korg/linux
0
0
Fork 0
mirror of https://mirrors.bfsu.edu.cn/git/linux.git synced 2024-11-26 21:54:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

io_uring: move struct io_kiocb from task_struct to io_uring_task

Browse Source 
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.

This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
This commit is contained in:
Jens Axboe 2024-11-03 10:23:38 -07:00
parent 6ed368cc5d
commit b6f58a3f4a
14 changed files with 48 additions and 36 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
include/linux/io_uring/cmd.h
View File

@ -110,7 +110,7 @@ static inline void io_uring_cmd_complete_in_task(struct io_uring_cmd *ioucmd,
static inline struct task_struct *io_uring_cmd_get_task(struct io_uring_cmd *cmd)
{
return cmd_to_io_kiocb(cmd)->task;
return cmd_to_io_kiocb(cmd)->tctx->task;
}
#endif /* _LINUX_IO_URING_CMD_H */

3
include/linux/io_uring_types.h
View File

@ -84,6 +84,7 @@ struct io_uring_task {
/* submission side */
int cached_refs;
const struct io_ring_ctx *last;
struct task_struct *task;
struct io_wq *io_wq;
struct file *registered_rings[IO_RINGFD_REG_MAX];
@ -625,7 +626,7 @@ struct io_kiocb {
struct io_cqe cqe;
struct io_ring_ctx *ctx;
struct task_struct *task;
struct io_uring_task *tctx;
union {
/* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */

2
io_uring/cancel.c
View File

@ -205,7 +205,7 @@ int io_async_cancel(struct io_kiocb *req, unsigned int issue_flags)
.opcode = cancel->opcode,
.seq = atomic_inc_return(&req->ctx->cancel_seq),
};
struct io_uring_task *tctx = req->task->io_uring;
struct io_uring_task *tctx = req->tctx;
int ret;
if (cd.flags & IORING_ASYNC_CANCEL_FD) {

2
io_uring/fdinfo.c
View File

@ -203,7 +203,7 @@ __cold void io_uring_show_fdinfo(struct seq_file *m, struct file *file)
hlist_for_each_entry(req, &hb->list, hash_node)
seq_printf(m, " op=%d, task_works=%d\n", req->opcode,
task_work_pending(req->task));
task_work_pending(req->tctx->task));
}
if (has_lock)

34
io_uring/io_uring.c
View File

@ -206,7 +206,7 @@ bool io_match_task_safe(struct io_kiocb *head, struct io_uring_task *tctx,
{
bool matched;
if (tctx && head->task->io_uring != tctx)
if (tctx && head->tctx != tctx)
return false;
if (cancel_all)
return true;
@ -407,11 +407,8 @@ static void io_clean_op(struct io_kiocb *req)
kfree(req->apoll);
req->apoll = NULL;
}
if (req->flags & REQ_F_INFLIGHT) {
struct io_uring_task *tctx = req->task->io_uring;
atomic_dec(&tctx->inflight_tracked);
}
if (req->flags & REQ_F_INFLIGHT)
atomic_dec(&req->tctx->inflight_tracked);
if (req->flags & REQ_F_CREDS)
put_cred(req->creds);
if (req->flags & REQ_F_ASYNC_DATA) {
@ -425,7 +422,7 @@ static inline void io_req_track_inflight(struct io_kiocb *req)
{
if (!(req->flags & REQ_F_INFLIGHT)) {
req->flags |= REQ_F_INFLIGHT;
atomic_inc(&req->task->io_uring->inflight_tracked);
atomic_inc(&req->tctx->inflight_tracked);
}
}
@ -514,7 +511,7 @@ static void io_prep_async_link(struct io_kiocb *req)
static void io_queue_iowq(struct io_kiocb *req)
{
struct io_kiocb *link = io_prep_linked_timeout(req);
struct io_uring_task *tctx = req->task->io_uring;
struct io_uring_task *tctx = req->tctx;
BUG_ON(!tctx);
BUG_ON(!tctx->io_wq);
@ -529,7 +526,7 @@ static void io_queue_iowq(struct io_kiocb *req)
* procedure rather than attempt to run this request (or create a new
* worker for it).
*/
if (WARN_ON_ONCE(!same_thread_group(req->task, current)))
if (WARN_ON_ONCE(!same_thread_group(tctx->task, current)))
atomic_or(IO_WQ_WORK_CANCEL, &req->work.flags);
trace_io_uring_queue_async_work(req, io_wq_is_hashed(&req->work));
@ -678,17 +675,17 @@ static void io_cqring_do_overflow_flush(struct io_ring_ctx *ctx)
}
/* must to be called somewhat shortly after putting a request */
static inline void io_put_task(struct task_struct *task)
static inline void io_put_task(struct io_kiocb *req)
{
struct io_uring_task *tctx = task->io_uring;
struct io_uring_task *tctx = req->tctx;
if (likely(task == current)) {
if (likely(tctx->task == current)) {
tctx->cached_refs++;
} else {
percpu_counter_sub(&tctx->inflight, 1);
if (unlikely(atomic_read(&tctx->in_cancel)))
wake_up(&tctx->wait);
put_task_struct(task);
put_task_struct(tctx->task);
}
}
@ -1207,7 +1204,7 @@ static inline void io_req_local_work_add(struct io_kiocb *req,
static void io_req_normal_work_add(struct io_kiocb *req)
{
struct io_uring_task *tctx = req->task->io_uring;
struct io_uring_task *tctx = req->tctx;
struct io_ring_ctx *ctx = req->ctx;
/* task_work already pending, we're done */
@ -1226,7 +1223,7 @@ static void io_req_normal_work_add(struct io_kiocb *req)
return;
}
if (likely(!task_work_add(req->task, &tctx->task_work, ctx->notify_method)))
if (likely(!task_work_add(tctx->task, &tctx->task_work, ctx->notify_method)))
return;
io_fallback_tw(tctx, false);
@ -1343,8 +1340,7 @@ static void io_req_task_cancel(struct io_kiocb *req, struct io_tw_state *ts)
void io_req_task_submit(struct io_kiocb *req, struct io_tw_state *ts)
{
io_tw_lock(req->ctx, ts);
/* req->task == current here, checking PF_EXITING is safe */
if (unlikely(req->task->flags & PF_EXITING))
if (unlikely(io_should_terminate_tw()))
io_req_defer_failed(req, -EFAULT);
else if (req->flags & REQ_F_FORCE_ASYNC)
io_queue_iowq(req);
@ -1403,7 +1399,7 @@ static void io_free_batch_list(struct io_ring_ctx *ctx,
}
io_put_file(req);
io_req_put_rsrc_nodes(req);
io_put_task(req->task);
io_put_task(req);
node = req->comp_list.next;
io_req_add_to_cache(req, ctx);
@ -2019,7 +2015,7 @@ static int io_init_req(struct io_ring_ctx *ctx, struct io_kiocb *req,
req->flags = (__force io_req_flags_t) sqe_flags;
req->cqe.user_data = READ_ONCE(sqe->user_data);
req->file = NULL;
req->task = current;
req->tctx = current->io_uring;
req->cancel_seq_set = false;
if (unlikely(opcode >= IORING_OP_LAST)) {

13
io_uring/io_uring.h
View File

@ -426,6 +426,19 @@ static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
ctx->submitter_task == current);
}
/*
* Terminate the request if either of these conditions are true:
*
* 1) It's being executed by the original task, but that task is marked
* with PF_EXITING as it's exiting.
* 2) PF_KTHREAD is set, in which case the invoker of the task_work is
* our fallback task_work.
*/
static inline bool io_should_terminate_tw(void)
{
return current->flags & (PF_KTHREAD | PF_EXITING);
}
static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
{
io_req_set_res(req, res, 0);

4
io_uring/msg_ring.c
View File

@ -89,8 +89,8 @@ static void io_msg_tw_complete(struct io_kiocb *req, struct io_tw_state *ts)
static int io_msg_remote_post(struct io_ring_ctx *ctx, struct io_kiocb *req,
int res, u32 cflags, u64 user_data)
{
req->task = READ_ONCE(ctx->submitter_task);
if (!req->task) {
req->tctx = READ_ONCE(ctx->submitter_task->io_uring);
if (!req->tctx) {
kmem_cache_free(req_cachep, req);
return -EOWNERDEAD;
}

4
io_uring/notif.c
View File

@ -89,7 +89,7 @@ static int io_link_skb(struct sk_buff *skb, struct ubuf_info *uarg)
/* make sure all noifications can be finished in the same task_work */
if (unlikely(notif->ctx != prev_notif->ctx ||
notif->task != prev_notif->task))
notif->tctx != prev_notif->tctx))
return -EEXIST;
nd->head = prev_nd->head;
@ -115,7 +115,7 @@ struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
notif->opcode = IORING_OP_NOP;
notif->flags = 0;
notif->file = NULL;
notif->task = current;
notif->tctx = current->io_uring;
io_get_task_refs(1);
notif->file_node = NULL;
notif->buf_node = NULL;

3
io_uring/poll.c
View File

@ -224,8 +224,7 @@ static int io_poll_check_events(struct io_kiocb *req, struct io_tw_state *ts)
{
int v;
/* req->task == current here, checking PF_EXITING is safe */
if (unlikely(req->task->flags & PF_EXITING))
if (unlikely(io_should_terminate_tw()))
return -ECANCELED;
do {

2
io_uring/rw.c
View File

@ -435,7 +435,7 @@ static bool io_rw_should_reissue(struct io_kiocb *req)
* Play it safe and assume not safe to re-import and reissue if we're
* not in the original thread group (or in task context).
*/
if (!same_thread_group(req->task, current) || !in_task())
if (!same_thread_group(req->tctx->task, current) || !in_task())
return false;
return true;
}

1
io_uring/tctx.c
View File

@ -81,6 +81,7 @@ __cold int io_uring_alloc_task_context(struct task_struct *task,
return ret;
}
tctx->task = task;
xa_init(&tctx->xa);
init_waitqueue_head(&tctx->wait);
atomic_set(&tctx->in_cancel, 0);

10
io_uring/timeout.c
View File

@ -300,16 +300,18 @@ static void io_req_task_link_timeout(struct io_kiocb *req, struct io_tw_state *t
{
struct io_timeout *timeout = io_kiocb_to_cmd(req, struct io_timeout);
struct io_kiocb *prev = timeout->prev;
int ret = -ENOENT;
int ret;
if (prev) {
if (!(req->task->flags & PF_EXITING)) {
if (!io_should_terminate_tw()) {
struct io_cancel_data cd = {
.ctx = req->ctx,
.data = prev->cqe.user_data,
};
ret = io_try_cancel(req->task->io_uring, &cd, 0);
ret = io_try_cancel(req->tctx, &cd, 0);
} else {
ret = -ECANCELED;
}
io_req_set_res(req, ret ?: -ETIME, 0);
io_req_task_complete(req, ts);
@ -643,7 +645,7 @@ static bool io_match_task(struct io_kiocb *head, struct io_uring_task *tctx,
{
struct io_kiocb *req;
if (tctx && head->task->io_uring != tctx)
if (tctx && head->tctx != tctx)
return false;
if (cancel_all)
return true;

2
io_uring/uring_cmd.c
View File

@ -61,7 +61,7 @@ bool io_uring_try_cancel_uring_cmd(struct io_ring_ctx *ctx,
struct io_uring_cmd);
struct file *file = req->file;
if (!cancel_all && req->task->io_uring != tctx)
if (!cancel_all && req->tctx != tctx)
continue;
if (cmd->flags & IORING_URING_CMD_CANCELABLE) {

2
io_uring/waitid.c
View File

@ -331,7 +331,7 @@ int io_waitid(struct io_kiocb *req, unsigned int issue_flags)
hlist_add_head(&req->hash_node, &ctx->waitid_list);
init_waitqueue_func_entry(&iwa->wo.child_wait, io_waitid_wait);
iwa->wo.child_wait.private = req->task;
iwa->wo.child_wait.private = req->tctx->task;
iw->head = &current->signal->wait_chldexit;
add_wait_queue(iw->head, &iwa->wo.child_wait);