io_uring: fix buffer auto-commit for multishot uring_cmd
PATCH 1/1 io_uring: fix buffer auto-commit for multishot uring_cmd
Commit 620a50c92700 ("io_uring: uring_cmd: add multishot support") added multishot uring_cmd support with explicit buffer upfront commit via io_uring_mshot_cmd_post_cqe(). However, the buffer selection path in io_ring_buffer_select() was auto-committing buffers for non-pollable files, which conflicts with uring_cmd's explicit upfront commit model.
This way consumes the whole selected buffer immediately, and causes failure on the following buffer selection.
Fix this by adding io_commit_kbuf_upfront() to identify operations that handle buffer commit explicitly (currently only IORING_OP_URING_CMD), and skip auto-commit for these operations.
--- io_uring/kbuf.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)
io_uring/kbuf.c +8 -1
@@ -155,6 +155,12 @@ static int io_provided_buffers_select(struct io_kiocb *req, size_t *len,return 1 ; } +/* So far, uring_cmd commits kbuf upfront, no need to auto-commit */ +static bool io_commit_kbuf_upfront (const struct io_kiocb *req) +{ +return req->opcode ==IORING_OP_URING_CMD ; +} +static struct io_br_sel io_ring_buffer_select (struct io_kiocb *req,size_t *len,struct io_buffer_list *bl,unsigned int issue_flags)
@@ -181,7 +187,8 @@ static struct io_br_sel io_ring_buffer_select(struct io_kiocb *req, size_t *len, sel.buf_list = bl; sel.addr =u64_to_user_ptr (buf->addr); -if (issue_flags &IO_URING_F_UNLOCKED || !io_file_can_poll (req)) { +if (issue_flags &IO_URING_F_UNLOCKED || (!io_file_can_poll (req) && + !io_commit_kbuf_upfront (req))) {/* * If we came in unlocked, we have no choice but to consume the * buffer here, otherwise nothing ensures that the buffer won't
Signature · 1 line
2.47.1
Jens Axboe
maintainerQuoted 38 lines · Ming Lei
Commit
620a50c92700("io_uring:uring_cmd: add multishot support") added multishoturing_cmdsupport with explicit buffer upfront commit viaio_uring_mshot_cmd_post_cqe(). However, the buffer selection path inio_ring_buffer_select()was auto-committing buffers for non-pollable files, which conflicts with uring_cmd's explicit upfront commit model.This way consumes the whole selected buffer immediately, and causes failure on the following buffer selection.
Fix this by adding
io_commit_kbuf_upfront()to identify operations that handle buffer commit explicitly (currently onlyIORING_OP_URING_CMD), and skip auto-commit for these operations.Cc Caleb Sander Mateos <[email protected]>Fixes 620a50c92700 ("io_uring: uring_cmd: add multishot support")Signed-off-by Ming Lei <[email protected]>--- io_uring/kbuf.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)io_uring/kbuf.c +6
-0@@ -155,6 +155,12 @@ static int io_provided_buffers_select(struct io_kiocb *req, size_t *len,return 1 ; } +/* So far, uring_cmd commits kbuf upfront, no need to auto-commit */ +static bool io_commit_kbuf_upfront (const struct io_kiocb *req) +{ +return req->opcode ==IORING_OP_URING_CMD ; +} +static struct io_br_sel io_ring_buffer_select (struct io_kiocb *req,size_t *len,struct io_buffer_list *bl,unsigned int issue_flags)
io_uring/kbuf.c static struct io_br_sel io_ring_buffer_select(struct io_kiocb *req, size_t *len, · line 194
Quoted 9 lines
io_uring/kbuf.c +2
-1@@ -181,7 +187,8 @@ static struct io_br_sel io_ring_buffer_select(struct io_kiocb *req, size_t *len, sel.buf_list = bl; sel.addr =u64_to_user_ptr (buf->addr); -if (issue_flags &IO_URING_F_UNLOCKED || !io_file_can_poll (req)) { +if (issue_flags &IO_URING_F_UNLOCKED || (!io_file_can_poll (req) && + !io_commit_kbuf_upfront (req))) {/* * If we came in unlocked, we have no choice but to consume the * buffer here, otherwise nothing ensures that the buffer won't
Might be cleaner to put this logic in a helper instead, ala:
static bool io_should_commit(req, issue_flags)
{
if (issue_flags & IO_URING_F_UNLOCKED)
return true;
if (!io_file_can_poll(req) && req->opcode == IORING_OP_URING_CMD)
return true;
return false;
}and just add the appropriate comments there, rather than add a new random helper.
Signature · 1 line
Jens Axboe