linux-kernel - [PATCH AUTOSEL 5.10 068/217] drm/amd/amdgpu: Fix incorrect logic to increment VCN doorbell index

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 22 Dec 2020 21:13:57 -0500
From:   Sasha Levin <sashal@...nel.org>
To:     linux-kernel@...r.kernel.org, stable@...r.kernel.org
Cc:     Bokun Zhang <Bokun.Zhang@....com>, Monk Liu <monk.liu@....com>,
        Alex Deucher <alexander.deucher@....com>,
        Sasha Levin <sashal@...nel.org>, amd-gfx@...ts.freedesktop.org,
        dri-devel@...ts.freedesktop.org
Subject: [PATCH AUTOSEL 5.10 068/217] drm/amd/amdgpu: Fix incorrect logic to increment VCN doorbell index

From: Bokun Zhang <Bokun.Zhang@....com>

[ Upstream commit 25a35065c066496935217748b1662a7fcf26ed58 ]

- The original logic uses a counter based index assignment,
  which is incorrect if we only assign VCN1 to this VF but no VCN0

  The doorbell index is absolute, so we can calculate it by
  using index variable i and j

Signed-off-by: Bokun Zhang <Bokun.Zhang@....com>
Reviewed-by: Monk Liu <monk.liu@....com>
Signed-off-by: Alex Deucher <alexander.deucher@....com>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
 drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
index b5f8f3d731cb0..8ecdddf33e18e 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
@@ -155,6 +155,13 @@ static int vcn_v3_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/*
+	 * Note: doorbell assignment is fixed for SRIOV multiple VCN engines
+	 * Formula:
+	 *   vcn_db_base  = adev->doorbell_index.vcn.vcn_ring0_1 << 1;
+	 *   dec_ring_i   = vcn_db_base + i * (adev->vcn.num_enc_rings + 1)
+	 *   enc_ring_i,j = vcn_db_base + i * (adev->vcn.num_enc_rings + 1) + 1 + j
+	 */
 	if (amdgpu_sriov_vf(adev)) {
 		vcn_doorbell_index = adev->doorbell_index.vcn.vcn_ring0_1;
 		/* get DWORD offset */
@@ -192,9 +199,7 @@ static int vcn_v3_0_sw_init(void *handle)
 		ring = &adev->vcn.inst[i].ring_dec;
 		ring->use_doorbell = true;
 		if (amdgpu_sriov_vf(adev)) {
-			ring->doorbell_index = vcn_doorbell_index;
-			/* NOTE: increment so next VCN engine use next DOORBELL DWORD */
-			vcn_doorbell_index++;
+			ring->doorbell_index = vcn_doorbell_index + i * (adev->vcn.num_enc_rings + 1);
 		} else {
 			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i;
 		}
@@ -216,9 +221,7 @@ static int vcn_v3_0_sw_init(void *handle)
 			ring = &adev->vcn.inst[i].ring_enc[j];
 			ring->use_doorbell = true;
 			if (amdgpu_sriov_vf(adev)) {
-				ring->doorbell_index = vcn_doorbell_index;
-				/* NOTE: increment so next VCN engine use next DOORBELL DWORD */
-				vcn_doorbell_index++;
+				ring->doorbell_index = vcn_doorbell_index + i * (adev->vcn.num_enc_rings + 1) + 1 + j;
 			} else {
 				ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + j + 8 * i;
 			}
-- 
2.27.0