[PATCH 02/10] drm/amdgpu: Add sdma instance specific functions
Alex Deucher
alexander.deucher at amd.com
Thu Mar 30 19:31:50 UTC 2023
From: Lijo Lazar <lijo.lazar at amd.com>
SDMA 4.4.2 supports multiple instances. Add functions to support
handling of each SDMA instance separately.
Signed-off-by: Lijo Lazar <lijo.lazar at amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang at amd.com>
Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 162 +++++++++++++----------
1 file changed, 91 insertions(+), 71 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 016813b295ac..6a971e15b4b0 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -94,7 +94,8 @@ static int sdma_v4_4_2_irq_id_to_seq(unsigned client_id)
}
}
-static void sdma_v4_4_2_init_golden_registers(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_init_golden_registers(struct amdgpu_device *adev,
+ uint32_t inst_mask)
{
u32 val;
int i;
@@ -418,13 +419,14 @@ static void sdma_v4_4_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
*
* Stop the gfx async dma ring buffers.
*/
-static void sdma_v4_4_2_gfx_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_gfx_stop(struct amdgpu_device *adev,
+ uint32_t inst_mask)
{
struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES];
u32 rb_cntl, ib_cntl;
int i, unset = 0;
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ for_each_inst(i, inst_mask) {
sdma[i] = &adev->sdma.instance[i].ring;
if ((adev->mman.buffer_funcs_ring == sdma[i]) && unset != 1) {
@@ -448,7 +450,8 @@ static void sdma_v4_4_2_gfx_stop(struct amdgpu_device *adev)
*
* Stop the compute async dma queues.
*/
-static void sdma_v4_4_2_rlc_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_rlc_stop(struct amdgpu_device *adev,
+ uint32_t inst_mask)
{
/* XXX todo */
}
@@ -460,14 +463,15 @@ static void sdma_v4_4_2_rlc_stop(struct amdgpu_device *adev)
*
* Stop the page async dma ring buffers.
*/
-static void sdma_v4_4_2_page_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_page_stop(struct amdgpu_device *adev,
+ uint32_t inst_mask)
{
struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES];
u32 rb_cntl, ib_cntl;
int i;
bool unset = false;
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ for_each_inst(i, inst_mask) {
sdma[i] = &adev->sdma.instance[i].page;
if ((adev->mman.buffer_funcs_ring == sdma[i]) &&
@@ -495,7 +499,8 @@ static void sdma_v4_4_2_page_stop(struct amdgpu_device *adev)
*
* Halt or unhalt the async dma engines context switch.
*/
-static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
+static void sdma_v4_4_2_inst_ctx_switch_enable(struct amdgpu_device *adev,
+ bool enable, uint32_t inst_mask)
{
u32 f32_cntl, phase_quantum = 0;
int i;
@@ -524,7 +529,7 @@ static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enabl
unit << SDMA_PHASE0_QUANTUM__UNIT__SHIFT;
}
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ for_each_inst(i, inst_mask) {
f32_cntl = RREG32_SDMA(i, regSDMA_CNTL);
f32_cntl = REG_SET_FIELD(f32_cntl, SDMA_CNTL,
AUTO_CTXSW_ENABLE, enable ? 1 : 0);
@@ -538,7 +543,6 @@ static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enabl
/* Extend page fault timeout to avoid interrupt storm */
WREG32_SDMA(i, regSDMA_UTCL1_TIMEOUT, 0x00800080);
}
-
}
/**
@@ -546,22 +550,24 @@ static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enabl
*
* @adev: amdgpu_device pointer
* @enable: enable/disable the DMA MEs.
+ * @inst_mask: mask of dma engine instances to be enabled
*
* Halt or unhalt the async dma engines.
*/
-static void sdma_v4_4_2_enable(struct amdgpu_device *adev, bool enable)
+static void sdma_v4_4_2_inst_enable(struct amdgpu_device *adev, bool enable,
+ uint32_t inst_mask)
{
u32 f32_cntl;
int i;
if (!enable) {
- sdma_v4_4_2_gfx_stop(adev);
- sdma_v4_4_2_rlc_stop(adev);
+ sdma_v4_4_2_inst_gfx_stop(adev, inst_mask);
+ sdma_v4_4_2_inst_rlc_stop(adev, inst_mask);
if (adev->sdma.has_page_queue)
- sdma_v4_4_2_page_stop(adev);
+ sdma_v4_4_2_inst_page_stop(adev, inst_mask);
}
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ for_each_inst(i, inst_mask) {
f32_cntl = RREG32_SDMA(i, regSDMA_F32_CNTL);
f32_cntl = REG_SET_FIELD(f32_cntl, SDMA_F32_CNTL, HALT, enable ? 0 : 1);
WREG32_SDMA(i, regSDMA_F32_CNTL, f32_cntl);
@@ -780,7 +786,8 @@ static void sdma_v4_4_2_init_pg(struct amdgpu_device *adev)
* Set up the compute DMA queues and enable them.
* Returns 0 for success, error for failure.
*/
-static int sdma_v4_4_2_rlc_resume(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_rlc_resume(struct amdgpu_device *adev,
+ uint32_t inst_mask)
{
sdma_v4_4_2_init_pg(adev);
@@ -795,7 +802,8 @@ static int sdma_v4_4_2_rlc_resume(struct amdgpu_device *adev)
* Loads the sDMA0/1 ucode.
* Returns 0 for success, -EINVAL if the ucode is not available.
*/
-static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_load_microcode(struct amdgpu_device *adev,
+ uint32_t inst_mask)
{
const struct sdma_firmware_header_v1_0 *hdr;
const __le32 *fw_data;
@@ -803,9 +811,9 @@ static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
int i, j;
/* halt the MEs */
- sdma_v4_4_2_enable(adev, false);
+ sdma_v4_4_2_inst_enable(adev, false, inst_mask);
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ for_each_inst(i, inst_mask) {
if (!adev->sdma.instance[i].fw)
return -EINVAL;
@@ -831,38 +839,41 @@ static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
}
/**
- * sdma_v4_4_2_start - setup and start the async dma engines
+ * sdma_v4_4_2_inst_start - setup and start the async dma engines
*
* @adev: amdgpu_device pointer
*
* Set up the DMA engines and enable them.
* Returns 0 for success, error for failure.
*/
-static int sdma_v4_4_2_start(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev,
+ uint32_t inst_mask)
{
struct amdgpu_ring *ring;
+ uint32_t tmp_mask;
int i, r = 0;
if (amdgpu_sriov_vf(adev)) {
- sdma_v4_4_2_ctx_switch_enable(adev, false);
- sdma_v4_4_2_enable(adev, false);
+ sdma_v4_4_2_inst_ctx_switch_enable(adev, false, inst_mask);
+ sdma_v4_4_2_inst_enable(adev, false, inst_mask);
} else {
/* bypass sdma microcode loading on Gopher */
if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP &&
- !(adev->pdev->device == 0x49) && !(adev->pdev->device == 0x50)) {
- r = sdma_v4_4_2_load_microcode(adev);
+ adev->sdma.instance[0].fw) {
+ r = sdma_v4_4_2_inst_load_microcode(adev, inst_mask);
if (r)
return r;
}
/* unhalt the MEs */
- sdma_v4_4_2_enable(adev, true);
+ sdma_v4_4_2_inst_enable(adev, true, inst_mask);
/* enable sdma ring preemption */
- sdma_v4_4_2_ctx_switch_enable(adev, true);
+ sdma_v4_4_2_inst_ctx_switch_enable(adev, true, inst_mask);
}
/* start the gfx rings and rlc compute queues */
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ tmp_mask = inst_mask;
+ for_each_inst(i, tmp_mask) {
uint32_t temp;
WREG32_SDMA(i, regSDMA_SEM_WAIT_FAIL_TIMER_CNTL, 0);
@@ -889,15 +900,16 @@ static int sdma_v4_4_2_start(struct amdgpu_device *adev)
}
if (amdgpu_sriov_vf(adev)) {
- sdma_v4_4_2_ctx_switch_enable(adev, true);
- sdma_v4_4_2_enable(adev, true);
+ sdma_v4_4_2_inst_ctx_switch_enable(adev, true, inst_mask);
+ sdma_v4_4_2_inst_enable(adev, true, inst_mask);
} else {
- r = sdma_v4_4_2_rlc_resume(adev);
+ r = sdma_v4_4_2_inst_rlc_resume(adev, inst_mask);
if (r)
return r;
}
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ tmp_mask = inst_mask;
+ for_each_inst(i, tmp_mask) {
ring = &adev->sdma.instance[i].ring;
r = amdgpu_ring_test_helper(ring);
@@ -1383,14 +1395,17 @@ static int sdma_v4_4_2_hw_init(void *handle)
{
int r;
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ uint32_t inst_mask;
+ /* TODO: Check if this is needed */
if (adev->flags & AMD_IS_APU)
amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, false);
+ inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
if (!amdgpu_sriov_vf(adev))
- sdma_v4_4_2_init_golden_registers(adev);
+ sdma_v4_4_2_inst_init_golden_registers(adev, inst_mask);
- r = sdma_v4_4_2_start(adev);
+ r = sdma_v4_4_2_inst_start(adev, inst_mask);
return r;
}
@@ -1398,22 +1413,27 @@ static int sdma_v4_4_2_hw_init(void *handle)
static int sdma_v4_4_2_hw_fini(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ uint32_t inst_mask;
int i;
if (amdgpu_sriov_vf(adev))
return 0;
+ inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
for (i = 0; i < adev->sdma.num_instances; i++) {
amdgpu_irq_put(adev, &adev->sdma.ecc_irq,
AMDGPU_SDMA_IRQ_INSTANCE0 + i);
}
- sdma_v4_4_2_ctx_switch_enable(adev, false);
- sdma_v4_4_2_enable(adev, false);
+ sdma_v4_4_2_inst_ctx_switch_enable(adev, false, inst_mask);
+ sdma_v4_4_2_inst_enable(adev, false, inst_mask);
return 0;
}
+static int sdma_v4_4_2_set_clockgating_state(void *handle,
+ enum amd_clockgating_state state);
+
static int sdma_v4_4_2_suspend(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -1650,15 +1670,39 @@ static int sdma_v4_4_2_process_srbm_write_irq(struct amdgpu_device *adev,
return 0;
}
-static void sdma_v4_4_2_update_medium_grain_clock_gating(
- struct amdgpu_device *adev,
- bool enable)
+static void sdma_v4_4_2_inst_update_medium_grain_light_sleep(
+ struct amdgpu_device *adev, bool enable, uint32_t inst_mask)
+{
+ uint32_t data, def;
+ int i;
+
+ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) {
+ for_each_inst(i, inst_mask) {
+ /* 1-not override: enable sdma mem light sleep */
+ def = data = RREG32_SDMA(i, regSDMA_POWER_CNTL);
+ data |= SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
+ if (def != data)
+ WREG32_SDMA(i, regSDMA_POWER_CNTL, data);
+ }
+ } else {
+ for_each_inst(i, inst_mask) {
+ /* 0-override:disable sdma mem light sleep */
+ def = data = RREG32_SDMA(i, regSDMA_POWER_CNTL);
+ data &= ~SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
+ if (def != data)
+ WREG32_SDMA(i, regSDMA_POWER_CNTL, data);
+ }
+ }
+}
+
+static void sdma_v4_4_2_inst_update_medium_grain_clock_gating(
+ struct amdgpu_device *adev, bool enable, uint32_t inst_mask)
{
uint32_t data, def;
int i;
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG)) {
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ for_each_inst(i, inst_mask) {
def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
data &= ~(SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
@@ -1672,7 +1716,7 @@ static void sdma_v4_4_2_update_medium_grain_clock_gating(
WREG32_SDMA(i, regSDMA_CLK_CTRL, data);
}
} else {
- for (i = 0; i < adev->sdma.num_instances; i++) {
+ for_each_inst(i, inst_mask) {
def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
data |= (SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
@@ -1688,45 +1732,21 @@ static void sdma_v4_4_2_update_medium_grain_clock_gating(
}
}
-
-static void sdma_v4_4_2_update_medium_grain_light_sleep(
- struct amdgpu_device *adev,
- bool enable)
-{
- uint32_t data, def;
- int i;
-
- if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) {
- for (i = 0; i < adev->sdma.num_instances; i++) {
- /* 1-not override: enable sdma mem light sleep */
- def = data = RREG32_SDMA(0, regSDMA_POWER_CNTL);
- data |= SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
- if (def != data)
- WREG32_SDMA(0, regSDMA_POWER_CNTL, data);
- }
- } else {
- for (i = 0; i < adev->sdma.num_instances; i++) {
- /* 0-override:disable sdma mem light sleep */
- def = data = RREG32_SDMA(0, regSDMA_POWER_CNTL);
- data &= ~SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
- if (def != data)
- WREG32_SDMA(0, regSDMA_POWER_CNTL, data);
- }
- }
-}
-
static int sdma_v4_4_2_set_clockgating_state(void *handle,
enum amd_clockgating_state state)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ uint32_t inst_mask;
if (amdgpu_sriov_vf(adev))
return 0;
- sdma_v4_4_2_update_medium_grain_clock_gating(adev,
- state == AMD_CG_STATE_GATE);
- sdma_v4_4_2_update_medium_grain_light_sleep(adev,
- state == AMD_CG_STATE_GATE);
+ inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
+
+ sdma_v4_4_2_inst_update_medium_grain_clock_gating(
+ adev, state == AMD_CG_STATE_GATE, inst_mask);
+ sdma_v4_4_2_inst_update_medium_grain_light_sleep(
+ adev, state == AMD_CG_STATE_GATE, inst_mask);
return 0;
}
--
2.39.2
More information about the amd-gfx
mailing list