[PATCH v2 2/2] UBIFS: fix fatal race condition

[hujianyang]

Hi Artem,

It's better than mine. I've tested some cases to reproduce this race
and found "> c->max_bud_bytes" is mostly hit in add_bud_to_log() but
'< c->min_log_bytes' is hardly hit. I think this race condition merely
occur and power cut frequently must be one of reasons.

Our fixes harm the performance a bit. Other process should wait for
write_mst_node() done before writing a new bud even if this commit is
not triggered by log lebs full. I don't know how to avoid this but
I think current version is enough. Flash is very fast.

I would like to test these patches now and report to you soon.

Thanks a lot!

Hu