io_uring/rw: keep CQE flags on iopoll requests when adding kbuf flags

io_do_iopoll() assigns the result of io_put_kbuf() to the request's CQE
flags upon completion. This overwrites any CQE flags that may have been
set by the opcode-specific layer. (For example, if __io_uring_cmd_done()
had set IORING_CQE_F_32, it would be cleared.) Switch the = to an |= so
the kbuf flags are added to the existing CQE flags rather than replacing
them. io_req_rw_complete() does the same with the io_put_kbuf() result.

Fixes: e26dca67fd ("io_uring: add support for IORING_SETUP_CQE_MIXED")
Reported-by: sashiko-bot@kernel.org
Link: https://sashiko.dev/#/message/20260827191705.D53C91F000E9%40smtp.kernel.org
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Anuj Gupta <anuj20.g@samsung.com>
Link: https://patch.msgid.link/20260902210200.2336720-1-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This commit is contained in:
Caleb Sander Mateos 2026-09-02 15:01:59 -06:00 committed by Jens Axboe
parent 6028b54388
commit 47ccc3f1c6

View File

@ -880,6 +880,7 @@ static int io_rw_init_file(struct io_kiocb *req, fmode_t mode, int rw_type)
kiocb->private = NULL;
kiocb->ki_flags |= IOCB_HIPRI;
req->iopoll_completed = 0;
req->cqe.flags = 0;
if (ctx->flags & IORING_SETUP_HYBRID_IOPOLL) {
/* make sure every req only blocks once*/
req->flags &= ~REQ_F_IOPOLL_STATE;
@ -1382,7 +1383,7 @@ int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin)
list_del(&req->iopoll_node);
wq_list_add_tail(&req->comp_list, &ctx->submit_state.compl_reqs);
nr_events++;
req->cqe.flags = io_put_kbuf(req, max(req->cqe.res, 0), NULL);
req->cqe.flags |= io_put_kbuf(req, max(req->cqe.res, 0), NULL);
if (!io_is_uring_cmd(req))
io_req_rw_cleanup(req, 0);
}