Re: [PATCH v2 4/4] btrfs: dump all space infos if we abort transaction due to ENOSPC

[Sweet Tea Dorminy <sweettea-kernel@dorminy.me>]

>>> +}
>> This function looks similar to btrfs_dump_space_info(), and the name and
>> callsite doesn't help distinguish it very much to me. It seems
>> potentially useful to print all the space_infos when one space_info
>> encounters a problem,
> 
> This is fine for trans abort dump, but may not be a good idea for
> enospc_debug output.
> 
> enospc_debug can be triggered way more frequent than trans abort, and
> the extra info of unrelated space info may just be noise.
>> and it seems potentially useful to print the block
>> group infos when we're dumping all the space infos already, so maybe the
>> two functions could be combined.
> 
> You mean block group infos? That can be very large for large fses.
> Thus it's avoided for most call sites other than btrfs_reserve_extent().

I appreciate the explanation, I now agree it makes sense to keep the 
dump-all function separate.
Maybe you could move the new dump-all function up beside the dump-one 
function, and name it btrfs_dump_all_space_infos()? I still find the 
name btrfs_dump_fs_space_info() a little hard to distinguish from the 
others, and prefer 'all' being in the function name, but it's only a 
mild preference.