1io_uring_queue_init(3) liburing Manual io_uring_queue_init(3)
2
6 io_uring_queue_init - setup io_uring submission and completion queues
7
9 #include <liburing.h>
10 int io_uring_queue_init(unsigned entries,
12 struct io_uring *ring,
13 unsigned flags);
14 int io_uring_queue_init_params(unsigned entries,
16 struct io_uring *ring,
17 struct io_uring_params *params);
18 int io_uring_queue_init_mem(unsigned entries,
20 struct io_uring *ring,
21 struct io_uring_params *params,
22 void *buf, size_t buf_size);
23
25 The io_uring_queue_init(3) function executes the io_uring_setup(2) sys‐
26 tem call to initialize the submission and completion queues in the ker‐
27 nel with at least entries entries in the submission queue and then maps
28 the resulting file descriptor to memory shared between the application
29 and the kernel.
30 By default, the CQ ring will have twice the number of entries as speci‐
32 fied by entries for the SQ ring. This is adequate for regular file or
33 storage workloads, but may be too small for networked workloads. The SQ
34 ring entries do not impose a limit on the number of in-flight requests
35 that the ring can support, it merely limits the number that can be sub‐
36 mitted to the kernel in one go (batch). if the CQ ring overflows, e.g.
37 more entries are generated than fits in the ring before the application
38 can reap them, then if the kernel supports IORING_FEAT_NODROP the ring
39 enters a CQ ring overflow state. Otherwise it drops the CQEs and incre‐
40 ments cq.koverflow in stuct io_uring with the number of CQEs dropped.
41 The overflow state is indicated by IORING_SQ_CQ_OVERFLOW being set in
42 the SQ ring flags. Unless the kernel runs out of available memory, en‐
43 tries are not dropped, but it is a much slower completion path and will
44 slow down request processing. For that reason it should be avoided and
45 the CQ ring sized appropriately for the workload. Setting cq_entries in
46 struct io_uring_params will tell the kernel to allocate this many en‐
47 tries for the CQ ring, independent of the SQ ring size in given in en‐
48 tries. If the value isn't a power of 2, it will be rounded up to the
49 nearest power of 2.
50 On success, io_uring_queue_init(3) returns 0 and ring will point to the
52 shared memory containing the io_uring queues. On failure -errno is re‐
53 turned.
54 flags will be passed through to the io_uring_setup syscall (see io_ur‐
56 ing_setup(2)).
57 The io_uring_queue_init_params(3) and io_uring_queue_init_mem(3) vari‐
59 ants will pass the parameters indicated by params straight through to
60 the io_uring_setup(2) system call.
61 The io_uring_queue_init_mem(3) variant uses the provided buf with asso‐
63 ciated size buf_size as the memory for the ring, using the IOR‐
64 ING_SETUP_NO_MMAP flag to io_uring_setup(2). The buffer passed to
65 io_uring_queue_init_mem(3) must already be zeroed. Typically, the
66 caller should allocate a huge page and pass that in to io_ur‐
67 ing_queue_init_mem(3). Pages allocated by mmap are already zeroed.
68 io_uring_queue_init_mem(3) returns the number of bytes used from the
69 provided buffer, so that the app can reuse the buffer with the returned
70 offset to put more rings in the same huge page.
71 On success, the resources held by ring should be released via a corre‐
73 sponding call to io_uring_queue_exit(3).
74
76 io_uring_queue_init(3) and io_uring_queue_init_params(3) return 0 on
77 success and -errno on failure.
78 io_uring_queue_init_mem(3) returns the number of bytes used from the
80 provided buffer on success, and -errno on failure.
81
83 io_uring_setup(2), io_uring_register_ring_fd(3), mmap(2), io_ur‐
84 ing_queue_exit(3)
85liburing-0.7 July 10, 2020 io_uring_queue_init(3)