io_uring: add napi busy polling support
Subsystems (3)
v7 cover io_uring: add napi busy polling support
This adds the napi busy polling support in io_uring.c. It adds a new napi_list to the io_ring_ctx structure. This list contains the list of napi_id's that are currently enabled for busy polling. This list is used to determine which napi id's enabled busy polling. For faster access it also adds a hash table.
When a new napi id is added, the hash table is used to locate if the napi id has already been added. When processing the busy poll loop the list is used to process the individual elements.
io-uring allows specifying two parameters:
- busy poll timeout and
- prefer busy poll to call of io_napi_busy_loop()
This sets the above parameters for the ring. The settings are passed with a new structure io_uring_napi.
There is also a corresponding liburing patch series, which enables this feature. The name of the series is "liburing: add add api for napi busy poll timeout". It also contains two programs to test the this.
Testing has shown that the round-trip times are reduced to 38us from 55us by enabling napi busy polling with a busy poll timeout of 100us. More detailled results are part of the commit message of the first patch.
Changes:
- V7:
- allow unregister with NULL value for arg parameter
- return -EOPNOTSUPP if CONFIG_NET_RX_BUSY_POLL is not enabled
- V6:
- Add a hash table on top of the list for faster access during the add operation. The linked list and the hash table use the same data structure
- V5:
- Refreshed to 6.1-rc6
- Use copy_from_user instead of memdup/kfree
- Removed the moving of napi_busy_poll_to
- Return -EINVAL if any of the reserved or padded fields are not 0.
- V4:
- Pass structure for napi config, instead of individual parameters
- V3:
- Refreshed to 6.1-rc5
- Added a new io-uring api for the prefer napi busy poll api and wire it to io_napi_busy_loop().
- Removed the unregister (implemented as register)
- Added more performance results to the first commit message.
- V2:
- Add missing defines if CONFIG_NET_RX_BUSY_POLL is not defined
- Changes signature of function io_napi_add_list to static inline if CONFIG_NET_RX_BUSY_POLL is not defined
- define some functions as static
Stefan Roesch (3):
io_uring: add napi busy polling support
io_uring: add api to set / get napi configuration.
io_uring: add api to set napi prefer busy poll
include/linux/io_uring_types.h | 10 ++ include/uapi/linux/io_uring.h | 12 ++ io_uring/io_uring.c | 297 +++++++++++++++++++++++++++++++++ io_uring/napi.h | 23 +++ io_uring/poll.c | 2 + io_uring/sqpoll.c | 17 ++ 6 files changed, 361 insertions(+) create mode 100644 io_uring/napi.h
base-commit: c0b67534c95c537f7a506a06b98e5e85d72e2b7d
Signature · 1 line
2.30.2
PATCH v7 1/3 io_uring: add napi busy polling support
This adds the napi busy polling support in io_uring.c. It adds a new napi_list to the io_ring_ctx structure. This list contains the list of napi_id's that are currently enabled for busy polling. The list is synchronized by the new napi_lock spin lock. The current default napi busy polling time is stored in napi_busy_poll_to. If napi busy polling is not enabled, the value is 0.
In addition there is also a hash table. The hash table store the napi id ond the pointer to the above list nodes. The hash table is used to speed up the lookup to the list elements.
The NAPI_TIMEOUT is stored as a timeout to make sure that the time a napi entry is stored in the napi list is limited.
The busy poll timeout is also stored as part of the io_wait_queue. This is necessary as for sq polling the poll interval needs to be adjusted and the napi callback allows only to pass in one value.
This has been tested with two simple programs from the liburing library
repository: the napi client and the napi server program. The client sends a request, which has a timestamp in its payload and the server replies with the same payload. The client calculates the roundtrip time and stores it to calcualte the results.
The client is running on host1 and the server is running on host 2 (in the same rack). The measured times below are roundtrip times. They are average times over 5 runs each. Each run measures 1 million roundtrips.
no rx coal rx coal: frames=88,usecs=33
Default 57us 56us
client_poll=100us 47us 46us
server_poll=100us 51us 46us
client_poll=100us+ 40us 40us
server_poll=100us
client_poll=100us+ 41us 39us
server_poll=100us+
prefer napi busy poll on client
client_poll=100us+ 41us 39us
server_poll=100us+
prefer napi busy poll on server
client_poll=100us+ 41us 39us
server_poll=100us+
prefer napi busy poll on client + server
--- include/linux/io_uring_types.h | 10 ++ io_uring/io_uring.c | 241 +++++++++++++++++++++++++++++++++ io_uring/napi.h | 23 ++++ io_uring/poll.c | 2 + io_uring/sqpoll.c | 17 +++ 5 files changed, 293 insertions(+) create mode 100644 io_uring/napi.h
include/linux/io_uring_types.h +10 -0
@@ -2,6 +2,7 @@#define IO_URING_TYPES_H #include <linux/blkdev.h> +#include <linux/hashtable.h> #include <linux/task_work.h> #include <linux/bitmap.h> #include <linux/llist.h>
@@ -274,6 +275,15 @@ struct io_ring_ctx {struct xarray personalities;u32 pers_next; +#ifdef CONFIG_NET_RX_BUSY_POLL +struct list_head napi_list;/* track busy poll napi_id */ +DECLARE_HASHTABLE (napi_ht,8 ); +spinlock_t napi_lock;/* napi_list lock */ + +unsigned int napi_busy_poll_to;/* napi busy poll default timeout */ +bool napi_prefer_busy_poll; +#endif +struct {/* * We cache a range of free CQEs we can use, once exhausted it
io_uring/io_uring.c +241 -0
295 lines not on this page · raw message
io_uring/napi.h new file +23 -0
@@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: GPL-2.0 */ + +#ifndef IOU_NAPI_H +#define IOU_NAPI_H + +#include <linux/kernel.h> +#include <linux/io_uring.h> +#include <net/busy_poll.h> + +#ifdef CONFIG_NET_RX_BUSY_POLL + +void io_napi_add (struct file *file,struct io_ring_ctx *ctx); +bool io_napi_busy_loop (struct list_head *napi_list,bool prefer_busy_poll); +void io_napi_merge_lists (struct io_ring_ctx *ctx,struct list_head *napi_list); + +#else + +static inline void io_napi_add (struct file *file,struct io_ring_ctx *ctx) +{ +} + +#endif +#endif
io_uring/poll.c +2 -0
@@ -15,6 +15,7 @@#include "io_uring.h" #include "refs.h" +#include "napi.h" #include "opdef.h" #include "kbuf.h" #include "poll.h"
@@ -629,6 +630,7 @@ static int __io_arm_poll_handler(struct io_kiocb *req,__io_poll_execute (req, mask);return 0 ; } +io_napi_add (req->file, req->ctx);if (ipt->owning) {/*
io_uring/sqpoll.c +17 -0
@@ -15,6 +15,7 @@#include <uapi/linux/io_uring.h> #include "io_uring.h" +#include "napi.h" #include "sqpoll.h" #define IORING_SQPOLL_CAP_ENTRIES_VALUE 8
@@ -168,6 +169,9 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) {unsigned int to_submit;int ret =0 ; +#ifdef CONFIG_NET_RX_BUSY_POLL +LIST_HEAD (local_napi_list); +#endif to_submit =io_sqring_entries (ctx);/* if we're handling multiple rings, cap submit size for fairness */
@@ -193,6 +197,19 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) ret =io_submit_sqes (ctx, to_submit);mutex_unlock (&ctx->uring_lock); +#ifdef CONFIG_NET_RX_BUSY_POLL +spin_lock (&ctx->napi_lock); +list_splice_init (&ctx->napi_list, &local_napi_list); +spin_unlock (&ctx->napi_lock); + +if (!list_empty (&local_napi_list) && +READ_ONCE (ctx->napi_busy_poll_to) >0 && +io_napi_busy_loop (&local_napi_list, ctx->napi_prefer_busy_poll)) { +io_napi_merge_lists (ctx, &local_napi_list); + ++ret; + } +#endif +if (to_submit &&wq_has_sleeper (&ctx->sqo_sq_wait))wake_up (&ctx->sqo_sq_wait);if (creds)
Signature · 1 line
2.30.2
Jens Axboe
maintainerQuoted 14 lines
include/linux/io_uring_types.h +1
-0@@ -2,6 +2,7 @@#define IO_URING_TYPES_H #include <linux/blkdev.h> +#include <linux/hashtable.h> #include <linux/task_work.h> #include <linux/bitmap.h> #include <linux/llist.h> @@ -274,6 +275,15 @@ struct io_ring_ctx {struct xarray personalities;
include/linux/io_uring_types.h struct io_ring_ctx { · line 285
Quoted 10 lines
include/linux/io_uring_types.h +8
-0u32 pers_next; +#ifdef CONFIG_NET_RX_BUSY_POLL +struct list_head napi_list;/* track busy poll napi_id */ +DECLARE_HASHTABLE (napi_ht,8 ); +spinlock_t napi_lock;/* napi_list lock */ + +unsigned int napi_busy_poll_to;/* napi busy poll default timeout */ +bool napi_prefer_busy_poll; +#endif
Minor thing, but I wonder if we should put this in a struct and allocate it if NAPI gets used rather than bloat the whole thing here. This doubles the size of io_ring_ctx, the hash above is 2k in size!
Quoted 13 lines
io_uring/io_uring.c +1
-0@@ -90,6 +90,7 @@#include "rsrc.h" #include "cancel.h" #include "net.h" +#include "napi.h" #include "notif.h" #include "timeout.h"
io_uring/io_uring.c static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p) · line 345
Quoted 10 lines
io_uring/io_uring.c +7
-0@@ -335,6 +336,14 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)INIT_WQ_LIST (&ctx->locked_free_list);INIT_DELAYED_WORK (&ctx->fallback_work, io_fallback_req_func);INIT_WQ_LIST (&ctx->submit_state.compl_reqs); + +#ifdef CONFIG_NET_RX_BUSY_POLL +INIT_LIST_HEAD (&ctx->napi_list); +spin_lock_init (&ctx->napi_lock); + ctx->napi_prefer_busy_poll = false; + ctx->napi_busy_poll_to =READ_ONCE (sysctl_net_busy_poll); +#endif
I think that should go in a io_napi_init() function, so we can get rid of these ifdefs in the main code.
Quoted 6 lines
io_uring/io_uring.c +1
-0static inline bool io_has_work (struct io_ring_ctx *ctx)@@ -2498,6 +2512,196 @@ static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,return ret <0 ? ret :1 ; } +#ifdef CONFIG_NET_RX_BUSY_POLL
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2525
Quoted 10 lines
io_uring/io_uring.c +10
-0+#define NAPI_TIMEOUT (60 * SEC_CONVERSION) + +struct io_napi_ht_entry { +unsigned int napi_id; +struct list_head list; + +/* Covered by napi lock spinlock. */ +unsigned long timeout; +struct hlist_node node; +};
Not strictly related to just this, but I think it'd be a good idea to add a napi.c file and put it all in there rather than in the core io_uring code. It really doesn't belong there.
Quoted 6 lines
io_uring/io_uring.c +6
-0+/* +* io_napi_add() - Add napi id to the busy poll list +* @file: file pointer for socket +* @ctx: io-uring context +* +* Add the napi id of the socket to the napi busy poll list and hash table.
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2547
Quoted 10 lines
io_uring/io_uring.c +10
-0+*/ +void io_napi_add (struct file *file,struct io_ring_ctx *ctx) +{ +unsigned int napi_id; +struct socket *sock; +struct sock *sk; +struct io_napi_ht_entry *he; + +if (!io_napi_busy_loop_on (ctx)) +return ;
I think io_napi_add() belongs in napi.h and should look ala:
static inline void io_napi_add(struct io_kiocb *req)
{
struct io_ring_ctx *ctx = req->ctx;
if (!io_napi_busy_loop_on(ctx))
return;
__io_napi_add(ctx, req->file);
}and put __io_napi_add() in napi.c
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2596
Quoted 12 lines
io_uring/io_uring.c +12
-0+static void io_napi_free_list (struct io_ring_ctx *ctx) +{ +unsigned int i; +struct io_napi_ht_entry *he; +LIST_HEAD (napi_list); + +spin_lock (&ctx->napi_lock); +hash_for_each (ctx->napi_ht, i, he, node) { +hash_del (&he->node); + } +spin_unlock (&ctx->napi_lock); +}
No need for the braces here for the loop.
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2666
Quoted 6 lines
io_uring/io_uring.c +6
-0+static void io_napi_blocking_busy_loop (struct list_head *napi_list, +struct io_wait_queue *iowq) +{ +unsigned long start_time =list_is_singular (napi_list) + ?0 + :busy_loop_current_time ();
No ternaries please. This is so much easier to read as:
unsigned long start_time = 0; if (!list_is_singular(napi_list)) start_time = busy_loop_current_time();
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2680
Quoted 13 lines
io_uring/io_uring.c +13
-0+do { +if (list_is_singular (napi_list)) { +struct io_napi_ht_entry *ne = +list_first_entry (napi_list, + struct io_napi_ht_entry, list); + +napi_busy_loop (ne->napi_id, io_napi_busy_loop_end, iowq, + iowq->napi_prefer_busy_poll,BUSY_POLL_BUDGET ); +break ; + } + }while (io_napi_busy_loop (napi_list, iowq->napi_prefer_busy_poll) && + !io_napi_busy_loop_end (iowq, start_time)); +}
This is almost impossible to read, please rewrite that in a way so that it's straight forward to understand what is going on.
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2702
Quoted 7 lines
io_uring/io_uring.c +7
-0+void io_napi_merge_lists (struct io_ring_ctx *ctx,struct list_head *napi_list) +{ +spin_lock (&ctx->napi_lock); +list_splice (napi_list, &ctx->napi_list); +io_napi_remove_stale (ctx); +spin_unlock (&ctx->napi_lock); +}
Question on the locking - the separate lock is obviously functionally correct, but at least for the arming part, we generally already have the ctx uring_lock at that point. Did you look into if it's feasible to take advantage of that? I
Quoted 33 lines
io_uring/io_uring.c +17
-0@@ -2510,6 +2714,9 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,struct io_rings *rings = ctx->rings;ktime_t timeout =KTIME_MAX ;int ret; +#ifdef CONFIG_NET_RX_BUSY_POLL +LIST_HEAD (local_napi_list); +#endif if (!io_allowed_run_tw (ctx))return -EEXIST ;@@ -2539,12 +2746,34 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,return ret; } +#ifdef CONFIG_NET_RX_BUSY_POLL + iowq.napi_busy_poll_to =0 ; + iowq.napi_prefer_busy_poll =READ_ONCE (ctx->napi_prefer_busy_poll); + +if (!(ctx->flags &IORING_SETUP_SQPOLL )) { +spin_lock (&ctx->napi_lock); +list_splice_init (&ctx->napi_list, &local_napi_list); +spin_unlock (&ctx->napi_lock); + } +#endif +if (uts) {struct timespec64 ts;if (get_timespec64 (&ts, uts))return -EFAULT ; + +#ifdef CONFIG_NET_RX_BUSY_POLL +if (!list_empty (&local_napi_list)) {
io_uring/io_uring.c static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, · line 2777
Quoted 10 lines
io_uring/io_uring.c +8
-0+io_napi_adjust_busy_loop_timeout (READ_ONCE (ctx->napi_busy_poll_to), + &ts, &iowq.napi_busy_poll_to); + } +#endif timeout =ktime_add_ns (timespec64_to_ktime (ts),ktime_get_ns ()); +#ifdef CONFIG_NET_RX_BUSY_POLL + }else if (!list_empty (&local_napi_list)) { + iowq.napi_busy_poll_to =READ_ONCE (ctx->napi_busy_poll_to); +#endif }
This is again a lot of ifdefs, please consider ways of getting rid of them.
io_uring/io_uring.c static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, · line 2795
Quoted 13 lines
io_uring/io_uring.c +9
-0@@ -2555,6 +2784,15 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, iowq.cq_tail =READ_ONCE (ctx->rings->cq.head) + min_events;trace_io_uring_cqring_wait (ctx, min_events); + +#ifdef CONFIG_NET_RX_BUSY_POLL +if (iowq.napi_busy_poll_to) +io_napi_blocking_busy_loop (&local_napi_list, &iowq); + +if (!list_empty (&local_napi_list)) +io_napi_merge_lists (ctx, &local_napi_list); +#endif +
And here.
io_uring/io_uring.c static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx) · line 2997
Quoted 10 lines
io_uring/io_uring.c +3
-0do { if (test_bit(IO_CHECK_CQ_OVERFLOW_BIT, &ctx->check_cq)) { finish_wait(&ctx->cq_wait, &iowq.wq);@@ -2754,6 +2992,9 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx)io_req_caches_free (ctx);if (ctx->hash_map)io_wq_put_hash (ctx->hash_map); +#ifdef CONFIG_NET_RX_BUSY_POLL +io_napi_free_list (ctx); +#endif
Put an io_napi_free_list() stub in napi.h with the actual thing in napi.c if CONFIG_NET_RX_BUSY_POLL is defined.
Quoted 32 lines
io_uring/sqpoll.c +10
-0@@ -15,6 +15,7 @@#include <uapi/linux/io_uring.h> #include "io_uring.h" +#include "napi.h" #include "sqpoll.h" #define IORING_SQPOLL_CAP_ENTRIES_VALUE 8@@ -168,6 +169,9 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) {unsigned int to_submit;int ret =0 ; +#ifdef CONFIG_NET_RX_BUSY_POLL +LIST_HEAD (local_napi_list); +#endif to_submit =io_sqring_entries (ctx);/* if we're handling multiple rings, cap submit size for fairness */ @@ -193,6 +197,19 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) ret =io_submit_sqes (ctx, to_submit);mutex_unlock (&ctx->uring_lock); +#ifdef CONFIG_NET_RX_BUSY_POLL +spin_lock (&ctx->napi_lock); +list_splice_init (&ctx->napi_list, &local_napi_list); +spin_unlock (&ctx->napi_lock); + + if (!list_empty(&local_napi_list) &&
io_uring/sqpoll.c static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) · line 215
Quoted 10 lines
io_uring/sqpoll.c +7
-0+ READ_ONCE(ctx->napi_busy_poll_to) > 0 && + io_napi_busy_loop(&local_napi_list, ctx->napi_prefer_busy_poll)) { +io_napi_merge_lists (ctx, &local_napi_list); + ++ret; + } +#endif +if (to_submit &&wq_has_sleeper (&ctx->sqo_sq_wait))wake_up (&ctx->sqo_sq_wait);if (creds)
Add a helper here too please.
Signature · 1 line
Jens Axboe
Quoted 16 lines · Jens Axboe
Quoted 16 lines
include/linux/io_uring_types.h +1
-0@@ -2,6 +2,7 @@#define IO_URING_TYPES_H #include <linux/blkdev.h> +#include <linux/hashtable.h> #include <linux/task_work.h> #include <linux/bitmap.h> #include <linux/llist.h> @@ -274,6 +275,15 @@ struct io_ring_ctx {struct xarray personalities;u32 pers_next;
Quoted 12 lines
include/linux/io_uring_types.h struct io_ring_ctx { · line 285
Quoted 8 lines
include/linux/io_uring_types.h +8
-0+#ifdef CONFIG_NET_RX_BUSY_POLL +struct list_head napi_list;/* track busy poll napi_id */ +DECLARE_HASHTABLE (napi_ht,8 ); +spinlock_t napi_lock;/* napi_list lock */ + +unsigned int napi_busy_poll_to;/* napi busy poll default timeout */ +bool napi_prefer_busy_poll; +#endif Minor thing, but I wonder if we should put this in a struct and allocate it if NAPI gets used rather than bloat the whole thing here. This doubles the size of
io_ring_ctx, the hash above is 2k in size!
I changed the hash table size to 16, so the size should no longer be a concern. The hash table was sized too big, its limited by the number of nic queues.
Quoted 11 lines
Quoted 11 lines
io_uring/io_uring.c +1
-0@@ -90,6 +90,7 @@#include "rsrc.h" #include "cancel.h" #include "net.h" +#include "napi.h" #include "notif.h"
Quoted 16 lines
io_uring/io_uring.c static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p) · line 345
Quoted 12 lines
io_uring/io_uring.c +7
-0#include "timeout.h" @@ -335,6 +336,14 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)INIT_WQ_LIST (&ctx->locked_free_list);INIT_DELAYED_WORK (&ctx->fallback_work, io_fallback_req_func);INIT_WQ_LIST (&ctx->submit_state.compl_reqs); + +#ifdef CONFIG_NET_RX_BUSY_POLL +INIT_LIST_HEAD (&ctx->napi_list); +spin_lock_init (&ctx->napi_lock); + ctx->napi_prefer_busy_poll = false; + ctx->napi_busy_poll_to =READ_ONCE (sysctl_net_busy_poll); +#endif I think that should go in a
io_napi_init()function, so we can get rid of these ifdefs in the main code.
The next version will add io_napi_init and io_napi_free.
Quoted 5 lines
Quoted 5 lines
io_uring/io_uring.c +0
-0static inline bool io_has_work (struct io_ring_ctx *ctx)@@ -2498,6 +2512,196 @@ static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,return ret <0 ? ret :1 ; }
Quoted 16 lines
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2525
Quoted 11 lines
io_uring/io_uring.c +11
-0+#ifdef CONFIG_NET_RX_BUSY_POLL +#define NAPI_TIMEOUT (60 * SEC_CONVERSION) + +struct io_napi_ht_entry { +unsigned int napi_id; +struct list_head list; + +/* Covered by napi lock spinlock. */ +unsigned long timeout; +struct hlist_node node; +};Not strictly related to just this, but I think it'd be a good idea to add a
napi.cfile and put it all in there rather than in the coreio_uringcode. It really doesn't belong there.
The next version adds a napi.c file and all napi functions are moved there.
Quoted 29 lines
Quoted 6 lines
io_uring/io_uring.c +6
-0+/* +* io_napi_add() - Add napi id to the busy poll list +* @file: file pointer for socket +* @ctx: io-uring context +* +* Add the napi id of the socket to the napi busy poll list and hash table. io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2547
Quoted 10 lines
io_uring/io_uring.c +10
-0+*/ +void io_napi_add (struct file *file,struct io_ring_ctx *ctx) +{ +unsigned int napi_id; +struct socket *sock; +struct sock *sk; +struct io_napi_ht_entry *he; + +if (!io_napi_busy_loop_on (ctx)) +return ;I think
io_napi_add()belongs innapi.hand should look ala:static inline void
io_napi_add(struct io_kiocb *req){ struct io_ring_ctx *ctx = req->ctx; if (!io_napi_busy_loop_on(ctx)) return; __io_napi_add(ctx, req->file); }
Quoted 2 lines
and put
__io_napi_add()innapi.c
I added the above function.
Quoted 6 lines
Quoted 6 lines
io_uring/io_uring.c +6
-0+static void io_napi_free_list (struct io_ring_ctx *ctx) +{ +unsigned int i; +struct io_napi_ht_entry *he; +LIST_HEAD (napi_list); +
Quoted 9 lines
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2596
Quoted 6 lines
io_uring/io_uring.c +6
-0+spin_lock (&ctx->napi_lock); +hash_for_each (ctx->napi_ht, i, he, node) { +hash_del (&he->node); + } +spin_unlock (&ctx->napi_lock); +}No need for the braces here for the loop.
Fixed in the next version.
Quoted 9 lines
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2666
Quoted 6 lines
io_uring/io_uring.c +6
-0+static void io_napi_blocking_busy_loop (struct list_head *napi_list, +struct io_wait_queue *iowq) +{ +unsigned long start_time =list_is_singular (napi_list) + ?0 + :busy_loop_current_time ();No ternaries please. This is so much easier to read as:
Quoted 5 lines
unsigned long start_time = 0; if (!list_is_singular(napi_list)) start_time = busy_loop_current_time();
Fixed in the next version.
Quoted 8 lines
Quoted 8 lines
io_uring/io_uring.c +8
-0+ do { + if (list_is_singular(napi_list)) { + struct io_napi_ht_entry *ne = + list_first_entry(napi_list, + struct io_napi_ht_entry, list); + + napi_busy_loop(ne->napi_id, io_napi_busy_loop_end, iowq, + iowq->napi_prefer_busy_poll, BUSY_POLL_BUDGET);
Quoted 9 lines
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2680
Quoted 5 lines
io_uring/io_uring.c +5
-0+break ; + } + }while (io_napi_busy_loop (napi_list, iowq->napi_prefer_busy_poll) && + !io_napi_busy_loop_end (iowq, start_time)); +}This is almost impossible to read, please rewrite that in a way so that it's straight forward to understand what is going on.
I rewrote the function in the next version of the patch series.
Quoted 4 lines
Quoted 4 lines
io_uring/io_uring.c +4
-0+void io_napi_merge_lists (struct io_ring_ctx *ctx,struct list_head *napi_list) +{ +spin_lock (&ctx->napi_lock); +list_splice (napi_list, &ctx->napi_list);
Quoted 9 lines
io_uring/io_uring.c static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx, · line 2702
Quoted 3 lines
io_uring/io_uring.c +3
-0+io_napi_remove_stale (ctx); +spin_unlock (&ctx->napi_lock); +}Question on the locking - the separate lock is obviously functionally correct, but at least for the arming part, we generally already have the ctx
uring_lockat that point. Did you look into if it's feasible to take advantage of that? I
Its not guaranteed to have the lock in all code paths when I checked.
Quoted 38 lines
Quoted 38 lines
io_uring/io_uring.c +21
-0@@ -2510,6 +2714,9 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,struct io_rings *rings = ctx->rings;ktime_t timeout =KTIME_MAX ;int ret; +#ifdef CONFIG_NET_RX_BUSY_POLL +LIST_HEAD (local_napi_list); +#endif if (!io_allowed_run_tw (ctx))return -EEXIST ;@@ -2539,12 +2746,34 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,return ret; } +#ifdef CONFIG_NET_RX_BUSY_POLL + iowq.napi_busy_poll_to =0 ; + iowq.napi_prefer_busy_poll =READ_ONCE (ctx->napi_prefer_busy_poll); + +if (!(ctx->flags &IORING_SETUP_SQPOLL )) { +spin_lock (&ctx->napi_lock); +list_splice_init (&ctx->napi_list, &local_napi_list); +spin_unlock (&ctx->napi_lock); + } +#endif +if (uts) {struct timespec64 ts;if (get_timespec64 (&ts, uts))return -EFAULT ; + +#ifdef CONFIG_NET_RX_BUSY_POLL +if (!list_empty (&local_napi_list)) { +io_napi_adjust_busy_loop_timeout (READ_ONCE (ctx->napi_busy_poll_to), + &ts, &iowq.napi_busy_poll_to); + } +#endif timeout =ktime_add_ns (timespec64_to_ktime (ts),ktime_get_ns ());
Quoted 9 lines
io_uring/io_uring.c static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, · line 2777
Quoted 5 lines
io_uring/io_uring.c +4
-0+#ifdef CONFIG_NET_RX_BUSY_POLL + }else if (!list_empty (&local_napi_list)) { + iowq.napi_busy_poll_to =READ_ONCE (ctx->napi_busy_poll_to); +#endif }This is again a lot of ifdefs, please consider ways of getting rid of them.
I added helper functions to eliminate the ifdefs.
Quoted 16 lines
io_uring/io_uring.c static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, · line 2795
Quoted 13 lines
io_uring/io_uring.c +9
-0@@ -2555,6 +2784,15 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, iowq.cq_tail =READ_ONCE (ctx->rings->cq.head) + min_events;trace_io_uring_cqring_wait (ctx, min_events); + +#ifdef CONFIG_NET_RX_BUSY_POLL +if (iowq.napi_busy_poll_to) +io_napi_blocking_busy_loop (&local_napi_list, &iowq); + +if (!list_empty (&local_napi_list)) +io_napi_merge_lists (ctx, &local_napi_list); +#endif +And here.
Quoted 14 lines
io_uring/io_uring.c static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx) · line 2997
Quoted 10 lines
io_uring/io_uring.c +3
-0do { if (test_bit(IO_CHECK_CQ_OVERFLOW_BIT, &ctx->check_cq)) { finish_wait(&ctx->cq_wait, &iowq.wq);@@ -2754,6 +2992,9 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx)io_req_caches_free (ctx);if (ctx->hash_map)io_wq_put_hash (ctx->hash_map); +#ifdef CONFIG_NET_RX_BUSY_POLL +io_napi_free_list (ctx); +#endif Put an
io_napi_free_list()stub innapi.hwith the actual thing innapi.cifCONFIG_NET_RX_BUSY_POLLis defined.
I added io_napi_free as a helper function.
Quoted 36 lines
Quoted 36 lines
io_uring/sqpoll.c +14
-0@@ -15,6 +15,7 @@#include <uapi/linux/io_uring.h> #include "io_uring.h" +#include "napi.h" #include "sqpoll.h" #define IORING_SQPOLL_CAP_ENTRIES_VALUE 8@@ -168,6 +169,9 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) {unsigned int to_submit;int ret =0 ; +#ifdef CONFIG_NET_RX_BUSY_POLL +LIST_HEAD (local_napi_list); +#endif to_submit =io_sqring_entries (ctx);/* if we're handling multiple rings, cap submit size for fairness */ @@ -193,6 +197,19 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) ret =io_submit_sqes (ctx, to_submit);mutex_unlock (&ctx->uring_lock); +#ifdef CONFIG_NET_RX_BUSY_POLL +spin_lock (&ctx->napi_lock); +list_splice_init (&ctx->napi_list, &local_napi_list); +spin_unlock (&ctx->napi_lock); + +if (!list_empty (&local_napi_list) && +READ_ONCE (ctx->napi_busy_poll_to) >0 && +io_napi_busy_loop (&local_napi_list, ctx->napi_prefer_busy_poll)) { +io_napi_merge_lists (ctx, &local_napi_list); + ++ret;
Quoted 8 lines
io_uring/sqpoll.c static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) · line 215
Quoted 6 lines
io_uring/sqpoll.c +3
-0+ } +#endif +if (to_submit &&wq_has_sleeper (&ctx->sqo_sq_wait))wake_up (&ctx->sqo_sq_wait);if (creds)Add a helper here too please.
I also added a helper function for the sqpoll busy poll.
PATCH v7 2/3 io_uring: add api to set / get napi configuration.
In reply to Stefan RoeschThis adds an api to register the busy poll timeout from liburing. To be able to use this functionality, the corresponding liburing patch is needed.
--- include/uapi/linux/io_uring.h | 11 ++++++++ io_uring/io_uring.c | 52 +++++++++++++++++++++++++++++++++++ 2 files changed, 63 insertions(+)
include/uapi/linux/io_uring.h +11 -0
@@ -516,6 +516,10 @@ enum {/* register a range of fixed file slots for automatic slot allocation */ IORING_REGISTER_FILE_ALLOC_RANGE =25 , +/* set/clear busy poll settings */ +IORING_REGISTER_NAPI =26 , +IORING_UNREGISTER_NAPI =27 , +/* this goes last */ IORING_REGISTER_LAST };
@@ -638,6 +642,13 @@ struct io_uring_buf_reg {__u64 resv[3 ]; }; +/* argument for IORING_(UN)REGISTER_NAPI */ +struct io_uring_napi { +__u32 busy_poll_to; +__u32 pad; +__u64 resv; +}; +/* * io_uring_restriction->opcode values */
io_uring/io_uring.c +52 -0
64 lines not on this page · raw message
Signature · 1 line
2.30.2
Re: [PATCH v7 2/3] io_uring: add api to set / get napi configuration.
Quoted 5 lines · Stefan Roesch
This adds an api to register the busy poll timeout from liburing. To be able to use this functionality, the corresponding liburing patch is needed.
Signed-off-by Stefan Roesch <[email protected]>Acked-by Jakub Kicinski <[email protected]>
Signature · 1 line
Ammar Faizi
PATCH v7 3/3 io_uring: add api to set napi prefer busy poll
In reply to Stefan RoeschThis adds an api to register and unregister the napi prefer busy poll setting from liburing. To be able to use this functionality, the corresponding liburing patch is needed.
--- include/uapi/linux/io_uring.h | 3 ++- io_uring/io_uring.c | 6 +++++- 2 files changed, 7 insertions(+), 2 deletions(-)
include/uapi/linux/io_uring.h +2 -1
@@ -645,7 +645,8 @@ struct io_uring_buf_reg {/* argument for IORING_(UN)REGISTER_NAPI */ struct io_uring_napi {__u32 busy_poll_to; -__u32 pad; +__u8 prefer_busy_poll; +__u8 pad[3 ];__u64 resv; };
io_uring/io_uring.c +5 -1
@@ -4251,15 +4251,17 @@ static int io_register_napi(struct io_ring_ctx *ctx, void __user *arg)#ifdef CONFIG_NET_RX_BUSY_POLL const struct io_uring_napi curr = { .busy_poll_to = ctx->napi_busy_poll_to, + .prefer_busy_poll = ctx->napi_prefer_busy_poll };struct io_uring_napi napi;if (copy_from_user (&napi, arg,sizeof (napi)))return -EFAULT ; -if (napi.pad || napi.resv) +if (napi.pad[0 ] || napi.pad[1 ] || napi.pad[2 ] || napi.resv)return -EINVAL ;WRITE_ONCE (ctx->napi_busy_poll_to, napi.busy_poll_to); +WRITE_ONCE (ctx->napi_prefer_busy_poll, !!napi.prefer_busy_poll);if (copy_to_user (arg, &curr,sizeof (curr)))return -EFAULT ;
@@ -4275,6 +4277,7 @@ static int io_unregister_napi(struct io_ring_ctx *ctx, void __user *arg)#ifdef CONFIG_NET_RX_BUSY_POLL const struct io_uring_napi curr = { .busy_poll_to = ctx->napi_busy_poll_to, + .prefer_busy_poll = ctx->napi_prefer_busy_poll };if (arg) {
@@ -4283,6 +4286,7 @@ static int io_unregister_napi(struct io_ring_ctx *ctx, void __user *arg) }WRITE_ONCE (ctx->napi_busy_poll_to,0 ); +WRITE_ONCE (ctx->napi_prefer_busy_poll, false);return 0 ;#else return -EOPNOTSUPP ;
Signature · 1 line
2.30.2