From: gregkh@linuxfoundation.org
To: linux-kernel@vger.kernel.org, stable@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Jens Axboe <axboe@kernel.dk>,
Pavel Begunkov <asml.silence@gmail.com>
Subject: [PATCH 5.11 09/36] io_uring/io-wq: return 2-step work swap scheme
Date: Wed, 10 Mar 2021 14:23:22 +0100 [thread overview]
Message-ID: <20210310132320.812008964@linuxfoundation.org> (raw)
In-Reply-To: <20210310132320.510840709@linuxfoundation.org>
From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
From: Pavel Begunkov <asml.silence@gmail.com>
commit 5280f7e530f71ba85baf90169393196976ad0e52 upstream
Saving one lock/unlock for io-wq is not super important, but adds some
ugliness in the code. More important, atomic decs not turning it to zero
for some archs won't give the right ordering/barriers so the
io_steal_work() may pretty easily get subtly and completely broken.
Return back 2-step io-wq work exchange and clean it up.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
fs/io-wq.c | 16 ++++++----------
fs/io-wq.h | 4 ++--
fs/io_uring.c | 26 ++++----------------------
3 files changed, 12 insertions(+), 34 deletions(-)
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -555,23 +555,21 @@ get_next:
/* handle a whole dependent link */
do {
- struct io_wq_work *old_work, *next_hashed, *linked;
+ struct io_wq_work *next_hashed, *linked;
unsigned int hash = io_get_work_hash(work);
next_hashed = wq_next_work(work);
io_impersonate_work(worker, work);
+ wq->do_work(work);
+ io_assign_current_work(worker, NULL);
- old_work = work;
- linked = wq->do_work(work);
-
+ linked = wq->free_work(work);
work = next_hashed;
if (!work && linked && !io_wq_is_hashed(linked)) {
work = linked;
linked = NULL;
}
io_assign_current_work(worker, work);
- wq->free_work(old_work);
-
if (linked)
io_wqe_enqueue(wqe, linked);
@@ -850,11 +848,9 @@ static void io_run_cancel(struct io_wq_w
struct io_wq *wq = wqe->wq;
do {
- struct io_wq_work *old_work = work;
-
work->flags |= IO_WQ_WORK_CANCEL;
- work = wq->do_work(work);
- wq->free_work(old_work);
+ wq->do_work(work);
+ work = wq->free_work(work);
} while (work);
}
--- a/fs/io-wq.h
+++ b/fs/io-wq.h
@@ -106,8 +106,8 @@ static inline struct io_wq_work *wq_next
return container_of(work->list.next, struct io_wq_work, list);
}
-typedef void (free_work_fn)(struct io_wq_work *);
-typedef struct io_wq_work *(io_wq_work_fn)(struct io_wq_work *);
+typedef struct io_wq_work *(free_work_fn)(struct io_wq_work *);
+typedef void (io_wq_work_fn)(struct io_wq_work *);
struct io_wq_data {
struct user_struct *user;
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2365,22 +2365,6 @@ static inline void io_put_req_deferred(s
io_free_req_deferred(req);
}
-static struct io_wq_work *io_steal_work(struct io_kiocb *req)
-{
- struct io_kiocb *nxt;
-
- /*
- * A ref is owned by io-wq in which context we're. So, if that's the
- * last one, it's safe to steal next work. False negatives are Ok,
- * it just will be re-punted async in io_put_work()
- */
- if (refcount_read(&req->refs) != 1)
- return NULL;
-
- nxt = io_req_find_next(req);
- return nxt ? &nxt->work : NULL;
-}
-
static void io_double_put_req(struct io_kiocb *req)
{
/* drop both submit and complete references */
@@ -6378,7 +6362,7 @@ static int io_issue_sqe(struct io_kiocb
return 0;
}
-static struct io_wq_work *io_wq_submit_work(struct io_wq_work *work)
+static void io_wq_submit_work(struct io_wq_work *work)
{
struct io_kiocb *req = container_of(work, struct io_kiocb, work);
struct io_kiocb *timeout;
@@ -6429,8 +6413,6 @@ static struct io_wq_work *io_wq_submit_w
if (lock_ctx)
mutex_unlock(&lock_ctx->uring_lock);
}
-
- return io_steal_work(req);
}
static inline struct file *io_file_from_index(struct io_ring_ctx *ctx,
@@ -8062,12 +8044,12 @@ static int io_sqe_files_update(struct io
return __io_sqe_files_update(ctx, &up, nr_args);
}
-static void io_free_work(struct io_wq_work *work)
+static struct io_wq_work *io_free_work(struct io_wq_work *work)
{
struct io_kiocb *req = container_of(work, struct io_kiocb, work);
- /* Consider that io_steal_work() relies on this ref */
- io_put_req(req);
+ req = io_put_req_find_next(req);
+ return req ? &req->work : NULL;
}
static int io_init_wq_offload(struct io_ring_ctx *ctx,
next prev parent reply other threads:[~2021-03-10 13:25 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-10 13:23 [PATCH 5.11 00/36] 5.11.6-rc1 review gregkh
2021-03-10 13:23 ` [PATCH 5.11 01/36] ACPICA: Fix race in generic_serial_bus (I2C) and GPIO op_region parameter handling gregkh
2021-03-10 13:23 ` [PATCH 5.11 02/36] io_uring: fix inconsistent lock state gregkh
2021-03-10 13:23 ` [PATCH 5.11 03/36] io_uring: deduplicate core cancellations sequence gregkh
2021-03-10 13:23 ` [PATCH 5.11 04/36] io_uring: unpark SQPOLL thread for cancelation gregkh
2021-03-10 13:23 ` [PATCH 5.11 05/36] io_uring: deduplicate failing task_work_add gregkh
2021-03-10 13:23 ` [PATCH 5.11 06/36] fs: provide locked helper variant of close_fd_get_file() gregkh
2021-03-10 13:23 ` [PATCH 5.11 07/36] io_uring: get rid of intermediate IORING_OP_CLOSE stage gregkh
2021-03-10 13:23 ` [PATCH 5.11 08/36] io_uring/io-wq: kill off now unused IO_WQ_WORK_NO_CANCEL gregkh
2021-03-10 13:23 ` gregkh [this message]
2021-03-10 13:23 ` [PATCH 5.11 10/36] io_uring: dont take uring_lock during iowq cancel gregkh
2021-03-10 13:23 ` [PATCH 5.11 11/36] media: cedrus: Remove checking for required controls gregkh
2021-03-10 13:23 ` [PATCH 5.11 12/36] nvme-pci: mark Kingston SKC2000 as not supporting the deepest power state gregkh
2021-03-10 13:23 ` [PATCH 5.11 13/36] parisc: Enable -mlong-calls gcc option with CONFIG_COMPILE_TEST gregkh
2021-03-10 13:23 ` [PATCH 5.11 14/36] arm64: Make CPU_BIG_ENDIAN depend on ld.bfd or ld.lld 13.0.0+ gregkh
2021-03-10 13:23 ` [PATCH 5.11 15/36] btrfs: export and rename qgroup_reserve_meta gregkh
2021-03-10 13:23 ` [PATCH 5.11 16/36] btrfs: dont flush from btrfs_delayed_inode_reserve_metadata gregkh
2021-03-10 13:23 ` [PATCH 5.11 17/36] iommu/amd: Fix sleeping in atomic in increase_address_space() gregkh
2021-03-10 13:23 ` [PATCH 5.11 18/36] scsi: ufs-mediatek: Enable UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL gregkh
2021-03-10 13:23 ` [PATCH 5.11 19/36] scsi: ufs: Add a quirk to permit overriding UniPro defaults gregkh
2021-03-10 13:23 ` [PATCH 5.11 20/36] misc: eeprom_93xx46: Add quirk to support Microchip 93LC46B eeprom gregkh
2021-03-10 13:23 ` [PATCH 5.11 21/36] scsi: ufs: Introduce a quirk to allow only page-aligned sg entries gregkh
2021-03-10 13:23 ` [PATCH 5.11 22/36] scsi: ufs: ufs-exynos: Apply vendor-specific values for three timeouts gregkh
2021-03-10 13:23 ` [PATCH 5.11 23/36] scsi: ufs: ufs-exynos: Use UFSHCD_QUIRK_ALIGN_SG_WITH_PAGE_SIZE gregkh
2021-03-10 13:23 ` [PATCH 5.11 24/36] drm/msm/a5xx: Remove overwriting A5XX_PC_DBG_ECO_CNTL register gregkh
2021-03-10 13:23 ` [PATCH 5.11 25/36] mmc: sdhci-of-dwcmshc: set SDHCI_QUIRK2_PRESET_VALUE_BROKEN gregkh
2021-03-10 13:23 ` [PATCH 5.11 26/36] HID: i2c-hid: Add I2C_HID_QUIRK_NO_IRQ_AFTER_RESET for ITE8568 EC on Voyo Winpad A15 gregkh
2021-03-10 13:23 ` [PATCH 5.11 27/36] ALSA: usb-audio: Add DJM750 to Pioneer mixer quirk gregkh
2021-03-10 13:23 ` [PATCH 5.11 28/36] ALSA: usb-audio: add mixer quirks for Pioneer DJM-900NXS2 gregkh
2021-03-10 13:23 ` [PATCH 5.11 29/36] HID: ite: Enable QUIRK_TOUCHPAD_ON_OFF_REPORT on Acer Aspire Switch 10E gregkh
2021-03-10 13:23 ` [PATCH 5.11 30/36] PCI: cadence: Retrain Link to work around Gen2 training defect gregkh
2021-03-10 13:23 ` [PATCH 5.11 31/36] ASoC: Intel: sof_sdw: reorganize quirks by generation gregkh
2021-03-10 13:23 ` [PATCH 5.11 32/36] ASoC: Intel: sof_sdw: add quirk for HP Spectre x360 convertible gregkh
2021-03-10 13:23 ` [PATCH 5.11 33/36] scsi: ufs: Fix a duplicate dev quirk number gregkh
2021-03-10 13:23 ` [PATCH 5.11 34/36] KVM: SVM: Clear the CR4 register on reset gregkh
2021-03-10 13:23 ` [PATCH 5.11 35/36] nvme-pci: mark Seagate Nytro XM1440 as QUIRK_NO_NS_DESC_LIST gregkh
2021-03-10 13:23 ` [PATCH 5.11 36/36] nvme-pci: add quirks for Lexar 256GB SSD gregkh
2021-03-10 21:59 ` [PATCH 5.11 00/36] 5.11.6-rc1 review Shuah Khan
2021-03-11 17:37 ` Greg KH
2021-03-10 23:53 ` Guenter Roeck
2021-03-11 17:37 ` Greg KH
2021-03-11 3:21 ` Naresh Kamboju
2021-03-11 17:38 ` Greg Kroah-Hartman
2021-03-11 4:09 ` Ross Schmidt
2021-03-11 17:39 ` Greg KH
2021-03-11 7:59 ` Jon Hunter
2021-03-11 17:39 ` Greg KH
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=20210310132320.812008964@linuxfoundation.org \
--to=gregkh@linuxfoundation.org \
--cc=asml.silence@gmail.com \
--cc=axboe@kernel.dk \
--cc=linux-kernel@vger.kernel.org \
--cc=stable@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.