Re: [PATCH] io_uring: support multiple rings to share same poll thread by specifying same cpu

[Pavel Begunkov <asml.silence@xxxxxxxxx>]

On 31/03/2021 16:59, Xiaoguang Wang wrote:
> We have already supported multiple rings to share one same poll thread
> by passing IORING_SETUP_ATTACH_WQ, but it's not that convenient to use.
> IORING_SETUP_ATTACH_WQ needs users to ensure that a parent ring instance
> has beed created firstly, that means it will require app to regulate the
> creation oder between uring instances.
> 
> Currently we can make this a bit simpler, for those rings which will
> have SQPOLL enabled and are willing to be bound to one same cpu, add a
> capability that these rings can share one poll thread by specifying
> a new IORING_SETUP_SQPOLL_PERCPU flag, then we have 3 cases
>   1, IORING_SETUP_ATTACH_WQ: if user specifies this flag, we'll always
> try to attach this ring to an existing ring's corresponding poll thread,
> no matter whether IORING_SETUP_SQ_AFF or IORING_SETUP_SQPOLL_PERCPU is
> set.
>   2, IORING_SETUP_SQ_AFF and IORING_SETUP_SQPOLL_PERCPU are both enabled,
> for this case, we'll create a single poll thread to be shared by rings
> rings which have same sq_thread_cpu.
>   3, for any other cases, we'll just create one new poll thread for the
> corresponding ring.
> 
> And for case 2, don't need to regulate creation oder of multiple uring
> instances, we use a mutex to synchronize creation, for example, say five
> rings which all have IORING_SETUP_SQ_AFF & IORING_SETUP_SQPOLL_PERCPU
> enabled, and are willing to be bound same cpu, one ring that gets the
> mutex lock will create one poll thread, the other four rings will just
> attach themselves to the previous created poll thread once they get lock
> successfully.
> 
> To implement above function, define below data structs:
>   struct percpu_sqd_entry {
>         struct list_head        node;
>         struct io_sq_data       *sqd;
>         pid_t                   tgid;
>   };
> 
>   struct percpu_sqd_list {
>         struct list_head        head;
>         struct mutex            lock;
>   };
> 
>   static struct percpu_sqd_list __percpu *percpu_sqd_list;
> 
> sqthreads that have same sq_thread_cpu will be linked together in a percpu
> percpu_sqd_list's head. When IORING_SETUP_SQ_AFF and IORING_SETUP_SQPOLL_PERCPU
> are both enabled, we will use struct io_uring_params' sq_thread_cpu and
> current-tgid locate corresponding sqd.

I can't help myself but wonder why not something in the userspace like
a pseudo-coded snippet below?

BTW, don't think "pid_t tgid" will work with namespaces/cgroups.



static std::vector<std::set<struct io_uring *>> percpu_rings;
static std::mutex lock;

int io_uring_queue_init_params_percpu(unsigned entries,
				      struct io_uring *ring,
				      struct io_uring_params *p);
{
	unsigned int cpu = p->sq_thread_cpu;
	std::unique_lock guard(lock);

	if (!(p->flags & IORING_SETUP_SQPOLL))
		return -EINVAL;
	if (percpu_rings.size() <= cpu)
		percpu_rings.resize(cpu + 1);

	p->flags &= ~IORING_SETUP_ATTACH_WQ;
	if (!percpu_rings[cpu].empty()) {
		struct io_uring *shared_ring = *percpu_rings[cpu].begin();
		p->wq_fd = shared_ring->ring_fd;
		p->flags |= IORING_SETUP_ATTACH_WQ;
	}

	int ret = io_uring_queue_init_params(entries, ring, p);
	if (!ret)
		percpu_rings[cpu].insert(ring);
	return ret;
}

void io_uring_queue_exit_percpu(struct io_uring *ring)
{
	std::unique_lock guard(lock);

	for (auto& cpu_set: percpu_rings)
		if (cpu_set.erase(ring))
			break;
	guard.release();
	io_uring_queue_exit(ring);
}

-- 
Pavel Begunkov