Find dnode's object number during disk recovery

[momosnn]

Hi. Does anybody know if there’s any way to retrieve a file’s object number from the on-disk dnode structure itself? As far as I can tell the only way to find the object number is to know the exact element offset within the containing dnode_phys_t array – is that correct, or am I missing something?

I’m trying to write an opensource recovery tool that works even with severely corrupt metadata.

During a disk scan, it seems easy to identify dnodes of types DMU_OT_PLAIN_FILE_CONTENTS and DMU_OT_DIRECTORY_CONTENTS – but without having the entire intact dnode_phys_t for the dataset I’m struggling to be able to reliably map the found file objects to their parent directories and also to be able to figure out the file and directory names (I have some approaches that work some of the time, but without knowing the object number it's never certain).

Is this metadata duplicated anywhere else?

Is there some other globally-unique identifier for a file/directory other than the object number?

Or is there any way determine if two found portions/blocks of a dnode_phys_t are adjacent to each other, or where they lie within the full array?

When given a file dnode, it seems the only things of value (for recovery) are:
• The parent object number (ZPL_PARENT)
• The file size (ZPL_SIZE)
• The generation number (ZPL_GEN)
A file doesn’t have a ZAP that I’m not aware of does it?

I’d very much appreciate it if somebody could point me in the right direction. Thanks for your time.