Merge branch 'drm-next-4.21' of git://people.freedesktop.org/~agd5f/linux into drm-next

New features for 4.21:
amdgpu:
- Support for SDMA paging queue on vega
- Put compute EOP buffers into vram for better performance
- Share more code with amdkfd
- Support for scanout with DCC on gfx9
- Initial kerneldoc for DC
- Updated SMU firmware support for gfx8 chips
- Rework CSA handling for eventual support for preemption
- XGMI PSP support
- Clean up RLC handling
- Enable GPU reset by default on VI, SOC15 dGPUs
- Ring and IB test cleanups

amdkfd:
- Share more code with amdgpu

ttm:
- Move global init out of the drivers

scheduler:
- Track if schedulers are ready for work
- Timeout/fault handling changes to facilitate GPU recovery

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181114165113.3751-1-alexander.deucher@amd.com

1 files changed, 216 insertions, 7 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
index 25d2f3e757f1..e05dc66b1090 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -90,6 +90,8 @@ static int psp_sw_fini(void *handle)
 	adev->psp.sos_fw = NULL;
 	release_firmware(adev->psp.asd_fw);
 	adev->psp.asd_fw = NULL;
+	release_firmware(adev->psp.ta_fw);
+	adev->psp.ta_fw = NULL;
 	return 0;
 }
 
@@ -118,21 +120,25 @@ int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
 static int
 psp_cmd_submit_buf(struct psp_context *psp,
 		   struct amdgpu_firmware_info *ucode,
-		   struct psp_gfx_cmd_resp *cmd, uint64_t fence_mc_addr,
-		   int index)
+		   struct psp_gfx_cmd_resp *cmd, uint64_t fence_mc_addr)
 {
 	int ret;
+	int index;
 
 	memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
 
 	memcpy(psp->cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp));
 
+	index = atomic_inc_return(&psp->fence_value);
 	ret = psp_cmd_submit(psp, ucode, psp->cmd_buf_mc_addr,
 			     fence_mc_addr, index);
+	if (ret) {
+		atomic_dec(&psp->fence_value);
+		return ret;
+	}
 
-	while (*((unsigned int *)psp->fence_buf) != index) {
+	while (*((unsigned int *)psp->fence_buf) != index)
 		msleep(1);
-	}
 
 	/* the status field must be 0 after FW is loaded */
 	if (ucode && psp->cmd_buf_mem->resp.status) {
@@ -191,7 +197,7 @@ static int psp_tmr_load(struct psp_context *psp)
 			PSP_TMR_SIZE, psp->tmr_mc_addr);
 
 	ret = psp_cmd_submit_buf(psp, NULL, cmd,
-				 psp->fence_buf_mc_addr, 1);
+				 psp->fence_buf_mc_addr);
 	if (ret)
 		goto failed;
 
@@ -258,13 +264,194 @@ static int psp_asd_load(struct psp_context *psp)
 			     psp->asd_ucode_size, PSP_ASD_SHARED_MEM_SIZE);
 
 	ret = psp_cmd_submit_buf(psp, NULL, cmd,
-				 psp->fence_buf_mc_addr, 2);
+				 psp->fence_buf_mc_addr);
+
+	kfree(cmd);
+
+	return ret;
+}
+
+static void psp_prep_xgmi_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd,
+					  uint64_t xgmi_ta_mc, uint64_t xgmi_mc_shared,
+					  uint32_t xgmi_ta_size, uint32_t shared_size)
+{
+        cmd->cmd_id = GFX_CMD_ID_LOAD_TA;
+        cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(xgmi_ta_mc);
+        cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(xgmi_ta_mc);
+        cmd->cmd.cmd_load_ta.app_len = xgmi_ta_size;
+
+        cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = lower_32_bits(xgmi_mc_shared);
+        cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = upper_32_bits(xgmi_mc_shared);
+        cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size;
+}
+
+static int psp_xgmi_init_shared_buf(struct psp_context *psp)
+{
+	int ret;
+
+	/*
+	 * Allocate 16k memory aligned to 4k from Frame Buffer (local
+	 * physical) for xgmi ta <-> Driver
+	 */
+	ret = amdgpu_bo_create_kernel(psp->adev, PSP_XGMI_SHARED_MEM_SIZE,
+				      PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
+				      &psp->xgmi_context.xgmi_shared_bo,
+				      &psp->xgmi_context.xgmi_shared_mc_addr,
+				      &psp->xgmi_context.xgmi_shared_buf);
+
+	return ret;
+}
+
+static int psp_xgmi_load(struct psp_context *psp)
+{
+	int ret;
+	struct psp_gfx_cmd_resp *cmd;
+
+	/*
+	 * TODO: bypass the loading in sriov for now
+	 */
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
+
+	memset(psp->fw_pri_buf, 0, PSP_1_MEG);
+	memcpy(psp->fw_pri_buf, psp->ta_xgmi_start_addr, psp->ta_xgmi_ucode_size);
+
+	psp_prep_xgmi_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr,
+				      psp->xgmi_context.xgmi_shared_mc_addr,
+				      psp->ta_xgmi_ucode_size, PSP_XGMI_SHARED_MEM_SIZE);
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+				 psp->fence_buf_mc_addr);
+
+	if (!ret) {
+		psp->xgmi_context.initialized = 1;
+		psp->xgmi_context.session_id = cmd->resp.session_id;
+	}
+
+	kfree(cmd);
+
+	return ret;
+}
+
+static void psp_prep_xgmi_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd,
+					    uint32_t xgmi_session_id)
+{
+	cmd->cmd_id = GFX_CMD_ID_UNLOAD_TA;
+	cmd->cmd.cmd_unload_ta.session_id = xgmi_session_id;
+}
+
+static int psp_xgmi_unload(struct psp_context *psp)
+{
+	int ret;
+	struct psp_gfx_cmd_resp *cmd;
+
+	/*
+	 * TODO: bypass the unloading in sriov for now
+	 */
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
+
+	psp_prep_xgmi_ta_unload_cmd_buf(cmd, psp->xgmi_context.session_id);
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+				 psp->fence_buf_mc_addr);
 
 	kfree(cmd);
 
 	return ret;
 }
 
+static void psp_prep_xgmi_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd,
+					    uint32_t ta_cmd_id,
+					    uint32_t xgmi_session_id)
+{
+	cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD;
+	cmd->cmd.cmd_invoke_cmd.session_id = xgmi_session_id;
+	cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id;
+	/* Note: cmd_invoke_cmd.buf is not used for now */
+}
+
+int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
+{
+	int ret;
+	struct psp_gfx_cmd_resp *cmd;
+
+	/*
+	 * TODO: bypass the loading in sriov for now
+	*/
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
+
+	psp_prep_xgmi_ta_invoke_cmd_buf(cmd, ta_cmd_id,
+					psp->xgmi_context.session_id);
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+				 psp->fence_buf_mc_addr);
+
+	kfree(cmd);
+
+        return ret;
+}
+
+static int psp_xgmi_terminate(struct psp_context *psp)
+{
+	int ret;
+
+	if (!psp->xgmi_context.initialized)
+		return 0;
+
+	ret = psp_xgmi_unload(psp);
+	if (ret)
+		return ret;
+
+	psp->xgmi_context.initialized = 0;
+
+	/* free xgmi shared memory */
+	amdgpu_bo_free_kernel(&psp->xgmi_context.xgmi_shared_bo,
+			&psp->xgmi_context.xgmi_shared_mc_addr,
+			&psp->xgmi_context.xgmi_shared_buf);
+
+	return 0;
+}
+
+static int psp_xgmi_initialize(struct psp_context *psp)
+{
+	struct ta_xgmi_shared_memory *xgmi_cmd;
+	int ret;
+
+	if (!psp->xgmi_context.initialized) {
+		ret = psp_xgmi_init_shared_buf(psp);
+		if (ret)
+			return ret;
+	}
+
+	/* Load XGMI TA */
+	ret = psp_xgmi_load(psp);
+	if (ret)
+		return ret;
+
+	/* Initialize XGMI session */
+	xgmi_cmd = (struct ta_xgmi_shared_memory *)(psp->xgmi_context.xgmi_shared_buf);
+	memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory));
+	xgmi_cmd->cmd_id = TA_COMMAND_XGMI__INITIALIZE;
+
+	ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id);
+
+	return ret;
+}
+
 static int psp_hw_start(struct psp_context *psp)
 {
 	struct amdgpu_device *adev = psp->adev;
@@ -292,6 +479,15 @@ static int psp_hw_start(struct psp_context *psp)
 	if (ret)
 		return ret;
 
+	if (adev->gmc.xgmi.num_physical_nodes > 1) {
+		ret = psp_xgmi_initialize(psp);
+		/* Warning the XGMI seesion initialize failure
+		 * Instead of stop driver initialization
+		 */
+		if (ret)
+			dev_err(psp->adev->dev,
+				"XGMI: Failed to initialize XGMI session\n");
+	}
 	return 0;
 }
 
@@ -321,7 +517,7 @@ static int psp_np_fw_load(struct psp_context *psp)
 			return ret;
 
 		ret = psp_cmd_submit_buf(psp, ucode, psp->cmd,
-					 psp->fence_buf_mc_addr, i + 3);
+					 psp->fence_buf_mc_addr);
 		if (ret)
 			return ret;
 
@@ -452,6 +648,10 @@ static int psp_hw_fini(void *handle)
 	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
 		return 0;
 
+	if (adev->gmc.xgmi.num_physical_nodes > 1 &&
+	    psp->xgmi_context.initialized == 1)
+                psp_xgmi_terminate(psp);
+
 	psp_ring_destroy(psp, PSP_RING_TYPE__KM);
 
 	amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);
@@ -479,6 +679,15 @@ static int psp_suspend(void *handle)
 	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
 		return 0;
 
+	if (adev->gmc.xgmi.num_physical_nodes > 1 &&
+	    psp->xgmi_context.initialized == 1) {
+		ret = psp_xgmi_terminate(psp);
+		if (ret) {
+			DRM_ERROR("Failed to terminate xgmi ta\n");
+			return ret;
+		}
+	}
+
 	ret = psp_ring_stop(psp, PSP_RING_TYPE__KM);
 	if (ret) {
 		DRM_ERROR("PSP ring stop failed\n");