[PATCH v4 5/7] drm/amdgpu/jpeg: sriov support for jpeg_v4_0_3
Samir Dhume
samir.dhume at amd.com
Tue Aug 8 16:26:20 UTC 2023
initialization table handshake with mmsch
Signed-off-by: Samir Dhume <samir.dhume at amd.com>
---
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c | 174 ++++++++++++++++++++---
1 file changed, 153 insertions(+), 21 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c
index ce2b22f7e4e4..33f04ea8549f 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c
@@ -26,6 +26,7 @@
#include "soc15.h"
#include "soc15d.h"
#include "jpeg_v4_0_3.h"
+#include "mmsch_v4_0_3.h"
#include "vcn/vcn_4_0_3_offset.h"
#include "vcn/vcn_4_0_3_sh_mask.h"
@@ -41,6 +42,7 @@ static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev);
static int jpeg_v4_0_3_set_powergating_state(void *handle,
enum amd_powergating_state state);
static void jpeg_v4_0_3_set_ras_funcs(struct amdgpu_device *adev);
+static void jpeg_v4_0_3_dec_ring_set_wptr(struct amdgpu_ring *ring);
static int amdgpu_ih_srcid_jpeg[] = {
VCN_4_0__SRCID__JPEG_DECODE,
@@ -160,6 +162,120 @@ static int jpeg_v4_0_3_sw_fini(void *handle)
return r;
}
+static int jpeg_v4_0_3_start_sriov(struct amdgpu_device *adev)
+{
+ struct amdgpu_ring *ring;
+ uint64_t ctx_addr;
+ uint32_t param, resp, expected;
+ uint32_t tmp, timeout;
+
+ struct amdgpu_mm_table *table = &adev->virt.mm_table;
+ uint32_t *table_loc;
+ uint32_t table_size;
+ uint32_t size, size_dw, item_offset;
+ uint32_t init_status;
+ int i, j, jpeg_inst;
+
+ struct mmsch_v4_0_cmd_direct_write
+ direct_wt = { {0} };
+ struct mmsch_v4_0_cmd_end end = { {0} };
+ struct mmsch_v4_0_3_init_header header;
+
+ direct_wt.cmd_header.command_type =
+ MMSCH_COMMAND__DIRECT_REG_WRITE;
+ end.cmd_header.command_type =
+ MMSCH_COMMAND__END;
+
+ for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+ jpeg_inst = GET_INST(JPEG, i);
+
+ memset(&header, 0, sizeof(struct mmsch_v4_0_3_init_header));
+ header.version = MMSCH_VERSION;
+ header.total_size = sizeof(struct mmsch_v4_0_3_init_header) >> 2;
+
+ table_loc = (uint32_t *)table->cpu_addr;
+ table_loc += header.total_size;
+
+ item_offset = header.total_size;
+
+
+ for (j = 0; j < adev->jpeg.num_jpeg_rings; j++) {
+ ring = &adev->jpeg.inst[i].ring_dec[j];
+ table_size = 0;
+
+ tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_LOW);
+ MMSCH_V4_0_INSERT_DIRECT_WT(tmp, lower_32_bits(ring->gpu_addr));
+ tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_HIGH);
+ MMSCH_V4_0_INSERT_DIRECT_WT(tmp, upper_32_bits(ring->gpu_addr));
+ tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_SIZE);
+ MMSCH_V4_0_INSERT_DIRECT_WT(tmp, ring->ring_size / 4);
+
+ if (j <= 3) {
+ header.mjpegdec0[j].table_offset = item_offset;
+ header.mjpegdec0[j].init_status = 0;
+ header.mjpegdec0[j].table_size = table_size;
+ } else {
+ header.mjpegdec1[j-4].table_offset = item_offset;
+ header.mjpegdec1[j-4].init_status = 0;
+ header.mjpegdec1[j-4].table_size = table_size;
+ }
+ header.total_size += table_size;
+ item_offset+= table_size;
+ }
+
+ MMSCH_V4_0_INSERT_END();
+
+ /* send init table to MMSCH */
+ size = sizeof(struct mmsch_v4_0_3_init_header);
+ table_loc = (uint32_t *)table->cpu_addr;
+ memcpy((void *)table_loc, &header, size);
+
+ ctx_addr = table->gpu_addr;
+ WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_ADDR_LO, lower_32_bits(ctx_addr));
+ WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_ADDR_HI, upper_32_bits(ctx_addr));
+
+ tmp = RREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_VMID);
+ tmp &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK;
+ tmp |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT);
+ WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_VMID, tmp);
+
+ size = header.total_size;
+ WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_SIZE, size);
+
+ WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_RESP, 0);
+
+ param = 0x00000001;
+ WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_HOST, param);
+ tmp = 0;
+ timeout = 1000;
+ resp = 0;
+ expected = MMSCH_VF_MAILBOX_RESP__OK;
+ init_status = ((struct mmsch_v4_0_3_init_header *)(table_loc))->mjpegdec0[i].init_status;
+ while (resp != expected) {
+ resp = RREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_RESP);
+
+ if (resp != 0)
+ break;
+ udelay(10);
+ tmp = tmp + 10;
+ if (tmp >= timeout) {
+ DRM_ERROR("failed to init MMSCH. TIME-OUT after %d usec"\
+ " waiting for regMMSCH_VF_MAILBOX_RESP "\
+ "(expected=0x%08x, readback=0x%08x)\n",
+ tmp, expected, resp);
+ return -EBUSY;
+ }
+ }
+ if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE && init_status != MMSCH_VF_ENGINE_STATUS__PASS) {
+ DRM_ERROR("MMSCH init status is incorrect! readback=0x%08x, header init status for jpeg: %x\n", resp, init_status);
+ }
+
+
+ }
+ return 0;
+}
+
+
/**
* jpeg_v4_0_3_hw_init - start and test JPEG block
*
@@ -172,31 +288,47 @@ static int jpeg_v4_0_3_hw_init(void *handle)
struct amdgpu_ring *ring;
int i, j, r, jpeg_inst;
- for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
- jpeg_inst = GET_INST(JPEG, i);
+ if (amdgpu_sriov_vf(adev)) {
+ r = jpeg_v4_0_3_start_sriov(adev);
+ if (r)
+ return r;
- ring = adev->jpeg.inst[i].ring_dec;
+ for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+ for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+ ring = &adev->jpeg.inst[i].ring_dec[j];
+ ring->wptr = 0;
+ ring->wptr_old = 0;
+ jpeg_v4_0_3_dec_ring_set_wptr(ring);
+ ring->sched.ready = true;
+ }
+ }
+ } else {
+ for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+ jpeg_inst = GET_INST(JPEG, i);
- if (ring->use_doorbell)
- adev->nbio.funcs->vcn_doorbell_range(
- adev, ring->use_doorbell,
- (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
- 9 * jpeg_inst,
- adev->jpeg.inst[i].aid_id);
+ ring = adev->jpeg.inst[i].ring_dec;
- for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
- ring = &adev->jpeg.inst[i].ring_dec[j];
if (ring->use_doorbell)
- WREG32_SOC15_OFFSET(
- VCN, GET_INST(VCN, i),
- regVCN_JPEG_DB_CTRL,
- (ring->pipe ? (ring->pipe - 0x15) : 0),
- ring->doorbell_index
- << VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
- VCN_JPEG_DB_CTRL__EN_MASK);
- r = amdgpu_ring_test_helper(ring);
- if (r)
- return r;
+ adev->nbio.funcs->vcn_doorbell_range(
+ adev, ring->use_doorbell,
+ (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+ 9 * jpeg_inst,
+ adev->jpeg.inst[i].aid_id);
+
+ for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+ ring = &adev->jpeg.inst[i].ring_dec[j];
+ if (ring->use_doorbell)
+ WREG32_SOC15_OFFSET(
+ VCN, GET_INST(VCN, i),
+ regVCN_JPEG_DB_CTRL,
+ (ring->pipe ? (ring->pipe - 0x15) : 0),
+ ring->doorbell_index
+ << VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
+ VCN_JPEG_DB_CTRL__EN_MASK);
+ r = amdgpu_ring_test_helper(ring);
+ if (r)
+ return r;
+ }
}
}
DRM_DEV_INFO(adev->dev, "JPEG decode initialized successfully.\n");
--
2.34.1
More information about the amd-gfx
mailing list