OS Updater improvements (#19914)

* Clear hashes before swapping

* add retry logic

* better retry logic

* actually fail on retries exceeded

launch_chffrplus.sh

@@ -146,6 +146,12 @@ function tici_init {
 
     if [[ "$SYSTEM_HASH" == "$SYSTEM_HASH_EXPECTED" && "$BOOT_HASH" == "$BOOT_HASH_EXPECTED" ]]; then
       echo "Swapping active slot to $OTHER_SLOT_NUMBER"
+
+      # Clean hashes before swapping to prevent looping
+      dd if=/dev/zero of="/dev/disk/by-partlabel/system$OTHER_SLOT" bs=1 seek="$SYSTEM_SIZE" count=64
+      dd if=/dev/zero of="/dev/disk/by-partlabel/boot$OTHER_SLOT" bs=1 seek="$BOOT_SIZE" count=64
+      sync
+
       abctl --set_active "$OTHER_SLOT_NUMBER"
 
       sleep 1
@@ -154,6 +160,12 @@ function tici_init {
       echo "Hash mismatch, downloading agnos"
       if $DIR/selfdrive/hardware/tici/agnos.py $MANIFEST; then
         echo "Download done, swapping active slot to $OTHER_SLOT_NUMBER"
+
+        # Clean hashes before swapping to prevent looping
+        dd if=/dev/zero of="/dev/disk/by-partlabel/system$OTHER_SLOT" bs=1 seek="$SYSTEM_SIZE" count=64
+        dd if=/dev/zero of="/dev/disk/by-partlabel/boot$OTHER_SLOT" bs=1 seek="$BOOT_SIZE" count=64
+        sync
+
         abctl --set_active "$OTHER_SLOT_NUMBER"
       fi
 

selfdrive/hardware/tici/agnos.py

@@ -76,19 +76,7 @@ def unsparsify(f):
       raise Exception("Unhandled sparse chunk type")
 
 
-def flash_agnos_update(manifest_path, cloudlog, spinner=None):
+def flash_partition(cloudlog, spinner, target_slot, partition):
-  update = json.load(open(manifest_path))
-
-  current_slot = subprocess.check_output(["abctl", "--boot_slot"], encoding='utf-8').strip()
-  target_slot = "_b" if current_slot == "_a" else "_a"
-  target_slot_number = "0" if target_slot == "_a" else "1"
-
-  cloudlog.info(f"Current slot {current_slot}, target slot {target_slot}")
-
-  # set target slot as unbootable
-  os.system(f"abctl --set_unbootable {target_slot_number}")
-
-  for partition in update:
   cloudlog.info(f"Downloading and writing {partition['name']}")
 
   downloader = StreamingDecompressor(partition['url'])
@@ -129,6 +117,37 @@ def flash_agnos_update(manifest_path, cloudlog, spinner=None):
     os.sync()
     out.write(partition['hash_raw'].lower().encode())
 
+
+def flash_agnos_update(manifest_path, cloudlog, spinner=None):
+  update = json.load(open(manifest_path))
+
+  current_slot = subprocess.check_output(["abctl", "--boot_slot"], encoding='utf-8').strip()
+  target_slot = "_b" if current_slot == "_a" else "_a"
+  target_slot_number = "0" if target_slot == "_a" else "1"
+
+  cloudlog.info(f"Current slot {current_slot}, target slot {target_slot}")
+
+  # set target slot as unbootable
+  os.system(f"abctl --set_unbootable {target_slot_number}")
+
+  for partition in update:
+    success = False
+
+    for retries in range(10):
+      try:
+        flash_partition(cloudlog, spinner, target_slot, partition)
+        success = True
+        break
+
+      except requests.exceptions.RequestException:
+        spinner.update("Waiting for internet...")
+        cloudlog.info(f"Failed to download {partition['name']}, retrying ({retries})")
+        time.sleep(10)
+
+    if not success:
+      cloudlog.info(f"Failed to flash {partition['name']}, aborting")
+      raise Exception("Maximum retries exceeded")
+
   cloudlog.info(f"AGNOS ready on slot {target_slot}")