io_uring: add set of tracing events
Subsystems (4)
RFC v3 1/1 io_uring: add set of tracing events
To trace io_uring activity one can get an information from workqueue and io trace events, but looks like some parts could be hard to identify via this approach. Making what happens inside io_uring more transparent is important to be able to reason about many aspects of it, hence introduce the set of tracing events.
All such events could be roughly divided into two categories:
* those, that are helping to understand correctness (from both kernel and an application point of view). E.g. a ring creation, file registration, or waiting for available CQE. Proposed approach is to get a pointer to an original structure of interest (ring context, or request), and then find relevant events. io_uring_queue_async_work also exposes a pointer to work_struct, to be able to track down corresponding workqueue events.
* those, that provide performance related information. Mostly it's about events that change the flow of requests, e.g. whether an async work was queued, or delayed due to some dependencies. Another important case is how io_uring optimizations (e.g. registered files) are utilized.
---
v1: add io_uring_link trace event
v2: Extend io_uring events, to include not only io_uring_link, but other
events to cover important parts of the functionality.
v3: Remove io_uring_import_iovec event.
Expose not only a failed request, but also all the failed links in the failing link event.
Move io_uring_add_to_prev into io_add_to_prev_work() and fix a comment typo.
Remove condition and add ret to io_uring_register.
fs/io_uring.c | 17 ++ include/Kbuild | 1 + include/trace/events/io_uring.h | 349 ++++++++++++++++++++++++++++++++ 3 files changed, 367 insertions(+) create mode 100644 include/trace/events/io_uring.h
fs/io_uring.c +17 -0
@@ -71,6 +71,9 @@#include <linux/sizes.h> #include <linux/hugetlb.h> +#define CREATE_TRACE_POINTS +#include <trace/events/io_uring.h> +#include <uapi/linux/io_uring.h> #include "internal.h"
@@ -483,6 +486,7 @@ static inline void io_queue_async_work(struct io_ring_ctx *ctx, } } +trace_io_uring_queue_async_work (ctx, rw, req, &req->work, req->flags);queue_work (ctx->sqo_wq[rw], &req->work); }
@@ -711,6 +715,7 @@ static void io_fail_links(struct io_kiocb *req) link = list_first_entry(&req->link_list, struct io_kiocb, list);list_del (&link->list); +trace_io_uring_fail_link (req, link);io_cqring_add_event (req->ctx, link->user_data, -ECANCELED );__io_free_req (link); }
@@ -2021,6 +2026,7 @@ static int io_req_defer(struct io_ring_ctx *ctx, struct io_kiocb *req, req->submit.sqe = sqe_copy;INIT_WORK (&req->work, io_sq_wq_submit_work); +trace_io_uring_defer (ctx, req, false);list_add_tail (&req->list, &ctx->defer_list);spin_unlock_irq (&ctx->completion_lock);return -EIOCBQUEUED ;
@@ -2279,6 +2285,8 @@ static bool io_add_to_prev_work(struct async_list *list, struct io_kiocb *req) ret = false; }spin_unlock (&list->lock); + +trace_io_uring_add_to_prev (req, ret);return ret; }
@@ -2327,6 +2335,7 @@ static int io_req_set_file(struct io_ring_ctx *ctx, const struct sqe_submit *s, }else {if (s->needs_fixed_file)return -EBADF ; +trace_io_uring_file_get (ctx, fd); req->file =io_file_get (state, fd);if (unlikely (!req->file))return -EBADF ;
@@ -2430,6 +2439,7 @@ static int io_queue_link_head(struct io_ring_ctx *ctx, struct io_kiocb *req,/* Insert shadow req to defer_list, blocking next IOs */ spin_lock_irq (&ctx->completion_lock); +trace_io_uring_defer (ctx, shadow, true);list_add_tail (&shadow->list, &ctx->defer_list);spin_unlock_irq (&ctx->completion_lock);
@@ -2488,6 +2498,7 @@ static void io_submit_sqe(struct io_ring_ctx *ctx, struct sqe_submit *s, s->sqe = sqe_copy;memcpy (&req->submit, s,sizeof (*s)); +trace_io_uring_link (ctx, req, prev);list_add_tail (&req->list, &prev->link_list); }else if (s->sqe->flags &IOSQE_IO_LINK ) { req->flags |=REQ_F_LINK ;
@@ -2626,6 +2637,7 @@ static int io_submit_sqes(struct io_ring_ctx *ctx, struct sqe_submit *sqes, sqes[i].has_user = has_user; sqes[i].needs_lock = true; sqes[i].needs_fixed_file = true; +trace_io_uring_submit_sqe (ctx, true, true);io_submit_sqe (ctx, &sqes[i], statep, &link, true); submitted++; }
@@ -2824,6 +2836,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,if (block_for_last && submit == to_submit) force_nonblock = false; +trace_io_uring_submit_sqe (ctx, force_nonblock, false);io_submit_sqe (ctx, &s, statep, &link, force_nonblock); }io_commit_sqring (ctx);
@@ -2906,6 +2919,7 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, ret =0 ; iowq.nr_timeouts =atomic_read (&ctx->cq_timeouts); +trace_io_uring_cqring_wait (ctx, min_events); do { prepare_to_wait_exclusive(&ctx->wait, &iowq.wq, TASK_INTERRUPTIBLE);
@@ -4057,6 +4071,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p) p->cq_off.cqes = offsetof(struct io_rings , cqes); p->features =IORING_FEAT_SINGLE_MMAP ; +trace_io_uring_create (ret, ctx, p->sq_entries, p->cq_entries, p->flags);return ret; err:io_ring_ctx_wait_and_kill (ctx);
@@ -4194,6 +4209,8 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,mutex_lock (&ctx->uring_lock); ret =__io_uring_register (ctx, opcode, arg, nr_args);mutex_unlock (&ctx->uring_lock); +trace_io_uring_register (ctx, opcode, ctx->nr_user_files, ctx->nr_user_bufs, + ctx->cq_ev_fd !=NULL , ret); out_fput:fdput (f);return ret;
include/Kbuild +1 -0
@@ -1028,6 +1028,7 @@ header-test- += trace/events/fsi_master_gpio.h header-test- += trace/events/huge_memory.h header-test- += trace/events/ib_mad.h header-test- += trace/events/ib_umad.h +header-test- += trace/events/io_uring.h header-test- += trace/events/iscsi.h header-test- += trace/events/jbd2.h header-test- += trace/events/kvm.h
include/trace/events/io_uring.h new file +349 -0
349 lines not on this page · raw message
Signature · 1 line
2.21.0
Jens Axboe
maintainerQuoted 16 lines · Dmitrii Dolgov
To trace
io_uringactivity one can get an information from workqueue and io trace events, but looks like some parts could be hard to identify via this approach. Making what happens insideio_uringmore transparent is important to be able to reason about many aspects of it, hence introduce the set of tracing events.All such events could be roughly divided into two categories:
* those, that are helping to understand correctness (from both kernel and an application point of view). E.g. a ring creation, file registration, or waiting for available CQE. Proposed approach is to get a pointer to an original structure of interest (ring context, or request), and then find relevant events.
io_uring_queue_async_workalso exposes a pointer towork_struct, to be able to track down corresponding workqueue events.
Quoted 5 lines
* those, that provide performance related information. Mostly it's about events that change the flow of requests, e.g. whether an async work was queued, or delayed due to some dependencies. Another important case is how
io_uringoptimizations (e.g. registered files) are utilized.
Looks good to me, I've applied it for some testing. Thanks!
Signature · 1 line
Jens Axboe
Quoted 3 lines · Jens Axboe
Looks good to me, I've applied it for some testing. Thanks!
Thank you!
Jens Axboe
maintainerQuoted 5 lines · Dmitry Dolgov
Quoted 3 lines · Jens Axboe
Looks good to me, I've applied it for some testing. Thanks!
Thank you!
Already saved me some time in verifying some activity, so well worth it!
Signature · 1 line
Jens Axboe