From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: linux-kernel@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
stable@vger.kernel.org, "Agarwal, Anchal" <anchalag@amazon.com>,
Jens Axboe <axboe@kernel.dk>, Sasha Levin <sashal@kernel.org>
Subject: [PATCH 5.4 01/65] io_uring: make sure async workqueue is canceled on exit
Date: Tue, 7 Jul 2020 17:16:40 +0200 [thread overview]
Message-ID: <20200707145752.491283624@linuxfoundation.org> (raw)
In-Reply-To: <20200707145752.417212219@linuxfoundation.org>
From: Jens Axboe <axboe@kernel.dk>
Track async work items that we queue, so we can safely cancel them
if the ring is closed or the process exits. Newer kernels handle
this automatically with io-wq, but the old workqueue based setup needs
a bit of special help to get there.
There's no upstream variant of this, as that would require backporting
all the io-wq changes from 5.5 and on. Hence I made a one-off that
ensures that we don't leak memory if we have async work items that
need active cancelation (like socket IO).
Reported-by: Agarwal, Anchal <anchalag@amazon.com>
Tested-by: Agarwal, Anchal <anchalag@amazon.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
fs/io_uring.c | 63 +++++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 63 insertions(+)
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 7fa3cd3fff4d2..e0200406765c3 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -267,6 +267,9 @@ struct io_ring_ctx {
#if defined(CONFIG_UNIX)
struct socket *ring_sock;
#endif
+
+ struct list_head task_list;
+ spinlock_t task_lock;
};
struct sqe_submit {
@@ -331,14 +334,18 @@ struct io_kiocb {
#define REQ_F_ISREG 2048 /* regular file */
#define REQ_F_MUST_PUNT 4096 /* must be punted even for NONBLOCK */
#define REQ_F_TIMEOUT_NOSEQ 8192 /* no timeout sequence */
+#define REQ_F_CANCEL 16384 /* cancel request */
unsigned long fsize;
u64 user_data;
u32 result;
u32 sequence;
+ struct task_struct *task;
struct fs_struct *fs;
struct work_struct work;
+ struct task_struct *work_task;
+ struct list_head task_list;
};
#define IO_PLUG_THRESHOLD 2
@@ -425,6 +432,8 @@ static struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
INIT_LIST_HEAD(&ctx->cancel_list);
INIT_LIST_HEAD(&ctx->defer_list);
INIT_LIST_HEAD(&ctx->timeout_list);
+ INIT_LIST_HEAD(&ctx->task_list);
+ spin_lock_init(&ctx->task_lock);
return ctx;
}
@@ -492,6 +501,7 @@ static void __io_commit_cqring(struct io_ring_ctx *ctx)
static inline void io_queue_async_work(struct io_ring_ctx *ctx,
struct io_kiocb *req)
{
+ unsigned long flags;
int rw = 0;
if (req->submit.sqe) {
@@ -503,6 +513,13 @@ static inline void io_queue_async_work(struct io_ring_ctx *ctx,
}
}
+ req->task = current;
+
+ spin_lock_irqsave(&ctx->task_lock, flags);
+ list_add(&req->task_list, &ctx->task_list);
+ req->work_task = NULL;
+ spin_unlock_irqrestore(&ctx->task_lock, flags);
+
queue_work(ctx->sqo_wq[rw], &req->work);
}
@@ -2201,6 +2218,8 @@ static void io_sq_wq_submit_work(struct work_struct *work)
old_cred = override_creds(ctx->creds);
async_list = io_async_list_from_sqe(ctx, req->submit.sqe);
+
+ allow_kernel_signal(SIGINT);
restart:
do {
struct sqe_submit *s = &req->submit;
@@ -2232,6 +2251,12 @@ static void io_sq_wq_submit_work(struct work_struct *work)
}
if (!ret) {
+ req->work_task = current;
+ if (req->flags & REQ_F_CANCEL) {
+ ret = -ECANCELED;
+ goto end_req;
+ }
+
s->has_user = cur_mm != NULL;
s->needs_lock = true;
do {
@@ -2246,6 +2271,12 @@ static void io_sq_wq_submit_work(struct work_struct *work)
break;
cond_resched();
} while (1);
+end_req:
+ if (!list_empty(&req->task_list)) {
+ spin_lock_irq(&ctx->task_lock);
+ list_del_init(&req->task_list);
+ spin_unlock_irq(&ctx->task_lock);
+ }
}
/* drop submission reference */
@@ -2311,6 +2342,7 @@ static void io_sq_wq_submit_work(struct work_struct *work)
}
out:
+ disallow_signal(SIGINT);
if (cur_mm) {
set_fs(old_fs);
unuse_mm(cur_mm);
@@ -3675,12 +3707,32 @@ static int io_uring_fasync(int fd, struct file *file, int on)
return fasync_helper(fd, file, on, &ctx->cq_fasync);
}
+static void io_cancel_async_work(struct io_ring_ctx *ctx,
+ struct task_struct *task)
+{
+ if (list_empty(&ctx->task_list))
+ return;
+
+ spin_lock_irq(&ctx->task_lock);
+ while (!list_empty(&ctx->task_list)) {
+ struct io_kiocb *req;
+
+ req = list_first_entry(&ctx->task_list, struct io_kiocb, task_list);
+ list_del_init(&req->task_list);
+ req->flags |= REQ_F_CANCEL;
+ if (req->work_task && (!task || req->task == task))
+ send_sig(SIGINT, req->work_task, 1);
+ }
+ spin_unlock_irq(&ctx->task_lock);
+}
+
static void io_ring_ctx_wait_and_kill(struct io_ring_ctx *ctx)
{
mutex_lock(&ctx->uring_lock);
percpu_ref_kill(&ctx->refs);
mutex_unlock(&ctx->uring_lock);
+ io_cancel_async_work(ctx, NULL);
io_kill_timeouts(ctx);
io_poll_remove_all(ctx);
io_iopoll_reap_events(ctx);
@@ -3688,6 +3740,16 @@ static void io_ring_ctx_wait_and_kill(struct io_ring_ctx *ctx)
io_ring_ctx_free(ctx);
}
+static int io_uring_flush(struct file *file, void *data)
+{
+ struct io_ring_ctx *ctx = file->private_data;
+
+ if (fatal_signal_pending(current) || (current->flags & PF_EXITING))
+ io_cancel_async_work(ctx, current);
+
+ return 0;
+}
+
static int io_uring_release(struct inode *inode, struct file *file)
{
struct io_ring_ctx *ctx = file->private_data;
@@ -3792,6 +3854,7 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
static const struct file_operations io_uring_fops = {
.release = io_uring_release,
+ .flush = io_uring_flush,
.mmap = io_uring_mmap,
.poll = io_uring_poll,
.fasync = io_uring_fasync,
--
2.25.1
next prev parent reply other threads:[~2020-07-07 15:28 UTC|newest]
Thread overview: 78+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-07-07 15:16 [PATCH 5.4 00/65] 5.4.51-rc1 review Greg Kroah-Hartman
2020-07-07 15:16 ` Greg Kroah-Hartman [this message]
2020-07-07 15:16 ` [PATCH 5.4 02/65] mm: fix swap cache node allocation mask Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 03/65] EDAC/amd64: Read back the scrub rate PCI register on F15h Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 04/65] usbnet: smsc95xx: Fix use-after-free after removal Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 05/65] sched/debug: Make sd->flags sysctl read-only Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 06/65] mm/slub.c: fix corrupted freechain in deactivate_slab() Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 07/65] mm/slub: fix stack overruns with SLUB_STATS Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 08/65] rxrpc: Fix race between incoming ACK parser and retransmitter Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 09/65] usb: usbtest: fix missing kfree(dev->buf) in usbtest_disconnect Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 10/65] tools lib traceevent: Add append() function helper for appending strings Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 11/65] tools lib traceevent: Handle __attribute__((user)) in field names Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 12/65] s390/debug: avoid kernel warning on too large number of pages Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 13/65] nvme-multipath: set bdi capabilities once Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 14/65] nvme: fix possible deadlock when I/O is blocked Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 15/65] nvme-multipath: fix deadlock between ana_work and scan_work Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 16/65] nvme-multipath: fix deadlock due to head->lock Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 17/65] nvme-multipath: fix bogus request queue reference put Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 18/65] kgdb: Avoid suspicious RCU usage warning Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 19/65] selftests: tpm: Use /bin/sh instead of /bin/bash Greg Kroah-Hartman
2020-07-07 15:16 ` [PATCH 5.4 20/65] tpm: Fix TIS locality timeout problems Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 21/65] crypto: af_alg - fix use-after-free in af_alg_accept() due to bh_lock_sock() Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 22/65] drm/msm/dpu: fix error return code in dpu_encoder_init Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 23/65] rxrpc: Fix afs large storage transmission performance drop Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 24/65] RDMA/counter: Query a counter before release Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 25/65] cxgb4: use unaligned conversion for fetching timestamp Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 26/65] cxgb4: parse TC-U32 key values and masks natively Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 27/65] cxgb4: fix endian conversions for L4 ports in filters Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 28/65] cxgb4: use correct type for all-mask IP address comparison Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 29/65] cxgb4: fix SGE queue dump destination buffer context Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 30/65] hwmon: (max6697) Make sure the OVERT mask is set correctly Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 31/65] hwmon: (acpi_power_meter) Fix potential memory leak in acpi_power_meter_add() Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 32/65] thermal/drivers/mediatek: Fix bank number settings on mt8183 Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 33/65] thermal/drivers/rcar_gen3: Fix undefined temperature if negative Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 34/65] kthread: save thread function Greg Kroah-Hartman
2020-07-07 15:31 ` J. Bruce Fields
2020-07-07 15:17 ` [PATCH 5.4 35/65] nfsd: clients dont need to break their own delegations Greg Kroah-Hartman
2020-07-07 15:31 ` J. Bruce Fields
2020-07-07 17:20 ` Sasha Levin
2020-07-07 17:29 ` Sasha Levin
2020-07-07 17:31 ` J. Bruce Fields
2020-07-07 15:17 ` [PATCH 5.4 36/65] nfsd4: fix nfsdfs reference count loop Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 37/65] nfsd: fix nfsdfs inode reference count leak Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 38/65] drm: sun4i: hdmi: Remove extra HPD polling Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 39/65] virtio-blk: free vblk-vqs in error path of virtblk_probe() Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 40/65] SMB3: Honor posix flag for multiuser mounts Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 41/65] nvme: fix identify error status silent ignore Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 42/65] nvme: fix a crash in nvme_mpath_add_disk Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 43/65] samples/vfs: avoid warning in statx override Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 44/65] i2c: algo-pca: Add 0x78 as SCL stuck low status for PCA9665 Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 45/65] i2c: mlxcpld: check correct size of maximum RECV_LEN packet Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 46/65] spi: spi-fsl-dspi: Fix external abort on interrupt in resume or exit paths Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 47/65] nfsd: apply umask on fs without ACL support Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 48/65] Revert "ALSA: usb-audio: Improve frames size computation" Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 49/65] SMB3: Honor seal flag for multiuser mounts Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 50/65] SMB3: Honor persistent/resilient handle flags " Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 51/65] SMB3: Honor lease disabling " Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 52/65] SMB3: Honor handletimeout flag " Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 53/65] cifs: Fix the target file was deleted when rename failed Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 54/65] MIPS: lantiq: xway: sysctrl: fix the GPHY clock alias names Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 55/65] MIPS: Add missing EHB in mtc0 -> mfc0 sequence for DSPen Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 56/65] drm/amd/display: Only revalidate bandwidth on medium and fast updates Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 57/65] drm/amdgpu: use %u rather than %d for sclk/mclk Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 58/65] drm/amdgpu/atomfirmware: fix vram_info fetching for renoir Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 59/65] dma-buf: Move dma_buf_release() from fops to dentry_ops Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 60/65] irqchip/gic: Atomically update affinity Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 61/65] mm, compaction: fully assume capture is not NULL in compact_zone_order() Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 62/65] mm, compaction: make capture control handling safe wrt interrupts Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 63/65] x86/resctrl: Fix memory bandwidth counter width for AMD Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 64/65] dm zoned: assign max_io_len correctly Greg Kroah-Hartman
2020-07-07 15:17 ` [PATCH 5.4 65/65] efi: Make it possible to disable efivar_ssdt entirely Greg Kroah-Hartman
2020-07-08 5:30 ` [PATCH 5.4 00/65] 5.4.51-rc1 review Naresh Kamboju
2020-07-08 8:41 ` Jon Hunter
2020-07-08 15:15 ` Greg Kroah-Hartman
2020-07-08 16:58 ` Jon Hunter
2020-07-09 9:30 ` Greg Kroah-Hartman
2020-07-08 15:00 ` Shuah Khan
2020-07-08 17:53 ` Guenter Roeck
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=20200707145752.491283624@linuxfoundation.org \
--to=gregkh@linuxfoundation.org \
--cc=anchalag@amazon.com \
--cc=axboe@kernel.dk \
--cc=linux-kernel@vger.kernel.org \
--cc=sashal@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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).