Improve UDF sector count

[Deterous]

Here's an example that redumper currently gets UDF sector count incorrect: PS3 IDU - 2014 Holiday Refresh (USA) (UDF-only BD-R)

sectors count (READ_CAPACITY): 4039168
sectors count (UDF): 4038915

The sector at 4038915 contains another Logical Volume Descriptor, followed by a Unallocated Space Descriptor, followed by a Terminating Descriptor, then a bunch of blank sectors, and ending in an AVDP.

[superg]

That is expected on BD-R as it has a trailing VDS.

I think we look at this from different angles.
The diagnostics output that I do is to hint user of different possibilities of final sectors count.
Unless you use --filesystem-trim, that will not be used and for the example you provided READ_CAPACITY is already good, isn't it?
The use case I am covering is that user dumps something, analyzes possible sectors count and then supplies --lba-end with what they think is good.

On the other hand, detection of trailing UDF structures is quite complicated, there can be AVDP, there can be VDS, there can be unallocated space etc.

[Deterous]

Currently filesystem trim is happening automatically if redumper detects Blu-ray profile mismatch?

[superg]

Yes, that is correct.

[Deterous]

Then it should be important to improve UDF detection for this case. It's certainly much more complicated than ISO9660

[superg]

The example you provided is not a profile mismatch case. Profile mismatch that I try to handle in redumper was meant to fix overdumping of all layers because capacity was reporting an incorrect number of sectors.

I definitely would want to have a better detection of UDF structures but that needs much more time.

[coderabbitai]

Related concern from PR #363 (dvdraw_edc branch): filesystem_search_size reads UDF VDS entries directly from fs_image via read_entry using a fixed FORM1_DATA_SIZE stride. This is correct for logical .iso images, but in raw dump mode (.sdram / .sbram), cfg.sector_size differs from FORM1_DATA_SIZE, so the read offsets are wrong and the function ends up parsing raw recording frames instead of extracted user data. This can cause UDF filesystem size detection to return an incorrect sector count and trim raw dumps incorrectly. When this is addressed in scope of this issue, the VDS read path inside filesystem_search_size should either be gated to non-raw images or go through the same logical-sector extraction routine used in the main dump loop.