From: Sasha Levin <sashal@kernel.org>
To: linux-kernel@vger.kernel.org, stable@vger.kernel.org
Cc: "Felix Kuehling" <felix.kuehling@amd.com>,
"Hao Zhou" <hao.zhou@amd.com>,
"Christian König" <christian.koenig@amd.com>,
"Alex Deucher" <alexander.deucher@amd.com>,
"Sasha Levin" <sashal@kernel.org>,
airlied@gmail.com, simona@ffwll.ch, Yunxiang.Li@amd.com,
matthew.auld@intel.com, tvrtko.ursulin@igalia.com,
amd-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org
Subject: [PATCH AUTOSEL 6.1 07/10] drm/amdgpu: Allow P2P access through XGMI
Date: Tue, 29 Apr 2025 19:53:12 -0400 [thread overview]
Message-ID: <20250429235316.538129-7-sashal@kernel.org> (raw)
In-Reply-To: <20250429235316.538129-1-sashal@kernel.org>
From: Felix Kuehling <felix.kuehling@amd.com>
[ Upstream commit a92741e72f91b904c1d8c3d409ed8dbe9c1f2b26 ]
If peer memory is accessible through XGMI, allow leaving it in VRAM
rather than forcing its migration to GTT on DMABuf attachment.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Hao (Claire) Zhou <hao.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 372c8d72c3680fdea3fbb2d6b089f76b4a6d596a)
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 30 ++++++++++++++++++++-
1 file changed, 29 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
index ab06cb4d7b358..4dcc7de961d08 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
@@ -42,6 +42,29 @@
#include <linux/pci-p2pdma.h>
#include <linux/pm_runtime.h>
+static const struct dma_buf_attach_ops amdgpu_dma_buf_attach_ops;
+
+/**
+ * dma_buf_attach_adev - Helper to get adev of an attachment
+ *
+ * @attach: attachment
+ *
+ * Returns:
+ * A struct amdgpu_device * if the attaching device is an amdgpu device or
+ * partition, NULL otherwise.
+ */
+static struct amdgpu_device *dma_buf_attach_adev(struct dma_buf_attachment *attach)
+{
+ if (attach->importer_ops == &amdgpu_dma_buf_attach_ops) {
+ struct drm_gem_object *obj = attach->importer_priv;
+ struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
+ return amdgpu_ttm_adev(bo->tbo.bdev);
+ }
+
+ return NULL;
+}
+
/**
* amdgpu_dma_buf_attach - &dma_buf_ops.attach implementation
*
@@ -53,12 +76,14 @@
static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf,
struct dma_buf_attachment *attach)
{
+ struct amdgpu_device *attach_adev = dma_buf_attach_adev(attach);
struct drm_gem_object *obj = dmabuf->priv;
struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
int r;
- if (pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
+ if (!amdgpu_dmabuf_is_xgmi_accessible(attach_adev, bo) &&
+ pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
attach->peer2peer = false;
r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
@@ -479,6 +504,9 @@ bool amdgpu_dmabuf_is_xgmi_accessible(struct amdgpu_device *adev,
struct drm_gem_object *obj = &bo->tbo.base;
struct drm_gem_object *gobj;
+ if (!adev)
+ return false;
+
if (obj->import_attach) {
struct dma_buf *dma_buf = obj->import_attach->dmabuf;
--
2.39.5
next prev parent reply other threads:[~2025-04-29 23:53 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-04-29 23:53 [PATCH AUTOSEL 6.1 01/10] scsi: target: iscsi: Fix timeout on deleted connection Sasha Levin
2025-04-29 23:53 ` [PATCH AUTOSEL 6.1 02/10] virtio_ring: Fix data race by tagging event_triggered as racy for KCSAN Sasha Levin
2025-04-29 23:53 ` [PATCH AUTOSEL 6.1 03/10] dma-mapping: avoid potential unused data compilation warning Sasha Levin
2025-04-29 23:53 ` [PATCH AUTOSEL 6.1 04/10] cgroup: Fix compilation issue due to cgroup_mutex not being exported Sasha Levin
2025-04-29 23:53 ` [PATCH AUTOSEL 6.1 05/10] scsi: mpi3mr: Add level check to control event logging Sasha Levin
2025-04-29 23:53 ` [PATCH AUTOSEL 6.1 06/10] net: enetc: refactor bulk flipping of RX buffers to separate function Sasha Levin
2025-04-29 23:53 ` Sasha Levin [this message]
2025-04-30 12:57 ` [PATCH AUTOSEL 6.1 07/10] drm/amdgpu: Allow P2P access through XGMI Alex Deucher
2025-04-29 23:53 ` [PATCH AUTOSEL 6.1 08/10] selftests/bpf: Mitigate sockmap_ktls disconnect_after_delete failure Sasha Levin
2025-04-29 23:53 ` [PATCH AUTOSEL 6.1 09/10] bpf: fix possible endless loop in BPF map iteration Sasha Levin
2025-04-29 23:53 ` [PATCH AUTOSEL 6.1 10/10] samples/bpf: Fix compilation failure for samples/bpf on LoongArch Fedora Sasha Levin
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=20250429235316.538129-7-sashal@kernel.org \
--to=sashal@kernel.org \
--cc=Yunxiang.Li@amd.com \
--cc=airlied@gmail.com \
--cc=alexander.deucher@amd.com \
--cc=amd-gfx@lists.freedesktop.org \
--cc=christian.koenig@amd.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=felix.kuehling@amd.com \
--cc=hao.zhou@amd.com \
--cc=linux-kernel@vger.kernel.org \
--cc=matthew.auld@intel.com \
--cc=simona@ffwll.ch \
--cc=stable@vger.kernel.org \
--cc=tvrtko.ursulin@igalia.com \
/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