[PATCH v4] drm/amd/amdgpu: fixing read wrong pf2vf data in SRIOV

[Chen, Horace]

[AMD Official Use Only]

Reviewed-by: Horace Chen <horace.chen at amd.com>

-----Original Message-----
From: Chen, JingWen <JingWen.Chen2 at amd.com>
Sent: Friday, January 14, 2022 2:54 PM
To: amd-gfx at lists.freedesktop.org
Cc: Liu, Monk <Monk.Liu at amd.com>; Chen, Horace <Horace.Chen at amd.com>; Skvortsov, Victor <Victor.Skvortsov at amd.com>; Chen, JingWen <JingWen.Chen2 at amd.com>
Subject: [PATCH v4] drm/amd/amdgpu: fixing read wrong pf2vf data in SRIOV

[Why]
This fixes 35709bd76d229a51b0c571a768b741650ec47828.
we should read pf2vf data based at mman.fw_vram_usage_va after gmc sw_init. patch 35709bd76 breaks this logic.

[How]
calling amdgpu_virt_exchange_data in amdgpu_virt_init_data_exchange to set the right base in the right sequence.

v2:
call amdgpu_virt_init_data_exchange after gmc sw_init to make data exchange workqueue run

v3:
clean up the code logic

v4:
add some comment and make the code more readable

Signed-off-by: Jingwen Chen <Jingwen.Chen2 at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c   | 20 +++++++-------------
 2 files changed, 8 insertions(+), 14 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index da3348fa7b0e..5610eae8ef0c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -2353,7 +2353,7 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
        }

        if (amdgpu_sriov_vf(adev))
-               amdgpu_virt_exchange_data(adev);
+               amdgpu_virt_init_data_exchange(adev);

        r = amdgpu_ib_pool_init(adev);
        if (r) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index 894444ab0032..07bc0f504713 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -625,20 +625,20 @@ void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
        adev->virt.fw_reserve.p_vf2pf = NULL;
        adev->virt.vf2pf_update_interval_ms = 0;

-       if (adev->bios != NULL) {
-               adev->virt.vf2pf_update_interval_ms = 2000;
+       if (adev->mman.fw_vram_usage_va != NULL) {
+               /* go through this logic in ip_init and reset to init workqueue*/
+               amdgpu_virt_exchange_data(adev);

+               INIT_DELAYED_WORK(&adev->virt.vf2pf_work, amdgpu_virt_update_vf2pf_work_item);
+               schedule_delayed_work(&(adev->virt.vf2pf_work), msecs_to_jiffies(adev->virt.vf2pf_update_interval_ms));
+       } else if (adev->bios != NULL) {
+               /* got through this logic in early init stage to get necessary flags,
+e.g. rlcg_acc related*/
                adev->virt.fw_reserve.p_pf2vf =
                        (struct amd_sriov_msg_pf2vf_info_header *)
                        (adev->bios + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));

                amdgpu_virt_read_pf2vf_data(adev);
        }
-
-       if (adev->virt.vf2pf_update_interval_ms != 0) {
-               INIT_DELAYED_WORK(&adev->virt.vf2pf_work, amdgpu_virt_update_vf2pf_work_item);
-               schedule_delayed_work(&(adev->virt.vf2pf_work), msecs_to_jiffies(adev->virt.vf2pf_update_interval_ms));
-       }
 }


@@ -674,12 +674,6 @@ void amdgpu_virt_exchange_data(struct amdgpu_device *adev)
                                if (adev->virt.ras_init_done)
                                        amdgpu_virt_add_bad_page(adev, bp_block_offset, bp_block_size);
                        }
-       } else if (adev->bios != NULL) {
-               adev->virt.fw_reserve.p_pf2vf =
-                       (struct amd_sriov_msg_pf2vf_info_header *)
-                       (adev->bios + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));
-
-               amdgpu_virt_read_pf2vf_data(adev);
        }
 }

--
2.30.2