From ea398e9d1c9c5a2a1db851fb43b4941efdccd0b7 Mon Sep 17 00:00:00 2001
From: jbtrystram <jbtrystram@redhat.com>
Date: Fri, 18 Jul 2025 15:18:40 +0200
Subject: [PATCH] cosa diff: add support for diffing metal images

This add two flags: `--metal` and `metal-part-table`.

The former will mount the disk partitions using guestfs tools to the
tmp-diff directory, then call git diff over the two mounted filesystems.
It requires multithreading because the FUSE mount call is blocking so
we mount the disks in two libreguest VMs that run in separate threads.

A simpler approach would have been to copy all the content to the
tmp-diff folder but that would copy a lot of data just for diffing. It
may be faster though.

The second flag `--metal-part-table` will simply show the partition
table for the two images.
---
 src/cmd-diff | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++
 src/deps.txt |   3 ++
 2 files changed, 104 insertions(+)

diff --git a/src/cmd-diff b/src/cmd-diff
index 9e655af6da..fc5799dc83 100755
--- a/src/cmd-diff
+++ b/src/cmd-diff
@@ -6,9 +6,12 @@ import shutil
 import subprocess
 import sys
 import tempfile
+import time
+from multiprocessing import Process
 
 from dataclasses import dataclass
 from enum import IntEnum
+import guestfs
 from typing import Callable
 
 sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
@@ -304,6 +307,100 @@ def diff_live_sysroot(diff_from, diff_to):
     git_diff(dir_from, dir_to)
 
 
+def get_metal_path(build_target):
+    metal_file = build_target.meta.get('images', {}).get('metal')
+
+    if not metal_file:
+        raise Exception(f"Could not find metal image for build {build_target.id}")
+    return os.path.join(build_target.dir, metal_file['path'])
+
+
+def diff_metal_partitions(diff_from, diff_to):
+    metal_from = get_metal_path(diff_from)
+    metal_to = get_metal_path(diff_to)
+    diff_cmd_outputs(['sgdisk', '-p'], metal_from, metal_to)
+
+
+def run_guestfs_mount(image_path, mount_target):
+    """This function runs in a background thread."""
+    g = None
+    try:
+        g = guestfs.GuestFS(python_return_dict=True)
+        g.set_backend("direct")
+        g.add_drive_opts(image_path, readonly=1)
+        g.launch()
+
+        # Mount the disks in the guestfs VM
+        root = g.findfs_label("root")
+        g.mount_ro(root, "/")
+        boot = g.findfs_label("boot")
+        g.mount_ro(boot, "/boot")
+        efi = g.findfs_label("EFI-SYSTEM")
+        g.mount_ro(efi, "/boot/efi")
+
+        # This is a blocking call that runs the FUSE server
+        g.mount_local(mount_target)
+        g.mount_local_run()
+
+    except Exception as e:
+        print(f"Error in guestfs process for {image_path}: {e}", file=sys.stderr)
+    finally:
+        if g:
+            g.close()
+
+
+def diff_metal(diff_from, diff_to):
+    metal_from = get_metal_path(diff_from)
+    metal_to = get_metal_path(diff_to)
+
+    mount_dir_from = os.path.join(cache_dir("metal"), diff_from.id)
+    mount_dir_to = os.path.join(cache_dir("metal"), diff_to.id)
+
+    for d in [mount_dir_from, mount_dir_to]:
+        if os.path.exists(d):
+            shutil.rmtree(d)
+        os.makedirs(d)
+
+    # As the libreguest mount call is blocking until unmounted, let's
+    # do that in a separate thread
+    p_from = Process(target=run_guestfs_mount, args=(metal_from, mount_dir_from))
+    p_to = Process(target=run_guestfs_mount, args=(metal_to, mount_dir_to))
+
+    try:
+        p_from.start()
+        p_to.start()
+        # Wait for the FUSE mounts to be ready. We'll check for a known file.
+        for i, d in enumerate([mount_dir_from, mount_dir_to]):
+            p = p_from if i == 0 else p_to
+            timeout = 60  # seconds
+            start_time = time.time()
+            check_file = os.path.join(d, 'ostree')
+            while not os.path.exists(check_file):
+                time.sleep(1)
+                if time.time() - start_time > timeout:
+                    raise Exception(f"Timeout waiting for mount in {d}")
+                if not p.is_alive():
+                    raise Exception(f"A guestfs process for {os.path.basename(d)} died unexpectedly.")
+
+        # Now that the mounts are live, we can diff them
+        git_diff(mount_dir_from, mount_dir_to)
+
+    finally:
+        # Unmount the FUSE binds, this will make the guestfs mount calls return
+        runcmd(['fusermount', '-u', mount_dir_from], check=False)
+        runcmd(['fusermount', '-u', mount_dir_to], check=False)
+
+        # Ensure the background processes are terminated
+        def shutdown_process(process):
+            process.join(timeout=5)
+            if process.is_alive():
+                process.terminate()
+                process.join()
+
+        shutdown_process(p_from)
+        shutdown_process(p_to)
+
+
 def diff_cmd_outputs(cmd, file_from, file_to):
     with tempfile.NamedTemporaryFile(prefix=cmd[0] + '-') as f_from, \
          tempfile.NamedTemporaryFile(prefix=cmd[0] + '-') as f_to:
@@ -356,6 +453,10 @@ DIFFERS = [
            needs_ostree=OSTreeImport.NO, function=diff_live_sysroot_tree),
     Differ("live-sysroot", "Diff live '/root.[ero|squash]fs' (embed into live-rootfs) content",
            needs_ostree=OSTreeImport.NO, function=diff_live_sysroot),
+    Differ("metal-part-table", "Diff metal disk image partition tables",
+           needs_ostree=OSTreeImport.NO, function=diff_metal_partitions),
+    Differ("metal", "Diff metal disk image content",
+           needs_ostree=OSTreeImport.NO, function=diff_metal),
 ]
 
 if __name__ == '__main__':
diff --git a/src/deps.txt b/src/deps.txt
index 8eaa711237..3b82b4a8b5 100644
--- a/src/deps.txt
+++ b/src/deps.txt
@@ -104,3 +104,6 @@ erofs-utils
 
 # Support for copr build in coreos-ci
 copr-cli
+
+# To mount metal disk images in cmd-diff
+python3-libguestfs