| Message ID | 1586429086-22975-9-git-send-email-xuyang2018.jy@cn.fujitsu.com |
|---|---|
| State | Rejected |
| Headers | show |
| Series | [v2,01/10] lapi/loop.h: Add fallback for loop ioctl and flag | expand |
Hi! This test fails for me, i.e. the attempt to enabe direct IO succeeds. I will have to have a closer look later on.
Hi Cyril > Hi! > This test fails for me, i.e. the attempt to enabe direct IO succeeds. I > will have to have a closer look later on. Now, I use 5.7.0-rc2 kernel to test, this case fails, I guess loop driver has changed its logic. I am trying. ps: when run ioctl_loop01.c, I also met another problem that /dev/loop0p1 and the corresponding sys file doesn't existed. dmesg report error "loop_reread_partitions: partition scan of loop0 () failed (rc=-16)". Also trying . >
Hi Cyril > Hi Cyril > > >> Hi! >> This test fails for me, i.e. the attempt to enabe direct IO succeeds. I >> will have to have a closer look later on. > Now, I use 5.7.0-rc2 kernel to test, this case fails, I guess loop > driver has changed its logic. I am trying. I write this test on fedora31(tmp directory use tmpfs), so I got a pseudo pass when setting dio. I misunderstood the meaning of the comment in the kernel. kernel commet as below: "We support direct I/O only if lo_offset is aligned with the logical I/O size of backing device, and the logical block size of loop is bigger than the backing device's and the loop needn't transform transfer." sb_bsize = bdev_logical_block_size(inode->i_sb->s_bdev); //get block size from backing device dio_align = sb_bsize - 1; I use logic block size of loop(1024 > 512) , so this test pass. Now, I think this test is meaningless and we can remove this. Best Regards Yang Xu > > ps: when run ioctl_loop01.c, I also met another problem that > /dev/loop0p1 and the corresponding sys file doesn't existed. dmesg > report error "loop_reread_partitions: partition scan of loop0 () failed > (rc=-16)". Also trying . >>
Hi Cyril > Hi Cyril > >> Hi Cyril >> >> >>> Hi! >>> This test fails for me, i.e. the attempt to enabe direct IO succeeds. I >>> will have to have a closer look later on. >> Now, I use 5.7.0-rc2 kernel to test, this case fails, I guess loop >> driver has changed its logic. I am trying. > > I write this test on fedora31(tmp directory use tmpfs), so I got a > pseudo pass when setting dio. > > I misunderstood the meaning of the comment in the kernel. > kernel commet as below: > "We support direct I/O only if lo_offset is aligned with the > logical I/O size of backing device, and the logical block > size of loop is bigger than the backing device's and the loop > needn't transform transfer." > > sb_bsize = bdev_logical_block_size(inode->i_sb->s_bdev); //get block > size from backing device > dio_align = sb_bsize - 1; > > I use logic block size of loop(1024 > 512) , so this test pass. > > Now, I think this test is meaningless and we can remove this. > > Best Regards > Yang Xu >> >> ps: when run ioctl_loop01.c, I also met another problem that >> /dev/loop0p1 and the corresponding sys file doesn't existed. dmesg >> report error "loop_reread_partitions: partition scan of loop0 () >> failed (rc=-16)". Also trying . Kernel commit d3ef55362("block: fix busy device checking in blk_drop_partitions") and commit cb6b771b05("block: fix busy device checking in blk_drop_partitions again") introduced this failure(ioctl_loop01, EBUSY error). more info see[1] [1]https://lkml.org/lkml/2020/4/23/380 >>>
diff --git a/runtest/syscalls b/runtest/syscalls index e9b0a8f60..f83561724 100644 --- a/runtest/syscalls +++ b/runtest/syscalls @@ -534,6 +534,7 @@ ioctl_loop03 ioctl_loop03 ioctl_loop04 ioctl_loop04 ioctl_loop05 ioctl_loop05 ioctl_loop06 ioctl_loop06 +ioctl_loop07 ioctl_loop07 ioctl_ns01 ioctl_ns01 ioctl_ns02 ioctl_ns02 diff --git a/testcases/kernel/syscalls/ioctl/.gitignore b/testcases/kernel/syscalls/ioctl/.gitignore index ec19b4901..3a3d49adc 100644 --- a/testcases/kernel/syscalls/ioctl/.gitignore +++ b/testcases/kernel/syscalls/ioctl/.gitignore @@ -12,6 +12,7 @@ /ioctl_loop04 /ioctl_loop05 /ioctl_loop06 +/ioctl_loop07 /ioctl_ns01 /ioctl_ns02 /ioctl_ns03 diff --git a/testcases/kernel/syscalls/ioctl/ioctl_loop07.c b/testcases/kernel/syscalls/ioctl/ioctl_loop07.c new file mode 100644 index 000000000..c58a92b3c --- /dev/null +++ b/testcases/kernel/syscalls/ioctl/ioctl_loop07.c @@ -0,0 +1,85 @@ +// 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 LOOP_SET_BLOCK_SIZE + * and LOOP_SET_DIRECT_IO. + * When blocksize is 1024(default align size is 512), set dio with + * 512 offset will fail. + */ + +#include <stdio.h> +#include <linux/loop.h> +#include <unistd.h> +#include <sys/types.h> +#include <stdlib.h> +#include <sys/mount.h> +#include "ioctl_loop_support.h" +#include "lapi/loop.h" +#include "tst_test.h" + +static char dev_path[1024], sys_loop_logicalpath[1024]; +static int dev_num, dev_fd, attach_flag, logical_block_size; + +static void verify_ioctl_loop(void) +{ + check_sys_value(sys_loop_logicalpath, 1024); + SAFE_IOCTL(dev_fd, BLKSSZGET, &logical_block_size); + if (logical_block_size == 1024) + tst_res(TPASS, "logical block size is 1024 by using BLKSSZGET ioctl"); + else + tst_res(TFAIL, "expected logical block size is 1024 but got %d by using BLKSSZGET ioctl", + logical_block_size); + 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"); +} + +static void setup(void) +{ + struct loop_info loopinfo; + + memset(&loopinfo, 0, sizeof(loopinfo)); + loopinfo.lo_offset = 512; + 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_logicalpath, "/sys/block/loop%d/queue/logical_block_size", dev_num); + tst_fill_file("test.img", 0, 1024 * 1024, 1); + tst_attach_device(dev_path, "test.img"); + attach_flag = 1; + + dev_fd = SAFE_OPEN(dev_path, O_RDWR); + TST_RETRY_FUNC(ioctl(dev_fd, LOOP_SET_STATUS, &loopinfo), TST_RETVAL_EQ0); + check_support_cmd(dev_fd, LOOP_SET_DIRECT_IO, 0, "LOOP_SET_DIRECT_IO"); + check_support_cmd(dev_fd, LOOP_SET_BLOCK_SIZE, 512, "LOOP_SET_BLOCK_SIZE"); + TST_RETRY_FUNC(ioctl(dev_fd, LOOP_SET_BLOCK_SIZE, 1024), TST_RETVAL_EQ0); +} + +static void cleanup(void) +{ + if (dev_fd > 0) + SAFE_CLOSE(dev_fd); + if (attach_flag) + tst_detach_device(dev_path); +} + +static struct tst_test test = { + .setup = setup, + .cleanup = cleanup, + .test_all = verify_ioctl_loop, + .needs_root = 1, + .needs_tmpdir = 1, + .needs_drivers = (const char *const []) { + "loop", + NULL + } +};
This is a basic ioctl test about loopdevice LOOP_SET_BLOCK_SIZE and LOOP_SET_DIRECT_IO. When logical blocksize is 1024, set dio with 512 offset will fail. Signed-off-by: Yang Xu <xuyang2018.jy@cn.fujitsu.com> --- runtest/syscalls | 1 + testcases/kernel/syscalls/ioctl/.gitignore | 1 + .../kernel/syscalls/ioctl/ioctl_loop07.c | 85 +++++++++++++++++++ 3 files changed, 87 insertions(+) create mode 100644 testcases/kernel/syscalls/ioctl/ioctl_loop07.c