From: "Kuehling, Felix" <Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
To: "amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org"
<amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org>,
"oded.gabbay-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org"
<oded.gabbay-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Cc: "Kuehling, Felix" <Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
Subject: [PATCH 7/9] drm/amdkfd: Fix and simplify sync object handling for KFD
Date: Tue, 6 Nov 2018 01:40:54 +0000 [thread overview]
Message-ID: <1541468434-29384-8-git-send-email-Felix.Kuehling@amd.com> (raw)
In-Reply-To: <1541468434-29384-1-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
The adev parameter in amdgpu_sync_fence and amdgpu_sync_resv is only
needed for updating sync->last_vm_update. This breaks if different
adevs are passed to calls for the same sync object.
Always pass NULL for calls from KFD because sync objects used for
KFD don't belong to any particular device, and KFD doesn't need the
sync->last_vm_update fence.
This fixes kernel log warnings on multi-GPU systems after recent
changes in amdgpu_amdkfd_gpuvm_restore_process_bos.
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 28 +++++-------------------
1 file changed, 5 insertions(+), 23 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index d005371..572ac5f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -395,23 +395,6 @@ static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm)
return 0;
}
-static int sync_vm_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
- struct dma_fence *f)
-{
- int ret = amdgpu_sync_fence(adev, sync, f, false);
-
- /* Sync objects can't handle multiple GPUs (contexts) updating
- * sync->last_vm_update. Fortunately we don't need it for
- * KFD's purposes, so we can just drop that fence.
- */
- if (sync->last_vm_update) {
- dma_fence_put(sync->last_vm_update);
- sync->last_vm_update = NULL;
- }
-
- return ret;
-}
-
static int vm_update_pds(struct amdgpu_vm *vm, struct amdgpu_sync *sync)
{
struct amdgpu_bo *pd = vm->root.base.bo;
@@ -422,7 +405,7 @@ static int vm_update_pds(struct amdgpu_vm *vm, struct amdgpu_sync *sync)
if (ret)
return ret;
- return sync_vm_fence(adev, sync, vm->last_update);
+ return amdgpu_sync_fence(NULL, sync, vm->last_update, false);
}
/* add_bo_to_vm - Add a BO to a VM
@@ -826,7 +809,7 @@ static int unmap_bo_from_gpuvm(struct amdgpu_device *adev,
/* Add the eviction fence back */
amdgpu_bo_fence(pd, &vm->process_info->eviction_fence->base, true);
- sync_vm_fence(adev, sync, bo_va->last_pt_update);
+ amdgpu_sync_fence(NULL, sync, bo_va->last_pt_update, false);
return 0;
}
@@ -851,7 +834,7 @@ static int update_gpuvm_pte(struct amdgpu_device *adev,
return ret;
}
- return sync_vm_fence(adev, sync, bo_va->last_pt_update);
+ return amdgpu_sync_fence(NULL, sync, bo_va->last_pt_update, false);
}
static int map_bo_to_gpuvm(struct amdgpu_device *adev,
@@ -911,7 +894,7 @@ static int process_sync_pds_resv(struct amdkfd_process_info *process_info,
vm_list_node) {
struct amdgpu_bo *pd = peer_vm->root.base.bo;
- ret = amdgpu_sync_resv(amdgpu_ttm_adev(pd->tbo.bdev),
+ ret = amdgpu_sync_resv(NULL,
sync, pd->tbo.resv,
AMDGPU_FENCE_OWNER_UNDEFINED, false);
if (ret)
@@ -2084,8 +2067,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
pr_debug("Memory eviction: Validate BOs failed. Try again\n");
goto validate_map_fail;
}
- ret = amdgpu_sync_fence(amdgpu_ttm_adev(bo->tbo.bdev),
- &sync_obj, bo->tbo.moving, false);
+ ret = amdgpu_sync_fence(NULL, &sync_obj, bo->tbo.moving, false);
if (ret) {
pr_debug("Memory eviction: Sync BO fence failed. Try again\n");
goto validate_map_fail;
--
2.7.4
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
next prev parent reply other threads:[~2018-11-06 1:40 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-11-06 1:40 [PATCH 0/9] KFD upstreaming Nov 2018, part 1 Kuehling, Felix
[not found] ` <1541468434-29384-1-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2018-11-06 1:40 ` [PATCH 1/9] drm/amdkfd: Replace mqd with mqd_mgr as the variable name for mqd_manager Kuehling, Felix
[not found] ` <1541468434-29384-2-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2018-11-06 21:32 ` Alex Deucher
2018-11-06 1:40 ` [PATCH 2/9] drm/amdkfd: Added Vega12 and Polaris12 for KFD Kuehling, Felix
[not found] ` <1541468434-29384-3-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2018-11-06 21:33 ` Alex Deucher
2018-11-06 1:40 ` [PATCH 3/9] drm/amdkfd: Adjust the debug message in KFD ISR Kuehling, Felix
[not found] ` <1541468434-29384-4-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2018-11-06 21:33 ` Alex Deucher
2018-11-06 1:40 ` [PATCH 4/9] drm/amdkfd: Workaround PASID missing in gfx9 interrupt payload under non HWS Kuehling, Felix
[not found] ` <1541468434-29384-5-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2018-11-06 21:34 ` Alex Deucher
2018-11-06 1:40 ` [PATCH 5/9] drm/amdgpu: Remove explicit wait after VM validate Kuehling, Felix
[not found] ` <1541468434-29384-6-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2018-11-06 21:35 ` Alex Deucher
2018-11-06 1:40 ` Kuehling, Felix [this message]
[not found] ` <1541468434-29384-8-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2018-11-16 20:19 ` [PATCH 7/9] drm/amdkfd: Fix and simplify sync object handling for KFD Kuehling, Felix
[not found] ` <972bbc9e-8f5d-9740-f400-4e1211e8e00f-5C7GfCeVMHo@public.gmane.org>
2018-11-17 16:01 ` Koenig, Christian
2018-11-06 1:40 ` [PATCH 6/9] drm/amdgpu: KFD Restore process: Optimize waiting Kuehling, Felix
[not found] ` <1541468434-29384-7-git-send-email-Felix.Kuehling-5C7GfCeVMHo@public.gmane.org>
2018-11-06 21:36 ` Alex Deucher
2018-11-06 1:40 ` [PATCH 8/9] drm/amdgpu: Fix KFD doorbell SG BO mapping Kuehling, Felix
2018-11-06 1:40 ` [PATCH 9/9] drm/amdkfd: change system memory overcommit limit Kuehling, Felix
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=1541468434-29384-8-git-send-email-Felix.Kuehling@amd.com \
--to=felix.kuehling-5c7gfcevmho@public.gmane.org \
--cc=amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
--cc=oded.gabbay-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org \
/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