Chris Webb [off-list ref] writes:

Return explicit error pointers out of bch2_fs_alloc() and handle them in
both callers, so the user instead sees

Very happy to re-spin this in a different shape if you prefer to do anything
differently. I haven't split any other failure cases away from -ENOMEM in
this patch: most of them seem well-characterised as 'out of memory' anyway,
at least at first glance.

I did wonder about factoring the superblock decryption out of bch2_fs_alloc
completely rather than changing the NULL return code, but I think that's not
possible as later parts bch2_fs_alloc already require it to be unlocked?

Signed-off-by: Chris Webb <chris@@arachsys.com>

*sigh* s/@@/@/ sorry!

Cheers,

Chris.