All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] io-wq: provide a way to limit max number of unbounded workers
@ 2021-08-27 17:38 Jens Axboe
  0 siblings, 0 replies; only message in thread
From: Jens Axboe @ 2021-08-27 17:38 UTC (permalink / raw)
  To: io-uring; +Cc: Johannes Lundberg

io-wq divides work into two categories:

1) Work that completes in a bounded time, like reading from a regular file
   or a block device. This type of work is limited based on the size of
   the SQ ring.

2) Work that may never complete, we call this unbounded work. The amount
   of workers here is just limited by RLIMIT_NPROC.

For various uses cases, it's handy to have the kernel limit the maximum
amount of pending unbounded workers. Provide a way to do with with a new
IORING_REGISTER_IOWQ_MAX_UNBOUND operation.

IORING_REGISTER_IOWQ_MAX_UNBOUND takes an integer and sets the max worker
count to what is being passed in, and returns the old value (or an error).
If 0 is being passed in, it simply returns the current value.

The value is capped at RLIMIT_NPROC. This actually isn't that important
as it's more of a hint, if we're exceeding the value then our attempt
to fork a new worker will fail. This happens naturally already if more
than one node is in the system, as these values are per-node internally
for io-wq.

Reported-by: Johannes Lundberg <johalun0@gmail.com>
Link: https://github.com/axboe/liburing/issues/420
Signed-off-by: Jens Axboe <axboe@kernel.dk>

---

diff --git a/fs/io-wq.c b/fs/io-wq.c
index 8da9bb103916..18698662d354 100644
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -1152,6 +1152,29 @@ int io_wq_cpu_affinity(struct io_wq *wq, cpumask_var_t mask)
 	return 0;
 }
 
+/*
+ * Set max number of unbounded workers, returns old value. If new_count is 0,
+ * then just return the old value.
+ */
+int io_wq_max_unbound(struct io_wq *wq, int new_count)
+{
+	int i, prev = 0;
+
+	if (new_count > task_rlimit(current, RLIMIT_NPROC))
+		new_count = task_rlimit(current, RLIMIT_NPROC);
+
+	rcu_read_lock();
+	for_each_node(i) {
+		struct io_wqe *wqe = wq->wqes[i];
+
+		prev = max_t(int, wqe->acct[IO_WQ_ACCT_UNBOUND].max_workers, prev);
+		if (new_count)
+			wqe->acct[IO_WQ_ACCT_UNBOUND].max_workers = new_count;
+	}
+	rcu_read_unlock();
+	return prev;
+}
+
 static __init int io_wq_init(void)
 {
 	int ret;
diff --git a/fs/io-wq.h b/fs/io-wq.h
index 308af3928424..d6631c5fca11 100644
--- a/fs/io-wq.h
+++ b/fs/io-wq.h
@@ -128,6 +128,7 @@ void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work);
 void io_wq_hash_work(struct io_wq_work *work, void *val);
 
 int io_wq_cpu_affinity(struct io_wq *wq, cpumask_var_t mask);
+int io_wq_max_unbound(struct io_wq *wq, int new_count);
 
 static inline bool io_wq_is_hashed(struct io_wq_work *work)
 {
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 53326449d685..0f827fbe8e6c 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -10233,6 +10233,22 @@ static int io_unregister_iowq_aff(struct io_ring_ctx *ctx)
 	return io_wq_cpu_affinity(tctx->io_wq, NULL);
 }
 
+static int io_register_iowq_max_unbound(struct io_ring_ctx *ctx,
+					void __user *arg)
+{
+	struct io_uring_task *tctx = current->io_uring;
+	__u32 new_count;
+
+	if (!tctx || !tctx->io_wq)
+		return -EINVAL;
+	if (copy_from_user(&new_count, arg, sizeof(new_count)))
+		return -EFAULT;
+	if (new_count > INT_MAX)
+		return -EINVAL;
+
+	return io_wq_max_unbound(tctx->io_wq, new_count);
+}
+
 static bool io_register_op_must_quiesce(int op)
 {
 	switch (op) {
@@ -10250,6 +10266,7 @@ static bool io_register_op_must_quiesce(int op)
 	case IORING_REGISTER_BUFFERS_UPDATE:
 	case IORING_REGISTER_IOWQ_AFF:
 	case IORING_UNREGISTER_IOWQ_AFF:
+	case IORING_REGISTER_IOWQ_MAX_UNBOUND:
 		return false;
 	default:
 		return true;
@@ -10406,6 +10423,12 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
 			break;
 		ret = io_unregister_iowq_aff(ctx);
 		break;
+	case IORING_REGISTER_IOWQ_MAX_UNBOUND:
+		ret = -EINVAL;
+		if (!arg || nr_args != 1)
+			break;
+		ret = io_register_iowq_max_unbound(ctx, arg);
+		break;
 	default:
 		ret = -EINVAL;
 		break;
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 45a4f2373694..bb6845e14629 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -309,6 +309,9 @@ enum {
 	IORING_REGISTER_IOWQ_AFF		= 17,
 	IORING_UNREGISTER_IOWQ_AFF		= 18,
 
+	/* set/get max number of unbounded workers */
+	IORING_REGISTER_IOWQ_MAX_UNBOUND	= 19,
+
 	/* this goes last */
 	IORING_REGISTER_LAST
 };
-- 
Jens Axboe


^ permalink raw reply related	[flat|nested] only message in thread

only message in thread, other threads:[~2021-08-27 17:38 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-08-27 17:38 [PATCH] io-wq: provide a way to limit max number of unbounded workers Jens Axboe

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.