From: Jens Axboe <axboe@kernel.dk>
To: io-uring@vger.kernel.org
Cc: Jens Axboe <axboe@kernel.dk>
Subject: [PATCH 2/2] io_uring: allow user configurable IO thread CPU affinity
Date: Thu, 17 Jun 2021 10:29:44 -0600 [thread overview]
Message-ID: <20210617162944.524917-3-axboe@kernel.dk> (raw)
In-Reply-To: <20210617162944.524917-1-axboe@kernel.dk>
io-wq defaults to per-node masks for IO workers. This works fine by
default, but isn't particularly handy for workloads that prefer more
specific affinities, for either performance or isolation reasons.
This adds IORING_REGISTER_IOWQ_AFF that allows the user to pass in a CPU
mask that is then applied to IO thread workers, and an
IORING_UNREGISTER_IOWQ_AFF that simply resets the masks back to the
default of per-node.
Note that no care is given to existing IO threads, they will need to go
through a reschedule before the affinity is correct if they are already
running or sleeping.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
fs/io-wq.c | 17 ++++++++++++
fs/io-wq.h | 2 ++
fs/io_uring.c | 51 +++++++++++++++++++++++++++++++++++
include/uapi/linux/io_uring.h | 4 +++
4 files changed, 74 insertions(+)
diff --git a/fs/io-wq.c b/fs/io-wq.c
index 2af8e1df4646..bb4d3ee9592e 100644
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -1087,6 +1087,23 @@ static int io_wq_cpu_offline(unsigned int cpu, struct hlist_node *node)
return __io_wq_cpu_online(wq, cpu, false);
}
+int io_wq_cpu_affinity(struct io_wq *wq, cpumask_var_t mask)
+{
+ int i;
+
+ rcu_read_lock();
+ for_each_node(i) {
+ struct io_wqe *wqe = wq->wqes[i];
+
+ if (mask)
+ cpumask_copy(wqe->cpu_mask, mask);
+ else
+ cpumask_copy(wqe->cpu_mask, cpumask_of_node(i));
+ }
+ rcu_read_unlock();
+ return 0;
+}
+
static __init int io_wq_init(void)
{
int ret;
diff --git a/fs/io-wq.h b/fs/io-wq.h
index af2df0680ee2..02299cdcf55c 100644
--- a/fs/io-wq.h
+++ b/fs/io-wq.h
@@ -128,6 +128,8 @@ void io_wq_put_and_exit(struct io_wq *wq);
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);
+
static inline bool io_wq_is_hashed(struct io_wq_work *work)
{
return work->flags & IO_WQ_WORK_HASHED;
diff --git a/fs/io_uring.c b/fs/io_uring.c
index d916eb2cef09..46a25a7cb70a 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -9983,6 +9983,43 @@ static int io_register_rsrc(struct io_ring_ctx *ctx, void __user *arg,
return -EINVAL;
}
+static int io_register_iowq_aff(struct io_ring_ctx *ctx, void __user *arg,
+ unsigned len)
+{
+ struct io_uring_task *tctx = current->io_uring;
+ cpumask_var_t new_mask;
+ int ret;
+
+ if (!tctx || !tctx->io_wq)
+ return -EINVAL;
+
+ if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
+ return -ENOMEM;
+
+ cpumask_clear(new_mask);
+ if (len > cpumask_size())
+ len = cpumask_size();
+
+ if (copy_from_user(new_mask, arg, len)) {
+ free_cpumask_var(new_mask);
+ return -EFAULT;
+ }
+
+ ret = io_wq_cpu_affinity(tctx->io_wq, new_mask);
+ free_cpumask_var(new_mask);
+ return ret;
+}
+
+static int io_unregister_iowq_aff(struct io_ring_ctx *ctx)
+{
+ struct io_uring_task *tctx = current->io_uring;
+
+ if (!tctx || !tctx->io_wq)
+ return -EINVAL;
+
+ return io_wq_cpu_affinity(tctx->io_wq, NULL);
+}
+
static bool io_register_op_must_quiesce(int op)
{
switch (op) {
@@ -9998,6 +10035,8 @@ static bool io_register_op_must_quiesce(int op)
case IORING_REGISTER_FILES_UPDATE2:
case IORING_REGISTER_BUFFERS2:
case IORING_REGISTER_BUFFERS_UPDATE:
+ case IORING_REGISTER_IOWQ_AFF:
+ case IORING_UNREGISTER_IOWQ_AFF:
return false;
default:
return true;
@@ -10137,6 +10176,18 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
ret = io_register_rsrc_update(ctx, arg, nr_args,
IORING_RSRC_BUFFER);
break;
+ case IORING_REGISTER_IOWQ_AFF:
+ ret = -EINVAL;
+ if (!arg || !nr_args)
+ break;
+ ret = io_register_iowq_aff(ctx, arg, nr_args);
+ break;
+ case IORING_UNREGISTER_IOWQ_AFF:
+ ret = -EINVAL;
+ if (arg || nr_args)
+ break;
+ ret = io_unregister_iowq_aff(ctx);
+ break;
default:
ret = -EINVAL;
break;
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 162ff99ed2cb..f1f9ac114b51 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -306,6 +306,10 @@ enum {
IORING_REGISTER_BUFFERS2 = 15,
IORING_REGISTER_BUFFERS_UPDATE = 16,
+ /* set/clear io-wq thread affinities */
+ IORING_REGISTER_IOWQ_AFF = 17,
+ IORING_UNREGISTER_IOWQ_AFF = 18,
+
/* this goes last */
IORING_REGISTER_LAST
};
--
2.32.0
prev parent reply other threads:[~2021-06-17 16:30 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-17 16:29 [PATCHSET 0/2] Allow io-wq user configurable CPU masks Jens Axboe
2021-06-17 16:29 ` [PATCH 1/2] io-wq: use private CPU mask Jens Axboe
2021-06-17 16:29 ` Jens Axboe [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210617162944.524917-3-axboe@kernel.dk \
--to=axboe@kernel.dk \
--cc=io-uring@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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.