drm/amdgpu: Implement new guest side VF2PF message transaction (v2)

- Refactor the driver code to use amdgpu_virt_read_pf2vf_data
  and amdgpu_virt_write_vf2pf_data instead of writing all code in
  one function (which is the old amdgpu_virt_init_data_exchange)

- Adding a new transaction method for VF2PF message between host
  and guest driver. Guest side will periodically update VF2PF
  message in the framebuffer.

  In the new header, we include guest ucode information, guest
  framebuffer usage, and engine usage

- Clean up the old macros since they will cause compile error if
  the new transaction method is used

v2: squash in build fix

Signed-off-by: Bokun Zhang <Bokun.Zhang@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

1 files changed, 189 insertions, 54 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index f76961d17246..d0aea5e39531 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -31,6 +31,12 @@
 #include "soc15.h"
 #include "nv.h"
 
+#define POPULATE_UCODE_INFO(vf2pf_info, ucode, ver) \
+	do { \
+		vf2pf_info->ucode_info[ucode].id = ucode; \
+		vf2pf_info->ucode_info[ucode].version = ver; \
+	} while (0)
+
 bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev)
 {
 	/* By now all MMIO pages except mailbox are blocked */
@@ -239,10 +245,10 @@ void amdgpu_virt_free_mm_table(struct amdgpu_device *adev)
 }
 
 
-int amdgpu_virt_fw_reserve_get_checksum(void *obj,
-					unsigned long obj_size,
-					unsigned int key,
-					unsigned int chksum)
+unsigned int amd_sriov_msg_checksum(void *obj,
+				unsigned long obj_size,
+				unsigned int key,
+				unsigned int checksum)
 {
 	unsigned int ret = key;
 	unsigned long i = 0;
@@ -252,9 +258,9 @@ int amdgpu_virt_fw_reserve_get_checksum(void *obj,
 	/* calculate checksum */
 	for (i = 0; i < obj_size; ++i)
 		ret += *(pos + i);
-	/* minus the chksum itself */
-	pos = (char *)&chksum;
-	for (i = 0; i < sizeof(chksum); ++i)
+	/* minus the checksum itself */
+	pos = (char *)&checksum;
+	for (i = 0; i < sizeof(checksum); ++i)
 		ret -= *(pos + i);
 	return ret;
 }
@@ -415,33 +421,188 @@ static void amdgpu_virt_add_bad_page(struct amdgpu_device *adev,
 	}
 }
 
-void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
+static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 {
-	uint32_t pf2vf_size = 0;
-	uint32_t checksum = 0;
+	struct amd_sriov_msg_pf2vf_info_header *pf2vf_info = adev->virt.fw_reserve.p_pf2vf;
+	uint32_t checksum;
 	uint32_t checkval;
-	char *str;
+
+	if (adev->virt.fw_reserve.p_pf2vf == NULL)
+		return -EINVAL;
+
+	if (pf2vf_info->size > 1024) {
+		DRM_ERROR("invalid pf2vf message size\n");
+		return -EINVAL;
+	}
+
+	switch (pf2vf_info->version) {
+	case 1:
+		checksum = ((struct amdgim_pf2vf_info_v1 *)pf2vf_info)->checksum;
+		checkval = amd_sriov_msg_checksum(
+			adev->virt.fw_reserve.p_pf2vf, pf2vf_info->size,
+			adev->virt.fw_reserve.checksum_key, checksum);
+		if (checksum != checkval) {
+			DRM_ERROR("invalid pf2vf message\n");
+			return -EINVAL;
+		}
+
+		adev->virt.gim_feature =
+			((struct amdgim_pf2vf_info_v1 *)pf2vf_info)->feature_flags;
+		break;
+	case 2:
+		/* TODO: missing key, need to add it later */
+		checksum = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->checksum;
+		checkval = amd_sriov_msg_checksum(
+			adev->virt.fw_reserve.p_pf2vf, pf2vf_info->size,
+			0, checksum);
+		if (checksum != checkval) {
+			DRM_ERROR("invalid pf2vf message\n");
+			return -EINVAL;
+		}
+
+		adev->virt.vf2pf_update_interval_ms =
+			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->vf2pf_update_interval_ms;
+		adev->virt.gim_feature =
+			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->feature_flags.all;
+
+		break;
+	default:
+		DRM_ERROR("invalid pf2vf version\n");
+		return -EINVAL;
+	}
+
+	/* correct too large or too little interval value */
+	if (adev->virt.vf2pf_update_interval_ms < 200 || adev->virt.vf2pf_update_interval_ms > 10000)
+		adev->virt.vf2pf_update_interval_ms = 2000;
+
+	return 0;
+}
+
+static void amdgpu_virt_populate_vf2pf_ucode_info(struct amdgpu_device *adev)
+{
+	struct amd_sriov_msg_vf2pf_info *vf2pf_info;
+	vf2pf_info = (struct amd_sriov_msg_vf2pf_info *) adev->virt.fw_reserve.p_vf2pf;
+
+	if (adev->virt.fw_reserve.p_vf2pf == NULL)
+		return;
+
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_VCE,      adev->vce.fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_UVD,      adev->uvd.fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MC,       adev->gmc.fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_ME,       adev->gfx.me_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_PFP,      adev->gfx.pfp_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_CE,       adev->gfx.ce_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_RLC,      adev->gfx.rlc_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_RLC_SRLC, adev->gfx.rlc_srlc_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_RLC_SRLG, adev->gfx.rlc_srlg_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_RLC_SRLS, adev->gfx.rlc_srls_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MEC,      adev->gfx.mec_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MEC2,     adev->gfx.mec2_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_SOS,      adev->psp.sos_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_ASD,      adev->psp.asd_fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_TA_RAS,   adev->psp.ta_ras_ucode_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_TA_XGMI,  adev->psp.ta_xgmi_ucode_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_SMC,      adev->pm.fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_SDMA,     adev->sdma.instance[0].fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_SDMA2,    adev->sdma.instance[1].fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_VCN,      adev->vcn.fw_version);
+	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_DMCU,     adev->dm.dmcu_fw_version);
+}
+
+static int amdgpu_virt_write_vf2pf_data(struct amdgpu_device *adev)
+{
+	struct amd_sriov_msg_vf2pf_info *vf2pf_info;
+	struct ttm_resource_manager *vram_man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
+
+	vf2pf_info = (struct amd_sriov_msg_vf2pf_info *) adev->virt.fw_reserve.p_vf2pf;
+
+	if (adev->virt.fw_reserve.p_vf2pf == NULL)
+		return -EINVAL;
+
+	memset(vf2pf_info, 0, sizeof(struct amd_sriov_msg_vf2pf_info));
+
+	vf2pf_info->header.size = sizeof(struct amd_sriov_msg_vf2pf_info);
+	vf2pf_info->header.version = AMD_SRIOV_MSG_FW_VRAM_VF2PF_VER;
+
+#ifdef MODULE
+	if (THIS_MODULE->version != NULL)
+		strcpy(vf2pf_info->driver_version, THIS_MODULE->version);
+	else
+#endif
+		strcpy(vf2pf_info->driver_version, "N/A");
+
+	vf2pf_info->pf2vf_version_required = 0; // no requirement, guest understands all
+	vf2pf_info->driver_cert = 0;
+	vf2pf_info->os_info.all = 0;
+
+	vf2pf_info->fb_usage = amdgpu_vram_mgr_usage(vram_man) >> 20;
+	vf2pf_info->fb_vis_usage = amdgpu_vram_mgr_vis_usage(vram_man) >> 20;
+	vf2pf_info->fb_size = adev->gmc.real_vram_size >> 20;
+	vf2pf_info->fb_vis_size = adev->gmc.visible_vram_size >> 20;
+
+	amdgpu_virt_populate_vf2pf_ucode_info(adev);
+
+	/* TODO: read dynamic info */
+	vf2pf_info->gfx_usage = 0;
+	vf2pf_info->compute_usage = 0;
+	vf2pf_info->encode_usage = 0;
+	vf2pf_info->decode_usage = 0;
+
+	vf2pf_info->checksum =
+		amd_sriov_msg_checksum(
+		vf2pf_info, vf2pf_info->header.size, 0, 0);
+
+	return 0;
+}
+
+void amdgpu_virt_update_vf2pf_work_item(struct work_struct *work)
+{
+	struct amdgpu_device *adev = container_of(work, struct amdgpu_device, virt.vf2pf_work.work);
+
+	amdgpu_virt_read_pf2vf_data(adev);
+	amdgpu_virt_write_vf2pf_data(adev);
+
+	schedule_delayed_work(&(adev->virt.vf2pf_work), adev->virt.vf2pf_update_interval_ms);
+}
+
+void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev)
+{
+	if (adev->virt.vf2pf_update_interval_ms != 0) {
+		DRM_INFO("clean up the vf2pf work item\n");
+		flush_delayed_work(&adev->virt.vf2pf_work);
+		cancel_delayed_work_sync(&adev->virt.vf2pf_work);
+	}
+}
+
+void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
+{
 	uint64_t bp_block_offset = 0;
 	uint32_t bp_block_size = 0;
-	struct amdgim_pf2vf_info_v2 *pf2vf_v2 = NULL;
+	struct amd_sriov_msg_pf2vf_info *pf2vf_v2 = NULL;
 
 	adev->virt.fw_reserve.p_pf2vf = NULL;
 	adev->virt.fw_reserve.p_vf2pf = NULL;
+	adev->virt.vf2pf_update_interval_ms = 0;
 
 	if (adev->mman.fw_vram_usage_va != NULL) {
+		adev->virt.vf2pf_update_interval_ms = 2000;
+
 		adev->virt.fw_reserve.p_pf2vf =
-			(struct amd_sriov_msg_pf2vf_info_header *)(
-			adev->mman.fw_vram_usage_va + AMDGIM_DATAEXCHANGE_OFFSET);
-		AMDGPU_FW_VRAM_PF2VF_READ(adev, header.size, &pf2vf_size);
-		AMDGPU_FW_VRAM_PF2VF_READ(adev, checksum, &checksum);
-		AMDGPU_FW_VRAM_PF2VF_READ(adev, feature_flags, &adev->virt.gim_feature);
-
-		/* pf2vf message must be in 4K */
-		if (pf2vf_size > 0 && pf2vf_size < 4096) {
-			if (adev->virt.fw_reserve.p_pf2vf->version == 2) {
-				pf2vf_v2 = (struct amdgim_pf2vf_info_v2 *)adev->virt.fw_reserve.p_pf2vf;
-				bp_block_offset = ((uint64_t)pf2vf_v2->bp_block_offset_L & 0xFFFFFFFF) |
-						((((uint64_t)pf2vf_v2->bp_block_offset_H) << 32) & 0xFFFFFFFF00000000);
+			(struct amd_sriov_msg_pf2vf_info_header *)
+			(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));
+		adev->virt.fw_reserve.p_vf2pf =
+			(struct amd_sriov_msg_vf2pf_info_header *)
+			(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB << 10));
+
+		amdgpu_virt_read_pf2vf_data(adev);
+		amdgpu_virt_write_vf2pf_data(adev);
+
+		/* bad page handling for version 2 */
+		if (adev->virt.fw_reserve.p_pf2vf->version == 2) {
+				pf2vf_v2 = (struct amd_sriov_msg_pf2vf_info *)adev->virt.fw_reserve.p_pf2vf;
+
+				bp_block_offset = ((uint64_t)pf2vf_v2->bp_block_offset_low & 0xFFFFFFFF) |
+						((((uint64_t)pf2vf_v2->bp_block_offset_high) << 32) & 0xFFFFFFFF00000000);
 				bp_block_size = pf2vf_v2->bp_block_size;
 
 				if (bp_block_size && !adev->virt.ras_init_done)
@@ -450,37 +611,11 @@ void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
 				if (adev->virt.ras_init_done)
 					amdgpu_virt_add_bad_page(adev, bp_block_offset, bp_block_size);
 			}
+	}
 
-			checkval = amdgpu_virt_fw_reserve_get_checksum(
-				adev->virt.fw_reserve.p_pf2vf, pf2vf_size,
-				adev->virt.fw_reserve.checksum_key, checksum);
-			if (checkval == checksum) {
-				adev->virt.fw_reserve.p_vf2pf =
-					((void *)adev->virt.fw_reserve.p_pf2vf +
-					pf2vf_size);
-				memset((void *)adev->virt.fw_reserve.p_vf2pf, 0,
-					sizeof(amdgim_vf2pf_info));
-				AMDGPU_FW_VRAM_VF2PF_WRITE(adev, header.version,
-					AMDGPU_FW_VRAM_VF2PF_VER);
-				AMDGPU_FW_VRAM_VF2PF_WRITE(adev, header.size,
-					sizeof(amdgim_vf2pf_info));
-				AMDGPU_FW_VRAM_VF2PF_READ(adev, driver_version,
-					&str);
-#ifdef MODULE
-				if (THIS_MODULE->version != NULL)
-					strcpy(str, THIS_MODULE->version);
-				else
-#endif
-					strcpy(str, "N/A");
-				AMDGPU_FW_VRAM_VF2PF_WRITE(adev, driver_cert,
-					0);
-				AMDGPU_FW_VRAM_VF2PF_WRITE(adev, checksum,
-					amdgpu_virt_fw_reserve_get_checksum(
-					adev->virt.fw_reserve.p_vf2pf,
-					pf2vf_size,
-					adev->virt.fw_reserve.checksum_key, 0));
-			}
-		}
+	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), adev->virt.vf2pf_update_interval_ms);
 	}
 }