syscall: Add io_uring related tests
PATCH v5 1/1 syscall: Add io_uring related tests
Added asynchronous I/O API tests for io_uring_setup(), io_uring_register() and io_uring_enter(). These tests intend to validate io_uring operations.
1. io_uring_setup() creates submission queue and completion queue to perform subsequent operations on the io_uring instance.
2. io_uring_register() registers user buffers in kernel for long term usese.
3. io_uring_enter() initiates I/O operations using the shared SQ and CQ queue.
--- include/lapi/io_uring.h | 42 ++++ testcases/kernel/syscalls/io_uring/Makefile | 7 + .../kernel/syscalls/io_uring/io_uring01.c | 214 ++++++++++++++++++ 3 files changed, 263 insertions(+) create mode 100644 testcases/kernel/syscalls/io_uring/Makefile create mode 100644 testcases/kernel/syscalls/io_uring/io_uring01.c
include/lapi/io_uring.h +42 -0
@@ -280,4 +280,46 @@ int io_uring_enter(int fd, unsigned int to_submit, unsigned int min_complete, }#endif /* HAVE_IO_URING_ENTER */ +/* +* Check whether the ioring related system calls are supported on +* the current kernel. These system calls are enabled by default +* on kernel version 5.1.0 or higher. But they also might have +* been backported as well. +*/ +void io_uring_setup_supported_by_kernel (void ) +{ +if ((tst_kvercmp (5 ,1 ,0 )) <0 ) { +TEST (syscall (__NR_io_uring_setup,NULL ,0 )); +if (TST_RET !=-1 ) +SAFE_CLOSE (TST_RET ); +else if (TST_ERR ==ENOSYS ) +tst_brk (TCONF , +"Test not supported on kernel version < v5.1" ); + } +} + +void io_uring_register_supported_by_kernel (void ) +{ +if ((tst_kvercmp (5 ,1 ,0 )) <0 ) { +TEST (syscall (__NR_io_uring_register,NULL ,0 )); +if (TST_RET !=-1 ) +SAFE_CLOSE (TST_RET ); +else if (TST_ERR ==ENOSYS ) +tst_brk (TCONF , +"Test not supported on kernel version < v5.1" ); + } +} + +void io_uring_enter_supported_by_kernel (void ) +{ +if ((tst_kvercmp (5 ,1 ,0 )) <0 ) { +TEST (syscall (__NR_io_uring_enter,NULL ,0 )); +if (TST_RET !=-1 ) +SAFE_CLOSE (TST_RET ); +else if (TST_ERR ==ENOSYS ) +tst_brk (TCONF , +"Test not supported on kernel version < v5.1" ); + } +} +#endif /* IO_URING_H__ */
testcases/kernel/syscalls/io_uring/Makefile new file +7 -0
@@ -0,0 +1,7 @@ +# SPDX-License-Identifier: GPL-2.0-or-later +# Copyright (C) 2020 ARM, Ltd. All rights reserved. + +top_srcdir ?= ../../../.. + +include $(top_srcdir)/include/mk/testcases.mk +include $(top_srcdir)/include/mk/generic_leaf_target.mk
testcases/kernel/syscalls/io_uring/io_uring01.c new file +214 -0
214 lines not on this page · raw message
Signature · 1 line
2.17.1
PATCH v5 1/1 syscall: Add io_uring related tests
Hi Vikas,
not a complete review, just a few notes.
...
Quoted 6 lines
+/*
+ * Check whether the ioring related system calls are supported on
+ * the current kernel. These system calls are enabled by default
+ * on kernel version 5.1.0 or higher. But they also might have
+ * been backported as well.
+ */
IMHO the function name is obvious, so please shorten the comment :).
Quoted 25 lines
include/lapi/io_uring.h +25
-0+void io_uring_setup_supported_by_kernel (void ) +{ +if ((tst_kvercmp (5 ,1 ,0 )) <0 ) { +TEST (syscall (__NR_io_uring_setup,NULL ,0 )); +if (TST_RET !=-1 ) +SAFE_CLOSE (TST_RET ); +else if (TST_ERR ==ENOSYS ) +tst_brk (TCONF , +"Test not supported on kernel version < v5.1" ); + } +} + +void io_uring_register_supported_by_kernel (void ) +{ +if ((tst_kvercmp (5 ,1 ,0 )) <0 ) { +TEST (syscall (__NR_io_uring_register,NULL ,0 )); +if (TST_RET !=-1 ) +SAFE_CLOSE (TST_RET ); +else if (TST_ERR ==ENOSYS ) +tst_brk (TCONF , +"Test not supported on kernel version < v5.1" ); + } +} + +void io_uring_enter_supported_by_kernel (void )
include/lapi/io_uring.h int io_uring_enter(int fd, unsigned int to_submit, unsigned int min_complete, · line 323
Quoted 10 lines
include/lapi/io_uring.h +10
-0+{ +if ((tst_kvercmp (5 ,1 ,0 )) <0 ) { +TEST (syscall (__NR_io_uring_enter,NULL ,0 )); +if (TST_RET !=-1 ) +SAFE_CLOSE (TST_RET ); +else if (TST_ERR ==ENOSYS ) +tst_brk (TCONF , +"Test not supported on kernel version < v5.1" ); + } +}
IMHO It's unlikely that somebody would backport just some syscalls, not all 3. I'd be just for testing io_uring_setup().
Then no setup function is needed (use .setup = io_uring_setup_supported_by_kernel in struct tst_test).
And according to man pages only io_uring_setup() returns a file descriptor. The other two takes it as a parameter, so SAFE_CLOSE(TST_RET) is wrong.
https://www.mankier.com/2/io_uring_setup
https://www.mankier.com/2/io_uring_register
https://www.mankier.com/2/io_uring_enter
testcases/kernel/syscalls/io_uring/io_uring01.c line 195
Quoted 9 lines
testcases/kernel/syscalls/io_uring/io_uring01.c new file +9
-0+ +static void run (unsigned int n) +{ +struct tcase *tc = &tcases[n]; +int ret =0 ; + + s =SAFE_MALLOC (sizeof (*s)); +if (!s) +return ;
All SAFE_* macros (and safe_* functions which use them) are here to avoid these checks in the code. They exit executing via tst_brk(TCONF, "..."), therefore use just
s = SAFE_MALLOC(sizeof(*s));
(drop the check).
testcases/kernel/syscalls/io_uring/io_uring01.c line 206
Quoted 11 lines
testcases/kernel/syscalls/io_uring/io_uring01.c new file +11
-0+ +memset (s,0 ,sizeof (*s)); + ret =setup_io_uring_test (s, tc); +if (ret) +tst_res (TFAIL |TTERRNO ,"io_uring_setup error" ); + + ret =submit_to_uring_sq (s, tc); +if (ret) +tst_res (TFAIL |TTERRNO ,"io_uring_submit error" ); +else +tst_res (TPASS ,"functionality of io_uring API is correct" );
nit: we prefer return in if instead of else:
if (submit_to_uring_sq(s, tc)) {
tst_res(TFAIL | TTERRNO, "io_uring_submit error");
return;
}tst_res(TPASS, "functionality of io_uring API is correct");
Kind regards,
Petr