| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20240324230116.1348576-82-sashal@kernel.org>
Date: Sun, 24 Mar 2024 18:51:58 -0400
From: Sasha Levin <sashal@...nel.org>
To: linux-kernel@...r.kernel.org,
stable@...r.kernel.org
Cc: Jens Axboe <axboe@...nel.dk>,
Sasha Levin <sashal@...nel.org>
Subject: [PATCH 6.6 081/638] io_uring: remove unconditional looping in local task_work handling
From: Jens Axboe <axboe@...nel.dk>
[ Upstream commit 9fe3eaea4a3530ca34a8d8ff00b1848c528789ca ]
If we have a ton of notifications coming in, we can be looping in here
for a long time. This can be problematic for various reasons, mostly
because we can starve userspace. If the application is waiting on N
events, then only re-run if we need more events.
Fixes: c0e0d6ba25f1 ("io_uring: add IORING_SETUP_DEFER_TASKRUN")
Signed-off-by: Jens Axboe <axboe@...nel.dk>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
io_uring/io_uring.c | 44 +++++++++++++++++++++++++++++---------------
1 file changed, 29 insertions(+), 15 deletions(-)
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index e22f6bf837f9d..049a163d0fdd0 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -1363,7 +1363,20 @@ static void __cold io_move_task_work_from_local(struct io_ring_ctx *ctx)
}
}
-static int __io_run_local_work(struct io_ring_ctx *ctx, struct io_tw_state *ts)
+static bool io_run_local_work_continue(struct io_ring_ctx *ctx, int events,
+ int min_events)
+{
+ if (llist_empty(&ctx->work_llist))
+ return false;
+ if (events < min_events)
+ return true;
+ if (ctx->flags & IORING_SETUP_TASKRUN_FLAG)
+ atomic_or(IORING_SQ_TASKRUN, &ctx->rings->sq_flags);
+ return false;
+}
+
+static int __io_run_local_work(struct io_ring_ctx *ctx, struct io_tw_state *ts,
+ int min_events)
{
struct llist_node *node;
unsigned int loops = 0;
@@ -1392,18 +1405,20 @@ static int __io_run_local_work(struct io_ring_ctx *ctx, struct io_tw_state *ts)
}
loops++;
- if (!llist_empty(&ctx->work_llist))
+ if (io_run_local_work_continue(ctx, ret, min_events))
goto again;
if (ts->locked) {
io_submit_flush_completions(ctx);
- if (!llist_empty(&ctx->work_llist))
+ if (io_run_local_work_continue(ctx, ret, min_events))
goto again;
}
+
trace_io_uring_local_work_run(ctx, ret, loops);
return ret;
}
-static inline int io_run_local_work_locked(struct io_ring_ctx *ctx)
+static inline int io_run_local_work_locked(struct io_ring_ctx *ctx,
+ int min_events)
{
struct io_tw_state ts = { .locked = true, };
int ret;
@@ -1411,20 +1426,20 @@ static inline int io_run_local_work_locked(struct io_ring_ctx *ctx)
if (llist_empty(&ctx->work_llist))
return 0;
- ret = __io_run_local_work(ctx, &ts);
+ ret = __io_run_local_work(ctx, &ts, min_events);
/* shouldn't happen! */
if (WARN_ON_ONCE(!ts.locked))
mutex_lock(&ctx->uring_lock);
return ret;
}
-static int io_run_local_work(struct io_ring_ctx *ctx)
+static int io_run_local_work(struct io_ring_ctx *ctx, int min_events)
{
struct io_tw_state ts = {};
int ret;
ts.locked = mutex_trylock(&ctx->uring_lock);
- ret = __io_run_local_work(ctx, &ts);
+ ret = __io_run_local_work(ctx, &ts, min_events);
if (ts.locked)
mutex_unlock(&ctx->uring_lock);
@@ -1620,7 +1635,7 @@ static int io_iopoll_check(struct io_ring_ctx *ctx, long min)
io_task_work_pending(ctx)) {
u32 tail = ctx->cached_cq_tail;
- (void) io_run_local_work_locked(ctx);
+ (void) io_run_local_work_locked(ctx, min);
if (task_work_pending(current) ||
wq_list_empty(&ctx->iopoll_list)) {
@@ -2463,7 +2478,7 @@ int io_run_task_work_sig(struct io_ring_ctx *ctx)
{
if (!llist_empty(&ctx->work_llist)) {
__set_current_state(TASK_RUNNING);
- if (io_run_local_work(ctx) > 0)
+ if (io_run_local_work(ctx, INT_MAX) > 0)
return 0;
}
if (io_run_task_work() > 0)
@@ -2531,7 +2546,7 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
if (!io_allowed_run_tw(ctx))
return -EEXIST;
if (!llist_empty(&ctx->work_llist))
- io_run_local_work(ctx);
+ io_run_local_work(ctx, min_events);
io_run_task_work();
io_cqring_overflow_flush(ctx);
/* if user messes with these they will just get an early return */
@@ -2569,11 +2584,10 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
trace_io_uring_cqring_wait(ctx, min_events);
do {
+ int nr_wait = (int) iowq.cq_tail - READ_ONCE(ctx->rings->cq.tail);
unsigned long check_cq;
if (ctx->flags & IORING_SETUP_DEFER_TASKRUN) {
- int nr_wait = (int) iowq.cq_tail - READ_ONCE(ctx->rings->cq.tail);
-
atomic_set(&ctx->cq_wait_nr, nr_wait);
set_current_state(TASK_INTERRUPTIBLE);
} else {
@@ -2592,7 +2606,7 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
*/
io_run_task_work();
if (!llist_empty(&ctx->work_llist))
- io_run_local_work(ctx);
+ io_run_local_work(ctx, nr_wait);
/*
* Non-local task_work will be run on exit to userspace, but
@@ -3283,7 +3297,7 @@ static __cold bool io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
if ((ctx->flags & IORING_SETUP_DEFER_TASKRUN) &&
io_allowed_defer_tw_run(ctx))
- ret |= io_run_local_work(ctx) > 0;
+ ret |= io_run_local_work(ctx, INT_MAX) > 0;
ret |= io_cancel_defer_files(ctx, task, cancel_all);
mutex_lock(&ctx->uring_lock);
ret |= io_poll_remove_all(ctx, task, cancel_all);
@@ -3642,7 +3656,7 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
* it should handle ownership problems if any.
*/
if (ctx->flags & IORING_SETUP_DEFER_TASKRUN)
- (void)io_run_local_work_locked(ctx);
+ (void)io_run_local_work_locked(ctx, min_complete);
}
mutex_unlock(&ctx->uring_lock);
}
--
2.43.0
Powered by blists - more mailing lists