Re: [PATCH v4] lib/dlock-list: Scale dlock_lists_empty()

[Andreas Dilger]

On Nov 7, 2017, at 4:59 AM, Jan Kara <jack@xxxxxxx> wrote:
> On Mon 06-11-17 10:47:08, Davidlohr Bueso wrote:
>> +	/*
>> +	 * Serialize dlist->used_lists such that a 0->1 transition is not
>> +	 * missed by another thread checking if any of the dlock lists are
>> +	 * used.
>> +	 *
>> +	 * CPU0				    CPU1
>> +	 * dlock_list_add()                 dlock_lists_empty()
>> +	 *   [S] atomic_inc(used_lists);
>> +	 *       smp_mb__after_atomic();
>> +	 *					  smp_mb__before_atomic();
>> +	 *				      [L] atomic_read(used_lists)
>> +	 *       list_add()
>> +	 */
>> +	smp_mb__before_atomic();
>> +	return !atomic_read(&dlist->used_lists);

Just a general kernel programming question here - I thought the whole point
of atomics is that they are, well, atomic across all CPUs so there is no
need for a memory barrier?  If there is a need for a memory barrier for
each atomic access (assuming it isn't accessed under another lock, which would
make the use of atomic types pointless, IMHO) then I'd think there is a lot
of code in the kernel that isn't doing this properly.

What am I missing here?

I don't see how this helps if the operations are executed like:

	 * CPU0				    CPU1
	 * dlock_list_add()                 dlock_lists_empty()
	 *   [S] atomic_inc(used_lists);
	 *					  smp_mb__before_atomic();
	 *       smp_mb__after_atomic();
	 *				      [L] atomic_read(used_lists)

or alternately like:

	 * CPU0				    CPU1
	 * dlock_list_add()                 dlock_lists_empty()
	 *					  smp_mb__before_atomic();
	 *   [S] atomic_inc(used_lists);
	 *       smp_mb__after_atomic();
	 *				      [L] atomic_read(used_lists)

then the same problem would exist, unless those functions/macros are somehow
bound to the atomic operations themselves?  In that case, what makes the use
of atomic_{inc,dec,read}() in other parts of the code safe without them?

Cheers, Andreas

Attachment: signature.asc
Description: Message signed with OpenPGP