Re: [Qemu-devel] aio: reg. smp_read_barrier_depends() in aio_bh_poll()

[Paolo Bonzini]

On 02/09/2016 20:23, Pranith Kumar wrote:
> If I understand you correctly, this is what might happen without the
> barrier():
> 
>             P0                        P1
>       ----------------------------------------
>       // bh = ctx->first_bh;  optimized       
>       if (ctx->first_bh) {
>       // next = ctx->first_bh->next;
>                                    lock(bh_lock);
>                                    new_bh->next = ctx->first_bh;
>                                    smp_wmb();  // this alone is not sufficient
>                                                // for Alpha
>                                    ctx->first_bh = new_bh;
>                                    unlock(bh_lock);
>       // bh = next;
>       bh = ctx->first_bh->next;
> 
>       if (bh) {do something}
>       }
> 
> Is this what might happen? If so, inserting a barrier() after the first load
> into bh will prevent the compiler from optimizing the load into bh since the
> compiler cannot optimize away loads and stores past the barrier().

Yes, this is what you can expect from a compiler.

> And on Alpha processors barrier() should really be smp_read_barrier_depends()
> to prevent this from happening because of it's memory model(issue a barrier
> after loading a pointer to shared memory and before dereferencing it).

Yes; the actual effect you could see on the Alpha it's even crazier.
Without the barrier, the "bh = ctx->first_bh" and "next = bh->next" can
be reordered.  If you take both threads into account, "next = bh->next"
can use a value from before P1's "ctx->first = new_bh".  Indeed, it can
use a value from before P1's smp_wmb() or before new_bh->next =
ctx->first_bh.  For example, "next" could load a NULL value.

FWIW, RISCv currently has the same memory model as the Alpha, but they
plan to fix it before they finalize the relevant specs!

Paolo

>> >
>> > So instead of smp_read_barrier_depends() you could load ctx->first_bh
>> > and bh->next with the consume memory order, but you do need _something_.
> OK, if the above situation is possible, then I think I understand the need for
> this barrier.