25aba572ca
There were quite a few reports in the community forum about Windows
VMs with SATA disks not working after upgrading to kernel 5.13.
Issue was reproducible during the installation of Win2019 (suggested
by Thomas), and it's already fixed in 5.15. Bisecting led to
io-wq: split bounded and unbounded work into separate lists
as the commit fixing the issue.
Indeed, the commit states
Fixes: ecc53c48c13d ("io-wq: check max_worker limits if a worker transitions bound state")
which is present as a backport in ubuntu-impish:
f9eb79f840052285408ae9082dc4419dc1397954
The first backport
io-wq: fix queue stalling race
also sounds nice to have and additionally served as a preparation for
the second one to apply more cleanly.
Signed-off-by: Fabian Ebner <f.ebner@proxmox.com>
73 lines
2.1 KiB
Diff
73 lines
2.1 KiB
Diff
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
|
|
From: Jens Axboe <axboe@kernel.dk>
|
|
Date: Tue, 31 Aug 2021 13:53:00 -0600
|
|
Subject: [PATCH] io-wq: fix queue stalling race
|
|
|
|
We need to set the stalled bit early, before we drop the lock for adding
|
|
us to the stall hash queue. If not, then we can race with new work being
|
|
queued between adding us to the stall hash and io_worker_handle_work()
|
|
marking us stalled.
|
|
|
|
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
[backport]
|
|
Signed-off-by: Fabian Ebner <f.ebner@proxmox.com>
|
|
---
|
|
fs/io-wq.c | 15 +++++++--------
|
|
1 file changed, 7 insertions(+), 8 deletions(-)
|
|
|
|
diff --git a/fs/io-wq.c b/fs/io-wq.c
|
|
index 6612d0aa497e..33678185f3bc 100644
|
|
--- a/fs/io-wq.c
|
|
+++ b/fs/io-wq.c
|
|
@@ -437,8 +437,7 @@ static bool io_worker_can_run_work(struct io_worker *worker,
|
|
}
|
|
|
|
static struct io_wq_work *io_get_next_work(struct io_wqe *wqe,
|
|
- struct io_worker *worker,
|
|
- bool *stalled)
|
|
+ struct io_worker *worker)
|
|
__must_hold(wqe->lock)
|
|
{
|
|
struct io_wq_work_node *node, *prev;
|
|
@@ -476,10 +475,14 @@ static struct io_wq_work *io_get_next_work(struct io_wqe *wqe,
|
|
}
|
|
|
|
if (stall_hash != -1U) {
|
|
+ /*
|
|
+ * Set this before dropping the lock to avoid racing with new
|
|
+ * work being added and clearing the stalled bit.
|
|
+ */
|
|
+ wqe->flags |= IO_WQE_FLAG_STALLED;
|
|
raw_spin_unlock(&wqe->lock);
|
|
io_wait_on_hash(wqe, stall_hash);
|
|
raw_spin_lock(&wqe->lock);
|
|
- *stalled = true;
|
|
}
|
|
|
|
return NULL;
|
|
@@ -519,7 +522,6 @@ static void io_worker_handle_work(struct io_worker *worker)
|
|
|
|
do {
|
|
struct io_wq_work *work;
|
|
- bool stalled;
|
|
get_next:
|
|
/*
|
|
* If we got some work, mark us as busy. If we didn't, but
|
|
@@ -528,12 +530,9 @@ static void io_worker_handle_work(struct io_worker *worker)
|
|
* can't make progress, any work completion or insertion will
|
|
* clear the stalled flag.
|
|
*/
|
|
- stalled = false;
|
|
- work = io_get_next_work(wqe, worker, &stalled);
|
|
+ work = io_get_next_work(wqe, worker);
|
|
if (work)
|
|
__io_worker_busy(wqe, worker, work);
|
|
- else if (stalled)
|
|
- wqe->flags |= IO_WQE_FLAG_STALLED;
|
|
|
|
raw_spin_unlock_irq(&wqe->lock);
|
|
if (!work)
|
|
--
|
|
2.30.2
|
|
|