src/queue: don't swallow -ETIME when SQEs were submitted
PATCH 1/2 src/queue: don't swallow -ETIME when SQEs were submitted
If _io_uring_get_cqe() submits SQEs and then times out waiting for completions, it returns the submit count instead of -ETIME:
1. The first enter submits the SQEs; because submit > 0 the kernel returns the submit count, not -ETIME, and it is stored in err.
2. On the next iteration the has_ts shortcut wants to report -ETIME, but the 'if (!err)' guard sees the non-zero submit count and keeps it, so -ETIME is dropped.
That contradicts io_uring_submit_and_wait_timeout(3) and io_uring_wait_cqes(3), which document -ETIME on timeout.
At these two sites (lines 113 and 118) err is only ever 0 or a positive submit count. A negative error from __io_uring_peek_cqe() or a prior enter breaks out of the loop before reaching here. So the change is functionally equivalent to dropping the err condition entirely; we change '!err' to 'err >= 0' so -ETIME is successfully synthesized whenever no CQE was seen.
The guards were added in 2f61e849 ("src/queue: don't wait twice if looping in _io_uring_get_cqe()") to carry the submit count across iterations for the partial-completion case (got some CQEs, no error); that case still returns the count because both sites remain guarded by !cqe.
--- src/queue.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
src/queue.c +2 -2
@@ -110,12 +110,12 @@ static int _io_uring_get_cqe(struct io_uring *ring,* timeout, so treat any timeout the same as -ETIME here. */ if (data->get_flags &IORING_ENTER_EXT_ARG_REG ) { -if (!cqe && !err) +if (!cqe && err >=0 ) err = -ETIME ; }else {struct io_uring_getevents_arg *arg = data->arg; -if (!cqe && arg->ts && !err) +if (!cqe && arg->ts && err >=0 ) err = -ETIME ; }break ;
Signature · 1 line
2.43.0
PATCH 2/2 test/timeout-swallow: verify -ETIME is not swallowed
Regression test for the previous commit. Submits an SQE and waits with a zero timeout for more completions than can arrive; the result must be -ETIME, not the positive submit count. Covers the normal EXT_ARG wait path and the registered-wait path, each skipped gracefully where unsupported.
--- test/Makefile | 1 + test/timeout-swallow.c | 117 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 118 insertions(+) create mode 100644 test/timeout-swallow.c
test/Makefile +1 -0
@@ -293,6 +293,7 @@ test_srcs := \ timerfd-short-read.c \ timeout.c \ timeout-new.c \ + timeout-swallow.c \ timestamp.c \ timestamp-bug.c \ truncate.c \
test/timeout-swallow.c new file +117 -0
117 lines not on this page · raw message
Signature · 1 line
2.43.0
Re: [PATCH 2/2] test/timeout-swallow: verify -ETIME is not swallowed
Quoted 35 lines · Prateek
Regression test for the previous commit. Submits an SQE and waits with a zero timeout for more completions than can arrive; the result must be
-ETIME, not the positive submit count. Covers the normalEXT_ARGwait path and the registered-wait path, each skipped gracefully where unsupported.Signed-off-by Prateek <[email protected]>--- test/Makefile | 1 + test/timeout-swallow.c | 117 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 118 insertions(+) create mode 100644 test/timeout-swallow.ctest/Makefile +1
-0@@ -293,6 +293,7 @@ test_srcs := \ timerfd-short-read.c \ timeout.c \ timeout-new.c \ + timeout-swallow.c \ timestamp.c \ timestamp-bug.c \ truncate.c \test/timeout-swallow.c new file +4
-0@@ -0,0 +1,117 @@ +/* SPDX-License-Identifier: MIT */ +/* +* Description: tests that io_uring_wait_cqes() and variants do not swallow +* -ETIME when loop-fetching CQEs if some SQEs were submitted.
test/timeout-swallow.c line 14
Quoted 10 lines
test/timeout-swallow.c new file +10
-0+*/ +#include <stdio.h> +#include <unistd.h> +#include <errno.h> +#include <sys/time.h> +#include "liburing.h" +#include "helpers.h" + +/* +* Test the normal -ETIME swallow path. (line 118 in queue.c)
This type of comment is useless as line number changes. Better to refer to function name in the API and describe the situation or not have anything at all. Other than that,
Signature · 1 line
Gabriel Krisman Bertazi
Quoted 19 lines · Prateek
If
_io_uring_get_cqe()submits SQEs and then times out waiting for completions, it returns the submit count instead of-ETIME:1. The first enter submits the SQEs; because submit > 0 the kernel returns the submit count, not
-ETIME, and it is stored in err.
2. On the next iteration thehas_tsshortcut wants to report-ETIME, but the 'if (!err)' guard sees the non-zero submit count and keeps it, so-ETIMEis dropped.That contradicts
io_uring_submit_and_wait_timeout(3)andio_uring_wait_cqes(3), which document-ETIMEon timeout.At these two sites (lines 113 and 118) err is only ever 0 or a positive submit count. A negative error from
__io_uring_peek_cqe()or a prior enter breaks out of the loop before reaching here. So the change is functionally equivalent to dropping the err condition entirely; we change '!err' to 'err >= 0' so-ETIMEis successfully synthesized whenever no CQE was seen.
Quoted 7 lines
The guards were added in
2f61e849("src/queue: don't wait twice if looping in_io_uring_get_cqe()") to carry the submit count across iterations for the partial-completion case (got some CQEs, no error); that case still returns the count because both sites remain guarded by !cqe.Signed-off-by Prateek <[email protected]>
Signature · 1 line
Gabriel Krisman Bertazi