Initial remote commit

misc_tools/offset_check.py

@@ -0,0 +1,91 @@
+CHUNK    = 128*512  # 64KB
+LV_START = 5120000*512
+BSIZE    = 4096
+
+# Block 5251104, virtual byte = 5251104 * 4096 = 21508521984
+# group=65638, chunk_idx=4, intra=0
+
+# Physical layout: each group of 5 virtual chunks occupies
+# 5*CHUNK bytes on the physical disk (md0)
+# group 65638 starts at:
+group = 65638
+group_phys_start = LV_START + group * 5 * CHUNK
+print(f'Group {group} physical start: {group_phys_start}')
+print()
+
+import struct
+
+with open('/dev/md0','rb') as f:
+    # Read all 5 chunks of this group
+    f.seek(group_phys_start)
+    all_chunks = f.read(5*CHUNK)
+
+for ci in range(5):
+    chunk = all_chunks[ci*CHUNK:(ci+1)*CHUNK]
+    nonzero = sum(1 for b in chunk if b != 0)
+    # Try first 32 bytes as directory
+    ino = struct.unpack_from('<I', chunk, 0)[0]
+    rec = struct.unpack_from('<H', chunk, 4)[0]
+    nl  = chunk[6]
+    ft  = chunk[7]
+    print(f'Chunk {ci}: nonzero={nonzero}/{CHUNK}', end='')
+    if 10 < ino < 500_000_000 and 8 <= rec <= 256 and 0 < nl < 32 and ft in (1,2,7):
+        name = chunk[8:8+nl].decode('utf-8',errors='replace')
+        print(f' POSSIBLE DIR: inode={ino} name={name!r}', end='')
+    print()
+
+# The data chunks (0-3) map to virtual chunks as follows:
+# Our translation: virtual chunk N -> physical chunk N (for N<4)
+# So physical chunks 0,1,2,3 = virtual chunks 0,1,2,3
+# And physical chunk 4 = the metadata chunk we skip
+
+# But what if our anchor points were right (FAT32 and LV)
+# yet the internal mapping within the LV is different?
+# What if the PERC uses a different chunk as metadata inside the LV?
+
+# Our two anchors:
+# FAT32 VBR at PERC virtual 2048 -> md0 sector 1664
+# LV start at PERC virtual 6400000 -> md0 sector 5120000
+
+# These are BEFORE the LV. Inside the LV, could the chunk order differ?
+# What if inside the LV the metadata chunk is chunk 0, not chunk 4?
+
+# Test: read chunk 0 of group 65638 as directory
+chunk0 = all_chunks[0:CHUNK]
+print()
+print('Chunk 0 as directory:')
+off = 0
+while off < BSIZE-8:
+    ino     = struct.unpack_from('<I', chunk0, off)[0]
+    rec_len = struct.unpack_from('<H', chunk0, off+4)[0]
+    name_len= chunk0[off+6]
+    ftype   = chunk0[off+7]
+    if rec_len < 8: break
+    if ino > 0 and name_len > 0:
+        name = chunk0[off+8:off+8+name_len].decode('utf-8',errors='replace')
+        tname = {1:'file',2:'dir',7:'link'}.get(ftype,'?')
+        print(f'  {tname} inode={ino} {name!r}')
+    off += rec_len
+
+# Test each chunk as potential /var directory
+print()
+for ci in range(5):
+    chunk = all_chunks[ci*CHUNK:(ci+1)*CHUNK]
+    off = 0
+    entries = []
+    while off < BSIZE-8:
+        ino     = struct.unpack_from('<I', chunk, off)[0]
+        rec_len = struct.unpack_from('<H', chunk, off+4)[0]
+        name_len= chunk[off+6]
+        ftype   = chunk[off+7]
+        if rec_len < 8: break
+        if 10 < ino < 500_000_000 and name_len > 0 and ftype in (1,2,7):
+            name = chunk[off+8:off+8+name_len].decode('utf-8',errors='replace')
+            if name.isprintable():
+                entries.append((ino, name, ftype))
+        off += rec_len
+    if entries:
+        print(f'Chunk {ci} has {len(entries)} directory entries:')
+        for ino, name, ftype in entries[:5]:
+            tname = {1:'file',2:'dir',7:'link'}.get(ftype,'?')
+            print(f'  {tname} inode={ino} {name!r}')