Hi!
> static inline void dec_pending(struct dm_io *io, int error)
> {
> static spinlock_t _uptodate_lock = SPIN_LOCK_UNLOCKED;
> unsigned long flags;
>
> if (error) {
> spin_lock_irqsave(&_uptodate_lock, flags);
> io->error = error;
> spin_unlock_irqrestore(&_uptodate_lock, flags);
> }
I think this spinlock is unnecessary. Yes, there can be a race, but it
doesn't matter. If there are two threads racing for io->error, what
happens? io->error will be non-null.
There are other places in the kernel where racing is explicitly allowed
in order to save a spinlock. We don't even have to take additional care
in this case.
