867..88a21400ae09 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> @@ -258,12 +258,15 @@ static int amdgpu_gem_object_open(struct drm_gem_object *obj,
>   
>   	amdgpu_vm_bo_update_shared(abo);
>   	bo_va = amdgpu_vm_bo_find(vm, abo);
> -	if (!bo_va)
> +	if (!bo_va) {
>   		bo_va = amdgpu_vm_bo_add(adev, vm, abo);
> -	else
> +		r = amdgpu_evf_mgr_attach_fence(&fpriv->evf_mgr, abo);
> +		if (r)
> +			goto out_unlock;
> +	} else {
>   		++bo_va->ref_count;
> +	}
>   
> -	amdgpu_evf_mgr_attach_fence(&fpriv->evf_mgr, abo);
>   	drm_exec_fini(&exec);
>   
>   	/* Validate and add eviction fence to DMABuf imports with dynamic

  reply	other threads:[~2026-03-11 12:26 UTC|newest]

Thread overview: 42+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-10 19:13 [PATCH 01/11] drm/amdgpu: revert to old status lock handling v4 Christian König
2026-03-10 19:13 ` [PATCH 02/11] drm/amdgpu: restructure VM state machine Christian König
2026-03-11  8:47   ` Khatri, Sunil
2026-03-12 12:49     ` Christian König
2026-03-12 10:07   ` Liang, Prike
2026-03-12 14:32   ` Tvrtko Ursulin
2026-03-16 13:44     ` Christian König
2026-03-16 14:26       ` Tvrtko Ursulin
2026-03-10 19:13 ` [PATCH 03/11] drm/amdgpu: fix amdgpu_userq_evict Christian König
2026-03-11  8:51   ` Khatri, Sunil
2026-03-13  7:25   ` Liang, Prike
2026-03-10 19:13 ` [PATCH 04/11] drm/amdgpu: completely rework eviction fence handling Christian König
2026-03-11 12:27   ` Khatri, Sunil
2026-03-13  8:00     ` Khatri, Sunil
2026-03-17  9:41     ` Christian König
2026-03-13  8:28   ` Liang, Prike
2026-03-17  9:57     ` Christian König
2026-03-17 11:21       ` Liang, Prike
2026-03-17 11:23         ` Christian König
2026-03-17 11:54           ` Liang, Prike
2026-03-10 19:13 ` [PATCH 05/11] drm/amdgpu: fix eviction fence and userq manager shutdown Christian König
2026-03-11 12:26   ` Khatri, Sunil
2026-03-13  9:35     ` Khatri, Sunil
2026-03-10 19:13 ` [PATCH 06/11] drm/amdgpu: fix adding eviction fence Christian König
2026-03-11 12:26   ` Khatri, Sunil [this message]
2026-03-10 19:13 ` [PATCH 07/11] drm/amdgpu: rework amdgpu_userq_wait_ioctl v3 Christian König
2026-03-12 16:34   ` Tvrtko Ursulin
2026-03-16 14:19     ` Christian König
2026-03-16 14:44       ` Tvrtko Ursulin
2026-03-17  7:05   ` Khatri, Sunil
2026-03-10 19:13 ` [PATCH 08/11] drm/amdgpu: make amdgpu_user_wait_ioctl more resilent v2 Christian König
2026-03-17  7:15   ` Khatri, Sunil
2026-03-10 19:13 ` [PATCH 09/11] drm/amdgpu: annotate eviction fence signaling path Christian König
2026-03-17  7:35   ` Khatri, Sunil
2026-03-10 19:13 ` [PATCH 10/11] drm/amdgpu: fix some more bug in amdgpu_gem_va_ioctl Christian König
2026-03-17  8:44   ` Khatri, Sunil
2026-03-17 11:08     ` Christian König
2026-03-10 19:13 ` [PATCH 11/11] drm/amdgpu: WIP sync amdgpu_ttm_fill_mem only to kernel fences Christian König
2026-03-17  8:59   ` Khatri, Sunil
2026-03-17 10:52     ` Christian König
2026-03-11  7:43 ` [PATCH 01/11] drm/amdgpu: revert to old status lock handling v4 Khatri, Sunil
2026-03-12  7:13 ` Liang, Prike

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=533e3f43-d375-44a4-a449-10a8c19aee28@amd.com \
    --to=sukhatri@amd.com \
    --cc=Alexander.Deucher@amd.com \
    --cc=Prike.Liang@amd.com \
    --cc=SRINIVASAN.SHANMUGAM@amd.com \
    --cc=Sunil.Khatri@amd.com \
    --cc=Yogesh.Mohanmarimuthu@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=ckoenig.leichtzumerken@gmail.com \
    --cc=tursulin@ursulin.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox