git
/
linux
mirror of https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

io_uring/kbuf: switch to storing struct io_buffer_list locally 

Currently the buffer list is stored in struct io_kiocb. The buffer list
can be of two types:

1) Classic/legacy buffer list. These don't need to get referenced after
   a buffer pick, and hence storing them in struct io_kiocb is perfectly
   fine.

2) Ring provided buffer lists. These DO need to be referenced after the
   initial buffer pick, as they need to get consumed later on. This can
   be either just incrementing the head of the ring, or it can be
   consuming parts of a buffer if incremental buffer consumptions has
   been configured.

For case 2, io_uring needs to be careful not to access the buffer list
after the initial pick-and-execute context. The core does recycling of
these, but it's easy to make a mistake, because it's stored in the
io_kiocb which does persist across multiple execution contexts. Either
because it's a multishot request, or simply because it needed some kind
of async trigger (eg poll) for retry purposes.

Add a struct io_buffer_list to struct io_br_sel, which is always on
stack for the various users of it. This prevents the buffer list from
leaking outside of that execution context, and additionally it enables
kbuf to not even pass back the struct io_buffer_list if the given
context isn't appropriately locked already.

This doesn't fix any bugs, it's simply a defensive measure to prevent
any issues with reuse of a buffer list.

Link: https://lore.kernel.org/r/20250821020750.598432-12-axboe@kernel.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This commit is contained in:
Jens Axboe 2025-08-20 20:03:39 -06:00
parent 461382a51f
commit 5fda512554
7 changed files with 55 additions and 74 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
include/linux/io_uring_types.h
View File

@ -674,12 +674,6 @@ struct io_kiocb {
/* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */
struct io_buffer *kbuf;
/*
* stores buffer ID for ring provided buffers, valid IFF
* REQ_F_BUFFER_RING is set.
*/
struct io_buffer_list *buf_list;
struct io_rsrc_node *buf_node;
};

6
io_uring/io_uring.c
View File

@ -1007,7 +1007,7 @@ void io_req_defer_failed(struct io_kiocb *req, s32 res)
lockdep_assert_held(&req->ctx->uring_lock);
req_set_fail(req);
io_req_set_res(req, res, io_put_kbuf(req, res, req->buf_list));
io_req_set_res(req, res, io_put_kbuf(req, res, NULL));
if (def->fail)
def->fail(req);
io_req_complete_defer(req);
@ -2025,11 +2025,11 @@ fail:
switch (io_arm_poll_handler(req, 0)) {
case IO_APOLL_READY:
io_kbuf_recycle(req, req->buf_list, 0);
io_kbuf_recycle(req, NULL, 0);
io_req_task_queue(req);
break;
case IO_APOLL_ABORTED:
io_kbuf_recycle(req, req->buf_list, 0);
io_kbuf_recycle(req, NULL, 0);
io_queue_iowq(req);
break;
case IO_APOLL_OK:

27
io_uring/kbuf.c
View File

@ -171,8 +171,8 @@ static struct io_br_sel io_ring_buffer_select(struct io_kiocb *req, size_t *len,
if (*len == 0 || *len > buf->len)
*len = buf->len;
req->flags |= REQ_F_BUFFER_RING | REQ_F_BUFFERS_COMMIT;
req->buf_list = bl;
req->buf_index = buf->bid;
sel.buf_list = bl;
sel.addr = u64_to_user_ptr(buf->addr);
if (issue_flags & IO_URING_F_UNLOCKED || !io_file_can_poll(req)) {
@ -186,8 +186,8 @@ static struct io_br_sel io_ring_buffer_select(struct io_kiocb *req, size_t *len,
* the transfer completes (or if we get -EAGAIN and must poll of
* retry).
*/
io_kbuf_commit(req, bl, *len, 1);
req->buf_list = NULL;
io_kbuf_commit(req, sel.buf_list, *len, 1);
sel.buf_list = NULL;
}
return sel;
}
@ -294,7 +294,6 @@ static int io_ring_buffers_peek(struct io_kiocb *req, struct buf_sel_arg *arg,
req->flags |= REQ_F_BL_EMPTY;
req->flags |= REQ_F_BUFFER_RING;
req->buf_list = bl;
return iov - arg->iovs;
}
@ -302,16 +301,15 @@ int io_buffers_select(struct io_kiocb *req, struct buf_sel_arg *arg,
struct io_br_sel *sel, unsigned int issue_flags)
{
struct io_ring_ctx *ctx = req->ctx;
struct io_buffer_list *bl;
int ret = -ENOENT;
io_ring_submit_lock(ctx, issue_flags);
bl = io_buffer_get_list(ctx, arg->buf_group);
if (unlikely(!bl))
sel->buf_list = io_buffer_get_list(ctx, arg->buf_group);
if (unlikely(!sel->buf_list))
goto out_unlock;
if (bl->flags & IOBL_BUF_RING) {
ret = io_ring_buffers_peek(req, arg, bl);
if (sel->buf_list->flags & IOBL_BUF_RING) {
ret = io_ring_buffers_peek(req, arg, sel->buf_list);
/*
* Don't recycle these buffers if we need to go through poll.
* Nobody else can use them anyway, and holding on to provided
@ -321,13 +319,16 @@ int io_buffers_select(struct io_kiocb *req, struct buf_sel_arg *arg,
*/
if (ret > 0) {
req->flags |= REQ_F_BUFFERS_COMMIT | REQ_F_BL_NO_RECYCLE;
io_kbuf_commit(req, bl, arg->out_len, ret);
io_kbuf_commit(req, sel->buf_list, arg->out_len, ret);
}
} else {
ret = io_provided_buffers_select(req, &arg->out_len, bl, arg->iovs);
ret = io_provided_buffers_select(req, &arg->out_len, sel->buf_list, arg->iovs);
}
out_unlock:
io_ring_submit_unlock(ctx, issue_flags);
if (issue_flags & IO_URING_F_UNLOCKED) {
sel->buf_list = NULL;
mutex_unlock(&ctx->uring_lock);
}
return ret;
}
@ -348,10 +349,12 @@ int io_buffers_peek(struct io_kiocb *req, struct buf_sel_arg *arg,
ret = io_ring_buffers_peek(req, arg, bl);
if (ret > 0)
req->flags |= REQ_F_BUFFERS_COMMIT;
sel->buf_list = bl;
return ret;
}
/* don't support multiple buffer selections for legacy */
sel->buf_list = NULL;
return io_provided_buffers_select(req, &arg->max_len, bl, arg->iovs);
}

16
io_uring/kbuf.h
View File

@ -63,11 +63,14 @@ struct buf_sel_arg {
};
/*
* Return value from io_buffer_list selection. Just returns the error or
* user address for now, will be extended to return the buffer list in the
* future.
* Return value from io_buffer_list selection, to avoid stashing it in
* struct io_kiocb. For legacy/classic provided buffers, keeping a reference
* across execution contexts are fine. But for ring provided buffers, the
* list may go away as soon as ->uring_lock is dropped. As the io_kiocb
* persists, it's better to just keep the buffer local for those cases.
*/
struct io_br_sel {
struct io_buffer_list *buf_list;
/*
* Some selection parts return the user address, others return an error.
*/
@ -107,13 +110,6 @@ struct io_mapped_region *io_pbuf_get_region(struct io_ring_ctx *ctx,
static inline bool io_kbuf_recycle_ring(struct io_kiocb *req,
struct io_buffer_list *bl)
{
/*
* We don't need to recycle for REQ_F_BUFFER_RING, we can just clear
* the flag and hence ensure that bl->head doesn't get incremented.
* If the tail has already been incremented, hang on to it.
* The exception is partial io, that case we should increment bl->head
* to monopolize the buffer.
*/
if (bl) {
req->flags &= ~(REQ_F_BUFFER_RING|REQ_F_BUFFERS_COMMIT);
return true;

46
io_uring/net.c
View File

@ -433,7 +433,6 @@ int io_sendmsg_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
if (req->opcode == IORING_OP_SENDMSG)
return -EINVAL;
sr->msg_flags |= MSG_WAITALL;
req->buf_list = NULL;
req->flags |= REQ_F_MULTISHOT;
}
@ -512,11 +511,11 @@ static inline bool io_send_finish(struct io_kiocb *req,
unsigned int cflags;
if (!(sr->flags & IORING_RECVSEND_BUNDLE)) {
cflags = io_put_kbuf(req, sel->val, req->buf_list);
cflags = io_put_kbuf(req, sel->val, sel->buf_list);
goto finish;
}
cflags = io_put_kbufs(req, sel->val, req->buf_list, io_bundle_nbufs(kmsg, sel->val));
cflags = io_put_kbufs(req, sel->val, sel->buf_list, io_bundle_nbufs(kmsg, sel->val));
if (bundle_finished || req->flags & REQ_F_BL_EMPTY)
goto finish;
@ -657,6 +656,7 @@ int io_send(struct io_kiocb *req, unsigned int issue_flags)
flags |= MSG_DONTWAIT;
retry_bundle:
sel.buf_list = NULL;
if (io_do_buffer_select(req)) {
ret = io_send_select_buffer(req, issue_flags, &sel, kmsg);
if (ret)
@ -682,7 +682,7 @@ retry_bundle:
sr->len -= ret;
sr->buf += ret;
sr->done_io += ret;
return io_net_kbuf_recyle(req, req->buf_list, kmsg, ret);
return io_net_kbuf_recyle(req, sel.buf_list, kmsg, ret);
}
if (ret == -ERESTARTSYS)
ret = -EINTR;
@ -795,18 +795,8 @@ int io_recvmsg_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
req->flags |= REQ_F_NOWAIT;
if (sr->msg_flags & MSG_ERRQUEUE)
req->flags |= REQ_F_CLEAR_POLLIN;
if (req->flags & REQ_F_BUFFER_SELECT) {
/*
* Store the buffer group for this multishot receive separately,
* as if we end up doing an io-wq based issue that selects a
* buffer, it has to be committed immediately and that will
* clear ->buf_list. This means we lose the link to the buffer
* list, and the eventual buffer put on completion then cannot
* restore it.
*/
if (req->flags & REQ_F_BUFFER_SELECT)
sr->buf_group = req->buf_index;
req->buf_list = NULL;
}
sr->mshot_total_len = sr->mshot_len = 0;
if (sr->flags & IORING_RECV_MULTISHOT) {
if (!(req->flags & REQ_F_BUFFER_SELECT))
@ -874,7 +864,7 @@ static inline bool io_recv_finish(struct io_kiocb *req,
if (sr->flags & IORING_RECVSEND_BUNDLE) {
size_t this_ret = sel->val - sr->done_io;
cflags |= io_put_kbufs(req, this_ret, req->buf_list, io_bundle_nbufs(kmsg, this_ret));
cflags |= io_put_kbufs(req, this_ret, sel->buf_list, io_bundle_nbufs(kmsg, this_ret));
if (sr->flags & IORING_RECV_RETRY)
cflags = req->cqe.flags | (cflags & CQE_F_MASK);
if (sr->mshot_len && sel->val >= sr->mshot_len)
@ -896,7 +886,7 @@ static inline bool io_recv_finish(struct io_kiocb *req,
return false;
}
} else {
cflags |= io_put_kbuf(req, sel->val, req->buf_list);
cflags |= io_put_kbuf(req, sel->val, sel->buf_list);
}
/*
@ -1038,6 +1028,7 @@ int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags)
flags |= MSG_DONTWAIT;
retry_multishot:
sel.buf_list = NULL;
if (io_do_buffer_select(req)) {
size_t len = sr->len;
@ -1048,7 +1039,7 @@ retry_multishot:
if (req->flags & REQ_F_APOLL_MULTISHOT) {
ret = io_recvmsg_prep_multishot(kmsg, sr, &sel.addr, &len);
if (ret) {
io_kbuf_recycle(req, req->buf_list, issue_flags);
io_kbuf_recycle(req, sel.buf_list, issue_flags);
return ret;
}
}
@ -1072,14 +1063,12 @@ retry_multishot:
if (ret < min_ret) {
if (ret == -EAGAIN && force_nonblock) {
if (issue_flags & IO_URING_F_MULTISHOT)
io_kbuf_recycle(req, req->buf_list, issue_flags);
io_kbuf_recycle(req, sel.buf_list, issue_flags);
return IOU_RETRY;
}
if (ret > 0 && io_net_retry(sock, flags)) {
sr->done_io += ret;
return io_net_kbuf_recyle(req, req->buf_list, kmsg, ret);
return io_net_kbuf_recyle(req, sel.buf_list, kmsg, ret);
}
if (ret == -ERESTARTSYS)
ret = -EINTR;
@ -1093,7 +1082,7 @@ retry_multishot:
else if (sr->done_io)
ret = sr->done_io;
else
io_kbuf_recycle(req, req->buf_list, issue_flags);
io_kbuf_recycle(req, sel.buf_list, issue_flags);
sel.val = ret;
if (!io_recv_finish(req, kmsg, &sel, mshot_finished, issue_flags))
@ -1178,7 +1167,7 @@ int io_recv(struct io_kiocb *req, unsigned int issue_flags)
{
struct io_sr_msg *sr = io_kiocb_to_cmd(req, struct io_sr_msg);
struct io_async_msghdr *kmsg = req->async_data;
struct io_br_sel sel = { };
struct io_br_sel sel;
struct socket *sock;
unsigned flags;
int ret, min_ret = 0;
@ -1198,6 +1187,7 @@ int io_recv(struct io_kiocb *req, unsigned int issue_flags)
flags |= MSG_DONTWAIT;
retry_multishot:
sel.buf_list = NULL;
if (io_do_buffer_select(req)) {
sel.val = sr->len;
ret = io_recv_buf_select(req, kmsg, &sel, issue_flags);
@ -1217,16 +1207,14 @@ retry_multishot:
ret = sock_recvmsg(sock, &kmsg->msg, flags);
if (ret < min_ret) {
if (ret == -EAGAIN && force_nonblock) {
if (issue_flags & IO_URING_F_MULTISHOT)
io_kbuf_recycle(req, req->buf_list, issue_flags);
io_kbuf_recycle(req, sel.buf_list, issue_flags);
return IOU_RETRY;
}
if (ret > 0 && io_net_retry(sock, flags)) {
sr->len -= ret;
sr->buf += ret;
sr->done_io += ret;
return io_net_kbuf_recyle(req, req->buf_list, kmsg, ret);
return io_net_kbuf_recyle(req, sel.buf_list, kmsg, ret);
}
if (ret == -ERESTARTSYS)
ret = -EINTR;
@ -1242,7 +1230,7 @@ out_free:
else if (sr->done_io)
ret = sr->done_io;
else
io_kbuf_recycle(req, req->buf_list, issue_flags);
io_kbuf_recycle(req, sel.buf_list, issue_flags);
sel.val = ret;
if (!io_recv_finish(req, kmsg, &sel, mshot_finished, issue_flags))

6
io_uring/poll.c
View File

@ -316,10 +316,10 @@ void io_poll_task_func(struct io_kiocb *req, io_tw_token_t tw)
ret = io_poll_check_events(req, tw);
if (ret == IOU_POLL_NO_ACTION) {
io_kbuf_recycle(req, req->buf_list, 0);
io_kbuf_recycle(req, NULL, 0);
return;
} else if (ret == IOU_POLL_REQUEUE) {
io_kbuf_recycle(req, req->buf_list, 0);
io_kbuf_recycle(req, NULL, 0);
__io_poll_execute(req, 0);
return;
}
@ -686,7 +686,7 @@ int io_arm_apoll(struct io_kiocb *req, unsigned issue_flags, __poll_t mask)
req->flags |= REQ_F_POLLED;
ipt.pt._qproc = io_async_queue_proc;
io_kbuf_recycle(req, req->buf_list, issue_flags);
io_kbuf_recycle(req, NULL, issue_flags);
ret = __io_arm_poll_handler(req, &apoll->poll, &ipt, mask, issue_flags);
if (ret)

22
io_uring/rw.c
View File

@ -579,7 +579,7 @@ void io_req_rw_complete(struct io_kiocb *req, io_tw_token_t tw)
io_req_io_end(req);
if (req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))
req->cqe.flags |= io_put_kbuf(req, req->cqe.res, req->buf_list);
req->cqe.flags |= io_put_kbuf(req, req->cqe.res, NULL);
io_req_rw_cleanup(req, 0);
io_req_task_complete(req, tw);
@ -648,7 +648,7 @@ static inline void io_rw_done(struct io_kiocb *req, ssize_t ret)
}
static int kiocb_done(struct io_kiocb *req, ssize_t ret,
unsigned int issue_flags)
struct io_br_sel *sel, unsigned int issue_flags)
{
struct io_rw *rw = io_kiocb_to_cmd(req, struct io_rw);
unsigned final_ret = io_fixup_rw_res(req, ret);
@ -662,7 +662,7 @@ static int kiocb_done(struct io_kiocb *req, ssize_t ret,
* from the submission path.
*/
io_req_io_end(req);
io_req_set_res(req, final_ret, io_put_kbuf(req, ret, req->buf_list));
io_req_set_res(req, final_ret, io_put_kbuf(req, ret, sel->buf_list));
io_req_rw_cleanup(req, issue_flags);
return IOU_COMPLETE;
} else {
@ -1024,10 +1024,10 @@ int io_read(struct io_kiocb *req, unsigned int issue_flags)
ret = __io_read(req, &sel, issue_flags);
if (ret >= 0)
return kiocb_done(req, ret, issue_flags);
return kiocb_done(req, ret, &sel, issue_flags);
if (req->flags & REQ_F_BUFFERS_COMMIT)
io_kbuf_recycle(req, req->buf_list, issue_flags);
io_kbuf_recycle(req, sel.buf_list, issue_flags);
return ret;
}
@ -1057,15 +1057,15 @@ int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
* Reset rw->len to 0 again to avoid clamping future mshot
* reads, in case the buffer size varies.
*/
if (io_kbuf_recycle(req, req->buf_list, issue_flags))
if (io_kbuf_recycle(req, sel.buf_list, issue_flags))
rw->len = 0;
return IOU_RETRY;
} else if (ret <= 0) {
io_kbuf_recycle(req, req->buf_list, issue_flags);
io_kbuf_recycle(req, sel.buf_list, issue_flags);
if (ret < 0)
req_set_fail(req);
} else if (!(req->flags & REQ_F_APOLL_MULTISHOT)) {
cflags = io_put_kbuf(req, ret, req->buf_list);
cflags = io_put_kbuf(req, ret, sel.buf_list);
} else {
/*
* Any successful return value will keep the multishot read
@ -1073,7 +1073,7 @@ int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
* we fail to post a CQE, or multishot is no longer set, then
* jump to the termination path. This request is then done.
*/
cflags = io_put_kbuf(req, ret, req->buf_list);
cflags = io_put_kbuf(req, ret, sel.buf_list);
rw->len = 0; /* similarly to above, reset len to 0 */
if (io_req_post_cqe(req, ret, cflags | IORING_CQE_F_MORE)) {
@ -1202,7 +1202,7 @@ int io_write(struct io_kiocb *req, unsigned int issue_flags)
return -EAGAIN;
}
done:
return kiocb_done(req, ret2, issue_flags);
return kiocb_done(req, ret2, NULL, issue_flags);
} else {
ret_eagain:
iov_iter_restore(&io->iter, &io->iter_state);
@ -1370,7 +1370,7 @@ int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin)
if (!smp_load_acquire(&req->iopoll_completed))
break;
nr_events++;
req->cqe.flags = io_put_kbuf(req, req->cqe.res, req->buf_list);
req->cqe.flags = io_put_kbuf(req, req->cqe.res, NULL);
if (req->opcode != IORING_OP_URING_CMD)
io_req_rw_cleanup(req, 0);
}