Initial remote commit

misc_tools/interleave_check.py

@@ -0,0 +1,80 @@
+CHUNK    = 128*512
+LV_START = 5120000*512
+BSIZE    = 4096
+
+# The directory data block is at virtual block 5251104
+# virt_byte = 5251104 * 4096 = 21508521984
+virt_byte = 5251104 * BSIZE
+
+group     = virt_byte // (5*CHUNK)
+in_group  = virt_byte % (5*CHUNK)
+chunk_idx = in_group // CHUNK
+intra     = in_group % CHUNK
+
+print(f'Virtual byte:  {virt_byte}')
+print(f'Chunk group:   {group}')
+print(f'Chunk index:   {chunk_idx}')
+print(f'Intra:         {intra}')
+print()
+
+# This is chunk_idx=4 - a PERC metadata chunk
+# The PERC stored filesystem data here
+# Physical location:
+# In our translation, we skip chunk_idx=4
+# But if the PERC stored data there, it's at:
+# phys = LV_START + group*5*CHUNK + chunk_idx*CHUNK + intra
+# (5 chunks per group including metadata chunk)
+phys_with_meta = LV_START + group*5*CHUNK + chunk_idx*CHUNK + intra
+print(f'Physical byte if 5-chunk groups: {phys_with_meta}')
+print(f'Physical sector: {phys_with_meta//512}')
+
+# Also try: physical = LV_START + group*4*CHUNK + ... but for chunk 4
+# In our current scheme chunk 4 has no mapping
+# But the data must be somewhere on disk
+
+# The PERC presents 5 virtual chunks = 5*64KB = 320KB per group
+# Of these, 4 are data and 1 is PERC metadata
+# But what if it's the OPPOSITE?
+# What if 4 are PERC metadata and only 1 is data? No - we verified 4/5 ratio
+
+# What if the metadata chunk IS stored on disk but at a different offset?
+# The PERC stores its own metadata in those chunk positions
+# For reading filesystem data that falls there, it must use the metadata chunk
+# storage differently
+
+# Let's read what's physically at that location
+# The virtual block 5251104 is in chunk group 'group' at chunk position 4
+# In the physical layout (5 chunks per group):
+# group X physical layout: [data0][data1][data2][data3][meta4]
+# Our translation: physical chunk = group*4 + chunk_idx (skipping meta)
+# So physical chunk for data chunks 0-3 = group*4 + 0,1,2,3
+# And physical chunk for meta = stored AFTER the 4 data chunks = group*4+4? 
+# But that would mean physical layout is: data0,data1,data2,data3,meta4
+# which IS 5 consecutive chunks on disk
+
+# So the physical byte for chunk_idx=4 would be:
+phys_consecutive = LV_START + group*5*CHUNK + 4*CHUNK + intra
+print(f'Physical byte (consecutive 5-chunk): {phys_consecutive}')
+
+# Read it
+with open('/dev/md0','rb') as f:
+    f.seek(phys_consecutive)
+    data = f.read(512)
+nonzero = sum(1 for b in data if b != 0)
+print(f'Nonzero bytes: {nonzero}/512')
+print(f'First 32: {data[:32].hex()}')
+
+# Also check: maybe the metadata chunk is interleaved differently
+# What if physical layout per group is: [meta0][data1][data2][data3][data4]?
+# i.e. metadata chunk is FIRST, not last?
+for meta_pos in range(5):
+    # If metadata is at position meta_pos within each group of 5 physical chunks
+    # Then for virtual chunk_idx=4 (the one we skip):
+    # virtual chunks 0,1,2,3 map to physical chunks skipping meta_pos
+    # virtual chunk 4 IS the metadata chunk
+    phys_test = LV_START + group*5*CHUNK + meta_pos*CHUNK + intra
+    with open('/dev/md0','rb') as f:
+        f.seek(phys_test)
+        data = f.read(64)
+    nonzero = sum(1 for b in data if b != 0)
+    print(f'meta_pos={meta_pos}: phys={phys_test} nonzero={nonzero} first8={data[:8].hex()}')