Addressed followup comments for MtFuji elba dpu platform (sonic-net#22058)

Addressed two action items from PR sonic-net#18536

Removed home dir from dpu_pensando_util.py for copying files from Pensando firmware container:
Earlier from Pensando firmware container, files were copied to first home dir /home/admin and then from there to shared directory /usr/share/sonic/device/.
Dissolved config_db.json into minigraph.xml, platform.json and init_cfg.json
Earlier config_db.json was copied from /usr/share/sonic/device/arm64-elba-asic-flash128-r0/config_db.json to /etc/sonic/config_db.json on first boot up post installation. Now on first boot, minigraph.xml and init_cfg.json gets copied to /etc/sonic and along with sonic default init_cfg.json, config_db.json is getting generated using sonic-cfggen command. This way, config_db.json will have flexibility for schema upgrades.
Bug fix:

Addressed slot id UNDEFINED issue for dpu_provisioning.sh and for DPU_STATE table entries
---------

Signed-off-by: Sahil Chaudhari <[email protected]>
Signed-off-by: Shantanu Shrivastava <[email protected]>

Author: SahilChaudhari

device/pensando/arm64-elba-asic-flash128-r0/Pensando-elba/port_config.ini

@@ -1,2 +1,2 @@
 #name          lanes alias speed autoneg fec
-Ethernet0 0,1,2,3 etp1 200000 on rs
+Ethernet0 0,1,2,3 etp1 200000 off rs

device/pensando/arm64-elba-asic-flash128-r0/default_sku

@@ -1 +1 @@
-Pensando-elba t1
+Pensando-elba t1-smartswitch

device/pensando/arm64-elba-asic-flash128-r0/platform.json

@@ -1,4 +1,12 @@
 {
-	"interfaces": {},
-	"DPU":{}
+    "interfaces": {
+        "Ethernet0": {
+            "index": "1",
+            "lanes": "0,1,2,3",
+            "breakout_modes": {
+                "1x200G": ["etp1"]
+            }
+        }
+    },
+    "DPU":{}
 }

device/pensando/arm64-elba-asic-flash128-r0/pre_reboot_hook

@@ -0,0 +1,26 @@
+#!/bin/bash
+
+LOG_MSG() {
+    echo "$1" | tee /dev/kmsg /dev/stderr /dev/console
+}
+
+LOG_MSG "Flushing file system buffers..."
+sync
+
+LOG_MSG "Flushing eMMC partitions..."
+for dev in $(lsblk -nd -o NAME | grep -E 'mmcblk'); do
+    LOG_MSG "Flushing /dev/$dev"
+    blockdev --flushbufs "/dev/$dev"
+done
+
+LOG_MSG "Ensuring all IOs are completed..."
+sync
+sleep 2  # Give time for completion
+
+LOG_MSG "Dropping caches..."
+LOG_MSG 3 > /proc/sys/vm/drop_caches
+LOG_MSG "Cache cleared."
+
+LOG_MSG "Pre-reboot flush and cache clear complete."
+LOG_MSG "Platform specific pre-shutdown is successful"
+

device/pensando/arm64-elba-asic-flash128-r0/system_health_monitoring_config.json

@@ -1,5 +1,5 @@
 {
-    "services_to_ignore": [],
+    "services_to_ignore": ["vlanmgrd", "vxlanmgrd"],
     "devices_to_ignore": ["psu","fan"],
     "user_defined_checkers": ["fetch_dpu_status"],
     "polling_interval": 60,
@@ -8,4 +8,4 @@
         "normal": "green",
         "booting": "amber_blink"
     }
-}
+}

files/dsc/dpu.init

@@ -62,6 +62,21 @@ function start_polaris()
     docker ps -a --format "{{.ID}}\t{{.Image}}" | grep "$IMAGE_NAME:$TAG" | awk '{print $1}' | xargs -I {} docker rm {}
 
     docker run -v $HOST_DIR_POLARIS/update:/update -v $HOST_DIR_POLARIS/sysconfig/config0:/sysconfig/config0 -v $HOST_DIR_POLARIS/sysconfig/config1:/sysconfig/config1 -v $HOST_DIR_POLARIS/obfl/a:/obfl -v $HOST_DIR_POLARIS/obfl:/var/log/obfl -v $HOST_DIR_POLARIS/data:/data -v $HOST_DIR_POLARIS/share:/share -v $HOST_DIR_POLARIS/share:/external -v $HOST_DIR_POLARIS/mnt/a:/ro -v /dev:/dev -v /sys:/sys --net=host --name=$CONTAINER_NAME_POLARIS --privileged $IMAGE_NAME:$TAG &
+
+    if [ -f /boot/first_boot ]; then
+        device="/usr/share/sonic/device"
+        platform=$(grep 'onie_platform=' /host/machine.conf | cut -d '=' -f 2)
+        echo "python3 -m pip install $device/$platform/sonic_platform-1.0-py3-none-any.whl"
+        python3 -m pip install $device/$platform/sonic_platform-1.0-py3-none-any.whl
+        rm /boot/first_boot
+    fi
+
+    sleep 5
+    INTERFACE="eth0-midplane"
+    if ip link show "$INTERFACE" &> /dev/null; then
+        echo "dhclient $INTERFACE"
+        /usr/sbin/dhclient $INTERFACE
+    fi
 }
 
 function start_dpu()

platform/pensando/sonic-platform-modules-dpu/dpu/service/dpu_provisioning.service

@@ -35,21 +35,81 @@ def run_cmd(cmd):
     return output
 
 def fetch_dpu_files():
+    api_helper_platform = apiHelper.get_platform()
+    platform = api_helper_platform if api_helper_platform != None else "arm64-elba-asic-r0"
     docker_id = apiHelper.get_dpu_docker_imageID()
-    cmd = "sudo docker cp {}:/tmp/fru.json /home/admin".format(docker_id)
-    run_cmd(cmd)
-    cmd = "sudo docker cp {}:/nic/bin/cpldapp /home/admin".format(docker_id)
-    run_cmd(cmd)
-    cmd = "sudo docker cp {}:/nic/lib/libpal.so /home/admin".format(docker_id)
-    run_cmd(cmd)
-    cmd = "sudo docker cp {}:/nic/lib/liblogger.so /home/admin".format(docker_id)
+    cmd = "sudo docker cp {}:/tmp/fru.json /usr/share/sonic/device/{}/fru.json".format(docker_id, platform)
     run_cmd(cmd)
-    cmd = "sudo docker cp {}:/nic/etc/VERSION.json /home/admin".format(docker_id)
+    cmd = "sudo docker cp {}:/nic/etc/VERSION.json /usr/share/sonic/device/{}/VERSION.json".format(docker_id, platform)
     run_cmd(cmd)
     cmd = "sudo docker cp {}:/usr/bin/mmc /usr/local/bin".format(docker_id)
     run_cmd(cmd)
+    try:
+        slot_id = apiHelper.run_docker_cmd("cpldapp -r 0xA").strip()
+        file = "/usr/share/sonic/device/{}/dpu_slot_id".format(apiHelper.get_platform())
+        with open(file, "w") as f:
+            f.write(slot_id)
+    except Exception as e:
+        log_err("failed to setup slot_id at platform dir due to {}".format(e))
+
+    try:
+        board_id = apiHelper.run_docker_cmd("cpldapp -r 0x80").strip()
+        file = "/usr/share/sonic/device/{}/dpu_board_id".format(apiHelper.get_platform())
+        with open(file, "w") as f:
+            f.write(board_id)
+    except Exception as e:
+        log_err("failed to setup board_id at platform dir due to {}".format(e))
+
+
+def setup_platform_components_json(slot_id):
+    try:
+        api_helper_platform = apiHelper.get_platform()
+        platform = api_helper_platform if api_helper_platform != None else "arm64-elba-asic-r0"
+        filename = "/usr/share/sonic/device/{}/platform_components.json".format(platform)
+        with open(filename, "r") as f:
+            data = json.load(f)
+
+        def replace_keys(obj):
+            if isinstance(obj, dict):
+                return {key.replace("-0", f"-{slot_id}"): replace_keys(value) for key, value in obj.items()}
+            elif isinstance(obj, list):
+                return [replace_keys(item) for item in obj]
+            return obj
+
+        updated_data = replace_keys(data)
+
+        with open(filename, "w") as f:
+            json.dump(updated_data, f, indent=4)
+        log_info("successfully updated platform_components.json")
+    except Exception as e:
+        log_err("failed to setup platform_components.json due to {}".format(e))
+
+def config_setup():
+    try:
+        from sonic_platform.chassis import Chassis
+        slot_id = Chassis().get_my_slot()
+    except Exception as e:
+        log_err("failed to get slot id due to {}".format(e))
+
+    try:
+        cmd = f'sonic-cfggen -a "{{\\"INTERFACE\\": {{\\"Ethernet0\\": {{}},\\"Ethernet0|18.{slot_id}.202.1/31\\": {{}}}}}}" --write-to-db'
+        run_cmd(cmd)
+    except Exception as e:
+        log_err("failed to set Ethernet0 ip due to {}".format(e))
+
+    setup_platform_components_json(slot_id)
+
+    try:
+        run_cmd("mkdir -p /host/images")
+        run_cmd("mkdir -p /data")
+        run_cmd("chmod +x /boot/install_file")
+    except Exception as e:
+        log_err("failed to setup fwutil due to {}".format(e))
+
 
 def set_onie_version():
+    api_helper_platform = apiHelper.get_platform()
+    platform = api_helper_platform if api_helper_platform != None else "arm64-elba-asic-r0"
     version = ''
     try:
         cmd = 'cat /host/machine.conf | grep -i onie_version'
@@ -66,36 +126,14 @@ def set_onie_version():
             version = "Not Available"
             pass
     try:
-        fru_file = "/home/admin/fru.json"
+        fru_file = "/usr/share/sonic/device/{}/fru.json".format(platform)
         data = json.load(open(fru_file))
         data["onie_version"] = version
         with open(fru_file, "w") as json_file:
             json.dump(data, json_file, indent=4)
     except:
         pass
 
-def cp_to_shared_mem():
-    api_helper_platform = apiHelper.get_platform()
-    platform = api_helper_platform if api_helper_platform != None else "arm64-elba-asic-r0"
-    cmd = "sudo cp /home/admin/fru.json /usr/share/sonic/device/{}/fru.json".format(platform)
-    run_cmd(cmd)
-    cmd = "sudo cp /home/admin/cpldapp /usr/share/sonic/device/{}/cpldapp".format(platform)
-    run_cmd(cmd)
-    cmd = "sudo cp /home/admin/libpal.so /usr/share/sonic/device/{}/libpal.so".format(platform)
-    run_cmd(cmd)
-    cmd = "sudo cp /home/admin/liblogger.so /usr/share/sonic/device/{}/liblogger.so".format(platform)
-    run_cmd(cmd)
-    cmd = "sudo cp /home/admin/VERSION.json /usr/share/sonic/device/{}/VERSION.json".format(platform)
-    run_cmd(cmd)
-
-def set_cpldapp():
-    cmd = "sudo cp /home/admin/cpldapp /usr/local/bin"
-    run_cmd(cmd)
-    cmd = "sudo cp /home/admin/libpal.so /lib/libpal.so"
-    run_cmd(cmd)
-    cmd = "sudo cp /home/admin/liblogger.so /lib/liblogger.so"
-    run_cmd(cmd)
-
 def set_ubootenv_config():
     cmd = "cat /proc/mtd | grep -e 'ubootenv' | awk '{print $1}' | tr -dc '0-9'"
     mtd_ubootenv = run_cmd(cmd)
@@ -111,7 +149,7 @@ def configure_iptable_rules():
         pass
 
 def pcie_tx_setup():
-    dpu_slot_id = int(run_cmd("cpldapp -r 0xA").strip(), 16)
+    dpu_slot_id = int(apiHelper.run_docker_cmd("cpldapp -r 0xA").strip(), 16)
     if dpu_slot_id == 6 or dpu_slot_id == 7:
         run_cmd("docker exec polaris /nic/tools/run-aacs-server.sh -p 9001 export SERDES_DUT_IP=localhost:9001")
         run_cmd("docker exec -e SERDES_ADDR=1:1-1:3f -e SERDES_DUT_IP=localhost:9001 -e SERDES_SBUS_RINGS=4 polaris aapl serdes -addr 1:39 -pre 0 -post 0 -atten 10")
@@ -131,13 +169,12 @@ def main():
         set_ubootenv_config()
     except:
         pass
-    time.sleep(10)
+    time.sleep(5)
     configure_iptable_rules()
     fetch_dpu_files()
+    config_setup()
     time.sleep(5)
     set_onie_version()
-    cp_to_shared_mem()
-    set_cpldapp()
     pcie_tx_setup()
 
 if __name__ == "__main__":

platform/pensando/sonic-platform-modules-dpu/dpu/utils/dpu_pensando_util.py

@@ -291,13 +291,11 @@ class DPUHealthUpdater():
             ]
             fvs_data.extend(fvs_control_plane_data)
 
-            self.delete_dpu_health_table_entries()
             for name, value in fvs_data:
                 self.db.hset(self.table, name, value)
 
         except Exception as e:
             log_err(f'Failed to refresh platform health status due to {e}, deleting db table')
-            self.delete_dpu_health_table_entries()
 
     def run(self):
         self.platform_health_stats = self._fetch_monitor_list()

platform/pensando/sonic-platform-modules-dpu/dpu/utils/dpu_provisioning.sh

@@ -30,6 +30,7 @@
 REBOOT_CAUSE_SOFTWARE = "Software causes"
 REBOOT_CAUSE_EXTERNAL = "External causes"
 RESET_CAUSE_PATH = "/sys/firmware/pensando/rstcause/this_cause"
+DOCKER_HWSKU_PATH = '/usr/share/sonic/platform'
 
 #cpld masks for system led
 SYSTEM_LED_GREEN   = 0x7
@@ -72,9 +73,6 @@ def __init__(self):
         self._api_helper = APIHelper()
         ChassisBase.__init__(self)
 
-        if not self._api_helper.is_host():
-            self._api_helper.setup_cpldapp()
-
         log_info("System chassis is ready")
 
     ##############################################
@@ -88,7 +86,7 @@ def initizalize_system_led(self):
     def get_status_led(self):
         cmd = "cpldapp -r {}".format(str(SYSTEM_LED_REG))
         try:
-            output = self._api_helper.runCMD(cmd)
+            output = self._api_helper.run_docker_cmd(cmd)
             reg_val = int(output, 16)
             if (reg_val & SYSTEM_LED_GREEN):
                 return self.STATUS_LED_COLOR_GREEN
@@ -361,8 +359,16 @@ def get_system_eeprom_info(self):
     def get_my_slot(self):
         cmd = "cpldapp -r 0xA"
         try:
-            slot_id = self._api_helper.runCMD(cmd)
-            return int(slot_id,16)
+            if self._api_helper.is_host():
+                slot_id = self._api_helper.run_docker_cmd(cmd)
+                return int(slot_id,16)
+            else:
+                slot_id_file = DOCKER_HWSKU_PATH + "/dpu_slot_id"
+                slot_id_hex = open(slot_id_file, "r").read()
+                if slot_id_hex:
+                    slot_id = int(slot_id_hex, 16)
+                    return slot_id
+            return -1
         except:
             return -1