[LTP] [PATCH v1 07/10] syscalls/ioctl_loop05: Add LOOP_SET_DIRECT_IO ioctl test
Yang Xu
xuyang_jy_0410@163.com
Mon Apr 6 09:53:04 CEST 2020
Hi Cyril
> Hi!
>> LOOP_SET_DIRECT_IO can updata a live loop device dio mode. It needs the
>> backing file also supports dio mode and the lo_offset is aligned with
>> the logical I/O size.
>>
>> It was introduced into kernel since 4.10
>> commit ab1cb278bc70 ("block: loop: introduce ioctl command of LOOP_SET_DIRECT_IO").
>>
>> Signed-off-by: Yang Xu <xuyang2018.jy@cn.fujitsu.com>
>> ---
>> runtest/syscalls | 1 +
>> testcases/kernel/syscalls/ioctl/.gitignore | 1 +
>> .../kernel/syscalls/ioctl/ioctl_loop05.c | 120 ++++++++++++++++++
>> 3 files changed, 122 insertions(+)
>> create mode 100644 testcases/kernel/syscalls/ioctl/ioctl_loop05.c
>>
>> diff --git a/runtest/syscalls b/runtest/syscalls
>> index 6e8d71d44..9644588f3 100644
>> --- a/runtest/syscalls
>> +++ b/runtest/syscalls
>> @@ -531,6 +531,7 @@ ioctl_loop01 ioctl_loop01
>> ioctl_loop02 ioctl_loop02
>> ioctl_loop03 ioctl_loop03
>> ioctl_loop04 ioctl_loop04
>> +ioctl_loop05 ioctl_loop05
>>
>> ioctl_ns01 ioctl_ns01
>> ioctl_ns02 ioctl_ns02
>> diff --git a/testcases/kernel/syscalls/ioctl/.gitignore b/testcases/kernel/syscalls/ioctl/.gitignore
>> index 039a5251c..f484d98d6 100644
>> --- a/testcases/kernel/syscalls/ioctl/.gitignore
>> +++ b/testcases/kernel/syscalls/ioctl/.gitignore
>> @@ -10,6 +10,7 @@
>> /ioctl_loop02
>> /ioctl_loop03
>> /ioctl_loop04
>> +/ioctl_loop05
>> /ioctl_ns01
>> /ioctl_ns02
>> /ioctl_ns03
>> diff --git a/testcases/kernel/syscalls/ioctl/ioctl_loop05.c b/testcases/kernel/syscalls/ioctl/ioctl_loop05.c
>> new file mode 100644
>> index 000000000..43bad6c18
>> --- /dev/null
>> +++ b/testcases/kernel/syscalls/ioctl/ioctl_loop05.c
>> @@ -0,0 +1,120 @@
>> +// SPDX-License-Identifier: GPL-2.0-or-later
>> +/*
>> + * Copyright (c) 2020 FUJITSU LIMITED. All rights reserved.
>> + * Author: Yang Xu <xuyang2018.jy@cn.jujitsu.com>
>> + *
>> + * This is a basic ioctl test about loopdevice.
>> + *
>> + * It is designed to test LOOP_SET_DIRECT_IO can updata a live
>> + * loop device dio mode. It need the backing file also supports
>> + * dio mode and the lo_offset is aligned with the logical I/O size.
>> + */
>> +#include <stdio.h>
>> +#include <unistd.h>
>> +#include <string.h>
>> +#include <stdlib.h>
>> +#include "ioctl_loop_support.h"
>> +#include "lapi/loop.h"
>> +#include "tst_test.h"
>> +
>> +#define DIO_MESSAGE "In dio mode"
>> +#define NON_DIO_MESSAGE "In non dio mode"
>> +
>> +static char dev_path[1024], sys_loop_diopath[1024];
>> +static int dev_num, dev_fd, attach_flag;
>> +
>> +static void check_dio_value(int flag)
>> +{
>> + struct loop_info loopinfoget;
>> +
>> + memset(&loopinfoget, 0, sizeof(loopinfoget));
>> +
>> + SAFE_IOCTL(dev_fd, LOOP_GET_STATUS, &loopinfoget);
>> + tst_res(TINFO, "%s", flag ? DIO_MESSAGE : NON_DIO_MESSAGE);
>> +
>> + if (loopinfoget.lo_flags & LO_FLAGS_DIRECT_IO)
>> + tst_res(flag ? TPASS : TFAIL, "lo_flags has LO_FLAGS_DIRECT_IO flag");
>> + else
>> + tst_res(flag ? TFAIL : TPASS, "lo_flags doesn't have LO_FLAGS_DIRECT_IO flag");
>> +
>> + check_sys_value(sys_loop_diopath, flag);
>> +}
>> +
>> +static void verify_ioctl_loop(void)
>> +{
>> + struct loop_info loopinfo;
>> +
>> + memset(&loopinfo, 0, sizeof(loopinfo));
>> +
>> + SAFE_IOCTL(dev_fd, LOOP_SET_DIRECT_IO, 0);
>> + check_dio_value(0);
>> + tst_res(TINFO, "Without setting lo_offset or sizelimit");
>> + SAFE_IOCTL(dev_fd, LOOP_SET_DIRECT_IO, 1);
>> + check_dio_value(1);
>> +
>> + tst_res(TINFO, "With offset equal to sector size");
>> + loopinfo.lo_offset = 512;
>
> We should use BLKSSZGET ioctl() to get the block size for direct I/O
> otherwise I would expect that it would break on one of the less common
> architectures. See also discussion at the end of man 2 open.
>
I see. Also, when testing LOOP_SET_BLOCK_SIZE ioctl, we can also call
this to check whether this ioctl succeed.
>> + safe_set_status(dev_fd, loopinfo);
>> + TEST(ioctl(dev_fd, LOOP_SET_DIRECT_IO, 1));
>> + if (TST_RET == 0) {
>> + tst_res(TPASS, "LOOP_SET_DIRECT_IO succeeded");
>> + check_dio_value(1);
>> + } else {
>> + tst_res(TFAIL | TTERRNO, "LOOP_SET_DIRECT_IO failed");
>> + }
>> +
>> + tst_res(TINFO, "With offset less than or unalign sector size");
>> + loopinfo.lo_offset = 256;
>> + safe_set_status(dev_fd, loopinfo);
>> +
>> + TEST(ioctl(dev_fd, LOOP_SET_DIRECT_IO, 1));
>> + if (TST_RET == 0) {
>> + tst_res(TFAIL, "LOOP_SET_DIRECT_IO succeeded unexpectedly");
>> + return;
>> + }
>> + if (TST_ERR == EINVAL)
>> + tst_res(TPASS | TTERRNO, "LOOP_SET_DIRECT_IO failed as expected");
>> + else
>> + tst_res(TFAIL | TTERRNO, "LOOP_SET_DIRECT_IO failed expected EINVAL got");
>> +
>> + loopinfo.lo_offset = 0;
>> + safe_set_status(dev_fd, loopinfo);
>> +}
>> +
>> +static void setup(void)
>> +{
>> + if (tst_fs_type(".") == TST_TMPFS_MAGIC)
>> + tst_brk(TCONF, "tmpfd doesn't support O_DIRECT flag, skip it");
>> +
>> + dev_num = tst_find_free_loopdev(dev_path, sizeof(dev_path));
>> + if (dev_num < 0)
>> + tst_brk(TBROK, "Failed to find free loop device");
>> +
>> + sprintf(sys_loop_diopath, "/sys/block/loop%d/loop/dio", dev_num);
>> + tst_fill_file("test.img", 0, 1024, 1024);
>> + tst_attach_device(dev_path, "test.img");
>> + attach_flag = 1;
>> + dev_fd = SAFE_OPEN(dev_path, O_RDWR);
>> + check_support_cmd(dev_fd, LOOP_SET_DIRECT_IO, 0, "LOOP_SET_DIRECT_IO");
>> +}
>> +
>> +static void cleanup(void)
>> +{
>> + if (dev_fd > 0)
>> + SAFE_CLOSE(dev_fd);
>> + if (attach_flag)
>> + tst_detach_device(dev_path);
>> + unlink("test.img");
>
> We don't have to remove the image here once the needs_tmpdir has been
> uncommented below, right?
Yes.
>
>> +}
>> +
>> +static struct tst_test test = {
>> + .setup = setup,
>> + .cleanup = cleanup,
>> + .test_all = verify_ioctl_loop,
>> + .needs_root = 1,
>> +// .needs_tmpdir = 1,
>
> Looks like a forgotten debug measure.
>
Yes, because some distribution use tmpfs in tmp directory, then this
case will TCONF(tmpfs doesn't supprt DIO). I am indecisive, or I should
use all_filesystem =1, so it will not report TCONF? Or, keep it?
>> + .needs_drivers = (const char *const []) {
>> + "loop",
>> + NULL
>> + }
>> +};
>> --
>> 2.23.0
>>
>>
>>
>>
>> --
>> Mailing list info: https://lists.linux.it/listinfo/ltp
>
More information about the ltp
mailing list